The Serum-Ascites Albumin Gradient Is Superior to the Exudate-Transudate Concept in the Differential Diagnosis of Ascites Bruce A. Runyon, MD; Agnes A. Montano, MD; Evangelos A. Akriviadis, MD; Mainor R. Antillon, MD; Michelle A. Irving, RN; and John G. McHutchison, MD
• Objective: To compare the serum-ascites albumin gradient to the exudate-transudate concept in the classification of ascites. • Design: Prospective collection of ascitic fluid data from patients with well-characterized causes of ascites. • Setting: Hepatology inpatient and outpatient ward and consult service of a large, urban hospital. • Patients: A total of 901 paired serum and ascitic fluid samples were collected from consecutive patients with all forms of ascites. • Interventions: None. • Main Outcome Measures: The utility of the serumascites albumin gradient and the old exudatetransudate concept (as defined by ascitic fluid total protein concentration [AFTP]) were compared for their ability in discriminating the cause for ascites formation. • Results: The albumin gradient correctly differentiated causes of ascites due to portal hypertension from those that were not due to portal hypertension 96.7% of the time. The AFTP, when used as defined in the old exudate-transudate concept, classified the causes of ascites correctly only 55.6% of the time. This resulted in part because the AFTP of most spontaneously infected samples (traditionally expected to be exudates) was low, and the AFTP of most cardiac ascites samples (traditionally expected to be transudates) was high. • Conclusions: The exudate-transudate concept should be discarded in the classification of ascites. The serum-ascites albumin gradient is far more useful than the AFTP as a marker for portal hypertension, but the latter remains a useful adjunct in the differential diagnosis of ascites.
Annals of Internal Medicine. 1992;117:215-220. From the University of Iowa, Iowa City, Iowa; and the University of Southern California, Los Angeles, California. For current author addresses, see end of text.
I raditionally, ascitic fluid has been routinely tested for cell count, bacterial culture, and total protein concentration. The cell count, particularly the absolute neutrophil count, is useful in the presumptive diagnosis of ascitic fluid infection (1). The culture has been used to confirm the presence of bacteria, resulting in continuation of empiric antibiotic treatment (1). The ascitic fluid total protein concentration (AFTP) has been used to classify specimens into broad categories of exudate or transudate (2-5). Samples were traditionally classified as exudates if the AFTP was 25 g/L or more and as transudates if the AFTP was less than 25 g/L; occasionally, 30 g/L was used as the level of discrimination (2-5). This exudate-transudate concept was based on the assumption that fluid that formed by "exudation" from an inflamed or tumor-laden peritoneal surface (for example, bacterial peritonitis, tuberculous peritonitis, pancreatitis, and peritoneal carcinomatosis) was high in protein. Fluid that "transuded" from a normal peritoneal surface because of an imbalance of Starling forces, as in cirrhosis, heart failure, and the nephrotic syndrome, was assumed to be low in protein. Many problems and exceptions have been noted, however, with the exudate-transudate concept. Many infected or malignancy-related samples have been reported to have protein concentrations in the transudate range, and many samples obtained from patients with cirrhosis or heart failure have had concentrations in the exudate range (Table 1) (5-10). Normal peritoneal fluid obtained from healthy women has an AFTP greater than 40 g/L, well into the exudative range (11, 12). Therefore, normal peritoneal fluid in the absence of infection, inflammation, or tumor is "exudative," contrary to the assumptions of the exudate-transudate concept. The AFTP appears to be relatively independent of peritoneal permeability and is determined almost completely by serum protein concentration (direct relationship) and by portal pressure (inverse relationship) (13). Also, the exudate-transudate concept makes no provision for those patients who have "mixed" ascites; that is, portal hypertension plus another cause of ascites formation (14). For example, 11% of patients with ascites and peritoneal carcinomatosis had underlying cirrhosis in one series (8), and 43% of patients with peritoneal tuberculosis had underlying cirrhosis in another series (15). In contrast, the serum-ascites albumin gradient (defined as the serum albumin concentration minus the ascitic fluid albumin concentration) has been proposed as a physiologically based alternative in the classification of ascites (Table 2) (13, 14, 17, 18). This "gradient" has been shown to correlate directly with portal © 1992 American College of Physicians
Downloaded From: http://annals.org/ by a University of Otago User on 12/14/2016
215
Table 1. Reported Utility of the Ascitic Fluid Total Protein Concentration in Classifying Samples as Exudates or Transudates* Variablet
AFTP > 25 g/L < 25 g/L
% Specimens traditionally predicted to be exudates Spontaneously infected cirrhotic ascites (273) Secondary bacterial peritonitis in patients with ascites (15) Malignancy-related (117) Tuberculous peritonitis (47) Specimens predicted to be transudates Sterile cirrhotic (202) Sterile cirrhotic at end of diuresis (44) Cardiac (33) Normal peritoneal fluid (132)
0 to 6 33 58 53 to 75 73 to 95 33 to 58 0 0
* AFTP = ascitic fluid total protein concentration. From references 6-25. t The numbers in parentheses are the numbers of samples tested.
p r e s s u r e such that patients with gradients of 11 g/L or greater h a v e b e e n s h o w n to have portal h y p e r t e n s i o n , w h e r e a s those with gradients less than 11 g/L d o not h a v e the disorder (13). B e c a u s e of the p r e s e n c e of portal h y p e r t e n s i o n , patients with mixed ascites h a v e a high gradient (14). Our s t u d y w a s c o n d u c t e d in a large n u m b e r of patients with all forms of ascites and w a s designed to determine w h e t h e r the serum-ascites albumin gradient is superior to the e x u d a t e - t r a n s u d a t e c o n c e p t (as defined b y the A F T P ) in the differential diagnosis of ascites. Methods Serum and ascitic fluid were obtained simultaneously (maximum interval 30 minutes) and were tested for albumin concentration using the bromcresol green method (26), and for AFTP by biuret reaction. Manual cell counts and differentials, cultures (in blood culture bottles), cytologic examinations, and tests for tuberculosis were done as previously described (27, 28). Specimens were collected serially in patients during follow-up such that specimens obtained before and during infec-
tion, at the beginning and end of diuresis, and at the beginning and end of a series of therapeutic paracenteses were tested and compared. Specimens were classified according to the presence or absence of portal hypertension for subsequent comparison with the albumin gradient. Specimens were categorized as related to portal hypertension if the wedged hepatic venous pressure minus inferior vena cava pressure was greater than 4 mm Hg; if varices were seen during endoscopy or laparoscopy; if liver biopsy samples showed cirrhosis, alcoholic hepatitis, or both; or if laparotomy or postmortem examination showed collaterals, chronic parenchymal liver disease, or both. Specimens were classified as unrelated to portal hypertension if portal pressure was normal, if no collaterals were present, and if no parenchymal liver disease was noted. The portal-hypertension-related group included sterile cirrhotic and infected cirrhotic samples as well as other samples, including mixed ascites, obtained from patients with portal hypertension. The sterile cirrhotic category included all sterile non-neutrocytic specimens (neutrophil concentration, < 0.25 x 109 cells/L) from patients with documented portal hypertension due to chronic parenchymal liver disease. The infected cirrhotic ascites category included all variants of peritonitis occurring in the presence of cirrhosis. The diagnosis of peritonitis was based on bacterial growth in an ascitic fluid culture or on an ascitic fluid neutrophil concentration of at least 0.25 x 109 cells/L. Subtypes of peritonitis included in the diagnosis were spontaneous bacterial peritonitis (1), culture-negative neutrocytic ascites (20), monomicrobial nonneutrocytic bacterascites (21), polymicrobial bacterascites (22), and secondary bacterial peritonitis (23). Mixed ascites was diagnosed when portal hypertension accompanied another cause of ascites formation, such as cirrhosis plus superimposed peritoneal carcinomatosis or cirrhosis plus peritoneal tuberculosis (14). Our 901 specimens were also classified into traditional exudate and transudate categories for subsequent comparison with AFTP. "Exudative" causes of ascites included malignancy, pancreatitis, tuberculosis, and peritonitis. Patients with mixed ascites who had an exudative cause of ascites formation were categorized as exudative. "Transudative" causes included sterile cirrhotic ascites, cardiac ascites, nephrotic syndrome, nephrogenous ascites, and myxedema ascites. Malignancy-related samples included peritoneal carcinomatosis, massive liver metastases, lymphoma, and hepatocellular carcinoma as determined by tumor location assessed at laparoscopy, laparotomy, or autopsy (8). The diagnosis of tuberculous peritonitis required mycobacterial growth in a culture of ascitic fluid or peritoneal biopsy with or without histologically proven granulomatous peritonitis (15). Pancreatic ascites was diagnosed when the ascitic fluid amylase level was at least twice that of the upper limit of normal for serum (19). Cardiac ascites was diagnosed when cardiac catheterization, ultrasound, or radionuclide ventriculogram showed heart failure (10). Nephrotic and nephrogenous ascites were diagnosed in patients with proven nephrotic syndrome or dialysis-related
Table 2. Reported Utility of the Serum-Ascites Albumin Gradient in Classifying Samples as Portal-Hypertension Related or Nonportal-Hypertension Related* Albumin Gradient > 11 g/L
Variablet
Albumin Gradient < 11 g/L
% Specimens predicted to be associated with portal hypertension Sterile cirrhotic (202) Cardiac (28) Massive liver metastases (20) Hepatocellular carcinoma (6) "Mixed" (12) Specimens predicted to be associated with normal portal pressure Peritoneal carcinomatosis (54) Tuberculous peritonitis without cirrhosis (1) * From references 13, 14, 17, 18. t The numbers in parentheses are the numbers of samples tested.
216
1 August 1992 • Annals of Internal Medicine
• Volume 117 • Number 3
Downloaded From: http://annals.org/ by a University of Otago User on 12/14/2016
84 to 97 80 to 100 100 83 71 to 100 93 to 100 100
Figure 1. Serum-ascites albumin gradient in ascites, classified by presence or absence of portal hypertension. Statistical comparisons are to the sterile cirrhotic group by unpaired Mest. CARD = cardiac ascites; Misc Non-PHT = miscellaneous-nonportalhypertension-related; Misc PHT = miscellaneous portal-hypertension-related; NS = not significant; PCA = peritoneal carcinomatosis. Mean ± SD bars are included as well as a horizontal line at 11 g/L, the threshold for portal hypertension. All groups differed significantly (P < 0.05) from the sterile cirrhotic samples by analysis of variance with the Dunnette test.
ascites, respectively (29). Myxedema ascites was diagnosed when abdominal fluid collected in patients with hypothyroidism (18). These specimens were classified in specific disease categories independent of albumin gradient or AFTP. We assessed the accuracy of a serum-ascites albumin gradient of at least 11 g/L in detecting portal hypertension and of an AFTP of at least 25 g/L in detecting an exudate. Accuracy was calculated as the sum of true positive results and true negative results divided by the total number of samples (30). This study was approved by our institutional review board. Participants signed forms indicating informed consent. Paired and unpaired Student Mests, the chi-square test, and ANOVA supplemented with the Dunnette test were used for statistical analysis. Data are presented as mean ± SD. A P value of less than 0.05 was considered statistically significant.
Results Of 1275 paired serum and ascitic fluid samples, 901 were obtained from 330 patients with a confirmed diagnosis. These 901 specimens were collected between February of 1983 and October of 1989 and are the subject of this analysis. The data for albumin gradient and AFTP are shown in Figures 1 and 2, respectively. The details of the classification of the specimens of Figure 1 are shown in Table 3. In our study sample, where 81.2% of the patients with ascites had cirrhosis, the albumin gradient accurately identified portal hypertension 96.7% of the time; that is, 871 (96.7%) of the 901 samples fit the expected pattern of high (> 11 g/L) gradient in patients with portal hypertension and low (< 11 g/L) gradient in patients without portal hypertension (see Figure 1). In contrast, the exudate-transudate concept identified only
55.6% correctly; only 501 (55.6%) of the 901 samples fit the expected pattern of AFTP less than 25 g/L in the transudates and of 25 g/L or greater in the exudates (Figure 2). The poor accuracy of the exudate-transudate concept in classifying ascites is largely explained by the large proportion (94.6%) of infected cirrhotic samples that were low in protein and by the large proportion (100%) of cardiac samples that were high in protein. Setting aside the exudate-transudate concept, AFTP does appear to separate most sterile cirrhotic ascites samples (AFTP, < 25 g/L) from cardiac ascites samples (AFTP, > 25 g/L) (see Figure 2), contrary to the old concept. Although the AFTP appears to separate the specimens from patients with peritoneal carcinomatosis alone from most transudative specimens (see Figure 2), only 3 (23.1%) of the 13 specimens with peritoneal carcinomatosis plus another cause for ascites formation (for example, cirrhosis or massive liver metastases) had an AFTP of 25 g/L or greater. Also, only 42% of 12 specimens from patients with peritoneal tuberculosis plus cirrhosis had an AFTP of 25 g/L or greater. Overall, only 30.2% of the 43 mixed ascites specimens had an AFTP of at least 25 g/L. Therefore, the presence of portal hypertension (with or without a superimposed exudative cause of ascites formation) usually results in a low AFTP concentration. Diuresis did not affect the gradient; the serum albumin increased in direct association with the ascitic fluid albumin concentration. In the 22 cirrhotic patients for whom samples were obtained at the beginning and end of diuresis, no difference was noted between the initial
1 August 1992 • Annals of Internal Medicine • Volume 117 • Number 3
Downloaded From: http://annals.org/ by a University of Otago User on 12/14/2016
217
(18.1 ± 3.8 g/L) and the final gradient (18.3 ± 3.3 g/L). In contrast, AFTP increased dramatically (from 9.3 ± 5.1 g/L to 15.8 ± 7.5 g/L; P < 0.0001) during diuresis, as in previous reports (9, 24, 25). Similarly, therapeutic paracentesis did not affect the gradient. In 67 patients for whom specimens were obtained at the beginning and end of a series of paracenteses, the initial gradient (18.2 ± 5 . 2 g/L) did not differ from the final gradient (18.3 ± 4.1 g/L). The cause of liver disease did not appear to affect the gradient; none of the 111 samples from patients with nonalcohol-related liver disease (with or without ascitic fluid infection) had a gradient less than 11 g/L compared with 2.9% of 647 samples from patients with alcoholic liver disease (chi-square = 2.25, not significant). Discussion In this large, prospectively collected series of patients with all forms of ascites, the serum-ascites albumin gradient was found to be far superior to the exudate-
transudate concept (as defined by AFTP) in the classification of ascites; accuracy for the two methods was 96.7% and 55.6%, respectively. The marked difference in accuracy can be explained by the factors that influence these parameters. The gradient correlates directly with only one physiologic factor, the portal pressure (13). In contrast, AFTP is influenced by serum protein concentration as well as by portal pressure (13). The AFTP level is directly related to serum protein concentration but is inversely related to portal pressure (13). In patients with cirrhosis, these two factors vary widely and predictably lead to highly variable ascitic fluid total protein concentrations. The elevated albumin gradient in patients with heart failure {see Figure 1) indicates that the gradient is probably a reflection of absolute portal pressure (which is high in patients with heart failure as a result of high right-sided heart pressures) rather than the gradient between the portal pressure and the caval pressure (which is usually normal in patients with heart failure until cardiac cirrhosis occurs) (31). The albumin gradient
Figure 2. Ascitic fluid total protein concentration (AFTP) in ascites of various types, classified by transudate and exudate. Statistical comparisons are to the sterile cirrhotic group by unpaired f-test. CARD = cardiac ascites; Misc Exud = miscellaneous exudative; Misc Trans = miscellaneous transudative; NS = not significant; PCA = peritoneal carcinomatosis. The miscellaneous transudative category included fulminant hepatic failure, chylous cirrhotic ascites, and nephrotic and nephrogenous ascites. The miscellaneous exudative category included patients with mixed ascites (if one of the causes was exudative), infected fulminant hepatic failure samples, massive liver metastases, hepatocellular carcinoma, infected cardiac ascites, chylous lymphoma-related ascites, tuberculous peritonitis, pancreatitis, chlamydia peritonitis, and secondary bacterial peritonitis in the absence of cirrhosis. Mean ± SD bars are included as well as a horizontal line at 25 g/L, the threshold for exudate. All groups except the infected cirrhotic samples differed significantly (P < 0.05) from the sterile cirrhotic samples by analysis of variance with the Dunnette test. 218
1 August 1992 • Annals of Internal Medicine • Volume 117 • Number 3
Downloaded From: http://annals.org/ by a University of Otago User on 12/14/2016
classifies cirrhotic ascites in the high-albumin-gradient category along with cardiac ascites. In contrast to the altered membrane permeability associated with meningitis or pleural infection, the permeability of the peritoneum to protein does not appear to change in the setting of spontaneously infected ascites (7). Low-protein ascites is a risk factor for the development of ascitic fluid infection, and the protein concentration does not change with the development of infection (6, 7). The fact that infected ascites is low in protein (when the exudate-transudate concept predicts it should be high in protein) explains a substantial portion of the inaccuracy of the AFTP in detecting exudative causes of ascites. The albumin gradient appears to retain its accuracy despite diuresis or therapeutic paracentesis, regardless of the cause of liver disease. In contrast to the results of Kajani and colleagues, nearly all our patients with alcoholic liver disease as well as those with nonalcoholic liver disease were found to have albumin gradients of at least 11 g/L (32). Because the serum-ascites albumin gradient is markedly superior to the AFTP concentration, the albumin gradient should replace the total protein level as the initial factor used to classify ascites. Ascitic fluid samples should be characterized as high gradient or low gradient rather than as transudate or exudate. In specific circumstances, the AFTP (used apart from the exudate-transudate concept) is of some value. Patients with low AFTP are known to be at high risk for spontaneous peritonitis (6). Prophylactic, poorly absorbed oral antibiotics have been shown to selectively decontaminate the gut and therefore to help prevent bacterial infections in this subgroup of patients, who are identified by AFTP (33). Also, AFTP can be helpful in discriminating spontaneous from secondary bacterial peritonitis (for example, ruptured viscus into ascitic fluid) (6, 23). Cardiac ascites is characterized as high in protein with a high albumin gradient (see Figures 1 and 2); a high AFTP level may suggest the possibility of heart failure as the cause of ascites in a patient with a high albumin gradient (10). Neither AFTP nor albumin gradient changed dramatically in this study when ascitic fluid infection developed, and neither can replace ascitic fluid cell count and culture in detecting bacterial infection. Similarly, neither AFTP nor albumin gradient can replace ascitic fluid cytologic or culture results for mycobacteria in confirming suspected peritoneal carcinomatosis or tuberculosis, respectively. In addition to assisting with classification of patients with ascites, the albumin gradient may be of value in predicting response to treatment. Patients with ascites related to portal hypertension (for example, those with cirrhosis) usually respond to dietary sodium restriction and diuretics (34). In contrast, patients with ascites unrelated to portal hypertension (for example, those with peritoneal carcinomatosis) are refractory to diuretic therapy (35). Because the albumin gradient is accurate in detecting portal hypertension, studies of its value in predicting those patients who will respond to treatment are needed.
Table 3 . Types of Ascitic Fluid Samples Collected Type
Samples n (%)*
Cirrhotic (with or without infection) Cardiac Miscellan eous portal-hypertension related Mixed Cirrh osis and tuberculous peritonitis Cirrh osis and hepatocellular carcinoma Mass ive liver metastases and peritoneal car cinomatosis Cirrh osis and peritoneal carcinomatosis Cirrh osis and pancreatitis Cirrh osis and the nephrotic syndrome Cirrh osis and chlamydia peritonitis Mass ive liver metastases and portal vein thr ombosis Fulminj ant hepatic failure Acute rlepatitis superimposed on cirrhosis Massiv
• Objective: To compare the serum-ascites albumin gradient to the exudate-transudate concept in the classification of ascites. • Design: Prospective collection of ascitic fluid data from patients with well-characterized causes of ascites. • Setting: Hepatology inpatient and outpatient ward and consult service of a large, urban hospital. • Patients: A total of 901 paired serum and ascitic fluid samples were collected from consecutive patients with all forms of ascites. • Interventions: None. • Main Outcome Measures: The utility of the serumascites albumin gradient and the old exudatetransudate concept (as defined by ascitic fluid total protein concentration [AFTP]) were compared for their ability in discriminating the cause for ascites formation. • Results: The albumin gradient correctly differentiated causes of ascites due to portal hypertension from those that were not due to portal hypertension 96.7% of the time. The AFTP, when used as defined in the old exudate-transudate concept, classified the causes of ascites correctly only 55.6% of the time. This resulted in part because the AFTP of most spontaneously infected samples (traditionally expected to be exudates) was low, and the AFTP of most cardiac ascites samples (traditionally expected to be transudates) was high. • Conclusions: The exudate-transudate concept should be discarded in the classification of ascites. The serum-ascites albumin gradient is far more useful than the AFTP as a marker for portal hypertension, but the latter remains a useful adjunct in the differential diagnosis of ascites.
Annals of Internal Medicine. 1992;117:215-220. From the University of Iowa, Iowa City, Iowa; and the University of Southern California, Los Angeles, California. For current author addresses, see end of text.
I raditionally, ascitic fluid has been routinely tested for cell count, bacterial culture, and total protein concentration. The cell count, particularly the absolute neutrophil count, is useful in the presumptive diagnosis of ascitic fluid infection (1). The culture has been used to confirm the presence of bacteria, resulting in continuation of empiric antibiotic treatment (1). The ascitic fluid total protein concentration (AFTP) has been used to classify specimens into broad categories of exudate or transudate (2-5). Samples were traditionally classified as exudates if the AFTP was 25 g/L or more and as transudates if the AFTP was less than 25 g/L; occasionally, 30 g/L was used as the level of discrimination (2-5). This exudate-transudate concept was based on the assumption that fluid that formed by "exudation" from an inflamed or tumor-laden peritoneal surface (for example, bacterial peritonitis, tuberculous peritonitis, pancreatitis, and peritoneal carcinomatosis) was high in protein. Fluid that "transuded" from a normal peritoneal surface because of an imbalance of Starling forces, as in cirrhosis, heart failure, and the nephrotic syndrome, was assumed to be low in protein. Many problems and exceptions have been noted, however, with the exudate-transudate concept. Many infected or malignancy-related samples have been reported to have protein concentrations in the transudate range, and many samples obtained from patients with cirrhosis or heart failure have had concentrations in the exudate range (Table 1) (5-10). Normal peritoneal fluid obtained from healthy women has an AFTP greater than 40 g/L, well into the exudative range (11, 12). Therefore, normal peritoneal fluid in the absence of infection, inflammation, or tumor is "exudative," contrary to the assumptions of the exudate-transudate concept. The AFTP appears to be relatively independent of peritoneal permeability and is determined almost completely by serum protein concentration (direct relationship) and by portal pressure (inverse relationship) (13). Also, the exudate-transudate concept makes no provision for those patients who have "mixed" ascites; that is, portal hypertension plus another cause of ascites formation (14). For example, 11% of patients with ascites and peritoneal carcinomatosis had underlying cirrhosis in one series (8), and 43% of patients with peritoneal tuberculosis had underlying cirrhosis in another series (15). In contrast, the serum-ascites albumin gradient (defined as the serum albumin concentration minus the ascitic fluid albumin concentration) has been proposed as a physiologically based alternative in the classification of ascites (Table 2) (13, 14, 17, 18). This "gradient" has been shown to correlate directly with portal © 1992 American College of Physicians
Downloaded From: http://annals.org/ by a University of Otago User on 12/14/2016
215
Table 1. Reported Utility of the Ascitic Fluid Total Protein Concentration in Classifying Samples as Exudates or Transudates* Variablet
AFTP > 25 g/L < 25 g/L
% Specimens traditionally predicted to be exudates Spontaneously infected cirrhotic ascites (273) Secondary bacterial peritonitis in patients with ascites (15) Malignancy-related (117) Tuberculous peritonitis (47) Specimens predicted to be transudates Sterile cirrhotic (202) Sterile cirrhotic at end of diuresis (44) Cardiac (33) Normal peritoneal fluid (132)
0 to 6 33 58 53 to 75 73 to 95 33 to 58 0 0
* AFTP = ascitic fluid total protein concentration. From references 6-25. t The numbers in parentheses are the numbers of samples tested.
p r e s s u r e such that patients with gradients of 11 g/L or greater h a v e b e e n s h o w n to have portal h y p e r t e n s i o n , w h e r e a s those with gradients less than 11 g/L d o not h a v e the disorder (13). B e c a u s e of the p r e s e n c e of portal h y p e r t e n s i o n , patients with mixed ascites h a v e a high gradient (14). Our s t u d y w a s c o n d u c t e d in a large n u m b e r of patients with all forms of ascites and w a s designed to determine w h e t h e r the serum-ascites albumin gradient is superior to the e x u d a t e - t r a n s u d a t e c o n c e p t (as defined b y the A F T P ) in the differential diagnosis of ascites. Methods Serum and ascitic fluid were obtained simultaneously (maximum interval 30 minutes) and were tested for albumin concentration using the bromcresol green method (26), and for AFTP by biuret reaction. Manual cell counts and differentials, cultures (in blood culture bottles), cytologic examinations, and tests for tuberculosis were done as previously described (27, 28). Specimens were collected serially in patients during follow-up such that specimens obtained before and during infec-
tion, at the beginning and end of diuresis, and at the beginning and end of a series of therapeutic paracenteses were tested and compared. Specimens were classified according to the presence or absence of portal hypertension for subsequent comparison with the albumin gradient. Specimens were categorized as related to portal hypertension if the wedged hepatic venous pressure minus inferior vena cava pressure was greater than 4 mm Hg; if varices were seen during endoscopy or laparoscopy; if liver biopsy samples showed cirrhosis, alcoholic hepatitis, or both; or if laparotomy or postmortem examination showed collaterals, chronic parenchymal liver disease, or both. Specimens were classified as unrelated to portal hypertension if portal pressure was normal, if no collaterals were present, and if no parenchymal liver disease was noted. The portal-hypertension-related group included sterile cirrhotic and infected cirrhotic samples as well as other samples, including mixed ascites, obtained from patients with portal hypertension. The sterile cirrhotic category included all sterile non-neutrocytic specimens (neutrophil concentration, < 0.25 x 109 cells/L) from patients with documented portal hypertension due to chronic parenchymal liver disease. The infected cirrhotic ascites category included all variants of peritonitis occurring in the presence of cirrhosis. The diagnosis of peritonitis was based on bacterial growth in an ascitic fluid culture or on an ascitic fluid neutrophil concentration of at least 0.25 x 109 cells/L. Subtypes of peritonitis included in the diagnosis were spontaneous bacterial peritonitis (1), culture-negative neutrocytic ascites (20), monomicrobial nonneutrocytic bacterascites (21), polymicrobial bacterascites (22), and secondary bacterial peritonitis (23). Mixed ascites was diagnosed when portal hypertension accompanied another cause of ascites formation, such as cirrhosis plus superimposed peritoneal carcinomatosis or cirrhosis plus peritoneal tuberculosis (14). Our 901 specimens were also classified into traditional exudate and transudate categories for subsequent comparison with AFTP. "Exudative" causes of ascites included malignancy, pancreatitis, tuberculosis, and peritonitis. Patients with mixed ascites who had an exudative cause of ascites formation were categorized as exudative. "Transudative" causes included sterile cirrhotic ascites, cardiac ascites, nephrotic syndrome, nephrogenous ascites, and myxedema ascites. Malignancy-related samples included peritoneal carcinomatosis, massive liver metastases, lymphoma, and hepatocellular carcinoma as determined by tumor location assessed at laparoscopy, laparotomy, or autopsy (8). The diagnosis of tuberculous peritonitis required mycobacterial growth in a culture of ascitic fluid or peritoneal biopsy with or without histologically proven granulomatous peritonitis (15). Pancreatic ascites was diagnosed when the ascitic fluid amylase level was at least twice that of the upper limit of normal for serum (19). Cardiac ascites was diagnosed when cardiac catheterization, ultrasound, or radionuclide ventriculogram showed heart failure (10). Nephrotic and nephrogenous ascites were diagnosed in patients with proven nephrotic syndrome or dialysis-related
Table 2. Reported Utility of the Serum-Ascites Albumin Gradient in Classifying Samples as Portal-Hypertension Related or Nonportal-Hypertension Related* Albumin Gradient > 11 g/L
Variablet
Albumin Gradient < 11 g/L
% Specimens predicted to be associated with portal hypertension Sterile cirrhotic (202) Cardiac (28) Massive liver metastases (20) Hepatocellular carcinoma (6) "Mixed" (12) Specimens predicted to be associated with normal portal pressure Peritoneal carcinomatosis (54) Tuberculous peritonitis without cirrhosis (1) * From references 13, 14, 17, 18. t The numbers in parentheses are the numbers of samples tested.
216
1 August 1992 • Annals of Internal Medicine
• Volume 117 • Number 3
Downloaded From: http://annals.org/ by a University of Otago User on 12/14/2016
84 to 97 80 to 100 100 83 71 to 100 93 to 100 100
Figure 1. Serum-ascites albumin gradient in ascites, classified by presence or absence of portal hypertension. Statistical comparisons are to the sterile cirrhotic group by unpaired Mest. CARD = cardiac ascites; Misc Non-PHT = miscellaneous-nonportalhypertension-related; Misc PHT = miscellaneous portal-hypertension-related; NS = not significant; PCA = peritoneal carcinomatosis. Mean ± SD bars are included as well as a horizontal line at 11 g/L, the threshold for portal hypertension. All groups differed significantly (P < 0.05) from the sterile cirrhotic samples by analysis of variance with the Dunnette test.
ascites, respectively (29). Myxedema ascites was diagnosed when abdominal fluid collected in patients with hypothyroidism (18). These specimens were classified in specific disease categories independent of albumin gradient or AFTP. We assessed the accuracy of a serum-ascites albumin gradient of at least 11 g/L in detecting portal hypertension and of an AFTP of at least 25 g/L in detecting an exudate. Accuracy was calculated as the sum of true positive results and true negative results divided by the total number of samples (30). This study was approved by our institutional review board. Participants signed forms indicating informed consent. Paired and unpaired Student Mests, the chi-square test, and ANOVA supplemented with the Dunnette test were used for statistical analysis. Data are presented as mean ± SD. A P value of less than 0.05 was considered statistically significant.
Results Of 1275 paired serum and ascitic fluid samples, 901 were obtained from 330 patients with a confirmed diagnosis. These 901 specimens were collected between February of 1983 and October of 1989 and are the subject of this analysis. The data for albumin gradient and AFTP are shown in Figures 1 and 2, respectively. The details of the classification of the specimens of Figure 1 are shown in Table 3. In our study sample, where 81.2% of the patients with ascites had cirrhosis, the albumin gradient accurately identified portal hypertension 96.7% of the time; that is, 871 (96.7%) of the 901 samples fit the expected pattern of high (> 11 g/L) gradient in patients with portal hypertension and low (< 11 g/L) gradient in patients without portal hypertension (see Figure 1). In contrast, the exudate-transudate concept identified only
55.6% correctly; only 501 (55.6%) of the 901 samples fit the expected pattern of AFTP less than 25 g/L in the transudates and of 25 g/L or greater in the exudates (Figure 2). The poor accuracy of the exudate-transudate concept in classifying ascites is largely explained by the large proportion (94.6%) of infected cirrhotic samples that were low in protein and by the large proportion (100%) of cardiac samples that were high in protein. Setting aside the exudate-transudate concept, AFTP does appear to separate most sterile cirrhotic ascites samples (AFTP, < 25 g/L) from cardiac ascites samples (AFTP, > 25 g/L) (see Figure 2), contrary to the old concept. Although the AFTP appears to separate the specimens from patients with peritoneal carcinomatosis alone from most transudative specimens (see Figure 2), only 3 (23.1%) of the 13 specimens with peritoneal carcinomatosis plus another cause for ascites formation (for example, cirrhosis or massive liver metastases) had an AFTP of 25 g/L or greater. Also, only 42% of 12 specimens from patients with peritoneal tuberculosis plus cirrhosis had an AFTP of 25 g/L or greater. Overall, only 30.2% of the 43 mixed ascites specimens had an AFTP of at least 25 g/L. Therefore, the presence of portal hypertension (with or without a superimposed exudative cause of ascites formation) usually results in a low AFTP concentration. Diuresis did not affect the gradient; the serum albumin increased in direct association with the ascitic fluid albumin concentration. In the 22 cirrhotic patients for whom samples were obtained at the beginning and end of diuresis, no difference was noted between the initial
1 August 1992 • Annals of Internal Medicine • Volume 117 • Number 3
Downloaded From: http://annals.org/ by a University of Otago User on 12/14/2016
217
(18.1 ± 3.8 g/L) and the final gradient (18.3 ± 3.3 g/L). In contrast, AFTP increased dramatically (from 9.3 ± 5.1 g/L to 15.8 ± 7.5 g/L; P < 0.0001) during diuresis, as in previous reports (9, 24, 25). Similarly, therapeutic paracentesis did not affect the gradient. In 67 patients for whom specimens were obtained at the beginning and end of a series of paracenteses, the initial gradient (18.2 ± 5 . 2 g/L) did not differ from the final gradient (18.3 ± 4.1 g/L). The cause of liver disease did not appear to affect the gradient; none of the 111 samples from patients with nonalcohol-related liver disease (with or without ascitic fluid infection) had a gradient less than 11 g/L compared with 2.9% of 647 samples from patients with alcoholic liver disease (chi-square = 2.25, not significant). Discussion In this large, prospectively collected series of patients with all forms of ascites, the serum-ascites albumin gradient was found to be far superior to the exudate-
transudate concept (as defined by AFTP) in the classification of ascites; accuracy for the two methods was 96.7% and 55.6%, respectively. The marked difference in accuracy can be explained by the factors that influence these parameters. The gradient correlates directly with only one physiologic factor, the portal pressure (13). In contrast, AFTP is influenced by serum protein concentration as well as by portal pressure (13). The AFTP level is directly related to serum protein concentration but is inversely related to portal pressure (13). In patients with cirrhosis, these two factors vary widely and predictably lead to highly variable ascitic fluid total protein concentrations. The elevated albumin gradient in patients with heart failure {see Figure 1) indicates that the gradient is probably a reflection of absolute portal pressure (which is high in patients with heart failure as a result of high right-sided heart pressures) rather than the gradient between the portal pressure and the caval pressure (which is usually normal in patients with heart failure until cardiac cirrhosis occurs) (31). The albumin gradient
Figure 2. Ascitic fluid total protein concentration (AFTP) in ascites of various types, classified by transudate and exudate. Statistical comparisons are to the sterile cirrhotic group by unpaired f-test. CARD = cardiac ascites; Misc Exud = miscellaneous exudative; Misc Trans = miscellaneous transudative; NS = not significant; PCA = peritoneal carcinomatosis. The miscellaneous transudative category included fulminant hepatic failure, chylous cirrhotic ascites, and nephrotic and nephrogenous ascites. The miscellaneous exudative category included patients with mixed ascites (if one of the causes was exudative), infected fulminant hepatic failure samples, massive liver metastases, hepatocellular carcinoma, infected cardiac ascites, chylous lymphoma-related ascites, tuberculous peritonitis, pancreatitis, chlamydia peritonitis, and secondary bacterial peritonitis in the absence of cirrhosis. Mean ± SD bars are included as well as a horizontal line at 25 g/L, the threshold for exudate. All groups except the infected cirrhotic samples differed significantly (P < 0.05) from the sterile cirrhotic samples by analysis of variance with the Dunnette test. 218
1 August 1992 • Annals of Internal Medicine • Volume 117 • Number 3
Downloaded From: http://annals.org/ by a University of Otago User on 12/14/2016
classifies cirrhotic ascites in the high-albumin-gradient category along with cardiac ascites. In contrast to the altered membrane permeability associated with meningitis or pleural infection, the permeability of the peritoneum to protein does not appear to change in the setting of spontaneously infected ascites (7). Low-protein ascites is a risk factor for the development of ascitic fluid infection, and the protein concentration does not change with the development of infection (6, 7). The fact that infected ascites is low in protein (when the exudate-transudate concept predicts it should be high in protein) explains a substantial portion of the inaccuracy of the AFTP in detecting exudative causes of ascites. The albumin gradient appears to retain its accuracy despite diuresis or therapeutic paracentesis, regardless of the cause of liver disease. In contrast to the results of Kajani and colleagues, nearly all our patients with alcoholic liver disease as well as those with nonalcoholic liver disease were found to have albumin gradients of at least 11 g/L (32). Because the serum-ascites albumin gradient is markedly superior to the AFTP concentration, the albumin gradient should replace the total protein level as the initial factor used to classify ascites. Ascitic fluid samples should be characterized as high gradient or low gradient rather than as transudate or exudate. In specific circumstances, the AFTP (used apart from the exudate-transudate concept) is of some value. Patients with low AFTP are known to be at high risk for spontaneous peritonitis (6). Prophylactic, poorly absorbed oral antibiotics have been shown to selectively decontaminate the gut and therefore to help prevent bacterial infections in this subgroup of patients, who are identified by AFTP (33). Also, AFTP can be helpful in discriminating spontaneous from secondary bacterial peritonitis (for example, ruptured viscus into ascitic fluid) (6, 23). Cardiac ascites is characterized as high in protein with a high albumin gradient (see Figures 1 and 2); a high AFTP level may suggest the possibility of heart failure as the cause of ascites in a patient with a high albumin gradient (10). Neither AFTP nor albumin gradient changed dramatically in this study when ascitic fluid infection developed, and neither can replace ascitic fluid cell count and culture in detecting bacterial infection. Similarly, neither AFTP nor albumin gradient can replace ascitic fluid cytologic or culture results for mycobacteria in confirming suspected peritoneal carcinomatosis or tuberculosis, respectively. In addition to assisting with classification of patients with ascites, the albumin gradient may be of value in predicting response to treatment. Patients with ascites related to portal hypertension (for example, those with cirrhosis) usually respond to dietary sodium restriction and diuretics (34). In contrast, patients with ascites unrelated to portal hypertension (for example, those with peritoneal carcinomatosis) are refractory to diuretic therapy (35). Because the albumin gradient is accurate in detecting portal hypertension, studies of its value in predicting those patients who will respond to treatment are needed.
Table 3 . Types of Ascitic Fluid Samples Collected Type
Samples n (%)*
Cirrhotic (with or without infection) Cardiac Miscellan eous portal-hypertension related Mixed Cirrh osis and tuberculous peritonitis Cirrh osis and hepatocellular carcinoma Mass ive liver metastases and peritoneal car cinomatosis Cirrh osis and peritoneal carcinomatosis Cirrh osis and pancreatitis Cirrh osis and the nephrotic syndrome Cirrh osis and chlamydia peritonitis Mass ive liver metastases and portal vein thr ombosis Fulminj ant hepatic failure Acute rlepatitis superimposed on cirrhosis Massiv
