Three-Year Outcomes of Neovascular Age-Related Macular Degeneration in Eyes That Do Not Develop Macular Atrophy or Subretinal Fibrosis

Purpose To report the 36-month treatment outcomes of eyes with neovascular age-related macular degeneration (nAMD) receiving vascular endothelial growth factor (VEGF) inhibitors in daily practice who did not develop either subretinal fibrosis (SRFi) or macular atrophy (MA). Methods This is a retrospective analysis of data from the Fight Retinal Blindness registry. Treatment-naïve eyes starting intravitreal injection of VEGF inhibitors for nAMD from January 1, 2010, to September 1, 2017, and did not have SRFI and MA at baseline were tracked. Results We identified 2478 eligible eyes, of which 1712 eyes did not develop SRFi or MA, 291 developed extrafoveal SRFI or MA, and 475 developed subfoveal SRFi or MA over 36 months. The estimated visual acuity stabilized from 6 months to 36 months in eyes that did not develop SRFI or MA with a mean (95% confidence interval [CI]) change in VA of −1 (−2, 0) letters, whereas eyes that developed extrafoveal (−3 [−5, −2] letters) or subfoveal (−10 [−11, −8] letters) SRFi or MA declined in vision in the same period. Eyes with no or extrafoveal SRFi or MA over 36 months were more likely to maintain their visual improvement from six months to 36 months (odds ratio [OR; 95% CI] = 2.3 [1.5, 3.3] for absence vs. subfoveal SRFi or MA, P ≤ 0.01 and OR = 2.0 [1.2, 3.4] for extrafoveal vs. subfoveal MA or SRFi, P = 0.01). Conclusions Treatment-naïve nAMD eyes receiving VEGF inhibitors maintain their initial six-month visual improvement over three years if they do not develop SRFI or MA. Translational Relevance The nAMD is still a major cause of blindness despite antiangiogenic treatments. We found that eyes that did not develop subretinal fibrosis or macular atrophy maintained their initial vision improvement for at least three years, suggesting that identifying treatments for these complications is the final barrier to achieving excellent outcomes in nAMD.


Introduction
The end-stage sequelae of neovascular age-related macular degeneration (nAMD), macular atrophy (MA), and subretinal fibrosis (SRFi) increase with time, are untreatable, and are associated with poor visual outcomes. [1][2][3][4][5] Both clinical trials and observational studies tend to find that visual acuity (VA) improves from baseline for six months after starting treatment with vascular endothelial growth factor (VEGF) inhibitors and then progressively declines thereafter in association with the development of foveal MA or SRFi, with final vision depending mainly on the presenting VA at the start of the treatment and the number of injections. [6][7][8] Older age, presenting VA, and type of choroidal neovascularization (CNV) may predict risk of progression to MA and SRFi under treatment more strongly than treatment strategy and frequency. 1,9,10 Few studies, if any, have investigated whether there is any other mechanism that causes visual loss in eyes with nAMD independently of these features. Here, we tested the hypothesis that eyes with nAMD treated with VEGF inhibitors continue to lose vision through unknown mechanisms, even if they do not develop SRFi or MA.

Design and Setting
This was a retrospective analysis of treatment-naïve eyes that had received intravitreal VEGF inhibitors for nAMD in routine clinical practice tracked in the prospectively designed observational database-The Fight Retinal Blindness! (FRB!) registry. The details of the FRB! database have been previously published. 11 Analyzed data are 100% completed because all fields must be filled out with in-range values before being accepted by the database. Participants in this analysis included patients from Australia, France, Ireland, Italy, the Netherlands, New Zealand, Singapore, Spain, and Switzerland. Institutional approval was obtained from the University of Sydney, the Royal Australian and New Zealand College of Ophthalmologists, the French Institutional Review Board (IRB) (Société Française d'Ophtalmologie IRB), the Mater Private Hospital IRB in Dublin, Ireland, the Fondazione IRCCS Cà Granda Ospedale Maggiore Policlinico, Milan, Comité de Ética de la Investigación con medicamentos de Euskadi (CEIm-E), and Agencia Española de Medicamentos y Productos Sanitarios (AEMPS), Singhealth Singapore, and the Cantonal Ethics Committee Zurich.
All patients gave their informed consent. Informed consent ("opt-in consent") was sought from patients in France, Ireland, Italy, the Netherlands, Singapore, Spain, and Switzerland. Ethics committees in Australia and New Zealand approved the use of "opt-out" patient consent. This study adhered to the Declaration of Helsinki's tenets and followed the Strengthening the Reporting of Observational Studies in Epidemiology statements for reporting observational studies. 12

Data Sources and Measurements
We analyzed data from the nAMD module of the FRB! outcomes registry. Data were obtained from each clinical visit, including the VA, the activity of the underlying choroidal neovascularization (CNV) lesion, the presence of SRFi or MA, treatment given, procedures, and ocular adverse events. Distance VA (uncorrected, corrected and pinhole if required) was measured in Snellen chart and converted as the number of letters read on a logarithm of the minimum angle of resolution (logMAR) VA standard ETDRS chart. 13 The activity of the CNV lesion was graded by the treating physician based on findings from clinical examination according to a definition provided in the data collection screen from optical coherence tomography (OCT) and dye-based fundus fluorescein angiography, alone or in combination, at each visit. Physician grading of MA and SRFi was implemented in April 2016 into FRB! to comply with the International Consortium for Health Outcomes Measurements (ICHOM) macular degeneration standard set and was recorded prospectively at each visit from then: these data were retrospectively entered for eyes with data entered before this date (n= 245 eyes). 14 No distinction was made between nonfibrotic scar and fibrotic scar in the grading, 15 because the diagnosis was based on the concordance between the appearance of SRFi on clinical examination, color fundus photography, and SD-OCT. At each visit, documentation of MA and SRFi was recorded according to the ICHOM standard set as: "Not present" or if present, based on location: "Extrafoveal" or "Subfoveal." 14 Repeat treatments were at the physician's discretion in consultation with the patient, thereby reflecting routine clinical practice.

Patient Selection and Definitions
Treatment-naïve eyes starting intravitreal injection of VEGF inhibitors of either aflibercept (2 mg Eylea; Bayer Healthcare, Leverkusen, Germany), bevacizumab (1.25 mg Avastin; Genetech Inc/Roche, Basel, Switzerland), or ranibizumab (0.5 mg Lucentis; Genetech Inc/Novartis) for nAMD from January 1, 2010, to September 1, 2017, thereby allowing the possibility of having at least 36 months of observations after the initial treatment, and who did not have SRFI or MA at baseline were tracked in the registry. Eyes were excluded if the grading of SRFI or MA was not entered at baseline.
To ensure that eligible eyes did not have SRFI or MA at presentation, the baseline grading of SRFi or MA was based on multimodal imaging at each visit from the start of the treatment to the three-month visit to detect eyes with undiagnosed SRFi or MA at the beginning of the treatment because of intense exudative signs and exclude other reasons of subretinal hyperreflective material such as fibrin or hemorrhage.
Three groups were defined based on the physician grading of SRFi or MA over 36 months of treatment: absence (i.e., eyes that did not develop SRFi or MA graded over 36 months), subfoveal SRFI or MA (i.e., eyes that developed subfoveal SRFi or MA graded over 36 months) and extrafoveal SRFI or MA (i.e., eyes that developed extrafoveal SRFi or MA graded over 36 months). Eye that developed first extrafoveal SRFi or MA and then progressed to subfoveal SRFi or MA over the period were included in the subfoveal SRFI or MA group for the analysis. Eyes that completed at least 1035 days of follow-up were defined as "completers." Eyes that did not complete 36 months of observations were defined as "non-completers." To investigate if vision declined after the initial visual improvement, we analyzed the mean change in VA from six months to 36 months. The initial visual improvement was considered as maintained if the VA change from six months to 36 months was more than −5 letters.

Outcomes
The main outcome was the estimated mean change in VA from 6 months at 36 months. Secondary visual outcomes included the estimated mean change in VA from baseline to 36 months, the proportion of eyes who maintained vision (VA change > −5 letters), lost ≥10, and ≥15 letters of vision from six months at 36 months, and the mean final VA. Other outcomes of interest were the median time to the development of MA and SRFi over 36 months, the baseline predictors of the development of MA, and SRFi over 36 months, the proportion of visits over 36 months in which the CNV lesion was graded as active, the median time to first grading of CNV inactivity over 36 months, the median time interval between injections over 36 months, the median number of visits and injections administered over 36 months, the rate of noncompletion, the median time and mean VA change to dropout, and the reason for discontinuation over 36 months.

Statistical Analysis
Descriptive data were summarized using the mean (standard deviation), median (first and third quartiles), and number (percentages) where appropriate. Calculation of crude visual outcomes over 36 months used the last-observation-carried-forward for non-completers.
We compared visual outcomes between SRFi or MA groups over three years using mixed-effects longitudinal generalized additive models with the interaction between the development and location of MA or SRFi during the treatment and time as the main predictor variable. Longitudinal models included all visits from completers and non-completers (all observations until the 36-month visit or dropout). The proportions of eyes who maintained vision, lost 10 letters and 15 letters of vision from six months at 36 months between groups were compared using logistic mixed-effects regression models. Longitudinal and logistic models included age, gender, VA, type of CNV lesion at baseline and lens status during the followup, and nesting of outcomes within practitioners and patients with bilateral disease as random effects. Generalized Poisson regression models were used to compare the number of injections and visits between groups over 36 months. Longitudinal, logistic and generalized Poisson models included age, gender, VA, type of CNV lesion at baseline and lens status during the follow-up, and nesting of outcomes within practitioners and patients with bilateral disease as random effects. Cox-proportional hazards models and Kaplan-Meier survival curves were used to assess and visualize the time to first grading of inactivity, first grading of SRFi and MA, and non-completion rates over 36 months.

Choroidal Neovascularization Activity Outcomes Over Three Years
Overall, the proportion of active visits in eyes completing 36 months was 55%, lower in eyes that developed extrafoveal SRFi or MA (43.3%) than eyes that did not develop (56%) or developed subfoveal SRFI or MA (56%) (P = 0.044). The median (Q1, Q3) time to first grading of inactivity was 119 (82, 385) days and was not significantly different between subgroups (Fig. 3).

Injections and Visits Over Three Years
The median (Q1, Q3) number of injections was 19 (14,25) over three years in completers with 8 (6, 10), 6 (3, 8) and 5 (3,8) median (Q1, Q3) injections yearly at first, second, and third year, respectively ( Table 2). The adjusted ratio (95% CI) of the number of injections and visits from the generalized Poisson regression model was similar between subgroups according to the development of SRFi or MA in eyes completing 36 months (

Outcomes of Eyes not Completing Three Years
The overall non-completion rate over 36 months was 45.4% (1125 eyes) and was more frequent in eyes that did not develop SRFi or MA than eyes that developed extrafoveal or subfoveal SRFi or MA (50% absence vs. 30% extrafoveal vs. 38% subfoveal, P < 0.01; see Supplementary Fig. S2). The mean VA at drop out was significantly better than the mean baseline VA each year of drop out ( Supplementary Fig. S3). The reasons for patients discontinuing treatment were tracked in 141 (14%) eyes. These were mainly not related to a poor outcome (71%, 100 eyes): treatment considered as successful 39% (54 eyes), patient transferred to another doctor 14% (20 eyes), death 14% (20 eyes), and medical contraindication 4% (six eyes).

Discussion
The present study reports that treatment-naïve nAMD eyes receiving VEGF inhibitors maintain their initial six-month visual improvement over three years of treatment in routine clinical practice if they do not develop SRFI or MA. This is not necessarily surprising, because SRFi and MA are well-known associations of poor long-term visual outcomes in treated nAMD eyes. 2,3,5,7,[15][16][17][18] The significance of our findings is that there is probably no other disease process that causes loss of the initial gains seen in eyes treated for nAMD. Although it would be helpful for this finding to be consolidated and extended in future studies, it appears that the prevention of MA and SRFi is the final obstacle to achieving better, enduring outcomes in nAMD.
Not surprisingly, eyes in the three groups were not comparable at baseline particularly regarding presenting VA. This may be possibly due to an increased amount of blood or fibrin at baseline in the incident SRFI or MA group. We tried to limit the inclusion of baseline SRFi or MA eyes in the study using multimodal imaging definition to differentiate these features with other causes of subretinal hyperreflective material and defined the status of baseline SRFi or MA on the first three months treatment visits.
Our results emphasize that the development or extension of MA or SRFi in the subfoveal region is associated with poor long-term visual outcomes. [15][16][17][18][19] Eyes that developed subfoveal SRFi or MA over 36 months had at least two to three lines difference in the final estimated mean VA change from baseline, were at least half as likely to maintain six-month visual improvement at 36 months and twice as likely to have a two-line or three-line VA loss from six months at 36 months than eyes that developed extrafoveal SRFi or MA over 36 months of treatment.
Approximately 20% and 25% of eligible eyes developed SRFI and MA respectively by 36 months from the start of the treatment. These cumulative rates of SRFi and MA were similar to those reported elsewhere. 9,16,17 The Macular atrophy in Pro re Nata versus Treat-and-Extend (MANEX) study reported that the incidence of new atrophy lesion in consecutive naïve treated nAMD eyes receiving VEGF inhibitors was approximately 19% and 22% at 2 and 3 years of treatment, respectively. 9 In the comparison of AMD Treatments Trials (CATT), non-geographic atrophy and scar rates were estimated to be approximately 12 to 19% and 16 to 20% at 2 years, respectively, depending on the type of drug and treatment pattern. 17,18 The FRB registry has implemented the ICHOM classification grading of SRFi and MA to standardize the diagnosis of these features and compare outcomes between different reports. Our results are derived from an extensive observational database with multiple practitioners grading the clinical features, which may be less precise than in reports from RCTs such as CATT. [16][17][18] However, most of the practitioners contributing data are retina specialists who have agreed to use the ICHOM multimodal grading definition of these features. These real-world findings also reflect how diagnosis and treatment decision would be made in daily clinical practice if an effective drug were developed for preventing or treating MA or SRFi.
The three-year visual real-world outcomes of VEGF inhibitors for nAMD were reasonably good (mean +3 letters improvement from baseline) with a median number of injections yearly of eight, six, and five during the first, second, and third year of treatment. Previous retrospective observational studies have reported poorer outcomes at three years. [20][21][22][23] It is difficult to compare our study to earlier reports because we included only eyes that had been diagnosed early without SRFi or MA when they started treatment.
There was no difference in treatment and visits frequency between the subgroups. However, we found that eyes with no SRFi or MA, which achieved the best visual outcomes, were more aggressively treated and monitored during the third year of treatment. This reinforces the idea that initial VA improvement in nAMD can be maintained with more intensive or proactive treatment approaches in clinical practice. 8 As previously reported in the literature, presenting VA was a significant predictive factor of MA and SRFi development in our study. 1-3 Type 3 CNV was associated with an increased risk of MA and type 2 CNV with an increased risk of SRFi, which has been confirmed in previous reports. 1,9,10,19 Loss to follow-up may introduce bias because eyes that discontinue may drop out because of poor outcomes or sometimes because of good response to treatment and stabilization of vision. The rate of noncompletion was, in fact, highest in eyes that did not develop SRFi or MA over 36 months of treatment. The mean VA of the eyes that dropped out tended to be better than the presenting VA when treatment had discontinued, suggesting that those eyes tended to have good visual outcomes. Most (70%) of tracked reasons for discontinuation were mainly not related to a poor outcome. Our estimated outcomes, particularly in eyes that did not develop SRFi or MA over the study, may be inferior to the actual results if patients with good vision tended to discontinue follow-up within 36 months.
We acknowledge several limitations that are mostly inherent in retrospective observational studies. Injection decisions in routine clinical practice are made without a guided management protocol, so they may vary among retinal specialists compared to RCTs. The grading of SRFi, MA, lesion type, and lesion activity may have interphysician variability. The FRB! registry receives data from a wide variety of international practices and practitioners. Thus we believe our data are fairly representative of clinicians worldwide, which may reduce potential bias caused by this variability to some extent. We also included nesting of outcomes within practitioners in our models to help account for these effects. This analysis's main strengths are its origi-nality and the large number of eyes that were studied over a long time period in daily clinical practice.
To conclude, our study suggests that SRFi and MA are the main retinal causes of the long-term (three-year) visual decline in vision in nAMD eyes in routine clinical practice. Early diagnosis and appropriate application of treatment regimens to prevent these features and their extension to the subfoveal region stabilize the visual improvement after the start of the treatment. Further research is warranted to determine whether these findings hold for longer periods. There is a need to develop new drugs with potential antifibrotic and neuroprotective effects, combined with VEGF inhibitors, to prevent or even treat these endstage features of nAMD and further improve visual outcomes and the quality of life of our patients.