Comparison of area under the curves for vancomycin from one- and two-compartment models using sparse data

Abstract

Vancomycin pharmacokinetics has been described by 1- and 2-compartment models. One-compartment models built from routine monitoring data, which were mainly trough samples, are commonly used to predict area under the curves (AUC), the useful indicator for vancomycin efficacy. The question stands whether AUCs from 1-compartment models with sparse data can sufficiently represent the true AUC. This study aimed to compare AUCs from 1- and 2-compartment models using sparse data. A previously published model was used to simulate full individual profiles for 100 patients. From these data, the reference AUC (AUCref) was calculated and two depleted datasets (trough-only and peak-trough) were also created. Both 1- and 2-compartment models were built from the depleted datasets using NONMEM®. AUC was calculated from concentration-time profiles of each model by linear trapezoidal method. Deviation of each AUC from the AUCref was examined from statistical and clinical perspectives. A two-compartment model from peak-trough data provided similar AUCs with the AUCref, but not that from trough-only data. The mean difference of AUCref and AUCs from the 2-compartment model with trough only data was up to 25.16% (p < 0.05) which were considered clinically significant. One-compartment models from both datasets could adequately estimate the AUCs with no significant differences (p > 0.05) from the AUCref. The mean differences were up to 4.38% and 6.23% for peak-trough and trough only data, respectively. Therefore, 1-compartment models from sparse data may be trustable to predict vancomycin AUC in clinical practice.