Characteristics and Principles of Adaptive Trial Designs#

Presenter : Jeongsun Park#

Introduction#

Conventional clinical trials = fixed sample trial design#

  • A single analysis is performed when the determined sample size is reached

  • Plan and design the trial based on what’s available regarding population, interventions, outcomes, and other parameters - some degree of uncertainty exists

  • The trials will be at risk if it is not planned or designed accurately

  • Unplanned ad hoc modifications (amendment of protocol) are common

Adaptive trial design = flexible with data-driven approach#

  • Use accumulated (interim) trial data to modify aspects of an ongoing trial (pre-planned)

  • Trials to be more responsive >> lower risks related to uncertainties at the planning phase

  • Can be applied from phase I to phase III

Unifying Property of Adaptive Trial Designs#

Key factor of adaptive trial design#

  • Use of accumulating interim data in the form of the interim analyses

  • Pre-specified plan to adapt the design

Interim Analyses#

  • “Comparative analyses”, “Unblinded analyses”

  • Sometimes involves breaking treatment groups and comparing study arms for efficacy and safety

  • Unbreaking the treatment groups is also possible (ex. Total events observed to re-estimate the sample size)

Flow of Adaptive Trial Designs#

  • Planning phase

      1. Review background information (ex. Relevant clinical context, historical data, etc)

      1. Create realistic scenarios and run statistical trial simulations

      1. Discuss and repeat simulations until needed

      1. Finalize trial design

      • Establish a pre-specified plan for interim analyses

      • ex. How the interim data will be used, who will perform the analyses, who will review the interim data, other statistical details

      • The interim analyses plan includes details on endpoint, statistical analyses, decision rules, timing and frequency of interim analyses

  • Operating phase

      1. Enroll subjects

      1. Go through the ‘burn-in period*’

      1. Conduct interim analyses

      1. Review the statistical rules if the target sample size is not met

      1. Adapt the protocol if required

      1. Determine if the trial can be terminated (ex. Target size reached, stopping rules are met)

      1. Conduct final analysis if ready

    • Monitoring for safety and adverse events must be conducted regardless the design of the study.

burn-in period: : A pre-specified initial phase where patients are enrolled and assigned to treatments using a fixed allocation ratio (e.g., 1:1) before any adaptive analyses or changes to the trial design occur.

: The burn-in period aims to collect sufficient data to allow for reasonable precision and to minimize the risk of making inappropriate adaptations based on early, potentially noisy data.

Advantages and Limitations of Adaptive Trial Designs#

Advantages#

  • Allow the trial to adjust to information that was not available at the design stage

  • Statistical efficiency; Improve statistical power to detect true treatment effect

  • Ethical advantages; the treatment can be halted early if efficacy is not established - minimize the number of subjects exposed to the less effective treatment

Limitations#

  • Increased flexibility in protocol design increases the statistical complexity

  • Not possible to derive closed-form expressions - trial simulation is required for evaluation of statistical properties (ex. Expected sample size, type I error rate, statistical power)

  • More time and effort is required to establish statistical planning

  • High-quality data is required for interim analysis - put burden on sites participating the trial since additional training and education is essential to meet the demands for an adaptive clinical trial

Principles of Adaptive Trial Designs#

A Priori Evaluation of Statistical Properties#

  • Regulatory agencies, ethics board, funding agencies require a priori evaluation of statistical properties (ex. Demonstration of expected type I error rate control)

  • Complex trials involve several goals - therefore often be more multi-faceted

  • Various features of adaptive trials (ex. Considering sample size, type I error rate, power, other performance characteristics, estimated treatment effects, number of patients assigned to each group) will be evaluated using trial simulations

  • Pre-specified plans should be clearly outlines in study documents (ex. Protocol, DSMB charter, SAP) - no ad hoc changes

Simulation-Guided Trial Planning#

  • Pros and cons can be determined by clinial trial simulation prior to selecting a specific design option or beginning the trial by enrolling subjects

  • Clinical designs doesn’t work like “one size fits all”

  • All designs have advantages and disadvantages - it is important using simulation-guided design process to consider multiple design candidates during the planning phase to choose the design would work most effectively

Proper Operational Oversight and Management#

  • High quality data that are as clean as possible must be entered by sites in timely manner to ensure interim analyses can be conducted properly - knowing the data can have impact on stake holders (ex. Sponsor, investigators, site staff, etc) therefore bias the study operations

  • Access to interim data is limited to minimize the operational bias and to maintain scientific credibility