Age-Related Maculopathy in a Randomized Trial of Low-Dose Aspirin Among US Physicians FREE

AGE-RELATED maculopathy (ARM) is the leading cause of vision loss in older Americans.¹ Despite its common occurrence, however, the pathophysiology of ARM is poorly understood and treatment options are limited.¹ Although laser photocoagulation^{2 - 4} and, more recently, photodynamic therapy^{5 - 6} have been shown to be of benefit for a few patients with advanced disease, laser treatment merely delays subsequent vision loss,^{2 ,7} and the long-term effects of photodynamic therapy remain to be determined. Moreover, other than cigarette smoking,^{8 - 10} and perhaps low levels of antioxidant nutrients,^{11 - 13} no avoidable risk factors have been identified. In the absence of an effective means of disease prevention, the importance of ARM as a source of ocular morbidity will only increase as the US population ages. For these reasons, identification of inexpensive, safe strategies to prevent the common forms of ARM is of particular public health importance.¹

Results of laboratory studies^{14 - 17} have shown that blood flow in the choroid of eyes with ARM is impaired, raising the possibility that vascular disease might contribute to the pathophysiological mechanism of ARM. Consistent with this hypothesis, a variety of epidemiologic studies have shown that the risk of ARM is increased for those with cardiovascular disease (CVD)^{18 - 21} or CVD risk factors such as elevated blood pressure^{18 ,20 ,22 - 26} and elevated lipid levels,^{25 ,27} although conflicting data have also been reported.^{25 ,28 - 32} If vascular disease is a contributing factor in ARM, then agents that enhance systemic circulation and decrease the risk of vascular events could also have a beneficial effect on choroidal circulation and decrease the risks of ARM. One possible agent is aspirin, which irreversibly inhibits platelet cyclooxygenase, resulting in a marked inhibition of platelet function.^{33 - 34}

The conduct of the Physicians' Health Study I (PHS I) presented a unique opportunity to examine the development of ARM in a large-scale trial of low-dose aspirin treatment. The PHS I was a randomized, double-masked, placebo-controlled trial of the use of aspirin and beta carotene in the prevention of CVD and cancer conducted among 22 071 US male physicians aged 40 to 84 years in 1982. The randomized aspirin component of the trial was terminated early, after an average of 5 years of treatment and follow-up, primarily because of a statistically extreme 44% reduced risk of first myocardial infarction. In this study we examine the effect of assignment to low-dose aspirin treatment on risks of ARM during the 5-year treatment period.

The PHS I was designed to evaluate use of low-dose aspirin (Bufferin; Bristol-Myers Products, New York, NY) (325 mg every other day) and beta carotene(50-mg supplement on alternate days) in the primary prevention of CVD and cancer among 22 071 apparently healthy US male physicians aged 40 to 84 years in 1982. Baseline information included height, weight, history of cigarette smoking, history of alcohol use, blood pressure, cholesterol level, history of diabetes mellitus, medication history, and multivitamin use. Annual questionnaires were sent to all participants to monitor their compliance with the treatment regimen and the occurrence of any relevant events.

The aspirin component of the PHS I was terminated early, on January 25, 1988, after an average follow-up of 60.2 months because of a statistically extreme 44% reduced risk of a first myocardial infarction among those taking aspirin.³⁵ (Adverse effects in the aspirin group included increased risks of upper gastrointestinal tract ulcers and bleeding problems, although the frequency and severity of these adverse effects were far lower than those reported in previous trials of aspirin.³⁵ ) The reported consumption of aspirin or other platelet-active drugs was 86% in the aspirin group and 14% in the placebo group.

Information concerning the occurrence of ARM during the first 7 years of the trial was requested on the 84-month questionnaire. Physicians were asked, "Have you ever had macular degeneration diagnosed in your right (left) eye?" If yes, they were requested to provide the month and year of the diagnosis. Subsequent annual questionnaires requested information on diagnoses during the preceding year. Signed permission to examine medical and hospital records pertaining to the diagnosis was also requested on the questionnaire and in separate follow-up mailings when necessary. Ophthalmologists and optometrists were contacted by mail and asked to complete an ARM questionnaire supplying information about the date of initial diagnosis of ARM, the best-corrected visual acuity at the time of diagnosis, and the date when visual acuity reached 20/30 or worse (if different from the date of initial diagnosis). Information was also requested about the pathological findings observed (drusen, retinal pigment epithelium [RPE] hypopigmentation/hyperpigmentation, geographic atrophy, RPE detachment, subretinal neovascular membrane, or disciform scar) when visual acuity was first noted to be 20/30 or worse and the date when exudative disease was first noted (defined by the presence of RPE detachment, subretinal neovascular membrane, or disciform scar). In addition, we asked whether there were other ocular abnormalities that would explain or contribute to visual loss and, if so, whether the ARM, by itself, was significant enough to cause best-corrected visual acuity to be reduced to 20/30 or worse.

Diagnoses of ARM were divided into 2 categories: (1) ARM, defined as a self-report confirmed by medical record evidence of an initial diagnosis of ARM subsequent to randomization but before January 25, 1988, and (2) ARM with vision loss, defined as in (1) but with vision loss to 20/30 or worse attributable to ARM.

The present study includes the 21 216 participants who did not report ARM at baseline and who were followed up for at least 7 years (ie, physicians who died during the first 7 years of follow-up and therefore did not respond to the 84-month questionnaire were excluded). As of November 1997, a total of 117 reports of ARM were confirmed as incident cases diagnosed before January 25, 1988. Of these, 57 cases were responsible for vision loss to 20/30 or worse.

Due to the early termination of the aspirin component of the PHS I, after an average of 60.2 months of follow-up, the number of incident cases of ARM observed during the aspirin treatment period was far lower than would have accrued without early termination. With the reduced number of end points, the study had good statistical power (95%) to detect a 50% reduction in the rate of total ARM but little power (<50%) to detect the most plausible 15% to 20% reduction in risk.

Incidence rates of confirmed ARM cases in the aspirin and placebo groups were calculated. Incidence rate was defined as the number of participants with ARM divided by the number of person-years of follow-up. An individual contributed person-time to total follow-up beginning at the time of randomization and continuing until the initial diagnosis of ARM or January 25, 1988, whichever came first. The relative risk (RR) of ARM in the aspirin group vs the placebo group was calculated using Cox proportional hazards regression models³⁶ adjusted for age and beta carotene treatment assignment. Because an extended duration of exposure to aspirin may be required to achieve an effect, the analyses were repeated excluding ARM cases diagnosed during the first year of treatment.

Stratified analyses and tests of interaction were performed to evaluate possible modification of the effect of aspirin on ARM by several baseline characteristics that have been suggested as possible risk factors, including cigarette smoking (ever or never); hypertension (systolic blood pressure ≥160 mm Hg or diastolic blood pressure ≥95 mm Hg or hypertension being treated currently) (yes or no); hypercholesterolemia (cholesterol level ≥261 mg/dL[≥6.75 mmol/L] or being treated currently) (yes or no); and alcohol use(daily, weekly, or rarely, with category values of 1, 2 and 3, respectively, for a test of linear trend over categories).

For each RR, 2-sided P values and 95% confidence intervals (CIs)³⁷ were calculated. Individuals, rather than eyes, were the unit of analysis because eyes were not examined independently, and participants were classified according to the status of the worse eye based on disease severity.

As expected in this large randomized trial, baseline characteristics that are possible risk factors for ARM were equally distributed between the 2 treatment groups (Table 1).

During an average of 60.2 months of follow-up, a total of 117 cases of ARM were confirmed by medical record review, including 57 cases with best-corrected visual acuity reduced to 20/30 or worse due to ARM. The most common manifestation of ARM for cases with vision loss (this information was not requested for cases without vision loss) was a combination of drusen and RPE changes, which was noted in 35% of cases (Table 2).

There were 51 cases of ARM in the aspirin group and 66 in the placebo group (RR, 0.77; 95% CI, 0.54-1.11; P = .16) (Table 3). The RR was similar when we considered only ARM cases with vision loss (RR, 0.78; 95% CI, 0.46-1.32; P = .36) (Table 4). The RRs tended to be lower in older men, but there was no statistically significant trend with age.

To assess a possible delayed effect of aspirin therapy, we also conducted analyses excluding ARM cases diagnosed during the first year of treatment. The RR estimates in these analyses were 0.80 (95% CI, 0.54-1.19; P = .27) for total ARM and 0.96 (95% CI, 0.54-1.70; P = .88) for ARM with vision loss, indicating no delayed effect of aspirin use during the 5-year treatment period.

In subgroup analyses, the effect of aspirin therapy on ARM did not differ markedly within categories of cholesterol, cigarette smoking, or alcohol use(Table 5). However, the effect of aspirin treatment seemed to differ according to reported hypertension at baseline. Among men who did not report hypertension at baseline, the RR was 0.95 (95% CI, 0.63-1.44; P = .82). For the subgroup of men who reported hypertension at baseline, there was a statistically significant 65% reduced risk of ARM (RR, 0.35; 95% CI, 0.15-0.83; P = .02), but the numbers were small.

Men assigned to low-dose aspirin treatment, compared with those assigned to placebo, had a statistically nonsignificant 23% reduced risk of total ARM during the 5-year treatment period. Findings were similar when we considered only cases of ARM that reduced visual acuity to 20/30 or worse. Thus, these randomized trial data tend to exclude any large beneficial effect of 5 years of low-dose aspirin treatment on ARM, although a smaller, but potentially important, beneficial effect cannot be ruled out.

Several possible limitations of the study merit consideration. The power of the study to detect any true benefit of aspirin treatment was limited by the small number of ARM end points, due largely to the early termination of the randomized aspirin component of the PHS I. In addition, the general good health of our study population might also have contributed to the small number of ARM end points. The total mortality rate in both groups combined during the 5-year aspirin treatment period of the PHS I was approximately 15% of that expected for a general population of white men of the same age distribution and period.³⁸ To the extent that mortality and ARM share common risk factors, a reduced number of ARM cases during this period might also be expected (thus limiting the interpretation of a comparison of incidence rates between this and other populations). With the reduced number of cases, the trial had sufficient power to detect a 50% reduction in risk but had insufficient power (<50%) to detect a risk reduction as small as that observed in the trial (20%-25%). Another possible limitation of the study is nonrandom or differential misclassification. This type of misclassification seems unlikely, however, because the reviewers of the medical records and the treating ophthalmologists and optometrists were all masked to the aspirin treatment assignment. Most study participants were aware of their aspirin treatment assignment when they were first requested to report on previous diagnoses of ARM (84-month questionnaire), but it seems unlikely that this knowledge would have materially affected the reporting of this end point. Random misclassification, which would underestimate any true effect of aspirin on ARM, was reduced by the use of medical records to confirm the self-reports of ARM. Confounding is unlikely in this large, randomized trial because, as expected, baseline characteristics were equally distributed between the aspirin and placebo groups. This equal distribution of risk factors provides reassurance that other potential confounders that were either unmeasured (eg, UV exposure^{39 - 40} ) or unknown were also likely to be evenly distributed between the 2 treatment groups. Finally, these findings from a population of healthy male physicians might not be generalizable to women or to the general male population.

In addition to its effect on study power, early termination of the trial also limited our ability to detect a possible delayed effect of aspirin treatment. Analyses that excluded cases of ARM occurring during the first year of treatment did not materially alter RR estimates. However, the average follow-up time of 5 years is relatively short, especially for a disease that is believed to develop slowly as a result of accumulated damage in the retina over several years. Detection of an effect of aspirin on ARM, if one exists, might require longer follow-up and exclusion of cases that develop during the first several years after initiation of therapy.

We found no evidence that the effect of aspirin treatment on ARM differed within baseline categories of cholesterol, cigarette smoking, or alcohol use. We did find, however, that aspirin treatment was associated with a significantly reduced risk of ARM among men who reported hypertension at baseline. Although these findings raise the possibility that aspirin therapy might have a differential impact in the subgroup of men with elevated blood pressure, results of subgroup analyses, particularly where there are no overall effects, must be interpreted with caution and might simply reflect the play of chance.

The few previous studies that examined aspirin use in ARM comprised patients with prevalent disease and were prompted out of concern for possible increased risk of hemorrhage for those who used aspirin. For example, the Macular Photocoagulation Study conducted an analysis of patients entering the study to determine whether those who regularly took aspirin had more hemorrhage than those who did not. The results of evaluations of 732 patients showed no difference between the 2 groups.⁴¹ That study also found no significant difference between users and nonusers of aspirin in the recurrence rate of macular degeneration after initially successful laser treatment.^{2 ,42} Several other anecdotal reports^{43 - 45} of increased risk of retinal hemorrhage associated with the use of aspirin or other antiplatelet agents in persons with ARM were limited by the absence of a comparison group and thus were impossible to interpret. Thus, our study seems to be the first to evaluate a possible beneficial effect of aspirin use in reducing risks of ARM. The Early Treatment Diabetic Retinopathy Study⁴⁶ examined a possible beneficial effect of aspirin therapy (650 mg/d) in diabetic retinopathy, and although the overall results of that trial indicated that aspirin use has no clinically important beneficial effects on the progression of retinopathy, the data also indicated no clinically important harmful effects either. Specifically, aspirin therapy had no effect on the occurrence,⁴⁶ or severity and duration,⁴⁷ of vitreous/preretinal hemorrhages and no effect on the occurrence of vitrectomy due to either vitreous hemorrhage or retinal detachment with or without vitreous hemorrhage.⁴⁸ In short, the Early Treatment Diabetic Retinopathy Study found no ocular contraindications to aspirin use in persons with diabetes mellitus (with mild to severe nonproliferative or early proliferative diabetic retinopathy) who require it for treatment of CVD or for other medical indications.

A consistent finding in histological studies^{14 ,49 - 51} of ARM is atrophy of the choriocapillaris, namely, a decrease in the number and diameter of capillaries in the macular region of the eye. Although some atrophy of the choroid occurs naturally with aging,^{51 - 52} the degradative process seems to be accelerated in ARM.⁵¹ A corresponding age-related decrease in choroidal blood flow^{53 - 55} also is accelerated in ARM^{15 ,56 - 58} and has recently been reported to be evident during the early, nonexudative stages of disease development.¹⁷ It remains unclear, however, whether this decreased choroidal perfusion is an important contributor to the development of ARM^{59 - 60} (and thus a potential avenue of disease prevention) or is merely a consequence of degradative changes in the outer retinal layers and RPE.^{61 - 64}

At the dose tested in the PHS I (325 mg every other day), aspirin irreversibly inhibits platelet cyclooxygenase, resulting in a rapid and marked inhibition of platelet function.^{33 - 34} An immediate decrease in platelet aggregability and risk of thrombosis is believed to underlie the protective effects of low-dose aspirin in CVD and could contribute to any beneficial effect that might be postulated for aspirin treatment in ARM. However, a more plausible mechanism in ARM might involve the long-term effects of platelet inhibition on initiation and progression of atherosclerosis. Platelet inhibition might limit platelet adherence or aggregation on vascular endothelium and existing plaque. In addition, activated platelets modify the chemotactic and adhesive properties of endothelial cells, which might be an important early pathophysiological event in atherogenesis.⁶⁵ Activated platelets also induce secretion of proinflammatory cytokines (eg, IL-8) that likely play an important role in vascular inflammation and atherogenesis⁶⁶ and can be stimulated to release substances that increase vessel wall permeability⁶⁷ and induce smooth muscle cell proliferation.⁶⁸ Thus, long-term inhibition of platelet function could reduce risks of ARM by altering the initiation and progression of atherosclerosis. In addition, because platelet activation frequently results from contact with subendothelial connective tissue exposed by endothelial denudation, our observation in subgroup analyses of a possible beneficial effect of aspirin use among those who reported hypertension at baseline is intriguing. Hypertension is believed to be a source of chronic injury to the arterial wall and is frequently associated with endothelial dysfunction and platelet activation. Although this subgroup finding might simply be a chance observation, it is also possible that men who reported hypertension at baseline might have been more likely to benefit from platelet inhibition than those who did not report hypertension. This possibility should be investigated in other populations.

In summary, these randomized trial data indicate that 5 years of low-dose aspirin treatment has no significant, large protective effect on the development of ARM. However, a smaller, but potentially important, beneficial effect could not be ruled out and requires testing in randomized trials of adequate power, such as the ongoing Women's Health Study of 39 876 female health professionals.⁶⁹

Accepted for publication January 12, 2001.

This study was supported by research grants HL 26490, HL 34595, CA 34944, CA 40360, and EY 06633 from the National Institutes of Health, Bethesda, Md.

Corresponding author and reprints: William G. Christen, ScD, Division of Preventive Medicine, Department of Medicine, Brigham and Women's Hospital, 900 Commonwealth Ave E, Boston, MA 02215-1204.