Guide · Oncology

Project Optimus and RP2D selection: the new dose-optimization playbook.

FDA's Project Optimus has redrawn the rules for dose selection in oncology. This guide is for biotech sponsors deciding how to move from first-in-human data to a defensible recommended phase 2 dose (RP2D) under the agency's current expectations.

1. Why Project Optimus exists

Project Optimus is the FDA Oncology Center of Excellence initiative to reform dose optimization and dose selection across oncology drug development. It responds to a decade of approvals — particularly targeted therapies and immuno-oncology agents — where the marketed dose was later found to be higher than necessary, driving tolerability problems, dose reductions and discontinuations after launch.

The core message is simple: sponsors are now expected to justify the dose taken into registration on the basis of efficacy, safety, pharmacokinetics and pharmacodynamics — not tolerability alone.

2. From MTD to RP2D under Project Optimus

For decades, the maximum tolerated dose (MTD) served as the default RP2D in oncology. That heuristic assumed a monotonic dose–response relationship inherited from cytotoxic chemotherapy — more drug, more effect, until toxicity forces a ceiling.

Modern oncology assets rarely behave that way. Kinase inhibitors, antibody-drug conjugates, bispecifics and cell therapies frequently reach a plateau of efficacy well below the MTD, while toxicity continues to rise. Under Project Optimus, the recommended phase 2 dose must be the dose that best balances efficacy, safety and long-term tolerability — and the sponsor must show the work.

3. Randomized dose-optimization designs

FDA has been explicit that randomized comparisons of two or more active doses are the preferred route to RP2D selection. In practice this usually means a dedicated dose-optimization cohort — sometimes framed as a “phase 1b/2” — that enrols enough patients per arm to characterise the efficacy–safety trade-off, not just to demonstrate activity.

  • Parallel dose arms with pre-specified decision rules based on response, durability and tolerability.
  • Backfill enrolment in escalation cohorts to enrich early PK and PD data before the randomized comparison opens.
  • Adaptive elements — cohort expansion, arm dropping, sample-size re-estimation — used sparingly and with a pre-agreed statistical analysis plan.

4. PK, PD and exposure–response evidence

Project Optimus raises the bar for the translational package supporting RP2D. Sponsors are expected to bring an integrated exposure–response analysis covering both efficacy and safety, anchored in prospectively collected PK sampling and, where feasible, target-engagement or pharmacodynamic biomarkers.

Modelling and simulation are no longer optional. Population PK models, ER models and dose-schedule simulations should be developed alongside the clinical program and presented at key regulatory interactions — not assembled retrospectively to defend a chosen dose.

5. Regulatory interactions

Pre-IND and end-of-phase 1 meetings are the leverage points. Sponsors should arrive with an explicit dose-optimization strategy, candidate doses, the rationale for the randomized comparison, and a written plan for the PK/PD data package that will support RP2D selection.

Where a program is intended for both US and EU markets, parallel FDA/EMA scientific advice is worth the effort. EMA has generally aligned with the Project Optimus direction, but expectations on comparator selection, exposure–response framing and post-authorisation commitments can differ in ways that shape the pivotal design.

6. Common pitfalls

  • Selecting a single dose from an escalation cohort of six patients and treating it as the RP2D.
  • Under-powering the dose-optimization cohort so that neither arm produces an interpretable efficacy–safety comparison.
  • Deferring PK/PD modelling until the pre-NDA package, when the data required to build a credible ER model was never collected.
  • Assuming a Breakthrough Therapy or Fast Track designation substitutes for a rigorous dose-optimization strategy — it does not.