revisi 4 diterima

8/2/2019 Revisi 4 diterima

1/30

www.medscape.com

Meta-Analysis

Predictors of Rebleeding After Endoscopic

Treatment for Bleeding Peptic UlcerP. Garca-Iglesias; A. Villoria; D. Suarez; E. Brullet; M. Gallach; F. Feu; J. P. Gisbert;A. Barkun; X. Calvet

Posted: 11/18/2011; Alimentary Pharmacology & Therapeutics. 2011;34(8):888-900. 2011 Blackwell Publishing

Abstract and Introduction

Abstract

Background Determining the risk of rebleeding after endoscopic therapy for pepticulcer bleeding (PUB) may be useful for establishing additional haemostaticmeasures in very high-risk patients.Aim To identify predictors of rebleeding after endoscopic therapy.Methods Bibliographic database searches were performed to identify studiesassessing rebleeding after endoscopic therapy for PUB. All searches and dataabstraction were performed in duplicate. A parameter was considered to be anindependent predictor of rebleeding when it was detected as prognostic bymultivariate analyses in 2 studies. Pooled odds ratios (pOR) were calculated forprognostic variables.Results Fourteen studies met the prespecified inclusion criteria. Pre-endoscopic

predictors of rebleeding were: (i) Haemodynamic instability: significant in 9 of 13studies evaluating the variable (pOR: 3.30, 95% CI: 2.574.24); (ii) Haemoglobinvalue: significant in 2 of 10 (pOR: 1.73, 95% CI: 1.142.62) and (iii) Transfusion:significant in two of six (pOR not calculable). Endoscopic predictors of rebleedingwere: (i) Active bleeding: significant in 6 of 12 studies (pOR: 1.70, 95% CI: 1.312.22); (ii) Large ulcer size: significant in 8 of 12 studies (pOR: 2.81, 95% CI: 1.984.00); (iii) Posterior duodenal ulcer location: significant in four of eight studies (pOR:3.83, 95% CI: 1.3810.66) and (iv) High lesser gastric curvature ulcer location:significant in three of eight studies (pOR: 2.86; 95% CI: 1.694.86).Conclusions Major predictors for rebleeding in patients receiving endoscopictherapy are haemodynamic instability, active bleeding at endoscopy, large ulcersize, ulcer location, haemoglobin value and the need for transfusion. These riskfactors may be useful for guiding clinical management in patients with PUB.

Introduction

Peptic ulcer bleeding (PUB) is a major cause of acute nonvariceal gastrointestinalbleeding. It is a common reason for emergency hospital admission and a majorcause of mortality, morbidity and health care expenditure.[1,2]
http://www.medscape.com/http://www.medscape.com/http://www.medscape.com/


2/30

In recent years, improvements in the treatment of PUB have reduced the risk ofrebleeding and death. Most severe bleeding is associated with high-risk endoscopicstigmata. In these patients, the combination of aggressive volume repletion andrapid correction of hypotension,[3] endoscopic therapy and intravenous proton-pumpinhibitors (PPI) improve outcomes.[410]

Recurrent haemorrhage is one of the most significant predictive factors formortality.[11,12] The risk of rebleeding in patients with clean ulcers is insignificant andrecurrence occurs mainly after endoscopic therapy of ulcers showing high-riskstigmata.[5,13] Identifying patients at very high risk for rebleeding within this populationmay allow targeting additional haemostatic measures in a cost-effective way. Forexample, routine use of second-look endoscopy in all endoscopically treatedpatients is not cost-effective [6] but becomes so when performed in those at very highrisk.[14] At the other end of the scale, identifying patients with the lowest risk ofrebleeding after therapy may also be useful: A previous study by our group showedthat highly selected patients could be safely discharged after endoscopic

therapy.[15] Even if not discharged immediately, patients at low risk of rebleedingafter therapeutic endoscopy are likely to benefit from an early switch to oral PPI anddischarge instead of maintaining the costly and uncomfortable hospitalisation for a72-h PPI perfusion that is currently recommended in all patients after endoscopictherapy.[6]

The scores currently available were not designed to predict rebleeding afterendoscopic therapy, and may not work well in this setting: [16] Most have beendeveloped in cohorts that included different causes of haemorrhage and patients notreceiving endoscopic treatment. In addition, many were designed on the basis of asingle study and their reproducibility has never been satisfactory.[1720] Finally, most

scores were designed to predict mortality or a combination of mortality andrebleeding instead of only rebleeding. Predictors of ulcer rebleeding differ from thosefor mortality: they are related more to the severity of haemorrhage and the ulcercharacteristics, whereas predictors for mortality are mainly comorbidity andrebleeding.[11,12,21]

In consequence, it is important from a clinical point of view to develop a specificscore for predicting rebleeding after endoscopic therapy. The first step in developingan accurate score is to identify the most important predictive variables. To ensuregeneral applicability, we aimed to determine which variables have been found to beprognostic in a range of populations, instead of deriving them from a single study.For this reason, the present study performed a systematic review and a meta-analysis to identify the most consistent predictors of recurrent haemorrhage inpatients after endoscopic therapy.

In addition, little attention has been paid to the heterogeneity in the definition of thedifferent prognostic variables. A secondary aim of the study was to address thisissue and to evaluate whether the definition or the cut-off value used could influence


3/30

a variable's prognostic value. Therefore we set out to determine how these variablesare described in the different studies and to establish whether a reliable cut-off valuefor predicting rebleeding can be established.

Methods

Search Strategy

A systematic review of the evidence was performed covering the period fromJanuary 1990 to June 2010. Relevant papers were identified by searching PubMed,the ISI Web of Knowledge and the Abstracts from the Digestive Disease Week(DDW) from 1990 to 2010 and the United European Gastroenterology Week(UEGW) from 1995 to 2009. The search was restricted to human studies in adultspublished in English, Spanish or French. Multiple search strategies were used.Search strings for each database were:

Pubmed (ulcer AND rebleeding AND factors (160 cites), ulcer AND bleeding ANDfailure AND endoscopy (52 cites), ulcer AND rebleeding AND endoscopy (366 cites),ulcer AND rebleeding AND endoscopy AND treatment (345 cites), ulcer ANDrebleeding AND risk (245 cites), (Ulcer*[ti] OR nonvariceal[ti]) AND (Bleed*[ti] ORrebleed*[ti] OR failure[ti]) AND endoscop*[ti] (365 cites).

ISI web of Knowledge:ulcer AND rebleeding AND factors (123 cites), ulcer andbleeding AND failure AND endoscopy (33 cites), ulcer AND rebleeding ANDendoscopy (181 cites), ulcer AND rebleeding AND endoscopy AND treatment (107cites), ulcer AND rebleeding AND risk (216 cites).

In addition, a fully recursive search of the references of the original studies andsignificant reviews in the personal databases of the authors was performed to findstudies not identified by the previous searches. Abstracts of the articles selected inthe search were reviewed separately by two of the authors (PG and AV) and thearticles that met the inclusion criteria were selected for further analysis.Discrepancies regarding the eligibility of a given study were solved by repeatedreview and discussion with a third author (XC).

Inclusion and Exclusion Criteria

The selection criteria were: (i) Inclusion of patients with peptic ulcer bleeding whopresented high-risk stigmata for rebleeding (Forrest Ia, Ib, IIa and IIb) and who had

received successful endoscopic treatment. (ii) Evaluation using multivariate analysisof predictors of rebleeding in patients who received endoscopic therapy. (iii) Studiesin humans, (iv) Studies published in English, Spanish or French. (v) Data notduplicated in another manuscript.

Dual publications were excluded; when multiple publications of the same patients'group were retrieved, only the more recent or complete version was included.Therefore, abstracts previous to published articles were not evaluated. Pharmaco-


4/30

dynamic studies, studies of treatment in upper gastrointestinal bleeding (UGIB) ofunspecified causes or from aetiologies other than peptic ulcer disease, were alsoexcluded.

Data Abstraction

The following data were abstracted for each of the studies: (i) date and setting of thetrial, (ii) demographics of the study population, (iii) definition of UGIB, (iv) definitionof rebleeding Including follow-up duration, (v) patient selection criteria, (vi) type ofassociated pharmacotherapy, (vii) modality of endoscopic haemostatic therapy, (viii)definition of the outcome measure, (ix) number and proportion of patients achievinginitial haemostasis, (x) number and proportion of patients experiencing rebleeding,(xi) predictors of rebleeding in the univariate analysis, (xii) predictors of rebleeding inthe multivariate analysis, (xiii) OR for each individual risk factor.

Validity Assessment

Two investigators assessed the quality of the studies independently, and a thirdresearcher was consulted in case of disagreement. Studies were graded using sixquality criteria, defined a priori, and adapted to observational studies from thosedescribed by Cook et al.[8,22] for clinical trials. Each component of the scoring systemwas graded '1' if the study met the quality criterion and '0' if it did not. Themethodological quality of each study included was also assessed using the criteriaused by the Third U.S. Preventive Services Task Force for cohort studies.[23] Thesecriteria rate the internal validity of a study as 'good' (all criteria met), 'fair' (not allcriteria met, but no fatal flaws), or 'poor' (fatal flaw in at least one of the criteria).

Statistical Methods

Although the PRISMA criteria[24] were primarily designed to evaluate meta-analysesof randomised trials, these recommendations were followed whenever applicable.

Pooled odds ratios (pOR) and their associated 95% confidence intervals (95%CI)were obtained for all variables selected as significant by at least two multivariateanalyses. pOR were calculated by including the raw data and the results of all thestudies regardless of whether the variable was significant. The analyses wereperformed using RevMan 5.2 (The Nordic Cochrane Centre, The CochraneCollaboration, Copenhagen, Denmark). The odds ratio (OR) was estimated usingthe MantelHaenszel method with a random effects model. Evidence for

heterogeneity between studies in these meta-analyses was assessed by computingI-squared statistics.[25] We considered an I2 of 0.5 as large-scale.

A previous meta-analysis[26] used a different statistical approach and calculated asummary odds ratios (sOR) combining the OR obtained in the multivariate analysisonly for the significant variables. As it excludes negative studies, this approach maycarry the risk of overestimating the prognostic value of the variable. We also


5/30

performed calculations with this method to evaluate both whether our resultscorrelate with the previous report and whether the two methods for determiningpOR/sOR were comparable. For this second analysis we performed random effectsmeta-analyses[27] for each predictor that was significant in two or more studies toobtain sOR. In some cases, ORs and confidence intervals were reconstructed from

other values such as log-OR, standard errors or P-values. Analyses were conductedin stata version 10 software (StataCorp LP, Texas, USA). These combined as ORsfrom significant studies were labelled sORs in contrast to the OR obtained using rawdata of all studies that were labelled as pOR.

Results

Results of the Search: Studies Excluded

The search identified 2193 citations. After reviewing the abstracts and additionalsearches, 33 articles suitable for analysis were detected. Nineteen of them were

excluded, for a variety of reasons. Ten articles included (either in addition orexclusively) patients without endoscopic treatment.[2,2836] Five did not perform or didnot report the results of multivariate analysis. [3741] One study determined risk factorsexclusively for mortality,[12] and another included a large proportion of patients withbleeding from aetiologies other than peptic ulcer.[42] Finally, one study determinedthe predictors of rebleeding after a second-look endoscopy,[43] and one used a scorecombining different prognostic variables but did not give values for individualprognostic variables in either the univariate or the multivariate analysis. [44] ThePRISMA flow-diagram of the study is shown in Figure 1.


6/30

Figure 1. PRISMA flow-diagram of the systematic review.

Studies Included


7/30

Fourteen studies (12 full articles and two abstracts) including 3609 patients whichevaluated the predictors of recurrent haemorrhage after endoscopic haemostatictherapy for PUB were finally included in the meta-analysis.[4558] Detailedcharacteristics of the studies are shown in Table 1. Three of the studies included asmall proportion (


8/30

2002 44 0.4 oral (E+HP) 8.6 .6

GSnchez2003

Spain R208

208/0 6572/34.6

PPI/H2-Biv

M/D(E; TC,E +

TC)

181/87.0

13/6.3

Lpez2003

Spain R336

0/336 6065/19.4

PPI/H2-Biv

M/D(E; TC,E +TC)

297/88.4

39/11.6

Thomopoulos2004

Greece

N/S191

79/112 58.336/18.9

PPIiv

M (E)154/80.6

37/19.4

Majid2007

Pakistan

N/S 74 0/74 5315/20.3

N/SD(E +APC)

74/100.0

16/21.6

Toki2008

N/S N/S200

N/SElderly

N/S N/SM/D (C+others)

N/S31/15.5

APC, argon plasma coagulation; C, hemoclip; D, dual therapy; E, epinephrine; ES,epinephrine + hypertonic saline; HP, heater probe; H2-B, H2-blocker; M,Monotherapy; MPE, multipolar electrocoagulation; N/S, not specified; P, prospective;Po, polydocanol; R, retrospective; T, thrombin; TC, thermocoagulation.

Table 2. Quality criteria and definition of rebleeding

Author

US

preventive

task force

criteria

Cook's

modified

criteria

Second-

lookCriteria for rebleeding

Villanueva1993

Fair 5 Yes*

Active bleeding at endoscopy orfresh blood in vomiting or aspirateplus either evidence ofhypovolaemia or haemoglobinrequiring transfusion

Park 1994 Fair 4 No Hematemesis + tachycardia orhypotension or transfusion of atleast 6 BU.

Brullet gastric1996

Fair 5 No

Fresh blood in vomiting oraspirate + hypovolaemia ( SBP>15 mmHg) after 6 h ortransfusion of >4 BU


9/30

Brulletduod.1996

Fair 5 NoSimilar to Brullet, gastric ulcer1996

Lin 1998 Fair 5

Fresh blood in vomiting oraspirate + hypovolaemia (SBP

30mmHg or DBP >15 mmHg afterstanding) or bloody stools

Thomopoulos2001

Fair 5 No

Tranfusion >5 BU in 24 h or 12BU in 48 h and recurrence ofbleeding with shock (SBP100/min)

Chung 2001 Fair 4 Yes

Fresh blood in vomiting oraspirate or bloody stools ,instability of vital signs, or Hb 2

mg/dL in 24 h

Huang 2002 Fair 6 No Similar to Lin 1998

Wong 2002 Fair 5 No

Fresh hematemesis, hypotension( SBP < 90 mmHg) + tachycardia(pulse rate < 110/min), melena,transfusion of >4 BU in 72 h.

G Snchez2003

Fair 5 No

Hematemesis or melena +Hematocrit >5 points in 24 h orhypovolaemia (>1000 mL/h or

3000 mL/12 h)Lpez 2003 Fair 4 No Similar to G Sanchez 2003

Thomopoulos2004

Fair 3 No

Fresh blood in vomiting and/ormelena + shock (SBP < 100mmHg , pulse rate >100/min) orHb 2 mg/dL in 24 h

Majid 2007 2 Abstract. Not described

Toki 2008 2 Abstract. Not described

BU, blood unit; HB, haemoglobin; Hto, haematocrit; SBP, systolic blood pressure.* In 105 of 233 patients. In all cases rebleeding was confirmed by endoscopy or surgery.

The design of the studies varied widely. Data regarding whether a second electiveendoscopy was performed are also shown in Table 2. Finally, the systematic reviewaccepted the definition of rebleeding of each individual study. Criteria for further


10/30

bleeding were relatively homogeneous and are described in Table 2. Furthermore,bleeding was confirmed by either endoscopy or surgery in all studies.


Author

US

preventive

task force

criteria

Cook'smodified

criteria

Second-


Villanueva1993

Fair 5 Yes*


Park 1994 Fair 4 No

Hematemesis + tachycardia or

hypotension or transfusion of atleast 6 BU.

Brullet gastric1996

Fair 5 No


Brulletduod.1996

Fair 5 NoSimilar to Brullet, gastric ulcer1996

Lin 1998 Fair 5

Fresh blood in vomiting oraspirate + hypovolaemia (SBP30mmHg or DBP >15 mmHg afterstanding) or bloody stools

Thomopoulos2001

Fair 5 No


Chung 2001 Fair 4 Yes

Fresh blood in vomiting oraspirate or bloody stools ,

instability of vital signs, or Hb 2mg/dL in 24 h


Wong 2002 Fair 5 No



11/30

G Snchez2003

Fair 5 No

Hematemesis or melena +Hematocrit >5 points in 24 h orhypovolaemia (>1000 mL/h or3000 mL/12 h)

Lpez 2003 Fair 4 No Similar to G Sanchez 2003

Thomopoulos2004

Fair 3 No




BU, blood unit; HB, haemoglobin; Hto, haematocrit; SBP, systolic blood pressure.* In 105 of 233 patients.

In all cases rebleeding was confirmed by endoscopy or surgery.


Author

US

preventive

task force

criteria

Cook's

modified

criteria

Second-


Villanueva

1993Fair 5 Yes*


Park 1994 Fair 4 NoHematemesis + tachycardia orhypotension or transfusion of atleast 6 BU.

Brullet gastric1996

Fair 5 No


Brulletduod.1996

Fair 5 No Similar to Brullet, gastric ulcer1996

Lin 1998 Fair 5

Fresh blood in vomiting oraspirate + hypovolaemia (SBP30mmHg or DBP >15 mmHg afterstanding) or bloody stools


12/30

Thomopoulos2001

Fair 5 No


Chung 2001 Fair 4 YesFresh blood in vomiting oraspirate or bloody stools ,instability of vital signs, or Hb 2mg/dL in 24 h


Wong 2002 Fair 5 No


G Snchez2003

Fair 5 No

Hematemesis or melena +Hematocrit >5 points in 24 h orhypovolaemia (>1000 mL/h or3000 mL/12 h)

Lpez 2003 Fair 4 No Similar to G Sanchez 2003

Thomopoulos2004

Fair 3 No




BU, blood unit; HB, haemoglobin; Hto, haematocrit; SBP, systolic blood pressure.* In 105 of 233 patients. In all cases rebleeding was confirmed by endoscopy or surgery.

Independent Factors Predicting Rebleeding

The risk factors that independently predicted rebleeding after endoscopic therapybased on the multivariate analysis are shown in Figure 2. Variables selected in atleast two multivariate analyses were further evaluated in detail. Factors that fulfilled

these criteria were haemodynamic instability, active bleeding at endoscopy, ulcersize, bleeding from an ulcer situated in the duodenal posterior wall or gastric highlesser curvature, transfusion and haemoglobin lower than 10 g/L.


13/30

Figure 2. Variables analysed and found significant in the different studies.

Haemodynamic Instability. Haemodynamic instability was analysed in 13 of 14studies, with a total of 3537 patients. The variable was significant in 9 of 13 studies.Haemodynamic instability was evaluated at admission in all the studies. Definition ofhaemodynamic instability was quite homogeneous across the different studies,allowing a pooled analysis of the OR. Most studies used a definition ofshock/haemodynamic instability/compromise that included a systolic blood pressure

100 b.p.m. andperipheral signs of shock. One study did not define this topic,[58] another used onlypulse rate[54] and another used a slightly different definition for shock (systolic bloodpressure 110 b.p.m.). [56] Despite this relative homogeneity in thecriteria for shock, the populations of the different studies and the methods forassessing shock were heterogeneous, as the rate of patients with haemodynamicinstability ranged from 10% to 74% depending on the study. Excluding the few


14/30

studies with extreme values, however, the range of haemodynamic instability wasmore reasonable, between 15% and 40%.

Data were available from 11 of 13 studies, including 3088 patients. Heterogeneitywas low (I2: 12%). pOR was 3.30, 95% CI 2.574.24 (Figure 3).

Figure 3. Meta-analysis of the studies evaluating hemodynamic instability.

Seven studies reported a significant OR from the multivariate analysis for calculatingthe sOR. Low heterogeneity was observed (I2: 7.2%). sOR was 3.05, 95% CI: 2.184.26 (Figure 4).


15/30

Figure 4. Comparative OR using conventional pOR and sOR of positivemultivariable analysis.


16/30

Active Bleeding at Endoscopy. Active bleeding was analysed in 12 studiesincluding 3220 patients. One abstract did not specify whether active bleeding wasanalysed [58] and one study included only patients with nonbleeding visiblevessel.[45] The variable was significant in the multivariate analysis in 6 of 12 studiesincluding 2091 patients. In this case, the definition of the variables was clearly

heterogeneous: some studies[52,53]

evaluated patients with spurting haemorrhageseparately, while others analysed spurting and oozing bleeding together.

Meta-analysis of spurting bleeding alone included four studies with a total of 745patients. The degree of heterogeneity was significant (I2: 64%). pOR was 3.30 (95%CI 1.557.03). Eleven studies analysed spurting and oozing bleeding together.Heterogeneity was moderate (I2: 31%) and the prognostic value of the variabledecreased: pOR: 1.70, 95% CI: 1.312.22 (Figures 5 and 6).

Figure 5. Meta-analysis evaluating active bleeding (both spurting and oozing) vs.no active bleeding.


17/30

Figure 6. Meta-analysis of the studies evaluating spurting bleeding vs. no activebleeding.

Regarding sOR, the respective values were: no significant heterogeneity ( I2 = 0.0%)

and an sOR of 5.33, 95% CI: 3.009.44 for spurting bleeding (two studies), andmoderate heterogeneity (I2 = 24.1%), and an sOR of 2.28, 95% CI: 1.423.67 forany active bleeding (Figure 4).

Large Ulcer Size. Large ulcer size was analysed in 12 of 14 studies including 2937patients, and proved significant in the multivariate analysis in 8 of 12 studies. Thedefinition of large ulcer varied, with ulcers larger than 1 cm. being included in fourstudies[46,50,53,56] and ulcers larger than 2 cm. being included in five.[4749,51,52] One studyused size as a continuous variable[54] and two others used cut-off values of 1.5cm.[57] and 3 cm.[58] respectively. Raw data analysis was possible for four studiesusing a cut-off of 1 cm, and for five studies using a cut-off of 2 cm. pOR was 2.31,95%CI 1.094.94 with significant heterogeneity (I2:58%) for the former studies; forthe latter, pOR was 2.81, 95% CI: 1.984.00 with lower heterogeneity (I2: 33%)(Figure 7).


18/30

Figure 7. Meta-analysis of the studies evaluating ulcer size.

SOR could only be calculated for ulcers larger than 2 cm. Heterogeneity was low -I2 = 17%- and the sOR was 2.31, 95% CI: 1.563.41 (Figure 4).

Posterior Duodenal Wall Ulcer. Posterior duodenal ulcer was analysed in eight ofthe 14 studies, including 2449 patients, and was significant in the multivariateanalysis in four. Six studies (including 1178 patients) presented enough data for themeta-analysis. The analysis showed significant heterogeneity (I2: 77%) and the

pOR was 3.83, 95%CI: 1.3810.66 (Figure 8).


19/30

Figure 8. Meta-analysis of the studies evaluating posterior duodenal ulcer location.

OR for calculating an sOR were available only from two studies. No heterogeneity

was observed (I

2

= 0.0%) and sOR was 6.15, 95%CI: 2.7713.68 (Figure 4).High Lesser Gastric Curvature Ulcer. This variable was significant in three of eightstudies, including 2705 patients. Raw data were available from six studies including953 patients. Low heterogeneity was found (I2: 15%) and the pOR was 2.86; 95%CI: 1.694.86; (Figure 9).

Figure 9. Meta-analysis of the studies evaluating high gastric lesser curvature ulcerlocation.

ORs for calculating a sOR were available from two studies, with no heterogeneity(I2:0.0%) and an sOR of2.92, 95%CI: 1.376.21 (Figure 4).

Haemoglobin value & transfusion. Transfusion was selected as significant in twoof the six studies analysing this variable. Three studies recorded the need fortransfusion after the diagnostic endoscopy and the remaining three considered onlythe period before endoscopy. In addition, one evaluated transfusion of any amount


20/30

and two the need for transfusion of 500 mL or more. Values for this variable werenot pooled because (i) in most studies the need for transfusion was recorded 12days after the diagnostic endoscopy, which limits its applicability as a prognosticvariable for deciding early additional measures, (ii) the criteria for evaluation oftransfusion were very heterogeneous.

Haemoglobin value was significant in two of 10 multivariable analyses. This variablewas selected only when transfusion was not evaluated. Haemoglobin was analysedas a continuous variable in four studies.[45,48,49,52] When a cut-off was selected, it was8.9 g/L in one study and 10 g/L in five. Data from these five studies using the 10 g/Lcut-off, and including 1810 patients, were pooled in a meta-analysis. Theheterogeneity was moderate (I2: 32%) and the pOR of 1.73, 95% CI: 1.142.62(Figure 10).

Figure 10. Meta-analysis of the studies evaluating haemoglobin value.

As an OR was available in only one study, the sOR could not be calculated.

Discussion

This systematic review identifies the presence of haemodynamic instability atadmission, active bleeding at endoscopy, and ulcer size and location as the mostconsistent prognostic variables for rebleeding in patients who received endoscopictreatment. These variables are shown to be prognostic despite the fact that the

studies included have a relatively heterogeneous design and often use differentdefinitions and cut-off points. Our findings suggest that their prognostic values arenot due to the particular design or population used in the individual studies.

A second point highlighted by the study is that most of the cut-offs used in thedifferent studies were established empirically or, at best, were taken from empiricallyestablished standards (such as the blood pressure and pulse cut-offs forhaemodynamic instability or the size of large vs. small ulcers). Whether cut-offs


21/30

specifically calculated for determining the rebleeding risk could have betterdiscriminative capacity remains to be determined.

The analysis also suggests that the cut-off used for haemoglobin (10 g/L) may beinadequate. Although the pooled analysis found borderline significant differences for

haemoglobin, transfusion seemed to achieve better discrimination, despite the factthat the criteria for evaluating transfusion varied widely between the differentstudies. Transfusion is probably a surrogate marker for haemoglobin values lowerthan 10 g/L, as it is usually indicated with haemoglobin values below 7 g/L to 9 g/L.This suggests that it might be more appropriate to apply a lower cut-off forhaemoglobin. It would be interesting to see whether assessing haemoglobin eitheras a continuous variable or with a cut-off value lower than the currently acceptedlevel of 10 g/L would be more useful for predicting rebleeding. In addition, thepresent data do not establish whether the haemoglobin value adds any significantprognostic capacity to the five variables mentioned above. We did not attempt toperform a meta-regression to ascertain this last point because the number of studies

included in the meta-analysis was insufficient.

The findings of our study confirm and complete those of the recent systematicreview by Elmunzer et al.[26] That review included 10 studies and 3524 patients andfound haemodynamic instability, active bleeding at endoscopy, large ulcer size,posterior duodenal ulcer and lesser gastric curvature ulcer and co-morbid illness tobe significant prognostic factors for ulcer rebleeding. Our review of 14 studiesconfirmed the prognostic value of all of the above except co-morbidity. Finally, ourdata suggest that haemoglobin level may also be an independent predictor ofrebleeding and may add prognostic value to the above variables.

The discrepancies between our study and the review by Elmunzer et al.[26] may bedue in part to the slightly different inclusion and exclusion criteria used. We alsoincluded one study [52] which was not selected by Elmunzer et al., and excluded twoadditional studies that did not fulfil our inclusion criteria: one that included a largeproportion of patients with low risk lesions who were not receiving endoscopictreatment [31] and another because multivariate analysis was not effectivelyperformed.[32]

Pooling only the significant values in the multivariate analysis, as Elmunzer et al. didin their study,[26] could be misleading; taking only positive studies carries a risk ofoverestimating the prognostic value of the selected variables. In consequence, the

reliability of this approach for calculating pORs is uncertain. To avoid this problemwe performed the meta-analysis using a more classical approach, examining theraw data from the univariate analysis of each study and including all studies,whether or not the variable was selected in the multivariate analysis to obtain apOR. This approach does not take into account the adjustment of independentpredictors for confounding that multivariate analysis provides, but allows theinclusion of both positive and negative studies. To compare the two methods, we


22/30

also calculated the sOR as Elmunzer et al.[26] had done, and found that it did notgenerally overestimate the risk: however, overestimation might have occurred whendata were pooled from a small number of studies as, for example, in the case ofposterior duodenal ulcer. In this case, the estimates of the sOR were far larger thanthose of the raw data meta-analysis.

Also, pooling the variables regardless of their definition may miss importantinformation and decrease the precision of the risk estimates. For this reason, wecarefully evaluated the definitions for each variable, pooling only studies usingsimilar definitions for a given variable. This approach leads us to suggest that thecut-off for ulcer size may be important and that larger ulcer size probably leads to ahigher risk of rebleeding. Our study also suggests that spurting and oozing bleedingmay have different prognostic values. Neither of these points were raised in theprevious analysis.

Many potential limitations of our meta-analysis have to be discussed. The first is that

the review included prospective and retrospective studies, an approach that mayhave limited the quality of data. There were two reasons for not includingprospective studies alone. First, very few studies (see Table 1) were trulyprospective, and by excluding retrospective ones we would have missed importantdata. Furthermore, most of the prognostic variables were 'hard' variables that areroutinely collected for clinical reasons in emergency wards and endoscopic units. Inaddition, no differences were found in the variables selected as prognostic variablesaccording to whether the study was prospective or retrospective (Table 1 and Figure1).

Table 1. Characteristics of the studies

AuthorCount

ry

Stu

dy

desi

gn

NGastric/Duo

denal

Age

(mea

n)

Wome

n

(n/%)

Drug

thera

py

Endosc

opic

treatme

nt

Initial

Hemos

thasis

(n/%)

Re-

Bleed

ing

(n/%)

Villanueva 1993

Spain N/S233

103/130 64.979/33.9

H2-Biv

M/D(E, E +Po; E +T)

217/93.1

57/24.5

Park

1994UK N/S

13

552/63 64

53/39

.3

H2-B

ivM (E)

127/94

.1

25/18

.5

Brulletgast1996

Spain R178

178/0 6557/32.0

H2-Biv

D (E +Po)

155/87.1

23/12.9

Brulletduod

Spain R120

0/120 6123/19.2

H2-Biv

D (E +Po)

88/73.3

18/15.0


23/30

1996

Lin 1998Taiwan

R100

N/S N/S N/SPPI/H2-Biv

M(MPE)

97/97.0

17/17.0

Thomopoulos2001

Greece

P427

162/319 58.684/19.7

H2-Biv

M (E)389/91.1

86/20.1

Chung2001

Korea N/S143

94/49 55.219/13.3

H2-Biv

M/D(C, ES;C +ES)

139/97.2

36/25.2

Huang2002

Taiwan

P1251

59/64 N/S 8/6.4H2-Biv

M (E)115/92.0

22/17.6

Wong2002

China P 1144

478/666 62.5 348/30.4

PPIoral

D(E+HP)

1128/98.6

110/9.6

GSnchez2003

Spain R208

208/0 6572/34.6

PPI/H2-Biv

M/D(E; TC,E +TC)

181/87.0

13/6.3

Lpez2003

Spain R336

0/336 6065/19.4

PPI/H2-Biv

M/D(E; TC,E +TC)

297/88.4

39/11.6

Thomopoulos2004

Greece

N/S191

79/112 58.336/18.9

PPIiv

M (E)154/80.6

37/19.4

Majid2007

Pakistan

N/S 74 0/74 5315/20.3

N/SD(E +APC)

74/100.0

16/21.6

Toki2008

N/S N/S200

N/SElderly

N/S N/SM/D (C+others)

N/S31/15.5


Table 1. Characteristics of the studies

Author Count Stu N Gastric/Duo Age Wome Drug Endosc Initial Re-


24/30

ry dy

desi

gn

denal (mea

n)

n

(n/%)

thera

py

opic

treatme

nt

Hemos

thasis

(n/%)

Bleed

ing

(n/%)

Villanueva 1993 Spain N/S 233 103/130 64.9 79/33.9 H2-Biv

M/D

(E, E +Po; E +T)

217/93.1 57/24.5

Park1994

UK N/S135

52/63 6453/39.3

H2-Biv

M (E)127/94.1

25/18.5

Brulletgast1996

Spain R178

178/0 6557/32.0

H2-Biv

D (E +Po)

155/87.1

23/12.9

Brulletduod1996

Spain R12

00/120 61

23/19

.2

H2-B

iv

D (E +

Po)

88/73.

3

18/15

.0

Lin 1998Taiwan

R100

N/S N/S N/SPPI/H2-Biv

M(MPE)

97/97.0

17/17.0

Thomopoulos2001

Greece

P427

162/319 58.684/19.7

H2-Biv

M (E)389/91.1

86/20.1

Chung

2001 Korea N/S

14

3 94/49 55.2

19/13

.3

H2-B

iv

M/D(C, ES;

C +ES)

139/97

.2

36/25

.2

Huang2002

Taiwan

P1251

59/64 N/S 8/6.4H2-Biv

M (E)115/92.0

22/17.6

Wong2002

China P1144

478/666 62.5348/30.4

PPIoral

D(E+HP)

1128/98.6

110/9.6

GSnchez2003

Spain R208

208/0 6572/34.6

PPI/H2-Biv

M/D(E; TC,E +

TC)

181/87.0

13/6.3

Lpez2003

Spain R336

0/336 6065/19.4

PPI/H2-Biv

M/D(E; TC,E +TC)

297/88.4

39/11.6

Thomopoulos

Greece

N/S191

79/112 58.336/18.9

PPIiv

M (E)154/80.6

37/19.4


25/30

2004

Majid2007

Pakistan

N/S 74 0/74 5315/20.3

N/SD(E +APC)

74/100.0

16/21.6

Toki2008 N/S N/S 200 N/S Elderly N/S N/S

M/D (C

+others)

N/S 31/15.5


A second important point is that the precision of endoscopic evaluation both formeasuring ulcer size and for establishing the ulcer position in the duodenum islimited.[59,60] In our case, the study confirms that these variables have prognostic

value even with the current measuring methods, which are relatively imprecise. Anadditional drawback in assessing ulcer site is the risk of a biassed assessment. Incase of doubt about the position of a duodenal ulcer during a difficult endoscopy, theendoscopist would suspect that profuse duodenal bleeding would come from thegastroduodenal artery and, therefore, would be located in the posterior wall. It wouldbe interesting to establish whether the prognostic value of these variables improvesas new methods for precise positioning and measurement of endoscopic lesionsbecome available.

Finally, a major reason for performing a meta-analysis is to avoid the problem of thepower limitations of individual studies. As nonsignificant variables were not

evaluated, we could have missed some potentially prognostic variables. However,since the objective of the analysis was to determine the most consistent and reliablevariables to predict rebleeding, we were interested in variables that were strongpredictors of rebleeding independently of patient population, supportive therapy andmedical treatment, and even if the criteria for defining the variables differed betweenthe studies. In addition, as shown in Figure 2, few other variables are likely to havesignificant prognostic value. The most promising, the presence of haematemesis,was not selected in any of the multivariate analyses, probably because it is asurrogate marker for haemodynamic instability. Additionally, many of the variablesdetected in the present review were also prognostic in the studies that we excludedbecause they did not use endoscopic treatment or because they included patientsbleeding from lesions other than peptic ulcer.[2,2833,37,42]

Since the literature review was finished, two additional articles evaluating theprognostic factors for rebleeding have been published.[61,62] The study by Hu etal.[61] included 175 patients and found older age >60 years, haemoglobin 12 mL as prognostic


26/30

variables. The second study[62] included an indeterminate number of patients withouthigh risk stigmata and was not suitable for inclusion in the review.

Future areas for research include determining the best method of measuring andcombining these prognostic variables to develop a reliable score for predicting

rebleeding after endoscopic treatment, and to validate the score in the setting ofappropriately treated patients. Validated scores would allow the evaluation of theefficacy and cost-effectiveness of different procedures according to the risk ofrebleeding of individual patients.

In conclusion, the major clinical parameters for predicting rebleeding after receivingendoscopic treatment are haemodynamic instability at admission and haemoglobinvalue. Major endoscopic parameters are active bleeding, large ulcer size and high-risk location. The use of these variables for stratifying risk in PUB patients may wellimprove the detection and management of patients at high risk for rebleeding.

References

1. UK Upper GI Bleeding Audit. UK Comparative Audit of Upper GastrointestinalBleeding and the Use of Blood. Available at:http://www.bsg.org.uk/clinical/general/uk-upper-gi-bleeding-audit.html.Accessed December 11, 2009.

2. Barkun A, Sabbah S, Enns R, et al. The Canadian Registry on NonvaricealUpper Gastrointestinal Bleeding and Endoscopy (RUGBE): endoscopichemostasis and proton pump inhibition are associated with improvedoutcomes in a real-life setting. Am J Gastroenterol2004; 99: 123846.

3. Baradarian R, Ramdhaney S, Chapalamadugu R, et al. Early intensive

resuscitation of patients with upper gastrointestinal bleeding decreasesmortality. Am J Gastroenterol2004; 99: 61922.

4. Adler DG, Leighton JA, Davila RE, et al. ASGE guideline: the role ofendoscopy in acute non-variceal upper-GI hemorrhage. GastrointestEndosc2004; 60: 497504.

5. Barkun A, Bardou M, Marshall JK. Consensus recommendations formanaging patients with nonvariceal upper gastrointestinal bleeding. AnnIntern Med2003; 139: 84357.

6. Barkun A, Bardou M, Kuipers E, et al. International ConsensusRecommendations on the Management of Patients with Non-Variceal UpperGastrointestinal Bleeding (ICON-UGIB). Ann Intern Med2010; 152: 10113.

7. Calvet X, Vergara M, Brullet E, et al. Addition of a second endoscopictreatment following epinephrine injection improves outcome in high-riskbleeding ulcers. Gastroenterology2004; 126: 44150.

8. Cook DJ, Guyatt GH, Salena BJ, et al. Endoscopic therapy for acutenonvariceal upper gastrointestinal hemorrhage: a meta-analysis. Gastroenterology1992; 102: 13948.


27/30

9. Leontiadis GI, Sharma VK, Howden CW. Proton pump inhibitor treatment foracute peptic ulcer bleeding. Cochrane Database Syst Rev2006; 1:CD002094.

10. Sung JJ, Barkun A, Kuipers EJ, et al. Intravenous esomeprazole forprevention of recurrent peptic ulcer bleeding: a randomized trial. Ann Intern

Med2009; 150: 45564.11. Marmo R, Koch M, Cipolletta L, et al. Italian registry on upper gastrointestinal

bleeding (ProgettoNazionaleEmorragie Digestive PNED 2). Predictingmortality in non-variceal upper gastrointestinal bleeders: validation of theItalian PNED Score and Prospective Comparison with the Rockall Score. AmJ Gastroenterol2010; 105: 128491.

12. Chiu PW, Ng EK, Cheung FK, et al. Predicting mortality in patients withbleeding peptic ulcers after therapeutic endoscopy. Clin GastroenterolHepatol2009; 7: 3116.

13. Forrest JA, Finlayson ND, Shearman DJ. Endoscopy in gastrointestinalbleeding. Lancet1974; 2: 3947.

14. Spiegel BM, Ofman JJ, Woods K, et al. Minimizing recurrent peptic ulcerhemorrhage after endoscopic hemostasis: the cost-effectiveness ofcompeting strategies. Am J Gastroenterol2003; 98: 8697.

15. Brullet E, Campo R, Calvet X, et al. A randomized study of the safety ofoutpatient care for patients with bleeding peptic ulcer treated by endoscopicinjection. Gastrointest Endosc2004; 60: 1521.

16. Brullet E, Calvet X, Campo R, et al. Clinical judgement versus risk scores forpredicting outcome in patients bleeding from high risk peptic ulcer.Preliminary results of a prospective multicenter study. Gastroenterology2009;136(Suppl. 1): A434.

17. Camellini L, Merighi A, Pagnini C, et al. Comparison of three different riskscoring systems in non-variceal upper gastrointestinal bleeding. Dig LiverDis2004; 36: 2717.

18. Church NI, Dallal HJ, Masson J, et al. Validity of the Rockall scoring systemafter endoscopic therapy for bleeding peptic ulcer: a prospective cohortstudy. Gastrointest Endosc2006; 63: 60612.

19. Enns RA, Gagnon YM, Barkun AN, et al. Validation of the Rockall scoringsystem for outcomes from non-variceal upper gastrointestinal bleeding in aCanadian setting. World J Gastroenterol2006; 12: 777985.

20. Vreeburg EM, Terwee CB, Snel P, et al. Validation of the Rockall risk scoringsystem in upper gastrointestinal bleeding. Gut1999; 44: 3315.

21. Rockall TA, Logan RF, Devlin HB, et al. Risk assessment after acute uppergastrointestinal hemorrhage. Gut1996; 38: 31621.

22. Barkun AN, Martel M, Toubouti Y, et al. Endoscopic hemostasis in pepticulcer bleeding for patients with high-risk lesions: a series of meta-analyses. Gastrointest Endosc2009; 69: 78699.

23. Harris RP, Helfand M, Woolf SH, et al. Current methods of the US PreventiveServices Task Force: a review of the process. Am J Prev Med2001;20(Suppl. 3): 2135.


28/30

24. Moher D, Liberati A, Tetzlaff J, et al. Preferred reporting items for systematicreviews and meta-analyses: the PRISMA statement. J Clin Epidemiol2009;62: 100612.

25. Higgins JP, Thompson SG, Deeks JJ, et al. Measuring inconsistency in meta-analyses. BMJ2003; 327: 55760.

26. Elmunzer BJ, Young SD, Inadomi JM, et al. Systematic review of thepredictors of recurrent hemorrhage after endoscopic hemostatic therapy forbleeding peptic ulcers. Am J Gastroenterol2008; 103: 262532.

27. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control ClinTrials1986; 7: 17788.

28. Jaramillo JL, Galvez C, Carmona C, et al. Prediction of further hemorrhage inbleeding peptic ulcer. Am J Gastroenterol1994; 89: 21358.

29. Lin HJ, Perng CL, Lee FY, et al. Clinical courses and predictors for rebleedingin patients with peptic-ulcers and non-bleeding visible vessels aprospective-study. Gut1994; 35: 138993.

30. Branicki FJ, Coleman SY, Fok PJ, et al. Bleeding peptic ulcer: a prospective

evaluation of risk factors for rebleeding and mortality. World J Surg1990; 14:2629.

31. Branicki FJ, Coleman SY, Lam TC, et al. Hypotension and endoscopicstigmata of recent hemorrhage in bleeding peptic ulcer: risk models forrebleeding and mortality. J Gastroenterol Hepatol1992; 7: 18490.

32. Al-Akeely MH, Alam MK, Al-Salamah SM, et al. Initial factors predictingrebleeding and death in bleeding peptic ulcer disease. Saudi Med J2004; 25:6427.

33. Hsu PI, Lin XZ, Chan SH, et al. Bleeding peptic-ulcer - risk-factors forrebleeding and sequential-changes in endoscopic findings. Gut1994; 35:

7469.34. Vidal PJ, Pagan PA, Rua GA, et al. Multivariate analysis of the risk ofhemorrhagic recurrence in gastroduodenal ulcer. Study of a prospectiveseries. Gastroenterol Hepatol2000; 23: 4227.

35. Guglielmi A, Ruzzenente A, Sandri M, et al. Risk assessment and predictionof rebleeding in bleeding gastroduodenal ulcer. Endoscopy2002; 34: 77886.

36. Bourienne A, Pagenault M, Heresbach D, et al. Multicenter prospective studyof prognostic factors of gastroduodenal ulcer hemorrhages. Reevaluation ofclinical and endoscopic factors in the era of endoscopichemostasis. Gastroenterol Clin Biol2000; 24: 193200.

37. Choudari CP, Rajgopal C, Elton RA, et al. Failures of endoscopic therapy for

bleeding peptic ulcer: an analysis of risk factors. Am J Gastroenterol1994;89: 196872.

38. Pundzius J. Clinical and endoscopic signs for the prediction of recurrentbleeding from gastroduodenal ulcers. Eur J Surg1994; 160: 68992.

39. Schilling D, Demel A, Nusse T, et al. Helicobacter pylori infection does notaffect the early rebleeding rate in patients with peptic ulcer bleeding aftersuccessful endoscopic hemostasis: a prospective single-centertrial. Endoscopy2003; 35: 3936.


29/30

40. Chow LW, Gertsch P, Poon RT, et al. Risk factors for rebleeding and deathfrom peptic ulcer in the very elderly. Br J Surg1998; 85: 1214.

41. Chen JJ, Changchien CS, Tai DI, et al. Success of endoscopic injectiontherapy in correlation with maximal one-day transfusionrequirement. Endoscopy1995; 27: 298303.

42. Travis AC, Wasan SK, Saltzman JR. Model to predict rebleeding followingendoscopic therapy for non-variceal upper gastrointestinal hemorrhage. JGastroenterol Hepatol2008; 23: 150510.

43. Chiu PW, Joeng HK, Choi CL, et al. Predictors of peptic ulcer rebleeding afterscheduled second endoscopy: clinical or endoscopicfactors? Endoscopy2006; 38: 7269.

44. Saeed ZA, Winchester CB, Michaletz PA, et al. A scoring system to predictrebleeding after endoscopic therapy of nonvariceal upper gastrointestinalhemorrhage, with a comparison of heat probe and ethanol injection. Am JGastroenterol1993; 88: 18429.

45. Thomopoulos KC, Theocharis GJ, Vagenas KA, et al. Predictors of

hemostatic failure after adrenaline injection in patients with peptic ulcers withnon-bleeding visible vessel. Scand J Gastroenterol2004; 39: 6004.

46. GarciaSanchez MV, Lopez VP, Gonzalez GA, et al. Factors associated withfailure of endoscopic therapy in gastric ulcer bleeding. GastroenterolHepatol2003; 26: 22733.

47. Lopez VP, GarciaSanchez MV, Gonzalez GA, et al. Failure of endoscopictherapy in upper gastrointestinal hemorrhage due to duodenal ulcers. RevEsp Enferm Dig2003; 95: 7009.

48. Brullet E, Campo R, Calvet X, et al. Factors related to the failure ofendoscopic injection therapy for bleeding gastric ulcer. Gut1996; 39: 1558.

49. Brullet E, Calvet X, Campo R, et al. Factors predicting failure of endoscopicinjection therapy in bleeding duodenal ulcer. Gastrointest Endosc1996; 43:1116.

50. Villanueva C, Balanzo J, Espinos JC, et al. Prediction of therapeutic failure inpatients with bleeding peptic ulcer treated with endoscopic injection. Dig DisSci1993; 38: 206270.

51. Wong SK, Yu LM, Lau JY, et al. Prediction of therapeutic failure afteradrenaline injection plus heater probe treatment in patients with bleedingpeptic ulcer. Gut2002; 50: 3225.

52. Chung IK, Kim EJ, Lee MS, et al. Endoscopic factors predisposing torebleeding following endoscopic hemostasis in bleeding peptic

ulcers. Endoscopy2001; 33: 96975.53. Thomopoulos KC, Mitropoulos JA, Katsakoulis EC, et al. Factors associated

with failure of endoscopic injection hemostasis in bleeding pepticulcers. Scand J Gastroenterol2001; 36: 6648.

54. Park KG, Steele RJ, Mollison J, et al. Prediction of recurrent bleeding afterendoscopic hemostasis in non-variceal upper gastrointestinal hemorrhage. BrJ Surg1994; 81: 14658.


30/30

55. Huang YS, Lin HJ, Fang YR, et al. Development and validation of a scoringsystem predicting failure of endoscopic epinephrine injection therapy inTaiwanese patients with bleeding peptic ulcers. Zhonghua Yi XueZaZhi(Taipei) 2002; 65: 14450.

56. Lin HJ, Tseng GY, Lo WC, et al. Predictive factors for rebleeding in patients

with peptic ulcer bleeding after multipolar electrocoagulation: a retrospectiveanalysis. J Clin Gastroenterol1998; 26: 1136.

57. Majid S, Ahmed A, Jafri W, Nasir A. Predictors of rebleed in actively bleedingduodenal ulcer after endoscopic intervention in a developing AsianCountry. Am J Gastroenterol2007; 102(Suppl. 2): S 520.

58. Toki M, Yamaguchi Y, Nakamura K, et al. Predictive risk factors porrebleeding and thirty day mortality after endoscopic hemostasis in elderlybleeding peptic ulcer patients. Gastrointest Endosc2008; 67: AB240.

59. Straker RJ, Bienvenu JC, Nord HJ. Endoscopic orientation within theduodenal bulb. Endoscopy1992; 24: 2667.

60. Sonnenberg A, Giger M, Kern L, et al. How reliable is determination of ulcer

size by endoscopy? BMJ1979; 2: 13224.61. Hu ML, Wu KL, Chiu KW, et al. Predictors of rebleeding after initial

hemostasis with epinephrine injection in high-risk ulcers. World JGastroenterol2010; 16: 54905.

62. Cheng CL, Lin CH, Kuo CJ, et al. Predictors of rebleeding and mortality inpatients with high-risk bleeding peptic ulcers. Dig Dis Sci2010; 9: 257783.

AcknowledgementsWe thank Michael Maudsley for his help with the English. Declaration of personalinterests:Xavier Calvet is a speaker for AstraZeneca y Almirall-Prodesfarma. He is

also a consultant and has received research support from AstraZeneca y Janssen-Cilag. Faust Feu is a speaker for AstraZeneca. He is also a consultant fromAstraZeneca. Alan N. Barkun is a speaker for Olympus Corporation. He is also aconsultant and has received research support from Olympus Corporation andAstraZeneca. P. Garca?Iglesias, A. Villoria, D. Suarez, E. Brullet, M. Gallach andJP. Gisbert have no conflicts of interest. Declaration of funding interests:CIBERehdwas funded by the Instituto de Salud Carlos III. This study was funded in part byresearch grants from the Societat Catalana de Digestologia and the InstitutionalResearch Committee of the Hospital Parc Tauli (Sabadell).

Alimentary Pharmacology & Therapeutics. 2011;34(8):888-900. 2011 Blackwell

Publishing

revisi 4 diterima

Documents