Prophylactic aspirin and risk of peptic ulcer bleedingBMJ 1995; 310 doi: http://dx.doi.org/10.1136/bmj.310.6983.827 (Published 01 April 1995) Cite this as: BMJ 1995;310:827
- John Weil, senior registrar in public health medicinea,
- Duncan Colin-Jones, consultant physicianb,
- Michael Langman, professor of medicinec,
- David Lawson, consultant physiciand,
- Richard Logan, reader in clinical epidemiologye,
- Michael Murphy, consultant epidemiologistf,
- Michael Rawlins, professor of clinical pharmacologyg,
- Martin Vessey, professor of public healthh,
- Paul Wainwright, research associatei
- a West Midlands Regional Health Authority, Birmingham
- b Queen Alexandra Hospital, Portsmouth PO6 5LY
- c Queen Elizabeth Hospital, University of Birmingham, Birmingham B15 2TH
- d Royal Infirmary, Glasgow G4 0SF
- e University of Nottingham, Nottingham NG7 2UH
- f Unit of Health Care Epidemiology, University of Oxford, Oxford OX2 6HE
- g University of Newcastle, Newcastle upon Tyne NE2 4HH
- h Radcliffe Infirmary, University of Oxford, Oxford OX2 6HE
- i Institute of Cancer Studies, University of Birmingham, Birmingham B15 2TJ
- Correspondence to: Professor Langman.
- Accepted 31 January 1995
Objective: To determine the risks of hospitalisation for bleeding peptic ulcer with the current prophylactic aspirin regimens of 300 mg daily or less.
Design: A case-control study with hospital and community controls.
Setting: Hospitals in Glasgow, Newcastle, Nottingham, Oxford, and Portsmouth.
Subjects: 1121 patients with gastric or duodenal ulcer bleeding matched with hospital and community controls.
Results: 144 (12.8%) cases had been regular users of aspirin (taken at least five days a week for at least the previous month) compared with 101 (9.0%) hospital and 77 (7.8%) community controls. Odds ratios were raised for all doses of aspirin taken, whether compared with hospital or community controls (compared with combined controls: 75 mg, 2.3 (95% confidence interval 1.2 to 4.4); 150 mg, 3.2 (1.7 to 6.5); 300 mg, 3.9 (2.5 to 6.3)). Results were not explained by confounding influences of age, sex, prior ulcer history or dyspepsia, or concurrent non-aspirin non-steroidal anti-inflammatory drug use. Risks seemed particularly high in patients who took non-aspirin non-steroidal anti-inflammatory drugs concurrently.
Conclusion: No conventionally used prophylactic aspirin regimen seems free of the risk of peptic ulcer complications.
This finding applies to doses in common use of 75, 150, and 300 mg daily
Despite these results, overall benefits of treatment are likely considerably to outweigh possible risks
The risks of peptic ulcer complications, particularly bleeding, are raised in association with aspirin1 2 3 4 and other non-steroidal anti-inflammatory drug use.5 6 7 8 Most information about aspirin use concerns doses of more than 300 mg daily and intermittent use.9 10 11 12 13 14 15 16 17 18 19 20 Less is known about risks associated with regular use of doses below 300 mg, which are increasingly being used as prophylaxis against vascular disease. As standard doses seem to be associated with substantial risks, it is important to know whether these lower doses are significantly safer. We report a case-control study conducted to see if important differences in the risks of hospital admission for ulcer bleeding existed between different non-steroidal anti-inflammatory drugs.
Patients and methods
Patients studied and methods used were as detailed elsewhere.8 Briefly, prior drug intake of all types was determined by questioning patients aged 60 and over admitted with haematemesis and melaena due to gastric and duodenal ulceration to hospitals in Glasgow (200), Newcastle (124), Nottingham (506), Oxford (144), and Portsmouth (170) between April 1987 and January 1991. Cases were as far as possible consecutive and were diagnosed by conventional clinical criteria. Findings were compared with those in age and sex matched hospital and community controls. Hospital controls were chosen from among all acute medical admissions (excluding patients with acute myocardial infarction, acute rheumatic diseases, and active non-bleeding ulcers); community controls were selected as the next person of the same sex and age (within five years) on the alphabetically ordered register of the same general practitioner as the case.
Standard questionnaires, in addition to seeking information about all prescribed and self administered drug intake, also asked (among other things) about alcohol consumption, smoking, and any history of gastrointestinal disease. Results were analysed by calculating odds ratios and 95% confidence intervals by means of unconditional logistic regression. Conditional methods were rejected because initial analysis suggested that similar results but with tighter confidence intervals were achieved by unconditional methods. This mainly arose because the matching code for triplets went missing at one of the centres, so reducing the number of triplets available. In addition, because the referent groups chosen were cases and controls not exposed to non-aspirin non-steroidal anti-inflammatory drugs or aspirin, a large number of triplets became incomplete because of the high overall rate of non-aspirin non-steroidal anti-inflammatory drug or aspirin use.
Of the 1144 cases intially included, 23 showed no evidence of ulcer haemorrhage on review and were therefore excluded. Results in the remaining 1121 were compared with those in the 1126 hospital controls selected and the 989 community controls who agreed to take part.
Table I shows that 144 (13%) cases were daily users of aspirin (defined as at least five days a week) compared with 101 (9%) of 1126 hospital controls and 77 (8%) of 989 community controls. In addition, 42 cases and 14 hospital and five community controls had been daily users of aspirin for less than a month. Odds ratios calculated by logistic regression to take account of confounding factors were consistently slightly greater in comparisons with community than with hospital controls, but all suggested raised risk (table II). Odds ratios increased progressively with aspirin dose in people who had been daily users for at least a month.
Daily users appeared to be at increased risk, by a doubling or more, if they had been users for less than a month compared with users for a month or more. There were also greater proportions of other regular or irregular users of aspirin among cases than among controls, defined respectively as used once to four times a week for more than a week and less than a month, and less than one day a week or only in the four days before admission. Table III shows that in all takers of aspirin, irrespective of dose, odds ratios compared with combined controls calculated by logistic regression tended to be greatest in those who had taken aspirin for a week or less and differed little in those who had taken it longer.
Table IV shows the numbers of aspirin takers according to formulation in cases and controls, but irrespective of the dose taken, with the odds ratios calculated by logistic regression compared with the combined controls. Odds ratios were significantly raised for all comparisons except for those with enteric coated aspirin and benorylate. Data on dosage (table V) showed that average daily aspirin doses were similar in cases for enteric coated and soluble aspirin, and for aspirin tablets, and slightly greater for other commercial varieties.
Tables VI and VII show the risks associated with daily aspirin use in the previous month together with non-aspirin non-steroidal anti-inflammatory drug use. The odds ratios were raised almost eightfold in those who took both drugs, the difference compared with users of aspirin alone being significant (P<0.05).
We found that no particular dose of aspirin between 75 mg and 300 mg daily currently used in cardiovascular prophylaxis is free of risk of causing bleeding from gastric or duodenal ulcers. Even very low (75 mg) doses of aspirin reportedly caused gastric bleeding in volunteers.21 Nevertheless, evidence that low doses of aspirin may or may not be associated with risks of peptic ulcer bleeding is limited.
In the atrial fibrillation, aspirin, anticoagulation (AFASAK) and thrombosis prevention trials there were three major episodes of upper gastrointestinal bleeding in, respectively, 336 and 907 recipients of aspirin, compared with no episodes in 336 and 932 subjects given placebo.22 23 In addition, in the Swedish aspirin low dose trial (SALT) there were nine episodes of severe gastrointestinal bleeding in subjects taking aspirin 75 mg daily and four in those taking placebo,24 whereas in the physicians' health study the relative risk of bleeding requiring transfusion (site unspecified) was 1.71 (95% confidence interval 1.09 to 2.69) in takers of aspirin 325 mg every other day compared with placebo.19
By contrast, in the research group on instability in coronary artery disease (RISC) study there were apparently no episodes of gastrointestinal bleeding in 399 recipients of 75 mg aspirin daily,25 and in the Dutch transient ischaemic attack trial there were no differences in the rates of gastrointestinal bleeding in those receiving 30 mg and 283 mg aspirin daily.26 Case-control studies have been claimed to overestimate the risks of non-steroidal anti-inflammatory drug associated upper gastrointestinal bleeding compared with randomised trials,27 but though we found higher risks for 150 mg and 300 mg than in some vascular prevention trials,18 19 they were no higher than those in the United Kingdom transient ischaemic attack study.28
Comparisons between the two data sources are, however, difficult. Randomised trials can give complete sets of unbiased information, but exclusion criteria employed at trial entry may limit generalisation to the population at large. Case-control studies, provided that their frame is wide enough, give data from a general base, though biases—for instance, in questioning and in control selection—may skew conclusions.
Our study was conducted in five major urban centres and included large numbers of cases and both hospital and community controls. Precautions were taken to minimise selection and information bias. Thus drug histories were checked against hospital and general practitioner records and found to be in substantial agreement.8 We believe that the design was robust, though there is evidence of selection bias, in that hospital controls were more likely to be exposed to aspirin than community controls. Even with this conservative bias the results showed a significantly increased risk of hospital admission for ulcer bleeding in association with all aspirin regimens used and compared with either set of controls.
Our main analyses took account of smoking and alcohol consumption as well as of prior histories of dyspepsia or peptic ulceration and of non-aspirin non-steroidal anti-inflammatory drug use. Further analysis of data showed that concurrent non-aspirin non-steroidal anti-inflammatory drug use roughly doubled risk. Table III indicates that risk tended to be greater when aspirin use had been of short duration. This increase associated with short term use may be explained by the taking of aspirin to alleviate symptoms arising from incipient peptic ulceration, but the substantially raised risk associated with daily aspirin use for less than a month (table II) suggests that there may be particular risks associated with prophylactic aspirin use early in treatment. These findings are consonant with results obtained elsewhere.5 8 28
Table IV suggests that enteric coated aspirin and the aspirin-paracetamol combination benorylate may be free of risk whereas the buffered perparation Alka-Seltzer may be associated with high risk. It should be noted, however, that confidence intervals were large and that because of its high buffering capacity Alka-Seltzer may have been used preferentially by people with gastric symptoms. There was no evidence of material differences in risk between aspirin tablets, soluble aspirin, and other commercial varieties.
Some 10000 episodes of ulcer bleeding occur in people aged 60 and over each year in England and Wales. Other data of ours suggest that some 3500 of these will be takers of non-aspirin non-steroidal anti-inflammatory drugs and, if our current figures are representative, 1700 or 17% of the total will be taking prophylactic aspirin compared with 8% of community controls. It may be deduced that 900 of the 10000 episodes could be associated with and ascribed to prophylactic aspirin use. A general change to low doses (75 mg) of aspirin would not eliminate risk but, again if our figures are soundly based, would reduce risk by about 40% compared with 300 mg doses and by 30% compared with 150 mg doses. It is unclear if substituting enteric coated aspirin would eliminate risk in view of evidence that non-steroidal anti-inflammatory drug use can cause haemorrhage, perforation, or stricture in the lower bowel.29 30 31 Taken overall, benefit from prophylactic drug use is likely substantially to outweigh risk, but refinements to dosage and delivery would clearly be valuable.
We are most grateful for the help given by our clinical colleagues and investigators Clare Clifford, Gail Faulkner, Gillian Paice, Shirley Powell, Ellen Thompson, and Shirley Wood and for the support of the Medical Research Council.