Drug-induced liver injury as a strong independent predictor of in-hospital mortality in tuberculous meningitis: potential age-related effect modification suggested in a large lifespan cohort.

Abstract

Drug-induced liver injury (DILI) is a common complication during anti-tuberculosis treatment for tuberculous meningitis (TBM), but its association with in-hospital mortality and age-specific effect modification remains unclear. We conducted a single-center retrospective cohort of non-HIV TBM patients hospitalized between October 2013 and January 2024. DILI was ascertained using a structured retrospective adjudication workflow incorporating biochemical thresholds, temporal relationship with anti-tuberculosis therapy, exclusion of competing etiologies, and clinical evolution. The primary analysis used multivariable Firth penalized logistic regression adjusted for age, hydrocephalus, immunodeficiency, albumin, cerebrospinal fluid (CSF) glucose, and hyponatremia. Inverse probability of treatment weighting (IPTW), MRI-adjusted analysis, and sensitivity analyses were performed. Restricted cubic splines (RCS) assessed age-DILI interaction. Incremental predictive value was assessed by comparing models with and without DILI. Among 1,574 patients, in-hospital mortality was 11.9% (187/1,574), and DILI occurred in 13.8% (217/1,574). DILI was independently associated with in-hospital mortality in the primary Firth model (adjusted odds ratio [aOR] 8.536, 95% confidence interval [CI] 5.970-12.245; P < 0.001), with consistent IPTW (weighted OR 8.699, 95% CI 6.010-12.591; P < 0.001) and MRI-adjusted estimates (aOR 8.617, 95% CI 6.015-12.388; P < 0.001). Median time from anti-tuberculosis treatment initiation to DILI onset was 23 days (IQR 18-29). Among DILI non-survivors, DILI preceded death in all cases (84/84), and no first DILI episode occurred within 48 h before death. Adding DILI increased the area under the receiver operating characteristic curve (AUC) from 0.661 to 0.805 (ΔAUC = 0.144). RCS suggested borderline overall age-DILI interaction ( P = 0.069), without non-linear interaction ( P = 0.835). DILI is a strong independent predictor of in-hospital mortality in TBM, improves mortality risk prediction, and shows an exploratory borderline trend toward age-related effect modification. Timing and sensitivity analyses suggest that this association is unlikely to be explained solely by terminal liver injury or early-death bias. These findings support intensive liver-function monitoring, prompt etiologic adjudication, and early protocolized regimen adjustment in TBM patients with DILI.