Introduction
Bleeding from gastroesophageal varices is the most common life-threatening complication
in patients with cirrhosis, being associated with mortality rates from 10 % to 50 %
per episode [1 ]
[2 ]. More than half of patients who survive the first episode suffer from recurrent
bleeding within 1 year [3 ]
[4 ]. Management of acute variceal bleeding (AVB) remains a clinical challenge with high
mortality, in spite of standardization in supportive and new therapeutic treatments
in the last two decades [5 ].
A beneficial effect on survival has been observed in parallel with introduction of
drugs that are capable of decreasing portal pressure, optimization of endoscopic therapy,
and use of antibiotics and interventional radiologic procedures. During the same period,
the 6-week mortality rate has decreased from approximately 40 % to 15 % [5 ]
[6 ].
Although overall survival has improved in recent years, mortality is still closely
related to failure to control the initial bleeding or early rebleeding, which occurs
in up to 30 % to 40 % of patients within the first 5 days after the index bleeding
episode [5 ]
[6 ]. As a consequence, many patients with cirrhosis and AVB still suffer from failure
to control bleeding and most of them die very early [2 ]. Therapeutic options include vasoactive drugs such as somatostatin, octreotide,
and terlipressin; endoscopic treatments such as sclerotherapy and band ligation; and
most recently, radiologic interventions, such as early-TIPS (transjugular intrahepatic
portosystemic shunt) placement. Therefore, the goal of this study was to evaluate
the efficacy and safety of the most-used endoscopic treatments for controlling AVB.
Methods
A meta-analysis and systematic review of published randomized controlled trials (RCTs)
were carried out.
Search strategy
Medline (PubMed), Embase, Cochrane library and manual searches were combined and last
performed on 16 March 2018. Key search terms were “esophageal and gastric varices,”
“esophageal varices,” “esophageal varix,” “oesophageal varices,” “oesophageal varix,”
“esophagogastric varices,” “esophagogastric varix,” “gastroesophageal varices,” “gastroesophageal
varix,” “oesophagogastric varices,” “oesophagogastric varix,” “oesophago-gastric varices,”
“oesophago-gastric varix,” “esophago-gastric varices,” “esophago-gastric varix,” upper
gastrointestinal bleeding,” “bleeding, upper gastrointestinal,” “upper digestive haemorrhage,”
“upper digestive hemorrhage,” “upperdigestive tract haemorrhage,” “upper digestive
tract hemorrhage,” “uppergastrointestinal haemorrhage,” “upper gastrointestinal hemorrhage,
“upper gastrointestinal tract bleeding,” “variceal bleeding,” “esophagus varices bleeding,”
“esophagus bleeding varix,” “esophagus varices haemorrhage,” “esophagus varices hemorrhage,”
and “esophagus varix bleeding”. MeSH terms and free-text terms, as well as variation
of root words were searched. Terms were combined within each database. The study has
been registered in PROSPERO database under code CRD42017058139.
Criteria for inclusion and exclusion of studies
Only RCTs were included. To reduce the risk of bias, strict inclusion and exclusion
criteria were defined prior to literature search. To be considered, a study had to
include patients exclusively with cirrhosis, patients with acute variceal bleeding,
have more than 10 patients in each arm, include only adults, and include treatments
performed in the first 24 to 48 hours after bleeding. Studies were excluded if they
included patients with hepatocellular carcinoma or other malignancies, use of portocaval
shunts or esophageal resection, recent use of balloon tamponade as first bleeding
control measure, placebo or elective treatment in one study arm. When two publications
existed covering the same study population, only the most recent was taken into account.
Endpoints
Endpoints were defined prior to the beginning of the meta-analysis. Main endpoints
were treatment efficacy for bleeding control and in-hospital mortality. Secondary
endpoints were rate of rebleeding from active bleeders at initial endoscopy, rate
of overall rebleeding, rate of overall mortality, and rate of adverse events (AEs)
related to each treatment.
Data extraction and assessment of quality
Two reviewers independently abstracted data from included articles (F.Q.O. and F.M.V.).
Disagreements were resolved by consensus of all authors. Extracted information included
patient population characteristics, intervention characteristics, comparator characteristics,
outcomes assessed, and study quality. The latter used the framework suggested by the
Cochrane Handbook for Systematic Reviews of Interventions (version 5·1·0), with evaluation
of the following trial characteristics: random sequence generator method, concealment
of treatment allocation, blinding of participants and personnel, blinding of outcome
assessment, and for selective reporting [7 ]. Intention-to-treat analysis and the funding source of the studies were also assessed.
The GRADE methodology [7 ] was used to define risk of bias for each of the outcomes that had available data.
Sources of support
This systematic review and meta-analysis was not supported by any grant.
Statistical analysis
We performed direct random effects model meta-analyses of head-to-head comparisons
for pooling effect sizes of reported comparisons and outcomes whenever enough data
were provided in published studies. No data imputation was done, and studies not reporting
information that allowed treatment-effect calculation were not included in the meta-analyses.
Summary effect for binary outcomes was calculated from risk ratios. Heterogeneity
was evaluated with the inconsistency test proposed by Higgins (I2 ), where values below 25 % were considered as low heterogeneity, and above 75 %, high
heterogeneity [8 ]
[9 ]. Publication bias was assessed with funnel plots of comparisons with seven or more
studies. Meta-analyses were carried out in Review Manager 5·3 (The Nordic Cochrane
Centre, The Cochrane Collaboration, 2011). Prediction intervals were calculated with
the Paule-Mandel estimator for tau squared and the Hartung-Knapp adjustment for random
effects model. Prediction interval calculations were done with the software “R” (v 3.5.0)
and package “Meta” (v 4.9 – 4) [10 ].
Results
Studies selection
A total of 8382 citations were screened, of them 6691 were evaluated after duplicates
were removed ([Fig. 1 ]). Of these, 69 were selected for full-text evaluation. Among them, only 36 randomized
trials [11 ]
[12 ]
[13 ]
[14 ]
[15 ]
[16 ]
[17 ]
[18 ]
[19 ]
[20 ]
[21 ]
[22 ]
[23 ]
[24 ]
[25 ]
[26 ]
[27 ]
[28 ]
[29 ]
[30 ]
[31 ]
[32 ]
[33 ]
[34 ]
[35 ]
[36 ]
[37 ]
[38 ]
[39 ]
[40 ]
[41 ]
[42 ]
[43 ]
[44 ]
[45 ]
[46 ] were identified that fulfilled the inclusion criteria ([Table 1 ]): seven studies compared sclerotherapy with vasoactive drugs (two studies with somatostatin,
three studies with octreotide, one with terlipressin and one with vasopressin plus
nitroglycerin); two studies compared ligation with vasoactive drugs (one with octreotide
and one with somatostatin); one study compared ligation with cyanoacrylate injection;
10 studies compared sclerotherapy with ligation; seven studies compared sclerotherapy
with the combination of sclerotherapy and vasoactive drugs (six with octreotide and
one with somatostatin); five studies compared ligation with sclerotherapy and ligation;
two studies compared ligation with ligation and vasoactive drugs (one with somatostatin
and with octreotide); one study compared sclerotherapy and octreotide with octreotide
alone; and one study compared ligation and octreotide with octreotide alone.
Fig. 1 Study selection flowchart.
Table 1
Demographic data from included studies.
Author (year)
Patients, n [a/b]
Mean age, years
Men, n %
Main cause of cirrhosis, n [a/b]
Child-pugh class c % [a/b]
Active bleeding % [a/b]
Follow-up for initial control of bleeding (hours)
Intervention axb
Sclerotherapy x VP + NG
Westaby (1989)
33/31
54.2
56.3
Alcohol 13/alcohol 22
36/32
100/100
12
Sclerotherapy x somatostatin
Shields (1992)
41/39
58
67.5
Alcohol 26/alcohol 28
41/64
61/69
120
Planas (1994)
35/35
57
71.4
Alcohol 28/alcohol 22
34/34
48.5/51.4
48
Sclerotherapy x octreotide
Sung (1993)
49/49
55.7
84.7
HBV 32/HBV 36
42/43
37/51
48
Sivri (2000)
36/30
47
24.2
Viral 8/viral 14
53/55
100/100
6
Bildozola (2000)
37/39
52.6
78.9
Alcohol 27/alcohol 28
8/13
48.6/38.5
12
Sclerotherapy x terlipressin
Escorsell (2000)
114/105
55.5
72.1
Alcohol 47/alcohol 41
31/32
42.9/35.2
48
Sclerotherapy x sclerotherapy + octreotide
Besson (1995)
101/98
56
76.4
Alcohol 93/alcohol 89
46/26
46.5/428
24
Shiha (1996)
96/93
49.6
81.5
HCV 45/HCV 44
12/15
100/100
168
Faraoqi (2000)
69/72
100/100
Not clearly stated
Zuberi (2000)
35/35
38.5
80.0
HBV 28/HBV 26
0/0
100/100
24
Shah (2005)
54/51
49.8
64.8
Viral 52/viral 49
26/21
44.4/45
Not clearly stated
Morales (2007)
28/40
51.8
66.2
HCV 14/HCV + alcohol 11
36/60
46/65
Not clearly stated
Sclerotherapy x sclerotherapy + somatostain
Avgerinos (1997)
101/104
58.6
70.7
Alcohol 59/alcohol 61
28/25
40.3/26.7
Not clearly stated
Octreotide + sclerotherapy x octreotide
Patsanas (2002)
15/15
51
70.0
Alcohol 8/viral 5
60/53
33/43
120
Sclerotherapy x ligation
Stiegmann (1992)
65/64
52.0
80.6
Alcohol 52/alcohol 53
20/19
20/22
8
Laine (1993)
38/39
46.0
75.3
Alcohol 30/alcohol 31
12,8/34,2
23/24
Not clearly stated
Gimson (1993)
49/54
51.4
55.3
Alcohol 24/alcohol 25
24/28
23/39
12
Lo (1995)
59/61
55.5
80.8
Viral 43/viral 41
47/49
25/29
72
Hou (1995)
67/67
60.6
79.9
Viral 47/viral 43
34/43
23/29.8
24
Lo (1997)
34/37
54.0
86.1
HCV 11 + alcohol 11/HBV 15
59/59
100/100
72
Shafqat (1998)
30/28
52.0
63.8
HCV 21/HCV 18
13/11
93/86
12
De la Peña (1999)
46/42
59.0
72.7
Alcohol 29/alcohol 29
28/24
47.8/42.8
Not clearly stated
Luz (2011)
50/50
52.3
72.0
Alcohol 19 + virus 19/alcohol 17
40/30
10/20
120
Sahu (2014)
103/111
Not clearly stated
Ligation x octreotide
Ximing (2013)
Not clearly stated
Ligation x somatostatin
Chen (2006)
62/63
53.2
76.0
Alcohol 24/alcohol 29
29/28
27.4/20.6
48
Ligation x cyanoacrylate injection
Ljubicic (2011)
21/22
58
72.1
Alcohol/ alcohol
19/41
52.4/90.9
24
Ligation x ligation + sclerotherapy
Laine (1996)
20/21
47
73.2
Alcohol 16/alcohol 15
45/43
20/19
Not clearly stated
Saeed (1997)
25/22
53.1
91.5
Alcohol 22/alcohol 16
16/41
28/18
Not clearly stated
Al traif (1999)
31/29
48.8
61.7
HCV 10/HCV 14
32/17
22.5/31
Not clearly stated
Djurdjevic (1999)
51/52
55.6
61.2
Alcohol 25/alcohol 28
23/19
23.5/19,2
Not clearly stated
Mansour (2017)
60/60
0.0
65.0
HCV 52/HCV 52
53/40
48
Ligation + octreotide x octreotide
Liu (2009)
51/50
41
81.2
55/48
35.2/34
72
Ligation x ligation + somatostatin
Sarin (2008)
24/23
43.6
74.0
40.0
Not clearly stated
Ligation x ligation + octreotide
Sung (1995)
47/47
57.0
71.3
Hepatitis 29/hepatitis 27
40.4/42.6
44.7/34.0
24
Studies characteristics
Only 32 RCTs had been published as full papers. Four trials were published as abstracts
[32 ]
[40 ]
[44 ]
[45 ]. In six studies [11 ]
[23 ]
[24 ]
[31 ]
[32 ]
[35 ], patients were included only if they had ongoing bleeding at time of initial endoscopy.
Alcoholic cirrhosis was the predominant cause of portal hypertension in 18 studies.
In contrast, cirrhosis due to viral hepatitis infection was the leading cause in 13
studies (patients from Asia, Brazil, and the Middle East). Otherwise, baseline characteristics
of the study populations, such as gender ratio, Child-Pugh class or mean age, were
comparable ([Table 1 ]). Only 11 of the 36 trials described separately the rebleeding rate of the different
treatment modalities in active bleeders at the time of endoscopy, i. e., 25 studies
analyzed together active and non-active bleeders (Supplementary Table 1 – Supporting Information ).
Risk of bias within trials
The included trials had risk of bias evaluated according to the Cochrane recommendations
for meta-analyses and systematic reviews ([Fig. 2 ]). None of the included trials were placebo-controlled. The randomization-method
of the majority of the trials was computer-generated random sequences, with only four
trials (abstracts) having no information about randomization. Eighteen trials had
low risk for concealment of treatment allocation. Blinding of outcome assessment was
not stated in any of the peer-reviewed articles.
Fig. 2 Risk of bias assessment of included studies. Green circles: low risk of bias for
a given quality assessment domain; blue circles: unclear risk of bias for a given
quality assessment domain; red circles: high risk of bias for a given quality assessment
domain.
Prediction intervals for random effects meta-analyses are presented in [Table 2 ].
Table 2
Prediction intervals for random-effects models.
Comparison
Outcome
Prediction interval
Meta-analysis and prediction interval interpretation
Ligation x sclerotherapy
Efficacy of bleeding control
0.91 to 1.29
Random effects meta-analysis statistically signiffican but prediction interval indicating
uncertainty on true effect size and direction
Overall rebleeding
0.25 to 1.99
Random effects meta-analysis statistically signiffican but prediction interval indicating
uncertainty on true effect size and direction
In-hospital mortality
Zero to infinity
No statistically significant differences in random-effects meta-analysis estimate
and prediction interval indicating complete uncertainty on true effect size and direction
Overall mortality
0.33 to 1.57
Random effects meta-analysis statistically signiffican but prediction interval indicating
uncertainty on true effect size and direction
Complications
0.18 to 0.47
Random effects meta-analysis statistically signifficant and prediction interval indicating
low uncertainty on true effect size and no uncertainty on true effect direction
Sclerotherapy x drug
Efficacy of bleeding control
0.99 to 1.17
No statistically significant differences in random-effects meta-analysis estimate
and prediction interval indicating low uncertainty on true effect size and direction
Overall rebleeding
0.71 to 1.06
No statistically significant differences in random-effects meta-analysis estimate
and prediction interval indicating low uncertainty on true effect size and direction
Rebleeding from active bleeders
0.92 to 1.48
No statistically significant differences in random-effects meta-analysis estimate
and prediction interval indicating some uncertainty on true effect size and direction
In-hospital mortality
0.43 to 1.42
No statistically significant differences in random-effects meta-analysis estimate
and prediction interval indicating low uncertainty on true effect size and direction
Overall mortality
0.41 to 1.49
No statistically significant differences in random-effects meta-analysis estimate
and prediction interval indicating some uncertainty on true effect size and direction
Complications
1.34 to 3.29
Random effects meta-analysis statistically signifficant and prediction interval indicating
some uncertainty on true effect size and no uncertainty on true effect direction
Sclerotherapy + drug x sclerotherapy
Efficacy of bleeding Control
1.04 to 1.31
Random effects meta-analysis statistically signifficant and prediction interval indicating
low uncertainty on true effect size and no uncertainty on true effect direction
Overall rebleeding
0.08 to 1.40
Random effects meta-analysis statistically signifficant and prediction interval indicating
some uncertainty on true effect size and direction
Rebleeding from active bleeders
0.02 to 3.45
Random effects meta-analysis statistically signiffican but prediction interval indicating
uncertainty on true effect size and direction
In-hospital mortality
0.52 to 1.32
No statistically significant differences in random-effects meta-analysis estimate
and prediction interval indicating low uncertainty on true effect size and direction
Overall mortality
0.64 to 1.28
No statistically significant differences in random-effects meta-analysis estimate
and prediction interval indicating low uncertainty on true effect size and direction
Ligation x ligation + sclerotherapy
Efficacy of bleeding control
0.70 to 1.45
No statistically significant differences in random-effects meta-analysis estimate
and prediction interval indicating high uncertainty on true effect size and direction
Overall rebleeding
0.60 to 1.48
No statistically significant differences in random-effects meta-analysis estimate
and prediction interval indicating some uncertainty on true effect size and direction
In-hospital mortality
0.04 to 13.65
No statistically significant differences in random-effects meta-analysis estimate
and prediction interval indicating high uncertainty on true effect size and direction
Overall mortality
0.06 to 14.32
No statistically significant differences in random-effects meta-analysis estimate
and prediction interval indicating high uncertainty on true effect size and direction
Complications
0.30 to 0.86
Random effects meta-analysis statistically signifficant and prediction interval indicating
low uncertainty on true effect size and no uncertainty on true effect direction
Ligation x drug
Efficacy of bleeding control
0.98 to 1.88
Random effects meta-analysis statistically signifficant and prediction interval indicating
some uncertainty on true effect size and little uncertainty on true effect direction
Overall rebleeding
Zero to infinity
No statistically significant differences in random-effects meta-analysis estimate
and prediction interval indicating complete uncertainty on true effect size and direction
Overall mortality
Zero to infinity
No statistically significant differences in random-effects meta-analysis estimate
and prediction interval indicating complete uncertainty on true effect size and direction
Ligation x ligation + drug
Overall rebleeding
Zero to infinity
No statistically significant differences in random-effects meta-analysis estimate
and prediction interval indicating complete uncertainty on true effect size and direction
Overall mortality
0.15 to 26.64
No statistically significant differences in random-effects meta-analysis estimate
and prediction interval indicating low uncertainty on true effect size and direction
Risk of bias across trials
With respect to risk of publication bias, funnel plots were generally symmetrical,
which indicates a low probability of publication bias in the present systematic review.
Comparison of sclerotherapy with vasoactive medications
Sclerotherapy was compared to somatostatin, octreotide, and vasopressin plus nitroglycerin
in seven trials. The rate of complications was significantly higher with sclerotherapy
(6 trials; relative risk [RR] 2.10; 95 % confidence interval [CI] 1.52 – 2.90; P < 0.00001; I2 = 0 %) when compared to vasoactive drug alone. There was no significant difference
in the other analyzed outcomes. The studies by Planas and Escorsell had no specific
description of in-hospital mortality, so they were not included in this analysis.
In patients with active bleeding at endoscopy, sclerotherapy was needed in 17 patients
to achieve active bleeding control in one of them when compared to vasoactive drug
alone (number needed to treat [NNT] 17, I2 = 0 %, P < 0.05).
Comparison of ligation with vasoactive medications
Only two trials [38 ]
[44 ] compared ligation with vasoactive medications (somatostatin and octreotide). However,
the study by Ximing was published as an abstract and did not have enough information
to be included in the analysis.
Comparison of sclerotherapy with ligation
Ligation was associated with significant improvement in bleeding control (10 trials;
RR 1.08; CI 1.02 – 1.15; P = 0.01; I2 = 49 %) compared to sclerotherapy ([Fig. 3 ]). The heterogeneity was potentially explained by the differences in two identified
subgroups: one formed by Lo [19 ], Hou [21 ], Lo [24 ], Squafat [27 ] and de la Peña [29 ] (mostly Asian studies) in which ligation was clearly superior to sclerotherapy and
another group of five trials (mostly Western trials) in which both techniques had
similar results.
Fig. 3 Forest plot of risk ratio for efficacy of bleeding control with ligation versus sclerotherapy.
Risk of overall rebleeding was statistically significantly higher (10 trials; RR 1.41
95 % CI 1.03 – 1.94; P = 0.03; I2 = 62 %) with sclerotherapy than with ligation. The high heterogeneity was explained
by the same reasons as mentioned above.
Overall mortality was 38 % higher in patients treated with sclerotherapy compared
to ligation (9 trials; RR 0.72 95 % CI 0.54 – 0.97; P = 0.03; I2 = 35 %) ([Fig. 4 ]). Overall mortality was not reported in the study by Laine et al [14 ].
Fig. 4 Forest plot of risk ratio for overall mortality with ligation versus sclerotherapy.
Rebleeding rate from active bleeders (only 1 trial) could not generate meta-analysis.
In-hospital mortality analysis (only 3 trials) had not shown statistical difference.
The rate of complications was significantly lower with ligation (8 trials; RR 0.29
95 %CI 0.20 – 0.44; P < 0.00001; I2 = 0 %) when compared to sclerotherapy.
Comparison of sclerotherapy and vasoactive medications with sclerotherapy alone
Efficacy of bleeding control was 17 % higher with the combination of sclerotherapy
and vasoactive drugs in comparison to sclerotherapy alone (7 trials; RR of 1.17; 95 %
CI 1.10 – 1.25; P < 0.00001; I2 = 25 %) ([Fig. 5 ]).
Fig. 5 Forest plot of risk ratio for efficacy of bleeding control with sclerotherapy and
vasoactive drug versus sclerotherapy alone.
Overall rebleeding was 66 % lower (6 trials, RR 0.34; 95 % CI 0.19 – 0.61; P = 0.0003; I2 = 42 %) with the association of sclerotherapy plus vasoactive drug compared to sclerotherapy
alone ([Fig. 6 ]).
Fig. 6 Forest plot of risk ratio for overall rebleeding with sclerotherapy and vasoactive
drug versus sclerotherapy alone.
Risk of rebleeding from active bleeders at initial endoscopy was 73 % lower (4 trials;
RR 0.27; 0.12 – 0.60; P = 0.001; I2 = 35 %) with the combination of sclerotherapy and vasoactive drugs compared to sclerotherapy
alone ([Fig. 7 ]).
Fig. 7 Forest plot of risk ratio for rebleeding from active bleeders comparing sclerotherapy
and vasoactive drug versus sclerotherapy alone.
In-hospital mortality and overall mortality did not show difference in effect.
Combining sclerotherapy with vasoactive drug in seven patients resulted in control
of active bleeding and reduced risk of rebleeding in one patient when compared to
sclerotherapy alone (NNT 7, I2 = 0 %, P < 0.05) . In addition, the combination of sclerotherapy with vasoactive drug was
needed in six patients to reduce rebleeding from active bleeders in one of them when
compared to sclerotherapy alone (NNT – 6, I2 = 0 %, P < 0.05).
Comparison of ligation with the combination of ligation and sclerotherapy
Five trials [22 ]
[26 ]
[28 ]
[30 ]
[46 ] had evaluated ligation versus the combination of ligation and sclerotherapy. Risk
of complications was significantly lower with ligation (5 trials, RR 0.58; 95 %CI
0.39 – 0.88; P = 0.01; I2 0 %) when compared to the combination of ligation and sclerotherapy. However, there
were no statistically significant differences among the other analyzed outcomes.
Comparison of ligation with cyanoacrylate injection
Only one trial [42 ] evaluated ligation with cyanoacrylate injection and, therefore, could not generate
meta-analysis. This trial showed no difference in efficacy of bleeding control, rebleeding
rate, or mortality rate with cyanoacrylate injection compared with endoscopic ligation.
Comparison of ligation with ligation and vasoactive drugs
Only two trials evaluated this treatment combination, one with somatostatin and other
with octreotide [20 ]
[39 ]. Among the outcomes analyzed, only overall rebleeding ([Fig. 8 ]) and in-hospital mortality ([Fig. 9 ]) generated meta-analysis, but with no significant statistical difference.
Fig. 8 Forest plot of risk ratio for overall rebleeding with ligation alone versus ligation
and vasoactive drug.
Fig. 9 Forest plot of risk ratio for in-hospital mortality with ligation alone versus ligation
and vasoactive drug.
Discussion
Thirty-six trials, including 3593 patients, evaluated treatments for AVB control.
Among them, 10 trials compared sclerotherapy with ligation, favoring ligation in terms
of efficacy of bleeding control, rebleeding, overall mortality, and rate of complications
in a statistically significant fashion. However, this comparison showed a moderate
heterogeneity.
The heterogeneity was potentially explained by the differences in two identified subgroups
as stated above (results chapter): one formed mostly by Asian studies [19 ]
[21 ]
[24 ]
[27 ]
[29 ] in which ligation was clearly superior to sclerotherapy and another formed mostly
by Western trials [13 ]
[14 ]
[15 ]
[43 ]
[45 ] in which both techniques had similar results. In the first subgroup of studies,
the main cause of cirrhosis was viral and in three of five trials, the sclerosant
used was tetradecyl sulfate with 50 % dextrose. In the second subgroup, the majority
of patients had cirrhosis secondary to excessive alcohol intake and only one study
used tetradecyl sulfate with 50 % dextrose as sclerosant. Moreover, the second subgroup
had higher percentages of active bleeders at initial endoscopy in all ligation arms
compared to the sclerotherapy arms, which was not noticed in the first subgroup of
studies. Prevalence of Child-Pugh C patients was similar in both subgroups.
Although ligation currently is considered the gold standard endoscopic method compared
to sclerotherapy, this meta-analysis could not demonstrate clearly the superiority
of one technique over the other, because there was a moderate heterogeneity (I2 = 49 %) among the studies included. We have no doubt that ligation is better than
sclerotherapy, but the advantage of ligation may not be in the bleeding episode, but
in the secondary prophylaxis with a faster and safer variceal eradication.
Sclerotherapy and vasoactive drugs combined were superior to sclerotherapy alone in
regard to efficacy of bleeding control, overall rebleeding rate, and rebleeding rate
from active bleeders in seven, six and four trials, respectively ([Fig. 5 ], [Fig. 6 ], [Fig. 7 ]). There is a compelling body of evidence that the combination of sclerotherapy and
vasoactive drugs is more effective than sclerotherapy alone in hemorrhage control.
This meta-analysis confirmed that with a highly significant statistical difference
and a low heterogeneity among the studies (7 trials; RR of 1.17; 95 % CI 1.10 – 1.25;
P < 0.00001; I2 = 25 %). However, there was no difference in respect to mortality in the meta-analysis
and in any individual RCT. It is interesting to note that none of these studies were
performed in North America (2 European, 1 Brazilian and 4 Asian trials).
Many previous trials and meta-analyses have shown that vasoactive drugs are better
than placebo, vasoactive drugs are similar to sclerotherapy, and the combination of
vasoactive drugs and sclerotherapy is superior to sclerotherapy alone [25 ]
[47 ]
[48 ]. A recent meta-analysis [47 ] even compared therapeutic interventions for AVB with placebo, which has been unacceptable
as a treatment option since the early 1990 s.
Another technique that generated a meta-analysis and is not performed anymore is the
combination of ligation and sclerotherapy. This therapy was abandoned due to a high
incidence of side effects, which was confirmed by our study; nonetheless in this meta-analysis,
it was demonstrated to be as effective as ligation alone in bleeding control, rebleeding,
and mortality.
In this study, when we analyzed separately active bleeders at the moment of initial
endoscopy, use of sclerotherapy with vasoactive drugs was superior to sclerotherapy
alone. In 439 patients from four studies, combined therapy reduced rebleeding by 22 %
(95 %CI 1.13 – 1.32) with no heterogeneity. We could not evaluate mortality in this
subgroup of patients because the studies, when quoting mortality, did not state this
outcome separately (they quoted mortality for both active and non-active bleeders).
Although most studies reported in the literature included patients with recent and
ongoing hemorrhage, it should be emphasized that therapies used after bleeding had
spontaneously stopped will have their results overestimated. Active bleeding at endoscopy
is a well-known risk factor for worse outcomes in patients with variceal as well as
non-variceal bleeding [3 ]. Only six studies of the 36 analyzed included only patients with active variceal
bleeding, four of them compared sclerotherapy with the combination of sclerotherapy
and octreotide. The other 30 RCTs pooled together the results of the different treatments
among active and non-active bleeders at time of endoscopy.
Notwithstanding we have done a meta-analysis with solely two studies comparing ligation
plus vasoactive drug versus ligation alone, this is the only available meta-analysis
grouping this treatment, which is recommended by AASLD, the American Society of Gastrointestinal
Endoscopy (ASGE) and EASL guidelines as the gold standard in management of variceal
bleeding (considered as level of evidence 1a, grade A recommendation) [49 ]
[50 ]
[51 ]. ESGE has no current guideline about this issue. Although that recommendation is
routinely used in clinical practice, just two Asian studies evaluated use of ligation
plus vasoactive drugs in comparison to ligation alone [20 ]
[40 ] and another trial compared ligation plus octreotide versus octreotide alone [42 ].
In the study by Sung et al., ligation and somatostatin was highly superior to ligation
alone in management of variceal bleeding. On the other hand, in the study by Sarin
et al., published as an abstract, the combination of ligation and octreotide did not
show an advantage over ligation alone. When performing a meta-analysis of both these
studies, there was clearly no benefit of combination therapy in terms of rebleeding
and mortality. It is important to note that no Western study evaluated the role of
ligation plus vasoactive drug in treatment of AVB.
There was no study evaluating use of early TIPS in AVB included in this meta-analysis.
The only study using early TIPS selected was excluded because all patients received
ligation or sclerotherapy in the first 24 hours, before randomization [52 ].
As in every meta-analysis, comparison of studies may have been impaired by differing
in-hospital follow-up, which in some studies is evaluated at 5 days and in others
at 6 weeks. In this study, those data have been pooled together as overall mortality.
Time until rebleeding occurred also varied among studies, ranging from 2 to 5 days
and was considered as one sole group. Notably, we excluded patients with hepatocellular
carcinoma, who comprise at least one-fifth of bleeders. However, this population of
patients has a worse response to any treatment and should be evaluated separately.
Furthermore, we also excluded a few articles that were not published in English due
to their unavailability, although their inclusion would not affect the final analysis.
Other possible limitations of our study are the unclear risk for concealment of treatment
allocation in 18 trials and high risk for blinding of participants/personnel in 10
trials. Meanwhile, the blinding of endoscopists and patients undergoing upper digestive
endoscopy is impossible, as in studies involving surgical interventions. Moreover,
only nine of 36 studies mentioned conflict of interest. In addition, prediction intervals
indicated a significant amount of uncertainty on treatment effect sizes and direction
for several of the meta-analytic comparisons performed, which means that many of the
research questions addressed are still unanswered. Larger and well-designed trials
are needed in this field.
On the other hand, studies using placebo as a treatment were not included because
there are well-established treatments available for AVB.
During the last decade, mortality rates with acute variceal bleeds have decreased.
Routine medical care varied with respect to use of diagnostic and/or therapeutic endoscopy,
balloon tamponade, resuscitation policy, and antibiotic prophylaxis for spontaneous
bacterial peritonitis.
Conclusion
In summary, the combination of sclerotherapy and vasoactive drugs is superior to sclerotherapy
or vasoactive drugs alone in management of variceal bleeding. Ligation was better
than sclerotherapy as a treatment option for variceal bleeding, although heterogeneity
of the results may invalidate this assumption. Although society guidelines recommend
the combination of endoscopic band ligation and vasoactive medications for treatment
of AVB, this statement could not be evidenced in the literature.
