Regulatory Science Symposium: Regulatory Aspects of Clinical Trial Design Session 3: Classic and Novel Designs Used in Regulatory Approvals (2018)

1. Presentation Title: Classical and Novel Designs Used in Regulatory Approval
2. Presented by Steven Snapinn, Vice President, Global Biostatistical Science, Amgen
3. Two-part presentation
4. Part 1: Classical Design Issues
   1. Phases of Drug Development
   2. Control Groups
   3. Controlling for Bias
   4.  Multi-Regional Clinical Trials
   5. Controlling for Multiplicity
      1. Phase I: Earliest studies, First-in-Human (FIH) Studies, healthy volunteers except in oncology trials, Single Ascending Doses (SAD), Multiple Ascending Doses (MAD), (for Oncology, 3+3 Design is used, most commonly used in Phase I)
      2. Phase II: Provide preliminary evidence of a drug's clinical efficacy, considered exploratory trials
      3. Design of Phase II Trials: Trying to find the optimal dosage, more flexibility in numerical values of Type I & II errors
      4. Decision-Making in Phase II: Internal decisions such as whether to move into   phase 3 program
         1. Ex. Of Type I Error: a decision to launch a phase 3 program with an ineffective drug
         2. Ex. Of Type II Error: an incorrect decision that a treatment has insufficient efficacy to move onto phase 3
      5. Bias and Random Error
         1. Concerns for Phase 3 trials
         2. Bias: a systematic tendency to incorrectly over/underestimate true treatment effect    Ex. Placing more ill patients into the clinical trials
         3. Random error: the difference between estimate and true effect (simply over/underestimating randomly)
      6. Avoiding Bias
         1. Randomization- to avoid conscious/unconscious tendency to allocate sicker patients to a specific group
         2. Blinding (3 types---Patient/Single-Blind, Investigator/Double-Blind, Sponsor/Triple-Blind)
         3. Most clinical trials are triple-blind.
         4. Blinding sponsor avoids "operational bias" which is making changes to protocol or how the drug is given during the trial to improve/influence the consequences of drugs
      7. Minimizing Random Error
         1. Bigger sample size is primary approach 
         2. Other ways: more homogenous patient pop & better process for collecting endpoints 
      8. Single-Arm Trials 
         1. Typically used in oncology trials 
         2. Problematic: You do not know how the patient would have reacted without the drug (This type of trial lacks a control) 
      9. Placebo Controls and Active Controls  
         1. Gold Standard for Demonstrating Efficacy and Safety 
         2. Biggest Problem: Not Ethnical to withhold medication for sick patients 
         3. 2 kinds of analysis: show the drug is better than active control (easiest) or show how active control is better than the placebo and then show that your drug is close in efficacy to the active control  
      10. Continuous Variables 
          1. Ex. Change in DBP from baseline to end of study 
      11. Ordinal Variables 
          1. Ex. Patient Assessment of Treatment Benefit via. Calculation of Score or Model-based Analysis  xii.
      12. Dichotomous/Binary Variables 
          1. Ex. Disappearance or shrinkage of tumor in oncology trials 
          2. Difficult to analysis because multiple ways (risk difference, relative risk, odds ratio) to do so 
          3. FDA provides guidance and meetings with FDA to address specific concerns 
      13. Analysis of Confirmatory Trials 
          1. Intention-to-Treat Principle 
             1. Includes those who stop taking the drug who can refuse to come back for measurements 
          2. Handling Missing Data 
             1. Analysis to guess the missing data  
          3. Intercurrent Events and Estimates 
             1. New revision to ICH E9 on Statistical Principles 
      14. Controlling for Multiplicity 
          1. Testing multiple hypothesis or multiple treatment methods/drugs 
      15. Interim Analyses
          1. Phase III Trials are monitored by an External Data Monitoring Committee (DMC) to protect patient safety 
      16. Multiregional Clinical Trials 
          1. US, European countries, Japan, China = big markets for pharmaceuticals industry 
          2. ICH Guidance Document (E17) 
          3. Intrinsic factor ie. DNA/genes and Extrinsic Factor ie. Diet  
      17. Examples of Regional Inconsistencies- Merit HR Study (inconsistent data in the US) 
      18. Examples of Regional Inconsistencies – Plato Study, drugs dealing with blood clots, trying to show that drug was superior to current car (was able to do in its target sites except Africa), look at those who took aspirin as the reasoning for inconsistency 
5. Novel Designs  
   1. Adaptive Designs 
   2. Platform and Umbrella Designs 
   3. Program Decision-Making Approach 
   4. Designs for Special Populations  
      1. Rationale for Adaptive Phase 2 and 3 Trials 
         1. Substantial uncertainty exists 
         2. Adaptive clinical trials are designed to take advantage of accumulating information and modify aspects of the study based on pre-determined decision criteria without undermining validity and integrity of the trial to increase success of trials 
      2. Characteristics of Adaptive Design 
         1. Clarity of goals 
         2. Frequent "looks" at data & potential modification of trial 
         3. Adaptive by "design" 
         4. Extensive use of simulation 
      3. Types of Adaptive Designs 
         1. More traditional: changing study eligibility criteria, stopping early for the sake of efficacy, blinded sample size re-estimation, increasing population sizes
         2. Newer: Doses dropping or stopping for futility, adaptive randomization, unblinded sample sizes (to make significant statistical difference), endpoint selection, seamless phases, platform trials, basket trials 
      4. Conditions that Favor or Do Not Favor Adaptive Design 
         1. Favor: Fast turnaround of data, slow enrollment relative to primary outcome measure, availability of biomarkers that are predictive of final outcome, broad dose range to study
         2. Do not favor: Opposite of the above 
      5. What Do Adaptive Designs Buy Us? 
         1. Increases efficiency and increases accuracy  
      6. What Do Adaptive Designs Cost Us?
         1. Require greater Set-Up Time & discussion  
         2. Greater External Spend 
      7. The Adaption Process 
      8. Simulation-Guided Trial Design 
         1. Proving that data is not falsely finding that drug is being positive 
      9. Longitudinal Modeling 
         1. Used for simulation 
      10. Operational Considerations 
          1. Randomization/IVRS capability evaluation- concern that patients were not adequately randomized 
      11. Program Decision-Making Approach 
          1. Design of entire clinical program including sample sizes for phase II and phase III studies; to maximize drug's expected net present value (eNPV) 
      12. Platform and Umbrella Designs 
          1. Platform: investigate multiple drugs for same disease 
             1. Can have one sponsor but rare 
          2. Basket trial: investigate one drug for multiple ailments 
      13. Other Innovative Designs 
          1. Pediatric Studies – Bayesian Borrowing Info from Adult Pop. For randomization 
          2. Small Populations/Orphan Drugs – Relaxed Type/Error Rate Requirement, Historical Controls 
          3. Biomarker-Based Designs – Determine efficacy in biomarker-sensitive pop and in broader pop at the same time 
             1. Oncology Trials: Genetic diff. And its relation to impact is not shown until later trials 

Accompanying text created by Annie Ly | Undergraduate Research Associate and Provost's First Generation Research Fellow | lyannie@usc.edu