Weighing Risk Versus Benefit
Many reports of epidemiologic studies focus on the strength of association, i.e., risk ratios and rate ratios. However, when trying to weigh decisions for an individual person it it important to consider:
- The individual's personal circumstances, i.e., what are their risks for the primary outcome?
- What adverse effects of the proposed treatment are they at risk for?
- What are the absolute benefits and absolute risks of the proposed therapy?
Is Low Dose Aspirin Beneficial?
A number of descriptive studies suggested that people who took aspirin regularly seemed to have a lower risk of myocardial infarction (heart attack). Observational studies suggested perhaps a 30% reduction in risk of myocardial infarction, but of course the subjects were not randomized, so there were concerns about unrecognized confounding. Several small clinical trials suggested similar reductions, but the sample sizes were too small to arrive at a solid conclusion. In the early 1980s The Physician's Health Study was conducted to test the hypothesis that 325 mg. of aspirin (one 'adult' sized aspirin) taken every other day would reduce mortality from cardiovascular disease (N. Engl. J. Med. 320:1238, 1989). Male physicians 40 to 84 years of age living in the US in 1980 were eligible to participate. Physicians were excluded if they had a personal history of myocardial infarction, stroke or transient ischemic attack; cancer; current gout; liver, renal or peptic ulcer disease; contraindication to aspirin consumption; current use of aspirin, platelet-active drugs or non-steroidal anti-inflammatory agents; intolerance to aspirin; or inability to comply with the protocol. Eligible subjects who met the inclusion criteria and who successfully completed a run-in phase were randomly assigned to receive aspirin or a placebo. Eventually 22,071 physicians were enrolled; 11,037 were assigned to aspirin, and 11,034 were assigned to placebo. The agents (aspirin and placebo) were identical in appearance and were mailed to the subjects. The recipients' treatment group was coded, and neither the subject nor the investigators knew which treatment group a given subject was in. The table below summarizes the number of 'events' that had occurred in each treatment group after about 5 years of follow up. The primary outcome of interest was myocardial infarction, but possible adverse effects of chronic aspirin use, such as stroke, ulcer disease, and bleeding problems were also recorded.
Endpoints |
Aspirin Group |
Placebo Group |
---|---|---|
|
(N=11,037) |
(N=11,034) |
Myocardial Infarction |
|
|
Fatal MI |
10 |
26 |
Non-fatal MI |
129 |
213 |
Total MI |
139 |
239 |
Ischemic Stroke |
91 |
82 |
Hemorrhagic Stroke |
23 |
12 |
Upper Gastrointestinal Ulcer |
169 |
138 |
Upper GI ulcer with bleeding |
38 |
22 |
Bleeding Problems |
2,979 |
2,249 |
Bleeding requiring blood transfusion |
48 |
28 |
The data clearly show a substantial decrease in the occurrence of both fatal and non-fatal myocardial infarctions among those randomized to aspirin compared to placebo. However, there also were an increased number of hemorrhagic strokes, ulcers, and bleeding problems. These results were controversial at the time. Most of the investigators wanted to continue the study to clarify whether there was an increased risk of stroke. However, the data safety and monitoring board for the study strongly recommended that the study be terminated, because the benefit of aspirin had been clearly demonstrated, and they felt it was unethical to withhold its use from half of the participants.
As an exercise, you can calculate the risk ratios comparing the aspirin and placebo groups on the different outcomes. How much do the risk ratios help you weigh the benefits and risks of aspirin therapy.
Your recommendations?
Based on the results of your analysis, what recommendations would you make regarding the benefits and risks of this regimen with low dose aspirin? Write down your recommendations, based on your analysis of this data, regarding the use of low dose aspirin.
An alternative approach?
Before you look at the feedback, consider whether there is an alternate way of looking at the data in the table. Would an alternate approach offer any advantages in weighting the risks and benefits?