Article Text

Download PDFPDF

Surgical shunts compared with endoscopic sclerotherapy for the treatment of variceal bleeding in adults with portal hypertension: a systematic review and meta-analysis
  1. Lu Tian,
  2. Yunjun He,
  3. Donglin Li,
  4. Hongkun Zhang
  1. Department of Vascular Surgery, The First Affiliated Hospital, College of Medicine, Zhejiang University, Hangzhou, China
  1. Correspondence to Dr Hongkun Zhang, Department of Vascular Surgery, The First Affiliated Hospital, College of Medicine, Zhejiang University, 79 Qingchun Road, Hangzhou, China; doczhk{at}


Aim Portal hypertension is a common complication of chronic liver disease and can cause variceal bleeding which is associated with high mortality. Choices for the treatment of variceal bleeding include surgical shunts and endoscopic sclerotherapy. The aim of this study was to compare the efficacy of surgical shunts and endoscopic sclerotherapy in treating variceal bleeding due to portal hypertension.

Design Systematic review and meta-analysis.

Setting Medline, PubMed, Cochrane and Google Scholar databases were searched until 12 February 2015, for relevant randomised control trials. Twenty studies with a total of 1540 participants were included.

Patients Patients with variceal bleeding due to portal hypertension.

Interventions Surgical shunts compared to endoscopic sclerotherapy.

Main outcome measures Rates of rebleeding, survival and hepatoencephalopathy, and length of hospital stay.

Results Pooled data for 17 studies showed that the rate of rebleeding was significantly more frequent with sclerotherapy compared with surgical shunt therapy (OR 3.99, 95% CI 2.98 to 5.33, p<0.001). The sclerotherapy patient group compared with the shunt group was less likely to develop hepatoencephalopathy (15 studies: pooled OR 0.53, 95% CI 0.31 to 0.91, p=0.021) and had shorter hospital stays (pooled mean difference−4.32, 95% CI− 7.97 to −0.66, p=0.021). No significant difference in the survival rate was observed between the two groups (seven studies: OR 1.01, 95% CI 0.63 to 1.62, p=0.964).

Conclusion This analysis indicated that the two types of treatment have similar mortality rates but differed with respect to rebleeding rate, incidence of hepatoencephalopathy and length of hospital stay.

  • surgical shunts
  • endoscopic sclerotherapy
  • portal hypertension
  • variceal bleeding
  • treatment

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Portal hypertension is a progressive complication of cirrhosis, which is the end stage of chronic liver disease.1 Portal hypertension is the most common complication of chronic liver disease and can result in death or the need for liver transplantation.2 Portal hypertension can also be caused by prehepatic abnormalities, posthepatic abnormalities and intrahepatic non-cirrhotic causes.3 Portal hypertension in cirrhosis can be due to an increase in resistance to portal flow and/or an increase in portal venous flow, resulting in a pathological increase in the portal venous pressure gradient between the portal vein and the inferior venous cava.1 The upper normal value of the hepatic venous pressure gradient is 5 mmHg.4 Portal hypertension becomes clinically significant when this gradient increases above 10 mmHg, at which time the patient is at risk of developing clinical signs of portal hypertension, such as ascites or oesophageal varices.1 4

Variceal bleeding is the most serious complication of portal hypertension.5 It is estimated that the yearly incidence of variceal bleeding in patients with cirrhosis is about 4% and the risk of bleeding increases to 15% when medium to large varices are present.6 The first rebleeding and frequently recurring bleeding episodes are associated with high mortality.7–9 The risk of rebleeding is high, at about 60% with a mortality rate of up to 33%.3 The type of treatment used to treat variceal bleeding, in part, depends on the different clinical stages in the natural history of portal hypertension.3 4 In patients with varices with a high risk of bleeding, prophylactic non-selective β-blockers with or without endoscopic ligation are recommended to prevent vessel rupture.3 7 In patients with acute variceal bleeding, standard treatments include endoscopic variceal sclerotherapy, ligation, banding and pharmacological management.3 Transjugular intrahepatic portosystemic shunts (TIPS) have also been used to treat bleeding varices due to cirrhosis and portal vein hypertension, and are associated with reductions in treatment failure and mortality.10 11 Other methods for treating acute variceal haemorrhage include extrahepatic portosystemic shunting, such as with portacaval, splenorenal or mesocaval shunts.4 12–19 However, some of these methods are associated with significant complications, such as a high rate of rebleeding and reintervention for occlusion, high mortality and a high incidence of encephalopathy.4 20 21

Although there are various pharmacological interventions for variceal bleeding, endoscopic sclerotherapy or surgical shunts are alternative treatments when bleeding cannot be controlled medically. The relative effectiveness of surgical shunts compared with endoscopic sclerotherapy is not well established.4 Therefore, the aim of this study was to compare, in a systematic review and meta-analysis, the effectiveness of variceal sclerotherapy and surgical shunting for the management of variceal bleeding due to portal hypertension.


Search strategy

The study was performed according to the PRISMA guidelines. Medline, PubMed, Cochrane and Google Scholar were searched until 12 February 2015 using the following search terms: shunts, sclerotherapy, variceal bleeding and portal hypertension. Eligible studies were randomised controlled trials or two-arm prospective studies that compared the efficacy of surgical shunts and sclerotherapy. Retrospective studies, letters, comments, editorials, case reports, proceedings, personal communications and single-arm studies were excluded. Studies were also excluded if they did not report quantitatively the outcomes of interest. The reference list of relevant studies was hand-searched. Studies were identified according to the search strategy by two independent reviewers. Where there was uncertainty regarding eligibility, a third reviewer was consulted.

Data extraction and quality assessment

The following information/data were extracted from studies that met the inclusion criteria: the name of the first author, year of publication, study design, number of participants in each group, participants’ ages and gender, and the major outcomes. The primary outcomes were rates of rebleeding, survival and hepatoencephalopathy, and length of hospital stay. The quality of the included studies was evaluated using the Cochrane Collaboration’s tool.22

Statistical analyses

Odds ratios (OR) with 95% confidence intervals (CI) for each individual study and for all the studies combined were calculated for dichotomous variables between patients in the sclerotherapy group and the shunt group. For studies where no events were observed in one or both arms, 0.5 was added to cells for that study to avoid a computational error. For rebleeding and hepatoencephalopathy, an OR >1 indicated that the shunt group was favoured, while, for survival rate an OR >1 indicated that the sclerotherapy group was favoured. The difference in means with the 95% CI between two groups was calculated for continuous outcome (length of hospital stay). Median, range and the size of a sample were used to estimate the mean and variance if data lacked a mean and SD.23 A χ2-based test of homogeneity was performed, and the Cochran’s Q inconsistency index (I2) statistics were determined. If the I2 statistic (>50%) indicated heterogeneity between studies, the random-effects model was used; otherwise, fixed-effects models were used. Pooled effects were calculated and a two-sided p value <0.05 was considered to indicate statistical significance.

Sensitivity analysis was carried out using the leave one-out approach to investigate the validity and robustness of the results. The leave-one-out approach involves performing a meta-analysis on each subset of the studies obtained by leaving out exactly one study. In addition, publication bias was assessed by applying Egger’s test to funnel plots.24 The absence of publication bias was indicated by the data points forming a symmetric funnel-shaped distribution and a one-tailed significance level of p>0.05. All analyses were performed using Comprehensive Meta-Analysis statistical software, version 2.0 (Biostat, Englewood, NJ, USA).


A total of 196 studies were identified through the database searches and other sources, and 129 remained for further evaluation after duplicates were removed (figure 1). Ninety-one articles were excluded for not meeting the inclusion criteria, and the full texts of 38 articles were assessed. Of these, 18 were excluded, the reasons for which are given in the PRISMA flow diagram shown in figure 1.

Twenty studies12–19 25–36 were eligible for this meta-analysis with 1540 patients, 778 of whom were in the sclerotherapy group and 762 in the shunt group (table 1). Patients’ ages ranged from 44 to 79 years, and the proportion of patients who  were male ranged from 51% to 92% in the sclerotherapy group and from 55% to 88% in the shunt group. The percentage of patients whose alcohol consumption had caused the variceal bleeding ranged from 22% to 90% in the sclerotherapy group and from 11% to 100% in the shunt group. Across all studies, the length of follow-up was over 1 year.

Table 1

Summary of the basic characteristics of randomised controlled trials selected for meta-analysis

Quality assessment

We assessed the quality of the studies using the Cochrane Collaboration’s tool and found that overall the studies had an unclear to high risk of performance and detection bias (see online supplementary figure 1A,B). This may reflect the fact that it is difficult and unethical to perform a double-blinded test with this type of surgery as both patients and physicians must be aware of the type of surgery to be performed, and its risks and complications. This analysis indicates that study data included in our analysis were of adequate quality.

Supplementary file 1


Seventeen studies provided rebleeding data and were included in the meta-analysis.12 15–19 26–36 There was no significant heterogeneity among the 17 studies (heterogeneity test: Q=29.176, I2=45.16%; figure 2) and so a fixed-effects model was used. The overall analysis revealed that rebleeding was significantly more frequent in sclerotherapy patients than in shunt patients (pooled OR 3.99, 95% CI 2.98 to 5.33, p<0.001).

Figure 2

Forest plots of meta-analysis of the rate of rebleeding.

Seven studies provided survival data and were included in the meta-analysis.12–14 18 25 29 34 There was no significant heterogeneity among the seven studies (heterogeneity test: Q=0.820, I2=0%; figure 3) and so  a fixed-effects model was used for the analysis. The overall analysis revealed no significant difference in the survival rate (less than 3 months) between the two groups (pooled OR 1.01, 95% CI 0.63 to 1.62, p=0.964).

Figure 3

Forest plots of meta-analysis of survival rate.

Fifteen studies provided hepatoencephalopathy data and were included in the meta-analysis.12 15 16 18 25–30 32–36 There was significant heterogeneity among the 15 studies (heterogeneity test: Q=40.51, I2=65.44%; figure 4) and so a random-effects model was used. The overall analysis revealed that patients in the sclerotherapy group were less likely to develop hepatoencephalopathy than those in the shunt group (pooled OR 0.53, 95% CI 0.31 to 0.91, p=0.021).

Figure 4

Forest plots of meta-analysis of the rate of hepatoencephalopathy.

Eight studies12 14 16 19 26 27 33 35 provided data on the length of hospital stay and were included in the meta-analysis. There was significant heterogeneity among the eight studies (heterogeneity test: Q=38.554, I2=81.84%; figure 5) and so a random-effects model was used. The overall analysis revealed that patients in the sclerotherapy group had shorter hospital stays than patients in the shunt group (pooled mean difference −4.32, 95% CI −7.97 to −0.66, p=0.021).

Figure 5

Forest plots of meta-analysis of the length of hospital stay.

Sensitivity analysis and publication bias

Sensitivity analyses were performed using the leave-one-out approach, in which the meta-analysis was performed with each study removed in turn (see online supplementary table 1). The direction and magnitude of combined estimates on rebleeding, survival rate and hepatoencephalopathy did not vary markedly on removal of any one study, indicating that the meta-analysis was robust and the data were not overly influenced by an individual study. However, for length of hospital stay, when we removed the study by Terés et al 35 the pooled mean difference became non-significant, indicating that the pooled estimates may have been overly influenced by this study.

Supplementary file 2

It was not possible to evaluate publication bias for the analysis of survival rate and length of hospital stay as too few studies were used. We found no evidence of publication bias for the rate of rebleeding (see online supplementary figure 2A) or hepatoencephalopathy (see online supplementary figure 2B).


We conducted a systematic review and meta-analysis in order to compare endoscopic sclerotherapy and surgical venous shunting in the treatment of portal hypertension. The rates of rebleeding, survival and hepatoencephalopathy, and length of hospital stay in the two groups were compared. We found that the rate of rebleeding was significantly higher in patients treated with endoscopic sclerotherapy than in patients who underwent surgical shunting procedures. Patients in the endoscopic sclerotherapy group were less likely to develop hepatoencephalopathy and had a shorter length of hospital stay compared with the surgical shunting group. The overall analysis revealed no significant difference in the survival rate between the two groups. The strength of our analysis is that it included only randomised controlled trials.

Our findings are comparable to the results of a previous meta-analysis by Khan et al 37 that compared the efficacy of surgical shunts (total surgical stunts: distal spleno-renal shunts (DSRS) or TIPS) and endoscopic therapy (sclerotherapy and/or banding) for the prevention of rebleeding in patients with cirrhosis.37 Their study included 22 trials with 1409 patients. The authors note that the included studies had methodological problems. Like our findings, the results of Khan et al showed that shunt therapy was associated with less rebleeding (OR 0.24; 95% CI 0.18 to 0. 30) compared with endoscopic therapy, but had a higher risk of hepatic encephalopathy (OR 2.07; 95% CI 1.59 to 2.69) and chronic encephalopathy (OR 2.09; 95% CI 1.20 to 3.62). However, in contrast to our results, they found there was no difference in length of hospital stay between the two treatment groups (weighed mean difference 0.78 days; 95% CI 1.48 to 3.05). Similar to us, Khan et al also found no difference in mortality between shunt therapy and endoscopic therapy (HR 1.00, 95% CI 0.82 to 1.21). They also found that the percentage of patients with shunt occlusion or dysfunction was 3.1% after total surgical stunts, 7.8% following DSRS and 59% following TIPS. We did not evaluate shunt occlusion or dysfunction in our study. The difference in findings between our study and that of Khan et al regarding length of hospital stay may be due to differences in the included studies, or the method of analysis.

The guidelines recommend that the first-line treatment for acute variceal haemorrhage, in conjunction with vasoactive pharmacological therapy (eg, non-selective β-blockers), is endoscopic variceal ligation, but state that sclerotherapy is an option when endoscopic variceal ligation is difficult.3 TIPS is recommended as an alternative procedure in patients who fail the standard combination therapy of endoscopic and pharmacological therapy, although TIPS is associated with high mortality.3 To prevent recurrence of bleeding, it is recommended that patients be prophylactically treated with non-selective β-blockers.3 Sclerotherapy decreases rebleeding and mortality rates, but is associated with serious complications, such as oesophageal strictures and bleeding from ulcers.3 As endoscopic variceal ligation is associated with better outcomes and fewer serious adverse effects, it is often performed instead of sclerotherapy.3

Our study has several limitations. We did not perform subgroup analysis of the various types of surgical shunts, and so the relative advantages of portacaval, splenorenal and mesocaval surgical shunts, if any, are unknown. In addition, we combined studies that reported findings on surgical shunts with those on TIPS. Combining data from these two different procedures may have confounded our results. We also did not analyse the various shunts according to emergent versus elective operations; such analysis might have revealed differences in their effectiveness in selective situations. The recommended first-line treatment for acute variceal haemorrhage is endoscopic variceal ligation plus pharmacotherapy. Variceal ligation has become more common than sclerotherapy and is associated with fewer side effects.38 This may, in part, explain why there were very few studies on endoscopic sclerotherapy after 2009. Nevertheless, it still is important to determine the efficacy of endoscopic sclerotherapy relative to other procedures for treating variceal bleeding, as this technique is recommended when variceal ligation is not possible.3 Further, sensitivity analysis indicated that our findings concerning length of hospital stay may have been overly impacted by one study. The substantial heterogeneity of the data of included studies also might have confounded our ability to estimate the overall association. A meta-regression based on study level covariates would have been ideal. However, most of the covariates analysed in this study were nominal, such as sex, aetiology or Child-Pugh class, thus making it harder to perform a meta-regression.

In conclusion, the results of this systematic survey and meta-analysis, comparing the outcomes of endoscopic sclerotherapy with surgical shunt procedures for the treatment of varices due to portal hypertension, revealed that survival rates after the two procedures were similar. The analysis found that surgical shunts are associated with lower rates of rebleeding, but higher rates of hepatoencephalopathy and longer length of hospital stay. These findings may help clinicians to select the optimum therapy for bleeding oesophageal or gastric varices caused by portal hypertension.

Main messages

  • Surgical shunt has a lower rate of rebleeding compared with endoscopic sclerotherapy.

  • Shunts showed higher rates of hepatoencephalopathy and longer hospital stays compared with sclerotherapy.

  • Sclerotherapy and  surgical shunts for treating variceal bleeding have similar survival rates.

Current research questions

  • How safe and effective is surgical shunt or endoscopic sclerotherapy compared with the other treatment methods such as variceal ligation?

  • Does time of intervention, elective or emergent, affect clinical outcome or recurrence?

  • Are there any changes in survival or complications when different locations of surgical shunts are chosen?



  • Contributors LT: guarantor of the integrity of the entire study, study design, literature research, data acquisition, manuscript editing. HZ: study concepts, definition of intellectual content, manuscript review. YH: literature research, manuscript preparation. DL: data analysis, statistical analysis.

  • Funding This study was sponsored by the Program of Education Department of Zhejiang province (grant code: Y201223954).

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.