Predicting malignant and tuberculous pleural effusions through demographics and pleural fluid analysis of patients

The differential diagnosis of malignant and tuberculous pleural effusion is frequently difficult.

The aim of our study is to determine the discrimination value of demographic parameters and different biological markers in pleural fluid.

In pleural fluid obtained from 106 patients with tuberculous, 250 with malignant and 218 with miscellaneous pleural effusion, clinical and analytical parameters were analysed, applying polytomous regression analysis and the receiver operating characteristic (ROC) curves.

The three groups could be differentiated using the measured markers. Age, tumour necrosing factor-alpha, lactate dehydrogenase (LDH), adenosine deaminase (ADA), C-reactive protein (CRP) and carcinoembryonic antigen (CEA) were significant predictors for discriminating tuberculous from malignant pleural effusions; nucleated cells, lymphocytes, cholesterol, LDH, ADA, CRP, CEA and CA15.3 distinguish between malignant and miscellaneous pleural effusions. The ROC areas (95% confidence interval) were, 0.973 (0.953, 0.992) for tuberculous, 0.922 (0.900, 0.943) for miscellaneous, and 0.927 (0.907, 0.948) for malignant pleural effusion. The polytomous model correctly classified a significantly high proportion of patients with tuberculosis (85.8%) and cancer (81.6%). The incorrect classification rate was 17.8%, which increased to 19.5% in the correction using bootstrap.

The results obtained to estimate the probability of tuberculous and malignant pleural effusion confirm that this model achieves a high diagnostic accuracy. This model should be applied to determine which patients with a pleural effusion of unknown origin would not benefit from further invasive procedures.