3 Results

3.1 Overview of Studies

A total of 204 papers were initially retrieved from the databases. After screening according to the predetermined inclusion and exclusion criteria, 13 PopPK models (M1-M13) published between 2010 and 2022 were retained in this review18,23-34. The screening process of the study is shown in Figure 1. Table 1 summarizes the demographic information of patients in the studies. The median number of subjects in each study was 37 (range, 6 to 335) with 38.46 % of the studies having numbers more than 50. With the exception of three studies that also included healthy volunteers18,28,30, the other studies included only patients with different pathological states such as obesity, immune deficiency, hematological malignancies, and pulmonary fibrosis. Nine18,23,25-28,30,31,34 studies included adults, three29,32,33 included children, and one included24 both. Of the 11 studies with oral formulations of posaconazole, three18,23-26,33 were with oral suspensions, four27,28,31,32 with delayed-release tablets, and one29 was on both oral suspension and delayed-release tablets. The two30,34remaining studies were conducted on intravenous formulations in the obese population and in critically ill patients treated with extracorporeal membrane oxygenation.

3.2 Model Building and Evaluation

Table 2 summarizes the information about model building and evaluation. The median number of the plasma samples used for modeling was 226 (55 to 5756). About half of the studies used sparsely sampled data from clinical TDM, with the rest of the rich data obtained mostly from PK studies. NONMEM software was used in all studies for modeling except in one study that used Monolix27. The deviation, reliability, and accuracy of the models were internally evaluated by goodness-of-fit (GOF), Jackknife technique, visual predictive check (VPC), and normalized prediction distribution errors (NPDE) or bootstrap. Almost all models exhibited satisfactory predictive performance and robustness in internal validation. Few studies had simulated dosing regimens based on the model and had proposed recommended doses for different conditions. Detailed recommended programs and target definitions are shown in Table 2.

3.3 Structural Model

Table 3 summarizes the characteristics of the final model, such as the type of structural model used, estimated pharmacokinetic parameters, model variability, and excluded and retained covariates. The PK characteristics of studies comprising oral suspensions and tablets were well described by the one-compartment model, while the two studies involving intravenous administration30,34 were better suited to the two-compartment model. With reference to absorption, six18,23,25,27,29,33 models were described in terms of first-order absorption and two18,33 with a lag time characterizing the absorption delay18,33. Out of the five27-29,31,32 studies using delayed-release tablets, two28,31 studies were described with sequential zero first-order absorption. The mode of absorption for the remaining five24,26,30,32,34 studies was not mentioned. The absorption rate constant (ka) was estimated from 11 oral administration studies with a median(range) of 0.494 h−1 (0.0396-1.26 h−1), five of which23,25,27,29,31 fixed it to a certain value according to the published literature. With the exception of four24,26,29,32 studies not mentioned the elimination of posaconazole, the remaining studies was best described by first-order elimination kinetics. Clearance (CL) and volume of distribution (V) varied considerably in the different models, with a median (range) for clearance of 14.95 L/h (7.3–195 L/h). The median (range) of V in the one-compartment model was found to be 1100 L (186–5280 L). In the two30,34 studies adopting the two-compartment model, V for the central compartment (V1) and peripheral compartment (V2) were estimated to be in the range of 26.2-150 L and 96.2-396 L, respectively.
The median (range) of inter-individual variability (IIV) of CL and V (or V1) was found to be 37.9% (21.8-87.8%) and 29.9 % (15.6-52.4%) respectively. Only four18,26-28 studies reported the inter-occasion variability (IOV) of related PK parameters18,26-28. The proportional, additive, or combined residual error was applied to the final models. The median (range) of the most widely used proportional residual error (coefficient of variation, % CV) was found to be 14.8% (1.79–53.8%).

3.4 Covariates

The stepwise covariate model (SCM) building exercise with forward inclusion, and backward elimination was the most commonly used method for building covariate models. The statistical criteria used in each study were slightly different. Multiple factors that potentially influenced the exposure of posaconazole were tested during modeling, and covariates such as weight, sex, age, total protein, incidence of diarrhea, use of drugs such as PPI, phenytoin, rifampin, fosamprenavir, nutritional supplements, and chemotherapeutic agents were retained in the final model of different studies to account for changes in PK parameters such as CL, V, and F.
In our review, the incidence of diarrhea and the use of PPI were the most common covariates included in the final model of six18,23-25,29,33 and five18,24,25,29,33 studies, respectively, with a negative effect on the bioavailability of posaconazole. Body weight appeared as a final covariate in 31% of the studies and also negatively correlated with posaconazole exposure. In addition, each of the other covariates such as the sex, age, total protein, and use of phenytoin were found in only one study.
To characterize the manner and extent of influence of the covariates on the corresponding models, we performed simulations of steady-state 24-hour plasma concentrations at different covariate levels. Since no covariates were included for model M627, M1132 and M1334, and incomplete information was available for M728, no simulation was performed for these models. According to the type of formulation, the models were divided into two groups for simulation: (a) oral suspension, (b) tablet or intravenous infusion. Tablets and intravenous formulations were placed together because they have similar plasma exposure. The simulation results have been shown in Figure 2. For most of the models, the effect of different covariate levels on the steady-state plasma concentration of posaconazole was clearly observable. Nevertheless, the effects of age in M123, gamma-glutamyl transferase (GGT) in M224, and weight and chemotherapy in M325 on the exposure of posaconazole seemed to be inconspicuous.

3.5 Dose Simulation

The therapeutic target and model-based dosing regimen adjustments are shown in Table 2. The simulation endpoint concentration of the final model in most studies was set as the minimum concentration of 0.7 mg/L for prophylaxis and 1.0 mg/L for treatment. To intuitively compare the exposure levels and attainment of posaconazole, we simulated the steady-state plasma concentration-time profiles at different dosing regimens for each model except M728, because there was not enough information to reproduce the model, and the results are shown in Figure 3. In the adult population using oral suspensions, only M518 could achieve the target concentration of 0.7 mg/L for prophylaxis at a dose of 200 mg thrice daily. On increasing the dose of posaconazole to 300 mg thrice daily or 400 mg thrice daily, more models were able to achieve posaconazole exposure for the prophylaxis or treatment. Nevertheless, M426 failed to meet the target exposure at three simulated doses. The pediatric population receiving 200 mg of oral suspension thrice daily could already reach the target concentration. At doses of 200, 300, and 400 mg daily, all models using tablet and intravenous formulations achieved the target concentrations.