Basket Trials and Umbrella Trials#

Presenter : HeeSeok Yoo#

Introduction#

  • Biomarker-guided clinical trials have become a cornerstone of precision medicine.

  • Basket and umbrella trials are types of master protocol trials using standardized screening and operating procedures.

  • These designs aim to improve efficiency in evaluating targeted interventions in biomarker-defined populations.

Basket Trials and Umbrella Trials#

Basket Trials#

  • Evaluate a targeted intervention across multiple diseases that share a common predictive biomarker.

  • Patients with different tumor histologies (e.g., lung, breast, colon) are grouped together if they harbor the same biomarker.

  • Can use single-arm or randomized designs depending on phase and objective.

Umbrella Trials#

  • Evaluate multiple targeted therapies in a single disease, with patients stratified into biomarker-defined subgroups.

  • Each subgroup receives a different intervention matched to its molecular profile.

  • May include randomization within arms and often involve a shared control arm.

Characteristics of Basket and Umbrella Trials#

Eligibility Criteria and Patient Grouping#

  • Basket Trials: Patients from different diseases are grouped by a shared predictive biomarker.

  • Umbrella Trials: Patients with the same disease are divided into subgroups by biomarker status.

  • Disease integration in basket trials assumes consistent drug activity across tumor types.

  • Biomarker-based stratification in umbrella trials requires biological plausibility for each subgroup.

Intervention Assignment and Control Group Choice#

  • Basket trials often assign a common intervention without a control group.

  • Umbrella trials typically assign different therapies per subgroup and more commonly include control arms (e.g., standard of care).

  • Use of shared control arms is encouraged for efficiency and comparability.

Comparison Against Other Biomarker-Guided Trials#

  • Basket and umbrella trials follow master protocols with predefined subgroups and standardized eligibility criteria.

  • Different from adaptive enrichment trials, which allow modification based on interim data.

  • Centralized screening and infrastructure improve implementation and coordination.

Key Design Considerations for Basket and Umbrella Trials#

Biological Plausibility#

  • Biological plausibility requires careful evaluations of existing clinical evidence and underlying biological assumptions.

  • Ensures that targeted interventions are appropriate for the biomarker-defined population.

  • Important to distinguish between driver mutations (causally linked to disease) and passenger mutations.

Accuracy of Biomarker Assays#

  • Misclassification due to false positives or false negatives reduces statistical power and may lead to incorrect conclusions.

  • False positive rates (FPRs) should be considered in trial planning, though often overlooked.

  • Particularly critical in exploratory (phase II) settings for correct candidate selection.

  • In basket trials, assay accuracy should be similar across tumor types.

Biospecimen Collection#

  • Standardized biospecimen collection is essential, especially in basket trials involving multiple histologies.

  • Sample quality, yield, and ease of acquisition should be consistent across tumor types.

  • Even in high-throughput centers, biopsy yield (e.g., 70% for lung adenocarcinoma) can be limited.

  • Liquid biopsy offers less invasive alternatives but must be validated and cannot replace histopathological examination.

Biomarker Prevalence#

  • Biomarker prevalence affects the size of the eligible patient pool and feasibility of recruitment.

  • Low prevalence may result in high screen failure rates and increased cost due to unnecessary testing.

  • Recruitment strategies must account for biomarker distribution to ensure adequate sample size within study duration.

Sample Size and Statistical Assumptions#

  • Sample size depends on trial phase and design (e.g., single-arm vs. randomized).

  • In basket trials, sample size may be calculated for the overall cohort using a single effect size.

  • This assumes a common predictive factor; heterogeneity in treatment effect can dilute efficacy signals.

  • In umbrella trials, sample size is calculated per subgroup; common control arms are recommended when feasible.

  • The predictive assumption must be valid—otherwise, an all-comers design may be more appropriate.

Randomization#

  • Predictive factors influence treatment response; prognostic factors affect outcomes regardless of treatment.

  • Without randomization, it is difficult to separate predictive from prognostic effects.

  • Randomization eliminates selection bias and balances measurable and unmeasurable variables.

  • Statistical adjustments may partially compensate when randomization isn’t feasible, but are limited by small sample sizes and only account for known factors.

  • Thus, randomization is strongly preferred for validity.

Conclusion#

  • Basket and umbrella trials are powerful tools for evaluating biomarker-guided therapies under a master protocol framework.

  • Their design requires careful attention to biological plausibility, assay accuracy, biomarker prevalence, biospecimen logistics, and statistical assumptions.

  • When well-executed, these trials enhance trial efficiency, reduce redundancy, and improve access to personalized treatment options.