Choose and defend the Phase 3 dose

What is at stake

The Phase 3 dose is one of the most expensive decisions in development, and in oncology FDA now expects it to rest on more than the maximum tolerated dose. A weak justification can mean more dose-finding before the pivotal trial can start.

Question of interest

Which dose and schedule should go into the registrational trial?

Population PK provides individual exposures. Exposure-response models for efficacy and safety predict outcomes at each candidate dose across the target population. The models complement the randomized dose comparison that FDA's oncology dosage-optimization guidance generally expects; they do not replace it.

What reviewers will ask

Is the exposure-response relationship confounded by disease severity?

Sicker patients often clear drug faster, especially for monoclonal antibodies, so low exposure can look like low efficacy. We test this with pre-specified covariate adjustment and case-matching, and lean on the randomized dose comparison where it exists.

Why this exposure metric?

Cavg, Cmin and Cmax are compared on criteria fixed in the analysis plan, and the choice is argued pharmacologically, not only by fit.

Does the benefit-risk hold in patients with high exposure?

We simulate exposure by covariate subgroup at each dose and show where each subgroup sits on both curves.

What you file with the agency

Dose-justification section of the end-of-Phase-2 briefing document
Population PK and exposure-response reports for Module 5.3.3.5
M15 assessment table for the dose question
Analysis datasets, define.xml and code for agency review

What we need from you

PK, efficacy and safety data (SDTM, ADaM or raw, with specifications)
Protocols, SAPs and any previous PK or E-R reports
The development context: target product profile, candidate doses, meeting dates

Exposure-response curves for efficacy and safety for the fictional compound PMX-101, with observed quartile response rates and the exposure ranges of two candidate doses. — Exposure-response for efficacy and safety.

What it may replace

Where randomized dose data and exposure-response data are adequate, the analysis can reduce the size or number of additional dose-ranging cohorts. It does not replace the randomized comparison of doses that FDA's oncology dosage-optimization guidance generally expects.

Scientific approach

Exposure. Population PK with steady-state Cavg, Cmin and Cmax per patient, using the dose history actually received, including interruptions and reductions.

Response. Logistic or ordinal models for response, time-to-event models for progression or discontinuation, and Emax or linear models for continuous biomarkers. The exposure metric is chosen by pre-specified criteria.

Confounding. Exposure and disease severity are often correlated, for example through clearance. We test this with covariate-adjusted models and case-matching sensitivity analyses, and report what the data can and cannot separate.

Decision. Simulated outcome distributions at each candidate dose, with parameter uncertainty propagated from bootstrap or SIR.

Relevant guidance: FDA guidance on optimizing the dosage of oncology drugs (Project Optimus); FDA exposure-response guidance; ICH E4; ICH M15.

Is this the decision in front of you?

Fixed price, paid in two milestones. A guaranteed date. Agency questions on this analysis answered at no extra cost. Pricing · How we work

Scope this package