Therapeutic drug monitoring of antibiotic agents: evaluation of predictive performance.

BACKGROUND

The precision of the population pharmacokinetic model used in therapeutic drug monitoring (TDM) is essential for successful dosage optimisation.

OBJECTIVE

To evaluate the predictive performance of pharmacokinetic models used in our hospital and to evaluate the possible impact of demographic characteristics or renal function on TDM accuracy.

METHODS

We compared a posteriori an adjusted concentration-time curve profile based on the first measured drug concentration with the second measured drug concentration. Linear regression models were used to compare predicted and observed drug serum concentrations, and to evaluate potential relationships between predictive performance and patients´ demographic/clinical features. Predictive performance of TDM was expressed using accuracy, precision, sensitivity and specificity.

RESULTS

One hundred and fifty-two patients were enrolled in the study. All pharmacokinetic models showed good predictive performance expressed by the coefficient of determination (r2) of 0.5642, 0.7263, 0.9001 and 0.9454 for continuous vancomycin, intermittent vancomycin, amikacin and gentamicin, respectively. Accuracy was 93.3%, 91.2%, 113.9% and 130.9% for continuous vancomycin, intermittent vancomycin, amikacin and gentamicin, respectively. Demographic characteristics or renal functions had no substantial impact on the accuracy of TDM.

CONCLUSION

We found the predictive performance of both aminoglycosides and vancomycin pharmacokinetic models to be satisfactory.