original article
The
new england journal
of
medicine
n engl j med
353;21
www.nejm.org november
24, 2005
2254
Beta-Blockers to Prevent Gastroesophageal
Varices in Patients with Cirrhosis
Roberto J. Groszmann, M.D., Guadalupe Garcia-Tsao, M.D., Jaime Bosch, M.D.,
Norman D. Grace, M.D., Andrew K. Burroughs, M.B., Ch.B., Ramon Planas, M.D.,
Angels Escorsell, M.D., Juan Carlos Garcia-Pagan, M.D., David Patch, M.B., B.S.,
Daniel S. Matloff, M.D., Hong Gao, M.D., Ph.D., and Robert Makuch, Ph.D.,
for the Portal Hypertension Collaborative Group
From the Veterans Affairs Connecticut
Healthcare System, West Haven (R.J.G.,
G.G.-T.); Yale University School of Medi-
cine, New Haven, Conn. (R.J.G., G.G.-T.,
H.G., R.M.); Hospital Clinic I Provincial de
Barcelona, Barcelona (J.B., A.E., J.C.G.-P.);
Brigham and Women’s Hospital, Boston
(N.D.G.); Faulkner Hospital, Jamaica Plain,
Mass. (N.D.G., D.S.M.); the Royal Free
Hospital and School of Medicine, London
(A.K.B., D.P.); and Hospital Germans Trias
I Pujol, Badalona, Spain (R.P.). Address
reprint requests to Dr. Groszmann at the
Yale University School of Medicine, VA CT
Healthcare System, Digestive Disease Sec-
tion/111H, 950 Campbell Ave., West Ha-
ven, CT 06516.
Drs. Groszmann and Garcia-Tsao contrib-
uted equally to the article.
N Engl J Med 2005;353:2254-61.
Copyright © 2005 Massachusetts Medical Society.
background
Nonselective beta-adrenergic blockers decrease portal pressure and prevent variceal
hemorrhage. Their effectiveness in preventing varices is unknown.
methods
We randomly assigned 213 patients with cirrhosis and portal hypertension (minimal
hepatic venous pressure gradient [HVPG] of 6 mm Hg) to receive timolol, a nonselective
beta-blocker (108 patients), or placebo (105 patients). The primary end point was the
development of gastroesophageal varices or variceal hemorrhage. Endoscopy and
HVPG measurements were repeated yearly.
results
During a median follow-up of 54.9 months, the rate of the primary end point did not
differ significantly between the timolol group and the placebo group (39 percent and
40 percent, respectively; P=0.89), nor were there significant differences in the rates of
ascites, encephalopathy, liver transplantation, or death. Serious adverse events were
more common among patients in the timolol group than among those in the placebo
group (18 percent vs. 6 percent, P=0.006). Varices developed less frequently among
patients with a baseline HVPG of less than 10 mm Hg and among those in whom the
HVPG decreased by more than 10 percent at one year and more frequently among those
in whom the HVPG increased by more than 10 percent at one year.
conclusions
Nonselective beta-blockers are ineffective in preventing varices in unselected patients
with cirrhosis and portal hypertension and are associated with an increased number of
adverse events. (ClinicalTrials.gov number, NCT00006398.)
abstract
Copyright © 2005 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at UNIVERSITY COLLEGE LONDON on December 10, 2008 .
n engl j med
353;21
www.nejm.org november
24, 2005
nonselective beta-blockers to prevent gastroesophageal varices
2255
onselective beta - adrenergic
blockers
reduce portal pressure through
a reduction in portal venous inflow
1,2
as
a result of a decrease in cardiac output (
b
1
-adrener-
gic blockade) and splanchnic blood flow (
b
2
-adren-
ergic blockade). Randomized, controlled trials have
demonstrated that nonselective beta-blockers pre-
vent variceal hemorrhage in patients with varices.
3
Decreasing portal pressure at earlier stages may
prevent gastroesophageal varices. In fact, an exper-
imental study demonstrated that beta-blockers pre-
vent the development of portosystemic collateral
vessels.
4
Therefore, we conducted a study to evalu-
ate the efficacy of nonselective beta-blockers in pre-
venting gastroesophageal varices and to assess
whether baseline and sequential measurements of
the hepatic venous pressure gradient (HVPG) are
useful in predicting the development of varices.
The study was an investigator-initiated, random-
ized, double-blind, placebo-controlled, clinical trial
conducted at four sites. The protocol was approved
by the institutional review board at each site, and all
patients gave written informed consent. Timolol
maleate (Blocadren) and placebo were provided by
Merck; Merck did not participate in any other aspect
of the study, including study design, data analysis,
and manuscript preparation.
patients
Patients were enrolled between August 1993 and
March 1999 and followed until September 2002.
Eligible patients had cirrhosis and portal hyperten-
sion, as defined by an HVPG of at least 6 mm Hg;
did not have gastroesophageal varices; and were
older than 18 years and younger than 75 years of
age. The diagnosis of cirrhosis was either biopsy-
proven or clinically suspected and confirmed by the
finding of an HVPG of 10 mm Hg or greater. The ab-
sence of gastroesophageal varices was determined
unanimously at endoscopy by two staff endosco-
pists who were present during the entire procedure
and who evaluated the procedure independently.
Exclusion criteria included ascites requiring diuret-
ics, hepatocellular carcinoma, splenic- or portal-
vein thrombosis, concurrent illnesses expected to
decrease life expectancy to less than one year, the
use of any drug or procedure affecting splanchnic
hemodynamics or portal pressure, primary biliary
cirrhosis or primary sclerosing cholangitis, con-
traindications to beta-blocker therapy, pregnancy,
or alcohol intake during the dose-titration phase.
Of 780 patients screened for varices, 490 (63 per-
cent) had none. Of these 490 patients, 213 (43 per-
cent) were included in the study. The remaining 277
were excluded for the following reasons: 92 de-
clined to participate, 79 had concomitant illnesses,
52 had a normal HVPG (less than 6 mm Hg), 21
could not tolerate the lowest dose of timolol, 15 had
an HVPG of less than 10 mm Hg and non–biopsy-
proven cirrhosis, 6 were lost to follow-up, 4 were
consuming alcohol during the titration phase,
4 were receiving treatment with interferon or phle-
botomy, 2 had primary biliary cirrhosis, and in 2,
efforts to measure the HVPG were unsuccessful.
titration of the dose
The dose of timolol (or placebo) to be used during
the study was determined for each patient before
randomization during a titration period in which
open-label timolol was administered orally. The
starting dose of timolol was 5 mg per day and was
increased by 5 mg every three days until one of the
following occurred: the resting heart rate was re-
duced by 25 percent from the baseline value, the
resting heart rate fell below 55 beats per minute,
a daily dose of 80 mg of timolol was reached, or
the patient could not tolerate a further increase in
the dose.
randomization
After the titration period, patients were randomly
assigned to receive timolol or an identical-appear-
ing placebo tablet. The randomization code was
generated by computer for each participating center.
Patients were stratified according to the cause of
cirrhosis (alcoholic vs. nonalcoholic) and baseline
HVPG (less than 10 mm Hg vs. 10 mm Hg or more).
An alcoholic cause was defined as a long-standing
history of alcohol ingestion exceeding 60 g per day.
In patients with a dual alcoholic and viral cause, the
classification of cirrhosis was based on the clinical
and histologic findings.
follow-up
Patients were assessed clinically at baseline, one and
three months after randomization, and every three
months thereafter. At each visit, the heart rate, pill
count, occurrence of adverse events, and alcohol
consumption were determined and blood was ob-
tained for hematologic and biochemical measure-
ments. To maintain study blinding, the patient’s
n
methods
Copyright © 2005 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at UNIVERSITY COLLEGE LONDON on December 10, 2008 .
n engl j med
353;21
www.nejm.org november
24
,
2005
The
new england journal
of
medicine
2256
heart rate was measured by the study nurse and not
by the investigators. At baseline and every year there-
after, upper endoscopy was performed and HVPG
was measured as described elsewhere.
5
According
to standard practice at the time, no patient received
antiviral therapy during the study.
end points
The primary end points were the development of
varices or variceal hemorrhage as identified unani-
mously at endoscopy by two staff endoscopists who
were present during the entire procedure and who
evaluated the procedure independently. Varices were
defined by the presence of one of the following:
large varices (at least 5 mm); small varices (less than
5 mm), confirmed by endoscopy 6 months later;
small varices (less than 5 mm) on one endoscopy,
with the patient’s declining to undergo comfirma-
tory endoscopy or an inability to perform confir-
matory endoscopy in the subsequent 12 months;
or gastric varices confirmed by endoscopic ultraso-
nography. Variceal hemorrhage was defined as any
hematemesis or melena in a patient in whom en-
doscopy showed active bleeding from an esopha-
geal or gastric varix, an esophageal or gastric varix
with an adherent clot, or varices but no other source
of bleeding. In addition, acute, clinically significant
bleeding as a result of portal hypertensive gastrop-
athy (defined by the need for a 2-unit transfusion,
a 6-point drop in the hematocrit, or a drop of more
than 20 mm Hg in systolic blood pressure with a
change in the patient’s posture) was considered a
primary end point.
Secondary end points were the development of
ascites or encephalopathy, liver transplantation, or
death. Data collection was terminated and treat-
ment was considered to have failed when a patient
reached the primary end point, underwent liver
transplantation, or died.
adverse events
An adverse event was any event that required a di-
agnostic or therapeutic intervention. All adverse
events, regardless of their possible association with
the disease or study treatment, were recorded. An
adverse event was judged severe if it was consid-
ered to endanger the health or safety of the patient.
data and safety monitoring board
Members of a data and safety monitoring board
were appointed by the National Institute of Diabe-
tes and Digestive and Kidney Diseases and met ev-
* Plus–minus values are means ±SD. There were no significant differences be-
tween groups. Race or ethnic group was self-reported. To convert values for
bilirubin to micromoles per liter, multiply by 17.1. To convert values for blood
urea nitrogen to micromoles per liter, multiply by 0.357. To convert values for
creatinine to micromoles per liter, multiply by 88.4. HCV denotes hepatitis C
virus, HBV hepatitis B virus, and INR international normalized ratio.
† The majority (37 of 51 [73 percent]) had been abstinent from alcohol for more
than one month; 5 others had evidence of mild alcoholic hepatitis at random-
ization (2 in the timolol group and 3 in the placebo group).
‡ The Child–Pugh score can range from 5 to 15, with higher scores indicating
more severe liver disease.
Table 1. Baseline Characteristics of the Patients.*
Characteristic
Timolol Group
(N=108)
Placebo Group
(N=105)
Age — yr
46±11 44±11
Male sex — no. (%) 70 (65) 56 (53)
Race or ethnic group — no. (%)
White 100 (93) 98 (93)
Black 3 (3) 1 (1)
Latin-American 3 (3) 2 (2)
Indian 2 (2) 3 (3)
Mideastern or Arabian 0 1 (1)
Cause of cirrhosis — no. (%)
Alcoholic† 26 (24) 25 (24)
Nonalcoholic 82 (76) 80 (76)
HCV 67 (62) 67 (64)
HBV 6 (6) 2 (2)
Cryptogenic 5 (5) 5 (5)
Other 4 (4) 6 (6)
Anti-HCV positivity — no. (%) 70 (65) 74 (70)
Child–Pugh score‡ 5.4±0.7 5.4±0.8
Child–Pugh class — no. (%)
A 98 (91) 91 (87)
B 10 (9) 14 (13)
Mean blood pressure — mm Hg 94±10 93±12
Heart rate — beats/min 75±11 74±10
Hemoglobin — g/dl 13.8±1.5 13.5±1.7
White-cell count — ¬10
¡3
/mm
3
5.7±2.8 5.7±2.1
Platelet count — ¬10
¡3
/mm
3
122±72
119±46
Total bilirubin — mg/dl 1.2±0.7 1.12±0.8
Albumin — g/dl 3.9±0.5 3.9±0.5
Prothrombin time — INR 1.34±2.53 1.34±2.52
Aspartate aminotransferase — U/liter 93±74 89±59
Alanine aminotransferase — U/liter 106±101 105±96
Alkaline phosphatase — U/liter 138±70 153±97
Serum sodium — mmol/iter 140±3 140±4
Blood urea nitrogen — mg/dl 14±8 15±10
Creatinine — mg/dl 0.9±0.2 0.9±0.2
HVPG — mm Hg 11.7±4.3 11.7±4.1
HVPG ≥10 mm Hg — no. (%) 67 (62) 67 (64)
Median follow-up — mo 52.7 57.9
Copyright © 2005 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at UNIVERSITY COLLEGE LONDON on December 10, 2008 .
n engl j med
353;21
www.nejm.org november
24, 2005
nonselective beta-blockers to prevent gastroesophageal varices
2257
ery six months to review the progress of the study
and accumulated data. According to the protocol,
one interim analysis was performed on October 26,
2000, after all patients had been enrolled. At that
time, the data and safety monitoring board was em-
powered to recommend termination of the study on
the basis of concern about safety or in the presence
of sufficient evidence to indicate that timolol was
statistically superior to placebo. The board voted
unanimously to recommend continuation of the
trial.
statistical analysis
We estimated that treatment with timolol would re-
duce the four-year cumulative probability of varices
from 50 percent (the rate without treatment)
6,7
to
30 percent, given a statistical power of 80 percent
to detect an absolute difference of 20 percent be-
tween the placebo and timolol groups at a two-sided
alpha level of 0.05. We estimated that the study
would require 193 patients, and we then increased
this amount by 10 percent to account for the loss of
patients to follow-up, yielding a total of 212 pa-
tients.
All analyses were conducted according to the
intention-to-treat principle. Qualitative variables
were compared by means of Fisher’s exact test.
Wilcoxon’s rank-sum test was used to compare
continuous variables or ordinal data. Actuarial prob-
abilities were calculated according to the Kaplan–
Meier method and compared with use of the log-
rank test. Data were censored when the primary end
point was reached, at the time of transplantation or
death, or at the time of the last visit, whichever oc-
curred first. A Cox proportional-hazards model was
used to identify the variables that best explained
the variability in the rates of primary end points,
treatment failure, and survival. Calculations were
performed with the use of the SAS statistical soft-
ware package.
A total of 213 patients underwent randomization:
108 were assigned to receive timolol, and 105 to re-
ceive placebo (110 at the Barcelona center, 52 at the
Connecticut center, 26 at the London center, and 25
at the Boston center). The median time from screen-
ing endoscopy to randomization was 29 days (range,
8 to 105). As shown in Table 1, the baseline charac-
teristics were similar in the two groups. There were
no significant differences between groups in the
proportion of patients with alcohol-induced cirrho-
sis or in the HVPG. The median Child–Pugh score
was 5 (range, 5 to 9; scores can range from 5 to 15,
with higher scores indicating more severe liver dis-
ease). The median HVPG was 11 mm Hg (range,
6 to 25), with 63 percent of the patients having an
HVPG of at least 10 mm Hg. The median duration
of follow-up was 54.9 months (range, 0 to 99.4).
The median daily dose of timolol was 10.8 mg
results
* No patient had isolated gastric varices as a primary end point. Among the pa-
tients in whom esophageal varices developed, five (four in the timolol group
and one in the placebo group) had concomitant gastric varices (three had
junctional and two had fundal varices).
† The total number in each group reflects the number of patients who did not
reach a primary end point (66 in the timolol group and 63 in the placebo
group).
‡ The reasons for transplantation were hepatocellular carcinoma in four pa-
tients, decompensated cirrhosis in two patients, and acute-on-chronic liver
failure in one patient.
§ In both patients, the reason for transplantation was hepatocellular carcinoma.
¶ Four deaths were related to infection followed by renal or liver dysfunction,
one was due to hepatocellular carcinoma, and five were unrelated to liver
disease. Ascites, encephalopathy, or both developed in 8 of the 10 patients
who died.
¿ Six deaths were related to infection followed by renal dysfunction, three were
due to hepatocellular carcinoma, two were due to liver failure, and four were
unrelated to liver disease. Ascites, encephalopathy, or both developed in 8 of
the 15 patients who died.
** Treatment failure was defined by the occurrence of the primary end point
(varices and variceal hemorrhage), transplantation, or death.
Table 2. Rates of Primary and Secondary End Points and Treatment Failure.
Variable
Timolol Group
(N=108)
Placebo Group
(N=105)
P
Value
no./total no. (%)
Primary end point*
42/108 (39)
42/105 (40) 0.89
Large varices 4 4
Confirmed small varices 27 30
Unconfirmed small varices 8 5
Variceal hemorrhage 2 3
Hemorrhage from portal hyper-
tensive gastropathy
10
Secondary end point†
22/66 (33)
22/63 (35) 1.00
Ascites 4 6
Death 3 2
Hepatic encephalopathy 3 2
Transplantation 1 0
Death 0 1
Ascites and encephalopathy 6 5
Transplantation 1 0
Death 5 5
Transplantation 7‡ 2§
Death 10¶ 15¿
Treatment failure**
59/108 (55) 59/105 (56) 0.89
Copyright © 2005 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at UNIVERSITY COLLEGE LONDON on December 10, 2008 .
n engl j med
353;21
www.nejm.org november
24
,
2005
The
new england journal
of
medicine
2258
(range, 1.25 to 80.0) in the timolol group, and the
median daily dose of the timolol placebo was 12.9 mg
(range, 1.25 to 80.0) in the placebo group (accord-
ing to titration). The dose had to be reduced in 29
patients (26 in the timolol group vs. 3 in the placebo
group, P<0.001), and the study medication was
withdrawn prematurely in 46 patients (25 in the
timolol group vs. 21 in the placebo group, P=0.62).
Adherence to treatment was considered adequate
if the pill count showed more than 70 percent ad-
herence; this degree of adherence was achieved in
86 patients in the timolol group (80 percent) and
88 patients in the placebo group (84 percent).
end points
The rates of the primary and secondary end points
and treatment failure are shown in Table 2. A total
of 84 patients reached the primary end point of var-
ices or variceal bleeding: 42 of 108 patients in the
timolol group and 42 of 105 patients in the placebo
group (39 percent vs. 40 percent, P=0.89) (Table 2
and Fig. 1). The rates of the primary end point (both
overall and for the timolol group) did not differ sig-
nificantly when patients who had a reduction in the
dose or stopped treatment were compared with
those who did not have a reduction in the dose or
discontinue treatment (data not shown). Hepato-
cellular carcinoma, which was not considered an end
point of the study, occurred in eight patients in the
timolol group and six patients in the placebo group.
A comparison of the 84 patients who reached
the primary end point with the 129 patients who did
not reach the primary end point revealed that the
following baseline variables differed at a P value of
less than 0.1: the Child–Pugh score, the white-cell
count, the aspartate aminotransferase level, the ala-
nine aminotransferase level, and the HVPG. On Cox
regression analysis, a baseline HVPG of 10 mm Hg
or more (P=0.005) and an elevated aspartate ami-
notransferase level (P=0.007) were independently
predictive of reaching the primary end point.
adverse events
The incidence of moderate or severe adverse events
was higher in the timolol group than in the placebo
group (48 percent [52 patients] vs. 32 percent [34
patients], P=0.02). Serious adverse events consid-
ered probably related to study medication occurred
in 20 patients (18 percent) in the timolol group
(7 had bradycardia, as defined by a heart rate of less
than 50 beats per minute; 5 had severe fatigue; 4 had
wheezing or shortness of breath, 2 had syncope; and
1 each had intermittent claudication and impotence)
and in 6 patients (6 percent) in the placebo group
(1 each had impotence, hypotension, depression,
heart failure, nodal rhythm, and bronchospasm)
(P=0.006)
.
None of the complications were fatal.
hemodynamics
Except at baseline, the heart rate was significantly
lower in the timolol group than in the placebo group
throughout the study (Fig. 2A). The average reduc-
tion in the heart rate from baseline was 17 percent
in the timolol group. Conversely, the HVPG did not
differ significantly between groups during the study
(Fig. 2B).
At baseline, an HVPG of 10 mm Hg or more was
associated with a significantly higher incidence of
the primary end point, as shown in Figure 3A.
HVPG measurements were repeated at one year in
154 patients (72 in the timolol group and 82 in the
placebo group). As compared with baseline values,
the HVPG decreased by a median of 1.45 mm Hg
among patients in the timolol group, as compared
with a decrease of only 0.5 mm Hg among patients
in the placebo group (P=0.07); the decrease in the
latter group was due solely to a drop in wedge pres-
sure. Reductions in the HVPG of more than 10 per-
cent (Fig. 3B), more than 15 percent, and more
than 20 percent were all associated with a signifi-
cantly lower incidence of the primary end point.
More important, a decrease in the HVPG of more
Figure 1. Kaplan–Meier Estimates of the Percentages of Patients Who Did Not
Reach the Primary End Point of Varices or Variceal Bleeding.
Cumulative percentages of patients who did not reach the primary end point
at 12, 24, 36, and 60 months were 91 percent, 86 percent, 79 percent, and 60
percent, respectively, in the timolol group and 97 percent, 82 percent, 78 per-
cent, and 57 percent, respectively, in the placebo group.
100
Patients Free from Primary
End Point (%)
80
90
70
60
40
30
10
50
20
0
0 12 24 36 48 60 72
Timolol
Placebo
Months
No. at Risk
Placebo
Timolol
20
20
37
34
46
46
65
72
99
97
75
89
105
108
P=0.89
Copyright © 2005 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at UNIVERSITY COLLEGE LONDON on December 10, 2008 .
