Treat-and-Extend Therapy Using Aflibercept for Neovascular Age-related Macular Degeneration: A Prospective Clinical Trial

FRANCIS CHAR DECROOS, DAVID REED, MURTAZA K. ADAM, DAVID SALZ, OMESH P. GUPTA, ALLEN C. HO, AND CARL D. REGILLO

PURPOSE: To determine the efficacy and durability of HOROIDAL NEOVASCULARIZATION (CNV) CAUSES aflibercept used in a treat-and-extend (TAE) regimen the vast majority of severe visual loss from for neovascular age-related macular degeneration neovascular age-related macular degeneration C 1 (NVAMD). (NVAMD). Intravitreal vascular endothelial growth fac- DESIGN: Multicenter, prospective, open label, noncom- tor (VEGF) blockade has revolutionized the treatment of parative, interventional study. NVAMD. Ranibizumab (Lucentis; Genentech, Inc) and METHODS: Forty eyes of 40 patients with treatment- bevacizumab (Avastin; Genentech, Inc) are anti-VEGF naı¨ve NVAMD were managed with a TAE regimen of agents that have been extensively studied.2–7 intravitreal aflibercept. The main endpoints were the Ranibizumab is currently approved by the U.S. Food and change in mean and median best-corrected visual acuity Drug Administration for treatment of NVAMD, while from baseline at years 1 and 2. Other endpoints included bevacizumab is used in an off-label fashion. mean number of annual injections and treatment More recently, the fully human, soluble recombinant intervals. VEGF receptor decoy aflibercept (Eylea; Regeneron Phar- RESULTS: Thirty-five (87.5%) and 31 patients maceuticals, Inc) was approved for treatment of NVAMD. (77.5%) completed year 1 and year 2, respectively. Aflibercept is a recombinant fusion protein consisting of The mean letter gain was 7.2 (P < .001) and 2.4 the extracellular domain of the human VEGF receptors 1 (P [ .269) letters at 1 and 2 years, respectively, (Ig2 domain) and 2 (Ig3 domain) fused to the Fc domain from a mean baseline of 58.9 letters (20/63 Snellen of human IgG1. The clinical efficacy of aflibercept for treat- equivalent). The median visual gain was 11.5 and 7.5 ment of CNV was established in trials evaluating afliber- letters at 1 and 2 years, respectively, from a median cept, namely the VEGF-Trap Eye: Investigation of baseline of 59.0 letters (20/63 Snellen equivalent). Efficacy and Safety in Wet AMD parts 1 and 2 (VIEW The mean number of injections was 8.0 and 6.5 during 1/2) studies, with reported results demonstrating that afli- the first and second year, respectively. Twelve-week or bercept administration was noninferior to ranibizumab longer treatment intervals were used in 35% and 38% dosed every 4 weeks. The VIEW 1/2 studies included a of patients during the first- and second-year time points, treatment arm that initially evaluated aflibercept injec- respectively. tions once every 8 weeks after 3 successive loading doses 8 CONCLUSION: Intravitreal aflibercept TAE therapy led given every 4 weeks. to significant visual improvement in eyes with NVAMD Although the pivotal trials demonstrated that serial at 1 year, with some loss in the visual gains at the end anti-VEGF treatments were effective for NVAMD, of year 2 that was not related to loss of exudative control. monthly follow-up was required except in the aforemen- TAE therapy with aflibercept is a rational strategy to tioned treatment arm within the VIEW 1/2 studies. reduce treatments and clinic evaluations over 2 years Mandated monthly follow-up can be cumbersome for pa- with satisfactory outcomes. (Am J Ophthalmol tients, caregivers, and treating physicians alike, and the 2017;180:142–150. Ó 2017 Elsevier Inc. All rights challenges presented by maintaining monthly follow-up reserved.) intervals generated interest in alternative intravitreal in- jection dosing strategies. One such strategy is the treat- and-extend (TAE) approach, initially described in 2007 by Spaide.9 TAE is a dosing paradigm in which the affected Supplemental Material available at AJO.com. eye is injected at each patient visit with a variable follow- Accepted for publication Jun 7, 2017. up interval that is tailored to individual treatment From the Retina Service, Wills Eye Hospital/Mid Atlantic Retina, 10 Philadelphia, Pennslyvania (F.C.D., D.R., M.K.A., D.S., O.P.G., response. The TAE regimen uses optical coherence to- A.C.H., C.D.R.); and Southeastern Retina Associates, Chattanooga, mography (OCT) based on detection of macular fluid to Tennessee (F.C.D.). direct management. Several retrospective11–16 and Inquiries to Carl D. Regillo, Director, Retina Service, Wills Eye 17–21 Hospital/Mid Atlantic Retina, 840 Walnut St, Suite 1020, Philadelphia, prospective series demonstrated that TAE using PA 19107; e-mail: [email protected] bevacizumab and/or ranibizumab for NVAMD reduced

142 © 2017 ELSEVIER INC.ALL RIGHTS RESERVED. 0002-9394/$36.00 http://dx.doi.org/10.1016/j.ajo.2017.06.002 the number of both follow-up visits and treatments intervals owing to the follow-up interval variability neces- compared with monthly dosing, while still achieving sitated by a treat-and-extend management paradigm. Best- satisfactory anatomic and functional outcomes.11,14 TAE corrected ETDRS visual acuity and CFT measured by employing aflibercept has not been prospectively spectral-domain OCT were reported at 7 prespecified characterized in a clinical trial, though retrospective time points: baseline, 1 month, 3 months, 6 months, series have examined aflibercept TAE in eyes previously 12 months, 18 months, and 24 months. treated with other anti-VEGF agents22,23 or in 15 combination therapy with other anti-VEGF agents. TREATMENT ADMINISTRATION: Study eyes fulfilling The TAE dosing regimen may be particularly suitable in inclusion criteria received an intravitreal 2 mg aflibercept conjunction with intravitreal administration of aflibercept, treatment at the initial visit and at each subsequent study as pharmacokinetic modeling of this treatment demon- visit. No specified loading doses were mandated. If all strated prolonged VEGF binding activity compared with extension criteria were not met 4 weeks after the baseline bevacizumab or ranibizumab.24 Prolonged binding may visit, the patient underwent repeat evaluation and treat- potentially correlate to fewer injections required for long- ment after 4 weeks. Until all extension criteria were term control of exudation in NVAMD. Before performing met, patients were subsequently followed and treated at a comparative study against bevacizumab or ranibizumab 4-week intervals. Extension criteria included absence of TAE, it is important to first prospectively investigate the macular fluid on OCT, defined as any intraretinal, subreti- efficacy and durability of aflibercept TAE monotherapy. nal, or sub–retinal pigment epithelial fluid; absence of The Aflibercept Treat and extend for Less frequent Admin- vision loss of 5 or more letters from the prior visit; absence istration Study (ATLAS) is the first multicenter, prospec- of new macular hemorrhage; and absence of increased tive study to help characterize the efficacy and durability lesion size or leakage on FA. Of note, FA was not of the TAE dosing regimen with intravitreal aflibercept mandated at any visit after the initial evaluation, but for NVAMD. could be performed at any visit at the discretion of the investigator. When a patient met extension criteria, the previous follow-up/treatment interval was increased by 2-week in- METHODS tervals to a maximum interval of 16 weeks. If a patient was being treated at an interval between 6 and 16 weeks, STUDY DESIGN: This prospective, open-label, multi- but failed to meet extension criteria at a particular visit, center, investigator-sponsored clinical study enrolled 40 treatment was administered and the follow-up interval patients. Prospective Institutional Review Board (IRB) was then reduced by 2-week increments until all extension approval was obtained (#20122054), all experimental pro- criteria were met once again. After NVAMD disease was cedures adhered to the tenets of the Declaration of deemed to be inactive based on fulfillment of all extension Helsinki, and all participants engaged in an informed con- criteria, the follow-up interval was then increased to a sent process and signed a written consent document prior maximum final interval of 2 weeks less than the subject’s to enrollment in the ATLAS (ClinicalTrials.gov Identifier: prior maximum treatment interval where disease reactiva- NCT01773954). The criteria for study inclusion included tion was noted. This limit was specified to minimize the patients of age greater than 55 years with treatment-naı¨ve, number of exudative recurrences. active, subfoveal NVAMD demonstrating macular fluid on spectral-domain OCT and CNV leakage on fluorescein OUTCOME MEASURES: The major endpoints of this angiography (FA). Only 1 eye for each patient demon- 2-year study were mean and median change from baseline strating a pretreatment best-corrected visual acuity in BCVA at years 1 and 2. Secondary outcomes included (BCVA) of 20/25 to 20/320 was eligible. Exclusion criteria change from baseline in CFT, the percentage of patients for the study included prior treatment for NVAMD, any with 20/40 or better Snellen equivalent vision, percentage concurrent progressive retinal disease, or prior vitrectomy of patients gaining or losing 15 or more letters of vision, in the study eye. Patients previously treated with systemic mean treatment interval, mean number of injections and anti-VEGF therapy, photodynamic therapy, or thermal distribution of treatment intervals at the first- and laser were also excluded. At each visit, patients underwent second-year visit, and ocular and systemic adverse events examination including best-corrected Early Treatment over 2 years. Additional outcomes included mean number Diabetic Retinopathy Study (ETDRS) visual acuity, slit- of treatments until first extension, etiologies of vision loss lamp biomicroscopy with a dilated fundus examination, in patients losing 15 or more letters, and cause of extension and spectral-domain OCT (Heidelberg Engineering, Hei- criteria not being met after a steady-state treatment inter- delberg, Germany). FA and fundus photography were val was achieved. A steady-state interval was defined as 2 or performed at baseline to confirm the presence of active more equal successive treatment intervals greater than or CNV. Longitudinal data for both BCVA and central foveal equal to 6 weeks in duration. Another outcome of interest thickness (CFT) were combined into 26 sequential 4-week was mean and median BCVA at 2 years in patients treated

VOL. 180 AFLIBERCEPT TREAT-AND-EXTEND THERAPY FOR NEOVASCULAR AMD 143 better Snellen equivalent visual acuity, 52.5% and 37.5% TABLE 1. Aflibercept Treat-and-Extend Study Population demonstrated 20/40 or better acuity at the 1- and 2-year Baseline Characteristics visit, respectively. Similarly, for the 35 patients who completed the year 1 Mean age (standard deviation) 81.3 (7.4) years visit and the 31 patients who completed the year 2 visit Sex, n (%) (per-protocol analysis), mean best-corrected ETDRS visual Male 17 (42.5%) acuity improved 9.6 (P < .001) and 1.8 letters (P ¼ .505, Female 23 (57.5%) Anti-VEGF treatment in other eye, 0 (0%) Figure 1, Left) at 1 and 2 years, respectively. For these n (percentage) same patients, the median best-corrected ETDRS visual acuity improved 15 and 8 letters at 1 and 2 years, respec- VEGF ¼ vascular endothelial growth factor. tively (Figure 1, Right). Of the 9 patients who lost 15 or more letters of vision by the 2-year visit, 1 patient developed culture-positive at 8-week or longer intervals and in patients treated at endophthalmitis (alpha-hemolytic streptococci) that 12-week or longer intervals. resulted in no light perception (NLP) vision. Of the other 8 patients who lost 15 or more letters of vision by the 2- STATISTICAL ANALYSIS: An intent-to-treat analysis was year visit, the etiology of vision loss was subretinal fibrosis performed using the last-observation-carried-forward in 1 patient, subfoveal geographic atrophy in 4 patients, (LOCF) method. A per-protocol analysis was performed and a 1-time vision decrease not associated with fluid, hem- for all patients present at a particular time point. Change orrhage, fibrosis, or atrophy in 2 patients. These 2 patients from baseline over time for both BCVA and CFT were with a 1-time vision decrease at month 24 demonstrated evaluated using a mixed model with an autoregressive 20/32 and 20/30 visual acuity, respectively, at the visit subse- (AR1) correlation structure to account for repeated mea- quent to conclusion of the trial. These visual gains were sures within subjects. Time was modeled as a class variable approximately 12 and 22 ETDRS letters from the 2-year with 26 levels. Analyses were performed using SAS 9.4 time point, respectively. Two of the 4 patients with 15 or (SAS Institute Inc, Cary, North Carolina, USA). more letters loss associated with subfoveal geographic atro- phy described above had preexisting, nonfoveal geographic atrophy at baseline. Finally, in 1 patient the etiology of vision loss could not be identified, with no hemorrhage, exudation, RESULTS fibrosis, atrophy, or worsening cataract noted upon review.

PATIENTS, FOLLOW-UP, AND TREATMENTS: Forty pa- ANATOMIC OUTCOMES: Intent-to-treat analysis demon- tients were enrolled across 4 clinical sites in 3 states of 2 strated that the mean CFT decreased by 143 and 139 mm, practices (Mid Atlantic Retina and Southeastern Retina respectively, at 1 and 2 years (all P < .001, Figure 2), from Associates). Of these patients, 35 (87.5%) completed the 343 mm at baseline (Table 2). Similarly, a per-protocol year 1 visit and 31 (77.5%) completed the year 2 visit. analysis demonstrated that the mean CFT decreased by The mean patient age at baseline was 81.3 years; 17 pa- 131 and 130 mm, respectively, at 1 and 2 years (all tients (42.5%) were male and 23 (57.5%) were female P < .001, Figure 2), from 343 mm at baseline. (Table 1). No patients experienced a myocardial infarc- tion, cerebrovascular accident (CVA), and/or transient TREATMENT INTERVALS: The mean number of injec- ischemic attack prior to enrollment. tions during the first year and the second year was 8.0 (standard deviation ¼ 1.6) and 6.5 injections (standard VISUAL ACUITY OUTCOMES: An intent-to-treat analysis deviation ¼ 2.6), respectively. The mean treatment inter- (LOCF) demonstrated that the mean ETDRS visual acuity val during the first year and the second year was 6.7 weeks improved 7.2 (P < .001) and 2.4 letters (P ¼ .269, Figure 1, (standard deviation ¼ 1.3) and 9.0 weeks (standard Left) at 1 and 2 years, respectively, from 58.9 (Snellen deviation ¼ 3.2), respectively. Eight-week or longer treat- equivalent ¼ 20/63) letters at baseline (Table 2). An ment intervals were being used in 71% and 75% of patients intent-to-treat analysis (LOCF) demonstrated that the me- during the end of year 1 and end of year 2 visit, respectively. dian ETDRS visual acuity improved 11.5 and 7.5 letters Twelve-week or longer treatment intervals were being used (Figure 1, Right) at 1 and 2 years, respectively, from 59.0 in 35% and 38% of patients during the end of year 1 and (Snellen equivalent ¼ 20/63) letters at baseline end of year 2 visit, respectively (Figures 3 and 4). Of the (Table 2). Eleven patients (27.5%) and 9 patients 40 enrolled patients, 7 (17.5%) and 9 (22.5%) had under- (22.5%) gained >_15 letters at the 1- and 2-year visit, gone at least 1 aflibercept treatment for NVAMD in the respectively, while 1 (2.5%) and 8 (22.5%) patients lost fellow eye by the first- and second-year visit, respectively. >_15 letters at the 1- and 2-year visit, respectively. From a A steady-state interval was defined as 2 or more equal suc- baseline of 17.5% of patients that demonstrated 20/40 or cessive treatment intervals greater than or equal to 6 weeks

144 AMERICAN JOURNAL OF OPHTHALMOLOGY AUGUST 2017 FIGURE 1. (Left) Change in mean best-corrected Early Treatment Diabetic Retinopathy Study (ETDRS) score from baseline. (Right) Change in median best-corrected ETDRS score from baseline. BCVA [ best corrected visual acuity.

TABLE 2. Aflibercept Treat-and-Extend Treatment Outcomes at 1 and 2 Years

Standard Visit Month Mean Deviation Median Minimum Maximum

ETDRS BCVA (letters) Baseline 58.9 11.4 59.0 35 79 1 63.5 13.1 66.0 23 80 3 65.6 14.3 69.0 24 85 6 64.1 14.5 66.0 26 86 12 66.0 16.0 70.5 17 89 18 61.3 18.7 66.0 10 88 24 61.1 18.9 66.5 12 85 CFT (mm) Baseline 343 140 322 92 746 1 218 79 210 82 437 FIGURE 2. Change in mean central foveal thickness (CFT) 3 199 68 195 67 363 from baseline. 6 206 91 194 58 513 12 209 82 198 56 533 18 219 90 201 68 535 ADDITIONAL OUTCOMES: 24 211 81 200 56 414 Of the 36 patients who main- tained follow-up long enough to meet extension criteria, BCVA ¼ best-corrected visual acuity; CFT ¼ central foveal the mean number of treatments until first extension was thickness; ETDRS ¼ Early Treatment Diabetic Retinopathy Study. 2.5 (standard deviation ¼ 1.8), which excluded 1 patient who never met extension criteria for the 2-year duration of the study. in duration. A steady-state treatment interval was achieved At 2 years, patients being treated at 8-week or longer in- for the 2-year duration of the study in 30 of 31 patients tervals (n ¼ 23) demonstrated a mean and median change (96.8%). Of the remaining 10 patients, 9 patients were in BCVA from baseline of 2.4 and 4.0 letters, respectively. lost to follow-up and 1 patient never met extension criteria At the same time interval, patients being treated at 12- for the 2-year duration of the study. Of these 30 patients, 15 week or longer intervals (n ¼ 12) demonstrated a mean (50%) did not demonstrate any subsequent recurrent macu- and median change in BCVA from baseline of 5.3 and 15 lar fluid, new macular subretinal hemorrhage, loss of 5 or letters, respectively. more letters of vision from a prior visit, or increased leakage on fluorescein angiography for the 2-year duration of the ADVERSE EVENTS: Two patients died during the study, study. For the remaining 15 patients, the reason for reduc- and 1 of these patients died owing to a fatal myocardial tion in treatment interval after a steady-state treatment in- infarction. Records were never released detailing the death terval was achieved is detailed in Table 3. of the second patient despite multiple record requests. All

VOL. 180 AFLIBERCEPT TREAT-AND-EXTEND THERAPY FOR NEOVASCULAR AMD 145 TABLE 3. Reason for Required Reduction in Treatment Interval After a Steady-State Treatment Intervala Achieved

N (%)

Total patientsb achieving steady-state treatment 30 (100%) interval for 2-year duration of trial Steady state maintained for 2-year duration 15 (50%) of trial One recurrence of macular fluid 3 (10%) Two recurrences of macular fluid 3 (10%) Three recurrences of macular fluid 1 (3%) One episode of >_5-letter loss of vision 5 (17%) (with no associated recurrence of fluid or hemorrhage) Two episodes of >_5-letter loss of vision 2 (6%) (with no associated recurrence of fluid or hemorrhage) One recurrence of macular fluid and 1 (3%) FIGURE 3. Distribution of patients at various treatment inter- concurrent >_5-letter loss of vision vals during the year 1 study visit. New subretinal hemorrhage 0 (0%) Increased leakage on fluorescein angiography 0 (0%)

aDefined as 2 or more equal successive treatment intervals greater than or equal to 6 weeks in duration. bExcludes 9 patients who were lost to follow-up prior to 2-year duration of trial and 1 patient who did not meet extension criteria for entire duration of trial.

detailed prior and 1 nonfatal CVA. Ocular and systemic adverse events are detailed in Table 4.

DISCUSSION

A TAE-STYLE DOSING REGIMEN FOR NVAMD CAN DIMINISH the need for intensive monthly follow-up and treatment as modeled in multiple pivotal clinical trials and still allow for similar visual and anatomic outcomes.18–21 ATLAS is FIGURE 4. Distribution of patients at various treatment inter- one of the first studies to prospectively report the efficacy vals during the year 2 study visit. of a TAE-style dosing regimen with aflibercept for treat- ment-naı¨ve NVAMD. In the present study, a mean gain of 7.2 letters and a median gain of 11.5 letters was demon- strated at 1 year, which was similar to other anti-VEGF serious adverse events were reported to the IRB and study agents in studies requiring either monthly dosing or as- sponsor, including the 2 deaths over the 2-year duration needed dosing with mandated monthly follow-up.4–6 of the trial. The IRB approved the reporting without Additionally, the visual gain at 1 year within ATLAS comment. Serious ocular adverse events included 1 patient was comparable to the 7.9- to 10.9-letter gain observed in with culture-positive endophthalmitis (alpha-hemolytic aflibercept-treated eyes at 52 weeks within the VIEW 1/2 streptococci). This patient developed NLP vision studies.8 The change in BCVA at 1 year within the present 21 months into the trial from this infection, and the subse- study was also similar to multiple other studies investi- quent BCVA results exclude these acuity measurements. gating other anti-VEGF agents used in a TAE fashion for At the study visit prior to developing endophthalmitis, NVAMD.18–20 In the Lucentis Compared to Avastin the patient had BCVA of 71 letters (20/40þ1 Snellen Study (LUCAS), 213 and 218 patients were treated with equivalent), improved from 60 letters at baseline. Serious bevacizumab and ranibizumab, respectively. In the systemic events included the 1 fatal myocardial infarction bevacizumab arm, a 7.9-letter gain was noted, which was

146 AMERICAN JOURNAL OF OPHTHALMOLOGY AUGUST 2017 bevacizumab and ranibizumab, respectively. One reason for TABLE 4. Systemic and Ocular Adverse Events at 2 Years the decline in BCVA from year 1 to year 2 within ATLAS Within the Aflibercept Treat-and-Extend Study Population may be owing to an increase in the percentage of eyes losing 15 or more letters of vision. We noted an increase from Adverse Event N (%) 2.5% to 22.5% of eyes losing 15 or more letters from the Ocular year 1 time point to the year 2 time point. For comparison, Endophthalmitis (culture-positive) 1 (2.5%) at 96 weeks within the VIEW 1/2 program, less than 10% of Endophthalmitis (culture-negative) 0 (0%) eyes lost 15 or more letters within all aflibercept groups.25 Systemic Likewise, within the LUCAS study less than 11% of eyes Death from any cause 2 (5%) lost 15 or more letters at 2 years.17 The reason for the Nonfatal myocardial infarction 0 (0%) higher rate of eyes losing 15 or more letters at 2 years within Nonfatal cerebrovascular event 1 (2.5%) Death from any vascular cause 1 (2.5%) ATLAS compared with the LUCAS and VIEW 1/2 is not Other serious related adverse systemic event 0 (0%) clear, but may be because of the impact of outliers on the smaller sample size of the present study. As detailed below, a handful of eyes losing 15 or more letters had substantial not statistically different compared with the 8.2-letter gain impact on the mean BCVA at 2 years. Supporting this of patients treated with ranibizumab. The mean number of assertion is the substantial divergence between mean and treatments required at 1 year in the ATLAS study was 8.0, median BCVA at the 2-year time point, which is suggestive which is comparable to the 8.9 and 8.0 treatments with that outliers in BCVA impacted the 2-year result. bevacizumab and ranibizumab, respectively, in the LUCAS Within this group of study eyes with severe vision loss at study at 1 year.20 2 years, 1 eye developed culture-positive endophthalmitis When compared with LUCAS at 1 and 2 years,17,20 we that led to NLP vision. Subgroup analyses of the 8 other observed that the treatment intervals in ATLAS were eyes that lost 15 or more letters at 2 years within ATLAS generally of longer duration. At the 1-year visit within revealed that the etiology of vision loss was a 1-time LUCAS, 46.7% of eyes treated with bevacizumab and decrease in vision that resolved at the subsequent clinic 32.9% of eyes treated with ranibizumab required treatment visit in 2 patients. Four eyes that lost more than 15 letters at 4-week intervals, compared with 6% of eyes requiring did so owing to fovea-involving geographic atrophy. This 4 week treatment in the present study. Similarly, 6% of geographic atrophy was new in 2 patients and progressed the ATLAS patients required a 4-week treatment interval from preexisting atrophy in 2 patients. Geographic atro- at 2 years, compared with 20%–27% in LUCAS. When phy, in particular subfoveal geographic atrophy incidence comparing the distributions of lengthier treatment inter- or progression during serial aflibercept treatment, has not vals, we observed that in ATLAS at 1 year, 8-week or yet been rigorously characterized. For comparison, within longer intervals were used in 71% of patients and 12- the Comparison of Age-Related Macular Degeneration week or longer intervals were used in 35% of patients. At Treatments Trials (CATT), the estimated cumulative inci- year 1 within LUCAS, the distribution of patients being dence of geographic atrophy involving the foveal center treated at these same intervals was 41.3% and 25.1% for was 2.7% at 2 years.26 The difference between the higher bevacizumab-treated eyes and 52.4% and 37.1% for observed rates of incident subfoveal geographic atrophy ranibizumab-treated eyes, respectively. At year 2 within within ATLAS may again be owing to the disparity of sam- ATLAS, the percentage of patients treated at 8-week or ple size between the present study and the CATT rather longer intervals or 12-week or longer intervals was 75% than owing to TAE dosing. and 38%, respectively. At year 2 within LUCAS, the distri- In addition to the eyes that lost 15 or more letters of bution of patients being treated at these same intervals was vision, we assessed the entire study population for loss of 46.5% and 33.1% for ranibizumab-treated eyes and 34.1% exudative control. Though we observed that 26% of pa- and 19.8% for bevacizumab-treated eyes, respectively. tients had at least 1 recurrence of macular fluid over 2 years, These findings suggest that there may be greater durability we noted only 1 recurrence of macular fluid that was asso- of effect with aflibercept TAE compared with bevacizumab ciated with a loss of 5 or more letters of vision. Similarly, we TAE and possibly ranibizumab TAE; however, a larger, did not observe any cases of a recurrent macular hemor- comparative randomized trial would be needed to rigor- rhage once a steady-state interval had been achieved. ously delineate any clinically meaningful differences. Next, we observed in subgroup analyses of all patients The BCVA gains noted in the ATLAS study at the 1- treated at intervals of 8 weeks or longer and 12 weeks or year time point decreased to a mean gain of 2.4 letters longer that the visual outcomes were similar to or better and a median gain of 7.5 letters from baseline at the 2- than the overall population at 2 years. These findings sug- year time point. For comparison, the LUCAS study gest that loss of visual gain at 2 years was not associated revealed a less significant decline in mean visual acuity with longer treatment intervals. Rather, it is more plausible from year 1 into year 2. At 2 years, the LUCAS investiga- that eyes that could not achieve longer treatment intervals tors reported a mean gain of 7.4 letters and 6.6 letters with likely had more severe disease, necessitating more frequent

VOL. 180 AFLIBERCEPT TREAT-AND-EXTEND THERAPY FOR NEOVASCULAR AMD 147 treatments, and these eyes possibly had more limited po- visits. Additionally, in ATLAS at 2 years we noted that tential for visual gain at baseline. 25% of patients were being treated at 14-week or longer in- It is reassuring to note that the longer treatment inter- tervals. Capped PRN mandating treatment at 12 weeks vals were not associated with decreased BCVA outcomes. may thus lead to more frequent treatment than is necessary This is important, as no current consensus exists as to the in a subgroup of patients. maximal treatment interval in TAE for NVAMD, though Conversely, for other patients, even 8-week aflibercept many prospective TAE studies employed a 12-week may represent an inadequate treatment regimen. In maximal interval, compared to the 16-week maximal inter- aflibercept-treated eyes with early persistent macular fluid, val in the present study.17,18,21 The ATLAS suggests that patients undergoing 4-week dosing demonstrated improved some patients being treated with TAE with aflibercept functional outcomes compared with patients undergoing may benefit from extension past 12 weeks with eventual 8-week dosing.30 In another study, intraretinal comparable visual outcomes. Similarly, no consensus fluid, subretinal fluid, and retinal pigment epithelial exists as to whether close monitoring can be substituted detachment on OCT were analyzed within the VIEW 1/2 for serial anti-VEGF treatment after a treatment interval studies. Investigators noted that recurrences were observed of 12 weeks or more is achieved. This question has not in all macular fluid compartments synchronized to eventual been rigorously investigated; however, less frequent every-8-week dosing, and importantly these recurrences follow-up and anti-VEGF therapy for eyes with NVAMD correlated with functional visual loss (Waldstein SM, has resulted in gradual vision loss over lengthy follow-up et al. IOVS 2014;55:ARVO E-Abstract 3959, and intervals in 2 independent series.27,28 Conversely, ref. 31). TAE-style dosing provides a potential tool to opti- continuous therapy for NVAMD with close follow-up mize appropriate patient monitoring and treatment, while demonstrated better visual outcomes over a similarly long minimizing both disease recurrence and the clinic visit follow-up interval.29 These studies suggest that meticulous burden that is otherwise unavoidable in real-world clinical follow-up is warranted even in eyes achieving lengthy scenarios. treatment intervals, as NVAMD is a chronic disease. The present study has several limitations. First, it is a Several findings of the ATLAS are helpful for educating noncontrolled, noncomparative, open-label study with a patients regarding the clinical course in an aflibercept relatively small sample size and higher dropout rate at treat-and-extend-style regimen. Patients met extension year 2, which makes mean visual results particularly suscep- criteria after a mean of 2.5 injections within the present tible to outlier visual changes. Additionally, no direct com- study, so many patients can be advised to expect to begin parison can be made between the efficacy of aflibercept extension within 2–4 months of beginning a TAE regimen TAE vs a similar dosing regimen with other anti-VEGF with aflibercept. Next, once a steady-state interval is main- agents such as ranibizumab or bevacizumab. Likewise, it tained, it is important for clinicians to maintain vigilance is important to realize that a direct comparison between for recurrence of macular fluid, which we observed in aflibercept TAE and other aflibercept dosing strategies, approximately a quarter of patients achieving steady state. such as capped PRN dosing, cannot be made within the Fortunately, the present study suggests that recurrence of present study. Next, in line with practice in a real-world macular fluid with concurrent vision loss of 5 or more let- clinical setting, all OCT images were reviewed by treating ters of vision is rare. retinal specialists rather than a reading center. However, it The aflibercept TAE dosing paradigm can be compared is important to note that although reading centers have to the capped pro re nata (PRN) strategy to optimize treat- been shown to demonstrate a slightly higher sensitivity ment for NVAMD used in the VIEW 1/2 studies after week for detecting macular fluid in comparison to site investiga- 52. Until week 52 within VIEW 1/2, the treatment arms tors, the reported differences are minimal32 and retreat- included 4-week aflibercept dosing and 8-week aflibercept ment decisions within trials are typically not dictated by dosing after 3 monthly loading doses. From week 52 until a reading center. Finally, ATLAS was not powered to week 96, patients were followed on a monthly basis for po- determine baseline factors associated with lengthier treat- tential treatment, with mandatory treatment at least every ment intervals. Despite these limitations within ATLAS, 12 weeks. During this capped PRN phase, a mean visual the present study is the first to report the efficacy and dura- gain of 7.6 letters was noted at 96 weeks for both the orig- bility of an aflibercept TAE dosing regimen for treatment- inal 4-week and 8-week dosing arms. Though these visual naı¨ve eyes with NVAMD over a lengthy follow-up interval. gains were 1–2 letters decreased from the 52-week This is particularly salient, as TAE remains a favored endpoint, the visual outcomes were accomplished with a method of management of NVAMD, according to the laudable average of 4.1–4.2 injections between weeks 52 recent American Society of Retinal Specialists 2016 Pref- and 96.8 TAE does have some advantages over this capped erences and Trends survey. In fact, more than 70% of PRN strategy. Capped PRN requires diligent monthly American retina specialists surveyed (n ¼ 1036) report us- follow-up, which may be challenging to implement over ing a TAE-style regimen to manage NVAMD.33 lengthy follow-up durations in everyday clinical practice, The present study demonstrated a relatively higher rate and the TAE paradigm does not require monthly clinic of systemic adverse events over 2 years compared to other

148 AMERICAN JOURNAL OF OPHTHALMOLOGY AUGUST 2017 series, specifically a death rate of 5% (n ¼ 2), a fatal To summarize, aflibercept therapy using a TAE regimen myocardial infarction rate of 2.5% (n ¼ 1), and a nonfatal led to significant visual and anatomic improvement in eyes CVA rate of 2.5% (n ¼ 1). This disparity may be attribut- with treatment-naı¨ve NVAMD at 1 year, although the vi- able to the impact of single cases on the small sample size of sual outcomes diminished to some degree at 2 years. The the ATLAS compared to other series. Of note, within the larger discrepancy in the mean and median visual acuity VIEW 1/2 program at 96 weeks, the death rate was 2.1%– at year 2 can be attributed to several outliers in this small 3.3%, the vascular death rate was 0.8%–1.8%, and the study, half of whom had central atrophy that appeared to CVA rate was 0.5%–0.8%.25 Similarly, an increase in account for the 3-or-more-line loss in BCVA. Overall, vascular adverse events was not observed in larger TAE se- exudative control was well maintained over the 2-year ries with other anti-VEGF agents.17,18 Within the LUCAS, course of the study, based on OCT results, and the rela- which used ranibizumab and bevacizumab in a TAE fashion tively low rate of visually significant fluid recurrences was over 2 years, the death rate was 5.9%–6.8%, the vascular coupled with successful treatment extension to 8 weeks death rate was 0.9%–1.4%, and the nonfatal CVA rate or more in approximately three-quarters of patients. A was 1.4%–1.8%.17 Larger, future prospective TAE series us- TAE treatment strategy with aflibercept in the manage- ing aflibercept will better demonstrate if the vascular ment of NVAMD allows for satisfactory outcomes in adverse event rate with aflibercept used in a TAE fashion most patients over 2 years, with the potential of reduced is more consistent with the rates reported in VIEW 1/2. treatments and office visits.

FUNDING SUPPORT: THIS WORK WAS SUPPORTED BY REGENERON (TARRYTOWN, NEW YORK, USA). FINANCIAL DISCLOSURES: Francis Char DeCroos: Regeneron (Research Support), Genentech (Research Support), Heed Foundation (Fellowship Support), Ronald G. Michels Foun- dation (Fellowship Support); David Reed: Regeneron (Research Support); Murtaza K. Adam: Regeneron (Research Support); David Salz: Regeneron (Research Support); Omesh P. Gupta: Regeneron (Research Support); Allen C. Ho: Regeneron (Research Support/Consultant), Genentech (Research Support/Consultant), Ophthotech (Research Support), Allergan (Research Support/Consultant), Alcon (Research Support/Consultant); Carl D. Regillo: Regeneron (Research Support), Genentech (Research Support/Consultant), Ophthotech (Research Support), Allergan (Research Support/Consultant), Alcon (Research Support/Consultant), Iconic (Research Support/Consultant). All authors attest that they meet the current ICMJE criteria for authorship. The authors thank Sarah Hegarty, MPhil, of Thomas Jefferson University for her assistance with the statistical analysis of the ATLAS trial.

REFERENCES 9. Spaide RF. The as-needed treatment strategy for choroidal neovascularization: a feedback-based treatment system. Am 1. Ferris FL 3rd, Fine SL, Hyman L. Age-related macular degen- J Ophthalmol 2009;148(1):1–3. eration and blindness due to neovascular maculopathy. Arch 10. Freund KB, Korobelnik JF, Devenyi R, et al. Treat-and- Ophthalmol 1984;102(11):1640–1642. extend regimens with anti-VEGF agents in retinal diseases: 2. Brown DM, Kaiser PK, Michels M, et al. Ranibizumab versus a literature review and consensus recommendations. Retina verteporfin for neovascular age-related macular degeneration. 2015;35(8):1489–1506. N Engl J Med 2006;355(14):1432–1444. 11. Gupta OP, Shienbaum G, Patel AH, Fecarotta C, Kaiser RS, 3. Rosenfeld PJ, Brown DM, Heier JS, et al. Ranibizumab for Regillo CD. A treat and extend regimen using ranibizumab neovascular age-related macular degeneration. N Engl J Med for neovascular age-related macular degeneration clinical 2006;355(14):1419–1431. and economic impact. Ophthalmology 2010;117(11): 4. Martin DF, Maguire MG, Ying GS, Grunwald JE, Fine SL, 2134–2140. Jaffe GJ. Ranibizumab and bevacizumab for neovascular 12. Shienbaum G, Gupta OP, Fecarotta C, Patel AH, Kaiser RS, age-related macular degeneration. N Engl J Med 2011; Regillo CD. Bevacizumab for neovascular age-related macu- 364(20):1897–1908. lar degeneration using a treat-and-extend regimen: clinical 5. Chakravarthy U, Harding SP, Rogers CA, et al. Ranibizumab and economic impact. Am J Ophthalmol 2012;153(3): versus bevacizumab to treat neovascular age-related macular 468–473.e1. degeneration: one-year findings from the IVAN randomized 13. Rayess N, Houston SK 3rd, Gupta OP, Ho AC, Regillo CD. trial. Ophthalmology 2012;119(7):1399–1411. Treatment outcomes after 3 years in neovascular age-related 6. Kodjikian L, Souied EH, Mimoun G, et al. Ranibizumab macular degeneration using a treat-and-extend regimen. Am J versus bevacizumab for neovascular age-related macular Ophthalmol 2015;159(1):3–8.e1. degeneration: results from the GEFAL noninferiority ran- 14. Oubraham H, Cohen SY, Samimi S, et al. Inject and extend domized trial. Ophthalmology 2013;120(11):2300–2309. dosing versus dosing as needed: a comparative retrospective 7. Krebs I, Schmetterer L, Boltz A, et al. A randomised double- study of ranibizumab in exudative age-related macular degen- masked trial comparing the visual outcome after treatment eration. Retina 2011;31(1):26–30. with ranibizumab or bevacizumab in patients with neovascu- 15. Arnold JJ, Campain A, Barthelmes D, et al. Two-year out- lar age-related macular degeneration. Br J Ophthalmol 2013; comes of ‘‘treat and extend’’ intravitreal therapy for neovascu- 97(3):266–271. lar age-related macular degeneration. Ophthalmology 2015; 8. Heier JS, Brown DM, Chong V, et al. Intravitreal aflibercept 122(6):1212–1219. (VEGF trap-eye) in wet age-related macular degeneration. 16. Rahimy E, Rayess N, Ho AC, Regillo CD. Treatment out- Ophthalmology 2012;119(12):2537–2548. comes for neovascular age-related macular degeneration

VOL. 180 AFLIBERCEPT TREAT-AND-EXTEND THERAPY FOR NEOVASCULAR AMD 149 patients with initial vision better than 20/40 using a treat- degeneration: ninety-six-week results of the VIEW studies. and-extend regimen. Retina 2015;36(5):875–880. Ophthalmology 2014;121(1):193–201. 17. Berg K, Hadzalic E, Gjertsen I, et al. Ranibizumab or bevaci- 26. Grunwald JE, Pistilli M, Daniel E, et al. Incidence and growth zumab for neovascular age-related macular degeneration of geographic atrophy during 5 years of Comparison of Age- according to the Lucentis Compared to Avastin Study Related Macular Degeneration Treatments Trials. Ophthal- treat-and-extend protocol: two-year results. Ophthalmology mology 2017;124(1):97–104. 2016;123(1):51–59. 27. Rofagha S, Bhisitkul RB, Boyer DS, Sadda SR, Zhang K. 18. Abedi F, Wickremasinghe S, Islam AF, Inglis KM, Seven-year outcomes in ranibizumab-treated patients in AN- Guymer RH. Anti-VEGF treatment in neovascular age- CHOR, MARINA, and HORIZON: a multicenter cohort related macular degeneration: a treat-and-extend protocol study (SEVEN-UP). Ophthalmology 2013;120(11):2292–2299. over 2 years. Retina 2014;34(8):1531–1538. 28. Maguire MG, Martin DF, Ying GS, et al. Five-year outcomes 19. Toalster N, Russell M, Ng P. A 12-month prospective with anti-vascular endothelial growth factor treatment of trial of inject and extend regimen for ranibizumab treat- neovascular age-related macular degeneration: the Compari- ment of age-related macular degeneration. Retina 2013; son of Age-Related Macular Degeneration Treatments Trials. 33(7):1351–1358. Ophthalmology 2016;123(8):1751–1761. 20. Berg K, Pedersen TR, Sandvik L, Bragadottir R. Comparison 29. Peden MC, Suner IJ, Hammer ME, Grizzard WS. Long-term of ranibizumab and bevacizumab for neovascular age-related outcomes in eyes receiving fixed-interval dosing of anti- macular degeneration according to LUCAS treat-and- vascular endothelial growth factor agents for wet age- extend protocol. Ophthalmology 2014;122(1):146–152. related macular degeneration. Ophthalmology 2015;122(4): 21. Wykoff CC, Croft DE, Brown DM, et al. Prospective trial of 803–808. treat-and-extend versus monthly dosing for neovascular age- 30. Jaffe GJ, Kaiser PK, Thompson D, et al. Differential response related macular degeneration: TREX-AMD 1-year results. to anti-VEGF regimens in age-related macular degeneration Ophthalmology 2015;122(12):2514–2522. patients with early persistent retinal fluid. Ophthalmology 22. Barthelmes D, Campain A, Nguyen P, et al. Effects of switch- 2016;123(9):1856–1864. ing from ranibizumab to aflibercept in eyes with exudative 31. Waldstein SM, Simader C, Staurenghi G, et al. Morphology age-related macular degeneration. Br J Ophthalmol 2016; and visual acuity in aflibercept and ranibizumab therapy for 100(12):1640–1645. neovascular age-related macular degeneration in the VIEW 23. Homer N, Grewal DS, Mirza RG, Lyon AT, Gill MK. trials. Ophthalmology 2016;123(7):1521–1529. Transitioning to intravitreal aflibercept following a previous 32. Toth CA, Decroos FC, Ying GS, et al. Identification of fluid treat-and-extend dosing regimen in neovascular age-related on optical coherence tomography by treating ophthalmolo- macular degeneration: 24-month results. Eye (Lond) 2015; gists versus a reading center in the Comparison of Age- 29(9):1152–1155. Related Macular Degeneration Treatments Trials. Retina 24. Stewart MW, Rosenfeld PJ. Predicted biological activity of 2015;35(7):1303–1314. intravitreal VEGF Trap. Br J Ophthalmol 2008;92(5): 33. Stone TW. ASRS Preferences and Trends (PAT) Survey. 667–668. American Society of Retinal Specialists; 2016. Available at: 25. Schmidt-Erfurth U, Kaiser PK, Korobelnik JF, et al. Intravi- http://www.asrs.org/content/documents/_asrspatslides2016 treal aflibercept injection for neovascular age-related macular finalsm.pdf, Accessed November 22, 2016.

150 AMERICAN JOURNAL OF OPHTHALMOLOGY AUGUST 2017