Predictive performance of single-trough Bayesian estimation for vancomycin across body mass index categories using Practical Antimicrobial Therapeutic Drug Monitoring (PAT) version 4.0.

Abstract

Background: This study evaluated whether body mass index (BMI) affects the predictive performance of single-trough Bayesian estimation for vancomycin (VCM) using Practical Antimicrobial Therapeutic Drug Monitoring (PAT) version 4.0.

Methods: This retrospective single-center study included adult patients who received VCM between April 2020 and March 2023 with at least two trough concentrations measured. Patients were categorized as underweight (BMI <18.5 kg/m²), normal-weight (18.5 ≤ BMI <25 kg/m²), or overweight (BMI ≥25 kg/m²). Using posterior parameters from the initial therapeutic drug monitoring (TDM) assessment, the second trough was predicted and compared with the observed value. Predictive performance was assessed by mean prediction error (ME), mean absolute error (MAE), and Spearman's rank correlation. Prediction deviation was defined as an observed-to-predicted ratio outside 0.8-1.2; associated factors were identified by logistic regression. Sensitivity analyses used continuous BMI, restriction to patients with stable renal function, and ΔSCr instead of ΔCLcr.

Results: Among 190 patients (45 underweight, 101 normal-weight, 44 overweight), BMI ranged from 14.2 to 49.3 kg/m², with 18 patients ≥30 kg/m² and 7 ≥35 kg/m². ME and MAE did not differ among BMI groups; correlation coefficients were 0.595, 0.617, and 0.711. Overall, 37.4% showed prediction deviation. Deviation was associated with variation in renal function, but not with BMI category, and this pattern persisted in sensitivity analyses.

Conclusions: In this cohort, BMI did not significantly affect the predictive performance of single-trough Bayesian estimation. Clinically relevant deviations occurred in approximately one-third of patients, indicating limitations of single-trough prediction, particularly when renal function is unstable.