Dig Dis I992:l0(suppl 1):56—64
Clínica Medica R. Univcrsitá di Palermo and Divisione di Medicina Ospedale V. Cervello, Palermo. Italy
Keywords Meta-analysis Randomized clinical trials Cirrhosis Bleeding P-Blockers Portal hypertension
Prevention of Upper Gastrointestinal Bleeding from Portal Hypertension in Cirrhosis: Rationale for Medical Treatment
Abstract We updated meta-analysis and critical descriptive analysis of randomized clinical trials (RCTs) assessing the value of Pblockers in preventing first bleeding (prophylactic) or rebleed ing (therapeutic) and on survival of patients with cirrhosis. Both the methods of Peto-Mantel-Haenszel and DerSimonian-Laird were used to assess the heterogeneity and obtain cumulative estimates of treatment effects: the L'Abbé plot was also used for a visual assessment of heterogeneity in the direc tion of treatment effect. Seven prophylactic and nine thera peutic RCTs were analysed. p-Blockers uniformly reduced the bleeding risk and revealed a trend toward improved survival in non-ascitic. well-compensated patients in both the prophy lactic and therapeutic sets of RCTs. Discordant results were found in patients with ascites or in poor functional condition.
Medical treatment for the prevention of bleeding or rebleeding in cirrhosis is based on drugs producing a sustained decrease in por tal hypertension. Although there is no linear relationship between the level of portal hyper tension and bleeding incidence, several stud ies (including one double-blind randomized
clinical trial of prophylactic p-blockers) have shown that bleeding does not occur or very rarely occurs below the threshold value of 12 mm Hg of hepatic venous pressure gra dient (HVPG) [1], Several drugs can decrease portal hyper tension in cirrhosis [2], However, only P-
Prof. Luigi Pagliaro Istituto Clinica Mcdica R Ospedale V. Cervello Via Trabucco 180 1-90146 Palermo (Italy)
© 1992 S. Karger AG, Basel 0257-2753/92/ 0I07-0056S2.75/0
Downloaded by: Kings's College London 137.73.144.138 - 11/16/2017 8:32:43 AM
Luigi Pagliaro Gennaro D ’A mico Fabio Tine Linda Pasta
Methods Retrieval and selection of RCTs, evaluation of effi cacy. assessment of heterogeneity and pooling of re sults were previously described [6]. Briefly, the relative risks of bleeding and death in treated patients as com pared to controls were assumed as measures of treat ment effect. Relative risks of bleeding and death in each individual trial were computed and pooled by the Peto-Mantel-Haenszel method, as described by Collins et al. [7], Heterogeneity was evaluated by x: statistical test and by visual evaluation in the graphic display provided by the L'Abbc plot [8], in which the coordi nates of each dot correspond to the outcome rates in each treatment group of each individual RCT. The diagonal in the plot corresponds to the equivalence between treatments. Dots below or above the equiva
lence line indicate that the treatment represented on the vertical axis is better or worse, respectively, than the treatment represented on the horizontal axis. Dis persion of dots below and above the equivalence line suggests heterogeneity in the direction of treatment effect and makes pooling unreliable. In this update we also report the absolute risk reduction in bleeding and death rate within each individual RCT, i.e. the differ ences in outcome rates between treated patients and controls. Weighted average of rate differences was used for measuring the cumulative absolute risk reduc tion of each set of RCTs [9]. The method provides a more conservative estimate of the cumulative thera peutic efficacy than the pooled relative risk computed by the method of Peto-Mantel-Haenszel when there is heterogeneity across trials [ 10].
Results RCT o f fi-Blockers for llte Prevention o f First Bleeding ('Prophylactic') (table 1, fig. 1. 2) The present update is based on seven RCTs [11-17], five of which [11-15] were included in our previous meta-analysis. Three RCTs [15-17] are reported as abstracts. All RCTs admitted patients with cirrhosis and esophageal varices. Cirrhosis was compen sated or mildly decompensated in most pa tients in three RCTs [12. 14. 15], In three oth ers [ 11.13. 17] 46.24 and 26% of the patients, respectively, were in Child C class, whereas no data were reported in the last RCT [16], Varices were large in all patients in three RCTs [12-14], large or medium-sized in two [11, 16], of any size in one [17], In another RCT [15], the presence of esophageal varices and HVPG > 12 mm Hg was requested for admission. These data and others, reported in table 1, show wide differences across RCTs (i.e.: in admission criteria, drug used, sample size, fol low-up and baseline bleeding and death risk, as reflected by bleeding and death rates in untreated controls).
57
Downloaded by: Kings's College London 137.73.144.138 - 11/16/2017 8:32:43 AM
blockers have been extensively studied from a pathophysiological point of view and assessed in clinical trials. Chronic administration of (Tblockers de creases portal pressure in about two thirds of patients with cirrhosis [3, 4], reaching the critical value of 12 mm Hg in some of them [5], This hemodynamic effect is probably due to a reduction in cardiac output and to unop posed a-adrcnergic splanchnic arteriolar con striction. both producing the combined result of a decrease in portal flow and pressure [2], It is of interest that both cardiac output and por tal blood flow are abnormally high in cirrho sis. playing a contributory role in portal hy pertension [I], This rationale has stimulated many clinical studies. Our meta-analysis [6] of randomized clini cal trials (RCTs) of treatments for the pre vention of first bleeding (‘prophylactic’) or rebleeding (‘therapeutic’) included five pro phylactic and nine therapeutic RCTs of Pblockers versus placebo. The present update includes three further RCTs and expanded data from two RCTs (originally reported in preliminary abstracts and) now available as final reports.
Fig. 1. Prophylactic ß-blockcrs for the prevention of bleeding. Analysis (L’Abbe plot) of the results of var ious RCTs published in the literature [11-17].«= Pro pranolol; □ = controls. Sample size: ■ s* 100; ■ 50-99; •
Clínica Medica R. Univcrsitá di Palermo and Divisione di Medicina Ospedale V. Cervello, Palermo. Italy
Keywords Meta-analysis Randomized clinical trials Cirrhosis Bleeding P-Blockers Portal hypertension
Prevention of Upper Gastrointestinal Bleeding from Portal Hypertension in Cirrhosis: Rationale for Medical Treatment
Abstract We updated meta-analysis and critical descriptive analysis of randomized clinical trials (RCTs) assessing the value of Pblockers in preventing first bleeding (prophylactic) or rebleed ing (therapeutic) and on survival of patients with cirrhosis. Both the methods of Peto-Mantel-Haenszel and DerSimonian-Laird were used to assess the heterogeneity and obtain cumulative estimates of treatment effects: the L'Abbé plot was also used for a visual assessment of heterogeneity in the direc tion of treatment effect. Seven prophylactic and nine thera peutic RCTs were analysed. p-Blockers uniformly reduced the bleeding risk and revealed a trend toward improved survival in non-ascitic. well-compensated patients in both the prophy lactic and therapeutic sets of RCTs. Discordant results were found in patients with ascites or in poor functional condition.
Medical treatment for the prevention of bleeding or rebleeding in cirrhosis is based on drugs producing a sustained decrease in por tal hypertension. Although there is no linear relationship between the level of portal hyper tension and bleeding incidence, several stud ies (including one double-blind randomized
clinical trial of prophylactic p-blockers) have shown that bleeding does not occur or very rarely occurs below the threshold value of 12 mm Hg of hepatic venous pressure gra dient (HVPG) [1], Several drugs can decrease portal hyper tension in cirrhosis [2], However, only P-
Prof. Luigi Pagliaro Istituto Clinica Mcdica R Ospedale V. Cervello Via Trabucco 180 1-90146 Palermo (Italy)
© 1992 S. Karger AG, Basel 0257-2753/92/ 0I07-0056S2.75/0
Downloaded by: Kings's College London 137.73.144.138 - 11/16/2017 8:32:43 AM
Luigi Pagliaro Gennaro D ’A mico Fabio Tine Linda Pasta
Methods Retrieval and selection of RCTs, evaluation of effi cacy. assessment of heterogeneity and pooling of re sults were previously described [6]. Briefly, the relative risks of bleeding and death in treated patients as com pared to controls were assumed as measures of treat ment effect. Relative risks of bleeding and death in each individual trial were computed and pooled by the Peto-Mantel-Haenszel method, as described by Collins et al. [7], Heterogeneity was evaluated by x: statistical test and by visual evaluation in the graphic display provided by the L'Abbc plot [8], in which the coordi nates of each dot correspond to the outcome rates in each treatment group of each individual RCT. The diagonal in the plot corresponds to the equivalence between treatments. Dots below or above the equiva
lence line indicate that the treatment represented on the vertical axis is better or worse, respectively, than the treatment represented on the horizontal axis. Dis persion of dots below and above the equivalence line suggests heterogeneity in the direction of treatment effect and makes pooling unreliable. In this update we also report the absolute risk reduction in bleeding and death rate within each individual RCT, i.e. the differ ences in outcome rates between treated patients and controls. Weighted average of rate differences was used for measuring the cumulative absolute risk reduc tion of each set of RCTs [9]. The method provides a more conservative estimate of the cumulative thera peutic efficacy than the pooled relative risk computed by the method of Peto-Mantel-Haenszel when there is heterogeneity across trials [ 10].
Results RCT o f fi-Blockers for llte Prevention o f First Bleeding ('Prophylactic') (table 1, fig. 1. 2) The present update is based on seven RCTs [11-17], five of which [11-15] were included in our previous meta-analysis. Three RCTs [15-17] are reported as abstracts. All RCTs admitted patients with cirrhosis and esophageal varices. Cirrhosis was compen sated or mildly decompensated in most pa tients in three RCTs [12. 14. 15], In three oth ers [ 11.13. 17] 46.24 and 26% of the patients, respectively, were in Child C class, whereas no data were reported in the last RCT [16], Varices were large in all patients in three RCTs [12-14], large or medium-sized in two [11, 16], of any size in one [17], In another RCT [15], the presence of esophageal varices and HVPG > 12 mm Hg was requested for admission. These data and others, reported in table 1, show wide differences across RCTs (i.e.: in admission criteria, drug used, sample size, fol low-up and baseline bleeding and death risk, as reflected by bleeding and death rates in untreated controls).
57
Downloaded by: Kings's College London 137.73.144.138 - 11/16/2017 8:32:43 AM
blockers have been extensively studied from a pathophysiological point of view and assessed in clinical trials. Chronic administration of (Tblockers de creases portal pressure in about two thirds of patients with cirrhosis [3, 4], reaching the critical value of 12 mm Hg in some of them [5], This hemodynamic effect is probably due to a reduction in cardiac output and to unop posed a-adrcnergic splanchnic arteriolar con striction. both producing the combined result of a decrease in portal flow and pressure [2], It is of interest that both cardiac output and por tal blood flow are abnormally high in cirrho sis. playing a contributory role in portal hy pertension [I], This rationale has stimulated many clinical studies. Our meta-analysis [6] of randomized clini cal trials (RCTs) of treatments for the pre vention of first bleeding (‘prophylactic’) or rebleeding (‘therapeutic’) included five pro phylactic and nine therapeutic RCTs of Pblockers versus placebo. The present update includes three further RCTs and expanded data from two RCTs (originally reported in preliminary abstracts and) now available as final reports.
Fig. 1. Prophylactic ß-blockcrs for the prevention of bleeding. Analysis (L’Abbe plot) of the results of var ious RCTs published in the literature [11-17].«= Pro pranolol; □ = controls. Sample size: ■ s* 100; ■ 50-99; •
