Development and validation of a scoring system for the identification of pleural exudates of cardiac origin.

BACKGROUND

Light's criteria misclassify about 30% of cardiac effusions as exudates, possibly leading to unnecessary testing. Our purpose was to derive and validate a scoring model to effectively identify these falsely categorized cardiac effusions, in the setting of natriuretic peptide lacking data.

METHODS

We retrospectively analyzed data from 3182 patients with exudative pleural effusions based on Light's criteria, of whom 276 had heart failure (derivation set). A scoring model was generated with those variables identified as independent predictors of cardiac effusions in a logistic regression analysis, and further evaluated in an independent population of 1165 patients.

RESULTS

The score consisted of age ≥75years (3 points), albumin gradient >1.2g/dL (3 points), pleural fluid lactate dehydrogenase <250U/L (2 points), bilateral effusions on chest radiograph (2 points), and protein gradient >2.5g/dL (1 point). At the best cutoff of ≥7 points, the score yielded 92% diagnostic accuracy, a likelihood ratio positive of 12.7 and a likelihood ratio negative of 0.39 for labeling cardiac effusions in the derivation sample. The respective figures in the validation sample were 87%, 6.5 and 0.33. Notably, the score had higher discriminatory properties than protein and albumin gradients in both the derivation (respective area under the curve - AUC - of 0.925, 0.825, and 0.801) and validation (respective AUC of 0.908 0.862 and 0.802; all p≤0.01) cohorts.

CONCLUSIONS

A simple scoring system can assist clinicians in accurately identifying false cardiac exudates when natriuretic peptides are not available.