Multi-database networkstudies in a Nordic context · Morten Andersen 2017-06-01 1 Morten Andersen...
Transcript of Multi-database networkstudies in a Nordic context · Morten Andersen 2017-06-01 1 Morten Andersen...
MortenAndersen 2017-06-01
1
Morten Andersen
Department of Drug Designand Pharmacology
Multi-database network studies in a Nordic context
Conflicts of interest
• Participate(d) in research projects funded by AstraZeneca, H. Lundbeck & Mertz, Novartis, Nycomed, Merck Sharp & Dohme and Pfizer, with grants paid to the institutions where I have beenemployed
• Personally receive(d) fees from Medicademy, the Danish Pharmaceutical Industry Association for teaching at pharmacoepidemiology courses
• Professorship in pharmacovigilance funded by the Novo Nordisk Foundation
2
MortenAndersen 2017-06-01
2
Combining observational data from multiple countries is attractive
Possibilities/advantages
• Increased power• Rare exposures• Rare outcomes• Subgroups• Validate results across
countries• Learn from between-
country differences
Challenges
• Permissions• Logistics• Analysis
3
The study will utilise high quality prescription databases and other national data sources, integrated at European level with advanced methods of harmonising data
CAncer Risk and INsulin analoGues
4
MortenAndersen 2017-06-01
3
CARING Study declaration of interests
The research leading to the results of this study has received funding from the European Community’s Seventh Framework Programme (FP-7) under grant agreement number 282526, the CARING project. The funding source had no role in study design, data collection, data analysis, data interpretation or writing of the presentation.
Research questions in CARING WP3
• To develop a common data model allowing the integration of information from differently structured databases
• To evaluate meta-analysis of aggregate patient data (APD) as compared with individual patient data (IPD) for the analysis of the association between exposure to different types of insulin and risk of cancer
6
MortenAndersen 2017-06-01
4
Data sources
STUDY PERIOD1989 2013
CARING
CPRD (UK)
Finland
Denmark
Norway
Sweden
• National registers from Denmark, Norway and Sweden and the UK CPRD
7
Prescribed Drug Register
Drug, dispensing form, strengthPackage, amount, dose textATC code, DDDsPatient and prescriber information
Hospital Electronic Records
Clinical informationMeasurements and lab dataInpatient medical treatmentDischarge summary
LISA
Education, income, employmentCountry of birthPlace of residence
Quality registers
Specific diagnoses, treatmentAge at onset of diseaseClinical and lab dataDisease severityRisk/prognosis assessmentBMI, smoking
Medical Birth Register
Identity of mother and childSocial factors, smoking, snuffMaternal, pregnancy, delivery and infant clinical information
Total Population Register
Dates of Immigration, emigration, death
Cause of Death Register
Date of deathCauses of deathInjury, poisoning, intent
Cancer Register
Date of cancer diagnosisType of malignancy, histologyTNM staging
Primary Care Electronic Records
Date of contact, diagnosesClinical and lab dataWeight, blood pressureBMI, smoking, alcoholPrescribed medicine
Patient Register
Inpatient admission and discharge datesOutpatient contact dateHospital, clinic, countyDiagnoses, procedure codes
Multi-Generation Register
Identity of relatives
Swedish health registers
Personal identity number
Laboratory data
Date, measurement, value, unit
8
MortenAndersen 2017-06-01
5
Data sources
STUDY PERIOD1989 2013
CARING
CPRD (UK)
Finland
Denmark
Norway
Sweden
• Nordic healthcare registers are relatively similar in content and structure, easily standardised
9
Data management in an epidemiologicalstudy
DATA IMPORT AND CLEANING
GENERATION OF ANALYSIS DATASET
STATISTICAL ANALYSIS
SIMPLIFIED:
RAW DATA CLEANED DATA ANALYSIS DATASET
RESULTS
DATABASE SOFTWARE
STATISTICAL SOFTWARE
10
MortenAndersen 2017-06-01
6
CARING study data challenge
Five complex databases with different medical classification systems – like five different studies
ICD7ICD8ICD9ICD10ICDO3NCSPATC
DK
FI
NO
SE
UK READCPRD PC
11
CARING study data solution
Alternative: decide on a common data model
ICD7ICD8ICD9ICD10ICDO3NCSPATC
DK
FI
NO
SE
UK READCPRD PC
12
MortenAndersen 2017-06-01
7
Overview of CDM data flow in the CARING project
DK NO SE UKFI
Nordic Data Model
CARING Common Data Model
Analysis datasets
Concept Dictionary
Statistics Denmark
13
Objective: No diagnoses and drugs hardcoded in statistical programs, all definitions contained in a database
CARING CDM – Concept Dictionaryand Breast Cancer Concept
MortenAndersen 2017-06-01
8
Multi-database analytic approaches
• Individual patient data meta-analysis• Pooled data in one place necessary• Adjustment for confounders common to all databases
• Aggregate data meta-analysis• Distributed analysis in separate databases possible• Adjustment for database-specific confounders (country-
optimized)
• Method paper: Preliminary results from 3 countries
15
Glargine versus human insulinGlargine versus human insulin(preliminary analysis)
MortenAndersen 2017-06-01
9
Detemir versus human insulin(preliminary analysis)
Other insulin versus human insulin – differentadjustment models (preliminary analysis)
18
MortenAndersen 2017-06-01
10
19
Pooled individual patient analysis ormeta-analysis on aggregate data?
Common confoundercovariates (+)
(-)
Power in each dataset (+)
(-)
Favors aggregate level Meta-analysis
Favors individual patient analysis
Nordic studies of SSRIs in pregnancy
20
MortenAndersen 2017-06-01
11
Association of SSRI/venlafaxine and cardiovascular birth defects(Selmer 2016 Pharmacoepidemiol Drug Saf)
21
• Fixed effects meta-analysis and individual-basedanalyses of a pooled dataset gave similar results
• Country-optimized adjustment attenuated the estimates more than adjustment for commoncovariates only
• When data are sparse, pooled data on the individuallevel are needed for adjusted analyses
Association of SSRI/venlafaxine and cardiovascular birth defects(Selmer 2016 Pharmacoepidemiol Drug Saf)
22
MortenAndersen 2017-06-01
12
Multi-database networks in pharmacoepidemiology
23
BIG(HEALTHCARE)
DATA
A common Nordic pharmacoepidemiology database for the future?
Approaches
• Distributed database network + analysis programs
• Distributed network + secure servers and commonanalysis platforms for individual-based data
• Common Nordic database, all in one place
24
MortenAndersen 2017-06-01
13
A common Nordic pharmacoepidemiology database for the future?
All approaches involve
• Collaboration on a common data model
• A common platform for data management and analysis
• A coordinated process for data access (legal, ethical, mutual agreements)
25
Opportunities in safety and effectivenessresearch using Nordic healthcare databases
• Pharmacovigilance: Systematic proactive surveillance, signal detection and risk evaluation using healthcaredatabases
• Comparative safety and effectiveness: Observationalstudies and pragmatic clinical trials linked to healthcare databases (including electronic healthrecords and quality registers)
• Developments in personalised/precision medicinelinking healthcare data to biobanks
26
MortenAndersen 2017-06-01
14
References
• Furu K, Wettermark B, Andersen M, Martikainen JE, Almarsdottir AB, Sørensen HT. The Nordic Countries as a cohort for pharmacoepidemiological research. Basic Clin PharmacolToxicol, 2010;106(2):86-94.
• Wettermark B, Zoega H, Furu K, Korhonen M, Nörgaard M, Sörensen HT, Almarsdottir AB, Andersen M, Andersson K, Bergman U, Englund C, Hallas J, Heli-Salmivaara A, Hoffmann M, Kieler H, Martikainen J, Mortensen M, Petzold M, Ringbäck Weitoft G, Skurtveit S, Wallach Kildemoes H, for the Nordic Pharmacoepidemiological Network (NorPEN). The Nordic prescription registries as a resource for pharmacoepidemio-logical research – A systematic literature review. Pharmacoepidemiol Drug Saf 2013;22(7):691-9.
• Bazelier MT, Eriksson I, de Vries F, Schmidt MK, Raitanen J, Haukka J, Starup-Linde J, De Bruin ML, Andersen M. Data management and data analysis techniques in pharmacoepidemiological studies using a pre-planned multi-database approach: a systematic literature review (review article). Pharmacoepidemiol Drug Saf, 2015 Sep;24(9):897-905. doi: 10.1002/pds.3828. Epub 2015 Jul 14.
• Selmer R. Haglund B, Furu K, Andersen M, Nørgaard M,Zoëga H, Kieler, H. Individual-based versus aggregate meta-analysis in multi-database studies of pregnancy outcomes: The Nordic example of selective serotonin reuptake inhibitors and venlafaxine in pregnancy. Pharmacoepidemiol Drug Saf. 2016 May 19. doi: 10.1002/pds.4033.
28
MortenAndersen 2017-06-01
15
29