Conditional approval: discussion points from the PSI conditional approval expert group

PHARMACEUTICAL STATISTICS

Pharmaceut. Statist. 2008; 7: 263–269

Published online 10 September 2007 in Wiley InterScience

(www.interscience.wiley.com) DOI: 10.1002/pst.309

Conditional approval: discussion points

from the PSI Conditional Approval

Expert Group

Kevin Carroll1,2, Hilary Chaudri Ross1,3, Dan Evans1,4, Lesley France1,2

Rob Hemmings1,5, Sara Hughes1,6, Frances Lynn1,7, Paul Mahoney1,8

and Alan Phillips1,9,*,y

1PSI Discussion Group on Conditional Approval2AstraZeneca, UK3Novartis Pharma AG, Switzerland4Pfizer, UK5MHRA, UK6GlaxoSmithKline, UK7Biogendic, UK8Roche, UK9 ICON Clinical Research, UK

The European Agency for the Evaluation of Medicinal Products has recently completed the

consultation of a draft guidance on how to implement conditional approval. This route of application is

available for orphan drugs, emergency situations and serious debilitating or life-threatening diseases.

Although there has been limited experience in implementing conditional approval to date, PSI

(Statisticians in the Pharmaceutical Industry) sponsored a meeting of pharmaceutical statisticians with

an interest in the area to discuss potential issues. This article outlines the issues raised and resulting

discussions, based on the group’s interpretation of the legislation. Conditional approval seems to fit well

with the accepted regulatory strategy in HIV. In oncology, conditional approval may be most likely

when (a) compelling phase II data are available using accepted clinical outcomes (e.g. progression/

recurrence-free survival or overall survival) and Phase III has been planned or started, or (b) when data

are available using a surrogate endpoint for clinical outcome (e.g. response rate or biochemical

measures) from a single-arm study in rare tumours with high response, compared with historical data.

The use of interim analyses in Phase III for supporting conditional approval raises some challenging

issues regarding dissemination of information, maintenance of blinding, potential introduction of bias,

ethics, switching, etc. Copyright # 2007 John Wiley & Sons, Ltd.

Keywords: conditional approval; oncology; HIV

yE-mail: [email protected]*Correspondence to: Alan Phillips, ICONClinical Research, UK.

Copyright # 2007 John Wiley & Sons, Ltd.

1. BACKGROUND

The European Agency for the Evaluation of Medic-inal Products has recently completed consultation of adraft guidance on how to implement conditionalapproval. This route of application is available for:

1. orphan drugs;2. emergency situations;3. seriously debilitating or life-threatening diseases.

Conditional approval implies that the clinical dataupon which an opinion on a conditional marketingauthorization is based will not be comprehensive.However, a requirement in the legislation is thatthe risk–benefit balance, as defined in Article1(28a) of Directive 2001/83/EC, should be posi-tive. The clinical database will need to besupplemented post-conditional approval for a fullapproval to be granted. This supplementation willform a part of the conditional marketing author-ization-specific obligations, and is, therefore,mandatory.

Although there is very limited experience inimplementing conditional approval to date, inpreparation for the release of the draft guidancedocument in December 2006, pharmaceuticalstatisticians with an interest in the subject matteror other related issues met in November 2006 todiscuss potential issues with the new legislation.The meeting was sponsored by PSI (Statisticians inthe Pharmaceutical Industry), a professional asso-ciation of statistics in the pharmaceutical industry.This article describes the issues raised and theresulting discussions, which were based on thegroup’s interpretation of the legislation. In parti-cular, how conditional approval was perceived tobe different from regular and exceptional ap-proval. Examples of how conditional approvalapplies to HIV and how it might also apply tooncology are also discussed.

2. REGULAR, EXCEPTIONAL ANDCONDITIONAL APPROVAL

Most applications for marketing authorizationcomprise the submission of comprehensive data

so that a full assessment of the risk benefit of thenew medicinal product can be made. In general,the data result from randomized active andplacebo-controlled clinical trials.

Until recently, the only other way of obtaining amarketing authorization in Europe was to applyfor an approval under exceptional circumstances.This route of application is available when it is notpossible to provide comprehensive data because ofspecific circumstances such as rarity of the diseasebeing studied, medical ethics preventing thecollection of such data, etc. When such a market-ing authorization is granted usually there is nospecific requirement to provide clinical trialupdates because it is accepted that clinical trialscannot be done, although increased pharmacov-igilance activities are likely to still be required.

Conditional approval is now available tosponsors as an alternative strategy for obtaininga marketing authorization. With conditionalapproval, a positive risk–benefit balance needs tobe established, but this can be based on prelimin-ary evidence. Crucially, and unlike approval underexceptional circumstances, it seems that thesponsor must be able to provide, in due course, afull data package confirming efficacy, safety and apositive risk benefit. Only then can full approvalbe granted. Furthermore, the conditional approvalwill be valid for 1 year, renewable annuallythereafter until full approval. Regulatory autho-rities have the option to suspend, revoke, with-draw or vary the authorization if a product isviewed as harmful or lacking therapeutic efficacybased on the post-conditional approval data. Asummary of the perceived differences betweenconditional and exceptional approval is given inTable I.

Conditional approval was perceived to beanalogous with the Food and Drug Administra-tion’s (FDAs) accelerated approval process, be-cause for accelerated approval companies need toshow that the drug is reasonably likely to offer aclinical benefit. Sponsors are required to provide, asa condition of the approval, confirmatory evidenceof efficacy and safety. If not marketing authoriza-tion can be withdrawn. Between 1992 and 2004, 22oncology drug applications were approved via the

Copyright # 2007 John Wiley & Sons, Ltd. Pharmaceut. Statist. 2008; 7: 263–269DOI: 10.1002/pst

264 K. Carroll et al.

accelerated approval route [1]. Interestingly as ofmid-2004, only six had completed confirmatorytrials and none had been removed from the market.Once available, removal of an anti-cancer medicinehas proven to be extremely difficult unless there isan emerging safety concern. Subsequently, it isconceivable that only a clear change in the risk–benefit profile, probably arising from a new safetyconcern would lead to the withdrawal of an EU-granted conditional marketing authorization. Withthis in mind, it is possible that the implementationof conditional approval in Europe will be cautious,at least in the first instance, with regard to theamount and strength of data on efficacy available.

Two therapeutic areas where the group thoughtconditional approval may be applicable were oncologyand HIV. In Sections 3 and 4, we discuss howconditional approval might be applicable to these areas.

3. APPLICATIONS OFCONDITIONAL APPROVAL INONCOLOGY

Within oncology, there are three potential scenar-ios where sponsor companies might consider usingconditional approval; namely the following:

1. Completion of phase II trials using acceptedclinical outcomes such as progression/recur-rence-free survival or overall survival (OS).

2. Completion of phase II trials using a surrogateendpoint for clinical outcome such as responserate or biochemical measures.

3. Results of a planned interim analysis in apivotal phase III trial.

In this section we will discuss the suitability ofconditional approval for each scenario.

3.1. Phase II trial using clinical endpoints

Consider a randomized, double-blind Phase IIclinical trial with a clinical endpoint such asprogression-free survival (PFS). Suppose that thetrial is designed to yield a high probability, say, atleast 90% power that the drug has a clinicallymeaningful effect. Furthermore, Phase III trialsare planned and ongoing. This scenario seems tobe an ideal candidate for conditional approval,assuming the risk/benefit ratio following the PhaseII trial is beneficial. Firstly, it satisfies the criteriaof a positive risk/benefit balance based on aconditional database. Secondly, supplemental datawill be available to support the application.

An example of such an application could beBRCA mutant (familial) breast cancer wherepresentation with untreated advanced diseaseis rare (55% of all advanced breast cancer).Suppose that a 2-year Phase II trial in 150 patientslooking for a 50% increase in PFS has providedsufficient data to show that the probabilitythe drug had an effect was at least 90%. A 4-yearPhase III confirmatory trial to show a 25%increase in survival would require 900 patients.Thus, the conditional approval route wouldseem a sensible approach for this life-threateningdisease.

3.2. Phase II trial using non-clinical endpoints

Instead of clinical endpoints, suppose that somesurrogate endpoint has been used such as responserate. Such response rate trials are not uncommonin refractory disease settings where all availabletherapeutic options have been exhausted. Thesetrials tend to be single arm without control sincespontaneous tumour regression is argued to beextremely rare. Provided that a high response rate

Table I. Conditional versus exceptional approval.

Conditional approval Exceptional approval

Granted before all data are available Comprehensive data cannot be provided (e.g. too rare)Authorization valid for one year (renewable) Annual reassessment of the risk/benefit balanceObligations: further clinical studies to verifybenefit/risk balance

Obligations: specific procedures, in particular, concerning safety

Data package: initialþ obligation ¼ regular Data package: initialþ obligations5regular

Conditional approval 265


is observed, conditional approval may be asuitable regulatory submission strategy. Rando-mized confirmatory trials could still be undertakensince the relationship between response rate andtreatment effect on clinical outcomes like PFS orOS is often unknown.

An example of such an application is Sutent(sunitinib malate) in metastatic renal cell carcino-ma [2]. A conditional approval was originallygranted based on two single-arm phase II trials inpatients having experienced failure to previouscytokine therapy. Core imaging laboratory as-sessed response rates were approximately 25% inboth trials, and investigator assessed responserates approximately 36%. Moreover, additionaldata, albeit uncontrolled, also supported thedurability of the response and presented effectson PFS and OS. A Phase III trial comparingSutent against Interferon-a as first-line treatmentwas ongoing at the time of the conditionalapproval and the company was given a specificobligation to provide response rates and durationof response from the 1st interim analysis, and PFS,OS and duration of response from the finalanalysis. The CHMP considered that data fromthis trial would provide comprehensive clinicaldata about the product. These data were subse-quently provided and the authorization amendedfrom conditional to a regular marketing author-ization.

At the meeting, there was some discussion as towhether companies would want to use the condi-tional approval route for the above scenario orwhether exceptional approval might be better.With exceptional approval, there is no burden toprovide confirmatory evidence. The rationale forexceptional approval might be that the highresponse rate (even though not proven to belinked to better clinical outcome) could make arandomized confirmatory trial in second-linepatients impossible.

3.3. Planned interim analysis of phase III

This third application discusses conditional ap-proval following an interim analysis of a Phase IIItrial. Although on the surface it seems that

conditional approval might be a viable strategy,we will show that the use of conditional approvalfollowing interim analyses is problematic.

Consider a randomized, double-blind Phase IIIclinical trial with a primary clinical endpoint suchas PFS or OS. These endpoints are commonly usedin oncology trials as outlined in the FDA guidancedocument entitled Clinical Trial Endpoints for theApproval of Cancer Drugs and Biologics [3]. If asignificant treatment effect is established on thisendpoint at a planned interim, then the null isrejected and the objective of the trial met. It isdifficult to see how conditional approval can applyunder such circumstances. The trial has met itsprimary endpoint and addressed its objective andso the logical course of action is to seek fullapproval via a regular application strategy. Thus,it would seem that conditional approval is reallyonly viable if the interim is based on somesecondary clinical endpoint or a biomarker end-point. However, such a scenario – an interim basedon a secondary or biomarker endpoint – is highlyproblematic which may explain, in part, why veryfew are seen in practice.

It is questionable whether supplemental datacan be provided to support any conditionalapproval application following unblinding/disse-mination of the interim results on the secondary/biomarker outcome. In the scenario described theintegrity of the study is likely to be jeopardizedfollowing the interim analysis; for example, if thereis sufficient evidence to support a conditionalapproval, it is very likely that the study design willbe affected. If the drug is good enough to beapproved and prescribed to patients matchingthose in the trial, is it really ethical to continuerecruitment/exposure to the control treatment?Patients (in consultation with their physicians)may wish to switch treatment regimens and, so, thedata collected following the interim analysis arelikely to be difficult to interpret and quite possiblydownwardly biased. Release of interim data in thisway might therefore serve only to make it difficult,if not impossible, to provide full data confirming atreatment benefit and thus would serve to circum-vent, rather than support, a conditional approvalstrategy.



4. APPLICATIONS OFCONDITIONAL APPROVAL IN HIV

Conditional approval seems to fit very well with theaccepted regulatory strategy in HIV. Indeed, theFDA issued a Guidance for Industry on clinicalconsiderations for accelerated and traditional ap-proval of HIV antiretroviral drugs in 2002 [4], therequirements for accelerated and conditional ap-proval having much in common. Likewise, theEMEA’s guideline on the development of medicinesfor treatment of HIV infection released at the endof 2005 has a section dedicated to early approval orconditional marketing authorization [5].

HIV antiretroviral therapy typically consists ofa cocktail of three or four drugs. Due to issues ofdevelopment of drug resistance, patients and theirdoctors do not wish to stop or switch a drug unlessabsolutely essential. Hence, clinical trials in HIVcan continue for a number of years until aftermarketing authorization has been gained andcommercial supplies of the drug are available inall relevant countries.

Plasma HIV RNA measurement (viral load) is awell-accepted surrogate marker in clinical trials [4].Typical data for FDA accelerated approvalapplications include demonstration of shorter-term reductions in HIV viral load, with subsequenttraditional approval based on evidence of dur-ability of viral load suppression. For example, inan FDA application focused on heavily pre-treated patients, accelerated approval applicationmay be based on 24-week viral load data frominterim analyses of randomized controlled trials(i.e. interim analysis occurs after full patient sethas completed 24 weeks of trial), with subsequenttraditional approval application based on 48-weekviral load data from the same trials, and with thetrials continuing still further.

In the past, non-inferiority appears to have beensufficient to meet FDA requirements for acceleratedapproval of HIV drugs [4]. However, EMEArequirements for application for conditional ap-proval [5] appear to include demonstration offulfilment of some unmet medical need (e.g. super-iority in terms of efficacy, safety or quality of lifemay be expected). This apparent difference could be

an artefact of the different times of introduction ofaccelerated and conditional approval routes incomparison with the development history of HIVantiretroviral drugs. The very limited number oftreatment options available to patients with HIVaround the time that accelerated approval wasintroduced, and the need for further drugs due todevelopment of resistance with existing drugs,meant that a demonstration of non-inferiority wassufficient. Several years on, with more treatmentoptions available to patients with HIV, it remainsto be seen whether FDA-accelerated approvalrequirements remain the same or move into linewith assumed EMEA requirements.

Whilst data from early interims are sometimeswithheld from the public arena until data from alater key interim are available, early interim resultsfor new regulatory filings (e.g. week 24) are oftenpublicly released prior to later data (e.g. week 48)being available, which are submitted as follow-updata. Despite issues regarding bias, this practiceappears to have been acceptable in the HIV arena.However, the potential for bias – particularlywhen the trials are open label – has receivedscrutiny recently. Indeed, there is published criti-cism of companies publicly announcing interimresults before key later data are available [6].

There was some discussion at the meeting as towhether or not it is ethical to withhold results frominterim analyses, whatever the disease in question.No consensus was reached, but the broad feelingwas that it was not because of the nature of thediseases being studied. This makes conditionalapproval a problematic strategy for the reasonsoutlined above.

5. ROLE OF STATISTICIANS

At the discussion group, the role of statisticians inconditional approval was reviewed. Four key areaswere identified. These included the following:

* Ensuring robust, well-designed, randomizedclinical trials.

* Interim analyses are appropriately planned, andIndependent Data Monitoring Committeesused whenever possible.



* Assisting with the assessment of the risk/benefitbalance so that an early significant treatmenteffect identified during an interim analysis is notlost with more mature data.

* Provide data and reassurance regarding truesurrogacy of intermediate endpoints used tosupport conditional approval applications.

To illustrate the above consider a randomized,double-blind, parallel-group two-treatment study.The study is designed to test the null hypothesis(H0) that the hazard ratio (HR) equals one, againstthe alternative hypothesis (H1) that the HR equals0.8 for PFS with 90% power using a two-sided 5%significance level. Then 844 events need to beobserved, which approximates to a 394 versus 450event split or better to achieve a two-sidedsignificance level of 5% or less. Further, an interimanalysis is planned after one-third of the events areobserved, and a significance level of 50.001 isrequired.

Suppose at the interim analysis the HR equals0.65 (p ¼ 0:0003). Everything looks good, butconcern is expressed that the interim analysis istoo early and, with two-thirds of the eventsremaining, the final results cannot be significant.This is a real issue in the adjuvant treatment wherethe percentage of patients with an event is oftenvery low. The statistician can provide reassurancethat the early significant treatment effect identifiedduring the interim analysis will not be lost withmore mature data. Observing a HR of 0.65 afterone-third of the events equates to an approximate111 versus 170 split of events. To lose significanceat the end of the trial, the remaining events wouldneed to be split 283 versus 280, which equates to aHR of 1.01 or worse. This seems rather unlikely.In fact the statistician can show:

1. Prob (Obs HR > 1:01 in outstanding 2/3 events/H1)¼ 0:0084:

2. Prob (Final HR ¼ P50:05/interim data andH1)¼ 99:7%:

3. Prob (Final HR ¼ P50:05/interim data)¼ 98:8%:

The probability estimate for point 2 utilizesconditional power, whereas point 3 is uncondi-tional (predictive) power.

Finally, it should be noted that the statisticianhas a significant role to play in trying to addressmethodological issues associated with conditionalapproval where a large number of questionsremain unanswered. For example, what are thesteps that can be taken to control for sources ofbias when interim results are known (or at leastknown to be positive) because of submissions toregulatory bodies. What are the implications forongoing/future studies with regard to the choice ofcontrol arm, selection of patient populations, etc.How should patients be recruited/data collectedbetween an interim analysis leading to submissionand agreement with Committee for MedicinalProducts for Human Use (CHMP) be handled?

6. CLOSING REMARKS

Conditional approval is now available to sponsorsas a viable regulatory strategy for:

1. orphan drugs;2. emergency situations;3. seriously debilitating or life-threatening diseases.

The clinical data upon which conditional market-ing authorization is based will not be comprehen-sive. However, sponsors will still need todemonstrate a positive risk–benefit ratio. As acondition of the marketing authorization, theclinical database will need to be supplementedwith post-conditional approval for a full approvalto be granted. This supplementation will form partof the conditional marketing authorization specificobligations, and is, therefore, mandatory.

Conditional approval seems to fit well with theaccepted regulatory strategy in HIV. HIV viralload is a well-accepted surrogate marker. In anHIV study, interim results from shorter-term timepoints seem to be routinely submitted for ap-proval, with longer-term data submitted as follow-up data. Results are often, but not always,disseminated in the public arena after the firstinterim analysis for regulatory application.

Opportunities exist for the use of conditionalapproval in oncology. However, the use of interimanalyses in Phase III for supporting conditional



approval (and, indeed in other indications) may beproblematic.

The current conditional approval legislationappears to only apply to the initial marketingapplication for a compound and does not seem tocover the potential issue with label variations afteran initial conditional approval is granted.

The data required to upgrade from a conditio-nal approval need to be carefully considered andplanned. It can be very problematic if not plannedahead. The generation of the supplemental dataaimed at converting to a normal marketingauthorization should not be unduly affected bythe submission of data for a conditional approval.

REFERENCES

1. Dagher R, Johnson J, Williams G, Keegan P, PazdurR. Accelerated approval of oncology products: adecade of experience. Journal of the National CancerInstitute 2004; 96(20):1500–1509.

2. Sutent Scientific Discussion. Available at: www.emea.europa.eu/humandocs/pdfs/EPAR/sutent/06870en6.pdf (accessed 21.06.2007).

3. Food and Drug Administration, Centre for DrugEvaluation and Research. Guidance for Industry:Clinical Trial Endpoints for the Approval of CancerDrugs and Biologics. Electronic document availableat: http://www.fda.gov/cder/guidance/7478fnl.htm(accessed 21.06.2007).

4. Food and Drug Administration, Centre for DrugEvaluation and Research. Guidance for Industry:Antiretroviral Drugs using Plasma HIV RNA Mea-surements – Clinical Considerations for Acceleratedand Traditional Approval. Electronic documentavailable at: http://www.fda.gov/cder/guidance/3647fnl.htm (accessed 21.06.2007).

5. European Medicines Agency, Committee for Medic-inal Products. Guideline on the Clinical Developmentof Medicinal Products for the Treatment of HIVInfection (CPMP/EWP/633/02, Rev. 1). Electronicdocument available at: http://www.emea.europa.eu/pdfs/human/ewp/063302en.pdf (accessed 25.04.2007).

6. Laessig KA, Lewis LL, Hammerstrom TS. TenofovirDF and emtricitabine vs. zidovudine and lamivudine:letter to the editor. New England Journal of Medicine2006; 254(23):2506–2507.



Conditional approval: discussion points from the PSI conditional approval expert group

Documents

Transcript of Conditional approval: discussion points from the PSI conditional approval expert group