Prepare a population PK analysis report and model package

What belongs in a population PK analysis report?

Describe the decision question, source studies, data preparation, model structure and development, evaluation, simulations and limitations. Deliver the files and version relationships needed to reconstruct the reported outputs. A model can be fit for one purpose without supporting every exposure estimate, population or proposed dosing scenario.

Before you begin

Population PK analyses supporting development or regulatory decisions. Depth depends on the question, data and lifecycle stage; this is not a universal model specification.

What you will prepare: A structured analysis report and traceable supporting-file inventory suitable for qualified scientific review.

Write the decision question before the model history

Define what the analysis is intended to inform: variability, a covariate effect, an exposure estimate or a proposed dosing scenario. Gather the analysis plan, source-study inventory, data assembly specifications, model run record, software versions, diagnostics and simulations. Identify the population and exposure range actually represented.

FDA's February 2022 guidance recommends a structured report with an executive summary, introduction, methods, results, discussion, conclusions and applicable appendices. Start the summary with clinically interpretable findings and their limits. A small change in an objective function is not, by itself, a clinical dosing recommendation.

Preserve the path from source studies to the final model

Describe data selection, pooling, dose/time reconstruction, covariate definitions and handling of missing or below-quantification observations. Explain exclusions and changes from the plan. Show model structure, variability terms, covariate relationships, estimation methods and the important development decisions; retain enough supporting files to reproduce them.

Report parameter estimates and uncertainty together with relevant diagnostics and model evaluations. Examine performance in the populations that drive the intended decision, not only in the pooled dataset. FDA describes fit-for-purpose validation and warns that one method is not sufficient for every component; use complementary evidence appropriate to the question.

Separate modeled scenarios from observed evidence

For each simulation identify the scenario, population, regimen, model version and treatment of variability or parameter uncertainty. Explain whether the scenario interpolates within the evidence or extrapolates outside it. Relate proposed dosing conclusions to the relevant efficacy and safety exposure-response information.

Fictional exercise: the fitted dataset contains no participants with severe organ impairment, but a simulation is described as confirming unchanged exposure in that group. The reviewer should label the extrapolation, inspect assumptions and uncertainty, and request the missing justification. A successful model run is not direct clinical evidence in an unstudied population.

Deliver the files that produced the report

Place the population PK report in 5.3.3.5 and coordinate companion datasets, model/control files, programs and definitions using the applicable technical specifications. Identify the base, final and key intermediate runs. Link source-study participant identifiers consistently through assembled datasets and derived exposure metrics.

Before handoff, have an independent qualified analyst trace a key reported output to its data and code versions. Remove broken local paths and explain file dependencies in the reviewer material. Record any reproducibility limitation explicitly. No specific software is required by the reporting guidance, and this writing checklist does not establish model adequacy.

Build a reproduction map for the decision-driving output

Start with the output that carries the clinical argument and trace it backward. This editorial map helps the analyst identify missing dependencies before delivery.

Review exercise: a simulation was rerun after a covariate correction, but the report still shows the earlier figure. Trace the figure to its run, resolve the authoritative output and reassess the discussion. A successful rerun does not update an exported report automatically.

Use the intrinsic-factor guide to compare dedicated evidence where relevant, and the supporting-file guide to document delivered dependencies. A reader should be able to tell which evidence underlies each proposed use of the model.

Frequently asked questions

Sources and revisions

Requirements, source recommendations and editorial preparation advice have different roles. Review the scope and revision of the source you use.

Related guides