of beta-blockers in the prevention of recurrent bleeding in A meta-analysis of 12 selected randomized trials was patients with cirrhosis and to assess the effect of this drug performed to assess the efficacy of beta-blockers in the on long-term survival by a new meta-analysis, including reprevention of rebleeding and the effect on long-term surcently published RCTs. vival in patients with cirrhosis. Five end points were assessed: rebleeding, variceal rebleeding, death, death MATERIALS AND METHODS from bleeding, and adverse events. Analyses were performed according to the intention-to-treat method. For Meta-analyses were conducted according to a predetermined protocol following the recommendations of Sacks et al., 23 but were not each end point, heterogeneity and treatment efficacy limited to English language RCTs. Four events were chosen as end were assessed by the Der Simonian and Peto methods.
points to estimate clinical efficacy: first rebleeding from any cause, When a significant difference was observed, sensitivity first rebleeding from esophageal varices, death, death from bleeding, analyses were performed by successive stratifications and one end point to estimate safety: the incidence of adverse events.
according to treatment duration, cause of initial bleeding, use of placebo, type of beta-blocker, type of publica-Literature Search tion, certainty of randomization, severity of cirrhosis, MEDLINE and manual search were combined because we had interval between index bleed and randomization, and previously demonstrated that MEDLINE search alone is not sensimethodological quality. Beta-blockers significantly intive enough. General reviews, references of published RCTs, letters creased the mean percentage of patients free of rebleedto pharmacological companies, and Current Contents were also used. ing (21% mean improvement rate, CI 95%: 10%-32%, P Γ΅ RCTs published in English, French, German, and Spanish were in- .001, relative risk 1.42), the mean percentage of patients cluded. free of variceal rebleeding (20% mean improvement rate, CI 95%: 11%-28%, P Γ΅ .001), the mean survival rate (5.4% Criteria of Inclusion mean improvement rate, CI 95%: 0%-11%, P Γ
.05, relative
To be included in this meta-analysis, a RCT had to fulfill the risk 1.27), the mean percentage of patients free of bleedfollowing criteria: it had to be published, clearly randomized, be pro- ing death (7.4%, CI 95%: 2%-13%, P Γ΅ .01, relative risk spective, include patients with cirrhosis and esophageal varices, in- 1.50). Five patients would need to be treated with betaclude patients enrolled after initial gastrointestinal bleeding, com- blockers to prevent one rebleeding episode, 14 treated pare a noncardioselective beta-blocker to placebo or no specific to prevent one death, and 13 treated to prevent one treatment, include patients with mean follow-up of more than 11 months, and use at least one clinical end point among the following: death from bleeding. There was no significant heterogegastrointestinal bleeding, gastrointestinal bleeding from esophageal neity among studies by both methods of analysis. In pavarices, and death. tients with esophageal varices, beta-blockers significantly increase the mean percentage of patients free of Criteria of Noninclusion rebleeding and the mean survival rate at 2 years. (HEPA-RCTs that were retrospective, not randomized, or that compared TOLOGY 1997;25:63-70.) different treatments without control groups were not included. RCTs that included only patients without cirrhosis, patients without initial From the 1 Service d'He Β΄pato-Gastroente Β΄rologie, Ho Λpital Pitie Β΄-Salpe Λtrie `re, 75651; 2 INian and Laird method. 27 To better ensure combinability, nine sensi-SERM U-24, Ho Λpital Beaujon, 92118 Clichy; and