Development of a clinical prediction model for the diagnosis of tuberculous pleural effusion in a resource-limited setting

Pleural tuberculosis is a common form of extrapulmonary tuberculosis but not many patients with tuberculous pleural effusion (TPE) are acid-fast bacilli (AFB) smear or culture positive due to its pathophysiology. High cost and limited availability can make adenosine deaminase (ADA) inaccessible. This study developed a clinical prediction model to diagnose TPE in adults using routinely collected clinical and laboratory data. A cross-sectional study was conducted at an academic medical centre in Manila, Philippines. Sociodemographic, clinical and laboratory data were collected via electronic health records. A multivariable logistic regression model was constructed with predictors selected through an Akaike information criterion algorithm. Bootstrapping was used for internal validation. Of 474 patients suspected to have TPE, the derivation cohort, only 55 (12%) were positive on pleural fluid (PF) AFB smear or culture, polymerase chain reaction, ADA (≥ 40 U/L) or pleural biopsy. Age, sex, steroid use, serous PF appearance, PF glucose and PF eosinophils were included in the model with acceptable calibration (Brier score 0.09 in derivation and 0.10 in validation) and modest discrimination (c-static 0.72 and 0.69 respectively). At an 11.4% threshold, the model had 75% (95% CI 49-83%) and 57% (47-76%) sensitivity and specificity respectively in the validation cohort. Positive and negative predictive values were 94% (91-97%) and 20% (15-26%) respectively. A prediciton model without ADA had high positive predictive ability for TPE. Work into making more accurate and accessible tests should be considered.

Recommended citation: Villanueva, C. A., Cruz, A. M., Cruz, K. M., Uy, T. C., Villejo, S. J., & Santiaguel, J. (2025). Development of a clinical prediction model for the diagnosis of tuberculous pleural effusion in a resource-limited setting. European Respiratory Journal, 66(Suppl 69), PA783. https://doi.org/10.1183/13993003.congress-2025.PA783
Download Paper