Basket Trial

A basket trial is a clinical trial design that tests a single investigational drug, or a fixed drug combination, across multiple diseases or disease subtypes that share a common molecular alteration or biomarker — rather than being organized around a single disease as a conventional trial is. The design is most developed in oncology, where it enrolls patients across different cancer types (different tissues of origin, or histologies) into one protocol whenever their tumors carry the specified genomic alteration the drug targets, treating the biomarker, not the cancer’s anatomical origin, as the organizing criterion for enrollment.

The U.S. Food and Drug Administration formally recognizes the basket trial as one of three related master protocol designs — alongside the umbrella trial and the platform trial — in its guidance Master Protocols: Efficient Clinical Trial Design Strategies To Expedite Development of Oncology Drugs and Biologics; Guidance for Industry, finalized and announced in the Federal Register on March 2, 2022. FDA defines a master protocol generally as a single protocol designed with multiple substudies, which may have different objectives, evaluating one or more investigational drugs across one or more diseases within a single overall study structure.

What makes a trial a basket trial

Per FDA’s guidance, a basket trial is designed to test a single investigational drug (or drug combination) in different populations defined by different cancers, disease stages, histologies, prior-therapy history, or genetic and other biomarkers. Three features distinguish the design in practice:

• One drug, many diseases. A single investigational product is under evaluation across the whole trial, not several.
• Biomarker-defined cohorts. Enrollment is organized by a shared molecular alteration or biomarker rather than by a shared disease — the same drug arm can include patients whose tumors originated in completely different organs.
• Single-arm, activity-estimating structure. Because a comparable control group across many different diseases is often impractical to assemble, basket trial substudies are typically single-arm, with overall response rate as the primary endpoint, generating early signal on whether the biomarker-matched treatment shows activity in each disease cohort before larger confirmatory studies are pursued.

Basket trial vs. umbrella trial

Basket and umbrella trials are frequently confused because both are master-protocol designs built around biomarker-matching, but the organizing axis is reversed:

• Basket trial: one drug, evaluated across multiple diseases that share a biomarker.
• Umbrella trial: multiple drugs, each matched to a different biomarker, evaluated within a single disease population.

An umbrella trial keeps the disease constant and varies the drug arm by the patient’s biomarker subtype (for example, several targeted-therapy arms all enrolling only non-small cell lung cancer patients, each arm defined by a different actionable mutation); a basket trial keeps the drug constant and varies the disease. The two are complementary designs answering different development questions, and some large precision-medicine programs combine elements of both.

Basket trial vs. platform trial

A platform trial is the third master-protocol design FDA’s guidance addresses: an ongoing, perpetual trial infrastructure in which treatment arms can be added or dropped over time as interim data accrue, often using response-adaptive randomization and a shared, persistent control arm. Where a basket trial is typically fixed at the outset — a defined set of biomarker-matched disease cohorts tested against one drug — a platform trial is built to run indefinitely, with new investigational agents entering and underperforming or completed arms exiting, all under one continuously operating protocol. A basket or umbrella trial can, in principle, be run using platform-style adaptive infrastructure, but “platform” describes the perpetual, arm-churning operational structure specifically, not the basket/umbrella enrollment logic.

A real example: NCI-MATCH

The NCI-MATCH (Molecular Analysis for Therapy Choice) trial, sponsored by the National Cancer Institute and launched in 2015, is a widely cited, well-documented basket trial in the peer-reviewed literature. It screened patients with treatment-refractory advanced solid tumors, lymphomas, or myeloma for actionable genomic alterations, then assigned patients — irrespective of tumor histology — to one of dozens of substudy treatment arms, each a Phase 2 evaluation of a therapy matched to a specific genomic alteration. Peer-reviewed reporting on the trial describes NCI-MATCH as distinguished from disease-specific protocols by its commitment to a basket trial strategy: grouping patients by shared molecular alteration rather than by cancer type or anatomic origin, a tumor-agnostic (or histology-agnostic) approach to treatment assignment.

Why the distinction matters for research administration

Correctly identifying a protocol as a basket, umbrella, or platform trial (or a hybrid of these) affects statistical planning, IRB review complexity, and regulatory strategy: master protocols typically involve substudy-level protocol amendments rather than whole-new-protocol submissions when a substudy is added, shared infrastructure (screening, safety monitoring, data management) across substudies, and endpoint and control-arm choices that differ meaningfully from a conventional single-disease randomized trial. Research administrators supporting a master-protocol trial should expect governance and amendment processes structured around the master protocol itself plus each substudy appendix, rather than a single monolithic protocol document.

For the broader design taxonomy this fits within — observational vs. interventional studies, RCTs, and where adaptive elements fit — see Clinical Study Design: The Major Types and How They Relate. For the practical mechanics of building out a trial’s endpoints, sample size, randomization, and statistical analysis plan, see Designing a Clinical Trial: Endpoints, Sample Size, Randomization, SAP.