Population pharmacokinetic analysis
Tezacaftor and tezacaftor-M1
In table 2 the estimated PK parameters of the final models of tezacaftor and its M1-metabolite are shown. The most stable model was created with a vague prior on Vc tez, with moderate informative priors on Vp tez, Qtez, D1 and Qtez-M1, and with informative priors on KA, Vc tez-M1 and Vp tez-M1. No prior information was applied on CL and its IIV, as the data were rich enough to estimate these parameters. No correlations of the covariates age, CF mutation or adherence were observed on CL.
Ivacaftor, ivacaftor-M1 and ivacaftor-M6
In table 2 the estimated PK parameters of the final models of ivacaftor (-M1/6) are shown. Values for Viva-M1 and Viva-M6 were fixed at 0.1*Viva, as we were unable to estimate them and no popPK models have been described for the metabolites in literature. A factor of 0.1 was chosen because basic lipophilic drugs such as ivacaftor often have a large V (>100L), whereas their more polar and acidic metabolites have volumes closer to 10 of 20L. (11) For ivacaftor-M6 the proportional error (RSE) in plasma samples was larger than DBS samples with values of 0.98 (225%) and 0.50 (10%), respectively. The most stable model was created with a vague prior on Vc iva, with moderate informative priors on Vp iva and Qiva, and with informative priors on KA and D1. No prior information was applied on CL and its IIV, as the data were rich enough to estimate these parameters. No correlations of the covariates age, CF mutation or adherence were observed on CL.
Model evaluation
RSE values of estimated parameters were generally low both for the typical PK parameters (≤29%) and the random effects (IIV on CL ≤38%). GOF plots (appendix – figure A1-3) and VPC plots (figure 3) demonstrate that the developed models adequately describe the observations. The robustness of the models was evaluated by a bootstrap analysis; its results are presented in table 2.