pkident: structural and practical identifiability for PK/PD models

pkident tells which parameters of a pharmacokinetic or pharmacodynamic ODE model, written in rxode2, the data can determine.

Workflow

1. Describe the model: pkpd_model() with the observed outputs, the doses (bolus_dose(), infusion_dose()) and the known quantities, or a reference_model().

2. Structure: structural_identifiability() decides, for continuous noise-free observation, which parameters and which combinations (such as CL/F) can be determined at all. Two independent methods must agree.

3. Design: practical_identifiability() gives the expected relative standard error of each parameter for a sampling design (experiment(), residual_error()); compare_designs() compares candidates.

4. Fit: classify_profiles() classifies the likelihood profiles of an nlmixr2 fit and links them to the structural result.

Every decision is local and says so; anything the methods cannot decide is reported as unresolved. as_table() turns any result into a data frame.

See also

Useful links:

• https://daikikumakura.github.io/pkident/

• https://github.com/DaikiKumakura/pkident

• Report bugs at https://github.com/DaikiKumakura/pkident/issues

Author

Maintainer: Daiki Kumakura 55353329+DaikiKumakura@users.noreply.github.com