Reimplementing a published population PK model
Rebuilding a published population PK model from its tables and supplement in your own software, and checking that it reproduces the published values, lets you ask new questions without the original data.
Definition
Reimplementing a published population PK (PopPK) model means rebuilding its structure, typical values, covariate effects and variability from the paper, its tables and its supplement (often NONMEM code) in software you control, for example R with rxode2 or Python. The original patient data are not needed.
Why do it
A reimplemented model can answer questions the paper did not: how exposure changes with another dosing interval, how it varies across a weight distribution, or how sensitive a conclusion is to a parameter reported with uncertainty.
How to check it
- Reproduce the typical values and the exposure summaries reported in the paper.
- Recompute key outputs with an independent implementation (a different solver or language) and compare.
- Record every inconsistency in the source, for example a table value that differs from its footnote, and test both versions as a sensitivity analysis.
What it does not show
Agreement with published values shows that the model was rebuilt correctly. It does not validate the model for predicting a new patient population or a new regimen. The mosunetuzumab reimplementation separates these two points, and the anti-PD-1 dosing analysis reimplements two models and checks them against an independent calculation.
Frequently asked questions
Do I need the original patient data to reimplement a PopPK model?
No. The structure, typical values, covariate effects and variability reported in the paper and supplement are enough to rebuild the model. The original data are needed only to validate or refit it.
If my model reproduces the published exposure, is it validated for new regimens?
No. It shows that the model was rebuilt correctly. Prediction for a different population or regimen needs separate evidence.