Hospitalization and diabetes mellitus as predictors of unfavorable tuberculosis treatment outcomes

BACKGROUND: Tuberculosis (TB) remains a global challenge, with 10.8 million cases and 1.25 million deaths annually. This study aimed to identify baseline characteristics of TB participants associated with treatment failure and death. METHODS: This retrospective cohort study enrolled adult participants with TB treated between 2015 and 2022 at a Brazilian referral center. Comprehensive baseline data (socioeconomics, demographics, clinicals, and laboratory tests) and monthly follow-up visits data were extracted from electronic medical records. Outcomes followed WHO 2021 criteria. Backward stepwise multivariable regression analysis was performed to identify predictors independently associated with failure and death. RESULTS: Among 485 participants (61.2% male, median age 38.0 years), 7.2% (N = 35) had unfavorable outcomes, comprising 1.8% death (N = 9) and 5.4% treatment failure (N = 26), the latter primarily driven by regimen changes due to adverse drug reactions (22/26 cases) rather than microbiological failure. Univariate analysis identified a higher proportion of unfavorable outcomes among people living with HIV (57.1% vs. 32.9%; p = 0.007), as well as in baseline hospitalization (48.6% vs. 14.0%; p < 0.001), disseminated TB (25.7% vs. 8.9%; p = 0.014), and alcohol (60.0% vs. 29.3%; p = 0.001) or tobacco consumption (48.6% vs. 21.8%; p = 0.002). Participants with unfavorable outcomes had lower hemoglobin (10.8 vs. 12.1 g/dL; p = 0.018), albumin (2.7 vs. 3.3 g/dL; p < 0.001), and lymphocyte count (1,111 vs. 1,561/µL; p = 0.003), as well as elevated C-reactive protein (9.54 vs. 4.93 mg/dL; p = 0.013), aspartate aminotransferase (37.0 vs. 24.0 U/L; p < 0.001) and gamma-glutamyl transferase (117.0 vs. 64.4 U/L; p = 0.001). Multivariate analysis identified baseline hospitalization (adjusted Odds Ratio [aOR] = 3.21, 95% confidence interval [CI] 1.42-7.26) and diabetes mellitus (DM) (aOR = 3.04, 95%CI 1.10-8.45) as independent predictors of unfavorable outcomes, whereas associations with alcohol use, and previous TB treatment were attenuated after adjustment, indicating possible confounding. CONCLUSION: DM and baseline hospitalization independently predict unfavorable TB outcomes, highlighting their importance for early risk stratification, while other factors, such as laboratory abnormalities, may primarily reflect secondary effects of disease severity captured by hospitalization. CLINICAL TRIAL: Not applicable.