loading page

Assessment of the nlmixr R-package for population pharmacokinetic modeling: A metformin case study
  • +3
  • Wen Yao Mak,
  • Qing Xi Ooi,
  • Cintia Cruz,
  • Irene Looi,
  • Kah Hay Yuen,
  • Joseph Standing
Wen Yao Mak
Hospital Pulau Pinang

Corresponding Author:[email protected]

Author Profile
Qing Xi Ooi
Pharmetheus AB
Author Profile
Cintia Cruz
Mahidol University
Author Profile
Irene Looi
Hospital Seberang Jaya
Author Profile
Kah Hay Yuen
Universiti Sains Malaysia
Author Profile
Joseph Standing
University College London
Author Profile


Aim: nlmixr offers first-order conditional estimation with or without interaction (FOCE or FOCEi) and stochastic approximation estimation-maximisation (SAEM) to fit nonlinear mixed-effect models (NLMEM). We modelled metformin’s population pharmacokinetics with flip-flop characteristics within nlmixr framework and investigated SAEM and FOCEi’s performance with respect to bias, precision, and robustness. Method: Compartmental pharmacokinetic models were fitted. The final model was determined based on the lowest objective function value and visual inspection of goodness-of-fit plots. To examine flip-flop pharmacokinetics, k_a values of a typical concentration-time profile based on the final model were perturbed and changes in the steepness of the terminal elimination phase were evaluated. The bias and precision of parameter estimates were compared between SAEM and FOCEi using stochastic simulations and estimations. For robustness, parameters were re-estimated as the initial estimates were perturbed 100-times and resultant changes evaluated. Results: A one-compartment model with transit compartment for absorption best described the data. At low n, Stirling’s approximation of n! over-approximated plasma concentration unlike the log-gamma function. Flip-flop pharmacokinetics were evident as the steepness of the terminal elimination phase changed with k_a. Mean rRMSE for fixed-effect parameters was 0.932. When initial estimates were perturbed, FOCEi estimates of k_a and food effect on k_a appeared bimodal and were upward biased. Discussion: nlmixr is reliable for NLMEM even if flip-flop is present but caution should be exercised when using Stirling’s approximation for n! in the transit compartment model. SAEM was marginally superior to FOCEi in bias and precision, but SAEM was superior against initial estimate perturbations.
26 Oct 2021Submitted to British Journal of Clinical Pharmacology
26 Oct 2021Submission Checks Completed
26 Oct 2021Assigned to Editor
07 Nov 2021Reviewer(s) Assigned
18 Jan 2022Review(s) Completed, Editorial Evaluation Pending
24 Jan 2022Editorial Decision: Revise Major
10 Apr 20221st Revision Received
11 Apr 2022Submission Checks Completed
11 Apr 2022Assigned to Editor
11 Apr 2022Review(s) Completed, Editorial Evaluation Pending
25 Apr 2022Reviewer(s) Assigned
26 Jul 2022Editorial Decision: Revise Minor
27 Jul 20222nd Revision Received
28 Jul 2022Submission Checks Completed
28 Jul 2022Assigned to Editor
28 Jul 2022Review(s) Completed, Editorial Evaluation Pending
08 Aug 2022Editorial Decision: Accept
30 Aug 2022Published in British Journal of Clinical Pharmacology. 10.1111/bcp.15496