Retraction Note: New proposal for the serum ascites albumin gradient cut-off value in Chinese ascitic patients

Serum ascites albumin gradient (SAAG) has been recognized as a reliable marker in the differential diagnosis of ascites. The etiological background of cirrhosis is rather different between western countries and eastern countries. The threshold of SAAG in Chinese ascitic patients has not been evaluated yet. The aim of this study was to define a new reasonable threshold of SAAG in Chinese ascitic patients. Adult patients with ascites admitted to the Shanghai Changzheng Hospital from Jan 2004 to Jun 2010 were retrospectively analyzed. The diagnostic criteria for cirrhotic ascites are clinical manifestations, radiological features and esophageal-gastric varicosis, or histopathology. Serum was detected by chemical method using a commercial kit. We used receiver operating characteristic (ROC) analysis to achieve maximal sensitivity and specificity of SAAG. The mean value of SAAG in portal-hypertension-related ascites was significantly higher than that in the non-portal-hypertension-related ascites (21.15 ± 4.38 g/L vs 7.48 ± 3.64 g/L, P = 0.002). The SAAG cut-off value under 12.50 g/L predicted portal hypertension ascites with the sensitivity of 99.20%, specificity of 95.10% and accuracy of 97.65%. SAAG is useful to distinguish portal-hypertension-related ascites and non-portal-hypertension-related ascites, and 12.50 g/L might present as a more reasonable threshold in Chinese ascitic patients. The virtual slide(s) for this article can be found here: http://www.diagnosticpathology.diagnomx.eu/vs/1602582638991860


Introduction
Ascites is a common manifestation of advanced liver disease. Besides, it can also occur in individuals with extrahepatic diseases such as abdominal malignancy, tuberculosis, and autoimmune diseases. Traditionally, ascites is classified as being either transudative or exudative based on the ascitic fluid total protein (AFTP) concentration [1], the ratio of ascitic fluid protein concentration to serum total protein concentration or lactic dehydrogenase (LDH) [2]. However, the serum ascites albumin gradient (SAAG), 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 in the past 20 years [3][4][5]. Hoefs [6] first introduces the SAAG and reports that SAAG can reflect the portal vein pressure and improve the accuracy of ascites identification. Thereafter, several investigators also demonstrate that the superiority of SAAG in distinguishing portal hypertensive ascites (SAAG > 11 g/L) and non-portal hypertensive ascites (SAAG < 11 g/L) [6][7][8][9]. Thus, SAAG 11 g/L has been considered as a reasonable threshold in clinical practice.
However, there are also some controversies for SAAG [10][11][12]. In 2002, Goran et al. [13] revise the threshold of SAAG and find that when using the cut-off value of 11 g/L, the diagnostic sensitivity is 97.56% and specificity is 46.34%, while using 15.86 g/L, the sensitivity and specificity could reach the maximum. According to Starling, SAAG is the best indicator of portal hypertension to distinguish nonportal hypertensive and portal hypertensive ascites [14]. However, the best threshold may call further research, the primary study reporting 11 g/L only included the ascites due to cirrhosis and tumor, except tuberculosis. In fact, tuberculous ascites are also frequently seen in clinical practice. Cirrhosis is mainly caused by alcohol abuse in Western countries, while in China by Hepatitis B virus infection. Besides, China is also an epidemic area for tuberculosis. To date, the SAAG in Chinese patients with ascites has not been evaluated yet. Hence, this study aimed to define a new cut-off value of SAAG with maximal sensitivity and specificity in Chinese patients with ascites.

Patients
Adult patients with ascites admitted to the Shanghai Changzheng Hospital from Jan 2004 to Jun 2010 were retrospectively analyzed. The diagnostic criteria for cirrhotic ascites are mainly based on clinical manifestations, radiological features and esophageal-gastric varicosis, or histopathology, excluding spontaneous bacterial peritonitis (SBP), malignancy and tuberculosis. Malignant ascites was diagnosed when the results of ascitic fluid cytologic study or peritoneal biopsy was positive. The diagnosis of tuberculous peritonitis required ascitic acid-fast stain or mycobacterial growth in a culture of ascites or peritoneal biopsy with or without histologically proven granulomatous peritonitis, and without tuberculous pericarditis. Serum and ascites fluid were obtained (maximum interval 24 hours) and were tested for albumin concentration and other related parameters, the albumin concentration was detected by chemical method using a commercial kit (P800, Roche), other parameters included ascite routine, protein, tumor marker, cast-off cells, ADA. The patients with unknown origin of ascites or without data of SAAG were excluded. We used cirrhotic ascites as Portalhypertension-related (PHT) ascites, and malignant or tuberculous ascites as Non-portal-hypertension-related (NON-PHT) ascites.

Statistical analysis
Continuous variables were tested for normal distribution and expressed as mean ± standard deviation (SD) and analyzed by Student's t test. Rates were analyzed by Chisquare test. ROC curve was used to assess the diagnostic value of SAAG. The statistical analysis was performed using SPSS 17.0 (SPSS Inc., Chicago, IL, USA) and P < 0.05 was considered statistically significant.

Patient characteristics
A total of 213 patients diagnosed with ascites were included in this study. The baseline characteristic is shown in (Table 1). The clinical findings of the patients with PTH group and NON-PTH group were similar with respect to age and sex ratio.

SAAG analysis
The mean SAAG level of all patients involved was 15.95 ± 7.82 g/L. The mean level of SAAG in the PHT group (cirrhotic ascites) was significantly higher than that in the NON-PHT group (malignant and tuberculous ascites) (21.15 ± 4.38 g/L vs 7.48 ± 3.64 g/L, P = 0.002). Besides, the SAAG levels of the patients with malignant or tuberculous ascites were both significantly lower than that of patients with cirrhotic ascites ( Table 2). In the PHT group, all patients had a SAAG of 11 g/L or greater. While in the NON-PHT group, 12 patients had a SAAG of 11 g/L or higher, and 69 patients had a SAAG less than 11 g/L ( Figure 1).

ROC curve
The ROC curve of SAAG is shown in Figure 2 (Figure 2). The area under the curve was 0.987, and the SAAG cut-  off value 12.50 g/L was the best threshold to predict PHT ascites with a sensitivity of 99.20%, a specificity of 95.10% and an accuracy of 97.65%. When using the SAAG cut-off value 11 g/L, its sensitivity and specificity were 100.00% and 85.19% with an accuracy of 94.37%. The accuracy and specificity using SAAG 12.50 g/L were significantly higher than those using SAAG 11 g/L (accuracy, 97.65% vs 94.37%,P<0.01; specificity, 95.10% vs 85.19%, P<0.01).

Discussion
Starling [14] recognizes that the protein content of edema fluid is a reflection of its osmotic pressure and that the osmotic pressure gradient between blood and interstitial fluid is a direct function of the corresponding capillary hydrostatic pressure gradient. According to his concept, the protein concentration gradient between serum and ascites reflects the portal pressure. Hoefs [6] first reports a liner correlation between the SAAG and portal venous pressure in 56 patients with chronic liver disease. More recently, this relationship is also confirmed by Rector and Reynolds [8], who report an excellent correlation between the SAAG and portal pressure. Using AFTP to distinguish exudate and transudate is influenced by serum protein concentration as well as portal pressure. In contrast, the SAAG correlates directly with only one physiologic factor, the portal pressure [6]. Portal pressure keeps stable, and therefore there is no or only minor change of SAAG despite diuresis or therapeutic paracentesis. Hence, it is possible to understand why the SAAG 11 g/L has been widely used to differentiate PHT Figure 1 The scatter of SAAG in classifying samples. In the PHT group, all patients had a SAAG of 11 g/L or greater. While in the NON-PHT group, 12 patients had a SAAG of 11 g/L or higher, and 69 patients had a SAAG less than 11 g/L.

Figure 2
The ROC curve of SAAG. The area under the curve was 0.987, and the SAAG cut-off value 12.50 g/L was the best threshold to predict PHT ascites with a sensitivity of 99.20%, a specificity of 95.10% and an accuracy of 97.65%. The accuracy and specificity using SAAG 12.50 g/L were significantly higher than those using SAAG 11 g/L (accuracy, 97.65% vs 94.37%,P<0.01; specificity, 95.10% vs 85.19%, P<0.01).
and NON-PHT ascites. However, Kajani et al. [15] report that when using the value of SAAG 11 g/L for differentiating PHT and NON-PHT ascites, the accuracy is only 87.5%. In addition, SAAG corrected with serum globulin level could also promote the accuracy to determine portal hypertension [16]. Cirrhosis, cancer and tuberculosis are the most common causes of ascites. As mentioned above, most studies about SAAG are performed in patients with cirrhotic or malignant ascites. However, its utility in tuberculous ascites has not been assessed yet. Therefore, our study involved 27 patients with tuberculous ascites. In this analysis, all the patients with cirrhotic ascites had SAAG of 11 g/L or greater, which was rather similar to the study by Shakil et al. [17]. In contrast to the study by Kajani et al. [15], the causes of cirrhosis did not appear to affect SAAG. The albumin gradient in 83 patients with viral-hepatitis-related cirrhosis (21.28 ± 4.22 g/L) was very near to that in the 49 patients with non-viralhepatitis-related cirrhosis (20.94 ± 4.67 g/L). In addition, the SAAG in 117 patients with nonalcoholic liver disease was not different from that in the 15 patients with alcoholic liver disease (21.26 ± 4.41 g/L vs 20.27 ± 4.17 g/L).
With regard to the discrimination between malignant and nonmalignant ascites, Pare et al. [7] report that SAAG less than 11 g/L is an excellent criterion for the diagnosis of malignant origin of ascites. Similar observations have also been made by Mauer et al. [5]. Based on the present experience, it appears that the criterion of SAAG less than 11 g/L for the distinction between malignant and nonmalignant ascites may be less specific than previously thought. In the present study, a SAAG less than 11 g/L was seen in only 43 of 54 patients with malignant ascites without metastatic liver involvement. Similarly, the gradient in the patients with malignant ascites also did not differ from the gradient in the patients with tuberculous ascites. It suggests that SAAG cannot further distinguish malignant ascites from tuberculous ascites.
Runyon et al. [9] report a very high accuracy of 96.7% for SAAG of 11 g/L based on 901 samples. In our study, when using SAAG 11 g/L, its accuracy was 94.37%, which was slightly lower than 96.7%, while the specificity was only 85.19%. This may be due to the different ascitic etiology between western and eastern countries. The threshold 11 g/L is based mainly on the prevalence of alcoholic cirrhosis in Western countries. In China, cirrhosis is mostly caused by HBV infection. Furthermore, the previous evaluation of diagnostic tests often uses sensitivity, specificity and accuracy, which often depend on the prevalence of study population. In fact, the ascitic prevalence is different in western and eastern countries. Moreover, the receiver operating characteristic (ROC) curve is currently recognized as the best way to measure the diagnostic information and decision-making. The cut-off value obtained by ROC curve has greater accuracy and clinical utility [18,19]. Our research achieved a new value of SAAG of 12.5 g/L by ROC curve. Compared with the previous SAAG of 11 g/L, the new SAAG had higher accuracy and specificity to distinguish PHT and NON-PHT ascites.
Based upon the data herein presented, we conclude that SAAG is useful to distinguish PHT and NON-PHT ascites, and 12.5 g/L might present as a more reasonable threshold in Chinese ascitic patients. However, further study is still needed to be done using larger samples.