Predictions of the Pharmacokinetics in Burn Injury Patients using Regression Models - Case Study with Levofloxacin

Owing to its excellent safety, tolerability, pharmacokinetic and pharmacodynamic profile levofloxacin is widely used. Although pharmacokinetics of levofloxacin was somewhat more variable in burn injury patients, it appeared to be comparable to healthy subjects or other patients. Linear regression model was established for C_max or C_min vs. [AUC_tau, CL and V_d] of levofloxacin using individual values from burn injury patients. Appropriate regression lines for C_max or C_min were subjected to internal and external validation on the ability to predict CL, V_d and AUC_tau parameters. The mean absolute error (MAE) and root mean square error (RMSE) of the predictions were used to judge the appropriateness of either C_max or C_min models. C_max models developed for levofloxacin showed moderate to strong correlations with the various parameters such as CL, V_d and AUC_tau. The C_min models showed strong correlation for CL and AUC_tau but not for V_d where the correlation was weak. Internal validation using data from individual burn patients showed RMSE of 13.47-25.42% for various predictions. External validation that used mean data from healthy subjects showed RMSE of 13.86-27.13%. Despite the pharmacokinetic variability, linear regression models using either C_max or C_min were established for levofloxacin rendering predictions of several key pharmacokinetic parameters. Although there was limitation of C_min model for predicting V_d, both models may be used as a prospective tool for the prediction of levofloxacin pharmacokinetics in burn care patients.