Nonselective β-blockers do not affect mortality in cirrhosis patients with ascites: Post Hoc analysis of three randomized controlled trials with 1198 patients

Use of nonselective β-blockers (NSBBs) as secondary prevention of variceal bleeding was introduced in 1981.1 Since then, numerous randomized trials and meta-analyses have documented NSBBs' efficacy in preventing variceal bleeding,2, 3 and today NSBBs are the standard pharmacological treatment for primary and secondary prevention of variceal bleeding.4, 5 They reduce portal pressure by decreasing cardiac output (β₁-blockade) and splanchnic blood flow (β₂-blockade).6

Surprisingly, in 2010 serious concerns about the safety of NSBB use in cirrhosis patients with refractory ascites were raised by a single-center observational study including 151 cirrhosis patients with refractory ascites.7 The investigators observed a median survival time of 5 months in the 77 NSBB users versus 20 months in the 74 nonusers. This study stirred an intense debate,8-11 and subsequent studies have not been able to clarify whether NSBBs are safe for use in patients with decompensated cirrhosis.11-16 In 2012, the seemingly contradictory findings concerning beneficial and harmful effects of NSBBs were united in the window hypothesis.17 It posits that NSBBs should only be administered within a certain window during disease progression that closes when the hyperdynamic circulation impairs the cardiac compensatory reserve so much that the cardioinhibitory effect of NSBBs compromises organ perfusion.17, 18 The problem now is to find a clinical event or a biomarker that tells clinicians that the window is closing. Suggestions have included refractory ascites,13 spontaneous bacterial peritonitis (SBP),16 low mean arterial pressure (MAP),12 the hepatorenal syndrome (HRS), and sepsis19; but the problem remains unsolved.

The aim of this study was to investigate whether NSBB treatment associates with increased mortality in cirrhosis patients as well as in subgroups of decompensated cirrhosis, e.g., in patients with refractory ascites who have been the focal point of the discussion of NSBB safety. We had access to data from three large multicenter randomized trials, and these data gave us a unique opportunity to pursue our aim using high-quality data on an international cohort of 1198 cirrhosis patients with ascites.20

Patients and Methods

Results

We analyzed data from 1188 patients, of whom 559 were NSBB users and 629 were nonusers at the time of randomization: 688 (58%) of the patients had cirrhosis due to alcoholism alone, 53 (4%) due to hepatitis B alone, and 161 (14%) due to hepatitis C alone; 97 (8%) had cirrhosis due to nonalcoholic steatohepatitis or cryptogenic cirrhosis, 83 (7%) had cirrhosis from both alcohol and hepatitis C, 33 (3%) had cirrhosis from autoimmune liver disease, 26 (2%) had cirrhosis from both alcohol and hepatitis B, 10 (1%) had cirrhosis from hemochromatosis, and the remaining 37 (3%) had cirrhosis from various other causes. The NSBB users were more likely than the nonusers to have a history of variceal bleeding (30% versus 13%) and low MAP (≤70 mm Hg, 13% versus 10%; 71-80 mm Hg, 34% versus 27%), but they were less likely to have Child-Pugh class C cirrhosis (24% versus 28%), refractory ascites (46% versus 52%), or serum sodium <135 mmol/L (28% versus 35%) (Table 1).

In total, 286 patients died during the follow-up (median follow-up for survival = 52.5 weeks), and the 52-week cumulative mortality was 23.2% for those who used NSBBs at randomization versus 25.3% for those who did not (Fig. 1). The impression that NSBB use was not associated with mortality persisted after confounder adjustment; the adjusted hazard ratio (HR) for NSBB use versus nonuse at randomization was 0.92 (95% confidence interval [CI] 0.72-1.18) (Table 2). When we examined the combined endpoint hospitalization or death, NSBB treatment appeared even more favorable: the 1-year cumulative risk was 57.1% for those who used NSBBs at randomization versus 63.9% for the others, and the adjusted HR was 0.83 (95% CI 0.71-0.97).

Among the 588 patients with refractory ascites, 258 used NSBBs at randomization and 330 did not. The 52-week cumulative mortality was 30.5% for NSBB users versus 30.9% for nonusers (adjusted HR = 1.02, 95% CI 0.74-1.39). In the group with diuretic-responsive ascites it was 17.0% for NSBB users versus 19.5% for nonusers (adjusted HR = 0.78, 95% CI 0.53-1.16) (Fig. 2). NSBBs were also not associated with increased mortality in any of the other subgroups we examined (Fig. 3). The pattern was the same when we examined the combined endpoint, i.e., hospitalization or death.

Of the 286 decedents, 226 (79%) died from cirrhosis-related causes, 33 (12%) from other known causes, and 27 (9%) from unknown causes. The adjusted HR of cirrhosis-related mortality for NSBB users versus nonusers was 1.00 (95% CI 0.76-1.31) (Table 3). In the subgroup of cirrhosis patients with refractory ascites it was 1.20 (95% CI 0.84-1.72), and in the subgroup of patients with diuretic-responsive ascites it was 0.75 (95% CI 0.48-1.15).

Discussion

This large observational study reflects actual clinical practice with respect to use of NSBBs among patients with decompensated cirrhosis. It showed that use of NSBBs did not increase the risk of death or hospitalization, either in the overall cohort or in the subgroup of patients with refractory ascites or in any of the other patient subgroups we considered.

The major strength of our study is the rigorous and detailed collection of clinically relevant data. These data showed that NSBB users and nonusers were very similar at the time of randomization with respect to cirrhosis severity. However, it is a limitation of our study that we do not know patients' hepatic venous pressure gradient. Nor do we know whether they had gastroesophageal varices or whether those who had varices were treated with ligation, but we can speculate that a large proportion of our NSBB users had large nonbleeding varices as only 30% of them had ever had variceal bleeding. By contrast, those who did not use NSBBs were more likely to have small nonbleeding varices or no varices. As a consequence of the lack of data on hepatic venous pressure gradient or gastroesophageal varices, we are unable to control for confounding by severity of portal hypertension. However, any uncontrolled confounding by portal hypertension is likely to bias our findings toward a higher mortality HR because NSBB users likely have more severe portal hypertension than nonusers. We cannot rule out that NSBB users were healthier in ways that are not described by our data; for example, the similar MAPs in NSBB users and nonusers could indicate that the NSBB-free MAP was higher in the NSBB users, who were therefore healthier. However, this is speculation, and we do not believe that a “healthy user bias” exists.22 Therefore, based on our data, it is more likely that NSBBs do not in fact increase mortality in advanced cirrhosis. Thus, our data support the use of NSBBs as described in the American Association for the Study of Liver Diseases guidelines from 2007,23 and they cannot confirm the recommendation in the newly published UK guidelines that NSBBs are discontinued at the time of SBP, renal impairment, and hypotension.24 Nor can they confirm that NSBBs can safely be continued despite SBP, renal impairment, and hypotension. Clinicians must weigh the risks and benefits of continued NSBB treatment in each patient, and we should emphasize that there was only 1 year of follow-up in the trials that we analyze here.

As many as 29% of patients stopped using NSBBs temporarily or permanently, and those who stopped taking NSBBs were clearly sicker than those who continued taking them. The major limitation of this study is that we do not know what would have happened if all patients had continued using NSBBs. There is no way to answer that question with our data—or, indeed, with any data, considering that our current data reflect actual clinical practice. Only the already existing double-blinded trials of patients randomized to NSBBs or placebo could provide an answer because clinicians are reluctant to discontinue a trial drug. None of them suggested that NSBBs increase mortality. For example, Pascal and Cales randomized 230 cirrhosis patients, 46% of whom had Child-Pugh class C cirrhosis, to propranolol or placebo.25 During the 2-year follow-up, 29 of 118 patients randomized to propranolol stopped treatment: 12 on their own initiative, two at the physician's request, 13 due to adverse effects, and the last two stopped inadvertently. The 2-year mortality risk was 28% in patients randomized to propranolol and 49% in patients randomized to placebo, and this difference was statistically significant. However, Pascal and Cales did not report the proportion of patients who had ascites, so we cannot rule out that a subgroup analysis of their patients with refractory ascites would have shown higher mortality in patients randomized to propranolol. Likewise, the Italian Multicentre Project for Propranolol in Prevention of Bleeding conducted a randomized trial including 174 cirrhosis patients with large varices to investigate the effect of propranolol versus placebo in primary prevention of variceal bleeding.26 Of the 85 patients randomized to propranolol, 25 (29%) stopped taking propranolol during the 30-month follow-up, 23 because of side effects and two because of noncompliance. The investigators concluded that there was no difference in 30-month mortality in the overall cohort or in subgroups of patients who had ascites (44% of the total cohort) or Child-Pugh class C cirrhosis (7% of the total cohort). Subgroup analyses indicated that propranolol did reduce bleeding risk in patients with Child-Pugh class A cirrhosis, but a later individual-patient meta-analysis of this and three other trials found that the NSBB effect on bleeding risk was greater in patients with a Child-Pugh score ≥8 than in patients with a lower Child-Pugh score.27 Nine other trials of NSBBs versus placebo have been conducted, but none of them included a greater proportion of Child-Pugh class C patients than the Pascal and Cales trial or conducted a subgroup analysis of patients with refractory ascites.28 A meta-analysis of the total 11 trials concluded that NSBBs reduce mortality by a nonsignificant amount.28 Trials of NSBBs versus variceal ligation have yielded similar results: patients offered ligation have the same mortality risk as those who are offered NSBBs,3 although they are not at risk of the possible adverse effects of NSBBs. Thus, NSBBs are generally safe in patients with decompensated cirrhosis, but there are no data to determine the safety of NSBBs in the most fragile patients. Therefore, we cannot claim to have shown that NSBBs are safe for all cirrhosis patients, but we can say that the way NSBBs were administered to patients in actual clinical practice was indeed safe.

Our findings conflict with those by Serste et al.7 and Mandorfer et al.,16 who found a higher mortality among NSBB users with refractory ascites and patients with SBP, respectively. These were single-center cohorts with longer follow-up and patients with more advanced disease evidenced by higher median MELD scores of 18 and 19, respectively. We can speculate that their patients continued using NSBBs past the suggested points where the effects of NSBBs may increase mortality. Another possibility is that their NSBB users had more severe cirrhosis to begin with than their nonusers did, and that was clearly the case in the study by Serste et al.7 It included 151 cirrhosis patients with refractory ascites, and the 1-year survival was only 19% in NSBB users versus 64% in nonusers. Our 1-year survival in patients with refractory ascites was 69% in both NSBB users and nonusers. In their study, NSBB users were more likely than nonusers to be Child-Pugh class C and to have a history of variceal bleeding and hepatic encephalopathy. The NSBB users also had a higher total bilirubin and a lower serum sodium, and, like others before us,8, 11, 15, 29-31 we question whether the investigators successfully managed to adjust for the substantial confounding disfavoring NSBB use. Their decision to use significance testing to identify confounding factors is not recommended,32 and it likely resulted in substantial residual confounding by cirrhosis severity. The study by Mandorfer et al. may also suffer from confounding,16 and they do not explain why they adjusted for presence of varices and Child-Pugh class but not for gender, cirrhosis etiology, and other markers of cirrhosis severity, although they were different for NSBB users and nonusers. Child-Pugh class has only three possible levels—A, B, and C—so there can be substantial residual confounding in their analysis. Finally, it must be remembered that, as a result of the randomized trials of NSBBs for primary or secondary prevention of variceal bleeding, NSBBs are nowadays given to cirrhosis patients with severe portal hypertension. These patients have higher mortality than patients with lesser portal hypertension,33 so our default expectation should be that NSBB users have higher mortality than nonusers. That expectation should temper the urge to change treatment recommendations on the basis of singular observational studies.

In conclusion, NSBBs did not increase all-cause or cirrhosis-related mortality in our overall cohort of cirrhosis patients with ascites or in any of the subgroups of more fragile patients we examined. NSBBs were frequently discontinued, primarily in patients who experienced stressful clinical events. We cannot clarify whether the decision to discontinue NSBBs saved these patients or had no impact at all, but our findings suggest that clinicians can continue to use NSBBs like they did when these randomized trials were conducted in 2006-2008.