A comparison between two strategies for monitoring hepatic function during antituberculous therapy

Abstract

Rationale: The optimum strategy for monitoring liver function during antituberculous therapy is unclear.

Objectives: To assess the value of the American Thoracic Society risk-factor approach for predicting drug-induced liver injury and to compare with a uniform policy of liver function testing in all patients at 2 weeks.

Methods: We conducted an observational study of adult patients undergoing therapy for active tuberculosis at a tertiary center. All patients had alanine transferase measurement at baseline and 2 weeks following commencement of therapy. Sensitivity, specificity, and positive and negative predictive values were used to assess strategies.

Measurements and main results: There were 288 patients included, and 21 (7.3%) developed drug-induced liver injury (57.1% "early" at 2 wk and 42.9% "late," after 2 wk). There were increased rates of individuals with HIV infection in the early drug-induced liver injury group compared with no drug-induced liver injury and late drug-induced liver injury groups (33% vs. 7.1% vs. 0%; P = 0.004). The American Thoracic Society algorithm had a sensitivity and specificity of 66.7 and 65.6%, respectively, for prediction of early and 22.2% and 63.7% for late drug-induced liver injury. The uniform monitoring policy had poor sensitivity but better specificity (22.2 and 82.1%) for prediction of late drug-induced liver injury.

Conclusions: In our urban, ethnically diverse population, a risk-factor approach is neither sensitive nor specific for prediction of drug-induced liver injury. A uniform policy of liver function testing at 2 weeks is useful for prompt identification of a subgroup who develop early drug-induced liver injury and may offer better specificity in ruling out late drug-induced liver injury.