211. Defining Optimal Sampling Times for Cefepime Therapeutic Drug Monitoring in Clinical Practice

Abstract

Background: Clinicians performing beta-lactam therapeutic drug monitoring (TDM) lack evidence on when levels should ideally be drawn after a dose. Herein, we define the optimal timing (i.e., optimal sampling) for cefepime using real-world TDM data to validate our approach. Methods: De-identified data from two centers performing routine cefepime TDM were extracted by InsightRX and served as an external validation cohort. Plasma cefepime was quantified using validated LC-MS/MS assays for TDM and dosing was protocolized at each site. CRRT and ECMO patients were included but other dialysis patients were not. Bias (MPE) and precision (RMSE) of a non-parametric prior were assessed. Multiple-model optimal (MM-opt) sampling strategies were estimated for the first 24 hours of treatment. To mirror clinical practice, one- and two-sample designs were evaluated. Dose and covariate values informed optimal sampling times. Bayesian PK exposures were compared using all samples, trough-only sampling, or using a single optimally timed sample. AUCs were calculated from the posteriors. For fT >MIC analysis, the MIC was fixed at 8 mg/L. We used Pmetrics 2.1.1 for R. Results: 116 patients (42% female; median age, CRCL, and weight: 62 years, 76 mL/min, and 80 kg, respectively) contributed 235 levels. The PK model demonstrated acceptable bias and precision (-6% MPE, 30.9 RMSE) as a prior for estimating exposures from the TDM data (Fig1). For a one-sample approach, the most common MM-opt sampling times varied (Fig2) but were often a mid-point or trough. In the two-sample approach, sample one was often a mid-point and sample two was often a trough (Fig3). First 24-hr AUC and fT>MIC did not significantly differ using all available samples for analysis vs. limiting sampling to a single optimized time point vs. limiting sampling to a trough-only approach (P >0.05 for all comparisons; Fig4). Conclusion: Optimal cefepime sampling times depended on dosing regimen, and renal disposition. When limited to a single sample, optimal sampling times for cefepime TDM were often midpoint/trough levels, but when two samples were obtained the optimal sampling times were often a mid-point followed by a trough. Estimation of PK and PK/PD exposures was not significantly worse when using a validated Bayesian prior and a trough-only sampling approach.