Study the past if you would define the future (Confucius)

Li et al¹ reported the results of an elegant study to estimate the prevalence and incidence of age-related macular degeneration (AMD) in the European Union (EU), where AMD is the leading cause of legal blindness and visual impairment. Importantly, the second objective was to predict future EU prevalence and incidence figures out to the year 2050.

HOW DID THEY DO IT?

The authors performed a systematic review and identified relevant articles in the literature: 22 prevalence but only four incidence studies. Since different studies used different AMD classification systems, the authors recategorised the data according to the Beckman Initiative for Macular Research system.² They then synthesised the data by meta-analysis to obtain prevalence and incidence estimates by AMD stage. Specifically, they used random-effects meta-analysis, which assumes that results vary between studies not just from random error but also from true variation (ie, ‘heterogeneity’, from differences in study protocols, regions, etc). Next, they performed random-effects meta-regression to investigate potential sources of heterogeneity.

In this way, the authors derived population prevalence estimates, expressed as percentages, for early/intermediate AMD, late AMD, neovascular AMD and geographic atrophy (GA). They reported incidence estimates in a similar way, but only for late AMD. To obtain estimates for the actual numbers of people with AMD, they applied their age-stratified prevalence and incidence estimates to current Eurostat population estimates. Similarly, they used Eurostat population projections to estimate how many EU individuals might be affected by AMD in the future.

WHAT WERE THE MAIN FINDINGS?

The current prevalence of early/intermediate AMD in EU adults aged 60 years and older was estimated at 25.3%, while that for late AMD was 2.4%. This corresponds to 67 million individuals with AMD, comprising 56.7 million with early/intermediate AMD and 10.2 million with late AMD. The annual incidence of late AMD was estimated at 1.4 per 1000 individuals, which corresponds to 400 000 new cases per year. The authors estimated that 77 million EU individuals will be affected by AMD in 2050, comprising 65.0 million with early/intermediate disease and 11.7 million with late disease. The incidence of late AMD was projected to reach 700 000 new cases per year in 2050. Importantly, these predicted increases are related purely to demographic changes in the EU, that is, ageing populations.

WHAT LIMITATIONS APPLY TO THIS STUDY?

Heterogeneity

Heterogeneity between the studies in the meta-analysis was very high. The I² measure (percentage of variance attributable to study heterogeneity) was 100% for early/intermediate AMD and 97% for late AMD. This decreases confidence that a combined estimate can be a meaningful description of the studies. However, the meta-regression steps helped address this: meta-regression by country and age decreased I² for late AMD to 20%, though it remained high for early/intermediate AMD at 95%. As the authors suggested, this may relate to study estimates differing widely depending on the classification system. However, meta-regression by classification system retained high heterogeneity for early/intermediate AMD (I² 99.7%). Indeed, for early/intermediate AMD, none of the characteristics analysed decreased heterogeneity substantially. Additional sources of heterogeneity are likely, for example, differences in AMD genetics, smoking and nutritional or other factors between study populations.

Overall, grading definitions for late AMD are relatively straightforward and consistent, leading to low heterogeneity and high confidence in the pooled estimates. By contrast, the boundaries around early AMD and intermediate AMD differ between classification systems; together with other protocol differences and genuine variation between countries, this was associated with greater heterogeneity.

Incidence analyses of late AMD

Only four studies were available, comprising just 7223 participants followed for a relatively short period (mean 3.3 years). The small number makes it difficult to ensure that they were representative of the EU population. Similarly, it makes it challenging to cover all strata, in terms of study region, time period and so on. For example, with two of the studies being from France, it is difficult to capture potential geographical differences. However, heterogeneity was lower than for the prevalence analyses.

Future projections

Both the age-stratified data and the population projections are estimates, so combining them is prone to error, which increases with longer time intervals. Also, the EU is experiencing a high degree of demographic change and population flux, and future events are difficult to predict. The estimates assume that age-specific prevalence and incidence percentages will remain stable. The authors have some justification, since they found no change in prevalence rates in studies initiated between 1982 and 2013. Similarly, a previous study did not find strong evidence of changes over time in age-specific AMD prevalence rates, in populations of European ancestry.³ A different meta-analysis suggested that the prevalence of early AMD in Europe has been stable in recent years, but that of late AMD has decreased.⁴ By contrast, data from the US Beaver Dam Eye Study suggested that the age-specific incidence of early/intermediate AMD has decreased substantially over recent generations,⁵ though the degree of participant age overlap was not ideal to make these comparisons.⁶ Finally, it remains possible that new preventative strategies will be discovered in the future. Hence, the projection estimates in the current study should perhaps be viewed as a benchmark scenario, accurate if we are unable to decrease the incidence of early AMD and of subsequent disease progression, through public health and medical interventions.

Eastern Europe

The terms Europe and the EU were often used interchangeably in this study, but the analysis pertained specifically to the EU. Notably, the most populous country in Europe is Russia, which is not in the EU. Even for eastern Europe, only two studies were identified in the systematic review, so the authors rightly pointed out that the estimates should be applied cautiously to eastern European countries.

IMPLICATIONS AND INSIGHTS

The estimates produced by this study are useful for two reasons: they have compelling public health implications and, through comparison with global data, may provide biological insights. The public health impact of 77 million EU individuals being affected by AMD in 2050 is profound. Hopefully, more effective prevention or longer-lasting treatment of neovascular AMD will exist by then, alongside new therapies that help prevent GA and slow its expansion. Clearly, more ophthalmologists will be required. Even so, without changes in care pathways, hospital eye services may struggle to meet this demand.

However, technology is likely to help. Artificial intelligence (AI) approaches to diagnosis, classification and prognostic tasks in AMD are already well underway as potential physician software aids.^7–13 In time, their role alongside humans will become established. Even in epidemiological studies of AMD, the prospect of combining smartphone-based telemedicine with AI-assisted disease classification may mean that future studies could be carried out on a massive scale. In this scenario, AI could provide an entirely consistent and uniform grading system that would address some of the issues highlighted in the meta-analysis.

Geographical differences between prevalence rates may provide clues to underlying disease mechanisms. Certainly, considerable variation in AMD prevalence remains between different populations, after accounting for age differences. AMD prevalence has been estimated at 12.3% in people with European ancestry, versus 10.4% (Hispanic), 7.5% (African) and 7.4% (Asian).¹⁴ The equivalent rates of early AMD were 11.2%, 9.9%, 7.1% and 6.8%, respectively, while those for late AMD were 0.5%, 0.3%, 0.3% and 0.4%. Hence, Europeans have substantially higher prevalence of early AMD, but only slightly higher prevalence of late AMD. This suggests that they are partially protected from progression of early to late disease, and/or those of Asian or African ancestry sometimes have disease characterised by more direct progression to late AMD (eg, polypoidal choroidal vasculopathy).

Notably, people of European ancestry have greatly higher prevalence of GA, at 1.1% versus 0.2% (Hispanic), 0.2% (Asian) and 0.1% (African).¹⁴ Understanding why these ancestries appear protected from GA could be important. Does it relate to genetic differences, such as protective variants in complement-related genes?¹⁵ In Asia, does it relate to high fish intake?¹⁶ As suggested recently,¹⁷ studying populations with low AMD prevalence despite prevalent risk factors may be informative. Protective environmental and diet-based approaches could be copied, and protective genetic profiles would demonstrate pathways for drug discovery.

CONCLUSIONS

This elegant study has provided important data with clear implications for public health. In the EU, in addition to expanded services for AMD, changes to infrastructure and care pathways will be required to meet increasing demand. Incorporating technology, including AI approaches, will be vital to maximise efficiency. This meta-analysis demonstrates the need for adoption of a standardised AMD classification system through international consensus. Ideally, this might incorporate an expanded set of phenotypes now recognised through multimodal imaging, such as reticular pseudodrusen and outer retinal atrophy.^18,19 However, it will also be important to capture heterogeneity in disease phenotypes between differences continents and ethnic groups. Indeed, beyond the EU, a shift in global patterns of AMD is occurring. The number of people with AMD world-wide is predicted to increase from 196 to 288 million by 2040.¹⁴ Notably, by this time, Asia is predicted to have over half of the world’s late AMD cases. AMD can no longer be considered primarily a European disease.