Drug exposure and minimum inhibitory concentration predict pulmonary tuberculosis treatment response.

BACKGROUND Prospective studies correlating pharmacokinetic/pharmacodynamic (PK/PD) indices to clinical responses are urgently needed. This study aimed to find clinically relevant PK/PD thresholds that can be used for treatment optimization. METHODS Pharmacokinetic sampling and minimum inhibitory concentration (MIC) measurements were performed for culture-confirmed tuberculosis patients. Classification and regression tree (CART) analysis was applied to obtain PK and/or PD thresholds for first-line drugs predictive of two-week/month culture conversion, treatment outcome determined at 6-8 months, acute kidney injury (AKI) and drug-induced liver injury (DILI). Least absolute shrinkage and selection operator (LASSO) logistic regression was used for model development and validation. RESULTS Finally, 168 and 52 patients with tuberculosis were included in development and validation cohort for analysis, respectively. Area under concentration-time curve (AUC)/MIC below CART-derived thresholds for pyrazinamide of 8.42, pyrazinamide of 2.79 or rifampicin of 435.45 were the predominant predictors of two-week culture conversion, two-month culture conversion or treatment success, respectively. Isoniazid AUC above 21.78 mg·h/L or rifampicin AUC above 82.01 mg·h/L were predictive of DILI or AKI during TB treatment. The predictive performance of trained LASSO models in validation cohort was evaluated by receiver operating characteristic curves and ranged from 0.625 to 0.978. CONCLUSIONS PK/PD indices and drug exposure of anti-TB drugs were associated with clinical outcome and adverse events. The effect of CART-derived thresholds for individualized dosing on treatment outcome should be studied in a randomized controlled trial.