Background
The prevalence of liver cirrhosis is estimated to be between 0.15 and 0.3% in European countries [
1]. The main causes are alcohol abuse, infection with viral hepatitis B and C as well as autoimmune liver diseases [
2]. A clinically relevant complication is the development of portal hypertension with all its clinical consequences such as ascites, spontaneous bacterial peritonitis and development of portosystemic collaterals. A progression rate of 12% has been reported for esophageal varices (EV) [
3]. Although the mortality of variceal hemorrhage has declined in the last decades, it is still very high with a six-week-mortality of up to 37% [
4], and a high recurrence rate after the first bleeding incident [
5]. Although repeated endoscopic controls of patients with an advanced liver fibrosis or liver cirrhosis are justified, it is an invasive diagnostic procedure with its own risks, and it is not always widely available in countries with lower health care standards. Therefore, non-invasive predictors for portosystemic collaterals are of high interest. Notably, the venous blood is drained from the gall bladder in part via small vessels directly into the liver. An additional venous blood drain flows via small veins towards the cystic duct and then with vessels from the common bile duct terminating in the portal venous system [
6].Therefore, the gall bladder should be directly affected by portal hypertension causing a thickened gall bladder wall due to impaired venous drainage. Here, we aim to determine whether non-inflammatory gall bladder wall thickness (GBWT) correlates with the presence of EV. To this end, we performed a retrospective endoscopic-ultrasonographic study correlating the presence of EV and GBWT with other non-invasive parameters for liver disease and portal hypertension.
Methods
In this study we retrospectively included all patients with chronic hepatic disease, who received an ultrasound of the abdomen either as an inpatient or outpatient in the Department of Gastroenterology and Gastrointestinal Oncology of the University Hospital of Goettingen between April 2015 and January 2016. Patients who had a cholecystectomy or complained of upper abdominal pain were excluded from the study. Gall stones and single gall bladder polyps without symptoms were no exclusion criteria. Of all patients who also had a documented upper endoscopy (median time interval 147 days), the following parameters were evaluated by ultrasound: The thickness of the gall bladder was measured twice after overnight fasting at two different locations and an average value was calculated. The spleen length was measured from a left lateral cross section. The diameter of the portal vein, the portal blood flow velocity and the liver size were measured. Ultrasound and endoscopy examinations were performed by experienced Gastroenterology trainees (> 3 years experience) and senior Gastroenterology consultants. The presence or absence of ascites was recorded. Additionally, clinical parameters such as the Child-Pugh-classification, laboratory results and upper endoscopic findings (presence of EV graded according the classification of Paquet) were obtained. Using the results of the cranio-caudal spleen diameter, gall bladder wall diameter and laboratory results, we calculated the ratio of platelet count to spleen diameter and the ratio of platelet count to gall bladder wall thickness. The statistical analysis was performed using the Mann Whitney U and Chi square test. Furthermore, variables with a P value < 0.1 from univariate analysis entered the multivariate binary logistic regression analysis and (receiver operating characteristic) ROC analysis was performed by SPSS Version 25 Mac OS. Since patient data were collected retrospectively and did not influence the diagnostic or therapeutic management of the patients, the ethic committee at the University Medical Centre Goettingen, Germany, was informed in written form about the study prior to data collection but did not request a separate ethical votum (24/7/15AN).
Discussion
Patients with compensated liver cirrhosis have a chance of up to 40% to develop EV [
8]. To avoid hemorrhage from EV, it is recommended to perform an upper endoscopy as soon as there are signs for the presence of liver cirrhosis in patients [
9,
10]. Therefore, many patients undergo upper endoscopy although they do not require treatment of EV (e.g. ligation) according to endoscopic classifications. While diagnostic gastroscopy itself is of low risk, low platelet counts as well as impaired coagulation parameters increase the risk of complications. Most patients prefer sedation during the procedure which is associated with additional risks. Therefore, more accurate non-invasive parameters for the presence of EV could be a valuable and clinically relevant tool. We based our study on non-invasive, standard diagnostic tests, which are routinely performed in patients with chronic liver disease: ultrasound, clinical and laboratory results were evaluated in terms of prediction of EV.
Because of its portal-venous blood supply, we assumed that the GBWT may predict the presence of portal hypertension and EV. An interesting study by Maruyama et al. also reported a lower sensitivity of 62% regarding the detection of large esophageal varices using the platelet count to spleen diameter ratio in 229 cirrhosis patients. In this study, the authors showed that a diameter of the left gastric vein -as a non-variceal collateral- of more than 5.35 mm had a sensitivity of 90% and a specificity of 62% for presence of large esophageal varices. Its sonographic detection was associated with a sensitivity of 84% for any esophageal varices and a sensitivity of 100% for large varices [
11]. However, further prospective studies are required to assess the value of portal vein velocity as non-invasive parameter for the presence of esophageal varices.
A small Chinese study showed a correlation between portal vein velocity and GBWT supporting the hypothesis that GBWT could also predict the presence of EV [
12]. From a pathophysiological point of view, GBWT may be a microcirculatory driven event caused by impaired portalvenous outflow before significant changes in portal vein velocity occur. However, the development of GBWT may also be caused by other factors such as the serum-ascites albumin gradient (SAAG) [
13].
Several studies have investigated non-invasive parameters as predictors for the presence of EV. A platelet count to spleen diameter ratio of 909 and less was associated with EV [
7]. The enlarged spleen is caused by portal hypertension and low platelets were also associated with a lowered thrombopoetin serum level due to reduced liver function [
14]. Chen performed a meta-analysis to confirm the usefulness of this ratio and calculated a summative sensitivity of the ratio of 84% with a specificity of 78% to predict EV. The sensitivity of this ratio was also influenced by etiology of the liver disease with a sensitivity of 92% in viral liver cirrhosis [
15]. Using the platelet count to spleen diameter ratio as previously described [
7], the sensitivity was somewhat lower with our dataset. The reasons might be the greater variety of causes for liver disease in our cohort than in previous evaluations. Another non-invasive method is the use of computed tomography (CT) imaging with a sensitivity of 90% and a specificity of 72% for the detection of EV [
16]. The higher sensitivity is traded against higher costs, exposure to irradiation, and the use of contrast agents. Other non-invasive measurements such as liver stiffness measurements are promising but further studies need to be performed. Meta-analysis of data so far collected by using transient elastography (FibroScan®) showed lower prognostic values for liver stiffness [
17]. A meta-analysis of studies using different modes of elastography techniques to measure spleen stiffness showed heterogeneous results to detect EV [
18]. The sensitivity of liver stiffness was 84% in predicting any varices, compared to 78% using the stiffness of the spleen as parameter. The specificity of the spleen stiffness was higher when compared to liver stiffness (76% versus 62%) [
18] [
16]. The use of capsule endoscopy to detect EV is also discussed in the literature [
19] but high costs and its semi-invasive nature need to be kept in mind. Because of those limitations of the aforementioned non-invasive methods, the use of GBWT could represent a novel and feasible clinical marker for the detection of EV.
Alcantara previously published a cut-off value of 4.35 mm for a thickened gall bladder wall and found a sensitivity of 60% and a specificity of 90% regarding the presence of EV in pediatric patients [
20]. The sensitivity was higher than in our study, although a higher value was used as cut-off. Other reasons for this difference might be the different patient cohorts, since de Alcantara based his study on data from children with various causes of cirrhosis such as biliary atresia and autoimmune hepatitis [
20]. In our study we used data from adult patients with chronic liver disease and common causes for cirrhosis in Western Europe, but lack of histological confirmation in almost half of them. The cut-off-value of 4 mm was arbitrarily set for univariate analysis and seems reasonable since a lower value could be measured in individuals that did not fast overnight with a higher rate of falsely positive cases. A higher cut-off value would lower sensitivity.
In addition, a lowered velocity within the portal vein seems reasonable in the presence of portal hypertension and esophageal varices, but existing data are still conflicting. An Indian study of sonographic parameters predicting esophageal varices in 56 patients showed a significant difference of mean portal vein velocity. In presence of esophageal varices the mean velocity was 14.77 cm/s and in absence of varices 17.66 cm/s [
21]. However, Li et al. did not report a significant difference in portal vein velocity between patients with and without esophageal varices. The mean velocity was 15.3 cm/s in healthy individuals, 14.2 cm/s in patients with first degree varices, 13.1 cm/s in second degree varices and 12.0 cm/s in third degree varices [
22]. In our study there was no difference in average portal vein velocity with 18.0 cm/s in both groups.
Our study has several limitations. First, ultrasonography and endoscopy were not always performed within a few days and may have biased results in case of rapidly changing endoscopy or ultrasonography findings. Furthermore, retrospective data collection could not establish a clear cause for chronic liver disease in almost 20% of patients. Secondly, although performed by experienced Gastroenterology trainees and consultants, GBWT measurements were performed by only one examiner, and inter-observer variability could thus not be accounted for. Third, we could not detect significant differences between small and large EV most likely due to the relatively low number of 3° EVs (n = 9) in the EV cohort.