Progression of geographic atrophy with subsequent exudative neovascular disease in age-related macular degeneration: AREDS2 Report 24

Abstract

Purpose

To examine whether the rate of geographic atrophy (GA) enlargement is influenced by subsequent exudative neovascular age-related macular degeneration (NV-AMD), hence, to explore indirectly whether non-exudative NV-AMD may slow GA enlargement.

Design

Post hoc analysis of a controlled clinical trial cohort.

Participants

Age-Related Eye Disease Study 2 (AREDS2) participants, aged 50–85 years.

Methods

Baseline and annual stereoscopic color fundus photographs were evaluated for (i) GA presence and area, and (ii) exudative NV-AMD presence. Two cohorts were constructed: eyes with GA at study baseline (prevalent cohort) and eyes that developed GA during follow-up (incident cohort). Mixed-model regression of the square root of GA area was performed according to the presence/absence of subsequent exudative NV-AMD.

Outcome measures

Change over time in square root of GA area.

Results

Of the 757 eyes in the incident GA cohort, over a mean follow-up period of 2.3 years (SD 1.2), 73 (9.6%) eyes developed subsequent exudative NV-AMD. GA enlargement in these eyes was significantly slower (0.20 mm/year, 95% CI 0.12–0.28) compared with the other 684 eyes that did not develop subsequent exudative NV-AMD (0.29 mm/year, 95% CI 0.27–0.30; p=0.037). Of the 456 eyes in the prevalent GA cohort, over a mean follow-up period of 4.1 years (SD 1.4), 63 (13.8%) eyes developed subsequent exudative NV-AMD. GA enlargement in these eyes was similar (0.31 mm/year, 95% CI 0.24–0.37) compared with the other 393 eyes that did not develop subsequent exudative NV-AMD (0.28 mm/year, 95% CI 0.26–0.29; p=0.37).

Conclusions

In eyes with recent GA, GA enlargement prior to the development of exudative NV-AMD appears slowed. This association was not observed in eyes with more long-standing GA, which have larger lesion sizes. Hence, perilesional non-exudative choroidal neovascular tissue (presumably present before the development of clinically apparent exudation) may slow enlargement of smaller GA lesions through improved perfusion. This hypothesis warrants further evaluation in prospective studies.

Précis

In age-related macular degeneration, geographic atrophy enlargement appears slower in the presumed presence of non-exudative macular neovascularization. Hence, perilesional non-exudative choroidal neovascular tissue may slow enlargement of smaller atrophic lesions by improving perfusion.

Introduction

Age-related macular degeneration (AMD) is the leading cause of legal blindness in developed countries.¹ Late AMD is the disease stage with substantial risk for visual loss. Late AMD has two forms, neovascular AMD (NV-AMD) and atrophic AMD. The defining lesion of atrophic AMD is geographic atrophy (GA), which refers to one or more circumscribed areas of atrophy in the macula.² In GA, atrophy of the retinal pigment epithelium (RPE) is typically accompanied by atrophy of the adjacent photoreceptors and choriocapillaris.^3,4 GA lesions enlarge and coalesce over time; however, enlargement rates vary between eyes, with multiple clinical, imaging, and genetic factors associated with faster or slower enlargement.^5,6 No treatments are routinely available in clinical practice to treat GA or slow its enlargement rate. However, potential strategies to slow GA enlargement are currently under investigation, including local inhibition of the complement system.^7,8 In these and previous clinical trials, GA enlargement rate is used as the primary outcome measure.^9–12

Importantly, GA and NV-AMD can coexist in the same eye. Indeed, in a previous study using the same dataset and cohorts of eyes as those analyzed in the current study, the risk of subsequent NV-AMD at four years after the new appearance of GA was 29%.⁵ The defining lesion of NV-AMD is macular neovascularization (MNV), accompanied by its exudative and fibrovascular sequelae. However, the advent of optical coherence tomography (OCT) angiography has led to renewed appreciation of the existence of non-exudative (also known as subclinical or quiescent) MNV.^13–15 Indeed, in a recent prospective study, the large majority of new cases of exudative MNV had pre-existing non-exudative MNV detectable by swept-source OCT angiography.¹⁶ If these findings are representative of the wider population of eyes with NV-AMD, it suggests that the large majority of NV-AMD cases likely exist as non-exudative MNV for a variable but potentially prolonged period of time before the onset of exudation.

In a previous study, we presented the hypothesis that the presence of non-exudative MNV adjacent to a GA lesion might slow GA enlargement.⁵ Other studies have reported potential associations between the presence of MNV (exudative and/or non-exudative) and decreased incidence and/or enlargement of macular atrophy.^{13,14,17–19} In those studies that assessed associations with non-exudative MNV, the major strength was direct visualization of the neovascular membrane by OCT angiography. However, the majority of these had very low numbers of study eyes. In this context, with such a high proportion of eyes with GA developing subsequent exudative MNV, it is important to characterize the natural history of GA in these eyes.

The AREDS2 was a multicenter, phase III, randomized controlled trial designed to assess the effects of nutritional supplements on the course of AMD (see Supplemental Appendix for AREDS2 Research Group).²⁰ It was based principally on color fundus photography; owing to the imaging technology available at the time, it did not incorporate OCT angiography. With five years of follow-up, many eyes with GA developed MNV over the course of the study, providing a unique opportunity to examine GA enlargement in these eyes. The aims of this current study were to use the AREDS2 dataset to: (i) identify eyes with GA that later developed exudative MNV and measure their GA enlargement rates in the period prior to exudative MNV; (ii) use the subsequent development of exudative MNV as a proxy for a period of presumed non-exudative MNV (since OCT angiography was not available for direct visualization of any eyes with non-exudative MNV); (iii) compare GA enlargement rates in eyes with and without subsequent exudative MNV; (iv) hence, explore indirectly whether GA enlargement may be slower in the presumed presence of non-exudative MNV.

Methods

Study participants and procedures

The study design for AREDS2 has been described previously.²¹ Briefly, 4203 participants (aged 50–85 years) were recruited at 82 retinal specialty clinics in the United States. Eligible participants were those with bilateral large drusen or late AMD in one eye and large drusen in the fellow eye. Institutional review board approval was obtained at each clinical site and written informed consent for the research was obtained from all study participants. The research was conducted under the Declaration of Helsinki and complied with the Health Insurance Portability and Accountability Act.

The AREDS2 participants were randomly assigned to lutein/zeaxanthin, docosahexaenoic acid (DHA) and eicosapentaenoic acid (EPA), the combination of lutein/zeaxanthin and DHA and EPA, or placebo. At baseline and annual study visits, comprehensive eye examinations were performed by certified study personnel using standardized protocols. Digital stereoscopic color fundus photography was performed at each study visit by certified photographers. The color fundus photography imaging protocol has been described previously.²² In addition, the AREDS2 ancillary study of FAF imaging was conducted at 66 clinic sites, according to the availability of imaging equipment and principally in later study years. ²³ The FAF imaging protocol has been described previously. ²³ In brief, FAF images were taken by certified photographers using the Heidelberg Retina Angiograph (Heidelberg Engineering, Heidelberg, Germany) or fundus cameras with autofluorescence capability.

Image analysis

The color fundus photographs were graded centrally by certified graders at the Fundus Photograph Reading Center at the University of Wisconsin.²² Calibrated stereoscopic images were viewed in a standardized digital viewing platform (ImageNet 2000; Topcon Corp., Tokyo, Japan) after color contrast and illumination adjustment. GA was defined as a lesion equal to or larger than drusen circle I-2 (diameter 433 mm, area 0.146 mm², i.e., 1/4 disc diameter and 1/16 disc area) at its widest diameter with at least two of the following features present: circular shape, sharp (well-demarcated) edges, and loss of the RPE (partial or complete depigmentation of the RPE, typically with exposure of underlying choroidal vessels). The configuration of GA was documented, using the definitions published by Sunness et al,²⁴ as either (1) small (single patch less than 1 disc area), (2) multifocal, (3) horseshoe or ring, (4) solid or unifocal, or (5) indeterminate. Planimetry tools were used to demarcate the area of GA within the AREDS grid in square millimeters. In the case of multi-focal GA, the areas were summed to yield a single value for analysis. The presence or absence of central macular involvement was recorded. Exudative NV-AMD was defined, as described previously²², by a history of treatment for NV-AMD and/or the presence of at least two of the following features on stereoscopic fundus photography: serous detachment of the sensory retina, hemorrhage, RPE detachment, fibrous tissue, or hard exudates. The FAF images were graded centrally by certified graders at the Fundus Photograph Reading Center at the University of Wisconsin for the presence of reticular pseudodrusen (RPD). The RPD grading protocol has been described previously.²³ In brief, RPD were defined as clusters of discrete round or oval lesions of hypoautofluorescence, usually similar in size, or confluent ribbon-like patterns with intervening areas of normal or increased autofluorescence; a minimum of 0.5 disc areas (approximately five lesions) was required.

Genotype analysis

As part of the AREDS2, 1826 participants gave consent for genotype analysis. Single nucleotide polymorphisms (SNPs) were analyzed using a custom Illumina HumanCoreExome array, as described previously.²⁵ The three loci selected for the current analyses were those identified from previous studies to be associated with altered speed of GA enlargement (rs10490924 at ARMS2, rs2230199 at C3, and rs73036519 at APOE).^5,26

Eye cohorts and statistical methods

The two cohorts of eyes with GA were the same as those constructed in a previous study of GA in the AREDS2 dataset.⁵ As described previously, two non-overlapping cohorts of eyes were constructed: a prevalent GA cohort and an incident GA cohort. The prevalent cohort comprised eyes with pre-existing GA at the AREDS2 baseline study visit, without prior or simultaneous exudative NV-AMD. The incident cohort comprised eyes with new GA that arose during the study follow-up period, without prior or simultaneous exudative NV-AMD. Eyes were included if they had at least two sequential visits with GA without simultaneous or prior exudative NV-AMD. Therefore, in cases of GA with subsequent development of exudative NV-AMD, at least two study visits were required prior to the development of exudative NV-AMD. In these cases, only the visits prior to the development of exudative NV-AMD were considered for GA assessment.

The unit of analysis was the eye. Regression analyses were performed using repeated-measures mixed-model regression with the square root of GA area as the primary outcome measure, according to the presence/absence of subsequent exudative NV-AMD, years from first appearance of GA, and their interaction term. The analyses were adjusted for square root of GA area at first appearance of the GA. To account for the correlation of measures of the same eye between visits, a first-order autoregressive covariance structure (AR[1]) was used. The regression analyses were performed separately for the prevalent and incident cohorts. As in previous analyses,⁵ the square root transformation of GA area was used in this study, as this decreases the dependence of enlargement rate on GA baseline size,²⁷ i.e. allows a fairer comparison of GA enlargement rates between eyes with different baseline sizes. However, in secondary analyses, the regression analyses were also performed without the square root transformation, but with adjustment for baseline lesion size.

For each cohort, characteristics were compared between the groups with and without subsequent development of exudative NV-AMD. For these comparisons, the chi-square and Fisher’s exact tests were used for categorical variables and the Wilcoxon two-sample test for continuous variables. The demographic characteristics were sex, age, and smoking status at study baseline. The clinical characteristics, defined at first appearance of GA on fundus photographs, were: GA area, central involvement, configuration, number of eyes with GA (i.e., unilateral or bilateral GA, defined as bilateral if GA was present in both eyes at any time during follow-up), and presence of RPD (only assessed for the subset of eyes with fundus autofluorescence imaging available at the time of first appearance of GA on fundus photographs). The genetic characteristics (only assessed for the subset of participants with genetic data available) were participant genotype at the three SNPs described above. All statistical analyses were performed with SAS (version 9.4, SAS Institute, Cary, NC). In these exploratory analyses, P-values < 0.05 were considered statistically significant.

Results

Participant cohorts: baseline characteristics

The numbers and baseline characteristics of the eyes with prevalent and incident GA are the same as those defined in a previous study.⁵ In brief, of the total of 8406 eyes of 4203 AREDS2 participants, 517 eyes of 411 participants already had GA without exudative NV-AMD at their baseline visit. Of these 517 eyes, 456 (88.2%) had at least two sequential visits with GA without simultaneous or prior exudative NV-AMD. These 456 eyes comprised the prevalent cohort and were eligible for subsequent analyses. Regarding the incident cohort, 1099 eyes of 883 participants without GA or exudative NV-AMD at their baseline visit developed new GA without simultaneous or prior exudative NV-AMD over the course of the 5-year trial. Of these 1099 eyes, 757 (68.9%) had at least two sequential visits with GA without simultaneous or prior exudative NV-AMD. These 757 eyes comprised the incident cohort and were eligible for subsequent analyses.

Of the 456 eyes of 367 participants in the prevalent cohort, over mean follow-up of 4.1 years (SD 1.4; Supplementary Table 1) comprising a mean of 5.0 annual visits (SD 1.5), 63 eyes (13.8%) of 59 participants developed exudative NV-AMD. Similarly, of the 757 eyes of 617 participants in the incident cohort, during mean follow-up of 2.3 years (SD 1.2) comprising a mean of 3.2 annual visits (SD 1.1), 73 eyes (9.6%) of 73 participants developed exudative NV-AMD.

The characteristics of the eyes in the two cohorts are shown in Table 1 separately according to the presence or absence of subsequent exudative NV-AMD. In the prevalent cohort, the two groups were not statistically different in terms of these characteristics. Importantly, baseline GA area was similar between the two groups. Mean age was numerically higher in the group with subsequent exudative NV-AMD, at 77.3 years (SD 5.3) versus 75.3 (SD 7.1) (p=0.09), respectively. However, previous analyses in the AREDS2 showed that age was not significantly associated with altered speed of GA enlargement.⁵ In the incident cohort, for the majority of these characteristics, the two groups were not statistically different. Baseline GA area was similar between the two groups. The proportion of former smokers was higher in the group with subsequent exudative NV-AMD (p=0.02 overall). However, previous analyses in the AREDS2 showed no significant difference in GA enlargement rates between former smokers and those who had never smoked.⁵ The proportion of eyes with central GA was numerically but not significantly lower in the group with subsequent exudative NV-AMD, at 23.3% versus 33.5% (p=0.08). The proportion of eyes with GA in the fellow eye was numerically but not significantly higher in the group with subsequent exudative NV-AMD, at 70.3% versus 61.1% (p=0.17).

Table 1.

Characteristics of the two cohorts with geographic atrophy, considered separately according to the presence of absence of subsequent exudative neovascular age-related macular degeneration.

Characteristic		Incident GA without subsequent exudative NV-AMD	Incident GA with subsequent exudative NV-AMD	P	Prevalent GA without subsequent exudative NV-AMD	Prevalent GA with subsequent exudative NV-AMD	P
Eyes, n		684	73		393	63
Age (years), mean (SD)		74.4 (SD 7.0)	75.5 (6.7)	0.21	75.3 (7.1)	77.3 (5.3)	0.09
Female sex, %		58.4	60.9	0.58	57.0	66.7	0.15
Smoking history, %				0.02			0.32
- Never		43.1%	28.8%		35.1%	44.4%
- Former		49.9%	65.8%		57.5%	50.8%
- Current		7.0%	5.5%		7.4%	4.8%
GA area (initial, mm2), mean (SD)		1.69 (2.5)	1.77 (2.6)	0.76	3.58 (4.3)	3.06 (4.0)	0.28
Central GA, %		33.5	23.3	0.08	33.3	33.3	1.00
GA configuration, %				0.35			0.48
- Small (single patch <1 DA)		58.8	57.5		34.4	39.7
- Multifocal		23.8	17.8		23.9	20.6
- Horseshoe or ring		2.5	2.8		9.4	4.8
- Solid		12.6	20.6		26.2	31.7
- Indeterminate		2.3	1.4		6.1	3.2
Fellow eye with GA, %		70.3	61.1	0.17	84.0	88.9	0.40
Genetic data available, n		350	39		136	29
ARMS2				0.90			0.98
rs10490924, % (0=G, 1=T)	0/0	35.7	35.9		36.0	37.9
	0/1	43.4	46.2		43.4	41.4
	1/1	20.9	17.9		20.6	20.7
C3				0.19			0.62
rs2230199, % (0=C, 1=G)	0/0	49.1	53.8		55.1	58.6
	0/1	42.6	30.8		36.0	37.9
	1/1	8.3	15.4		8.8	3.4
APOE				0.29			0.94
rs73036519, % (0=G, 1=C)	0/0	49.7	43.6		54.4	51.7
	0/1	42.3	41.0		34.6	37.9
	1/1	8.0	15.4		11.0	10.3
RPD grading available, n		184	17		*	*
RPD present, %		81.0	82.4	0.98	*	*	*

Abbreviations: DA = disc area; GA = geographic atrophy; NV-AMD = neovascular age-related macular degeneration; RPD = reticular pseudodrusen; SD = standard deviation.

^*No meaningful data available, since only 22 out of the 456 eyes had grading available for reticular pseudodrusen presence (as fundus autofluorescence imaging was very rare at the AREDS2 baseline study visit).

Geographic atrophy enlargement rate, according to subsequent development of exudative neovascular age-related macular degeneration

The results of the regression analyses are shown in Table 2. In the incident cohort, a significant interaction (p=0.037) was observed between GA enlargement rate (with the square root transformation) and the subsequent development of exudative NV-AMD. The GA enlargement estimate was 0.29 mm/year (95% CI 0.27–0.30) for the group without subsequent exudative NV-AMD. By contrast, the estimate was 0.20 mm/year (0.12–0.28) for the group with subsequent exudative NV-AMD. Similar results were obtained in analyses without the square root transformation: a significant interaction (p=0.004) was observed, and the GA enlargement estimates were 1.13 mm²/year (1.05–1.20) and 0.62 mm²/year (0.29–0.96), respectively.

Table 2.

Mixed-model regression of geographic atrophy area according to the presence or absence of subsequent exudative neovascular age-related macular degeneration.

	Number of eyes (participants)	Change per year in square root of GA area			Change per year in GA area
		Estimate (mm)	95% CI (mm)	P*	Estimate (mm)	95% CI (mm)	P*
Incident GA without subsequent exudative NV-AMD	684 (561)	0.29	0.27–0.30	0.037	1.13	1.05–1.20	0.004
Incident GA with subsequent exudative NV-AMD	73 (73)	0.20	0.12–0.28		0.62	0.29–0.96
Prevalent GA without subsequent exudative NV-AMD	393 (321)	0.28	0.26–0.29	0.37	1.41	1.31–1.51	0.87
Prevalent GA with subsequent exudative NV-AMD	63 (59)	0.31	0.24–0.37		1.38	1.10–1.75

Abbreviations: CI = confidence interval; GA = geographic atrophy; NV-AMD = neovascular age-related macular degeneration.

^*P values for interaction between presence/absence of subsequent exudative neovascular age-related macular degeneration and years from first appearance of geographic atrophy.

However, in the prevalent cohort, no significant interaction was observed (p=0.37). The GA enlargement estimate was 0.28 mm/year (0.26–0.29) for the group without subsequent exudative NV-AMD, and 0.31 mm/year (0.24–0.37) for the group with subsequent exudative NV-AMD. Similar results were obtained in analyses without the square root transformation: no significant interaction was observed (p=0.87), and the GA enlargement estimates were 1.41 mm²/year (1.31–1.51) and 1.38 mm²/year (1.01–1.75), respectively.

In separate analyses of the incident cohort, the analyses were repeated while limiting follow-up to three years or two years (i.e. three or two years after first emergence of GA). This was done because of the low numbers of eyes in the incident cohort with four or five years of follow-up (Supplementary Table 1). The results are shown in Table 3. In the analyses limiting follow-up to three years, a significant interaction (p=0.041) was observed between GA enlargement rate (with square root transformation) and subsequent exudative NV-AMD. The GA enlargement estimate was 0.29 mm/year (0.27–0.31) for the group without subsequent exudative NV-AMD and 0.21 mm/year (0.13–0.29) for the other group. In the analyses limiting follow-up to two years, again, a significant interaction (p=0.031) was observed. The GA enlargement estimates were 0.30 mm/year (0.27–0.32) and 0.20 mm/year (0.11–0.28), respectively.

Table 3.

Mixed-model regression of geographic atrophy area according to the presence or absence of subsequent exudative neovascular age-related macular degeneration, in the incident geographic atrophy cohort, according to different follow-up periods.

	Number of eyes (participants)	Change per year in square root of GA area			Change per year in GA area
		Estimate (mm)	95% CI (mm)	P*	Estimate (mm)	95% CI (mm)	P*
Full duration of follow-up included
Incident GA without subsequent exudative NV-AMD†	684 (561)	0.29	0.27–0.30	0.037	1.13	1.05–1.20	0.004
Incident GA with subsequent exudative NV-AMD†	73 (73)	0.20	0.12–0.28		0.62	0.29–0.96
Follow-up limited to 3 years only
Incident GA without subsequent exudative NV-AMD	684 (561)	0.29	0.27–0.31	0.041	1.12	1.03–1.20	0.007
Incident GA with subsequent exudative NV-AMD	73 (73)	0.21	0.13–0.29		0.65	0.32–0.98
Follow-up limited to 2 years only
Incident GA without subsequent exudative NV-AMD	684 (561)	0.30	0.27–0.32	0.031	1.08	0.98–1.18	0.010
Incident GA with subsequent exudative NV-AMD	73 (73)	0.20	0.11–0.28		0.60	0.25–0.95

Abbreviations: CI = confidence interval; GA = geographic atrophy; NV-AMD = neovascular age-related macular degeneration.

^*P values for interaction between presence/absence of subsequent exudative neovascular age-related macular degeneration and years from first appearance of geographic atrophy.

^†Same results as in Table 2 (shown for comparison).

Discussion

Main findings, potential explanations, and implications

In this exploratory retrospective analysis of participants from the AREDS2, a large prospective clinical trial, the rate of GA enlargement was variably associated with the subsequent development of exudative NV-AMD, depending on whether eyes had GA present at baseline. In the incident cohort, slower GA enlargement was observed in eyes with subsequent development of exudative NV-AMD than those without. However, this was not found in the prevalent cohort, whose eyes had larger baseline lesion sizes. These findings may be consistent with the idea that the presence of non-exudative MNV near a GA lesion can slow GA enlargement only in its local vicinity, i.e., enabling a detectable overall slowing of the enlargement of smaller GA lesions but no detectable slowing in larger GA lesions. Importantly, this relies on using subsequent exudative MNV as a proxy for the existence of presumed non-exudative MNV during the time interval that GA enlargement was assessed.

In the incident cohort, the difference between GA enlargement rates according to subsequent NV-AMD was relatively large, with estimates of 0.20 versus 0.29 mm/year. The estimate of 0.29 mm/year is similar to that from previous analyses of all eyes with incident GA in the AREDS2 cohort, irrespective of subsequent NV-AMD development (0.28 mm/year, 95% CI 0.27–0.30).⁵ By contrast, the estimate of 0.20 mm/year demonstrated by the subset of eyes developing subsequent exudative NV-AMD was statistically distinct (p=0.037) and 29% lower than this overall rate.

Importantly, we were unable to account for the slow GA enlargement rates observed for these eyes on the basis of any clinical, imaging, or genetic factors. Previous analyses of the AREDS2 dataset, along with those from other studies, have demonstrated multiple factors associated with altered enlargement rate.^5,6 Factors associated with faster GA enlargement include GA presence in the fellow eye, GA characteristics (intermediate size, no central involvement, and multifocal configuration), genetic variants (including ARMS2, C3, and APOE genotype), and possibly current smoking status. However, in the incident GA cohort, no significant differences in these factors were observed between the two groups. Small non-significant differences in the proportions with central GA and with bilateral GA might lead to expectations of slightly faster and slightly slower GA enlargement, respectively, in the group with subsequent exudative NV-AMD. Since these differences were small and in opposite directions, one might expect minimal overall effect.

However, similar findings were not observed in the prevalent cohort. First, it is possible that the positive findings in the incident cohort are not real (i.e. type 1 error). Related to this, the number of eyes in the incident cohort with GA followed by exudative NV-AMD and with follow-up of at least two years was 30 (compared to 49 in the prevalent cohort). Conversely, it is possible that the negative findings in the prevalent cohort are not real (i.e. type 2 error). However, this seems unlikely because not even any numerical difference in the appropriate direction was observed between the two estimates. Third, it is possible that both findings were real and a genuine reason exists for the different results between the two cohorts. The main difference between the two cohorts was smaller baseline GA size in the incident cohort. Hence, the potential effect of non-exudative MNV on GA enlargement (if real) might be relatively small and localized¹⁸; this effect might therefore have been sufficient to generate a signal for smaller GA lesions but not for larger ones. For example, in one previous report¹³, non-exudative MNV was observed along one margin of GA and was associated with slower GA enlargement selectively at that edge. Additionally, in another study with a larger mean baseline GA size than that of our incident cohort (7.2 mm² vs 1.8 mm², respectively), the protective effect of non-exudative MNV on GA enlargement was localized to the GA area adjacent to the non-exudative MNV but, globally, any protective effects were undetectable.¹⁸ Therefore, the fact that non-exudative MNV might cover a larger proportion of the vulnerable area adjacent to GA in newer and smaller GA lesions might potentially account for the difference.

However, alternative reasons might explain these results. A genuine association might be present without causation: one or more unknown risk factors might be shared between increased risk of exudative NV-AMD and slower GA enlargement. In addition, a genuine association might be present through reverse causation: it is possible that slower GA enlargement might increase the risk of exudative NV-AMD, through unknown mechanisms.

Potential mechanisms

A potential explanation for the slower enlargement of incident GA that subsequently develops exudative MNV is the presence of non-exudative MNV. A large majority of cases of exudative MNV are thought to exist as non-exudative MNV prior to the onset of exudation.¹⁶ Non-exudative MNV is thought to be comprised primarily of type 1 MNV and occasionally type 3 MNV.^16,28 The presence of type 1 MNV has also been associated with decreased incidence of GA.^17,29,30 In GA, the region of abnormal RPE has been shown to extend beyond the visible boundary of RPE atrophy on fundus autofluorescence imaging, and these abnormalities are thought to precede the development of RPE atrophy.^31–33 It has therefore been hypothesized that type 1 MNV can penetrate Bruch’s membrane to form a capillary-like network in the sub-RPE space, providing nutrients and oxygen to the RPE, enhancing the interchange between the RPE and choroid, and potentially compensating for areas of choriocapillaris degeneration.^34,35 Indeed, a recent case study of non-exudative type 1 MNV using histopathologic correlation observed that the MNV in the sub-RPE space appeared structurally similar to native choriocapillaris, with similar fenestrations, caveolae, and density.³⁶ Moreover, the outer retina overlying the non-exudative MNV was preserved in terms of retinal thickness and preserved lamination. In contrast, type 2 MNV has not been reported to occur as non-exudative MNV.³⁰ According to reports characterizing MNV subtypes in exudative NV-AMD, approximately 38–40% of newly diagnosed, treatment-naïve NV-AMD cases are comprised of type 1 MNV.^30,37

While the mechanisms underlying known associations between GA enlargement rates and AMD genotype have not been fully elucidated, potential links between GA enlargement rates and non-exudative MNV presence have been suggested. The C3 variant rs2230199, which is associated with increased risk of late AMD,²⁵ is also associated with slower GA enlargement.^5,38 Conversely, the APOE variant rs73036519, which is associated with decreased risk of late AMD,²⁵ is associated with faster GA enlargement.⁵ This might suggest that these variants could lead to increased risk of non-exudative MNV, which would lead in turn to slower GA enlargement. However, the C3 genotype data observed in the current study do not seem sufficient to account for the differences observed.

The results of the Filly phase II trial of pegcetacoplan (a pegylated C3 inhibitor peptide) showed that local administration of this C3 inhibitor caused a dose-dependent decrease in GA enlargement.⁷ However, the results were also consistent with a dose-dependent increase in the development of exudative NV-AMD.³⁹ The authors hypothesized that local C3 inhibition may decrease phagocytic attack by innate immune cells, thereby protecting the RPE cells and photoreceptors, but potentially also permitting the choriocapillaris endothelium to develop MNV more easily. Conversely, one might consider whether slower GA enlargement may have occurred as a direct result of NV-AMD induction³⁹, which would support the ideas discussed in the current study. However, in sensitivity analyses of the Filly trial that excluded the eyes that developed exudative NV-AMD, the results were similar to the primary results. This argues against the idea that exudative MNV was entirely responsible for the slower GA enlargement in the trial, though it is possible that some eyes developed non-exudative MNV (leading to slower GA enlargement) but remained undiagnosed. Overall, these considerations demonstrate the complexity of local complement regulation and highlight the potentially delicate balance between GA and NV-AMD.

Comparison with literature

The results of this study are consistent with those from several recent studies reporting an association between the presence of non-exudative MNV and slower GA enlargement. Heiferman et al. described a case of GA associated with non-exudative MNV that enlarged at a much slower rate (without square root transformation) than four cases of GA without non-exudative MNV.¹⁴ In a study characterizing GA associated with non-exudative MNV, all five cases with subfoveal or parafoveal non-exudative MNV (confirmed by OCT angiography) were associated with foveal sparing.¹³ Moreover, in 13 out of 14 cases, the area of non-exudative MNV was spared from atrophy at the last follow-up study visit. Pfau et al. showed quantitatively, in seven eyes, that the presence of non-exudative type 1 MNV was associated with greatly decreased localized enlargement of atrophy (i.e. to involve the area overlying the MNV).¹⁸ They demonstrated similar results for 10 eyes with exudative type 1 MNV. Similarly, other recent studies have demonstrated a lower GA incidence and/or slower GA enlargement in the setting of exudative type 1 MNV. In one retrospective study, all nine eyes with mature exudative NV-AMD (having received at least 50 anti-VEGF injections) that demonstrated apparent resistance to atrophy development and enlargement were associated with type 1 MNV.¹⁷ Furthermore, using en-face OCT images overlaid with OCT angiography, Christenbury et al. reported that, out of 11 eyes with exudative type 1 MNV that developed macular atrophy, nine developed atrophy eccentric to the MNV (as opposed to overlying the MNV).¹⁹

Study strengths and limitations

To our knowledge, this is the largest study of eyes with GA followed by exudative NV-AMD, comprising 136 eyes in both cohorts. However, the number of these eyes with at least two years of follow-up was lower, particularly in the incident cohort, at 30 (incident cohort) and 49 (prevalent cohort). For comparison, one of the largest previous studies examined 17 eyes with a combination of macular atrophy and MNV (either exudative or non-exudative).¹⁸ In addition, the prospective nature of data acquisition, including uniform reading center grading of all images by masked graders, is an important strength, since the data were not subject to observer bias. Most previous studies have been retrospective. The eyes in this dataset were well characterized for factors known to be associated with altered speed of GA enlargement, including multiple clinical, imaging, and genetic factors. Importantly, the square root transformation was used to allow a fairer comparison of GA enlargement rates between eyes with different baseline sizes.^5,27

Similar to all previous studies in this area, this analysis was limited by its retrospective (unplanned) nature. Directly related to this, the most important limitation is that this study relied on using subsequent exudative MNV as a proxy for a period of presumed non-exudative MNV. Hence, there was no direct diagnosis or visualization of non-exudative MNV using OCT angiography or other imaging in any case. For similar reasons, the potential duration of presumed non-exudative MNV is unknown. In addition, the number of eyes with RPD grading available was low; however, in the subset with grading available in the incident GA cohort, the proportions with RPD present were very similar between the two groups.

Together with increasing recognition of non-exudative MNV as a clinical entity, this study suggest that some comments made in a previous report of the same dataset should be updated. In particular, in AREDS2 Report 16, the authors stated: “It was also possible to exclude from the analyses those eyes with simultaneous or prior neovascular AMD, so as to examine more accurately the behavior of “pure GA” (i.e., without any possible influence from coexistent neovascular disease and/or anti-vascular endothelial growth factor therapy). In a similar way, it was also possible to analyze the emergence of neovascular disease from pre-existing GA, specifically in eyes without a history of neovascular AMD.”⁵ Specifically, the authors were able to exclude eyes with exudative neovascular AMD from the analyses.

Conclusions

In this large but exploratory analysis of participants from the AREDS2, in eyes with recent appearance of GA, the subsequent appearance of exudative NV-AMD was associated with slower GA enlargement. This difference could not be explained by the clinical, imaging, and genetic characteristics measured in this study. Interestingly, similar findings were not observed in the prevalent cohort. The results for the prevalent cohort might represent the biological truth, with those for the incident cohort arising through type 1 error. Alternatively, if both are true, the difference might be reconciled by considering differences in baseline lesion size. This would be consistent with the idea that non-exudative NV-AMD may slow GA enlargement in its vicinity only, i.e., locally but not globally. However, this study relied on using subsequent exudative NV-AMD as a proxy for a period of presumed non-exudative NV-AMD, i.e., non-exudative NV-AMD was not visualized or diagnosed directly. These findings are important to help understand the natural history of GA followed spontaneously by exudative NV-AMD, which occurs commonly in AMD. The results may also be important to help interpret the findings of ongoing clinical trials with a high rate of exudative NV-AMD following GA. The findings warrant further evaluation in prospective studies with direct visualization of non-exudative and exudative MNV alongside GA measurements over time.