Background
This study focuses on desipramine pharmacokinetics, aiming to develop a semi-mechanistic model with a liver compartment, analyze BQL methods, and assess the impact of covariates. Handling below limit of quantification (BQL) data in population pharmacokinetic analyses is crucial for accurate parameter estimation [1].
Methods
Data were obtained from seven phase 1 clinical drug-drug interaction studies (desipramine only control arm). BQL handling was conducted using four of Beal’s [2] recommended methods: M1 (discard BQL data), M3 (estimate BQL likelihood), M5 (set all BQL observations to the lower limit of quantification (LLOQ)/2), and M6 (set the first BQL observation to LLOQ/2 and discard subsequent BQL). Method performance and precision were evaluated using parameter estimation and visual predictive checks (VPCs).
Results
Data had 22% (n = 315) of the desipramine samples available as BQL. Notably, 14% non-quantifiable BQL were identified in the absorption phase, while 86% were non-quantifiable in the elimination phase. A semi-mechanistic model was employed to evaluate BQL methods, with M1 and M3 yielding similar estimates and improved VPCs compared to M5 and M6. M3 and M5 showed reduced bias during elimination compared to M1 and M6, while performance in the absorption phase was comparable across all methods. The M1 method performed similarly to M3 and was chosen for covariate testing. Influential covariates for desipramine pharmacokinetics were identified using stepwise covariate modelling (SCM), and included fat-free mass (FFM), formulation, race, and bilirubin. Three formulations were used (generic, Pertrofran® and Norpramin®). Subjects taking Pertofran® had a decreased desipramine area under the concentration curve to infinity (AUCinf), while those on Norpramin® experienced an increase. Non-Caucasian subjects exhibited significantly higher drug exposures.
Conclusion
We developed a semi-mechanistic pharmacokinetic model for desipramine, highlighting the importance of BQL data in parameter estimation. Omitting BQL (M1) was sufficient for the base model development (without covariates).
References
- Keizer RJ, Jansen RS, Rosing H, Thijssen B, Beijnen JH, Schellens JHM, et al. Incorporation of concentration data below the limit of quantification in population pharmacokinetic analyses. Pharmacol Res Perspect. 2015 Mar 1;3(2):1–15.
- Beal SL. Ways to fit a PK model with some data below the quantification limit. J Pharmacokinet Pharmacodyn. 2001;28(5):481–504.
