Is It Meaningful To Use Biochemical Parameters To Discriminate Between Transudative and Exudative Pleural Effusions?* Santiago Romero-Candeira, MD; Luis Hernández, MD; Susana Romero-Brufao, MD; David Orts, MD; Cleofé Fernández, MD; and Concepción Martín, MD Objectives: The usefulness of biochemical criteria to separate pleural transudates from exudates is controversial, and the limitations of Light s criteria in patients receiving diuretic therapy is of general concern. We evaluated the added value of biochemical criteria to clinical judgment for separating transudates from exudates. Setting: A community teaching hospital. Design: A prospective, observational study for the evaluation of diagnostic tests. Patients and measurements: In 249 consecutive patients referred for diagnostic thoracentesis, two physicians classified the pleural effusion as transudate or exudate based on all available information just before performing the thoracentesis. The sensitivity, specificity, and accuracy of the clinical presumption were compared with those of Light s criteria, and serum-pleural fluid albumin and protein gradients. The combined accuracy of biochemical and clinical criteria was also assessed. Results: The accuracy of Light s criteria (93%) was significantly higher than that showed by the initial clinical presumption (84%; p < 0.0) and that of the alternative biochemical criteria: serum-pleural fluid albumin gradient (87%; p < 0.03) and serum-pleural fluid protein gradient (86%; p < 0.0). In patients receiving diuretic therapy, the accuracy of Light s criteria was 83% (60 of 72 cases), neither different to that of the albumin gradient (88%; 63 of 72 cases) nor to that of the protein gradient (86%; 62 of 72 cases). When these alternative biochemical criteria were applied exclusively to patients receiving diuretics who were thought to have a transudative effusion by clinical criteria, but having fluid identified as exudative by Light s criteria, the results did not improve significantly. Using a multiparametric model for the population receiving diuretics, the accuracy was greater (93%; 67 of 72 cases) than that of Light s criteria but without reaching statistical significance (p 0.2). Conclusions: Light s criteria are significantly superior to the clinical presumption to separate pleural transudates from exudates. In patients receiving diuretics, Light s criteria lose accuracy, which is similar to that showed by the use of alternative biochemical criteria alone or combined with clinical judgment. (CHEST 2002; 22:524 529) Key words: biochemical criteria; exudates; pleural effusions; transudates Abbreviations: CHF congestive heart failure; LDH lactate dehydrogenase It is generally admitted that defining a pleural effusion as a transudate limits the differential diagnosis to a small number of disorders. It also ends the need for further diagnostic workup of the pleural *From Servicio de Neumología. Hospital General Universitario de Alicante, Spain. Manuscript received August 2, 200; revision accepted June 5, 2002. Correspondence to: Santiago Romero-Candeira, MD, C/Italia, n 30, Esc. 2 a, Dcha., 03003 Alicante, Spain; e-mail: romero_san@ gva.es effusion itself. 3 As a consequence, cost and morbidity are dramatically reduced. Occasionally ( 5%), neoplastic effusions may be transudative, For editorial comment see page 505 but in these cases the classification remains useful. The occurrence of a transudative effusion in a patient with lung cancer in the absence of other potential causes of pleural effusion indicates the presence of pulmonary atelectasis or early lymphatic 524 Clinical Investigations
obstruction, 4,5 findings that may lead the following steps in the diagnostic workup. The use of Light s criteria to separate transudates from exudates has been generally admitted as the first step in the study of a pleural effusion of unknown cause. 6 However, concerns about the usefulness of this approach are growing. First, alternative biochemical parameters, such as the cholesterol level, 7 9 or complementary criteria, such as the serum-effusion albumin gradient when the patient is receiving diuretic therapy, were recommended. 0, More recently, the doubts are about their actual utility and that of other biochemical alternatives. 2 Do biochemical criteria improve more than confound the initial diagnosis based on clinical judgment? In the present study, we prospectively investigated the information added to the initial presumptive diagnosis by the use of Light s criteria to separate transudates from exudates. We also assessed the convenience of turning to alternative criteria, such as the serum-pleural fluid albumin gradient alone or combined with clinical criteria, in patients receiving diuretic therapy. Finally, because in patients with congestive heart failure (CHF) who have undergone diuresis the increases in the pleural concentration of albumin are similar to those of total protein, 3 we also analyzed the convenience of using a more cost-efficient parameter, the serum-pleural fluid protein gradient, as an alternative in these patients. Materials and Methods Consecutive patients referred to our service for diagnostic thoracentesis from May 999 to March 2000 were prospectively studied. The following studies were performed on pleural fluid samples: glucose, protein, albumin, lactate dehydrogenase (LDH), cholesterol, cell count, differential cell count, bacterial culture, acid-fast bacilli smear and culture, and cytology. Only the results of the first thoracentesis were considered. A simultaneous sample of serum was drawn at the time of thoracentesis to measure glucose, total protein, albumin, cholesterol, and LDH levels. Further studies, including pleural biopsy, were done at the discretion of the primary physician. Biochemical parameters were determined using a selective discrete multichannel analyzer (Hitachi 747; Hitachi; Tokyo, Japan). Total protein concentration was estimated using the biuret method. LDH level was measured using a kinetic ultraviolet optimized standard method conforming to the recommendations of the Deutsche Gesellschaft for Klinische Chemie; the upper normal limit (serum) is defined as 460 IU/L. Albumin pleural concentration was determined by a nephelometric method using the Array Protein System (Beckman Instruments; Brea, CA). The final clinical diagnosis was based on the results of all diagnostic studies, but blind of the pleural biochemical parameters, and the response to treatment at the end of the follow-up. All patients treated with diuretics on a daily basis were classified as the diuretic therapy group, irrespective of the dose used. Effusions were considered malignant if one of the following criteria was met: () demonstration of malignant cells at cytologic examination or in a biopsy specimen, or (2) histologically proven primary malignancy with exclusion of any other cause known to be associated with pleural effusion. A pleural effusion was considered to be parapneumonic when there was an acute febrile illness with purulent sputum and pulmonary infiltrate in the absence of malignancy or diseases causing transudates. Tuberculous pleurisy was diagnosed with a positive culture finding for Mycobacterium tuberculosis in pleural fluid, pleural biopsy, or sputum, and/or from the presence of caseous granulomas in pleural biopsy. A diagnosis of pulmonary embolus or infarction was made when there was a strong clinical suspicion and a highprobability perfusion scan or abnormal angiographic findings. CHF was determined by an enlarged heart, pulmonary venous congestion on radiograph, peripheral edema, response to CHF treatment, and the absence of malignancy or pulmonary infiltrates associated with an inflammatory process. The nephrotic syndrome was diagnosed when the patient had proteinuria, edema, and hypoalbuminemia. Effusions were considered secondary to liver cirrhosis when clinical or laboratory evidence of hepatic damage with portal hypertension or hypoalbuminemia occurred in the absence of heart failure or malignancy. The effusions secondary to diseases that directly involve pleural surfaces were considered exudates, and the rest were considered transudates. Effusions secondary to lung cancer without evidence of pleural invasion, and increased negative pleural pressure (trapped lung), were considered transudates. Because in our experience it has been the rule, effusions secondary to pulmonary embolism were considered exudates. When two diseases causing different types of pleural effusion were present in the same patient at the moment of the thoracentesis, the patient was not excluded from the study if one condition appeared to be clearly responsible for the effusion at the end of the follow-up. The following biochemical parameters were estimated and calculated: () the criteria of Light et al 6 (namely, pleural fluid/serum protein ratio, pleural fluid/serum LDH ratio, pleural fluid LDH concentration); (2) albumin gradient (serum albumin concentration minus pleural effusion albumin concentration); and (3) protein gradient (serum total protein concentration minus pleural effusion total protein concentration). 3 When separating transudates from exudates, cutoff points recommended in the literature were used. 9,0,3 The clinical presumption of the nature of the effusion (transudate or exudate) was based on all available information obtained just before performing thoracentesis, and was compared with that obtained from biochemical criteria. Two physicians (the physician who referred the patient, usually an internist or a pulmonologist, and the pulmonologist who performed the thoracentesis) were separately asked to classify the effusion as transudate or exudate. A second opinion was asked of this latter physician after performing thoracentesis. The sensitivity and specificity for all diagnostic criteria were computed using Bayes theorem and were compared using the 2 test. 4 Receiver-operating characteristic curves were generated using commercial software (SPSS for Windows, V 9.0; SPSS; Chicago, IL). To identify independent parameters of greater classification impact, we obtained a multiparametric model considering the pretest probability (clinical presumption), Light s criteria, the albumin gradient, and the protein gradient. A p value 0.05 was considered significant. Results Two hundred seventy-three patients with pleural effusion were evaluated. In 7 patients, despite extensive evaluation, the cause of the pleural effusion was indeterminate. Fourteen patients had an www.chestjournal.org CHEST / 22 / 5/ NOVEMBER, 2002 525
Table Cause of Pleural Effusions Variables No. (%) Exudates 85 (74) Malignant 8 (33) Infectious 59 (24) Parapneumonic 36 Empyema Tuberculosis Aspergillosis Pulmonary embolism 6 (2) Traumatic or postsurgery 28 () Systemic lupus erythematosus 5 (2) Asbestos exposure 3 () Pleuropericarditis 2 () Postsurgery chylothorax Transudates 64 (26) Congestive heart failure 44 (8) Cirrhosis 5 (2) Nephrotic syndrome 0 (4) Trapped lung (atelectasis) 4 (2) Ventricle-pleural shunt Total 249 exudative pleural effusion according to Light s criteria, and patients had an exudative pleural effusion according to the serum-effusion albumin gradient. These 7 patients, and additional patient (exudate by both criteria) who had an iatrogenic pneumothorax during thoracentesis, were excluded from the study. Six patients in whom pleural fluid albumin measurements were not done were also excluded. The mean ( SD) age of the 249 remaining patients (5 men and 98 women) with an effusion of known cause was 6 7 years (range, 4 to 90 years). The final diagnosis is shown in Table. The effusions were definitively classified as transudates in 64 patients (26%) and as exudates in 85 patients (74%). In three patients with atelectasis due to lung cancer, and in one patient with a trapped lung secondary to tuberculosis, the effusion was considered a transudate. The suspected causes to indicate a thoracentesis in the 44 patients with CHF were as follows: fever, patients; no improvement of the effusion with diuretic therapy, patients; risk of venous thrombosis, 7 patients; malignancy, 5 patients; other associated diseases, 7 patients (bronchiectasis [n 3], AIDS [n 2], bronchial asthma [n ], scleroderma [n ]); presence of a focal pulmonary infiltrate, 2 patients; and previous exudative effusion of unknown etiology, patient. Table 2 shows the percentage of correctly classified transudates and exudates and the accuracy of each criteria used. Light s criteria showed a significant higher sensitivity and accuracy than serumeffusion albumin and protein gradients. Sixteen patients (25%) with transudates (2 patients receiving diuretic therapy) were misclassified as exudates by Light s criteria. The albumin gradient misclassified nine patients (4%) with transudative effusions (five patients receiving diuretic therapy) and the protein gradient in six patients (four patients receiving diuretic therapy). When using the criteria proposed by Light et al, not only the accuracy but the sensitivity and specificity as well were significantly higher (p 0.05 to p 0.0) than that showed by each initial clinical presumption (simple concordance index between observers, 94%; index, 0.80). The physician who performed the thoracentesis changed his initial presumption 38 times (5%), but the accuracy did not change. Seventy-two patients were receiving diuretic therapy when the thoracocentesis was performed. In these patients, the accuracy of Light s criteria was 83% (60 of 72 cases), 88% for albumin gradient (63 of 72 cases), and 86% for protein gradient (62 of 72 cases). The differences were not significant (Table 3). When, following one recommendation, alternative biochemical criteria (albumin or protein gradient) were applied to those patients receiving diuretics thought to have a transudative effusion by clinical criteria but the fluid was identified as exudative by Light s criteria, the accuracy improved but not sig- Table 2 Transudates and Exudates Correctly Classified by Different Criteria in 249 Patients* Criteria Sensitivity, Exudates Specificity, Transudates Accuracy Light s criteria 84/85 (99.5) 48/64 (75) 232/249 (93) Albumin gradient 62/85 (88) 55/64 (86) 27/249 (87) Protein gradient 55/85 (84) 58/64 (9) 23/249 (86) Clinical presumption A 73/85 (94) 36/64 (56) 209/249 (84) Clinical presumption B 72/85 (93) 37/64 (58) 209/249 (84) Clinical presumption plus postthoracentesis 69/85 (9) 39/64 (6) 208/249 (84) *Data are presented as No. of correct classifications/total patients (%). Significantly lower (p 0.00) with respect to Light s criteria. Significantly lower (p 0.0) with respect to Light s criteria. Significantly higher (p 0.03) with respect to Light s criteria. Significantly lower (p 0.05) with respect to Light s criteria. 526 Clinical Investigations
Table 3 Transudates and Exudates Correctly Classified by Different Criteria in 72 Patients Receiving Diuretic Therapy* Criteria Sensitivity, Exudates Specificity, Transudates Accuracy Light s criteria 3/3 (00) 29/4 (7) 60/72 (83) Albumin gradient 27/3 (87) 36/4 (88) 63/72 (88) Protein gradient 26/3 (84) 36/4 (88) 62/72 (86) Clinical presumption A 26/3 (84) 27/4 (66) 53/72 (74) Clinical presumption B 26/3 (84) 27/4 (66) 53/72 (74) Clinical presumption plus postthoracentesis 29/3 (94) 26/4 (63) 55/72 (76) *Data are presented as No. of correct classifications/total patients (%). Significantly lower (p 0.0) with respect to Light s criteria. nificantly (Table 4). There were no significant differences between the area under the receiver-operating characteristic curve of protein and albumin serumpleural fluid gradients with respect to final diagnosis for the overall population (0.946 vs 0.907), for the population not receiving diuretics (0.962 vs 0.899), and for the population receiving diuretics (0.94 vs 0.935, respectively). Using a multiparametric model (Table 5), the accuracy for the overall population was 96% (235 of 244 cases), not significantly different from that obtained by Light s criteria alone (93%). Independent parameters with the greater classification impact were as follows: Light s criteria (p 0.00), clinical presumption B (p 0.00), and protein gradient (p 0.00). For the population not receiving diuretics, the accuracy was 97% (7 of 77 cases), also similar to that obtained by Light s criteria (97%; 72 of 77 cases). The independent parameter with the greater classification impact was Light s criteria (p 0.00). For the population receiving diuretics, the multiparametric model had an accuracy of 93% (67 of 72 cases), higher but without statistical significance (p 0.2) than that obtained by Light s criteria alone (83%; 60 of 72 cases). The independent parameters with the greater classification impact were as follows: clinical presumption B (p 0.022) and protein gradient (p 0.032). Deleting Light s criteria in this last population, the accuracy of the multiparametric model did not change (93%; 67 of 72 cases). Discussion Is it meaningful to use biochemical parameters to discriminate between transudative and exudative pleural effusions? The answer to this question has been traditionally considered affirmative; however, as far as we know, the present study has been the first designed to address this question. The results of a preliminary analysis that included the first 08 patients of this series 5 cast some doubts about the superiority of the biochemical criteria over the clinical presumption to separate transudative from exudative pleural effusions. However, once extended with an appropriate number of patients, the present study shows a significant superiority of Light s criteria (the biochemical criteria most commonly used) over clinical presumption or the other biochemical criteria herein tested. It is important to emphasize, however, that Light s criteria did not provide perfect separation for these diagnostic categories despite its superior discriminatory power. Using Light s criteria, 5 to 30% of transudates are misclassified as exudates (25% in the Table 4 Transudates and Exudates Correctly Classified Using Albumin and Protein Gradients Combined With Clinical Presumption in 72 Patients Receiving Diuretic Therapy* Criteria Sensitivity, Exudates Specificity, Transudates Accuracy p value Light s criteria 3/3 (00) 29/4 (7) 60/72 (83) Albumin gradient plus Clinical presumption A 30/3 (97) 33/4 (80) 63/72 (88) 0.54 Clinical presumption B 30/3 (97) 32/4 (78) 62/72 (86) 0.79 Clinical presumption plus postthoracentesis 3/3 (00) 33/4 (80) 64/72 (89) 0.43 Protein gradient plus Clinical presumption A 29/3 (94) 34/4 (83) 63/72 (88) 0.54 Clinical presumption B 30/3 (97) 34/4 (83) 64/72 (89) 0.43 Clinical presumption plus postthoracentesis 3/3 (00) 34/4 (83) 65/72 (90) 0.33 *Data are presented as No. of correct classifications/total patients (%). www.chestjournal.org CHEST / 22 / 5/ NOVEMBER, 2002 527
Table 5 Data Obtained Using a Multiparametric Model* Variables Odds Ratio Confidence Interval p Value Overall population Clinical presumption 3.32 22.9 5. 03.9 0.00 Light s criteria 4.237 69.2 7.5 635.3 0.00 Protein gradient.822 0.2 0. 0.5 0.00 Albumin gradient 0.00.0.0.0 0.5 Coefficient 2.656 4.2 0.8 EXP [0 ((3.32 cp) (4.237 lp) (.822 pg) (0.00 ag) 2.656)] Population not receiving diuretics Clinical presumption 3.40 30.0. 803.9 0.05 Light s criteria 5.035 53.7.3 2098.6 0.00 Protein gradient 2.40 0. 0.0 0.6 0.0 Albumin gradient 0.00.0.0.0 NS Coefficient 4.256 92.4 0.0 EXP [0 ((3.40 cp) (5.035 lp) (2.40 pg) (0.00 ag) 4.256)] Population receiving diuretics Clinical presumption 2.632 3.9.5 3.9 0.05 Light s criteria 8.246 38.0 0.0 NS Protein gradient 2.0 0. 0.0 0.8 0.05 Albumin gradient 0.003.0.0.0 0.05 Coefficient 0.64 0.9 NS EXP [0 ((2.632 cp) (8.246 lp) (2.0 pg) (0.003 ag) 0.64)] Population receiving diuretics (excluding Light s criteria) Clinical presumption (cp) 2.780 6..7 55.4 0.05 Protein gradient (pg) 2.92 0. 0.02 0.7 0.05 Albumin gradient (ag) 0.004.0.0.0 0.05 Coefficient 9.442 260.5 0.0 EXP [0 ((2.780 cp) (2.92 pg) (0.004 ag) 9.442)] *cp clinical presentation; lp Light s criteria; pg protein gradient; ag albumin gradient; EXP e (the base of the natural logarithms). present study). Apart from the potential presence of unsuspected exudative disease, most cases of false exudative effusion by Light s criteria occur in patients receiving diuretic therapy, 6 as our results also confirm. To overcome that limitation, the use of alternative criteria, such as albumin and protein gradients in patients receiving diuretic therapy, has been recommended. 0,3 Our results also show that the indiscriminate use of these alternative criteria in all patients receiving diuretic therapy does not improve the accuracy. Because only some selected patients could benefit from this approach, it has been recommended to use these alternative biochemical criteria only in patients receiving diuretics who are thought to have a transudative effusion by clinical criteria but the fluid is identified as exudative by Light s criteria. Following that recommendation, the accuracy improved but without reaching statistical significance. Neither did the use of a multiparametric analysis (that included clinical presumption, together with total protein and albumin serum-pleural fluid gradients) change the results obtained following the previous recommendation significantly. These results show that the use of clinical judgment is of limited value for doing the selection. As became evident in the present study, clinical judgment, as Light s criteria, misclassify a higher proportion of transudates than exudates; similarly, a high percentage of those patients misclassified were receiving diuretic treatment. One likely explanation for these results is the frequent suspicion, not always confirmed, of a nonapparent cause of exudate, when a patient, initially thought to have a transudate, fails to respond to diuretic therapy. All of this decreases the added value of the above-mentioned combined approaches. However, because the number of patients receiving diuretics herein included was limited and some combined criteria showed a trend toward statistical significance, we cannot rule out that it could be reached with a higher number of patients. The use of an alternative biochemical criteria, however, in no case must convey an additional cost; as the results of the present study show, the protein gradient, with a similar accuracy to that of albumin but without requiring the determination of any 528 Clinical Investigations
additional parameter to those used for calculating Light s criteria, would appear to be the alternative of choice. Other parameters, such as cholesterol or bilirubin levels, have not been included in the present analysis because previous comparative studies performed in our clinical setting 7 demonstrated that they are not superior to Light s criteria in the separation of transudates and exudates. The use of sensitive inflammatory markers as soluble leukocyte selectin may require further analysis. 8 In conclusion, Light s criteria add significant information to that provided by clinical judgment for classifying pleural effusions in transudates and exudates. In patients receiving diuretics, Light s criteria lose accuracy that is not significantly lower than that showed by alternative biochemical criteria. The recommended use of alternative biochemical parameters together with clinical judgement in patients receiving diuretics results in a better accuracy than that of Light s criteria, but without statistical significance. References Peterman TA, Speicher CE. Evaluating pleural effusions: a two-stage laboratory approach. JAMA 984; 252:05 053 2 Good JT, Taryle DA, Maulitz RM, et al. The diagnostic value of pleural fluid ph. Chest 990; 78:55 59 3 Bartter T, Santarelli R, Akers SM, et al. The evaluation of pleural effusion. Chest 994; 06:209 24 4 Sahn SA. Malignancy metastatic to the pleura. Clin Chest Med 998; 9:35 36 5 Fernández C, Martín C, Aranda I, et al. Malignant transient pleural transudate: a sign of early lymphatic tumoral obstruction. Respiration 2000; 67:333 336 6 Light RW, Macgregor MI, Luchsinger PC, et al. Pleural effusions: the diagnostic separation of transudates and exudates. Ann Intern Med 972; 77:507 53 7 Hamm H, Brohan U, Bohmer R, et al. Cholesterol in pleural effusions: a diagnostic aid. Chest 987; 92:296 302 8 Valdes L, Pose A, Suarez J, et al. Cholesterol: a useful parameter for distinguishing between pleural transudates and exudates. Chest 99; 99:097 02 9 Romero S, Candela A, Martín C, et al. Evaluation of different criteria for the separation of pleural transudates from exudates. Chest 993; 04:399 404 0 Roth BJ, O Meara TF, Cragun WH. The serum-effusion albumin gradient in the evaluation of pleural effusions. Chest 990; 98:546 549 Light RW. Useful tests on the pleural fluid in the management of patients with pleural effusions. Curr Opin Pulm Med 999; 5:245 249 2 Carlenton RA. Dichotomous disservice? Ann Intern Med 997; 26:589 59 3 Romero S, Fernández C, Martín C, et al. Influence of diuretics on the concentration of proteins and other components of pleural transudates in patients with heart failure. Am J Med 200; 0:68 686 4 Rosner B. Fundamentals of biostatistics, 4th ed. Belmont, CA: Duxbury Press, 995; 4 85 5 Fernández C, Orts D, Sánchez J, et al. Is it meaningful to use biochemical parameters to discriminate between transudative and exudative pleural effusions [abstract]? Am J Respir Crit Care Med 2000; 6:A69 6 Burgess LJ, Maritz FJ, Taljaard JJF. Comparative analysis of the biochemical parameters used to distinguish between pleural transudates and exudates. Chest 995; 07:604 609 7 Romero S, Martinez A, Hernandez L, et al. Light s criteria revisited: consistency and comparison with new proposed alternative criteria for separating pleural transudates from exudates. Respiration 2000; 67:333 336 8 Longmore LH, Gallup RA, Worley BD, et al. Soluble leukocyte selectin in the analysis of pleural effusions. Chest 200; 20:362 368 www.chestjournal.org CHEST / 22 / 5/ NOVEMBER, 2002 529