Notes on model-informed drug development
New guidance, labels where a model carried a dose or replaced a study, and what they change for a development program. Every claim links to its source.
One dose, two Phase 3 trials, two verdicts
Bayer took one dose of asundexian into two Phase 3 trials. It lost to apixaban in AF. Last week it won FDA approval in stroke.
Read the article →
Half the exposure is not half the effect
A fatty breakfast before the dose halves exposure to Merck's oral PCSK9 drug. Once a week, it costs ~1 point of LDL-C lowering (model).
Read →
The first human dose, informed by a QSP model
FDA's June 2026 draft guidance describes how a quantitative systems pharmacology model can inform a MABEL-based first-in-human dose, and when a start above MABEL may be justified.
Read →ICH M15 is final, and every package now follows it
ICH M15 reached Step 4 in January 2026 and FDA published it as final guidance in June 2026. Every package we deliver now follows its framework.
Read →pmxstudio.ai is open for projects
The studio now takes on population PK, exposure-response, MIDD and submission-package work, each on a fixed price and a fixed timeline.
Read →Project Optimus and the case for early exposure-response
Project Optimus has moved dose optimization from a late-stage afterthought to an early-development expectation.
Read →How the work is split between agents and the modeler
A pharmacometric analysis is a sequence of distinct tasks, and each one rewards a different kind of automation.
Read →No articles on this topic yet.
Follow the notes on LinkedIn
Two short notes a week, each with its slides and its sources.