Regulatory Science Virtual Symposium: “Study Design for Clinical Trials: Types and Trends” Session 6: Statistical Challenges of Adaptive and Innovation Study Designs (2023)

1. Outline
   1. Overview of drug development
   2. Interim analyses and data monitoring committees
   3. Adaptive clinical trials
      1. Data monitoring issues specific to adaptive trials
   4. Other innovative study designs
2. Phases of drug development: phase 1
   1. Includes safety and dosing, First-in-human (FIH) studies, Single Ascending Doses (SAD), M Ascending Doses (MAD), 3+3 Design in oncology
3. Phases of drug development: phase 2
   1. Provide preliminary evidence of a drug’s clinical efficacy and attempt to identify an optimum dose
   2. Try to find the right dose
   3. Sample size is chosen to control error rate
   4. Decision made by comparisons of individual treatment groups
   5. Decision making:
      1. Phase 2 trials are typically designed to make decisions
      2. Decisions made based on statistical hypothesis testing
4. Phases of drug development: Phase 3
   1. Confirmatory trials designed to provide definitive evidence of efficacy and safety
      1. Benefit-risk analysis
   2. Increased scrutiny from regulatory agencies
      1. Incorrect conclusions from Exploratory (Phase I and II) trials put sponsor at risk
      2. Incorrect conclusions from Confirmatory (Phase III) Trials put the consumer a risk
5. Single-arm trials
   1. No internal control but often an external (historical) control
   2. Terminology: open-label trial vs uncontrolled (single-arm) trial
6. Placebo controls and active controls
   1. The standard for demonstrating efficacy and safety
   2. Rationale for active controls: placebo unethical, evaluation of comparative efficacy (reimbursement)
   3. Superiority vs non-interiority with active controls
      1. Non-inferiority aim to show new treatment no substantially worse than standard treatment
7. Interim analyses
   1. Many phase 3 trials are monitored by an external data monitoring committee (DMC) to protect patient safety
   2. Group sequential designs allow interim stopping for efficacy or futility
      1. Require statistical stopping rules that control Type I error rate
8. Overview of data monitoring committee
9. Some current issues with DMCs
   1. Program-wide DMCs has a broad overview of safety and efficacy but has potential for efficacy based on “meta-analysis”
   2. Stopping for futility
      1. Controversial; Can save resources and lead to early stopping for unexpected harm with respect to efficacy variable
   3. Adaptive clinical trials
10. Adaptive clinical trials
    1. Allow modifications to the protocol based on interim data (must conform to basic principles)
    2. Adaptive dose-finding
    3. Changing sample size based on estimated effect size
    4. Seamless Phase 2-3 designs
    5. Changing the focus to a marker-positive subgroup
11. Adaptive dose-finding
    1. Phase 2 trial benefit from more innovative design and analysis approaches
    2. Change allocation of dose during the process
12. Adaptive sample size re-estimation
    1. Start with a small trial to detect a great big event
    2. Sample size are calculated to provide adequate power to detect a specific magnitude of treatment effect
    3. Blinded sample size re-estimation
    4. Unblinded (adaptive) sample size re-estimation
13. Changing focus to a marker-positive subgroup
    1. Genetic marker may predict which patients will benefit from treatment
    2. Adaptive approach: interim analysis evaluating if the marker appears to be predictive
14. Adaptive trials and DMCs
    1. Many different types of adaptations (blinded vs. unblinded)
    2. Some unblinded adaptations fall outside the DMC’s typical responsibilities
       1. Examples
       2. Focus on efficacy, not safety
    3. DMC is the natural group to take responsibility
15. Composition of DMC
    1. Skill requirements may differ from adaptive decision-making
    2. DMC committees will typically have a statistician but may want someone with specific expertise in adaptive clinical trials
    3. Long commitment for a seamless phase 2/3 design
16. Need for the sponsor involvement
    1. Some adaptive decisions may be too important or complex for the sponsor to allow to be made by an external committee
       1. Allowing sponsor to be unblinded may pose risk to trial integrity
    2. Gallo (2006) recommendations
       1. Minimal sponsor representation
       2. Not otherwise involved in trial
       3. Access to results only at time of adaptation
       4. Appropriate firewalls to ensure limited access to results
17. Other innovative design
    1. Platform trials: investigate multiple drugs in same indication or disease condition
    2. Basket trials: investigate one drug in multiple indications/subtypes of disease conditions
18. The intention-to-treat principle and estimands
    1. The intention-to treat principle: includes all randomized subjects, complete follow-up regardless of protocol violations, symptom trials (e.g., LDL lowering, pain) vs outcome trials (looking for long-term survival)
    2. Revision to ICH E9 regarding estimands
       1. Intercurrent event: something that happens during course of trial which influences the ability to assess outcome
       2. Require sponsor to define estimand, how intercurrent event handled in analysis and relate to scientific question being posed
       3. Want to know: Are you interested in treatment policy estimand (corresponds to intention to treat), or some other estimand (e.g., hypothetical estimand)?
19. Estimands
    1. Recent addendum to ICH E9, statistical principles for clinical trials
    2. Components:
       1. Population, variable (or endpoint), how to account for intercurrent events, population-level summary
20. Intercurrent event
    1. Event that occurs after treatment initiation and either preclude observation of variable or affects its interpretation
21. Recommended estimands:
    1. Treatment policy strategy, composite strategy, hypothetical strategy, principal stratum strategy
22. Questions?

Accompanying text created by Roxy Terteryan, RKS Project Administrator, SC CTSI (atertery@usc.edu) and Cyan Tan, Student Worker