Improving the predictive accuracy of identifying exudative effusions.

Link to article at PubMed

Improving the predictive accuracy of identifying exudative effusions.

Chest. 2013 Sep 5;

Authors: Kummerfeldt CE, Chiuzan CC, Huggins JT, Divietro ML, Nestor JE, Sahn SA, Doelken P

Abstract
ABSTRACT BACKGROUND Application of Light's criteria results in misclassification of some transudative effusions as exudative, particularly due to congestive heart failure (CHF). We sought to determine if the serum to pleural fluid albumin (SF-A) and serum to pleural fluid protein (SF-P) gradients increased the predictive accuracy to correctly identify exudative effusions. METHODS We retrospectively analyzed 1,153 consecutive patients who underwent a diagnostic thoracentesis at the Medical University South Carolina (MUSC). Univariable logistic regression analyses were used to determine the statistical significance of pleural fluid tests that correctly identified exudative effusions. Tests with significant diagnostic accuracy were combined in multivariable logistic regression models, with calculation of areas under the curve (AUC) to determine their predictive accuracy. The predictive capability of the best model was compared to Light's criteria and other test combinations. RESULTS Pleural fluid LDH, SF-A, and SF-P gradients had a significant effect on the probability of identifying exudative pleural effusions. When combined together in a multivariable logistic regression, LDH (odds ratio 14.09 [95% CI, 2.25 to 85.50]), SF-A (odds ratio 7.16 [95% CI, 1.24 to 41.43]) and SF-P gradients (odds ratio 6.83 [95% CI, 1.56 to 27.88]) had an AUC of 0.92 (95% CI, 0.85 to 0.98). CONCLUSION Application of Light's criteria, not uncommonly, misclassifies CHF transudative effusions as exudates. In cases where no cause for an exudative effusion can be identified or CHF is suspected, the sequential application of the fluid LDH, followed by the SF-P and then the SF-A gradients may assist in re-classifying pleural effusions as transudates.

PMID: 24008773 [PubMed - as supplied by publisher]

Leave a Reply

Your email address will not be published.