Proton pump inhibitors versus H2-antagonists: a meta-analysisof their ef®cacy in treating bleeding peptic ulcer

J . P. GISBERT*, L. GONZAÂ LEZ*, X. CALVET  , M. ROQUEÂ à , R. GABRIEL§, & J. M. PAJARES*

*Department of Gastroenterology and §Department of Clinical Epidemiology, University Hospital `La Princesa', Madrid,

Spain;  Servei de Medicina, Corporacio SanitaÁria Parc TaulõÂ, Sabadell, Spain; and àCenter d'Estudis, Programes i Serveis

Sanitaris (CEPSS), Barcelona, Spain

Accepted for publication 26 February 2001

SUMMARY

Purpose: To evaluate whether proton pump inhibitors

are more effective than H2-antagonists (H2-A) for the

treatment of bleeding peptic ulcer.

Data sources: PubMed database until January 2000.

Study selection: Comparative randomized trials of proton

pump inhibitors (omeprazole, lansoprazole, or pantop-

razole) vs. H2-A (cimetidine, ranitidine or famotidine).

Data extraction: Meta-analysis combining the odds ratios

(OR) of the individual studies in a global OR (Peto

method).

Outcomes evaluated: Persistent or recurrent bleeding,

need for surgery, or mortality.

Data synthesis: Eleven studies ful®lled the inclusion

criteria and contained data for at least one of the

planned comparisons. Persistent or recurrent bleeding

was reported in 6.7% (95% CI: 4.9±8.6%) of the

patients treated with proton pump inhibitors, and in

13.4% (95% CI: 10.8±16%) of those treated with H2-A

(OR 0.4; 95% CI: 0.27±0.59) (v2-homogeneity test, 18;

P1 � 0.09). Surgery was needed in 5.2% (95% CI:

3.4±6.9%) of the patients treated with proton pump

inhibitors, and in 6.9% (95% CI: 4.9±8.9%) of the

patients treated with H2-A (OR 0.7; 95% CI:

0.43±1.13). Respective percentages for mortality were

1.6% (95% CI: 0.9±2.9%) and 2.2% (95% CI: 1.3±3.7%)

(OR 0.69; 95% CI: 0.31±1.57).

Sub-analysis: Five studies evaluated the effect of both

therapies given in bolus injections on persistent or

recurrent bleeding rate, which was 6% (95% CI:

3.6±8.3%) and 8.1% (95% CI: 5.3±10.9%), respectively

(OR, 0.57; 95% CI: 0.31±1.05). Persistent or recurrent

bleeding in high risk patients (Forrest Ia, Ib and IIa)

occurred in 13.2% (95% CI: 7.9±8%) of the patients

treated with proton pump inhibitors and in 34.5%

(27±42%) of those treated with H2-A (OR 0.28; 95% CI:

0.16±0.48). In patients not having endoscopic therapy,

persistent or recurrent bleeding was reported, respect-

ively, in 4.3% (95% CI: 2.7±6.7%) and in 12% (95% CI:

8.7±15%) (OR 0.24; 95% CI: 0.13±0.43). Less marked

differences were observed in patients having adjunct

endoscopic therapy: 10.3% (95% CI: 6.7±13.8%) and

15.2% (11.1±19.3%) (OR 0.59; 95% CI: 0.36±0.97).

Moreover, the signi®cance disappeared in this group

when a single outlier study was excluded.

Conclusions: Proton pump inhibitors are more effective

than H2-A in preventing persistent or recurrent bleed-

ing from peptic ulcer, although this advantage seems to

be more evident in patients not having adjunct sclerosis

therapy. This bene®cial effect seems to be similar or

even more marked in patients with Forrest Ia, Ib or IIa

ulcers. However, proton pump inhibitors are not more

effective than H2-A for reducing surgery or mortality

rates. Nevertheless, the data are too scarce and

heterogeneous to draw de®nitive conclusions, and

further comparative trials are clearly warranted.

Correspondence to: Dr J. P. Gisbert, Playa de MojaÂcar 29. Urb. Bonanza, 28669 Boadilla del Monte, Madrid, Spain.E-mail: gisbert@meditex.es

Aliment Pharmacol Ther 2001; 15: 917±926.

Ó 2001 Blackwell Science Ltd 917

INTRODUCTION

Acute upper gastrointestinal haemorrhage is a major

cause of morbidity and mortality, with peptic ulcer being

the most frequent source of bleeding.1 It has been

estimated that approximately 2±3% of duodenal ulcer

(DU) patients who are not receiving antisecretory therapy

are likely to develop haemorrhage during each year of

follow-up study, giving a cumulative risk of haemor-

rhage after 5 years of approximately 10±14%.2 Patients

with bleeding ulcers account for an overall mortality

rate that has remained around 5±10% for the past

50 years, despite improved medical and surgical treat-

ments, the development of diagnostic and therapeutic

endoscopy, and the availability of intensive care units.3

Data from in vitro studies suggest that clotting proceeds

more ef®ciently and the dissolution of clots by proteo-

lytic enzymes occurs more slowly at high pH levels.4±7

Both acid and pepsin alter coagulation by interfering

with the intrinsic and extrinsic coagulation system,

®brinogen polymerization, and platelet aggregation.4

Therefore, pharmaco-therapy of bleeding ulcers attempts

to improve the microenvironment at the bleeding point

by keeping the gastric pH above the proteolytic range

for pepsin to stabilize the clotting process.

Traditionally, pharmacological treatment for bleeding

peptic ulcer has included H2-receptor antagonists, but

these drugs have shown only marginal bene®cial effects

or no effect at all when compared with placebo.1, 8±10

The lack of a clear bene®cial effect of H2-antagonists

could be due, at least in part, to the limited control of

gastric pH. This is because at the conventional recom-

mended doses of these drugs, intragastric pH cannot be

maintained higher than 4.0 for a long period in patients

with a bleeding peptic ulcer.11±13

On the other hand, intravenous proton pump inhib-

itors produce consistently high intragastric pH values in

patients with bleeding peptic ulcers, whereas cimetidine

or ranitidine is less effective.14±17 However, as the half-

life of omeprazole in the circulation is short, it has been

suggested that it needs to be given in continuous

infusion instead of in bolus injections. Nevertheless,

although some placebo-controlled trials have demon-

strated a bene®cial effect of proton pump inhibitors on

bleeding peptic ulcer, other studies have demonstrated

no effect in outcome of bleeding when comparing

proton pump inhibitors and placebo.18±21

The prevention of even a low effect is a worthwhile aim

for a condition that is common.9 However, in order to

reliably detect such a moderate effect on persistent or

recurrent bleeding, need for surgery, or mortality, the

randomization of a large number of patients is required.

Most of the studies comparing proton pump inhibitors

vs. H2-antagonists have a relatively small sample size,

thus lacking the power to reliably demonstrate the

differences between the two regimens. Although several

well-designed randomized clinical trials comparing the

two drugs have been published, no formal meta-analysis

has been performed to date. Therefore, our aim was to

conduct a meta-analysis of the studies comparing

proton pump inhibitors and H2-antagonists for the

treatment of bleeding peptic ulcer.

PATIENTS AND METHODS

Search strategy

Bibliographical searches were performed in the PubMed

(Internet) database, including studies available until

January 2000, looking for the following words (all ®elds):

bleeding, cimetidine, ranitidine, famotidine, omeprazole,

lansoprazole, pantoprazole, proton pump inhibitor, pep-

tic ulcer. References of reviews on the treatment of

bleeding peptic ulcer, and from the articles selected for

the study, were also examined in search of articles

meeting inclusion criteria. Articles published in any

language were included. The bibliographic searches were

performed independently by two different reviewers.

Selection criteria

The selection criteria were as follows: (i) articles had to

report comparative randomized trials; (ii) they had to

include at least one branch of treatment consisting of

a histamine H2-antagonist (cimetidine, ranitidine or fa-

motidine, which were pooled together in the present

meta-analysis) and another branch with a proton pump

inhibitor (omeprazole, lansoprazole, or pantoprazole, also

pooled together); (iii) studies had to evaluate these

therapies in patients with bleeding gastroduodenal ulcer;

(iv) the effect of therapy had to be evaluated by means of

at least one of the following variables: persistent or recur-

rent bleeding, need for surgery, or mortality; (v) only stud-

ies that clearly stated information about the number of

treated patients in each therapeutic group were included.

Publications identi®ed as duplicates were excluded.

The quality of the studies was assessed using a score

proposed by Jadad et al. based on three items:22

918 J. P. GISBERT et al.

Ó 2001 Blackwell Science Ltd, Aliment Pharmacol Ther 15, 917±926

(i) randomization (one point if yes, or two points if the

method to generate the sequence of randomization

was described and was appropriate); (ii) double-

blinding (one point if yes, or two points if the method

of double-blinding was described and was appropriate);

and (iii) description of withdrawals and dropouts (one

point).

Sub-analyses were planned a priori to compare

studies in sub-groups depending on: omeprazole dose

(40 mg/12 h and 40 mg/8 h), schemes of drug adminis-

tration (bolus injections and continuous infusion),

endoscopic (Forrest) classi®cation [active bleeding (Ia,

Ib) and non-active bleeding (IIa, IIb, IIc)],23 and the

application of concomitant endoscopic therapy. All

these factors have been reported to in¯uence, in some

studies, the risk of persistent or recurrent bleeding or the

effect of proton pump inhibitor on bleeding peptic ulcer.

Statistics

The outcomes considered in this meta-analysis were

persistent or recurrent bleeding, need for surgery, and

mortality. The homogeneity of effects throughout

studies was appraised using a homogeneity test based

on the v2-test. Due to the low power of this test, a

minimum cut-off P-value of 0.20 was established as a

threshold of homogeneity: lower values indicated het-

erogeneity, and prevented us from relying on the

combination of the study results. Meta-analysis was

performed combining the Peto odds ratios (OR) of the

individual studies in a global OR, under the assumption-

free model (or ®xed effects model). We chose to use the

®xed effects model to obtain more precision on the

estimates, because the con®dence intervals are nar-

rower than with the random effects model. Signi®cance

and 95% con®dence intervals (95% CI) are provided for

the combined OR. All calculations were performed with

the freeware program Review Manager 4.1. The statis-

tical methods and formulae are described in the

Handbook of the Cochrane Collaboration and the

RevMan User Guide.24

RESULTS

Description of studies

Eleven studies ful®lled the inclusion criteria and con-

tained data for at least one of the planned compar-

isons.15, 16, 25±33 Detailed characteristics of the studies

are shown in Table 1, where it can be seen that

omeprazole was the proton pump inhibitor used in all

but in one study (in which lansoprazole was prescribed),

whilst both cimetidine and ranitidine were given in

different studies.

Effect of therapy on persistent or recurrent bleeding

Ten out of the 11 studies gave information about the

effect of therapy on persistent or recurrent bleeding, and

in the remaining study we contacted and obtained the

data from the authors.29 Persistent bleeding was

de®ned, depending on the study, as: (a) more than

2.5 L of blood or more than 4 units blood/day

necessary to maintain haemoglobin level above

10 g/L;26, 27, 33 or (b) persistence of bleeding during

the ®rst 48 h or 72 h by clinical and endoscopic

criteria.27±29 Recurrent bleeding was de®ned,

depending on the study, as: (a) blood in the stomach

or a fresh blood clot or bleeding in the ulcer

base at endoscopy;15, 25, 26, 30±32 (b) haematemesis/

melena;16, 27, 31, 32 (c) haemodynamic and clinical

evidence of hypovolemia;16, 31, 32 or (d) decrease in

haemoglobin requiring transfusion, in all cases during

the same hospitalization and after bleeding was initially

stopped by both clinical and analytical criteria.16, 31

In total, 681 patients were treated with proton pump

inhibitors, and 671 with H2-antagonists. The results of

these studies are graphically summarized in Figure 1.

Thus, persistent or recurrent bleeding was reported in

6.7% (95% CI: 4.9±8.6%) of the patients treated with

proton pump inhibitors, and in 13.4% (95% CI:

10.8±16%) of those treated with H2-antagonists.

The OR for the effect of proton pump inhibitors vs.

H2-antagonists on persistent or recurrent bleeding was

0.4 (95% CI: 0.27±0.59). The v2 homogeneity test was

18 (P2 � 0.09), thus indicating mild heterogeneity and

suggesting the need to perform sub-analyses depending

on different factors.

Effect of therapy on the need for surgery

The results of the eight studies comparing the effects of

proton pump inhibitors (618 patients) vs. H2-antag-

onists (621 patients) on the need for surgery are

summarized in Figure 2. Surgery was needed in 5.2%

(95% CI: 3.4±6.9%) of the patients treated with proton

pump inhibitors, and in 6.9% (95% CI: 4.9±8.9%) of the

patients treated with H2-antagonists. The OR for the

PROTON PUMP INHIBITORS VS. H 2 -ANTAGONISTS IN BLEEDING PEPTIC ULCER 919

Ó 2001 Blackwell Science Ltd, Aliment Pharmacol Ther 15, 917±926

Ta

ble

1.

Ch

ara

cter

isti

cso

fst

ud

ies

com

pa

rin

gp

roto

np

um

pin

hib

ito

rsv

s.H

2-a

nta

go

nis

tsfo

rth

etr

eatm

ent

of

ble

edin

gp

epti

cu

lcer

Ref

eren

ceD

rug

DU

/GU

/oth

erle

sio

nE

nd

osc

op

icth

era

py

Qu

ali

tysc

ore

Per

sist

ent*

or

recu

rren

tb

leed

ing

Su

rger

yM

ort

ali

ty

Bru

nn

eret

al.

19

90

26

Om

epra

zole

DU

(6/1

9),

GU

(11

/19

),o

ther

(2/1

9)

No

23

/19

*N

AN

A

Ra

nit

idin

eD

U(9

/20

),G

U(8

/20

),o

ther

(3/2

0)

No

17

/20

*N

AN

A

Ca

rdi

etal

.1

99

73

3O

mep

razo

leN

AN

o2

1/2

1*

1/2

00

/21

Ra

nit

idin

eN

AN

o7

/24

*2

/17

0/2

4

La

ben

zet

al.

19

97

25

Om

epra

zole

DU

(10

/20

),G

U(1

0/2

0)

Fo

rres

tI,

IIa

33

/20

NA

NA

Ra

nit

idin

eD

U(1

0/2

0),

GU

(10

/20

)F

orr

est

I,II

a2

/20

NA

NA

La

na

set

al.

19

95

16

Om

epra

zole

DU

(13

/28

),G

U(1

5/2

8)

No

26

/28

1/2

82

/28

Ra

nit

idin

eD

U(1

5/2

3),

GU

(8/2

3)

No

9/2

35

/23

2/2

3

Lin

etal

.1

99

71

5O

mep

razo

leD

U(7

/13

),G

U(3

/13

),o

ther

(3/1

3)

No

32

/13

NA

NA

Om

epra

zole

DU

(9/1

3),

GU

(4/1

3)

No

2/1

3N

AN

A

Cim

etid

ine

DU

(8/1

3),

GU

(5/1

3)

No

5/1

3N

AN

A

Cim

etid

ine

DU

(7/1

3),

GU

(6/1

3)

All

pa

tien

ts2

/13

NA

NA

Lin

etal

.1

99

83

0O

mep

razo

leD

U(2

8/5

0),

GU

(21

/50

),o

ther

(1/5

0)

All

pa

tien

ts3

2/5

00

/50

0/5

0

Cim

etid

ine

DU

(19

/50

),G

U(2

7/5

0),

oth

er(4

/50

)A

llp

ati

ents

12

/50

0/5

02

/50

Mic

hel

etal

.1

99

43

2L

an

sop

razo

leD

U(2

0/3

8),

GU

(18

/38

)F

orr

est

Ia4

8/3

85

/38

1/3

8

Ra

nit

idin

eD

U(2

1/3

7),

GU

(16

/37

)F

orr

est

Ia1

1/3

79

/37

2/3

7

Ort

õÂet

al.

19

95

28

Om

epra

zole

NA

No

22

/26

79

/26

72

/26

7

Ra

nit

idin

eN

AN

o4

/25

21

1/2

52

2/2

52

Pe r

ezF

lore

set

al.

19

94

27

Om

epra

zole

DU

(31

/38

),G

U(7

/38

)N

o1

1/3

8*

1/3

80

/38

Ra

nit

idin

eD

U(3

2/4

3),

GU

(11

/43

)N

o3

/43

*1

/43

0/4

3

Uri

ba

rren

aet

al.

19

94

29

Om

epra

zole

DU

(66

/13

1),

GU

(37

/13

1),

oth

er(2

8/1

31

)F

orr

est

Ia,

Ib,

IIa

15

/13

1*

6/1

31

2/1

31

Cim

etid

ine

DU

(66

/15

1),

GU

(46

/15

1),

oth

er(3

9/1

51

)F

orr

est

Ia,

Ib,

IIa

11

/15

1*

6/1

51

5/1

51

Vil

lan

uev

aet

al.

19

95

31

Om

epra

zole

DU

(30

/45

),G

U(1

2/4

5),

oth

er(3

/45

)A

llp

ati

ents

31

1/4

39

/45

3/4

5

Ra

nit

idin

eD

U(2

1/4

1),

GU

(19

/41

),o

ther

(1/4

1)

All

pa

tien

ts9

/38

9/4

11

/41

DU

,d

uo

den

al

ulc

er;

GU

,g

ast

ric

ulc

er;

NA

,n

on

-av

ail

ab

led

ata

.Q

ua

lity

sco

re,

pro

po

sed

by

Jad

ad

etal

.,b

ase

do

nth

ree

item

s:2

2(i

)ra

nd

om

iza

tio

n(1

po

int,

or

2p

oin

tsif

the

met

ho

dw

as

ap

pro

pri

ate

);(i

i)d

ou

ble

-bli

nd

ing

(1p

oin

t,o

r2

po

ints

ifth

em

eth

od

wa

sa

pp

rop

ria

te);

an

d(i

ii)

des

crip

tio

no

fw

ith

dra

wa

lsa

nd

dro

po

uts

(1p

oin

t).

Am

ore

com

ple

tev

ersi

on

of

the

tab

le,

incl

ud

ing

do

sea

nd

sch

eme

of

ad

min

istr

ati

on

of

bo

thp

roto

np

um

pin

hib

ito

ra

nd

H2

-an

tag

on

ists

an

dF

orr

est

cla

ssi®

cati

on

,is

av

ail

ab

leo

nre

qu

est

fro

mth

ea

uth

ors

(E-m

ail

:g

isb

ert@

med

itex

.es)

.

920 J. P. GISBERT et al.

Ó 2001 Blackwell Science Ltd, Aliment Pharmacol Ther 15, 917±926

effect of proton pump inhibitors vs. H2-antagonists on

the need for surgery was 0.7 (CI 95%: 0.43±1.13).

Results of the meta-analysis were homogeneous

(P3 � 0.78).

Effect of therapy on mortality

The effect on mortality of the two therapies evaluated in

this meta-analysis can be seen in Figure 3, where the

eight studies selected are represented graphically, inclu-

ding 618 patients treated with proton pump inhibitors

and 621 treated with H2-antagonists. Mortality was

1.6% (95% CI: 0.9±2.9%) in patients treated with proton

pump inhibitors and 2.2% (1.3±3.7%) in those receiving

H2-antagonists. The OR for the effect on the mortality

was 0.69 (95% CI: 0.31±1.57). Results obtained with

different studies were homogeneous (P4 � 0.63).

Sub-analysis on the ef®cacy of therapy

on bleeding peptic ulcer

Dose of proton pump inhibitors. The initial intention was

to perform separate comparisons for the different doses

of omeprazole; however, in only three studies were the

omeprazole doses of 40 mg/8 h used (vs. H2-antago-

nists in bolus injections), so the number of studies was

considered too small to combine and compare the

results. When we again attempted to perform separate

comparisons with omeprazole, doses of 40 mg/12 h, we

found only three studies, and in each fewer than 30

patients were treated with those doses, thus precluding

adequate comparison of the study results.

Scheme of administration of the drugs. Five studies

evaluated the effect of both therapies given in bolus

injections on persistent or recurrent bleeding rate (402

patients received proton pump inhibitors, and 369

received H2-antagonists). Re-bleeding was detected in

6% (95% CI: 3.6±8.3%) and 8.1% (95% CI: 5.3±10.9%),

respectively, with an OR for this comparison of 0.57

(95% CI: 0.31±1.05%). When evaluating the need for

surgery, also with bolus injections of both drugs, it was

necessary in 5.3% (95% CI: 3±7.5%) of the patients

treated with proton pump inhibitors, and in 7.2%

(95% CI: 4.6±9.9%) of those receiving H2-antagonists,

with an OR for this comparison of 0.68 (95% CI:

Figure 1. Proton pump inhibitors vs. H2-antagonists. Effect on persistent or recurrent bleeding of peptic ulcer.

Figure 2. Proton pump inhibitors vs. H2-antagonists. Effect on the need for surgery of bleeding peptic ulcer.

PROTON PUMP INHIBITORS VS. H 2 -ANTAGONISTS IN BLEEDING PEPTIC ULCER 921

Ó 2001 Blackwell Science Ltd, Aliment Pharmacol Ther 15, 917±926

0.37±1.26%). Finally, the mortality rate was 1.8%

(95% CI: 0.9±3.8%) and 1.4% (95% CI: 0.6±3.2%),

respectively, with bolus injections of proton pump

inhibitors and H2-antagonists, with an OR of 1.25

(95% CI: 0.40±3.97%). Results obtained in different

studies with bolus injections were homogeneous for all

comparisons. Other comparisons, evaluating other

schemes of administration, were precluded due to the

low number of studies in each group.

Endoscopic (Forrest) classi®cation. Although the initial (a

priori) intention was to perform sub-analyses (plural)

comparing the effect of the therapy on active ulcer

bleeding (Forrest Ia, Ib) and non-active ulcer bleeding

(Forrest IIa, IIb, IIc), the number of studies was

considered too small to combine and compare the

results. Therefore, we classi®ed (a posteriori, ad hoc

analysis) the studies into those with high or low risk of

persistent or recurrent bleeding. Thus, persistent or

recurrent bleeding in high risk patients (Forrest Ia, Ib

and IIa, that is, with active bleeding or non-bleeding

visible vessel) occurred in 13.2% (95% CI: 7.9±18%) of

the patients treated with proton pump inhibitors and in

34.5% (95% CI: 27±42%) of those treated with H2-

antagonists (OR 0.28; 95% CI: 0.16±0.48). The het-

erogeneity of the results (P5 � 0.02) disappeared when

we excluded the study of Villanueva et al., in which

adjunct endoscopic therapy was performed (OR 0.15;

95% CI: 0.08±0.3; P6 � 0.64).31

Endoscopic therapy. In patients who did not have

endoscopic therapy, persistent or recurrent bleeding

was reported in 4.3% (95% CI: 2.7±6.7%) of those

treated with proton pump inhibitors and in 12%

(95% CI: 8.7±15%) of those treated with H2-antagonists

(OR 0.24; 95% CI: 0.13±0.43). The corresponding

®gures for patients having endoscopic therapy (during

the treatment) were 10.3% (95% CI: 6.7±13.8%) and

15.2% (95% CI: 11.1±19.3%) (OR 0.59; 95% CI:

0.36±0.97%; Figure 4). The homogeneity of the results

(P7 � 0.27) was higher when we excluded the study of

Lin et al., in which heater probe thermocoagulation or

multipolar electrocoagulation instead of sclerosis was

performed (OR 0.77; 95% CI: 0.04±1.35%; P8 � 0.78).30

DISCUSSION

Pharmacological treatment for bleeding peptic ulcer has

traditionally included H2-antagonists drugs, with con-

troversial results. It remains unknown whether more

recent and powerful antisecretory drugs, namely proton

pump inhibitors, are superior to H2-antagonists in the

treatment of bleeding peptic ulcer. Several well-designed

randomized clinical trials comparing the two drugs have

been published, but the results have been contradictory,

and to date no formal meta-analysis has been per-

formed.

In the present meta-analysis, 11 studies ful®lled the

inclusion criteria and compared the effect of proton

pump inhibitors vs. H2-antagonists on bleeding peptic

ulcer (Table 1).15, 16, 25±33 When combining the results

of the studies (Figure 1) it was observed that persistent

or recurrent bleeding was less frequent with proton

pump inhibitors (6.7%) than with H2-antagonists

(13.4%) (OR 0.4; 95% CI: 0.27±0.59%). Surgery was

needed in 5.2% of the patients treated with proton

pump inhibitors, and in 6.9% of those receiving

H2-antagonists (Figure 2), which indicates a trend

towards better results with the proton pump inhibitors,

though this difference was not statistically signi®cant.

Nevertheless, the available data on the need for surgery

may lack the power to reliably demonstrate a number of

moderate but clinically important effects (the OR was

0.7, but con®dence interval limits ranged from to

Figure 3. Proton pump inhibitors vs. H2-antagonists. Effect on the mortality for bleeding peptic ulcer.

922 J. P. GISBERT et al.

Ó 2001 Blackwell Science Ltd, Aliment Pharmacol Ther 15, 917±926

0.43±1.13). This is due in part to the relatively small

size of the individual trials, and in part to the very low

proportion of patients needing surgery in most of the

studies (Table 1). Finally, a trend towards lower mor-

tality was also found with proton pump inhibitors

(1.6%) in comparison with H2-antagonists (2.2%;

Figure 3), but this difference was not statistically

signi®cant either (OR 0.69; 95% CI: 0.31±1.57). A

recent review of the literature on randomized controlled

studies evaluating proton pump inhibitors (vs. placebo,

somatostatin, or H2-antagonists) in patients with peptic

ulcer bleeding found that four of the 16 studies included

showed a statistically signi®cant decrease in overall

re-bleeding rate, and four studies also showed a signi®-

cantly decreased surgery rate, but none demonstrated a

signi®cant mortality reduction.34

Pharmacological studies in healthy subjects and

patients with ulcers have shown that infusions with

antisecretory drugs are superior to the bolus injec-

tions.35±38 The degree of acid inhibition produced by

omeprazole is closely related to the area under the

curve, but not to the plasma concentration at any given

point in time.39 Furthermore, because the half-life of

omeprazole in the circulation is short, it has been

suggested that it needs to be given more frequently than

every 8 h (a commonly used schedule) or continuously,

although other studies report similar intragastric pH

values with omeprazole given in bolus injections and in

continuous infusion.17, 40, 41 The number of studies

included in our meta-analysis using continuous infu-

sion was considered too small to combine and compare

the results. However, we found that when both drugs

were prescribed in bolus injections, persistent or recur-

rent bleeding was observed less frequently with proton

pump inhibitors (6%) than with H2-antagonists (8.1%),

although this difference did not reach statistical signi-

®cance (OR 0.57; 95% CI: 0.31±1.05). The need for

surgery, again with bolus injections of both drugs, was

also less frequent in the group treated with proton pump

inhibitors (5.3% vs. 7.2%), but the difference was not

statistically signi®cant. After bolus injections of raniti-

dine (50 mg/8 h), hourly gastric pH values ¯uctuate

widely from 7.6 to 1.6, and even with more frequent

bolus injections of ranitidine (50 mg/4 h), control of

intragastric pH is not achieved.16, 35 Therefore, if the

goal is to achieve a near neutral pH for 24 h to avoid

pepsinogen activation, ranitidine falls far short of

achieving the objective.16 However, omeprazole

prescribed in bolus injections can reduce gastric

secretion to the point of anacidity, keeping gastric pH

values generally > 6.16

Endoscopic diagnosis (e.g. Forrest classi®cation) may

be useful for selecting patients for whom proton pump

inhibitors may be bene®cial. Patients with low-risk

lesions (e.g. clean-based ulcers, or ones with ¯at

pigmented spots) generally recover regardless of ther-

Figure 4. Proton pump inhibitors vs. H2-antagonists. Effect on persistent or recurrent bleeding of peptic ulcer without (top) and with

(bottom) adjunct endoscopic therapy.

PROTON PUMP INHIBITORS VS. H 2 -ANTAGONISTS IN BLEEDING PEPTIC ULCER 923

Ó 2001 Blackwell Science Ltd, Aliment Pharmacol Ther 15, 917±926

apy. Our results show that persistent or recurrent

bleeding in high risk patients (Forrest Ia, Ib and IIa, e.g.

with active bleeding or non-bleeding visible vessel) was

less frequent with proton pump inhibitors than with

H2-antagonists (13.2% vs. 34.5%; OR 0.28; 95%

CI: 0.16±0.48). This difference was higher than that

previously calculated with all studies (OR for all Forrest

stages: 0.4), suggesting that the bene®cial effect of

proton pump inhibitors (against H2-antagonists) is

similar or even more marked in patients with Forrest I

or IIa ulcers.

Nevertheless, we should bear in mind the possible

confounding role of adjunct endoscopic therapy, which

has been proven effective in the treatment of bleeding

ulcers.42 The study of Villanueva et al., in which

endoscopic therapy was performed and showed no

bene®t of omeprazole over H2-antagonists in Forrest Ia,

Ib and IIa patients, contrasts markedly with the other

studies (in which, excepting the study of Lin et al.,

endoscopy therapy was not performed).15, 31 The results

were no longer heterogeneous when the study of

Villanueva et al. was excluded from the analysis.31 In

this respect, our meta-analysis shows that in patients

who do not have endoscopic therapy, persistent or

recurrent bleeding is less frequent with proton pump

inhibitors than with H2-antagonists (4.3% vs. 12%; OR

0.24) (Figure 4). However, the difference with these two

drugs on patients who underwent endoscopic therapy

was less clear (10.3% vs. 15.2%; OR 0.59; Figure 4).

Furthermore, the differences were no longer statistically

signi®cant (and homogeneity increased) when we

excluded the study of Lin et al., in which heater probe

thermocoagulation or multipolar electrocoagulation

instead of sclerosis was performed.30 These results

suggest that the bene®cial effect of proton pump

inhibitors on bleeding peptic ulcer, when compared

with H2-antagonists, may be mainly or exclusively

observed in those patients not having adjunct endo-

scopic therapy, and that when sclerosis is performed the

advantage of proton pump inhibitors could be limited or

even lost altogether. In other words, injection therapy

may be so effective that the additional effect of the

antisecretory drugs may be minimal. Because endo-

scopic therapy for bleeding peptic ulcers is routinely

performed (when indicated) in most centres, the role of

proton pump inhibitors may be considered less import-

ant in the future if these data are con®rmed. Most of

the patients with an actively bleeding ulcer or a visible

vessel have a very low risk of re-bleeding after

endoscopic therapy and will not therefore bene®t from

any adjunct therapy. Nonetheless, recurrence of bleed-

ing, even with endoscopic therapy, is a signi®cant

problem in selected patients with bleeding peptic ulcer.1

The risk factors for re-bleeding after endoscopic therapy

have been previously described: ulcer greater than

2 cm; haemodynamic instability; advanced age; active

haemorrhage during endoscopy; and location high in

the lesser curvature.43, 44 Measures leading to a reduc-

tion in bleeding rate in this group of patients could

result in a better outcome, although this point has not

been speci®cally investigated. Further prospective stud-

ies are needed to compare endoscopic therapy vs.

combined therapy in this sub-group of patients with

high risk of re-bleeding.

How do we explain the fact that results are better with

proton pump inhibitors than with H2-antagonists in

bleeding peptic ulcer? Some authors have demonstrated

that intravenous omeprazole produces consistently

high intragastric pH values in patients with bleeding

peptic ulcers, whereas H2-antagonists are less

effective.14±16, 41 This loss of effectiveness of ranitidine

may be due to tolerance (tachyphylaxis), which is

known to occur in response to repetitive doses of

H2-antagonists, but has never been found with proton

pump inhibitors.45, 46 As stated above, both acid and

pepsin alter coagulation by interfering with the coagu-

lation system, ®brinogen polymerization, and platelet

aggregation.4 Therefore, parenteral proton pump inhib-

itors seem to be more effective than H2-antagonists in

keeping the gastric pH above the proteolytic range for

pepsin in bleeders. Furthermore, omeprazole can heal

96% of ranitidine-resistant peptic ulcerations, and

several case reports have been published where intra-

venous omeprazole was able to stop the bleeding from

peptic lesions that did not respond to ranitidine

treatment.10 26, 46±51 Finally, because intravenous proton

pump inhibitors are more expensive than intravenous

H2-antagonists, further studies are needed to evaluate

the cost±effectiveness relationship of these two drugs.

The present meta-analysis has several limitations.

First, the signi®cant variability between studies makes

combination of the results dif®cult. For example, there

was a marked variability between studies with respect

to location of peptic ulcer, dose of proton pump

inhibitors and H2-antagonists, scheme of administration

of the drugs, Forrest classi®cation of the bleeding ulcer,

and concomitant endoscopic therapy. Second, the

relatively low number of studies (and patients in each

924 J. P. GISBERT et al.

Ó 2001 Blackwell Science Ltd, Aliment Pharmacol Ther 15, 917±926

study) included in the meta-analysis sometimes preclu-

ded adequate comparisons of outcomes, depending on

the status of important variables such as dose and

scheme of administration of proton pump inhibitors, or

Forrest classi®cation. Furthermore, comparisons

depending on Forrest classi®cation were performed

a posteriori (ad hoc analysis) and were not planned

a priori. Finally, overall analysis must be quali®ed

because of the poor data that are available (see quality

scores of the studies in Table 1).

In summary, the results of the present meta-analysis

indicate that proton pump inhibitors are more effective

than H2-antagonists in preventing persistent or recur-

rent bleeding from peptic ulcer, although this advantage

seems to be restricted to those patients who do not have

adjunct sclerosis therapy. Nevertheless, the data are too

scarce and heterogeneous to draw de®nitive conclusions

and further randomized clinical trials comparing proton

pump inhibitors and H2-antagonists are clearly

warranted.

ACKNOWLEDGEMENTS

We are indebted to Brenda Ashley and Michael

Maudsley for assistance with the English.

This study was not funded by any pharmaceutical

company.

REFERENCES

1 Laine L, Peterson WL. Bleeding peptic ulcer. N Engl J Med

1994; 331: 717±27.

2 Mignon M, Penston JG, Deltenre M, Ruszniewski P, Dobrill G.

Natural history of duodenal ulcer disease: are we at a turning

point? Gastroenterol Int 1994; 7: 95±113.

3 Gilbert DA. Epidemiology of upper gastrointestinal bleeding.

Gastrointest Endosc 1990; 36: S8±13.

4 Green FW Jr, Kaplan MM, Curtis LE, Levine PH. Effect of acid

and pepsin on blood coagulation and platelet aggregation. A

possible contributor to prolonged gastroduodenal mucosal

hemorrhage. Gastroenterology 1978; 74: 38±43.

5 Patchett SE, O'Donoghue DP. Pharmacological manipulation

of gastric juice: thrombelastographic assessment and impli-

cations for treatment of gastrointestinal haemorrhage. Gut

1995; 36: 358±62.

6 Patchett SE, Enright H, Afdhal N, O'Connell W, O'Donoghue

DP. Clot lysis by gastric juice: an in vitro study. Gut 1989; 30:

1704±7.

7 Chaimoff C, Creter D, Djaldetti M. The effect of pH on platelet

and coagulation factor activities. Am J Surg 1978; 136:

257±9.

8 RollhauSeries C, Fleischer DE. Nonvariceal upper gastroin-

testinal bleeding: an update. Endoscopy 1997; 29: 91±105.

9 Collins R, Langman M. Treatment with histamine H2 antag-

onists in acute upper gastrointestinal hemorrhage. Implica-

tions of randomized trials. N Engl J Med 1985; 313: 660±6.

10 Falk A, Darle N, Haglund U, Tornqvist A. Histamine2-receptor

antagonists in gastroduodenal ulcer haemorrhage. Scand J

Gastroenterol 1985; 110(Suppl.): 95±100.

11 Reynolds JR, Walt RP, Clark AG, Hardcastle JD, Langman MJ.

Intragastric pH monitoring in acute upper gastrointestinal

bleeding and the effect of intravenous cimetidine and raniti-

dine. Aliment Pharmacol Ther 1987; 1: 23±30.

12 Peterson WL, Barnett C, Feldman M, Richardson CT. Reduc-

tion of twenty-four-hour gastric acidity with combination

drug therapy in patients with duodenal ulcer. Gastroenterol-

ogy 1979; 77: 1015±20.

13 Merki HS, Witzel L, Kaufman D, et al. Continuous intravenous

infusions of famotidine maintain high intragastric pH in

duodenal ulcer. Gut 1988; 29: 453±7.

14 Labenz J, Borsch G. Role of Helicobacter pylori eradication in

the prevention of peptic ulcer bleeding relapse. Digestion

1994; 55: 19±23.

15 Lin HJ, Lo WC, Perng CL, Wang K, Lee FY. Can optimal acid

suppression prevent rebleeding in peptic ulcer patients with a

non-bleeding visible vessel: a preliminary report of a ran-

domized comparative study. Hepatogastroenterology 1997;

44: 1495±9.

16 Lanas A, Artal A, Blas JM, Arroyo MT, Lopez-Zaborras J, Sainz

R. Effect of parenteral omeprazole and ranitidine on gastric pH

and the outcome of bleeding peptic ulcer. J Clin Gastroenterol

1995; 21: 103±6.

17 Kiilerich S, Rannem T, Elsborg L. Effect of intravenous infu-

sion of omeprazole and ranitidine on twenty-four-hour

intragastric pH in patients with a history of duodenal ulcer.

Digestion 1995; 56: 25±30.

18 Khuroo MS, Yattoo GN, Javid G, et al. A comparison of

omeprazole and placebo for bleeding peptic ulcer. N Engl J Med

1997; 336: 1054±8.

19 Hasselgren G, Lind T, Lundell L, et al. Continuous intravenous

infusion of omeprazole in elderly patients with peptic ulcer

bleeding. Results of a placebo-controlled multicenter study.

Scand J Gastroenterol 1997; 32: 328±33.

20 Schaffalitzky de Muckadell OB, Havelund T, Harling H, et al.

Effect of omeprazole on the outcome of endoscopically treated

bleeding peptic ulcers. Randomized double-blind placebo-

controlled multicentre study. Scand J Gastroenterol 1997; 32:

320±7.

21 Daneshmend TK, Hawkey CJ, Langman MJ, Logan RF, Long

RG, Walt RP. Omeprazole versus placebo for acute upper

gastrointestinal bleeding: randomised double blind controlled

trial. Br Med J 1992; 304: 143±7.

22 Jadad AR, Moore RA, Carroll D, et al. Assessing the quality of

reports of randomized clinical trials: Is blinding necessary?

Controlled Clin Trials 1996; 17: 1±12.

23 Foster DN, Miloszewski KJ, Losowsky MS. Stigmata of recent

haemorrhage in diagnosis and prognosis of upper gastroin-

testinal bleeding. Br Med J 1978; 1: 1173±7.

PROTON PUMP INHIBITORS VS. H 2 -ANTAGONISTS IN BLEEDING PEPTIC ULCER 925

Ó 2001 Blackwell Science Ltd, Aliment Pharmacol Ther 15, 917±926

24 Clarke M, Oxman AD, eds. Cochrane Reviewers's Handbook

4.0. In: Review Manager (RevMan) 4.0. Oxford: The Cochrane

Collaboration, 1999.

25 Labenz J, Peitz U, Leusing C, Tillenburg B, Blum AL, Borsch G.

Ef®cacy of primed infusions with high dose ranitidine and

omeprazole to maintain high intragastric pH in patients with

peptic ulcer bleeding: a prospective randomised controlled

study. Gut 1997; 40: 36±41.

26 Brunner G, Chang J. Intravenous therapy with high doses of

ranitidine and omeprazole in critically ill patients with bleeding

peptic ulcerations of the upper intestinal tract: an open

randomized controlled trial. Digestion 1990; 45: 217±25.

27 Perez Flores R, Garcia Molinero MJ, Herrero Quiros C, et al.

The treatment of upper digestive hemorrhage of peptic origin:

intravenous ranitidine versus intravenous omeprazole. Rev

Esp Enferm Dig 1994; 86: 637±41.

28 Orti E, Canelles P, Quiles F, et al. Does the antisecretory agent

used affect the evolution of upper digestive hemorrhage? Rev

Esp Enferm Dig 1995; 87: 427±30 (in Spanish).

29 Uribarrena R, Bajador E, Simon MA, Sebastian JJ, Gomollon F.

Omeprazole and cimetidine in the treatment of upper digestive

hemorrhage. Rev Esp Enferm Dig 1994; 86: 878±83 (in

Spanish).

30 Lin HJ, Lo WC, Lee FY, Perng CL, Tseng GY. A prospective

randomized comparative trial showing that omeprazole pre-

vents rebleeding in patients with bleeding peptic ulcer after

successful endoscopic therapy. Arch Intern Med 1998; 158:

54±8.

31 Villanueva C, Balanzo J, Torras X, et al. Omeprazole versus

ranitidine as adjunct therapy to endoscopic injection in act-

ively bleeding ulcers: a prospective and randomized study.

Endoscopy 1995; 27: 308±12.

32 Michel P, Duhamel C, Bazin B, et al. Lansoprazole versus

ranitidine in the prevention of early recurrences of digestive

hemorrhages from gastroduodenal ulcers. Randomized dou-

ble-blind multicenter study. Gastroenterol Clin Biol 1994; 18:

1102±5 (in French).

33 Cardi M, Muttillo IA, Amadori L, et al. Intravenous omepraz-

ole versus ranitidine in the treatment of hemorrhagic duode-

nal ulcer: a prospective randomized study. Ann Chir 1997;

51: 136±9 (in French).

34 Bustamante M, Stollman N. The ef®cacy of proton-pump

inhibitors in acute ulcer bleeding: a qualitative review. J Clin

Gastroenterol 2000; 30: 7±13.

35 Ballesteros MA, Hogan DL, Koss MA, Isenberg JI. Bolus or

intravenous infusion of ranitidine: effects on gastric pH and

acid secretion. A comparison of relative ef®cacy and cost. Ann

Intern Med 1990; 112: 334±9.

36 Heiselman DE, Hulisz DT, Fricker R, Bredle DL, Black LD.

Randomized comparison of gastric pH control with intermittent

and continuous intravenous infusion of famotidine in ICU

patients. Am J Gastroenterol 1995; 90: 277±9.

37 Ostro MJ, Russell JA, Soldin SJ, Mahon WA, Jeejeebhoy KN.

Control of gastric pH with cimetidine: boluses versus primed

infusions. Gastroenterology 1985; 89: 532±7.

38 Peterson WL, Richardson CT. Sustained fasting achlorhydria:

a comparison of medical regimens. Gastroenterology 1985;

88: 666±9.

39 Cederberg C, Andersson T, Skanberg J. Omeprazole phar-

macokinetics and metabolism in man. Scand J Gastroenterol

1989; 24(Suppl. 166): 33±40.

40 Brunner G, Luna P, Thiesemann C. Drugs for pH control in

upper gastrointestinal bleeding. Aliment Pharmacol Ther

1995; 9: 47±50.

41 Artal A, Lanas A, Barrao ME, Moliner FJ, Blas JM, Lopez J.

Evaluation of intravenous ranitidine and omeprazole effect on

the 24-hour gastric ph-metry in duodenal ulcer hemorrhage.

Rev Esp Enferm Dig 1996; 88: 191±6 (in Spanish).

42 Cook DJ, Guyatt GH, Salena BJ, Laine LA. Endoscopic therapy

for acute nonvariceal upper gastrointestinal hemorrhage: a

meta-analysis. Gastroenterology 1992; 102: 139±48.

43 Brullet E, Calvet X, Campo R, Rue M, Catot LI, Donoso LI.

Factors predicting failure of endoscopic injection therapy

in bleeding duodenal ulcer. Gastrointest Endosc 1996; 43:

111±6.

44 Brullet E, Campo R, Calvet X, Coroleu D, Rivero E, Simo Deu J.

Factors related to the failure of endoscopic therapy for bleed-

ing gastric ulcer. Gut 1996; 39: 155±8.

45 Wilder-Smith CH, Merki HS. Tolerance during dosing with

H2-receptor antagonists. An overview. Scand J Gastroenterol

1992; 193(Suppl.): 14±9.

46 Merki HS, Wilder-Smith CH. Do continuous infusions of

omeprazole and ranitidine retain their effect with prolonged

dosing? Gastroenterology 1994; 106: 60±4.

47 Brunner G, Creutzfeldt W. Omeprazole in the long-term

management of patients with acid-related diseases resistant

to ranitidine. Scand J Gastroenterol 1989; 166(Suppl.):

101±5.

48 Collier DS, Crampton J, Everett WG. Acute haemorrhagic

gastritis controlled by omeprazole. Lancet 1989; 1: 776

(Letter).

49 Hetzel DJ, Bonnin M. Long term management of hemorrhagic

esophagitis with cimetidine and omeprazole. Aust N Z J Med

1986; 16: 226±8.

50 Vezzadini P, Tomassetti P, Marrano D, Labo G. Life-threaten-

ing gastrointestinal hemorrhage with omeprazole. Dig Dis Sci

1988; 33: 766±7 (Letter).

51 Gabbrielli M, Pennati P, Trallori G, Vannozzi G. Use of intra-

venous omeprazole in emergency cases of gastroduodenal

hemorrhage. Am J Gastroenterol 1992; 87: 1229 (Letter).

926 J. P. GISBERT et al.

Ó 2001 Blackwell Science Ltd, Aliment Pharmacol Ther 15, 917±926