Low-Dose Aspirin and Upper Gastrointestinal Bleeding in Primary Versus Secondary Cardiovascular Prevention
A Population-Based, Nested Case–Control Study
Background—The benefit–risk profile of low-dose aspirin in primary prevention of cardiovascular disease is unclear. We sought to quantify upper gastrointestinal bleeding (UGIB) risk associated with low-dose aspirin in secondary versus primary prevention patients.
Methods and Results—We performed a population-based nested case–control study using The Health Improvement Network (THIN) Database between 2000 and 2007. We identified 2049 cases of UGIB and 20 000 controls, frequency-matched to the cases on age, sex, and calendar year, who were subdivided into primary (without previous cardiovascular disease) and secondary (with previous cardiovascular disease) prevention populations. We estimated the relative risk of UGIB associated with the use of low-dose aspirin by multivariate logistic regression. The UGIB risk in patients taking low-dose aspirin relative to nonusers was significantly higher in the primary (adjusted relative risk, 1.90; 95% confidence interval, 1.59–2.26) than in the secondary (relative risk, 1.40; 95% confidence interval, 1.14–1.72; P value for the difference=0.0014) prevention cohort. However, as the baseline risk of UGIB was lower in the primary than in the secondary prevention cohort, numbers needed to harm per 1 year of low-dose aspirin use were 601 and 391 for primary and secondary prevention, respectively.
Conclusions—The relative risk of UGIB in patients taking low-dose aspirin is higher when used for primary than for secondary cardiovascular disease prevention, but this difference is more than compensated by the lower baseline risk in the primary prevention population. Such estimates are important for an assessment of the net clinical benefit in primary prevention.
Cardiovascular disease (CVD), including ischemic heart disease and cerebral vascular accidents, is the leading cause of morbidity and mortality in North America and Europe.1 Because of the prominent role of thrombosis in the pathogenesis of CVD, antiplatelet therapy, primarily with low-dose aspirin (75–300 mg/d), is recommended for its prevention.2,3 Even taking into account serious adverse events associated with the long-term use of low-dose aspirin (primarily major bleeding), the net clinical benefit is clearly in favor of low-dose aspirin use in the secondary prevention setting.3
The absolute benefit of low-dose aspirin for primary prevention (individuals without a history of CVD) is less clear, and its use remains controversial. Recommendations from the US Preventive Services Task Force are sex and age specific because both of these factors play important roles in determining the risk of CVD and bleeding (mainly upper gastrointestinal bleeding [UGIB]).4 Therefore, knowledge of sex- and age-specific risks of UGIB associated with the use of low-dose aspirin in patients without previous CVD is crucial for estimating the benefit–risk ratio of low-dose aspirin use in primary CVD prevention. Data on this topic are, however, very limited, and extrapolations from data in secondary prevention to primary prevention cohorts may be arbitrary.
Therefore, we sought to estimate sex- and age-specific UGIB risks associated with the use of low-dose aspirin in a large primary prevention cohort (comprising individuals without a history of CVD) and, separately, in a large secondary prevention cohort (individuals with a history of CVD), arising from the same general population.
Source Population and Case Ascertainment
A cohort study with nested case–control analysis was performed using data extracted from The Health Improvement Network (THIN) UK primary care database. Data on >3 million patients are systematically recorded by participating primary care practitioners as part of their routine patient care and sent to THIN, anonymized, for use in research projects. The study was approved by an institutional review committee (research ethics committee reference number: 09/H0305/90).
THIN contains patient information such as demographic factors, consultation rates, referrals, hospitalizations, laboratory test results, and prescriptions issued by primary care practitioners, including details on the dose regimens and duration of treatment. Prescriptions written by primary care practitioners are generated and recorded automatically in the database using a coded drug dictionary (Multilex).5
The study population comprised all individuals recorded in THIN who were 40 to 84 years of age between January 1, 2000, and December 31, 2007, and who were registered with a practice. Individuals were included in the study population only if they had been enrolled with their primary care practitioner for ≥2 years and had a computerized prescription history of ≥1 year. Patients with a recorded diagnosis of cancer, esophageal varices, Mallory–Weiss disease, alcohol abuse, liver disease, or coagulopathy before the start date, which is defined as the date they became eligible to enter the study, were excluded. The population was further subdivided into a primary prevention cohort (patients without a history of CVD) and a secondary prevention cohort (those with a history of CVD, including myocardial infarction, unstable angina, stable angina, cerebral vascular accidents, and peripheral arterial disease).
All members of the study population were followed up from the start date until the earliest of the following end points: a recorded diagnosis of UGIB (case detection) or of an exclusion factor; reaching the age of 85 years; death; or the end of the study period. The computerized patient profiles, including free-text comments, of all potential cases were reviewed manually. Individuals were considered to be UGIB cases if they had no exclusion criteria in the 2 months after the recorded UGIB; if they had not been discharged from hospital in the previous month; if their bleed was located in the stomach or duodenum; and the clinical diagnosis of the underlying lesion was peptic ulcer, erosions, or inflammation of gastric or duodenal mucosa. Patients with bleeding in the esophagus or lower gastrointestinal tract were excluded.
This study is an extension of an earlier study that demonstrates the validity of our method of UGIB case ascertainment.6 In our previous study, using a random sample, a positive predictive value of >95% was obtained for our definition of UGIB after comparison with primary care practitioners’ paper-based records (although THIN is anonymized, there is a unique ID created for each individual in the database that makes a link to additional data possible).6,7 After manual review of computerized records, the final number of confirmed cases of UGIB was 2049. The index date was the date of UGIB diagnosis for cases. A date during the study period was generated at random for each of the potential controls. If the random date was included in his or her eligible person-time (follow-up period), we marked that person as an eligible control and used the random date as the index date. All patients with a CVD diagnosis before the index date were considered the secondary prevention cohort so that, when estimating the relative risk (RR) of UGIB comparing those with versus those without a CVD history, a CVD event occurred before an UGIB event (happening on the index date) can contribute to the RR appropriately. A random selection of 20 000 controls was performed from the pool of eligible controls, frequency-matched to cases by age, sex, and calendar year. This sampling procedure enables the probability of being selected as a control to be proportional to the person-time at risk (incidence-density sampling, the method of choice for obtaining unbiased results).8
Estimation of Baseline Risk of UGIB
Although the main analysis, enrolling all UGIB cases with nested case–control analysis, can only provide RR estimates, we performed a subcohort analysis to estimate the baseline risk, which is required to estimate the absolute risk difference. A subcohort of aspirin nonusers without a history of CVD, enrolled in THIN from 2000 to 2007 (n=121 057), was used to estimate baseline risk of UGIB, ascertained by the same computerized algorithm as in the main analysis.
Computerized prescription records were used to assess the use of low-dose aspirin (75–300 mg/d), clopidogrel, oral anticoagulants, nonsteroidal anti-inflammatory drugs (NSAIDs; both cyclooxygenase-2 inhibitors [coxibs] and traditional NSAIDs), oral corticosteroids, selective serotonin reuptake inhibitors, nitrates, histamine-2 receptor antagonists, and proton pump inhibitors before the index date. Drug exposure was classified into 2 categories based on the expected pharmacological effects: current use was defined as use lasting until the index date or ending in the 30 days before the index date, and past use was defined as use ending 31 to 365 days before the index date. However, in a previous study,9 we found that the risk of UGIB among NSAID users starts to decrease quite markedly ≈7 days after treatment cessation, reflecting the reversible nature of the inhibition of cyclooxygenase by NSAIDs. Therefore, for NSAID use, current use was defined as use lasting until the index date or ending in the 7 days before the index date, recent use was defined as use ending 8 to 90 days before the index date, and past use was defined as use ending 91 to 365 days before the index date. For all medications studied, the reference group (nonuse) was defined as no exposure to the drug in the year before the index date. To reduce any lasting effects from the previous medication after recent switching from low-dose aspirin to clopidogrel or vice versa, monotherapy with low-dose aspirin or clopidogrel was defined as current use of the drug in question with no use of the other antiplatelet drug in the previous year.
A nested case–control analysis was performed to estimate the age-, sex-, and calendar year–adjusted RR of UGIB associated with the use of low-dose aspirin in the overall population, the primary versus secondary prevention cohort, and in sex- and age-specific groups separately within each cohort. Fully multivariate-adjusted RRs and 95% confidence intervals (CIs) were calculated using unconditional logistic regression. An interaction term (product of low-dose aspirin use and a history of CVD) was included to examine the difference in RRs in the primary versus secondary prevention cohort. Adjustment was made for age, sex, calendar year, number of primary care practitioner visits and hospitalizations in the year before the index date, history of peptic ulcer disease, smoking status, alcohol consumption, and use of oral corticosteroids, selective serotonin reuptake inhibitors, oral anticoagulants, NSAIDs, low-dose aspirin, clopidogrel, nitrates, histamine-2 receptor antagonists, and proton pump inhibitors before the index date.
Risk of UGIB Associated With Low-Dose Aspirin Use in the Primary Prevention Cohort
The adjusted RR of UGIB in current users of low-dose aspirin for primary prevention of CVD was 1.90 (95% CI, 1.59–2.26) compared with patients not taking low-dose aspirin in the year before the index date. There were only 3 controls and no cases taking clopidogrel in the primary prevention cohort; therefore, their RR could not be estimated. In this cohort of patients without a history of CVD, there was no significant difference in the RRs of UGIB associated with low-dose aspirin use across different age (P=0.901) or sex (P=0.829) groups (Table 1).
Risk of UGIB Associated With Low-Dose Aspirin Use in the Secondary Prevention Cohort
The adjusted RR of UGIB in current users of low-dose aspirin for secondary prevention of CVD was 1.40 (95% CI, 1.14–1.72) compared with patients not taking low-dose aspirin in the year before the index date. In these patients with a history of CVD, there was no significant difference in the RRs of UGIB associated with low-dose aspirin use across different age (P=0.932) or sex (P=0.9067) groups (Table 2).
Interaction Between History of CVD and the RR of UGIB Associated With Low-Dose Aspirin Use
The RR of UGIB in current users of low-dose aspirin for primary prevention of CVD was significantly higher than that in current users of low-dose aspirin for secondary prevention (P value for the difference=0.0014; Tables 1 and 2). Patients with a history of CVD were more likely to use a proton pump inhibitor than those without such history (14.4% versus 4.9%; P=0.0009, data not shown). However, the RR of UGIB associated with low-dose aspirin was still significantly greater in the primary prevention cohort (RR, 2.07; 95% CI, 1.71–2.52) than in the secondary prevention cohort (RR, 1.43; 95% CI, 1.13–1.82; P value for the difference=0.0014, data not shown) when the analysis was restricted to proton pump inhibitor nonusers.
Specific Indications of Low-Dose Aspirin Use for Secondary Prevention and the RR of UGIB
The RR of UGIB associated with low-dose aspirin use was 1.91 (95% CI, 1.43–2.55) in patients with cerebral vascular accidents or peripheral arterial disease and 1.14 (95% CI, 0.83–1.55) in those with ischemic heart disease (P value for the difference between the 2 groups=0.0029). Patients with previous ischemic heart disease were more likely to use a proton pump inhibitor than those with previous cerebral vascular accidents or peripheral arterial disease (19% versus 13%; P=0.0014, data not shown). The differential effect of low-dose aspirin between patients with previous ischemic heart disease and those with cerebral vascular accidents or peripheral arterial disease remained evident among proton pump inhibitor nonusers (P=0.044, data not shown). Note that the borderline P value should be interpreted cautiously in the context of multiple testing.
Comparison of Aspirin Primary Versus Secondary Prevention Cohort
Table 3 summarizes the distribution of patient characteristics with a history of CVD versus those without a history of CVD in a random sample of 10 000 subjects without UGIB from our study population. These subjects meet the same criteria as our controls in the nested case–control analysis except that they were unmatched to the cases of UGIB. Therefore, their distribution of patient characteristics is representative of the general population where cases and controls were ascertained. Compared with patients in the primary prevention population, patients in the secondary prevention cohort were older and more likely to have a history of ulcers, be smokers, and use oral corticosteroids, oral anticoagulants, NSAIDs, clopidogrel, and acid-suppressing agents such as proton pump inhibitors and histamine-2 receptor antagonists. The RR of UGIB comparing the primary versus secondary prevention cohorts in aspirin nonusers was 3.46 (95% CI, 2.94–4.06; data not shown).
Estimation of Absolute Risk Differences in the Primary Versus Secondary Prevention Cohort
Using our subcohort of aspirin nonusers without a history of CVD (n=121 057), the baseline risk of UGIB was estimated to be 1.85 cases per 1000 person-years in the primary prevention cohort. The baseline risk in aspirin nonusers in the secondary prevention cohort was thus 6.40 cases per 1000 person-years (1.85×3.46, where 3.46 is the RR of UGIB comparing the primary versus secondary prevention cohorts estimated among aspirin nonusers in our study population). With these baseline risk estimates and the differential RRs in those with versus without a history of CVD, the numbers needed to harm per year of aspirin use were 601 in the primary and 391 in the secondary prevention cohort.
We found that use of low-dose aspirin was associated with a significantly higher RR of UGIB in the primary prevention than in the secondary prevention population. The RRs of UGIB among low-dose aspirin users were relatively constant across different age and sex groups in both of the 2 CVD prevention populations.
Randomized controlled trials have reported a wide range of risk estimates for extracranial bleeding associated with use of low-dose aspirin in primary CVD prevention, with RRs ranging from 1.22 to 4.0710–14 and a pooled estimate of 1.70 (95% CI, 1.17–2.46).15 Of note, all these RRs came from selected populations, such as male doctors, female health professionals, or patients with risk factors for CVD. They are therefore not directly comparable with our population-based estimates. Reported extracranial bleeding risks associated with low-dose aspirin use for secondary CVD prevention have also been heterogeneous and have had wide CIs.16 A meta-analysis of aspirin secondary prevention trials reported a pooled estimate of gastrointestinal bleeding RR of 2.69 (95% CI, 1.25–5.76), but only postcerebral vascular accidents trials contributed to this estimate.3 Another meta-analysis of secondary prevention trials noted that the RR of major bleeding was 2.73 (95% CI, 1.75–4.27) in trials enrolling patients who had a history of cerebral vascular accidents and 1.58 (95% CI, 0.78–3.19) among those with ischemic heart disease.17 In our data set, we also found the RR of UGIB associated with low-dose aspirin use in patients with cerebral vascular accidents or peripheral arterial disease was significantly higher than that in patients with ischemic heart disease (P=0.0029).
It is important to take into account concomitant use of other antiplatelet agents when comparing gastrointestinal toxicity of aspirin across studies. Dual-antiplatelet therapy with low-dose aspirin and clopidogrel is commonly prescribed in secondary CVD prevention, and previous studies have shown that the UGIB risk associated with use of dual-antiplatelet therapy is greater than that with low-dose aspirin alone.18,19 Therefore, we have reported separately the risk estimates for UGIB associated with use of low-dose aspirin alone and low-dose aspirin plus clopidogrel, compared with nonuse of both drugs. We found that the RR estimates for dual-antiplatelet therapy were consistently higher than those associated with either low-dose aspirin or clopidogrel monotherapy across all subgroups. Most secondary prevention studies have not distinguished the UGIB risks associated with low-dose aspirin monotherapy from those associated with aspirin plus other antithrombotic therapy. Hence, the RRs for UGIB associated with low-dose aspirin reported in these studies would be expected to be greater—and less reliable—than our results.
Our findings demonstrate that, compared with the primary prevention population, patients in the secondary prevention cohort were, on average, older, more likely to have a previous history of ulcers, to be smokers, and to use oral corticosteroids, oral anticoagulants, NSAIDs and clopidogrel, which are all variables associated with an increased risk of UGIB. However, these patients were also more likely to use acid-suppressing agents, such as proton pump inhibitors and histamine-2 receptor antagonists, which are expected to lower the risk of UGIB. These factors are unlikely to explain the differential effects of low-dose aspirin between the primary and secondary prevention populations because they were all adjusted for in the regression models. Also, the interaction between a history of CVD and the risk of UGIB associated with low-dose aspirin use remained significant (P=0.0014) when the analysis was restricted to proton pump inhibitor nonusers. Therefore, other alternative explanations for this need to be found.
One explanation may be related to the categorization of primary and secondary prevention. As the definition of secondary prevention in this study is based primarily on the occurrence of a previous thrombotic event, which may be contributed, to some extent, by a thrombotic tendency, individuals in the secondary prevention cohort would probably share some protection from bleeding based on the shift of the hemostatic balance as compared with the primary prevention cohort. For example, higher levels of von Willebrand factor20 or a higher tendency to platelet aggregation21—prompting thrombosis but protecting from bleeding—in the secondary prevention cohort might explain the slightly higher RR associated with use of low-dose aspirin in the primary prevention cohort compared with the secondary prevention cohort. This hypothesis, which should be confined mostly to the ischemic heart disease subgroup, may now deserve attention and further confirmation in other data sets.
In our analysis, the RR of UGIB among low-dose aspirin users seems to be relatively constant across different age and sex groups in both of the 2 CVD prevention populations. Therefore, the greater excess risk associated with low-dose aspirin use in the primary prevention cohort cannot be directly explained by the younger age distribution in this population. A meta-analysis on sex difference in aspirin effect in the setting of primary prevention found differential benefits; more myocardial infarction prevention in males and more stroke prevention in females, although the bleeding risks were similar in both sex groups.22
Valid and precise estimates of absolute risk associated with aspirin use in primary and secondary prevention are important. Our analysis showed the numbers needed to harm per year of aspirin use were 601 in the primary and 391 in the secondary prevention cohort. To put these numbers into perspective, although taking aspirin increases UGIB risk by 90% for a patient without previous CVD and 40% for a patient with such history (based on RRs), a doctor will see 1 case of UGIB in a patient prescribed low-dose aspirin for a year after treating 391 patients with previous CVD and 601 without such history (based on numbers needed to harm). This is because the 90% and 40% increase in the risk of UGIB were relative to different baseline risks, lower in primary than in secondary prevention cohort. Our risk estimates provide valuable data arising from a same-source population in calculating risk–benefit ratios in a real practice setting, whereas those reported in the literature were from selected populations in randomized controlled trials, without differentiation of primary prevention from secondary prevention populations, or inferred from different populations.
A major strength of this study is its large sample size. In addition, the similarity of data in THIN and those from UK national statistics supports the notion that THIN data are accurate and representative of the UK population.23–25 Furthermore, our method of ascertaining cases of UGIB has been validated in a previous study: 95% of UGIB cases from a random sample were confirmed on direct consultation with the primary care practitioners.6,7 A limitation of THIN is that it captures only recorded prescription medications. It is likely that some patients were using over-the-counter low-dose aspirin or NSAIDs and may have been misclassified as nonusers in this study. However, in a previous study, the influence of under-recording of aspirin use on estimates of risk was investigated and found to be minimal.26 In addition, although we adjusted for a comprehensive list of potential confounders, our findings should be interpreted with caution of possible residual confounding or confounding by unmeasured factors.
In conclusion, although the RR of UGIB in patients taking low-dose aspirin seems to be higher for primary than for secondary CVD prevention, the lower baseline risk of UGIB in the primary prevention than in the secondary prevention more than compensates for the higher RR associated with low-dose aspirin. Within the secondary prevention cohort, low-dose aspirin users with previous cerebral vascular accidents or peripheral arterial disease may have a significantly higher risk of UGIB than those with ischemic heart disease. We did not find significant difference in the RR of UGIB associated with low-dose aspirin across different age and sex groups in either the primary or secondary prevention cohort.
WHAT IS KNOWN
Upper gastrointestinal bleeding associated with the use of low-dose aspirin in patients without previous cardiovascular disease is crucial for estimating the benefit–risk ratio of low-dose aspirin use for primary cardiovascular disease prevention.
Data on this topic are, however, very limited.
WHAT THE STUDY ADDS
The relative risk of upper gastrointestinal bleeding in patients taking low-dose aspirin is higher when used for primary than for secondary cardiovascular disease prevention, but the lower background bleeding risk in the primary prevention population more than compensates for this difference.
We thank Susan Bromley of Oxford PharmaGenesis Ltd, who provided editorial assistance funded by AstraZeneca R&D, Mölndal, Sweden.
Sources of Funding
The study was supported by an unrestricted research grant from AstraZeneca R&D, Mölndal, Sweden.
Dr Rodríguez works for Spanish Centre for Pharmacoepidemiologic Research (CEIFE), which has received research funding and consultancy fees from AstraZeneca R&D, Mölndal, Sweden. The other authors report no conflicts.
- Received July 22, 2013.
- Accepted October 16, 2013.
- © 2013 American Heart Association, Inc.
- Baigent C,
- Blackwell L,
- Collins R,
- Emberson J,
- Godwin J,
- Peto R,
- Buring J,
- Hennekens C,
- Kearney P,
- Meade T,
- Patrono C,
- Roncaglioni MC,
- Zanchetti A
- 5.↵MULTILEX for primary care. In: First Data Bank, 2010. http://www.fdbhealth.co.uk/solutions/multilex/. Accessed January 25, 2010.
- Hansson L,
- Zanchetti A,
- Carruthers SG,
- Dahlöf B,
- Elmfeldt D,
- Julius S,
- Ménard J,
- Rahn KH,
- Wedel H,
- Westerling S
- Peto R,
- Gray R,
- Collins R,
- Wheatley K,
- Hennekens C,
- Jamrozik K,
- Warlow C,
- Hafner B,
- Thompson E,
- Norton S
- 16.↵Antithrombotic Trialists' Collaboration. Collaborative meta-analysis of randomised trials of antiplatelet therapy for prevention of death, myocardial infarction, and stroke in high risk patients. BMJ. 2002;324(7329):71–86.
- García Rodríguez LA,
- Lin KJ,
- Hernández-Díaz S,
- Johansson S
- Hallas J,
- Dall M,
- Andries A,
- Andersen BS,
- Aalykke C,
- Hansen JM,
- Andersen M,
- Lassen AT
- Elwood PC,
- Renaud S,
- Sharp DS,
- Beswick AD,
- O’Brien JR,
- Yarnell JW