Development and evaluation of the new predictive models in tuberculous pleuritis

Abstract

Different pleural fluid biomarkers have been found useful in the discrimination between tuberculous pleural effusion (TPE) and non-TPE, with interferon gamma (IFN-γ) showing the highest single marker diagnostic accuracy. The aim of the present study was to develop predictive models based on clinical data and pleural fluid biomarkers, other than IFN-γ, which could be applied in differentiating TPE and non-TPE. Two hundred and forty two patients with newly diagnosed pleural effusion were prospectively enrolled. Upon completion of the diagnostic procedures, the underlying disease was identified in 203 patients (117 men and 86 women, median age 65 years; 44 patients with TPE and 159 with non-TPE) who formed the proper study group. Pleural fluid level of ADA, IFN-γ, IL-2, IL-2sRα, IL-12p40, IL-18, IL-23, IP-10, Fas-ligand, MDC, and TNF-α was measured and then ROC analysis and multivariate logistic regression were used to construct the predictive models. Two predictive models with very high diagnostic accuracy (AUC > 0.95) were developed. The first model included body temperature, white blood cell count, pleural fluid ADA and IP-10. The second model was based on age, sex, body temperature, white blood cell count, pleural fluid lymphocyte percentage, and IP-10 level. We conclude that two new predictive models based on clinical and laboratory data demonstrate very high diagnostic performance and can be potentially used in clinical practice to differentiate between TPE and non-TPE.