Recent U.S. Regulatory Efforts on Complex Innovative Clinical Trial Design - PowerPoint Presentation

1. Recent U.S. Regulatory Efforts on Complex Innovative Clinical Trial DesignJohn Scott, Ph.D.Director, Division of BiostatisticsOffice of Biostatistics and EpidemiologyCenter for Biologics Evaluation and ResearchFDA

2. DisclaimerThis presentation reflects the views of the author and should not be construed to represent FDA’s views or policieswww.fda.gov

3. OutlineFDA adaptive design guidanceFDA complex and innovative trial design (CID) interactions guidanceCID pilot review programPilot program examples

4. Adaptive design guidance

5. BackgroundFDA published a draft guidance in 2010 on Adaptive Design Clinical Trials for Drugs and BiologicsIntent was to encourage use of AD trials that fit well into regulatory decision-makingClassified designs into “Well-Understood” or “Less Well-Understood”Seen by some as inherently discouraging of designs in the latter categoryWhat was less well understood in 2010 shouldn’t remain so forever….

6. New adaptive design guidanceNew guidance finalized in November 2019Replaces 2010 draftMoves away from categorization as well-understood or less well-understoodFocuses on key principles in design, conduct, analysis, and reportingExpands discussion on technical aspects such as estimation, simulations, Bayesian methodsAdds clarity on what should be submitted

7. Regulatory principles for adaptive designsChance of erroneous conclusions should be adequately controlledEstimation of treatment effects should be sufficiently reliableDetails of design should be completely pre-specifiedTrial integrity should be appropriately maintained

8. CID INTERACTIONS Guidance

9. CID Interactions guidanceDraft issued September, 2019, finalized December, 2020Satisfies, with the AD guidance, a 21st Century Cures Act mandateDiscusses interacting with FDA on CID proposalsSome detail on specific design considerationswww.fda.gov

10. ScopeDoesn’t precisely define CIDMoving targetGenerally, CID designs are those that have rarely or never been used to support approval in a given indicationOften need for trial simulations for operating characteristicsPrimary focus is confirmatory designsDoes not identify which specific designs are appropriate for regulatory purposesInstead provides structure for communicationwww.fda.gov

11. CID includes:Complex Adaptive Designs (frequentist or Bayesian)Seamless designsFor example, combining dose-finding and hypothesis confirmation in a trial Adaptations to multiple design features such as treatment arm selection, patient allocation, or endpoint selectionFormal incorporation of “prior” information Placebo augmentation using external controls or other data sources (e.g. real world data)Leveraging or borrowing strength from information internal or external to the trialwww.fda.gov

12. Interacting with FDAMeeting availabilityExtra meetings may be available to discuss trial simulations, especially for serious and life-threatening conditionsPilot programCommon elements to be submitted for reviewBayesian featuresPriorsDecision criteriaSimulations (covered in detail in AD guidance)www.fda.gov

13. CID Pilot program

14. CID Pilot Program motivationTo be successful with an innovative proposal, sponsors may need:Robust regulatory feedbackHigh-level buy-inTo encourage the use of innovative designs, FDA needs:Case studies that we can talk about publiclyHence, the CID Pilot Program

15. Pilot review programJoint CDER/CBER effort Sponsors submit designs have the opportunity to engage with regulatory staff on designs via two meetings Agency will select up to 2 submissions per quarteruses the design as a case study for continuing education and information sharingMeetings led by Biostatistics groups (CDER/OTS/OB or CBER/OBE/DB) with participation from all relevant disciplines Five year duration

16. Eligibility criteriaThe sponsor must have a pre-IND or IND number for the medical product(s) included in the CID proposal with the intent of implementing the CID in the pilot program application The proposed CID is intended to provide substantial evidence of effectiveness to support regulatory approval of the medical productThe trial is not a first in human study, and there is sufficient clinical information available to inform the proposed CIDThe sponsor and FDA are able to reach agreement on the trial design information to be publicly disclosed

17. Evaluation of proposalsNeed for simulations to assess trial design operating characteristics Therapeutic needTrial design appropriateness for CID Pilot Meeting ProgramLevel of innovation of the trial designValue proposition of the CID

18. Websitehttps://www.fda.gov/CIDpilot

19. Manuscripthttps://doi.org/10.1177/17407745211050580

20. CID Pilot program examples

21. Case example 1Population: Duchenne muscular dystrophyRandomized, double-blind, placebo-controlled, phase 2/3 trialPrimary endpoint change in dystrophin levels from baseline to a specified timepointImportant secondary endpoint: Change from baseline in a clinical outcome assessmentEndpoints analyzed via a Bayesian repeated measures model with multiple interim analyses

22. Case example 1Bayesian adaptive design with the following potential adaptations:Stop the trial for efficacy or safetyModify the sample sizeDrop an armPool dosesChange randomization ratioAlso proposed to explore placebo augmentation with historical controlsRegulatory discussions focused on areas needing clarification and simulation space

23. Case example 2Randomized, double-blind, placebo-controlled, master protocol to evaluate multiple interventions across multiple pain conditionsPrimary endpoint pain relief from baseline measured on a numerical scaleBayesian mixed-model repeated measures analysis to allow for:Borrowing patient information from placebo groups within a pain conditionBorrowing treatment effect information across pain conditions for a given intervention

24. Case example 2Possible adaptationsStop for futilityModify sample sizeAdd or remove armsRegulatory discussions focused on:Potential drift in placebo responseExchangeability assumptions across pain conditionsMissing data frequently encountered in chronic pain trials

25. Case example 3Population: Systemic lupus erythematosusRandomized, double-blind, Bayesian adaptive designPatients randomized to placebo or one of three dosesPrimary endpoint: Response index at a specified timepointPrimary analysis was Bayesian hierarchical model with dynamic borrowing to estimate response ratesFeaturesResponse adaptive randomizationInterim analyses for futility and to inform dose and endpoint selection for future studiesRegulatory discussions focused on simulationsNuisance parameters and additional scenarios

26. Case example 4Population: oncology setting with unmet needMulticenter randomized trial with both internal and external control armsPrimary endpoint: Progression-free survivalAnalyzed with internal control arm comparisonKey secondary outcomes including overall survival would incorporate external control information with a Bayesian dynamic borrowing approachRegulatory discussions focused on:Patient populationModel assumptionsIncorporation of propensity score as a covariate

27. Related FDA initiativesBenefit-Risk AssessmentPatient-Focused Drug DevelopmentModel-Informed Drug DevelopmentReal-World EvidenceRare Disease initiativeswww.fda.gov

28. www.fda.gov