Aspirin Use and Risk of Cataract in Posttrial Follow-up of Physicians' Health Study I

Background  In Physicians' Health Study I, randomized trial results indicated no major beneficial effect of 5 years of low-dose aspirin treatment on total cataract (relative risk [RR], 0.94; 95% confidence interval [CI], 0.79-1.13) or cataract extraction (RR, 0.81; 95% CI, 0.65-1.01) during the period of treatment.

Objective  To examine the effect of assigned aspirin treatment and posttrial, self-selected aspirin use on the risk of age-related cataract over the 15 years of follow-up of Physicians' Health Study I.

Methods  Participants were 20 968 US male physicians enrolled in Physicians' Health Study I who did not report cataract at baseline. At 7 years, after termination of the randomized aspirin component of the trial, self-selected aspirin use was computed from annual questionnaires. The main outcome measures were age-related cataract and extraction of age-related cataract, defined as an incident, age-related lens opacity responsible for a reduction in best-corrected visual acuity to 20/30 or worse based on self-report confirmed by medical record review.

Results  During a median of 14.9 years of follow-up, there were 2081 cataracts and 1198 cataract extractions. Overall, the age- and beta carotene–adjusted RR of cataract in men assigned to aspirin compared with those assigned to placebo was 1.09 (95% CI, 1.00-1.18). For cataract extraction, the RR was 1.09 (95% CI, 0.98-1.22). During a median posttrial follow-up of 7.9 years, a total of 1225 incident cataracts and 635 cataract extractions were documented. The multivariate RR of cataract in men who reported using aspirin frequently(≥180 days per year) at 7 years compared with nonusers (0-13 days per year) was 1.20 (95% CI, 1.03-1.40). For cataract extraction, the multivariate RR was 1.22 (95% CI, 0.98-1.51). Results for diagnosis and extraction of cataract subtypes were similar.

Conclusions  Analyses based on randomized aspirin assignment indicated no long-term benefit of 5 years of low-dose aspirin treatment on total cataract or cataract extraction. Posttrial, observational data also indicated no decreased risk of cataract in aspirin users and suggested a small increased risk of cataract in aspirin users. Further randomized trial data to investigate the effect of longer term treatment with low-dose aspirin are being collected as part of the ongoing Women's Health Study, a randomized trial of low-dose aspirin and vitamin E among 39 876 apparently healthy, postmenopausal US female health professionals.

A POSSIBLE ROLE for aspirin in reducing risks of cataract was first suggested in studies of patients with arthritis or diabetes mellitus^1,2 and in a study of patients undergoing cataract surgery.³ Most subsequent observational epidemiologic studies, however, have generally shown no association between self-selected aspirin use and cataract.^4-14 Results from 2 of 3 randomized trials that included an evaluation of the aspirin-cataract hypothesis also indicate no apparent benefit of aspirin in reducing risks of cataract. These trials include an investigation of 5139 British physicians followed for an average of 6 years¹⁵ and a study of 3711 participants in the Early Treatment Diabetic Retinopathy Study followed for 5 to 7 years.¹⁶

In the third trial, comprising 21 000 participants in the Physicians' Health Study I (PHS I), results excluded a large benefit of 5 years of low-dose aspirin treatment on cataract development during the period of treatment but could not rule out a modest benefit on cataract extraction.^17,18 Subgroup analyses also indicated a modest protective effect of low-dose aspirin on posterior subcapsular cataract, a particularly disabling subtype. Interpretation of these results, however, is limited by the smaller than expected number of cataract end points, a consequence of the early termination of the randomized aspirin component of PHS I after 5 years of treatment and follow-up due primarily to a statistically extreme benefit on the risk of a first myocardial infarction(MI).¹⁹

In this article we present the results of posttrial follow-up for cataract in PHS I. Specifically, we examine possible long-term effects of 5 years of randomized aspirin treatment by extending the intention-to-treat analysis to include cataracts and cataract extractions documented during 15 years of follow-up. We also examine the relation of self-selection of aspirin use at the end of the randomized aspirin component of the trial with the risk of subsequent cataract and cataract extraction.

A detailed description of the subjects and methods of PHS I has been presented elsewhere.^20,21 Briefly, PHS I was a randomized, double-blind, placebo-controlled trial of aspirin, 325 mg (Bufferin, Bristol-Myers Products, New York, NY), and beta carotene, 50 mg (Lurotin, BASF Corp, Mt Olive, NJ), in reducing risks of cardiovascular disease and cancer among 22 071 US male physicians aged 40 to 84 years in 1982. Eligible participants had no history of MI, stroke, transient cerebral ischemia, or cancer (except nonmelanoma skin cancer). Baseline information included height, weight, history of cigarette smoking, history of alcohol use, blood pressure, cholesterol levels, history of diabetes mellitus, history of arthritis, and history of multivitamin use. Information on a personal history of cataract was also obtained at baseline. Annual follow-up questionnaires were sent to all participants asking about their compliance with assigned treatment and the occurrence of any relevant events, including cataract.

On January 25, 1988, the aspirin component of PHS I was terminated early, primarily because of a statistically extreme benefit on the risk of a first MI.²⁰ At that time, the average length of follow-up was 60.2 months. The beta carotene component of the trial continued uninterrupted until its scheduled completion on December 31, 1995. Since that date, study participants have been followed as an observational cohort, and we have continued to document all study end points, including cataract. In this report, we include all incident cataracts and cataract extractions that were documented to have occurred before December 2, 1997, with a median follow-up for cataract of 14.9 years (interquartile range, 14.1-15.1).

For the analysis of long-term effects of 5 years of aspirin treatment, we used the initial aspirin and placebo groups as assigned by randomization and compared the incidence of cataract over the 15-year follow-up period.

For the posttrial observational analyses, we compared the incidence of cataract according to self-selected frequency of aspirin use following the termination of the aspirin component of the trial on January 25, 1988. After that date, all physicians were asked whether they preferred active aspirin or placebo to be included in their calendar packs. Detailed information on self-selected use of aspirin and other platelet-active drugs in the posttrial period continued to be collected on annual follow-up questionnaires.

The primary exposure in these observational analyses was aspirin use as reported on the 7-year follow-up questionnaire since all participants had entered the posttrial period (past January 25, 1988) by this time. On this questionnaire, participants were asked how many days they had taken the white pills from their calendar packs over the past 12 months, with possible answer categories of 0, 1-13, 14-30, 31-60, 61-90, 91-120, 121-180, and 180+ days. Participants were also asked on how many days they had taken additional aspirin or medication containing aspirin, not including the white pills in the calendar packs, using the same response categories. Total aspirin use was estimated from both the white pill and reported outside use. Categories were then collapsed into 4 groups (0-13, 14-120, 121-179, and ≥180 days in the past year) and 2 groups (<180 vs ≥180 days in the past year) of aspirin use.

Following the report of a cataract diagnosis or extraction, a written consent identifying the treating ophthalmologist or optometrist was obtained. Ophthalmologists and optometrists were contacted by mail and requested to complete a cataract questionnaire supplying information about the presence of lens opacities, date of diagnosis, visual acuity loss, cataract extraction, other ocular abnormalities that could explain visual acuity loss, cataract type, and origin (including age-related, traumatic, congenital, inflammatory, or surgery- or steroid-induced). Ophthalmologists and optometrists were given the option to provide the requested information by supplying copies of the relevant medical records. Medical records were obtained for more than 92% of participants reporting cataract.

End points included incident cataract and extraction of incident cataract. Cataract was defined as a self-report confirmed by medical record review to be initially diagnosed after randomization, age-related in origin (congenital cataracts and those due to trauma, steroids, intraocular inflammation, or surgery were excluded), with best-corrected visual acuity of 20/30 or worse and with no alternate ocular disease to explain the visual acuity loss. In the presence of alternate ocular disease, a lens opacity was considered a cataract if in the judgment of the ophthalmologist or optometrist the opacity was of sufficient severity to reduce visual acuity to 20/30 or worse when considered alone. Extraction was defined as the surgical removal of an incident cataract. Three cataract subgroups were also defined; any nuclear sclerosis(NS), any cortical opacity, and any posterior subcapsular component (PSC). Information on cataract subtypes was available for 99.1% of confirmed cases.

Analyses of the effect of 5 years of randomized aspirin treatment on the occurrence of cataract over the entire 15-year follow-up period were based on person-years from the time of randomization to the time of cataract or December 2, 1997, whichever was earliest, in an intention-to-treat analysis. These analyses included the 20 968 physicians who did not report a diagnosis of cataract at baseline. Incidence rate ratios and 95% confidence intervals(CIs) were calculated using Cox proportional hazards models that adjusted for age (years) and beta carotene treatment assignment.²² Analyses were also conducted that examined the occurrence of cataract in consecutive 2-year periods. An interaction term with length of follow-up was used to test for a trend of the rate ratio over time and to evaluate the adequacy of the proportional hazards assumption over time. Data were analyzed separately for cataract diagnoses, extractions, and subtypes.

We used an observational analysis to compare the occurrence of cataract with frequency and duration of aspirin use in 19 228 physicians who were alive and free of a confirmed diagnosis of cataract at the time of the year-7 questionnaire. Self-selected aspirin use between year 6 and year 7, as reported on the year-7 questionnaire (described in the "Observational Analyses" section in the "Definition of Aspirin Use" section), was considered the primary exposure variable. Cataract end points accrued from follow-up year 7 through December 2, 1997, a median follow-up of 7.9 years (interquartile range, 7.5-8.1 years). To assess the effect of the duration of aspirin use, participants were further grouped according to whether they had been randomly assigned to receive aspirin in the previous 5 years. Aspirin use after the year-7 questionnaire was not considered.

The Mantel Haenszel χ² trend test and linear regression were used to assess the association of participant characteristics with ordered categories of self-selected aspirin use at 84 months. Cox proportional hazards models were used to calculate the relative risk (RR) of cataract and 95% CIs while adjusting for age (years); smoking (never, past, current); alcohol use(daily, weekly, monthly, rarely); history of diabetes; history of hypertension(systolic blood pressure of 160 mm Hg or greater, diastolic blood pressure of 95 mm Hg or greater, or history of treatment for high blood pressure); history of high cholesterol levels (reported high cholesterol, reported blood cholesterol levels of 6.72 mmol/L [260 mg/dL] or higher, or history of treatment with cholesterol-lowering medication); history for arthritis, obesity (body mass index [calculated as weight in kilograms divided by the square of height in meters] of 27.8 or higher), and physical activity (reported vigorous exercise once per week or more); and parental history of MI. All of these variables were assessed at the 7-year follow-up except obesity and parental history of MI, which were assessed at baseline. We also adjusted for cardiovascular end points (MI, ischemic stroke, hemorrhagic stroke, transient cerebral ischemia, coronary revascularization, and angina) that were documented to have occurred prior to the 7-year follow-up. Interaction terms were used to test for possible interaction of randomized aspirin assignment and posttrial aspirin use in Cox regression models.

In all analyses, for each RR, we calculated the 2-sided P value and 95% CI.²³ 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 as defined by the occurrence of cataract surgery or, in the absence of cataract surgery, by an earlier date of diagnosis. When the 2 eyes had the same date of diagnosis, the eye with the worse visual acuity at the most recent eye examination was designated the worse eye. When the worse eye was excluded because of visual acuity loss attributed to other ocular abnormalities or a cause that was not age-related, the fellow eye was considered for classification.

The baseline characteristics of the aspirin and placebo treatment groups were virtually identical (Table 1). To examine the long-term effects of 5 years of randomized aspirin treatment, we extended the intention-to-treat analysis to include all cataracts and cataract extractions documented during the 15 years of extended follow-up. During this period, 2081 cataracts and 1198 cataract extractions were documented.

There were 1084 cataracts in the aspirin group and 997 in the placebo group. The overall RR of cataract, comparing men assigned to aspirin with those assigned to placebo, was 1.09 (95%CI, 1.00-1.18). For cataract extraction, 627 were documented in the aspirin group and 571 in the placebo group. The RR comparing the aspirin and placebo groups was 1.09 (95% CI, 0.98-1.22).

Figure 1 shows the number of cataracts (Figure 1A) and cataract extractions (Figure 1B) in the aspirin and placebo groups, and the RRs and 95% CIs for random assignment to aspirin in consecutive 2-year periods and over the total follow-up period. For both cataract and cataract extraction, the 2-year RRs show no clear trend over time and, except for one interval for cataract extraction, their CIs always include 1. Thus, there was no evidence of a long-term benefit of 5 years of aspirin treatment on total cataract or cataract extraction.

Results for specific subtypes also indicated no long-term benefit of 5 years of aspirin treatment. Among the 2081 confirmed cataracts, 1836 (88.2%) had NS, 844 (40.6%) had a PSC, and 781 (37.5%) had a cortical opacity. The RRs for the diagnosis of subtypes, comparing the aspirin and placebo groups, were NS, 1.11 (95% CI, 1.01-1.21); PSC, 1.03 (95% CI, 0.90-1.18); and cortical opacity, 1.10 (95% CI, 0.96-1.27). The RRs for extraction of subtypes were NS, 1.10 (95% CI, 0.98-1.24) (n = 1076); PSC, 1.04 (95% CI, 0.89-1.22) (n= 623); and cortical opacity, 1.27 (95% CI, 1.06-1.52) (n = 471). The patterns of RRs over time for diagnosis and extraction of subtypes (data not shown) were broadly similar to those for total cataract and total cataract extraction, and none of the trends over time were statistically significant.

After the end of the aspirin component of the randomized trial, 72% of participants chose to take aspirin more than 120 days per year, and 61% reported taking aspirin at least 180 days per year (66% of those who had been randomized to the aspirin group reported taking aspirin ≥180 days per year compared with 55% of those who had been randomized to the placebo group). Men who reported more frequent use of aspirin were slightly older and were more likely to have been randomly assigned to aspirin (Table 2). They also reported more alcohol use, hypertension, high cholesterol levels, multivitamin use, and physical activity. They were also more likely to report a family and personal history of MI, coronary bypass surgery, and angina. Frequent users of aspirin were less likely to report a history of hemorrhagic stroke and history of arthritis.

Following the completion of the aspirin arm of the study, 1225 incident cataracts and 635 cataract extractions were documented during a median follow-up of 7.9 years. Men who reported use of aspirin for at least 180 days per year compared with nonusers (0-13 days per year) had a statistically significant 20% increased risk of total cataract after adjustment for age and beta carotene treatment assignment (Table 3). Risks were not increased for men in the 2 intermediate groups of aspirin use. Regarding subtypes, men who reported using aspirin at least 180 days per year when compared with nonusers had significantly increased risks of NS (RR, 1.22; 95% CI, 1.04-1.43), PSC (RR, 1.33; 95% CI, 1.04-1.70), and cortical opacity(RR, 1.61; 95% CI, 1.23-2.11) after adjustment for age and beta carotene treatment assignment. These estimates changed little after further adjustment for other cataract risk factors.

For total cataract extraction, men in the 3 higher categories of aspirin use had nonsignificant 10% to 20% increased risks of cataract extraction after adjustment for age and beta carotene treatment assignment (Table 4). The test of linear trend over categories of aspirin use also failed to attain significance (P = .08). In analysis of subtypes, men in the 3 higher categories of aspirin use had an approximate 2-fold increased risk of extraction of cortical cataract compared with nonusers. Adjustment for other possible risk factors for cataract had no material impact on these RR estimates.

To examine the effect of duration of aspirin use, we repeated these observational analyses stratifying on randomized aspirin treatment assignment in the years before the self-selected use. A 2-level variable for self-selected aspirin use in the posttrial period (≥180 vs <180 days per year) was used to conserve statistical power. As given in Table 5 and Table 6, RRs for total cataract and total cataract extraction during the posttrial period were generally higher among those who had been randomly assigned to the aspirin group vs the placebo group. This pattern was observed for both total cataract and cataract extraction, as well as for specific subtypes. However, none of the interactions of randomized aspirin assignment and posttrial aspirin use attained statistical significance.

The previously published final results for the randomized aspirin component of PHS I, based on 501 cases of cataract diagnosed during the 5-year treatment period, indicated that 5 years of low-dose aspirin therapy was unlikely to have a major benefit on total cataract (RR, 0.94; 95% CI, 0.79-1.13) during the period of treatment.¹⁸ The data were consistent, however, with a modest reduction in the risk of cataract extraction (RR, 0.81; 95% CI, 0.65-1.01), as well as a possible beneficial effect of aspirin on the PSC subtype (RR, 0.74; 95% CI, 0.57-0.98). In the present report, we examined a possible delayed effect of aspirin treatment by extending the intention-to-treat analysis to include all 2081 cases of cataract and 1198 cataract extractions documented during the 15 years of extended follow-up. We found evidence neither for a long-term benefit of 5 years of low-dose aspirin treatment on total cataract or cataract extraction nor for a beneficial effect on any of the cataract subtypes, including posterior subcapsular cataract. Our observational analyses of posttrial, self-selected aspirin use also failed to indicate a decreased risk of cataract in aspirin users and suggested a possible increased risk for aspirin users. Together with the results of completed trials, these data indicate that low-dose aspirin treatment for up to 6 years is unlikely to alter materially the risk of cataract during the period of treatment or in the years following this duration of aspirin treatment.

Several possible limitations of this study need to be considered. Random misclassification of the cataract end point was reduced by the use of medical records to confirm the self-reports and by the use of strict diagnostic criteria that included reduction in best-corrected visual acuity to 20/30 or worse due to cataract. Nonrandom misclassification was unlikely in the randomized comparisons since medical records were reviewed without knowledge of aspirin treatment assignment, and study participants and treating ophthalmologists and optometrists were unaware of aspirin treatment assignment. In the observational analyses, however, nonrandom misclassification would have occurred if knowledge of the participant's self-selected aspirin use influenced a treating physician's likelihood of diagnosing cataract or, more plausibly, if aspirin users had more medical contacts, and thus were more likely to have an existing cataract diagnosed. Finally, confounding is unlikely in the randomized comparisons since, as expected, baseline characteristics were equally distributed between the aspirin and placebo groups (Table 1). In the observational analyses, men who elected to use aspirin differed in several respects from nonusers of aspirin (Table 2). Although we did control for a number of measured confounders, other potential confounders that were either unmeasured or unknown may have contributed to the findings of the observational analyses.

Our study is the first to examine the possible long-term effects of randomized aspirin treatment on cataract. The totality of evidence from 3 completed randomized trials that evaluated the aspirin-cataract hypothesis clearly show no major beneficial effect of 5 to 6 years of aspirin therapy over a range of doses during the period of treatment. In addition to PHS I,^17,18 these trials include a study of 5139 British physicians who were allocated to 500 mg of aspirin daily or placebo and followed for an average of 6 years¹⁵ and a study of 3711 diabetic patients who were assigned to 650 mg of aspirin daily or placebo and followed for an average of 5 years.¹⁶ The present report extends our previous findings from PHS I by showing that 5 years of randomized aspirin therapy has no beneficial effect on the rate of cataract development during the posttrial period. Whether longer duration aspirin treatment can materially reduce risks of cataract during and following the period of treatment remains unknown. Age-related cataract develops slowly as a result of accumulated biochemical and biophysical damage in the lens over a number of years,²⁴ and the average length of treatment of 5 to 6 years in these trials may have been insufficient to materially reduce risks of cataract.

The results of posttrial, observational analyses of self-selected aspirin use in PHS I suggest a small increased risk of cataract for those who use aspirin regularly. These findings contrast somewhat with the trial results for PHS I, which ruled out a large benefit but were consistent with a possible small decreased risk of cataract extraction and the PSC subtype among those assigned to aspirin. Reasons for the apparently discrepant findings are unclear. In the observational analyses, adjustment for a range of cataract risk factors had little impact on RR estimates, suggesting that residual confounding is unlikely to explain the excess risk. However, unmeasured and therefore uncontrolled confounding (eg, steroid use), as well as confounding by indication may have contributed to the findings. It is also possible that the apparent excess risk in the observational analyses is due, at least in part, to surveillance bias. Those who chose to use aspirin regularly tended to be older and to report more hypertension, elevated cholesterol levels, personal history of coronary artery disease, and parental history of MI. These men may also have had more medical contacts (we have no information on the number of medical contacts for study participants) and may therefore have been more likely to have an existing cataract detected. None of these possibilities, however, can explain the finding that RRs associated with self-selected aspirin use in the posttrial period were generally higher for those randomized to aspirin during the treatment period (Table 5 and Table 6). If real, this finding would suggest a deleterious effect of long-term aspirin treatment on risk of cataract. Finally, it is possible that the small excess risk observed in the observational analyses and the modest but nonsignificant beneficial effect found for aspirin during the randomized treatment period^17,18 are simply a result of chance.

Experimental evidence supports a possible protective effect of aspirin against cataract. Plausible mechanisms include acetylation of lens proteins and inhibition of glycation,^25-27 improved glucose tolerance,^28-30 and an indirect antioxidant effect.³¹ To the best of our knowledge, however, there is no experimental evidence of a mechanism through which aspirin might accelerate the rate of cataract development.

In summary, these results from a large cohort of US male physicians indicate that 5 years of low-dose aspirin treatment has no long-term beneficial effect on risks of cataract over an extended follow-up period of 15 years. If aspirin therapy is capable of reducing risks of cataract, it seems that a duration of treatment in excess of the 5 to 6 years tested in completed trials will be required before a beneficial effect can emerge. Toward this end, important new data to address this and other vision-related hypotheses are being collected as part of the ongoing Women's Health Study, a randomized, double-blind, placebo-controlled trial of aspirin and vitamin E in the prevention of cancer and cardiovascular disease, conducted among 39 876 apparently healthy, postmenopausal US female health professionals.

Accepted for publication July 26, 2000.

Corresponding author and reprints: William G. Christen, ScD, 900 Commonwealth Ave E, Boston, MA 02215-1204.