Modeling global pricing and launching of new drugs · Modeling Global Pricing and ... his helps in...

DOCTORAL THESIS

Modeling Global Pricing and Launching of New Drugs

Author: Borja García Lorenzo Las Palmas de Gran Canaria, April 2014

DOCTORADO EN ECONOMÍA: APLICACIONES A LAS FINANZAS Y SEGUROS, A LA ECONOMÍA SECTORIAL, AL MEDIO AMBIENTE Y A LAS

INFRAESTRUCTURAS.


Tesis doctoral presentada por D. Borja García Lorenzo

Dirigida por Dra. Beatriz González López-Valcárcel

La Directora, El Doctorando,

Las Palmas de Gran Canaria, abril de 2014

Existimos porque alguien piensa en nosotros,

y no al revés

Acknowledgments

Foremost, I would like to express my sincere gratitude to my advisor Dr. Beatriz

González López-Valcárcel for the continuous support of my Ph.D study and research, for

his patience, motivation, enthusiasm, constant feedback and immense knowledge. His

guidance helped me in all the time of research and writing of this thesis.

My sincere thanks also go to Dr. Izabela Jelovac and Dr. Margaret Kyle for offering

me the opportunities to enjoy my visiting scholars in the Groupe d’Analyse et Théorie

Economique (GATE) and the Toulouse School of Economics (TSE) respectively, for their

encouragement, insightful comments, and hard questions. I also thank Carlos J. Pérez for

his helps in the field of decision theory.

I thank my fellow officemates in the University of Las Palmas de Gran Canaria

(ULPGC): Reinaldo, Hicham, Rubén, Teresa and Federico for the stimulating discussions,

for the hard days we were working together, and for all the fun we have had in the last

years.

I gratefully acknowledge the funding received towards my PhD from the Canarian

Agency for Research, Innovation and Information Society of the Canarian Government

(ACIISI). Also, I thank IMS for providing the data for the empirical section, particuarly to

Miguel Martínez.

Last but not the least, I would like to thank my parents Roque and Pepa for

supporting my education without regard, and Naira, for her understanding, even so she

has not a Ph.D, she has supported me as if she was one. Friends around me have been a

great support to reach this moment. Thank you all.

Contents

List of Figures ............................................................................................................... XV

List of Tables .............................................................................................................. XVII

List of abbreviates ....................................................................................................... XIX

Introduction ................................................................................................................... 1

1 Chapter 1: Global Pricing and Launching of New Drugs: What Does the Theory Say?

What Do the Empirical Models Show? ........................................................................... 7

1.1 Introduction ................................................................................................................. 7

1.2 What does the theory say? ........................................................................................... 9

1.2.1 Launching as a result of bargaining process: trade‐offs between pricing and

launching .................................................................................................................................... 9

1.2.2 How the ERP is affecting the bargaining results in pricing and launching? ................ 10

1.2.3 Which role does PT play in the pharmaceutical market? ............................................ 13

1.2.4 How asymmetric information on quality of drugs may affect drug pricing and

launching? ................................................................................................................................ 14

1.2.5 Are important the headquarters location and the contacts among firms when

pricing drugs? ........................................................................................................................... 16

1.2.6 Which effects do arise in pricing and innovation when countries apply internal RP? 17

1.3 What do the empirical models show? ......................................................................... 20

1.3.1 Samples, Variables and Methods ................................................................................ 20

1.3.2 Factors influencing prices ............................................................................................ 24 1.3.2.1 Drug Characteristics ............................................................................................................ 24 1.3.2.2 Competition and substitutes .............................................................................................. 26 1.3.2.3 Regulation Characteristics .................................................................................................. 28 1.3.2.4 Country Characteristics ...................................................................................................... 30 1.3.2.5 Firm Characteristics ............................................................................................................ 31

1.3.3 Factors influencing launching ...................................................................................... 33 1.3.3.1 Drug Characteristics ............................................................................................................ 33 1.3.3.2 Competition and substitutes .............................................................................................. 34 1.3.3.3 Regulation Characteristics .................................................................................................. 34 1.3.3.4 Country Characteristics ...................................................................................................... 37 1.3.3.5 Firm Characteristics ............................................................................................................ 38

1.4 Discussion .................................................................................................................. 39

XII Modeling Global Pricing and Launching of New Drugs

2 Chapter 2: External Reference Pricing and Pharmaceutical Cost‐Containment. ...... 47

2.1 Introduction ............................................................................................................... 47

2.2 The model .................................................................................................................. 53

2.3 Price Setting and Sequential Launch ........................................................................... 63

2.3.1 The firm is trusted by the health agency ..................................................................... 63

2.3.2 The firm states the number of QALYs above the true value ....................................... 75

2.4 Comparing Policies: CEA vs. ERP .................................................................................. 75

2.5 Conclusions ................................................................................................................ 78

3 Chapter 3: Global Pricing and Launching of New Drugs. An Econometric Approach 81

3.1 Introduction ............................................................................................................... 81

3.2 Data description ......................................................................................................... 82

3.3 Replicating Danzon and Epstein (2008) ....................................................................... 85

3.3.1 The D&E model ............................................................................................................ 85

3.3.2 Data ............................................................................................................................. 86

3.3.3 Comparison of results ................................................................................................. 87 3.3.3.1 Launch equation ................................................................................................................. 87 3.3.3.2 Launch price equation ........................................................................................................ 89

3.4 Replicating Verniers et al. (2011) ................................................................................ 91

3.4.1 The Verniers et al. model ............................................................................................ 91

3.4.2 Data ............................................................................................................................. 94

3.4.3 Comparison of results ................................................................................................. 94 3.4.3.1 Launch window equation ................................................................................................... 94 3.4.3.2 Launch price equation ........................................................................................................ 97

3.5 New Pricing and Launching Model (NPLM) .................................................................. 98

3.5.1 The Model ................................................................................................................... 98

3.5.2 Data ........................................................................................................................... 102

3.5.3 Results ....................................................................................................................... 102 3.5.3.1 Launch delay equation ..................................................................................................... 102 3.5.3.2 Relative launch price equation ......................................................................................... 104

3.5.4 Discussion .................................................................................................................. 108

3.6 Conclusions .............................................................................................................. 110

Conclusions and further research ............................................................................... 113

A. Appendix A .......................................................................................................... 117

A.1. Method of Review ................................................................................................... 117

A.1.1 Search strategy .......................................................................................................... 117

A.1.2 Selection and Exclusion Criteria ................................................................................ 117

A.1.3 Search results ............................................................................................................ 118

B. Appendix B .......................................................................................................... 139

B.1 Decision tree and proofs ........................................................................................... 139

C. Appendix C .......................................................................................................... 147

Contents XIII

C.1. Variable definitions of Danzon and Epstein (2008) ................................................... 147

C.2. Variables definitions of Verniers et al. (2011) ........................................................... 162

C.3. Variable definitions of the NPLM ............................................................................. 169

C.4. Selection of parametric model ................................................................................. 173

Resumen en español .................................................................................................. 179

Motivación ........................................................................................................................ 179

Objetivos ........................................................................................................................... 183

Planteamiento y metodología ............................................................................................ 185

Revisión de la literatura ......................................................................................................... 185

Modelo teórico ....................................................................................................................... 187 Fijación de precios y lanzamiento secuencial ................................................................................... 200

Modelo empírico .................................................................................................................... 205 Réplica del modelo de Danzon y Epstein (2008) .............................................................................. 205 Réplica del modelo de Verniers et al. (2011) .................................................................................... 207 Nuevo modelo precio y lanzamiento (NPLM) ................................................................................... 210

Resultados ......................................................................................................................... 217

Revisión de la literatura ......................................................................................................... 217

Modelo teórico ....................................................................................................................... 219

Modelo empírico .................................................................................................................... 222

Conclusiones ...................................................................................................................... 228

References ................................................................................................................. 235

Contents XV

List of Figures FIGURE 2.1 OPTIMAL COUNTRY LAUNCH SEQUENCE UNDER A) ....................................................................................... 71

FIGURE 2.2 OPTIMAL COUNTRY LAUNCH SEQUENCE UNDER B) ....................................................................................... 72

FIGURE 2.3 OPTIMAL COUNTRY LAUNCH SEQUENCE UNDER C) ....................................................................................... 73

FIGURE 2.4 OPTIMAL COUNTRY LAUNCH SEQUENCE UNDER D) ...................................................................................... 74

FIGURE 2.5. ERP VS. CEA ..................................................................................................................................... 77

FIGURE A.1 FLOW DIAGRAM OF LITERATURE SCREENING PROCESS ............................................................................... 119

FIGURE B.1 DECISION TREE .................................................................................................................................. 140

FIGURE C.1. DENSITY FUNCTION OF DELAY IN MONTHS. RETAIL MARKET ..................................................................... 174

FIGURE C.2. DENSITY FUNCTION OF DELAY IN MONTHS. HOSPITAL MARKET ................................................................. 174

FIGURE C.3. WEIBULL, GAMMA. RETAIL MARKET ..................................................................................................... 175

FIGURE C.4. WEIBULL, INVERSE GAUSSIAN. RETAIL MARKET ....................................................................................... 175

FIGURE C.5. GOMPERTZ, GAMMA. RETAIL MARKET .................................................................................................. 175

FIGURE C.6.GOMPERTZ, INVERSE GAUSSIAN. RETAIL MARKET ..................................................................................... 176

FIGURE C.7. WEIBULL, GAMMA. HOSPITAL MARKET ................................................................................................. 176

FIGURE C.8. WEIBULL, INVERSE GAUSSIAN. HOSPITAL MARKET .................................................................................. 177

FIGURE C.9. GOMPERTZ, GAMMA. HOSPITAL MARKET .............................................................................................. 177

FIGURE C.10. GOMPERTZ,INVERSE GAUSSIAN. HOSPITAL MARKET .............................................................................. 177

FIGURA 4.1 DIAGRAMA DE FLUJO DEL PROCESO DE SELECCIÓN DE LA LITERATURA ............................................................ 187

FIGURA 4.2 ÁRBOL DE DECISION ............................................................................................................................ 199

FIGURA 4.3 REGIONES DE SECUENCIAS ÓPTIMAS DE LANZAMIENTO SI A) ........................................................................ 203

FIGURA 4.4 FUNCIÓN DE DENSIDAD DE RETRASO EN EL LANZAMIENTO EN MESES. MERCADO AMBULATORIO ........................ 212

FIGURA 4.5. RESIDUOS DE COX‐SNELL PARA EL MODELO WEIBULL, GAMMA. MERCADO AMBULATORIO ............................. 212

FIGURA 4.6 TRADE‐OFF ENTRE CEA Y ERP .............................................................................................................. 220

Contents XVII

List of Tables TABLE 2.1 PPR FOR OPTIMAL COUNTRY LAUNCH SEQUENCE (PI , QI) ................................................................................ 69 TABLE 2.2 COUNTRY LAUNCH SEQUENCE IN FIGURE 2.2 .............................................................................................. 71 TABLE 2.3. OPTIMAL COUNTRY LAUNCH SEQUENCE IN FIGURE 2.3 ................................................................................ 72 TABLE 2.4. OPTIMAL COUNTRY LAUNCH SEQUENCE IN FIGURE 2.4 ................................................................................ 73 TABLE 2.5. OPTIMAL COUNTRY LAUNCH SEQUENCE IN FIGURE 2.5 ................................................................................ 74 TABLE 3.1 DESCRIPTIVE STATISTICS. RETAIL MARKET .................................................................................................... 83 TABLE 3.2 DESCRIPTIVE STATISTICS. HOSPITAL MARKET ................................................................................................ 84 TABLE 3.3 RELATIVE PRICES PEARSON CORRELATION. RETAIL AND HOSPITAL MARKET ......................................................... 85 TABLE 3.4. BIVARIATE TEST. ERP VS. NO ERP ........................................................................................................... 85 TABLE 3.5. BIVARIATE TEST. EMA VS. NO EMA ........................................................................................................ 85 TABLE 3.6. LAUNCH DELAY EQUATION OF THE NPLM ................................................................................................ 103

TABLE 3.7. RELATIVE LAUNCH PRICE EQUATION OF THE NPLM .................................................................................... 106

TABLE A.1 OVERVIEW OF THEORETICAL STUDIES ....................................................................................................... 120

TABLE A.2. OVERVIEW OF EMPIRICAL STUDIES .......................................................................................................... 121

TABLE B.1. HEALTH AGENCY SURPLUS OF COUNTRY C................................................................................................ 143

TABLE B.2. HEALTH AGENCY SURPLUS OF COUNTRY D ............................................................................................... 143

TABLE C.1. LAUNCH EQUATION: D&E VS. UPDATED MODEL ....................................................................................... 153

TABLE C.2. LAUNCH PRICE EQUATION: D&E VS. UPDATED MODEL............................................................................... 157

TABLE C.3. LIST OF COUNTRIES. VERNIERS ET AL. VS. UM .......................................................................................... 165

TABLE C.4. LAUNCH WINDOW AND LAUNCH PRICE EQUATIONS. VERNIERS ET AL. VS. UPDATED MODEL ............................. 167 TABLE C.5. VARIABLE CLASSIFICATION OF THE NPLM ................................................................................................ 171 TABLE C.6. PROBIT SELECTION EQUATION OF NPML ................................................................................................. 171 TABLE C.7. GOODNESS OF FIT OF PARAMETRIC MODELS. RETAIL MARKET ....................................................................... 174 TABLE C.8. GOODNESS OF FIT OF PARAMETRIC MODELS. HOSPITAL MARKET ................................................................... 176 TABLA 4.1. PPR PARA LA SECUENCIA ÓPTIMA DE LANZAMIENTO (PI , QI) ......................................................................... 202 TABLA 4.2. REGIONES DE SECUENCIAS ÓPTIMAS DE LANZAMIENTO A) ........................................................................... 204 TABLA 4.3.BONDAD DEL AJUSTE DE LOS MODELOS PARAMÉTRICOS. MERCADO AMBULATORIO .......................................... 212 TABLA 4.4. ECUACIÓN DEL PRECIO RELATIVO DE LANZAMIENTO ................................................................................... 225 TABLA 4.5. ECUACIÓN DE RETRASO EN EL LANZAMIENTO ............................................................................................ 227

List of abbreviates

AIFA Italian National Agency for Drug Administration and Control Prices

AME Average Marginal Effect

ATC Anatomic Therapeutic Chemical Classification System

CHEPA Centre for Health Economics and Policy Analysis

CEA Cost-Effectiveness Analysis

CRES Centre de Recerca en Economia i Salut

DDD Defined Daily Dosage

D&E Danzon & Epstein

EMA European Medical Agency

ERP External Reference Pricing

EU European Union

FDA Food and Drug Administration

GDP Gross Domestic Product

GLS Generalized Least Squares

GPRM Global Price Reporting Mechanism

HHI Hirschmand-Herfindähl Index

HTA Health Technology Assessment

XX Modeling Global Pricing and Launching of New Drugs

ICER Incremental Cost-Effectiveness Ratio

IMF International Monetary Fund

IMR Inverse Mills Ratio

LSE London School of Economics

MES Minimum Efficacy Standard

MEPS Medical Expenditure Panel Survey

MLIC Middle and Low Income Country

NBER The National Bureau of Economic Research

NCE New Chemical Entities

NGO No-Governmental Organization

NICE National Institute for Health and Care Excellence

OECD Organisation for Economic Cooperation and Development

OF Objective Function

OLS Ordinary Least Squares

OTC Over-the-Counter

PC Price Cap

PE Public Expenses

PI Parallel Importer

PPI Producer Price Indexes

PPP Purchasing Power Parities

PPR Preliminary Results

Contents XXI

PRISMA Preferred Reporting Items for Systematic Reviews and

Meta-Analyses

PT Parallel Trade

QALY Quality-adjusted Life Years

R&D Research & Development

RP Reference Price

SU Standard Unit

UK United Kingdom

UM Updated Model

US United States

WtP Willingness to pay

3SLS Three-stage least squares

Introduction Pharmaceuticals are sold in a global market. This characteristic implies a specific

bargaining procedure between pharmaceutical firms and countries’ health agencies. On

the one hand, these firms make strategic decisions when launching medicines in different

countries and to maximize their global profits; and on the other hand, countries’ health

agencies implement pricing policies in order to control their pharmaceutical expenditure

and to guarantee access to medicines.

From the perspective of national health insurances, pricing policies within the

pharmaceutical market are a key factor in controlling public expenditure (Scherer, 1993,

Lobo, 2014)1. Particularly, the total pharmaceutical bill: across the Organisation for

Economic Cooperation and Development (OECD) countries in 2009, this bill is estimated

to have accounted for around 19% of health spending. In relation to the overall economy,

pharmaceutical spending accounts for an average 1.5% of GDP in OECD countries.

However, the dispersion around this average is high, pharmaceutical spending accounts

for less than 1% of GDP in Norway and Denmark, while it reaches close to 2.5% of GDP

in Greece, Hungary and the Slovak Republic. Expenditure on pharmaceuticals is

predominantly financed through third-party payers in most OECD countries – either

through the public health insurance, which accounts for around 60% of the total on

average, or through private insurance coverage, leaving an average of more than a third

of the total to be charged to households (OECD, 2011).

From the pharmaceutical industry view, pricing and launching a new drug is a

complex task directly connected to R&D policy, industrial policy and healthcare policy.

Hence, pricing and launching are major strategic decisions. In many countries, the price is

agreed with health care insurance providers (public or private). National pricing policies

1 The case of Spain as an example of the price regulation LOBO, F. 2014. La Intervención de Precios de los Medicamentos en España, Madrid, Springer..


2

and strategies are essential elements in setting prices and making medicines available,

since drug pricing should contribute to enhancing social welfare and take into account the

interests of the industry, consumers and public insurers. Therefore, encouragement must

be provided to develop new medicines, make them available to consumers and, at the

same time, control pharmaceutical expenditure.

Pricing and launching involve trade-offs between public welfare and private profits,

between the interests of the manufacturer and those of the country. When countries set a

drug price, they risk the possibility of not providing it at the time they desire, which may

have consequences for the health and the welfare of the population (Lichtenberg, 2005).

In turn, a firm that delays the launch of a medicine in a country is also delaying the profits

to be derived from this country. However, in an increasingly globalized world, national

pricing/launching of drugs has become in fact an international matter and

interdependencies across countries should be taken into account. Both companies and

countries must act locally but think globally. Due to mechanisms like external reference

pricing (ERP, henceforth) and parallel trade (PT, henceforth) (Danzon et al., 2005,

Danzon and Epstein, 2008, Garcia Mariñoso et al., 2011), setting the price of a drug in a

particular country influences other countries’ pricing and launching. The use of ERP by

countries may make a firm apply international pricing strategies that may harm countries’

welfare. On the one hand, the firm may set a single price2, which may benefit high-price3

countries but harm low-price ones. On the other hand, the firm may either attempt to set

high4 prices in the first countries to avoid low-prices in later launches via ERP, or delay

launches in low-price countries to avoid spill-over effects. These strategies can harm low-

price countries, and may even harm high-price ones (Garcia Mariñoso et al., 2011).

Among existing drug pricing policies, most countries in the industrialized world

have implemented either Cost-Effectiveness Analysis (CEA, henceforth) or ERP at some

2 Two factors contribute to price uniformity between different markets: a) the threats of parallel imports, and b) the use of international reference pricing DANZON, P. M. & TOWSE, A. 2003. Differential Pricing for Pharmaceuticals: Reconciling Access, R&D and Patents. International Journal of Health Care Finance and Economics, 3, 183-205..

3 In the long run, consumers from high price countries will be worse off if this lower price results in lower than expected returns on R&D, and hence fewer new medicines than they would have been willing to pay for DANZON, P. M. 1997. Price Discrimination for Pharmaceuticals: Welfare Effects in the US and the EU. International Journal of the Economics of Business, 4, 310-322..

4 This company strategy will not work if the high-price country revises its prices downwards after launch DANZON, P. M. & TOWSE, A. 2003. Differential Pricing for Pharmaceuticals: Reconciling Access, R&D and Patents. International Journal of Health Care Finance and Economics, 3, 183-205, DANZON, P. M. 1997. Price Discrimination for Pharmaceuticals: Welfare Effects in the US and the EU. International Journal of the Economics of Business, 4, 310-322.

Introduction

3

point in time with the aim of controlling pharmaceutical expenditure, while still ensuring

access to medicines, mainly in on-patent medicines (Espin J et al., 2011, Rawlins, 2012).

In this thesis, ERP is defined as “the practice of setting a price cap for

pharmaceuticals, based on ex-manufacturer5 prices of identical or comparable products in

other countries” (Garcia Mariñoso et al., 2011). Most countries use ERP as a

pharmaceutical pricing strategy. The use of ERP as a mechanism to set pharmaceutical

prices is quite widely applied: 24 of the 30 OECD countries (Espin J et al., 2011) and

approximately 24 of the 28 EU Member States (Leopold et al., 2012) have used it.

However, ERP is not applied homogeneously in every country. There are a wide variety of

methods to design a foreign price index (Leopold et al., 2012, Espin J et al., 2011). It

mainly depends on each country’s basket, the type of prices collected6, the method used

(the lowest price, the average price, a percentage of the previous ones, etc.) and whether

a weighted-index7 is used or not. We also note that some countries take into account ERP

as a complementary pricing policy together with other pricing policies to help to make the

price decision, and therefore it is not exclusively applied as a blind pricing policy8. ERP is

used because of its simplicity at a technical and analytical level; collecting price

information abroad does not require a huge effort. Furthermore, ERP users think that the

prices taken as reference are roughly right, suitable or fair. However, they recognise that

it is difficult to assess if the resulting prices are appropriate, efficient or optimal in

accordance with any objective criterion. Additionally, if referencing countries set their

prices too high or too low, then any country later applying the ERP method may run the

risk of repeating the same mistake (Espin J et al., 2011).

CEA in health economics aims to estimate the ratio between the cost of a health-

related intervention and the benefit it produces in terms of the number of years lived in full

health by the beneficiaries. Cost is measured in monetary units, while benefit needs to be

expressed in gain of health measured by quantitative values. However, unlike cost– 5 Prices are ex-manufacturer prices.

6 Current price vs. price at launch

7 The most widely method used for new drugs is through non-weighted measures; such methods will not help to achieve the target of obtaining a comparable average level of prices. The application of weighted price indexes, comparable and useful as reference to the rest of countries, has been proposed DANZON, P. M. & CHAO, L. W. 2000. Cross-national price differences for pharmaceuticals: How large, and why? Journal of Health Economics, 19, 159-195..

8 Espín et al. state that “regulators might not always be able or willing to “impose” a certain price, but instead use the price computed as a benchmark or reference for negotiations, often alongside other criteria, such as cost-plus, internal or therapeutic pricing”.


4

benefit analysis, the benefits do not have to be expressed in monetary terms. In

pharmaeconomics, it is usually expressed in quality-adjusted life years (QALYs)9

(National Institute for Health and Care Excellence (NICE), 2010). The incremental cost-

effectiveness ratio (ICER) is the ratio between the difference in costs and the difference in

benefits of two interventions. A firm knows this threshold for a given country; therefore, it

conducts CEA and calculates the number of QALYs gained if the drug were provided in

one country. Since the firm is aware of both threshold and number of QALYs, it offers the

country the drug at a certain price. However, the firm may upwardly distort the number of

QALYs to obtain greater profits. Then, it is the country that may revise the firm’s CEA

applying its own CEA to estimate a fair price. However, this CEA requires resources and

consequently an investment of money by the country.

On the whole, the trade-offs previously mentioned have driven theoretical and

empirical research, particularly in recent years. Besides, interdependences among

markets due to the implementation of ERP and the presence of PT may change pricing

and launching strategy and lead to small price differences globally.

First of all, this thesis aims to provide an overall perspective of original theoretical

and empirical analyses of global interdependencies with respect to drug pricing and

launching worldwide. Secondly, it attempts to find the main factors influencing both launch

prices and launch of new drugs. Then, a theoretical model based on a bargaining model

between a pharmaceutical firm and two countries’ health agencies is provided, which

aims to analyse the convenience of applying ERP instead of CEA as a cost-containment

policy on pharmaceutical expenditure. Ultimately, the thesis develops an empirical model

that aims at analyzing the trade-off between pricing and launching and the impact of ERP

policy on pricing and launching.

This thesis is divided into three chapters. Each chapter deals with a specific area

ofresearch; therefore, each chapter has its own introduction, development and

conclusions. The first chapter describes the current state of knowledge in this area, the

second chapter develops a theoretical model and the third chapter provides evidence

9 The QALY is a measure of disease burden, including both the quality and the quantity of life lived. The QALY model requires utility independent, risk neutral, and constant proportional trade-off behaviour. The QALY is based on the number of years of life that would be added by the intervention. Each year in perfect health is assigned the value of 1.0 down to a value of 0.0 for being dead. If the extra years are not lived in full health, for example if the patient loses a limb, or goes blind or has to use a wheelchair, then the extra life-years are given a value between 0 and 1 to account for this.

Introduction

5

based on the development of an empirical model.

Chapter 1 provides a systematic survey of original scientific studies up to April

2012 based on PRISMA (Preferred Reporting Items for Systematic reviews and Meta-

Analyses). This chapter reviews what we know about the main factors influencing both

launch prices and launch of new drugs, and also, whether there are any common general

patterns that could be derived from economic models to explain the strategic games

played by governments, public insurers and pharmaceutical companies. Finally, it

evaluates if there is any empirical evidence of these factors in OECD countries and if ERP

and PT make the markets inseparable in launching and pricing.

Chapter 2 suggests a game based on a bargaining model involving sequential

launching of one drug by one firm across two countries based on a “take-it-or-leave-it-

offer” procedure that has been developed under asymmetric information (Muthoo, 1999).

The model is based on the best pharmaceutical firm strategy to sequentially launch a new

product, given the size of the countries, the CEA cost, the launch delay cost and the

prices set by each pricing policy. In brief, the country chooses either to apply ERP without

any additional cost or use CEA, with its corresponding investment of money. Then, the

firm chooses its optimal country launch sequence. We introduce two important

innovations that are particularly noteworthy: firstly, we include different types of countries

depending on their ERP. Thus, we differentiate the type of country based on their formula

of foreign prices to implement ERP criteria, either the minimum or the average price

observed. Secondly, we introduce the launch delay cost and the cost related to applying

CEA to check if the firm declares the true QALY of the drug or not.

Chapter 3 evaluates two studies (Danzon and Epstein, 2008, Verniers et al., 2011)

where econometric models of pricing and launching have been applied. The same data

treatment and methodology conducted by the two studies have been implemented with

our database. On the one hand, we replicate the study of Danzon and Epstein published

in 2008 (Danzon and Epstein, 2008) with our database containing more recent data, and

we compare our results with theirs. Danzon and Epstein (Danzon and Epstein, 2008) data

cover the years 1992-2003, while our data cover the period 2004-2010. Then, we look at

the study of Verniers et al. published in 2011 (Verniers et al., 2011) to check if the results

have changed due to the use of more recent data (2010 vs. 2008), and if the results are

robust for the choice of the list of countries. Verniers et al. (Verniers et al., 2011) applied


6

their model to a large dataset of countries, both rich and poor; however, we restrict our

application to developed countries. Afterwards, this thesis develops an empirical model

that focuses on the analysis of the trade-off between pricing and launching as well as the

impact of ERP policy on pricing and launching controlling for molecules, regulation and

country characteristics. It develops two equations consisting of a launch delay equation

and a launch price equation. The contribution of this chapter to the previous literature

studied involves an analysis of data at presentation level10, the consideration of the

relative launch price11 as an endogenous variable in the launch price equation, the study

of the launch delay as a duration time variable and the analysis of the inpatient market.

Additionally, we introduce the country size and the country purchasing power as

additional explanatory variables in the same model. We use data from IMS Health

database on 56 new molecules launched in 20 countries belonging to 11 therapeutic

classes, all of them approved through the centralised procedure by the EMA, during the

study period, 2004-2010. We have collected yearly inpatient and outpatient sales in

euros at ex-manufacturer price and unit volume (IMS SU). The number of molecule-

presentation-country observations are 1 and 1 for the launch equation and the launch

price equation respectively.

10 We define two products with the same presentation when both products belong to the same molecule i and have the same quantity of active ingredient per standard unit (SU) (see definition of SU in Chapter 3 section 3.3.2).

11 Defined in Appendix C.3.

1 Chapter 1: Global Pricing and Launching of New Drugs: What Does the Theory Say? What Do the Empirical Models Show?

1.1 Introduction

The total pharmaceutical bill across OECD countries in 2009 is estimated to have

accounted for around 19% of health spending. In relation to the overall economy,

pharmaceutical spending accounts for 1.5% of GDP on average in OECD countries.

However, the dispersion around this average is high, pharmaceutical spending accounts

for less than 1% of GDP in Norway and Denmark, while it reaches close to 2.5% of GDP

in Greece, Hungary and the Slovak Republic. Expenditure on pharmaceuticals is

predominantly financed through third-party payers in most OECD countries – either

through the public health insurance, which accounts for around 60% of the total on

average, or through private insurance coverage, leaving an average of more than a third

of the total to be charged to households. Hence, pricing policies within the pharmaceutical

market are a key factor in controlling public expenditure in this field (OECD, 2011).

Pharmaceutical price regulation is high on policy agendas in many countries,

either because countries have just reformed, intend to reform or question their practices.

This is why this chapter proposes to review what we know about the main factors

influencing both launch prices and launch of new drugs: firms’ pricing strategies,

regulators’ pricing policies, and the empirical evidence in OECD countries.

Pricing and launching in the pharmaceutical industry is a complex task directly

connected to R&D policy, industrial policy and healthcare policy. From the perspective of

8 Modeling Global Pricing and Launching of New Drugs

the pharmaceutical companies, pricing and launching are their major strategic decisions.

The price may be agreed with health care insurance providers (public or private). At this

point, national pricing policies and strategies are important elements in setting prices and

making medicines available. Drug pricing should contribute to enhancing social welfare,

taking into account the interests of the national industry, consumers and public insurers.

Therefore, encouragement must be provided to developing new medicines, making them

available to consumers, yet, at the same time, controlling pharmaceutical expenditure.

Pricing and launching a medicine involve trade-offs between public welfare and private

profits, between the interests of the manufacturer and those of the country. A firm that

delays the launch of a medicine in a country is also delaying the profits to be derived from

this country. However, in an increasingly globalized world, national pricing/launching of

drugs has become in fact an international matter and interdependencies across countries

should be taken into account. Both companies and countries must act locally but think

globally. Due to mechanisms like external reference pricing (ERP) and parallel trade (PT)

(Danzon et al., 2005, Danzon and Epstein, 2008, Garcia Mariñoso et al., 2011), setting

the price of a drug in a particular country influences pricing and launching in other

countries. For instance, by delaying the launch of a new drug in a low-price country, a

company manages to prevent this low price from over spilling into other countries through

reference pricing.

The trade-offs we mentioned have driven theoretical and empirical research

particularly in recent years. What are the main factors influencing both launch prices and

launch of new drugs? Are there any common general patterns that could be derived from

economic models to explain the strategic games played by governments, public insurers

and pharmaceutical companies? Is there any empirical evidence of these factors in OECD

countries? Do ERP and PT make markets inseparable in launching and pricing?

This chapter tries to answer these questions. We performed a systematic review of

the literature from the period 1995-April 2012 and synthesize the main facts, ideas and

results from them. Our search covers theoretical and empirical models. Details of the

search can be found in Appendix A. The rest of this chapter is organized as follows. In

Section 1.2, the theoretical studies are discussed. Section 1.3 examines the empirical

studies identified, followed by a discussion in Section 1.4.

Chapter 1: Global Pricing and Launching of New Drugs: What Does the Theory Say?

What Do the Empirical Models Show?

9

1.2 What does the theory say?

In this section, we ask whether there are common general patterns that could be

derived from economic models to explain the strategic games played by governments,

public insurers and pharmaceutical companies. Table A.1 summarizes the main

characteristics of the articles that met the selection criteria for theoretical studies (see

Appendix A).

We firstly describe under which conditions pricing and launching occurs and,

according to the paper of Danzon et al., which factors determine the drug pricing and

launching (Danzon et al., 2005). More recently, Danzon and Epstein (Danzon and

Epstein, 2008), based on the theoretical model developed by Danzon et al. (Danzon et

al., 2005), include more explanatory factors. Both papers show intuitively how these

factors affects drug pricing and launching. In the literature, we find further papers that

theoretically examine and predict the former factors.

Each paper assumes different hypothesis to develop their corresponding

theoretical models. Mainly, these hypotheses concern the number and types of firms and

countries considered; the design of objective functions; the type of drug sold; the

information available and the timing of decisions. The common hypothesis considered by

the most theoretical papers is that markets are not separable, which implies that both

pricing and launching occur worldwide.

1.2.1 Launching as a result of bargaining process: trade-offs between pricing and launching

Two papers analyse the conditions under which drug launching and pricing arise.

Danzon et al. focus on launching (Danzon et al., 2005) and later, Danzon and Epstein

anlayze pricing (Danzon and Epstein, 2008). The first paper considers a market where a

pharmaceutical firm sells a drug in at least more than one market (country), which are not

separable. One country, which regulates prices, offers a price and the firm can accept it or

refuse it. The authors assume that the government has a reservation or maximum offer

price that depends positively on the price of existing products, the ICER and on the

country per capita income. On the other hand, the firm has also a reservation or minimum

ask price below which it will not launch in the country. The ask price depends negatively


on the country size and positively on the country per capita income. This means that as

long as the country size is large, the regulator will have greater bargaining power to

negotiate prices with the pharmaceutical industry. By contrast, a high GDP per capita is

related to a high willingness to pay for a drug. Also, the ask price depends positively on

the country propensity for spillovers due to the use of ERP policy and parallel exports.

This means that whether the country is a potential referenced country (see 2.2. below) or

a potential parallel exporter (see 2.3 below), the firm will increase its reservation price.

The bargaining results in the launch of the product if the country’s maximum offer price

equals or exceeds the firm’s minimum ask price. If this condition is not met the delay

occurs, moreover, the greater this difference, the longer the delay in launch. Danzon and

Epstein (Danzon and Epstein, 2008) work under the same hypothesis as Danzon et al.

(Danzon et al., 2005). In this case, they also contemplate the bargaining results in price

and add more variables as explanatory factors of pricing and launching. One of them is

the regulatory regime, which can be an internal reference price (RP) or ERP; both are

expected to positively affect prices because substitute prices, either at home or abroad,

are expected to increasing prices (see 2.2. and 2.6. below). The other variable added, the

firm’s location, is measured by the fixed costs, which are expected to be lower if the

launching firm is located in the country analysed (see 2.5. below). Now, the bargaining

results in a price agreed within the range between the offer price of the country and ask

price of the firm. Then the launch is likely to occur when the offer price of the country

equals or exceeds the ask price of the firm. The authors underline that the trade-offs

between price and delay are expected to differ across markets and across products within

markets.

1.2.2 How the ERP is affecting the bargaining results in pricing and launching?

Both Danzon et al. and Danzon and Epstein consider that the propensity of being

a reference country may positively affect the price of a drug (Danzon et al., 2005, Danzon

and Epstein, 2008). This argument is supported by the extended use of ERP by countries

as a cost-containment policy (see Leopold et al. (Leopold et al., 2012)). Basically, ERP

consists of setting a price cap for pharmaceuticals, based on prices of identical or

comparable products in other countries. Despite the method of calculation (see Leopold et

al., Richter, Stargardt and Schreyögg (Leopold et al., 2012, Richter, 2008, Stargardt and

Schreyögg, 2006), we think that, whether a country is taken as reference by other



11

countries when pricing medicines, it is reasonable to consider that firm’s incentive will be

to set high prices in the reference country. As discussed by Richter (Richter, 2008), the

firm is better off launching its drug in high-price countries first, to influence prices in other

countries to its advantage. In this sense, García-Mariñoso et al. (Garcia Mariñoso et al.,

2011) directly analyse the effects of an ERP policy on a referencing country on the

negotiation in this country and, furthermore, the incentive of the referencing country to

apply ERP. To go into this idea, the authors consider one pharmaceutical firm, one on-

patent drug and two countries operating a positive list of reimbursed pharmaceuticals,

where patients pay a fixed and exogenous co-payment. Countries differ in the population

size and the level of co-payment. A model of negotiation process as a Nash bargaining

game is designed through which the authors compare independent price negotiations to

the situation in which one country (the referencing country) engages in ERP. Two different

scenarios are analysed, under “weak threats”12 if the drug is not reimbursed, or under

“tough threats”13 if the drug is banned. In the case of ERP with weak threats, when the

referencing country engages in ERP, the price negotiated in the referenced country

increases. The total surplus generated by the negotiation between the referenced country

and the firm increases14. This shows that the implicit negotiation power of the firm is

higher when the referencing country engages in ERP as compared with independent

negotiations. As Danzon et al. and Danzon and Epstein had suspected (Danzon and

Epstein, 2008, Danzon et al., 2005), García-Mariñoso et al. (Garcia Mariñoso et al., 2011)

show the fact that the referencing country engaged in ERP policy harms the referenced

country in terms of high price and lower outputs. The same authors also examine the

incentives to apply ERP policy rather than independent negotiations. Under the same

hypothesis stated above, they state that a country has an incentive to engage in ERP if its

co-payment levels are high when compared with the referenced country. This preference

decreases as the size of the referencing country increases, and as co-payments of both

countries converge. First, the referencing country size increases the ERP negotiated

12 Under “weak threats”, if the negotiations fail, the firm can still sell the drug at any price of its choice, but with no subsidy.This assumption is motivated by the fact that, in Europe, price-negotiating agencies have a minor role in the authorization of drugs.

13 Some countries outside Europe, such as Brazil or Canada, are known to threaten the firms with not authorizing drug sales if negotiations fail or if the firm does not accept ERP.

14 In this model, ERP is based on the price of a single reference country. However, results are highly sensitive to the modalities of the ERP.


between the referenced country and firm in two ways. The pie to be shared between both

parties is larger, and the firm has a stronger disagreement payoff while the disagreement

payoff of the referenced country remains the same. Second, as the negotiated price is

shown to be increasing with the patients’ co-payment (see (Jelovac, 2008)), if the co-

payment in the referenced country decreases with respect to the referencing country then,

the ERP will decrease and therefore the difference between the price independently

negotiated and the ERP will decrease. Then, they conclude that only small countries

should be observed to engage in ERP and/or ERP should be based on prices in large

countries (or a large group of countries). The analysis yields an analogous prediction if

one substitutes “large country” by ‘small co-payment country’ and vice versa.

The authors further extend their analysis to account for competition between the

firm’s pharmaceutical product and a therapeutic substitute that is already present on the

market in both countries. This extension adds realism, particularly, it makes the weak

threats scenario compatible with the observation that, in most European markets, being

excluded from the public funding may be almost as bad as being banned, as sales out of

the positive list of reimbursed drugs are negligible if subsidized therapeutic substitutes are

available. Now, two drugs, 1 and 2, with similar therapeutic indications are considered.

Each drug is produced by a different firm (firm 1 and firm 2). Both drugs are off patent and

one is the generic substitute of the other15. The consumer perceives them to be different

but face the same co-payment, although this co-payment may differ among countries. The

drug 2 is already listed in both the referenced and the referencing markets. The two drugs

are horizontally differentiated á la Hotelling (see (Hotelling, 1929)). As the independent

price negotiations lead to a higher price in the referencing country, the firm will reject low

prices in the referenced countries knowing that they will face a price cap in the

referencing one. Main results continue to hold in this extension: ERP benefits the

referencing country and harms the referenced country as well as the firm.

On the other hand, under tough threats (see footnote 13), the firm suffers a

harsher punishment in the case that negotiations fail (drug is banned). The main result

with weak threats remains, i.e., ERP benefits the referencing country and harms the firm,

15 Lobo and Feldman have also modelled the role of trademarks, advertising and generic names on competition FELDMAN, R. & LOBO, F. 2013. Competition in prescription drug markets: the roles of trademarks, advertising, and generic names. European Journal of Health Economics, 14, 667-675, LOBO, F. & FELDMAN, R. 2013. Generic Drug Names and Social Welfare. Journal of Health Politics Policy and Law, 38, 573-597..



13

but the referenced country is not affected by the ERP.

1.2.3 Which role does PT play in the pharmaceutical market?

As mentioned earlier in the section above, the firm prefers firstly launching its drug

in high-price countries to influence prices in other countries to its advantage. This

strategic behaviour may be not so useful if PT exists. As Danzon et al. and Danzon and

Epstein commented, to be a parallel exporter country might influence positively on prices

(Danzon et al., 2005, Danzon and Epstein, 2008). The reason has been clearly explained

by Richter (Richter, 2008). We note that Richter considers that PT implies a loss of

income for the firm, since it stops selling a certain amount at a higher price than in the

absence of PT, the author includes PT as variable into objective function firm. As

expected, in order to compensate this loss of income, the firm will have to increase the

price. The author proposes a mathematical optimization problem for the firm, considering

that the firm sells a drug across all countries and over all time periods and the lowest

price country will be the parallel exporter one. This is quantified as the sum of the

differences between the lowest price among all countries and the price stated in each

country, multiplied by the quantity lost16 in the parallel importer country.

Ganslandt and Maskus (Ganslandt and Maskus, 2004) go further and, not only

consider PT as potential loss of income for the original manufacturer, but also investigate

how PT firms behave and how the presence of PT affects equilibrium prices in the parallel

importer (PI) and exporter countries. They develop a simple model of parallel imports in

which an original manufacturer competes in its home market (Sweden) with PI firms and

all firms set prices simultaneously. The authors suppose that the quantity to trade is

exogenously given, and it is all sold. This idea is supported by the fact that in high-price

markets the PI quantity rarely exceeds 10%, except in a few major products. A model of

two countries is considered, with a high-income and a low-income country. The high-

income country is unregulated, the low-income country has a price cap set by the

government and the drug sold is an on-patent drug without substitutes. A marginal and a

fixed cost of engaging in PT also exist. Given the quantity chosen by PI firms, the profit-

maximizing price is calculated. The authors compare a model under a limited PI quantity

16 The quantity lost is calculated multiplying the market share lost by the quantity sold in the parallel importer country.


to another model that allows unlimited PI quantity. Under an unlimited quantity of PI, an

“arbitrage-free” price arises, and consequently a price convergence from the high-income

country to the low-income one. However, under a limited quantity of PI, the most real case

as stated above, the manufacturing firm has an incentive to adapt to PT rather than to set

an arbitrage-free price in its market. In this case, the equilibrium price of high-income

country converges to the low-income one plus a variable trade cost17. As expected, there

is an effect of competition, whereby the equilibrium price in the high-income country falls

in the number of PI firms. Also, the authors identify that the equilibrium number of PI firms

increases in the size of the market but decreases in the low-income country price and in

the fixed and variable trade cost.

1.2.4 How asymmetric information on quality of drugs may affect drug pricing and launching?

One of the factors influencing pricing and launching considered by Danzon et al.

and Danzon and Epstein, has been the ICER that affects positively drug prices (Danzon

et al., 2005, Danzon and Epstein, 2008). This measure may be interpreted as a

price/quality indicator of the drug. In the literature, under different hypotheses, we can

observe that information about quality matters. Two papers have considered the

asymmetric information about either the quality drugs or the demand of quality drugs, to

analyse the pricing and launching drugs (Atella et al., 2012, García-Mariñoso and Olivella,

2012). Atella et al. (Atella et al., 2012) propose a model of asymmetric information on the

quality of the drugs, to find out how two types of regulatory regimes, one focused on

quality and another on price control, affect drug prices, and furthermore, how the price

regulation affects, ultimately, the quality of the drugs. On the other hand, García-Mariñoso

and Olivella (García-Mariñoso and Olivella, 2012) propose a sequential launching and

analyse how the informational spillovers, issued from the asymmetric information about

the quality drugs, affects the drug pricing. The informational spillovers are defined as the

claim of lower prices by one country, generated by the knowledge about lower prices in

other countries. The notion that low prices may overspill to other countries even in the

absence of PT or ERP is here introduced, different from the previous research.

Atella et al. (Atella et al., 2012) compare two regulatory regimes. Under the first

17 The price in the low-income country is taken as given.



15

regime the government fixes a minimum efficacy standard (MES), this regime

corresponds to the regulatory structure of the pharmaceutical market in the United States.

Under the second regime, in addition to a MES, the government fixes a price cap (PC);

this regime corresponds to the structure in many other countries as Italy. The model

considers two countries that differ in their demand for drug efficacy, and one firm, which

produces two types of drug, low and high-efficacy, but it cannot distinguish between high

and low-type buyers. Regarding the paper of Atella et al. (Atella et al., 2012), this paper

introduces patient co-payments, the discount’s factor on profits and on quality from the

country.

Whether the government fixes a MES that binds the low efficacy drug, then the

efficacy and the price of the low efficacy drug increases to meet the MES and to cover the

drug marginal cost. Instead, the efficacy of the high-efficacy drug is not affected and its

price may be lower. However, if the regulation imposes too high efficacy standards, low-

type buyers would be excluded form the market. There exists a minimum efficacy

threshold that optimally balances the higher R&D costs with the higher efficacy drugs

delivered to low-type buyers. This optimal level is just below the level that excludes low-

type buyer from the market.

Whether a PC that binds on the high-type drug is considered (but not on the low-

type drug), the firms respond producing high-type drugs lower in quality at a price

correspondingly lower. Under both regimes, the efficacy and the price of the low efficacy

drug increases to meet the MES and to cover marginal cost respectively, however, the

efficacy of the high-type drugs may be undermined if the government fixes a PC that

binds the high-type drugs, but also that the consumer of high-type drugs will save money.

Finally, the net welfare will depend upon how binding is the price control and on the

relative size of the two groups of buyers.

It has been shown above how different types of regulation may affect on drug

quality and pricing under asymmetric information about the buyers. Garcia-Mariñoso and

Olivella (García-Mariñoso and Olivella, 2012) assume the asymmetric information on the

other side. The countries (i.e. the buyers) do not know about the type of the firm, thus, the

firm may be of high or low quality, and this information is in the hands of the firm. As the

authors propose a sequential launching, the countries will have their prior beliefs and


there may exist informational spillovers. Thus, whether a low price is fixed in the country

where the drug is first launched, this reveals private information (the quality of the firm

concerning the production and distribution costs) to subsequent players concerning the

price, and therefore, the following countries will also demand for low prices. Now, low

prices may overspill to other countries even in the absence of PT or ERP. The reason is

that countries that would in principle make generous price offers whether observe the firm

accepting a low price elsewhere, they might change their mind and become aggressive.

Along a dynamic game, now it is the firm, which accepts or rejects the offer from the

country thus, the game is based on a “take it or leave it offer”. Countries may be

aggressive or non-aggressive18.

According to the firm strategic behaviour, although information spillovers can be

avoided by launching in all countries simultaneously, the firm will prefer to delay if (from

more to less expected) (i) the firm is sufficiently patient (high discount’s factor19); (ii) the

aggressive country has a sufficient population; (iii) co-payments differ enough across

countries; and (iv) countries have relatively pessimistic priors on quality. Interestingly, the

authors present a counterargument to the statement that delay only occurs in small

countries, thus it could happen that the country that suffers delay is the largest in size (as

long as the rest of the factors mentioned go in the right direction) (García-Mariñoso and

Olivella, 2012).

1.2.5 Are important the headquarters location and the contacts among firms when pricing drugs?

Different from other studies, Cabrales and Jiménez-Martín (Cabrales and

Jimenez-Martin, 2007) consider that the firm is located in the countries analysed. There

are two countries, one of them regulates prices under ERP (the referencing country) and

the other does not (the referenced country). One of the main contributions is that the firm

profits are now maximized together with the consumer surplus by the regulator. The

authors compare the maximizing price in two situations, when headquarters are located in

18 The aggressiveness will positively depend on the co-payment; the larger is the co-payment, the more aggressive the price offers will be. We already mentioned that if a country has observed a low price acceptation in a previous country, it will update its beliefs and become aggressive. The third factor is dynamic and forward-looking. Being aggressive today may lead the firm to reject the price offer in order to avoid the aggressiveness of future agencies.

19 The discount factor is the factor by which a future cash flow must be multiplied in order to obtain the present value. The higher the discount factor is, the greater the present value is assessed.



17

the regulated country or in the unregulated one. This theoretical model predicts that the

price set by the regulator is slightly higher for the local multinational than for the foreign

one, and as the size of the referenced country grows with respect to the referencing

country, the price of the foreign multinational converges to the local multinational one. In

this regard, if we assume that the countries that are stricter regulators were relatively

small in size; this would imply that they could not influence substantially the prices in

favour of the local multinationals.

Not only the location of firms matters, but also the contacts between firms

competing in the same markets. The multimarket contact theory implies that more

contacts between firms competing in the same markets may induce more collusion. This

collusion support prices above the equilibrium prices. At this regard, Coronado et al.

(Coronado et al., 2007) try to predict the effect of multimarket contact structure on the

equilibrium prices under two different regimes, the price regulation and the free pricing,

and ultimately to know if price regulation may affect multimarket contact effect. For this,

the authors propose a game infinitely repeated where prices are set simultaneously. The

firms can collude and support prices above the equilibrium prices. In case of deviation,

the firms will be penalized reverting to the equilibrium prices. It is also supposed that the

maximum sustainable price (in collusion) depends positively on the discount factor, i.e.,

the future profits are more valuable, and therefore the short run benefits from deviation

are accordingly less preferred. Taking into account the hypothesis above described, the

model predicts that the effect of multimarket contact structure increases the equilibrium

prices but this effect is undermined in regulated countries.

In summary, the theoretical models predict that both the firm location and the

multimarket contact not only affect drug pricing but also, their effects depends on price

regulation regimes.

1.2.6 Which effects do arise in pricing and innovation when countries apply internal RP20?

Danzon and Epstein (Danzon and Epstein, 2008), extending the paper of Danzon

20 Internal reference pricing, as opposed to the ERP, compares product prices within a single country.


et al. (Danzon et al., 2005), they consider as influencing factor the RP as regulatory

regime being expected to positively affect on prices. At this regard, two papers examine,

on the one hand, how the RP policy affects the equilibrium prices and the firm pricing

strategies (Miraldo, 2009), on the other hand, how the RP policy influences the intensity of

research and the introduction of new pioneer in the market (Bardey et al., 2010). Both

papers compare the outputs under no regulation and under RP policy. Furthermore, other

authors have deeply studied the RP policy from an international perspective (Lopez-

Casasnovas and Puig-Junoy, 2000, Puig-Junoy, 2010a, Puig-Junoy, 2010b) and

particularly the Spanish case (Mestre-Ferrandiz, 2003b, Moreno-Torres et al., 2009, Puig-

Junoy, 2007)

The model developed by Miraldo (Miraldo, 2009) considers two pharmaceutical

firms, a continuum of consumers uniformly distributed and a market of drugs horizontally

differentiated à la Hotelling (Hotelling, 1929). Each firm produces two distinct variety of

drug. Each consumer is assumed to have a most preferred drug that is given by her

location on the line segment. Indeed, the constant marginal cost of distance is the loss in

utility incurred by a consumer.

Miraldo studies the explicit RP formulations and considers a different timing of

implementation of the policy. The author analyses a finite dynamic game, in which

duopolists compete by non-cooperatively setting prices in two subsequent periods. In the

first stage, the two pharmaceutical firms set the prices for one variety. At the beginning of

the second, the government fixes the RP level, and then, the firms set prices for the

second variety of drugs. Therefore, at the last stage, the firms’ profits depend via demand

not only on the pricing strategies but also on the RP level fixed previously by the

government.

The RP policy is introduced as reimbursement scheme21. Miraldo shows that

under the RP policy, the equilibrium prices are at least as high as the equilibrium prices

without RP. As a main contribution, when both RP rules are compared, the minimum and

the weighted average, the author states that firm set higher prices at first stage when the

21 In countries, such as Germany and Spain, where pharmaceuticals are reimbursed through a RP system, patients are typically reimbursed a lump sum amount for any homogeneous pharmaceutical cluster, independently of the drug variety bought. There are several criteria to cluster drugs and the replicated model applies to countries that use chemical and therapeutic criteria. The first criterion clusters drugs with the same active ingredient and therefore refers to patent expired drugs. The second criterion clusters drugs that have the same therapeutic function and therefore, within the same cluster one can find patent protected drugs.



19

RP is calculated as a weighted average than when the minimum policy is applied. The

firm pricing strategy in the second period (when already the RP level has been fixed) will

depend on the weights of each price. If they are high (i.e., for sufficiently high and low

values of the weights), the minimum policy makes firms to fix lower prices than the

weighted average rule. But if the weights are similar, results will be ambiguous; they will

depend on consumers’ preferences, on the degree of horizontal differentiation and on the

discount factor. Anyway, in a symmetric market, in order to avoid higher prices, the

regulator should implement a policy where the reference pricing consists of the minimum

observed price.

In turn, Bardey et al. (Bardey et al., 2010) evaluate the long run impact of RP on

pharmaceutical innovation and on health expenditure. The paper is based on a dynamic

model with three players: the firms (innovators/producers), the regulator and the

consumers22. Both horizontal and vertical differentiations are considered. Vertical

differentiation has different levels (therapeutic classes). To simplify, there are two levels,

C and N, designing respectively current and new. Patients obtain utility from the treatment

of the drug, perceive its side effects, pay a price for the drug and receive a

reimbursement. They also assume that drugs are produced at zero cost and when a drug

is introduced in the level N, producers of level C have no sales. The price negotiations are

developed à la Rubinstein (see Rubinstein (Rubinstein, 1982)). The firms choose the level

of research investment, and then negotiate introductory prices for new drugs with the

regulator. The innovation process is deterministic and can discover a new product either

in the same level as existing products (horizontal innovation or follower), or in a superior

level (vertical innovation or pioneer). There exist a cost of bringing an innovation.

Bardey et al. authors compare how the dynamic of innovation behaves without RP

and under RP. Thus, in terms of delay of introduction, the application of RP yields the

delay of followers, and the delay of pioneers if only if the price is above some threshold.

In the long run, allowing innovation to occur in level C (prior to the discovery of the first

level N drug), the follower may be never introduced (short sequence), or it can be

introduced before the pioneer of level C (long sequence). Again, both sequences are

22 The relation between the patient and the physician is considered a relation of perfect agency; therefore they are viewed as a single agent, the consumer.


compared without RP and under RP. In this case, there exist a threshold under which, an

equilibrium with a short sequence exists, and above which, an equilibrium with a long

sequence so does. The introduction of RP makes all prices be smaller and makes this

threshold increase.

When RP is applied, there may be countervailing effects on the introductory time

of the pioneer. On the one hand, the profitability of a pioneer is reduced because its price

decreases when the follower is introduced. Instead, the delay of introduction of the

follower increases, allowing the producer of a pioneer to benefit from a longer period of

monopoly situation. The global effect of RP is thus ambiguous. Two antagonist effects of

the RP regulation are found: a decrease in price reduces the incentives to create pioneer

drugs; but however, the introduction of followers is delayed, which gives positive

incentives to launch pioneers. Consequently, the net effect within a class is ambiguous.

1.3 What do the empirical models show?

Theoretical models and frameworks explain the strategic games played by

governments, public insurers and pharmaceutical companies, and undoubtedly they also

suggest factors influencing both launch prices and launch of new drugs. This section

shows empirical evidence on these factors.

The articles retrieved are first classified according to the two outcome variables

evaluated, price and launch; and secondly, the impact factors influencing each outcome

variable are explored and classified into drug, competition, regulation, country and firm

characteristics.

1.3.1 Samples, Variables and Methods

Table A.1 and Table A.2 summarize the main characteristics of the 17 studies that

met the selection searching criteria. The papers retrieved examine the determinants of

pricing or launching new drugs focusing on drug, brand competition, regulation, country or

firm effects (Borrell, 2007, Cabrales and Jimenez-Martin, 2007, Coronado et al., 2007,

Danzon and Epstein, 2008, Danzon et al., 2011, Danzon et al., 2005, Heuer et al., 2007,

Kanavos and Vandoros, 2011, Kyle, 2006, Kyle, 2007, Lanjouw, 2005, Kanavos P and

Costa-Font J, 2005, Verniers et al., 2011, Danzon and Chao, 2000, Timur et al., 2011,

Atella et al., 2012). Two studies analyse both pricing and launching (Danzon and Epstein,



21

2008, Verniers et al., 2011), but only one considers pricing and launching simultaneously

(Verniers et al., 2011).

Five articles have developed a theoretical model as a basis for empirical analysis

(Cabrales and Jimenez-Martin, 2007, Coronado et al., 2007, Danzon and Epstein, 2008,

Danzon et al., 2005, Atella et al., 2012). Empirical papers have developed multivariate

causal models, controlling for variables mainly classified according to drug, competition,

regulation, country and firm characteristics.

Regarding pricing, most papers explain the price level within a given period of time

(Borrell, 2007, Cabrales and Jimenez-Martin, 2007, Coronado et al., 2007, Danzon et al.,

2011, Kanavos and Vandoros, 2011, Kanavos P and Costa-Font J, 2005, Danzon and

Chao, 2000, Timur et al., 2011, Atella et al., 2012), although two take both prices at

launch time as dependent variable (Danzon and Epstein, 2008, Verniers et al., 2011).

Prices are generally taken at ex-manufacturer level (Cabrales and Jimenez-Martin, 2007,

Coronado et al., 2007, Danzon and Epstein, 2008, Danzon et al., 2011, Kanavos and

Vandoros, 2011, Verniers et al., 2011, Danzon and Chao, 2000) but are also taken at

wholesale (Borrell, 2007, Kanavos P and Costa-Font J, 2005) and retail (Kanavos and

Vandoros, 2011, Timur et al., 2011) levels. In every article, log transformations of the

price are used. The articles that develop a theoretical model as a basis for empirical

analysis estimate a reduced-form equation of the theoretical form (Cabrales and Jimenez-

Martin, 2007, Coronado et al., 2007, Danzon and Epstein, 2008, Danzon et al., 2005).

Different estimation methods are employed depending on the nature of the data, with

ordinary least squares (OLS) being most commonly employed (Borrell, 2007, Danzon and

Epstein, 2008, Danzon et al., 2011, Ganslandt and Maskus, 2004, Kanavos P and Costa-

Font J, 2005, Danzon and Chao, 2000, Atella et al., 2012), while Danzon and Epstein use

the generalized least squares (GLS) (Danzon and Epstein, 2008) and Kanavos and

Costa-Font apply two step least squares (2SLS) and two step generalized least squares

(2SGLS) (Kanavos P and Costa-Font J, 2005). Other estimating methods applied include

differences-in-differences regressions (Danzon et al., 2011), a Heckman selection model

with a first-stage clog-log regression (Danzon and Epstein, 2008) and Tobit regressions

(Kanavos P and Costa-Font J, 2005, Verniers et al., 2011).

Launching studies have analysed how quickly a drug is made available in a given


country, using diverse measures to evaluate the launch of a medicine into a market.

These include logit and probit models, measuring the probability of a new drug being

launched in a given country within a certain period23 after the first appearance (Heuer et

al., 2007, Lanjouw, 2005), while other papers have analysed this question using discrete-

time hazard models, examining the probability of placing a new drug on the market in time

t when it has failed to enter by month t-1 after global launch (Danzon and Epstein, 2008,

Danzon et al., 2005, Kyle, 2007, Lanjouw, 2005, Kyle, 2006). This probability will be

called henceforth, launch hazard. Only Verniers et al. studies as the dependent variable

the launch window, i.e., the difference, in months, between the first worldwide launch and

the subsequent launch in a specific country (Verniers et al., 2011). Kyle estimates a

negative-binomial model to estimate the number of countries where the drug has been

launched (Kyle, 2007).

Samples vary widely across papers. In the main, three types of data are used:

data (collected yearly, monthly or quarterly) at product level (Borrell, 2007, Cabrales and

Jimenez-Martin, 2007, Coronado et al., 2007, Danzon et al., 2011, Kanavos P and Costa-

Font J, 2005, Verniers et al., 2011, Atella et al., 2012), molecule level (Danzon and

Epstein, 2008, Danzon et al., 2005, Heuer et al., 2007, Kyle, 2006, Kyle, 2007, Lanjouw,

2005, Timur et al., 2011) or both (Kanavos and Vandoros, 2011, Danzon and Chao,

2000), including: i) Sales series by volume and value (euros/dollars); ii) Launch dates; iii)

Aggregated data, including medicine, market, country and firm characteristics. In all the

empirical articles, the impacts of these characteristics on prices and launches are

evaluated through econometric models. The numbers of molecules/products, the

countries selected and the period of time analysed vary considerably. The data sources

used are in most cases IMS databases, occasionally complemented with other

databases. These differences in samples are presented in detail in Table A.2.

Different specifications of competition and substitutes were found, with articles

using therapeutic substitutes defined either at the molecule (Danzon and Chao, 2000,

Danzon and Epstein, 2008, Timur et al., 2011) or the product level (Danzon et al., 2011,

Kyle, 2006, Verniers et al., 2011). Most papers define the therapeutic class according to

the Anatomical Therapeutic Chemical (ATC) Classification System at different levels (1, 3

and 4) (WHO, 2012). Danzon and Epstein simply distinguish between “superior” and

23 Eight months, two and ten years.



23

“inferior” subclasses (Danzon and Epstein, 2008) (see also (Berndt et al., 2007)).

Many empirical papers introduce the price regulation level of countries as an

explanatory variable when studying medicines pricing or launching. Both the criteria and

the definitions used to classify the price regulation level may vary slightly among papers,

ranging from a simple differentiation between whether countries regulate or not (Heuer et

al., 2007, Kyle, 2007, Kanavos P and Costa-Font J, 2005, Verniers et al., 2011) to

whether extensive regulation is applied, as in Lanjouw (Lanjouw, 2005), to other types of

classification which divide the degrees of regulation into two (less regulated and more

regulated) as Coronado et al. (Coronado et al., 2007) or three (low, medium and high) as

Cabrales and Jiménez-Martín (Cabrales and Jimenez-Martin, 2007).

Apart from the country fixed effects, country characteristics are mainly related to

variables based on demographics (Heuer et al., 2007, Kyle, 2006, Kyle, 2007, Lanjouw,

2005, Verniers et al., 2011) and income (Borrell, 2007, Cabrales and Jimenez-Martin,

2007, Danzon et al., 2011, Heuer et al., 2007, Kyle, 2006, Kyle, 2007, Lanjouw, 2005,

Verniers et al., 2011). The papers selected mainly use two demographic variables, the

population size and the population structure (Kyle, 2007, Lanjouw, 2005) while income

measures are defined as GDP or GDP per capita (Borrell, 2007, Cabrales and Jimenez-

Martin, 2007, Danzon et al., 2011, Heuer et al., 2007, Kyle, 2006, Kyle, 2007, Lanjouw,

2005, Verniers et al., 2011). Together with income level, income distribution is also

included as a country characteristic (Borrell, 2007, Danzon et al., 2011, Lanjouw, 2005).

The amount of R&D expenditure has been measured as a percentage of GDP (Lanjouw,

2005). Membership of the EMA24 (European Medical Agency) has also been included in

the econometric analysis and it is considered in the empirical contribution of this thesis

(see chapter 3) as a country characteristic (Verniers et al., 2011, Lanjouw, 2005).

Firm characteristics can also influence drug prices and market entry. In the papers

retrieved, these features at company level mainly concern the firm’s size (Cabrales and

Jimenez-Martin, 2007, Coronado et al., 2007, Danzon et al., 2005), experience (Danzon

et al., 2005, Kyle, 2006, Kyle, 2007), type (Cabrales and Jimenez-Martin, 2007) and

location (Danzon and Epstein, 2008, Kyle, 2006, Kyle, 2007, Verniers et al., 2011,

24 Before EMEA.


Cabrales and Jimenez-Martin, 2007, Coronado et al., 2007, Danzon et al., 2005). Firm

size is constructed as the total corporation sales corrected by excluding sales of the

product under analysis (Cabrales and Jimenez-Martin, 2007, Coronado et al., 2007,

Danzon et al., 2005). Experience was measured through the firm’s worldwide outpatient

sales at the beginning of the study period (Danzon et al., 2005), the number of years it

has been active in the country (Kyle, 2006) and the number of countries in which it has

launched any medicine (Kyle, 2007). The firm type variable is strongly related to firm

location. The simplest criterion considers whether or not the headquarter is located in the

country (Danzon and Epstein, 2008, Danzon et al., 2005, Kyle, 2006, Kyle, 2007, Verniers

et al., 2011); other papers distinguish among local non-multinational, local multinational

and international multinational firms (Cabrales and Jimenez-Martin, 2007) or, among local

originator, solo licensee and local co-marketer firms (Danzon and Epstein, 2008).

1.3.2 Factors influencing prices

1.3.2.1 Drug Characteristics

The longer a product has been in the market, the lower its price tends to be; thus,

new medicines enjoy price premiums compared to those already in the market. At the

same time, this is consistent with the hypothesis that newer molecules offer improved

therapeutic quality25, which is associated with high prices (Cabrales and Jimenez-Martin,

2007, Coronado et al., 2007, Danzon and Chao, 2000, Kanavos and Vandoros, 2011,

Timur et al., 2011, Atella et al., 2012). The same is true with respect to the strength of the

medicine and the number of different presentations (Danzon and Chao, 2000, Danzon

and Epstein, 2008, Timur et al., 2011), which are associated with higher prices. This is

coherent with the generally accepted positive relationship between strength and

therapeutic value, when a medicine contains more quantity of active ingredient and when

adding perceived value in health, it is generally accept to pay more. Contrary, price

decreases with pack size (Danzon and Chao, 2000, Timur et al., 2011) and it is also in

line with Verniers et al. (Verniers et al., 2011), who finds that the higher the required daily

dosage of a medicine, the lower the price per gram of this medicine. These results are

consistent with economies of scale in packaging.

25 There exist differences among therapeutic definitions.



25

Whether we look at abroad, a higher mean global price of the medicine positively

affects the drug price, and moreover, the more countries in which the molecule is present,

the greater this positive effect is. Note, however, that when there is no average global

price26, i.e. when the drug is very innovative, the price is higher than when a global

reference exists (Cabrales and Jimenez-Martin, 2007). Furthermore, Verniers et al. find

an inverted U-shaped effect of the launch window on the launch price in which launch

price is highest at moderate delay. Considering on-patent products, the authors explain

that, at moderate launch windows, the firm can still make money under patent protection.

At very long delay, regulator and firm will agree more easily a relatively low launch price

as a prelude to generic competition. However, at very short delay, the firm will accept a

lower launch price more easily enjoying the full life of patent protection in order to

recuperate the R&D investment. This hypothesis about low prices at very short delay may

face the widely used ERP, the presence of PT and the spillover effects.(Verniers et al.,

2011).

Some authors have highlighted differences in therapeutic categories (Danzon and

Chao, 2000, Verniers et al., 2011). Moreover, the significant country-therapeutic category

interaction implies that therapeutic category effects differ across countries, owing to such

factors as category-specific differences in medical norms, insurance, regulation and OTC

share (Danzon and Chao, 2000).

The results published by these studies suggest that, in general, different degrees

of regulation do not distort the effects of the above attributes. However, Danzon et al.

(Danzon and Chao, 2000) have found that some effects may change depending on the

country regulation level. We mention here some examples. Regarding therapeutic

innovation, there is a tighter correlation between medicine prices and quality in the USA

than in Italy, where price controls are applied. Furthermore, molecule age elasticity is

most negative (2.66 or greater) in France, Italy and Japan, which have the strictest price

regulation. The effect of drug strength is significantly negative in Japan, where the

practice of polypharmacy27 may reduce the value of strong pills. In turn, the price is

independent of the number of presentation forms in the USA and Canada, but for other

26 There is no mean global price because the medicine has been launched in a single country.

27 Prescribing several different drugs simultaneously.


countries, as mentioned above, the elasticity is significantly positive. This is consistent

with the hypothesis that introducing line extensions is one means of achieving a price

increase in countries that do not permit higher prices for established products. The price

elasticity with respect to the number of forms is largest in Japan, which presents the

greatest price reduction over the product life cycle and hence where there exist strong

incentives to introduce new forms and thus obtain a higher price (Danzon and Chao,

2000). In addition, Danzon and Chao show that the global diffusion28 as an indicator of

therapeutic value obtains higher prices in unregulated markets, although this effect is

insignificant or small at best in less regulated markets such as the UK, Canada and

Germany, but is significantly negative in strictly regulated countries such as France, Italy

and Japan (Danzon and Chao, 2000). Besides, Coronado et al. show that the product

market share is expected to be significantly positive only in the least regulated countries.

1.3.2.2 Competition and substitutes

In general, therapeutic substitutes do not appear to exert competitive pressure on

price. Danzon and Chao (Danzon and Chao, 2000) show that, competition from

therapeutic substitute molecules appears to have small significant negative effects in

France, Italy, Germany and the UK, but the interpretation is unclear. The most plausible

explanation is that regulators use implicit reference pricing, setting prices for new

products based on prices of established products (Danzon and Chao, 2000). Timur et al.

do not find significant effect on price from therapeutic substitutes, however, different from

Danzon and Chao (Danzon and Chao, 2000). Timur et al. estimate the pool of data and

do not show specific results for countries. In this case, Timur et al. suggest that substitute

molecules with a higher price might not receive reimbursement, and substitution is not

always possible in regulated countries because of prescribing or consumption

preferences. Furthermore, the effect of delayed entry for therapeutic substitutes has been

also analysed, and the coefficients obtained for the USA and Canada imply that

successive molecules enter at lower prices, however these lower prices are not

sufficiently low to fully erode the first mover’s advantage. Other country interactions are

generally positive, but significant only in Germany.

On the contrary, only Verniers et al. (Verniers et al., 2011) find that competition

28 The number of countries where the medicine has already been sold.



27

drives down launch prices when estimating the pool of data. In this case, this different

result can be supported by the use of a richer sample. Furthermore, Danzon and Epstein

collect data for superior and inferior medicines. They state that while the prices of

competitors are positively related to launch prices, the prices of inferior medicines do not

affect those of superior ones, and vice versa, which means that dynamic competition

between subclasses is based on non-price product attributes (Danzon and Epstein,

2008).

For middle and low income countries (MLICs), Lanjouw finds that competition from

other originator medicines does not appear to be effective at reducing prices in retail

channels in these countries (Lanjouw, 2005).

Most papers found generic competition to negatively affect prices (Cabrales and

Jimenez-Martin, 2007, Coronado et al., 2007, Danzon and Chao, 2000, Danzon and

Epstein, 2008, Danzon et al., 2011, Timur et al., 2011). However, Danzon an Epstein

(Danzon and Epstein, 2008) only find that the effect of generic prices in inferior class,

which may reflect a selection effect: late entrants in inferior subclasses launch only if they

expect to receive high prices relative to competing generics. In turn, Cabrales and

Jiménez-Martín, and Coronado et al. (Cabrales and Jimenez-Martin, 2007, Coronado et

al., 2007) find a significant and negative effect in a large majority of countries. However,

the presence of generics on a market does not mean that brand name products will

reduce their prices. According to Coronado et al., in some cases, the presence of

generics will have the impact of concentrating brand name products over the inelastic

portion of the demand, which will then increase the price of these products. Hence, the

expected sign of the number of generics will be positive29. Interestingly, Danzon and Chao

(Danzon and Chao, 2000) find that generic competition is significant in unregulated or

less regulated markets but regulation undermines generic competition in strict regulatory

systems. As Coronado et al. (Coronado et al., 2007), Danzon and Chao find positive

effects of generic competition but they explain that multi-source suppliers in these

countries are usually licensed co-marketers rather than competing generic manufacturers

or minor ‘‘new’’ products that enter to obtain a higher regulated price. Danzon et al.

(Danzon et al., 2011) find that in MLICs the number of generic competitors only weakly 29 The US, Germany, The Netherlands, the UK and France.


affects prices to retail pharmacies, may be because uncertain quality leads to competition

on brand rather than price. Contrary, tendered procurement attracts multi-national generic

suppliers and significantly reduces prices for originators and generics, compared to prices

to retail pharmacies.

Only Kanavos and Vandoros (Kanavos and Vandoros, 2011) find that generics is

non-significant. They point out the existence of the generics paradox, particularly in the

US, where prices of off-patent originator brands do not decline post-patent expiry, but,

rather, increase faster than prices of in-patent originator brands.

1.3.2.3 Regulation Characteristics

When examining price regulation regimes through explicit regulation variables, the

most of the literature does not find significant influences on prices (Kanavos P and Costa-

Font J, 2005, Verniers et al., 2011, Kanavos and Vandoros, 2011, Danzon and Epstein,

2008, Atella et al., 2012). Only Kanavos et al. find that countries with “free-pricing”

systems (the US and Germany) present higher prices significantly positive. As explicit

price regulation policy, they only find that the explicit use of HTA (Health Technology

Assessment) has a significant negative effect on prices (Kanavos and Vandoros, 2011).

In this sense, Atella et al. proposes that rather than higher or lower prices, under a price

control regime (as in Italy), there is greater price variability than in a free price regime (as

in the US) (Atella et al., 2012).

Although not introducing explicit regulation variables in the econometric models,

Danzon and Epstein and Atella et al. (Atella et al., 2012, Danzon and Epstein, 2008)

retrieve information from price regulation characteristics. In this sense, Atella et al. find a

positive relationship between quality and price in a free price regime (as in the US), but

also find a negative relationship between drug price and drug quality in a price control

regime (as in Italy), which suggests that price regulations have created perverse

incentives (Atella et al., 2012). Also, Danzon and Epstein interestingly observe that launch

prices increase with the level of the lowest price previously received in other high-price

EU countries, whereas the effects of a previous launch in low-price EU countries are

insignificant. This is also true for non-EU countries, but only for superior medicines. This

result is consistent with the hypothesis that launching first in high-price EU markets can

influence prices in low-price ones. That evidence about a sequential launch prices

validates the theory that a launch delay in low-price markets may ultimately yield higher



29

prices in these markets through spillovers from higher-price ones (Danzon and Epstein,

2008). This theory is shown by Stargardt and Schreyögg (Stargardt and Schreyögg,

2006). They develop an analytical model in their paper, which analyses direct and indirect

impact due to the use of ERP from the referenced to the referencing country. The authors

estimate the impact of drug price changes in Germany30 on drug prices in other countries

using ERP in the former EU-15. The authors use the formulas applied by each

referencing country and then, they calculate the partial differential of these formulas with

respect to a 1 euro price reduction in Germany. They do not only know the formula (the

average, the minimum, a percentage, etc. see Leopold (Leopold et al., 2012)) but also the

basket of countries used by each referencing country. They differentiate the direct impact

(caused by Germany to other countries) and the indirect impact (caused by the

referencing countries that have taken Germany in their baskets, to other referencing

countries). The authors state that the relationship between the direct and indirect impact

of a price change depends mainly on the scheme applied to set prices. For instance, the

price is either determined by the lowest of foreign prices (e.g. Portugal), the average of

foreign prices (e.g. Ireland) or a weighted average of foreign prices (e.g. Italy). If the

respective drug is marketed in all referenced countries and prices are regularly updated, a

price reduction of €1.00 in Germany will reduce prices in the former EU-15 countries from

€0.15 in Austria to €0.36 in Italy. Whether we distinguish between direct and indirect

impact, almost more than (about) the 50% of the total impact in Austria comes from the

indirect impact (0.08 euros), however, only 1% of the total impact in Italy is due to the

indirect impact (0.03 euros). Both Austria and Italy include 14 and 12 countries in their

ERP scheme respectively, but we observe how Italy, which uses a weighted average, is

less harmed by undue indirect impact. Thus, to avoid the negative effects of ERP and

determine prices in order to reduce the direct and indirect impact of individual countries, a

weighted formula of prices containing as many countries as possible should be used.

Surprisingly, the literature does not find a direct effect on pricing when ERP is

applied. This result is robust among different specifications, but it could be explained by

the database collected. In Verniers et al. (Verniers et al., 2011), the database was limited

to the drugs launched as of February 1994 but the ERP had not been widely applied by 30 Germany is chosen because is one of the largest pharmaceutical markets in the world, it is characterised by relatively high prices and it is referenced by most referencing countries scheme.


that time (Leopold et al., 2012). In Kanavos and Vandoros (Kanavos and Vandoros,

2011), the prices of branded originators do not correspond to launch prices; therefore,

competition can also lead to downward pressure of off-patent originator brands, but the

ERP is applied on new drugs at launch time. On the other hand, when the ERP is found to

have effect on launching, the database comprises prices at launch time between 1995

and 2005.

Even more, Kanavos and Costa-Font have found that PT does not influence prices

downwards in importing countries (Kanavos P and Costa-Font J, 2005). Furthermore,

Danzon and Epstein show that the presence of PT is insignificant for superior medicines,

but significant and negative for inferior ones, indicating that the presence of PT reduces

launch prices mainly for late entrants into older subclasses (Danzon and Epstein, 2008).

Furthermore, Danzon et al. make an interesting contribution for MLICs concerning

price regulation policies. They compare the two different ways to provide medicine in

MLICs: tendered procurement mechanism by NGOs and standard retail channels. In

these terms, they find that originator brands purchased through tendered procurement

mechanisms by NGOs tend to lower originator prices, compared to those obtained

through standard retail channels. These large procurement effects may reflect not only

price-competitive tendering but also a greater willingness of originators to grant discounts

to a separate distribution channel that targets lower income customers and is less prone

to price spillovers to other countries (Danzon et al., 2011).

1.3.2.4 Country Characteristics

In most studies, country characteristics are included in the econometric models in

order to capture price differences among countries. The variable most commonly included

is that of country GDP per capita (Cabrales and Jimenez-Martin, 2007, Danzon and

Epstein, 2008, Borrell, 2007, Danzon et al., 2011). This is coherent with the generally

accepted positive relationship between wealth and greater willingness to pay. Thus, the

country-level fixed effects controlled by GDP per capita show that although the lower-

income EU countries such as Spain, Portugal and Greece regulate medicine prices, they

have relatively high medicine prices with respect to GDP, whereas higher-income EU

countries have lower medicine prices relative to their per capita GDP. Then, Danzon and

Epstein suggest that ERP has contributed to the price convergence of medicines among

EU countries relative to GDP (Danzon and Epstein, 2008). In turn, Kanavos and



31

Vandoros (Kanavos and Vandoros, 2011) state that as we move towards newer

molecules over time by launch date, there is upward price convergence across the study

countries overall. This is partly explained by ERP and the launch sequence for new

products, whereby new products are first launched in less-regulated countries followed by

price-regulated countries. This launch sequence influences in part the final price in price-

regulated countries. According to the convergence above mentioned, Cabrales and

Jiménez-Martín (Cabrales and Jimenez-Martin, 2007) observe that the US does not

present higher prices than other countries. Contrary to the conventional wisdom about

countries regulatory regimes, the fixed effect of the US is significantly lower than that of

Canada (but not for all specifications), France or Italy. The authors interpret that if

average prices in the US are higher than in other countries it is not because other

countries engage in “free riding regulation”, but because the US per-capita income is

higher, in fact, in many cases, the US pays less, not more, than countries of similar

income or lower income (such as Eastern European countries). The authors also interpret

that as the US market size is larger and more competitive than the other countries, it

provides with some protection with respect to similarly rich countries. This contradictory

result may be result of the innovations introduced with respect to the previous empirical

literature in the subject. Mainly, they estimate pricing equations separately for each

country, they do not restrict the sample in any way and identify the effect of time-invariant

variables by following a two-stage procedure (Cabrales and Jimenez-Martin, 2007).

For MLICs, Borrell shows evidence of a persistent positive relationship between

drug prices and per capita income in MLICs (Borrell, 2007). In addition, the income

distribution within countries has also been considered by Danzon et al. and Borrell

(Borrell, 2007, Danzon et al., 2011). Borrell suggests that income effects alone are

unlikely to achieve affordable prices in low-income countries. Thus, although per capita

income effects are positive, the negative effect of the income distribution implies that the

poorest countries face with the highest relative prices. Moreover, skewed income

distributions appear to exacerbate high drug prices relative to per capita incomes in

MLICs (Danzon et al., 2011), however, Borrell reports ambiguous effects (Borrell, 2007).

1.3.2.5 Firm Characteristics

Some studies have examined how certain firm characteristics may influence prices


(Cabrales and Jimenez-Martin, 2007, Coronado et al., 2007, Danzon and Epstein, 2008,

Kanavos and Vandoros, 2011, Verniers et al., 2011). Coronado et al. show that the firm

size have a high significant positive effect on prices, indicating that large companies enjoy

higher prices either because its products are of higher quality or perceived as such

(Coronado et al., 2007). However, Cabrales and Jiménez-Martín find this effect significant

and negative, but small (Cabrales and Jimenez-Martin, 2007). As the definition variable

and the sample is the same, except the number of countries considered, we think that the

use of more variables concerning the firm characteristics may undermine this issue.

Another question widely mentioned concerns the type and location of the firm.

Danzon and Epstein do not find price premium for medicines launched by local firms

(Danzon and Epstein, 2008). In addition, Cabrales and Jiménez-Martín firstly check

differences in location for multinational firms between local and foreign multinationals, and

contrary to conventional wisdom, in most countries, is found that countries do not

distinguish between local and foreign multinationals. However, Verniers et al., different

from former results, find that firms obtain higher launch prices in their domestic market

than they do in foreign ones (Verniers et al., 2011). These different findings may be due to

differences in samples. Verniers et al. have collected a richer sample of countries where

we can find more MLICs where usually headquarters are not held, while the most of

countries collected by Cabrales and Jiménez-Martín and Danzon and Epstein are middle

and high-income countries (Cabrales and Jimenez-Martin, 2007, Danzon and Epstein,

2008). Additionally, Cabrales and Jiménez-Martín also compare between local non-

multinational firms and multinational firms and find that local non-multinational firms tend

to have lower prices than any multinational (Cabrales and Jimenez-Martin, 2007).

Furthermore, as in the theoretical models, Coronado et al. analyse the existence

of a multimarket contact effect and try to find whether more contacts between firms

competing in the same markets may induce more collusion. They empirically show that

multimarket contacts have a positive influence on prices for the firm in less regulated

countries31, and unstable effects in regulated countries32. This suggests that in more

regulated markets there exist distortions that interact with market forces. For instance, the

product market share is significant and positive only in the least regulated countries.

31 The US, Canada, Germany, the Netherlands and the UK.

32 France, Spain, Italy and Japan.



33

Moreover, reducing prices in more competitive markets compared to the existing level

may discourage entry and may have a negative dynamic effect in the development of the

industry. This may help predict the undesirable effects of public interventions (Coronado

et al., 2007).

1.3.3 Factors influencing launching

1.3.3.1 Drug Characteristics

As expected, potential prices and volumes positively affect launching (Danzon et

al., 2005), although according to Danzon and Esptein, the volume is not significant

(Danzon and Epstein, 2008). This insignificant effect of volume contrasts with significant

positive effects in Danzon et al. (Danzon et al., 2005). These different findings may reflect

differences in sample countries and drugs, in addition to the use of more detailed

measures of country-class prices and other characteristics. As previously found for prices,

the speed of launch increases with the medicine’s importance33 but falls with its age (Kyle,

2007). Furthermore, peculiarities have been found depending on the type of medicine

evaluated. Thus, there are significant differences among therapeutic classes (Danzon et

al., 2005, Verniers et al., 2011), for instance, there is a higher probability of never

launching for inferior medicines than for superior ones (Danzon and Epstein, 2008).

Interestingly, Verniers et al. observe an U-shaped effect of launch price on the

launch delay; launch price is highest at moderate launch delay. As expected, at very long

launch delay, we will expect a relatively low launch price as a prelude to generic

competition. This relationship shows the trade-off for the pharmaceutical firm between the

price and the launch delay (Verniers et al., 2011).

Regarding the international context, from the first global launch the launch hazard

pattern is first decreasing and then increasing34. This idea is generally accepted; there is

a threshold at which the firm will not be worried about the spillover effects from the ERP

and the presence of PT. On the other hand, the number of countries where the medicine

has already been launched affects significant and positively, with the exception of prior 33 Drug’s share of Medline citation for therapeutic class.

34 Quadratic effect.


launch in the three lowest price EU countries, Spain, Portugal and Greece. This pattern

confirms that firms delay the launch in low-price EU countries until it has taken place in

higher-price ones (although this is not the case for inferior medicines). Furthermore, a

prior launch in a high-price country has a stronger effect on launching in a low-price one

than vice versa. This effect is even larger when the countries are both EU members

(Danzon and Epstein, 2008).

1.3.3.2 Competition and substitutes

With respect to market factors, the launch hazard is positively affected by

competitor prices (Danzon and Epstein, 2008, Kyle, 2006, Kyle, 2007), although it does

not seem to influence when the endogenous variable is studied as launch window instead

of launch hazard (Verniers et al., 2011). These differences may come from the definition

of the variable to measure the competition. Verniers et al. construct a Herfindahl–

Hirschman index35 for each drug in each country, however, Danzon and Epstein use the

competitor prices and Kyle includes the measure by Djankov et al. (Djankov et al., 2002).

Furthermore, cross-class price effects are insignificant, indicating that competition occurs

within subclasses rather than between subclasses, and that dynamic competition is driven

by product characteristics other than price (Danzon and Epstein, 2008).

Concerning generic competition, Danzon and Epstein (Danzon and Epstein, 2008)

find that the effects of number of generic competitors are negative but statistically

insignificant, providing further evidence that availability of older, cheaper generic

substitutes is not a significant deterrent to the launch of new brand products, even in older

subclasses where generics are more numerous, possibly because generic substitution is

mostly within molecules rather than between molecules.

1.3.3.3 Regulation Characteristics

The question of regulation has been widely analysed. The most common finding is

that price regulation tends to produce a launch delay (Danzon et al., 2005, Kyle, 2006,

Kyle, 2007, Lanjouw, 2005). When price regulation reduces prices below the level

expected given a country’s per capita income, this problem is exacerbated and launch

35 This index is constructed by summing the squared market shares (MS) (based on revenues in the IMS Health data) of the m drugs in the same ATC4 category as drug i at the time of launch of drug i in country j.



35

delays may extend even to high-income countries. Controlling for expected price, the

models show that such delays have been observed in countries with strict regulation and

in those traditional parallel exporters, either performing negative country fixed effect

(Danzon et al., 2005), or directly introducing dummy variables concerning price regulation

(Kyle, 2006, Kyle, 2007, Lanjouw, 2005) or via the average price competitors in the

country (Danzon and Epstein, 2008).

Various degrees of regulation have been explored. Lanjouw (Lanjouw, 2005)

examine separately high and low-income countries. For high-income countries, all price

regulation – whether moderate or extensive – tends to reduce the probability of a

medicine being launched within two years after first launch, while for lower-income

countries, extensive price controls clearly lower the probability of new medicines reaching

consumers quickly. On the other hand, moderate price control does not appear to have a

significant influence on entry in this case. However, as in the richer countries, the effect of

moderate price regulation depends on a country’s income level. For example, in low-

income countries the existence of a national formulary positively affects launch hazard

(which is not the case for the higher-income countries). One would expect its direct effect

to be negative, but within the lower-income country group this variable may be acting as a

proxy for bureaucratic competence.

On the other hand, Kyle (Kyle, 2007) and Verniers et al. (Verniers et al., 2011)

show that entry actually appears more likely in countries using internal RP. Both papers

show that direct price controls are not significant factors on launch delay or do affect

negatively the drug launch hazard. Therefore, there is some evidence that indirect

controls may be preferable to direct ones from the standpoint of attracting new medicines.

Kyle (Kyle, 2007) also suggests that the effect of price controls is not restricted to

an individual market, but affects the launch of a medicine in other markets as well. From

this idea, Heurer et al. (Heuer et al., 2007) introduce the ERP as explanatory factor and

find that countries using ERP to determine their prices present a significantly lower

probability of launch within the first eight months. Specifically, the two forms of

international comparison based on a formula of foreign prices – either determining prices

directly from such a formula, or using it as an informal basis for the decision – have

impacts that are negative but different. This difference may derive from the fact that, if a


strict rule is applied, there is less room to negotiate higher prices and therefore

companies delay the launch of new products. Moreover, Kanavos and Vandoros

(Kanavos and Vandoros, 2011) state that the ERP may yield that new products are first

launched in less-regulated countries followed by price-regulated countries. However,

Verniers et al. do not find the use of ERP as significant factor affecting the launch delay.

These different results may come from the use of different samples. Heurer et al. analyse

the data for the former EU-15, where the use of ERP is more pronounced (see Leopold et

al. (Leopold et al., 2012)), whereas the sample used by Verniers et al. includes 50

countries worldwide where the use of ERP might not be so orthodox (see Espín J. et al.

(Espin J et al., 2011)).

Furthermore, when considering the presence of PT, no significant effects have

been reported on launch hazard in the importing country; PT risk is more likely to lead to

non-launch or launch delay in the parallel export country (Danzon and Epstein, 2008).

Some studies have examined how the role of the EMA as a centralized

mechanism for marketing approval within Europe has affected the drug launch. Kyle firstly

analyses the countries price ranking and finds that a negative and significant effect

implies that high-price markets are less attractive for launch, although this effect is

quantitatively small. When an interaction term with the post-1995 period is included

considering the beginning of the EMA period, which implies a PT more widespread, it

appears that launch hazard in higher-price countries is more likely. Parallel traders can

essentially arbitrage price differences across countries in the EU, entry into high-price

countries should be more attractive, and entry into low-price ones less so, as imports of

drugs from low-price countries could crowd out sales in higher-price markets (Kyle, 2007).

Moreover, there is some evidence that the establishment of the EMA in 1995 succeeded

in speeding up access to new medicines for consumers, increasing the launch hazard

within 2 years (Lanjouw, 2005). Also, Verniers et al. finds that firms launch faster in

countries belonging to the EMA zone (Verniers et al., 2011).

Additionally, Lanjouw (Lanjouw, 2005) evaluates different levels of patent rights or

protection as factor affecting launch hazard for MLICs. They concluded that short-term

product and long-term process patent regimes both tend to encourage faster launches

(Lanjouw, 2005, Verniers et al., 2011). Verniers et al. (Verniers et al., 2011) reach the

same conclusion.



37

1.3.3.4 Country Characteristics

Population and GDP per capita are both expected to positively affect drug launch.

We might think that a country with a large population may have a large potential demand

and a high-income country would be willing to pay a high price. Therefore, both may get a

rapid access to medicines. However, when these effects are econometrically analysed

their estimated influence depends on the sample and the specification used. Using a

sample in which all countries are large and relatively wealthy, Kyle (Kyle, 2006) reveals

that population and per capita GDP are not especially important determinants. But when

Kyle (Kyle, 2007) analyses a sample of countries with more variance, as expected, high-

income countries and countries with larger populations are likely to have earlier launches.

However, the coefficient of GDP per capita becomes negative and statistically

insignificant while more detailed variables of price regulation are introduced in the model,

such as internal RP36 or pharmacoeconomic evidence37. Additionally, Kyle also shows that

a medicine that has previously been launched in a high-price market is much more likely

to enter an additional market than one that was previously launched in a low-price market.

Also, Danzon et al. (Danzon et al., 2011) confirm that the resulting interconnectedness

across countries contributes to the launch delay or non-launch of new medicines in low-

price EU countries, as firms seek to avoid spillovers to prices in higher-price EU countries.

Similar results are found when Heurer et al., with a wide sample (60 countries),

test only for general country characteristics such as GDP per capita, drug expenditure,

and population size and structure in addition to health indicators. A higher GDP per capita

and a larger population indicate a large potential market and therefore have a positive

impact on the probability of a launch within eight months. However, when variables for the

regulatory scheme such as the use of ERP or the application of internal RP are included,

the effects for GDP per capita and population become insignificant (Heuer et al., 2007).

So, these two important country characteristics that are very relevant to price

bargaining with the industry could be undermined by the country pricing policy. However,

Verniers et al. that takes into account 50 countries worldwide find that population size is 36 Therapeutic class reference pricing: a reference-pricing scheme, in which the patient is responsible for paying the price difference between his chosen drugs and a reference drug defined by the government.

37 Pharmacoeconomic evidence presents the cost effectiveness of a treatment with a new drug as the ratio of the cost of treatment (including the drug price, hospital stays, surgery, and so on) to relevant measures of its effect.


statistically significant and negatively affects the delay launch, even including variables

concerning the country regulatory scheme. But in this case no variable regarding the

country wealth was included (Verniers et al., 2011).

Interestingly, Lanjouw (Lanjouw, 2005) analyses the MLICs and includes

regulatory variables, find that the existence of a larger population and a higher level of

GDP per capita increase the likelihood of a country having a medicines available on the

market within two years after the first launch. Furthermore, the age composition of a

country’s population is a significant determinant of the speed of medicine launch. Drugs

are more likely to reach the market in countries with many children and those with a high

proportion of elderly persons. Among the high-income countries, having a larger

proportion of children seems to be the most important factor. Income distribution also

appears to be an important determinant of market entry. When interacted with the GDP

per capita, is statistically significant and show a pronounced pattern across high-income

and MLICs. Then, a lower-income country is more likely to get new drugs if it is unequal,

ensuring that it has a wealthy “elite”. On the other hand, a high-income country is better

off with a more equal distribution as this generates the largest “middle class”. Equality

becomes less important as average income increases.

1.3.3.5 Firm Characteristics

All the papers examined show that domestic firms tend to enter the market with

short delays (Danzon and Epstein, 2008, Danzon et al., 2005, Kyle, 2006, Kyle, 2007,

Verniers et al., 2011). Similarities between a firm’s home market and the potential launch

market greatly increase the probability of launch; the existence of a common border and

language provides about half the advantage of domestic status (Kyle, 2006). But Kyle

shows medicines invented by firms headquartered in price-controlled countries are less

likely to be introduced in additional markets and therefore price regulation level matters

(Kyle, 2007).

Danzon and Epstein interestingly differ among type of local firms. They differ

among local originator firm, local licensee and local co-marketers38. The drugs produced

38 Local Originator identifies a molecule’s originator corporation launching in its country of domicile; Solo Licensee identifies a locally-domiciled, licensed corporation that launched the molecule in at least one country by itself; and Local Co-marketer identifies a locally-domiciled, licensee corporation that launched together with another firm in its home country and did not launch alone in any country.



39

by local originator firms have a significantly greater local advantage than compounds that

are simply sold by local licensee firms or local co-marketers (Danzon and Epstein, 2008).

The firm’s experience also seems to contribute to the earlier entry of medicines into the

market (Danzon et al., 2005, Kyle, 2006, Kyle, 2007). However, Kyle points out firms with

many medicines in their portfolios tend to launch in fewer countries (Kyle, 2007).

1.4 Discussion

This review tries to know what are the main factors influencing both launch prices

and launch of new drugs at international level. An important conclusion is that pricing and

launching a medicine are two inseparable issues in the global market.

The results predicted from the theoretical models explain the strategic games

played by governments, public insurers and pharmaceutical companies. Although the

models proposed do not reflect the whole reality of the global pricing and launching

process and differ in the hypothesis assumed, they do provide insight into the main issues

that the literature has significantly considered. We distinguish two types of factors. Firstly,

we have observed factors that directly affect drug pricing and launching, such as the

regulation pricing policy, the presence of PT and the firm characteristics. And secondly,

we have also found other determinants that not only impact directly but also indirectly,

affecting the measure in which the first types of factors influence drug pricing and

launching, such as country size and the level of co-payments.

The price regulation policies seem to be one of the most robust factors affecting

the drug pricing and launching. The literature has particularly underlined the direct price

controls such as the use of the ERP (Garcia Mariñoso et al., 2011), the indirect price

controls such as the internal RP (Miraldo, 2009) and the MES + PC, which include both

types (Atella et al., 2012). At this regard, the use of the ERP affects referenced countries.

Their prices increase as consequence of the engagement to the ERP by the referencing

country (Garcia Mariñoso et al., 2011). In turn, the internal RP has been found influencing

positively the firm pricing strategies. Firms set higher prices at first stage when they

anticipate that the RP pricing policy will be applied in the second stage. As in the case of

ERP, the extent of this effect will basically depend on the rule used by the regulator for


calculate the RP level, either the minimum or the weighted average (Miraldo, 2009). Price

regulation regimes not only affect prices, they also may affect R&D and the quality of

drugs. In these terms, Bardey et al., under different hypothesis than Miraldo (Miraldo,

2009), find that the RP yields a decrease in price, this decreasing in price reduces the

incentives to create pioneer drugs; however, the introduction of followers is delayed,

which gives positive incentives to launch pioneers. The net effect within a class is

ambiguous (Bardey et al., 2010). In turn, the quality of drugs may be affected by

regulation regimes. When a price cap is applied together with a MES, whether the

regulator fixes a price cap that binds the high-type quality drugs and a MES that only

binds the low-type quality ones, then not only the price of the high-type drug will decrease

but also its efficacy will be negatively affected (Atella et al., 2012).

The presence of PT appears to be also an important factor. The model proposed

by Ganslandt and Maskus (Ganslandt and Maskus, 2004) predicts that the price in high-

income countries (the parallel importers) converges to the low-income ones (plus a trade

cost). However, Heurer et al. (Richter, 2008) suggest that PT should be considered as a

loss of income for the original manufacturer. Intuitively, in order to compensate this loss of

income, the firm should increase the drug price in the parallel importer country.

The firms as economic agents can also influence prices and launching. Cabrales

and Jiménez-Martín predict that the price set by the regulator is weakly higher for a local-

multinational than for a foreign one (Cabrales and Jimenez-Martin, 2007). Not only the

location of firms matters, but also the contacts between firms competing in the same

markets. Thus, an increasing in the multimarket contact structure induces higher levels of

collusion and therefore higher prices. However, this effect will be undermined in countries

with price regulation (Coronado et al., 2007).

The country size and the level of co-payments affect direct and indirectly the drug

pricing and launching. Both have been widely used as variable to design the theoretical

models. The country size seems to be one of the most important determinants in the

theoretical models.

Concerning the price regulation, the preferences to engage to ERP policy

decrease as the size of the referencing country and as the level of co-payments of

referencing and referenced countries converge (Garcia Mariñoso et al., 2011). When a

regulation regime based on a direct PC in addition to a MES is applied, as the MES



41

makes improve the low-type quality drug and increases its price, but makes worsen the

efficacy in high-type drugs and decreases its price, then the net welfare will depend on

the size of both drug buyers (Atella et al., 2012). Also, the size of the country parallel

importer influences positively the number of PI firms, consequently, the drug price

decreases in the parallel importer country (the high-income country) (Ganslandt and

Maskus, 2004). Furthermore, the population size is also taken into account when firms

decide where to firstly launch a drug in the presence of informational spillovers

concerning asymmetric information on drug quality. Although the information spillovers

can be avoided by launching in all countries simultaneously, the preference to delay

launch will depend positively on the relative country size and on the difference in the

levels of co-payments between countries (García-Mariñoso and Olivella, 2012). Besides,

the population size is also taken into account by the firms. Given that a local non-

multinational receive a weak price premium with respect to a multinational39 in a regulated

country, as the country size of the foreign multinational increases, its price converges to

that of the local multinational (Cabrales and Jimenez-Martin, 2007).

From the perspective of the regulator, concerning the application of ERP, the

literature recommends to engage to ERP only to small countries and/or apply an ERP

based on prices in large countries (or large group of countries); the same applies if one

substitutes “large country” by “small co-payment country” and vice versa (Garcia

Mariñoso et al., 2011). Also, a minimum RP level is advised in order to avoid major

increasing price level with respect to the average RP (Miraldo, 2009). And furthermore, a

MES policy together with a PC should be applied when the welfare lost due to the high-

type drug buyers is not large enough to undermine the welfare gained due to the low-type

drug (Atella et al., 2012).

From the perspective of the firm, the loss of income coming from PT should be

considered and the firm should set a higher price than without PT. However, since the

ERP is widely used by countries and PT does exist, the markets are inseparable. Thus,

the highest drug price is not always the best option for the firm, nor a low drug price

always the worst option in a given country. What may have been an optimal pricing

strategy in a single country is no longer optimal when considering the ERP and PT 39 Either local or foreign.


(Ganslandt and Maskus, 2004, Richter, 2008). Furthermore, in the presence of

informational spillovers on quality, the launch delay should occur in the non-aggressive

country. Regarding the firm location, multinational firms should settled down in large

countries in order to better compete with local multinational firms (Cabrales and Jimenez-

Martin, 2007).

On the other hand, the empirical literature collects an amount of econometric

models which have identified and measured the influence of the most significant factors

affecting prices and launches of medicines in different countries. The use of different

samples may prevent from doing comparisons.

Demographic and income country features, and regulation regimes, seem to be

the most important factors affecting drug pricing and launching in the empirical literature.

Moreover, the drug characteristics as strength, packsize and presentation forms are

significantly related to the price, but it is the therapeutic value, which robustly affect

pricing and launching. The firm location turns up an important factor for the launching

decision but price premiums due to headquarters location appear ambiguous. Generic

competition generally drives down prices; however, there are some different effects,

which deserve some comments. In turn, the brand competition factors do not appear to

exert competitive pressure on prices.

The country size, the GDP per capita and income distribution are three country

characteristics shown as the most significant factors influencing pricing and launching.

Ceteris paribus, higher-income countries pay higher drug prices (Cabrales and Jimenez-

Martin, 2007, Danzon and Epstein, 2008) and have a more rapid access to medicines

than lower-income countries (Danzon et al., 2011, Heuer et al., 2007, Kyle, 2007,

Lanjouw, 2005); furthermore, populated countries also enjoy a higher probability of launch

(Heuer et al., 2007, Kyle, 2007, Lanjouw, 2005, Verniers et al., 2011). Additionally, an

unequal income distribution positively affects the drug launch hazard in lower-income

countries via a wealthy “elite”. On the other hand, a more equal distribution makes the

same effect in high-income countries via a largest “middle class”. However, we should

note that the country pricing policies might undermine the effects of these important

factors. Contrary, interestingly, Cabrales and Jiménez-Martín observe lower prices in the

US than in other countries as Canada, France or Italy, and contrary to the conventional

wisdom about countries regulatory regimes (Cabrales and Jimenez-Martin, 2007).



43

In the empirical review, the most common finding is that price regulation tends to

produce launch delay. Such as delays have been observed in countries with strict

regulation and in traditional parallel exporters. Particularly, Heurer et al. show that

countries using the ERP have a lower probability of launch (Heuer et al., 2007), and

Stargardt and Schreyögg state that the use of ERP have positive direct and indirect

impact on the referencing countries. At this regard, they propose to use as many

countries as possible in the formula (Stargardt and Schreyögg, 2006), which partially

coincides with the recommendations from the theoretical model of García-Mariñoso et al.

(Garcia Mariñoso et al., 2011). However, Stargardt and Schreyögg further recommend

avoiding countries using ERP, in order to prevent undue impacts; and integrate the

market volumes of the referenced countries into the index in order to avoid high prices

and launch delays in countries with small markets (Stargardt and Schreyögg, 2006). In

turn, Kyle shows that there is evidence from the standpoint of attracting new medicines

that indirect price controls such as RP may be preferable to the direct ones such as

pharmacoeconomic evidence or price freeze (Kyle, 2007). Furthermore, the belonging to

the EMA make launches more likely, particularly in higher-price countries (Kyle, 2007,

Verniers et al., 2011). On the other hand, as explicit price regulation, only the explicit use

of HTA has a significant effect on prices (Kanavos and Vandoros, 2011).

The therapeutic quality, the strength, the pack size, the number of presentations

number of the drug and the product life cycle are very significant factors on pricing. The

therapeutic quality, the strength and the number of presentations make increase the price,

the pack size and the product life cycle affect negatively (Cabrales and Jimenez-Martin,

2007, Coronado et al., 2007, Danzon and Chao, 2000, Kanavos and Vandoros, 2011,

Timur et al., 2011, Danzon and Epstein, 2008). Also, the literature importantly considers

the sequential launch of a drug and the effect on its price. Commonly, the higher are the

prices previously set, the higher will be the drug price in a country. We note that when

there is no average global price40, what can be interpreted as a drug innovative character,

the price is higher than when previous prices exist (Cabrales and Jimenez-Martin, 2007).

Therapeutic innovation also influences positively the launching. It has been observed that

a prior launch in a high-income country has a stronger effect on launching in a low-income

40 There is no mean global price because there are no countries to be calculated.


country than vice versa. Also, the fact of launching first in high-price EU markets positively

affects launch prices in the low-price ones via ERP (Danzon and Epstein, 2008).

Although the papers analysed appear to agree on drug characteristics, differences in

therapeutic categories (Danzon and Epstein, 2008, Danzon et al., 2005, Kyle, 2007,

Lanjouw, 2005, Verniers et al., 2011).

From the perspective of the firms, it has been shown that all multinational firms

obtain a price advantage. By contrast, domestic firms tend to enter the market with short

delays (Danzon and Epstein, 2008, Danzon et al., 2005, Kyle, 2006, Kyle, 2007, Verniers

et al., 2011), however, it is not clear that they receive price premiums (Cabrales and

Jimenez-Martin, 2007, Danzon and Epstein, 2008, Verniers et al., 2011). These last

results do not support the theoretical model that predict that local-multinational firms

receive price premiums compared to foreign multinational ones (Cabrales and Jimenez-

Martin, 2007). Only Verniers et al. find that domestic firms obtain higher prices in their

domestic markets, but they do not make differences between multinational and non-

multinational firms. In any case, the literature shows that the contact among firms in

unregulated markets induces higher prices through price collusion (Coronado et al.,

2007).

Most papers find generic competition to negatively affect prices (Cabrales and

Jimenez-Martin, 2007, Coronado et al., 2007, Danzon and Chao, 2000, Danzon and

Epstein, 2008, Danzon et al., 2011, Timur et al., 2011). Interestingly, price regulation is

found to undermine generic competition in strict regulatory systems. However, the

presence of generics may increase the price of brand names products (Cabrales and

Jimenez-Martin, 2007, Coronado et al., 2007).

In general, except for the case of tendering procurement in MLICs, brand

competition does not seem to exert a significant competitive pressure on prices, to the

contrary, several authors find that brand competitor prices affect positively launch hazard

(Danzon and Epstein, 2008, Kyle, 2006, Kyle, 2007). Only Danzon and Epstein indicate

that competition occurs within subclasses rather than between subclasses, and that

dynamic competition is driven by product characteristics other than price (Danzon and

Epstein, 2008).

Contrary to the theoretical model that predicts that the presence of PT reduces

prices in high-income countries (Ganslandt and Maskus, 2004), the empirical literature



45

does not find robust effects on it (Kanavos P and Costa-Font J, 2005). By contrast, PT

risk is more likely to lead to non-launch or launch delay in the parallel export while it has a

lower impact on the importing country.

2 Chapter 2: External Reference Pricing and Pharmaceutical Cost-Containment.

2.1 Introduction

Pharmaceuticals are sold on a global market. This characteristic gives rise to a

specific bargaining procedure between pharmaceutical firms and countries’ health

agencies. On the one hand, a firm makes strategic decisions to sequentially launch

medicines in different countries and to maximize global profits; and on the other hand,

countries’ health agencies implement pricing policies in order to control their

pharmaceutical expenditure and yet guarantee access to medicines.

Among existing drug pricing policies, most countries in the industrialized world

have implemented either Cost-Effectiveness Analysis (CEA) or External Reference

Pricing (ERP) at some point in time with the aim of controlling pharmaceutical expenditure

but still ensuring access to medicines, mainly for on-patent medicines (Espin J et al.,

2011, Rawlins, 2012).

According to the OECD, ERP, also referred to as External Price Benchmarking or

International Reference Pricing, is defined as “the practice of comparing pharmaceutical

prices across countries” and it is further indicated that, “there are various methods applied

and different country baskets used” (Paris et al., 2008). In this thesis, we use the ERP

definition from Garcia-Mariñoso et al. (Garcia Mariñoso et al., 2011): “ERP consists of

setting a price cap for pharmaceuticals, based on ex-manufacturer prices of identical or

comparable products in other countries”.

ERP is not applied homogeneously in every country. There are a wide variety of

methods used to design a foreign price index (Leopold et al., 2012, Espin J et al., 2011). It


mainly depends on each country’s basket, date of prices41, the method used (the lowest

price, the average price, a percentage of the previous ones, etc.) and whether a

weighted-index42 is used or not. We also note that some countries take into account ERP

as a complementary pricing policy together with other pricing policies to help to make the

price decision, thus it is not exclusively applied as a blind pricing policy43. ERP is used

because of its simplicity at technical or analytical level; it does not require a huge task to

collect price information abroad. Furthermore, ERP users think that those prices taken as

a reference are approximately right, suitable or fair. However, they mention that it is

difficult to assess if the resulting prices are appropriate, efficient or optimal in accordance

with any objective criterion. Therefore, if referencing countries set their prices too high or

too low, then any country later applying the ERP method may run the risk of repeating the

same mistake (Espin J et al., 2011).

The basic trade-off faced by a pharmaceutical firm is the following. If a firm delays

a launch in a referencing (low-price) country, it will also delay profits that could be derived

from this country. However, on the positive side, it avoids this low price from overspilling

into other countries due to ERP strategy or parallel trade (Danzon and Epstein, 2008,

Danzon and Towse, 2003, Garcia Mariñoso et al., 2011). By contrast, when countries set

a drug price, they risk the possibility of not providing the drug at the time they desire,

which may have consequences for the health and the welfare of the population

(Lichtenberg, 2005). The use of ERP by countries may make firms apply international

pricing strategies that can harm countries’ welfare. On the one hand, a firm may set a

single price44 which may benefit the high-priced45 countries but harm the low-priced ones,

41 Current price vs. price at launch

42 The most widely method used for new drugs is through non-weighted measures; such methods will not help to achieve the target of obtaining a comparable average level of prices. The application of weighted price indexes, comparable and useful as reference to the rest of countries, has been proposed DANZON, P. M. & CHAO, L. W. 2000. Cross-national price differences for pharmaceuticals: How large, and why? Journal of Health Economics, 19, 159-195..

43 Espín et al. state that “regulators might not always be able or willing to “impose” a certain price, but instead use the price computed as a benchmark or reference for negotiations, often alongside other criteria, such as cost-plus, internal or therapeutic pricing”

44 Two factors contribute to price uniformity between different markets: a) threats of parallel imports, and b) the use of international reference pricing DANZON, P. M. & TOWSE, A. 2003. Differential Pricing for Pharmaceuticals: Reconciling Access, R&D and Patents. International Journal of Health Care Finance and Economics, 3, 183-205..

45 In the long run, consumers from high price countries will be worse off if this lower price results in lower expected returns on R&D, and hence fewer new medicines than they would have been willing to pay for DANZON, P. M. 1997. Price Discrimination for Pharmaceuticals: Welfare Effects in the US and the EU. International Journal of the Economics of Business, 4, 310-322..

Chapter 2: External Reference Pricing and Pharmaceutical Cost-Containment 49

and on the other hand, the firm may either attempt to set high46 prices in first countries to

avoid low prices in later launches via ERP, or delay launches in low-priced countries to

avoid the spill-over effects. These strategies may harm low-priced countries, and they

may even harm high-priced ones (Garcia Mariñoso et al., 2011). Also, another strategy

exists for firms to avoid spill-over effects from ERP. This consists of setting high prices

and granting confidential rebates or discounts to referenced countries. This strategy

allows firms to guarantee lower prices in referenced countries and avoids information

spill-overs of low prices to referencing countries (Espin J et al., 2011).

Most countries use ERP as a pharmaceutical pricing strategy. The use of ERP as

a mechanism to set pharmaceutical prices is quite widely applied: 24 of the 30 OECD

countries (Espin J et al., 2011) and approximately 24 of the 28 EU Member States

(Leopold et al., 2012) have used it.

CEA in health economics aims to estimate the ratio between the cost of a health-

related intervention and the benefit it produces in terms of the number of years lived in full

health by the beneficiaries. Cost is measured in monetary units, while benefit needs to be

expressed as a gain in health measured by quantitative values. However, unlike cost–

benefit analysis, the benefits do not have to be expressed in monetary terms. In

pharmaeconomics, it is usually expressed in quality-adjusted life years (QALYs)47

(National Insitute for Health and Care Excellence (NICE), 2010). The ICER is the ratio

between the difference in costs and the difference in benefits of two interventions. So, an

example in which the costs and gains, respectively, are 140,000 euros and 3.5 QALYs,

would yield an ICER of 40,000 euros per QALY. Commonly, each country has a different

threshold to pay for one QALY. If we suppose that such a country has a threshold of

30,000 euros per QALY, any drug which has an ICER of more than 30,000 euros per

additional QALY gained is likely to be rejected and any drug which has an ICER of less

than or equal to 30,000 euros per extra QALY gained is likely to be accepted as cost-

effective (WHO, 2003). However, drugs do not always yield a single ICER. We note that

other authors have studied in depth CEA according to Bayesian models (Negrin and

46 This company strategy will not work if the high-price country revises its prices downwards after launch

47 The QALY is a measure of disease burden, including both the quality and the quantity of life lived. The QALY model requires utility independent, risk neutral, and constant proportional tradeoff behaviour. The QALY is based on the number of years of life that would be added by the intervention. Each year in perfect health is assigned the value of 1.0 down to a value of 0.0 for being dead. If the extra years are not lived in full health, for example if the patient looses a limb, or goes blind or has to use a wheelchair, then the extra life-years are given a value between 0 and 1 to account for this.


Vazquez-Polo, 2006, Negrin and Vazquez-Polo, 2008, Negrin et al., 2010, Moreno et al.,

2010)

A pharmaceutical firm knows this threshold48. It would also carry out a CEA and

obtain the number of QALYs gained if its drug were provided in one country. Since the

firm is aware of both threshold and number of QALYs, it offers the country the drug at a

certain price just below the threshold. However, the firm may upwardly distort the number

of QALYs to obtain greater profits. Then, it is the country that may revise the firm’s CEA

applying its own CEA to reveal a fair price. However, this CEA requires resources and

consequently an investment of money by the country.

Previous literature has developed games based on bargaining models between

pharmaceutical firms and countries’ health agencies to set drug prices in an international

context. Garcia-Mariñoso et al. (Garcia Mariñoso et al., 2011) examine the effects of

using ERP by referencing countries on referenced countries’ welfare via a bargaining

model. They find that a country has an incentive to engage in ERP if its co-payment levels

are high when compared to other countries. This preference decreases as the relative

size of the country engaging in ERP increases. They also find that these effects harm

referenced countries’ welfare. Furthermore, García-Mariñoso and Olivella (García-

Mariñoso and Olivella, 2012) present a negotiation model based on a “take-it-or-leave-it”

procedure that examines the conditions under which a firm can use launch delay and the

consequent information spill-overs49 to reject low prices. The notion that low prices may

overspill to other countries even in the absence of parallel trade or ERP is introduced here

differently from previous research.

Furthermore, other theoretical papers (Jelovac and Houy, 2013, Richter, 2008)

deal with pharmaceutical firms’ strategies and countries’ pharmaceutical pricing policies

but are not based on bargaining models. In this regard, Richter (Richter, 2008) proposes

48 We note that this threshold does not have to be a single threshold. There exists currently an interesting discussion about the social value of a QALY that determines such a threshold DONALDSON, C., BAKER, R., MASON, H., JONES-LEE, M., LANCSAR, E., WILDMAN, J., BATEMAN, I., LOOMES, G., ROBINSON, A. & SUGDEN, R. 2011. The social value of a QALY: raising the bar or barring the raise? BMC health services research, 11, 8, MASON, H., JONES�LEE, M. & DONALDSON, C. 2009. Modelling the monetary value of a QALY: a new approach based on UK data. Health Economics, 18, 933-950, NIHR, H. 2010. Weighting and valuing quality-adjusted life-years using stated preference methods: preliminary results from the Social Value of a QALY Project. Health Technology Assessment, 14, PINTO-PRADES, J. L., LOOMES, G. & BREY, R. 2009. Trying to estimate a monetary value for the QALY. Journal of Health Economics, 28, 553-562, ROBINSON, A., GYRD-HANSEN, D., BACON, P., BAKER, R., PENNINGTON, M. & DONALDSON, C. 2013. Estimating a WTP-based value of a QALY: The ‘chained’approach. Social Science & Medicine, 92, 92-104.

49 Information spillovers are essentially the demand for lower prices in a country generated by the knowledge about lower prices in other countries.


a mathematical optimization problem for a firm with examples in which international price

dependencies play an important role. This model can help countries to understand the

implication of their ERP policies on a global repeated pricing game. On the other hand,

Jelovac and Houy (Jelovac and Houy, 2013) analyse the timing decisions of

pharmaceutical firms to launch a new drug in countries using ERP. When all countries

reference the prices in all other countries and in all previous periods of time, then there is

no withdrawal of drugs in any country, and in any period of time and there is no incentive

to delay the launch of a drug in any country. However, these results do not hold when the

countries only reference a subset of all countries or when the reference is only on the

latest period prices.

Concerning empirical studies, Danzon and Epstein (Danzon and Epstein, 2008)

interestingly observe that launch prices increase with the level of the lowest price

previously received in other high-price EU countries, whereas the effects of a previous

launch in low-price EU countries are insignificant. This result is consistent with the

hypothesis that launching first in high-price EU markets can influence prices in low-price

ones. This evidence about sequential launch prices validates the theory that a launch

delay in low-price markets may ultimately yield higher prices in these markets through

spillovers from higher-price ones. This theory is shown by Stargardt and Schreyögg

(Stargardt and Schreyögg, 2006). They analyse the direct and indirect impact of the use

of ERP from the referenced to the referencing country. They estimate the impact of drug

price changes in Germany50 on drug prices in other countries using ERP in the former

EU-15. The authors use the formulas applied by each referencing country and then, they

calculate the partial differential of these formulas with respect to a 1 euro price reduction

in Germany. The authors state that the relationship between the direct and indirect impact

of a price change depends mainly on the scheme applied to set prices. Thus, to avoid the

negative effects of ERP and determine prices in order to reduce the direct and indirect

impact of individual countries, a weighted formula of prices containing as many countries

as possible should be used.

Surprisingly, only one out of three studies finds a direct effect on pricing when

ERP is applied. These results could be explained by the different databases collected.

50 Germany is chosen because it is one of the largest pharmaceutical markets in the world, it is characterised by relatively high prices and it is referenced by most cross-referencing countries schemes.


The only time ERP was found to have an effect on launching, the database comprised

prices at launch time between 1995 and 2005. However, in Verniers et al. (Verniers et al.,

2011), the database was limited to drugs launched as of February 1994, though ERP had

not been widely applied at that time (Leopold et al., 2012). In Kanavos and Vandoros

(Kanavos and Vandoros, 2011), the prices of branded originators do not correspond to

launch prices; therefore, competition can also lead to downward pressure on off-patent

originator brands, however ERP is only applied on new drugs at launch time.

In addition, Heurer et al. (Heuer et al., 2007) find that countries using ERP to

determine their prices present a significantly lower probability of launch within the first

eight months. Moreover, Kanavos and Vandoros (Kanavos and Vandoros, 2011) state

that the ERP may result in new products being launched first in less-regulated countries

and, then, followed by launches in price-regulated ones. However, Verniers et al. do not

find the use of ERP as significant factor affecting the launch delay. These different results

may come from the use of different samples. Heurer et al. (Heuer et al., 2007) analyse

the data for the former EU-15, where the use of ERP is more pronounced (see Leopold et

al. (Leopold et al., 2012)), whereas the sample used by Verniers et al. includes 50

countries worldwide, where the use of ERP might not be so common (see Espín J. et al.

(Espin J et al., 2011))

We suggest a game based on a bargaining model involving the sequential

launching of one pharmaceutical by one firm across two countries based on a take-it-or-

leave-it-offer procedure that is developed under asymmetric information (Muthoo, 1999).

Each country has a different ICER or willingness to pay (WtP, henceforth) for one QALY

provided by a new on-patent medicine offered by one firm as monopoly producer. The

firm claims that the medicine produces certain number of QALYs Yw, and offers the

medicine at ex-manufacturer price to the country with a high WtP, and at price to

the country with low WtP. In turn, the countries communicate its price based on their

previously chosen pricing policy. In brief, each country chooses between applying ERP

without any additional cost and CEA with its corresponding investment. Then, the firm

choses its optimal country launch sequence. We introduce two important innovations that

are particularly worth mentioning: firstly, we include two different types of countries

depending on their ERP. Thus, we differentiate the type of country based on their formula

of foreign prices to implement ERP criteria, either the minimum or the average price

pF pF


observed. Secondly, we introduce the launch delay cost and the cost related to applying

CEA for countries to check if the firm declares the true QALY of the drug or not.

This chapter is organized as follows. In Section 2, we present the model. In

Section 3, we solve the game and calculate the best strategy for the firm. In Section 4,

using a given optimal country launch sequence, we compare countries’ welfare under

both types of pricing policies, CEA and ERP. Finally, we draw some conclusions.

2.2 The model

Players

Consider two countries, C and D. Each country i (i = C, D) submits a price based

on its pricing policy either CEA or ERP, which has been previously chosen. Each country i

has a WtP for one QALY achieved by providing one first-line on-patent pharmaceutical.

We also consider a firm that behaves as monopoly producer of this first-line on-patent

pharmaceutical, decides the order of countries and sells the medicine to the countries

during two periods (in stage 3 and 4, see Timing later on). The firm is not located in any of

the countries i (i = C, D). This drug is aimed at treating chronic conditions without any co-

payment. It is authorized via a centralized procedure51; thus, countries are not able to not

authorize this drug. They are just able to not list it for reimbursement. The drug has

already been launched in j countries (j = 1,2,3…J, ) at price . Hence, the players

of the game are one pharmaceutical company and one health agency in each country i.

Consumers

Consumers in each country are defined as those patients that may be treated by

the drug sold by the pharmaceutical firm. Hence, the demand for this drug in a country i (i

= C, D) is defined as the prevalence rate of the chronic diseases for which it may be

51 In the European Union, the centralized procedure is required for biotechnology products as well as for orphan medicines but is optional for other pharmaceuticals. In all cases, the European Commission approves market authorization, following the recommendations of the EMA. Market authorization can be also achieved via the mutual recognition procedure where firms demand authorization in one Member State and file for mutual recognition in other countries. However, our analysis only focuses on the centralized procedure, that is the most prevalent.

j i pj


(2.1)

prescribed. This prevalence rate and therefore such demand, is given exogenously by qi52

and represents the country size.

The drug is provided without co-payment; therefore, consumers will not be

charged for consuming it. The health agency of each country will be the only payer.

Consequently, we assume that the in-patients’ total demand for the drug is exogenous

and given by qi.

Health Agencies

Health agencies care about the gross benefit provided by this drug, represented

by the number of QALY achieved, associated to the prevalence rate qi of its country,

denoted by Bi, then

with

Each country has a different ICER threshold for one QALY and this is defined as

WtP. We note that this threshold may represent a measure of opportunity cost or the

consumption value of health (Claxton et al., 2013). Countries do state their WtP.

Particularly, country C has a low WtP denoted by , and country D has a high WtP

denoted by . Y is the number of QALYs of the drug. Therefore, the Bi represents the

monetary value of health.

52 We note that qi does not significantly change over the years as the medicine is prescribed for patients with chronic conditions.

Bi WtPi Y qi

WtPWtP if country D

WtP if country C

WtPC

WtPD


(2.2)

Countries care about public expenses (PE) since patients do not have to pay for

the drug.

with

r unit cost of delay one period for one patient

PEi piqi i rqi ia

pi

piCEA WtPiY if CEA

with Y YF if the firm is honest

YiCEA otherwise

with WtPWtP if country D

WtP if country C

pIh if ERP with h , min

i 1 if launch delay in country i

0 otherwise

i 1 if country i applies CEA

0 otherwise

YiCEA YF


PE comprises the amount of money paid for the medicine piqi and, the launch

delay cost and the CEA cost where appropriate.

The countries i will pay a price pi that depends on two types of pricing policies. On

the one hand, countries may apply CEA, and on the other hand, countries may use ERP.

If the country i applies CEA, it will pay a price equals to its WtP per each

QALY yielded by a unit of the drug. Each country states their WtP ( for country C;

for country D). The number of QALYs may be either equal to the QALYs

proposed by the firm if the firm states the true QALYs of the drug, or less than , i.e.,

, if the firm states a number of QALYs above its true value.

When country i (i = C, D) applies ERP, two types of player depending on the ERP

formula’s aggressiveness are considered. First, we have country C, defined as the more

aggressive one, whose ERP formula consists of setting a price cap on the new product

taking as its price the lowest price of the product at launch time in the set of countries

where the drug has been already launched (min). Then, we have country D, whose ERP

formula consists of setting a price cap taking the average price ( ) of the drug at launch

time in the set of countries where this pharmaceutical has been already launched. In this

model, the price set when countries i apply ERP is denoted by . Then,

country C will choose the minimum price between the minimum international price ( )

and any price previously set. In turn, country D will choose the average international price

( ), being pI pI

min (see Assumption 2 later on). In case of suffering a delay in the

launch, country D will take into account into the average of any price previously set (in this

model, the price already set in country C). Both types of pricing policies and their

respective strategies are formally explained in Appendix B.

Countries i may experience launch delay if this drug is previously launched in one

of the other countries i. We define the unitary launch delay cost ri for country i (i = C, D) as

the health cost of not providing the medicine to one patient during one period. Therefore,

the total launch delay cost for a country i (i = C, D) is defined as .

piCEA

WtP

WtP Y YF

YF

YiCEA

pIh (h min,)

pImin

pI

rqi


(2.3)

(2.4)

(2.5)

(2.6)

We note that countries i (i = C, D) applying CEA invest an amount of money to

verify if the number of QALY stated by the firm YF is true or not. This investment is

assumed equal across countries with a monetary amount a.

We assume that the objective of a health agency is to the maximize benefits

provided by the drug and to reduce the public expenses (PE) associated with its

purchase. Therefore, the objective function (OF) of the health agency can be written as,

OFi Bi PE (WtPiY pi )qi i rqi ia

Now, let us define several assumptions.

Assumption 1

The WtP per QALY of country D is strictly higher than that of country C.

Prices and proposed by the firm to country C and D, respectively, are defined as

the WtPi times the QALYs stated by the firm,

Then, the price proposed by the firm to country D is higher than the offered to

WtPD WtPWtPC WtP

pF

pF

pFWtPCYF

pF WtPDYF

pF pF


(2.7)

(2.8)

(2.9)

(2.10)

country C

Assumption 2

There is price variability among the J countries where the drug has been already

marketed. Therefore, the average international price is higher than the minimum

international price.

Assumption 3

Given a country i applying CEA, the QALYs revealed by the CEA carried out by the

country will not be higher than that proposed by the firm.

Particularly, if the firm is honest and, therefore, declares the true number of QALYs , then

the QALYs revealed by the CEA carried out by the country is equal to that proposed by

the firm. However, if the firm is not honest and therefore does not give the true number of

QALYs, the QALYs resulting from the CEA is lower than that suggested by the firm.

pF pF

PI PI

min

YiCEA YF if the firm declares the true number of QALYs

YiCEA YF otherwise

YF YiCEA


(2.11)

(2.12)

(2.13)

Consequently, the price resulting from the application of CEA will not be higher than that

proposed by the firm,

piCEA piF if the firm trusted by the health agency

piCEA piF otherwise

Then, owing to the scientific evidence showed by the CEA, assuming that the marginal

cost (mc) of producing the pharmaceutical is zero (mc=0) and given that the firm is profit

maximizing, the firm will always sell at CEA price53.

Assumption 4

Given both countries i applying CEA, the number of QALYs revealed by the research,

YiCEA,, will be the same for both countries.

Assumption 5

The countries’ beliefs about the firm’s honesty are the same for both countries i (i = C,D).

Assumption 6

Under assumptions 3 and 5, the price expected by the country i (i = C,D) when applying

CEA is, 53 We note that the firm knows the countries i (i = C,D) WtP, therefore if it is too low (below a given threshold), the firm does not even initiate negotiations to launch in that country.

Pr(YF YiCEA) , i

Pr(YF YiCEA) 1 , i


(2.14)

(2.15)

(2.16)

(2.17)

Then, under Assumption 1 and 5, the price expected by country D is larger than the price

expected by country C,

Assumption 754

If the price proposed by the firm is higher (lower) than the international reference price,

, then the price discovered by country when applying CEA will also be higher

(lower). Therefore, the CEA expected price will not be lower (higher) than the

international reference price.

Assumption 855

The average between the average international reference price and the country C price is

approximately equal to the average international reference price. Therefore, it is assumed

that J is large enough that the average price is not affected by a further observation.

54 This assumption guarantees the trade-off between choosing CEA and ERP. If country i applies CEA, it will pay a higher price to avoid having the drug launch delayed.

55 The Assumption 8 makes simpler the results and does not affect the conclusions obtained.

piF (1) piCEA E pi CEA

E pD CEA E pC CEA

pF

pIh piCEA

E pi CEA

If pF pIh piCEA pI

h E pi CEA pI

h

If pF pIh piCEA pI

h E pi CEA pI

h


(2.18)

(2.19)

Given p I

pj

j1

J

J

then,

pj pC

j1

J

J 1

p I with pC

p Imin

pF

pCEAC

Assumption 9

The price difference between the international reference price and the price proposed by

the firm is the same regardless of the type of country. Thus, the incentive to apply ERP is

also the same regardless of the type of country (high or low WtP), more formally

The Firm

The pharmaceutical industry is characterized by high fixed costs (F) (Mestre-

Ferrandiz, 2012, Mestre-Ferrandiz, 2013) and low mc (mc=0). Hence, F > 0 stands for the

fixed costs of R&D, safety approval process and marketing the drug in all countries. They

are fixed and independent of the number of people or countries that use the drug. The

firm sells the drug to country i (i = C, D) at price pi. This price pi is the maximum price at

which the firm and the health agencies agree56. If the country’s price comes from a CEA

policy, under assumption 3, the firm will always sell at CEA price. However, if the country

price comes from an ERP policy, the firm will be able to accept or refuse it. Should it

refuse, the firm will delay launch in such a country. Therefore the firm commits to

launching the drug and to satisfying the whole demand in this country (qi) at price pi.

Selling this pharmaceutical product without subsidization is not considered as an option.

Thus, we assume that the objective of a monopoly producer of a medicine is to maximize

56 We are aware that purchaser bodies such as hospital or pharmacy bodies may achieve discounts from this price pi, but they are not considered in this thesis.

pImin p

F pI

pF


(2.20)

the accumulated profits function during the length of the sales (two periods, stage 3 and

4, see timing), which can be written as,

OFF pitqit FiC

D

t1

2

with

qit 0 if the drug is not marketed in country i in time t

Also, we assume that the firm is not located in any of the countries i (i = C,D).

Timing

The timing of this game is as follows. The game has 4 stages. In stage 1,

countries C and D choose their pricing policies, CEA or ERP, and the firm proposes a

price for the drug. In stage 2, countries communicate their prices according to their pricing

policies. In stage 3, as launching is sequential (say launch first in C and then in D or vice

versa), the firm chooses the country launch sequence, i.e., it chooses between delaying in

country D or delaying in country C, and sells the drug in the first country of the sequence.

In stage 4, the firm sells the drug in the second country. Since the firm is profit

maximazing, the firm sells in both countries i (i=C, D).

A priori, notice that country i (i=C, D) can choose between a pricing policy that

eventually requires an investment of money, CEA, and another pricing policy with no

charges, ERP. However, applying ERP, the country may have the drug launch delayed if

the firm the international reference price is lower than de price proposed, whereas,

applying CEA, the country risks not making a useful investment if the firm gives the true

number of QALYs.

In order to show the timing more clearly, see the decision tree in Appendix B.


Assumption 10

Drug launching is sequential and the firm keeps selling the drug in stage 4 to the country

where the drug has previously been launched.

Assumption 11

If only one country i (i= C,D) applies ERP and the international reference price is lower

than the price proposed by the firm, the firm will punish such a country i by delaying

launch in it. If both countries i (i= C,D) apply ERP and the international reference prices

are lower than the prices proposed by the firm, the firm will delay launch only in the

country i (i= C,D) according to the country that offers the lowest income.

2.3 Price Setting and Sequential Launch

Players maximize their objective function and we solve the game applying backward

induction. For both cases, when the firm claims a number of QALYs above the true

number of QALYs, and when it claims the true QALYs of the drug, then, we can solve for

each country’s pricing policy: i) no countries apply ERP, ii) only country D applies ERP, iii)

only country C applies ERP and, iv) both countries apply ERP, we calculate the conditions

of the optimal country launch sequence for the firm, either first country C and second

country D, or vice versa. The proof can be checked in Appendix B. In Section 4, we will

compare the countries’ surplus for each pricing policy, given the optimal country launch

sequence.

2.3.1 The firm is trusted by the health agency

i) No countries apply ERP

Since both countries C and D apply CEA, countries C and D will pay and

respectively according to (2.11). Since the firm is trusted by countries i (i= C,D),

under assumption 1, countries C and D will equivalently pay

and correspondingly.

pCCEA

pDCEA

pF

pF


(2.21)

(2.22)

(2.23)

Notice that, under assumption 11, the incomes of the country where the drug has been

first launched are multiplied by two.

Concerning the health agencies’ surpluses, we note that the price considered by

countries C and D when applying CEA is an expected price according to assumptions 5

and 6, since both countries have uncertainty about the number of QALYs stated by the

firm.

the firm’s profits are,

and the health agencies’ surplus are57,

OFC (WtPCYF E pC CEA

)qC a if C, D (WtPCYF p

F)qC a rqC otherwise

OFD (WtPDYF E pD CEA

)qD a rqD if C, D (WtPDYF pF )qD a otherwise

57 Note that health agencies do not know if the firm has given the true number of QALYs or not. Therefore, they only know the expected price.

OF 2p

FqC pFqD F if C, D

pFqC 2pFqD F otherwise


(2.24)

PPR58 1

{C, D} if

pF

pF

qD

qC

{C, D} otherwise

If no countries apply ERP, the firm chooses to delay launch in country D, if and only if the

price ratio country C to country D is greater than the size ratio country D to country C,

otherwise the firm will delay launch in country C. Since the price ratio is less than unity,

then a necessary condition for the sequence {C,D} is that country C must be larger than

the D’s.

ii) Only country D applies ERP

Country D decides to apply ERP in stage 1. If the average international price is

lower than the price proposed by the firm, i.e., , under assumption 11, the firm

delays the launch in country D. Since the pricing policies are set ex-ante, if ,

country D will pay in any case. However, as country C applies CEA, it will pay .

Since the firm gives the true number of QALYs, under assumption 1, it will equivalently

pay .

Concerning the health agencies’ surpluses, we note that the price considered by

country C when applying the CEA is an expected price according to assumptions 5 and 6,

since the country C has uncertainty about the number of QALYs given by the firm.


58 Preliminary result.

pI pF

pI pF

pI pCCEA

pF


(2.26)

(2.28)

(2.27)

(2.25) OFF 2p

FqC p

I qD F if C, D p

FqC 2p I

qD F otherwise

and the health agencies’ surpluses are,

PPR 2

{C, D} if p

F

pI

qD

qC

{C, D} otherwise

If only country D applies ERP, the firm will choose to delay launch in country D if and only

if the price ratio country C to country D is larger than the size ratio country D to country C,

otherwise the firm will delay launch in country C. Specifically, the price ratio is the ratio

between the CEA price and the average international reference price. Since the price ratio

is less than unity, then a necessary condition for the sequence {C,D} is that country C

must be larger than the D’s.

iii) Only country C applies ERP

Country C decides to apply ERP in stage 1. If the minimum international price is

lower than the price proposed by the firm, i.e., . The firm delays the launch in

OFC (WtPCYF E[pC ]CEA)qC a if C, D (WtPCYF E[pC ]CEA)qC a rqC otherwise

OFD (WtPDYF pI

)qD rqD if C, D (WtPDYF pI

)qD otherwise

pImin p

F


(2.30)

(2.29)

(2.31)

(2.32)

country C. Since the pricing policies are set ex-ante, if , the country C will pay

in any case. In turn, country D applies CEA and it will pay . Since the firm

gives the true number of QALYs, under assumption 1, it will pay the equivalent of .

Concerning the health agencies’ surpluses, we note that the price considered by the

country D when applying CEA is an expected price according to assumptions 5 and 6,

since country D is uncertain about the number of QALYs stated by the firm.


OFF 2pI

minqC pFqD F if C, D pI

minqC 2pFqD - F otherwise


PPR 3

{C, D} if pI

min

pF

qD

qC

{C, D} otherwise

pImin p

F

pImin pDCEA

pF

OFC (WtPCYF pI

min )qC if C, D (WtPCYF pI

min )qC rqC otherwise

OFD (WtPDYF E[pD ]CEA)qD a rqD if C, D (WtPDYF pF )qD a otherwise


(2.35)

(2.34)

(2.33)

If only country C applies ERP, the firm chooses to delay launch in country D if and only if

the price ratio country C to country D is larger than the size ratio country D to country C,

otherwise the firm will delay launch in country C. Specifically, if the price ratio is the ratio

between the CEA price and the minimum international reference price. Since the price

ratio is less than unity, then a necessary condition for the sequence {C,D} is that country

C must be larger than the D’s.

iv) Both countries C and D apply ERP

Both countries decide to apply ERP in stage 1. If one of the reference international

prices is lower than the price proposed by the firm, i.e., or , the firm

delays the launch in such country. In the event that both countries hold such a condition,

under Assumption 11, the firm will sell the drug to one of the countries i at stage 3. Since

the pricing policies are set ex-ante, both countries C and D will pay international reference

prices and respectively.


OFF 2pI

minqC pIqD F if C, D

pIminqC 2pI

qD F otherwise


pI pF pI

min pF

pImin pI

OFC (WtPCYF pI

min )qC if C, D (WtPCYF pI

min )qC rqC otherwise

OFD (WtPDYF pI


)qD otherwise


(2.36)

PPR 4

{C, D} if pI

min

pI qD

qC

{C, D} otherwise

If both countries apply ERP, the firm chooses to delay launch in country D if, and only if,

the price ratio between country C and country D is larger than the size ratio country D to

country C, otherwise the firm will delay launch in country C. Specifically, the price ratio is

the ratio between the minimum and the average international reference price. Since the

price ratio is less than unity, then C country must be larger than D country.

In Table 2.1, we summarize the PPR 1, 2, 3 and 4. We show for each combination

of pricing policies i), ii), iii) and iv), under which conditions (pD, pC, qC, qD) the firm chooses

its optimal country sequence. In brief, we note that there is a trade-off between prices and

country sizes under each pair of pricing policies.

Table 2.1 PPR for optimal country launch sequence (pi , qi)

We define the four sets of possible combinations of firm offer prices and ERP:

a)

b) min

IFIF ppandpp

c)

d) pF pI

and pF pI

min

pF pI and p

F pI

min

pF pI and p

F pI

min

Sequence / Pricing

Strategy

i) No countries apply ERP

ii) Only country D

applies ERP

iii) Only country C

applies ERP

iv) Both countries

apply C and D

{C,D}

{D,C}

pF

pF

qD

qC

pF

pI

qD

qC

pImin

pF

qD

qC

pImin

pI qD

qC

pF

pF

qD

qC

pF

pI

qD

qC

pImin

pF

qD

qC

pImin

pI qD

qC


In the following figures (Figure 2.1, 2.2, 2.3 and 2.4), we show under which pair of

pricing policies ( i), ii), iii) and iv) ) and which pairs of price ratios and country size ratios

the firm chooses its optimal country sequence to launch the drug. Below each figure, we

present additional tables (Table 2.2, 2.3, 2.4 and 2.5) containing detailed information

relating to each figure. Each figure shows the optimal launching sequence under the

regions of price ratios and country size ratios.


a) If

Figure 2.1 Optimal country launch sequence under a)

Table 2.2 Country Launch Sequence in Figure 2.1

Region/ Policies

A B X59 E F

i) No ERP {C,D} {C,D} {D,C} {D,C} {D,C} ii) D ERP {C,D} {C,D} {C,D} {C,D} {D,C} iii) C ERP {C,D} {D,C} {D,C} {D,C} {D,C} iv) Both ERP {C,D} {C,D} {C,D} {D,C} {D,C}

59 Under pI

pF pF pI

pF pI and p

F pI

min


b) If pF pI

and pF pI

min

Figure 2.2 Optimal country launch sequence under b)

Table 2.3. Optimal Country Launch Sequence in Figure 2.2

Region/ Policies

A B X60 E F

i) No ERP {C,D} {C,D} {C,D} {C,D} {D,C} ii) D ERP {C,D} {C,D} {C,D} {D,C} {D,C} iii) C ERP {C,D} {C,D} {D,C} {D,C} {D,C} iv) Both ERP {C,D} {D,C} {D,C} {D,C} {D,C}

60 Under pI

pF pF pI


c) If

Figure 2.3 Optimal country launch sequence under c)


Region/ Policies

A B X61 E F

i) No ERP {C,D} {D,C} {D,C} {D,C} {D,C} ii) D ERP {C,D} {C,D} {C,D} {D,C} {D,C} iii) C ERP {C,D} {C,D} {D,C} {D,C} {D,C} iv) Both ERP {C,D} {C,D} {C,D} {C,D} {D,C}

61 Under pI

pF pF pI

pF pI and p

F pI

min


d) If pF pI

and pF pI

min

Figure 2.4 Optimal country launch sequence under d)


Region/ Policies

A B X62 E F

i) No ERP {C,D} {C,D} {D,C} {D,C} {D,C} ii) D ERP {C,D} {C,D} {D,C} {C,D} {D,C} iii) C ERP {C,D} {D,C} {C,D} {D,C} {D,C} iv) Both ERP {C,D} {C,D} {C,D} {D,C} {D,C}

62 Under pI

pF pF pI


2.3.2 The firm states the number of QALYs above the true value

This part of the tree has a similar solution to that solved above. For reasons of

brevity, we just highlight the differences.

In this case, if the country applies ERP, as in the previous case, it will not be able

to discover the real value of the drug and it will pay the international reference price ,

which may be higher or lower than the price revealed when applying CEA. However, if the

country decides to apply CEA, it will reveal a lower number of QALYs than those

proposed by the firm . Consequently, the country will pay a lower price than the

price proposed by the firm . Therefore, on the one hand, it is now less likely for the firm

to accept the international reference prices than in the case of being honest, which

implies that countries applying ERP will be more likely to experience launch delays. Also,

the regions under which the firm chooses its optimal launching sequence change (Figures

2.2, 2.3, 2.4 and 2.5). On the other hand, the expected value of the drug price for the

countries will be higher when the firm states a number of QALYs above its true

value than when it does not. The implications of this issue will be explained in section 4.

2.4 Comparing Policies: CEA vs. ERP

In this section, given the optimal countries launch sequence by the firm under PPR

1, 2, 3 and 4, we compare the countries’ welfare under each pricing policy, CEA and ERP,

to know which of them is more convenient for countries. Thus, since we have solved the

problem by backward induction, we have carried out this comparison for each country

given the optimal country launch sequence for the firm. We have made this comparison in

three steps. Firstly, we have compared the best outcome for each country under the same

pricing policy and under the same country launch sequence, i.e., using CEA (ERP) under

{C, D} and {D, C}, respectively. Then, in a second step, we have compared the best

outcomes between ERP and CEA for each country launch sequence. In a third step, we

have compared the best pricing policy under each country launch sequence. Thus, we

have Condition 1. The proof can be checked in Appendix B.

pIh

YiCEA

YFpiCEA

pF

pIh

E[pi ]CEA


(2.38)

(2.39)

(2.40)

Condition 1

If country i does not suffer from delay launch under any pricing policy or, under both

pricing policies, country i will be better off applying ERP when the unitary cost of carrying

out CEA is higher than the difference between the international reference pricing and

the expected price of country i under CEA (henceforth, the price difference). In addition,

the smaller the population size is, the more attractive will be the use of ERP, since the

unitary cost of CEA increases.

If country i suffers from delay launch when applying ERP but not under the use of CEA,

the delay cost (r) will make CEA more attractive,

Analogously, if country i suffers from delay launch when applying CEA but not under ERP,

the delay cost (r) will make ERP more attractive,

In order to show graphically Condition 1, we plot the following figure,

pIh

a

qi

pIh E[pi ]CEA

a

qi

pIh E[pi ]CEA r

a

qi

pIh E[pi ]CEA r


Figure 2.5. ERP vs. CEA

According to Figure 2.5, we observe under which conditions regarding the unitary

cost of applying CEA and the price difference, the country chooses either ERP or CEA.

Either if both pricing policies (CEA and ERP) are applied without experiencing any delay

launch, or both suffering from a delay launch, the country will be better off applying ERP

only if the unitary cost of applying CEA is higher than the price difference.

Intuitively, we note that since the unitary cost of applying CEA decreases, to keep

the application of ERP beneficial, the price difference should be smaller, either because

the international reference price is lower or the expected price of the country i under CEA

increases. On the other hand, if the unitary cost of applying CEA increases, to maintain

the benefits of applying CEA, the price difference should be larger, either because the


international reference price goes up or the expected price of the country i under CEA

decreases.

However, when we compare both pricing policies, on the one hand, if only the

country applying CEA suffers from launch delay, even though the price difference is

higher than the unitary cost of applying CEA, the unitary delay cost associated with CEA

may compensate this higher difference and make ERP more worthwhile. On the other

hand, if only the country applying ERP experiences launch delay, and the unitary cost of

applying CEA is higher than the price difference, the delay cost of applying ERP may

offset a high unitary cost of CEA and make CEA more attractive than ERP for the country.

Then, when the price difference is negative, i.e., the expected price of the country i

under CEA is higher than the international reference price and there is also a launch

delay when applying CEA, ERP will always be chosen by the country. Similarly, if there is

no delay applying CEA, intuitively, the unitary cost of applying CEA must be higher than

the unitary delay cost induced by applying ERP to offset the negative price difference and

therefore to keep ERP attractive, despite the delay launch.

Importantly, we note that the expected price will be higher when the firm

declares a number of QALYs above its true value than when it does not, which implies

that ERP will be more attractive for the countries when the firm is not trusted by countries

i (i = C, D).

2.5 Conclusions

Using a model where one firm sells an on-patent drug to two countries, which

differ in their WtP (ICER), population size, pricing policy (ERP vs. CEA) and ERP formula,

one of our main results is that the optimal country launch sequence depends on the

relative price and the relative country size. The relative price depends on the countries’

pricing policy (ERP or CEA) and the ERP formula.

Given the optimal country launch sequence, our other overall result is that a

country is better off applying ERP instead of CEA if the unitary cost of CEA is higher than

the price difference, i.e., the difference between the international reference price and the

expected price of the country under CEA. The cost of delaying may affect this decision if

only one of the pricing policies is applied with delay. Thus, if ERP is applied with delay,

E[pi ]CEA


the delay cost will make it less attractive with respect to CEA, and analogously, the same

applies for CEA with respect to ERP.

Basically, the higher the cost of CEA and the lower the international reference

price is, then the more attractive the use of ERP is. In brief, we have compared two

pricing policies: one of them, ERP, does not require any investment, and the other, CEA,

needs an investment of money. In these terms, the smaller the population size is, the

more attractive the use of ERP will be, since the unitary cost of CEA increases.

The application of ERP will be more attractive than CEA when the firm is not

honest, however, it will be more likely to experience launch delays when it is not honest.

Therefore, the convenience, in this case, will depend on how many more QALYs above

the true number the firm states its drug has and the delay cost.

We accept that a wider variety of factors than already used in the model that may

affect the bargaining process. For example, the formulas used to apply ERP may be other

than the average or the minimum, more than two periods could be considered, firms may

offer one single price or launch simultaneously and the effectiveness revealed by CEA

could be different among countries. Also, we accept that assumption 9 constrains the firm

strategy since one price pF is set, the other is implicitly set as well. Besides, other factors

such as the population age structure or the lobbying activity of the pharmaceutical

industry (Abraham, 2002) may also need to be considered.

However, we consider that this chapter has provided insights into the way a

country’s WtP and its pricing policies affect the optimal launch sequence of a firm.

Additionally, given an optimal launch sequence, we propose under what conditions,

regarding country size and pricing policy, it is better off applying CEA or applying ERP for

country i (i = C, D).

3 Chapter 3: Global Pricing and Launching of New Drugs. An Econometric Approach

3.1 Introduction

Pharmaceuticals are sold in a global market that involves a specific bargaining

procedure between pharmaceutical firms and countries’ health agencies. On the one

hand, firms sequentially launch medicines in different countries to maximize global profits,

therefore pricing and launching are their major strategic decisions. On the other hand,

countries’ health agencies implement pricing policies to control their pharmaceutical

expenditure and to guarantee access to medicines. Indeed, pharmaceutical price

regulation is high on policy agendas in several countries, either because countries have

just reformed, intend to reform or question their practices (see Chapter 1 section 1.1).

Among existing drug pricing policies, most countries in the industrialized world have

implemented ERP at some time with the aim of controlling their pharmaceutical

expenditure and ensuring access to medicines, mainly in on-patent medicines (see

Chapter 2 section 2.1).

In this chapter, we aim to analyze the trade-off between pricing and launching and

the impact of ERP policy on pricing and launching from an empirical point of view. We

develop a model that focuses on both issues, controlling for molecules, regulation and

country characteristics. We replicate the study of Danzon and Epstein, published in 2008,

and the study of Verniers et al. published in 2011. Thus, we aim to test how the situation

has changed applying the same methodology to more recent data, and in the case of the

second study, to a different list of countries.

The previous literature concerning the trade-off between pricing and launching has

been already discussed in detail in Chapter 1 in Section 1.2.3 at a theoretical level and in

Section 1.2.4 from an empirical point of view. Furthermore, literature concerning the


impact of ERP policy has also been theoretically and empirically discussed in Chapter 1 in

sections 1.2.2 and 1.4.2, respectively. Additionally, in chapter, 2 we developed a

theoretical model that analyses the convenience of applying ERP as an alternative to

CEA as a cost-containment policy on pharmaceutical expenditure. Particularly, chapter 2,

section 2.1, provides insights into the implementation of ERP policy.

In this chapter, we develop a two-equation empirical model consisting of a launch

delay equation and a relative launch price equation. Previously, we replicate two studies

(Danzon and Epstein, 2008, Verniers et al., 2011) using our database to compare their

results with those obtained from our updated data and different list of countries.

We use data from IMS Health database on 56 new molecules launched in 20

countries belonging to 11 therapeutic classes, all of them approved through the

centralised procedure by the EMA, during the study period, 2004-2010. We have

collected yearly inpatient and outpatient sales in euros at ex-manufacturer price and unit

volume (IMS SU).

Our contribution to the previous literature analysed in Chapter 1, section 1.3,

consists of the analysis of data at presentation level63, the consideration of the relative

launch price64 as an endogenous variable in the relative launch price equation, the study

of the launch delay as a duration time variable and the analysis of the inpatient market.

Additionally, we introduce country size and country purchasing power as additional

explanatory variables.

3.2 Data description

In Tables 3.1, 3.2, 3.3, 3.4 and 3.5, we show the descriptive statistics of our

database. 75% of the countries belong to the EMA and 70% of the countries apply ERP.

No all molecules have been launched in all countries. In the retail market countries

belonging to the EMA experience shorter launch delays and pay lower relative launch

price on average than countries out of the EMA. In the hospital market, the pattern of

launch delays is similar to the retail market, while countries belonging to the EMA and

63 We define two products with the same presentation when both products belong to the same molecule i and have the same quantity of active ingredient per standard unit (see definition of standard unit in Chapter 3 section 3.3.2).

64 Defined later in Appendix C.3.

Chapter 3: Global Pricing and Launching of New Drugs. An Econometric Approach 83

countries out of it pay the same relative launch prices on average. Both the launch delays

and the relative launch prices show high variability. Countries pay higher relative launch

prices in the hospital market than in the retail one, however, no correlation have found

between relative launch prices for the retail and hospital market.

In both retail and hospital markets, we do not find statistical significant differences

in relative launch prices neither between the countries that apply ERP and countries that

do not, nor between countries belonging to the EMA and countries that do not. However,

statistical significant differences are found when statistical differences in launch delays

are analysed. Then, countries applying ERP present significant longer launch delays on

average while countries belonging to the EMA experience significant shorter launch

delays on average.

Table 3.1 Descriptive statistics. Retail market

Retail market

Number of Molecules Launched

Mean Relative Price

SD Relative Price

Mean Delay in Months

SD Delay in Months

ERP

EMA 5.470597 51.26997 11.82316 11.73209 Austria 48 7.50811 12.50254 *

Belgium 24 0.9734428 19.43561 *

Czech Republica 32 6.151763 18.99198 *

Denmark 52 7.387367 10.90712

Finland 36 5.091636 11.36712 *

France 31 4.115586 17.85784 *

Germany 58 11.03968 9.974775

Hungarya 33 1.468164 18.55194 *

Italy 19 4.34083 20.33394 *

Netherlands 31 1.011494 6.705952 *

Norway 35 1.883906 8.131455 *

Polanda 28 1.959413 13.76601 *

Spain 23 0.873803 18.75988 *

Sweden 46 10.72532 6.21954

United Kingdom 37 1.449288 8.15

Non EMA 7.450721 40.28605 16.31287 16.42291 Australia 34 11.41437 22.50628 *

Canada 41 8.664639 16.12424


Retail market


Mean Relative Price

SD Relative Price


SD Delay in Months

ERP

Japan 27 4.778401 34.7982 *

Switzerland 47 10.82668 15.38667 *

United States 62 7.842861 6.343363

a:CountriesthatjointtheEMAduringthestudyperiod;*:CountriesapplyingERP.

Table 3.2 Descriptive statistics. Hospital market

Hospital market


Mean Relative Price

SD Relative Price


SD Delay in Months

ERP

EMA 12.43664 71.40279 11.42848 11.38087 Austria 44 14.05914 12.04966 *

Belgium 22 18.32661 16.71733 *

Czech Republica 31 16.08996 19.83333 *

Denmark 52 11.27124 11.44528

Finland 35 15.11979 9.185088 *

France 30 12.74718 17.39222 *

Germany 56 12.88087 10.01222

Hungarya 28 15.39054 20.26437 *

Italy 19 10.99761 22.70952 *

Netherlands 31 8.352865 7.192381 *

Norway 35 13.02232 7.750926 *

Polanda 25 8.212163 13.00641 *

Spain 22 12.09223 18.39067 *

Sweden 46 16.21124 7.570139

United Kingdom 36 4.784038 8.495

Non EMA 13.32901 60.59304 16.25211 16.42291 Australia 34 20.39585 21.96126 *

Canada 36 9.132268 15.19402

Japan 27 2.832625 32.43333 *

Switzerland 45 16.75053 15.06879 *

United States 62 11.91477 4.975661

a:CountriesthatjointtheEMAduringthestudyperiod;*:CountriesapplyingERP.


Table 3.3 Relative prices Pearson correlation. Retail and hospital market

Retail market Hospital market Retail market 1

Hospital market -0.0192 1

p-value 0.7349

Table 3.4. Bivariate Test. ERP vs. No ERP

Price Delay Group / Market Retail market Hospital market Retail market Hospital market

Mean Mean Mean Mean ERP 5.015412 11.16336 16.19557 17.42601 NO ERP 8.090827 13.10127 9.252808 9.14058 p-value 0.2155 0.5698 0.0000 0.0000

Table 3.5. Bivariate Test. EMA vs. No EMA

Price Delay Group / Market Retail market Hospital market Retail market Hospital market

Mean Mean Mean Mean EMA 5.470597 12.48235 11.82316 13.53388 NO EMA 7.450721 12.48235 16.31287 17.3578 p-value 0.4284 0.9945 0.0000 0.0000

3.3 Replicating Danzon and Epstein (2008)

In this section, we replicate the study by Danzon and Epstein (Danzon and

Epstein, 2008) (D&E, henceforth) with our updated database to compare their results with

those obtained from our more recent data. D&E data cover the years 1992-2003 for 111

molecules in 15 countries belonging to 12 therapeutic classes. Our data cover the period

2004-2010 for 71 molecules in 20 countries belonging to 11 therapeutic classes. The

same data treatment and methodology conducted by D&E have been implemented on our

database.

3.3.1 The D&E model

D&E estimate separately a launch proportional hazard and a launch price

equation. They first estimate a launch proportional hazard model based on a clog-log

regression with molecule-clustered standard errors. They also estimate a random effects

clog-log model to test for molecule-level heterogeneity. Subsequently, the marginal


effects for each independent variable are calculated. Particularly, the marginal effects for

continuous covariates are calculated as the Average Marginal Effect (AME). By contrast,

the marginal effect for a categorical covariate is the discrete change from the base level

(x=0) to the level referring to the presence of the attribute (x=1).

Then, a launch price model is estimated under a Heckman two-step procedure

(Heckman, 1979) to account for possible selection bias produced by the correlation

between the propensity to launch and the launch price. In the first-stage, the launch

proportional hazard model described above is estimated. Then, the Inverse Mills Ratio

(IMR) (Lee, 1983)65 is calculated and introduced as a control variable in the launch price

equation. This equation is estimated by OLS with molecule-clustered standard errors. Its

dependent variable is the logarithm of the volume-weighted mean price for each

molecule-country-year observation. To account for unobserved molecule characteristics,

they also report results from a GLS molecule random effects estimator.

3.3.2 Data

D&E use data from IMS Health’s Midas database on drugs in 15 countries for 12

therapeutic classes, all of which experienced the launch of a new subclass shortly before

or during the study period, 1992-2003. D&E collect quarterly data on outpatient sales at

ex-manufacturer prices and unit volume (IMS SU)66. After the data were screened for

internal consistency, revenue was adjusted for inflation using country-quarter specific

Producer Price Indexes (PPI) available from the International Monetary Fund (IMF), with

2003 as the base year, and converted to US dollars using the average 2003 country-

specific exchange rate. Price per dose for each drug was calculated on a quarterly basis

as the ratio of total revenues to SU sold67.

65 The IMR for molecule i in country j and time t, Mijt, is calculated using the predicted probability of launch p̂ijt from a

clog-log regression as Mijt [1( p̂ijt )]

( p̂ijt ), where is the standard Normal density function and is the standard

Normal distribution function.

66 The IMS SU is a proxy for a dose for each formulation e.g. one tablet or capsule, 5ml. for liquids. The IMS price data for the US do not reflect off-invoice discounts given by manufacturers to health plans and hence are upward biased for manufacturer net revenues.

67 Multiple form-3 level formulations are combined (e.g. tablets and capsules, possibly of different strength) in a given country and quarter into a single observation and define the price as the volume-weighted average price per unit. Identical forms that were launched by different co-marketing companies were also averaged.

[] ()


We also use data from IMS Health database. However, we only consider new

launch drugs. Our data corresponds to 20 countries for 11 therapeutic classes during the

period 2004-2010, all of them approved by the centralised procedure of the EMA. We

have also collected outpatient sales at ex-manufacturer price. Differently from D&E, our

frequency is yearly. The same procedure as D&E is applied for the inflation adjustment

using the PPI available from the IMF; however, we use 2005 as the base year.

Furthermore, since we collected the prices in euros, they have been converted into USD

dollars applying the exchange rates from the IMF. The drug price has been calculated as

in D&E. Although D&E distinguish between Superior and Inferior molecules, we have not

taken into account such a difference, since we have only included new launch drugs,

therefore, we will compare the D&E results from Superior subclasses with our regression

results, as our new launch drugs may never belong to an Inferior (old) subclass as

defined by D&E. D&E classify countries into three categories: high-price EU, low-price EU

and high-price non-EU, with Germany being the reference country. Since the sample of

countries is not exactly the same, the number of countries by category where the

molecule has been previously launched cannot be directly compared.

In Appendix C.1, we report the variable definitions and why certain variables have

not been included in our updated model. In Tables C.1 and C.2, we show the comparison

between the results from D&E and our model.

3.3.3 Comparison of results

3.3.3.1 Launch equation

The sample of D&E is larger than in our updated model (23.400 vs. 3609) because

their study period was longer, the number of molecule-level clusters was larger (111 vs.

71), and as we mentioned our data are yearly and the data of D&E are quarterly. This

could have implications for the results. Results commented below are shown in Table C.1.

In the D&E model, the number of high-price EU and non-EU countries where the

molecule is launched and competitor drug prices appear as robust predictors of the

launch delay of a molecule. Therefore, the more the molecule is spread in high-price

countries, the higher is the propensity to launch, while firms seem to delay launch in low-

price EU countries until launch has occurred in high-price EU countries. In our updated


model, we find similar results, on the one hand, as we only find effects in high-price EU

countries68, we may state that firms delay launch even in high price non-EU countries until

launch has occurred in high-price EU countries. On the other hand, we have not been

able to include the competitor drug prices in the updated model as we explain in detail in

the Appendix C.1. Furthermore, the launch delay is significantly related to the time

elapsed since the global launch in both models. However, this negative effect is offset

when the launch delay is very long under the D&E model (positive quadratic effect), while

in the updated model, the quadratic effect is not statistically significant.

With the aim of analyzing the spillover effects, D&E consider the number of

countries where the drug has already been launched as an explanatory variable of the

propensity of launch. The authors distinguish between different groups of countries using

specific country dummies, e.g. Germany or UK make up one group, and Sweden and the

Netherlands another. In the updated model, since both studies have different countries,

we have grouped the five D&E variables into three, according to an alternative

specification of D&E69. Though variables in D&E and our updated model are not exactly

comparable, we observe that the marginal effect of having launched in high-price EU and

high-price non-EU countries is positive and statistically significant, while the marginal

effect of having launched in low-price EU is significant and negative. These results also

support the hypothesis that launch in low-priced EU countries is adversely affected by the

risk of spillovers to higher-price EU and non-EU countries through ERP and PT. We may

conclude that firms delay launch in low-price EU countries until launch has occurred in

higher-price EU and non-EU countries. Our results support more strongly this hypothesis

than the D&E model, due to the negative effect of launching in a low-price EU country on

the propensity of launch.

Concerning the types of firm, in D&E the three types of firm considered are

statistically significant and positive. Therefore, launch is more likely for molecules from

firms that are domiciled domestically. Indeed, the marginal effect of being a Local

Originator firm on the propensity of launch is greater than being a Solo Licensee or Co-

marketer. This evidence supports the hypothesis that Local Originators have local R&D

68 In Table C.1 only the variable Num Already Launched (Spain, Portugal, Greece) is not statistically significant. See explained in the next paragraph how countries have been grouped in D&E and in our updated model.

69 This specification distinguishes among high-price EU, low-price EU and high-price non-EU countries.


activities, and are, thus, likely to be larger, have greater regulatory expertise and are

viewed as more valuable to the local economy than Co-marketers, with Solo Licensees in

the middle. In the updated model, being a Local Corporation, either Originator or Solo

Licensee, does not present differences in launching.

When introducing the country-fixed effects, taking Germany as reference, we

observe a similar pattern in both models with marginal effects statistically significant and

negative. Indeed, Germany enjoys earlier drug launches on average than the rest of

countries. In both models, therapeutic class Fixed-effects are included and are significant.

However D&E do not report these results, so we have not compared them.

Regarding alternative specifications, the clog-log with random effects estimates

are generally similar to the clog-log estimates with robust clustered standard errors. The

country-fixed effects are mostly not statistically significant under random effects.

3.3.3.2 Launch price equation

The sample of D&E has a similar size to our sample (950 vs. 762) because

although our study period is shorter and the number of molecule-level clusters is smaller

than that of D&E (109 vs. 71), we analyse the data at pharmaceutical presentation level,

while D&E conduct their study at molecule level. Both R-Squared are high (0.89 vs. 0.87).

Results commented on below are shown in Table C.2. D&E focus on estimates that

include GDP per capita; excluding GDP changes mainly the country fixed-effects. Year

fixed-effects are also included. However, D&E omit the therapeutic class effects because

they are highly collinear with competitor prices and because of the order of entry within

class. Since our model does not include either the competitor prices variable or the order

of entry within class variable (see Appendix C.1.), there will not be any collinear problems.

We have been able then to include the therapeutic class effects in our updated model.

First, we observe that launch prices are influenced by the propensity to launch70,

since the IMR presents a statistically significant coefficient; however, the launch delay

does not affect the drug pricing in either D&E or the updated model. Concerning other

explanatory factors affecting launch prices, according to the D&E model, the price of

established products in a country is significantly related to the launch price of new 70 D&E do not report the coefificent for the IMR in Table C.2, but its significance is reported in the text.


products. Also, the order in which a molecule is launched in each country-subclass

influences positively the launch price. However, as in the launch equation, we have not

been able to include these variables in our updated model as we explain in detail in

Appendix C.1.

In both models, launch prices increase with the minimum price previously set in

other high-price EU country. However, the effect on launch prices of the minimum price

previously set in high-price non-EU countries is different under each model, positive under

the D&E model and negative under ours. Only our updated model reports a significant

and positive effect from a minimum price set in low-price EU countries. Furthermore, if

prices in high-price EU countries are missing, it also affects positively the launch price.

This could indicate that countries setting the launch price with no reference in the EU may

pay high launch prices. We note that when we estimate with random effects, the effects

from low-price EU and high-price non-EU countries become statistically significant. Then,

we observe that the lowest price previously set in other low-price EU country affects

positively the launch price but more slightly than the effect from the high-price EU country

price. Interestingly, both the minimum price previously set in a high-price non-EU country

and the absence of a minimum price from high-price non-EU countries affect negatively

the launch price. This may indicate that spillover effects also occur either between EU and

non-EU countries or among non-EU countries. Furthermore, since the effect of the former

is greater, it may show that the net effect on the launch price of a previous launch in at

least one high-price non-EU country is positive. On the other hand, the absence of a

minimum price from low-price EU countries means countries do not have a reference from

this type of countries, and therefore it cannot be included into their reference basket,

which results in paying higher prices than if prices from low-price EU countries were

available. The results from the updated model with clustered standard errors support the

occurrence of spillover effects from high-price EU countries to low-price EU countries, and

therefore, that the ERP only concerns EU countries. However, the updated model with

random effects supports the suggestion that spillover effects occur in all directions.

In both models, per capita income does not seem to statistically affect the launch

price. Concerning the type of firms, under our updated model the drugs sold by a Solo

Licensee firm obtain lower launch prices than drugs sold by other types of firms. D&E do

not find any significant effect related to the type of firm.


Furthermore, regarding the product’s characteristics, the effect of strength, as

expected, is slightly statistically significant and positive, while the packsize affects

negatively the launch price. Under the updated model, strength seems to not have any

significant effect on the launch price. However, we have not included the variables related

to packsize in the updated model as we explain in detail in the Appendix C.1. When

introducing the administration route, the more robust result in both models is that

injectable drugs are statistically more expensive than other types such as oral solid

formulations.

In the launch price equation, both models found some launch price differences

among countries. The observed pattern is that all significant coefficients are negative;

Germany seems to present higher prices on average than the rest of the countries. The

random effects in the D&E model show that some country dummies are positive.

However, our updated model does not support the hypothesis that firms sell drugs at

single price in order to avoid spillovers effects.

As mentioned earlier, D&E do not include the therapeutic class fixed-effects due to

collinear problems with variables. Since our updated model does not include either the

competitor prices variable or the order of entry within class variable, there will not be any

collinear problems. We have been able to include the therapeutic class effects in our

updated model. We have taken the ATC-A (Alimentary tract and metabolism) as

reference and we have found some significant fixed-effects; however, in none of the

models year fixed-effects were significant.

3.4 Replicating Verniers et al. (2011)

In this section, the methodology conducted in a study published by Verniers et al.

in 2011 (Verniers et al., henceforth) (Verniers et al., 2011) is applied to our database to

compare whether results have changed due to the use of more recent data (2010 vs.

2008) and if the results are still robust using a different choice of the list of countries. They

applied their model to a set of a large set of countries, rich and poor, and we restrict our

application to developed countries.

3.4.1 The Verniers et al. model

Verniers et al. consider, on the one hand, the launch window of drug i in country j


(3.1)

(b)

( ), defined as the difference, in months, between the first worldwide launch and the

subsequent launch in the specific country j. The launch price is defined as the natural-

logarithm-transformed of the ex-manufacturer price at launch per gram of drug i in country

j ( ). Verniers et al. consider that censoring occurs for drug-country combinations for

which we do not observe a launch at the end of the observation window. Censoring time

(Cij) is defined as the time between the drug- and country-specific launch date and the

end of the observation period. Since the actual values of and are not observed

because right censoring is present, observed values are denoted by and such

that,

Moreover, we only observe the observations for which and thus .

The structural equations are:

LWij* 1LPij

* 2 (LPij*)2 'Zij1 uij1

LPij* 1LWij

* 2 (LWij*)2 'Zij 2 uij 2

where Zij1 and Zij1 are defined as additional explanatory variables. Zij1 comprises the

country size, the health expenditure per capita and the use of certain pricing policies such

as the ex-manufacturer price regulation, the profit control, the ERP, the internal RP and

the pharmaco-economic regulation. Also, it comprises the strength of patent protection,

the EMA and the firm’s home country variables. Additionally, it comprises the four

dimensions identified by Hofstede (Hofstede, 1984, Hofstede, 2001): uncertainty

avoidance, masculinity, individualism and power distance. The variable of competition and

the variable of summer are also comprised in this Zij1 variable. Zij2 includes the same

variables of Zij1 except from the summer and the EMA variable. However, it further

includes the inflation rate and the daily dosage (DDD).

LWij*

LPij*

LWij* LPij

*

LWij LPij

LWij* Cij LPij LPij

*

(a)

(3.2) (b)

(a)

* *

ij ij ij ij

ij ij

LW LW if LW C

LW C otherwise


Following Garen (1984) (Garen, 1984), Verniers et al. consider the launch window

and the launch price as endogenous variables. Therefore, the firm and the regulator may

both decide a launch window with the goal of influencing the launch price and select the

level of launch price also with the goal of influencing the launch window. The omitted

variables in the error terms of the launch window and launch price equations include non-

observable strategic variables used by the firm and the regulator to select the optimal

value for the launch window and launch price, respectively. These strategic variables

would be expected to correlate with the launch price and the launch window,

correspondingly.

Verniers et al., to account for the endogeneity between the launch prices and

launch window, estimate a system of simultaneous equations using a three-stage least

squares (3SLS) procedure, as in Bayus et al. (Bayus et al., 2007). Additionally, the

authors correct for right-censoring and selectivity using the procedure described in Vella

(Vella, 1993) or Wooldridge (Wooldridge, 2002). Random country effects are included in

the equations to account for the fact that there are repeated observations across

countries for most drugs.

On the one hand, to estimate the structural launch window equation, they first

estimate the reduced form of the launch price equation by a Tobit regression of the

second type (to account for the fact that we only observe prices if the drug has already

been launched). This launch price equation contains two variables that influence launch

price but not launch window, namely the defined DDD and the inflation rate, which serve

as instruments for the launch price in the launch window equation. The generalized

residuals of the reduced launch price equation are added to the launch window equation

as a correction term. However, we only use one instrument in our updated model, since

we have not been able to calculate DDDs in our database71. On the other hand, to

estimate the structural launch price equation, Verniers et al. first estimate the reduced

form of the launch window equation by a Tobit regression of the first type (to account for

right censoring). This launch window equation contains two variables that influence

launch window but not launch price, namely, summer and ema, which serve as

instruments for the launch window in the launch price equation. In this case, we have

71 For some molecules of our database, the DDD depends on patient characteristics. Therefore, a unique DDD for each molecule could not be used.


included both instruments in our model. The generalized residuals of the reduced launch

window equation as a correction term are added to the launch price equation.

3.4.2 Data

Verniers et al. collect data from the IMS Health database on drugs in 50 countries

(see Table C.3 in Appendix C.2) for 5 therapeutic classes, all of which experienced a

launch during the study period, 1994-2008. They have collected yearly data on outpatient

sales at ex-manufacturer prices. Price per gram in US dollars for each drug has been

calculated. To make drug prices comparable across countries, the drug prices in local

currencies were converted to US dollars using the currency conversion rate at launch.

We also use data from IMS Health database. However, we only consider the new

launch drugs in 20 developed countries for 11 therapeutic classes during the study period

2004-2010, all of them approved by the centralised procedure of the EMA. We have also

collected outpatient sales yearly at ex-manufacturer price. Since we have collected the

prices in euros, euros have been converted into US dollars applying the exchange rates

from the IMF. Finally, the drug price has been calculated as done by Verniers et al. Thus,

we also use the price per gram in US dollars for each in order to make results

comparable.

In Appendix C.2, we report the variable definitions. We distinguish among those

variables that we define as in Verniers et al and those that we cannot use or we define

differently. In Table C.4, we show the results from both models to be compared.

3.4.3 Comparison of results

3.4.3.1 Launch window equation

The sample of Verniers et al. is larger than our sample (1711 vs. 505) because

their study period is longer and the sample of countries is larger (50 vs. 20). As expected,

in both models, launch price affects negatively the launch window, i.e., the higher the

price a country pays for a drug, the shorter the delay the country will suffer. Also, both

models find a positive quadratic effect that offsets the above mentioned negative effect

(U-shaped effect). Furthermore, both models find a positive and significant coefficient of

the selectivity variable, suggesting endogeneity of prices in the launch equation. This


finding may indicate that health regulators act strategically in delaying market access for

expensive drugs, which is against the interests of the drug company.

Concerning the regulation variables, although researchers have not examined the

direct effect of profit control on launch window, Verniers et al. argue that it may slow

market access. However, in our updated model, profit control regulation seems to affect

the launch window negatively. Indeed, the only country applying profit control in our

dataset is the UK, which does not suffer particularly from long launch delays. ERP72 may

show counterintuitive effects (Hunter, 2005). First, when a country applies ERP, firms will

try to gain market access as early as possible to minimize the number of reference

countries. Second, ERP may push prices upward rather than downward. Typically,

regulators that seek early drug access are more willing to agree to higher prices. Thus,

the likelihood of a reference country having a high price is higher early in the life cycle

than it is later on, as we discussed in chapter 2. Consequently, the reference set of a

country is likely to contain a greater number of countries with high prices early in the life

cycle as compared to later in the life cycle. Our updated model shows a significant and

negative coefficient for this regulation variable; therefore, it confirms the hypothesis

proposed by Verniers et al. As said by Verniers et al., typically, therapeutic referencing

delays launch because the administrative procedure requires an examination of

therapeutic similarities, delaying market access. However, and contrary to the results

obtained by Verniers et al., our updated model reports a negative and significant

coefficient for this variable. Therefore, those countries using therapeutic reference pricing

experience shorter launch delays.

Pharmacoeconomic evidence, in addition to the clinical evidence required to gain

therapeutic approval from institutes such as the FDA (Food and Drug Administration) or

EMA, also requires evidence on the cost effectiveness of the drug in the local population,

and it must be submitted according to complicated administrative procedures. This

requirement often causes a delay in market access similar to therapeutic reference pricing

(Wilking et al., 2005) as we discusses in chapters 1 and 2. Results reported by our

updated model seem to support the results and the hypothesis proposed by Verniers et

al.

72 This variable is named by Verniers et al.as Cross-country reference pricing.


Concerning the strength of patent protection, it is known that high strength of

patent protects the firm from bio-equivalent price competition. Thus, a higher strength of

patent protection in a country may yield quicker access to drugs. In both models, those

countries with strong patent protection show shorter launch delays.

Other country characteristics considered are population size, health expenditure

per capita and dummies for the firm’s home country. The bargaining power given by the

population size is shown in our updated model since the effect of this variable is

significant and negative. Therefore, it supports the hypothesis and the results shown by

the Verniers et al. model. Regarding health expenditure per capita, Verniers et al. propose

that firms may be more eager to launch in countries with high health expenditures per

capita, as these countries may have a more favourable attitude towards new drugs.

However, higher health expenditures per capita could lower health regulators’ aspirations

to provide quick market access to new drugs (Comanor and Schweitzer, 2007). In both

models, the effect of health expenditure per capita on the launch window is significant and

positive, supporting the second idea proposed by Verniers et al. Concerning a firm’s

location, under the Verniers et al. model, firms with a greater familiarity with the home

market's therapeutic needs or health regulators' favouritism toward these firms may lead

to a faster launch (Kyle, 2006). In this case, both models support the hypothesis

described above.

A variable exclusively affecting the launch window is the dummy equal to one if the

country belongs to the EMA. According to Verniers et al., belonging to the EMA should

affect negatively the launch window. Although market access and price negotiations take

place at country level, the drug approval process in Europe is centralized. It is expected

that launch windows in EMA countries are shorter than those in non- EMA countries

because of differences in administrative efficiencies. Again, the Verniers et al. model

supports this hypothesis while our updated model shows a high significant positive effect

(countries not belonging to the EMA enjoy shorter launch delays than countries belonging

to the EMA). This could be due to the differences in the sample of countries. In our

database, only countries not belonging to the EMA are high-price countries. In the

Verniers et al. database there are a lot of low-price countries and very low-price countries.

Out of the four dimensions identified by Hofstede (Hofstede, 1984, Hofstede,

2001)) concerning a country’s national culture- uncertainty avoidance, masculinity,


individualism and power distance (defined in Appendix C.2) – only the masculinity and the

power distance present the same effects in both models and support the hypothesis

stated by Verniers et al., the more masculine the society is and the more bureaucratic it is

(higher power avoidance), the longer the launch window is. Verniers et al. expected and

showed that, on the one hand, low subjective health perceptions (high level of uncertainty

avoidance) may encourage health regulators to allow prompt access to new drugs and to

be less price sensitive, and on the other hand, countries showing a greater satisfaction

toward health care and spending more money on healthcare (high level of individualism)

enjoy shorter launch windows than collectivist countries. However, our updated model

presents the opposite effects.

Finally, the drug therapeutic fixed-effects (not reported) seem to be statistically

significant under the Verniers et al. model and in our updated model.

3.4.3.2 Launch price equation

Regarding the explanatory factors of the launch price equation, under our updated

model, the launch window does not seem to affect the launch price, while the Verniers et

al. model finds a significant and positive effect (negative for the quadratic term) as

expected under their hypothesis. Indeed, Verniers et al. propose an inverted U-shaped

effect of launch window on launch price in which launch price is highest for moderate

launch windows. For these moderate launch windows, a firm can still make money under

patent protection if the price is high enough to make up for local market entry

expenditures. For very short launch windows, a firm will accept a lower launch price more

easily because the drug enjoys a full lifetime under patent protection, so the firm can

recover R&D expenditure and gains resources for international market access

immediately. For very long launch windows, a firm and a health regulator will agree more

easily on a relatively low launch price as a prelude to generic competition.

Regarding the regulation variables, neither of the models find any significant effect

on launch prices, except from the strength of patent protection, where only the Verniers et

al. model finds a negative and significant effect as expected: stronger patent protection

may impose a downward pressure on launch prices because pharmaceutical firms can be

more lenient on prices if there is sufficient time left under patent protection to recover

R&D expenditure.


Among other country characteristics considered such as population size, health

expenditure per capita and the firm home’s country, only the firm home’s country seems

to be statistically significant and positive, but only under the Verniers et al. model,

supporting their hypothesis. As previously mentioned, a greater familiarity with the home

market's therapeutic needs or health regulators' favouritism toward these firms may lead

to a higher launch price (Wagner and McCarthy, 2004). The drug therapeutic fixed-effects

(not reported) are not statistically significant.

3.5 New Pricing and Launching Model (NPLM)

3.5.1 The Model

We estimate the launch delay and the relative launch price equations separately,

and each of them is estimatedfor retail and for hospital distribution channels. We have

also tried to estimate a system of both equations to account for endogeneity. However,

the available instrument of the relative launch price equation is weak73.

We use a parametric duration model of the hazard of launching in time t, given the

observed explanatory variables, with right-censored data to model the launch delay of the

molecule i in country j, which is defined as the time elapsed in months from the first global

launch of the molecule i and its launch in country j. We have specified the shape of the

hazard rate, i.e. its time-dependency, with a Weibull distribution that assumes a

monotonic hazard with respect to time. Since we have not been able to observe all

variables affecting the launch delay, we have controlled for the unobserved heterogeinity

introducing a gamma frailty distribution for the random error term, The model selection

has followed the method proposed by Kiefer (Kiefer, 1988) (see Appendix C.4). We

estimate a right-censored model since all the drugs in our data set were launched

between January 2004 and December 2010; however, not all drugs had been launched in

all 20 countries by the end of our observed period. Therefore, our data contain right-

censored observations. Our parametric duration model for the hazard of launching does

not allow the use of time-varying covariates74; instead we have used the data collected in

73 We selected the variable inflation as the instrument for the launch price in the launch equation. The correlation between inflation and launch price was weak (0.02), therefore, we should not use it as instrument.

74 Since we reject the null hypothesis of the log-rank test, then the assumption of proportional hazard is not satisfied, we should not neither incorporate time-varying nor use the standard Cox regression. Furthermore, the extended Cox model allows incorporating time-varying covariates but we should not use it since it does not allow incorporating censoring.


the base year75. Thus, we have:

where , the subindex i=molecule, j=country, h is the hazard rate of launching, Xij

are the covariates, the covariates’ parameters, t the time elapsed until the launch of

molecule i occurs in country j, p the shape parameter76, U a random variable and the

variance of the frailty77 (Keele, 2007, Jenkins, 2008). The covariates Xij are the relative

launch price at molecule level, the logarithm of the country size (population), the logarithm

of the public health expenditure per capita, the logarithm of the pharmaceutical

expenditure per capita, the dummy for the firm’s headquarters location in the launching

country, the dummy for belonging to the EMA and therapeutic fixed-effects at ATC-1 level.

These variables are defined in detail in Appendix C.3.

Regarding the relative launch price equation, we use OLS with molecule-

presentation-clustered standard errors to model the log of the relative launch price of

molecule i, product k in country j at the time t, conditional on launching. The relative

launch price is defined as the price ratio between the launch price of molecule i, product k

in country j at the time t, and the launch price of molecule i, product k in country g at the

first global launch time 0. To account for unobserved molecule characteristics, we also

report results from a GLS (Generalized Least Squares) random effects estimator. To

account for possible selection bias produced by the correlation between the propensity to

launch and the launch price, we also estimate a Heckman selection model with a first-

stage probit regression (Heckman, 1979). Then, we have the probit selection equation:

75 For these covariates, such as country population, the GDP per capita, health and pharmaceutical expenditure per capita variables, we use the data for the base year (2004). These covariates are in the model to control for differences in country sizes, wealth and expenditure, which are well represented with the data collected for the base year 2004.

76 The shape parameter p determines whether the hazard is increasing, decreasing, or constant over time.

77 By testing the hypothesis = 0 using a likelihood ratio test, we determine whether we need to worry about unobserved heterogeneity.

h(t, X) p(t)p1[U]

ij eXij

(3.3)


P(Lijkt 1) 0 ijkt Xijkt Uijkt

where the subindex i=molecule, j=country, k=product and t=year. Then, Lijkt is a

dummy variable equal to 1 if the molecule i has been launched in country j at time t.

Then, Xijkt is the vector of explanatory variables and Uijkt is the random term.

The explanatory variables in this model are: the launch delay of molecule i in

country j, the logarithm of the GDP per capita of country j at time t, country size

(population) of country j at time t, public heath expenditure per capita of country j at time t,

pharmaceutical expenditure per capita of country j at time t, the dummy for the use of

ERP policy in country j, the firm’s home country=1 if the headquarters’ firm of launching

the molecule i is located in country j, EMA=1 if the country j belongs to the EMA at time t

and the ATC fixed-effects of molecule i. These explanatory variables are defined in more

detail in Appendix C.3. See results in Table C.6 Appendix C.3.

The relative launch price equation is:

RPijkt Pijkt

Pizg0

0 'Zijkt Vijkt

where the subindex i=molecule, k=product, j=country, g=country where first global

launched occurred and t=year. Then, is the relative launch price, is the launch

price in country j at time t, and is the launch price in country g at the first global launch

time 078. Then, is the vector of explanatory variables and is the random term.

The explanatory variables in this model are the same that in the probit selection

model besides the square of the launch delay, the year fixed-effect and the IMR from the

selection probit equation (see Appendix C.3. for definition).

78 When two countries or more experience the first global launch, i.e., country g is represented by two or more countries, then is calculated as the weighted-volume average of the price of each country g.

RPijkt Pijkt

Pijko

Zijkt Vijkt

Pijko

(3.5)

(3.4)


We mainly focus on the effect of launch price on the launch delay and vice versa,

and on the effect of the application of ERP on both the launch price and the launch delay.

We expect that countries paying high prices experience short launch delays (see Chapter

1, sections 2.1 and 3.3.1) and in turn, that countries suffering from launch delays pay

lower relative launch prices (see Chapter 1, sections 2.1 and 3.2.1). We also expect that

applying ERP is effective, so countries applying this pricing policy suffer from longer delay

launch (see Chapter 1 sections 2.2 and 3.3.3) and pay lower relative launch prices (see

Chapter 1, sections 2.2 and 3.2.3)

To show this, we have controlled for country characteristics regarding the

bargaining power of the country, such as the country size and the country purchasing

power (GDP) (see Chapter 1 sections 3.2.4 and 3.3.4). We expect that countries with a

large population experience short launch delays and pay lower prices for new drugs, while

countries with high-incomes have quicker market access but are expected to pay higher

prices. Other country features concern health investment and the attitude of firms towards

countries with a high public health and pharmaceutical expenditure per capita. We expect

that the increase in these variables may shorten the launch delay and mean countries pay

high prices. Since countries with high public health expenditure should be more worried

about the availability of drugs in their market, they are expected to pay higher prices and

try to have the new drug available as soon as possible. Similarly, countries with a high

pharmaceutical expenditure per capita are willing to pay higher prices and therefore, firms

are interested in launching earlier in those countries.

Also, we have controlled for the firm’s headquarters location. We expect that

countries hosting the firm’s headquarter experience shorter launch delays and pay higher

prices. We think that hosting the firm’s headquarter could generate other types of profits

for the country that mean the country will have the drug available in its market as soon as

possible and at a higher relative launch price (e.g. employment, incomes from taxes, etc.

; see Chapter 1, sections 2.5 and 3.3.5). Furthermore, since our database contains

countries belonging to the EMA and other countries outside of it, and given that drug

approval processes take varying times depending on the organization (the FDA, the EMA,

etc.), we expect that the EMA dummy variable affects the launch delay (see Chapter 1

section 3.3.3)


Additionally, we have also controlled for the therapeutic class fixed-effects and the

time trend; however, the country fixed effects have not been included because the

variable of greatest interest, the one measuring the use of price controls, has little within

country variation.

3.5.2 Data

We used the database described above (see section 3.1 of this chapter). The price

per SU for each product k was calculated on a yearly basis as the ratio of total revenues

to SU sold. Two products k are considered the same as they present the same quantity of

SU and the same administration route. For each molecule-country providing two or more

identical products with different packsizes79, the volume-weighted average price was

calculated. Also, we note that the USA sales were collected at inpatient and outpatient

level through different sales channels (e.g. drugstores, foodstores, mail service, etc. for

inpatient sales, and hospital, non-federal hospitals, home health care, etc. for

outpatients). Furthermore, we note that sales from Denmark, the Netherlands and

Sweden were jointly collected by IMS (inpatient and outpatient). The relative launch price

for the whole database was calculated as noted in the above section 3.5.1. Besides the

variables already defined in the previous section 3.5.1, we used additional variables such

as the relative launch price and the pharmaceutical expenditure per capita. Also, we

introduce country size and the GDP per capita as additional explanatory variables in the

same model. In Appendix C.3 the variable definitions and classifications are reported in

Table C.3.

3.5.3 Results

3.5.3.1 Launch delay equation

Now, we present the results from the analysis of the duration of the launch delay

from the molecule global launch. This analysis has been conducted by the estimation of a

Weibull model controlling for unobserved heterogeinity. Results are shown in Table 3.6.

In the analysis of the retail sales, the model reports a statistically significant hazard

ratio slightly lower than unity for the relative launch price of the molecule. Higher relative

79 E.g. Tablets 150MG 28 and Tablets 150MG 56.


launch prices affect negatively the probability of having the molecule launched. This

unexpected result indicates that countries that experience longer launch delays pay

Table 3.6. Launch delay equation of the NPLM

Retail Hospital Variables Hazard Ratio Hazard Ratio

Relative Price (at molecule level) 0.9934*** 0.9949*** [0.0017] [0.0014] Log of Country size in 2004 (population) 0.8754** 0.8746** [0.0492] [0.0467] Log of GDP per capita in 2004 4.2414** 3.2867** [2.9975] [2.1143] Log of Health Expenditure per capita in 2004 0.4584 0.6324 [0.2581] [0.3465] Log of Pharmaceutical Expenditure pc in 2004 0.8243 0.5657 [0.3031] [0.2016] IRP 0.3838*** 0.3712*** [0.0714] [0.0659] Firm’s home country 1.0816 1.2849 [0.3358] [0.3963] EMA 1.6443*** 1.5136*** [0.2479] [0.1955] ATC-1 B 2.5813 2.9381 [0.6169] [0.7006] C 0.5827 0.6983 [0.1693] [0.2024] D 2.6959 2.9717 [0.8045] [0.9475] G 0.3353 0.3347 [0.1011] [0.1037] J 0.7837 0.6636 [0.1686] [0.1343] L 0.6780 0.5443 [0.1329] [0.1032] M 0.4678 0.2891 [0.1683] [0.0972] N 1.4022 1.2241 [0.2861] [0.2531] R 13.2524 13.9722 [7.1829] [7.7269] S 0.6492 0.9355 [0.2343] [0.3175] T 0.6209 0.8516 [0.3703] [0.3427] V 0.8353 0.7506 [0.2777] [0.2233] Observations 732 921 AIC 38.1443 37.9427 Significance(sign.)levels(two‐sided):*:p<0.10;**:p<0.05;***:p<0.01.;[]:standarderror;n.r.:no‐reported;‐:no‐included;RC:referencecategory.SeedefinitionofvariablesinAppendixC.3.


higher prices than countries with a shorter market access. However, we should remark

that the extent of this effect is negligible and ultimately we may interpret that different

relative launch prices do not yield large differences in launch delays. Furthermore, as

expected, the use of ERP policy generates a lower probability of launch. This may

indicate that firms try to avoid the spillover effects by delaying launch in countries rather

than using ERP. As some of the countries applying ERP policy are potential exporter

countries, this could be another reason why firms delay launch in these countries.

Regarding other country characteristics, it is observed that larger countries suffer

from longer launch delays. This unexpected result shows that the bargaining power due to

country size (population size) has no effect on a quick market access. The factors that

seem to have an important positive effect on the probability of launching a molecule are

the GDP per capita and belonging to the EMA. However, other indicators such as public

health or pharmaceutical expenditure do not affect the probability of launch. Even, the

firms with their headquarters in the launching country do not launch in it first instead of in

others.

At molecule level, the model reports statistically significant differences among

different ATC-1 classes, taking the A-class as the reference category.

When we analyse hospital sales, we do not observe major changes with respect to

retail sales. Factors influencing the launch delay and their signs remain the same. We

should just mention that GDP per capita is now statistically significant at 10% (very close

to the 5%). There is no huge change in the extent of the influence of the significant

covariates. As in the analysis of retail sales, the model reports statistically significant

differences at ATC-1 level.

3.5.3.2 Relative launch price equation

When we analyse the retail sales, the relative launch price is conditioned by the

launching, since the IMR80 coefficient is statistically significant. Regarding the variables on

80 Following Heckman HECKMAN, J. J. 1979. Sample selection bias as a specification error. Econometrica: Journal of the econometric society, 153-161., the IMR of molecule i in country j and time t, IMRijt , is calculated using the score of

launching from the probit regression as where is the standard Normal density and

is the standard Normal distribution function.

X ' IMRijt (X ')

(X ')() ()


which we focus our study, we note that the launch delay seems to have no statistically

significant effect on the relative launch price. This result may indicate that the firms have

prioritized to avoid the spillover effects from ERP and PT as opposed to making profits

from some late sales at lower prices. Indeed, it seems that countries no longer benefit

from lower prices in exchange for having the product available with a certain delay.

Furthermore, we observe that the use of ERP does not affect significantly the relative

launch price. Particularly, it seems that countries applying ERP policy do not pay lower

relative launch prices than countries that do not use it. This unexpected result shows that

the use of ERP is not effective, either because some countries may directly take the

reference price, or because other countries may not apply ultimately ERP as a unique

criterion, or because firms do not sell at that price (see Chapter 2 section 2.1).


Table 3.7. Relative launch price equation of the NPLM

Retail Hospital OLS w/ Robust Clustered SEs Normal Random Effects OLS w/ Robust Clustered SEs Normal Random Effects Delay -0.0722 0.0053 -0.0726 0.0137 [0.1296] [0.1199] [0.1695] [0.1490] Delay*delay 0.0023 0.0008 0.0000 1.41e-06 [0.0021] [0.0008] [0.0000] [6.87e-06] Log of Country size (population) -8.6599* -5.8246*** -10.1434** -4.2420* [4.7556] [1.9964] [4.4531] [2.4337] Log of GDP per capita -9.5648 -2.6522 7.7304 1.7151 [9.7722] [16.5960] [12.6634] [19.4184] Log of Health Expenditure per capita 24.6213* 14.3926*** 18.0517* 4.1666 [13.9677] [12.7490] [9.8605] [14.9830] Log of Pharmaceutical Expenditure pc 50.7958* 37.3478 70.6921** 29.2789 [27.4052] [13.0590] [31.7399] [18.2923] IRP -5.8320 -2.7864 -8.9012 -4.8740 [4.1272] [3.3182] [6.0281] [4.3355] Firm’s home country -4.1632 -4.1037 -19.7583* -5.1026 [3.5854] [5.8000] [11.5931] [8.3115] Year 2004 RC RC RC RC RC RC RC RC 2005 0.6371 -2.8228 3.6965 -1.4028 [2.3904] [7.3313] [3.3774] [9.1615] 2006 -0.9426 -4.1618 4.4947 1.0742 [2.8771] [8.1672] [4.3652] [10.8644] 2007 -3.6626 -6.0497 -3.1937 -0.3632 [5.1028] [8.4271] [5.5596] [11.4744] 2008 -4.4736 -8.0594 -1.3471 1.0891 [5.0712] [8.8661] [5.6821] [12.3694] 2009 -2.1013 -8.1843 14.1190 1.3525 [3.6430] [9.2742] [11.2567] [13.2468] 2010 -16.3579 -16.5441 -7.6186 -4.3132 [10.4757] [10.2073] [9.7540] [14.6237] IMR 60.7956* 45.6041*** 80.9075** 25.3029 [33.2356] [16.1445] [36.1181] [23.8379] Constant -19.4495 -303.1349** -648.5746* -210.4026 [87.4215] [140.4957] [329.3245] [207.2856] Observations 1334 1334 1369 1369


Retail Hospital OLS w/ Robust Clustered SEs Normal Random Effects OLS w/ Robust Clustered SEs Normal Random Effects Number of Molecule-presentation-level Clusters

69 69 70 70

R-squared 0.1776 0.1744 0.1952 0.1952 Significance(sign.)levels(two‐sided):*:p<0.10;**:p<0.05;***:p<0.01.;[]:standarderror;n.r.:no‐reported;‐:no‐included;RC:referencecategory.SeedefinitionofvariablesinAppendixC.3.


Furthermore, other country characteristics such as pharmaceutical and health

public expenditure per capita do seem to affect significantly the relative launch price.

Indeed, as expected, countries with high pharmaceutical and public health expenditures

per capita pay higher relative launch prices. In addition, the results show that countries

with a high bargaining power, since they have a large population, pay lower relative

launch prices on average (see Chapter 2 section 2.1). As we account for unobserved

molecule characteristics, we also report results from a GLS random effects estimator. The

results slightly change when we estimate this alternative specification. Particularly, only

the country size and public health expenditure per capita remain as significant factors

influencing the relative launch prices.

When we analyse the hospital sales, we observe that, compared to the retail

market, significant results remain. Furthermore, in this analysis, the firm’s headquarters’

location has a slightly significant and negative effect on the relative launch price.

Therefore, drugs launched by firms with their headquarters in the launching country set

lower prices than drugs launched by firms with their headquarters outside the launching

country. This unexpected effect will be discussed later on in this chapter, and it can be

compared with the results reported in the literature in Chapter 1 Section 3.2.5. Similar to

the retail market, when we report the results from a GLS random effects estimator, the

only significant effect that remains is the country size, the rest of variables affecting the

relative launch price become insignificant. Even, the IMR does not affect significantly the

relative launch price, only being significant in the retail market for both specifications and

in the hospital market for the OLS molecule-clustered estimate.

3.5.4 Discussion

Our contribution to the previous literature analysed in Chapter 1, sections 2 and 3,

firstly consists of the analysis of the database at presentation level, the analysis of the

relative launch price as endogenous variable in the launch price equation, the study of the

launch delay as a duration time variable and the analysis of the inpatients market. In this

chapter, we have carried out an analysis of the trade-off between pricing and launching

and the impact of ERP policy on both pricing and launching.

In this regard, we have observed that the launch delay does not significantly affect

the relative launch price; however, the relative launch price does affect launch delay, but

the extent of the influence is quite low. In addition, the results show that the use of ERP


makes countries experience longer launch delays but does not lead to paying lower

relative launch prices. These results may have several implications on the bargaining

process. Indeed, we may think that firms do not want to play the game in which, countries

reject a firm’s offer knowing that over the time they will obtain lower prices. Besides, we

observe that firms delay launches in countries using ERP policy; however, these countries

do not necessarily pay lower prices. This last result may indicate that ERP policy is not

effective in “pricing terms” but it is in “launching terms”. It seems that firms do not accept

lower prices in exchange for delaying launches from countries applying ERP. These

results may suggest that firms basically delay launches because countries probably

cannot afford to have the product available straight from the global launch, and even not

having the product. In contrast to previous literature, where firms sometimes delay launch

to avoid spillover effect, in our study, we show that firms may use a more aggressive

strategy, which does not allow countries to have the products available with a launch

delay in exchange for paying lower relative launch prices. Under this strategy, firms

would avoid the spillover effects from ERP policy and PT, though they would lose profits

from sales in those countries where the molecule is not ultimately launched.

Furthermore, among other country characteristics, we observe that the bargaining

power of country size is effective to obtain lower prices; however, this country

characteristic does not seem to be an influencing factor on achieving shorter launch

delays, even more, unexpectedly, countries with a large country size find lower

probabilities to have a product launched. In the same line, we have observed that GDP

per capita does not affect the relative launch price; however, other country characteristics,

more specifically ones affecting pharmaceutical consumption, such as pharmaceutical

and health public expenditure per capita, affect positively the relative launch price. Exactly

the opposite effect occurs for the launch delay. The pharmaceutical and the public health

expenditure do not seem to result in countries experiencing shorter launch delays.

Indeed, what does make countries have products available in the short-term is a high

level of wealth per capita. We may say that wealthy countries have the products available

in the short-term, and the countries that ultimately pay high relative launch prices are

those that allocate large budgets to public health and the pharmaceutical expenditure.

On the basis of the results, firms neither make discounts nor launch in the short-

term in those countries where they have their headquarters. Only in the hospital market

do we observe that countries obtain lower prices from this type of firms than from foreign


ones. So far, the previous literature has either not found any significant price premiums

for local firms or has found a positive significant and expected effect. Note that these

studies collected data not only in high-income but also in low-income countries where

firms’ headquarters are not usually located, therefore, the positive effect found could be

due not exclusively to the firm’s location but also to the country’s wealth. Furthermore,

countries belonging to the EMA enjoy shorter market access on average than countries

outside of the EMA regime.

3.6 Conclusions

Conclusions from replicating the D&E model

The updated model presents some differences from the D&E model. We must note

that our sample starts in 2004, immediately after the D&E sample finishes. Also, the lists

of countries and products are not exactly the same in both models. This may justify

differences in results. The D&E model is robust among different specifications, while our

updated model presents some alterations. The most remarkable differences in results

between the D&E model and the updated one concern the spillover effects and the effects

of the type of firms.

The updated model finds that the number of low-price EU countries affects

negatively the propensity of launch, which may confirm that low-price countries are

suffering longer launch delays, while the D&E model does not find any significant effect.

Furthermore, the updated model finds that the minimum price set in the low-price EU

affects positively the launch price, which may show that spillover effects also exist among

low-price countries, while the D&E model does not. Moreover, the minimum price set in

high-price non-EU countries presents a negative effect on the launch price (positive in the

D&E model) which may indicate that spillover effects occur among EU and high-price

non-EU countries. To have no reference prices from low-price EU and high-price non-EU

countries only seems to have effect in the updated model. No references from low-price

EU countries yields a positive effect due to two different situations, either that spillover

also occurs among low-price countries, or that being first means paying higher prices. No

references from high-price non-EU countries may indicate either that these countries are

also taken as reference by EU countries or that launch prices in the EU could be higher

than the launch price out of the EU.


Furthermore, according to the D&E model, Local Corporations enjoy a higher

propensity to launch, but our model does not find any significant effect from this

characteristic. The other way around occurs for the launch price, where the D&E model

does not indicate any significant effect on launch prices while the updated model shows

that Solo Licensee firms obtain lower launch prices. Both models present significant

country-fixed effects in the launch and launch price equation.

Conclusions from replicating the Verniers et al. model

The same data treatment and methodology conducted by Verniers et al. have been

implemented for our database. Differences in the list of countries and drugs studied, and

in the time period covered may justify some of the differences in the results above.

Verniers et al. find evidence of endogeneity of both launch prices and launch delays. We

only find the effect in a single way; launch delay is negatively affected by launch prices

but not the other way around.

The effects from some regulatory policies do not seem to coincide. Some

regulatory policies traditionally positively affecting the launch window, such as profit

control or the therapeutic reference pricing, are not significant in the updated model.

However, both models show that the use of pharmaco-economic evidence regulation

leads to longer launch delays. Also, in both models, the stronger the strength of patent is,

the shorter launch delays are. Only our updated model presents a significant and

expected effect for the use of ERP regulation. On the other hand, the regulatory policies

do not show significant effects on the launch price under any models, except for the

strength of patent protection; stronger patent protection imposes a downward pressure on

launch prices.

Results from other country characteristics like population size and health

expenditure per capita, when statistically significant (only in the launch window equation)

are concordant in both models. Among other country characteristics, countries that host a

firm’s headquarters of the firm launching the drug experience shorter launch delays under

both models. This effect is not significant in the launch price equation under the updated

model but positive under the Verniers et al. model, supporting their hypothesis that firms

settled in the launch country enjoy higher prices. Furthermore, belonging to the EMA

shows the opposite effects. According to Verniers et al. countries belonging to the EMA


show shorter launch delays, however, our updated model reports that these countries

experience longer launch delays than in countries outside of the EMA.

Finally, both models present significant therapeutic class fixed-effects. However,

the four dimensions identified by Hofstede (Hofstede, 1984, Hofstede, 2001) concerning a

country’s national culture present different effects in both models.

Conclusions of the NPLM

Under the NPLM, the pricing and launching seem to be no longer related to each

other. Differences exist in prices across countries but not due to the launch delay. Firms

do not accept lower prices in exchange for delaying launches, even from countries

applying ERP policy, therefore, ERP policy seems to not be effective in “pricing terms” but

is in “launching terms”. These results may lead to several implications in the bargaining

process. We suggest that firms basically delay launches because countries probably

cannot afford to have the product available straight from the global launch, and ultimately

end up not having the product launched. While the firms often delay launch to avoid

spillover effects, under our study, we show that the firms may conduct a more aggressive

strategy that does not allow countries to pay lower prices in exchange for experiencing

longer launch delays. Under this strategy the firms would avoid the spillover effects from

IRP policy and PT, but they would also lose profits from sales in countries where the

molecule is not ultimately launched.

Regarding other country characteristics, our study shows that wealthy countries

have the products available in a shorter period, but the countries that ultimately pay high

relative launch prices are those that allocate large budgets to public health and

pharmaceutical expenditure. Countries belonging to the EMA seem to enjoy shorter

launch delays than the countries outside of it; however, there are no significant price

differences between countries under the EMA regime and countries outside of it.

In general, the results in the retail market and the hospital market do not show

huge differences, but we highlight the firm with headquarters in the launching country

bargains lower prices with the country concerned just in the hospital market.

Conclusions and further research

Our systematic review shows that demographic and income country features, and

regulation regimes, seem to be the most important factors affecting drug pricing and

launching. However, price regulation can undermine the effects of these important factors.

Drug characteristics like strength, packsize and presentation forms are significantly

related to the price. Also, supported by previous studies, belonging to the EMA’s

therapeutic category robustly affects the launch delay and the launch price. The

therapeutic value is shown in the previous literature as a robust factor-influencing drug

pricing and launching. Additionally, firm location turns out to be an important factor for the

launching decision; however, price premiums due to headquarters location appear

ambiguous.

When replicating the D&E model with more recent data, we find some new

patterns compared with the D&E results concerning the spillover effects and the type of

firms. Therefore, we determine that low-price countries suffer longer launch delays and

spillover effects also exist among low-price countries: spillover effects also occur

between EU and high-price non-EU countries. Furthermore, Local Corporations do not

have a higher propensity to launch any longer; however Solo Licensee firms do obtain,

nowadays, lower launch prices.

Replicating Verniers et al. with more recent data and a different list of countries

yields new outcomes. The most important outcome is that the endogeneity of both launch

prices and launch delays found by Verniers et al. is no longer found; only the launch delay

is negatively affected by the launch price. The effects of some regulatory policies on the

launch delay do not generally seem to coincide. Again, similar to the D&E, in our updated

model, firm location loses its effect on drug launching and pricing.

One of the most important conclusions of this doctoral thesis is that, in contrast to

previous models, pricing and launching seem to be no longer related to each other.


Differences in prices exist across countries but not due to the launch delay. Firms do not

accept lower prices in exchange for delaying launches, even from countries applying ERP

policies, therefore, ERP seems to not be effective in “pricing terms” but it is in “launching

terms”. These results may hold several implications for the bargaining process. We

suggest that firms basically delay launches because countries probably cannot afford to

have the product available straight from the global launch, or ultimately not having the

product launched. While firms used to delay launch to avoid spillover effects, in our study,

we show that firms conduct a more aggressive strategy that does not allow countries to

pay lower prices in exchange for experiencing longer launch delays. Under this strategy,

firms would avoid the spillover effects from ERP and PT, but they would also lose profits

from sales in countries where the molecule is not ultimately launched.

Regarding other country characteristics, being a large country helps to have more

rapid market access and obtain lower prices. Furthermore, wealthy countries have the

products available within a shorter period, but the countries that ultimately pay high

relative launch prices are those that allocate large budgets to public health and

pharmaceutical expenditure. Firm location no longer affects the price and neither does the

launch delay. Countries belonging to the EMA seem to enjoy shorter launch delays than

the countries outside of it; however, there are no significant price differences between

countries inside the EMA’s regime and countries outside of it.

In general, the results in the retail market and the hospital market do not show

huge differences, but we highlight the firm with headquarters in the launching country

bargains lower prices with the country concerned just in the hospital market.

In view of the overall perspective concerning the main factors influencing launch

prices and launch of new drugs based on theoretical studies, we mainly distinguish two

types of factors. Firstly, we have observed factors that directly affect drug pricing and

launching, such as the presence of PT, the firm’s characteristics and the regulation pricing

policies, for example ERP, internal RP or MES + PC. Secondly, our review shows other

determinants that not only impact directly but also indirectly, affecting the measures of the

first types of factors that influence drug pricing and launching, such as country size and

the level of co-payments.

According to our theoretical model, given the optimal country launch sequence, we

conclude that the smaller the population size is, the more attractive the use of ERP will

Conclusions and further research 115

be, since the unitary cost of CEA increases. Note that ERP does not require any

investment, however CEA does. Therefore, the use of ERP is helpful to relatively small

countries compared to the use of CEA. This result confirms some statements on this

issue that have not been previously shown. ERP is a low-cost pricing policy; however, we

have now more information about why countries apply this type of pricing policy fully

aware that it may be not fair.

We also conclude that the optimal country launch sequence depends on the

relative prices and the relative country sizes. There is a trade-off between price and

volume, which affects the country launch sequence. In addition, the relative price depends

on the countries’ pricing policies, ERP and CEA, and subsequently, on the ERP formula.

Particularly, a country is better off applying ERP instead of CEA, if the difference between

the international reference price and the expected price of country i under CEA is not

higher than the unitary cost of CEA. This result is affected by the delay cost if only one of

the pricing policies is applied with delay.

From the perspective of the regulator, concerning the application of ERP, the

previous theoretical literature recommends only small countries to engage in ERP and/or

apply ERP based on prices in large countries (or large group of countries); the same

applies if one substitutes “large country” by “small co-payment country” and vice versa.

Also, a minimum RP level is recommended to avoid major increases in price level with

respect to the average RP. Furthermore, an MES policy together with a PC should be

applied when the welfare loss due to high-type drug buyers is not large enough to

outweigh the welfare gained due to the low-type drug.

From the perspective of the firm, the literature recommends that the loss of income

coming from PT should be taken into account and the firm should set a higher price than

without PT. However, since ERP is widely used by countries and PT does exist, the

markets are inseparable. Therefore, the highest drug price is not always the best option

for the firm and the lowest drug price is not always the best option in a given country.

What may have been an optimal pricing strategy in a single country is no longer optimal

when considering ERP and PT.

Our theoretical model shows that there is a trade-off between prices and volumes

affecting the country launch sequence. Also, it provides information about ERP, an issue


widely discussed in the literature. In any case, we accept that a greater variety of factors

may exist than already used in the model, which affect the bargaining process. Further

theoretical models may need to be designed, for example, ERP formulas may be other

than the average or the minimum, firms may offer one single price, etc. Besides, other

factors such as the population age structure or the lobbying activity of the pharmaceutical

industry may also be taken into account.

Despite the limitations caused by the lack of data, this is the first study that

indicates that pricing and launching a medicine are no longer two inseparable issues.

However, we consider that further analysis is needed concerning the use of larger

samples at country and drug level, in order to achieve more accurate comparisons.

Particularly, in the analysis of ERP, further research should take into account the

interdependencies that occur between countries due to ERP. We have controlled for this

pricing policy through a dummy variable, but richer insights could be obtained by

collecting data on the reference basket of each country that applies it.

Behind the idea that pricing and launching are no longer related to each other,

prices may be converging to a single price, at least, under the EMA countries. An

interesting avenue of future research could be the analysis of price convergence through

time series analysis, particularly, based on the cointegration theory. This work requires

long-term data of pricing and launching. This further study may demonstrate some of the

hypotheses stated in this thesis about the future lines of pharmaceutical industry strategy.

A limitation of the data concerning the hospital market is that we cannot take into

account price discounts due to the volume purchased. Another interesting line of

research would be to test the benefits of centralized purchasing. There are countries

traditionally applying this type of purchasing through regional and national hospital bodies,

such as Denmark or Norway. Particularly, in Spain, a new mechanism has been

implemented in the Public Sector Procurement Act, which was introduced in 2013. A

centralized frame agreement for drug procurement has been established. This new

mechanism is a further step within a major health sector reform to reduce pharmaceutical

expenditure and ensure the sustainability of the Spanish Health System. In this context,

interrupted time series analysis may be applied to test if this new drug purchasing policy

is being effective. Again, gathering data directly from hospitals seems to be the toughest

issue.

A. Appendix A

A.1. Method of Review

A.1.1 Search strategy

The following main databases were searched in April 2012, using combinations of key

terms in both UK and US English and addressing the period from January 1995 to April

2012: PubMed, EconLit and Web of Knowledge. The search was then complemented by

another, using the same keywords, in the following scientific centres advised by experts:

the NBER (The National Bureau of Economic Research), the University of York, the

CHEPA (Centre for Health Economics and Policy Analysis), the LSE (London School of

Economics) Health, the CRES (Centre de Recerca en Economia i Salut) at University

Pompeu Fabra, and the School of Economics at Erasmus University Rotterdam. Finally,

some documents referenced by experts were also included. We searched separately for

“ ("pharmaceutical* pric*") and (new or launch* or patent), ("drug* pric*") and [new or

launch* or patent) and ("medicine* pric*") and (new or launch* or patent).

A.1.2 Selection and Exclusion Criteria

The references found in the search were systematically assessed by reviewing the

title, abstract and publication type to identify relevant articles. All relevant articles meeting

the inclusion criteria were retrieved, following a two-stage selection strategy.

In the first stage, selection criteria were applied to all the studies under review. In

the second stage, the studies selected in the first stage were individually screened to

evaluate the theoretical and empirical contributions made in each case. In the first stage,

we only included studies satisfying the criteria stated in Figure A.1. Each article was

sequentially evaluated against the four criteria, from the first to the fourth: i) to be an


original article, ii) to be published in scientific journal, iii) to be written in English language

and iv) to be focused on patented drug pricing or launching. As soon as a criterion was

not met, the article was excluded.

In the second stage, the full text of each article selected was retrieved. We

selected only original theoretical contributions developing new analytical insights and

original suggestions, and empirical papers developing a causal model focusing on the

price of drug or on a measure of drug launch as dependent variable accounting for a

cross-country sample

A.1.3 Search results

The selection process and the number of articles excluded and retained at each

stage are summarized in the PRISMA (Preferred Reporting Items for Systematic reviews

and Meta-Analyses) in Figure A.1 (Moher et al., 2009). The search delivered a total of

1183 articles. Then, we applied the selection and exclusion criteria in first and second

stage as above explained. A total of 65 articles met all four criteria in the first stage of our

selection strategy. These 65 studies were evaluated for inclusion in the second stage.

Finally, 22 full articles were retrieved. Twelve of these had a theoretical focus and 17

were empirical.

Appendix A 119

Figure A.1 Flow diagram of literature screening process


Table A.1 Overview of theoretical studies

Reference Focus variable Focus effects Number of countries

Atella et al. (2008) Price Drug and Regulation 2

Bardey et al. (2010) Price Competition and Regulation Undetermined

Cabrales and Jiménez-Martín (2007)* Price Firm 2

Coronado et al. (2007)* Price Competition and Regulation Undetermined

Danzon and Epstein (2008)* (WP) Price and Launch Drug, Market, Regulation, Country

and Firm

2

Danzon et al. (2005) Launch Drug, competition, Regulation,

Country and Firm

Undetermined

Ganslandt and Maskus (2004) Price Competition and Regulation 2

García-Mariñoso and Olivella (2012) Price and Launch Drug, Regulation and Country 2

García-Mariñoso et al. (2011) Price Regulation and Country 2

Miraldo (2009) Price Competition and Regulation Undetermined

Richter (2008) Price and Launch Competition and Regulation Undetermined

Stargardt and Schreyögg (2006) Price Regulation and Country 15

*: Working paper (WP)

Appendix A 121

Table A.2. Overview of empirical studies

Reference Dependent Variable Focus Effecta Independent Variables Sample Period Data Source TFc

Atella et al. (2008)**

Price level (n.a.) Regulation

* (-) QI (drug quality indicator, defined as one over the $/QALY measure) US (dummy variable which equals 1 if the price is referred to an U.S. brand name) * (+) US*QI

181 products

Only Patent products

n.a.

MEPSc / AIFAc

x

Borrell (2007) Price level (wholesale) Regulation

* (+) Original drug in patent regime * (+) Original drug in no-patent regime Generics after patent expiration Country Mean income Country income inequality (GINI,%) * (+) Number of doses per day Efficacy of the drug Adverse reactions of the drug Years in the US market (1-12 years) * (-) 5 * (-) 7 * (-) 8 * (-) 9 (n.a.) Therapeutic Fixed Effects

15 products

All type of products (NCE, Patent, off-patent,generics)

1995-2000

IMS HEALTH

Cabrales and Jiménez-Martín

(2007)** Price level (ex-factory) Firms

The market share of all national products in the ATC4 market HHI-local (The Hirschmand-Herfindähl concentration index for national ¯rms in the ATC4 market) * (-) Firm size New (A dummy taking one if the product was first

All products at ATC-4


1998-2003

IMS HEALTH

x



observed in the previous year) * (+) Dummy of absence of a global price of reference * (+) Average global price of the molecule in US real $ * (+) A dummy taking one if the corporation is local-non multinational A dummy taking one if the corporation is local but multinational Number of identified generics in the market Berry index (it measures the degree of specialization of the corporation) * (+) Fpc (Fraction of public consumption in GDP) GDP per capita (Fraction of public consumption in GDP) * (+) GDP per capita in US dolars Reg (Level of regulation: 1 low, 2 medium, 3 high) * (-) Fpc*Reg2 * (-) Fpc*Reg3 * (+) GDP*Reg2 * (+) GDP*Reg3 * (-) Molage (Time elapsed since the molecule was launched to December 31, 2003) Log of the number of market a molecule is present Cesnormol (A dummy taking one if the molecule was launched before January 1, 1991)

Appendix A 123


Censorlag (Censormol_lag_1) * (-) Generic (A dummy taking one if the product is generic)

Coronado et al. (2007)** Price level (ex-factory) Regulation

HHI (Herfindähl-Hirschman Index) *(+) Firm size lag *(-) New product lag * (+) Global Price in USD for product j belonging to firm i Binary variable, taking 1 if product j is a compound of molecules * (+) Market share of product j in market k * Number of generic products in market k Time elapsed up to 2003 since molecule (market) k was launched Corporation share in market k excluding product j's share Binary variable, taking 1 if molecule age is censored in the sample Binary variable, taking 1 if product j was launch date is censored in the sample Binary variable, taking 1 if product j is a generic) Weighted average multimarket contact variable for firm i in market k Alternative weighted average multimarket contact variable in market k

All products at ATC-4.


1998-2004

IMS HEALTH

x

Danzon and Chao

(2000a)b Price level (ex-factory) Drug and Brand

Competition

Strength Molecule Age (months from the first product launch each country to September 1992) Form codes (The number of distinct

171 molecules / 5690 products

All type of products (NCE, Patent,

off-patent,generics)

10/1991-09/1992

IMS HEALTH



formulation of strengths in the molecule) Global penetration (number of countries in which the molecule is available out of the seven countries in sample) Packsize Number of manufacturers of the molecule Therapeutic substitute molecules (ATC3) Therapeutic substitute molecules entry lag ( lag in months between each molecule launch date and the first launch of the molecule, ATC3)

Danzon and Chao

(2000b)b Price level (ex-factory) Brand Competition and

Regulation

Strength Molecule Age (months from the first product launch each country to September 1992) Form codes (The number of distinct formulation or strengths in the molecule) Packsize Number of manufacturers of the molecule Generic Entry Lag (lag in months between the product’s own launch date and the launch date of the first product in the molecule) Therapeutic substitute molecules (ATC3) Products per Therapeutic Substitute Molecule Therapeutic substitute molecules entry lag ( lag in months between each molecule launch date and the first launch of the molecule, ATC3)

171 molecules / 5690 products


10/1991-09/1992

IMS HEALTH

Danzon and

Epstein (2008)**

Price at launch (ex-factory) and Launch window

Brand Competition and Regulation

S=Superior ; I=Inferior * (+) S Expected Drug Price Superior Brands (lag) Superior Brand’s Price Missing

375 molecules


QI/1992-QIV/2003

IMS HEALTH

Appendix A 125


* I (-) Expected Drug Price Inferior Brands (lag) Inferior Drug Price Missing Expected Drug Volume (lag) Number of Generic Manufacturer in Superior Subclass Number of Generic Manufcaturer in Inferior Subclass * I(-) Generics’ Price Missing * I(+)First Brand Launch in Country-Subclass * S(+) I(+) Second Brand Launch in Country-Subclass * S(+) I(+) Third or Fourth Brand Launch in Country-Subclass * S(+) Min Own Price in High-price EU Missing

* S(+) I(+) Min Own Price in High-Price EU Min Own Price in Low-price EU Missing Min Own Price in Low-Price EU * I(-) Min Own Price High-price non-EU Missing * S(+) Min Own Price in Hi-Price non-EU Number of low-price EU countries a molecule has already launched Number of high-price EU countries a molecule has already launched Number of high-price non-EU countries a molecule has already launched Number of molecules in Superior Subclass



* S(-) Number of molecules in Inferior Subclass PIc Share in Subclass * S(+) GDP per Capita

* S(-) Country-Specific Quarterly Producer Price Index * S(+) Strenght * S(-) I(-) Pack size Form: Oral Solid Delayed * S(+) I(+) Form: Injectable Form: Other * S(-) I(-) Time since Global Launch * S(+) I(+) Time since Global Launch Squared * I(+) First Global Launch before 1990 First Global Launch in [1996-end] * S(+) I(+) Launch by Local Originator Corporation * S(+) I(+) Launch by Solo Licensee Corporation * S(+) I(+) Launch by Local Co-marketer Corporation Exchange rate ( US to EUR) * S() I() Country fixed effects (n.a.) ATC fixed effects

Danzon et al.

(2005) Hazard Launch Regulation *(+) Expected Drug Price *(+) Expected Drug Volume

85 molecules

Only NCE (New Chemical Entities)

09/1994-09/1998

IMS HEALTH

x

Appendix A 127


*(+) Firm’s Global Launch Experience (sales)

*(+) The originator firm’s home country *(+) GDP per capita * Country fixed effects * ATC fixed effects

Danzon et al. (2011)** Price level (ex-factory) Regulation and Country

* (-) IMS*GENERIC indicator ( IMS: a dummy taking one if the drug is sold through standard retail channels; GENERIC: a dummy taking one if the the generic is present in a country-year) * (-) GPRM*BRAND indicator (GPRM: a dummy taking one if the drug is procured by NGOs; a dummy taking one if the drug is the originator is present in a country-year) * (-) GPRM*GENERIC indicator ( GPRM: a dummy taking one if the drug is procured by NGOs; a dummy taking one if a generic is present in a country-year) * (+) Per capita income country * (-) GINI coefficient GINI missing indicator HIV prev. (HIV country prevalence rate) * (-) The number of tender generic products in the same therapeutic class-country-year * (-) The number of retail generic products in the same therapeutic class-country-year The number of originator products in the same therapeutic class-country-year * (+) Originator molecule flag


01/2004-06/2008

IMS HEALTH / GPRMc



* (+) Generic molecule flag

Heuer et al. (2007)** Hazard Launch Regulation

Country GDP per capita Size of the country population A dummy taking one if the country use ERP A dummy taking one if the country use other direct price controls such as cost-effectiveness, etc.) A dummy taking one if the country use RP * (-) A dummy taking one if the country use ERP explicitily * (-) A dummy taking one if the country use ERP as a basis for their decision making criteria

35 molecules

Only NCE

01/1995-12/2005

IMS HEALTH

Kanavos and Costa-Font

(2005) Price level (wholesale) Regulation

Total market size, defined as sales for all products Market share of each PI product within each product market and each importing country Average Euclidean Distance of latitude and longitude between each importing and exporting country capitals * (-) Exchange rate $ * (-) Purchasing Power Parities in importing country * (-) Market shares of generics consumption in a country (i ) * (-) Dummy variable for introduction of the clawback; Price regulation (Dummy variable for price regulation defined as the intervention of third party payer (national insurance company) or the government in terms of setting price of each product ( j )

19 molecules


QI/1997-QIV/2002

IMS HEALTH x

Kanavos and Price level (ex-factory and * (+) Number of years since molecule’s launch in 68 molecules / 100 products 2004,2007 IMS HEALTH

Appendix A 129


Vandoros (2011)

retail) the local market Age squared generics (dummy variable 1 if there is a generic competitor present in the market) * (+) Dummy variable for United States * (+) Dummy variable for United Kingdom * (+) Dummy variable for Mexico Dummy variable indicating the impact of Health Technology Assessment being explicitly used as a policy measure Dummy variable. Indicates the presence of reference pricing * (+) Dummy variable. Indicates the presence of free pricing Dummy variable. Indicates the explicit use of ERP) Exchange rate * () Therapeutic fixed effects

All type of products (NCE, Patent,

off-patent,generics)

Kyle (2006) Hazard Launch Firms

* (+) Drug importance (drug’s share of stock of Medline citations for class) * (+) Number of countries launched (Number of countries where the molecule has been launched) * (-) Number of countries launched squared * (+) Multinational (Firm has launched drugs in 10+ countries) * (+) Domestic firm (taking 1 if headquarters are located in the country) * (-) Portfolio (total number of firm’s drug)

1482 molecules

Only NCE and patent products

1980-2000

PJB Pc



* (+) Common language Common border Common regulations * (+) Country experience (Count of firm’s other drugs launched in country) Country-class experience (count of firm’s drug in country-class market) * (+) Experience years (number of years firm has marketed in country) * (-) Price controls (dummy variable: country using price controls) Population (country population) Population squared GDP per capita * (-) Number of new drug in the market (Count of drugs in market launched less than 5 years ago) * (+) Number of new drug in the market squared (Count of drugs in market launched less than 5 years ago) * (-) Number of old drugs in market (Count of drugs in market launched more than 5 years ago) * (+) Number of old drugs in market squared (Count of drugs in market launched more than 5 years ago) Number of potential competitors (Count of drugs launched in class elsewhere in the world) Number of domestic incumbents Number of foreign incumbents Therapeutic fixed effects

Appendix A 131


(n.a.) Country fixed effects (n.a.) Year fixed effects

Kyle (2007) Hazard Launch Regulation

* (+) Number of drugs in the market (Count of drugs in therapeutic classcount market) Number of potential entrants * (+) Drug importance (Drug's share of stock of Medline citations for class) * (+) Number of countries launched in ( Number of countries where the molecule has been launched) * (+) Prior launch in a high-price country * (+) Prior launch in a low-price country * (-) Firm is headquartered in a price-controlled country * (-) Portfolio ( total number of firm’s drug) * (+) Domestic firm (taking 1 if headquarters are located in the country ) * (+) International experience (Count of countries in which firm has launched any drugs) * (+) Country experience ( Count of firm’s other drugs launched in country) * (-) Price freeze * (-) Price controls * (-) Supply-side controls * (-) Price Rank * (+) Price Rank*post-1995 period Prescribing budgets

1444 molecules

Only NCE and patent products

1980-1999

PJB P



* (+) Therapeutic class reference pricing * (-) Pharmacoeconomic evidence * (+) Population (country population) * (-) GDP per capita * (-) Corruption score * (-) Market competition index * (-) Entry cost as percentage of GDP per capita (n.a) Therapeutic fixed effects (n.a.) Year fixed effects

Lanjouw (2005)** Hazard Launch Regulation and Country

MLICsc equation: Short process patent (< 15 years) * (+) Add Long process (only) patents Add Long process & product patents Some Price Control (dummy=1 if country has a formal price control mechanism but it is not extensive) * (-) Extensive Price Control (dummy = 1 if price control covers most of the market and/or is viewed as particularly restrictive) * (-) Essential Drug List Dummy =1 for national adoption of an EDL Standard Treatment Guidelines (dummy = 1 for national adoption of standard treatment guidelines) * (+) Health Expenditure Private Share of Health Expenditure R&D Expenditure * (+) Country Population

782 molecules


1982-2001

IMS HEALTH

Appendix A 133


* (+) GDPcapita * (+) Gini Coefficient * (-) Gini Coefficient x GDPcapita * (+) Percentage Population over 65 years * (-) Percentage Population between 15-64 years Population Growth GDP Growth Radios per capita 1990 (Average radios per person) Growth Radio 90-95 * (+) Doctors Rate 1990 Growth Doctors 90-95 (n.a.) ATC fixed-effect (n.a.) year of first launch High-income Countries * (+) Short process patent (< 20 years) * (+) Add Long process and/or product patents * (-) Some Price Control (dummy=1 if country has a formal price control mechanism but it is not extensive) * (-) Extensive Price Control (dummy = 1 if price control covers most of the market and/or is viewed as particularly restrictive) * (+) Some Price Control x GDPcapita * (-) Essential Drug List Dummy =1 for national adoption of an EDL Standard Treatment Guidelines (dummy = 1 for national adoption of



standard treatment guidelines) Standard Treatment Guidelines National Formulary EMA * (-) Health Expenditure Private Share of Health Expenditure R&D Expenditure * (+) Country Population * (-) GDPcapita * (+) Gini Coefficient * (-) Gini Coefficient x GDPcapita Percentage Population over 65 years * (-) Percentage Population between 15-64 years * (-) Population Growth GDP Growth Doctors Rate 1990 Growth Doctors 90-95 (n.a.) ATC fixed-effect (n.a.) year of first launch

Timur et al. (2011) Price level (retail) Drug and Brand

Competition

* (-) Molecule age ( the number of years since the first product launch of each molecule in each country) * (+) Strength Form code ( the number of different product formulations for each molecule) * (-) Packsize

124 molecules


1994-2003

IMS HEALTH

Appendix A 135


* (-) Diffused molecules ( the number of sample countries in which the molecule is available) * (-) Generic competition ( number of manufacturers of the molecule) Therapeutic substitute molecule

Verniers et al. (2011)

Price at launch (ex-factory) and Launch window Regulation

Price equation * (+) Launch window * (-) Launch window squared Selectivity variable (generalizedresiduals) Ex-manufacturer price regulation Profit control regulation Cross-country reference pricing regulation Therapeutic reference pricing regulation Pharamco-economic evidence regulation Strenght of patent protection * (+) Population size Health expenditure per capita Uncertainity avoidance (country's national culture) Masculinity (country's national culture) Individualism ( country's national culture) Power distance ( country's national culture) * (+) Competition ( a Herfindahl–Hirschman index foreachdrugineachcountry)

58 products

All type of products, not including generics (NCE, Patent, off-patent)

02/1994-06/2008

IMS HEALTH



* (+) Firm’s home country * (+) Daily dosage (DDD defined by the WHO) Inflation * () ATC fixed effects Launch equation: * (+) Launch price * (-) Launch price squared * (+) Selectivity variable (generalizedresiduals) Ex-manufacturer price regulation * (+) Profit control regulation Cross-country reference pricing regulation * (+) Therapeutic reference pricing regulation * (+) Pharamco-economic evidence regulation * (+) Strenght of patent protection * (-) Population size * (+) Health expenditure per capita * (-) Uncertainity avoidance (country's national culture) * (+) Masculinity (country's national culture) * (-) Individualism ( country's national culture) * (+) Power distance ( country's national culture) Competition (aHerfindahl–Hirschmanindexforeachdrugineachcountry)

Appendix A 137


* (-) Firm’s home country * (-) Summer (dummyvariablethatcaptureswhetherthelaunchofdruginacountryoccurredinsummer)* (-) EMA (dummy = 1 when a country is a member of the EMA) Inflation * () ATC fixed effects

a. Focus Effects have been classified according to the paper: i) Drug characteristics such as pack size, strenght etc. ii) Brand competition such as number of brand therapeutic competitors iii) Regulation characteristics such as level of regulation (high, medium or low), type of regulation policy applied (reference pricing, external reference pricing, profit controls, etc.), etc. iv) Country characteristics such as demographic features, socioeconomics indicators, etc. v) Firm features such as firm location, type of firm (multinational or local), etc. b. Details on individual variables and information on variables having significant influence have not been displayed in this Table 2 because this paper carries out country by country regressions. Please, see the paper. c. TF: Theoretical Framework; GPRM: WHO’s Global Price Reporting Mechanism; MEPS: Medical Expenditure Panel Survey; AIFA: Italian National Agency for Drug Administration and Control Prices; PJB P: PJB Publications; MLIC: Middle and Low Income Countries; PI: Parallel Importer *: Statistically significant at 5%. **: Working papers (WP); n.a.: non-available

B. Appendix B

B.1 Decision tree and proofs

ERP countries formulas

pERPminC

pImin min p

1, p

2...p

j...p

J if {C,D}

pImin min p

1, p

2...p

j...p

J, p

D if {D,C}

pERPD

pI

pj

J

j1

pC

J 1 if {C,D}

pI

pj

J

j1

J

if {D,C}


Figure B.1 Decision tree

Appendix B 141

Proof of PPR 1

We compare the firm’s profits when the firm delays the launch in country D

(sequence {C, D}) to delay launch in country C (sequence {D, C}). Then,

Proof of PPR 2

We compare the firm’s profits when the firm delays launch in country D (sequence

{C, D}) to delaying launch in country C (sequence {D, C}). Then,

Proof of PPR 3



OF C,D OF D,C

2pFqC pFqD F 2pFqD p

FqC F

p

F

pF

qD

qC

OF C,D OF D,C

2pFqC pI

qD F 2pIqD p

FqC F

p

F

pI

qD

qC


Proof of PPR 4



Proof of Condition 1

Below, we present table B.6 and B.7. Both tables display the health agencies

surplus under each country pricing policy and under each country sequence chosen by

the firm. Then, for each country, we first compare the health agency surplus of choosing

CEA to that of the ERP under the same country sequence. Second, we do the same

under different country sequences,

OF C,D OF D,C

2pIminqC pFqD F 2pFqD pI

minqC F

pImin

pF

qD

qC

OF C,D OF D,C

2pIminqC pI

qD F 2pIqD pI

minqC F

pImin

pI qD

qC

Appendix B 143

Table B.1. Health Agency surplus of country C

Country C {C,D} {D,C}

i) and iii) CEA OFC (WtPYF pF)qC a OFC (WtPYF p

F)qC a r

ii) and iv) ERP OFC (WtPYF pImin )qC OFC (WtPYF pI

min )qC r

Table B.2. Health Agency surplus of country D

Country D {C,D} {D,C}

i) and iii) CEA OFD (WtPYF pF )qD a OFD (WtPYF pF )qD a r

ii) and iv) ERP OFD (WtPYF pI )qD OFD (WtPYF pI

)qD r

Country C

So, we firstly operate for country C under sequence {C,D}. Notice that the surplus

under i) CEA and iii) CEA is the same. Analogously, the surplus under ii) ERP and iv)

ERP is also the same. Then, it is enough to compare only one of the surpluss when

applying the CEA to one of the surplus applying the ERP. So, under the sequence {C,D}:

ii) ERP i) CEA

(WtPYF pImin )qC (WtPYF p

F)qC a

a

qi

pImin p

F

Then, we also operate for country C under sequence {D,C}. Notice that the surplus

under i) CEA and iii) CEA is the same. Analogously, the surplus under ii) ERP and iv)

ERP is the same. Therefore, we just need to compare one of the surplus when applying

CEA to one of them applying the ERP.


iv) ERP i) CEA

(WtPYF pImin )qC r (WtPYF p

F)qC a r

a

qi

pImin p

F

Second, we compare the health agency surplus of choosing CEA to that of the

ERP under the different country sequences. Since i) CEA and iii) CEA present the same

surplus under {C,D} and {D,C} we choose only one of each sequence to compare, for

example, i) CEA. As ii) ERP and iv) ERP also present the same surplus under {C,D} and

{D,C}, we can choose only one of each sequence to make the comparison, for example,

ii) ERP.

ii) ERP{D,C} i) CEA{C,D}

(WtPYF pImin )qC r (WtPYF p

F)qC a

a

qC

pImin p

F r

qC

ii) ERP{C,D} i) CEA{D,C}

(WtPYF pImin )qC (WtPYF p

F)qC a r

a

qC

pImin p

F r

qC

Appendix B 145

Country D

So, we firstly operate for country D under sequence {C,D}. Notice that the surplus

under i) CEA and ii) CEA is the same. Analogously, the surplus under iii) ERP and iv)

ERP is also the same. Then, it is enough to compare only one of the surpluss when

applying CEA to one of the surplus applying ERP. So, under the sequence {C,D}:

iii) ERP i) CEA

(WtPYF pI )qD r (WtPYF pF )qD a r

aqD

pI pF

Then, we also operate for country D under sequence {D,C}. Notice that the

surpluss under i) CEA and ii) CEA are the same. Analogously, the surpluss under iii) ERP

and iv) ERP are the same. Therefore, it is enough to compare only one of the surplus

when applying CEA to one of the surplus applying ERP.

iii) ERP i) CEA

(WtPYF pI )qD (WtPYF pF )qD a

aqD

pI pF

Second, we compare the health agency surplus of choosing CEA to that of the

ERP under the different country sequences. Since i) CEA and ii) CEA present the same

surplus under {C,D} and {D,C}, we choose only one of each sequence to compare, for

example, i) CEA. As iii) ERP and iv) ERP also present the same surplus under {C,D} and


{D,C}, we can choose only one of each sequence to make the comparison, for example,

iii) ERP.

iii) ERP{D,C} i) CEA{C,D}

(WtPYF pI )qD (WtPYF pF )qD a r

aqC

pI pF

rqC

iii) ERP{C,D} i) CEA{D,C}

(WtPYF pI )qD r (WtPYF pF )qD a

aqC

pI pF

rqC

C. Appendix C

C.1. Variable definitions of Danzon and Epstein (2008)

Log Avg Price of Superior Brands (Lag 1Q): The (log lagged) unweighted average

price of competitor brand (originator and licensed, including parallel imports) products in

the same therapeutic class as a comprehensive measure of the direct effect of price

regulation on the expected price for a new drug. We have not included this variable in the

updated model because our database has only new launches, differently to the D&E

database, therefore, we have not been able to measure it, and additionally, because we

do not distinguish between superior and inferior drugs.

Superior Brand’s missing D.V.: A dummy variable equals to 1 whether the product k

has no superior brand competitors in country j in year t, and equals to 0 otherwise. We

have not included this variable in the updated model because our database has only new

launches, differently to the D&E database, and additionally, because we do not

distinguish between superior and inferior drugs.

Log Avg Price of Inferior Brands (Lag 1Q): The (log lagged) unweighted average price

of competitor brand (originator and licensed, including parallel imports) products in the

same therapeutic class as a comprehensive measure of the direct effect of price

regulation on the expected price for a new drug. We have not included this variable in the

updated model because our database has only new launches, differently to the D&E

database, therefore, we have not been able to measure it, and additionally, because we

do not distinguish between superior and inferior drugs.

Inferior Brand’s missing D.V.: A dummy variable equals to 1 whether the product k has

no inferior brand competitors in country j in year t, and equals to 0 otherwise. We have not


included this variable in the updated model because our database has only new launches,

differently to the D&E database, and additionally, because we do not distinguish between

superior and inferior drugs.

Num Generic Manufs per Molc in Superior Subclass: Number of Generics

Manufacturers per Molecule in country j for superior subclass. We have not included this

variable since we only deals with on-patent molecules.

Num Generic Manufs per Molc in Inferior Subclass: Number of Generics

Manufacturers selling the drug s in country j for inferior subclass. We have not included

this variable since we only deals with on-patent molecules.

Generics Brand’s missing D.V: A dummy variable equals to 1 whether the product k has

no generics competitors in country j in year t, and equals to 0 otherwise. We have not

included this variable since we only deals with on-patent molecules.

No molecule in Superior Subclass: Since we do not distinguish between superior and

inferior drugs, we have not included this variable.

No molecule in Inferior Subclass: Since we do not distinguish between superior and

inferior drugs, we have not included this variable.

Time Since Global Launch (Yrs): number of years elapsed since the drug s global

launch until the drug s launch in country j. Note that under the updated model as we have

collected the launch time in month/year, we have considered each launch the first day of

the month.

Time Since Global Launch squared (Yrs): variable Time Since Global Launch squared.

First Global Launch Before 1990 D.V.: An indicator (=1) for molecules launched before

1990 controls for their relatively old age. Molecules launched during 1990-1995 are the

reference category (=0). Since the D&E database has not only the new launches but also

the drugs already launched into the market, they could differ between old and new

molecules, however, as our database has only new launches between 2006 and 2010, we

have not considered any drug as “old drug”.

First Global Launch in (1996-end) D.V.: An indicator for molecules launched since 1996

tests for effects of the EMA regime. Molecules launched during 1990-1995 are the

Appendix C 149

reference category.

Num Already Launched (UK, Germany): number of countries between the UK and

Germany where the drug s has already been launched at the time of launch of drug s in

country j. Under the updated model81 this variable considers the number of high-price EU

countries where the drug s has already been launched at the time of launch of drug s in

country j.

Num Already Launched (Sweden, Netherlands): number of countries (Sweden,

Netherlands) where the drug s has already been launched at drug s launch time in

country j. Under the updated model81 this variable is included in the variable “Num

Already Launched (UK, Germany)”.

Num Already Launched (Italy, France): number of countries (Italy, France) where the

drug s has already been launched at drug s launch time in country j. Under the updated

model81 this variable considers the number of low-priced EU countries where the drug s

has already been launched at the time of launch of drug s in country j.

Num Already Launched (Spain, Portugal, Greece): number of countries (Spain,

Portugal, Greece) where the drug s has already been launched at drug s launch time in

country j. Under the updated model81 this variable is included in the variable “Num

Already Launched (Italy, France)”.

Num Already Launched (Canada, Japan, Switzerland, USA): number of countries

(Spain, Portugal, Greece) where the drug s has already been launched at drug s launch

time in country j. Under the updated model81 this variable considers the number of high-

priced non-EU countries where the drug s has already been launched at the time of

launch of drug s in country j.

Any PI Share in Subclass D.V.: A dummy variable equals to 1 if any % of total sales of

drug s, is sold by parallel importers. Since we do not have information concerning parallel

trade, we have not included this variable.

First Brand Launch in Ctry-Subclass D.V.: A dummy variable equals to 1 whether the

81 Since our database has some different countries than that of D&E, we have grouped the countries by price level and their belonging to the EU, as D&E made for the variables concerning the Minimum Price (eg., Log Min Own Price in Hi-Price EU).


launch of drug s in country j was the first entrant in country-subclass. Since our database

do not have the drugs already launched into the market, we have not included this

variable.

Second Brand Launch in Ctry-Subclass D.V.: A dummy variable equals to 1 whether

the launch of drug s in country j was the second entrant in country-subclass. Since our

database do not have the drugs already launched into the market, we have not included

this variable.

Third or Fourth Brand Launch in Ctry-Subclass D.V.: A dummy variable equals to 1

whether the launch of drug s in country j was the third or fourth entrant in country-

subclass. Since our database do not have the drugs already launched into the market, we

have not included this variable.

Log Min Own Price in Hi-Price EU (Lag 1Q): Logarithm of the lowest price received for

the molecule m in any country classified as high-price EU country where launch has

already occurred.

Log Min Own Price in Low-Price EU (Lag 1Q): Logarithm of the lowest price received

for the molecule m in any country classified as low-price EU country, where launch has

already occurred.

Log Min Own Price in Hi-Price non-EU (Lag 1Q): Logarithm of the lowest price received

for the molecule m in any country classified as high-price non-EU country where launch

has already occurred.

High-price EU Min Own Price Missing D.V.: A dummy variable equals to 1 whether the

molecule m has not been previously launched in any country classified as high-price EU

country.

Low-price EU Min Own Price Missing D.V.: A dummy variable equals to 1 whether the

molecule m has not been already launched in any country classified as low-price EU

country; otherwise equals to 0.

High-price non-EU Min Own Price Missing D.V.: A dummy variable equals to 1 whether

the molecule m has not been already launched in any country classified as high-price

non-EU country; otherwise equals to 0.

Appendix C 151

Launched by Local Originator D.V.: A dummy variable equals to 1 when identifying a

molecule’s originator corporation launching in its country of domicile; otherwise equals to

0.

Launched by Solo Licensee D.V.: A dummy variable equals to 1 when identifying a

locally-domiciled, licensed corporation that launched the molecule in at least one country

by itself.

Launched by Comarketed D.V.: A dummy variable equals to 1 when identifying a

locally-domiciled, licensee corporation that launched together with another firm in its

home country and did not launch alone in any country; otherwise equals to 0.

We note that the last three variables concerning the categories of firm are not

mutually exclusive; for example, a molecule with a small Local Originator could also have

a Solo Licensee or a Local Co-marketer as a marketing partner, all from the same

country. Therefore, all of them are included in the model and there is no common

reference category. We also note that we have not included the variable “Launched by

Co-marketed” in the updated model because there was no observation presenting this

characteristic.

USD to (ECU or Euro) Exchange Rate: Exchange rate USD/EURO.

Country-Specific Quarterly Producer Price Index: The Quarterly Producer Price Index

of country j in year t (price deflator).

Avg Pack Size (Up to 100): The average pack size of the available presentations (with

pack size lower than 100) in country j in year t. We have not included this variable

because we have found most observations missing for this variable.

Pack Size > 100 D.V.: A dummy variable equals to 1 whether the average pack size of

the available presentations, is greater than 100. We have not included this variable

because we have found most observations missing for this variable.

Avg Pill Strength (g): The average strength measured in grams of the available

presentations in country j in year t.

Form: Oral Solid Delayed D.V.: A dummy variable equals to 1 whether the form of the

presentation is Oral Solid.


Form: Injectable D.V.: A dummy variable equals to 1 if the form of the presentation is

Injectable

Form: Other: A dummy variable equals to 1 if the form of the presentation is not neither

oral solid nor injectable

Log GDP per Capita: Logarithm of the Gross Domestic Product (GDP) of country j in

year t.

Country Fixed Effects: A dummy variable for each country j, to capture other country

specific factors that may affect launch delay and launch prices, as bureaucratically-driven

delays and parallel export risk. Germany is excluded as the reference country.

Year Fixed Effects: A dummy variable for each year t, included in some specifications.

Omitted in tables because its lack of significance.

Appendix C 153

Table C.1. Launch equation: D&E vs. Updated model

Coefficients Marginal Effects Clog-log with robust

clustered SEs Clog-log with Normal

REs Clog-log with robust


REs Variables D&E UM D&E UM D&E UM D&E UM

Log Avg Price of Superior Brands (Lag 1Q) 0.1138** ‐ 0.1442*** ‐ 0.0053* ‐ 0.0086*** ‐

[0.0552] ‐ [0.0476] ‐ [0.0027] ‐ [0.0032] ‐

Log Avg Price of Inferior Brands (Lag 1Q) 6.58E‐02 ‐ 0.0894* ‐ 3.10E‐03 ‐ 0.0053* ‐

[0.0612] ‐ [0.0506] ‐ [0.0029] ‐ [0.0031] ‐

Log Total Volume of All Drugs in Class (Lag 1Q) ‐0.0738 ‐ ‐0.0202 ‐ ‐0.0034 ‐ ‐0.0012 ‐

[0.0527] ‐ [0.0581] ‐ [0.0027] ‐ [0.0035] ‐ Num Generic Manufs per Molc in Superior Subclass (Lag 1Q)

‐0.004 ‐‐0.0079

‐‐0.0002

‐‐0.0005

‐

[0.0045] ‐ [0.0059] ‐ [0.0002] ‐ [0.0004] ‐ Num Generic Manufs per Molc in Inferior Subclass (Lag 1Q)

‐0.0016 ‐‐0.0014

‐‐0.0001

‐‐0.0001

‐

[0.0021] ‐ [0.0017] ‐ [0.0001] ‐ [0.0001] ‐

No Molecules in Superior Subclass D.V. 0.2585 ‐ 0.1416 ‐ 0.0135 ‐ 0.0089 ‐

[0.1848] ‐ [0.1947] ‐ [0.0105] ‐ [0.0129] ‐

No Molecules in Inferior Subclass D.V. ‐0.5880*** ‐ ‐0.5529 ‐ ‐0.0210*** ‐ ‐0.0264 ‐

[0.2181] ‐ [0.4677] ‐ [0.0074] ‐ [0.0181] ‐

Time Since Global Launch (Yrs) ‐0.6240*** ‐0.6185*** ‐0.4540*** ‐1.4765*** ‐0.0291*** ‐0.0930*** ‐0.0271*** ‐1.4765***

[0.0578] [0.2109] [0.0536] [0.1461] [0.0046] [0.0312] [0.0049] [0.1461]

Time Since Global Launch Squared (Yrs) 0.0231*** 0.0471 0.0158*** 0.1100*** 0.0011*** 0.0070 0.0009*** 0.1100***

[0.0027] [0.0398] [0.0030] [0.0280] [0.0002] [0.0060] [0.0002] [0.0280]

First Global Launch Before 1990 D.V. ‐0.0034 ‐ ‐0.2426 ‐ ‐0.0002 ‐ ‐0.0131 ‐

[0.1931] ‐ [0.2860] ‐ [0.0090] ‐ [0.0149] ‐

First Global Launch in [1996-end] D.V. ‐0.0497 ‐ ‐0.1018 ‐ ‐0.0023 ‐ ‐0.0058 ‐

[0.1479] ‐ [0.1907] ‐ [0.0066] ‐ [0.0108] ‐

Num Already Launched (UK, Germany) 0.5935*** 0.1137*** 0.4902*** 0.1351*** 0.0290*** 0.0171*** 0.0301*** 0.1351***

[0.0920] [0.0167] [0.0785] [0.0151] [0.0071] [0.0023] [0.0072] [0.0151]

Num Already Launched (Sweden, Netherlands) 0.5079*** ‐ 0.3935*** ‐ 0.0245*** ‐ 0.0239*** ‐

[0.0705] ‐ [0.0749] ‐ [0.0052] ‐ [0.0059] ‐

Num Already Launched (Italy, France) 0.2688*** ‐0.2323*** 0.3057*** ‐0.1625*** 0.0126** ‐0.0349*** 0.0185*** ‐0.1625***

[0.0986] [0.0519] [0.0840] [0.0340] [0.0055] [0.0072] [0.0058] [0.0340]







Num Already Launched (Spain, Portugal, Greece) 0.067 ‐ ‐0.0409 ‐ 0.0031 ‐ ‐0.0024 ‐

[0.0661] ‐ [0.0654] ‐ [0.0030] ‐ [0.0040] ‐

Num Already Launched (Canada, Japan, Switzerland, USA) 0.1907*** 0.2293*** 0.1321** 0.2378*** 0.0089*** 0.0344*** 0.0079** 0.2378***

[0.0637] [0.0366] [0.0560] [0.0282] [0.0030] [0.0050] [0.0035] [0.0282]

Any PI Share in Subclass D.V. 0.0161 ‐ 0.067 ‐ 0.0008 ‐ 0.0041 ‐

[0.1536] ‐ [0.1510] ‐ [0.0072] ‐ [0.0092] ‐

Launch by Local Originator Corporation D.V. 1.3954*** 0.1504 1.5681*** 0.1603 0.1286*** 0.0233 0.1822*** 0.1603

[0.2676] [0.1099] [0.1626] [0.1788] [0.0458] [0.0177] [0.0420] [0.1788]

Launch by Solo Licensee Corporation D.V. 0.5481*** 0.4494 0.5646*** 0.0012 0.0328** 0.0752 0.0424*** 0.0012

[0.1765] [0.4546] [0.1551] [0.4065] [0.0130] [0.0841] [0.0157] [0.4065]

Launch by Local Co-marketer Corporation D.V. 0.5592*** ‐ 0.3463* ‐ 0.0340** ‐ 0.0239 ‐

[0.1786] ‐ [0.1984] ‐ [0.0145] ‐ [0.0160] ‐

USD to (ECU or Euro) Exchange Rate 0.0945 ‐14.1664** ‐0.192 ‐21.722*** 0.0044 ‐2.1310*** ‐0.0115 ‐21.722***

[0.4523] [1.4340] [0.4216] [1.1970] [0.0209] [0.2154] [0.0255] [1.1970]

UK D.V. ‐0.243 ‐ 0.0540*** ‐0.1616 ‐0.0074 ‐0.0101 ‐0.0092 ‐0.009 ‐0.0074

[0.2096] [0.0907] [0.2000] [0.2250] [0.0085] [0.0155] [0.0111] [0.2099]

Netherlands D.V. ‐0.8790*** 0.0202 ‐0.7894*** +0.0892 ‐0.0276*** 0.0035 ‐0.0343*** 0.2115

[0.2465] [0.1088] [0.2194] [0.2324] [0.0082] [0.0190] [0.0106] [0.2250]

Sweden D.V. ‐0.7520** 0.1008 ‐0.5688** 0.0779 ‐0.0249** 0.0179 ‐0.0269** 0.0892

[0.2968] [0.0718] [0.2307] [0.2094] [0.0099] [0.0129] [0.0116] [0.2324]

France D.V. ‐1.2645*** ‐0.8017*** ‐1.2202*** ‐0.5192** ‐0.0340*** ‐0.1153*** ‐0.0452*** ‐0.5192

[0.2103] [0.1610] [0.2304] [0.2574] [0.0078] [0.0215] [0.0110] [0.2574]

Greece D.V. ‐1.2725*** ‐ ‐1.0673*** ‐ ‐0.0341*** ‐ ‐0.0418*** ‐

[0.2637] ‐ [0.2496] ‐ [0.0086] ‐ [0.0116] ‐

Italy D.V. ‐0.9829*** ‐0.9402*** ‐0.8500*** ‐0.7843*** ‐0.0296*** ‐0.1305*** ‐0.0361*** ‐0.7843**

[0.2439] [0.1960] [0.2173] [0.3042] [0.0084] [0.0237] [0.0109] [0.3042]

Portugal D.V. ‐1.9128*** ‐ ‐1.7246*** ‐ ‐0.0405*** ‐ ‐0.0536*** ‐

[0.2319] ‐ [0.2522] ‐ [0.0084] ‐ [0.0115] ‐

Spain D.V. ‐0.8052*** ‐0.8177*** ‐0.6327*** ‐0.6889** ‐0.0261*** ‐0.1171*** ‐0.0292*** ‐0.6889**

[0.2007] [0.1738] [0.2181] [0.2869] [0.0079] [0.0223] [0.0112] [0.2869]

Canada D.V. ‐1.0640*** ‐0.4958*** ‐0.9515*** ‐0.2299 ‐0.0310*** ‐0.0769*** ‐0.0389*** ‐0.2299

Appendix C 155






[0.2341] [0.1228] [0.2103] [0.2231] [0.0084] [0.0190] [0.0108] [0.2231]

Japan D.V. ‐2.4793*** ‐1.2956*** ‐2.5477*** ‐0.8370*** ‐0.0436*** ‐0.1639*** ‐0.0613*** ‐0.8370***

[0.2353] [0.2266][0.2339]

[0.2794] [0.0083]

[0.0229][0.0114]

[0.2794]

Switzerland D.V. ‐1.0800*** ‐0.2259* ‐0.9100*** ‐0.1484 ‐0.0313*** ‐0.0373* ‐0.0378*** ‐0.1484

[0.2644] [0.1266] [0.2540] [0.2192] [0.0089] [0.0203] [0.0119] [0.2192]

USA D.V. ‐0.9119*** ‐0.0848 ‐0.9440*** ‐0.3566 ‐0.0283*** ‐0.0144*** ‐0.0387*** n.e.

[0.2876] [0.1109] [0.2336] [0.2189] [0.0085] [0.0188] [0.0104] n.e.

Brazil D.V. ‐1.1626*** ‐ ‐0.9768*** ‐ ‐0.0326*** ‐ ‐0.0395*** ‐

[0.2411] ‐ [0.2211] ‐ [0.0084] ‐ [0.0111] ‐

Mexico D.V. ‐1.3041*** ‐ ‐1.1008*** ‐ ‐0.0345*** ‐ ‐0.0426*** ‐

[0.2765] ‐ [0.2448] ‐ [0.0086] ‐ [0.0115] ‐

Austria D.V.

‐‐0.1206*

‐ ‐0.0957no sig

‐ ‐0.0204*** ‐ ‐0.0957

‐ [0.0694] ‐ [0.2108] ‐ [0.0117] ‐ [0.2108]

Belgium D.V. ‐ ‐1.0137*** ‐ ‐0.6231** ‐ ‐0.1381*** ‐ ‐0.6231**

‐ [0.1833] ‐ [0.2820] ‐ [0.0221] ‐ [0.2820]

Czech Republic D.V. ‐ ‐0.7642*** ‐ ‐0.6421** ‐ ‐0.1110*** ‐ ‐0.6421**

‐ [0.1831] ‐ [0.2552] ‐ [0.0235] ‐ [0.2552]

Denmark D.V. ‐ ‐0.0376 ‐ ‐0.0265 ‐ ‐0.0064 ‐ ‐0.0265

‐ [0.0615] ‐ [0.2013] ‐ [0.0106] ‐ [0.2013]

Finland D.V. ‐ ‐0.2771*** ‐ ‐0.1990 ‐ ‐0.0452** ‐ ‐0.1990

‐ [0.1304] ‐ [0.2342] ‐ [0.0209] ‐ [0.2342]

Germany D.V. RC RC RC RC RC RC RC RC

RC RC RC RC RC RC RC RC

Hungary D.V. ‐ ‐0.9181*** ‐ ‐0.5569** ‐ ‐0.1282*** ‐ ‐0.7843**

‐ [0.1966] ‐ [0.2464] ‐ [0.0238] ‐ [0.3042]

Norway D.V. ‐ ‐0.0440 ‐ 0.0225 ‐ ‐0.0075 ‐ 0.0225

‐ [0.1135] ‐ [0.2272] ‐ [0.0195] ‐ [0.2272]

Poland D.V. ‐ ‐0.5140*** ‐ ‐0.4264* ‐ ‐0.0794*** ‐ ‐0.4264*

‐ [0.1758] ‐ [0.2564] ‐ [0.0251] ‐ [0.2564]







Australia D.V. ‐ ‐0.6305*** ‐ ‐0.3451 ‐ ‐0.0946*** ‐ n.e.

‐ [0.1544] ‐ [0.2377] ‐ [0.0213] ‐ n.e.

Constant ‐1.1015 9.5227*** ‐1.6217 15.842*** ‐ ‐ ‐ ‐

[0.9652] [1.0730] [1.0217] [1.0061] ‐ ‐ ‐ ‐

Num Observations 23,400 3609 23,400 3609 23,400 3609 23,400 3609

Number of Molecule- level Clusters 111 71 111 71 111 71 111 71

Model Log-Likelihood ‐3071.2 ‐1385.6 ‐3045.7 ‐1292.5 ‐ ‐ ‐ ‐

Mean of Dependent Variable 0.0378 0.09543 0.0378 0.09543 0.0378 0.09543 0.0378 0.09543 Significance(sign.)levels(two‐sided):*:p<0.10;**:p<0.05;***:p<0.01.;[]:standarderror;n.r.:no‐reported;‐:no‐included;RC:referencecategory;D&E:DanzonandEpstein,2008results;UM:Updatedmodelresults;D.V.:Dummyvariable;IMR:InverseMillsRatio

Appendix C 157

Table C.2. Launch Price equation: D&E vs. Updated model

OLS w/ Robust Clustered SEs Normal Random Effects Coefficients Coefficients Log of GDP

included Log of GDP

included Log of GDP

NOT included

Log of GDP NOT

included

Log of GDP included

Log of GDP included

Log of GDP NOT

included

Log of GDP NOT

included Variables D&E UM D&E UM D&E UM D&E UM

Superior Brands' Price Missing D.V. ‐0.0615 n.a. ‐0.0636 n.a. ‐0.0644 n.a. ‐0.0685 n.a. [0.1879] n.a. [0.1886] n.a. [0.1134] n.a. [0.1138] n.a. Log Avg Price of Superior Brands (Lag 1Q) 0.1574*** n.a. 0.1586*** n.a. 0.1244*** n.a. 0.1283*** n.a. [0.0394] n.a. [0.0395] n.a. [0.0201] n.a. [0.0201] n.a. Inferior Brands' Price Missing D.V. 0.1093 n.a. 0.114 n.a. 0.004 n.a. 0.0157 n.a. [0.1523] n.a. [0.1541] n.a. [0.1118] n.a. [0.1121] n.a. Log Avg Price of Inferior Brands (Lag 1Q) 0.0393 n.a. 0.0387 n.a. 0.0844*** n.a. 0.0830*** n.a. [0.0398] n.a. [0.0397] n.a. [0.0206] n.a. [0.0206] n.a. Generics' Price Missing D.V. 0.0674 n.a. 0.0689 n.a. 0.0095 n.a. 0.0115 n.a. [0.1168] n.a. [0.1165] n.a. [0.0758] n.a. [0.0761] n.a. Log Avg Price of Generics in Class (Lag 1Q) 0.0292 n.a. 0.0294 n.a. 0.0169 n.a. 0.0162 n.a.

[0.0295] n.a. [0.0296] n.a. [0.0194] n.a. [0.0195] n.a.

Time Since Global Launch (Yrs) ‐0.0427 0.2350 ‐0.0413 0.2297 ‐0.0391 0.1718 ‐0.0367 0.1685

[0.0265] [0.2131] [0.0265] [0.2129] [0.0260] [0.1216] [0.0261] [0.1212]

Time Since Global Launch Squared (Yrs) 0.0028 ‐0.0709 0.0027 ‐0.0697 0.0005 ‐0.0519 0.0003 ‐0.0515**

[0.0018] [0.0580] [0.0018] [0.0577] [0.0018] [0.0236] [0.0018] [0.0237]

First Brand Launch in Ctry-Subclass D.V. 0.1998 n.a. 0.2034 n.a. 0.2132 n.a. 0.2164* n.a.

[0.1656] n.a. [0.1656] n.a. [0.1307] n.a. [0.1312] n.a.

Second Brand Launch in Ctry-Subclass D.V. 0.3496*** n.a. 0.3486*** n.a. 0.2819*** n.a. 0.2770*** n.a.

[0.0824] n.a. [0.0824] n.a. [0.0749] n.a. [0.0751] n.a. Third or Fourth Brand Launch in Ctry- Subclass D.V. 0.2412*** n.a. 0.2399*** n.a. 0.1859*** n.a. 0.1816*** n.a.

[0.0611] n.a. [0.0613] n.a. [0.0555] n.a. [0.0557] n.a.

High-price EU Min Own Price Missing D.V. 0.2170*** 3.6913*** 0.2138** 3.6687*** 0.0649 3.5098*** 0.0631 3.4941***

[0.0821] [0.8528] [0.0821] [0.8521] [0.0575] [0.4077] [0.0578] [0.4010] Log Min Own Price in Hi-Price EU (Lag 1Q) 0.2179*** 0.5896*** 0.2174*** 0.5920*** 0.1000*** 0.5606** 0.1012*** 0.5652***

[0.0623] [0.0993] [0.0622] [0.0993] [0.0261] [0.0665] [0.0262] [0.0654]

Low-price EU Min Own Price Missing D.V. ‐0.0185 0.7836 ‐0.0161 ‐0.2816 0.0251 1.0856*** 0.03 1.0728***



included Log of GDP

included Log of GDP

NOT included

Log of GDP NOT

included

Log of GDP included

Log of GDP included

Log of GDP NOT

included

Log of GDP NOT


[0.0524] [0.5238] [0.0527] [0.3580] [0.0483] [0.3777] [0.0485] [0.3724] Log Min Own Price in Low-Price EU (Lag 1Q) ‐0.0243

0.1286‐0.0234

.7813‐0.0221

0.2028***‐0.0188

0.2018***

[0.0394] [0.0824] [0.0390] [0.0660] [0.0279] [0.0715] [0.0280] [0.0702] High-price non-EU Min Own Price Missing D.V. 0.1054 0.0295 0.1076 0.0538 0.1019* ‐0.7910** 0.1049* ‐0.7322**

[0.0649] [0.4630] [0.0651] [0.4523] [0.0554] [0.3258] [0.0556] [0.3217] Log Min Own Price in Hi-Price non-EU (Lag 1Q) 0.2682*** 0.0173 0.2677*** 0.0162 0.1433*** ‐0.1467*** 0.1425*** ‐0.1442***

[0.0522] [0.0560] [0.0519] [0.0567] [0.0251] [0.0309] [0.0252] [0.0521]

Any PI Share in Subclass D.V. 0.022 n.a. 0.0232 n.a. 0.0207 n.a. 0.0236 n.a.

[0.0795] n.a. [0.0786] n.a. [0.0650] n.a. [0.0653] n.a.

Log GDP per Capita 0.862 0.7603 ‐ - 1.8350** 2.6070 ‐ ‐

[0.9232] [2.744] ‐ - [0.7426] [1.6017] ‐ ‐ Launch by Local Originator Corporation D.V. 0.0311 ‐0.1260 0.0332 ‐0.1271 0.0693 ‐0.1400 0.0739 ‐0.1413

[0.1009] [0.1064] [0.1011] [0.1054] [0.0694] [0.1480] [0.0696] [0.1486]

Launch by Solo Licensee Corporation D.V. ‐0.0742 ‐0.7741*** ‐0.0708 ‐0.7695 ‐0.0543 ‐0.2689 ‐0.0452 ‐0.2805

[0.0770] [0.2911] [0.0768] [0.2904] [0.0623] [0.3203] [0.0625] [0.3213] Launch by Local Co-marketer Corporation D.V. 0.0031

n.a.0.0038

n.a.0.0265

n.a.0.0288

n.a.

[0.0971] n.a. [0.0980] n.a. [0.0782] n.a. [0.0785] n.a.

USD to (ECU or Euro) Exchange Rate ‐0.1475 0.0745 ‐0.1373 ‐11.9437 ‐0.0265 2.5506 0.0042 ‐8.2194

[0.6339] [10.8946] [0.6234] [8.0278] [0.4524] [10.4129] [0.4542] [8.0960] Country-Specific Quarterly Producer Price Index ‐0.0089**

‐0.0226*‐0.0065

‐0.0139‐0.0088**

‐0.0194**‐0.0038

‐0.0117

[0.0044] [0.0113] [0.0039] [0.0091] [0.0044] [0.0091] [0.0039] [0.0078]

Avg Pack Size (Up to 100) ‐0.0118*** n.a. ‐0.0118*** n.a. ‐0.0094*** n.a. ‐0.0093*** n.a.

[0.0017] n.a. [0.0017] n.a. [0.0010] n.a. [0.0010] n.a.

Pack Size > 100 D.V. ‐1.1996*** n.a. ‐1.2014*** n.a. ‐0.9891*** n.a. ‐0.9930*** n.a.

[0.1880] n.a. [0.1888] n.a. [0.1114] n.a. [0.1118] n.a.

Avg Pill Strength (g) 0.4791* 6.69e‐08 0.4928** 6.41e‐08 0.0577 7.88e‐08 0.0766 7.48e‐08

Appendix C 159


included Log of GDP

included Log of GDP

NOT included

Log of GDP NOT

included

Log of GDP included

Log of GDP included

Log of GDP NOT

included

Log of GDP NOT


[0.2452] [7.09e‐08] [0.2457] [7.14e‐08] [0.2737] [1.66e‐07] [0.2738] 1.62e‐07

Form: Oral Solid D.V. ‐0.1014 1.0132*** ‐0.1092 1.0067** 0.1059 1.0828 0.0926 1.0512

[0.1994] [0.3894] [0.1988] [0.3903] [0.1823] [1.0963] [0.1829] [1.0632]

Form: Injectable D.V. 2.0793*** 2.2366*** 2.0865*** 2.2307*** 1.7654*** 2.8240** 1.7827*** 2.7787***

[0.3009] [0.6784] [0.3025] [0.6798] [0.0885] [1.1136] [0.0887] [1.0812]

Form: Other ‐0.0523 0.6951*** ‐0.0551 0.6929 0.0793 0.1795 0.0734 0.1724

[0.1683] [0.5717] [0.1663] [0.5697] [0.1702] [1.2235] [0.1708] [1.1869]

IMR n.r. 5.9640*** n.r. 5.8563*** n.r. 2.3759 n.r. 2.4453

n.r. [2.2320] n.r. [2.2150] n.r. [1.8211] n.r. [1.8211]

UK D.V. ‐0.3218*** ‐0.6239*** ‐0.2751*** ‐0.5530*** ‐0.2829*** ‐0.5943*** ‐0.1861** ‐0.5316***

[0.0957] [0.1362] [0.0827] [0.1198] [0.0875] [0.1994] [0.0784] [0.1962]

Netherlands D.V. ‐0.0707 ‐0.7173** ‐0.0734 ‐0.3280*** ‐0.0669 ‐0.6824** ‐0.0711 ‐0.3305*

[0.1034] [0.2771] [0.1022] [0.1195] [0.0798] [0.2910] [0.0802] [0.1946]

Sweden D.V. ‐0.1746 ‐0.5008** ‐0.0541 ‐0.2797 ‐0.3307** ‐0.4641** ‐0.0743 ‐0.2667

[0.1684] [0.1931] [0.0859] [0.1018] [0.1335] [0.2164] [0.0844] [0.1793]

France D.V. ‐0.2055* ‐0.3881* ‐0.2382** ‐0.5045** ‐0.1272 ‐0.3579 ‐0.1956** ‐0.4619**

[0.1117] [0.2225] [0.1098] [0.2037] [0.1003] [0.2296] [0.0969] [0.2212]

Greece D.V. 0.2805 n.a. ‐0.3932*** n.a. 1.1945** n.a. ‐0.2393** n.a.

[0.7187] n.a. [0.1198] n.a. [0.5893] n.a. [0.1050] n.a.

Italy D.V. ‐0.19 ‐0.0913 ‐0.3485*** ‐0.3319 0.0835 ‐0.1937 ‐0.2521*** ‐0.4067

[0.1922] [0.2619] [0.1031] [0.2120] [0.1664] [0.2931] [0.0970] [0.2625]

Portugal D.V. 0.373 n.a. ‐0.3098*** n.a. 1.2153** n.a. ‐0.2341** n.a.

[0.7279] n.a. [0.1107] n.a. [0.5961] n.a. [0.1077] n.a.

Spain D.V. 0.1941 ‐0.1741 ‐0.2498** ‐0.4009** 0.7733** ‐0.1734 ‐0.1710* ‐0.3766

[0.4840] [0.2483] [0.0992] [0.1890] [0.3931] [0.2726] [0.0928] [0.2433]

Canada D.V. 0.0439 ‐0.7178*** 0.0363 ‐0.4194** 0.0338 ‐0.6016** 0.0197 ‐0.3318

[0.1078] [0.2653] [0.1054] [0.1616] [0.0923] [0.2613] [0.0925] [0.2023]

Japan D.V. 0.1482 0.0281 0.5852*** ‐0.1411 ‐0.3752 ‐0.1481 0.5573*** ‐0.2937

[0.5017] [0.2205] [0.2021] [0.1590] [0.3967] [0.2511] [0.1221] [0.2339]



included Log of GDP

included Log of GDP

NOT included

Log of GDP NOT

included

Log of GDP included

Log of GDP included

Log of GDP NOT

included

Log of GDP NOT


Switzerland D.V. ‐0.1417 ‐0.4329 0.2091** 0.0725 ‐0.6109* ‐0.3742 0.1389 0.0820

[0.4039] [0.3313] [0.0911] [0.1263] [0.3162] [0.3355] [0.0888] [0.1845]

United States D.V. 0.1838 ‐0.5044 0.5272*** 0.3065** ‐0.35 ‐0.4903 0.3833*** 0.2424

[0.4014] [0.5798] [0.1347] [0.1475] [0.3125] [0.4857] [0.0987] [0.1833]

Brazil D.V. 1.2852 n.a. ‐0.3136*** n.a. 3.1815** n.a. ‐0.2182** n.a.

[1.6940] n.a. [0.1111] n.a. [1.3787] n.a. [0.0940] n.a.

Mexico D.V. 0.9276 n.a. ‐0.2477** n.a. 2.3375** n.a. ‐0.1610* n.a.

[1.2631] n.a. [0.1215] n.a. [1.0153] n.a. [0.0960] n.a.

Austria n.a. ‐0.3591* n.a. ‐0.1733 n.a. ‐0.3222 n.a. ‐0.1562

n.a. [0.1835] n.a. [0.1249] n.a. [0.1993] n.a. [0.1716]

Belgium n.a. ‐0.0142 n.a. ‐0.0247 n.a. ‐0.0153 n.a. ‐0.0193

n.a. [0.1682] n.a. [0.1680] n.a. [0.2316] n.a. [0.2324]

Czech Republic n.a. 0.8090 n.a. ‐0.2704 n.a. 0.6424 n.a. ‐0.3248

n.a. [0.6225] n.a. [0.2980] n.a. [0.6308] n.a. [0.2090]

Denmark n.a. ‐0.6089** n.a. ‐0.3469*** n.a. ‐0.4866** n.a. ‐0.2537

n.a. [0.2334] n.a. [0.1258] n.a. [0.2279] n.a. [0.1768]

Finland n.a. ‐0.2701** n.a. ‐0.2677** n.a. ‐0.2217 n.a. ‐0.2193

n.a. [0.1224] n.a. [0.1219] n.a. [0.1862] n.a. [0.1870]

Germany n.a. RC n.a. RC n.a. RC n.a. RC

n.a. RC n.a. RC n.a. RC n.a. RC

Hungary n.a. 1.9018 n.a. 0.1550 n.a. 1.5965 n.a. 0.0256

n.a. [1.2119] n.a. [0.1473] n.a. [0.9888] n.a. [0.2063]

Norway n.a. ‐1.4516 n.a. ‐0.0860 n.a. ‐1.2198 n.a. 0.0085

n.a. [1.0064] n.a. [0.2214] n.a. [0.7818] n.a. [0.1987]

Poland n.a. 2.4426 n.a. 0.2093 n.a. 2.1297* n.a. ‐0.2193**

n.a. [1.4799] n.a. [0.1641] n.a. [1.2562] n.a. [0.1870]

Australia n.a. ‐1.052* n.a. ‐0.7313 n.a. ‐1.0074*** n.a. ‐0.7155***

n.a. [0.6270] n.a. [0.4934] n.a. [0.2623] n.a. [0.1927]

Year-2004 n.r. RC n.r. RC n.r. RC n.r. RC

Appendix C 161


included Log of GDP

included Log of GDP

NOT included

Log of GDP NOT

included

Log of GDP included

Log of GDP included

Log of GDP NOT

included

Log of GDP NOT


n.r. RC n.r. RC n.r. RC n.r. RC

Year-2005 n.r. 0.51902 n.r. 0.6427* n.r. 0.1677 n.r. 0.2886

n.r. [0.3511] n.r. [0.3326] n.r. [0.2964] n.r. [0.2878]

Year-2006 n.r. 0.5425 n.r. 0.7641** n.r. 0.1461 n.r. 0.3550

n.r. [0.3424] n.r. [0.3183] n.r. [0.2901] n.r. [0.2633]

Year-2007 n.r. 0.5001 n.r. 0.0434** n.r. 0.3443 n.r. ‐0.0653

n.r. [0.4724] n.r. [0.3596] n.r. [0.4547] n.r. [0.3766]

Year-2008 n.r. 0.2902 n.r. ‐0.6273 n.r. 0.4507 n.r. ‐0.3859

n.r. [0.9149] n.r. [0.7221] n.r. [0.8639] n.r. [0.6966]

Year-2009 n.r. ‐0.0634 n.r. ‐0.5090 n.r. 0.1636 n.r. ‐0.2447

n.r. [0.4793] n.r. [0.3927] n.r. [0.4468] n.r. [0.3719]

Year-2010 n.r. Omitted n.r. Omitted n.r. Omitted n.r. Omitted

n.r. Omitted n.r. Omitted n.r. Omitted n.r. Omitted

Constant ‐7.0696 ‐29.6973 1.2244 8.9033 ‐16.8192** ‐28.0345 0.8265 6.7188

[8.7738] [25.9383] [0.8798] [6.3573] [7.1791] [22.1928] [0.7247] [6.5143]

Log GDP per Capita Included? Yes Yes No No Yes Yes No No

Year Fixed Effects Included? Yes Yes Yes Yes Yes Yes Yes Yes

Observations 950 762 950 762 950 762 950 762

Number of Molecule-level Clusters 109 71 109 71 109 71 109 71

R-squared 0.89 0.87 0.89 0.87 0.87 0.86 0.87 0.86

Mean of Dependent Variable 0.74 3.77 0.74 3.77 0.74 2.84 0.74 2.84 Significance(sign.)levels(two‐sided):*:p<0.10;**:p<0.05;***:p<0.01.;[]:standarderror;n.r.:no‐reported;‐:no‐included;RC:referencecategory;D&E:DanzonandEpstein,2008results;UM:Updatedmodelresults;D.V.:Dummyvariable;IMR:InverseMillsRatio.


C.2. Variables definitions of Verniers et al. (2011)82

Launch window: the launch window of drug i in country j is the difference (in months)

between the month in which the drug was first launched anywhere in the world and the

month in which the drug was launched in country j. The month of launch is the first month

in which sales of the new drug are non-zero.

Launch price: The launch price of drug i in country j is the natural logarithm of the ex-

manufacturer price at launch (the selling price charged by the manufacturer to the

wholesaler) in US dollars per gram. To make drug prices comparable across countries,

the drug prices in local currencies are converted to US dollars using the currency

conversion rate at launch provided by the IMF.

Ex-manufacturer price regulation: A dummy variable that indicates the presence (= 1)

or absence (= 0) of a direct restriction of price levels by the regulator.

Profit control regulation: A dummy variable that indicates the presence (= 1) or absence

(= 0) of a threshold on the profits that drug companies can obtain.

Cross-country reference pricing regulation: A dummy variable that indicates the

presence (= 1) or absence (= 0) of the ERP in the country pricing policy.

Therapeutic reference pricing regulation: A dummy variable that indicates whether

health regulators generate a reference price for a cluster of therapeutically similar drugs,

above which price the patient is surcharged (= 1), or no such price is generated (= 0)

Pharmacoeconomic evidence regulation: A dummy variable that indicates whether

health regulators ask for some proof of the drug's cost effectiveness before launch (=1) or

not (=0).

Population size: The population size at the time of launch, measured by the natural

logarithm of the number of inhabitants of country j.

Health Expenditure: The natural logarithm of health expenditures per capita in country j

at the time of launch.

82 Variables preceded by (*) have not been included in the replicated model.

Appendix C 163

Strength of patent protection regulation: an index based on levels of patent laws

ranging from 0 to 5 for each country, representing weak to strong patent protection

(Ginarte and Park, 1997, Park and Wagh, 2002).

*Competition: the Herfindahl–Hirschman index for drug i in country j. This index is

constructed by summing the squared market shares (MS) (based on revenues in the IMS

Health data) of the m drugs in the same ATC4 category as drug i at the time of launch of

drug i in country j. A high Herfindahl-Hirschman index indicates that there is little

competition for drug i in country j. We have not included this variable in the updated

model because our database has only new launches, differently to the Verniers et al.

database (Verniers et al., 2011).

Firm’s home country: A dummy variable that indicates if the company's headquarters is

located in the country of launch j (=1) and 0 otherwise.

Selectivity variable: It is defined as the generalized residuals of the launch reduced

equation for the price equation, and the generalized residuals of the price equation for the

launch equation.

Summer: a dummy variable that captures whether the launch of drug i in country j

occurred in July or August for countries in the Northern Hemisphere or in January or

February for countries in the Southern Hemisphere.

EMA: a dummy variable that has the value 1 if a drug was launched in a country j that is

part of the EMA.

*Daily dosage: A drug i's defined daily dosage in grams is the assumed average

maintenance dose per day of a drug used for its main indication in adults (WHO

definition).

Inflation: the inflation rate (annual percentage change in GDP deflator) in country j at the

time of launch from the World Bank.

Anatomical therapeutic classes: Dummy variables for the therapeutic classes to which

drug i belongs. Verniers et al. treat the therapeutic class A10BG as the base category.

We treat the therapeutic class A as the reference category.

Power Distance: This quantitative index reflects the country’s degree to which a culture


believes how institutional and organizational power should be distributed (equally or

unequally) and how the decisions of the power holders should be viewed (challenged or

accepted). In other words, people in high power distance cultures are much more

comfortable with a larger status differential than low power distance cultures. See

(Hofstede, 1984, Hofstede, 2001, Hofstede, 2012)

Individualism: This quantitative index describes the country’s degree to which a culture

relies on and has allegiance to the self or the group. See (Hofstede, 1984, Hofstede,

2001, Hofstede, 2012).

Masculinity: This quantitative index indicates the country’s degree to which a culture

values such behaviours as assertiveness, achievement, acquisition of wealth and caring

for others, social supports and the quality of life. See (Hofstede, 1984, Hofstede, 2001,

Hofstede, 2012).

Uncertainty Avoidance: This quantitative index refers to the extent to which a culture

feels threatened by ambiguous, uncertain situations and tries to avoid them by

establishing more structure. See (Hofstede, 1984, Hofstede, 2001, Hofstede, 2012).

Appendix C 165

Table C.3. List of countries. Verniers et al. vs. UM

Verniers et al. 2011 UM World Region and Countries North America Canada ✓Mexico Puerto Rico US ✓Western Europe Austria ✓Belgium ✓Denmark ✓Finland ✓France ✓Germany ✓Greece Ireland Italy ✓Luxemburg Netherlands ✓Norway ✓Portugal Spain ✓Sweden ✓Switzerland ✓UK ✓South America Argentina Brazil Chile Colombia Ecuador Peru Uruguay Venezuela Oceania Australia ✓New Zealand Asia Japan ✓Korea Philippines Eastern Europe Czech Republic ✓Estonia Hungary ✓Latvia Lithuania Poland ✓Slovakia Africa and the Middle East Egypt Jordan Kuwait


Verniers et al. 2011 UM World Region and Countries Lebanon Morocco Saudi Arabia South Africa Tunisia United Arabic Emirates UM: Updated Model; ✓: Countries included in the UM

Appendix C 167

Table C.4. Launch Window and Launch Price equations. Verniers et al. vs. Updated Model

Hypothesized effect by

Verniers et al.

Launch Window Equation Hypothesized effect by

Verniers et al.

Launch price Equation

Variables Verniers et al. UM Verniers et al. UM

Constant ‐41.90*** ‐454.9407*** 3.19*** ‐21.4562

[5.99] [41.3960] [0.86] [47.2688]

Launch price ‐ ‐5.65*** ‐31.9941*** ‐ ‐

[0.8] [1.5060] ‐ ‐

Launch price^2 + 0.33*** 0.1684*** ‐ ‐

0.04 [0.0482] ‐ ‐

Launch window ‐ ‐ + 0.03*** 0.0672

‐ ‐ [5.10 x 10‐3] [0.1769]

Launch window^2 ‐ ‐ ‐ ‐1.79 x 10‐4*** ‐0.0015

‐ ‐ [5.89 x 10‐5] [0.0028]

Selectivity variable 2.77*** 29.8786*** ‐2.32 0.0108

[0.69] [1.3277] [2.81] [0.0959]

Ex-manufacturer price regulation + 3.75 12.0945 ‐ ‐0.14 1.3937

[2.55] [10.0227] [0.09] [11.0171]

Profit control regulation + 16.07*** ‐74.8528*** ‐ ‐0.14 ‐2.7865

[3.02] [10.1429] [0.11] [2.2248]

Cross-country reference pricing regulation ‐ ‐3.44 ‐74.2094*** + 0.06 ‐3.8411

[2.45] [5.1783] [0.12] [8.7102]

Therapeutic reference pricing regulation + 4.19** ‐36.1606*** ‐ ‐0.13 ‐2.0732

[1.92] [5.2288] [0.09] [8.0221]

Pharmaco-economic evidence regulation + 3.40* 19.3136*** ‐ ‐0.03 0.2929

[1.76] [5.1848] [0.09] [3.7278]

Strength of patent protection ‐ ‐5.96*** ‐70.0080*** ‐ ‐0.07* ‐3.2159

[1.89] [10.490] [0.09] [13.6724]

Population size ‐1.98** ‐73.3468*** 0.07 ‐2.2274

[0.79] [6.7747] [0.04] [3.2402]

Health expenditures per capita 19.23*** 113.5521*** 7.37 x 10‐3 5.0556

[1.75] [5.5933] [0.09] [11.6811]

Uncertainty avoidance ‐0.20*** 0.3786*** ‐1.45 x 10‐3 0.0141

[0.06] [0.0691] [2.82 x 10‐3] [0.0656]


Hypothesized

effect by Verniers et al.

Launch Window Equation Hypothesized effect by

Verniers et al.

Launch price Equation

Variables Verniers et al. UM Verniers et al. UM

Masculinity 0.25*** 0.2802*** ‐1.71 x 10‐3 0.0063

[0.05] [0.0812] [2.33 x 10‐3] [0.0157]

Individualism ‐0.37*** 8.7132*** ‐3.62 x 10‐4 0.2624

[0.07] [0.8794] [3.22 x 10‐4] [0.4944]

Power distance 0.33*** 18.6091*** ‐6.04 x 10‐3 0.5755

[0.08] [1.8344] [3.82 x 10‐3] [0.9586]

Competition (reverse-scored) 2.57 n.a. 0.65*** n.a.

[2.38] n.a. [0.24] n.a.

Firm's home country ‐6.34*** ‐13.1345*** 0.44* ‐0.1680

[2.41] [1.7095] [0.23] [1.2266]

Summer ‐1.96* 1.8075* ‐ ‐

[1.14] [0.9471] ‐ ‐

EMA ‐4.01* 14.4236*** ‐ ‐

[2.14] [6.5522] ‐ ‐

Daily dosage ‐ ‐ ‐3.08* n.a.

‐ ‐ [0.18] n.a.

Inflation ‐ ‐ 9.79 x 10‐3 0.0844

‐ ‐ [8.29 x 10‐3] [0.0711]

Anatomical therapeutic classes n.r. *** n.r. *** n.r. *** n.r. ***

Observations 1711 505 1711 505

Adjusted R-Squared 0.26 0.53 0.66 0.38 Significance(sign.)levels(two‐sided):*:p<0.10;**:p<0.05;***:p<0.01.;[]:Standarderror;n.r.:no‐reported;n.a.:non‐available;‐:notincluded;UM:Updatedmodel

Appendix C 169

C.3. Variable definitions of the NPLM

Relative launch price: The relative launch price is the price ratio between the launch

price of molecule i, product k in country j at the time t, and the launch price of molecule i,

product k in country g at the first global launch time 0. Being country g the country where

the first global launching of molecule i has occurred. Both prices, numerator and

denominator, are measured at the ex-manufacturer price per standard unit.

Delay: the launch delay of molecule i in country j is the difference (in months) between

the month in which the drug was first launched anywhere in the world and the month in

which the drug was launched in the country j. The month of launch is the first month in

which sales of the new drug are non-zero.

External Reference Pricing (ERP): A dummy variable indicating if country j applies the

ERP (= 1) or (= 0) otherwise.

Log GDP per Capita: Logarithm of the Gross Domestic Product (GDP) of country j in

year t in Euros.

Log Population size (POP): The population size at time t, measured by the natural

logarithm of the number of inhabitants of country j.

Log Health Expenditure per capita (Exp_PublicHealth_percapita): The natural

logarithm of public health expenditures per capita in country j at time t in Euros.

Log Pharmaceutical Expenditure per capita (Exp_Pharma_percapita): The natural

logarithm of pharmaceutical expenditures per capita in country j at time t in Euros.

Log GDP per Capita (GDP2004): Logarithm of the Gross Domestic Product (GDP) of

country j in year 2004 in Euros.

Log Population size (POP2004): The population size in year 2004, measured by the

natural logarithm of the number of inhabitants of country j.

Log Health Expenditure per capita (HPE2004): The natural logarithm of public health

expenditures per capita in country j in year 2004 in Euros.

Log Pharmaceutical Expenditure per capita (PE2004): The natural logarithm of

pharmaceutical expenditures per capita in country j in year 2004 in Euros.


EMA: A dummy variable that has the value 1 if a drug was launched in a country j that is

part of the EMA at the time of launching.

Home launched (HOME): A dummy variable that indicates if the company's headquarters

launching the molecule i is located in the country of launch j (=1) and 0 otherwise.

Inflation: the inflation rate (annual percentage change in GDP deflator) in country j at the

time of launch. Source: World Bank.

ATC: Dummy variables for the therapeutic classes to which molecule i could belong to an

ATC at level one (ATC-1). The therapeutic class A is used as reference category. The

therapeutic classes included are: A- Alimentary tract and metabolism, B- Blood and blood

forming organs, C- Cardiovascular system, D- Dermatologicals, G- Genito-urinary system

and sex hormones, J- Antiinfectives for systematic use, L- Antineoplastic and

immunomodulating agents, M- Musculo-skeletal system, N- Nervous system, R-

Respiratory system, S- Sensory organs and V- Various

Year Fixed Effects: A dummy variable for each year t. Reference year 2004.

Appendix C 171

Table C.5. Variable classification of the NPLM

Level / Time Time Invariant Time Variant

Molecule - ATC fixed effects (ATC)

Country

- External Reference Pricing (ERP) - Country Fixed Effect (CFE) - GDP per capita 2004 (GDP2004) - Population size 2004 (POP2004) - Pharmaceutical Expenditure per capita (HPE2004). - Public Health Expenditure per capita (PE2004).

- GDP per capita (GDP). - Population size (POP) - Pharmaceutical Expenditure per capita (Exp_Pharma_pc). - Public Health Expenditure per capita (Exp_PubHealth). - EMA (EMA)83.

Molecule-Country

- Delay (DELAY) - Home launched (HOME)

Molecule-Country-Presentation

-Relative price (RP)

Table C.6. Probit selection equation of NPML

Variables Retail HospitalDelay -0.0033*** -0.0022 [0.0007] [0.0008] Log of Country size (population) -0.0763*** -0.0634 [0.0193] [0.0159] Log of GDP per capita -0.7109 -0.4016 [0.5320] [0.4553] Log of Health Expenditure per capita 1.0860*** 0.8069 [0.3885] [0.3180] Log of Pharmaceutical Expenditure pc 0.5164*** 0.6106 [0.1083] [0.0960] IRP 0.0959** 0.0305 [0.0374] [0.0392] Firm’s home country -0.1846*** -0.2765

83 The EMA variable has been included in the launch equation of the NPLM as time invariant variable since only few countries of our database got into the EMA during the study period.


Variables Retail Hospital

[0.0431] [0.0724] ema -0.2062*** -0.1657*** [0.0545] [0.0448] ATC A RC RC RC RC B -0.1834 -0.1478 [0.1658] [0.1249] C -0.0884 -0.1172 [0.1647] [0.1130] D -0.2234 -0.2476 [0.2056] [0.1260] G 0.1596 0.1589 [0.2248] [0.2095] J 0.1335 0.1676 [0.1760] [0.1489] L 0.0089 0.0105 [0.1756] [0.1401] M -0.1961 -0.0296 [0.2007] [0.1596] N 0.1631 0.1565 [0.2224] [0.2057] R -0.5232 -0.4502 [0.1494] [0.1191] S 0.1277 0.1722 [0.1713] [0.1384] T 0.2546 0.2735 [0.1499] [0.1146] V 0.4071 0.3979 [0.1539] [0.1253] Constant -4.0329 -5.843 [2.6845] [2.4505] Observations 6021 7585 Number of Molecule-presentation-level Clusters

69 71

R-squared 0.0463 0.0436 Significance(sign.)levels(two‐sided):*:p<0.10;**:p<0.05;***:p<0.01.;[]:Standarderror

Appendix C 173

C.4. Selection of parametric model

We have followed the method proposed by Kiefer (Kiefer, 1988). Basically, in a

first step, according to the empirical density function of the endogenous variable, as

shown in Figure C.1 and Figure C.2, in general terms, we have a monotonically

decreasing function. Then, the more suitable models are the Weibull and Gompertz

parametric models. These two models are often used with data presenting monotonically

failure rates, either increasing or decreasing. Since we have not been able to observe all

variables affecting the launch delay, it is likely to have unobservable heterogeneity. If so,

it would lead to a bias in the inferences concerning the relationship of dependency. The

most common way to control for the unobservable heterogeneity consists in introducing a

parametric distribution for the random error term and thus, to estimate the parameters of

the function which generates such error term. There is no a strict pattern to chose that

distribution. Based on the previous literature, the two distribution most used are the

Gamma (Lancaster, 1992, Klein and Moeschberger, 1997) and the Inverse Gaussian

(StataCorp, 2009). We estimate four different parametric models. In Table C.7 and C.8 we

show the AIC for measuring the goodness of fit and the statistical tests of the unobserved

heterogeneity. Another way to have an accurate selection model is based on the analysis

of the Cox-Snell generalized residuals. Then, if the model has been correctly selected, the

Cox-Snell residuals should present a form of a unit exponential function84. The Cox-Snell

residuals are shown in Figure C.3, C.4, C.5 and C.6 (for retail market), and Figure C.7,

C.8, C.9 and C.10 (for hospital market).

84 f (x) ex if x 0

0 otherwise

with 1


Figure C.1. Density Function of Delay in Months. Retail Market

Figure C.2. Density Function of Delay in Months. Hospital Market

Table C.7. Goodness of fit of parametric models. Retail market

Weibull with Gamma

Heterogeneity

Weibull with Inverse Gaussian

Heterogeneity

Gompertz with Gamma

Heterogeneity

Gompertz with Inverse Gaussian

Heterogeneity AIC 38.14430201 38.142709 38.13897749 38.10271705

Heterogeneity (Ho= No heterogeneity) =29.32

p=0.000

=26.21

p=0.000

=81.28

p=0.000

=8.63

p=0.002

1

21

21

21

2

Appendix C 175

Figure C.3. Weibull, Gamma. Retail market

Figure C.4. Weibull, Inverse Gaussian. Retail market

Figure C.5. Gompertz, Gamma. Retail market


Figure C.6.Gompertz, Inverse Gaussian. Retail market

Table C.8. Goodness of fit of parametric models. Hospital market

Weibull with Gamma

Heterogeneity

Weibull with Inverse Gaussian

Heterogeneity

Gompertz with Gamma

Heterogeneity

Gompertz with Inverse Gaussian

Heterogeneity AIC 37.94271225 37.94223474 37.94238153 37.90821115


p=0.000

=36.17

p=0.000

=100.29

p=0.000

=14.54

p=0.000

Figure C.7. Weibull, Gamma. Hospital Market

1

21

21

21

2

Appendix C 177

Figure C.8. Weibull, Inverse Gaussian. Hospital Market

Figure C.9. Gompertz, Gamma. Hospital Market

Figure C.10. Gompertz,Inverse Gaussian. Hospital Market

Resumen en español

Motivación

Los medicamentos se venden en un mercado globalizado. Esta característica

implica un proceso de negociación específico entre las empresas farmacéuticas y las

agencias de salud de los países. Por un lado, la empresa toma decisiones estratégicas

para lanzar los fármacos en diferentes países y para maximizar sus beneficios globales, y

por otro lado, las agencias de salud de los países implementan políticas de fijación de

precios con el fin de controlar el gasto farmacéutico y garantizar el acceso a los

medicamentos a la población.

Desde la perspectiva de los seguros nacionales de salud, las políticas de fijación

de precios en el mercado farmacéutico son un factor clave en el control del gasto público.

En particular, el gasto farmacéutico de los países pertenecientes a la Organización de

Cooperación y Desarrollo Económicos (OCDE) en 2009 se calcula que representa

alrededor del 19 % del gasto sanitario. En relación con la economía global, el gasto

farmacéutico supone en promedio un 1,5 % del PIB (Producto Interior Bruto) en los

países de la OCDE. Sin embargo, la dispersión en torno a este promedio es alta, para

algunos supone menos del 1 % del PIB como Noruega y Dinamarca, mientras que para

otros alcanza cerca del 2,5 % del PIB, en países como Grecia, Hungría y la República

Eslovaca. El gasto farmacéutico se financia principalmente a través de terceros

pagadores en la mayoría de los países de la OCDE - ya sea a través del seguro público

de salud, lo que representa en media alrededor de un 60 % del total, o por medio de la

cobertura de los seguros privados, dejando en media más de un tercio del total sobre los

hogares (OECD, 2011).

Desde la visión de la industria farmacéutica, los precios y el lanzamiento son tareas

complejas y conectadas directamente con la política de I+D (Investigación y Desarrollo),

industrial y sanitaria. Por lo tanto, los precios y el lanzamiento son sus principales

decisiones estratégicas. En muchos países, el precio se negocia con los proveedores de

seguros de salud (públicos o privados). En este punto, las políticas y estrategias de

fijación de precios son elementos esenciales en el establecimiento de los mismos y en el

acceso a los medicamentos. La fijación de precios de los medicamentos debe contribuir a

mejorar al bienestar social, teniendo en cuenta los intereses de la industria, los

180 Modelización de los Precios y Lanzamiento Global de Nuevos Medicamentos

consumidores y las aseguradoras públicas. Por lo tanto, se debe promocionar el

desarrollo de nuevos medicamentos, poniéndolos a disposición de los consumidores y

controlando el gasto farmacéutico.

El precio y el lanzamiento de un medicamentos envuelven un equilibrio entre el

bienestar público y los beneficios privados, entre los intereses de los fabricantes y los del

país. Cuando los países establecen un precio, corren el riesgo de no proporcionar el

medicamento en el momento que desean, lo que puede producir consecuencias para la

salud y el bienestar de la población (Lichtenberg, 2005). A su vez, la empresa, al retrasar

el lanzamiento de un medicamento en un país, está retrasando también la obtención de

los beneficios que se derivan de la venta en dicho país. Sin embargo, en un mundo cada

vez más globalizado, los precios/lanzamiento de medicamentos se ha convertido en una

cuestión internacional y por lo tanto, los flujos de información entre los países deberían

ser considerados. Tanto las empresas como los países deben actuar localmente pero

pensar globalmente. Debido a la existencia de mecanismos como los precios de

referencia externa (ERP, de ahora en adelante) y el comercio paralelo (PT, en adelante)

(Danzon and Epstein, 2008, Danzon et al., 2005, Garcia Mariñoso et al., 2011), la fijación

del precio de un medicamento en un país influye en los precios y lanzamientos de otros

países. El uso del ERP por determinados países puede provocar que la empresa aplique

determinadas estrategias de precios que puedan perjudicar el bienestar de otros países.

Por un lado, la empresa puede establecer un precio único85 que puede beneficiar a los

países con tradicionalmente precios elevados86, pero dañar a los países que

tradicionalmente han pagado precios bajos. Por otro lado, la empresa puede tratar de fijar

precios altos87 en los primeros países para evitar precios bajos en los últimos países a

través del ERP, o bien, retrasar el lanzamiento en países de precios bajos para evitar

efectos colaterales en el resto países. Concretamente, estas estrategias pueden

perjudicar a los países de precio bajo, e incluso también pueden perjudicar a los países

de precio alto (Garcia Mariñoso et al., 2011).

85 Dos factores contribuyen a la uniformidad de precios entre los distintos mercados: a) las amenazas de las importaciones paralelas, y b) el uso del ERP DANZON, P. M. & TOWSE, A. 2003. Differential Pricing for Pharmaceuticals: Reconciling Access, R&D and Patents. International Journal of Health Care Finance and Economics, 3, 183-205..

86 En el largo plazo los consumidores de los países de precio alto estarán peor si estos precios más bajos producen menores beneficios esperados para la I+D y, por tanto, un menor número de nuevos medicamentos de lo que hubieran estado dispuestos a pagar DANZON, P. M. 1997. Price Discrimination for Pharmaceuticals: Welfare Effects in the US and the EU. International Journal of the Economics of Business, 4, 310-322. .

87 Esta estrategia empresarial no funcionará si el país de alto precio revisa sus precios a la baja después del lanzamiento.

Resumen en español 181

Entre las políticas de precios de medicamentos, la mayoría de los países del

mundo industrializado han puesto en práctica en algún momento bien el análisis coste-

efectividad (CEA, en adelante) o el ERP con el objetivo de controlar el gasto farmacéutico

y garantizar el acceso a los medicamentos, particularmente en los medicamentos bajo

patente (Espin J et al., 2011, Rawlins, 2012).

En esta tesis, el ERP se define como " la práctica de fijar un precio máximo sobre

los fármacos, con base en los precios de fabricante de productos idénticos o similares en

otros países" (Garcia Mariñoso et al., 2011). La mayoría de los países utilizan el ERP

como política de fijación de precios de medicamentos. El uso de este sistema tiene una

aplicación bastante amplia: 24 de los 30 países de la OCDE (Espin J et al., 2011) y

aproximadamente 24 de los 28 Estados miembros de la Unión Europea lo han utilizado

(Leopold et al., 2012). Sin embargo, el ERP no se aplica de forma homogénea en todos

los países. Hay una amplia variedad de métodos para diseñar el índice de precios

externos de referencia (Espin J et al., 2011, Leopold et al., 2012). Depende

principalmente de la elección de los países de referencia, el tipo de precios tenidos en

cuenta88, el método utilizado (el precio más bajo, el precio medio, un porcentaje de los

anteriores, etc.) y si se utiliza un índice ponderado89 o no. También es importante señalar

que algunos países consideran el ERP como una política de precios complementaria a

otras políticas de precios con el fin de tomar la decisión final, por lo que no se aplica

exclusivamente como una política de precios a ciegas90. El ERP se utiliza debido a su

simplicidad a nivel técnico y analítico; recopilar información sobre los precios en el

extranjero no requiere una gran tarea. Los usuarios del ERP piensan que esos precios se

toman como referencia son más o menos correctos, adecuados o justos. Sin embargo,

reconocen que es difícil evaluar si los precios resultantes son apropiados, eficientes u

óptimos de acuerdo a cualquier criterio objetivo. Podemos pensar que si los países que

son tomados como referencia fijan sus precios demasiado altos o bajos, entonces

cualquier país después de aplicar el método de ERP puede correr el riesgo de repetir el

88 Precio actual vs. precio en el momento de lanzamiento

89 El método más frecuente llevado a cabo para fijar el precio de nuevos medicamentos se realiza a través de medidas no ponderadas; estos métodos no ayudan a alcanzar el objetivo de obtener un nivel medio de precios comparable. Se propone la aplicación de índices de precios ponderados con el fin de establecer un precio que sea comparable y útil como referencia para el resto de países DANZON, P. M. & CHAO, L. W. 2000. Cross-national price differences for pharmaceuticals: How large, and why? Journal of Health Economics, 19, 159-195..

90 Espín et al. afirman que "los reguladores no siempre podrían ser capaces o estar dispuestos a "imponer" un determinado precio, pero en su lugar sí utilizar un precio calculado como punto de referencia para las negociaciones junto con otros criterios, tales como el cost-plus o los precios de referencia interna".


mismo error (Espin J et al., 2011).

A su vez, el CEA en economía de la salud tiene como objetivo estimar la relación

entre el coste de una intervención relacionada con la salud y el beneficio que produce en

términos de la cantidad de años vividos con buena salud por los beneficiarios. El coste se

mide en unidades monetarias mientras que el beneficio necesita ser expresado en

valores cuantitativos relacionados con el aumento de la salud. Sin embargo, a diferencia

del análisis de coste-beneficio, los beneficios no tienen que expresarse en términos

monetarios. En farrmacoeconomía se expresa habitualmente en años de vida ajustados

por calidad (AVAC)91 (National Institute for Health and Care Excellence (NICE), 2010). El

ratio incremental de coste-efectividad (ICER) es la relación entre la diferencia de los

costes y la diferencia de los beneficios de dos intervenciones. La empresa conoce este

umbral en un determinado país; por lo tanto, la empresa lleva a cabo el CEA y calcula el

número de AVACs ganados si el medicamento se proporciona en un solo país. Dado que

la empresa es consciente tanto de umbral y el número de AVACs, ofrece al país el

medicamento a un precio determinado. No obstante, la empresa puede distorsionar al

alza el número de AVACs para obtener mayores beneficios. Entonces, es el país quién

podría revisar el CEA de la empresa y aplicar su propio CEA para revelar el precio justo.

Es importante señalar que este CEA requiere de recursos e inversión monetaria por parte

del país.

En general, las ventajas y desventajas antes mencionadas han impulsado la

investigación teórica y empírica particularmente en los últimos años. Además, la

interdependencia entre los mercados debido a la implementación del ERP y la presencia

del PT puede haber cambiado la estrategia de precios y lanzamiento, y podría llevar a

observar menores diferencias de precios en entre los países.

La tesis en primer lugar presenta una revisión sistemática que tiene como objetivo

ofrecer una perspectiva global de los análisis teóricos y empíricos originales de las

interdependencias globales con respecto a los precios de los medicamentos y al

lanzamiento, así como encontrar cuáles son los principales factores que influyen tanto en

91 El AVAC sirve para medir la carga de una enfermedad, incluyendo tanto la calidad como la cantidad de vida que se vive. El modelo de AVAC requiere una utilidad independiente, neutral al riesgo y una visión del trade-off constante y proporcional. El AVAC se basa en el número de años de vida ajustados por calidad que aportaría la intervención al paciente. Cada año, en perfecto estado de salud se le asigna el valor de 1,0 hasta un valor de 0,0 si se está muerto. Si los años de más no se viven con plena salud, por ejemplo, si el paciente pierde una extremidad, o si es ciego o tiene que usar una silla de ruedas, los años de vida adicionales se les asigna un valor entre 0 y 1 para tomar en cuenta esta carga.


los precios como en los retrasos de lanzamiento de nuevos fármacos. Una vez alcanzado

el objetivo anterior, la tesis desarrolla una parte teórica y una parte empírica. De esta

manera, se desarrolla un modelo teórico basado en un modelo de negociación entre la

industria farmacéutica y dos agencias de salud de dos países cualesquiera. Este modelo

tiene como objetivo analizar la conveniencia de la aplicación del ERP en lugar del CEA

como política de contención del gasto farmacéutico. En última instancia, la tesis

desarrolla un modelo empírico que tiene por objeto el análisis del trade-off entre el precio

y el retraso en el lanzamiento de los fármacos, y el impacto de la política de ERP sobre

ambos, precio y momento de lanzamiento.

Esta tesis está organizada en tres capítulos. Cada capítulo podría suponer una

investigación individual, por lo que cada uno está dotado de introducción propia,

desarrollo y conclusiones. El primer capítulo articula lo que ya se conoce, el segundo

capítulo desarrolla un modelo teórico y el tercer capítulo muestra la evidencia basada en

el desarrollo de un modelo empírico.

Objetivos

El capítulo 1 proporciona una revisión sistemática basados en la metodología

PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses) (Moher

et al., 2009) de estudios científicos originales. Los trade-offs que hemos mencionado

anteriormente han impulsado la investigación teórica y empírica particularmente en los

últimos años. En este capítulo se revisa lo que sabemos sobre los principales factores

que influyen en los precios y en el retraso en el lanzamiento de nuevos fármacos, y si

existen además patrones generales que pudieran derivarse de modelos económicos con

el objetivo de explicar los intereses estratégicos de los gobiernos, las aseguradoras

públicas y la industria farmacéutica, si hay alguna evidencia empírica sobre dichos

factores en los países de la OCDE, y si el ERP y el PT hacen que los mercados sean

inseparables a la hora del lanzamiento y la fijación del precio.

El capítulo 2 sugiere un juego basado en un modelo de negociación en virtud de

lanzamiento secuencial de un nuevo medicamento por una farmacéutica en dos países

cualesquiera basado en un procedimiento de take-it-or-leave-it-offer bajo información

asimétrica (Muthoo, 1999). El modelo se basa en la mejor estrategia de la empresa


farmacéutica a la hora de lanzar secuencialmente un nuevo producto, teniendo en cuenta

el tamaño de los países, el coste de llevar a cabo un CEA por parte de los países, el

coste del retraso en el lanzamiento y los precios fijados por cada una de las políticas de

precios. De manera breve, el país elige entre aplicar el ERP sin ningún coste adicional y

el CEA con su correspondiente inversión. La empresa acepta o rechaza un precio y elige

su secuencia de países óptima de lanzamiento. Se introducen dos innovaciones

importantes que requieren especial atención: en primer lugar, incluimos dos diferentes

tipos de países en función del método de ERP aplicado, ya sea el precio mínimo o el

precio medio observado. En segundo lugar, introducimos el coste del retraso en el

lanzamiento y el coste relacionado con la aplicación del CEA para comprobar si la

empresa declara el verdadero valor de AVACs del medicamento.

El capítulo 3 replica en primer lugar dos estudios empíricos (Danzon and Epstein,

2008, Verniers et al., 2011) donde se han aplicado modelos econométricos de fijación de

precios y lanzamiento. El mismo tratamiento de datos y metodología llevada a cabo por

los dos estudios se han aplicado a nuestra base de datos. Por un lado, se replica el

estudio de Danzon y Epstein publicado en 2008 (Danzon and Epstein, 2008) con nuestra

base de datos que cuenta con datos más recientes, y se comparan nuestros resultados

con los suyos. Los datos de Danzon y Epstein (Danzon and Epstein, 2008) abarcan los

años 1992 a 2003, mientras que nuestros datos cubren el período 2004-2010. De manera

similar, se replica el estudio de Verniers et al. publicado en 2011 (Verniers et al., 2011)

para comprobar si los resultados han cambiado debido a la utilización de datos más

recientes (2010 vs. 2008) y si los resultados son robustos a la elección de la lista de

países. Verniers et al. (Verniers et al., 2011) aplica su modelo a una base de datos que

contiene un gran conjunto de países, ricos y pobres. Sin embargo, en esta tesis el

modelo se restringe a países desarrollados. Luego, se desarrolla un modelo empírico

centrado en el análisis del trade-off entre los precio y el retraso en el lanzamiento, así

como el impacto de la política de ERP en los precios y el lanzamiento, controlando por

las características de moléculas, de regulación y de país. Desarrollamos dos ecuaciones,

una para el retraso en el lanzamiento y la otra para el precio de lanzamiento. Utilizamos

los datos de la base de datos de IMS Health para 70 nuevas moléculas lanzadas en 20

países y pertenecientes a 11 clases terapéuticas diferentes, todas ellas aprobadas


mediante el procedimiento centralizado de la EMA92 (Agencia Europea del Medicamento)

durante el período de estudio, 2004-2010. Se toman las ventas ambulatorias y

hospitalarias en euros a precio de fabricante y en unidad de volumen (IMS SU) con datos

anuales. La contribución de este capítulo a la literatura previa analizada consiste en el

análisis de los datos a nivel de presentación farmacéutica93, la consideración del precio

relativo de lanzamiento94 como variable endógena en la ecuación de precio de

lanzamiento, el estudio del retraso en el lanzamiento como variable de tiempo de

duración y el análisis del mercado hospitalario. Además, se introduce el tamaño del país

y el poder adquisitivo del país como variables explicativas adicionales en el modelo.

Planteamiento y metodología

Revisión de la literatura

En el capítulo 1 se lleva a cabo una revisión sistemática de la literatura de 1995 a

abril de 2012, y se sintetizan los principales hechos, ideas y resultados encontrados. La

búsqueda cubre los modelos teóricos y empíricos. El resto de este capítulo está

organizado de la siguiente manera. En un primer lugar se discuten los estudios teóricos.

Luego se examinan los estudios empíricos y finalmente se recoge una discusión.

La estrategia de búsqueda se llevó a cabo en abril de 2012 en las siguientes

bases de datos: PubMed, EconLit y Web of Knowledge, usando diferentes

combinaciones de palabras clave, tanto en inglés británico como americano, y abarcando

el período comprendido entre enero de 1995 a abril de 2012. Esta búsqueda se

complementó con otra adicional, utilizando las mismas palabras clave en NBER (The

National Bureau of Economic Research), la Universidad de York, el CHEPA (Centre for

Health Economics and Policy Analysis), la LSE (London School of Economics) Health, el

CRES (Centre de Recerca en Economia i Salut), y la Facultad de Economía de la

92 Antes EMEA.

93 Se definen dos productos con la misma presentación farmacéutica cuando ambos productos pertenecen a la misma molécula i y contienen la misma cantidad de principio activo por standard unit (SU). Véase la definición de SU en la sección 4.4 de este resumen.

94 Se define en el anexo de este resumen


Erasmus University Rotterdam. Por último, también se incluyeron algunos documentos

referenciados por expertos. Se hicieron búsquedas por separado utilizando la siguientes

palabras clave: “("pharmaceutical* pric*") and (new or launch* or patent), ("drug* pric*")

and [new or launch* or patent) and ("medicine* pric*") and (new or launch* or patent)”.

Las referencias encontradas en la búsqueda sistemática fueron evaluadas

mediante la revisión del título, el resumen y el tipo de publicación con el fin de identificar

los artículos relevantes. Todos los trabajos pertinentes que cumplieron los criterios de

inclusión fueron clasificados siguiendo una estrategia de selección de dos etapas.

En la primera etapa se aplicaron los criterios de selección para todos los estudios

analizados. En la segunda etapa, los estudios escogidos en la primera fueron

seleccionados individualmente para evaluar las aportaciones teóricas y empíricas

realizadas en cada caso. En la primera etapa, sólo se incluyeron los estudios que

cumplieron los criterios establecidos en la Figura 4.1. Cada artículo fue secuencialmente

evaluado en base a cuatro criterios, desde el primero al cuarto: ser un artículo original,

estar publicado en una revista científica, estar escrito en lengua inglesa y estar enfocado

en el precio y lanzamiento de medicamentos bajo patente. A medida que algún criterio no

se cumplía, el artículo era excluido.


Figura 4.1 Diagrama de flujo del proceso de selección de la literatura

En la segunda etapa, se evaluó el texto completo de cada artículo seleccionado.

Se escogieron únicamente aportaciones teóricas originales en el desarrollo de nuevas

perspectivas de análisis y trabajos empíricos que desarrollasen un modelo causal cuya

variable endógena fuera el precio o el lanzamiento de un medicamento para una muestra

de al menos dos o más países.

Finalmente, la revisión consta de 22 artículos, 12 basados en modelos teóricos y

17 en modelos empíricos.

Modelo teórico

Se consideran dos países, C y D. Cada país i (i = C, D) y una sóla empresa

negocian la adquisición de un medicamento bajo patente. Cada país reacciona a la oferta

de la empresa y presenta un precio basado en su política de precios entre el CEA y el

ERP que ha elegido previamente. Cada país i tiene una disposicón a pagar (WtP, de aquí

en adelante) por AVAC por un nuevo medicamento y poniéndolo a disposición de la


población. Se considera también una empresa que se comporta como productor

monopolista del nuevo fármaco bajo patente que vende a las agencias de salud de los

países. La empresa no está localizada en niguno de los países i y decide el orden de

estos para el lanzamiento del medicamento. La empresa vende el medicamento durante

dos periodos. Este fármaco está destinado al tratamiento de enfermedades crónicas y no

requiere de ningún copago. El medicamento está autorizado mediante procedimiento

centralizado95. Por lo tanto, los países no son capaces de no autorizar el medicamento.

Ellos son solamente capaces de no introducir el medicamento en las listas de reembolso.

El medicamento ya ha lanzado en J (j= 1,2,3...J) países. Así, los jugadores del juego son

la empresa farmacéutica y una agencia de salud en cada país i.

Consumidores

Los consumidores en cada país se definen como aquellos pacientes que pueden

ser tratados por el medicamento vendido por la empresa farmacéutica. Por lo tanto la

demanda de un medicamento en un país i (i = C, D) se define como la tasa de

prevalencia de las enfermedades crónicas para las que el medicamento está indicado.

Esta tasa de prevalencia y por ende, la demanda, está representada exógenamente por

qi96

y representa el tamaño del país.

El medicamento se suministra sin copago alguno, por lo tanto, los consumidores

no tendrán que pagar por él. La agencia de salud de cada país será el único pagador de

la medicina. En consecuencia, se supone que la demanda total del fármaco es exógena y

está representado por qi.

Agencias de Salud

Las agencias de salud deben tener en cuenta el beneficio bruto proporcionado por

el medicamento, representado por el número de AVACs que genera, función de la tasa

de prevalencia qi de su país. Se denota por Bi,

95 En la UE, el proceso centralizado es necesario para productos biotecnológicos como para medicamentos huérfanos y otros fármacos opcionales. En cualquier caso, la Comisión Europea aprueba la autorización a tenor de los recomendaciones de la EMA. La autorización se puede obtener via procedimiento de reconocimiento mutuo donde las empresas solicitan autorización a un Estado Miembro y se archiva para el reconocimiento en otros países de la UE. No obstante, nuestro análisis solamente considera el procedimiento centralizado, que es el más prevalente.

96 Se debe señalar que qi no cambia significativamente a lo largo de los años ya que el medicamento está indicado para pacientes crónicos.


con

WtP WtP si país D

WtP si país C

Cada país tiene un umbral de ICER diferente para un AVAC, ese umbral se define

como su WtP. Este umbral podría representar una medida del coste de oportunidad o el

valor del consumo de salud (Claxton et al., 2013). Los países declaran su WtP. En

particular, el país C tiene una WtP baja expresada por , y el país D tiene una alta

WtP expresado por . Por lo tanto, el Bi representa el valor monetario de la salud.

Los países deben tener en cuenta los gastos públicos (PE) ya que los pacientes

no pagan por el medicamento y la empresa se encuentra fuera del país i (i = C, D).

pi

WtPiY si CEA

con Y YF si el país se cree a la empresa

YiCEA en otro caso

with YiCEA YF

con WtPWtP si el país D

WtP si el país C

pIh si ERP con h ,min

Bi WtPi Y qi

WtPC

WtPD

PEi piqi i rqi ia

(4.1)

(4.2)

con


i 1 si retraso en el lanzamiento en el país i

0 en otro caso

i 1 si el país i aplica CEA

0 en otro caso

r coste de retrasar un periodo la provisión del medicamento para un paciente

El PE está compuesto por la cantidad de dinero pagado por el medicamento piqi y

el coste del retraso en el lanzamiento si correspondiese.

Los países i pagarán un precio pi que depende de los tipos de políticas de precio

aplicados. Los países pueden aplicar el CEA o el ERP.

Si el país i aplica el CEA pagará un precio igual a su WtP por cada unidad

AVAC proporcionado por el medicamento. Cada país declara su WtP ( para el país

C; para el país D). El número de AVAC puede ser igual a los AVACs propuestos

por la empresa si la empresa declara el número verdadero de AVACs, o menor que ,

es decir, , si la empresa declara un número de AVACs por encima de su verdadero

valor.

Cuando los países aplican el ERP, se consideran dos tipos de jugador

dependiendo de la agresividad de la fórmula del ERP. En primer lugar, tenemos el país

C, que se define como el más agresivo, cuyo fórmula del ERP consistirá en el

establecimiento de un precio máximo del nuevo producto igual al precio más bajo del

producto en el momento de su lanzamiento en el conjunto de países en los que el

medicamento ya ha sido lanzado (min). Por su parte, el país D, aplica una fórmula de

ERP consistente en fijar un precio máximo calculado como el precio de lanzamiento

medio del medicamento en el conjunto de países en los que el fármaco ha sido ya

lanzados ( ). En este modelo, el precio fijado cuando los países i aplican el ERP se

denota por .Por lo tanto, el país C escogerá el precio mínimo entre el

precio mínimo internacional ( pImin ) y cualquier precio establecido previamente. A su vez,

el país D escogerá el precio promedio internacional ( pI ). En caso de sufrir retraso en el

lanzamiento, el país D tendrá en cuenta a la hora de calcular su promedio cualquier

piCEA

WtP

WtP YF

YF

YiCEA

pIh (h min,)


(4.5)

(4.4)

precio fijado previamente (en este modelo, el precio ya establecido en el país C). De

manera formal,

pERPminC

pImin min p

1, p

2...p

j...p

J si {C, D}

pImin min p

1, p

2...p

j...p

J, p

D en otro caso

pERPD

pI

pj

J

j1

pC

J 1 si {C,D}

pI

pj

J

j1

J

en otro caso

Cuando el país i aplica el CEA, está invirtiendo una cantidad de dinero a para

verificar si el número de AVACs del medicamento YF declarados por la empresa es cierto

o no. Esta inversión se define como cCEAi, siendo igual para todos los países.

El país i sufre un retraso en el lanzamiento si el medicamento se lanza

previamente en uno de los países i. Definimos el coste unitario del retraso en el

lanzamiento ri para el país i como el coste que supone en salud no suministrar el

medicamento a los pacientes durante un período. Por lo tanto, el coste total del retraso

en el lanzamiento para el país i se define como .

Suponemos que el objetivo de una agencia de salud es maximizar los beneficios

proporcionados por el medicamento menos los gastos públicos (PE) asociados a su

compra. Por lo tanto, la función objetivo (OF) de la agencia de salud se puede escribir

como,

OFi Bi PE (WtPiY pi )qi i rqi ia

rqi

(4.3)


(4.9)

(4.6)

(4.8)

(4.7)

Ahora, definiremos algunos supuestos que posibiliten el desarrollo y la resolución

del problema.

Supuesto 1

La WtP por AVAC de el país D es estrictamente mayor que la del país C.

Los precios y prouestos por la empresa a los países C y D respectivamente se

definen como el producto de WtPi por el número de AVACs declarados por la empresa

pFWtPCYF

pF WtPDYF

Por lo tanto, dado que la empresa maximiza sus beneficios, el precio propuesto por la

empresa al país D es estrictamente superior al ofrecido al país C

Supuesto 2

Existe variabilidad de los precios entre los J países donde el medicamento se ha lanzado

previamente. Por lo tanto, el precio medio internacional es estrictamente mayor que el

precio mínimo internacional.

WtPD WtPWtPC WtP

pF

pF

pF

pF

pF pF


(4.12)

(4.13)

(4.11)

(4.10)

Supuesto 3

Dado un país que aplica el CEA, los AVACs revelados por el CEA llevado a cabo por el

país no serán superiores a los declarados por la empresa

Concretamente, si la empresa es honesta y por lo tanto declara el verdadero número de

AVACs, los AVACs revelados por el CEA llevado a cabo por el país serán iguales a los

declarados por la empresa. A su vez, si la empresa no es honesta y por lo tanto no

declara el verdadero número de AVACs, los AVACs revelados por el CEA serán

estrictamente inferiores a los sugeridos por la empresa.

YiCEA YF si la empresa declara el verdadero número de AVACs

YiCEA YF en otro caso

De esta manera, el precio que resulta de la aplicación del CEA no será superior al precio

propuesto por la empresa.

piCEA piF si la empresa declara el verdadero número de AVACs

piCEA piF en otro caso

PI PI

min

YF YiCEA


(4.14)

(4.15)

(4.16)

Por lo tanto, debido a evidencia científica mostrada por el CEA, asumiendo que los

costes marginales (mc) de producción son 0 (mc=0) y dado que la empresa maximiza

sus beneficios, el precio del CEA será siempre aceptado por la empresa97.

Supuesto 4

En caso de que ambos países i lleven a cabo el CEA, el número de AVACs revelado por

ambas investigaciones, YiCEA, será idéntico para ambos países.

Supuesto 5

La creencia de los países sobre la honestidad de la empresa a la hora de declarar el

número de AVACs que ofrece el nuevo medicamento es igual a e idéntica para ambos

países i (i = C,D).

Supuesto 6

Bajo los supuestos 3 y 5, el precio esperado por el país i (i = C,D) si llevara a cabo el

CEA es,

97 La empresa conoce el WtP de los países i (i = C,D), por lo tanto si el WtP es demasiado bajo (por debajo de un umbral dado), la empresa ni siquiera iniciará las negociaciones para lanzar en ese país.

Pr(YF YiCEA) , i

Pr(YF YiCEA) 1 , i

piF (1) piCEA E pi CEA


(4.17)

(4.18)

(4.19)

Por lo tanto, bajo los supuestos 1 y 5, el precio esperado por el país D es estrictamente

superior al precio esperado por el país C.

Supuesto 798

Si el precio propuesto por la empresa es superior (inferior) que el precio de referencia

internacional , el precio revelado por el país cuando lleva a cabo el CEA será

también superior (inferior). Por lo tanto, el precio esperado del CEA no será

inferior (superior) al precio de referencia internacional.

Supuesto 899

El precio medio entre el precio medio de referencia internacional y el precio del país C es

aproximadamente igual al precio medio de referencia internacional. Por lo tanto, se

asume que el conjunto de J países es lo suficientemente grande que el precio medio no

se ve afectado por precios posteriores.

98 Este supuesto garantiza el trade-off entre la elección del CEA y el ERP. Si el país i aplica el CEA, pagará un precio más alto para evitar tener acceso al medicamento con retraso.

99 El supuesto 8 hace más sencillo los resultados y no afecta a las conclusiones obtenidas.

E pD CEA E pC CEA

pF

pIh piCEA

E pi CEA

If pF pIh piCEA pI

h E pi CEA pI

h

If pF pIh piCEA pI

h E pi CEA pI

h


(4.20)

(4.21)

Dado p I

pj

j1

J

J

entonces,

pj pC

j1

J

J1

p I con pC

p Imin

pF

pCEAC

Supuesto 9

La diferencia de precios entre el precio de referencia internacional y el precio propuesto

por la empresa es el mismo independientemente del tipo de país. De esta manera, el

incentivo para aplicar el ERP es el mismo independientemente del tipo de país (con WtP

baja o alta). De manera formal,

La empresa

Dado que la industria farmacéutica se caracteriza por presentar altos costes fijos

(F) y bajos costes marginales (mc=0), en este modelo asumimos que los costes

marginales de producir el medicamento son cero (Berndt et al., 1996, Cockburn and Anis,

1998, Suslow, 1996). Así, F> 0 representa los costes fijos de I+D, del proceso de

aprobación y comercialización de los medicamentos en todos los países. Estos costes

son fijos e independientes del número de personas o países que utilicen los

medicamentos. La empresa vende el medicamento al país i (i = C, D) al precio pi. Este

precio pi es el precio máximo al cual las agencias de salud y la empresa farmacéutica

acuerdan pagar y vender el nuevo fármaco100. Si el precio del país proviene de aplicar un

CEA, dada la evidencia científica, la empresa aceptará dicho precio. A su vez, si el precio

del país proviene de aplicar el ERP, la empresa podrá aceptarlo o rechazarlo. En el caso

de rechazarlo, la empresa retrasará el lanzamiento en dicho país. En todo caso,

independientemente del momento, la empresa se compromete a poner en marcha el

100 Somos conscientes que las asociaciones organizadas de compra como asociaciones de hospitales o de farmacias podrían obtener determinados descuentos al precio pi, no obstante, no han podido ser considerados en esta tesis.

pImin p

F pI

pF


(4.22)

medicamento y servir a toda la demanda en este país (qi) a precio pi. La venta del

medicamento sin subvenciones no se considera una opción alternativa. De esta manera,

se supone que el objetivo del productor monopolista de la medicina es maximizar la

función de los beneficios acumulados por ventas (dos periodos) y se puede escribir

como,

OFF pitqit FiC

D

t1

2

con

qit 0 si el medicamento no se comercializa en el país i en el año t

Se asume también que la empresa no está localizada en ninguno de los países i (i

= C,D).

El timing

El timing de este juego se desarrolla de la siguiente manera. El juego tiene 4

etapas. En la etapa 1, los países C y D eligen sus políticas de fijación de precios: CEA o

ERP y la empresa propone un precio para el nuevo medicamento. En la etapa 2, los

países comunican sus precio de acuerdo a sus políticas de precio. En la etapa 3, dado

que el lanzamiento es secuencial (digamos primero el país C y luego el país D, o

viceversa), la empresa elige la secuencia de lanzamiento, es decir, si retrasar el

lanzamiento en el país D o en el país D, y vende el medicamento al primer país de la

secuencia. En la etapa 4, la empresa vende el medicamento en el segundo país. Dado

que la empresa maximiza sus beneficios, vende siempre en ambos países i (i=C, D).

A priori, se debe señalar que el país i (i =C, D) puede elegir entre una política de

fijación de precios que finalmente requiere una inversión monetaria, el CEA, y otra

política que no implica ningún tipo de inversión, el ERP. Sin embargo, aplicando el ERP

el país corre el riesgo de que su precio sea rechazado por la empresa y por lo tanto sufra

algún retraso en el lanzamiento. A su vez, aplicando el CEA, el país se arriesga a realizar

una inversión inútil si la empresa declara el verdadero número de AVACs. Para mostrar


el timing de manera más clara, véase el árbol de decisión en la figura 4.2.

En general, los jugadores maximizan su función objetivo y resuelven el juego

aplicando inducción hacia atrás. En ambos casos, cuando la empresa declara el

verdadero número de AVACs y cuando declara un número inferior al mismo, el juego se

resuelve para cada par de políticas de fijación de precios posible: i) ningún país aplica el

ERP, ii) solamente el país D aplica el ERP, iii) solamente el país C aplica el ERP y, iv)

ambos países aplican el ERP. De esta manera, se calculan las condiciones de la

secuencia óptima de lanzamiento para el empresa, bien primero el país C y luego D, o

viceversa. Luego, dada la secuencia óptima de lanzamiento, se compara el output de los

países para cada política de precios.

En este resumen se muestra cómo se ha operado para un par de las cuatro

posibles combinaciones de políticas de precios. Concretamente se muestra a

continuación el par ii) donde solamente el país D aplica el ERP. En primer lugar, cuando

la empresa declara un número de AVACs por encima de su valor verdadero, y en

segundo lugar, cuando la empresa declara el número verdadero de AVACs.

Supuesto 10

El lanzamiento del medicamento es secuencial y la empresa sigue vendiendo en la Etapa

4 al país donde ya vendió previamente en la Etapa 3.

Supuesto 11

Si solamente uno de los países i (i = C,D) aplica el ERP y la empresa rechaza su precio,

la empresa retrasará el lanzamiento en dicho país i. Sin embargo, si ambos países

aplican ERP y la empresa rechaza sus precios, la empresa elegirá su secuencia óptima

de lanzamiento de acuerdo al país que le ofrezca mayor ingreso en virtud del precio pi y

la cantidad a vender qi.


Figura 4.2 Árbol de decision


(4.23)

Fijación de precios y lanzamiento secuencial

El problema se resuelve para cada combinación possible de pol´íticas de precio i)

ningún país aplicael ERP, ii) solamente el país D aplica el ERP, iii) solamente el país D

aplica el ERP, y iv) ambos países aplican el ERP. En este resumen, a modo de ejemplo,

resolveremos el caso ii).

La empresa declara un número de AVACs de acuerdo a su valor verdadero

ii) Únicamente el país D aplica ERP

El país D decide aplicar ERP en la etapa 1. Si el precio medio internacional es

inferior al precio propuesto por la empresa, es decir, pI pF , bajo el supuesto 11, la

empresa retrasa el lanzamiento en el país D. Dado que las políticas regulatorias de

precio se fijan ex-ante, si pI pF , el país D pagará pI

en cualquier caso. A su vez, el

país C lleva a cabo el CEA y pagará pF. Dado que la empresa declara el verdadero

número de AVACs, bajo el supuesto 1, pagará .

En relación al excedente de las agencias de salud, se debe señalar que el precio

considerado por el país C cuando aplica el CEA es un precio esperado de acuerdo a los

supuesos 5 y 6, dado que el país C tiene incertidumbre sobre el número de AVACs

declarados por la empresa.

los beneficios de la empresa son,

OFF 2p

FqC p

I qD F if C, D p

FqC 2p I

qD F en otro caso

y el excedente de las agencias de salud son,

pF


(4.24)

(4.25)

(4.26)

OFC (Wtp

CYF E[pC]CEA)qC a if C, D

(WtPCYF E[pC ]CEA)qC a rqC en otro caso

OFD (WtpDYF pI


)qD en otro caso

PPR101

{C, D} if p

F

pI

qD

qC

{D,C} en otro caso

Si únicamente el país D aplica el ERP, la empresa opta por retrasar el lanzamiento en el

país D si y sólo si el ratio de precios entre el país C y el país D es mayor que el ratio de

tamaños de los países entre D y C, en otro caso la empresa retrasará el lanzamiento en

el país C. La relación de precios es el ratio entre el precio del CEA y el precio medio de

referencia internacional. Dado que la relación de los precios es menor que la unidad,

entonces, una condición necesaria para que la empresa elija la secuencia {C, D} es que

el país C debe ser estrictamente mayor que el país D.

En la Tabla 4.1 se resumen los PPR correspondientes a cada uno de los pares de

políticas. Se muestra para cada par de políticas de fijación de precio i), ii), iii) and iv), bajo

qué condiciones (pD, pC, qC, qD) la empresa elige su secuencia óptima de lanzamiento. Se

pone de manifiesto que siempre existe un trade-off entre los precios y los tamaños de los

países.

101 PPR: Resultado preliminar


(4.27)

Tabla 4.1. PPR para la secuencia óptima de lanzamiento (pi , qi)

En la siguiente figura, Figura 4.3, se muestra bajo qué par de política de precios

(i), ii), iii) and iv)) y qué par de ratios de precios y ratios de tamaño del país, la empresa

elige su secuencia óptima de lanzamiento, atendiendo a las cuatro diferentes relaciones

que existen entre los precios que propone la empresa y los precios del ERP:

a) pF pI y p

F pI

min

b) pF pI

y pF pI

min

c) pF pI y p

F pI

min

d) pF pI y p

F pI

min

Debajo de la figura se presenta una tabla anexa, Tabla 4.2, que contiene de

manera detallada la información correspondiente a cada zona. A título ilustrativo se

muestra aquí un ejemplo atendiendo a la relación de precios

a) Si pF pI y p

F pI

min

Que implica

pFmin

pF

p

F

pF

pImin

pI

pF

pI si

pF

pF

pImin

pI

pFmin

pF

pImin

pI

pF

pF

p

F

pI en otro caso

Secuencia / Política

i) Ningún país aplica ERP

ii) Solamente el país D

aplica ERP

iii) Solamente el país C

aplica ERP

iv) Ambos países aplican

ERP

{C,D}

{D,C}

pF

pF

qD

qC

pF

pI

qD

qC

pImin

pF

qD

qC

pImin

pI qD

qC

pF

pF

qD

qC

pF

pI

qD

qC

pImin

pF

qD

qC

pImin

pI qD

qC


Bajo esta secuencia de ratios de precio, la Figura 4.3 muestra la secuencia óptima

de lanzamiento para cada par de políticas y para cada región de combinación de ratios

de precios y tamaño de los países.

Figura 4.3 Regiones de secuencias óptimas de lanzamiento si a)


Tabla 4.2. Regiones de secuencias óptimas de lanzamiento a)

Región/ Políticas A B X102 E F i) No ERP {C,D} {C,D} {D,C} {D,C} {D,C} ii) D ERP {C,D} {C,D} {C,D} {C,D} {D,C} iii) C ERP {C,D} {D,C} {D,C} {D,C} {D,C} iv) Both ERP {C,D} {C,D} {C,D} {D,C} {D,C}

El país i no se cree el número de AVACs declarado por la empresa

Esta parte del árbol se soluciona de manera similar a la parte anteriormente

resuelta. En este resumen únicamente destacamos las diferencias.

En este caso, si el país aplica ERP, al igual que en el caso anterior, el país no

será capaz de descubrir el valor real del medicamento y pagará por tanto el precio

internacional de referencia pIh que podría ser superior o inferior al precio revelado en

caso de aplicar el CEA. Sin embargo, si el país decide aplicar el CEA revelará un número

de AVACs YiCEA menor o igual que el declarado inicialmente por la empresa YF . De esta

manera, el país pagará un precio piCEA inferior al precio pF propuesto por la empresa.

Por lo tanto, por un lado es ahora menos probable que la empresa acepte los precios de

referencia internacional pIh que si hubiera sido creída por el país i (i = C, D), lo que

implica que los países que aplican ERP tendrán más probabilidades de experimentar

retrasos en el lanzamiento. Además, las regiones bajo las cuáles la empresa elige su

secuencia óptima de lanzamiento cambian (Figura 4.2). Por otro lado, el valor esperado

del precio del medicamento E[pi ]CEA será superior cuando la empresa declare un

número de AVACs por encima de su valor verdadero que cuando declare su valor

verdadero.

Dada la secuencia óptima de lanzamiento para la empresa, se compara el

bienestar de los países en el marco de cada política de precios, el CEA y el ERP, con el

objetivo de saber cuál de ellos es más conveniente para los países. Por lo tanto, ya que

se ha resuelto el problema por inducción hacia atrás, se ha llevado a cabo esta

comparación para cada país teniendo en cuenta la secuencia óptima de lanzamiento

102 Bajo pI

pF pF pI


para la empresa. Esta comparación se ha llevado a cabo en tres pasos. En primer lugar,

se ha comparado el mejor resultado para cada país bajo la misma política de precios y en

la misma secuencia de lanzamiento del país, es decir, utilizando el CEA (ERP) bajo las

secuencias {C, D} y {D, C}, respectivamente. Luego, en un segundo paso, se han

comparado los mejores resultados entre el ERP y el CEA para cada secuencia de

lanzamiento. En un tercer paso, se ha comparado la mejor política de precios para cada

secuencia de lanzamiento.

Modelo empírico

En el capítulo 3 desarrollamos un modelo empírico basado en dos ecuaciones,

una ecuación de precio relativo de lanzamiento y otra ecuación de retraso en el

lanzamiento. Previamente, llevamos a cabo la réplica de dos trabajos con nuestra base

de datos, con el objetivo de saber si el uso de una base de datos más reciente y una lista

diferente de países podría influir en los resultados aún utilizando la misma metodología.

Hemos utilizado los datos de la base de datos de IMS Health para 70 nuevas

moléculas lanzadas en 20 países y pertenecientes a 11 clases terapéuticas diferentes,

todas ellas aprobadas mediante el procedimiento centralizado de la EMA103 durante el

período de estudio, 2004-2010. Se han tomado las ventas ambulatorias y hospitalarias en

euros a precio de fabricante y en unidad de volumen (IMS SU) con datos anuales. Los

ingresos se ajustaron por la inflación usando el IPP (Índice de Precios del Productor)

específicos disponibles en el FMI (Fondo Monetario Internacional) tomando 2005 como

año base.

Réplica del modelo de Danzon y Epstein (2008)

Replicamos el estudio de Danzon y Epstein (Danzon and Epstein, 2008) (D&E, a

partir de ahora) con nuestra base de datos con el objetivo de comparar nuestros

resultados, obtenidos de datos más recientes, con lo suyos. La base de datos de D&E

abarca los años 1992-2003 para 11 moléculas pertenecientes a 12 clases terapéuticas

en 15 países que experimentaron el lanzamiento de una nueva subclase poco antes o

durante el período de estudio. D&E recogen datos trimestrales sobre ventas ambulatorias

103 Antes EMEA


en euros y en volumen (IMS SU) a precio de fabricante104. D&E ajustan por inflación

usando el IPP disponibles en el FMI, con 2003 como año base y convertido a dólares

utilizando el promedio del tipo de cambio específico para cada país en 2003. Se calcula

el precio por dosis de cada fármaco como el ratio entre los ingresos totales y las SU

vendidas en cada trimestre105. El mismo tratamiento de datos y metodología llevados a

cabo por D&E se han aplicado a nuestra base de datos con el objetivo de obtener una

comparación lo más precisa posible, aunque con ciertas limitaciones que comentaremos

en esta sección.

D&E estiman por separado una ecuación de lanzamiento y una ecuación de

precio de lanzamiento. Estiman en primer lugar el modelo de lanzamiento basado en una

regresión clog-log, ajustando los errores estándar por grupos de moléculas. También

estiman un modelo clog-log de efectos aleatorios para tener en cuenta la heterogeneidad

a nivel de molécula. Posteriormente, calculan los efectos marginales para cada variable

independiente. En particular, los efectos marginales para las variables continuas se

calcularon como el Average Marginal Effect (AME). El efecto marginal de las variables

categóricas se calculó como el cambio discreto del nivel base (x = 0) al nivel de

referencia que supone la presencia del atributo (x = 1).

El modelo de precio de lanzamiento se estima en virtud del procedimiento en dos

pasos de Heckman (Heckman, 1979) para tener en cuenta el posible sesgo de selección

producido por la correlación entre la probabilidad de lanzamiento y el precio. En la

primera etapa se estima un modelo lanzamiento basado en una regresión cloglog de

acuerdo a la ecuación de lanzamineto descrita anteriormente. Se calcula la razón inversa

Mills (IMR) (Lee, 1983)106 y se introduce como variable de control en la ecuación precio

104 La unidad definida como la SU de IMS representa la dosis minima de administración de cada formulación como por ejemplo un comprimido o cápusla para sólidos, o 5ml. para líquidos. Los datos de precios de IMS para los EEUU no reflejan descuentos fuera de facture dados por los fabricantes a determinados planes de salud y por lo tanto están sesgados al alza al contabilizar los ingresos netos del fabricante.

105 Diferentes presentaciones farmacéuticas aparecen en un determinado país y trimestre (ej: pastillas y capsulas, posiblemente con diferente cantidad de principio activo). Por lo tanto, para convertir las observaciones existentes en un país y un trimestre en una única observación se define el precio medio ponderado por volumen por unidad. Las presentaciones farmacéuticas idénticas que han sido lanzado por diferentes empresas que comparten la venta del medicamento se incluyeron también en el cálculo de dicho precio. 106 La IMR para la molécula i in el país j y en el momento t, Mijt, se calcula usando la probabilidad predicha de lanzamiento

de la regresión clog-log como , donde es la función de densidad de la Normal

estándar y la función de distribución de la Normal estándar.

p̂sjt Msjt [1( p̂sjt )]

( p̂sjt )[]

()


de lanzamiento. Esta ecuación de precio se estima por mínimos cuadrados ordinarios

(MCO) con errores estándar agrupados por moléculas. La variable dependiente es el

logaritmo del precio promedio ponderado por volumen para cada observación molécula-

país-año. Para tener en cuenta las características de la molécula no observadas D&E

también ofrecen los resultados del estimador de mínimos cuadrados generalizados

(MCG) con efectos aleatorios en la molécula.

En el tratamiento de datos se aplica el mismo procedimiento de D&E para el

ajuste de la inflación utilizando el IPP disponible en el FMI, sin embargo, se toma 2005

como año base. Las ventas en euros han sido convertidas a dólares aplicando el tipo de

cambio del FMI. El precio de los medicamentos se ha calculado de acuerdo a la

metodología de D&E. Aunque D&E distinguen entre moléculas superiores e inferiores, en

este trabajo no hemos tenido en cuenta esta diferencia ya que solamente hemos incluido

el lanzamiento de nuevos medicamentos, por lo tanto, compararemos los resultados de

D&E de la subclase de moléculas superiores con los resultados de nuestra regresión,

entendiendo que los nuevos medicamentos bajo patente no pueden pertenecer a la

subclase de moléculas inferiores definida por D&E. D&E clasifican los países en tres

categorías: de precio alto en UE (Unión Europea), de precio bajo en la UE y de precio

alto fuera a la UE, siendo Alemania el país de referencia. Dado que la muestra de países

no es exactamente la misma, el número de países por cada categoría en los que la

molécula ha sido lanzadas previamente no puede ser comparado de manera directa.

Réplica del modelo de Verniers et al. (2011)

Replicamos el estudio de Verniers et al. (Verniers et al., 2011) con nuestra base

de datos para comparar si los resultados han cambiado debido a la utilización de datos

más recientes (2010 vs. 2008) y si los resultados son robustos a la elección de la lista de

países. Verniers et al. aplicaron su modelo a un gran conjunto de países, ricos y pobres,

mientras que nosotros restringimos nuestra réplica a países desarrollados.

Verniers et al. tienen en cuenta, por un lado, el retraso en el lanzamiento del

medicamento i en el país j ( ), que se define como la diferencia medida en meses

entre el primer lanzamiento a nivel mundial y el lanzamiento posterior en el país

específico j. El precio de lanzamiento se define como logaritmo natural transformado del

precio de fabricante por cada gramo del medicamento i en el país j en el momento de su

LWij*


(a)

(b)

(4.29)

(4.28)

lanzamiento ( ). Verniers et al. consideran que se produce censura para las

medicamentos de los países para los que el lanzamiento al final del periodo de

observación no se ha producido. De esta manera, el tiempo de censura (Cij ) se define

como el tiempo entre la fecha de lanzamiento del medicamento en cada país y el final del

período de observación. Dado que los valores reales de y no se observan

debido a la existencia de censura por la derecha, los valores observados de y se

indican de tal manera que,

LWij LWij* if LWij

* Cij

LWij Cij otherwise

Además, solamente observamos las observaciones con LWij* Cij y por tanto,

LPij LPij*

Las ecuaciones estructurales son:

Donde Zij1 y Zij2 son definidas como variables explicativas. Las variables Zij1 y Zij2

comprenden el tamaño del país, el gasto en sanidad per cápita y el uso de determinadas

políticas regulatorias de precios como regulación directa del precio de fabricante, control

de beneficios, el uso del ERP, el uso del sistema de RP (precios de referencia) y la

evaluación económica como cuarta barrera. Además, comprenden el grado de protección

de la patente, la pertenencia a la EMA y la localización de la empresa. La variable de

competencia y la variable verano están también comprendidas en la variable Zij1. Zij2

LPij*

LWij* LPij

*

LWijLPij

LWij* 1LPij

* 2 (LPij*)2 'Zij1 uij1

LPij* 1LWij

* 2 (LWij*)2 'Zij 2 uij 2

(b)

(a)


incluye las mismas variables que Zij1 excepto la variable verano y la variable EMA. Sin

embargo, incluye la tasa de inflación y la DDD.

De acuerdo a Garen (Garen, 1984), Verniers et al. consideran el retraso de

lanzamiento y el precio de lanzamiento como variables endógenas. Por lo tanto, la

empresa y el regulador pueden decidir el retraso del lanzamiento con el objetivo de influir

en el precio de lanzamiento, y seleccionar el precio de lanzamiento con el objetivo de

influir en el retraso en del lanzamiento. Las variables omitidas contenidas en los términos

de error de ambas ecuaciones incluyen variables estratégicas no observables utilizadas

por la empresa y el regulador para elegir el valor óptimo del retraso y del precio,

respectivamente. Se puede esperar que estas variables estratégicas estarían

correlacionadas con el precio de lanzamiento y el retraso del lanzamiento

correspondientemente.

Verniers et al. para tener en cuenta la endogeneidad entre los precios y los

retrasos en el lanzamiento estiman un sistema de ecuaciones simultáneas usando un

procedimiento de mínimos cuadrados en tres etapas (MC3E), como en Bayus et

al.(Bayus et al., 2007). Además, los autores corrigen la censura a la derecha y el sesgo

de selección utilizando el procedimiento descrito en Vella (Vella, 1993) o Wooldridge

(Wooldridge, 2002). Incluyen los efectos aleatorios de los países en ambas ecuaciones

para tener en cuenta el hecho de que existen observaciones repetidas en todos los

países para la mayoría de los medicamentos.

Por un lado, para estimar la ecuación estructural de lanzamiento, Verniers et al.

primero estiman la forma reducida de la ecuación de precio de lanzamiento a través de

una regresión Tobit de tipo II (para tener en cuenta el hecho de que sólo observamos

precios si el medicamento ya se ha lanzado). Esta ecuación de precio contiene dos

variables que influyen únicamente en el precio pero no en el retraso de lanzamiento, la

dosis diaria definida (DDD) y la tasa de inflación, que sirven como instrumentos para el

precio en la ecuación de lanzamiento. Los residuos generalizados de la ecuación

reducida de precio se añaden a la ecuación de lanzamiento como término de corrección.

Sin embargo, nosotros solamente usamos un instrumento en el modelo ya que no hemos

sido capaces de calcular la DDD para la mayoría de las moléculas. Por otro lado, para

estimar la ecuación estructural del precio, Verniers et al. estiman primero la forma

reducida de la ecuación de lanzamiento a través de una regresión Tobit de tipo I (para

tener en cuenta de la censura a la derecha). Esta ecuación de lanzamiento contiene dos


variables que influyen sobre el lanzamiento pero no sobre los precios, la variable summer

y la EMA, que sirven como instrumentos para el retraso en el lanzamiento en la ecuación

de precio. En este caso se han incluido los dos instrumentos en el modelo replicado. Los

residuos generalizados de la ecuación reducida de lanzamiento se añaden como término

de corrección a la ecuación de precio.

Verniers et al . recopilan la base de datos IMS Health sobre medicamentos en de

5 grupos terapéuticos en 50 países durante el período de estudio, 1994-2008. Recogen

datos anuales sobre las ventas ambulatorias a precios de fabricante. Calculan el precio

por gramo en dólares estadounidenses para cada fármaco. Para hacer que los precios de

medicamentos sean comparables entre países, los precios en moneda local se

convirtieron a dólares utilizando el tipo de cambio vigente en el momento del

lanzamiento.

En nuestra réplica también utilizamos los datos de la base de datos de IMS

Health. Sin embargo, sólo tenemos en cuenta el lanzamiento de los nuevos

medicamentos en 20 países desarrollados para 11 clases terapéuticas durante el período

de estudio 2004-2010, todos ellos aprobados por el procedimiento centralizado de la

EMA. También hemos recogido las ventas ambulatorias por año a precio de fabricante.

Los precios en euros han sido convertidos a dólares USD aplicando el tipo de cambio del

Fondo Monetario Internacional. Por último, el precio del medicamento se ha calculado

como Verniers et al., por lo tanto, también se utiliza el precio por gramo en dólares

estadounidenses con el fin de hacer comparables los resultados.

Nuevo modelo precio y lanzamiento (NPLM)

Estimamos las ecuaciones de retraso en el lanzamiento y precio relativo de

lanzamiento por separado, y cada una de ellas se estima para el mercado ambulatorio y

el mercado hospitalario. Hemos tratado de estimar un sistema de ecuaciones para tener

en cuenta la endogeneidad entre el retraso en el lanzamiento y el precio, sin embargo, el

instrumento disponible para la ecuación de precio es débil107.

Usamos un modelo paramétrico de duración del riesgo de lanzamiento de un

medicamento en el tiempo t dadas las variables explicativas observadas, con los datos

107 Seleccionamos la variable inflación como instrumento del precio en la ecuación de lanzamiento. La correlación entre la inflación y el precio de lanzamiento es muy baja (0.02), por lo tanto no deberíamos usarla como instrumento.


censurados a la derecha para modelizar el retraso del lanzamiento de la molécula i en el

país j, que se define como el tiempo transcurrido entre el mes del primer lanzamiento a

nivel mundial de la molécula i y su lanzamiento en el país j. Se especifica la forma de la

tasa de riesgo, es decir, su dependencia del tiempo, con una distribución de Weibull que

asume una relación monótona respecto al tiempo. Puesto que no hemos sido capaces de

observar todas las variables que afectan al retraso en el lanzamiento hemos controlado

por la heterogeneidad no observada introduciendo una distribución gamma para el

término aleatorio del error. La selección del modelo ha seguido el método propuesto por

Kiefer (Kiefer, 1988). En un primer paso, de acuerdo con la función de densidad de la

variable endógena, como se muestra en las Figura 4.4 y 4.5, en términos generales,

tenemos una función monótonamente decreciente. En base a estas formas los modelos

más adecuados son los modelos paramétricos Weibull y Gompertz. Estos dos modelos

son de uso frecuente con datos que presentan tasa de fracaso monótonas, ya sea en

aumento o en disminución. La forma más común para el control de la heterogeneidad no

observable consiste en la introducción de una distribución paramétrica para el término de

error aleatorio y por lo tanto, estimar los parámetros de la función que genera tal término

de error. No hay un patrón estricto para elegir esa distribución. Basados en la literatura

previa, las dos distribuciones más utilizadas son la Gamma (Klein and Moeschberger,

1997, Lancaster, 1992) y la inversa de Gauss (StataCorp, 2009). De acuerdo a esto,

estimamos cuatro modelos paramétricos diferentes. En las Tabla 4.2 se muestra el AIC

(Criterio de Información de Akaike) para medir la bondad de ajuste de los modelos y las

pruebas estadísticas de la heterogeneidad no observada. Otra forma de selección de los

modelos se basa en el análisis de los residuos generalizadas de Cox-Snell. Si el modelo

ha sido seleccionado correctamente los residuos de Cox-Snell deberían presentar una

forma cercana a la función exponencial unitaria. Las figuras y tablas mostradas aquí en

relación a la selección del modelo de lanzamiento se refieren únicamente al mercado

ambulatorio. Idéntico proceso y resultado arroja el mercado hospitalario.


Figura 4.4 Función de densidad de retraso en el lanzamiento en meses. Mercado

ambulatorio

Tabla 4.3.Bondad del ajuste de los modelos paramétricos. Mercado ambulatorio

Weibull con Heterogeneidad

Gamma

Weibull con Heterogeneidad

Gaussiana Inversa

Gompertz con Heterogeneidad

Gamma

Gompertz con Heterogeneidad

Gaussiana Inversa AIC 38.14430201 38.142709 38.13897749 38.10271705


p=0.000

=26.21

p=0.000

=81.28

p=0.000

=8.63

p=0.002

Figura 4.5. Residuos de Cox-Snell para el modelo Weibull, Gamma. Mercado ambulatorio

1

21

21

21

2


(4.30)h(t, X) p(t)p1[U]

De acuerdo al proceso descrito anteriormente se estima un modelo de duración

paramétrico Weibull con heterogeneidad Gamma. Concretamente, estimamos un modelo

censurado a la derecha ya que todas los medicamentos en nuestro conjunto de datos se

lanzaron entre enero de 2004 y diciembre de 2010, sin embargo no todos los fármacos

se habían lanzado en los 20 países a finales de nuestro período de observación. Por lo

tanto, nuestros datos contienen observaciones censuradas a la derecha. El modelo

paramétrico de duración para el riesgo de lanzamiento no tiene en cuenta el uso de

covariables que varíen en el tiempo108, en su lugar se han utilizado los datos recogidos

en el año base109. Por lo tanto, tenemos:

donde el subíndice i=molécula, j=país, h es la tasa de riesgo de poner en

marcha, Xij son las covariables, los parámetros de las covariables, t el tiempo

transcurrido hasta el lanzamiento de la molécula i en el país j, p el parámetro de la forma

funcional110, U la variable aleatoria y la varianza de la heterogeneidad no observada111

(Jenkins, 2008, Keele, 2007). Las covariables Xij incluyen el precio relativo de

lanzamiento a nivel de molécula, el logaritmo del tamaño del país (población), el

logaritmo del gasto público en salud per cápita, el logaritmo del gasto farmacéutico per

cápita, la variable dicotómica que señala la ubicación de la sede de la empresa en el país

de lanzamiento, la variable dicotómica pertenecer a la EMA y los efectos fijos de la clase

terapéutica a la que pertenece el medicamento a nivel ATC-1 (Categoría Anatómica

Terapéutica-1).

108 Ya que rechazamos la hipótesis nula del lg-rank test, el supuesto de proporcionalidad no se cumple, por lo tanto no deberíamos incluir variables cambiantes en el tiempo ni usar el modelo de regression de Cox. El modelo de Cox extendido permite incorporar variables que cambian en el tiempo pero no deberíamos usarlo ya que no permite incorporar censura. 109 Para las covariables que atienden a características del país como población, PIB per cápita, gasto en salud y gasto farmacéutico per cápita, usamos el dato del primer año o año base (2004). Estas variables se incluyen en el modelo para controlar por las diferencias entre tamaño, riqueza y capacidad de gasto de los países, bien representadas por los datos del año 2004.

110 El parámetro p de la forma funcional determina si la probabilidad crece, decrece o permanece constante a lo largo del tiempo.

111 Contrastando la hipótesis de usando el test de razón de verosimilitudes, se puede determinar si se debe controlar por la heterogeneidad no observada.

ij eXij


En lo que respecta a la ecuación del precio relativo de lanzamiento, se aplican

MCO con errores estándar agrupados por molécula-presentación para estimar el

logaritmo del precio relativo de lanzamiento de la molécula i, producto k en el país j en el

momento t, condicionado al lanzamiento. Para tener en cuenta las características no

observadas de la molécula también calculamos los resultados de las estimación por MCG

con estimador de efectos aleatorios. Para incluir el posible sesgo de selección producido

por la correlación entre la propensión al lanzamiento y el precio también se estima un

modelo de selección de Heckman con un probit en la primera etapa (Heckman, 1979)112.

El precio relativo de lanzamiento se define como el ratio de precios entre el precio de

lanzamiento de la molécula i, producto k en el país j en el momento t, y el precio de

lanzamiento de la molécula i, producto k en el país g en el momento del primer

lanzamiento global 0. De esta manera tenemos la siguiente ecuación probit de selección:

P(Lijkt 1) 0 ijkt Xijkt Uijkt

donde el subíndice i=molecule, j=país, k=product y t=año. Lijkt es una variable

dicotómica igual a 1 si la molecula i ha sido lanzada en el país j en el año t. Xijkt es el

vector de variables explicativas y Uijkt .

Las variables explicativas de este modelo son: el retraso en el lanzamiento de la

molécula i en el país j, el logaritmo del PIB per cápita del país j en el año t, el tamaño del

país (población) del país j en el momento t, los gastos de salud pública por habitante del

país j en el año t, el gasto farmacéutico per cápita del país j en el año t, la variable

dicotómica que determina el uso de la política de ERP en el país j, la ubicación de la

sede de la empresa en el país j de la molécula i, la pertenencia a la EMA del país j en el

momento t y los efectos fijos del ATC de la molécula i.

112 De acuerdo con Heckman HECKMAN, J. J. 1979. Sample selection bias as a specification error. Econometrica: Journal of the econometric society, 153-161., la IMR de la molécula i en el país j en el momento t, IMRijt, se calcula usando el score

del lanzamiento de la regresión probit como donde es la función de densidad de la

normal estándar y es la función de distribución de la normal estándar.

X ' IMRijt (X ')

(X ')()

()

(4.31)


RPijkt Pijkt

Pizg0

0 'Zijkt Vijkt

donde el subíndice i=molécula, k=producto, j=país, g=país del primer lanzamiento a nivel

mundial y t = año. es el precio de lanzamiento relativa, es el precio de

lanzamiento en el país j en el momento t, y es el precio de lanzamiento en el país g

en el momento del primer lanzamiento global 0, es el vector de variables explicativas

y Vijkt es el término aleatorio de error.

Las variables explicativas de este modelo son las mismas que en la ecuación

probit de selección además del cuadrado del retraso en el lanzamiento, la tendencia

durante el periodo observado y la IMR de la ecuación probit de selección.

El modelo estudia el efecto del precio de lanzamiento sobre retraso de

lanzamiento y viceversa. Además, pretende observar el efecto de la aplicación de la

política de ERP tanto sobre el precio y retraso del lanzamiento. Se espera que los países

que pagan precios altos experimenten menores retrasos en el lanzamiento y a su vez,

que los países que sufren mayores retrasos en el lanzamiento paguen menores precios

relativos de lanzamiento. También se espera que el uso del ERP sea eficaz, por lo que

los países que apliquen esta política de precios sufran mayores retrasos en el

lanzamiento y paguen menores precios de lanzamiento.

Para mostrar lo comentado en el párrafo anterior hemos controlado por las

características del país en cuanto a la capacidad de negociación del mismo, como su

tamaño y poder adquisitivo (PIB). Se espera que los países con gran volumen de compra

experimenten menores retrasos en el lanzamiento y paguen precios más bajos para los

nuevos medicamentos, mientras que los países con mayor poder adquisitivo, se espera

que también tengan un acceso al mercado más rápido, pero a precios más altos. Otras

características de los países que hemos tomado en cuenta son el gasto en salud y el

gasto farmacéutico público per cápita. Se espera que el aumento de estas variables

pueda acortar el retraso en el lanzamiento e implica pagar precios más altos. Esta idea

RPijkt Pijkt

Pijko

Zijkt

(4.32)


se basa en que dado que los países con un elevado gasto en salud pública deberían

estar más preocupados por la disponibilidad de los medicamentos en su mercado, se

espera que paguen precios más altos y traten de tener el nuevo fármaco disponible lo

antes posible. A su vez, los países con un elevado gasto farmacéutico per cápita se

presume que estarían dispuestos a pagar precios más altos y, por lo tanto, las empresas

estarían interesadas en lanzar los medicamentos con prioridad en estos países.

Además, hemos controlado por la ubicación de la sede de la empresa. Se espera

que los países que acogen la sede de la empresa experimenten menores retrasos y

paguen precios más altos. Creemos que la localización de la sede de la empresa en el

país de lanzamiento podría generar otro tipo de beneficios para el país que motiven que

el país tenga disponible el medicamento en su mercado tan pronto como sea posible y a

un precio relativo de lanzamiento superior (por ejemplo, el empleo, los ingresos

procedentes de los impuestos, etc.). Además, dado que la base de datos contiene países

pertenecientes a la EMA y países ajenos, y sabiendo que los procesos de aprobación de

medicamentos no tardan los mismo dependiendo del órgano correspondiente (la Food

and Drug Administration (FDA), la EMEA, etc.), se espera que la variable dicotómica

EMA afecte al retraso en el lanzamiento del nuevo fármaco.

También hemos controlado por los efectos fijos de la clase terapéutica a la que

pertenece el medicamento a nivel ATC-1 y por la tendencia temporal, sin embargo, los

efectos fijos de país no se han incluido porque la variable de mayor interés, que mide el

uso de los controles sobre precios, tiene poca variación dentro del país.

Hemos utilizado la base de datos de IMS Health descrita previamente al comienzo

de la sección empírica de esta Tesis. El precio por SU para cada producto k se calculó

sobre una base anual como el ratio de los ingresos totales y las SU vendidas. Dos

productos k son considerados idénticos si presentan la misma cantidad de SU y la misma

vía de administración. Para cada combinación molécula-país que proporciona dos o más

productos idénticos pero con diferentes packsizes113, se calcula el precio promedio

ponderado por volumen.

113 Ej: Pastillas 150MG 28 and Pastillas 150MG 56.


Resultados

Revisión de la literatura

Los resultados obtenidos de la revision de modelos teóricos explican las

estrategias utilizadas por los gobiernos, los aseguradores públicos y las compan´ías

farmacéuticas. Aunque los modelos propuestos difieren en las hipótesis asumidas y no

reflejan la realidad del lanzamiento de medicamentos y de la fijación de precios en todo

su conjunto, aportan información sobre las principales cuestiones que la literature ha

analizado. Distinguimos dos tipos de factores: primero, los factores que afectan

directamente al precio y al lanzamiento como las políticas de regulación de precios, la

existencia de PT y las características de las compañías farmacéuticas. Por otro lado,

otros factores que no solamente afectan directa sino también indirectamente al

lanzamiento y al precio influyendo en el grado de influencia de los primeros, como el

tamaño del país y el nivel de los copagos farmacéuticos.

La literatura destaca la regulación directa de los precios como el uso del ERP, el

control indirecto como el RP interno y el MES+PC (Estándar Mínimo de Eficacia+Control

de Precios) que incluye ambos tipos, directo e indirecto. El resultado de estas tres

políticas es muy sensible a su diseño. El ERP afecta al precio de los países que son

tomados de referencia mientras que el nivel elegido en el diseño del RP afecta al alza los

precios fijados por la industria y al lanzamiento de nuevos fármacos. El esquema del

MES+PC puede afectar no solamente al precio, sino también a la inversión en I+D y a la

calidad del fármaco.

La presencia de PT hace converger los precios entre los países importadores y

exportadores. Las pérdidas procedentes del PT deben ser consideradas por las

empresas como una pérdida de ingresos a la hora de hacer su balance. Se propone que

las empresas exportadoras incrementen sus precios en los países importadores.

Las empresas como agentes económicos tienen una importante influencia en el

lanzamiento y en los precios de los medicamentos. Las empresas que tienen su sede en

el país de lanzamiento parecen fijar precios en ese país ligeramente superiores al resto

de países. Además, un incremento en el contacto entre empresas que venden en

diferentes países induce a altos niveles de colusión y por tanto a precios superiores,


aunque este efecto colusivo parece ser menor en aquellos países donde los precios

están altamente regulados.

El tamaño del país y el nivel de copago afectan de manera indirecta a algunos de

los factores anteriormente descritos. La preferencia de aplicar ERP se reduce cuando

nivel de copago entre el país de referencia y el país que referencia converge. Cuando el

MES+PC mejora la calidad de un medicamento de baja calidad e incrementa su precio,

pero reduce la calidad del fármaco de alta calidad y reduce su precio, el resultado final

dependerá del tamaño de los grupos de compradores de calidad alta y de calidad baja. El

tamaño del país importador influye positivamente en el número de empresas

importadoras y por ende en el precio del medicamento en el país importador. Asimismo,

los precios en un país entre una empresa local y una multinacional extranjera convergen

cuando el tamaño del país de la multinacional extranjera aumenta. Además, ante la

posibilidad de filtración de información entre países sobre la calidad de un producto,

aunque la empres podría evitar esta situación lanzando simultáneamente el fármaco, la

preferencia de la empresa sobre dónde retrasar el lanzamiento depende positivamente

del tamaño relativo de los países y de los niveles de copago.

La literatura empírica recoge modelos econométricos que han identificado y

medido la influencia de los factores que afectan al lanzamiento y al precio de manera

significativa. El uso de diferentes muestras debe hacer cauta la comparación.

Características demográficas y económicas de los países, y los niveles de

regulación del mercado farmacéutico, son los factores más importantes. No obstante, la

regulación de los precios podría debilitar los efectos de estos factores importantes. Con

el apoyo de estudios anteriores, la pertenencia a la EMA y la innovación terapéutica

afecta fuertemente el retraso en el lanzamiento y el precio de lanzamiento

respectivamente. Las características del medicamento como la cantidad de principio

activo, el tamaño y la forma de presentación, se han mostrado significativamente

relacionados con el precio en la literatura anterior. El valor terapéutico aparece en la

literatura anterior como un factor influyente y robusto en el precio y lanzamiento de

nuevos medicamentos. La localización de la empresa es importante a la hora de lanzar

un medicamento pero su efecto sobre los precios parece ambiguo.

El tamaño del país, el PIB per capita y la distribución de la renta son tres factores

que afectan de manera robusta al lanzamiento y al precio. Los países con rentas más


altas pagan precios más altos y tienen un acceso más rápido a los medicamentos;

además, los países más poblados también disfrutan de una mayor probabilidad de

lanzamiento. Asimismo, una distribución de la renta menos equitativa afecta

positivamente a la probabilidad de lanzamiento en países de renta baja a través de una

élite poderosa. Una distribución de la renta más equitativa tiene el mismo efecto en

países de renta alta a través de una clase media pudiente.

El resultado más común de esta revisión demuestra que la regulación de precios

tiende a producir retrasos en el lanzamiento. Esto ha sido observado en países de

regulación estricta y en países tradicionalmente exportadores. Algunos ejemplos estudian

el uso del ERP, de la evaluación económica como cuarta barrera o la congelación de

precios.

La calidad terapéutica, la concentración del principio activo y el número de

presentaciones afectan positivamente a los precios. El pack size y el ciclo de vida del

producto lo hace negativamente.

Modelo teórico

La Figura 4.6 presenta la relación entre el coste unitario de aplicar el CEA y la

diferencia de precios entre el precio internacional de referencia y el precio esperado. De

acuerdo a esta relación, el país i (i = C,D) podrá elegir su política de precios más

conveniente, bien el ERP o el CEA.


(4.33)

Figura 4.6 Trade-off entre CEA y ERP

Si el país i no sufre retraso en el lanzamiento bajo ninguna de las política de precios, o

bien, bajos las dos políticas, el país i estará mejor aplicando ERP cuando el coste unitario

del CEA sea mayor que la diferencia entre el precio internacional de referencia y el precio

esperado del país i bajo el CEA. Cuanto menor sea el tamaño de la población, más

atractivo será el ERP ya que el coste unitario del CEA aumenta.

a

qi

pIh E[pi ]CEA


(4.34)

(4.35)

Si el país i sufre retraso en el lanzamiento cuando aplica ERP, pero no si aplica CEA, el

coste del retraso (r) hará más atractiva el CEA,

a

qi

pIh E[pi ]CEA r

De manera análoga, si el país i sufre retraso en el lanzamiento cuando aplica CEA, pero

no si aplica ERP, el coste del retraso (r) hará más atractiva el ERP,

a

qi

pIh E[pi ]CEA r

Intuitivamente, desde que el coste unitario del CEA decrece, para que ERP siga

siendo conveniente, la diferencia de precios debería ser menor, bien porque el precio

internacional de referencia baja o el precio esperado del país i cuando aplica CEA

aumenta. De otro modo, si el coste unitario de aplicar CEA crece, para que CEA siga

siendo ventajoso, la diferencia de precios debería aumentar, bien porque el precio

internacional de referencia crece o porque el precio esperado de el país i cuando aplica

CEA decrece.

Sin embargo, cuando comparamos ambas políticas de precio y solamente el país i

que aplica CEA sufre retraso en el lanzamiento, incluso siendo la diferencia de precios

más alta que el coste unitario de aplicar CEA, el coste unitario del retraso del CEA podría

compensar esta diferencia y hacer más ventajoso el ERP. De manera análoga, si

solamente el país i que aplica ERP sufre retraso en el lanzamiento, incluso siendo que el

coste unitario de aplicar CEA superior a la diferencia de precios, el coste unitario del

retraso del ERP podría compensar esta diferencia y hacer más atractivo el CEA.

Cuando la diferencia de precios es negativa, es decir, el precio esperado del país i

aplicando CEA es superior que el precio internacional de referencia, y el país que aplica


el CEA sufre retraso en el lanzamiento, el ERP será siempre preferido por el país i.

El precio esperado será superior cuando el país no se crea el número de

AVACs declarado de la empresa que cuando se lo crea. Esto implica que ERP será más

atractivo para los países cuando el país no se crea el número de AVACs.

Modelo empírico

Cuando replicamos el modelo de D&E, el número de países de precio alto de

dentro y fuera de la UE donde el medicamento se ha lanzado anteriormente, y los precios

de los competidores, son los factores más robustos que afectan al lanzamiento de

medicamentos. Las empresas retrasan el lanzamiento en países de precio bajo hasta

que el medicamento está disponible en los países de precio alto. Nuestro modelo

encuentra resultados similares. Encontramos incluso que las empresa retrasan el

lanzamiento en países de precio alto fuera de la UE hasta lanzarlos en países de precio

alto de la UE. D&E encuentran que dependiendo de los diferentes tipos de empresas:

empresa local que produce medicamentos originales, empresa local que solamente

vende con licencia y empresa comercializadora junto con otra, afectan de manera

significativa a la hora del lanzamiento. La primera tiene mayor probabilidad de

lanzamiento que las otras dos. Nuestro modelo no encuentra efectos significativos de

estas características. D&E no presentan los efectos de las dummies de país aunque

reporta que son significativas. Nuestro modelo muestra que Alemania disfruta de

menores retrasos en el lanzamiento de medicamentos en media que el resto de países.

En ambos modelos, el de D&E y el nuestro, el precio mínimo previamente fijado

en otros países presenta efectos diferentes sobre el precio de lanzamiento. Los países

que fijan su precio de lanzamiento sin ninguna referencia de la UE pagan precio más

altos. La transmisión de información sobre los precios entre países para fijar sus propios

precios ocurre entre países dentro de la UE, fuera de la UE y entre países de la UE y

fuera de ella. Sin embargo, D&E muestran que esta relación solamente se sucede entre

los países de la UE. A diferencia que en la ecuación de lanzamiento, D&E no encuentran

ningún efecto sobre el precio del tipo de empresa que venda el fármaco. Sin embargo,

nuestro modelo encuentra que las empresas que venden solamente con licencia

presentan precios más bajos que el resto de empresas. Ninguno de los modelos no

encuentra ningún efecto de la renta per cápita sobre los precios. La concentración del

E[pi ]CEA


principio activo afecta positivamente al precio en el modelo de D&E pero nuestro modelo

no encuentra efecto significativo de esta variable. En ambos modelos, la vía de

administración presenta un resultado robusto, siendo los medicamentos inyectables más

caros que otro tipo de fármacos como los orales. En ambos modelos también, Alemania

presenta en media precios superiores al resto de países.

Nuestra réplica de Verniers et al. con datos más recientes y una selección

diferente de países arroja nuevos resultados. El resultado más importante es que el

carácter endógeno de precios de lanzamiento y del retraso encontrado por Verniers et al.

no se observa con nuestros datos, sólo el retraso en el lanzamiento se ve afectado

negativamente por el precio de lanzamiento en una dirección. Los efectos de algunas

políticas regulatorias sobre el retraso en el lanzamiento no parecen cumplirse de manera

general. Algunas políticas de regulación que tradicionalmente afectan al retraso en el

lanzamiento, como el control sobre los beneficios o los precios de referencia interna, no

son significativas en el modelo que se replica. Sin embargo, ambos modelos muestran

que el uso de la evaluación económica como cuarta barrera provoca mayores retrasos en

el lanzamiento. Además, en ambos modelos también, mientras más fuerte es la

protección sobre la patente, los países disfrutan de menores retrasos en el lanzamiento.

Sólo el modelo que se replica presenta un efecto significativo y esperado para el uso del

ERP. Por otro lado, las políticas de regulación no muestran efectos significativos en el

precio de lanzamiento bajo ninguno de los modelos, excepto la protección de la patente;

en este caso una mayor protección de patentes supone una presión a la baja sobre los

precios de lanzamiento. Una vez más, igual que con el modelo de D&E, la ubicación de la

empresa ya no tiene efecto sobre el lanzamiento y la fijación de precios de nuevos

medicamentos.

En nuestro modelo, el precio no presenta efecto significativo sobre el lanzamiento.

El uso del ERP afecta negativamente a la probabilidad de lanzamiento. Sin embargo, y a

diferencia de estudios anteriores, los países de mayor tamaño (mayor población) parece

sufrir mayores retrasos. Este resultado no esperado muestra que el poder de negociación

debido al tamaño del país no tiene efecto sobre un acceso más rápido a los

medicamentos. Los factores que tienen un importante efecto positivo en la probabilidad

de lanzamiento son la renta per cápita y la pertenencia a la EMA. Sin embargo, otros

indicadores como el gasto público en salud y en el gasto farmacéutico. Tampoco el

hecho de que la empresa esté localizada en el país de lanzamiento parece no estar


relacionado con un lanzamiento precoz.

Bajo nuestro modelo, el retraso en el lanzamiento no afecta significativamente al

precio relativo de lanzamiento. Este resultado podría indicar que las empresas priorizan

evitar los efectos del ERP y del PT frente a poder generar ingresos de esas ventas más

tardías a precios más bajos. De hecho, los países ya no se benefician de pagar precios

inferiores a cambio de acceder más tarde a los medicamentos. Además, observamos que

el ERP no afecta al precio de lanzamiento. Los países que aplican ERP no pagan precios

relativos más bajos que aquellos países que no aplican esta política de precios. Este

resultado no esperado muestra que el uso de ERP no es efectivo, bien porque algunos

países podrían tomar el precio de referencia directamente, porque otros países podrían

no aplicar ERP en último término como criterio exclusivo, o porque las empresas no

están dispuestas a vender a los precios que resultan de aplicar ERP. Además, otras

características como el gasto público en salud y el gasto farmacéutico per cápita afectan

significativamente el precio relativo de lanzamiento. De hecho, los países con un alto

gasto público en salud y un gasto farmacéutico pagan precios relativos superiores.

También, los países con mayor población, que disfrutan de un mayor poder de

negociación, pagan precios más bajos en medio. En general, los resultados para el

mercado ambulatorio y el mercado hospitalario no muestran grandes diferencias.

Resumen en castellano 225

Tabla 4.4. Ecuación del precio relativo de lanzamiento

Ambulatorio Hospitalario

MCO - ESs Robustos Agrupados Efectos aleatorios MCO ESs Robustos Agrupados Efectos aleatoriosRetraso -0.0722 0.0053 -0.0726 0.0137 [0.1296] [0.1199] [0.1695] [0.1490] Retraso*Retraso 0.0023 0.0008 0.0000 1.41e-06 [0.0021] [0.0008] [0.0000] [6.87e-06] Log del Tamaño del País (población) -8.6599* -5.8246*** -10.1434** -4.2420*

[4.7556] [1.9964] [4.4531] [2.4337] Log del PIB per cápita -9.5648 -2.6522 7.7304 1.7151

[9.7722] [16.5960] [12.6634] [19.4184] Log del Gasto en Salud per cápita 24.6213* 14.3926*** 18.0517* 4.1666

[13.9677] [12.7490] [9.8605] [14.9830] Log del Gasto Farmacéutico per cápita 50.7958* 37.3478 70.6921** 29.2789

[27.4052] [13.0590] [31.7399] [18.2923] ERP -5.8320 -2.7864 -8.9012 -4.8740 [4.1272] [3.3182] [6.0281] [4.3355] País sede de la empresa -4.1632 -4.1037 -19.7583* -5.1026

[3.5854] [5.8000] [11.5931] [8.3115] Año 2004 RC RC RC RC RC RC RC RC 2005 0.6371 -2.8228 3.6965 -1.4028 [2.3904] [7.3313] [3.3774] [9.1615] 2006 -0.9426 -4.1618 4.4947 1.0742 [2.8771] [8.1672] [4.3652] [10.8644] 2007 -3.6626 -6.0497 -3.1937 -0.3632 [5.1028] [8.4271] [5.5596] [11.4744] 2008 -4.4736 -8.0594 -1.3471 1.0891 [5.0712] [8.8661] [5.6821] [12.3694]


Niveldesignificación(amboslados):*:p<0.10;**:p<0.05;***:p<0.01.;[]:ErrorEstándar;n.r.:nomostrado;‐:noincluido;RC:categoríadereferencia

Ambulatorio Hospitalario MCO - ESs Robustos Agrupados Efectos aleatorios MCO ESs Robustos Agrupados Efectos aleatorios 2009 -2.1013 -8.1843 14.1190 1.3525 [3.6430] [9.2742] [11.2567] [13.2468] 2010 -16.3579 -16.5441 -7.6186 -4.3132 [10.4757] [10.2073] [9.7540] [14.6237] IMR 60.7956* 45.6041*** 80.9075** 25.3029 [33.2356] [16.1445] [36.1181] [23.8379] Constant -19.4495 -303.1349** -648.5746* -210.4026 [87.4215] [140.4957] [329.3245] [207.2856] Observaciones 1334 1334 1369 1369 Número de Molécula-presentación-grupo 69 69 70 70

R-cuadrado 0.1776 0.1744 0.1952 0.1952


Tabla 4.5. Ecuación de retraso en el lanzamiento

Ambulatorio Hospitalario Variables Razón de riesgo Razón de riesgo

Precio relativo (a nivel de molécula) 0.9934*** 0.9949***

[0.0017] [0.0014] Log del tamaño del país en 2004 (población) 0.8754** 0.8746**

[0.0492] [0.0467] Log del PIB per cápita en 2004 4.2414** 3.2867**

[2.9975] [2.1143] Log del gasto en salud per cápita en 2004 0.4584 0.6324

[0.2581] [0.3465] Log del gasto farmacéutico per cápita en 2004 0.8243 0.5657

[0.3031] [0.2016] ERP 0.3838*** 0.3712*** [0.0714] [0.0659] País sede de la empresa 1.0816 1.2849

[0.3358] [0.3963] EMA2004 1.6443*** 15136*** [0.2479] [0.1955] ATC-1 B 2.5813 2.9381 [0.6169] [0.7006] C 0.5827 0.6983 [0.1693] [0.2024] D 2.6959 2.9717 [0.8045] [0.9475] G 0.3353 0.3347 [0.1011] [0.1037] J 0.7837 0.6636 [0.1686] [0.1343] L 0.6780 0.5443 [0.1329] [0.1032] M 0.4678 0.2891 [0.1683] [0.0972] N 1.4022 1.2241 [0.2861] [0.2531] R 13.2524 13.9722 [7.1829] [7.7269] S 0.6492 0.9355 [0.2343] [0.3175] T 0.6209 0.8516 [0.3703] [0.3427] V 0.8353 0.7506 [0.2777] [0.2233] Observaciones 732 921 AIC 38.1443 37.9427 Niveldesignificación(amboslados):*:p<0.10;**:p<0.05;***:p<0.01.;[]:ErrorEstándar;RC:categoríadereferencia


Conclusiones

Existen similitudes y disparidades de nuestros resultados empíricos en relación

con la literatura empírica anterior. La revisión sistemática muestra que las características

demográficas e ingresos de los países, y los regímenes de regulación, parecen ser los

factores más determinantes que afectan a los precios y al lanzamiento de los nuevos

medicamentos. No obstante, la regulación de los precios podría debilitar los efectos de

estos factores importantes. Con el apoyo de estudios anteriores, la pertenencia a la EMA

y la innovación terapéutica afecta fuertemente el retraso en el lanzamiento y el precio de

lanzamiento respectivamente. Las características del medicamento como la cantidad de

principio activo, el tamaño y la forma de presentación, se han mostrado

significativamente relacionados con el precio en la literatura anterior. Estas últimas

características del medicamento no se han incluido en el NPLM ya que hemos analizado

los datos a nivel de presentación. El valor terapéutico, que aparece en la literatura

anterior como un factor influyente y robusto en el precio y lanzamiento de nuevos

medicamentos, tampoco fue incluido en el NPLM debido a la disponibilidad de datos. A

diferencia de estudios anteriores, no hemos ecnontrado influencia de la localización de la

empresa.

Al replicar el modelo de D&E con datos más recientes, se encuentran algunos

nuevos patrones acerca de los spillover effects y el tipo de empresas que venden el

nuevo medicamento. Los países con precios bajos sufren mayores retrasos en el

lanzamiento y que también existen spillover effects entre los países que tradicionalmente

tienen precios bajos. Además, se observa que los spillover effects se producen entre

países pertenecientes a la UE y países fuera de la misma. Por otra parte, parece que las

empresas locales no gozan ya de una mayor probabilidad de lanzamiento en su país,

pero sin embargo las empresas que disfrutan de la licencia de venta exclusiva de un

nuevo medicamento obtienen precios de lanzamiento más bajos.

Nuestra réplica de Verniers et al. con datos más recientes y una selección

diferente de países arroja nuevos resultados. El resultado más importante es que el

carácter endógeno de precios de lanzamiento y del retraso encontrado por Verniers et al.

no se observa con nuestros datos, sólo el retraso en el lanzamiento se ve afectado

negativamente por el precio de lanzamiento en una dirección. Los efectos de algunas

políticas regulatorias sobre el retraso en el lanzamiento no parecen cumplirse de manera


general. Algunas políticas de regulación que tradicionalmente afectan al retraso en el

lanzamiento, como el control sobre los beneficios o los precios de referencia interna, no

son significativas en el modelo que se replica. Sin embargo, ambos modelos muestran

que el uso de la evaluación económica como cuarta barrera provoca mayores retrasos en

el lanzamiento. Además, en ambos modelos también, mientras más fuerte es la

protección sobre la patente, los países disfrutan de menores retrasos en el lanzamiento.

Sólo el modelo que se replica presenta un efecto significativo y esperado para el uso del

ERP. Por otro lado, las políticas de regulación no muestran efectos significativos en el

precio de lanzamiento bajo ninguno de los modelos, excepto la protección de la patente;

en este caso una mayor protección de patentes supone una presión a la baja sobre los

precios de lanzamiento. Una vez más, igual que con el modelo de D&E, la ubicación de la

empresa ya no tiene efecto sobre el lanzamiento y la fijación de precios de nuevos

medicamentos.

Una de las conclusiones más importantes es que, a diferencia de los modelos

anteriores, la fijación de precios y el lanzamiento parecen no estar relacionados entre sí.

Existen diferencias de precios entre países, pero no debido al retraso en el lanzamiento.

Las empresas no aceptan precios más bajos a cambio de lanzamientos tardíos, incluso

de los países que aplican la política de ERP, por lo tanto, la política de precios basada en

el ERP parece no ser efectiva en términos de “precios", pero lo hace en términos de

“lanzamiento". Estos resultados pueden conllevar implicaciones en el proceso de

negociación. Sugerimos que las empresas, básicamente, retrasan el lanzamiento porque

los países probablemente no pueden permitirse el lujo de tener el producto disponible

inmediatamente después del primer lanzamiento a nivel mundial, y en última instancia no

acceder al producto. Mientras que las empresas solían retrasar el lanzamiento para evitar

determinados efectos secundarios, en nuestro estudio mostramos que las empresas son

capaces de llevar a cabo una estrategia más agresiva que no permita a los países pagar

precios más bajos a cambio de experimentar mayores retrasos en el lanzamiento. Bajo

dicha estrategia, las empresas evitarían los efectos derivados de la política de ERP y del

PT, pero sin embargo perderían los beneficios de las ventas asociadas a los países

donde el medicamento se lanza.

En cuanto a otras características del país, tener una gran demanda potencial

ayuda a tener un acceso más rápido al mercado y obtener precios más bajos. Los países

ricos tienen los productos disponibles en un corto plazo, pero los países que en última


instancia pagan altos precios de lanzamiento son los que asignan grandes presupuestos

para la salud y gasto farmacéutico. La ubicación de la empresa parece no afectar al

precio ni retraso en el lanzamiento. Los países pertenecientes a la EMA parecen disfrutar

de menores retrasos que los países fuera de ella, sin embargo no encontramos

diferencias de precios significativas entre los países EMA y el resto.

En general, los resultados para el mercado ambulatorio y el mercado hospitalario

no muestran grandes diferencias.

Dada la perspectiva global sobre los principales factores que influyen en los

precios y el lanzamiento de nuevos fármacos basados en estudios teóricos, se distinguen

principalmente dos tipos de factores. En primer lugar, aquellos que inciden directamente

en los precios y el lanzamiento, tales como la presencia de PT, las características de las

empresas y las políticas de regulación de precios, por ejemplo, el ERP, el RP interno y el

MES+PC. En segundo lugar, la revisión arroja otros determinantes que no sólo tienen un

impacto directo sino también indirecto, afectando a la medida en la cual los anteriores

factores influyen en el precio y lanzamiento de los nuevos fármacos, tales como el

tamaño del país y el nivel de copagos.

De acuerdo con nuestro modelo teórico, dada una secuencia óptima de

lanzamiento, llegamos a la conclusión de que a medida que el país es más pequeño y

por lo tanto su demanda potencial, el coste unitario del CEA crece y el uso del ERP es

más atractivo. Debemos tener en cuenta que el ERP no requiere ningún tipo de inversión,

sin embargo, el CEA sí. El uso del ERP es útil para los países relativamente pequeños en

comparación con el uso de CEA. Este resultado confirma algunas declaraciones sobre

este tema en la literatura anterior que no se han demostrado aún. El ERP es una política

de precios de bajo coste.

Se concluye también que la secuencia óptima de lanzamiento de la empresa

depende de los precios relativos y el tamaño relativo de los países. Hay un trade-off entre

precio y volumen de venta que afecta directamente a la secuencia de lanzamiento. A su

vez, el precio relativo depende de la política de precios (ERP o CEA) y, posteriormente,

de la fórmula y composición del ERP (precio mínimo o precio medio). Concretamente, un

país se encuentra en una mejor situación aplicando el ERP en lugar del CEA, si y sólo si

la diferencia entre el precio de referencia internacional y el precio esperado del país i en

el marco del CEA no es mayor que el coste unitario del CEA. Este resultado se ve


afectado por el coste del retraso en el lanzamiento. Si uno de los países sufre retraso en

el lanzamiento, el coste del retraso afectaría negativamente al beneficio de aplicar

cualquiera de las dos políticas en el país en cuestión.

Desde la perspectiva del regulador, la literatura anterior recomienda aplicar el

ERP sólo a los países pequeños y/o basado en los precios de países grandes (o un gran

número de países); lo mismo aplica si uno sustituye "país grande" por "pequeño co-

pago" y viceversa. Además, se recomienda un nivel mínimo de RP para evitar un nivel de

precios cada vez mayor con respecto al promedio del RP. Por otra parte, conviene aplicar

una política mixta de MES+PC cuando el bienestar perdido por los compradores de

medicamentos de tipo alto114 no es lo suficientemente grande como para anular el

bienestar ganado por los compradores de tipo bajo115.

Desde la perspectiva de la empresa, dado que el ERP se utiliza frecuentemente

por los países donde se practica el PT, los mercados son inseparables. Un claro ejemplo

es el caso del PT. La literatura sugiere que la pérdida de ingresos procedentes del PT

debería ser contabilizada por la empresa que debería establecer por tanto un precio más

alto que en ausencia de importaciones paralelas. Por lo tanto, el precio más alto para un

medicamento no es siempre la mejor opción para la empresa, ni el precio más bajo es

siempre la peor opción para un país determinado. Lo que podría ser una estrategia de

precios óptima en un solo país podría dejar de serlo cuando ERP y PT entran en juego.

Nuestro modelo teórico muestra que existe un trade-off entre el precio y el

volumen que afecta a la secuencia de lanzamiento por parte de la empresa. Se

proporciona información acerca de un tema ampliamente señalado por la literatura

previa. El uso del ERP parece ser beneficioso para los países pequeños en comparación

con el uso de CEA. Además de ser una política de precios de bajo coste tenemos ahora

más información acerca de por qué los países aplican este tipo de política de precios

siendo totalmente conscientes de que podría no ser una política óptima. En cualquier

caso, somos conscientes de que pueda haber otros factores que afecten a la

negociación. De esta manera se podrían diseñar otros modelos teóricos que contengan

114 Compradores de medicamentos de alta eficacia (ver ATELLA, V., BHATTACHARYA, J. & CARBONARI, L. 2012. Pharmaceutical Price Controls and Minimum Efficacy Regulation: Evidence from the United States and Italy. Health Services Research, 47, 293-308.).

115 Compradores de medicamentos de baja eficacia (ver ibid.).


por ejemplo fórmulas de ERP diferentes a la media o el mínimo, o modelos donde las

empresas ofreciesen un precio único a todos los países, etc. Además, otros factores

como la estructura de edad de la población o el lobby de la industria farmacéutica

también podrían tenerse en cuenta. No obstante, la inclusión de estas características

simultáneamente complicaría excesivamente el modelo.

A pesar de las limitaciones originadas por la falta de datos, este es el primer

estudio que indica que la fijación de precios y el lanzamiento de un medicamento no son

dos cuestiones inseparables. No obstante, consideramos que sería conveniente realizar

otros análisis que cuenten con muestras más grandes en número de medicamentos y

países con el fin de lograr una comparación más precisa. Concretamente, en el análisis

del ERP, una futura investigación podría tener en cuenta las interdependencias que se

producen entre los países debidas al ERP. Se ha analizado el efecto de esta política de

precios a través de una variable dicotómica, pero se podrían obtener puntos de vista más

interesantes mediante la recopilación de datos sobre el conjunto de países de referencia

que cada país considera.

Detrás de la idea de que la fijación de precios y el lanzamiento de nuevos

fármacos ya no están relacionados entre sí, los precios podrían estar convergiendo a un

único precio, al menos, en los países de la EMA. Una futura línea de investigación podría

ser analizar si los precios de lanzamiento de nuevos medicamentos han convergido

durante los últimos años a través del análisis de series temporales, en particular, basado

en la teoría de cointegración. Este trabajo requiere de datos de precios y lanzamientos

que abarquen periodos largos de tiempo. Este nuevo estudio podría mostrar algunas de

las hipótesis planteadas en esta tesis acerca de las líneas de estrategia de la industria

farmacéutica.

Si atendemos al mercado hospitalario, no podemos olvidar de los descuentos

sobre precios que se producen en este ámbito, debidos a los grandes volumenes de

compras. Una futura investigación podría dedicarse a contrastar la eficacia de las

compras centralizadas. Hay países como Dinamarca o Noruega que aplican desde hace

tiempo este tipo de compra a nivel regional y nacional. Concretamente, en España, la

articulación de estas adquisiciones se efectúa a través de un nuevo mecanismo previsto

en la Ley de Contratos del Sector Público, mediante la adopción de un acuerdo marco

centralizado de adquisiciones de medicamentos y productos sanitarios. Este nuevo


mecanismo da un paso más en la racionalización y cohesión en el Sistema Nacional de

Salud, y supone el cumplimiento de las principales medidas de la reforma sanitaria para

reducir el gasto farmacéutico, entre otros, y garantizar la sostenibilidad del Sistema

Nacional de Salud. En este caso, se podría contrastar si esta nueva política de compra

de medicamentos ha sido efectiva con un diseño quasi-experimental. Una vez más, la

recopilación de los datos directamente de los hospitales parece ser el problema más

difícil.

References

ABRAHAM, J. 2002. The pharmaceutical industry as a political player. Lancet, 360, 1498-1502.

ATELLA, V., BHATTACHARYA, J. & CARBONARI, L. 2012. Pharmaceutical Price Controls and Minimum Efficacy Regulation: Evidence from the United States and Italy. Health Services Research, 47, 293-308.

BARDEY, D., BOMMIER, A. & JULLIEN, B. 2010. Retail price regulation and innovation: reference pricing in the pharmaceutical industry. J Health Econ, 29, 303-16.

BAYUS, B. L., KANG, W. & AGARWAL, R. 2007. Creating Growth in New Markets: A Simultaneous Model of Firm Entry and Price*. Journal of Product Innovation Management, 24, 139-155.

BERNDT, E. R., COCKBURN, I. M. & GRILICHES, Z. 1996. Pharmaceutical Innovations and Market Dynamics: Tracking Effects on Price Indexes for Antidepressant Drugs. Brookings Papers on Economic Activity: Microeconomics 1996,.

BERNDT, E. R., DANZON, P. M. & KRUSE, G. B. 2007. Dynamic competition in pharmaceuticals: cross-national evidence from new drug diffusion. Managerial and Decision Economics, 28, 231-250.

BORRELL, J. R. 2007. Pricing and patents of HIV/AIDS drugs in developing countries. Applied Economics, 39, 505-518.

CABRALES, A. & JIMENEZ-MARTIN, S. 2007. The Determinants of Pricing in Pharmaceuticals: Are U.S. Prices Really Higher than Those of Canada? SSRN eLibrary.

CLAXTON, K., MARTIN, S., SOARES, M., RICE, N., SPACKMAN, E., HINDE, S., DEVLIN, N., SMITH, P. C. & SCULPHER, M. 2013. Methods for the Estimation of the NICE Cost Effectiveness Threshold. York: Centre for health Economics, University of York.

COCKBURN, I. M. & ANIS, A. H. 1998. Hedonic Analysis of Arthritis Drugs. National Bureau of Economic Research Working Paper Series, No. 6574.

COMANOR, W. S. & SCHWEITZER, S. O. 2007. Determinants of drug prices and expenditures. Managerial and Decision Economics, 28, 357-370.

CORONADO, F. J., JIMÉNEZ-MARTÍN, S. & MARÍN, P. L. 2007. Multimarket Contact in Pharmaceutical Markets. SSRN.

DANZON, P. M. 1997. Price Discrimination for Pharmaceuticals: Welfare Effects in the US and the EU. International Journal of the Economics of Business, 4, 310-322.

DANZON, P. M. & CHAO, L. W. 2000. Cross-national price differences for pharmaceuticals: How large, and why? Journal of Health Economics, 19, 159-195.

DANZON, P. M. & EPSTEIN, A. J. 2008. Effects of Regulation on Drug Launch and Pricing in Interdependent Markets. National Bureau of Economic Research, Inc, NBER Working Papers: 14041.

236 Modelling Global Pricing and Launching of New Drugs

DANZON, P. M., MULCAHY, A. W. & TOWSE, A. K. 2011. Pharmaceutical Pricing in Emerging Markets: Effects of Income, Competition and Procurement. National Bureau of Economic Research Working Paper Series, No. 17174.

DANZON, P. M. & TOWSE, A. 2003. Differential Pricing for Pharmaceuticals: Reconciling Access, R&D and Patents. International Journal of Health Care Finance and Economics, 3, 183-205.

DANZON, P. M., WANG, Y. R. & WANG, L. 2005. The impact of price regulation on the launch delay of new drugs - evidence from twenty-five major markets in the 1990s. Health Economics, 14, 269-292.

DJANKOV, S., LA PORTA, R., LOPEZ-DE-SILANES, F. & SHLEIFER, A. 2002. The Regulation of Entry. Quaterly Journal of Economics, 117.

DONALDSON, C., BAKER, R., MASON, H., JONES-LEE, M., LANCSAR, E., WILDMAN, J., BATEMAN, I., LOOMES, G., ROBINSON, A. & SUGDEN, R. 2011. The social value of a QALY: raising the bar or barring the raise? BMC health services research, 11, 8.

EKELUND, M. & PERSSON, B. 2003. Pharmaceutical pricing in a regulated market. Review of Economics and Statistics, 85, 298-306.

ESPIN J, ROVIRA J & OLRY DE LABRY A 2011. WHO/HAI Project on Medicine Prices and Availability. Review Series on Pharmaceutical Pricing Policies and Interventions. Working Paper 1: External Reference Pricing.

FELDMAN, R. & LOBO, F. 2013. Competition in prescription drug markets: the roles of trademarks, advertising, and generic names. European Journal of Health Economics, 14, 667-675.

GANSLANDT, M. & MASKUS, K. E. 2004. Parallel imports and the pricing of pharmaceutical products: Evidence from the European Union. Journal of Health Economics, 23, 1035-1057.

GARCIA MARIÑOSO, B., JELOVAC, I. & OLIVELLA, P. 2011. External referencing and pharmaceutical price negotiation. Health Economics, 20, 737-756.

GARCÍA-MARIÑOSO, B. & OLIVELLA, P. 2012. Informational spillovers and strategic launch delay of pharmaceutical drugs. Mimeo.

GAREN, J. 1984. The returns to schooling: A selectivity bias approach with a continuous choice variable. Econometrica: Journal of the econometric society, 1199-1218.

GINARTE, J. C. & PARK, W. G. 1997. Determinants of patent rights: A cross-national study. Research Policy, 26, 283-301.

GRABOWSKI, H., VERNON, J. & DIMASI, J. A. 2002. Returns on research and development for 1990s new drug introductions. Pharmacoeconomics, 20, 11-29.

HECKMAN, J. J. 1979. Sample selection bias as a specification error. Econometrica: Journal of the econometric society, 153-161.

HEUER, A., MEJER, M. & NEUHAUS, J. 2007. The National Regulation of Pharmaceutical Markets and the Timing of New Drug Launches in Europe. Kiel advanced studies working papers, No. 437.

HOFSTEDE, G. 1984. Culture's consequences: International differences in work-related values, sage.

HOFSTEDE, G. 2012. Hofstede Centre [Online]. Available: http://geert-hofstede.com/the-hofstede-centre.html.

HOFSTEDE, G. H. 2001. Culture's consequences: Comparing values, behaviors, institutions and organizations across nations, Sage.

HOTELLING, H. 1929. Stability of competition. Economic Journal, 39, 41-57.

References 237

HUNTER, D. 2005. Guaranteed future pain and suffering: The recent research on drug price controls. Heritage Foundation WebMemo.

JELOVAC, I. 2008. On the relationship between the negotiated price of pharmaceuticals and the patients’ co-payment. Center for Economic Development, Hong Kong University of Science and Technology WP 0208.

JELOVAC, I. & HOUY, N. 2013. Drug launch timing and international reference pricing. GATE Lyon St Etienne WP 1301.

JENKINS, S. P. 2008. Survival Analysis

. Unpublished manuscript: Institute for Social and Economic Research, University of Essex, Colchester. Ohio State University.

KANAVOS P & COSTA-FONT J 2005. Pharmaceutical parallel trade in Europe: stakeholder and competition effects. Economic Policy, 20, 751-798.

KANAVOS, P. G. & VANDOROS, S. 2011. Determinants of branded prescription medicine prices in OECD countries. Health Economics Policy and Law, 6, 337-367.

KEELE, L. 2007. Cross validation tests for frailty models. Unpublished manuscript, Ohio State University.

KIEFER, N. M. 1988. Economic duration data and hazard functions. Journal of economic literature, 26, 646-679.

KLEIN, J. P. & MOESCHBERGER, M. L. 1997. Statistics for biology and health. Survival analysis: techniques for censored and truncated data.

KYLE, M. K. 2006. The role of firm characteristics in pharmaceutical product launches. RAND Journal of Economics, 37, 602-618.

KYLE, M. K. 2007. Pharmaceutical price controls and entry strategies. Review of Economics and Statistics, 89, 88-99.

LANCASTER, T. 1992. The econometric analysis of transition data, Cambridge university press.

LANJOUW, J. O. 2005. Patents, Price Controls, and Access to New Drugs: How Policy Affects Global Market Entry. National Bureau of Economic Research Working Paper Series, No. 11321.

LEE, L.-F. 1983. Generalized econometric models with selectivity. Econometrica: Journal of the econometric society, 507-512.

LEOPOLD, C., VOGLER, S., MANTEL-TEEUWISSE, A. K., DE JONCHEERE, K., LEUFKENS, H. G. & LAING, R. 2012. Differences in external price referencing in Europe: a descriptive overview. Health Policy, 104, 50-60.

LICHTENBERG, F. R. 2005. The Impact of New Drug Launches on Longevity: Evidence from Longitudinal, Disease-Level Data from 52 Countries, 1982–2001. International Journal of Health Care Finance and Economics, 5, 47-73.

LOBO, F. 2014. La Intervención de Precios de los Medicamentos en España, Madrid, Springer.

LOBO, F. & FELDMAN, R. 2013. Generic Drug Names and Social Welfare. Journal of Health Politics Policy and Law, 38, 573-597.

LOPEZ-CASASNOVAS, G. & PUIG-JUNOY, J. 2000. Review of the literature on reference pricing. Health Policy, 54, 87-123.

LU, Z. J. & COMANOR, W. S. 1998. Strategic pricing of new pharmaceuticals. Review of Economics and Statistics, 80, 108-118.

MASON, H., JONES�LEE, M. & DONALDSON, C. 2009. Modelling the monetary value of a QALY: a new approach based on UK data. Health Economics, 18, 933-950.

MESTRE-FERRANDIZ, J. 2003b. Reference prices: The Spanish way. Investigaciones Económicas, 27, 125-149.

238 Modelling Global Pricing and Launching of New Drugs

MESTRE-FERRANDIZ, J. 2012. The R&D Cost of a New Medicine London: Office of Health Economics.

MESTRE-FERRANDIZ, J. 2013. Increasing R&D costs for new medicines. Chemistry & Industry, 77, 44-44.

MIRALDO, M. 2009. Reference pricing and firms' pricing strategies. J Health Econ, 28, 176-97.

MOHER, D., LIBERATI, A., TETZLAFF, J., ALTMAN, D. G. & GROUP, A. T. P. 2009. Reprint—Preferred Reporting Items for Systematic Reviews and Meta-Analyses: The PRISMA Statement. Physical Therapy, 89, 873-880.

MORENO, E., GIRON, F. J., VAZQUEZ-POLO, F. J. & NEGRIN, M. A. 2010. Optimal healthcare decisions: Comparing medical treatments on a cost-effectiveness basis. European Journal of Operational Research, 204, 180-187.

MORENO-TORRES, I., PUIG-JUNOY, J. & BORRELL, J. R. 2009. Generic Entry into the Regulated Spanish Pharmaceutical Market. Review of Industrial Organization, 34, 373-388.

MUTHOO, A. 1999. Bargaining Theory with Applications, Cambridge University Press. NATIONAL INSITUTE FOR HEALTH AND CARE EXCELLENCE (NICE). 2010.

Measuring effectiveness and cost effectiveness: the QALY [Online]. NICE. Available: http://www.nice.org.uk/newsroom/features/measuringeffectivenessandcosteffectivenesstheqaly.jsp.

NATIONAL INSTITUTE FOR HEALTH AND CARE EXCELLENCE (NICE). 2010. Measuring effectiveness and cost effectiveness: the QALY [Online]. NICE. Available: http://www.nice.org.uk/newsroom/features/measuringeffectivenessandcosteffectivenesstheqaly.jsp.

NEGRIN, M. A. & VAZQUEZ-POLO, F. J. 2006. Bayesian cost-effectiveness analysis with two measures of effectiveness: the cost-effectiveness acceptability plane. Health Economics, 15, 363-372.

NEGRIN, M. A. & VAZQUEZ-POLO, F. J. 2008. Incorporating model uncertainty in cost-effectiveness analysis: A Bayesian model averaging approach. Journal of Health Economics, 27, 1250-1259.

NEGRIN, M. A., VAZQUEZ-POLO, F. J., MARTEL, M., MORENO, E. & GIRON, F. J. 2010. Bayesian Variable Selection in Cost-Effectiveness Analysis. International Journal of Environmental Research and Public Health, 7, 1577-1596.

NIHR, H. 2010. Weighting and valuing quality-adjusted life-years using stated preference methods: preliminary results from the Social Value of a QALY Project. Health Technology Assessment, 14.

OECD 2011. Health at a Glance 2011: OECD Indicators. Paris. PARIS, V., DOCTEUR, E. & MOÏSE, P. 2008. Pharmaceutical pricing polices in a global

market. OECD Health Policy Studies. Paris. PARK, W. G. & WAGH, S. 2002. Index of patent rights. Economic freedom of the world:

2002 annual report, 33-43. PINTO-PRADES, J. L., LOOMES, G. & BREY, R. 2009. Trying to estimate a monetary

value for the QALY. Journal of Health Economics, 28, 553-562. PUIG-JUNOY, J. 2007. The impact of generic reference pricing interventions in the statin

market. Health Policy, 84, 14-29. PUIG-JUNOY, J. 2010a. Policies encouraging price competition in the generic drug

market: Lessons from the European experience. Gaceta Sanitaria, 24, 193-199.

References 239

PUIG-JUNOY, J. 2010b. Impact of European Pharmaceutical Price Regulation on Generic Price Competition A Review. Pharmacoeconomics, 28, 649-663.

PUIG-JUNOY, J. 2012. Lecturas sobre Economía del Sector Biofarmacéutico, Madrid, Springer Healthcare.

PUIG-JUNOY, J. & GONZÁLEZ LÓPEZ-VALCÁRCEL, B. 2014. Launch prices for new pharmaceuticals in the heavily regulated and subsidized Spanish market. Health Policy, Forthcoming.

RAWLINS, M. D. 2012. Crossing the fourth hurdle. Br J Clin Pharmacol, 73, 855-60. RICHTER, A. 2008. Assessing the impact of global price interdependencies.

Pharmacoeconomics, 26, 649-659. ROBINSON, A., GYRD-HANSEN, D., BACON, P., BAKER, R., PENNINGTON, M. &

DONALDSON, C. 2013. Estimating a WTP-based value of a QALY: The ‘chained’approach. Social Science & Medicine, 92, 92-104.

RUBINSTEIN, A. 1982. Perfect Equilibrium in a Bargaining Model. Econometrica, 50, 97-109.

SCHERER, F. M. 1993. PRICING, PROFITS, AND TECHNOLOGICAL-PROGRESS IN THE PHARMACEUTICAL-INDUSTRY. Journal of Economic Perspectives, 7, 97-115.

SCHERER, F. M. 2004. The pharmaceutical industry-Prices and progress. New England Journal of Medicine, 351, 927-932.

STARGARDT, T. & SCHREYÖGG, J. 2006. Impact of Cross-Reference Pricing on Pharmaceutical Prices: Manufacturers’ Pricing Strategies and Price Regulation. Applied Health Economics and Health Policy, 5, 235-247.

STATACORP 2009. Stata Statistical Software. College Station, TX: StataCorp LP. SUSLOW, V. Y. 1996. Measuring Quality Change in the Market for Anti-Ulcer Drug. In:

HELMS, R. M. & (ED) (eds.) In Competitive Strategies in the Pharmaceutical Industry. Washington, DC: American Enterprise Institute: AEI Press.

TIMUR, A., PICONE, G. & DESIMONE, J. 2011. Has the European union achieved a single pharmaceutical market? International Journal of Health Care Finance and Economics, 11, 223-244.

VELLA, F. 1993. A simple estimator for simultaneous models with censored endogenous regressors. International Economic Review, 441-457.

VERNIERS, I., STREMERSCH, S. & CROUX, C. 2011. The global entry of new pharmaceuticals: A joint investigation of launch window and price. International Journal of Research in Marketing, 28, 295-308.

WAGNER, J. L. & MCCARTHY, E. 2004. International differences in drug prices. Annu. Rev. Public Health, 25, 475-495.

WHO 2003. Making Choices in Health: WHO Guide to Cost-Effectiveness Analysis Geneva, World Health Organization.

WHO. 2012. ATC/DDD Index 2012 [Online]. Available: http://www.whocc.no/atc_ddd_index/.

WILKING, N., JÖNSSON, B. & I STOCKHOLM, H. 2005. A pan-European comparison regarding patient access to cancer drugs. Karolinska Institutet.

WOOLDRIDGE, J. M. 2002. Econometric analysis of cross section and panel data, The MIT press.

Modeling global pricing and launching of new drugs · Modeling Global Pricing and ... his helps in...

Documents

Transcript of Modeling global pricing and launching of new drugs · Modeling Global Pricing and ... his helps in...