3.3 Population pharmacokinetic analysis
A one-compartment pharmacokinetic model with first-order elimination adequately describes the data. The inter-individual variability and residual variability were all characterized by exponential model.
The analysis of the covariate model confirmed the ALB and EM were the most significant covariates for the clearance (CL) and there were no covariates that had a significant effect on volume of distribution (V). The typical value of CL, V and oral bioavailability (F) of voriconazole obtained in the final model are 1.52 L/h, 35.7 L and 0.909, respectively. The oral absorption rate constant (KA) was fixed to a value of 1.19h-1.22, 23 The following formulations are included in the final model: CL (L/h) = 1.52×1.42(EM-1) × e0.911×(ALB/34.57)× e0.254; V (L) = 35.7; KA (h-1) = 1.19 (fixed); F = 0.909. The PPK parameter estimates of base model and final model are summarized in Table 3.
The GOFs from the final model displayed that the individual-predicted concentration (IPRED) and population-predicted concentration (PRDE) versus detection values (DV) of the final model are close to Y=X (figure 1A and B). Besides, the most conditional weighted residuals (CWRES) are symmetrical distribution on both sides of the line Y=0 and most of points are within ±2(figure 1C and D). The final model’s VPC results are presented in Figure 2. Most of the measured values of voriconazole are covered within 90% of the prediction interval (PI), and the predicted value is 5%-95%, demonstrating that the final model possesses good predictive capacity. The results of the final model NPDE statistical tests are as follows: t-test P = 0.77, Fisher test P = 0.343, Shapiro-Wilks test P = 0.117, comprehensive test P = 0.352. The statistical test results combined with the diagnostic chart reveal that the final model exhibited a normal distribution, with excellent stability and predictive performance (Figure 3). In the final model bootstrap verification, 1000 resampled bootstrap data sets were run successfully for 864 times. More importantly, the PPK parameter estimates of the final model showed similar results to the bootstrap results (Table 3), indicating the final model with good robustness and stability.