Bayesian estimation of pharmacokinetic parameters of vancomycin in patients with decreasing renal function

In clinical settings, decrease of renal function is frequently observed in patients treated with vancomycin (VCM). In this study, mutable covariates models (MCMs) were constructed to analyze the pharmacokinetics (PK) of VCM with Bayesian estimation, considering time-dependent decreases in creatinine clearance in 23 patients with decreasing renal function. The predicted mean percentage error (MPE) of VCM concentrations analyzed with a conventional fixed covariates model (FCM) was -19.1%, whereas the MPE was improved to 2.5% by applying MCM. Furthermore, a probable lag time between fluctuations in VCM clearance (CL(VCM) ) and serum creatinine (S(cr) ) was analyzed by MCM, MCM(Lag1d) , and MCM(Lag2d) , which considered lag times of 0, 1, and 2 days, respectively. Compared with FCM, all MCMs improved fitness with the significantly decreased root mean square percentage error (RMSPE) and MPE. However, RMSPE and MPE analyzed with MCM were not significantly different from those with MCM(Lag1d) and MCM(Lag2d) , indicating that lag times between alterations in CL(VCM) and S(cr) were obscure in these patients. Collectively, these results suggest that PK parameters of VCM were more accurately calculated by MCMs than by conventional FCM, and that VCM dosages calculated by FCM would be overestimated by approximately 20% in patients with decreasing renal function.