Incidence of macular atrophy following untreated neovascular age-related macular degeneration: Age-Related Eye Disease Study Report 40

Abstract

Purpose

To report the natural history of untreated neovascular age-related macular degeneration (NV), concerning risk of subsequent macular atrophy.

Design

Prospective cohort within a randomized, controlled trial of oral micronutrient supplements.

Participants

Age-Related Eye Disease Study (AREDS) participants, aged 55–80 years, who developed NV during follow-up (1992–2005), prior to the advent of anti-VEGF therapy.

Methods

Stereoscopic color fundus photographs were collected at annual study visits and graded centrally for features of late AMD. Incident macular atrophy after NV was examined by Kaplan-Meier analysis and proportional hazards regression.

Main outcome measures

Incident macular atrophy following NV, including risk of central involvement.

Results

Of the 4,757 AREDS participants, 708 eyes (627 participants) developed NV during follow-up and were eligible for analysis. The cumulative risks of incident macular atrophy after untreated NV were 9.6% (standard error 1.2%), 31.4% (2.2%), 43.1% (2.6%), and 61.5% (4.3%) at two, five, seven, and 10 years, respectively. This corresponded to a linear risk of 6.5%/year. The cumulative risk of central involvement was 30.4% (3.2%), 43.4% (3.8%), and 57.0% (4.8%) at first appearance of atrophy, two years, and five years, respectively. Geographic atrophy (GA) in the fellow eye was associated with increased risk of macular atrophy after NV (HR 1.70, 1.17–2.49; p=0.006). However, higher 52-SNP AMD Genetic Risk Score was not associated with increased risk of macular atrophy after NV (hazard ratio 1.03, 95% CI 0.90–1.17; p=0.67). Similarly, no significant differences were observed according to the SNPs CFH rs1061170, CFH rs10922109, ARMS2 rs10490924, or C3 rs2230199.

Conclusions

The rate of incident macular atrophy following untreated NV is relatively high, increasing linearly over time and affecting half of eyes by eight years. Hence, factors other than anti-VEGF therapy are involved in atrophy development, including natural progression to GA. Comparison with studies of treated NV suggests it may not be necessary to invoke a large effect of anti-VEGF therapy on inciting macular atrophy, though a contribution remains possible. Central involvement by macular atrophy is present in approximately one third of eyes at the time atrophy develops (similar to pure GA) and increases linearly to half of eyes at three years.

Précis

In the Age-Related Eye Disease Study, the incidence of macular atrophy following neovascular age-related macular degeneration was high at 31% (five years) and 62% (ten years), despite absence of anti-VEGF therapy.

Introduction

Age-related macular degeneration (AMD) is the most common cause of legal blindness in developed countries and is predicted to affect 196 million people worldwide by 2020.^1–3 Late AMD, the stage associated with severe visual loss, occurs in two forms, neovascular AMD (NV) and geographic atrophy (GA). Untreated NV has a very poor prognosis: 76% of eyes have 20/200 or worse visual acuity three years after NV development.⁴ The advent of anti-VEGF therapy revolutionized NV treatment. In 2006, the Minimally Classic/Occult Trial of the Anti-VEGF Antibody Ranibizumab in the Treatment of Neovascular Age-Related Macular Degeneration (MARINA) demonstrated that eyes with NV treated with ranibizumab had a nearly 10-fold decreased risk of severe vision loss at two years, compared with eyes that received sham.⁵ The efficacy of anti-VEGF for NV was corroborated in many clinical trials and became the standard of care.^6–8

However, long-term follow-up of these clinical trial participants demonstrated that eyes with NV receiving anti-VEGF therapy had high rates of subsequent macular atrophy. The Comparison of AMD Treatments Trials (CATT) reported that the proportion of eyes with incident macular atrophy five years after starting anti-VEGF therapy for NV was 38%.⁹ In addition, using a different definition for macular atrophy, the Seven-Year Observational Update of Macular Degeneration Patients Post-MARINA/ANCHOR and HORIZON Trials (SEVEN-UP), which assessed eyes that were originally treated with ranibizumab in these trials, found the proportion of eyes with macular atrophy at seven years to be 98%; indeed, by this point, the macular atrophy had become the primary anatomic determinant of visual outcomes.¹⁰

This led some authors to question whether anti-VEGF therapy itself might increase the tendency to develop macular atrophy, which would have potential implications for clinical practice. In the Inhibit VEGF in Age-Related Choroidal Neovascularization (IVAN) trial, the authors reported a possible increased risk of macular atrophy in eyes treated with anti-VEGF for two years with a monthly regimen compared with a pro re nata (PRN) regimen (odds ratio 1.47, 95% CI 1.03–2.11).¹¹ However, recent reanalyses using a specially designed grading protocol did not replicate these findings (odds ratio 1.01 per treatment cycle, 95% CI 0.91–1.13).¹² Similarly, in the CATT, the authors observed increased risk of macular atrophy (hazard ratio 1.59, 1.17–2.16) in eyes treated for two years with monthly versus PRN anti-VEGF¹³, though no significant difference was observed at five years (i.e., three years after the end of the two-year clinical trial).⁹ In post hoc analysis of the HARBOR study, monthly versus PRN anti-VEGF therapy for two years was not associated with a significant difference (hazard ratio 1.29, 95% CI 0.99–1.68).¹⁴

However, in these and similar studies of NV treated with anti-VEGF therapy, regarding macular atrophy following NV, it is extremely difficult to separate out the relative contributions of (i) natural progression to GA, as the final common pathway in AMD disease progression (i.e., independent of NV presence), (ii) NV-associated macular atrophy (i.e., through NV/exudation-related damage to the retinal pigment epithelium (RPE) and nearby cells), and (iii) potential contribution of anti-VEGF therapy.

The natural history of untreated NV would provide valuable information regarding the potential contribution of anti-VEGF therapy, since removing component (iii) would enable an understanding of the natural progression from NV to macular atrophy through components (i) and (ii) only. The Age-Related Eye Disease Study (AREDS) was a multicenter prospective study of the clinical course of AMD and age-related cataract, as well as a phase III RCT designed to assess the effects of nutritional supplements on AMD and cataract progression previously.¹⁵ Since the AREDS occurred prior to the advent of anti-VEGF therapy, this study provided an opportunity to evaluate long-term natural history data.

The primary aim of this study was to analyze the risk of incident macular atrophy following untreated NV. Secondary aims included examining the risk of central involvement by incident macular atrophy at first appearance of the atrophy, subsequent progression to central involvement over time, and potential risk factors for macular atrophy.

Methods

Study population and procedures for the Age-Related Eye Disease Study

The study design for the AREDS has been described previously.¹⁵ In short, 4,757 participants aged 55 to 80 years were recruited between 1992 and 1998 at 11 retinal specialty clinics in the United States. Based on fundus photographic gradings, best-corrected visual acuity, and ophthalmoscopic evaluations, participants were enrolled into one of several AMD categories. Of the 4,757 participants, 3,640 took part in the AMD trial. The participants were randomly assigned to placebo, antioxidants, zinc, or the combination of antioxidants and zinc. The primary endpoint was progression to late AMD (defined as NV or central GA).

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 tenets of the Declaration of Helsinki and pre-dated the Health Insurance Portability and Accountability Act.

Questionnaires administered at the baseline and subsequent study visits collected information that included medications, adverse events and treatment compliance. At baseline and annual study visits, comprehensive eye examinations were performed by certified study personnel using standardized protocols, and stereoscopic color fundus photographs (CFP) were captured. Progression to late AMD was defined by the study protocol based on the grading of CFP, as described previously. Additional ‘event photographs’ were taken at any visit in which a ≥10 letter reduction in best corrected visual acuity was noted, compared to the baseline visit, or when the clinical examination suggested NV or central GA. CFPs were graded using a standardized protocol by certified graders at a reading center (Fundus Photographic Reading Center, University of Wisconsin). GA/macular atrophy was defined as a sharply demarcated, usually circular zone of partial or complete depigmentation of the RPE, typically with exposure of underlying large choroidal blood vessels, that must be as large as circle I-1 (1/8 disk diameter, i.e., approximately 215μm). An area of RPE atrophy within or adjacent to a subretinal fibrous scar was not considered macular atrophy. The CFPs were deemed to be of gradable quality in 99.4% of the 48,998 sets of CFPs evaluated in the first 10 years of the study.¹⁶ All of the CFPs were graded independently, and the graders were masked to the grades from previous visits. Contemporaneous replicate grading exercises showed 96% exact agreement between graders, with weighted kappa values of 0.71–0.73 for GA suggesting precise grading.¹⁶

The AREDS cohort was followed between 1992 and 2005, prior to the United States Food and Drug Administration approval of ranibizumab for NV. Hence, the participants were offered standard of care treatment for NV by their local retinal specialists. For the early years of the AREDS follow-up, this generally comprised no active treatment; rarely, participants might undergo laser photocoagulation, following the results of the Macular Photocoagulation Study (MPS).¹⁷ In the later years of follow-up, again, this normally meant no active treatment; in some rare cases, participants might receive photodynamic therapy (PDT), following publication of the Treatment of Age-related macular degeneration with Photodynamic therapy (TAP) and Verteporfin in Photodynamic Therapy (VIP) studies.^18,19

Eligibility criteria and statistical analysis

The unit of analysis was at the eye level. In the AREDS participants, eyes that developed NV without prior/simultaneous GA during study follow-up were eligible for analysis. Hence, eyes that developed NV and GA/macular atrophy simultaneously (i.e., both NV and GA graded positive for the first time in the same CFP) were excluded from analysis (since the primary outcome was incident macular atrophy subsequent to NV). In addition, eyes that underwent macular laser photocoagulation and/or PDT for NV, prior to NV positive grading on CFP at an AREDS study visit, were also excluded.

The rates of incident macular atrophy after NV were examined by Kaplan-Meier survival analysis. In addition, in these eyes with incident macular atrophy after NV, the rates of atrophy progression to central involvement were calculated by Kaplan-Meier survival analysis. Study eyes that underwent macular laser photocoagulation and/or PDT for NV were censored at the time of treatment to prevent confounding with respect to the development of macular atrophy.

Further Kaplan-Meier survival analyses were performed with stratification by (i) AMD genotype and (ii) fellow eye GA status. Regarding AMD genotype, the following genetic characteristics were pre-specified for analysis: AMD Genetic Risk Score (GRS)²⁰, CFH rs1061170 (Y402H) risk alleles, CFH rs10922109 protective alleles (the lead variant at this locus in a large genome-wide association study (GWAS) of late AMD²⁰), ARMS2 rs10490924 risk alleles (the locus with highest attributable risk of late AMD), and C3 rs2230199 risk alleles. The AMD GRS is a weighted risk score for late AMD, based on 52 independent variants at 34 loci identified in a large GWAS²⁰; it was calculated for each participant according to methods described previously.²⁰ Participants can be divided into three groups (0–2); group 0 participants are those with a GRS below the mean GRS of a control population without late AMD, while group 1 and 2 participants are those with a higher GRS (below and above the median GRS of a large population with late AMD, respectively). Regarding fellow eye GA status, the analyses were limited to those participants with one but not both eyes in the current study; fellow eye GA status (present or absent) was defined at the time of incident NV in the study eye. In addition, the demographic characteristics of participants (age, sex, educational level, and smoking status, as well as AREDS treatment assignment) were analyzed separately according to presence/absence of progression to macular atrophy, as potential risk factors.

For all of these variables, statistical testing was performed using proportional hazards regression. For analyses of the demographic characteristics, the full study population was used (n=708 eyes, taking into account correlation between the eyes of an individual); for those of the genetic characteristics, the study population was restricted to those participants with genotype data available (n=433 eyes, taking into account correlation between the eyes of an individual); for those of the fellow eye GA status, the analyses were limited to those participants with one but not both eyes in the current study (n=546 participants/eyes). All analyses were conducted using SAS version 9.4 (SAS Inc, Cary NC).

Results

A total of 1,042 eyes developed NV during the AREDS follow-up. Of these, 265 eyes (25.4%) were excluded because of prior or simultaneous GA/macular atrophy and an additional 69 eyes (6.6%) were excluded because of laser photocoagulation prior to NV detection. The remaining 708 eyes (67.9%) of 627 participants were considered at risk of incident macular atrophy and eligible for subsequent analyses. The baseline characteristics of these eyes are shown in Table 1. The number of study eyes censored during follow-up because of laser photocoagulation and/or PDT for NV was 239 (comprising 130 eyes with treatment at the same study visit as NV detection and 109 eyes with treatment after NV detection).

Table 1.

Demographic and genetic characteristics of the study population

	All eyes (n=708 for demographic characteristics; n=433 for genetic characteristics)	Eyes that did not progress to macular atrophy after neovascular AMD (n=504; n= 293)	Eyes that did progress to macular atrophy after neovascular AMD (n=204; n=140)
Age at neovascular AMD (years: mean, SD)	75.5 (5.7)	76.3 (5.4)	73.6 (5.8)
Female (%)	62.0	61.7	62.7
Education level (%)
High school or less	41.8	40.5	45.1
Some college	28.1	29.6	24.5
College graduate	30.1	30.0	30.4
Smoking status (%)
Never	38.8	38.9	38.7
Former	51.3	51.6	50.5
Current	9.9	9.5	10.8
Body mass index (%)
≤25	33.1	35.3	27.5
26–30	40.5	39.3	43.6
>30	26.4	25.4	28.9
AREDS treatment assignment (%)
Placebo	27.7	28.4	26.0
Antioxidants	25.7	25.4	26.5
Zinc	24.6	23.2	27.9
Antioxidants & zinc	22.0	23.0	19.6
Follow-up time from neovascular AMD to last visit (years: mean, SD)	3.2 (3.4)	2.1 (2.8)	6.1 (3.1)
AMD Genetic Risk Score (%)
Group 0	9.2	9.2	9.3
Group 1	36.5	39.6	30.0
Group 2	54.3	51.2	60.7
CFH rs1061170 alleles (%)
0	16.4	16.0	17.1
1	45.5	45.4	45.7
2	38.1	38.6	37.1
CFH rs10922109 alleles (%)
0	66.5	67.6	64.3
1	30.3	29.7	31.4
2	3.2	2.7	4.3
ARMS2 rs10490924 alleles (%)
0	32.6	34.1	29.3
1	47.1	47.4	46.4
2	20.3	18.4	24.3
C3 rs2230199 alleles (%)
0	8.8	8.9	8.6
1	40.9	39.2	44.3
2	50.3	51.9	47.1

Abbreviations: AMD=age-related macular degeneration; AREDS=Age-Related Eye Disease Study; SD=standard deviation

During mean follow-up of 3.2 years (SD 3.4), of the 708 eyes at risk, 204 eyes (28.8%) developed incident macular atrophy, at a mean interval of 3.8 years (SD 2.4) from the time of NV detection. By Kaplan-Meier analysis, the cumulative risk of developing macular atrophy after untreated NV was 9.6% (standard error (SE) 1.2%), 31.4% (SE 2.2%), 43.1% (SE 2.6%), and 61.5% (SE 4.3%) at two, five, seven, and 10 years, respectively (Figure 1). The time point at which 50% of eyes were affected was 8.0 years (95% confidence interval (CI) 7.3–9.6). The cumulative progression to macular atrophy over time appeared strikingly linear. Considered in this way, the progression corresponded (by linear regression) to a linear risk of 6.5% per year.

Figure 1.

Kaplan-Meier curve of the progression of eyes to incident macular atrophy following neovascular age-related macular degeneration, without treatment.

Of the 204 eyes with incident macular atrophy after NV, the proportion with central involvement at first appearance of the atrophy was 30.4% (62 eyes). Of the 142 eyes whose incident macular atrophy was non-central at first appearance, the proportion that progressed to central involvement (over mean follow-up of 7.6 years (SD 1.9)) was 22.5% (32 eyes). Hence, the total number of eyes that developed incident central macular atrophy at any point after untreated NV was 94 (representing 13.3% of all eyes at risk).

By Kaplan-Meier analysis, out of the 204 eyes with incident macular atrophy after NV, the cumulative risk of developing central involvement by atrophy was 30.4% (SE 3.2%) at baseline (i.e., at first appearance of macular atrophy), 43.4% (SE 3.8%) at two years, and 57.0% (SE 4.8%) at five years (Figure 2). The time point at which 50% of eyes had central involvement was 3.1 years (95% CI 2.0–5.1). Excluding those eyes with central involvement at baseline, this corresponded (by linear regression) to a linear risk of 8.8% per year. Finally, considering progression to macular atrophy and progression to central involvement simultaneously, the Kaplan-Meier analysis of the cumulative risk of incident central macular atrophy following NV is shown in Figure 3.

Figure 2.

Kaplan-Meier curve of the progression of eyes with incident macular atrophy (after neovascular AMD, without treatment) to central involvement.

Figure 3.

Kaplan-Meier curve of the progression of eyes to incident central macular atrophy following neovascular age-related macular degeneration, without treatment.

Potential risk factors for the development of macular atrophy after untreated NV were analyzed by proportional hazards regression, including demographic characteristics, AREDS treatment assignment, AMD genotype, and fellow eye GA status (Table 2). No demographic characteristic was significantly associated with progression to macular atrophy: age (p=0.81), sex (p=0.66), smoking status (p=0.41), education level (p=0.37), or body mass index (0.20). Similarly, no significant difference was observed according to AREDS treatment assignment (p=0.66).

Table 2.

Potential risk factors for progression to macular atrophy following untreated neovascular age- related macular degeneration: results of proportional hazards regression

	Hazard ratio	95% confidence interval	P
Age (per year)*	1.00	0.96–1.03	0.81
Male	0.93	0.69–1.27	0.66
Education level
High school or less (reference)
Some college	0.84	0.59–1.19	0.33
College graduate	0.86	0.61–1.20	0.37
Smoking status
Never (reference)
Former	0.96	0.70–1.32	0.79
Current	1.23	0.75–2.04	0.41
Body mass index
≤25 (reference)
26–30	1.22	0.85–1.74	0.28
>30	1.28	0.88–1.88	0.20
AMD Genetic Risk Score†	1.03	0.90–1.17	0.67
CFH rs1061170 alleles
0 (reference)
1	0.89	0.57–1.39	0.62
2	0.88	0.56–1.41	0.60
CFH rs10922109 alleles
0 (reference)
1	1.02	0.71–1.47	0.92
2	0.81	0.42–1.53	0.51
ARMS2 rs10490924 alleles
0 (reference)
1	0.99	0.67–1.47	0.98
2	0.97	0.60–1.55	0.89
C3 rs2230199 alleles
0 (reference)
1	0.98	0.54–1.78	0.94
2	0.78	0.42–1.45	0.44
Geographic atrophy in fellow eye‡	1.70	1.17–2.49	0.006

Abbreviations: AMD=age-related macular degeneration

^*n=708 eyes for age and the other demographic characteristics, with all characteristics considered simultaneously in the same multivariable model

^†n=433 eyes for all genetic characteristics, with each considered separately in univariate models; the AMD Genetic Risk Score was treated as a continuous variable

^‡n=546 eyes for fellow eye analysis, including only those participants with one but not both eyes in the study population

Participants with a higher AMD GRS were not significantly more or less likely to develop macular atrophy after NV, compared to those with a lower GRS (Table 2). The hazard ratio associated with higher GRS was 1.03 (95% CI 0.90–1.17, p=0.67). Similarly, no significant differences were observed according to the individual loci examined: CFH rs1061170, CFH rs10922109, ARMS2 rs10490924, or C3 rs2230199.

However, participants whose fellow eye had GA (at the time of incident NV in the study eye) were significantly more likely to progress to macular atrophy (Table 2; Figure 4). For example, the cumulative risk of developing macular atrophy by five years after NV was 43.9% (SE 6.8%) for participants whose fellow eye had GA, versus 26.8% (SE 2.6%) for participants whose fellow eye did not have GA. According to proportional hazards regression, the hazard ratio for macular atrophy associated with GA in the fellow eye was 1.70 (1.17–2.49, p=0.006). In additional regression analyses that considered all variables simultaneously in the same model, the results were similar, i.e., fellow eye GA status was significant (p=0.01, hazard ratio 1.81, 1.13–2.91) and all other variables were non-significant (p>0.4 for each).

Figure 4.

Kaplan-Meier curve of the progression of eyes to incident macular atrophy following neovascular age-related macular degeneration, without treatment, stratified by the presence or absence of geographic atrophy in the fellow eye (at the time of neovascular age-related macular degeneration in the study eye).

Discussion

The AREDS represents the largest cohort of prospective data on the natural history of untreated NV, with respect to the development of macular atrophy, to our knowledge. Since anti-VEGF therapy has become the standard of care for NV, future data on untreated NV are unlikely, making this dataset potentially unique in history.

This study demonstrates that, without anti-VEGF therapy, incident macular atrophy occurs in approximately one third of eyes within five years of NV, and that one half of eyes are affected by eight years. Interestingly, the cumulative development of macular atrophy appears strikingly linear over this long time period of a decade without NV treatment. This prolonged linearity may represent contributions from component (i) described above (i.e., natural progression to GA, as the final common pathway in AMD disease progression, independent of NV presence) and perhaps also component (ii) (i.e., macular atrophy from NV/exudation-related damage to RPE and nearby cells), though separating their contributions is very difficult.

However, the AREDS grading definitions did not consider areas of RPE atrophy within or adjacent to subretinal fibrous scars as macular atrophy, and 131 of the 708 study eyes had a positive grading for subretinal fibrous scars at the time of incident NV. Hence, it is possible that these percentages are underestimates of the true proportions with macular atrophy. For these reasons, we repeated the Kaplan-Meier analyses with censoring of eyes at the point of subretinal fibrosis formation. Of the 577 study eyes, 81 progressed to macular atrophy and 496 did not. The rates of progression to macular atrophy were much lower than in the original analyses. Hence, many of the eyes with subretinal fibrosis did subsequently receive a positive grade for macular atrophy, even though the area within or adjacent to the fibrosis was not considered; for example, of the 131 eyes described above, 44 still received a subsequent grade of macular atrophy. These considerations may also emphasize the contribution of component (ii) to macular atrophy.

Additional findings in this study relate to central involvement, which is important for determining visual potential. This study shows that central involvement is already present at first occurrence of the atrophy in approximately one third of eyes. This is similar to the proportion with central involvement by pure GA (i.e., without preceding NV), as observed in the AREDS2 (i33% of eyes).²¹

Another point of comparison between these AREDS data (on macular atrophy after untreated NV) with AREDS2 data (on pure GA without prior NV) is the speed of progression to central involvement in eyes whose incident atrophy was non-central. In the current study, the rates of central involvement were 18.7% (SE 4.0%) by two years and 35.1% (SE 5.8%) by four years. The equivalent values for pure GA in AREDS2 were 32% and 57%. Although direct comparison between the two datasets is difficult, the lower rate of progression to central involvement for macular atrophy than pure GA might support the idea that enlargement of atrophy could be slower in the presence of NV.^21–23 However, alternative explanations include different distribution of starting locations for the macular atrophy versus pure GA and the fact that, in the AREDS, atrophy would not be graded as central in the presence of a central subretinal fibrous scar.

An additional finding in this study was the increased risk of progression to macular atrophy according to the presence of GA in the fellow eye. This is consistent with similar findings for treated NV in the CATT^9,13 and the HARBOR study¹⁴. This result, and its similarity to related findings for treated NV, may emphasize the relative contribution of component (i). However, the genetic analyses did not support the hypothesis that higher genetic load (as regards risk of late AMD) is associated with increased risk of macular atrophy after untreated NV. Together with the other results, this means that few risk factors are available to predict the likelihood of atrophy. The genetic results might potentially argue against the contribution of component (i) to the development of atrophy. However, all of the study eyes had already demonstrated the capability of progressing to late AMD and the large majority (91%) were from participants with a high GRS; this may help explain why AMD genotype is not a distinguishing feature in predicting progression to macular atrophy after untreated NV but is for predicting GA. We are not aware of previous studies that have analyzed the AMD GRS in this way. The only study to have examined genotype in macular atrophy after (treated) NV was the CATT, where the authors considered four SNPs; the data at five years suggested significant results for ARMS2 (hazard ratio 1.8 (1.2–2.7) for TT versus GG) and TLR3 (hazard ratio 0.5 (0.3–0.9) for TT versus CC).⁹

Comparison with literature

The purpose of comparing these findings with those in the literature (Table 3) was to examine similarities and differences between the behavior of eyes with NV that is treated (with anti-VEGF therapy) versus untreated (i.e., natural history). This should provide insights into the question of whether anti-VEGF therapy may cause and/or contribute to macular atrophy.

Table 3.

Proportion of study eyes with incident macular atrophy following neovascular age-related macular degeneration at specified time points: current study (of untreated neovascular disease) and comparison with literature (for neovascular disease treated with anti-VEGF therapy)

	Age-Related Eye Disease Study (no anti-VEGF therapy)	Comparison of Age- Related Macular Degeneration Treatments Trials	Inhibit VEGF in Age- Related Choroidal Neovascularization (revised grading 2019)	HARBOR	SEVEN-UP open- label extension study (following MARINA, ANCHOR, and HORIZON)
Eligibility criteria	Macular NV. No pre-existing/ simultaneous macular atrophy (current study)	NV with subfoveal involvement and no central macular atrophy	Macular NV without central macular atrophy	Subfoveal NV	Per MARINA, ANCHOR, and HORIZON trials
Minimum size definition for macular atrophy	215 μm	250 μm	175 μm	250 μm	175 μm
Imaging modalities used to define macular atrophy	CFP	FA and CFP	FA and/or CFP, aided by OCT	FA	FAF and red-free/FA
Proportion of study eyes with incident macular atrophy at specified time points from study baseline
2 years	9.6%	17%	25%*	29%	-†
5 years	31.4%	38%	-	-	-
7 years	43.1%	-	-	-	98%†
10 years	61.5%	-	-	-	-

^*Total of 35%, including macular atrophy that was already present at baseline

^†But, in the subset of eyes with fluorescein angiography available at two years and seven years, 95% (two years) and 100% (seven years) Abbreviations: CFP=color fundus photograph; FA=fluorescein angiography; FAF=fundus autofluorescence; NV=neovascular age-related macular degeneration; OCT=optical coherence tomography

The CATT randomized 1,185 participants with NV to either ranibizumab or bevacizumab, administered either monthly or PRN, over two years.²⁴ Masked graders at a reading center assessed the presence of macular atrophy at two years and five years. The cumulative proportions of eyes with incident macular atrophy were 17% and 38%, respectively (Table 3).^9,13 These are higher than the equivalent values in the current study of 10% and 31%. However, the proportions in the current study may have been underestimates for three reasons: first, the AREDS grading definitions did not count atrophy that was within or adjacent to subretinal fibrous scars. Second, the sensitivity of detection was likely higher in the CATT, which used both fluorescein angiography (FA) and CFP.^9,13 Third, eyes that underwent laser photocoagulation or PDT for NV were censored in the current study. Overall, these considerations suggest that anti-VEGF therapy may make a relatively minor contribution to the risk of macular atrophy after NV. Alternatively, another possibility is that anti-VEGF therapy did make an important contribution to macular atrophy incidence in the CATT, but the rates are similar to those in the AREDS because, in the AREDS, the effect of removing component (iii) is compensated by increased effects of component (ii), i.e., increased NV/exudation-related RPE damage.

Like the CATT, the IVAN was a controlled trial of different anti-VEGF drugs and regimens: 610 participants were randomized to ranibizumab or bevacizumab, administered monthly or PRN.¹¹ Following recent regrading using a revised protocol¹², the proportion of eyes with incident macular atrophy at the primary endpoint of two years was 25% (Table 3). This is higher than the proportion observed in the current study of untreated NV (10%). However, in the IVAN, the size threshold for defining atrophy was much lower at 175 μm, and atrophy was defined using FA and/or CFP, assisted by optical coherence tomography (OCT).

In the HARBOR study, a recent post hoc analysis reported rates of incident macular atrophy of 21% at one year and 29% at two years (Table 3).¹⁴ Again, this is higher than the equivalent rates in the current study, though sensitivity for detecting atrophy was likely higher through the use of FA.

In the SEVEN-UP study, post hoc analysis reported macular atrophy rates (combining pre-existing and incident cases) as high as 98% of eyes at a mean of 7.3 years after enrollment.¹⁰ However, important contributing factors to this high rate include the imaging modality used (fundus autofluorescence) and the lower minimum size requirement (175 μm). In addition, the sample size was small (60 eyes) and may not be representative of the original study cohorts.

Strengths and limitations

The main strength of this study relates to its unique position in history, where a cohort of eyes with NV was followed longitudinally, prior to the advent of anti-VEGF injections. The study benefitted from large sample size and long follow-up time, permitting meaningful Kaplan-Meier analyses over a long period. Additional strengths include comprehensive data collection at set time-points and standardized reading center evaluation of images by masked graders using a uniform definition of macular atrophy.

One important limitation was that macular atrophy was assessed on CFP only, rather than by modern multimodal approaches including OCT and fundus autofluorescence. However, since the widespread use of these imaging modalities post-dated the advent of anti-VEGF therapy, we presume that no AMD dataset will ever achieve the combination of untreated NV and multimodal imaging. Similarly, it would be ideal to compare treated and untreated NV in the same cohort of participants, for more direct comparisons of macular atrophy formation. Again, this is unlikely to occur. Finally, our dataset does not permit explicit considerations of intralesional versus extralesional macular atrophy (since NV lesion boundaries can not be assessed on CFP alone).

Conclusions

In summary, the AREDS demonstrates that the rate of incident macular atrophy in untreated NV is relatively high, occurring in approximately one third of eyes within five years of NV and one half by eight years. Hence, when the potential effect of anti-VEGF on progression to macular atrophy is removed, a relatively high rate of progression to macular atrophy is still observed. Factors other than anti-VEGF therapy are therefore involved in atrophy formation. This includes natural progression to GA (as the final common pathway in AMD disease progression, independent of NV presence); indeed, this study observed increased risk of macular atrophy in participants with pure GA in the fellow eye.

The progression rates in this study may be underestimates owing to detection methods and grading definitions. With this in mind, comparison with studies of treated NV suggests that it may not be necessary to invoke a large effect of anti-VEGF therapy on the development of macular atrophy. However, a smaller contribution of anti-VEGF therapy to macular atrophy remains possible, particularly if, in the current study, the effect of removing anti-VEGF therapy was exchanged for the effect of increased NV/exudation-associated RPE damage. Finally, in this study, central involvement by macular atrophy was present at the outset in about one third of eyes and rose to one half by three years. These data may be useful, since central macular atrophy is increasingly likely to become the main determinant of visual acuity and legal blindness following NV successfully treated by anti-VEGF therapy.

Abbreviations

AMD

age-related macular degeneration

AREDS

Age-Related Eye Disease Study

CATT

Comparison of Age-Related Macular Degeneration Treatments Trials

CFP

color fundus photograph

CI

confidence interval

FA

fluorescein angiography

GA

geographic atrophy

GRS

genetic risk score

GWAS

genome-wide association study

HR

hazard ratio

IVAN

Inhibit VEGF in Age-Related Choroidal Neovascularization

MARINA

Minimally Classic/Occult Trial of the Anti-VEGF Antibody Ranibizumab in the Treatment of Neovascular Age-Related Macular Degeneration

MPS

Macular Photocoagulation Study

NV

neovascular age-related macular degeneration

OCT

optical coherence tomography

PDT

photodynamic therapy

PRN

pro re nata

RPE

retinal pigment epithelium

SD

standard deviation

SE

standard error

SEVEN-UP

Seven-Year Observational Update of Macular Degeneration Patients Post-MARINA/ANCHOR and HORIZON Trials

TAP

Treatment of Age-related macular degeneration with Photodynamic therapy study

VEGF

vascular endothelial growth factor

VIP

Verteporfin in Photodynamic Therapy study