Aliment Pharmacol Ther 2010; 32: 831–839
Background Aspirin is widely used to modify the risk of recurrent vascular events. It is, however, associated with increased upper gastrointestinal bleeding risk. The influence of Helicobacter pylori on this risk is uncertain.
Aim To determine the influence of H. pylori on upper gastrointestinal bleeding risk in patients taking aspirin.
Methods MEDLINE and EMBASE databases were searched. All studies providing data regarding H. pylori infection in adults taking aspirin and presenting with upper gastrointestinal bleeding were included.
Results A total of 13 studies that included 1 case–control, 10 cohort studies and 2 randomized-controlled trials (RCTs) were analysed. The case–control study (n = 245) determined H. pylori to be a significant independent risk factor for upper gastrointestinal bleeding. The cohort studies were heterogeneous, varying in inclusion criteria, doses and duration of aspirin used, mode of H. pylori testing and causative GI pathology considered. Comprising 5465 patients, H. pylori infection was tested for in 163 (0.03%) aspirin users with upper gastrointestinal bleeding. The RCTs yielded no significant results.
Conclusions The current data are not sufficient to allow meta-analyses. The widely held belief that H. pylori is a risk factor for upper gastrointestinal bleeding in regular aspirin users is not supported by the very limited evidence available.
Aspirin is used for a wide range of medical conditions including stroke, myocardial infarction and peripheral vascular disease. When used for the secondary prevention of occlusive vascular events, it reduces the relative risk by one quarter in high-risk patients.1 Aspirin, however, is not without risk. It is associated with increased intra- and extra-cranial bleeding, with gastrointestinal (GI) haemorrhage the most common life-threatening extracranial site for bleeding. Long-term maintenance doses of aspirin increases this risk by over two-fold.2, 3
There are other factors contributing to the risk of gastrointestinal haemorrhage in patients taking aspirin. Concurrent medications such as non-aspirin nonsteroidal anti-inflammatory drugs (NANSAIDs), corticosteroids or anticoagulants increase bleeding risk.4 In addition, advanced age is a major risk factor for complicated peptic ulcer disease. Ageing is associated with diminished epithelial cell turnover and reduced capacity to repair the gastric mucosa.5 Prostaglandin levels in the gastric mucosa also decrease.6 As a consequence, the integrity of the gastric mucosal surface becomes impaired and progressively susceptible to damage by factors that can overwhelm the protective barriers of the stomach. Older people are therefore at higher risk of upper gastrointestinal bleeding (UGIB).
Infection with Helicobacter pylori is an independent risk factor for peptic ulcer disease. Although it is commonly asserted that H. pylori infection will increase the risk of aspirin-associated GI side effects, the current evidence for this has not been formally reviewed. Furthermore, the impact and influence of H. pylori infection on the incidence of UGIB in patients taking aspirin remain uncertain.
The mechanisms through which aspirin and H. pylori exert their gastric damage differ. Aspirin causes ulcers through both local toxic effects exposing luminal acid to the epithelial cells and, following absorption, the systemic depletion of prostaglandins, which are key to promoting the defence mechanisms of the stomach. In contrast, H. pylori survives in grooves between epithelial cells under the protective gastric mucosal layer of the stomach and causes local damage by inducing inflammatory infiltration. Helicobacter pylori may promote gastric mucosal prostaglandin secretion by up to 50% to maintain its preferred environmental conditions.7–9 Given these differences in ulcer genesis and antagonistic effects on prostaglandin synthesis, H. pylori and aspirin may not have a synergistic effect on ulcer formation.
Studies examining the effect of H. pylori infection on UGIB risk in patients taking NANSAIDs yield conflicting results. Even if H. pylori is contributory to the development of peptic ulcer disease in users of NANSAIDs, it has been suggested that it may be protective against complications of peptic ulcer disease.10
Through its preference for inhibiting the cyclo-oxygenase (COX)-1 enzyme, aspirin differs in its actions from other NANSAIDs. Thus, results from NANSAID studies cannot be reliably extrapolated to guide aspirin prescribing. Hence, it would be valuable to ascertain the effect of H. pylori in this defined patient group. Identifying whether H. pylori is a contributory risk factor to UGIB in patients taking aspirin is therefore key to potentially minimizing side-effects and improving the delivery of aspirin and possibly other antiplatelet therapies in future.
A broad and inclusive literature search was undertaken for all antiplatelet therapies. However, given that only one study described antiplatelet therapy use other than aspirin,11 this systematic review focuses on determining the influence of H. pylori on the UGIB risk in patients taking aspirin.
MEDLINE and EMBASE databases were searched by exploding the MESH headings ‘platelet aggregation inhibitors’, ‘aspirin’ and ‘dipyridamole’ and combining with OR. The search was performed on the 29th January 2010 and included all studies in the MEDLINE and EMBASE databases at that time. Studies with ‘clopidogrel’ were identified using this as the keyword in the title or abstract and merged with the initial search term, again using OR. For bleeding, the MESH headings ‘haemorrhage’ and ‘gastrointestinal haemorrhage’ were exploded and combined with OR. The results for bleeding and antiplatelet use were combined with AND and further combined with the results of an exploded ‘Helicobacter pylori’ search. The results from the two databases were merged and the duplicates filtered, leaving 414 studies for review.
At this point, EF and CG independently screened the titles and abstracts for eligibility using a structured checklist with audit trail for included/excluded articles. Articles which published data referring to the routine clinical use of aspirin, the occurrence of upper GI bleeding and the testing of H. pylori in adult humans were sought for inclusion in the review. Articles that were not clinical trials and those not written in English were excluded. Full articles were retrieved and reviewed for all titles where abstracts were not available and for all abstracts that potentially appeared to fulfil the inclusion criteria.
A total of 37 studies were examined in detail. Of these, 22 were excluded as no direct data regarding the prevalence of H. pylori in patients using solely aspirin (with no other NANSAIDs) and the incidence of bleeding could be extracted (Figure 1). Two case–control studies were found to describe the same population of participants,12, 13 with the second examining genetic factors in a smaller subcohort and therefore only the first paper was considered. Two papers were also published on subjects recruited from overlapping cohorts.14, 15 The more recent paper reporting the overall cohort was included in the review. Core data from eligible studies were recorded on a standard proforma for subsequent analysis.
Our searches identified 13 studies from which the prevalence of H. pylori in users of aspirin presenting with UGIB could be examined. These 13 studies comprised 1 case–control study, 10 cohort studies and 2 randomized controlled trials (RCTs). In this review, we describe these studies; the dose, frequency and duration of aspirin used, antisecretory therapy use and determination of H. pylori status across the trials.
Consecutive, current users of low-dose aspirin (LDA) admitted with UGIB were prospectively recruited from two hospital sites in North East Spain over a 2-year period. Inclusion criteria were witnessed melaena/haematemesis with evidence of peptic ulcers on oesophagoduodenoscopy (OGD); and the use of LDA i.e. ≤ 325 mg/day, taken daily (≥5 days/week) for at least 15 days prior to admission. Patients who were taking both LDA and NANSAIDs were excluded. Controls were LDA users, with no UGIB, matched for similar age, gender and extent of aspirin use as the cases, and were recruited from out-patient cardiology and neurology clinics.
Patients were deemed H. pylori positive if either the serology or urea breath test was positive. Although both tests were performed in the majority of patients, 30.2% had only one test performed and this was predominantly serological testing (95% of single tests). The concordance of both tests, however, was high (Table 1). The prevalence of H. pylori infection, according to the authors’ criteria, was 89.8% in cases and 68.7%, (P = 0.0001) in controls. Despite matching for age and gender, controls used antisecretory treatment more frequently than cases. Past histories of UGIB, alcohol use and lung disease were more common in cases (Table 2). Following logistic regression, the authors determined that H. pylori infection was a risk factor for ulcer bleeding in low-dose aspirin users.
|Variable||Cases (n = 98) n (%)||Controls (n = 147) n (%)|
|Helicobacter pylori infection||88/98 (89.8)||102/147 (69.4)|
|Positive serology||83/96 (86.5)||93/145 (64.1)|
|Positive urea breath test||63/78 (80.7)||61/97 (62.9)|
|Cases (%)||Controls (%)|
|History of lung diseases||14.4||6.6|
|Ca channel blocker use||40.8||27.2|
Cohort studies11, 15–23
Ten cohort studies were examined in detail. Five recruited patients who had clinical or endoscopic evidence of UGIB. Four studies included patients who had been investigated for upper GI symptoms, but one of these reported only the results of patients found to have gastric ulcers on investigation. The tenth study was a prospective cohort study of patients taking LDA for ischaemic heart disease.
Helicobacter pylori prevalence. The five cohort studies11, 15–18 where UGIB was the inclusion criterion comprised 1353 patients in total (Table 3). The prevalence of H. pylori was not specified in one study and only determined in 190/238 (79.8%) patients in another. Therefore, in total, 975 patients presenting with UGIB in these studies were tested for H. pylori. Of these, 626 (64.2%) were positive. The number of patients using aspirin, as the only non-steroidal anti-inflammatory drug, was 230 (17.0%). However, in only 133 patients with UGIB who were taking aspirin could the prevalence of H. pylori be determined – 68.4% of aspirin users who presented with bleeding were H. pylori positive.
|Study n||Inclusion criteria||UGIB n (%)||Helicobacter pylori infection in UGIB cases n (%)||Aspirin use in UGIB cases n (%)||Helicobacter pylori infection in aspirin users with UGIB n (%)|
|Adamopoulos et al.16 330||Clinical UGIB||330 (100)||–||103 (31.2)||51/76 tested (67.1)|
|Nakashima et al.17 238||Bleeding PU on OGD||238 (100)||156/190 tested (82.1)||18 (7.6)||10/16 tested (62.5)|
|Nakayama et al.15 285||Bleeding PU on OGD||285 (100)||221 (77.5)||41 (14.4)||30 (73.2)|
|Tsesmeli et al.18 110||Clinical UGIB||110 (100)||32 (29.1)||23 (20.9)||–|
|Liu et al.11 390||Bleeding PU on OGD||390 (100)||217 (55.6)||45 (11.5)||–|
|Ng et al.19 217||GU on OGD||100 (46.1)||84 (84.0)||21 (21.0)||–|
|Akhtar et al.20 552||Upper GI symptoms||296 (53.6)||–||51 (17.2)||–|
|Pilotto et al.21 2251||>65 years, OGD*||225 (10.0)||–||8 (8.0)||–|
|Pilotto et al.22 189||>65 years, OGD*, LDA users||6 (3.2)||6 (100)||6 (100)||6 (100)|
|Serrano et al.23 903||LDA use||41 (4.5)||24/24 tested (100)||41 (4.5)||24/24 tested (100)|
A total of 3209 patients were recruited to cohort studies where the inclusion criterion was upper GI symptoms;19–22 627 had clinical or endoscopic evidence of UGIB. One hundred and six of these were tested for H. pylori, of whom 90 (84.9%) were positive. The prevalence of H. pylori in aspirin users, who presented with bleeding, however, could be determined only from one study. In this study, of the six patients with bleeding seen on endoscopy, all were taking LDA and all were H. pylori positive.
For the tenth study,23 where LDA use was the inclusion criterion, during the mean 45-month follow-up, 41 patients developed UGIB. Twenty four of these 41 (58.5%) were tested for H. pylori and all 24 were positive. However, 19 of the total 903 patients included in this study were taking NANSAIDs in addition to LDA, and information whether the 41 patients with bleeding complications included anyone on dual therapy was not specified.
Across all the cohort studies, only five provided sufficient data to determine the prevalence of H. pylori infection in aspirin users with UGIB. The studies have different inclusion criteria and given that only 163 aspirin users with UGIB were tested for H. pylori, a meta-analysis is not appropriate. Furthermore, it is impossible to determine bleeding risk in H. pylori positive patients vs. H. pylori negative patients in those taking regular aspirin.
Aspirin dose used. For studies where the inclusion criteria were clinical or endoscopic evidence of UGIB, two did not specify the dose of aspirin patients were taking (Table 4). The other three included patients taking LDA; the definition of LDA, however, was not specified in one study, was 100 mg/day in a second and less than or equal to 300 mg/day in the third. Where upper GI symptoms were the principal inclusion criterion, two studies did not specify the dose of aspirin used, one used 75–300 mg daily and the fourth used a range of doses, which was dichotomised at the 300 mg dose. In this study, 9/106 (8.5%) users of aspirin <300 mg/day had UGIB compared with 9/42 (21.4%) users >300 mg/day. For the LDA study, again a range of aspirin doses were used, with the majority either taking 125 mg or 200 mg daily. The dose of aspirin used by those patients who experienced complications with bleeding, however, was not specified.
|Study||Dose of aspirin||Frequency of aspirin||Duration of aspirin|
|Adamopoulos et al.16||–||71/71 occasional (prn in previous week) 24/24 acute (reg, <1 month) 8/8 chronic (reg, >1 month)||Acute vs. chronic|
|Nakashima et al.17||LDA (not defined)||–||–|
|Nakayama et al.15||≤300 mg/day||Daily||–|
|Tsesmeli et al.18||100 mg/day||Daily||–|
|Liu et al.11||–||–||–|
|Ng et al.19||–||–||–|
|Akhtar et al.20||–||–||–|
|Pilotto et al.21||9/106 <300 mg/day 9/42 >300 mg/day||8/36 acute (prn or reg >5 days, <30 days) 10/112 chronic (reg >1 month)||Acute vs. chronic|
|Pilotto et al.22||75–300 mg/day*||Daily||>3 months|
|Serrano et al.23||75 mg/day*n=3|
100 mg/day n=27
125 mg/day n=341
150 mg/day n=89
200 mg/day n=416
250 mg/day n=8
325 mg/day n=19
Frequency of aspirin use. For the bleeding inclusion studies, two included patients who were taking daily aspirin, two did not specify the frequency of aspirin use and the fifth categorized aspirin use as occasional, acute or chronic (with most of these patients being only occasional users). In the GI symptom studies, two did not specify the frequency of aspirin use, one stated daily aspirin use and the other divided use into acute (including as required use) and chronic. Most of the participants who had bleeding in this study were acute users of aspirin. In the LDA study, all were daily users.
Duration of aspirin use. The majority of studies did not report the duration of prior aspirin use. As outlined previously, there was some indication in two studies, where use was defined as acute or chronic. In one other study the inclusion criteria specified that aspirin use had to be greater than 3 months.
Antisecretory therapy use. Concomitant antisecretory therapy use cannot be easily extracted from the published data. In the majority of studies, it is not reported and where it has been, the use by patients with UGIB and those taking aspirin cannot be specifically determined.
Determination of Helicobacter pylori status. All studies used several modes of testing, either individually or in combination, to determine the H. pylori status of individuals and deemed the patient to be positive for the infection if any one of the tests was positive (Table 5). Five included the use of serological testing. None specified the concordance of the H. pylori test results, but two11, 15 did state that if the rapid urease test, culture or breath tests were negative, serological testing was performed. The number of additional patients considered to be H. pylori positive as a result of serological testing was not detailed.
|Study||Serology||Breath test||Histology||Culture||Rapid urease test|
|Adamopoulos et al.16||–||–||✓||✓||✓|
|Nakashima et al.17||✓||–||–||✓||–|
|Nakayama et al.15||✓||✓||–||–||✓|
|Tsesmeli et al.18||–||✓||✓||–||✓|
|Liu et al.11||✓||–||–||✓||✓|
|Ng et al.19||–||–||✓||✓||–|
|Akhtar et al.20||✓||–||Biopsies taken, test not specified|
|Pilotto et al.21||–||–||✓||–||✓|
|Pilotto et al.22||–||–||✓||–||✓|
|Serrano et al.23||✓||✓||✓||–||✓|
Bleeding criteria. The ten cohort studies used a mix of clinical and endoscopic definitions for UGIB (Table 6). The pathologies found on OGD also varied. One study included only gastric ulcers, whereas the majority included both gastric and duodenal ulcers, with two of these additionally including acute gastroduodenal mucosal lesions, gastritis and oesophagitis. Three studies, however, also included bleeding which had occurred as a result of non-peptic ulcer pathologies e.g. Mallory Weiss tears and varices.
|Study||Bleeding diagnosis||UGI pathology|
|Adamopoulos et al.16||Signs & symptoms of UGIB||All*|
|Nakashima et al.17||Bleeding PU on OGD||PU|
|Nakayama et al.15||Bleeding PU on OGD||PU|
|Tsesmeli et al.18||H/M requiring admission||All*|
|Liu et al.11||Bleeding lesions on OGD||PU|
|Ng et al.19||H/M or drop in Hb >2 g/dL with raised urea or bleeding GU on OGD||GU|
|Akhtar et al.20||UGIB symptoms||All*|
|Pilotto et al.21||H/M or drop in Hb >3 g/dL and/or bleeding lesions on OGD||PU|
|Pilotto et al.22||Bleeding lesions on OGD||PU/inflammation|
|Serrano et al.23||H/M requiring admission||PU/inflammation|
Randomized controlled trials24, 25
Chan et al.24 included data on both aspirin and naproxen use but, for the purposes of this review, only the aspirin data were examined. Both RCTs were of identical design and therefore the results have been combined. Patients diagnosed with UGIB on endoscopy and who were taking LDA and determined as H. pylori positive by rapid urease test or histology testing were included in the studies. In total, 312 patients were recruited. At least 8 weeks of proton pump inhibitor (PPI) treatment was prescribed to heal the ulcers and then the patients were randomly allocated to receive either eradication therapy or long-term maintenance PPI treatment (n = 156 in each group). At 6 months, the PPI group had two recurrent bleeds compared with four bleeds in the eradication group. One of the patients in the eradication group did, however, commence taking a NANSAID in addition to the prescribed aspirin during the course of the study. In the study by Chan et al.,24 the group allocated to eradication treatment had a higher proportion of patients with previous symptomatic ulcers (30%) and UGIB (26%) compared with the PPI treatment group, (24% and 18% respectively).
After adjusting for confounding factors such as previous UGIB, alcohol and the use of gastroprotective medication, the study by Lanas et al.12 suggests that H. pylori is a significant risk factor for UGIB in patients taking long-term LDA. The prevalence of H. pylori infection, however, was determined by serology only in just fewer than 30% patients. Serological evidence of H. pylori indicates previous infection by the bacterium, not necessarily current infection, although in this study serology appeared to correlate strongly with urea breath test results.
The prevalence of H. pylori infection in the study (controls 69%) was higher than what may be expected in a developed world sample. Helicobacter pylori infection is present in approximately 50% people worldwide, varying from 20% to 40% in the developed world to over 80% in some developing countries. It may be that the higher prevalence seen in this study differs from epidemiological studies as a result of selection bias; both the cases and controls are taking LDA for vascular disease secondary prevention. The risk factors associated with vascular disease are more prevalent in populations with poorer socioeconomic factors – a major risk factor for the childhood contraction of H. pylori. Thus, participants potentially have a higher prevalence of H. pylori infection compared with the overall population in that area.
With the cohort studies, the prevalence of H. pylori in users of aspirin presenting with UGIB varied from 62% to 100%, but the total number of patients involved is small. It is also impossible to interpret this apparently high prevalence of H. pylori infection without reference to a control group and the data do not allow for comparative testing.
Moreover, the variety of inclusion criteria, doses or duration of aspirin taken and H. pylori testing used make comparisons difficult. The inclusion of serological testing inflates the proportion of participants who are H. pylori positive, as seen in the case–control study, as this reflects positive results for both active and previous H. pylori infection. Although some studies specified that this was only used in cases where H. pylori eradication had not previously been prescribed, it does not account for eradication through other antibiotic use or spontaneous clearance of the infection. In fact, the prevalence of H. pylori in aspirin users with UGIB was less than that in the overall study population in two studies,15, 17 but again, this involves only small numbers and is unlikely to be of statistical significance.
The aspirin dose used varied in the studies and the doses at which bleeding occurred was not reported in all but one of the studies,21 but these limited data are consistent with previous studies where higher doses of aspirin are associated with an increased bleeding risk.26
The bleeding risk associated with acute vs. chronic use is important to address, given that with chronic use, the stomach may partially adapt to the local toxic effects of aspirin,27 and thus more acute use may be associated with increased bleeding risk. However, there is insufficient evidence within this review to determine an increased UGIB risk with acute aspirin use.
Although PPI treatment can reduce the incidence of UGIB with aspirin, it is associated with Clostridium difficile diarrhoea and, more recently, has been found to attenuate the antithrombotic effects of clopidogrel in some patients. Its impact on patients taking aspirin who have H. pylori and subsequent bleeding risk cannot be determined from this review, but would be important to clarify in future.
With the inclusion of non-ulcer pathologies as a cause of UGIB in some studies, the incidence of UGIB may be overestimated.
Given the heterogeneous nature of these cohort studies, and the lack of explicit data regarding H. pylori prevalence in aspirin users presenting with UGIB in most studies, the currently available evidence is not sufficient to support a meta-analysis and thus is inadequate to allow conclusions to be drawn regarding the impact of H. pylori on UGIB risk in aspirin users.
The results from the randomized controlled trials are equivocal owing to their small sample size and do not guide our clinical practice further. PPI therapy reduces the risk of bleeding from peptic ulcers caused by aspirin in patients who are H. pylori negative.28Helicobacter pylori is thought to impair the gastric mucosa’s adaptation to long-term aspirin use, but it may also potentiate the effects of PPI therapy thereby reducing overall bleeding risk. Its interaction with aspirin is certainly complex and poorly defined.
Unfortunately, occlusive vascular disease is common. With aspirin being key to its secondary prevention, and H. pylori potentially one of the few modifiable risk factors for UGIB, it is essential that we understand the impact of this organism on upper gastrointestinal bleeding risk in patients taking regular aspirin.
Declaration of personal interests: None. Declaration of funding interests: This systematic review was undertaken as part of the Kuck Clinical Research Training Fellowship awarded to Dr Emma Fletcher by Research Into Ageing and the British Geriatrics Society, grant number 313.