Table 1.
Statistical analysis of the parameters used to classify transudates and exudates, expressed as percentage (95% confidence intervals)
| Fluid total protein | Fluid cholesterol | Fluid LDH | Fluid:serum total protein | Fluid LDH + fluid:serum total protein | Light's criteria | |
| Cut off | 28 g/l | 0.8 mmol/l | 130 U/l | 0.4 | ||
| Sensitivity | 83 (69 to 96) | 87 (69 to 96) | 90 (73 to 98) | 90 (73 to 98) | 100 (91 to 100) | 90 (73 to 98) |
| Specificity | 75 (53 to 90) | 67 (45 to 84) | 79 (58 to 93) | 79 (58 to 93) | 71 (51 to 80) | 67 (45 to 84) |
| Efficiency | 80 (66 to 89) | 78 (64 to 88) | 85 (73 to 93) | 85 (73 to 93) | 87 (78 to 92) | 80 (66 to 89) |
| PPV | 81 (63 to 93) | 76 (59 to 89) | 84 (67 to 95) | 84 (67 to 95) | 81 (66 to 89) | 77 (60 to 90) |
| NPV | 78 (56 to 93) | 80 (56 to 94) | 86 (65 to 97) | 86 (65 to 97) | 100 (87 to 100) | 84 (61 to 96) |
| Youden's index | 0.58 | 0.53 | 0.69 | 0.69 | 0.71 | 0.57 |
Cut off values were obtained from ROC analysis. Sensitivity is TP/(TP+FN), specificity is TN/(TN+FP), efficiency is (TP+TN)/(TP+TN+FP+FN), where TP is the number of true positive diagnoses (number of exudates correctly diagnosed), TN is the number of true negative diagnoses (number of transudates correctly diagnosed), FP is the number of false positive diagnoses (number of transudates undiagnosed), and FN is the number of false negative diagnoses (the number of exudates undiagnosed). Youden's index can be used to compare the false negative and false positive rates of different tests. For fluid LDH, the probability of a false negative, Pr(FN) is 3/30 and the probability of a false positive, Pr(FP) is 5/24. Youden's index = 1−(3/30+5/24)=0.69. The higher the value, the lower the false negative and false positive rates.
LDH, lactate dehydrogenase; ROC, receiver operator characteristic.