Dose selection in oncology development has become a focal point for both FDA and EMA, with each agency signaling that the historical default of maximum tolerated dose is no longer sufficient justification for late-stage and registration doses. For regulatory and clinical development teams, understanding precisely where these two agencies agree, where they diverge, and how each has exercised its expectations in practice is now a prerequisite for designing phase 1 programs that will withstand pre-IND, scientific advice, and review-stage scrutiny on both sides of the Atlantic.
The analysis below examines the structural and procedural differences between FDA's Project Optimus framework and EMA's approach as expressed through its oncology evaluation guideline, compares how each agency treats key concepts such as MTD, DLT, RP2D, and optimal biologically active dose, and reviews documented agency responses and precedents that illustrate how these expectations have translated into real development and review decisions.
Want Rhizome's help on your own question? Try it for free.
FDA Project Optimus vs EMA: how the two agencies diverge on phase 1 oncology dose optimization
For most of oncology's history, the phase 1 objective was singular: escalate to the maximum tolerated dose (MTD) and carry the highest safely administered dose forward. That paradigm was built around cytotoxic chemotherapy, where more drug generally meant more tumor kill. It fits poorly with modern targeted agents, antibody-drug conjugates (ADCs), and immuno-oncology drugs, where a dose well below the MTD can deliver similar activity with materially less toxicity 3233. FDA and EMA have both concluded that dose selection must change. Where they differ is in how prescriptively, how early, and through what mechanism they enforce it.
The short version: FDA has turned dose optimization into a named, structured, pre-approval expectation through Project Optimus and a dedicated finalized guidance, and it has begun making that expectation bite at the review stage. EMA reaches a substantively similar scientific position, but it does so through successive revisions of its anticancer evaluation guideline rather than a standalone program, and it retains MTD/DLT/RP2D as core phase 1 concepts while layering "optimal biologically active dose" thinking on top for targeted agents.
What FDA's Project Optimus actually asks for
Project Optimus is the FDA Oncology Center of Excellence initiative that reworks how oncology doses are chosen. Its central demand is to move away from the MTD default and toward the dose that best balances efficacy and tolerability for the intended population, supported by pharmacokinetic (PK), pharmacodynamic (PD), and pharmacogenomic data, and discussed with FDA early in development 65323366. The position is codified in FDA's guidance "Optimizing the Dosing of Human Prescription Drugs and Biological Products for the Treatment of Oncologic Diseases," issued in draft in January 2023 and finalized in August 2024 3233.
The operative expectations for a phase 1 (and early phase 2) program are:
- Compare multiple dosages in patients, not just one. FDA's guidance says more than one dosage should be evaluated in clinical trials to characterize activity, safety, and tolerability and to reduce uncertainty in selecting the dose 3233. The Friends of Cancer Research position that helped shape the debate goes further, arguing the pre-registrational dose-finding study should ideally be randomized and compare at least two doses to confirm the dose taken into the registrational trial 65.
- Abandon MTD as the automatic answer. FDA states plainly that the MTD approach is rooted in older cytotoxic drugs and is a poor fit for modern targeted therapies, because a lower-than-MTD dose may match efficacy with fewer toxicities 3233.
- Optimize before the registrational trial, not after. FDA encourages sponsors to plan for dose optimization early, including early formal meetings, and stresses that carrying an ill-optimized dose into a pivotal trial can undermine the ability to demonstrate the drug's true benefit 3365. Deferring the question to a post-marketing requirement (PMR) is discouraged because those studies are lengthy and difficult to run once a product is already on the market 66.
- Use PK/PD and exposure-response, anchored to ICH E4. FDA frames the guidance as complementing ICH E4, "Dose-Response Information to Support Drug Registration," the shared international dose-response framework 3233.
The initiative has not been frictionless. The National Cancer Institute publicly objected that FDA appeared to be asking for additional dose-level studies before clinical effectiveness and tolerability had even been established, arguing that dose optimization should be handled in phase 2 only after clear evidence of efficacy and tolerability 66. That tension, front-loaded dose work versus evidence-gated dose work, is the live debate within the FDA framework itself.
What EMA expects, and how it says it
EMA has no program branded like Project Optimus. Its expectations live in the "Guideline on the (clinical) evaluation of anticancer medicinal products," revised repeatedly (Revision 4 in 2013, Revision 5 in 2018, Revision 6 published 2024-01-23, with the 2020 draft in between) 565560. Read across those revisions, EMA's scientific position converges closely with FDA's, but the emphasis and the default assumptions differ.
Key EMA positions:
- MTD/DLT/RP2D remain the baseline phase 1 objectives. EMA still says toxicity is an acceptable phase 1 endpoint and that the main objective is to identify dose-limiting toxicities (DLT), the MTD, and the recommended phase 2 dose (RP2D) 56. Earlier text described the RP2D as "usually one dose step below MTD" 60. This MTD-anchored framing is retained more explicitly than in FDA's current guidance.
- But MTD is expressly recognized as inadequate for targeted agents. The guideline states that "there are cases where dose escalation to MTD is not adequate in order to define the recommended dose," and that in those cases escalation should be based on PD and safety data in relevant animal models plus human PK/PD data, with mechanism-based PK/PD modelling 55.
- "Optimal biologically active dose" is the EMA analogue to dose optimization. For molecularly targeted agents, EMA says dose-finding should aim to determine an optimal biologically active dose, defined as the dose at which the optimal biological response on a predefined marker is achieved and giving more does not further improve outcomes, while minimizing the dose needed to reach that maximum PD effect so as to limit toxicity 4955. It prefers a combination of PK/PD endpoints and clinical response endpoints (e.g. objective response or progression-free survival) alongside safety to establish that dose 55.
- Exposure-response and population PK/PD are encouraged, not mandated as randomized comparisons. EMA encourages exposure-efficacy and exposure-safety analysis/modelling in randomized phase 2 trials to support phase 3 dose selection, and encourages population PK/PD studies and longitudinal PD modelling (e.g. tumor shrinkage as a continuous variable) 6253. Where no informative PD endpoint exists, EMA concedes that dose finding "essentially relies on toxicity and tolerability" 44.
- Within-patient dose escalation is permitted in low-toxicity settings to reduce the number of patients exposed to inactive doses, subject to DLT assessment 56.
The practical contrast: EMA expects a scientifically justified, PK/PD-informed dose and is receptive to biologically active dose concepts, but it does not, in guideline text, impose FDA's specific structural expectation of a randomized, multi-dose comparison completed before the registrational trial. Its lever is scientific advice and assessment judgment applied case by case, rather than a named program with an accompanying finalized guidance.
Head-to-head
| Dimension | FDA (Project Optimus) | EMA (anticancer evaluation guideline) |
|---|---|---|
| Vehicle | Named OCE initiative plus finalized guidance (draft 2023, final Aug 2024) 3233 | Successive revisions of the anticancer evaluation guideline (Rev 5 2018, Rev 6 2024) 5556 |
| MTD stance | MTD paradigm treated as outdated for targeted/IO agents; steer away from it 3233 | MTD/DLT/RP2D retained as baseline objectives, but MTD "not adequate" for many targeted agents 5655 |
| Core expectation | Evaluate multiple dosages, ideally randomized comparison of ≥2 doses 323365 | Determine "optimal biologically active dose" via PK/PD, biomarkers, exposure-response 4955 |
| Timing | Optimize before the registrational trial; avoid deferring to PMR 3366 | Refine dose/schedule through phase 2; exposure-response modelling to support phase 3 dose selection 62 |
| Data emphasis | PK, PD, pharmacogenomics; complements ICH E4 3233 | Non-clinical justification for first dose; PK/PD, population PK/PD, biomarkers; ICH E4 lineage 5662 |
| Enforcement style | Explicit, program-driven, increasingly applied at review | Case-by-case via scientific advice and assessment |
Both ultimately trace to the same international backbone, ICH E4 on dose-response information, which is why the scientific substance converges even though the regulatory packaging differs 3233.
Precedents: how each agency has responded in practice
Sotorasib (Lumakras), FDA, 2021 to 2025. This is the defining Project Optimus-era precedent. FDA accepted 960 mg once daily as effective and generally well tolerated and agreed it was a safe and effective dose, but the review record repeatedly states FDA did not consider 960 mg to be optimized 1920. In the exposure-response analysis, 960 mg was only numerically superior to 180, 360, and 720 mg for objective response rate, tumor size change, progression-free survival, and overall survival, with none of the differences statistically significant 1920. The PK showed saturable absorption and similar steady-state exposure from 180 mg up to 960 mg, and FDA judged the efficacy exposure-response analysis confounded by baseline disease status and therefore inconclusive for dose selection 192324. In the colorectal program, the 960 mg objective response rate (26%) was numerically higher than 240 mg (6%), yet the analyses did not identify a positive exposure-response relationship and could not differentiate the two doses given overlapping exposures, so acceptance of 960 mg rested primarily on clinical benefit-risk rather than a demonstrated optimal dose 252228. The dose question persisted into a later 2025 review cycle for the program 25. Sotorasib is the clearest illustration of FDA approving a dose while formally flagging it as un-optimized, exactly the outcome Project Optimus is designed to prevent going forward.
Imatinib (Gleevec), FDA, historical. The intellectual seed of the current policy is visible in FDA's own early review of imatinib, which noted that tyrosine kinase inhibitors are theoretically cytostatic rather than cytotoxic and that "the paradigm of dose escalation until the maximum tolerated dose may not be appropriate in these agents" 8. FDA articulated the core rationale for dose optimization two decades before it became a formal program.
Pivekimab sunirine, FDA, recent. More recent targeted/ADC review documents still frame early development around determining the MTD and RP2D through escalation and expansion, while exploring markedly different cohort dose levels (for example a selected cohort dose of 0.45 against 0.015 in the first escalation cohort), showing the transition from pure MTD selection toward wider dose exploration is still in progress even under the new expectations 4.
Pediatric dose optimization, FDA ODAC, June 2023. FDA took the dose-optimization question to its advisory apparatus: the Pediatric Oncology Subcommittee of the Oncologic Drugs Advisory Committee met on June 16, 2023 to discuss dosage optimization of new drug and biological products for pediatric cancer patients, describing dosage optimization as an integral aspect of oncology drug development 77.
EMA side. EMA's "response" is expressed through guideline evolution rather than a marquee case or program. Across Revisions 5 and 6 of the anticancer evaluation guideline, EMA moved from RP2D described as one step below MTD toward explicit optimal-biologically-active-dose language and encouragement of exposure-response modelling to support later-phase dose selection 60554962. EMA has convened oncology-specific discussions (for example on the challenges of bringing immuno-oncology products to market, including trial design and population selection) and channels dose questions through scientific advice, but the datasets reviewed do not show an EMA program named or structured like Project Optimus 41.
What this means for a phase 1 oncology program
A sponsor planning a first-in-human oncology study for a targeted, ADC, or immuno-oncology agent should now design the dose-escalation and dose-optimization strategy together rather than sequentially. For FDA, that increasingly means building in a randomized comparison of at least two candidate doses and generating the PK/PD, exposure-response, and tolerability evidence to justify the selected dose before the registrational trial, with the dose plan raised at early formal meetings 323365. For EMA, the emphasis falls on a scientifically reasoned optimal biologically active dose supported by PK/PD, biomarker, and exposure-response data, with MTD/DLT/RP2D still acknowledged as the starting scaffold and dose refinement expected through phase 2 555662. Because both frameworks descend from ICH E4, a well-constructed exposure-response and biologically-active-dose package will travel across both jurisdictions, but the FDA program is the one currently most likely to convert an un-optimized dose into a review problem, as the sotorasib record shows 1920.
Open questions worth pursuing further with primary sources include the exact scope of the finalized August 2024 FDA guidance, how EMA scientific advice has handled specific randomized dose-comparison requests, and whether ICH will produce a harmonized oncology dose-optimization standard that formally aligns the two agencies.