Oncology sponsors are increasingly bringing two unapproved agents into the clinic together. When they do, they have to show FDA why codevelopment is justified, how each agent's contribution will be isolated, and how a combination dose will be selected. FDA's expectations are spread across several final and draft guidances rather than one document. Gaps in any of them can lead to clinical hold questions, a redesigned dose-escalation scheme, or a confirmatory trial that cannot show contribution of components.
The analysis below draws together the relevant FDA and ICH sources: the codevelopment guidance, the draft guidance on contribution of individual drugs in novel oncology combinations, the dosage optimization and expansion cohort guidances, and ICH S9 for nonclinical support. It sets out what each one asks a sponsor to address at the IND stage and in early-phase trial design.
FDA expectations for INDs combining two investigational oncology agents: contribution of components and dose finding
When a sponsor wants to take two unapproved agents into the clinic together for cancer, FDA's expectations come from several overlapping documents rather than one guidance. The foundation is the final guidance Codevelopment of Two or More New Investigational Drugs for Use in Combination. Sitting on top of it are:
- the 2025 draft Development of Cancer Drugs for Use in Novel Combination – Determining the Contribution of the Individual Drugs' Effects
- the final oncology dosage optimization guidance
- the 2022 expansion cohort guidance
- ICH S9 and its Q&A, for nonclinical support 124135802165.
This article summarizes what those documents ask a sponsor to show at the IND stage and in early-phase design.
1. Is codevelopment justified? The threshold criteria
The codevelopment guidance covers two or more drugs that have not been developed for any indication and are intended to be used together. It does not cover combinations of approved drugs, or a new drug combined with an approved drug 124125. FDA states plainly that codeveloping drugs generally produces less information about each drug's safety, effectiveness, and dose-response than developing them separately 126.
For that reason, FDA says codevelopment should ordinarily be reserved for programs that meet all of the following criteria 265:
| Criterion | What FDA expects |
|---|---|
| Serious disease | The combination is intended to treat a serious disease or condition 265 |
| Strong biological rationale | The disease or tumor biology makes the combination plausible. One example is targeting different steps in a cell-proliferation pathway, or reducing resistance 127265 |
| Evidence of advance | Nonclinical or short-term biomarker evidence suggests a significant advance over available therapy and superiority to the individual agents 26597 |
| Compelling reason not to develop separately | For example, monotherapy rapidly produces resistance or has very limited activity 265 |
FDA recommends consulting the agency before starting clinical development 265. For cancer combinations specifically, the 2025 draft says a novel combination should be considered for development when the necessity of each drug is supported by a strong biologic rationale, including nonclinical characterization of each drug and early clinical evidence. Relevant evidence includes nonclinical or short-term biomarker findings that a drug with limited expected monotherapy activity potentiates its partner 46.
2. Nonclinical package to support the combination IND
Biological rationale and contribution in models
Sponsors should support the rationale in a relevant model. An in vivo model is preferred, but an in vitro model is acceptable. The model should compare the combination with each individual drug and show substantial combination activity. It should also show greater activity, a more durable response, and/or a better toxicity profile than either drug alone 12867. Such models may also inform the relative doses of the two agents 67.
Toxicology: oncology differs from the general M3(R2) default
Outside oncology, ICH M3(R2) recommends a nonclinical combination toxicity study to support clinical trials of two early-stage entities. When it is warranted, the study generally uses one relevant species, and its duration matches the clinical trial up to 90 days 64.
For drugs intended to treat patients with advanced cancer, ICH S9 takes a different position:
- Each drug should first be well studied individually in toxicology, and the combination rationale should be provided before the clinical study 65.
- Dedicated combination toxicology studies are generally not warranted for advanced-cancer combinations, particularly when human toxicity profiles are already characterized 65.
- If at least one agent's human toxicity profile is not characterized, provide a combination pharmacology study. It should show increased activity without a substantial increase in toxicity on limited endpoints such as mortality, clinical signs, and body weight. The available information then determines whether dedicated combination toxicology is warranted 65.
The S9 Q&A adds several clarifications 66909192:
- "Well studied individually" means toxicology sufficient to support clinical study of each drug alone 6690.
- When human safety data for a component are absent or very limited, consider a combination pharmacology study in addition to single-agent toxicology 6690.
- For drugs that are inactive in animal species, relevant in vitro tests or an understanding of target biology can support the assessment 90.
- The rationale should, where feasible, include evidence of increased antitumor activity from pharmacology studies or the literature 91.
- If the available clinical and nonclinical data cannot establish a safe starting dose for the combination, a dedicated combination toxicology study may be needed 92.
When human safety data for the individual drugs are limited, the codevelopment guidance likewise notes that additional nonclinical data may be needed before human combination studies. In that case the nonclinical combination studies should support initial human dosing and include PK/ADME, toxicokinetic, and relevant biomarker or target-interaction data 6869.
3. IND structure for two investigational agents
The codevelopment guidance ties IND structure to the sponsor's development intent 188:
- Committed to codevelopment: By the time combination clinical studies begin, one IND should cover all drugs in the combination 189.
- Undecided: Preliminary proof-of-concept combination studies may run under one drug's IND, provided an individual-drug IND is in effect for each drug. If the sponsor then pursues the combination, it can use that IND as the combination IND or submit a new one 190.
- Also developing either drug as monotherapy: File a separate IND for the combination and for each individually developed drug 191.
The combination IND may cross-reference the individual-drug INDs. If an individual-drug IND belongs to another sponsor, written authorization to cross-reference is required 192. Where separate INDs exist, each must have its own annual report, and the reports should cross-reference one another 194. FDA encourages contact with the review division as early as possible, for example at a pre-IND meeting. Phase 3 design should be discussed at an end-of-phase-2 meeting 195196.
4. Phase 1: sequencing and dose finding for the combination
Characterize each agent first where possible
FDA's approach depends on the case. It does not require both drugs to complete development alone before they are combined. Still, whenever possible, each drug's safety should first be characterized in phase 1 as it would be for a single agent: maximum tolerated dose, dose-limiting toxicity, and pharmacokinetics. If even short-term monotherapy is unethical or infeasible, nonclinical combination data should support initial human dosing 156. Early human studies should characterize safety and PK for both the individual drugs and the combination, so that phase 2 dosing is supported 267.
Setting the combination starting dose and escalation
- Use the individual drugs' phase 1 safety data, when available, as the primary basis for the combination's starting dose, escalation intervals, and dose-response doses. If those data are not available, nonclinical combination data may be needed 157.
- If feasible, evaluate each drug's dose-response to pick the doses to explore in combination. If the drugs cannot be given alone, assess different doses of each in combination 158264.
- The safety assessment should account for prior human experience and for possible PK, PD, and toxicologic interactions, including shared target-organ toxicity and narrow safety margins 268.
Clinical pharmacology expectations
Each drug should be studied much as if it were being developed separately. That includes PK, bioavailability, mass balance, the effects of organ impairment, food, and drug interactions, and exposure-response. In vitro metabolism and transporter data should guide whether in vivo drug-drug interaction studies are needed 264. If an interaction is found and its mechanism (PK, PD, or overlapping toxicity) is unclear, FDA urges studies to understand it 287288.
Dosage optimization: moving beyond the MTD
FDA's final oncology dosage optimization guidance cautions that the MTD paradigm was developed for cytotoxic chemotherapy. For many targeted agents, a lower dose may give similar activity with fewer toxicities 80. Its recommendations carry over directly to a combination program:
- Compare multiple dosages in trials designed to assess activity, safety, and tolerability. A randomized, parallel dose-response design is preferred 81.
- Select dosages using more than DLTs. Relevant data include PD, low-grade symptomatic toxicities, dose modifications, tolerability, activity, and preliminary dose- and exposure-response relationships 219.
- Include a PK sampling plan in every protocol. In dose-finding trials, it should characterize PK after the first dose and at steady state at each dosage evaluated 218.
- When expansion cohorts compare dosages or schedules, randomize to two or more regimens and justify a sample size able to detect clinically important differences 222.
- For a new combination, consider existing dose- and exposure-response data. If the data do not support the proposed dosages, conduct additional dose finding, where quantitative modeling may help 285.
- Discuss dosage optimization plans with FDA at formal meetings, including early in clinical development 223. Plan to identify an optimized dosage before the marketing application 224.
FDA's master-protocol guidance gives a concrete expectation for novel-combination substudies. Identify an RP2D for each individual drug when each may have antitumor activity. Dose finding is appropriate if the combination's RP2D is not established. Before efficacy evaluation, safety data should be available from at least six subjects treated at the proposed combination dosage, and the dose-finding results should be submitted to the IND 290. That guidance also asks for a summary of safety, pharmacology, and preliminary efficacy for each investigational drug, plus the combination rationale and any evidence of synergy 291.
The 2025 contribution-of-effect draft does not address safety or dosing considerations for combination trials. It defers those topics to the codevelopment guidance 240289.
5. Expansion cohorts that test the combination
Under the 2022 expansion cohort guidance, a cohort testing an investigational drug with another investigational or approved drug should begin only after each investigational drug's preliminary single-agent safety profile has been characterized at its recommended phase 2 dose. The protocol should justify the combination scientifically and include a monitoring plan for overlapping or potentially synergistic toxicities 21.
The same guidance sets out IND-level expectations for first-in-human multiple expansion cohort trials:
- Cohort content: a scientific rationale, endpoints, eligibility, monitoring plan, and statistical justification of sample size for each cohort. Type I error should be addressed where cohorts are compared 22.
- Safety reporting: cumulative safety submissions more frequent than annually, at an interval agreed with FDA 2228.
- Cover letter: the original IND cover letter should clearly identify the trial as a first-in-human multiple expansion cohort trial 23.
- Independent oversight: an IDMC or another independent entity for all such protocols. Its role includes real-time review of serious adverse events and assessment of each cohort's safety and futility 2627.
6. Demonstrating contribution of each component
Codevelopment guidance: phase 2 design scenarios
Phase 2 should establish each drug's contribution where needed and possible, show that the combination is effective, and optimize dosing for phase 3. A factorial design is generally preferred 15998. The recommended design depends on how much each drug can do alone:
| Scenario | Recommended design |
|---|---|
| Both drugs active and can be given separately | Four-arm comparison: AB vs A vs B vs placebo or SOC. An adaptive variant can stop poorly performing monotherapy arms early 9816099 |
| One drug minimally active alone | That drug generally need not be studied alone beyond initial phase 1 safety work. A three-arm A vs AB vs SOC design can show contribution 161 |
| Neither drug can be given separately for an effectiveness assessment | Compare AB with SOC (or an SOC add-on design) for proof of concept 161266 |
2025 oncology draft on contribution of effect
The July 2025 draft defines "contribution of effect" as showing that each drug contributes to the overall treatment effect observed in the population 135136. It applies to the following 137138:
- combinations of two or more investigational drugs
- an investigational drug combined with a drug approved for another indication
- drugs approved for different indications
It excludes add-on trials against a drug approved for the same indication 137138. Key positions:
- A randomized factorial trial is highly recommended when feasible. The traditional design includes the combination arm, each monotherapy arm, and SOC if SOC is not one of the monotherapies 109139.
- Randomized evidence is preferred. Earlier randomized trials using response rate or another direct measure of anticancer effect are preferred over comparisons with external data 140108.
- Alternatives to full factorial designs may be appropriate when monotherapy activity is limited, or when preliminary data suggest a large combination effect on a well-characterized outcome. The trials used to isolate contribution must still be adequate and well controlled 129.
- External data have defined limits. Complete, high-quality patient-level data from a contemporaneous or earlier trial may supplement or replace single-agent arms. Published summary-level results generally serve only to generate hypotheses 131. Plans to use external data should be discussed with FDA in advance 131144.
- Nonclinical evidence helps but has limits. It can reduce uncertainty, but clinical information is generally more compelling because preclinical models do not necessarily predict clinical activity 133143.
The draft is nonbinding and not for implementation. It was issued as a draft in July 2025, so check whether a final version has since been published. It encourages consultation with the review division as early as possible, for example at a pre-IND meeting, and frequently during complex programs 135109.
Practical IND checklist
Drawing the documents together, an IND for an early-phase oncology trial combining two investigational agents should be ready to address the following:
- Justification: where neither drug has been developed for any indication (the codevelopment guidance's scope), why the codevelopment criteria are met, including serious disease, biological rationale, evidence of advance, and why the agents cannot be developed separately 265.
- Nonclinical package: single-agent toxicology for each drug, plus a combination pharmacology study or combination toxicology where human data for a component are lacking or a safe combination starting dose cannot otherwise be set 656692.
- IND structure: a single combination IND or separate INDs, with cross-reference authorizations in place 189191192.
- Phase 1 design: single-agent characterization where feasible. The combination starting dose and escalation should rest on individual-drug data, with interaction-focused safety monitoring 156157268.
- Dosage optimization: comparison of more than one dosage, PK at each dose level, and early discussion of the strategy with FDA 81218223.
- Expansion cohorts: combination cohorts only after single-agent RP2D safety is established. For first-in-human multiple expansion cohort trials, add independent safety oversight and an agreed cumulative safety reporting cadence 212628.
- Contribution-of-effect plan: a prospective strategy using a factorial design or a justified alternative, aligned with FDA before external data are relied on 98109131.
Questions worth taking further with Rhizome include how specific CDER oncology review divisions have applied these principles in recent approvals, and how the 2025 draft's contribution standards compare with EMA's approach to fixed and free combinations.