Interventions for Primary Prevention of Esophageal Variceal Bleeding
SEE ARTICLE ON PAGE 1657
Potential conflict of interest: Nothing to report.
Abbreviations
-
- NSBB
-
- nonselective β-blockers
-
- VBL
-
- variceal band ligation
Current guidelines recommend that patients with cirrhosis undergo endoscopic screening for varices, unless patients have liver stiffness less than 20k Pa and platelet count greater than 150 × 109/L.1 Screening for varices presumes that we can offer an intervention to reduce risk of first variceal hemorrhage and, ideally, death. Because risk of bleeding increases with variceal size, guidelines recommend that patients with medium-large esophageal varices receive endoscopic variceal band ligation (VBL), nonselective β-blockers (NSBB), or carvedilol as primary preventive therapy to reduce first variceal hemorrhage.1
Previous meta-analyses show that NSBB and VBL each reduce first variceal hemorrhage compared with no active treatment, with VBL also reducing mortality.2, 3 A meta-analysis of 19 randomized trials showed that VBL is superior to NSBB at preventing variceal bleeding, with no difference in mortality.4
In this issue of Hepatology, Sharma et al5 state that “although NSBB and VBL are effective and comparable for prevention of variceal bleeding, neither therapy has been shown to significantly decrease mortality.” They performed a network meta-analysis to add more evidence beyond previous direct-evidence meta-analyses and compared the efficacies of different interventions.
Network Meta-Analyses
Network meta-analysis is an increasingly popular way to extend the use of data from randomized controlled trials, combining direct evidence from head-to-head comparisons of interventions with indirect evidence from studies that compare different interventions with a common comparator.6 This technique may be especially useful to compare interventions that have not been studied head-to-head in randomized trials. For example, if one wishes to compare intervention A to intervention B and each intervention has been compared in randomized trials to placebo P (i.e., A versus P and B versus P), the indirect comparison is made by subtracting the meta-analytic estimate of studies of treatment A versus P from the estimate of studies of treatment B versus P.6 This indirect estimate is combined with a direct estimate from randomized head-to-head comparisons of A versus B (if available) in a network meta-analysis, producing a mixed estimate of effect size.
An assumption in performing indirect comparisons is that sets of studies (e.g., A versus P studies and B versus P studies) are not substantively different other than in the assigned study interventions.6 If there are differences in the distribution of characteristics that may affect outcome (e.g., disease severity, comorbidities, comparator dose), indirect comparison may not be valid to perform. Statistical testing should be done to determine whether differences between direct and indirect estimates are greater than expected by chance,6 as done by Sharma et al. If inconsistency is found, authors should explore possible explanations and may consider not synthesizing the data as a network meta-analysis.6
Current Study: Summary and Context of Results
NSBB Versus Placebo/No Intervention
Variceal Bleeding
NSBBs showed no significant reductions in both the direct meta-analysis of four trials and the network meta-analysis mixed estimate by Sharma et al.5 In contrast, a previous direct-evidence meta-analysis of eight randomized trials comparing NSBB to placebo or inactive treatment for medium-large esophageal varices reported significant benefit of NSBB.2 Sharma et al. included fewer direct-comparison studies throughout their review than previous meta-analyses, in part due to more stringent inclusion criteria (e.g., no abstracts, 1-year or longer follow-up).
Mortality
NSBB did not show any benefit in the direct estimate by Sharma et al.5 or previous direct-evidence meta-analysis.2 However, Sharma et al. found indirect evidence that favored NSBB. Thus, their mixed estimate, combining direct and indirect evidence, showed a trend toward benefit of NSBB versus placebo/no intervention (upper bound of 95% confidence interval was 1.00),5 whereas based on indirect evidence only, NSBB plus nitrates showed reduced mortality versus placebo/no intervention.5
VBL Versus No Intervention
Variceal Bleeding
VBL reduced variceal bleeding on direct and mixed estimates, which is consistent with the previous direct-evidence meta-analysis.3
Mortality
VBL reduced mortality in the direct estimate (which is consistent with the previous direct-evidence meta-analysis3) but resulted in a non-significant reduction in the mixed estimate.
VBL Versus NSBB
Variceal Bleeding
VBL was superior to NSBB in the direct and mixed estimate, which is consistent with the previous direct-evidence meta-analysis.4
Mortality
Direct comparison showed an non-significant trend toward higher mortality with VBL, whereas the indirect estimate was significantly lower than the direct estimate, resulting in no suggestion of difference in the mixed estimate. Potential explanations for this inconsistency between direct and indirect estimates were not explored. Previous direct evidence meta-analyses also showed no significant difference for VBL versus NSBB.4, 7
Carvedilol
Based on more limited evidence, mixed estimates indicated that carvedilol had less variceal bleeding than placebo or NSBB without differences in mortality versus any comparator.
Adverse Events
Adverse events necessitating discontinuation were more common with NSBB (10%) than VBL (1%), whereas serious adverse events were more common with VBL (4%) than NSBB (1%). Curiously, most serious adverse events with VBL did not appear to result in discontinuation of VBL.
Sharma et al. suggested that NSBB may be favored over VBL due to the lower risk of serious complications, although their data showed that the less serious side effects associated with NSBB were much more likely to result in discontinuation of therapy. Interestingly, physicians rank serious events such as perforation as significantly more important in choosing a therapy for primary prevention of variceal bleeding than do patients.8 Patients place the greatest importance on more common, non-life-threatening side effects seen with NSBB, such as shortness of breath and fatigue.8 Thus, when patients with cirrhosis were given standardized educational materials about cirrhosis, varices, bleeding risk, NSBB, VBL, and side effects—and were told that NSBB and VBL are equally effective—64% chose VBL over NSBB.8
NSBB: Effect on Mortality?
NSBB have been proposed to have a mortality benefit, because, unlike VBL, they affect portal hypertension and may provide benefit beyond an effect on variceal hemorrhage by reducing other complications of cirrhosis.1 Direct-evidence meta-analyses of trials for secondary prevention of variceal bleeding (i.e., all patients had previous bleeding and decompensated cirrhosis) suggest that mortality is reduced with NSBB versus placebo/no active treatment2 and with NSBB plus nitrates versus VBL.7 In contrast, available direct evidence2, 4, 5, 7 does not demonstrate that a mortality benefit of NSBB extends to patients without previous variceal bleeding, who generally have less severe liver disease than those who have bled. With the addition of indirect evidence, the analysis from Sharma et al. suggests that NSBB plus nitrates or perhaps even NSBB alone may have a mortality benefit versus placebo/no intervention in primary prevention.
Further primary prevention studies that stratify patients based on severity of liver disease would be useful, although demonstrating a mortality benefit in patients with less severe liver disease may be difficult. Additional trials assessing carvedilol also are important to determine whether its more reliable reduction in hepatic venous pressure gradient translates into greater clinical efficacy than NSBB, as suggested by the results for variceal bleeding from Sharma et al.
Effect of Current Study on Practice
Sharma et al. concluded that NSBB may be preferred for primary prevention of variceal bleeding, particularly in early or compensated liver disease.5 I would not draw this conclusion from the results of their network meta-analysis. NSBB were not better than placebo/no intervention at preventing variceal bleeding. A trend toward reduced mortality with NSBB versus placebo/no intervention was seen, but NSBB-nitrate combination therapy ranked much higher than NSBB for reduction of mortality.5 More importantly, both VBL and carvedilol were better than NSBB at preventing variceal bleeding without differences in mortality for these treatments versus NSBB.
Current recommendations for primary prevention with VBL or NSBB or carvedilol still appear to be acceptable—although the results of Sharma et al. suggest that NSBB plus nitrates might be preferred to NSBB monotherapy. It is important to educate patients regarding interventions and their differences, such as daily medication versus periodic endoscopy, and common bothersome side effects versus infrequent serious events. Choices regarding primary preventive therapy should then be based on patient preferences using a shared decision-making approach.
Potential conflict of interest
Nothing to report.
