Cardiovascular disease is the leading cause of morbidity and mortality in Europe and North America 1. The overall incidence of myocardial infarction is around 1 to 3 cases per 1,000 person-years 12. However, a given person's risk for myocardial infarction varies significantly with a number of factors including age, sex, blood pressure, cigarette smoking, diabetes, cholesterol levels, and a previous diagnosis of cardiovascular disease 3. This risk variability is so widely accepted that the specific cardiovascular risk for a given person can be formally estimated using equations currently available on the web 4.

Long-term use of low-dose aspirin reduces the risk of myocardial infarction by about 25%–30% irrespective of the baseline risk 567, therefore this relative change has a greater absolute impact on persons with an elevated underlying risk. It has been estimated that treating patients with a moderately high baseline risk with aspirin would prevent up to 4 myocardial infarctions per 1,000 persons treated per year, while treating low risk patients would prevent less than 1 event per 1,000 persons treated per year 8.

The cardiovascular benefits from low-dose aspirin have to be balanced against the potential gastrointestinal harm. Observational studies have estimated that the incidence of serious upper gastrointestinal tract complications (UGIC; i.e., bleeding or perforation) in the general population is on the order of 1 case per 1,000 person-years, with a case fatality rate around 5–10% 9. Aspirin use, even at doses below 300 mg/day and despite modified release formulations, has been associated with a two to threefold increased risk of developing UGIC 10. Based on these data, the average incidence of UGIC would be close to 2–3 cases per 1,000 person-years for aspirin users, and about 1–2 of these cases might be attributable to aspirin. Clinical trials, typically conducted in low-risk subjects, have also estimated an overall excess of 1–2 cases of UGIC per 1,000 patients treated with aspirin per year 8.

After considering the absolute increase in bleeding complications associated with aspirin, several studies have suggested that aspirin use for prevention of cardiovascular events might be harmful for groups of patients at low risk for coronary heart disease (< 5 events per 1,000 person-years) and clearly beneficial for high-risk groups (≥15 events per 1,000 person-years) 31112. While these studies considered various cardiovascular risk factors to determine the coronary heart disease risk for each individual, they did not take into account the independent gastrointestinal risk factors to determine individualized UGIC risks.

Estimates of the specific UGIC risk for each person would add a valuable piece of information to individual therapeutic decisions and could optimize the overall risk-benefit profile of aspirin. Therefore, we describe below i) the distribution of the main risk factors for UGIC among populations of prescription aspirin users, ii) the estimated baseline risk of UGIC within levels of these factors, and iii) the estimated excess risk of UGIC that could be attributed to aspirin within subgroups of aspirin users with specific baseline risks of UGIC.

The prevalence of prescription aspirin use was 4.6% in the GPRD and 3.4% in BIFAP. Over 95% of the aspirin use recorded in the GPRD and 59% of the use in BIFAP was for cardioprotection. In both data sources the prevalence of aspirin use increased substantially with age (Figure 1).

The distribution of gastrointestinal risk factors among cardioprotection aspirin users is presented in Figure 2. Overall, the proportion of aspirin users above 60 years of age was 88% in the GPRD and 79% in BIFAP. The proportion of males was 52% in the GPRD and 54% in BIFAP. The proportion of aspirin users with a recorded history of gastrointestinal ulcer was 3.8% in the GPRD and 5.9% in BIFAP. The proportion of aspirin users with a recorded past episode of UGIC was 1.2% in the GPRD and 1.9% in BIFAP. The proportion of aspirin users concomitantly using NSAIDs was 14.8% in the GPRD and 13.2% in BIFAP. NSAID use was not included in the figure to simplify the presentation.

Figure 3 depicts how the estimated incidence rate of UGIC increases with age and with past history of ulcer both for males and females. These rates would be around four times higher among NSAID users and would double with aspirin use. Taking into account the distribution of baseline gastrointestinal risk factors among aspirin users in each population, the estimated average incidence rate of UGIC among aspirin users was 12 cases per 1,000 person-years in the GPRD and 10 in the BIFAP. However, this rate can be more than 20 cases per 1,000 person-years in certain high-risk groups. For example, considering, age, sex, ulcer history and use of NSAIDs as stratifying factors, the expected UGIC incidence rate would be above 20 per 1,000 person-years in 14% and 11% of aspirin users, based on GPRD and BIFAP, respectively. Moreover, in a smaller fraction of the population, for example males 70 and older with a recent history of peptic ulcer and who take aspirin and NSAIDs concomitantly, the risk might exceed 100 cases per 1,000 person-years.

The number of cases that could be attributable to aspirin among aspirin users according to age, sex, and ulcer history is also presented in Figure 3. For cardioprotection aspirin users overall, the estimated excess risk of UGIC was 6 and 5 cases per 1,000 users per year in the GPRD and BIFAP, respectively. These figures would be larger in high-risk groups. For example, patients with baseline incidence rates over 20 per 1,000 person-years, such as males 70 years and older with a past history of uncomplicated peptic ulcer, would develop over 40 cases per 1,000 person-years if they were treated with low-dose aspirin; thus, more than 20 cases per 1,000 aspirin users with these risk factors per year could be attributable to their use of aspirin.

The number of extra cases attributable to aspirin depends on the assumptions regarding incidence rates, distribution of risk factors, and relative risks used throughout the analyses. Estimates were particularly sensitive to the value used for the relative risk of UGIC associated with aspirin use. Figure 4 presents the number of extra cases potentially attributable to aspirin according to age, sex, and ulcer history for a plausible range of relative risks associated with low-dose aspirin (i.e., from 1.5 to 3).

Based on our estimates, the expected incidence of UGIC in the control groups of the Vioxx Gastrointestinal Outcomes Research (VIGOR) trial 21, the Celecoxib Long-term Arthritis Safety Study (CLASS) 22, and the Therapeutic Arthritis Research and Gastrointestinal Event Trial (TARGET) 23 would have been 7, 9.5 and 8.1 cases per 1,000 person-years, respectively; the actual reported annualized rates were 8.6, 14.5, and 9.1 per 1,000 person-years, respectively (Table 1).

The prevalence of cardioprotective aspirin use among adults in year 2000 was 4.6% in the UK and 2.0% in Spain according to European databases. The prevalence increased substantially with age and was slightly higher in men than in women, so that use was above 10% among men over 65 years of age. These data agree well with other studies from the UK and Spain that reported a prevalence of regular aspirin use of 8–9% in elderly populations 2425.

In the US, according to the population-based National Health and Nutrition Examination Survey (NHANES) 26, close to 9% of adults were using an aspirin containing product regularly in the 2000–2002 period, although the prevalence of aspirin use was above 30% in men over 65 years of age. Similarly, a contemporaneous US survey specifically designed to provide information on drug use reported a prevalence of aspirin use that ranged from 10% in women and men aged 18–44 years of age to 39% in men over 65 27. The higher prevalence of aspirin use in the US surveys might be due not only to actual differences in aspirin use or aspirin users in the US and Europe, but also to methodological aspects. For example, the European databases used for the current study included only prescription aspirin, focused on single component products used for cardioprotection and were based on prospectively recorded prescriptions; while the US surveys included prescription and nonprescription aspirin used regularly for any indication and both as single component or as part of multiple component products, and were based on subject recall. Of note, the lack of information on over the counter aspirin use would have a limited impact on this particular study as we were interested in users of aspirin for cardioprotection, and i) most regular aspirin users in the UK and Spain seek prescriptions in order to benefit from the drug coverage offered by the universal medical care, and ii) even if we had missed a fraction of cardioprotective aspirin users, those included in our population would probably be representative in terms of UGIC risks and, therefore, valid for the objectives of the study.

The specific UGIC risk for a given person according to the most relevant gastrointestinal risk factors can be calculated using the estimates provided in this paper. In addition, these estimates can be used to determine the expected number of UGIC cases in a particular group of individuals during a given period of time. For example, our estimates of the incidence of UGIC in the control groups of VIGOR 21, CLASS 22, and TARGET 23 trials were strikingly close to the observed ones, despite the distinct methodological characteristics of the trials.

This paper also provides estimates for the absolute impact of aspirin treatment on persons with different underlying UGIC risk profiles. The excess risk attributable to aspirin in the general population has traditionally been suggested to be around 1–2 cases per 1,000 patients treated with low-dose aspirin per year 8. However, this figure largely applies to a middle-aged group of people with no prior history of peptic ulcer and may substantially underestimate the bleeding risk among aspirin users in a real-life setting, where persons over 65 are precisely the group most likely to be taking low-dose aspirin for its cardioprotective properties. We estimated an overall excess risk of 5 cases per 1,000 aspirin users per year. Yet this is an average figure; the excess risk varied in parallel to the baseline UGIC incidence rate and ranged from less than 1 case to over 10 extra cases per 1,000 aspirin users per year for patients with low or high underlying gastrointestinal risk, respectively.

Therefore, age, together with other major risk factors for UGIC such as prior history of gastrointestinal ulcer, male gender, or concomitant use of NSAIDs, should be considered when balancing the risks and benefits of low-dose aspirin for a particular subgroup of patients. Estimates of the absolute impact of aspirin treatment can assist patients and health care providers in evaluating the magnitude of the trade-offs involved in taking aspirin for a particular individual. Nonetheless, the decision making process will still face the challenge of weighing the benefits against the potential harm. That is, the number of cardiovascular events prevented cannot be directly compared with the number of bleeding events caused, as they are associated with different degrees of disability and mortality. Moreover, each person might weigh these beneficial and harmful effects differently 11.

Some of the estimates for the distribution of gastrointestinal risk factors among aspirin users presented in this article were based on relatively small numbers and, therefore, there is a level of uncertainty around them. Regarding the effect estimates, although we assigned reliable relative risks to each gastrointestinal risk factor (i.e., they were based on pooled estimates with tied confidence intervals), we had to make assumptions (e.g. that the relative risk for aspirin is constant across certain groups with different baseline UGIC risks) that, while reasonable based on current knowledge, might be proven partly wrong in future studies. For simplicity, we restricted the number of gastrointestinal risk factors to those with greatest potential impact due to their high prevalence and/or relative risk; less frequent (e.g. use of anticoagulants, other comorbidities) and weaker (e.g. smoking or alcohol drinking) risk factors for UGIC were not considered. Also, while information on other risk factors such as Helicobacter pylori infection or genetic predisposition is rarely available at present, these and other markers will probably become important determinants of a person UGIC risk in the future. Similarly, we did not include other benefits of aspirin treatment, such as the secondary prevention of serious vascular events other than myocardial infarction 56 or the potential reduction in the risk of colon cancer 28 and cognitive decline in the elderly 29; nor other potential areas of harm, such as the increased risk of hemorrhagic strokes 56 and uncomplicated peptic ulcers 30. However, aspirin effects on the risks of myocardial infarction and UGIC are the most common, firmly established and serious and, therefore, would probably drive the decisions 12. Finally, we did not discuss interventions that might modify the balance, such as use of a proton pump inhibitor.

The net impact of aspirin on the risk of UGIC for a given person depends on the underlying gastrointestinal risk. The clinical decision about placing a person into low-dose aspirin for cardioprotection should consider not only the cardiovascular but also the gastrointestinal risk factors for each individual. This paper presents estimates of the UGIC risk for aspirin users with a wide range of gastrointestinal risk profiles.

The authors have served as consultants for manufacturers of non-steroidal anti-inflammatory drugs. The Pharmacoepidemiology Program at the Harvard School of Public Health and CEIFE are partially supported by training and research grants from various pharmaceutical companies. The current manuscript was not directly financed by any organization.

SHD participated in the design of the study, performed the statistical analysis and drafted the manuscript. LAGR conceived the study, participated in its design, and contributed to the interpretation of the results and writing of the manuscript. All authors read and approved the final manuscript.