A comparative analysis of the biochemical parameters used to distinguish between pleural exudates and transudates. Using the criteria of Light et al., 94.5% of the effusions were correctly classified, yielding a sensitivity and specificity of 99.3% and 87.6%, respectively. When the combination of pleural fluid cholesterol level and lactate dehydrogenase (LDH) level was used, the specificity and accuracy were found to be higher than that
A comparative analysis of the biochemical parameters used to distinguish between pleural exudates and transudates. Using the criteria of Light et al., 94.5% of the effusions were correctly classified, yielding a sensitivity and specificity of 99.3% and 87.6%, respectively. When the combination of pleural fluid cholesterol level and lactate dehydrogenase (LDH) level was used, the specificity and accuracy were found to be higher than that
A comparative analysis of the biochemical parameters used to distinguish between pleural exudates and transudates. Using the criteria of Light et al., 94.5% of the effusions were correctly classified, yielding a sensitivity and specificity of 99.3% and 87.6%, respectively. When the combination of pleural fluid cholesterol level and lactate dehydrogenase (LDH) level was used, the specificity and accuracy were found to be higher than that
A comparative analysis of the biochemical parameters used to
distinguish between pleural exudates and transudates AnNnN YILMAZ, IxNun KiIi TUNABOYU, EsrN AKKAYA AND BinoI BAYRAMGRLER Department of Pulmonology, SSK Sreyyapapa Center for Chest Diseases and Thoracic Surgery, Istanbul, Turkey A comparative analysis of the biochemical parameters used to distinguish between pleural exu- dates and transudates YILMAZ A, TUNABOYU K, AKKAYA E, BAYRAMGRLER B. Respirology 2000; 5: 363367 Objective: The aim of the present study was to compare the various parameters used to identify exudates. Methodology: The study included 255 patients with pleural effusions. According to aetiological diagnosis, 105 pleural effusions were labelled as transudates and 150 were labelled as exudates. Results: Using the criteria of Light et al., 94.5% of the effusions were correctly classied, yielding a sensitivity and specicity of 99.3% and 87.6%, respectively. Use of the pleural uid/serum biliru- bin ratio produced results of 92.9%, 90.7%, and 96.2%, respectively. Using pleural uid cholesterol level yielded results of 95.7%, 95.3%, and 96.2%, respectively. When the combination of pleural uid cholesterol level and lactate dehydrogenase (LDH) level was used, the specicity and accuracy were found to be higher than that using the criteria of Light et al. We found that there was no signicant difference among the parameters with respect to accuracy. Conclusion: When the accuracy and cost are considered, differentiation of pleural exudates and transudates can be achieved only by pleural uid cholesterol level or LDH level; and when two para- meters were used together, the accuracy and specicity were higher than that using the criteria of Light et al. Key words: differentiation, exudates, pleural effusion, transudates. patients with transudates. 1,57 This cast some doubt on the universal applicability of these criteria. Several alternative criteria have been proposed, and some of these seem to have higher diagnostic accuracy for identifying transudates. These include use of the pleural uid total cholesterol concentration, pleural uid serum albumin gradient, and pleural uid/ serum bilirubin ratio. 2,6,810 Controversy exists as to which method is more accurate. 1,5,6,9,10 The purpose of the present study was to evaluate a large number of patients prospectively to compare the relative utility of the various parameters in identifying exudates and nd the most suitable parameter with respect to ef- ciency and cost. METHODS Patient selection Two hundred and fty-ve pleural effusions from consecutive patients admitted to SSK Sreyyapapa Respirology (2000) 5, 363367 INTRODUCTION Pleural effusions evolve in the course of a variety of diseases. A correct diagnosis of the underlying disease is essential to rational management. 1 The rst step in the diagnosis of pleural effusions is the distinction between exudates and transudates. 24 The current cri- teria for differentiating exudates from transudates in pleural effusions were established by Light et al. in 1972. Sensitivity and specicity, calculated from their data, were 99% and 98%, respectively. 3 However, several recent reports have shown that the low speci- city of the criteria of Light et al. may lead to unwar- ranted invasive interventions in up to 2030% of Correspondence: Dr Adnan Yilmaz, Zmrtevler Atatrk Cad. Abant Apt. No: 30, Kat:281530, Maltepe, Istanbul, Turkey. Email: elim@rt.net.tr Received 13 March 2000; accepted for publication 13 July 2000. Center for Chest Diseases and Thoracic Surgery, Istanbul, Turkey, were studied prospectively between November 1996 and April 1997. Considering the aeti- ology of the effusions as the gold standard for the classication of pleural uid, 2,8 105 pleural uid samples were labelled as transudates and 150 were labelled as exudates. Among the 105 patients with transudates, 33 were women and 72 were men, with an average age of 55.8 (range 2876). Among the 150 patients with exudates, 25 were women and 125 were men, with an average age of 39.6 (range 1475). The diagnosis of the disease causing the effusion was considered to be conrmed when the following conditions were met: 2,8 1. Congestive heart failure (CHF): presence of an enlarged heart with clinical or echocardiographic evidence of cardiac disfunction, and one or more of the following alterations: elevated venous pressure, oedema, tachycardia, or ventricular gallop. Patients suspected of having respiratory infections, pul- monary emboli, or persistence of the effusion after adequate treatment of the cardiac insufciency, were excluded. 2. Liver cirrhosis (LC): clinical and laboratory evi- dence of hepatic damage with portal hypertension or hypoalbuminaemia and an absence of purulent sputum, malignancy, or pulmonary inltrates. 3. Nephrotic syndrome: It was diagnosed when there was proteinuria, oedema, and hypoalbu- minaemia and an absence of purulent sputum, malignancy, or pulmonary inltrates. 4. Tuberculosis: presence of tuberculous granulo- mas in pleural biopsy specimen or positive smear or culture of acid-fast bacilli. 5. Parapneumonic effusion: clinically and radio- logically conrmed pneumonia with evidence of bac- terial invasion of the effusion. 6. Empyema: presence of pus cells in the pleural cavity. 7. Neoplastic: cytological and/or histological demonstration of pleural involvement. 8. Rheumatoid arthritis: clinical and laboratory ndings of disease. Study design Samples of venous blood and pleural uid in each patient were obtained on the same day shortly after hospital admission. Only the results of the rst thora- centesis were considered. The following biochemical parameters were determined on samples of the pleural uid and the venous blood: total protein, albumin, total cholesterol, total bilirubin, and lactate dehydrogenase (LDH). Total protein levels were mea- sured by the biuret method. Lactate dehydrogenase was measured using a kinetic UV optimized standard method (the upper normal limit for serum is dened at 460 IU/L). Cholesterol levels were measured with enzymatic colorimetric method (CHOD-PAP). Biliru- bin levels were measured by modied Van den Bergh and Mller method. For the laboratory classication of pleural uids, the following criteria were evaluated: pleural uid (PF) to serum (S) protein ratio > 0.5, PF to S LDH ratio > 0.6, PF LDH> 307 IU/L (the cut-off value of pleural uid LDH is two-thirds of the upper normal limit for the serum LDH), PF cholesterol > 60 mg/dL, and PF to serum bilirubin ratio > 0.6. 9,11 Three cut-off points ( 0.3, 0.35, and 0.4) were adopted for PF to S cholesterol ratio. 2 A cut-off point of 1.2 g/dL was used for pleural uid serum albumin gradient. 2 Statistical analysis The usefulness of each biochemical parameter for identifying exudates was evaluated using Bayesian methods to measure the following: sensitivity, TP/ (TP+ FN); specicity, TN/(TN+ FP); accuracy (TP+ TN)/(TP+ TN+ FP+ FN); positive predictive value, TP/TP+ FP; negative predictive value, TN/TN+ FN where TP is the number of true positive diagnoses, FP the number of false positive diagnoses, TN the number of true negative diagnoses, and FN the number of false negative diagnoses. These indices were compared using McNemars exact test for cor- related proportions. 12 RESULTS According to the causal disease, 105 pleural uid samples were labelled as transudates and 150 were labelled as exudates (Table 1). The criteria of Light et al. Using these criteria for separating exudates from transudates, 14 of 255 were misclassied (accuracy 94.5%). Among the 150 exudates, one was misclassi- ed (sensitivity 99.3%). Clinical diagnosis of this case was tuberculosis. Thirteen patients with transudates were not correctly classied (specicity 87.6%). Among the transudates falsely classied as exudates, eight were CHF, two were liver cirrhosis, and three were nephrotic syndrome. 364 A Yilmaz et al. Table 1 Causes of 255 pleural effusions Cause No. Transudates 105 Congestive heart failure 60 Liver cirrhosis 32 Nephrotic syndrome 13 Exudates 150 Tuberculosis 85 Malignant effusion 36 Parapneumonic effusion 22 Empyema 5 Rheumatoid arthritis 2 Ratio of pleural uid protein to serum protein Using a dividing line of 0.5, 242 of 255 pleural uids were correctly classied (accuracy 94.9%). Six patients with exudates were misclassied as transu- dates (sensitivity 96%). Among the misclassied exudates, two were tuberculosis, two were parapneu- monic effusion, and two were neoplastic effusion. Seven patients with transudates were falsely classied as exudates (specicity 93.3%). Of these patients, four were CHF, two were liver cirrhosis, and one was nephrotic syndrome. Pleural uid lactate dehydrogenase Using 307 IU as a cut-off point for separating exudates from transudates, 9,11 241 of 255 patients were cor- rectly classied (accuracy 94.5%). Ten patients with exudates and four patients with transudates were falsely classied (sensitivity 93.3%, specicity 96.2%). Among these, four were tuberculosis, six were neo- plastic, two were CHF, one was liver cirrhosis, and one was nephrotic syndrome. Ratio of pleural uid lactate dehydrogenase to serum lactate dehydrogenase Using a dividing line of 0.6, 242 of 255 were correctly classied (accuracy 94.9%). Five of 150 patients with exudates were misclassied using this parameter (sensitivity 96.6%). Among the exudates falsely classi- ed as transudates, three were tuberculosis and two were neoplastic. Eight of 105 patients with transu- dates were mistakenly classied as exudates (speci- city 92.3%). Among these patients, six were CHF and two were nephrotic syndrome. Pleural uid serum albumin gradient Using a dividing line of 1.2 g/dL, 242 of 255 pleural uid samples (efciency 94.9%) were obtained. Five of 150 patients with exudates were falsely classied (sensitivity 96.6%). Among these patients, two were tuberculosis, two were neoplastic, and one empyema. Eight of 105 cases with transudates were misclassied (specicity 92.3%). Among the misclassied transu- dates, three were CHF, three were liver cirrhosis, and two were nephrotic syndrome. Pleural uid cholesterol Drawing a dividing line in the concentration of pleural uid cholesterol of 60 mg/dL, 244 of 255 pleural uid samples were correctly classied (accuracy 95.7%). Seven of 150 patients with exudates were misclassied (sensitivity 95.3%). Among seven patients, two were tuberculosis, one was empyema, and four were neoplastic. Four of 105 patients with transudates were falsely classied as exudates (speci- city 96.2%). All patients were CHF. Ratio of pleural uid cholesterol to serum cholesterol Using a dividing line of 0.3, 2 237 of 255 pleural uid samples were correctly classied (accuracy 92.9%). Seven of 150 exudates were mistakenly classied as transudates (sensitivity 95.3%). Among these falsely classied exudates, three were tuberculosis, one was parapneumonic effusion, and three were neoplastic. Eleven of 105 transudates were misclassied (speci- city 89.5%). Among the transudates incorrectly clas- sied were six pleural uids due to CHF, three due to liver cirrhosis, and due secondary to nephrotic syn- drome. Using a dividing line of 0.4, 2 230 of 255 patients were correctly classied (accuracy 90.2%). Twenty-two of 150 exudates were falsely classied (sensitivity 85.3%). Three of 105 transudates were incorrectly classied (specicity 97.1%). Among the pleural uids incorrectly classied, two were CHF, one was liver cirrhosis, 13 were tuberculosis, four were parapneumonic, and ve were neoplastic. Ratio of pleural uid bilirubin to serum bilirubin Using a dividing line of > 0.6, 237 of 255 pleural uid samples were correctly classied (accuracy 92.9%). Fourteen of 150 exudates were misclassied (sensi- tivity 90.6%). Among these falsely classied exudates, seven were tuberculosis, two were parapneumonic, one was empyema, and four were neoplastic. Four of 105 transudates were falsely classied as exudates (specicity 96.2%). Of these patients, three were CHF and one was liver cirrhosis. Pleural uid cholesterol and pleural uid lactate dehydrogenase Using these parameters together, 247 of 255 patients were correctly classied (accuracy 96.8%). Among the exudates, one tuberculosis patient was falsely classi- ed as transudate (sensitivity 99.3). Seven of 105 transudates were misclassied (specicity 93.3%). Among these patients, ve were CHF, one was liver cirrhosis, and one was nephrotic syndrome. The statistical results of these parameters are sum- marized in Table 2. When these indices were com- pared using McNemars exact test, there was no signicant difference among the parameters with respect to accuracy. DISCUSSION Separation of exudates from transudates remains a useful initial step in determining the cause of a pleural effusion. 2 Historically, specic gravity was used to separate these two entities. 13 Later, a pleural uid protein level of 30 g/L was used. 14 Many mis- classications were made resulting in patients endur- ing unnecessary investigations. Chandrasekhar et al. The differentiation of pleural uids 365 thus proposed the use of absolute values of pleural uid LDH in making this distinction. 15 A combination of two parameters was used by Light et al. in 1972. Sensitivity and specicity, calculated from their data, were 99 and 98%, respectively. 3 Recently, however, several prospective studies were unable to reproduce the excellent results obtained by Light et al. In most of them, using the criteria of Light et al., the sensitiv- ity remained high, greater than 95%, but the speci- city did not surpass 78%. 1,5,6 Roth et al. showed that limitation of the criteria of Light et al. could be overcome by measuring the serumeffusion albumin gradient. 6 In this study, a sensitivity and specicity of 95% and 100%, respec- tively, was demonstrated. Burgess et al., however, did not agree with these results. 2 From the results of this study, the sensitivity and specicity were 87% and 92%, respectively. Hamm et al. 1 and Valdes et al. 5 sug- gested the use of cholesterol levels in distinguishing between transudates and exudates. The sensitivity and specicity for cholesterol level were found as 100% and 95%, respectively. 1 Romero et al., however, did not agree with these results. 9 Sensitivity and specicity, calculated from these data, were 72% and 85%, respectively. Meisel et al. investigated the pleural uid/serum bilirubin ratio and yielded a sensitivity and specicity of 96% and 83%, respectively. 10 In the study by Burgess et al., a sensitivity and specicity of 81% and 61%, respectively, were obtained. 2 From the results of the present study, the criteria of Light et al. remain the most sensitive (99.3%) for exu- dates. However, unlike the results obtained by Light et al. 3 in their initial study, the specicity was much lower at 87.6%. Using one of the proposed alternative criteria, the pleural uid cholesterol level, we found that the specicity was higher than when using the criteria of Light et al. However, sensitivity of this cri- teria was lower than when using the criteria of Light et al. Using the combination of pleural uid choles- terol and pleural uid LDH obtained higher sensitiv- ity than using either alone. The specicity and accuracy of this combination were found to be higher than when using the criteria of Light et al. When the ratio of pleural uid/serum bilirubin level was used, the sensitivity and accuracy were lower than when using the criteria of Light et al. However, these criteria had higher specicity. When it was compared with pleural uid cholesterol level, it had a lower sen- sitivity and accuracy but specicity was similar. The present results showed that there was no signicant difference among the parameters with respect to accuracy. We conclude that differentiation of pleural exu- dates and transudates can be achieved only by pleural uid cholesterol or LDH level measurement. When pleural uid cholesterol and LDH level measurement were used together, this combination has an ef- ciency higher than that achieved by the criteria of Light et al. The proposed combination has advan- tages in that a contemporary blood sample is not required and that chemical tests are reduced, thereby lowering the cost of the diagnostic procedure. REFERENCES 1 Hamm H, Brohan U, Bohmer R, Missmahl HP. Choles- terol in pleural effusions: A diagnostic aid. Chest 1987; 92: 296302. 2 Burgess LJ, Maritz FJ, Taljaard JJF. Comparative analysis of the biochemical parameters used to distinguish between pleural transudates and exudates. Chest 1995; 107: 16049. 3 Light RW, MacGregor MI, Luchsinger PC, Ball WC. Pleural effusions: The diagnostic separation of transu- dates and exudates. Ann. Intern. Med. 1972; 77: 50713. 4 Peterman TA, Speicher CE. Evaluating pleural effusions: A two-stage laboratory approach. JAMA 1984; 252: 10513. 5 Valdes L, Pose A, Suarez J et al. Cholesterol: A useful para- meter for distinguishing between pleural exudates tran- sudates. Chest 1991; 99: 1097102. 6 Roth BJ, OMeara TF, Cragun WH. The serum-effusion albumin gradient in the evaluation of pleural effusions. Chest 1990; 98: 5469. 7 Hirsch A, Rufe P, Nebut Met al. Pleural effusion: Labo- ratory tests in 300 cases. Thorax 1979; 34: 10612. 8 Costa M, Quiroga T, Cruz E. Measurement of pleural uid cholesterol and lactate dehydrogenase: A simple and 366 A Yilmaz et al. Table 2 Summary of statistical results Sensitivity Specicity Accuracy Positive Negative predictive predictive value value Pleural uid/serum protein 96.0 93.3 94.9 95.3 94.2 Pleural uid lactate dehydrogenase 93.3 96.2 94.5 97.2 91.0 Pleural uid/serum lactate dehydrogenase 96.7 92.4 94.9 94.8 95.1 Light et al. 99.3 87.6 94.5 91.9 99.0 Serum/pleural uid albumin 96.7 92.4 94.9 94.8 95.1 Pleural uid/serum bilirubin 90.7 96.2 92.9 97.1 87.8 Pleural uid cholesterol 95.3 96.2 95.7 97.3 93.5 Pleural uid/serum cholesterol 0.3 95.3 89.5 92.9 92.9 93.1 Pleural uid/serum cholesterol 0.35 92.0 93.3 92.5 95.2 89.1 Pleural uid/serum cholesterol 0.4 85.3 97.1 90.2 97.7 82.3 Pleural uid cholesterol + pleural uid lactate dehydrogenase 99.3 93.3 96.9 95.5 99.0 accurate set of indicators for separating exudates from transudates. Chest 1995; 108: 12603. 9 Romero S, Candela A, Martin C, Hernandez L, Trigo C, Gil J. Evaluation of different criteria for the separation of pleural transudates from exudates. Chest 1993; 104: 399404. 10 Meisel S, Shamis A, Thajer Met al. Pleural uid to serum bilirubin concentration ratio for the separation of tran- sudates from exudates. Chest 1990; 98: 1414. 11 Anderson WM, Light RW. Invasive diagnostic proce- dures. In: George RB, Light RW, Matthay AM, Matthay RA (eds). Chest Medicine. Williams & Wilkins, Baltimore, 1995; 94. 12 Dawson-Saunders B, Trapp RG. Basic and clinical statis- tics. Appleton & Lange, New Jersey, 1990; 1545. 13 Paddock EC. The diagnostic signicance of serous uids in disease. N. Engl. J. Med. 1940; 223: 10105. 14 Leuallen EC, Carr DT. Pleural effusion: A statistical study of 436 patients. N. Engl. J. Med. 1955; 252: 7983. 15 Chandrasekhar AJ, Palatao A, Dubin A, Levine H. Pleural uid lactic acid dehydrogenase activity and protein content. Arch. Intern. Med. 1969; 128: 4850. The differentiation of pleural uids 367