by Henry I. Bussey, Pharm.D.
The assistance of Robert Talbert, Pharm.D. and Roger Lyons, M.D. in developing this posting is gratefully acknowledged.
December 11, 2018
It is critically important to recognize that this discussion is about “primary prevention” (preventing a first stroke or heart attack in at risk individuals). This discussion is NOT about “secondary prevention” (preventing a subsequent heart attack or stroke in someone who has already experienced one of these events) or about preventing complications of various procedures such as surgery or stent placement. The data supporting aspirin for secondary prevention is much stronger and not being questioned here.
Four Recent Studies Feed an Ongoing Controversy:
Whether the benefits of aspirin in primary prevention out-weigh the risks has been an ongoing controversy for at least the last decade as documented by several postings on ClotCare.org. The first of 4 recent studies is the focus of the most recent ClotCare posting in August, 2018. That study was a pooled analysis of 10 earlier studies of aspirin for primary prevention. The title of the posting is “Is It Time to Tailor the Dose of Aspirin for Primary Prevention of Cardiovascular Events and Cancer?”. Then, the very next month, three new large, randomized, placebo-controlled studies provided new data that raise the question “Should We Stop Using Aspirin for Primary Prevention?”. [2-6] The 3 new studies are known as the ASCEND, ARRIVE, and ASPREE trials. Each trial involved a different study population and details will be summarized below. Following the publication of the 3 new studies, Paul Ridker, M.D. provided an editorial which concluded that “…the benefit-risk ratio for prophylactic aspirin in current practice is exceptionally small” and he suggested that a better choice would be to add a statin. Dr. Ridker also pointed out that of the three new studies and 11 others published since 2008, only 1 showed a difference in all-cause mortality and that difference favored placebo over aspirin in the ASPREE trial. His comments suggest that smoking cessation and the aggressive treatment of diabetes mellitus, high blood pressure, and elevated cholesterol (especially with statins) lower cardiovascular (CV) risk to such a level that aspirin provides little further benefit.
Taken together, the 4 new studies certainly question the use of aspirin for primary prevention and they also raise questions about possible harms that have not been well documented previously. Even so, there are issues of study design, study execution, and data analysis that may undermine the conclusions. Therefore, the following discussion will be divided into 3 parts:
First let’s look at the key conclusions from the pooled analysis of 10 earlier studies and see how those findings relate to key findings of the 3 recent randomized trials.
Second, let’s examine some of the key limitations of these studies.
Third, let’s review key aspects of each of the 3 new randomized trials.
First – Key conclusions of the pooled analysis and how they relate to findings in the 3 new randomized trials. [Relevant findings of the new studies are listed in brackets]
In the pooled analysis of 10 earlier trials, which included 117,279 patients, the authors suggested the following. 
- Low dose aspirin (75 mg to 100 mg daily) was effective only in patients who weighed 70 Kg or less. [ASCEND, ARRIVE, and ASPREE used an aspirin dose of 100 mg daily. [2-6] And in the ASCEND trial, in patients with diabetes, those weighing < 70 Kg did NOT show more benefit and, in fact, the authors indicated that the trend was in the opposite direction (but data were not presented).]
- Aspirin may increase the risk of sudden death in those weighing less than 50 Kg. [In the ASPREE trial of aspirin in the elderly (70 years and older), all-cause mortality was increased in the aspirin group but this harm was due mainly to more cancer deaths. Sudden death was not evaluated.][4-6]
- Enteric coated or sustained-release aspirin – or alternate day dosing – may be less effective. [ASCEND, ARRIVE, and ASPREE trials used only enteric coated tablets given once daily.][2-6]
- “High dose” aspirin (> 300 mg) is effective in those weighing more than 70 Kg but may be harmful in those weighing less than 60 Kg. [ASCEND, ARRIVE, and ASPREE did not use “high dose” aspirin.][2-6]
- Twice daily dosing may be more effective. [ASCEND, ARRIVE, and ASPREE used once a day dosing.][2-6]
- Low dose aspirin does not alter the rate of cancer in those who weigh 70 Kg or more. [ASPREE, conducted in individuals 70 years and older, found an increase in overall mortality with aspirin which was due mainly to more cancer deaths – including an increase in colorectal cancer. [4-6] ]
- Low dose aspirin increases the rate of cancer in those weighing less than 70 Kg or more. [See comment on ASPREE in #6 above.]
- Aspirin (any dose) appeared to reduce the incidence of colorectal cancer at 20 years of follow-up in those who weighed 50 to 80 Kg with no clear benefit outside of that range. [See comment on ASPREE in #6 above; also it is possible that the 5 to 7 year duration of the new studies may have been too short to show any beneficial effect on reduction of cancer.]
- Younger (< 50 years old) women with diabetes who took aspirin had a > 2 fold increase in breast cancer and a > 4 fold increase in overall cancer rates. [See comment on ASPREE in #6 above.]
- Low dose aspirin increased major bleeding up to a body weight of 90 Kg. [ASCEND, ARRIVE, and ASPREE each reported an increase in some type of bleeding although fatal bleeding and hemorrhagic stroke sometimes were not increased.][2-6]
- Major bleeding was increased with high dose aspirin and this risk increased even further in those who weighed more than 90 Kg. [See note in #10 above; “high dose aspirin” was not evaluated in any of the 3 new randomized trials.]
Second – Key study limitations that may undermine the conclusions.
- Each study was quite large; the pooled analysis had more than 100,000 patients while ASCEND, ARRIVE, and ASPREE enrolled 12,500  to 19,114 patients [4-6]. Such large studies carry the risk of identifying event rate differences that are statistically significant but not clinically important. For example, a gastrointestinal bleeding rate that is 0.5% higher with aspirin over a 5-yr period may be statistically significant, but such a small increase in bleeding risk may not be clinically important; especially if those events were mainly mild bleeds?
- ASCEND, ARRIVE, and ASPREE used an aspirin dose of 100 mg once daily administered as an enteric coated tablet. The pooled analysis suggested that such a dose is too little for many (most?) patients, that the dose may be too infrequent (considering that 10 to 15% of one’s platelets may be produced daily after aspirin has left the blood stream), and that an enteric coated product may compromise any effect due to limited and variable absorption.
- Use of composite endpoints and assessing multiple “other” endpoints. With a composite endpoint, event rates with several of the endpoints may “dilute” or obscure a benefit or harm in another end point. For example, if the composite endpoint is heart attack + cardiovascular death + stroke + TIA (transient ischemic attack), the finding of no significant reduction in the composite event rate may obscure the fact that heart attacks were reduced.
- Repeated significance testing. Most statistical tests are designed to see if a given, pre-specified endpoint is different than what would be expected to occur based on chance alone. If one employs such a test to examine multiple endpoints, the likelihood of finding an erroneous “statistical” difference increases with each additional endpoint tested. In several of these studies, the correct adjustment for a statistical “p” value was not made. For example, mortality rates of 5.9% vs. 5.2% over almost 5 years were found to be statistically significant, but when the test was adjusted for multiple comparisons, the difference was no longer statistically significant.
- Patient adherence and “intention–to-treat” analysis”. Treatment studies can be analyzed by “on treatment” or “per protocol” analysis (evaluating only those events that happened among individuals actually adhering to the assigned treatment), or by “intention to treat” analysis (evaluating events that happened based on assigned treatment without regard to whether the patient was adhering to the treatment). Intention-to-treat analysis is usually preferred because it is thought to be more indicative of what happens in usual practice when treating a given group of individuals, and it is a more statistically rigorous analysis. But, a significant effect of treatment may be missed if patients do not adhere to the assigned treatment and/or if they switch from one treatment arm of the study to the other arm of the study. In ASCEND, ARRIVE, and ASPREE, adherence was poor (approximately 60 – 70%) and at least in one study a decline in aspirin use in the aspirin study group was identified as was an increase in non-study aspirin use in the placebo group.
- Location of studies. The outcomes evaluated in these studies are influenced by diet, lifestyle, and other factors that vary from country to country. Each of these trials was conducted largely or exclusively outside of the U.S. which could mean that the results do not apply to individuals in the U.S.
- Low baseline risk. The likelihood of demonstrating a beneficial effect of a given treatment becomes less at lower levels of baseline risk. The observed event rates in these studies were less than one-half of the predicted risk. This lower than expected rate of events may have been due, somehow, to the selection criteria or, perhaps, to changes that have occurred over time. Changes such as smoking cessation, more aggressive treatment of high blood pressure, and more aggressive treatment of lipid disorders (and especially the use of statins) may explain the lower than anticipated event rates.
- Each study involved a different patient population. Benefits and risks of treatment in ASCEND, ARRIVE, and ASPREE may be similar or conflicting when comparing results across studies. It is not surprising that “healthy elderly” (> 70 years old) persons in Australia may respond differently than a 40 year old patient with diabetes mellitus living in the U.K. Therefore, it would seem reasonable for the clinician to keep in mind how various populations appeared to respond. For that reason, the three studies will be summarized below individually.
Third, let’s review key aspects of each of the 3 new randomized trials.
- ASCEND :
- Population: 15,480 patients with diabetes mellitus (DM), 40 years of age or older, living in the U.K. The predicted 5-year risk of a vascular event was low (< 5%) in 40% of the study participants, moderate (5-10%) in 42%, and high (>10%) in only 17%.
- Follow-up: mean of 7.4 years
- Benefit: First vascular event (excluding intracranial hemorrhage – ICH) – 8.5% with ASA vs. 9.6% with placebo (p=0.01) with no difference in vascular deaths, gastrointestinal (GI) cancer, or other cancers.
- Harm: First major bleed 4.1% with aspirin vs. 3.2% with placebo (p=0.003) – 41.3% of bleeds were GI, 21.1% “sight threatening”, and 17.2% ICH. Fatal bleeds (0.2%) and hemorrhagic stroke (0.3%) were not different between groups.
- Bottom line: Is a 1.1% lower vascular event rate over 7.4 yrs worth a 0.9% increase in major bleeding with no difference in vascular deaths, fatal bleeds, hemorrhagic stroke, or cancer?
- Notes: Analysis did not adjust statistics for multiple comparisons, adherence was 70% in both groups with aspirin use declining over time in the aspirin group while non-study aspirin increased over time in the placebo group. 75% of patients were taking a statin and those weighing < 70 kg did NOT show greater benefit and, in fact, the trend was in the opposite direction (per the authors but data were not presented).
- ARRIVE 
- Population: 12,546 moderate risk (10-20% 10-yr risk) in men 55 years and older or women 60 years and older – anticipated event rate was 17%. Those with DM and high bleeding risk were excluded. Study participants were enrolled mainly from Western Europe with only 4% from U.S.
- Follow-up: median of 60 months
- Benefits: The composite endpoint of heart attack or myocardial infarction (MI) + cardiovascular (CV) death + unstable angina + stroke or transient ischemic attack (TIA) was NOT significantly lower with aspirin (4.29% vs. 4.48%) but approximately 30% of patients failed to complete the study. When data were analyzed “per protocol” in the 60% of patients who were at least 60% adherent, the difference in the composite endpoint approaches statistical significance (3.40% vs. 4.19%, difference of 0.79%, p=0.0756) and the difference in MI is statistically different (0.98% vs. 1.84%, difference = 0.86%, p=0.0014).
- Harm: GI Bleeding rates were higher with aspirin (0.97% vs. 0.46%, difference = 0.51%, p<0.0007). Most bleeding events were mild and there was no difference in fatal bleeds.
- Bottom line: There was no difference in benefit by intention-to-treat analysis, and GI bleeding increase was small (0.51% over 5 years). But if one looks at patients who were at least 60% adherent, there was a statistically significant 0.86% reduction in MI.
- Notes: Approximately 30% of patients failed to complete the study and 43% were taking a statin. Mortality was not different – 160 patients (2.55%) with aspirin vs 161 (2.57%) with placebo. The observed event rate (<4.5%) was much less than the predicted 17%.
- ASPREE – this study was published in 3 parts with one reporting CV events and bleeding, another reporting mortality, and the third assessing “disability-free survival”. [4-6]
- Population: 19,114 “healthy elderly” 70 years and older, 90% white, 87% from Australia. 70% had 2 or more risk factors for CV disease.
- Follow-up: Median 4.7 yrs – stopped early because of lack of benefit.
- Benefits: There was no difference in CV disease. MI and Ischemic Stroke were lower with aspirin but the difference was not statistically significant. The composite endpoint of death + dementia + physical disability was not different (21.5 with aspirin vs. 21.2 with placebo).
- Harm: Deaths were actually higher with aspirin (12.7 vs. 11.1 per 1,000 pat-yrs) due mainly to an increase in cancer deaths (6.7 vs. 5.1, a difference of 1.6 per 1,000 pat-yrs.). Even colorectal cancer mortality was higher (0.8 vs. 0.5 per 1,000 pat-yrs). Major bleeding was higher with aspirin (8.6 vs 6.2 per 1,000 pat-yrs, difference = 2.4 major bleeds per 1,000 pat-yrs, p<0.001). Almost half of all major bleeds were GI. ICH was 50% higher with aspirin at 2.5 vs 1.7 per 1,000 pat-yrs, difference = 0.8 per 1,000 pat-yrs. Fatal hemorrhage was not different.
- Bottom line: No clear benefit vs. a risk of major bleeding including almost 1 more ICH per 1,000 pat-years. Mortality (including cancer-related and even colorectal cancer-related deaths) were statistically higher with aspirin.
- Notes: This study was stopped early because of lack of benefit. The difference in overall mortality becomes non-significant if the p value is adjusted for multiple comparisons. Mortality rates were lower than expected and when compared to a matched general population, all-cause mortality in study patients was approximately one-third of the matched population and cancer-related mortality was 49% of the matched population. Adherence in the final year of the trial was 62.1% in the aspirin group vs. 64.1%.
Conclusion: The risk of bleeding and, in some patients, the risk of a higher mortality rate due to cancer-related deaths, makes it rather difficult to recommend adding aspirin to other currently used measures for primary preventiondev. However, these studies have a number of potential flaws and we do not have data to say whether a weight-based aspirin dosing regimen or a different aspirin product would yield better results. But with very low event rates in the 4 to 5% range over several years, it is hard to imagine that a different aspirin dosing regimen and/or aspirin product would produce results great enough to warrant taking on the increased risk of bleeding and/or cancer in primary prevention. Even so, a given clinician may encounter selected higher risk patients in whom the perceived benefit of aspirin may warrant its use. In such an instance, I would use an unproven approach and modify the dosing regimen in 3 ways. First, I would use a chewable aspirin formulation to minimize GI irritation and perhaps improve absorption. Second, I would use a twice a day schedule to inhibit the function of the platelets released during the day after the day’s first dose of aspirin has been eliminated (as discussed in the pooled analysis).  Third, I would use a total daily dose larger than 100 mg., especially in patients who weigh more than 70 Kg.
- Rothwell PM, Cook NR Gaziano JM, et al. Effects of aspirin on risks of vascular events and cancer according to bodyweight and dose: analysis of individual patient data from randomized trials. Lancet 2018; 392:387-99. dx.doi.org/10.1016/S0140-6736(18)31133-4
- ASCEND Study Collaborative Group. Effect of aspirin for primary prevention in persons with diabetes mellitus. N Engl J Med 2018; 379:1529-39. DOI: 10.1056/NEJMoa1804988
- Gaziano JM, Brotons C, Coppolecchia R et al. Use of aspirin to reduce risk of initial vascular events in patients at moderate risk of cardiovascular disease (ARRIVE): a randomized, double-blind, placebo-controlled trial. Lancet 2018; 392:1036-1046. Dx.doi.org/10.1016/S0140-6736(18)31924-x
- McNeil JJ, Wolfe R, Woods RL, et al. Effect of aspirin on cardiovascular events and bleeding in the healthy elderly. N Engl J Med 2018; 379:1509-18. DOI: 10.1056/NEJMe1812000
- McNeil JJ, Nelson MR, Wods RL, et al. Effect of aspirin on all-cause mortality in the healthy elderly. N Engl J Med 2018; 379:1519-28. DOI:10.1056/NEJMe1812000
- McNeil JJ, Woods RL, Nelson MR, et al. Effect of aspirin on disability-free survival in the healthy elderly. N Engl J Med 2018; 379:1499-508. DOI: 1056/NEJMe1812000
- Ridker PM. Should aspirin be used for primary prevention in the post-statin era? N Engl J Med 2018; 379:1572-74. DOI: 10.1056/NEJMe1812000