|
Pharmacological Treatment of Portal
Hypertension:
An Evidence-Based Approach
Gennaro D'Amico, M.D., Luigi Pagliaro, M.D., and Jaime Bosch,
M.D., From Department of Liver
Transplantation and Hepatobiliary Medicine, Royal Free Hospital,
London, United Kingdom
Semin Liver Dis 19(4):475-505, 1999. © 1999 Thieme Medical
Publishers
http://www.medscape.com/viewarticle/416617_print
Abstract and Introduction
Abstract
Continuing advances in the knowledge of the pathophysiology of
portal hypertension result in the progressive expansion of the
spectrum of drugs with a potential role for clinical practice,
with objectives that now tend to include the prevention of the
enlargement or even the development of esophageal varices. This
systematic review summarizes the evidence of efficacy of drug
therapy for portal hypertension and draws recommendations for
clinical practice. Although there is not yet enough evidence to
support the treatment for the prevention of the development or
enlargement of varices, nonselective beta-blockers are the
first-choice therapy to prevent the first bleeding in patients
with medium or large-sized varices and rebleeding in patients
surviving a bleeding episode. The clinical role of
isosorbide-5-mononitrate either alone or in association with
beta-blockers still remains unsettled. Vasoactive drugs are
generally effective and safe in controlling acute variceal
bleeding, although the evidence is not equivalent for each of
them.
Introduction
The continuous development of the knowledge in pathophysiology
of portal hypertension has substantially expanded the spectrum
of potentially effective drugs. The first step in the
pharmacological treatment of portal hypertension was the
reduction of the portal blood inflow. This is abnormally
increased because of splanchnic and systemic vasodilatation that
result in the hyperdynamic circulation. This causes a
progressively increasing amount of blood to enter in the
splanchnic bed and hence in the portal vein. Several
vasoconstrictors have been shown to be effective in reducing
portal blood inflow and, as a consequence, portal pressure.
Because the hyperdynamic circulation is accompanied and
maintained by hypervolemia, a low-sodium diet and spironolactone
have also been shown to reduce portal pressure, even in
compensated patients without sodium retention. Most importantly,
in the last decade it has been established that, contrary to
what was previously thought, the increase in hepatic vascular
resistance is not only a mechanical consequence of the
distortion of the hepatic structure but also the result of a
dynamic component due to the active contraction of vascular
smooth muscle cells in the liver vasculature and of
extravascular cells, such as myofibroblasts or activated hepatic
stellate cells surrounding the sinusoidal endothelial cells.[1]
This dynamic component is enhanced by endothelin and
a-adrenergic stimulation and is reduced by nitric oxide,
prostacyclin, and several vasodilating drugs. Thus, vasodilators
have become another important category of drugs for portal
hypertension.
Combination therapy (associating a nonselective beta-blocker
with a nitrovasodilator) and the identification of hemodynamic
markers of long-term clinical efficacy of the treatment is
opening the clinical scene to the tailoring of drug therapy to
the individual patient, similarly to what is currently done to
treat arterial hypertension. In that regard, it is now well
established that the risk of bleeding from varices is nil when
the portal pressure gradient (usually evaluated as the hepatic
vein pressure gradient [HVPG]) is reduced to 12 mm Hg or below.
Even without reaching this target, the risk of bleeding is
extremely low when the HVPG is decreased by at least 20% from
its baseline value.[2,3]
Although the pathophysiological basis for pharmacological
treatment of portal hypertension is discussed elsewhere in this
issue, here the evidence of clinical efficacy of drugs is
summarized and recommendations for clinical practice are drawn.
Objectives Of Treatment
The principal objectives of pharmacological treatment for portal
hypertension have been for many years the prevention of bleeding
or rebleeding and the control of acute bleeding, aiming at
improving survival by reducing bleeding-related deaths. More
recently, the hypothesis that drug therapy may prevent varices
from developing or becoming large enough to be at risk of
rupture has been tested in one trial reported in abstract and in
another still in progress.
Here, we review the body of evidence of clinical effectiveness
of drug therapy for portal hypertension according to the
baseline risk in different clinical scenarios to provide
reliable estimates of the treatment effects for clinical
practice.
Review Of Randomized Clinical Trials
To synthesize the evidence of efficacy of treatments, we pooled
data from relevant groups of trials according to a random
effects model and reported treatment effects as absolute rate
difference (ARD) and number of patients needed to be treated (NNT
= 1/absolute rate difference).[4-6] The number of patients
needed to be treated to save a harmful treatment effect (NNH =
1/ absolute rate difference of harmful events due to treatment)
is also reported when appropriate. ARD is always calculated as
the event rate in the experimental group minus the event rate in
the control group. In this way, a negative value indicates a
beneficial effect of the experimental treatment. Ninety-five
percent confidence intervals (CI) are also reported. When the
confidence limits are both negative, the favorable experimental
treatment effect is statistically significant, whereas an
entirely positive CI indicates a significantly harmful effect of
the experimental treatment. Cumulative meta-analysis is used to
graphically represent the modification of the pooled estimate of
the treatment effect at any time a new randomized clinical trial
(RCT) has been reported.
Recommendations for clinical practice are derived from the
evidence of efficacy of treatments. They are graded in six
levels according to the strenght of evidence (validity and
consistency of the studies), from A1 to C2 (Table 1). This
grading of recommendations has been recently suggested by a
Consensus Conference of the American College of Chest
Physicians.[7]
Prevention Of First Bleeding
Relevant information from the natural history of portal
hypertension has been recently thoroughly reviewed,[8] and the
following key points may be considered as a basis for the
definition of the objectives of treatment before the first
bleeding.
Information on the incidence of esophageal varices is
insufficient. One study reported a 10-year incidence of 90%[5-];
however, in that study the 5-year cumulative proportion of
patients with varices was near 50% but half of these patients
had varices when they entered the study. Thus, the net 5-year
incidence was near 25% (about 5% per year). After 5 years of
observation only 50 of the 532 patients originally included were
still at risk, and confidence intervals of estimates were very
large after that point. More recently, we reported cumulative
data from two prospective studies of the natural history of
cirrhosis performed at our unit including 1,123 patients without
varices when first diagnosed. After a 10-year follow-up, the
cumulative proportion of patients with esophageal varices was
44% with 123 patients still at risk at the beginning of the 10th
year of observation.[8] The incidence of varices was quite
constant and near to 4.4% per year. Because the incidence of
varices observed in the former study[9] during the first 5 years
was very close to that observed at our unit and no other data
are available, we should assume the best available estimate of
the incidence of varices is between 4 and 5% per year. This
would require a sample size of 170 patients per group in a trial
for the prevention of varices with a 5-year follow up, to show a
50% difference between the treatment and control group (a = 0.05
and b = 0.80).
From transversal studies we also know that esophageal varices
develop only in patients with HVPG >10 mm Hg, although we do not
know the incidence of varices in patients with HVPG above this
level or whether the risk of developing varices increases with
increasing values of HVPG above this threshold value. It may be
expected that worsening of liver function and continuing
exposure to alcohol are related to the risk of developing
varices, but it has also been shown, in a prospective study,
that HVPG increases with worsening of liver function and
continuing alcohol abuse and decreases when liver function
improves and with interrupting alcohol abuse.[10] Thus, to the
best of our knowledge, HVPG is the only risk factor for
developing esophageal varices and should be the parameter to
select patients for preprimary prophylaxis of esophageal varices.
Once developed, varices increase in size from small to large in
approximately 12% patients per year in the first 2 years after
endoscopic diagnosis of varices. This is the median value from
the four studies reporting this information.[8-11]
The following have been reported as risk indicators of bleeding
from esophageal varices: variceal size, red marks on the
variceal wall, Child-Pugh class, HVPG increasing with time,
variceal pressure, and variceal wall tension. No bleeding has
been reported in patients with HVPG <12 mm Hg, and this value is
considered the threshold value for variceal bleeding.[8]
The 2-year bleeding incidence is near 30% in patients with
medium or large-sized varices and near 10% in patients with
small varices.[12]
Nonselective beta-blockers, nitrates, and spironolactone are the
only drugs assessed for primary prophylaxis of variceal
bleeding. Although several risk factors for variceal bleeding
have been identified, clinical trials assessing pharmacological
treatments for prophylaxis of first bleeding selected patients
mainly on the basis of variceal size.[13] Information on the
treatment effect according to the presence of ascites is also
reported in some studies. Thus, the prevention of first variceal
bleeding has been assessed in clinical scenarios mainly
identified by the variceal size and clinical decompensation.
Nonselective Beta-Blockers
As discussed elsewhere in this issue, these drugs reduce portal
pressure through a reduction in portal and collateral blood
flow. This is a consequence in part of the reduction of cardiac
output through the blockade of b1-cardiac receptors and in part
of the splanchnic vasoconstriction due to the blockade of
b2-splanchnic receptors with unopposed a-adrenergic activity.
Twelve RCTs assessing nonselective beta-blockers for the
prophylaxis of the first variceal bleeding and 1 for the
prevention of varices have been reported.[14-26] One[23] is not
included in this review because although it included 319
patients, the results cannot be interpreted in a clinically
useful way. In fact, in this study, patients with chronic liver
disease were included, with or without cirrhosis, or with or
without unknown varices (endoscopy in 45%, barium swallow test
in 25%, and no information at all in 30%). Overall, only 11
patients bled and, although a very heterogeneous patient
population was included, no attempt was done to perform a
reliable subgroup analysis.
Nine of the remaining 11 RCTs were described in detail in
previous meta-analyses.[13,27] Propranolol was assessed in nine
studies and nadolol in two. Placebo or nonactive treatment was
given to the control groups in all trials. The principal
characteristics and results of the 11 studies, which included
1,189 patients, are reported in Table 1 and Figure 1. The
methodological quality of the seven trials published as full
reports was considered to be fair to good.[27] In nine RCTs, the
bleeding rate was reduced and in one was not modified by
beta-blockers. One RCT,[19] considered an outlier[27] because
propranolol was associated with a significant increase in
bleeding risk, was excluded from the pooled estimate of the
treatment effect for bleeding (Table 2, Fig. 1). Overall, the
bleeding rate was 25% in controls and 15% in treated patients,
after a median follow-up of 24 months. The ARD was 210% (CI, 216
to 25; NNT = 10). Also, a small reduction in mortality,
approaching statistical significance (Table 2, Fig. 1), was
found (ARD = 24%; CI, 29% to 0%)
.
FIG. 1. Cumulative meta-analysis (random effects model) of
randomized clinical trials of beta-blockers compared with
placebo or nonactive treatment for prevention of variceal
bleeding in cirrhosis. The treatment effects on bleeding (top)
and mortality (bottom) are reported. Solid circles represent the
absolute risk difference (treated minus controls) obtained by
pooling the results of every new trial with all those reported
previously. The horizontal bars represent the 95% confidence
intervals of each estimate. The vertical line represents the
zero difference (the equivalence between the treatments).
Negative values (on the left of the equivalence line) favor
beta-blockers; positive values favor control treatment (CT). A
significant bleeding risk reduction was evident after the second
study and was consistently confirmed by adding all the
subsequent studies; the important death risk reduction shown by
the first three studies was reduced by the others but it is
consistent and nearly significant.
In the following sections, beta-blockers efficacy is assessed
according to the most important clinical characteristics of
patients included in the RCTs to extract more sound information
for clinical practice.
The Patient with Varices, Independent of Size
In four RCTs, 374 patients with varices of any size were
included (Table 3). In two of these RCTs, however, only patients
with HVPG, respectively, >10 mm Hg19 or >12 mm Hg21 were
included. It is surprising that the baseline risk (the incidence
of bleeding in untreated patients) in these two studies was
similar or even lower compared with the two studies selecting
patients only on the presence of varices. Also, excluding the
outlier trial of Colman et al.,[19] there is a significant
heterogeneity across these studies with a bleeding risk
reduction ranging from 0 to 227%. If we accept this
heterogeneity as a qualitative one (i.e., in the size and not in
the direction of the treatment effect), the pooled risk
reduction is 214%, but it fails to reach statistical
significance with a CI including 0 (from 230 to 0) (Table 2).
Overall mortality rate was also not significantly reduced (ARD =
29%; CI 217 to 0), without heterogeneity across trials.
The Patient with Medium or Large-Sized Varices, Independent of
Signs of Clinical Decompensation of Cirrhosis
Six of 11 RCTs of beta-blockers included patients with medium or
large-sized varices.[14-18,24] Moreover, in two RCTs including
patients with varices of any size, a separate bleeding analysis
was reported for patients with medium or large varices.[20,21]
The experimental drug was propranolol in six studies and nadolol
in two. A total of 811 patients were assessable in these eight
studies. The mean weighted bleeding rate in control patients was
30% after a mean follow-up near to 2 years. The corresponding
bleeding rate in treated patients was 14% (ARD = 216%; CI from
224% to 28%; NNT = 6). No significant treatment effect on
mortality was detected (Table 3; Fig. 2).
FIG. 2. Cumulative meta-analysis (random effects model) of
randomized clinical trials of beta-blockers compared with
placebo or nonactive treatment for prevention of variceal
bleeding in cirrhosis with medium or large esophageal varices.
The treatment effects on bleeding (top) and mortality (bottom)
are reported. Solid circles represent the absolute risk
difference (treated minus controls) obtained by pooling the
results of every new trial with all those reported previously.
The horizontal bars represent the 95% confidence intervals of
each estimate. The vertical line represents the zero difference
(the equivalence between the treatments). Negative values (on
the left of the equivalence line) favor beta-blockers; positive
values favor control treatment (CT). An important bleeding risk
reduction with beta-blockers is consistently confirmed in all
the studies, whereas the favorable effect on mortality shown in
the first studies disappeared by increasing the number of
studied patients.
FIG. 3. Meta-analysis (random effects model) of randomized
clinical trials of IMN compared with propranolol (first and
second panels, top) and of IMN plus nadolol (Nad) compared with
nadolol alone (third and fourth panels, bottom), for the
prevention of first variceal bleeding in cirrhosis. Solid
circles represent the absolute risk difference (treated minus
controls) for each trial and the overall estimates. The
horizontal bars represent the 95% confidence intervals of each
estimate. The vertical line represents the zero difference (the
equivalence between the treatments). Negative values (on the
left of the equivalence line) favor beta-blockers; positive
values favor control treatment (CT). No differences were found
either for bleeding or for mortality in both groups of RCTs.
Four of these eight studies
were pooled in a meta-analysis based on individual patient data,
including 589 patients.[28] The mean 2-year baseline risk of
bleeding was significantly reduced from 27% to 17% by the
treatment (ARD = 210%; CI, 217% to 23%; NNT = 10) (Table 3). No
reduction in overall mortality was found with the treatment,
although mortality from bleeding was reduced from 14% in control
patients to 7% in treated patients (ARD = 27%; CI, 212% to 22%;
NNT = 14).
The Patient with Medium or Large Varices, Without Ascites
The meta-analysis based on individual patient data above
reported[28] is the largest available dataset providing
information on this point and includes all the relevant
information. Overall, 280 patients without ascites were included
in the four RCTs. The 2-year bleeding rate was 25% in the
control patients and 12% in treated patients (ARD = 213%; CI,
222% to 24%; NNT = 8) (Table 3). No mortality data were
reported.
The Patient with Medium or Large Varices with Ascites
Again, the relevant information comes from the meta-analysis
based on individual patient data, reported above.[28] In the
four trials included, there were 305 patients with ascites. The
2-year bleeding rate was 31% in the control patients and 22% in
treated patients (ARD = 29%; CI, 218% to 0%; NNT = 11) (Table
3). Therefore, either the absolute or the relative risk
reduction in patients with ascites is appreciably lower than in
patients without ascites. Moreover, the ARD in patients with
ascites fails to reach statistical significance because the CI
includes 0. However, when assessed by the log-rank test on the
Kaplan-Meyer cumulative estimates, the 2-year bleeding risk
reduction is statistically significant also in patients with
ascites (from 41% to 27%, p = 0.002). No mortality data were
reported.
The Patient with Small Varices
Among four RCTs including patients with varices of any size,
three[20-22] reported separate results in 166 patients with
small varices and showed no benefit from propranolol (n = 81)
compared with placebo (n = 85). No mortality data were reported
in this subset of patients. More recently, preliminary results
from a trial of nadolol compared with placebo only in patients
with small varices has been reported.[25] After a mean follow-up
of 16 months, 2 of 54 patients given placebo and none of 50
given nadolol bled. The difference was not significant. No
difference in mortality was observed between the treatment
groups (four deaths with placebo and three with nadolol).
FIG. 4. Meta-analysis (random effects model) of randomized
clinical trials of terlipressin compared with placebo for acute
variceal bleeding: treatment effect on overall mortality. Top,
traditional meta-analysis; bottom, cumulative meta-analysis.
Solid circles represent the absolute risk difference (treated
minus controls) for each trial and the overall estimates. The
horizontal bars represent the 95% confidence intervals of each
estimate. The vertical line represents the zero difference (the
equivalence between the treatments). Negative values (on the
left of the equivalence line) favor terlipressin; positive
values favor placebo. Cumulative meta-analysis shows that
mortality was consistently reduced since the first study and
that the effect reached statistical significance with the third
study in 1990.
When the results on bleeding
of this RCT are pooled with two [20,21] of the three above
mentioned studies (the third[22] reports only the relative risk
of bleeding and not the number of observed events), the bleeding
rate was 7 in 100 with placebo and 2 in 91 with nadolol or
propranolol (ARD = 25%; CI, from 211% to 2%).
The Patient Without Varices
There is only one trial reported in abstract form[26] and
another still in progress.[29] In the study reported in abstract
form and not yet published as a full report, 207 patients among
whom 41% were without varices and 59% with small varices were
randomized to propranolol (n = 103) or placebo (n = 104). After
a mean follow-up of 18 months, the actuarial proportion of
patients with large varices was 41% in the propranolol group and
17% in the placebo group (p < 0.01). The authors did not provide
any plausible explanation for this surprising result, and any
conclusion must be delayed until the study is published as a
full report. For the moment, in the absence of reliable
information, there is no reason to prescribe propranolol for the
prevention of varices.
Contraindications and Side Effects
Several contraindications remarkably reduce the number of
patients suitable for beta-blocker treatment. The most frequent
are chronic pulmonary obstructive disease, heart failure or
heart diseases in which heart function may be worsened by
beta-blockers, atrioventricular heart block, and peripheral
arterial insufficiency. Sinusal bradycardia and
insulin-dependent diabetes are relative contraindications.
Because of these contraindications, 15-20% of patients are
excluded from beta-blockers. If these contraindications are
carefully checked, the incidence of side effects among treated
patients is in the order of 15%. The most frequent are dyspnea,
bronchospasm, asthenia, and reduction of sexual activity. No
more than 5% of side effects require treatment discontinuation.
Organic Nitrates
Short-acting (nitroglycerin [NTG]) or long-acting (isosorbide
dinitrate, or isosorbide-5-mononitrate [IMN]) organic nitrates
produce vasodilatation, mainly venous. They act by forming
nitric oxide that induces a reduction of intracellular calcium
concentration.[30] Several studies have shown that organic
nitrates reduce portal pressure in cirrhotic patients. Moreover,
the association of NTG with vasopressin enhances the reduction
in portal pressure while counteracting the adverse systemic
hemodynamic effects.[31] Similarly, adding IMN to propranolol
enhances the reduction of portal pressure and allows a
significant portal pressure reduction also in nonresponders to
propranolol.[32]This is probably because IMN decreases hepatic
resistance and prevents the increase in portal collateral
resistance caused by propranolol.
Among long-acting nitrates, only IMN has been evaluated in
clinical trials for the prevention of variceal bleeding.
IMN Compared with Beta-Blockers
In one RCT, 118 patients with varices of any size were
randomized to IMN (20 mg three times a day; n = 57) or
propranolol (maximum tolerated dose or to reduce heart rate to
55 beats/min; n = 61).[33] The bleeding rate was low, and no
difference was found between the two groups after a median
follow-up of 29 months. It was concluded that IMN is equivalent
to propranolol, although the study appeared to be underpowered.
After 7 years the bleeding incidence was still low (30/118
patients), and there were no differences between the two
treatment groups. However, there was an increased mortality in
the IMN group that was significant among the patients over 50;
these patients were 90 of the 118 randomized (76%). In the IMN
group there was also a significantly higher mortality without
bleeding. If this increase in mortality reflected a harmful
long-term effect of IMN or a beneficial effect of propranolol
remains to be clarified.
In another small RCT34 including 30 patients with ascites, 6 of
15 patients given IMN (40 mg twice a day) and none of 15 given
nadolol (to reduce heart rate by 25%) bled (p = 0.02). No
differences in mortality were found. Sodium excretion was almost
halved by IMN (201 6 28 vs. 114 6 15 mEq/min, p < 0.05).
Pooling the results of these two RCTs, a nonsignificant increase
in bleeding rate and mortality with IMN was found (Table 1, Fig.
3). However, the number of patients included in the two RCTs is
insufficient to draw conclusions even by meta-analysis.
Beta-Blockers and IMN Compared with Beta-Blockers
Three RCTs have been reported,35-37 only one as full report[35]
(Table 2). Two studies used nadolol[35,36] and one propranolol.[37]
The first open study included 146 patients, all but 6 with
medium or large-sized varices. Overall, 13 of 74 (18%) patients
bled in the nadolol group and 7 of 72 (10%) bled in the
combination therapy group (p = 0.17). If only variceal bleeding
is considered, corresponding figures were 11 of 74 (15%) and 4
of 72 (6%; p = 0.064). Both differences were statistically
significant when assessed by the log-rank test on Kaplan-Meyer
cumulative rates. Deaths were 14 (19%) and 8 (11%), respectively
(p = 0.19). The incidence of side effects was 8 of 74 (11%) in
the nadolol group and 37 of 72 (51%) in the combination therapy
group (p < 0.00001). Side effects were severe enough to require
treatment withdrawal in three (4%) and eight (11%) patients,
respectively (p = 0.10). The incidence of ascites was similar in
the treatment groups after 6 months of follow-up,[38] but no
information is reported in the long term.[35]
The other two studies[36,37] were double blind and placebo
controlled. Of these, one[36] was a small single-center study
including [57] patients with large varices and red color sign.
Ten of 27 patients (37%) in the placebo-nadolol group and 5 of
30 (17%) in the IMN (40 mg twice a day) nadolol group bled
within the 2-year study period (p = 0.08) (Table 2). Variceal
bleeding was, respectively, 8 of 27 (30%) and 4 of 30 (13%; p =
0.13). Deaths were three and five, respectively. Overall, 7 of
27 patients (26%) in the placebo and nadolol group and 16 of 30
(53%) in the IMN and nadolol group developed side effects (p =
0.03). Among them, 4 (15%) and 12 (40%) patients, respectively,
had to be withdrawn from the treatment (p = 0.034). A
statistically insignificant higher incidence of ascites was
found with the combination therapy (4/27 vs. 8/30), and the
study was prematurely interrupted, because of a statistically
significant excess mortality in patients treated with IMN and
nadolol in a parallel identical trial for the prevention of
rebleeding.[39]
In the third multicenter study,[37] 349 patients were included
(199 with large varices): 174 were given propranolol with
placebo and 175 propranolol with IMN. The 2-year bleeding rate
was low and similar in the treatment groups (10% and 13%,
respectively), even when only the 199 patients with large
varices were considered (13% in both groups). Side effects were
significantly more frequent in the propranolol and IMN group
(37% vs. 25%, p < 0.05). There were no differences in the number
of patients developing or worsening ascites (16% vs. 17%).
Neither treatment caused deleterious effects on renal function.
Overall, 552 patients were included in the three studies.
Bleeding rate was 15% in the beta-blockers-treated patients and
10% in the combination therapy patients (ARD = 25%, CI, 216% to
6%). Mortality rate was 10% in both groups (Table 2, Fig. 3).
Side effects were remarkably more frequent with the combination
therapy.
Spironolactone and Beta-Blockers Compared with Beta-Blockers
Alone
Spironolactone, by preventing or reducing the increase in plasma
volume that sustains the hyperdynamic circulation in portal
hypertension, lowers HVPG in patients with cirrhosis.40 Its
association with nadolol has been compared with nadolol alone
for the prevention of first variceal bleeding and of ascites in
compensated cirrhotic patients in an RCT still in progress.[41]
Preliminary results on the first 67 patients show no differences
in the bleeding or in the ascites incidences (Table 1).
Beta-Blockers Compared with Endoscopic Therapy
Two studies[20,22] with a three-arm design (placebo or no
treatment vs. propranolol vs. sclerotherapy) included 226
patients in the arms comparing propranolol with sclerotherapy.
Patients were not selected for variceal size. No differences
were found in the risk of variceal bleeding (ARD = 25%; CI, 214%
to 4%). However, in one study20 a significant favorable effect
with propranolol was found when bleeding from any portal
hypertensive source were considered (2/43 vs. 9/42; p < 0.03).
In one recent study,42 propranolol was compared with variceal
band ligation in 89 patients with medium or large varices with
red signs and showed a significant benefit with ligation.
However, this study has been criticized[42]b because of the
unusually high rate of bleeding with propranolol; lower mean
dose of propranolol with respect to previous RCTs (70 vs. 123
mg/day[28]), and lack of information on bleeding from
nonvariceal sources. Furthermore, no differences in mortality
were observed. Therefore, these results should be validated in
further studies before they are accepted for clinical practice.
Monitoring of Treatment Response
In prospective cohort studies and clinical trials[2,3,42-45] it
has been shown that the risk of bleeding is virtually abolished
when HVPG or the variceal pressure is reduced by $20% of
baseline values or HVPG is reduced below 12 mm Hg. Failure to
reach a treatment response according to one of these parameters
indicates the need for a different therapy. Although the
efficacy of such hemodynamic monitoring has not been directly
proven in an RCT, it may be predicted in patients treated for
the prevention of rebleeding, whose baseline risk is 60% in 1
year. However, it is largely uncertain if it may be effective in
patients treated for prevention of first bleeding, whose
baseline risk is much lower, being in the order of 15% per year
in the presence of large varices.
Conclusions and Recommendations for Clinical Practice
Patients without varices should be screened for appearance of
varices every 2 years (grade C1). They should not receive
treatment for prevention of development of varices (grade C2).
Patients with small varices should be screened for enlargement
of varices every year (grade C2). At present, there is not
evidence to recommend treatment for prevention of variceal
bleeding in these patients, who have a low risk of bleeding
(grade A2).
Patients with medium or large varices should be treated with a
nonselective beta-blocker (grade A1) either in the absence of
ascites (grade A1) or in the presence of ascites (grade A2). The
expected benefit is higher in patients without ascites.
Available evidence does not support the use of combination
therapy with nonselective beta-blockers and IMN in the
prevention of firts bleeding, because no benefit has been shown
from this therapy in comparison with beta-blockers alone (grade
A2). This seems to depend on the low risk of bleeding, even with
large varices, in patients treated with beta-blockers. This
conclusion might change in the future if more accurate
indicators of the risk of bleeding will allow to select patients
with higher risk.
Patients with medium or large varices with contraindications or
who are intolerant to beta-blockers may be treated with IMN
(grade B2). Patients aged over 50 should be strictly monitored
for liver and renal function (grade B2).
Treatment Of Acute Bleeding
The following key points of the clinical course of variceal
bleeding[8] should be considered when defining the principal
objectives for treatment.
Hemorrhage spontaneously stops in as much as 40-50% of
patients.[13]
Immediate mortality because of uncontrolled bleeding is about 8%
and occurs within 1-2 days from admission. This is the median
rate from eight studies including 1,488 patients. Bacterial
infection is independently associated with failure to control
bleeding.[46] No other prognostic indicators of this very early
death risk have been identified.
Rebleeding occurs in 40% of patients within 6 weeks. About half
of rebleeding episodes occur within 1 week from start of
bleeding. Indicators of early rebleeding are poorly defined: Low
albuminemia, gastric varices, high blood urea nitrogen, active
bleeding at emergency endoscopy, and HVPG >16 mm Hg measured the
first or second day after hospital admission, have been
proposed.
Six-week mortality is about 30%. The most consistently reported
death risk indicators are Child-Pugh classification or its
components, age, active alcohol abuse, blood urea nitrogen or
creatininemia, active bleeding on endoscopy, early rebleeding,
and HVPG >20 mm Hg.
Based on these observations, the principal objectives of therapy
for acute variceal bleeding are control of bleeding, prevention
of early rebleeding, and reduction of mortality. Because the
bleeding is a stop and start event, there has been some
difficulty in the past to distinguish the initial bleeding from
early rebleeding either in clinical practice or in clinical
trials. Therefore, it is often difficult to extract from
clinical trials a reliable measure of the treatment effect on
each of these two end points. Moreover, in some trials, a
unified end point has been considered: failure to control
bleeding or early rebleeding within 5 days. Two recent large
consensus conferences[47,48] proposed the following definitions
for time events in variceal bleeding:
The time of admission to the first hospital the patient is taken
to is time zero.
Bleeding is the occurrence of hematemesis and/or melena.
Clinically significant bleeding is any bleeding requiring two
units of blood or more within 24 hours of time zero, together
with a systolic blood pressure <100 mm Hg or a postural change =
20 and/or pulse rate = 100/min at time zero.
The acute bleeding episode is represented by an interval of 48
hours from time zero with no evidence of clinically significant
bleeding between 24 and 48 hours. Evidence of any bleeding after
48 hours is the first rebleeding episode.
Failure to control acute variceal bleeding within 6 hours is
defined as any of the following three factors: transfusion of
four units of blood or more, inability to achieve an increase in
systolic blood pressure by 20 mm Hg or to 70 mm Hg or more,
and/or pulse reduction to less than 100/min or a reduction of
10/min from baseline pulse rate.
Failure to control acute variceal bleeding after 6 hours is
defined as any of the following four factors: the occurrence of
hematemesis; reduction of blood pressure = 20 mm Hg from the
6-hour point, increase in pulse rate = 20/min from the 6-hour
point on two consecutive readings an hour apart, and/or
transfusion of two units of blood or more (over and above the
previous transfusions) required to increase the hematocrit = 27%
or hemoglobin = 9 g/dL.
Because these definitions were not adopted in most published
trials, we consider the following measures of treatment
efficacy, to extract more homogeneous data from RCTs: failure to
control bleeding within 24-48 hours of treatment, rebleeding
after bleeding control before definitive therapy for long-term
prevention of rebleeding, and mortality within 6 weeks or during
the stay in hospital.
The drugs assessed for acute bleeding are vasopressin and its
analogue terlipressin and somatostatin and its analogue
octreotide. Vasopressin and terlipressin have been associated
with NTG in some trials.
Vasopressin
Vasopressin causes a marked splanchnic vasoconstriction and
hence reduces portal blood flow and portal pressure.[49] It also
reduces portocollateral blood flow and variceal pressure.[50]
However, it also causes systemic vasoconstriction that may
result in serious complications such as myocardial ischemia or
infarction, arrhythmias, mesenteric ischemia, limb ischemia, and
cerebrovascular accidents.
Ten RCTs assessing vasopressin for acute variceal bleeding have
been published.[13] In nine of these studies, the time of
infusion of vasopressin or control treatment ranged from 12 to
48 hours and the assessment of failure to control bleeding was
done during the drug infusion (need to change therapy) or at the
end of the infusion (continuing bleeding). Only in one RCT,[51]
vasopressin was infused over a 20-minute period and treatment
failure was assessed 60 minutes after starting therapy. Details
of these RCTs have been recently reviewed,[13] and because no
other trials have been subsequently reported, they will not be
discussed further here. Briefly, four RCTs including 157
patients compared vasopressin with nonactive treatment. Although
failure to control bleeding was reduced from 82% in control
patients to 50% in treated patients (ARD = 232%; CI, from 259%
to 26%; NNT = 3), no differences in mortality were found. Side
effects in vasopressin-treated patients ranged from 32% to 64%
and were severe enough to require withdrawal of therapy in 25%
of cases. Mortality from complications of vasopressin was 3 of
85 treated patients. Intraarterial was compared with intravenous
vasopressin in three RCTs including 73 patients and no
differences were found either in the rates of failure to control
bleeding or in mortality. Therefore, the use of intraarterial
vasopressin was abandoned.
Aiming at reducing side effects, vasopressin has been combined
with NTG and compared with vasopressin alone in three RCTs
including 176 patients. Failure rate in controlling bleeding was
57% with vasopressin alone and 35% with the combination (ARD =
222%; CI, 236% to 28%; NNT = 4.5) without differences in
mortality (34% and 33%, respectively).
Although the only double-blind study[50] did not show
significant differences in side effects between the treatments,
the pooled incidence of side effects in the three RCTs was 61%
wih vasopressin and 31% with the combination therapy (ARD =
230%; CI, 254% to 26%). The number of patients needed to be
treated with vasopressin and NTG to save a harmful event (NNH)
was 3. However, the number of patients in which treatment was
withdrawn because of side effects was not significantly
different between the treatment groups (ARD = 210%; CI, 227% to
7%).
Among the 10 RCTs, rebleeding rate was only reported in 2
studies comparing intraarterial with intravenous vasopressin and
no differences were found.
Terlipressin
This synthetic analogue of vasopressin is composed of three
glycyl residues and lysine-vasopressin. When administered
intravenously it causes immediate vasoconstriction by intrinsic
activity followed by prolonged hemodynamic effects because it is
slowly converted into vasopressin by enzymatic cleavage of the
glycyl residues. Like vasopressin, it causes portal and
portocollateral blood flow reduction and parallel decrease in
portal and variceal pressure. It was speculated that the low
blood levels consequent to the slow release of the active agent
would result in less frequent side effects. Clinical studies[13]
have consistently shown less frequent and severe side effects
with terlipressin than with vasopressin even when vasopressin is
associated with NTG. Terlipressin, unlike vasopressin, does not
activate fibrinolysis and has a longer biological activity that
allows its administration every 4-6 hours.
To assess its efficacy, terlipressin was compared with placebo
or nonactive treatment in five RCTs,[265,66] to vasopressin in
five,[57-61] to somatostatin in three,[55,62-64] to octreotide
in 265,[265,66] (see next section), to sclerotherapy in
1,[265,66] and to esophageal tamponade in 3.[265,66]
Terlipressin Compared with Placebo or Nonactive Treatment
Four[52-54,56] of the five available RCTs are published as full
reports (Table 4) and were double-blind placebo-controlled
studies. The fifth is reported as a letter[55] and was an open
three-arm trial comparing terlipressin with somatostatin and
with conventional therapy (i.e., no active drugs, no
sclerotherapy). Overall, 256 patients were included. Treatment
duration ranged from 8 to 36-48 hours. In one study, treatment
was started at the patient's home if variceal bleeding was
suspected.[56] Overall, the rate of failure to control bleeding
was 50% in control patients and 26% in treated patients (ARD =
224%; CI, 236 to 213; NNT = 4) (Table 4). Rebleeding rate was
reported only in three studies and was difficult to interpret
because no information on treatments for prevention of
rebleeding was reported and some patients, after failure of the
experimental treatment, were treated by endoscopic
sclerotherapy. With these limitations, the overall estimate did
not show a significant difference between treatment groups.
Mortality was significantly and remarkably reduced with
terlipressin (ARD = 218%; CI, 228% to 27%; NNT = 6). The
reduction in mortality was found in all the studies but was
clinically important in three and statistically significant in
one.
Some of these studies have been criticized. One[52] included
other therapies, the timing of which is unclear, and in
another[56] no difference in the transfusion requirement at the
end of the 12-hour therapy was found between the two groups,
although a significant mortality reduction was found in Child C
patients (which were 81% of the trial population).[71] However,
cumulative meta-analysis (Fig. 4) shows that since the first
RCT, the reduction in mortality with terlipressin was about 20%,
remained stable by adding the following studies, and became
statistically significant after the third, when only 141
patients had been randomized. It is extremely unlikely that this
result comes from some inapparent bias affecting mortality
always in the same direction across five RCTs, four of which
were double-blind placebo-controlled studies.
Terlipressin Compared with Vasopressin
Five unblinded trials[57-61]
compared terlipressin with vasopressin, which was associated
with transdermal or sublingual nitroglycerin in two studies.[59,61]
Relevant characteristics and results of these studies are
reported in Table 4. No statistically significant differences
were found in the rate of failure to control bleeding,
rebleeding, or mortality. There was a significant heterogeneity
for the failure to control bleeding that disappeared (x2 for
heterogeneity, 5.26; p = 0.15; ARD = 22; CI, 217% to 13%) by
removing an outlier RCT57 where an extremely high failure rate
with vasopressin (91%) was found (far out of the range of all
other RCTs of vasopressin[13]).
The larger[61]
of these RCTs reported detailed information on side effects,
which were significantly less frequent and less severe with
terlipressin compared with vasopressin plus transdermal
nitroglycerin.
Terlipressin Compared with Emergency Endoscopic Variceal
Sclerotherapy
There is only one RCT comparing terlipressin with endoscopic
variceal sclerotherapy (EVS) (Table 4). It is reported in
abstract form[67]
and included 219 patients. Failure of treatment was defined as
failure to control bleeding or early rebleeding within 5 days
after initial control of bleeding. Treatment failures were 36 of
114 patients (32%) with EVS and 39 of 105 patients (37%) with
terlipressin (p = 0.39). Failure rate in control of bleeding
was, respectively, 17% and 21% (p = 0.45), rebleeding within 5
days 13% and 15% (p = 0.67), and 6-week mortality 16% and 23% (p
= 0.19). Side effects were 29% with EVS and 30% with
terlipressin and were serious in 5% and 6%, respectively.
Although the full report of the study is needed to draw
definitive conclusions, these results indicate that terlipressin
may be considered an equivalent alternative to emergency
sclerotherapy for acute variceal bleeding.
Terlipressin Compared with Balloon Tamponade
Three RCTs[68-70]
including 141 patients compared terlipressin with esophageal
tamponade. Although a trend was found toward a higher failure
rate in the control of bleeding with terlipressin, the
difference was not significant (ARD = 8%; CI, 28% to 24%). Early
rebleeding rate (4-7 days) and hospital mortality were similar
with the two treatments (Table 4). These results may not be
considered as a conclusive indication of equivalence between the
two treatments because of the insufficient number of patients
included in the three RCTs. However, further studies would not
be justified because it is now well established that esophageal
tamponade is a temporary salvage therapy for otherwise
uncontrollable bleeding.
Somatostatin
Natural somatostatin is a tetradecapeptide that causes
splanchnic vasoconstriction within a few seconds after
intravenous administration. The hormone has no direct effect on
splanchnic vascular smooth muscle at low levels,[72]
but it increases the vascular tone by inhibiting secretion of
gut-derived vasodilatory peptides, such as glucagon, vasoactive
intestinal peptide, and substance P.[73-75]
It causes splanchnic vasoconstriction, thereby decreasing portal
and collateral blood flow and portal pressure without the
adverse effects of vasopressin on the systemic circulation.[75]
However, vasoconstriction is probably not confined to the
splanchnic circulation because bolus injections cause a
transient increase in mean arterial pressure and systemic
vascular resistance.[76]
Somatostatin consistently reduces azygous blood flow, suggesting
an effect on portocollateral circulation.[77]
Bolus injections of somatostatin cause rapid and marked
reduction of both portal pressure and azygous blood flow,
greater than those induced by continuous infusions.[76]
Somatostatin has been compared with placebo or nonactive
treatment in eight RCTs,[55,78-84]
with vasopressin in seven,[85-91]
with terlipressin in three,[55,62-64]
with endoscopic sclerotherapy in four,[92-95]
and with esophageal tamponade in two.[96,97]
Somatostatin Compared with Placebo or Nonactive Treatment
Three double-blind placebo-controlled trials of somatostatin[78-80]
including 290 patients showed contrasting results with a
favorable treatment effect in one[79]
not confirmed in the others.[78,80]
It is of interest that one[78]
of the two negative RCTs found an unusually high rate of
spontaneous bleeding cessation (83%, the highest ever reported)
and the other[80]
did not assess failure to control bleeding. None of the three
studies found any beneficial effect on survival.
Four unblinded RCTs compared somatostatin with a nonactive
treatment[55,81-83]
and showed a trend to ward benefit from somatostatin. A recent
double-blind placebo-controlled study,[84]
showed that somatostatin administered before emergency variceal
sclerotherapy makes the endoscopic procedure significantly more
easy and reduces failure to control bleeding.
Overall, failure to control bleeding was assessed in seven RCTs
(Table 5, Fig. 5) including 552 patients. The pooled estimate of
treatment effect shows a significant qualitative heterogeneity
(i.e., in the direction, not in the size of the effect), which
disappears by removing the study by Valenzuela et al.,[78]
which seems to be an outlier because of the very high rate of
bleeding control in the placebo group. The remaining six studies
(468 patients) show a significant reduction of failure to
control bleeding with somatostatin (ARD = 217%; CI, 229% to 26%;
NNT = 6) (Fig. 5). This result did not change in a sensitive
analysis by also removing the most favorable trial82 (ARD =
214%; CI, 224% to 24%; NNT = 7) (Fig. 5). It is important to
note that the two double-blind placebo-controlled RCTs[79,84]
consistently showed a marked and significant benefit from
somatostatin (Table 5). Rebleeding rate after control of
bleeding is reported in three studies (Table 5), but the time of
assessment was different, and details on other therapies after
control of bleeding were insufficient, making pooling
meaningless. Mortality is reported in eight RCTs (638 patients),
homogeneously showing no treatment effect (Table 5).
FIG. 5. Meta-analysis (random effects model) of randomized
clinical trials of somatostatin compared with placebo or
nonactive treatment for acute variceal bleeding: treatment
effect on failure to control bleeding. Top, traditional
meta-analysis including all the available RCTs; middle,
cumulative meta-analysis by excluding the Valenzuela et al.1
RCT78 and showing significant benefit from somatostatin since
the last two studies; bottom, sensitivity meta-analysis
(cumulative) by excluding the trial with the most favorable
treatment effect82 from the set of RCTs shown in the midle
panel. Solid circles represent the absolute risk difference
(treated minus controls) for each trial and the overall
estimates. The horizontal bars represent the 95% confidence
intervals of each estimate. The vertical line represents the
zero difference (the equivalence between the treatments).
Negative values (on the left of the equivalence line) favor
somatostatin; positive values favor placebo or nonactive
treatment (CT).
Somatostatin Compared with Vasopressin
There are seven RCTs including 301 patients[85-91]
(Table 5). Failure to definitively control bleeding is not
different between the two treatments (ARD = 29%; CI, 225% to
8%), whereas a recent meta-analysis98 showed that immediate
control of bleeding is significantly higher with somatostatin.
The lack of benefit from somatostain in the definitive control
of bleeding is explained by the higher rebleeding rate after
initial control (ARD = 18%; CI, 8% to 27%). Mortality is almost
equal (ARD = 21%; CI, 211% to 9%). However, side effects are
significantly reduced with somatostatin (from 57% to 10%: ARD =
247%, CI, 269% to 225%; NNH = 2) as well as major side effects
requiring drug dosage reduction or treatment withdrawal (from
20% to 2%: ARD = 218%, CI, 230% to 26 %; NNH = 6) (Fig. 6). In
conclusion, the overall balance indicates that somatostatin is
equivalent in efficacy to vasopressin, with significantly less
frequent and less severe side effects.
FIG. 6. Cumulative meta-analysis (random effects model) of
randomized clinical trials of somatostatin compared with
vasopressin for acute variceal bleeding: total (top) and major
(bottom) side effects. Solid circles represent the absolute risk
difference (treated minus controls) for each trial and the
overall estimates. The horizontal bars represent the 95%
confidence intervals of each estimate. The vertical line
represents the zero difference (the equivalence between the
treatments). Negative values (on the left of the equivalence
line) favor somatostatin; positive values favor vasopressin. A
marked and consistent reduction of total and major side effects
was found with somatostatin since the first studies.
Somatostatin Compared with Terlipressin
Somatostatin has been compared with terlipressin in three
studies,[55,62-64]
two of which were double blind placebo controlled,62-64
including 302 patients. Overall, no differences were found for
failure to control bleeding, rebleeding, and mortality (Table
5). Total side effects were not significantly reduced with
somatostatin from 29% to 21%, and major side effects (requiring
withdrawal of treatment or specific therapy) were 4% in both
treatment groups. Most frequently reported side effects with
somatostatin were bradycardia, hypertension, hyperglicemia, and
diarrhea and with terlipressin abdominal cramps, diarrhea,
bradycardia, hypertension, arrythmias, and angina.
Somatostatin Compared with Emergency Sclerotherapy
Four RCTs,[92-95]
including 367 patients compared somatostatin with sclerotherapy.
Two[92,95]
are still in abstract form (Table 5); one[95]
included patients after control of bleeding and compared the two
treatments for 6-week rebleeding and mortality. No significant
differences were found in failure to control bleeding (ARD = 7%;
CI, 24% to 18%) (Fig. 7), rebleeding (ARD = 4%; CI, 25% to 13%),
and mortality (ARD = 4%; CI, 23% to 12%) (Table 5), although
complications were significantly less frequent and less severe
with somatostatin. It should be remarked that early rebleeding
was assessed before starting definitive treatment for long-term
prevention of rebleeding in the three RCTs available for this
outcome (including 305 patients), and although assessment times
varied widely across studies (6 weeks,[95]
7 days,[94]
5 days[93]),
no heterogeneity in the treatment effect was found (Table 5).
FIG. 7. Cumulative meta-analysis (random effects model) of
randomized clinical trials of somatostatin (top) or octreotide
(bottom) compared with emergency sclerotherapy for acute
variceal bleeding: treatment effects on failure to control
bleeding. Solid circles represent the absolute risk difference
(treated minus controls) for each trial and the overall
estimates. The horizontal bars represent the 95% confidence
intervals of each estimate. The vertical line represents the
zero difference (the equivalence between the treatments).
Negative values (on the left of the equivalence line) favor
somatostatin or octreotide; positive values favor sclerotherapy.
A not significant trend favoring sclerotherapy compared with
somatostatin was found (top). The equivalence of sclerotherapy
and octreotide shown by the overall estimate in the bottom panel
is not acceptable for clinical practice because of a
statistically significant heterogeneity both in the direction
and in the size of the effect across trials.
Somatostatin Compared with Balloon Tamponade
Two RCTs[96,97]
failed to show differences between somatostatin and balloon
tamponade for variceal bleeding (Table 5), although they were
largely underpowered. However, side effects were significantly
less frequent and less severe with somatostatin.
Octreotide
Octreotide is a synthetic octapeptide which by a shared
four-amino acid segment retains most of the effects of
somatostatin with a longer half-life. The affinity of octreotide
for the somatostatin S2 receptors is 3- to 10-fold higher than
the natural hormone.[99]
Thus, octreotide mimicks part of the effects of somatostatin.
Octreotide infusions, at the doses used empirically to treat
variceal hemorrhage or higher, have not been found to decrease
portal pressure.[100,101]
Bolus injections of octreotide produce marked but transient
decreases in HVPG and variceal pressure, and a rapid
desensitization to the drug has been hypothesized.[100]
Like somatostatin, bolus injections of octreotide cause a
transient increase in mean arterial pressure and systemic
vascular resistance, suggesting a systemic effect.[100,102]
The more consistently reported hemodynamic effect of octreotide
is to prevent the increase in portal pressure associated with
volume replacement after hemorrhagic shock[103]
or induced by a meal.[104]
Therefore, the net effect of octreotide on portal and systemic
hemodynamics remains unclear.
Octreotide has been compared with placebo in four RCTs,[105-108]
two still in abstract form[107,108]
(Table 6). In the larger study,[107]
sclerotherapy was used in octreotide or placebo failures. In the
others, octreotide or placebo was given immediately after
performing sclerotherapy[105,108]
or band ligation.[106]
Two RCTs were double blind.[105,107]
A total of 639 patients were included in the four RCTs. Overall,
failure to control bleeding was significantly reduced from 44%
with placebo to 29% with octreotide (ARD = 215%; CI, 226% to
23%; NNT = 7). It may be noted that in the only study using
octreotide or placebo as initial treatment,107 no benefit from
octreotide was found, whereas when sclerotherapy or ligation
were performed before or at the same time of octreotide
administration, a significant benefit was found in two studies[105,106]
and near significant in the third[108]
(Table 6). These results suggest that octreotide may improve the
outcome of endoscopic therapy but has no or little effect if
used alone. However, further studies are needed to draw
conclusions for clinical practice. No effects of octreotide on
rebleeding or death were found (Table 6).
Octreotide was better than vasopressin for control of bleeding
in two RCTs,[109,110]
and equivalent to terlipressin in other two[65,66]
(Table 6). Side effects were less frequent and severe with
octreotide than with either vasopressin or terlipressin, but the
difference was significant only for vasopressin.
Octreotide has also been compared with emergency sclerotherapy
in four RCTs,[111-114]
including 341 patients (Table 6, Fig. 7). Two are only available
in abstract and were still ongoing at the time of the report.[112,113]
No differences were found for failure to control bleeding,
rebleeding, or mortality in the four studies; however, they were
underpowered and also their meta-analysis was difficult to
interpret because of a significant heterogeneity for the effect
on failure to control bleeding and on rebleeding (Fig. 7).
Therefore, no conclusions may be drawn from these studies at
present. There is another RCT recently reported comparing
octreotide with emergency sclerotherapy in 100 patients with
schistosomal portal hypertension.[115]
In Child classes A and B failure to control bleeding was 10%
with octreotide and 0% with sclerotherapy (p > 0.05), whereas in
class C was 42% and 6%, respectively (p < 0.05).
Only one RCT compared octreotide with balloon tamponade,116
showing a nonsignificant trend in favor of tamponade for control
of bleeding (Table 6).
Finally, in a double-blind placebo-controlled pragmatic RCT,[117]
including 262 patients, octreotide was administred
subcutaneously (100 mg three times a day) over a period of 15
days, aiming at reducing early rebleeding in patients treated
with beta-blockers or sclerotherapy for long-term prevention of
rebleeding. Sixty-four patients did not receive long-term
prophylaxis because of intolerance to beta-blockers, low
compliance, or too advanced liver disease, and no effect of
octreotide was found in this group. Among 198 patients treated
for long-term prophylaxis, 15-day rebleeding rate was reduced
from 26% to 16% (p = 0.05) with octreotide.
Conclusions and Recommendations for Clinical Practice
There is now a large number of RCTs, several of fair to good
quality, showing that pharmacological treatment is effective in
controlling variceal bleeding. Most importantly, for
terlipressin and somatostatin there is growing evidence that
they are equivalent to emergency sclerotherapy with a remarkably
lower incidence and severity of side effects. Moreover,
terlipressin significantly reduces mortality when compared with
placebo. Although there is uncertainty on the mechanisms of
action of somatostatin and octreotide and on their hemodynamic
effects, RCTs have shown that somatostatin is more effective
than placebo or nonactive treatment and equivalent to
vasopressin, terlipressin, sclerotherapy, and balloon tamponade,
whereas data are still insufficient for octreotide. Compared
with placebo, octreotide has been consistently effective when
associated with endoscopic therapy but not if used as a single
therapy. It seems to be equivalent to vasopressin and
terlipressin, but the available RCTs are few and underpowered,
and the encouraging results of meta-analysis should be confirmed
before this conclusion is assumed for clinical practice. The
results of RCTs comparing octreotide with sclerotherapy are not
convincing because of significant heterogeneity that is
difficult to explain with two of four studies still available
only in abstract. Serious side effects are acceptably few with
terlipressin or vasopressin plus nitroglycerin and not
significantly different from those of somatostatin or
octreotide. This evidence suggests the following recommendations
for clinical practice.
-
Pharmacological treatment should be started immediately when
variceal bleeding is suspected, even before endoscopic
confirmation of diagnosis (grade A1).
-
Terlipressin should be the first choice treatment because it
may reduce mortality (grade A2).
-
Somatostatin may be considered alternative to terlipressin
(grade A2). Additive bolus injections are advisable for
patients actively bleeding on endoscopy.
-
Octreotide improves the efficacy of emergency sclerotherapy
in controlling bleeding. Therefore, it should be given to
patients initially treated with sclerotherapy or band
ligation (grade A1), whereas, at present, there is no
evidence supporting its use as a single therapy.
-
If
vasopressin is used, it should be associated with
nitroglycerin (grade A1).
-
Endoscopic sclerotherapy may be considered a further step
after terlipressin or somatostatin failure (grade A1).
Prevention Of Rebleeding
Patients surviving a first episode of variceal bleeding have a
very high risk of rebleeding and death. Median rebleeding
incidence in untreated controls of 20 RCTs reported since 1981
of nonsurgical treatment for prevention of recurrent bleeding
was 63% within 1 to 2 years.[13]
The corresponding mortality figure is 33%. For this reason, in
two recent consensus conferences[47,48]
it has been agreed that all patients surviving a variceal
bleeding need active treatments for prevention of rebleeding.
Pharmacological treatment for prevention of rebleeding has been
based on nonselective beta-blockers since the late 1980s.
Recently, their association with IMN has been proposed.
Beta-Blockers
The hemodynamic basis for the use of nonselective beta-blockers
in the prevention of rebleeding and its contraindications and
side effects are the same as for the prevention of first
bleeding.
Beta-Blockers Compared with Placebo or Nonactive Treatment
Twelve RCTs including 809 patients have been reported[118-129]
(Table 7, Fig. 8). One trial compared propranolol with atenolol
and with placebo[125]
and is analyzed as two separate RCTs. One trial assessed nadolol[123]
and the remaining propranolol. Overall, beta-blockers reduced
the rebleeding rate from 63% in controls to 42% (ARD = 221%; CI,
230% to 213%; NNT = 5) and mortality from 27% to 20% (ARD = 27%;
CI, 212% to 22%; NNT = 14). A recent meta-analysis of these RCTs
showed also that beta-blockers significantly reduce mortality
fom bleeding.[130]
FIG. 8. Cumulative meta-analysis (random effects model) of
randomized clinical trials of beta-blockers compared with
placebo or nonactive treatment for prevention of rebleeding in
cirrhosis. The treatment effects on rebleeding (top) and
mortality (bottom) are reported. Solid circles represent the
absolute risk difference (treated minus controls) obtained by
pooling the results of every new trial with all those reported
previously. The horizontal bars represent the 95% confidence
intervals of each estimate. The vertical line represents the
zero difference (the equivalence between the treatments).
Negative values (on the left of the equivalence line) favor
beta-blockers; positive values favor control treatment (CT). A
significant rebleeding risk reduction with beta-blockers was
evident since the fourth study in 1986 and was consistently
confirmed thereafter (top). The smaller beneficial effect on
mortality became statistically significant after seven studies
and was consistently confirmed thereafter.
Beta-Blockers Compared with EVS
There are 10 RCTs including 862 patients[129,131-140]
(Table 7, Fig. 9). Although the treatment for rebleeding was not
the same in the two groups in several RCTs, in which emergency
sclerotherapy was allowed in patients randomized to EVS and not
in those randomized to beta-blockers,[13]
no significant differences were found between the two treatments
either for rebleeding rate (ARD = 7%; CI, 22% to 17%) or for
mortality (ARD = 2%; CI, 25% to 8%) (Table 7). There is a nearly
significant heterogeneity for the effect on rebleeding, which
disappears by separately analyzing the treatment effect for
rebleeding from varices or from any portal hypertensive source.[141]
Because beta-blockers may reduce the risk of rebleeding from
either varices or other sources, the effect on overall
rebleeding is reported here (Table 7, Fig. 9). Side effects were
significantly less frequent and severe with beta-blockers (ARD =
222%; CI, 238% to 26; NNH = 4).[141]
FIG. 9. Cumulative meta-analysis (random effects model) of
randomized clinical trials of beta-blockers compared with
sclerotherapy for the prevention of rebleeding in cirrhosis. The
treatment effects on rebleeding (top) and mortality (bottom) are
reported. Solid circles represent the absolute risk difference
(treated minus controls) obtained by pooling the results of
every new trial with all those reported previously. The
horizontal bars represent the 95% confidence intervals of each
estimate. The vertical line represents the zero difference (the
equivalence between the treatments). Negative values (on the
left of the equivalence line) favor beta-blockers; positive
values favor sclerotherapy. A nonsignificant trend toward a
beneficial effect of sclerotherapy over beta-blockers was found,
but mortality was consistently equal with the two treatments
since the fourth study.
Beta-Blockers Compared with EVS Plus Beta-Blockers
Beta-blockers have been compared with the combination of EVS and
beta-blockers in three studies,[142-144]
one still in abstract,[144]
including 277 patients (Table 7). They consistently show a
significantly higher rebleeding risk with beta-blockers (ARD =
19%; CI, 8% to 30%) without significant differences in mortality
(ARD = 15%; CI 21% to 32%).
It is of interest that 10 RCTs comparing EVS with the
combination of EVS and beta-blockers showed that the rebleeding
rate was significantly lower with the combination therapy
without differences in mortality,[13]
suggesting that patients rebleeding under beta-blockers or EVS
might benefit from associating the two treatments.
Combined Pharmacological Therapy
At present, the only combination therapy assessed in clinical
trials is the association of nonselective beta-blockers with IMN.
This drug combination has been compared with beta-blockers
alone, with endoscopic therapy, and with transjugular
intrahepatic partasystemic shunt (TIPS). Overall, seven RCTs
have been reported,[43,145-150]
but only one as a full report.[43]
Beta-Blockers Plus IMN Compared with Beta-Blockers Alone
There are two RCTs reported in abstract[145,146]
(Table 7) including 199 patients. One is double blind was
placebo controlled.[146]
Neither shows benefit from the combination therapy. However, the
full report of the two studies is needed to draw conclusions. In
fact, although no significant heterogeneity was found, the
results are largely inconsistent, particularly for survival,
with one study[146]
showing a statistically significant excess mortality with the
combination therapy. In both studies, side effects were
significantly more frequent with the combination therapy (mostly
headache and weakness).
Beta-Blockers Plus IMN Compared with Endoscopic Therapy
The combination therapy has been compared with sclerotherapy in
one RCT[43]
and with band ligation in 2.[147,148]
Only one of these studies is published as a full report.[43]
Propranolol was used in one RCT[148]
and nadolol in two.[43,147]
A marked reduction of rebleeding with the combination therapy
was found, which was statistically significant compared with
sclerotherapy and nearly significant with band ligation (Table
7). The risk reduction was similar in the three studies (range,
221% to 228%), suggesting that they may be combined to assess
the overall combination therapy effect size compared with
endoscopic treatment. A total of 199 patients were included in
the three studies. Rebleeding rate was reduced from 50% with
endoscopic treatment to 25% with the combination therapy (ARD =
225%; CI, 23 to 212; NNT = 4), whereas mortality was similar
with the two treatments (20% vs. 18%) (Table 7).
In one study comparing the combination therapy with
sclerotherapy,[43]
HVPG was measured before and 3 to 4 months after starting
therapy in 31 of 43 patients in each treatment group. None of 14
patients (12 drug therapy, 2 sclerotherapy) with HVPG reduction
to below 12 mm Hg and only 1 of 17 with HVPG reduction of >20%
of baseline (13 drug therapy, 4 sclerotherapy) rebled. By
contrast, 23 patients (8 drug therapy, 15 sclerotherapy) of 45
(18 and 27, respectively) with HVPG reduction of 20% or less
rebled. This result confirms previous observations from a
prophylactic trial of propranolol and from prospective cohort
studies[39-40,42]
suggesting that pharmacological treatment should be targeted to
achieve an HVPG reduction >20% or to below 12 mm Hg.
Propranolol plus IMN has been compared with sclerotherapy (in
Child-Pugh A and B patients) and with shunt surgery (Child-Pugh
C patients) in one study still reported in abstract[149]
and showing no significant differences either for rebleeding or
for mortality between treatment groups (Table 7). However, a
full understanding of the results must wait the full report of
the study.
Beta-Blockers Plus IMN Compared with TIPS
There is only one study including 91 patients with Child-Pugh
score >7 reported in abstract form.[150]
The rebleeding rate was higher with the pharmacological therapy
(Table 7), but there was no difference in mortality, and
encephalopathy was significantly more frequent and severe with
TIPS (requiring hospital admissions in 23 vs. 9 occasions). The
estimated cost of TIPS therapy (16,903 U.S.$) was more than
twice that of drug therapy.
Conclusions and Recommendations for Clinical Practice
Available evidence from RCTs shows that nonselective
beta-blockers significantly reduce the risk of rebleeding and
mortality. They are equivalent to endoscopic therapy with
significantly less frequent and severe side effects. The
association of beta-blockers and sclerotherapy is superior to
either therapy alone. Although it is unclear whether the
combination of beta-blockers with IMN is superior to
beta-blockers alone, it is superior to endoscopic sclerotherapy
and probably to endoscopic band ligation. The combination of
beta-blockers with IMN has not been compared with the
association of beta-blockers and endoscopic therapy. Benefit
from drug therapy is expected to improve by tailoring the
individual patient treatment to achieve HVPG reduction >20% of
baseline or below 12 mm Hg. This evidence suggests the following
recommendations for clinical practice.
-
Patients
surviving a bleeding episode should be treated with
nonselective beta-blockers if they were previously untreated
(grade A1). Those intolerant or who have contraindications
to beta-blockers should be treated with endoscopic therapy
(band ligation is superior to sclerotherapy; see above)
(grade A1).
-
Whenever
possible, the hemodynamic effect of beta-blockers should be
monitored. (This recommendation is indirectly drawn from
RCTs that were not designed to prove the clinical efficacy
of this intervention; therefore, it may not be graded
according to the grading system for recommendations used in
this review.) If a reduction of HVPG >20% or below 12 mm Hg
is not achieved, IMN may be added.
-
The
association of a beta-blocker with endoscopic sclerotherapy
or band ligation may be used after beta-blocker failure
(grade A1).
-
Patients
with severe or repeat rebleeding while treated with
beta-blockers associated with endoscopic therapy or IMN
should be considered for "rescue" therapy with TIPS
(preferably in candidates to liver transplantation) or shunt
surgery.
Abbreviations Used
ARD: absolute rate difference
CI: confidence interval
EVS: endoscopic variceal scleratherapy
HVPG: hepatic vein pressure gradient
IMN: isosorbide-5-mononitrate
NTG: nitroglycerin
NNH: number of patients needed to save a harmful treatment
effect
NNT: number of patients needed to treat
RCT: randomized clinical trials
TIPS: transjugular intrahepatic partasystemic shunt
Tables
Table 1.
Table 2.
Table 3.
Table 4.
Table 5.
Table 6.
 < |
