Hematochemical hallmarks as markers of pulmonary TB severity: A multicenter cross-sectional study

Abstract

Background: Identifying accessible and reliable biomarkers for tuberculosis (TB) severity is crucial for improving patient management. This study evaluates hematological findings as potential indicators of TB severity in a large multicenter Italian cohort.

Methods: This retrospective, multicenter, cross-sectional study analyzed hematological parameters (hemoglobin, white blood cells, inflammatory indices, hepatorenal function, albuminuria) in 577 TB patients from 10 Italian centers (2018-2023). Severe TB was defined by at least two criteria: TIMIKA score > 60, sputum conversion time > 21 days, or need for oxygen supplementation. Statistical analyses included receiver operating characteristic curve (AUC) evaluation, calibration curves, and clinical utility.

Results: Of the patients, 30.3 % were classified as severe, 60.2 % as non-severe, and 9.5 % as uncertain. AUC values for predicting severe TB ranged from 0.51 to 0.56 across hematological variables. Anemia and elevated CRP demonstrated sensitivities of 0.71 and 0.74, respectively. Models using continuous or categorical hematological variables achieved AUCs of 0.61 and 0.65, showing poor calibration and limited clinical utility in the 30-60 % threshold range.

Conclusions: Hematological markers, while rapid and cost-effective, demonstrated limited discriminative ability for TB severity. Further studies are required to develop reliable predictive models, integrating additional clinical and molecular data.

Keywords: Anemia; Disease progression; Hematochemical markers; Prognostic factors; TB control.

Fig. 1

Summary forest plot with the area under the receiver operating characteristic curve (AUC) with 95 % confidence interval (CI) for each hematological variable.

Fig. 2

Calibration curves (A) and decision curves (B) for the risk prediction models. Higher net benefit suggests higher clinical utility. Model 1: model including the hematological variables as continuous variables. Model 2: model including the hematological variables as clinically relevant categories.