Geographic Atrophy in Age-Related Macular Degeneration

Abstract

Purpose

To examine disease progression in age-related macular degeneration (AMD) at 2 distinct stages, progression to geographic atrophy (GA) versus GA expansion, by comparison of the risk and protective factors at each stage.

Design

Perspective.

Subjects

Individuals at risk of GA or with GA.

Main Outcome Measures

Progression to GA and GA expansion rate.

Methods

Critical synthesis of the literature on risk and protective factors, both environmental and genetic, for progression to GA versus GA expansion in AMD.

Results

Comparison of the risk and protective factors demonstrates partially overlapping but partially distinct risk and protective factors for progression to GA versus GA expansion. Some factors are shared (i.e., operating in the same direction at both stages), others are not shared, and others seem to operate in different directions at each stage. Risk variants at ARMS2/HTRA1 increase both risk of progression to GA and GA expansion rate, presumably through the same mechanism. By contrast, risk and protective variants at CFH/CFHR alter risk of GA but not GA expansion rate. A risk variant at C3 increases risk of GA but is associated with slower GA expansion. In environmental factors, cigarette smoking is associated with increased risk of GA and faster GA expansion, whereas increased age is associated with the former but not the latter. The Mediterranean diet is associated with decreased progression at both stages, although the food components with the largest contributions seem to differ between the 2 stages. Some phenotypic features, such as reticular pseudodrusen and hyperreflective foci, are associated with increased progression at both stages.

Conclusions

Analysis of the risk and protective factors for progression to GA and GA expansion demonstrates partially overlapping but partially distinct elements at each stage: some are shared, some are relevant to 1 stage only, and some even seem active in opposite directions at each stage. Aside from ARMS2/HTRA1, the overlap between the genetic risk factors for the 2 stages is minimal. This suggests that the biologic mechanisms differ at least partially between the 2 disease stages. This has implications for therapeutic approaches and suggests that treatment aimed at the underlying disease processes may need to be tailored by stage.

Financial Disclosure(s)

Proprietary or commercial disclosure may be found after the references.

“I tell thee,” said madame …, “that although it is a long time on the road, it is on the road and coming. I tell thee it never retreats, and never stops. I tell thee it is always advancing.”¹

In age-related macular degeneration (AMD), the development of geographic atrophy (GA) differs partially but fundamentally from its subsequent expansion: in its time-course, its risk factors, and its underlying mechanisms. Across the life cycle of GA, its conception and gestation differ in notable ways from its infancy and adulthood. Its birth is over 50 years in the making, whereas its expansion across the macula is typically accomplished in less than a decade; the interval from incidence of noncentral GA to central involvement is typically around 3 years.²

“It does not take a long time to strike a man with Lightning,”said Defarge. “How long,” demanded madame, composedly, “does it take to make and store the lightning? Tell me.”¹

Similarities and differences between GA incidence and GA expansion can be evaluated by comparison of genetic and environmental risk factors between the 2 stages (Table 1). This has important implications not only for potential differences in the underlying biological mechanisms but also for whether partially distinct therapeutic approaches may therefore be required at each stage.

Table 1.

Partially Distinct and Partially Overlapping Risk and Protective Factors for Progression to GA (Incidence) Versus Progression of GA (Expansion)

	Progression to GA (Incidence)	GA Progression (Expansion)
Shared	ARMS2/HTRA1 genotype: minor allele at rs10490924 (risk factor)	ARMS2/HTRA1 genotype: minor allele at rs10490924 (faster progression)
	Cigarette smoking (risk factor)	Cigarette smoking (faster progression)
	Mediterranean dietary pattern (protective factor)	Mediterranean dietary pattern (slower progression)
	Phenotypic characteristics including reticular pseudodrusen and hyperreflective foci (risk factors)	Phenotypic characteristics including reticular pseudodrusen and hyperreflective foci (faster progression)
Not Shared	Age (risk factor)	Age (no association)
	CFH genotype/haplotype (risk or protective factors)	CFH genotype/haplotype (no association)
	Multiple other genetic variants, including C2/CFB and CFI (risk or protective factors)	Multiple other genetic variants, including C2/CFB and CFI (no association)
	Genetic variants at PRMT6 and LSS (no altered risk at genome-wide significance)	Genetic variants at PRMT6 and LSS (altered progression rate)
	Mediterranean diet food components: -higher fish intake (protective factor; strong interaction with CFH genotype) -higher whole grain intake (protective factor)	Mediterranean diet food components: -higher fish intake (no association) -higher whole grain intake (no association)
	Mediterranean diet food components: -higher whole fruit intake (possible protective factor) -lower red meat intake (no association) -moderate alcohol intake (possible protective factor) -higher MUFA: SFA intake ratio (no association)	Mediterranean diet food components (association with slower progression for each): -higher whole fruit intake -lower red meat intake -moderate alcohol intake -higher MUFA: SFA intake ratio
Opposite Direction	C3 genotype: minor allele at rs2230199 (risk factor)	C3 genotype: minor allele at rs2230199 (slower progression)
	APOE genotype: minor allele at rs73036519 (protective factor)	APOE genotype: minor allele at rs73036519 (faster progression)
Uncertain	AREDS supplements (no significant effect, with outcome as progression to GA or central GA)	AREDS supplements (nominally significant slower progression for antioxidants plus zinc vs. placebo, in post hoc analyses with low power)
	Local complement inhibition at C3 by pegcetacoplan (nominally significant decreased progression from iRORA to cRORA, in post hoc analyses of a phase II trial)	Local complement inhibition at C3 by pegcetacoplan or C5 by avacincaptad pegol (slower progression for each)

AREDS = Age-Related Eye Disease Study; cRORA = complete retinal pigment epithelium and outer retinal atrophy; GA = geographic atrophy; iRORA = incomplete retinal pigment epithelium and outer retinal atrophy; MUFA = monounsaturated fatty acid; SFA = saturated fatty acid.

Metaphors from other fields may be helpful: in a forest fire, factors that increase the risk of the first ignition may have some similarities but important differences from those that lead to faster spread of the fire at its leading edges. Related to this, the mechanisms responsible for the first ignition may be quite different from those responsible for subsequent spread. Importantly, therefore, different fire-fighting approaches may be required at each stage. Similarly, in large-scale uprisings like the French Revolution, factors that increase the risk of isolated insurrection may have some similarities but important differences from those that make insurrection spread more quickly.3, 4, 5 Again, therefore, partially distinct approaches may be effective in addressing uprising at each stage.

In some scenarios, as in the spread of revolution, the nature of the inciting event has implications for the rate of subsequent spread because each new unit added must undergo its own conversion to the cause (akin to cell death). Any attempt at treatment must take into account the original event and underlying grievances (perhaps related to AMD genotype and phenotype, for example). In other scenarios, as in a forest fire, the nature of the original event may be largely irrelevant for the purposes of subsequent spread. In these cases, treatments that are agnostic to the original event might still be effective. Knowing whether GA behaves in one or other of these ways, therefore, has important implications for appropriate therapeutic strategies. For example, 1 cross-sectional cluster analysis of patients with GA reported 3 partially distinct GA subtypes: (1) one characterized by higher frequency of soft drusen and GA central involvement and higher complement-based genetic risk score (GRS); (2) another characterized by lower frequency of soft drusen or reticular pseudodrusen (RPD) and low AMD GRS; and (3) a third characterized by higher frequency of RPD, lower frequency of GA central involvement, and higher ARMS2 GRS and extracellular matrix GRS.⁶ However, longitudinal cluster analysis of GA in the Age-Related Eye Disease Study 2 (AREDS2) did not replicate these findings; no strong genotype-phenotype correlations were observed and the results were not consistent with distinct GA subtypes arising from different pathway-based genetic etiologies.⁷

A further distinction is important: some forest fires are triggered by a single ignition event, whereas others may arise from multiple ongoing ignitions at different locations, which gradually spread and merge. A similar distinction is relevant in considering the phenomenon of GA expansion, although this difference is not captured in most natural history studies or clinical trials.^8,9 Geographic atrophy expansion occurs through a combination of contiguous enlargement at the circumference of established lesions and incidence of new lesions. However, the relative contribution of these 2 factors likely differs widely between individuals (and even between different time points of the same eye). At 1 extreme, some eyes with GA may have only 1 incidence event, followed by contiguous enlargement only, such that the GA remains unifocal throughout its life cycle. At the other extreme, other eyes may have multiple incidence events early on and on an ongoing basis, such that the GA is highly multifocal in early years before lesion confluence leads to fewer foci. Importantly, the risk factors and mechanisms underlying each type of expansion may differ partially; the risk factors for ongoing incidence events seem likely to resemble more closely the risk factors for the first incidence event.^8,9

The spatiotemporal behavior of these different types will vary and, importantly, the optimum approaches to slowing expansion may vary according to these types. Therefore, models attempting personalized predictions of GA expansion rate (either globally as a mm/year rate or spatially on a pixel-by-pixel basis) should be more accurate if they are trained on data that makes this distinction. Similarly, analysis of risk factors of expansion should be more meaningful if this distinction is considered, such that risk factors are considered separately for contiguous enlargement versus the incidence of additional foci. Most importantly, studying the treatment effect of potential therapies will be more meaningful when this distinction is prespecified in the analyses.

Age

Higher age is an established risk factor for GA incidence. Meta-analysis of multiple studies from different countries has shown that the prevalence of GA increases exponentially with age—from 0.7% at 70 years, to 2.9% at 80 years, to 11.3% at 90 years in populations of European ancestry.¹⁰ By contrast, higher age is not a significant risk factor for faster GA expansion.^2,11,12 In the AREDS2 cohort of 1616 eyes with GA (without previous neovascular AMD), age was not significantly associated with altered GA expansion rate (in multivariable analyses using the square root transformation).² By analogy, conception might take longer, on average, to occur with older parental age; however, once this is attained, the subsequent rate of pregnancy does not vary according to parental age.

This may seem surprising: age represents an accumulation of multiple environmental exposures, other age-related biological changes, and even accumulated genetic and epigenetic changes. Therefore, one might imagine GA tending to progress more quickly with advanced age, perhaps owing to increased tendency to inflammation; decreased reserve in the retinal pigment epithelium (RPE), choriocapillaris, and photoreceptors; and to multiple other adverse consequences. One possible explanation is that some of these abnormalities may correlate more closely with AMD stage than with age alone. In this way, the degree of macular aging changes might already be captured by age of GA onset, such that age has no additional capacity to predict GA enlargement rate. We might imagine that GA develops when a particular threshold of macular aging changes is reached. This threshold could be similar between individuals; someone who developed GA at 60 years had a faster rate of macular aging changes, whereas someone who developed GA at 90 years had a slower rate of macular aging changes but to the same threshold. In this way, the degree of macular aging changes would be similar at GA onset between the 2 individuals, despite their age difference, and GA enlargement would therefore proceed at a similar rate, irrespective of age.

However, the contrast might also point to important differences between the mechanisms underlying GA incidence and expansion. If GA tends to progress at similar rates in individuals aged 60 and 90 years, then aging changes (at the molecular, cellular, and genetic/epigenetic levels¹³) might not be centrally implicated in mechanisms of GA expansion. Related to this, previous studies have investigated the potential efficacy of drugs like sirolimus, which have demonstrated anti-aging properties in extending the lifespan of worms, flies, and mice through the mTOR pathway.^14,15 However, clinical studies of subconjunctival or intravitreal sirolimus in GA did not seem to slow GA enlargement.^16,17 Overall, it remains difficult to distinguish between these 2 possible explanations, and further studies are recommended.

Genetic Factors

Genome-wide association studies (GWAS) have been conducted separately for late AMD prevalence and GA expansion.^18,19 Interestingly, the results do not overlap at all. A large GWAS of late AMD identified 52 independently associated genetic variants across 34 loci.¹⁸ These loci relate to genes involved in biological pathways, including the complement system, lipid metabolism and transport, extracellular matrix maintenance and remodeling, angiogenesis, and cell survival.^18,20 By contrast, a GWAS of GA expansion revealed only 2 loci with genome-wide significance (likely related to the genes PRMT6 and LSS), neither of which was observed in the GWAS of late AMD.¹⁹

For late AMD prevalence, of the 34 loci, the 2 with highest population attributable risk are at CFH and ARMS2/HTRA1.¹⁸ Of these 2, 1 appears to be associated with faster GA expansion, whereas the other does not. For ARMS2/HTRA1, the minor allele at rs10490924, which is highly associated with increased risk of GA incidence, is also highly associated with faster GA expansion in most studies.^2,11,19,21 In multivariable analyses of the AREDS2 cohort of 771 eyes with GA and genetic data, ARMS2/HTRA1 risk alleles were very strongly associated with faster GA expansion.²¹ However, the relative contribution of the 2 types of expansion described earlier is unknown here. By contrast, for CFH, none of the common risk or protective haplotypes appear to confer altered risk of faster or slower GA expansion.^2,11,19 In the AREDS2, CFH genotype was not even nominally associated with altered GA expansion rate.² Hence, the minor allele at ARMS2/HTRA1 increases both the risk of GA development and of that GA, once present, expanding faster—presumably through the same mechanism, according to Occam's razor. By contrast, CFH haplotypes increase or decrease the risk of developing GA but do not alter the subsequent expansion rate.

According to the revolution metaphor, widespread grievances against the political elite might increase both the likelihood of initial insurgence and speed of subsequent spread, whereas fragmentation of communities might increase the former but not the latter. Similarly, in the case of forest fires, severe heat might increase both the risk of initial ignition and rate of spread, through similar mechanisms, whereas human activity might increase the former but not the latter.

In fact, CFH genotype may be less relevant for disease progression at even earlier severity stages. In analyses of the AREDS and AREDS2 cohorts using multivariable models that included baseline AMD severity level and demographic factors, the minor allele at rs10922109 (the lead variant at the CFH locus in a large GWAS and a protective variant) was highly associated with altered risk of progression to late AMD in the AREDS,¹⁸ which comprised a wide spectrum of AMD severity at baseline, but not even nominally in the AREDS2, for which the eligibility criterion was at least intermediate AMD.²² This was also true for progression to GA specifically. By contrast, the minor allele at ARMS2/HTRA1 was highly associated with progression to late AMD (and GA specifically) in both the AREDS and AREDS2.

In some cases, we observe not only lack of overlap for risk at the 2 disease stages but even associations in opposite directions (Table 1). This is most striking for C3 (rs2230199); the minor allele at this locus has consistently been associated with increased risk of late AMD, including GA.¹⁸ However, in multivariable analyses of the AREDS2, the same allele was highly associated with slower GA expansion²; this finding has been observed elsewhere also.²³ Hence, it appears that C3 risk alleles, which increase systemic complement activation levels,^24,25 may drive disease pathogenesis over many decades toward the point of GA incidence but, at the point of established GA, encourage slower expansion. This appears paradoxical and difficult to explain. However, by analogy, in a forest fire, low wind conditions may increase the likelihood of initial ignition but lead to slower spread, or a thunderstorm might increase the risk of initial ignition through lightning strike but lead to slower spread through wet timber.

Other important observations have come from recent randomized controlled trials of local complement inhibition, at either the C3 or C5 level, as discussed below. In both cases, local complement inhibition has slowed GA expansion, although the degrees of slowing have been relatively modest, and treatment has led to increased risk of neovascular AMD.26, 27, 28, 29

It is difficult to reconcile these 3 sets of observations: (1) no difference in GA expansion rates according to CFH haplotype; (2) slower GA expansion rates with a C3 genotype that increases systemic complement activation and increases risk of GA; and (3) slowing of GA expansion with local C3 (or C5) inhibition. From the first observation, we might conclude that the complement system is involved in the risk of progression to GA but not in GA expansion. However, from the other 2 observations, this appears not to be the case; rather, it seems likely that the complement system is involved at both stages but in different ways. One possibility is that the complement system might represent the central process driving disease progression at both stages but in different ways at each stage (perhaps through partially distinct effector mechanisms or targeting different cell types). This might help explain why local C3 inhibition is effective at slowing GA enlargement,^26,30 despite the first observation, as well as why C3 genotype appears to have different effects at different disease stages. Another possibility is that the complement system might represent the central process driving disease progression up to GA incidence but might be a contributing rather than a central process for GA expansion. This might help explain why the efficacy of local C3 inhibition for slowing GA enlargement is not even greater, as well as the failure of some previous trials of complement inhibitors for GA expansion.^24,26,30,31

In particular, the complement system has multiple effector mechanisms; the most well-known are the membrane attack complex (which can lead to cell lysis but also sublytic effects), the opsonin C3b and its fragments (which lead to phagocytosis of opsonized material, but also immunomodulation), and the anaphylatoxins C3a and C5a (which lead to chemotaxis, inflammation, and immunomodulation).^32,33 However, C3 and its fragments are thought to play highly multifaceted roles in AMD, even beyond these mechanisms. These have been reviewed in detail and include inflammasome activation, recruitment of microglia/macrophages to the subretinal space, intracellular C3 mechanisms within the RPE (including activation of the mTOR pathway), and aberrant lysosomal turnover.^24,32 Given this complex situation and the lack of animal or other models that fully recapitulate human disease, understanding which of these different roles played by C3 are those most highly implicated in AMD, and whether the dominant mechanisms may differ by disease stage, has not been fully elucidated.^34,35

It is even possible that a degree of complement signaling may be required to maintain retinal health and homeostasis. Hence, whereas prolonged excess complement activation may lead to chronic local inflammation and contribute to AMD progression,³⁶ a degree of complement signaling may be required for photoreceptor and neuronal health through noncanonical roles of the complement system in cell survival and immune regulation.^37,38 For the membrane attack complex, in some cells and scenarios, sublytic membrane attack complex levels are thought to activate transcription factors that favor cell survival and resistance to apoptosis.³⁹ For C3, there is some evidence from animal models that it has an important role in maintaining retinal integrity and homeostasis during aging,⁴⁰ although these observations are complicated by a background of abnormal complement-dependent neuronal pruning during development in C3 knockout mice.⁴¹

The idea of C3 potentially acting as a survival factor in some situations might help explain the paradoxical findings for C3 genotype by disease stage. Despite this, it remains difficult to reconcile the observations of slower GA expansion with a high-risk C3 genotype and slower GA expansion with local C3 inhibition. However, substantial differences may exist between the complement status of an individual through genetic variation versus pharmacologic treatment³²; in this case, the former is systemic, continuous, and life-long, whereas the latter is local, intermittent, and therefore likely partial. Overall, further research from both the basic science and clinical domains is required to help reveal the precise roles played by the complement system in AMD, both according to patient genotype and to disease stage.

The genetic factors described earlier have been considered in terms of associations with global GA expansion rates. However, 2 additional factors have important influences on visual acuity and prognosis: (1) GA location at the point of incidence and (2) preferential GA expansion locally. The location most commonly affected by GA at incidence is the parafoveal region.^2,21,42 In the AREDS2 cohort of eyes with incident GA, 67% of cases did not involve the central macular zone at the time of incidence, whereas the remaining 33% did.² In the same cohort, the mean proximity of GA to the central macula (i.e., the shortest distance from the nearest pixel with GA to the macular center point) was 459 μm. In analyses of potential associations between AMD genotype (specifically 18 single nucleotide polymorphisms at 9 loci strongly associated with risk of late AMD) and risk of central involvement at GA incidence, no significant association was observed.² Few other studies have analyzed potential associations between genotype and GA location. Hence, although the location of GA at the point of incidence is important for visual acuity and prognosis, this characteristic does not seem to be strongly genetically determined. The predilection for the parafoveal region may relate more strongly to anatomical factors, such as the presence of high-risk lesions (particularly soft drusen and basal linear deposits) in both the fovea and parafoveal regions, but the relative absence of protective xanthophyll carotenoid pigment and Müller glia cells in the parafoveal area.^42,43 Indeed, dark adaptation studies and histological analyses show that, in AMD, rod photoreceptor degeneration is prominent in a parafoveal ring surrounding the fovea long before GA occurs.⁴³

Regarding local GA expansion, GA tends to expand preferentially in some directions more than others. In particular, analyses of an AREDS GA cohort have demonstrated that local GA expansion rates tend to be incrementally higher with increasing eccentricity across the macula, after adjustment for baseline GA area.⁴⁴ Hence, expansion into the fovea tends to be slowest, whereas expansion into the peripheral macula tends to be fastest. This likely explains the phenomenon of relative foveal sparing, including the common occurrence of GA configurations (e.g., horseshoe or donut) that tend to wrap around the fovea before the fovea itself finally becomes involved in later years.^2,45,46 Potential mechanisms for faster expansion with increasing macular eccentricity overlap with the anatomic factors discussed above; they may include greater vulnerability of rods than cones to atrophy expansion (with rod concentrations increasing gradually away from the central macula), the effects of decreased macular pigment density away from the fovea, or other potential factors.^43,44

Overall, the effects of GA on visual acuity and prognosis relate to the interplay of the following: (1) location of GA at incidence, likely related more to anatomic or stochastic factors than to AMD genotype; (2) altered global GA expansion rates, partly explained by genetic and environmental factors; and (3) preferential GA expansion by macular eccentricity, likely related partly to anatomic factors. In analyses of the AREDS2, visual acuity at the time of GA incidence was worse with GA that had central involvement or larger size.² The rate of visual acuity decline was numerically faster according to some characteristics (e.g., GA that was central, or had intermediate lesion size, or had a horseshoe/ring configuration).² In analyses of the Chroma/Spectri trials, visual acuity decline was faster in unifocal than multifocal GA; faster GA expansion was also associated with faster decline in visual acuity.^47,48

Environmental Factors

"I am not old, but my young way was never the way to age."¹

Cigarette smoking has consistently been demonstrated as a strong risk factor for the development of late AMD, including GA.^22,49,50 Evidence for altered GA expansion rate according to smoking status has been less consistent.¹¹ However, in 1 of the largest studies, multivariable analyses of the AREDS2 cohort of 1616 eyes with GA revealed a nominally significant association between smoking status (current vs. former vs. never) and faster GA expansion.² Similarly, in longitudinal phenotypic cluster analyses of the same cohort, positive smoking status and faster GA expansion rate characterized a prominent cluster division.⁷ Hence, this factor appears to be active and operating in the same direction at both disease stages, perhaps like heat in the context of forest fire initiation and spread.

“La destinée des nations dépend de la manière dont elles se nourrissent”⁵¹

Like smoking, dietary pattern has consistently been identified as a strong risk or protective factor for the development of late AMD and particularly GA.^52,53 Comprehensive longitudinal analyses using detailed data on diet, genetics, and AMD status have been conducted in the AREDS and AREDS2.⁵⁴ In both cohorts, closer adherence to a Mediterranean-like dietary pattern was very strongly associated with decreased risk of progression to late AMD; this was particularly true for GA. Similarly, regarding GA expansion, multivariable analyses of an AREDS2 cohort of 1155 eyes with GA showed that closer adherence to a Mediterranean-like dietary pattern was strongly associated with slower GA enlargement.⁵⁵ Therefore, like smoking, this environmental factor appears to be active and operating in the same direction at both disease stages (Table 1).

However, further insights into the foods chiefly responsible for these associations can be provided by analyses at the level of the 9 components of the Mediterranean diet index.⁵⁶ Importantly, in evaluation of the AREDS and AREDS2, these analyses isolated the contribution of each component by adjusting for the overall dietary pattern excluding the component under study.^54,55 For progression to GA, the dietary component with the strongest protective association was higher fish intake, with weaker protective associations observed for whole grains and possibly whole fruits and moderate alcohol.⁵⁴ By contrast, for GA expansion, no meaningful association with faster or slower expansion was observed according to fish intake; instead, the components with the strongest associations for slower GA expansion were higher intake of whole fruits, lower intake of red meat, moderate alcohol intake, and higher healthy-to-unhealthy lipid intake ratio.⁵⁵ Hence, a Mediterranean-like dietary pattern is strongly associated with slower progression to GA and slower GA expansion, but for partially distinct reasons at each stage. This in turn suggests that the biologic mechanisms underlying GA development may be partially distinct from those driving GA expansion and emphasizes that some differences in therapeutic approaches may be required at each stage.

Furthermore, a very strong gene–diet interaction has been observed in association with GA development, but not in association with GA expansion.^54,55 In the AREDS, the association between closer adherence to a Mediterranean-like dietary pattern and decreased progression to late AMD was found only in those with a common protective allele at CFH.⁵⁴ At the food component level, an extremely strong interaction was present; the association between higher fish intake and decreased progression to GA was present only in those with the CFH protective allele. By contrast, for GA expansion, in the absence of any significant association between GA expansion and either CFH genotype or fish intake, no such interaction is possible.

The relevant foods, together with their constituent nutrients and potential effects on the intestinal microbiome,^57,58 may provide important insights into which biologic mechanisms are most implicated at each disease stage. Overall, the beneficial effects of the Mediterranean diet are thought to include anti-inflammatory properties, anti-oxidative properties, altered lipid metabolism, improved vascular endothelial health, improved mitochondrial energetics, and neural protection.⁵⁹ For progression to GA, the strong finding for fish and its powerful interaction with the protective CFH genotype suggest that, for many patients, the dominant biologic mechanisms at this stage may include those related to innate immunity and inflammation, as discussed below. For GA expansion, the findings suggest that the dominant mechanisms at this stage may include oxidative stress, lipid metabolism, and perhaps vascular dysfunction. For example, the beneficial effects of higher whole fruit intake on systemic health are thought to relate to flavonoids and other antioxidants decreasing oxidative stress and likely improving vascular health.⁶⁰

For progression to GA, the strong genotype dependence argues that the biological mechanism underlying potential protection from fish intake may be closely related to the complement system. The protective allele at CFH is frequently characterized by the CFHR3/1 deletion.⁶¹ These 2 CFH-related proteins are known to compete with CFH (and its splice variant FHL-1) for binding to important epitopes, such as malondialdehyde (MDA).^62,63 Malondialdehyde is a common lipid peroxidation product that accumulates in many pathophysiological processes, including AMD; it is an important source of oxidative stress and of enhanced complement activation.⁶² CFH is a major MDA-binding protein that is thought to block the uptake of MDA-modified proteins by macrophages and MDA-induced proinflammatory effects.⁶² Therefore, the deletion means that CFH can bind to its appropriate epitopes, leading to less excess complement activation and less chronic inflammation. However, the food–gene interaction observed suggests that this protective effect is particularly powerful when combined with high dietary intake of fish. Fish flesh is known to contain multiple natural antioxidants, including enzymes (e.g., catalase and superoxide dismutase), carotenoids, peptides, amino acids (e.g., taurine), and phenolic compounds (e.g., tocopherols like vitamin E, and the ubiquinones), as well as omega-3 polyunsaturated fatty acids like docosahexaenoic acid and eicosapentaenoic acid that also have antioxidant capacity.^64,65 Hence, the pairing of high fish intake, with many antioxidants, and a protective CFH genotype, with decreased tendency to complement activation and inflammation, may work strongly in combination to decrease risk of progression to GA, by decreasing both the stimulus for and the response to oxidative stress. However, further research to investigate interactions between fish nutrients and complement regulation is appropriate to understand which particular nutrients may be most important.

Genetic and environmental factors have been considered simultaneously in some studies. For example, pooled analysis of 3 population-based studies of AMD led to the development of a prediction model to calculate the risk of progression to late AMD.⁶⁶ The model included the factors age, sex, 26 single nucleotide polymorphisms in late AMD risk genes, smoking, body mass index, and baseline AMD phenotype. However, when the same model was applied to GA expansion rate in 2 of the same population-based studies, the correlation was very poor, and not even a nominally significant association was observed (although GA expansion was considered without square root or other transformation).¹²

Clinical and Imaging Factors

A comprehensive review of phenotypic or imaging factors is outside the scope of this review. However, RPD, also known as subretinal drusenoid deposits, and hyperreflective foci are worthy of discussion. In both the AREDS and AREDS2, RPD presence was associated with significantly increased risk of progression to late AMD, despite adjustment for other characteristics of AMD severity.⁶⁷ This was particularly true for progression to GA. Similarly, in both the AREDS and AREDS2, RPD presence was also associated with significantly faster GA expansion.²¹ This is generally consistent with the results of other smaller studies.^11,68,69 Likewise, hyperreflective foci have been reported as a significant risk factor for both progression to late AMD and faster GA expansion.68, 69, 70, 71 Finally, GA presence in 1 eye is associated in the fellow eye with both increased risk of GA incidence (if GA is absent) and faster GA expansion (if GA is present).^2,11,72,73 Hence, these phenotypic factors have associations operating in the same direction at both disease stages.

According to Occam's razor, this suggests that similar mechanisms are responsible for linking the risk feature to faster progression at both stages. Reticular pseudodrusen, for example, might represent a precursor lesion for GA at the time points of both GA incidence and GA expansion.²¹ Geographic atrophy is diagnosed on OCT by a combination of RPE loss and photoreceptor degeneration,⁷⁴ and retinal areas affected by RPD typically show decreased RPE cell density and a thin outer retina (particularly in the outer nuclear layer).75, 76, 77 Considering RPD in this way could explain faster progression at both disease stages through local effects. Other explanations might include innate immune cell activity or choroidal abnormalities, with each believed to be involved in the pathogenesis of RPD but also with GA development and GA expansion.^75,78,79

Potential Treatments

The AREDS oral supplements are commonly used at more severe stages of AMD. For individuals with large drusen in ≥ 1 eye or advanced AMD in 1 eye, they decrease the risk of progression to advanced AMD (i.e., neovascular AMD or central GA).⁸⁰ However, this treatment effect is weighted strongly toward decreased risk of neovascular AMD. In the AREDS category 3 and 4 participants, the odds ratio (OR) was highly significant for neovascular AMD (OR, 0.62; 99% confidence interval [CI], 0.43–0.90; P = 0.001) but not significant for either central GA (OR, 0.75; 99% CI, 0.45–1.24; P = 0.13) or any GA (OR, 1.08; 99% CI, 0.70–1.65). Regarding a potential treatment effect of the AREDS supplements in slowing GA expansion, post hoc analyses of a subset of AREDS participants with GA have been performed; these included adjustment for baseline GA area but no square root transformation, and power was low, with only 68 participants between the 4 treatment arms.⁸¹ The statistical test for an overall treatment effect was not significant, although the individual test of antioxidants plus zinc versus placebo was nominally significant (P = 0.05), with slower enlargement in participants randomized to antioxidants plus zinc. In the AREDS2, randomization to either lutein/zeaxanthin or docosahexaenoic acid/eicosapentaenoic acid did not lead to significantly slower GA enlargement (in the context of all participants also taking the AREDS supplements).² Hence, AREDS supplements seem to have relatively little, if any, effect on decreasing progression to GA and a possible but uncertain effect on slowing GA expansion.

Local complement inhibition has attracted great interest as a potential therapy to slow GA expansion, as discussed earlier.⁹ Randomized controlled trials of local inhibition, at both the C3 and C5 level, have observed significantly slower decrease GA expansion, although the treatment effects reported in phase III trials have been relatively modest.26, 27, 28, 29 For progression to GA, in post hoc analyses of 1 phase II trial of a C3 inhibitor, macular lesions > 500 μm beyond GA margins at baseline were analyzed by OCT.⁸² Monthly treatment was nominally significant for decreased progression from incomplete to complete RPE and outer retinal atrophy. Hence, it appears that local complement inhibition might be effective and operating in the same direction at both disease stages. However, for progression to GA, the findings were from post hoc analyses of a single phase II trial and relate to a very narrow spectrum of disease progression (when incomplete RPE and outer retinal atrophy is already present, rather than progression from the medium or large drusen stage). In addition, importantly, the 2 sets of findings overlap so should not necessarily be considered additive. If the potential decrease in progression to complete RPE and outer retinal atrophy contributes to lower GA area measurements at later time points in the clinical trial, this effectively downgrades the treatment effect related to slower enlargement at GA margins.^8,9 As discussed above, for any potential treatment, we must consider that decreased ongoing progression to GA will seem to cause slower GA expansion if we do not analyze contiguous enlargement at the circumference of established lesions separately from the addition of new foci.

Conclusions

Analysis of the risk and protective factors for progression to GA and GA expansion demonstrates partially overlapping but partially distinct elements at each stage: (1) some are shared, (2) some are relevant at 1 stage only, and (3) some even seem active in opposite directions at each stage. This indicates that the underlying biologic mechanisms differ at least partially between the 2 stages. This in turn suggests that therapeutic approaches need to be tailored to the stage. Where a risk factor is shared between stages, as for ARMS2/HTRA1 risk alleles or cigarette smoking, Occam's razor suggests that the factor is increasing risk through the same mechanism throughout. Where a risk factor is relevant at 1 stage only, as for age or CFH genotype, it suggests that the mechanism operating during 1 stage is not highly implicated in pathogenesis at the other stage. Where a factor increases risk at 1 stage but decreases risk at the other, as for C3 genotype, we must either invoke 2 separate mechanisms or consider 1 mechanism with opposing effects at each stage. Such an integrated view of risk and protective factors operating differentially by stage, including interactions between diet and genotype, will also improve prognostic predictions across the life cycle of GA development and expansion.

Regarding potential therapeutic approaches, these considerations suggest that complement modulation is likely to be more effective at relatively early disease stages, whereas future therapies targeting ARMS2/HTRA1 might be effective at both decreasing GA development and slowing GA expansion. In addition, complement modulation will presumably be most appropriate in those with a high complement pathway-based polygenic risk score (or other evidence of local or systemic complement dysregulation), whereas ARMS2/HTRA1 modulation will presumably be most appropriate in those with risk alleles at that locus. In these ways, early therapeutic approaches can be tailored to the underlying disease processes (i.e., by analogy, to the type of ignition event or the underlying grievances). Even after the time point of GA incidence, some of these might need to be continued, either in the case of ongoing relevance (e.g., for ARMS2/HTRA1) or to decrease ongoing incidence events (toward multifocality). In addition, after GA incidence, some additional therapeutic approaches that are agnostic to the incident event could also be appropriate.

Importantly, alongside pharmacological interventions, these data indicate that lifestyle modification may be effective at all stages. Smoking cessation may be helpful at any stage. Similarly, the Mediterranean diet is strongly associated with decreased progression to intermediate AMD, decreased progression to GA, and slower GA expansion.^54,55 However, pending further research, dietary recommendations on the most relevant food components for each stage might be tailored in the future to disease stage and CFH genotype. Meanwhile, AREDS-style supplements remain appropriate alongside a healthy diet because associations with decreased progression to late AMD seem additive, without redundancy or other interaction.^54,83 The 2 are complementary, because supplements preferentially decrease risk of neovascular AMD, whereas a healthy diet is associated preferentially with decreased risk of GA.

Manuscript no. XOPS-D-23-00031R2.