BACKGROUND

Model-informed precision dosing (MIPD) is a promising tool used to ensure therapeutic antimicrobial concentrations. Model selection and sampling strategy might lead to different pharmacokinetic (PK) parameter estimates. Herein, we assess the predictive performance for cefepime PK in two models implemented within the InsightRX software using differing sampling approaches.

METHODS

Historic cefepime PK data and individual Bayesian estimates in predominantly critically ill patients, some of whom had extracorporeal support, served as the reference standard. Two population PK models (A; B) were evaluated using four sampling scenarios: (i) trough only, (ii) midpoint only, (iii) trough + midpoint, and (iv) peak + midpoint + trough. The median prediction error (MPE) and median absolute prediction error (MAPE) were calculated for clearance (CL) and volume of central compartment (V). Predicted categorical achievement of ≥70% time that the free drug concentration was greater than the minimum inhibitory concentration [fT>MIC] was compared.

RESULTS

MAPE and MPE for CL and V resulted in variability that was dependent on model and sampling strategy. Both models' overall MPE and MAPE for CL were <±20 and <30% for all tested scenarios, respectively, with a low MPE of -2.4% to 4.4% on CL for sampling scenario 4. For V, MPE and MAPE were >±20 and >30% for the majority of test scenarios across both models, respectively. When excluding patients with extracorporeal support, MPE/MAPE for V decreased to 3.7-4.8/23.3%-34.5% and -7.9-2.5/25.2%-29.6% for model A and B, respectively. Using each model and sampling scheme, only four patients had discordant predicted achievement of ≥70% fT>MIC.

CONCLUSIONS

These two population PK models and all sampling scenarios demonstrated acceptable prediction of cefepime PK parameters and pharmacodynamic exposures; therefore, they demonstrated suitability for utilizing MIPD for cefepime therapeutic drug monitoring.