Stress ulcer prophylaxis in the new millennium: A systematicreview and meta-analysis

Paul E. Marik, MD; Tajender Vasu, MD; Amyn Hirani, MD; Monvasi Pachinburavan, MD

I n 1969, Skillman et al (1) reporteda clinical syndrome of lethal“stress ulceration” in seven of 150(5%) consecutive intensive care

unit (ICU) patients. These patients had in

common respiratory failure, hypotension,and sepsis. Pathologic examination dem-onstrated multiple superficial ulcers thatwere confined to the gastric fundus. Fol-lowing this report, these authors per-formed a randomized controlled study inwhich 100 critically ill ICU patients atrisk of stress ulceration were randomizedto either antacid prophylaxis (titrated tokeep the gastric pH above 3.5) or no pro-phylaxis (2). Two of 51(4%) treated pa-tients had significant gastrointestinal(GI) bleeding as compared to 12 of 49(25%) control patients (p � .005). Sub-sequent studies confirmed this findingand two meta-analyses published by Cooket al (3, 4) demonstrated that both hista-

mine-2 receptor blockers (H2RBs) andsucralfate decreased the risk of bleedingfrom stress ulceration when compared toa placebo. Stress ulcer prophylaxis (SUP)become regarded as the standard of carein patients admitted to the ICU, and thisintervention is currently endorsed bymany professional bodies (5, 6). The uni-versal use of SUP has been reinforcedwith the adoption of “ventilator bundles.”Currently The Joint Commission and theInstitute for Healthcare Improvementrecommend universal SUP as a core“quality” measure for mechanically ven-tilated patients (7). Estimates indicatethat approximately 90% of critically illpatients admitted to the ICU receive some

From the Division of Pulmonary and Critical CareMedicine (PEM); Eastern Virginia Medical School, Nor-folk, VA; Division of Pulmonary and Critical Care Med-icine (TV, AH, MP), Thomas Jefferson University, Phil-adelphia, PA.

The authors have no potential conflicts of interestto disclose.

For information regarding this article, E-mail:marikpe@evms.edu

Copyright © 2010 by the Society of Critical CareMedicine and Lippincott Williams & Wilkins

DOI: 10.1097/CCM.0b013e3181f17adf

Background: Recent observational studies suggest that bleed-ing from stress ulceration is extremely uncommon in intensivecare unit patients. Furthermore, the risk of bleeding may not bealtered by the use of acid suppressive therapy. Early enteral tubefeeding (initiated within 48 hrs of intensive care unit admission)may account for this observation. Stress ulcer prophylaxis may,however, increase the risk of hospital-acquired pneumonia andClostridia difficile infection.

Objective: A systematic review of the literature to determinethe benefit and risks of stress ulcer prophylaxis and the moder-ating effect of enteral nutrition.

Data Sources: MEDLINE, Embase, Cochrane Register of Con-trolled Trials, and citation review of relevant primary and reviewarticles.

Study Selection: Randomized, controlled studies that evaluatedthe association between stress ulcer prophylaxis and gastroin-testinal bleeding. We included only those studies that compared ahistamine-2 receptor blocker with a placebo.

Data Extraction: Data were abstracted on study design, studysize, study setting, patient population, the histamine-2 receptorblocker and dosage used, the incidence of clinically significantgastrointestinal bleeding, hospital-acquired pneumonia, mortal-ity, and the use of enteral nutrition.

Data Synthesis: Seventeen studies (which enrolled 1836 pa-tients) met the inclusion criteria. Patients received adequateenteral nutrition in three of the studies. Overall, stress ulcerprophylaxis with a histamine-2 receptor blocker reduced the riskof gastrointestinal bleeding (odds ratio 0.47; 95% confidenceinterval, 0.29–0.76; p < .002; I2 � 44%); however, the treatmenteffect was noted only in the subgroup of patients who did not

receive enteral nutrition. In those patients who were fed enterally,stress ulcer prophylaxis did not alter the risk of gastrointestinalbleeding (odds ratio 1.26; 95% confidence interval, 0.43–3.7).Overall histamine-2 receptor blockers did not increase the risk ofhospital-acquired pneumonia (odds ratio 1.53; 95% confidenceinterval, 0.89–2.61; p � .12; I2 � 41%); however, this complica-tion was increased in the subgroup of patients who were fedenterally (odds ratio 2.81; 95% confidence interval, 1.20–6.56;p � .02; I2 � 0%). Overall, stress ulcer prophylaxis had no effecton hospital mortality (odds ratio 1.03; 95% confidence interval,0.78–1.37; p � .82). The hospital mortality was, however, higherin those studies (n � 2) in which patients were fed enterally andreceived a histamine-2 receptor blocker (odds ratio 1.89; 95%confidence interval, 1.04–3.44; p � .04, I2 � 0%). Sensitivityanalysis and meta-regression demonstrated no relationship be-tween the treatment effect (risk of gastrointestinal bleeding) andthe classification used to define gastrointestinal bleeding; neitherthe Jadad quality score nor the year of the study was reported.

Conclusions: The results of this meta-analysis suggest that, inthose patients receiving enteral nutrition, stress ulcer prophylaxismay not be required and, indeed, such therapy may increase therisk of pneumonia and death. However, because no clinical studyhas prospectively tested the influence of enteral nutrition on therisk of stress ulcer prophylaxis, our findings should be consideredexploratory and interpreted with some caution. (Crit Care Med2010; 38:000–000)

KEY WORDS: stress ulcer prophylaxis; histamine receptorblocker; proton pump inhibitor; enteral nutrition; systematicreview; meta-analysis

1Crit Care Med 2010 Vol. 38, No. 11

form of SUP (8). This practice has nowextended outside the ICU with up to 52%of non-ICU patients receiving SUP (9, 10).Furthermore, although proton-pump in-hibitors (PPIs) have never been demon-strated to reduce the rate of bleedingfrom stress ulceration (as compared to aplacebo), these agents are commonly pre-scribed for the prevention of this condi-tion (11, 12). However, a recent meta-analysis did not find strong evidence that“PPIs were different from H2RBs in termsof stress-related GI bleeding prophylaxis,pneumonia, and mortality among pa-tients admitted to ICUs” (13).

SUP is not without risks. Acid suppres-sive therapy is associated with increasedcolonization of the upper gastrointestinaltract with potentially pathogenic organ-ism. This has been demonstrated to in-crease the risk of hospital-acquired pneu-monia (HAP) (10). Furthermore, gastricacid is an important defense against theacquisition of Clostridium difficilespores, and the use of acid suppressivetherapy has been linked to an increasedrisk of Clostridium difficile infection (14–16). Furthermore, these agents have im-portant interactions with other drugs aswell as having agent-specific side effects.

Most of the clinical trials on which thecurrent recommendations are based wereperformed in the 1980s and early 1990swhen it was common to keep ICU pa-tients nil-per-os and when the early ini-tiation of enteral nutrition was uncom-mon. It has been suggested that patientsreceiving enteral alimentation have alower incidence of stress ulceration thanunfed patients (17). In animal models,enteral alimentation has been demon-strated to protect the gastric mucosafrom stress-related gastric mucosal dam-age (18, 19). We postulated that SUP mayhave no added benefits in ICU patientsreceiving enteral nutrition, and indeed,such therapy may have an unfavorablerisk-benefit profile. We, therefore, per-formed a meta-analysis to assess the ef-fect of SUP on the risk of GI bleeding,grouping the studies by those that usedor did not use enteral nutrition. The lat-ter group included patients receiving par-enteral nutrition and those who were ini-tially nil-per-os and then transitioned toan oral diet as well as studies that usedinadequate enteral nutrition. Our sec-ondary aims were to determine the effectof SUP on the incidence of HAP and mor-tality. As almost of all the placebo-controlled clinical trials reported to datehave investigated the role of H2RBs in

preventing stress ulceration, with onlyone study investigating a PPI, (12) weconfined our meta-analysis to random-ized controlled trials (RCTs) that com-pared a H2RB to a placebo (or control).

METHODS

Identification of Trials

Our aim was to identify all relevant ran-domized controlled trials that evaluated therole of H2RBs in the prevention of stress ul-ceration. We restricted this analysis to humanadults; there was no restriction, however, as tothe type of patient or the setting where thestudy was performed or the language of thepublication. We used a multi-method ap-proach to identify relevant studies for thisreview. All authors independently searchedthe National Library of Medicine’s MEDLINEdatabase for relevant studies in any languagepublished from 1966 to September 2009 byusing the following Medical Subject Headings(MeSH) and keywords: stress ulcer prophy-laxis, cimetidine, ranitidine or famotidine andcritical care or intensive care, and randomizedcontrolled trial (publication type). In addition,we searched Embase and the Cochrane Data-base of Systematic Reviews. Bibliographies ofall selected articles and review articles thatincluded information on SUP were reviewedfor other relevant articles. In addition, wesearched the “gray literature” to avoid anyreporting bias; this included studies publishedin abstract only. This search strategy was doneiteratively, until no new potential citationswere found on review of the reference lists ofretrieved articles. We performed this meta-analysis according to the guidelines proposedby the QUOROM group (20).

Study Selection and DataExtraction

Only randomized, placebo-controlled stud-ies that evaluated the role of a H2RB in pre-venting bleeding from stress ulceration wereincluded in the meta-analysis. In those studieswith a third or fourth treatment arm (sucral-fate, PPI, or pirenzepine), only patients whoreceived the H2RB and a placebo were in-cluded in the analysis. The primary end-pointwas the incidence of clinically significant GIbleeding (as defined in each study). If thestudy did not report the incidence of clinicallysignificant bleeding, the incidence of bleedingas determined by endoscopy was used. Sec-ondary end-points included the incidence ofHAP (as defined in each study) and hospitalmortality. All authors independently ab-stracted data from all studies by using a stan-dardized form. Data were abstracted on studydesign, study size, study setting, patient pop-

ulation, the H2RB used and its dosage, inci-dence of clinically significant GI bleeding,HAP, hospital mortality, and whether the pa-tients received enteral nutrition. For the pur-poses of our meta-analysis, we included stud-ies in the enteral nutrition subgroup if theauthors specifically reported that �50% ofenrolled patients received enteral nutrition.Patients were considered to be nil-per-os ifthey had a nasogastric tube that was placed togravity drainage and/or the total gastric out-puts were being measured. Attempts weremade to contact the primary authors for miss-ing data elements.

All reviewers independently assessed allo-cation concealment and the likelihood of biasto determine the methodologic quality of theincluded trials. The allocation concealmentwas ranked as adequate, uncertain, or inade-quate, and the likelihood of bias was scored onthe Jadad 5-point scale, which contains twoquestions each on randomization and maskingand one question on the reporting of dropoutsand withdrawals (21). Any disagreement be-tween reviewers was resolved by consensus.

Data Analysis

Statistical analysis was performed by usingReview Manager 5.023 (Cochrane Collabora-tion, Oxford, U.K.) and Comprehensive Meta-analysis 2.0 (Biostat, Englewood, NJ). We as-sessed heterogeneity among studies by usingthe Cochran Q statistic, with p � .10 indicat-ing significant heterogeneity, (22) and I2 withsuggested thresholds for low (25–49%), mod-erate (50–74%), and high (�75%) values (23,24). We used a random effects model if the Qstatistic was significant; otherwise we used afixed effects model. Summary effects estimatesare presented as odds ratio (OR) with 95%confidence interval (CI); we considered p �.05 (two-sided) as significant. Subgroup anal-ysis was performed by grouping the studies byenteral nutrition or no enteral nutrition.Summary estimates are presented as OR with95% CI. The relationship among the treat-ment effect (reduced risk of GI bleeding) andthe date of study publication and the Jadadscore was assessed by meta-regression. Thepresence of publication bias was assessed vi-sually with a funnel plot. We performed asensitivity analysis to determine the effect ofthe definition of GI bleeding and the Jadadscore on the treatment effect (risk of GI bleed-ing). In addition, we performed meta-regres-sion to determine the relationship betweenthe treatment effect and the year the study wasreported.

RESULTS

The initial search strategy generated56 citations; of these 21 were excludedbecause they did not include a placebo

2 Crit Care Med 2010 Vol. 38, No. 11

arm (compared two or more SUP agents),seven were excluded because they evalu-ated the pharmacokinetics/pharmacody-namics of H2RBs/PPIs or they did notreport the end-point of interest (20 stud-ies). An additional nine studies were iden-tified from the bibliographies of the se-lected articles and review articles. Nostudies were identified that were pub-lished in abstract only. The 17 studiesincluded in the meta-analysis enrolled atotal of 1836 patients between the years1980 and 2004 (12, 25–40). These studiesare summarized in Table 1, and themethodologic quality of the studies isprovided in Table 2. In three studies, pa-tients received adequate enteral nutrition(29, 34, 37).

The incidence of clinically significantbleeding was reported in 16 studies; inone of these studies the source of bleed-ing was confirmed by endoscopy (37).The study by Peura et al (27) evaluatedthe incidence of endoscopic signs ofbleeding. Overall, SUP with a H2RB re-duced the risk of GI bleeding (OR 0.47;95% CI, 0.29–0.76; p � .002; I2 � 44%);however, the treatment benefit was notedonly in the subgroup of patients who didnot receive enteral nutrition (Fig. 1). Inthose studies in which patients were fedenterally, SUP did not alter the risk of GIbleeding (OR 1.26; 95% CI, 0.43–3.7).The incidence of HAP was reported innine studies. Overall H2RBs did not in-crease the risk of HAP (OR 1.53; 95% CI,

0.89–2.61; p � .12; I2 � 41%), however,this complication was increased in thesubgroup of patients who were fed enter-ally (OR 2.81; 95% CI, 1.20–6.56; p �.02; I2 � 0%; Fig. 2). Mortality was re-ported in 14 studies. Overall, SUP had noeffect on hospital mortality (OR 1.03;95% CI, 0.78–1.37; p � .82; Fig. 3). Thehospital mortality was, however, higherin those studies (n � 2) in which patientswere fed enterally and received a H2RB(OR 1.89; 95% CI, 1.04–3.44; p � .04,I2 � 0%). Visual inspection of the funnelplots failed to reveal a publication bias(Fig. 4). Meta-regression demonstratedno relationship between the treatmenteffect (risk of GI bleeding) and the yearthe study was reported (Fig. 5). Similarly,there was no relationship between thetreatment effect and the classificationused to define neither GI bleeding nor theJadad score (data not shown).

DISCUSSION

The results of this meta-analysis sug-gest that in patients who are fed enter-ally, SUP does not reduce the risk ofbleeding from stress ulceration. Further-more, in patients receiving SUP, our datasuggest that enteral feeding may increasethe risk of pneumonia and death. Theresults of our meta-analysis are sup-ported by recent observational studiesthat have demonstrated that the risk ofclinically significant bleeding from stressulceration is very low (approximately 1%)in ICU patients and that SUP does notalter this risk. Faisy et al (41) compared

Table 1. Randomized controlled trials comparing stress ulcer prophylaxis with placebo: Characteristics of studies included in meta-analysis

Author (reference) Year n Agent Dose Type ICU Enteral Feeds*

Halloran (25) 1980 50 Cimetidine 300 mg/every 4 hrs Head injury NZinner (26) 1981 200 Cimetidine 300 mg/every 6 hrs SICU NPeura (27) 1985 39 Cimetidine 300 mg/every 6 hrs MICU NCheadle (28)† 1985 195 Cimetidine 200 mg/every 6 hrs Postabdominal surgery Nvan den Berg (29) 1985 28 Cimetidine 20 mg/kg/24 hr GICU YGroll (30) 1986 221 Cimetidine 300 mg q 6 GICU NReusser (31) 1990 40 Ranitidine‡ 50 mg/every 6 hrs Neurosurgery NKarlstadt (32)‡ 1990 87 Cimetidine infusion at 50 mg/hr GICU¶ NRuiz-Santana (33) 1991 49 Ranitidine 50 mg/every 6 hrs GICU � TPN NApte (34) 1992 34 Ranitidine 50 mg/every 6 hrs Tetanus and tracheostomy YMetz (35) 1993 167 Ranitidine infusion at 6.25 mg/hr head injury NMartin (36) 1993 131 Cimetidine infusion at 50–100 mg/hr GICU¶ NBen-menachem (37)§ 1994 200 Cimetidine Infusion at 900 mg/day MICU YBurgess (38) 1995 4 Ranitidine infusion at 6.25 mg/hr Head injury NChan (39) 1995 101 Ranitidine 50 mg/every 6 hrs Neurosurgical NHanisch (40)†† 1998 114 Ranitidine 50 mg/every 8 hrs SICU¶ NKantorova (12)** 2004 146 Famotidine 40 mg/every 12 hrs SICU¶ N

SICU, surgical intensive care unit; MICU, medical intensive care unit; GICU, mixed intensive care unit; TPN, ●●.*�50% of patients received enteral nutrition; †also randomized to ●● NG or no NG ●●; ‡antacid; ¶high risk patients; §also included a sucralfate arm;

**also included a PPI and sucralfate arm; ††also included a pirenzepine arm.

Table 2. Quality assessment of studies included in meta-analysis

AuthorClinicianBlinding

Intentionto Treat

AllocationConcealment

Jadad Score(0–5)

Halloran (25) Yes Yes Adequate 5Zinner (26) No Yes Adequate 3Peura (27) Yes Yes Uncertain 4Cheadle (28) Yes Yes Adequate 5van den Berg (29) Yes Yes Uncertain 4Groll (30) Yes Yes Uncertain 4Reusser (31) No Yes Uncertain 3Karlstadt (32) Yes Yes Uncertain 3Ruiz-Santana (33) No Yes Adequate 3Apte (34) No Yes Uncertain 2Metz (35) Yes Yes Adequate 5Martin (36) Yes Yes Adequate 5Ben-Menachem (37) No Yes Adequate 3Burgess (38) Yes Yes Adequate 5Chan (39) Yes Yes Uncertain 4Hanisch (40) Yes Yes Adequate 5Kantorova (12) Yes Yes Adequate 5

3Crit Care Med 2010 Vol. 38, No. 11

the rate of clinically significant GI bleed-ing during two sequential time periods.During the first phase all patients (n �736) received SUP, whereas SUP waswithheld during the second period (n �737). Although the patients during thesecond phase of the study were sicker(higher SAPS II score), the rate of overt(1.9% vs. 1.6%) and clinically significantbleeding (1.4% vs. 1.1%) as well as theuse of blood products was similar be-tween the two time periods. During bothtime periods, patients received early en-teral feeding (within 48 hrs of ICU admis-

sion). Zandstra and Soutenbeek (42) re-ported that one of 183 patients (0.6%)receiving prolonged mechanical ventila-tion without any SUP developed stressulcer-related bleeding. Erstad et al (43)conducted a prospective study on 543 pa-tients and reported clinically significantgastrointestinal bleeding rates were sim-ilar for those patients with and withoutappropriate SUP. Kantorova et al (12)performed a randomized, placebo-con-trolled study in critically ill patients athigh risk for stress-related GI bleeding(mechanical ventilation �48 hrs and co-

agulopathy) in which they comparedthree SUP regimens (omeprazole, famo-tidine, and sucralfate) with placebo. Theoverall bleeding rate was 1% with no sig-nificant difference between treatmentgroups (bleeding rate was 1% in the pla-cebo group). More aggressive resuscita-tion and the early initiation of enteralnutrition were postulated to account forthe low incidence of bleeding from stressulceration in these studies (41).

Enteral nutrients buffer acid and mayact as a direct source of mucosal energy,induce the secretion of cytoprotectiveprostaglandins and mucu,s and improvemucosal blood flow (18, 19). Mucosal im-munity may be enhanced via stimulationof the gut-associated lymphoid tissue. Inaddition, it has been postulated thatstress triggers vagal stimulation of thestomach through central nervous systempathways; these pathways may be bluntedby enteral nutrition (44, 45). Bonten et al(46) demonstrated that continuous en-teral nutrition was more likely to raisegastric pH to �3.5 than patients receiv-ing H2RBs or PPIs. Rodent restraintmodels have demonstrated that enteralnutrition provides better protectionagainst stress ulceration than the intra-gastric administration of an antacid orsucralfate or intravenous administrationof cimetidine (18, 47–49). In a retrospec-tive analysis of prospectively collecteddata, Raff et al (50) demonstrated thatearly (within 12 hrs posttrauma) enteralnutrition was more effective in prevent-ing over upper GI bleeding than cimeti-dine and antacids (3.3% vs. 8.3%, p �.05) in a cohort of 526 severely burnedpatients. Patients in the cimetidine groupreceived antacids if the intragastric pHdropped below 3.5 with all the patientsreceiving parenteral nutrition. Similarly,in a cohort of 146 severely burned pa-tients, Choctaw et al (51) reported majorupper GI bleeding in 30% of patients whoreceived “the usual diet” compared with3% (p � .05) in patients who received acontinuous infusion of an elemental diet.Pingleton et al (17) reported similar find-ings in 43 ventilated patients; 14 of 20patients receiving antacids and seven ofnine patients receiving cimetidine hadevidence of upper GI bleeding; however,none of the patients (n � 14) receivingenteral alimentation had evidence ofbleeding.

An intriguing finding of this study wasthe observation that the incidence of HAPwas increased in the subgroup of patientswho received both a H2RB and enteral

Figure 1. Effect of SUP on the risk of GI bleeding. Studies are grouped by the use or nonuse of enteralnutrition. Weight is the relative contribution of each study to the overall treatment effect (OR and 95%confidence interval) on a log scale assuming a random effects model. SUP, stress ulcer prophylaxis; GI,gastrointestinal; OR, odds ratio; H2RB, histamine-2 receptor blocker; CI, confidence interval.

Figure 2. Effect of SUP on the risk of HAP. Studies are grouped by the use/nonuse of enteral nutrition.Weight is the relative contribution of each study to the overall treatment effect (OR and 95%confidence interval) on a log scale assuming a random effects model. SUP, stress ulcer prophylaxis;HAP, hospital-acquired pneumonia; OR, odds ratio, H2RB, histamine-2 receptor blocker; CI, confi-dence interval.

4 Crit Care Med 2010 Vol. 38, No. 11

nutrition. Patients who received a H2RBand were not fed enterally did not have anincreased risk of HAP. Gastric microbialgrowth is pH dependent (52). Normally,the fasting stomach maintains sterility bymaintaining an acid pH, and an increasein pH may allow the stomach to becomecolonized. Both H2RBs and enteral feed-ing increase gastric pH and gastric colo-nization (53). Bonten et al (46) demon-strated that the combination of acidsuppressive therapy and enteral feedingresulted in a significantly higher pH thaneither intervention alone, and this wasassociated with an increased rate of gas-tric colonization. This observation likelyexplains the increased risk of HAP in thepatients receiving a H2RB and enteralfeeds. The increased risk of HAP may ex-plain the increased mortality in thisgroup of patients. In a large prospectivepharmacoepidemiologic cohort study in-volving hospitalized non-ICU patients,Herzig et al (10) demonstrated that acid-suppressive medication was associatedwith a 30% increased odds ratio of devel-oping HAP. The results of our meta-analysis suggest that enteral nutritionprovides adequate protection againststress ulceration and that the addition ofacid suppressive therapy serves only toincrease the risk of HAP and the risk ofdying.

It should be noted that in many of thestudies in which patients did not receiveenteral tube feeds, “oral feedings” wererestarted as soon as the “patient’s condi-tion permitted” or on day 2–3 after sur-gery. Delayed oral feeding, therefore,would appear not to protect againststress ulceration. This is supported bythe study of Choctaw et al (51) whereinthe risk of serious bleeding from stressulceration was significantly lower inthe patients who received a continuousintragastric infusion of an elementaldiet as compared with the patients whoreceived the usual diet.

The strength of our review includesthe rigorous attempt to identify all rele-vant studies and the inclusion of onlyrandomized controlled studies. However,our meta-analysis has many of the limi-tations that apply to meta-analyses ingeneral, including lack of homogeneity ofpatient populations and the use of some-what different criteria for the major end-points (for the diagnosis of clinically sig-nificant bleeding and HAP) (54 –56).Although we could not use standardizedend-points for both bleeding and HAP (asthe required data were not collected and

Figure 3. Effect of SUP on the mortality. Studies are grouped by the use/nonuse of enteral nutrition.Weight is the relative contribution of each study to the overall treatment effect (OR and 95%confidence interval) on a log scale assuming a fixed effects model. SUP, stress ulcer prophylaxis; OR,odds ratio, H2RB, histamine-2 receptor blocker; CI, confidence interval.

Figure 4. Funnel plot for potential publication bias. SE, standard error; OR, odds ratio.

Regression of Date on Log odds ratio3.00

io

2.50

2.00

1.50

og o

dds

rat

1.00

0.50

0.00

Lo -0.50

-1.00

-1.50

Date

-2.00

1977 1980 1983 1986 1989 1992 1995 1998 2001 2004

Figure 5. Meta-regression of treatment effect (reduced risk of gastrointestinal bleeding) and date ofstudy publication.

5Crit Care Med 2010 Vol. 38, No. 11

reported), the same diagnostic criteriawere used in both the treatment and con-trol groups in each study, making it likelythat any differences were real. Further-more a sensitivity analysis did not dem-onstrate a relationship between the clas-sification used to define GI bleeding andthe treatment effect. A major limitationof our meta-analysis is that adequate en-teral nutrition was provided in only threeof the studies. Furthermore, in the ab-sence of an individual patient data anal-ysis or a prespecified standardized proto-col for administering enteral nutritionthat would not introduce a systematicbias, it is likely that the patients whoreceived enteral nutrition varied in theiracuity of illness compared with the non-enteral nutrition group. This may explainthe finding of increased mortality, andeven perhaps HAP, in that subgroup. Thestudies included in our meta-analysisspanned a long period of time (25 yrs),during which both the standard of care aswell the conduct of clinical trials haschanged enormously. However, meta-regression failed to demonstrate a rela-tionship between the year of publicationof the study and the treatment effect.Despite these limitations, the results ofour meta-analysis are supported by exper-imental animal models and more recentobservational studies.

In conclusion, the results of this meta-analysis suggest that in those patientswho received enteral tube feeds, SUP maynot be required. Indeed, in these patients,SUP is likely to increase the risk of com-plications. Furthermore, the results ofour study suggest that the use of SUP inthe ICU should no longer be used as a“quality” indicator. However, becauseno clinical study has prospectivelytested the influence of enteral nutritionon the risk of SUP, our findings shouldbe considered exploratory and inter-preted with some caution. Additionalstudies are required to confirm thefindings of our meta-analysis.

REFERENCES

1. Skillman JJ, Bushnell LS, Goldman H, et al:Respiratory failure, hypotension, sepsis andjaundice. A clinical syndrome associated withlethal hemorrhage from acute stress ulcer-ation of the stomach. Am J Surg 1969; 117:523–530

2. Hastings PR, Skillman JJ, Bushnell LS, et al:Antacid titration in the prevention of acutegastrointestinal bleeding: A controlled, ran-domized trial in 100 critically ill patients.N Engl J Med 1978; 298:1041–1045

3. Cook DJ, Witt LG, Cook RJ, et al: Stress ulcerprophylaxis in the critically ill: A meta-analysis. Am J Med 1991; 91:519–527

4. Cook DJ, Reeve BK, Guyatt GH, et al: Stressulcer prophylaxis in critically ill patients. Re-solving discordant meta-analyses. JAMA1996; 275:308–314

5. Dellinger RP, Levy MM, Carlet JM, et al:Surviving sepsis Campaign: Internationalguidelines for management of severe sepsisand septic shock: 2008. Crit Care Med 2008;36:296–327

6. ASHP therapeutic guidelines on stress ulcerprophylaxis. ASHP Commission on Thera-peutics and approved by the ASHP Board ofDirectors on November 14, 1998. Am JHealth System Pharm 1999; 56:347–379

7. Centers for Medicare and Medicaid ServicesQuality Measures Compendium, Division ofQuality, Evaluation and Health Outcomes,Vol. 2, December 2007. http://www.cms.hhs.gov/MedicaidSCHIPQualPrac/Downloads/pmfinalaugust06.pdf 2. Accessed March 13,2009

8. Daley RJ, Rebuck JA, Welage LS, et al: Pre-vention of stress ulceration: Current trendsin critical care. Crit Care Med 2004; 32:2008–2013

9. Janicki T, Stewart S: Stress-ulcer prophylaxisfor general medical patients: A review of theevidence. J Hosp Med 2007; 2:86–92

10. Herzig SJ, Howell MD, Ngo LH, et al: Acid-suppressive medication use and the risk forhospital-acquired pneumonia. JAMA 2009;301:2120–2128

11. Jung R, MacLaren R: Proton-pump inhibitorsfor stress ulcer prophylaxis in critically illpatients. Ann Pharmacother 2002; 36:1929–1937

12. Kantorova I, Svoboda P, Scheer P, et al:Stress ulcer prophylaxis in critically ill pa-tients: A randomized controlled trial.Hepato-Gastroenterology 2004; 51:757–761

13. Lin PC, Chang CH, Hsu PI, et al: The efficacyand safety of proton pump inhibitors vs his-tamine-2 receptor antagonists for stress ul-cer bleeding prophylaxis among critical carepatients: A meta-analysis. Crit Care Med2010; 38:1197–1205

14. Cunningham R, Dale B, Undy B, et al: Protonpump inhibitors as a risk factor for Clostrid-ium difficile diarrhoea. J Hosp Infect 2003;54:243–245

15. Dial S, Alrasadi K, Manoukian C, et al: Risk ofClostridium difficile diarrhea among hospitalinpatients prescribed proton pump inhibi-tors: Cohort and case-control studies. CanMed Assoc J 2004; 171:33–38

16. Louie TJ, Meddings J: Clostridium difficileinfection in hospitals: Risk factors and re-sponses. Can Med Assoc J 2004; 171:45–46

17. Pingleton SK, Hadzima SK: Enteral alimen-tation and gastrointestinal bleeding in me-chanically ventilated patients. Crit Care Med1983; 11:13–16

18. Ephgrave KS, Kleiman-Wexler RL, Adair CG:Enteral nutrients prevent stress ulceration

and increase intragastric volume. Crit CareMed 1990; 18:621–624

19. Shorr LD, Sirinek KR, Page CP, et al: Therole of glucose in preventing stress gastricmucosal injury. J Surg Res 1984; 36:384–388

20. Moher D, Cook DJ, Eastwood S, et al: Im-proving the quality of reports of meta-analyses of randomised controlled trials: TheQUOROM statement Quality of Reporting ofMeta-analyses. Lancet 1999; 354:1896–1900

21. Jadad AR, Moore RA, Carroll D, et al: Assess-ing the quality of reports of randomized clin-ical trials: Is blinding necessary? ControlledClinical Trials 1996; 17:1–12

22. Berlin JA, Laird NM, Sacks HS, et al: A com-parison of statistical methods for combiningevent rates from clinical trials. Stat Med1989; 8:141–151

23. Higgins JP, Thompson SG: Quantifying het-erogeneity in a meta-analysis. Stat Med 2002;21:1539–1558

24. Higgins JP, Thompson SG, Deeks JJ, et al:Measuring inconsistency in meta-analyses.BMJ 2003; 327:557–560

25. Halloran LG, Zfass AM, Gayle WE, et al: Pre-vention of acute gastrointestinal complica-tions after severe head injury: A controlledtrial of cimetidine prophylaxis. Am J Surg1980; 139:44–48

26. Zinner MJ, Zuidema GD, Smith P, et al: Theprevention of upper gastrointestinal tractbleeding in patients in an intensive care unit.Surg Gynecol Obstet 1981; 153:214–220

27. Peura DA, Johnson LF: Cimetidine for pre-vention and treatment of gastroduodenalmucosal lesions in patients in an intensivecare unit. Ann Intern Med 1985; 103:173–177

28. Cheadle WG, Vitale GC, Mackie CR, et al:Prophylactic postoperative nasogastric de-compression. A prospective study of its re-quirement and the influence of cimetidine in200 patients. Ann Surg 1985; 202:361–366

29. van den Berg B, van Blankenstein M: Preven-tion of stress-induced upper gastrointestinalbleeding by cimetidine in patients on assistedventilation. Digestion 1985; 31:1–8

30. Groll A, Simon JB, Wigle RD, et al: Cimeti-dine prophylaxis for gastrointestinal bleedingin an intensive care unit. Gut 1986; 27:135–140

31. Reusser P, Gyr K, Scheidegger D, et al: Pro-spective endoscopic study of stress erosionsand ulcers in critically ill neurosurgical pa-tients: Current incidence and effect of acid-reducing prophylaxis. Crit Care Med 1990;18:270–274

32. Karlstadt RG, Iberti TJ, Silverstein J, et al:Comparison of cimetidine and placebo forthe prophylaxis of upper gastrointestinalbleeding due to stress-related gastric muco-sal damage in the intensive care unit. J In-tensive Care Med 1990; 5:26–32

33. Ruiz-Santana S, Ortiz E, Gonzalez B, et al:Stress-induced gastroduodenal lesions andtotal parenteral nutrition in critically ill pa-tients: Frequency, complications, and thevalue of prophylactic treatment. A prospec-

6 Crit Care Med 2010 Vol. 38, No. 11

tive, randomized study. Crit Care Med 1991;19:887–891

34. Apte NM, Karnad DR, Medhekar TP, et al:Gastric colonization and pneumonia in intu-bated critically ill patients receiving stressulcer prophylaxis: A randomized, controlledtrial. Crit Care Med 1992; 20:590–593

35. Metz CA, Livingston DH, Smith JS, et al:Impact of multiple risk factors and ranitidineprophylaxis on the development of stress-related upper gastrointestinal bleeding: Aprospective, multicenter, double-blind, ran-domized trial. The Ranitidine Head InjuryStudy Group. Crit Care Med 1993; 21:1844–1849

36. Martin LF, Booth FVM, Karlstadt RG, et al:Continuous intravenous cimetidine de-creases stress-related upper gastrointestinalhemorrhage without promoting pneumonia.Crit Care Med 1993; 21:19–30

37. Ben-Menachem T, Fogel R, Patel RV, et al:Prophylaxis for stress-related gastric hemor-rhage in the medical intensive care unit. Arandomized, controlled, single-blind study.Ann Intern Med 1994; 121:568–575

38. Burgess P, Larson GM, Davidson P, et al:Effect of ranitidine on intragastric pH andstress-related upper gastrointestinal bleedingin patients with severe head injury. Dig DisSci 1995; 40:645–650

39. Chan KH, Lai EC, Tuen H, et al: Prospectivedouble-blind placebo-controlled randomizedtrial on the use of ranitidine in preventingpostoperative gastroduodenal complicationsin high risk neurosurgical patients. J Neuro-surg 1995; 82:413–417

40. Hanisch EW, Encke A, Naujoks F, et al: Arandomized, double-blind trial for stress ul-cer prophylaxis shows no evidence of in-creased pneumonia. Am J Surg 1998; 176:453–457

41. Faisy C, Guerot E, Diehl JL, et al: Clinicallysignificant gastrointestinal bleeding in criti-cally ill patients with and without stress-ulcer prophylaxis. Intensive Care Med 2003;29:1306–1313

42. Zandstra DF, Stoutenbeek CP: The virtualabsence of stress-ulceration related bleedingin ICU patients receiving prolonged mechan-ical ventilation without any prophylaxis. Aprospective cohort study. Intensive Care Med1994; 20:335–340

43. Erstad BL, Camamo JM, Miller MJ, et al: Im-pacting cost and appropriateness of stress ulcerprophylaxis at a university medical center. CritCare Med 1997; 25:1678–1684

44. Hernandez DE: Neuroendocrine mecha-nisms of stress ulceration: Focus on thyro-tropin-releasing hormone (TRH). Life Sci1986; 39:279–296

45. Ephgrave KS, Scott DL, Ong A, et al: Aregastric, jejunal, or both forms of enteral feed-ing gastroprotective during stress? J SurgRes 2000; 88:1–7

46. Bonten MJ, Gaillard CA, van Tiel FH, et al:Continuous enteral feeding counteracts pre-ventive measures for gastric colonization inintensive care unit patients. Crit Care Med1994; 22:939–944

47. Mabogunje OA, Andrassy RJ, Isaacs H Jr, etal: The role of a defined formula diet in theprevention of stress-induced gastric mucosal

injury in the rat. J Pediatr Surg 1981; 16:1036–1039

48. Harju E, Sajanti J: The protective effect ofnutrients against stress-induced gastric ul-cers in the rat. Surg Gynecol Obstet 1987;165:530–534

49. Lally KP, Andrassy RJ, Foster JE, et al: Eval-uation of various nutritional supplements inthe prevention of stress-induced gastric ul-cers in the rat. Surg Gynecol Obstet 1984;158:124–128

50. Raff T, Germann G, Hartmann B: The valueof early enteral nutrition in the prophylaxisof stress ulceration in the severely burnedpatient. Burns 1997; 23:313–318

51. Choctaw WT, Fujita C, Zawacki BE: Preven-tion of upper gastrointestinal bleeding inburn patients: A role for ‘elemental’ diet.Arch Surg 1980; 115:1073–1076

52. Gray J, Shiner M: Influence of gastric pH ongastric and jejunal flora. Gut 1967;8:574–781

53. Spilker CA, Hinthorn DR, Pingleton SK: In-termittent enteral feeding in mechanicallyventilated patients. The effect on gastric pHand gastric cultures. Chest 1996; 110:243–248

54. Arriagada R, Pignon JP: Is meta-analysis ametaphysical or a scientific method? Chest2000; 118:832–834

55. Bigby M, Williams H: Appraising systematicreviews and meta-analyses. Arch Dermatol2003; 139:795–798

56. Egger M, Smith GD: Meta-Analysis. Poten-tials and promise. Brit Med J 1997; 315:1371–1374

7Crit Care Med 2010 Vol. 38, No. 11