Retrospective population pharmacokinetic analysis of vancomycin serum concentrations using non-linear mixed effects modeling

Date of Completion

January 2000


Health Sciences, Pharmacology|Statistics|Health Sciences, Pharmacy




Introduction. Vancomycin is a glycopeptide antibiotic active against numerous gram-positive bacteria, including methicillin-resistant Staphylococcus aureus. The dosing regimens developed for vancomycin continue to result in high interpatient variability in serum drug concentrations. Variability in serum vancomycin concentrations is of concern for both efficacy and safety reasons. Therefore, it was desirable to describe the differences in vancomycin disposition, based on patient characteristics, to provide more accurate and precise predictions of observed serum concentrations. ^ Methods. Data pertaining to vancomycin were collected retrospectively from two tertiary clinical settings and analyzed using the population approach. A population pharmacokinetic (PPK) model was developed from these data. PPK model development included determination of a suitable compartmental model and subsequent incorporation of patient characteristics to describe differences in vancomycin disposition. A breakdown procedure was performed to ensure all model parameters were statistically significant at an α-level of at least 0.005. The final PPK model underwent several validation checks, including cross-validation, posterior predictive check and objective function profiling. ^ Results. Data were collected from 129 patients, encompassing a wide range of characteristics, including age (16–94 years), weight (41–208 kg), serum creatinine (0.4–6.8 mg/dL), estimated creatinine clearance (6–300 mL/min), and disease states (33 congestive heart failure, 27 cancer and 10 chronic renal failure patients). A two-compartmental pharmacokinetic model was best-fitted to these data. Estimated creatinine clearance, chronic renal failure, patient age, and congestive heart failure were associated with total systemic clearance. The medical conditions cancer and congestive heart failure were found to increase the central volume of distribution. Hispanic patients displayed a significantly larger peripheral volume of distribution. Additionally, significant inter-occasion variability was noted in the estimated peripheral volumes of distribution. ^ Conclusion. The population approach, involving non-linear mixed effects modeling, was applied to the sparse concentration-time data obtained from routine clinical care records. Differences in observed concentrations between individuals were partially explained through associated differences between those individuals. Individualized dosing schedules may, therefore, be based on these observable differences prior to initial administration of vancomycin. ^