A comprehensive diagnostic model for tuberculous PE: integration of clinical and immunological biomarkers

Background Tuberculous pleurisy represents a prevalent form of extrapulmonary tuberculosis and constitutes a significant diagnostic challenge in clinical practice, particularly in endemic regions where it accounts for approximately 20–30% of all PEs. The nonspecific clinical presentation and the limitations of conventional diagnostic methods—including the low sensitivity and prolonged turnaround time of pleural fluid culture—often result in delayed diagnosis and treatment, potentially compromising patient outcomes. Objective To develop and validate a comprehensive, evidence-based predictive model that integrates readily available clinical and immunological biomarkers to enhance the early and accurate diagnosis of tuberculous pleurisy, thereby supporting timely clinical decision-making and optimized patient management. Methods We conducted a retrospective cohort study of 523 consecutive patients presenting with PE at a tertiary care center between 2010 and 2021, including 375 patients with confirmed tuberculous pleurisy and 148 with non-tuberculous effusions. Demographics, biochemical markers (adenosine deaminase, lactate dehydrogenase, C-reactive protein, D-dimer), immunological parameters (T-SPOT.TB, T-cell subsets), and clinical indicators were systematically evaluated. Statistical analyses encompassed descriptive and comparative tests, multivariate logistic regression, and receiver operating characteristic curve analysis. Missing data were addressed using median imputation (Rivalta test, 4.4%) and case-wise deletion (T-cell subsets, 14.9%). Results Five independent predictors were significantly associated with tuberculous pleurisy: younger age, elevated adenosine deaminase, lower C-reactive protein, positive Rivalta test, and positive T-SPOT.TB result. The multivariate logistic regression model demonstrated excellent discriminative performance (pseudo R 2 = 0.450) and strong model fit. Receiver operating characteristic analysis identified an optimal adenosine deaminase cutoff of 30.5 U/L, yielding a sensitivity of 7.1%. Model robustness was further confirmed through rigorous internal validation. Conclusion This study presents a robust and clinically applicable diagnostic model that effectively distinguishes tuberculous pleurisy from non-tuberculous PEs by integrating multiple routinely available biomarkers. The model offers a practical, cost-effective tool for early diagnosis, with the potential to improve therapeutic timeliness and patient outcomes across varied healthcare settings.