The association between aspirin use and neovascular age-related macular degeneration (AMD) has garnered much attention in the lay press and led to many patients with AMD asking their ophthalmologists about the advisability of continued aspirin use. This commentary seeks to critically assess the available literature on the topic and provide guidance for physicians discussing the subject with patients.
In a February 2013 issue of JAMA Internal Medicine , researchers from the Blue Mountains Eye Study examined data on 2389 participants who were followed prospectively for up to 15 years with fundus photographs that were graded for AMD. Liew and associates found that persons who were regular aspirin users, defined as aspirin use at least once per week, had an approximately 2.5-fold higher risk of developing neovascular AMD (odds ratio [OR] = 2.46, 95% confidence interval [CI] = 1.25-4.83) after adjustment for confounders including cardiovascular disease, body mass index, blood pressure, age, and sex. The 15-year cumulative incidence of neovascular AMD was 9.3% in users and 3.7% in nonusers, a relatively small magnitude of risk. They did not find an association between aspirin use and incident geographic atrophy (GA). These findings were consistent with a cross-sectional study published in 2012, which found that daily aspirin users were more likely to have neovascular AMD (OR = 2.22, 95% CI = 1.61-3.05). Both studies found a dose-response relationship, with more frequent aspirin use associated with greater risk. A third study published in JAMA in 2012 also found an association between regular aspirin use 10 years prior to retinal examination and incident neovascular AMD (hazard ratio [HR] = 2.20, 95% CI 1.20-4.15) but not GA.
The strengths of the study by Liew and associates include the large size of the cohort and the quality of risk factor and outcome measurement. Aspirin use was ascertained in a structured interview using a questionnaire. Serial fundus photography was performed and photographs were graded for AMD using standardized criteria. There was detailed information available on potential demographic and clinical confounders. The longitudinal design allows for a temporal association to be claimed; the exposure to aspirin occurred before the development of AMD.
However, there are some limitations. The main limitation of the study is that of any nonrandomized study of a potential therapeutic drug: the possibility for residual confounding from variables that are not included in the analysis. For example, regular aspirin use could be related to increased systemic inflammation, whether that be manifested as cardiovascular disease or musculoskeletal pain, and several lines of evidence indicate that inflammation is a prominent part of AMD pathophysiology. If systemic inflammation was not controlled for in the analyses, then aspirin would seem to be associated with AMD when in reality it might only be a “marker” for the true causal agent, systemic inflammation. The authors were diligent in trying to minimize confounding by examining models with additional covariates (including inflammatory markers) and using propensity score adjustment, but it is not possible to eliminate confounding completely in a nonrandomized study because unmeasured confounding variables can always exist. Also, data were missing for 44% of the cohort for follow-up greater than 15 years, and this could lead to sampling bias. Sampling bias would occur if those participants who were lost to follow-up differed systematically from the rest of the sample. For sampling bias to contribute to a false-positive association in this study, the participants lost to follow-up would have to include a disproportionate number of people who were either (1) more likely to use aspirin and less likely to develop AMD or (2) less likely to use aspirin and more likely to develop AMD than that observed in the population who completed follow-up. Although either of these scenarios is possible, there is evidence that patients with cardiovascular disease are more likely to be lost to follow-up because of increased mortality but are also more likely to be aspirin users and more likely to develop AMD. If this was the case for the Liew and associates study, then the missing data would have actually led to an underestimation of the observed association between aspirin and incident AMD. Recall bias occurs when there is intentional or unintentional recall, and thus reporting, about the exposure of an association in one outcome group compared to the other. Aspirin use was determined based on the questionnaire administered at the baseline visit and before the occurrence of the outcome of neovascular AMD. Generally this design should reduce the likelihood that participant recall bias would significantly affect the results.
The recent findings also need to be evaluated in the context of previous studies on this subject ( Table ). The results have been inconsistent and there are many reasons that could explain these disparities. In particular, the 2 prospective randomized trials, which represent the most rigorous study design to examine this relationship, have not found an association between aspirin use and AMD. These 2 trials did not ascertain AMD directly through grading of fundus photographs but rather used medical record review to confirm the diagnosis. The period of aspirin exposure was also shorter than that which has been associated with neovascular AMD in the longitudinal cohorts. The association between aspirin use and AMD has been primarily observed for neovascular AMD and has not been seen for GA. The 2 randomized trials did not look at these 2 advanced subtypes of AMD separately. Aspirin dose has not been documented in any of the studies except for the randomized controlled trials. Differences in aspirin dose could have an effect on the association. Finally, differences in the patient populations in each study and their underlying AMD risk could also account for some of the variability in findings.
|Study||Design||Follow-up (y)||Number of Events||Aspirin Ascertainment||AMD Ascertainment||Outcome Examined||Point Estimate (95% CI)||P Value|
|Klein et al, Arch Ophthalmol 2001||Population-based incidence study||5||229 incident cases of ARM||Interview with questionnaire; baseline aspirin used for analyses; doses not recorded||Fundus photographs||Incidence of early ARM||OR = 1.08 (0.79-1.48)||NS|
|Christen et al, Arch Ophthalmol 2001||Randomized, placebo-controlled trial||5||117 incident cases of ARM||Randomized to aspirin 325 mg every other day vs placebo||Self-report with medical record confirmation||Incidence of any ARM||RR = 0.77 (0.54-1.11)||NS|
|AREDS Study Research Group, Ophthalmology 2005||Prospective cohort study||5||534 cases of neovascular AMD; 223 cases of GA||Interview with questionnaire; baseline aspirin used for analyses; doses not recorded||Fundus photographs||Incident neovascular AMD and GA||Bilateral drusen at baseline: OR = 1.30 for neovascular AMD and OR = 1.00 for GA. |
Unilateral advanced AMD in 1 eye at baseline: OR = 0.99 for neovascular AMD and OR = 0.25 for GA.
|Christen et al, Ophthalmology 2009||Randomized, placebo-controlled trial||10||245 incident cases of visually significant AMD||Randomized to aspirin 100 mg every other day vs placebo||Self-report with medical record confirmation||Incidence of visually significant AMD (BCVA 20/30 or worse)||HR = 0.82 (0.64-1.06)||NS|
|Rudnicka et al, Eye (Lond) 2010||Case-control||N/A||81 cases of late AMD (GA or neovascular AMD)||Interview with questionnaire; doses not recorded||Fundus photographs||Late AMD (both GA and neovascular AMD)||OR = 0.61 (0.23-1.57)||NS|
|Klein et al, JAMA 2012||Longitudinal, population-based study||10||512 incident cases of early AMD and 117 incident cases of late AMD||Interview with questionnaire; aspirin use from serial visits used for time-varying covariate analyses; doses not recorded||Fundus photographs||Incident late AMD||Neovascular AMD: HR = 2.20 (1.20-4.15) GA: HR = 0.66 (0.25-1.95)||Neovascular AMD: P = .01 |
GA: P = .46
|De Jong et al, Ophthalmology 2012||Population-based, cross-sectional study||N/A||1706 early AMD cases; 108 neovascular AMD cases; 49 GA cases||Interview with questionnaire; doses not recorded||Fundus photographs||Neovascular AMD||2.22 (1.61-3.05)||<.001|
|Liew et al, JAMA Intern Med 2013||Longitudinal, population-based cohort||15||63 incident cases of neovascular AMD||Interview with questionnaire; baseline aspirin used for analyses; doses not recorded||Fundus photographs||Neovascular AMD||2.46 (1.25-4.83)||.01|
|Cheung et al, Br J Ophthalmol 2013||Population-based, cross-sectional study||N/A||179 cases of early AMD||Interview with questionnaire; baseline aspirin used for analyses; doses not recorded||Fundus photographs||Early AMD||1.38 (0.89-2.14)||NS|