Build the best clinical strategy possible to get your products to patients faster

Designing a Phase 1 oncology trial to satisfy both FDA and EMA

Methodology

This briefing synthesizes the harmonized and region-specific guidance that governs early oncology development. I searched ICH guidelines (S9 nonclinical anticancer, E4 dose-response, E8(R1) general considerations, E6 GCP, E17 multi-regional trials), EMA scientific guidelines (first-in-human/early clinical trials, evaluation of anticancer medicinal products, dose finding), EudraLex Volume 10 (EU Clinical Trials Regulation 536/2014 application and safety-reporting rules), and EMA CHMP/SAWP committee records for the scientific-advice mechanism. Because FDA's Project Optimus and its first-in-human expansion-cohort guidance are not primary-source datasets here, I cross-referenced FDA's positions through RAPS Regulatory Focus and DIA Global Forum reporting, and I checked the current methodological consensus in PubMed (2020 to present). Scope is limited to Phase 1 (first-in-human dose escalation, dose optimization, and expansion cohorts) for anticancer drugs and biologics in patients with advanced cancer. Where the two agencies converge I say so; where FDA-specific expectations rest on secondary reporting rather than the agency's own text, that limitation is flagged.

The short answer: there is no separate "FDA protocol" and "EMA protocol." A single, well-justified, risk-based, ICH-compliant protocol will satisfy both, provided you (1) build the nonclinical and starting-dose rationale to ICH S9 and the EMA FIH guideline, (2) design for dose optimization rather than a single MTD (FDA Project Optimus), (3) use a modern model-based escalation method, and (4) get both agencies' input early, ideally in parallel. The differences are mostly procedural (IND vs CTA/CTIS) rather than scientific.

1. Engage both agencies early, and preferably in parallel

Early dialogue is where FDA-EMA alignment is actually built, before the protocol is locked.

• FDA encourages a pre-IND meeting, and for a first-in-human trial with multiple expansion cohorts FDA specifically says sponsors should request a pre-IND meeting to discuss the plan ⁵². FDA also wants dose-optimization plans raised early, even at the pre-IND stage, and has signaled it is very open to early feedback as data emerge ⁵⁸.
• EMA offers scientific advice / protocol assistance through CHMP with the Scientific Advice Working Party (SAWP) preparing the procedure and CHMP adopting the advice letter ²⁶². It is intended to shape clinical development strategy before the programme is too far advanced ²⁶², and EMA frames it as "early and enhanced scientific and regulatory support" ²⁷⁰²⁷¹.
• The FDA-EMA Parallel Scientific Advice (PSA) program (launched 2005) lets sponsors get concurrent feedback from both agencies on the same questions at the same time ¹⁷⁴⁶³. It is most valuable for innovative products, rare diseases, pediatric programs, and unmet need ¹⁷⁴⁶³. Practical tips from the reporting: frame questions around issues where the agencies do not already align, gather all data in advance, resolve internal company positions before the joint meeting, and mind request timing (July requests can be slowed by EMA August staffing) ¹⁷⁴⁶³. The benefit is simultaneous feedback, a clear explanation of any divergence, and avoidance of duplicated studies ^174175178.

Practical sequence: preclinical package to ICH S9 → PSA / pre-IND / EMA scientific advice on the integrated protocol and dose-optimization plan → IND (US) and CTA via CTIS (EU) → seamless FIH trial.

2. Nonclinical foundation and starting-dose justification

Both agencies require a scientifically justified, risk-based starting dose. This is the most technically prescriptive part of the package.

ICH S9 (anticancer, applies in patients with advanced cancer):

• For most anticancer products intended for advanced cancer, nonclinical studies of 3 months' duration are considered sufficient to support marketing; for initial clinical trials the nonclinical data must be sufficient to support the clinical dose and schedule and to identify potential toxicity ⁴⁷⁶⁵.
• The starting dose should be scientifically justified using all available nonclinical data (PK, PD, toxicity) and expected to be pharmacologically active yet reasonably safe ⁴⁷. A common small-molecule approach is 1/10 of the STD10 in rodents, or 1/6 of the HNSTD when a non-rodent is the most relevant species ⁶⁶.
• S9 is written for patients with advanced cancer and does not apply to healthy-volunteer studies; if healthy volunteers are used, ICH M3 applies instead ⁶⁷.

EMA first-in-human / early-phase guideline:

• Determine the NOAEL from nonclinical safety studies and convert the NOAEL exposure in the most relevant species to a human-equivalent exposure ¹⁵⁴¹⁵⁵.
• Also determine the MABEL from nonclinical pharmacology (including ex vivo / in vitro human tissue where feasible) to estimate the minimal anticipated biological effect level, PAD, and/or ATD ¹⁵⁴¹⁵⁵.
• For patients, target a starting dose expected to have minimal pharmacological effect while being reasonably safe; a higher starting dose needs an explicit rationale ¹⁷².

Alignment point: for a cytotoxic small molecule, an S9-based STD10/HNSTD starting dose is standard; for a molecularly targeted agent, immunomodulator, or biologic with potential for exaggerated pharmacology, add a MABEL-based estimate and use the lower of the estimates as the EMA guideline expects.

3. Study population

• ICH S9 and the EMA anticancer guidance frame early oncology trials in patients with advanced cancer, not healthy volunteers ⁶⁷¹⁵³.
• FDA (as reported) wants the population for rapid FIH expansion-cohort trials limited to serious diseases with no available curative therapy, as a risk-mitigation safeguard ⁵².
• ICH E8(R1): choose a population whose disease and molecular/genetic profile matches the scientific question, with eligibility criteria that reflect the objectives and are well documented ^194198201.

4. Trial architecture: an integrated, seamless FIH design

Both regions now expect early oncology trials to be run as integrated, "learn-as-you-go" protocols rather than rigid Phase 1/2/3 silos.

• EMA notes FIH/early-phase trials are increasingly run under integrated protocols combining parts (e.g., single ascending dose, multiple ascending dose, food effect). Emerging data must drive continuation of dosing, initiation of subsequent parts, and dose selection for later components; each part must be scientifically justified and the protocol must describe escalation, transitions, and decision criteria ^160161165.
• FDA (as reported) expects a single, seamless protocol that moves from dose escalation into expansion cohorts, potentially compressing the traditional phase sequence into one continuous trial ⁵²⁵³⁵⁴. Expansion cohorts gather Phase 2-like information: antitumor activity, safer doses in specific populations, alternate doses/schedules, combinations, and biomarker value ⁵². For antitumor-activity assessment in a non-randomized cohort, FDA accepts a Simon two-stage design or similar to limit exposure to ineffective treatment ⁵⁴. FDA also expects real-time infrastructure: streamlined logistics, real-time data review, and prompt sharing of interim results with investigators, IRBs, and regulators ⁵².
• ICH E17 applies if the trial is multi-regional: use a single protocol with prospective regulatory agreement, plan for intrinsic/extrinsic (ethnic) factors in dose selection and analysis, pre-specify any pooling strategy, and seek early scientific consultation across regions ^295296297.

5. Dose-escalation methodology

The current consensus, reflected in the literature and both agencies' direction of travel, is to abandon 3+3 as the default.

• Model-based (CRM) and model-assisted (BOIN) designs have better operating characteristics than 3+3: better MTD/RP2D selection, fewer patients at suboptimal doses, and faster escalation ^231233239252. BOIN is repeatedly highlighted as simple, transparent, and rigorous while performing comparably to CRM ^239250252.
• The EMA anticancer guidance accepts toxicity as a dose-finding endpoint in Phase 1 single-agent trials (defining DLTs and the dose to bring forward), expects only the smallest necessary number of patients per level, and encourages population PK/PD modelling ^70133134. It cautions that for molecularly targeted and immunomodulating agents, dose selection should not rely only on classical cycle-1 cytotoxic DLT logic; later and lower-grade toxicities over prolonged treatment matter, and if MTD is not definable, escalation can be guided by PD, safety, and human PK/PD modelling ¹⁴².

6. Dose optimization: FDA Project Optimus (and EMA's parallel view)

This is the single most important recent change and the area where an under-designed trial will fail at FDA.

• Project Optimus is FDA's initiative to move oncology away from the "more is better" / MTD mindset toward earlier, more rigorous dose finding that defines the therapeutic window, tolerability, and efficacy ⁵⁸. FDA advises planning dose optimization early (from pre-IND) ⁵⁸.
• The Friends of Cancer Research position reported alongside it: a pre-registrational dose-finding study should ideally be randomized, compare at least two doses, and confirm the dose for the registrational trial; use preclinical data to narrow the range, consider interim assessments and intra-patient escalation, and size the study for benefit-risk assessment ⁵⁸. The rationale: an unoptimized registrational dose can obscure a drug's true benefit, while a better-tolerated dose improves adherence and access ⁵⁸.
• FDA also wants patient-reported tolerability data (not just clinician-reported AEs) collected as "regulatory-grade" evidence to support dose finding ¹⁹³.
• EMA is aligned in substance: the recommended dose should rest on the totality of PK, PD, activity, and safety data with a documented rationale ⁷⁷; RP2D for targeted/immuno agents should reflect an integrated assessment of adverse reactions across the whole treatment period, not just cycle-1 DLTs ¹⁴²; and exposure-response modelling is encouraged to support later dose selection ¹⁴³¹⁴⁴. For non-cytotoxic compounds, toxicity may not be the right endpoint and biomarker/population-PK/PD approaches may be needed ⁷². (For CAR-T and similar, classical dose-finding is less applicable and the dose range must be justified from nonclinical, clinical, product, and disease-specific factors such as antigen expression and tumour load ⁸⁶.)

Design implication: build in a randomized, multi-dose optimization step (two or more doses) with integrated safety, PK, PD, biomarker, and preliminary-efficacy readouts, plus PROs. This satisfies Project Optimus and maps cleanly onto EMA's "totality of data" expectation.

7. Objectives and endpoints

• Primary: safety and tolerability (DLTs), characterization of the dose-toxicity and dose-exposure relationships, and identification of the dose(s) to carry forward ^70133134.
• PK: determine main PK parameters; study food effect for oral drugs; use population PK/PD modelling ⁷⁰.
• PD / biomarkers: EMA expects tumour samples to be integral unless otherwise justified, with adequate analytical validity for any biomarker used early ^133134138. Phase 0 microdose work can inform distribution, receptor binding, CNS penetration, and early activity ^133134137.
• Preliminary antitumor activity in expansion cohorts (Simon two-stage or similar for non-randomized activity assessment) ^54135136.
• ICH E4: characterize dose-response early using PK/PD to avoid carrying ineffective or excessive doses forward ²⁹³⁰⁵⁰.

8. Safety management, stopping rules, and reporting

Protocol-level safety design (EMA FIH guideline):

• Pre-specify and justify dose-escalation steps, maximum exposure, and the calculation methods in the protocol; escalation decisions must incorporate emerging clinical data and pre-defined criteria ¹⁵⁴¹⁵⁵.
• Stopping rules must be in the protocol ¹⁶¹¹⁷². A stopping criterion should include clinical exposure equivalent to the NOAEL exposure in the most sensitive nonclinical species (adjusted by safety factors, PK-based), plus rules reflecting animal toxicity and PD knowledge; safety events must trigger rapid communication and action before the next dose ¹⁶².

Reporting obligations differ procedurally by region but both follow ICH E2F:

• EU (CTR 536/2014): expedited SUSAR reporting to EudraVigilance/CTIS with the defined minimum data set, an annual safety report (DSUR, ICH E2F format) via CTIS, and Reference Safety Information maintained as a clearly separated section of the Investigator's Brochure for an unauthorized IMP ^{272274275279283286}.
• US: IND safety reporting and the IND annual report / DSUR (per ICH E6 GCP, the sponsor expedites SUSARs to authorities and informs investigators/IRBs) ¹⁰⁴¹⁰⁶. A single DSUR can generally serve both regions.

9. Quality, GCP, and operational readiness

• ICH E6 GCP governs sponsor and investigator responsibilities, protocol compliance, oversight, and a proportionate risk-based quality management system ^9193100101.
• ICH E8(R1) quality-by-design: start from clear objectives, identify critical-to-quality factors, keep the design operationally feasible and proportionate, avoid unnecessary procedures and data, engage stakeholders (including patients and regulators) early, and build in periodic review and adaptation ^{194195197198199213}. Early-phase designs may be more flexible or adaptive but must stay fit for purpose ^201204206.
• FDA's expansion-cohort expectations add an operational bar: infrastructure for streamlined logistics, real-time review of emerging data, and prompt interim sharing with investigators, IRBs, and regulators ⁵².

10. Regulatory submission mechanics (the main US/EU divergence)

The scientific dossier (nonclinical package, starting-dose rationale, protocol, IB) is essentially common; the packaging and procedure differ.

Bottom line and a design checklist

A single ICH-compliant protocol works for both regions if you:

• Get parallel or coordinated advice (FDA pre-IND + EMA SAWP, ideally via PSA) on the integrated protocol and dose-optimization plan ^5258174262.
• Justify the starting dose to ICH S9 (STD10/HNSTD) and add a MABEL estimate per the EMA FIH guideline for targeted/biologic agents; take the more conservative value ^4766154155.
• Enroll patients with advanced serious cancer without curative options ^5267153.
• Run an integrated/seamless design: dose escalation → randomized dose optimization → expansion cohort(s) ^5254160165.
• Use a model-based/model-assisted escalation (BOIN or CRM), not 3+3 ^231239252.
• Build a randomized ≥2-dose optimization step with integrated safety/PK/PD/biomarker/efficacy data and patient-reported tolerability to satisfy Project Optimus and EMA's totality-of-data standard ^5877142193.
• Put pre-specified stopping rules and escalation criteria in the protocol; run a common DSUR with region-specific SUSAR routing ^154162272279.
• Apply ICH E8(R1) quality-by-design and E6 GCP, and if multi-regional, ICH E17 with prospective cross-regional agreement ^{93194199295297}.
• File an IND (US) and a CTA via CTIS (EU, Part I + Part II, IMPD) from the shared dossier ¹⁶¹⁹²⁴.