Pharmacological versus endoscopic therapy in the prevention of variceal hemorrhage: And the winner is…


  • Jaime Bosch,

    Corresponding author
    1. Hospital Clinic, Liver Unit, Centro de Investigación Biomédica en Red de Enfermedades Hepáticas y Digestivas (CIBERehd), University of Barcelona, Spain
    • Hepatic Hemodynamic Laboratory, Hospital Clinic, CIBERehd, C/Villarroel 170, Barcelona 08036, Spain
    Search for more papers by this author
  • Guadalupe Garcia-Tsao

    1. Digestive Diseases Section and Liver Center, Yale University, New Haven, CT
    Search for more papers by this author

  • See Article on Page 825

  • Potential conflict of interest: Dr. Bosch is a consultant for Nicox. He is also a consultant for and received grants from Astellas.

Preventing hemorrhage from esophageal varices has been attempted in many different ways over the years.1 Historically, this was first approached by means of prophylactic portacaval surgical shunts. These procedures were very useful in preventing hemorrhage, but, as demonstrated in the first randomized controlled trials (RCTs) in the field of hepatology, shunting was associated with unacceptable rates of disabling encephalopathy and decreased survival. These results proscribed the further use of prophylactic shunt surgery (when thousands of prophylactic shunts had already been performed all over the world).1 The same restriction on prophylactic surgical shunts has been extended to the use of transjugular intrahepatic portosystemic shunts when this new type of portosystemic shunt became clinically available.2, 3


EBL, endoscopic band ligation; EIS, endoscopic injection sclerotherapy; HVPG, hepatic vein pressure gradient; NSBB, nonselective beta-blocker; RCT, randomized controlled trial.

The second type of therapy to prevent first variceal hemorrhage was endoscopic injection sclerotherapy (EIS). However, even though initial reports from single-center studies were enthusiastic, larger cooperative trials tempered this enthusiasm, and given uncertain results from meta-analyses and a high rate of complications, it became clear that EIS was not the answer either.1

From a pathophysiological point of view, it is clear that a therapeutic strategy which ameliorates portal hypertension—the primary factor leading to the formation, dilatation, and rupture of esophageal varices—should be more effective than a local therapy aimed at varices themselves. Even if varices are effectively “eradicated”, recurrence is likely if portal pressure is not reduced.4 Because of this, the advent of nonselective beta-blockers (NSBBs), mainly propranolol and nadolol, in the pharmacological treatment of portal hypertension appeared to solve the issue. In carefully conducted RCTs, these drugs were shown to be highly effective in decreasing the risk of variceal bleeding by about 50%5, 6 and in decreasing the risk of bleeding from portal hypertensive gastropathy.7 The beneficial effect of NSBBs has been linked to their ability to decrease the portal pressure gradient, assessed clinically through measurement of the hepatic vein pressure gradient (HVPG).8 Many studies have shown that prevention of variceal hemorrhage (first or recurrent) is maximal when HVPG is reduced to values ≤12 mm Hg or by at least 20% from baseline values (the so-called “HVPG responders”).9, 10 Importantly, HVPG responders have a lower probability of developing other complications of portal hypertension that are frequently associated with varices, such as ascites, encephalopathy, and spontaneous bacterial peritonitis.10–12 As such, and unlike local therapies, therapies that correct portal hypertension can change the natural history of cirrhosis and improve survival.10, 11 The main inconvenience of NSBBs is that approximately 15% of patients may have absolute or relative contraindications to therapy and that another 15% require dose reduction or discontinuation due to its common side-effects (e.g., fatigue, weakness, shortness of breath) that resolve upon discontinuation but precludes patients from using these drugs.13

For years, prevention of variceal bleeding became the domain of NSBBs, the only innovation being the addition of low doses of mild vasodilators, such as isosorbide mononitrate to enhance portal pressure reduction, and the investigation of other drugs and drug combinations such as carvedilol,14–16 the combination of propranolol and prazosin,17 and more recently the use of statins,18 which had a greater portal pressure–reducing effect than NSBBs. Also, efforts were directed at obviating repeat HVPG measurements and at better defining a “good response” in order to reduce the “grey zone” of patients that are HVPG nonresponders but who do not bleed during follow-up.19

The panorama changed when endoscopic band ligation of varices (EBL) emerged as a new local therapy that, compared to EIS, was safer and probably more effective.20 This prompted the performance of prospective studies to assess its efficacy in preventing first variceal hemorrhage. Initial studies comparing EBL versus no treatment deserve no comment. Such studies were unjustified (and are considered unethical) because an effective therapy (NSBB) already existed at the time. Eighteen RCTs have compared EBL versus NSBBs, and even though overall results show a benefit of EBL in preventing first variceal hemorrhage,21, 22 when trials with an appropriate treatment allocation and a longer follow-up are analyzed, this borderline benefit of EBL disappears.23 Additionally, no differences in first variceal bleeding rates are observed in studies that include 100 patients or more (Fig. 1). This suggests that small, single-center RCTs with a shorter follow-up may have had more biases, which is unavoidable in this type of RCT that cannot be performed or assessed blindly. Clearly, the field would have benefited more from the performance of two thoughtfully designed, adequately powered, large multicenter clinical trials than from many small studies that create confusion when combined in a meta-analysis and dilute the results of good studies.

Figure 1.

Meta-analysis of randomized controlled trials (RCTs) comparing endoscopic band ligation (EBL) versus nonselective beta-adrenergic blockers (NSBBs). The RCTs included are grouped according to sample size (A ≥100 patients, B <100 patients) and type of publication (C, only abstracts). As seen in (A), in the relatively large trials there is no advantage of EBL over NSBB, and an initial trend toward better results with EBL observed with earlier trials disappears in more recent trials. The apparent advantage of EBL (D) results from (B) underpowered trials or from (C) studies reported only as abstracts. CI, confidence interval; RR, relative risk. (Figure modified from Bosch et al.4).

Because both NSBBs and EBL appear to be equally effective in preventing first variceal hemorrhage, side-effects of each become an important issue. Although the quantity of side-effects is greater with NSBBs than with EBL,21, 22 it is the quality of side-effects that is more important. Whereas no lethal side-effects have been reported with the use of NSBBs in these trials (or over the past 25 years), three deaths resulting from EBL-induced bleeding ulcers were reported in these 18 trials.21, 22 It is counterintuitive that a therapy that is supposed to prevent bleeding can actually cause bleeding. Ultimately, whether bleeding is from varices or from an EBL-induced ulcer makes no difference to the patient. The issue of which is the best treatment for primary prophylaxis has not yet been settled, and there are centers that perform EBL as the predominant therapy for varices, whereas others prefer the more rational approach of starting with NSBBs and switching to EBL if there is intolerance to NSBBs.

The apparent lack of benefit of a therapy that acts by reducing portal pressure (use of NSBBs) compared to a local therapy (undergoing EBL) is probably due to the fact that only up to 50% of patients treated with NSBBs are HVPG responders. A drug that could increase the rate of HVPG response would most probably tip the balance in favor of pharmacological treatment. As shown by the study by Tripathi et al. in this issue of HEPATOLOGY,24 carvedilol could be such a drug.

The RCT by Tripathi et al. is the first to show a statistically significant benefit for a beta-blocker compared to EBL in the prevention of first variceal hemorrhage.24 It should be noted that this is the first study using carvedilol, an NSBB with intrinsic anti–alpha-adrenergic vasodilatory activity. It thereby mimics the combination of propranolol and prazosin, which has dramatic portal pressure–reducing effects.17 In a controlled hemodynamic study, carvedilol (mean dose, 31 mg/day) was shown to lead to a greater reduction in HVPG than propranolol, although by causing systemic vasodilatation, it decreased mean arterial pressure and worsened sodium retention.15 Therefore, there is a good rationale for the use of carvedilol in portal hypertension at the doses used in this study (maximum, 12.5 mg/day) that were not associated with hypotension. Carvedilol had not been previously tested, probably because of the reluctance of industry to invest in diseases with a low number of patients (e.g., cirrhosis) compared to other more common (e.g., cardiovascular) diseases. The study by Tripathi et al. is an investigator-initiated study not supported by industry.

Unfortunately, the trial has several problems that limit its conclusions: First, the trial was underpowered and thus its results are not very robust. It has been calculated that in this setting, a sample size of about 400-500 patients is necessary.25 Second, the study lacks measurements of HVPG and it is therefore unclear whether the benefit of carvedilol was due to a greater HVPG response or whether it was due to a failure of EBL. Given the lower dose of carvedilol used compared to the previous hemodynamic study,15 it is possible that carvedilol caused a less marked HVPG reduction. Analysis of the data shows, however, that the results with carvedilol rank among the best results reported for prophylactic drug therapy (10% first hemorrhage rate) whereas their results with EBL rank among the worst (23% rate). The poor effects of EBL could have been due to the fact that the median time from randomization to EBL was 21 days (with a maximum of 171 days) and in some patients, bleeding occurred in this period. Although this was clearly unfavorable for EBL, it reflects an additional practical problem with EBL particularly in countries where due to scheduling issues the lag time from screening to therapeutic endoscopy are usually this long. On the other hand, the promising results obtained with carvedilol in this study should be confirmed in specific studies, preferably through double-blind RCTs that compare carvedilol with NSBBs and that include HVPG measurements.

In summary, it is clear that carvedilol has emerged as a new effective agent in the treatment of portal hypertension that appears to be more effective than endoscopic therapy. However, further studies are required to assess whether carvedilol is a better option than standard NSBBs. Such studies should define the optimal dosage, assess the correlation between hemodynamic effects and clinical benefit, and confirm whether carvedilol could be a safe and useful drug in other settings. Such settings include prevention of recurrent hemorrhage and treatment of patients on NSBBs who cannot tolerate the drug or who fail to achieve an adequate HVPG response. There is no clear winner in the contest between current therapies for the prevention of first variceal hemorrhage, but certainly the emphasis placed on EBL is fading as new evidence accumulates and new drugs are introduced.


We acknowledge the help of Dr. J. G. Abraldes and Dr. A. Berzigotti in the preparation of this article. The CIBERehd is funded by the Spanish Instituto de Salud Carlos III, Ministry of Science and Innovation. Grant support: P30 DK034989, FIS PI06/0623 and Fundación BBVA.