Revisiting Light's criteria: a validated blood-free triple combination matches diagnostic accuracy in over 7000 patients

Background Light's criteria remain the standard for distinguishing exudative from transudative pleural effusions, but require serum sampling and lack specificity. We assessed whether a pleural fluid-only approach could match the diagnostic accuracy. Methods We analysed 7280 diagnostic thoracenteses from a single centre, divided into derivation (n=5000) and validation (n=2280) cohorts. We compared Light's criteria with a triple (protein >3 g·dL −1 , lactate dehydrogenase (LDH) >250 IU·L −1 or cholesterol >55 mg·dL −1 ) and a double (LDH >250 IU·L −1 or cholesterol >55 mg·dL −1 ) combination using sensitivity, specificity, likelihood ratios and area under the curve (AUC). AUCs were assessed using the DeLong method with multiple imputations from a mixed model. McNemar's test examined discordant classifications. Results The triple combination showed no significant AUC difference versus Light's criteria in either cohort and had equivalent sensitivity (99% versus 98% in derivation; both 98% in validation). In the derivation cohort, McNemar's test showed a small but statistically significant excess of false negative exudates with the triple combination (p