Should Hypertensive Patients Take Aspirin? A Report from the Food and Drug Administration

It is clearly established that patients who have had a coronary event such as a myocardial infarction (MI) or a bypass operation should take low‐dose aspirin to prevent recurrences. The benefits of aspirin in patients at lower levels of risk are more controversial because the prevention of MI must be balanced against the risk of bleeding, both in the gut and in the brain to produce hemorrhagic stroke. The issue of whether hypertensive patients should be advised to take aspirin routinely appeared to have been resolved by the Hypertension Optimal Treatment (HOT) study, in which nearly 19,000 hypertensive patients were randomized to take aspirin 75 mg/d or placebo and followed for 5 years. ¹ The result, which was published in 1998, concluded that aspirin reduced major cardiovascular events by 15% (p=0.03) and all MIs by 36% (p=0.002). There was no effect on the rates of stroke. To put the benefits in perspective, the HOT trial showed that low‐dose aspirin prevented between one and two major cardiovascular events per 1000 patient‐years (equivalent to 100 patients treated for 10 years). Put another way, 176 patients would need to be treated for 4 years to prevent one event. The conclusion of this analysis was that “association of acetylsalicylic acid with antihypertensive therapy can therefore be recommended, provided that blood pressure is well controlled and the risk of gastrointestinal and nasal bleeding is carefully assessed.” In keeping with this result, the recent seventh report of the Joint National Committee on the Prevention, Detection, Evaluation, and Treatment of High Blood Pressure stated that “low‐dose aspirin therapy should be considered only when [blood pressure] is controlled because of the increased risk of hemorrhagic stroke when the hypertension is not controlled.” ² So there, one might think, the matter should rest.

Indeed it might have, if the US Food and Drug Administration (FDA) had not agreed to review a “citizen's petition” put forward by Bayer HealthCare, the leading manufacturer of aspirin, stating that “aspirin should be recommended for individuals who A) have a 10% or greater risk of [coronary heart disease] events over 10 years or B) for whom there is a positive benefit as judged by their healthcare providers.” A similar request had been made previously, and the Cardiorenal Advisory Committee of the FDA reviewed the evidence in 1989 but concluded that there was insufficient evidence to support a primary prevention claim. At that time there were only two primary prevention studies, published in 1988 and 1989, both using male physicians as subjects. The first was the British Doctors' Study, ³ which found no overall benefit in 5138 physicians treated with 500 mg aspirin or placebo for 6 years and also noted a higher risk of gastrointestinal bleeding. The second was the Physicians' Health Study, ⁴ which randomized 22,000 physicians to 325 mg aspirin or placebo every other day. The trial was terminated prematurely because there was a marked reduction in fatal and nonfatal MIs.

Since that time three other primary prevention trials have been published: The Thrombosis Prevention Trial (TPT), ⁵ the Primary Prevention Project (PPP), ⁶ and HOT. These five studies, which included a total of 55,000 participants, were reviewed in a meta‐analysis by the US Preventive Services Task Force in 2002, ⁷ which estimated the risks and benefits of aspirin according to the patients' level of risk for coronary heart disease (CHD). Four of the five studies (the exception being the British Doctors' Study) showed a reduction in MIs, but none found any difference in overall mortality. The analysis assumed that the number of MIs prevented would be proportional to the patients' risk of MI and that the risks of gastrointestinal bleeding and hemorrhagic strokes would be the same in low‐ and high‐risk patients. The estimate that led to the Task Force recommending aspirin in patients at moderate risk (10% over 10 years) was that over a 5‐year period, if 1000 such patients took aspirin, 14 MIs would be prevented at a cost of one hemorrhagic stroke and three major gastrointestinal bleeds. The American Diabetes Association came to similar conclusions: “Consider aspirin therapy as primary prevention strategy in high‐risk men and women with type 1 or type 2 diabetes.” ⁸ The American Heart Association was more cautious with its statement: “Additional data in primary prevention are needed for complete assessment of aspirin's benefit‐to‐risk ratio.” ⁹

Proponents of the approval of aspirin for primary prevention in moderate‐risk patients made several points at the recent FDA hearing. The first had to do with the current and projected morbidity of CHD. First, although there has been a well publicized reduction in CHD mortality over the past 20 years, the number of new cases has not decreased. Second, the current epidemic of obesity will result in a marked increase of new cases during the coming years. Third, 20% of cases present as sudden death. Fourth, congestive heart failure, which is a major outcome of CHD, is one of the only chronic diseases for which incidence and mortality has increased every year over the past 25 years.

Given these arguments and recommendations from established authorities, it might be expected that the FDA Advisory Committee would readily agree to the citizen's petition. In fact, the Committee, of which I am a member, voted 11 to three against the proposed label change. There were a number of reasons that suggested the evidence was not quite as robust as originally proposed. These bear going into in some detail because the arguments have to do with how the data from clinical trials are analyzed and presented. First, when the actual 10‐year CHD risk in the primary prevention trials was examined, only one (TPT) fell into the “moderate risk” group (10%–20% over 10 years); the others were all less than 10%. TPT showed a 32% reduction of nonfatal MIs with aspirin and a smaller reduction of all MIs (29%). In high‐risk patients, aspirin reduces the chances of any serious vascular event by about 25% and also reduces the risk of strokes and mortality, ¹⁰ a very different picture from the primary prevention studies, which, it was argued by the FDA reviewers, makes it hard to extrapolate from one risk group to the other. Most hypertensive patients are in the moderate‐ or high‐risk groups, especially if they are older than age 60 years.

An analysis presented by the Antithrombotic Trialists' Collaboration pooled the data from the five primary prevention trials and grouped the individual participants into three risk groups: low (<10% risk of CHD over 10 years), moderate (10%–19%), and high (>20% risk). As expected, the vast majority was in the low‐risk group. For the moderate‐risk group, the reduction of MI was similar to the overall results (a 35% reduction), but the increase in the number of strokes (33%) was much higher than for the whole sample (a 9% decrease in the low‐risk group).

One problem was that with the exception of the TPT, where there was a marginal reduction in all MIs (p<0.04), none of the other primary prevention trials achieved their primary aim, which in each case was to either reduce cardiovascular mortality or a combination of fatal and nonfatal cardiovascular end points (MI and stroke, fatal and nonfatal). Even in PHS, which was stopped prematurely because of a dramatic 44% reduction in MIs, there was no suggestion of any effect on mortality, which was the stated primary end point. The PPP was also terminated early, with the statement that “aspirin lowered the frequency of all the end points.” However, the primary end point (major fatal and nonfatal cardiovascular events) was not achieved (although the relative risk was 0.71, the confidence interval was 0.48–1.04, indicating a lack of significance). Similarly, although MIs were reduced (the relative risk was 0.69), the effect was not significant. What was significant was “total cardiovascular events or diseases,” which included items such as angina and transient ischemic attacks. Claiming positive results when the primary end point is not met may seem reasonable, but it may also lead to trouble. As stated in a leading textbook on clinical trials, “Excluding randomized participants or observed outcomes or response variables can lead to biased results of unknown magnitude or direction.” ¹¹ Several examples come to mind of unexpected positive findings that were not part of the original hypothesized end points, such as the Prospective Randomized Amlodipine Survival Evaluation (PRAISE) ¹² and Evaluation of Losartan in the Elderly Study (ELITE) ¹³ trials, which led to larger trials that had negative outcomes—ELITE II ¹⁴ and PRAISE‐2. Interestingly, although the findings of PRAISE‐2 (a study of the effects of amlodipine on mortality on patients with heart failure) were presented at the American College of Cardiology meeting in 2000, the final results have yet to be published.

Another issue concerned the type of MI that was prevented by aspirin. As the other analyses had stressed, the FDA analysis showed that there was a consistent reduction of nonfatal MI, but no effect on fatal MI; however, the relative numbers of fatal MIs were relatively small (16% of the total). The statement in the HOT publication that aspirin reduced overall cardiovascular events by 15% and all MIs by 36% is actually misleading because these analyses excluded silent MIs. The HOT trialists had in fact collected data on silent MIs, which were diagnosed by comparing electrocardiograms (ECGs) taken at the beginning and end of the study. Their justification for excluding silent MIs from the final analysis was “to make the results comparable with those of other intervention studies and since 14% of the ECGs could not be obtained, silent myocardial infarctions were analyzed separately.” As shown in one of the tables in the original publication, when silent MIs were included in the category of “all myocardial infarctions,” there was no longer any significant effect of aspirin. The numbers of silent MIs in HOT were not trivial: 39% of the total number, and there was actually a bigger number of them in the aspirin group than in the placebo group (75 vs. 57). Because we have no way of knowing whether our patients are going to have a nonfatal and nonsilent MI as opposed to some other cardiovascular event, the advantages of taking aspirin for patients who have not yet declared which type of event they are going to experience are marginal at best.

So the results of the HOT study regarding the effects of aspirin stand or fall according to whether silent MIs are included or not. Because the silent MI data were collected and analyzed, it seems that downplaying their significance in the HOT report was a post hoc decision. The data were included in one of the tables of the paper ¹ (Table VI), but the statements in the discussion that the investigation “provides very clear evidence of a substantial beneficial action of acetylsalicylic acid on fatal and nonfatal acute myocardial infarctions,” and that “inclusion of silent myocardial infarction among events limited the benefits of acetylsalicylic acid” are disingenuous. The inclusion of silent MIs did not “limit” the benefits—it obliterated them.

The main argument for including silent MIs is that they do predict an adverse outcome. Thus in the Framingham Heart Study, 25% of MIs were discovered by examination of routine ECGs, which had either caused no symptoms or atypical ones that had not led to a clinical diagnosis of MI. Follow‐up of these subjects showed that the unrecognized infarctions were just as likely as the recognized ones to lead to death, heart failure, or strokes, ¹⁵ and in the Multiple Risk Factor Intervention Trial (MRFIT), 25% of MIs were recognized only by ECG changes, with no history of hospitalization. These silent MIs carried a similar prognosis to those that had been clinically diagnosed. ¹⁶

The issue of the effects of aspirin in women was also a concern. Only two of the five studies included any women (HOT and PPP). In the larger of these, the HOT study, there was no effect of aspirin on fatal or nonfatal MIs. The Women's Health Study should help answer this question

Most of the studies did not address the issue of hypertension. TPT included patients with uncontrolled hypertension (up to a blood pressure of 170/100 mm Hg), but the average systolic pressure at entry was 139 mm Hg. A subgroup analysis of TPT found that there was no benefit from aspirin in patients whose systolic pressure was >145 mm Hg, but there was a 25% reduction in MIs in patients with pressures between 130 and 145 mm Hg. ¹⁷ The HOT study indicated that in people whose blood pressure is well controlled, aspirin produces an overall net benefit if silent MIs are excluded, although it was quite small. There was benefit in patients whose blood pressure was very well controlled (a diastolic pressure below 85 mm Hg), but not in those in whom it remained high. ¹⁸ Unfortunately, the corresponding systolic pressures were not given. There was also a substantial benefit in patients who had impaired kidney function or who entered the study with very high systolic pressures (>180 mm Hg). The authors of the study recommended that low‐dose aspirin be recommended in patients whose blood pressures are well controlled, particularly if they are at high risk for heart disease and also if they have some degree of kidney damage.

The decision to recommend that a symptomless patient take a medication for the rest of his or her life that has a very small chance of preventing an adverse event and an even smaller chance that it may cause another type of adverse event is not one to be undertaken lightly. Neither patients nor doctors are used to evaluating risk in quantitative terms, but it is something we are going to have to get better at. As the practice of medicine evolves toward prevention of disease rather than treating acute catastrophes, the decision process becomes one of shades of gray as opposed to black and white. This requires better communication of risk to our patients, for whom it may not be intuitively obvious that a 10‐year risk of a heart attack of 15% is considered moderate, while a 25% risk is considered high. If our advice is to be followed, it will need to be a shared decision made by both the patient and the doctor.