Expansion Cohort Trial Design in Oncology Trials

An expansion cohort is a group of additional subjects enrolled into an ongoing first-in-human (FIH) oncology trial, at a dose selected from the dose-escalation portion of that same trial, to generate the safety, pharmacokinetic, pharmacodynamic, and preliminary anti-tumor activity data that a small, sequentially-escalating dose-finding cohort cannot produce on its own. Because the expansion cohort is written into the trial’s original protocol rather than started as a new study, it lets a sponsor move from dose-finding into a larger, disease- or biomarker-defined population without closing out the FIH trial and opening a separate one — this is what “seamless Phase 1/2 design” means in practice, and it is the specific operational mechanism this guide covers.

What makes something an expansion cohort, operationally

Three features distinguish a true expansion cohort from an ordinary dose-escalation cohort:

• It follows dose selection, not dose-finding. Subjects are enrolled at a dose (or small number of doses) already identified as tolerable and pharmacologically active from the escalation portion — they are not part of the algorithm deciding what the next dose should be.
• It is written into the original protocol. The same Investigational New Drug (IND) application and protocol that governed dose-escalation is amended, not replaced, to open the expansion cohort(s) — this is the structural feature that makes the design “seamless” rather than two sequential, separately-initiated studies.
• Its purpose is broader than the escalation portion’s. Where dose-escalation exists to characterize the dose-toxicity relationship, an expansion cohort exists to gather a fuller safety database at a candidate dose, refine pharmacokinetic/pharmacodynamic characterization, and generate preliminary efficacy signal in a defined population — objectives that a 3-6 subject escalation cohort is not designed to answer.

FDA finalized guidance directly on this design: “Expansion Cohorts: Use in First-in-Human Clinical Trials to Expedite Development of Oncology Drugs and Biologics — Guidance for Industry,” announced in the Federal Register on March 2, 2022 (finalizing an August 2018 draft). It sets out what makes a drug candidate suitable for an expansion-cohort strategy, what an IND submission needs to contain before each new cohort opens, when a sponsor should engage FDA, and the safety safeguards expected for subjects enrolled once escalation has ended but before a full Phase 2 program exists.

How this differs from traditional Phase 1 dose-escalation

Traditional Phase 1 oncology dose-escalation exists to answer one question: how much of the drug can be given before toxicity becomes unacceptable. The two dominant designs for answering it are the rule-based 3+3 design (cohorts of 3 subjects, escalate on 0/3 dose-limiting toxicities (DLTs), expand to 6 on 1/3, stop or de-escalate on 2/3 or worse) and the model-based Continual Reassessment Method (CRM), which fits a pre-specified dose-toxicity model and re-estimates it after each patient or cohort. See CASRAI’s 3+3 Design vs. Continual Reassessment Method (CRM) guide for the full mechanics and the comparative accuracy evidence between them.

An expansion cohort begins where that process ends. Once escalation identifies a maximum tolerated dose (MTD) or a recommended dose for expansion, the trial does not simply stop — it enrolls further subjects at that dose to answer questions escalation was never designed to answer: is the safety profile consistent in a larger sample, does activity appear in a specific tumor type or biomarker-defined population, and is the pharmacokinetic/pharmacodynamic picture from a handful of escalation subjects representative of a broader population. The practical distinction for anyone reading a protocol: dose-escalation is adaptive (each cohort’s dose depends on the last cohort’s toxicity data), while an expansion cohort’s dose is fixed going in — the adaptive question has already been answered.

Seamless Phase 1/2 design: why this doesn’t require a new study

FDA’s own regulation governing drug development phases, 21 CFR 312.21, states directly that Phase 1, Phase 2, and Phase 3 “are not necessarily conducted in strict sequence” and may overlap — the phases describe the purpose and scale of an investigation, not a rigid administrative sequence of separate studies. Expansion-cohort trials are the clearest practical expression of that flexibility in oncology: the same protocol and IND that opened with a 3+3 or CRM dose-escalation portion is amended to add one or more expansion cohorts once a dose is selected, rather than closing the FIH study and filing a new Phase 2 protocol from scratch.

The practical advantage is time. A conventional path — complete Phase 1, write a clinical study report, design and submit a new Phase 2 protocol, wait for site activation — inserts months of administrative lag between a promising dose-escalation signal and the start of expansion enrollment. A seamless design keeps the same sites, the same IRB-approved consent infrastructure (amended rather than re-built), and the same regulatory file open throughout, which is precisely why FDA’s Oncology Center of Excellence has encouraged the approach for drugs addressing serious, unmet oncology need. It is not a shortcut around safety oversight — FDA’s 2022 expansion-cohort guidance requires the same IND safety-reporting and cohort-opening justification regardless of how quickly the trial moves.

Multiple parallel expansion cohorts: by indication and by biomarker

Most modern expansion-cohort trials do not run a single expansion arm — they open several in parallel, each testing the same dose in a different, pre-specified population. Two organizing principles are common, often combined within one trial:

• By indication (tumor type): parallel cohorts enroll subjects with different tumor histologies at the same dose, letting a sponsor screen for which disease setting shows the strongest early activity signal before committing to a single Phase 2/3 indication.
• By biomarker: parallel cohorts enroll subjects selected for a shared molecular alteration or biomarker status regardless of tumor histology (a tumor-agnostic approach), or stratify a single tumor type by biomarker subgroup.

A well-known biomarker-stratified example outside the expansion-cohort structure specifically, but built on the same “route patients to a matched arm from one master protocol” logic, is NCI-MATCH (Molecular Analysis for Therapy Choice), which screened patients with treatment-refractory solid tumors, lymphomas, and myeloma for actionable genomic alterations and assigned them to dozens of substudy arms matched by alteration rather than tumor histology.

Each parallel cohort in a genuine expansion-cohort trial is still governed by the single IND that opened the FIH study — cohorts are added through protocol amendments, each independently justified to FDA under the 2022 guidance, not spun up as separate trials. FDA’s guidance specifically expects a sponsor to justify, cohort by cohort, why the accumulated safety database supports opening it, since a drug can be simultaneously accumulating exposure across several parallel cohorts rather than one linear escalation path.

FDA Project Optimus: why dose selection for expansion is under new scrutiny

Expansion-cohort trials sit downstream of a dose decision, which makes the dose itself consequential — and FDA’s Project Optimus initiative, run by the Oncology Center of Excellence, was created specifically to reform how that dose gets chosen. Historically, oncology dose selection defaulted to the maximum tolerated dose (MTD) identified in escalation, on the assumption that more drug means more benefit. Project Optimus pushes sponsors away from an MTD-only default toward an optimal biological dose supported by pharmacokinetic/pharmacodynamic, safety, and preliminary efficacy evidence — commonly by evaluating more than one candidate dose (typically the MTD plus at least one lower dose) in a randomized comparison before selecting the dose that expansion cohorts, and eventual registration studies, will use.

FDA issued draft dose-optimization guidance in January 2023 and finalized it in August 2024, alongside a companion Oncology Dosing Tool Kit from the Oncology Center of Excellence. For expansion-cohort trial design specifically, the practical consequence is that the “dose selected from escalation” step this guide describes above is no longer simply “the highest tolerated dose” by default — sponsors are expected to show their expansion dose was chosen with efficacy and tolerability in view, not toxicity alone. See CASRAI’s dedicated guide, FDA Project Optimus: Oncology Dose Optimization, for the full guidance history and what it changes about dose-finding upstream of expansion.

Expansion cohorts vs. basket and umbrella trials

These three design types are frequently confused because all three enroll multiple, parallel, disease- or biomarker-defined patient groups under one governing document — but they answer different questions and sit in different places in a drug’s development timeline. FDA’s Master Protocols guidance (also finalized and announced in the Federal Register on March 2, 2022 — the same day as the expansion-cohort guidance, but a distinct document) formally defines the other two:

• Expansion cohort — one drug, one dose (already selected), multiple parallel populations, nested inside a single first-in-human trial’s IND as it moves from dose-finding into a larger safety/activity dataset. See CASRAI’s Expansion Cohort dictionary entry.
• Basket trial — one drug (or investigational regimen) tested across multiple diseases that share a common biomarker, independent of any dose-escalation relationship — the organizing question is “does this biomarker-matched drug work across tumor types,” not “what dose should we use.” NCI-MATCH is a commonly cited example. See CASRAI’s Basket Trial dictionary entry.
• Umbrella trial — one disease, multiple drugs, with subjects assigned to a biomarker-matched drug arm under a shared master screening protocol; Lung-MAP (SWOG S1400, NCT03851445), a master protocol in previously treated advanced non-small cell lung cancer, is a real, sponsor-independent example. See CASRAI’s Umbrella Trial dictionary entry.

A rule of thumb: if the organizing question is “what dose,” it is a dose-escalation/expansion-cohort structure. If the organizing question is “which biomarker-matched patients respond to this drug across diseases,” it is a basket trial. If the question is “which of several drugs works best for this one disease, matched by biomarker,” it is an umbrella trial. A single trial can combine elements — a dose-escalation portion feeding into biomarker-stratified expansion cohorts is functionally an expansion-cohort design with basket-like population stratification layered on top, rather than a basket trial in FDA’s formal sense, since dose selection (not biomarker-drug matching across diseases) is still the organizing event.

Frequently asked questions

Does an expansion cohort require a new IND?

No. A genuine expansion cohort is added by amending the same IND and protocol that governed dose-escalation, which is the structural feature that makes the design “seamless.” FDA’s 2022 expansion-cohort guidance sets out what that protocol amendment needs to justify before a new cohort opens — including the safety database supporting the dose and, cohort by cohort, why that population is appropriate to enroll next.

Can expansion cohorts support accelerated approval?

Single-arm expansion-cohort data has supported accelerated approval for some oncology drugs in serious, unmet-need settings, based on a surrogate or intermediate clinical endpoint reasonably likely to predict clinical benefit — but this is a case-by-case regulatory determination governed by FDA’s accelerated approval pathway generally, not an automatic consequence of using an expansion-cohort design. A sponsor pursuing this route should engage FDA specifically on endpoint and confirmatory-trial expectations well before expansion enrollment completes.

How is the dose for an expansion cohort chosen?

Historically, sponsors defaulted to the maximum tolerated dose (MTD) identified in escalation. Under FDA’s Project Optimus framework, sponsors are now expected to support the expansion dose with pharmacokinetic/pharmacodynamic, safety, and preliminary efficacy evidence — commonly by comparing the MTD against at least one lower dose — rather than carrying the MTD forward by default.

Is an expansion cohort the same as a Phase 2 trial?

Not formally. An expansion cohort is nested within the same first-in-human trial and IND that began with dose-escalation, and functions to bridge dose-finding into a larger dataset. A conventional Phase 2 trial is a separate study with its own protocol, typically initiated once a recommended dose is already established. In practice, a well-powered expansion cohort can generate data that substitutes for some of what a traditional Phase 2 would otherwise need to establish — which is the point of the seamless design — but the two are not interchangeable regulatory categories.

Related CASRAI resources

• 3+3 Design vs. Continual Reassessment Method (CRM) — the dose-escalation methodology an expansion cohort follows.
• FDA Project Optimus: Oncology Dose Optimization — how the dose an expansion cohort uses is now chosen.
• Expansion Cohort — the operational dictionary definition.
• Basket Trial and Umbrella Trial — the two master-protocol designs most often confused with expansion cohorts.
• Clinical Research hub — the full clinical-research cluster.