Identifying patients at high risk of late AMD, particularly neovascular AMD, is important from both public health and clinical perspectives as this would facilitate detection of disease before the onset of irreversible visual loss enabling earlier intervention. Of the 16 risk factors identified in our systematic review and meta-analysis, age, current smoking, cataract surgery, and potentially family history were strongly and consistently associated with late AMD. All of these are easily assessed through discussions with patients and do not entail a lengthy medical history taking or laboratory evaluations. Other significant factors with a lower strength of association (risk estimates generally 1.5 or less) were BMI, hypertension, a history of cardiovascular disease and plasma fibrinogen. All of these factors are associated with cardiovascular disease and are also likely to be measured and monitored in the primary care setting.
Our meta-analysis confirmed the increased risks of late AMD associated with advancing age (especially for the oldest age groups of 80 years and over), current cigarette smoking, and previous cataract surgery. The relationship of cataract surgery and late AMD was previously reported in a pooled analysis of 3 studies in different continents [26
]. Although this association could reflect shared risk factors and the fact that both are diseases that affect the ageing eye, there is concern that surgery may predispose the operated eye to the development of neovascular AMD. We further confirmed the association with family history of AMD, consistent with the growing recognition of major genes involved in AMD (e.g., complement factor H and C3, LOCS 3877, HTRA 1) [21
Vascular diseases, including myocardial infarction, stroke, angina, and hypertension are thought to be pathogenic factors for the development of late AMD. Our meta-analysis showed that while the magnitude of the ORs were inconsistent across studies, the pooled estimates for case-control studies were statistically significant for both cardiovascular disease (OR 2.20; 95% CI 1.49 - 3.26) and hypertension (OR 1.48, 95% CI 1.22 - 1.78). The association of diabetes and late AMD is also less consistent, and while prospective studies showed a significant association, this was not significant in the cross-sectional or case-control studies. The relationship between higher BMI and late AMD could be explained by shared risk factors (e.g., hypertension), or potential unmeasured confounders (e.g., nutritional factors). Our analysis also indicated that, based only on US data, people of European origin were at increased risk compared to other ethnic groups but at present the evidence for risks associated with other specific ethnic groups is inadequate. These are areas of future research.
Our study has several limitations. Although we selected only studies that reported some adjustment for confounding factors, we could not ascertain the appropriateness or completeness of adjustment in the studies. Second, the data included in some studies may have been too crude and also subject to measurement error. For example, a 'history of cardiovascular disease' may encompass a spectrum of conditions, from asymptomatic angina to myocardial infarction; this was not often specified. Such potential measurement errors would likely dilute effects in the meta-analysis. Third, we did not consider ocular factors (e.g., presence of large drusen), that have been found to be strong predictors of progression to late AMD, as we felt they could not be easily ascertained by family physicians. Neither did we consider dietary factors, such as the consumption of vegetables rich in carotenoids (lutein and zeaxanthin) or zinc and antioxidant vitamin supplementation, fish or omega-3 fatty acids in the diet, or glycemic index, because the methods for estimating such risk factors in the setting of primary health care is difficult without access to specialised personnel and resources [6
]. Fourth, despite strong associations of AMD with genetic factors, we felt that genetic testing was not readily available in general clinical practice. Fifth, we included only English language articles because a preliminary assessment did not identify any non-English language articles that fitted our inclusion criteria. Nonetheless, many studies included in this review were from non-white populations (e.g., Chinese, Malay Asians) and thus, we believe our results can be generalisable to different populations in different countries around the world. Sixth, we only included articles that were identified in the Medline and Cochrane databases. Expanding the search to EMBASE may have identified additional articles; however, given the extensive hand searching of bibliographies and the experience of the authors we feel it is unlikely any relevant articles were missed. Seventh, the assessment of the quality of the publications was performed as part of this study to provide supplementary evidence of the internal and external validity of the data. However, ultimately, we decided to present the data based on the type of study design reported in the publication: cohort, cross sectional and case control. Large, epidemiological, cohort studies had the advantage over other study designs in that they removed any temporal or causal ambiguity as the exposure precedes the disease and if follow-up is not biased selection bias is less of a problem than in other study designs [90
]. Finally, whether the study findings could be used as prognostic information to refer patients with higher risk of AMD requires further research. The observed odds ratios were generally small, and there are limited interventions to prevent AMD.