Packages/Model audit
Model takeover and audit

Find out if an inherited model will hold up

Scope this package
Delivery
Up to 6 weeks from a clean dataset
Price
Fixed, quoted after the call
Typical model risk
Rated in the audit

What is at stake

Models change hands: a CRO switch, a licensing deal, a team that moved on. Before an inherited model supports a decision or a filing, someone has to check that it reproduces and that it fits the new question.

Question of interest

Can we rely on the model we inherited for the decision we now need to make?

The audit reproduces the inherited model from its delivered code and data, assesses it against the M15 framework for its intended context of use, and identifies what is needed to make the evidence reviewable.

What reviewers will ask

Does the model reproduce from the delivered files?

We re-run it in a clean environment and compare estimates against pre-set tolerances.

Was the dataset built correctly?

Where source data exist, we re-derive it independently and compare record by record.

Is it fit for the new context of use?

The assessment is written for the decision you face now, not the one the model was built for.

What you receive

  • Gap report with an M15 assessment table for the intended use
  • Remediation plan with a fixed price per item
  • Reproduction record for the model and the dataset

What we need from you

  • The model files, code, analysis datasets and report you received
  • Source data, if available, for dataset re-derivation
  • The decision or submission the model now needs to support
Reported versus reproduced parameter estimates for an inherited model with a five percent tolerance band; one covariate effect does not reproduce.
Reproduction check.
Scientific approach

Reproduce. Re-run the final model in a pinned environment and compare estimates, standard errors and objective function value against the report, with pre-set tolerances.

Re-derive. Rebuild the analysis dataset from source data independently and compare record by record.

Assess. Evaluate structure, covariate selection, diagnostics and uncertainty against current practice and the intended context of use.

Report. A gap report that separates what blocks reliance from what is good practice, so remediation is proportionate.

Relevant guidance: ICH M15; FDA and EMA population PK guidance.

Is this the decision in front of you?

Fixed price, paid in two parts: at the start and when you receive the gap report. A guaranteed date. Pricing · How we work

Scope this package