Master Protocol Guide

When Shared Control Helps

• •Reduced total enrollment—a single control arm serves multiple experimental arms, avoiding duplication of control patients across separate trials
• •Increased power—more control patients accumulate over time, improving precision of treatment effect estimates for each arm
• •Operational efficiency—shared sites, shared data management, and a single DSMB reduce overhead

When Shared Control Is Risky

• •Temporal trends—if the standard of care changes during the trial, early and late control patients may differ systematically, biasing comparisons for arms that enter later
• •Non-concurrent bias—in platform trials, arms added after the original start share control patients who were enrolled before the new arm existed. This non-concurrent control can introduce bias if patient characteristics or care evolve
• •Multiplicity inflation—sharing a control arm creates correlation among test statistics, which can complicate FWER control

Platform Trial Control Strategies

The Zetyra platform trial calculator implements three strategies for handling non-concurrent control data (see Platform Trial docs for details):

• •concurrent_only—uses only control patients enrolled during the same period as the experimental arm (conservative, avoids temporal bias)
• •pooled_naive—pools all control patients across all periods without adjustment (maximizes power, but susceptible to temporal drift)
• •pooled_adjusted—for binary/continuous endpoints, includes non-concurrent control at 50% weight; for survival, excludes non-concurrent control entirely (balances power and bias)

When Borrowing Is Appropriate (Basket Trials)

Bayesian hierarchical models (BHM) and exchangeability-nonexchangeability (EXNEX) models borrow information across baskets under the assumption that treatment effects are related. This is powerful when:

• •Baskets share a common molecular mechanism of action (e.g., same genomic alteration targeted by the same drug)
• •Individual baskets have small sample sizes where standalone inference is imprecise
• •EXNEX provides a safety valve—baskets with outlier responses are automatically down-weighted via the non-exchangeable component

When Independence Is Safer (Umbrella & Platform Trials)

Umbrella and platform trials test different treatments, so borrowing across arms is generally inappropriate:

• •Each arm tests a distinct drug with its own mechanism—no reason to expect similar effect sizes
• •Borrowing between arms with different drugs could inflate type I error for ineffective treatments that happen to share an arm with an effective one
• •Each arm-vs-control comparison is self-contained and can proceed independently

Regulatory Perspective

Regulators generally accept information borrowing when the statistical model and borrowing rationale are pre-specified in the protocol and statistical analysis plan. The FDA 2022 guidance emphasizes that the degree of borrowing should be justified by biological plausibility, and that operating characteristics (type I error, power) should be evaluated via simulation under both homogeneous and heterogeneous scenarios.

FDA Guidance (2022)

The FDA’s Master Protocols: Efficient Clinical Trial Design Strategies to Expedite Development of Oncology Drugs and Biologics (2022) provides formal recommendations for all three design types:

• •Pre-specify all sub-studies, endpoints, analysis methods, and decision rules before enrollment begins
• •Simulate operating characteristics under multiple scenarios, including null, alternative, and mixed (some arms effective, some not)
• •For basket trials with borrowing, demonstrate type I error control under heterogeneous response patterns
• •For platform trials, address non-concurrent control bias and describe the chosen mitigation strategy

EMA Considerations

The EMA has not issued master-protocol-specific guidance but applies general adaptive design principles (CHMP/EWP/2459, 2007). Key EMA expectations include pre-specification of adaptations, independence of the DSMB, and careful justification of any information borrowing or shared control designs.

What Regulators Want to See

1. Woodcock J, LaVange LM (2017). Master protocols to study multiple therapies, multiple diseases, or both. New England Journal of Medicine, 377(1), 62–70.
2. FDA (2022). Master protocols: Efficient clinical trial design strategies to expedite development of oncology drugs and biologics. Guidance for industry.
3. Berry SM, Broglio KR, Groshen S, Berry DA (2013). Bayesian hierarchical modeling of patient subpopulations: Efficient designs of phase II oncology clinical trials. Clinical Trials, 10(5), 720–734.
4. Park JJH, Siden E, Zorber MJ, Heidebrink JL, Harari O, et al. (2019). Systematic review of basket trials, umbrella trials, and platform trials: a landscape analysis of master protocols. Trials, 20, 572.
5. Saville BR, Berry SM (2016). Efficiencies of platform clinical trials: A vision of the future. Clinical Trials, 13(3), 358–366.
6. Royston P, Parmar MKB, Qian W (2011). Novel designs for multi-arm clinical trials with survival outcomes, with an application in ovarian cancer. Statistics in Medicine, 30(2), 149–170.
7. Neuenschwander B, Wandel S, Roychoudhury S, Bailey S (2016). Robust exchangeability designs for early phase clinical trials with multiple strata. Pharmaceutical Statistics, 15(2), 123–134.
8. Cunanan KM, Iasonos A, Shen R, Begg CB, Gönen M (2017). An efficient basket trial design. Statistics in Medicine, 36(10), 1568–1579.