The objective was 1) to determine the usefulness of different criteria in the differential diagnosis between exudate and transudate in pleural effusion, 2) to evaluate albumin gradient changes in pleural effusion fluids characterized as transudates in patients who do and do not receive diuretic therapy, 3) to define the specificity of pleural effusions of neoplastic etiology. All patients with pleural effusion admitted to the hospital between January 15 and August 15 1994 were evaluated consecutively. Serum and pleural effusion, total protein, LDH, albumin and cholesterol levels were measured and the etiologic diagnosis of the pleural effusion (gold standard) was established. Out of the total of 112 evaluated patients, 7 were excluded because it was impossible to reach a final diagnosis. Based on the etiologic diagnosis, 47 patients (44.8%), average age of 69.6 +/- 12.07, had pleural effusions defined as transudate and 58 patients (55.2%), average age of 66.5 +/- 14.26, had pleural effusions defined as exudate. Sixty-six percent of the transudates were secondary to heart failure, while 40% of the exudates were of neoplastic origin. Using the criteria of Light et al, we obtained a diagnostic accuracy (DA) of 82.7% (CI 95% 73.1-90.0)%. However, when the cut-off point was modified according to Valdez and the value of cholesterol in pleural effusion and its relation to serum cholesterol was added, the DA rose to 90.2 (83.2-96.0)% (p < 0.05). The effusion-serum cholesterol ratio demonstrated 100 (85.1-100)% sensitivity for neoplastic effusions, whereas for non-neoplastic exudative effusions the sensitivity was 89 (73.2-96.8)%. The tests, however, showed only 17.4 (6.56-33.6)% specificity. The albumin gradient (the difference between serum and pleural effusion albumin) did no vary in patients with transudates who received diuretics, allowing a correct diagnosis of transudate in 93 (82.4-97.8)% of the cases. However, in patients who were taking diuretics, the classic criteria of protein index defined correctly only 66 (53.4-82.1)% of the cases (p < 0.05). It can be concluded that the variation of cut-off points originally established by Light et al. and the addition of cholesterol determination in pleural effusion and its relation to the serum cholesterol level allowed us to increase the DA. This appears to be the best way to differentiate a transudate from an exudate. The relation between pleural effusion and serum cholesterol levels showed a very low specificity for the differentiation of neoplastic and non-neoplastic exudative pleural effusions. Unlike the pleural effusion-serum total protein ratio, the albumin gradient allowed us to establish the correct diagnosis of transudate even in patients taking diuretics.