An easy-to-use Nomogram for Predicting Antituberculosis Drug- Induced Hepatotoxicity in Hospitalized Patients

Abstract

Abstract Background & aims: Anti-tuberculosis (TB) drug-induced hepatotoxicity (ATDH) is one of the most deleterious side effects of anti-TB treatment. However, published data regarding prediction and prevalence of ATDH among TB patients were lacking. Methods: Hospitalized TB patients with retrospective data were extracted from the West China Hospital, Sichuan University. The predictors for ATDH were acquired by the least absolute shrinkage and selection operator (Lasso) and multivariate logistic regression methods were used to obtain the features related to ATDH. Results: 279 ATDH cases and 725 non-ATDH controls were enrolled in the training group, and 150 patients were included in the validation group, including 33 ATDH.. C-index for discriminating between those with and without ATDH was 0.656 (95% CI: 0.619, 0.693) in training set, and the independent validation with 1000 bootstrap replications was 0.742 (95% CI: 0.652, 0.831). The decision curve analysis indicated that the threshold probability of ATDH was 15–67%. Conclusion: Our nomogram was a valuable tool to identify patients at high risk of ATDH in hospitalized TB patients and could improve the management of ATDH.