This week, a headline-making study in the New England Journal of Medicine sought to cast doubts on long-established science that says daily aspirin can be a broadly-effective anti-aging tonic. I’m writing this response because I think that this new, small study has to be viewed in the context of many larger studies over many decades that together make a solid case for aspirin’s benefits.
Aspirin has two kinds of effects: First, aspirin thins the blood, reduce clotting, which lowers the risk of most kinds of heart attacks and stroke (ischemic) while raising the risk of bleeding ulcers and hemorrhagic stroke. Second, aspirin lowers the level of systemic inflammation, which reduces risk of heart disease, stroke, most cancers, and Alzheimer’s disease.
Historically, daily low-dose aspirin began to be prescribed broadly to middle-aged and older adults in the 1960s as the medical establishment theorized about the first effect. This led to a grand natural experiment—tens of millions of older people taking low-dose aspirin. Studies comparing these people with matched populations who didn’t take aspirin have shown lower rates of all-cause mortality, Alzheimer’s dementia, and of cancer and probably of heart disease as well. These studies are based on millions of tabulated deaths. The current study is based on 1052 total deaths in the aspirin group and the placebo group, and the difference between the two was barely statistically significant in the direction against aspirin.
Summary of past studies
Eidelman, JAMA, 2003: Summarizing 5 trials, they found aspirin was associated with a 32% reduction in the incidence of first heart attacks. Statistical significance was 2 chances in 100,000 (p<0.00002).
Methods A computerized search of the English literature from 1988 to the present revealed 5 published trials: the Physicians’ Health Study (22,071 participants), the British Doctors’ Trial (5,139), the Thrombosis Prevention Trial (5,085), the Hypertension Optimal Treatment Study (18,790), and the Primary Prevention Project (4,495).
Results Among the 55,580 randomized participants (11,466 women), aspirin was associated with a statistically significant 32% reduction in the risk of a first MI and a significant 15% reduction in the risk of all important vascular events, but had no significant effects on nonfatal stroke or vascular death.
Conclusions The current totality of evidence provides strong support for the initial finding from the Physicians’ Health Study that aspirin reduces the risk of a first MI. For apparently healthy individuals whose 10-year risk of a first coronary event is 10% or greater, according to the US Preventive Services Task Force and the American Heart Association, the benefits of long-term aspirin therapy are likely to outweigh any risks.
Rothwell, The Lancet 2011: Summarizing 8 trials, they found aspirin was associated with a 21% reduction in the incidence of all cancers. Statistical significance was 1 chances in 10,000 (p<0.0001).
In eight eligible trials (25,570 patients, 674 cancer deaths), allocation to aspirin reduced death due to cancer (pooled odds ratio [OR] 0·79, 95% CI 0·68–0·92, p=0·003). On analysis of individual patient data, which were available from seven trials (23,535 patients, 657 cancer deaths), benefit was apparent only after 5 years’ follow-up (all cancers, hazard ratio [HR] 0·66, 0·50–0·87; gastrointestinal cancers, 0·46, 0·27–0·77; both p=0·003). The 20-year risk of cancer death (1634 deaths in 12 659 patients in three trials) remained lower in the aspirin groups than in the control groups (all solid cancers, HR 0·80, 0·72–0·88, p<0·0001; gastrointestinal cancers, 0·65, 0·54–0·78, p<0·0001), and benefit increased (interaction p=0·01) with scheduled duration of trial treatment (≥7·5 years: all solid cancers, 0·69, 0·54–0·88, p=0·003; gastrointestinal cancers, 0·41, 0·26–0·66, p=0·0001). The latent period before an effect on deaths was about 5 years for oesophageal, pancreatic, brain, and lung cancer, but was more delayed for stomach, colorectal, and prostate cancer. For lung and oesophageal cancer, benefit was confined to adenocarcinomas, and the overall effect on 20-year risk of cancer death was greatest for adenocarcinomas (HR 0·66, 0·56–0·77, p<0·0001). Benefit was unrelated to aspirin dose (75 mg upwards), sex, or smoking, but increased with age—the absolute reduction in 20-year risk of cancer death reaching 7·08% (2·42–11·74) at age 65 years and older.
Wang, Journal of Alzheimer’s 2015: Summarizing 11 trials, they found aspirin was associated with a 49% reduction in the incidence of dementia. Statistical significance was less than 1 chances in a billion (p<0.0000000005).
Objective: Alzheimer’s disease, the most prevalent dementia, is a prominent source of chronic illness in the elderly. Laboratory evidence suggests that nonsteroidal anti-inflammatory drugs (NSAIDs) might prevent the onset of Alzheimer’s disease. Since the early 1990s, numerous observational epidemiological studies have also investigated this possibility. The purpose of this meta-analysis is to summarize and evaluate available evidence regarding exposure to nonaspirin NSAIDs and risk of Alzheimer’s disease using meta-analyses of published studies. Methods: A systematic search was conducted using Medline, Biological Abstracts, and the Cochrane Library for publications from 1960 onwards. All cross-sectional, retrospective, or prospective observational studies of Alzheimer’s disease in relation to NSAID exposure were included in the analysis. At least 2 of 4 independent reviewers characterized each study by source of data and design, including method of classifying exposure and outcome, and evaluated the studies for eligibility. Discrepancies were resolved by consensus of all 4 reviewers. Results: Of 38 publications, 11 met the qualitative criteria for inclusion in the meta-analysis. For the 3 case-control and 4 cross-sectional studies, the combined risk estimate for development of Alzheimer’s disease was 0.51 (95% CI = 0.40–0.66) for NSAID exposure. In the prospective studies, the estimate was 0.74 (95% CI = 0.62–0.89) for the 4 studies reporting lifetime NSAID exposure and it was 0.42 (95% CI = 0.26–0.66) for the 3 studies reporting a duration of use of 2 or more years. Conclusions: Based on analysis of prospective and nonprospective studies, NSAID exposure was associated with decreased risk of Alzheimer’s disease. An issue that requires further exploration in future trials or observational studies is the temporal relationship between NSAID exposure and protection against Alzheimer’s disease.
Problems with the present study
Because of small numbers and short duration, the result of the study was only marginally significant (p<0.05). The aspirin group had higher cancer rates and lower heart attack rates than placebo.
Typically, doctors advise patients to start low-dose aspirin around age 50, but this study was with patients more than 70 years old who had no cardiovascular symptoms by age 70. Most people by age 70 have had some cardiovascular diagnosis before age 70, so this is an unrepresentative sample. The study fails to address the question, how many deaths and how many diseases could be avoided between the ages of 50 and 70? This is the period in life when inflammation is most active, and a great deal of destruction of the body’s veins, joints, and nervous system happens during these years. Excluding those with a history of heart disease during those ages is excluding just the people most likely to be helped by aspirin. Of course, when you’re 50 and considering whether to start on aspirin, you may not know whether you’re lucky enough (or have the right genes) to be in the group that will do fine for the next 20 years without it.
This table breaks the composite test group into sub-groups according to various criteria. Dots to the right of the line mean “aspirin was worse”, and to the left mean “aspirin was better”. Among the subgroup in the US, aspirin was better. Among people who had never taken aspirin before, aspirin was better. Among people within fairly wide limits of a “normal” weight range, aspirin was better.
Why are we seeing this?
Scientists are only human, and their environment, their preconceptions, and their incentives shape the way that statistics are handled. In my experience, it is not difficult to make a small effect look like a (p<0.05) effect by making consistent choices in the way the data are treated, none of which are suspect or dishonest. If the group had come up with the conventional and accepted conclusion based on such a small study, there would have been no prominent publication, no headlines, probably no follow-on grant. So they had every incentive to perform the analysis in a way that makes the results appear more interesting than they are.