Risk factors for thrombosis in tuberculosis patients admitted to a tuberculosis-dedicated intensive care unit: a retrospective cohort study.

BACKGROUND

Patients with tuberculosis in the intensive care unit (ICU) face an elevated risk of thrombosis; however, the contributing factors remain incompletely understood. This study aimed to identify independent risk factors for thrombus formation, evaluate the incremental predictive value of inflammatory biomarkers [C-reactive protein (CRP), D-dimer (DDR), and interleukin-6 (IL-6)], and explore the determinants of IL-6 levels in critically ill tuberculosis patients.

METHODS

This retrospective cohort study consecutively enrolled 168 tuberculosis patients admitted to the ICU, including 101 with thrombosis and 67 without. Demographic, clinical, and laboratory data were collected. Univariate and multivariable binary logistic regression analyses were performed to identify independent risk factors. The area under the receiver operating characteristic curve (AUC) and DeLong test were used to compare five logistic models: a base model [age, activated partial thromboplastin time(APTT), and non-TB bacterial/fungal infection status] and four extended models incorporating ln-transformed CRP, DDR, or IL-6, individually or in combination. Multivariable linear regression analysis was employed to identify factors independently associated with IL-6 levels.

RESULTS

Advanced age (OR = 1.057, p = 0.001) and prolonged APTT (OR = 1.053, p = 0.041) were independent predictors of thrombus formation. In the multivariable model, fungal co-infection (OR = 3.185, p = 0.006) and non-TB bacterial co-infection (OR = 0.336, p = 0.038) also reached statistical significance. Age-stratified analysis revealed that among patients aged <70 years, only age was independently associated with thrombosis (OR = 1.050, p < 0.05), whereas no other parameter demonstrated independent predictive value. The base model yielded an AUC of 0.753; the addition of CRP, DDR, or IL-6 did not produce a statistically significant improvement in discriminatory performance (all p > 0.05). In multivariable linear regression, only CRP was independently associated with IL-6 levels (β = 0.423, p < 0.001).

CONCLUSION

Age and APTT are independent risk factors for thrombus formation in patients with tuberculosis. Fungal co-infection confers additional thrombotic risk, whereas non-TB bacterial co-infection exhibits a paradoxical protective effect, the underlying mechanisms of which warrant further investigation. A parsimonious model based on age, APTT, and co-infection status demonstrated moderate predictive accuracy (AUC = 0.753), and the incorporation of CRP, DDR, or IL-6 failed to enhance its discriminatory ability. CRP serves as a reliable surrogate marker for IL-6-driven inflammation in this population. Large-scale prospective studies are warranted to validate and extend these findings.