Long-term outcomes of ranibizumab vs. aflibercept for neovascular age-related macular degeneration and polypoidal choroidal vasculopathy

To evaluate the long-term outcomes of ranibizumab (RBZ) vs. aflibercept (AFL) in treatment-naïve eyes with typical neovascular age-related macular degeneration (nAMD) and polypoidal choroidal vasculopathy (PCV). This multicenter, retrospective, matched-cohort analysis was conducted on data up to 4 years of follow-ups. The primary outcome was the visual acuity (VA) change from baseline. The secondary outcomes included the number of injections, proportion of eyes without a yearly injection, and the number of eyes with treatment switching. Subgroup analyses were performed for typical nAMD and PCV. Typical nAMD was defined as nAMD other than PCV. We included VA-matched 215 eyes of 209 patients (131 and 84 eyes with RBZ and AFL, respectively). The crude mean VA changes from baseline were + 6.7 vs. + 2.6, + 2.1 vs. − 0.4, − 1.3 vs. − 1.8, and − 2.2 vs. − 5.0 letters in the RBZ and AFL groups, at 1, 2, 3, and 4 years, respectively (p > 0.05). The adjusted predicted VA by linear mixed model, proportion of eyes stratified by VA, and the survival curve for significant vision loss were comparable during the 4-year follow-up (p > 0.05). The mean number of injections were similar between the RBZ and AFL groups (2.9 vs. 3.0, respectively, p = 0.692). The subgroup analysis for typical nAMD and PCV showed similar results between the groups. The visual outcomes did not differ between RBZ and AFL during 4 years with comparable numbers of injections. Our study reflects the long-term, real-world clinical practice and treatment pattern of two treatments for typical nAMD and PCV.


Patients.
We enrolled treatment-naïve eyes with newly diagnosed nAMD that started treatment with either RBZ (Lucentis; Genentech, Inc., CA/Novartis, Basel, Switzerland; 0.5 mg/0.05 mL) or AFL (Eylea; Regeneron, Inc., NJ/Bayer, Leverkusen, Germany; 2 mg/0.05 mL), from March 1, 2007 to June 31, 2017. Eyes were included only when the same drug was maintained without switching for at least 1 year after the initial treatment. In South Korea, AFL became available and was funded in 2013. Therefore, only eyes that started treatment after 2013 were included for patient matching between the two treatment groups. Consequently, eyes that started treatment between January 2013 and June 2015 were included in the present study. Eyes were matched for baseline visual acuity (VA), and a matching ratio of 1:2 (RBZ to AFL) was used to maintain the maximum number of subjects as possible.
A total of 863 treatment-naïve eyes from 819 patients were identified. The Seoul National University Bundang Hospital, Asan Medical Center, and Kim's Eye Hospital cohorts included 367 eyes of 347 patients, 128 eyes of 114 patients, and 368 eyes of 358 patients, respectively. From this population, 215 eyes of 209 patients (131 eyes with RBZ and 84 eyes with AFL) were finally included in the analysis after attrition by inclusion criteria and patient matching. A flow chart of the study population is presented in Fig. 1.
The treatment regimen in this study varied depending on the preference of practitioners, with either the PRN or T&E regimens adopted. The first injection date was regarded as the baseline. Labeled usage, which was reimbursed through Korean National Health Insurance, consisted of a bimonthly injection after three loading dose injections for AFL and a monthly injection after three initial injections for RBZ. For eyes with insufficient response to RBZ or AFL, the treatment could be switched by a clinician.
Patient evaluation and grouping. In all patients, baseline ophthalmic examinations included VA measurement in decimals, dilated fundus examination, fluorescein angiography, indocyanine green angiography www.nature.com/scientificreports/ (ICGA), and optical coherence tomography, were performed. VA measurements with refractive error correction were conducted at every visit. The long-term longitudinal follow-up results until 4 years after the initial treatment were evaluated. The VA in decimals was converted into the Early Treatment Diabetic Retinopathy Study letter scores for arithmetic comparison. Eyes were primarily analyzed by treatment group, that is, either RBZ or AFL. For the subgroup analysis, eyes were divided into typical nAMD and PCV eyes. PCV was defined using the following diagnostic criteria: nodular hyperfluorescence of the polyps on ICGA, hypofluorescent halo around the nodules, abnormal vascular channels supplying the polyps (branching vascular networks (BVN)), and orange subretinal nodules on fundus photography corresponding to the polyps on ICGA, as diagnosed in the EVEREST study report 2 18 . nAMD other than PCV was regarded as typical nAMD, which included classic choroidal neovascularization (CNV), occult CNV, and retinal angiomatous proliferation.
Study outcomes. The primary outcomes were the mean VA changes from baseline. Additional visual parameters, including the adjusted prediction of mean VA, the proportion of eyes stratified by VA, and the survival analysis without significant vision loss, were evaluated. VA stratification was evaluated by calculating the proportion of eyes with VA ≥ 70 letters (Snellen's equivalent = 20/40, the threshold of driving vision in the United States) and VA ≤ 35 letters (Snellen's equivalent = 20/200, legally blind). Significant vision loss in the survival analysis was defined as losing 10 letters from baseline at a certain point of the follow-up year. Secondary outcomes included the mean number of injections, the proportion of eyes without a yearly injection, and the number of eyes where the treatment was switched. Subgroup analyses were performed for typical nAMD and PCV. An evaluation of completion rate and a comparison between completers and non-completers were performed. Completion was defined as completing the follow-up until the end of the observation period, regardless of the yearly injection count or treatment switching.

Statistical analysis.
All analyses of the demographics and outcomes were based on the eye as the unit of analysis. A nearest-neighbor strategy-based matching of 1:2 ratio was implemented. Baseline VA was considered as the matching condition. A linear mixed-effect model was used to compensate for the loss to follow-up (LTFU). The chi-square test was used to compare the categorical variables between groups. Continuous variables were compared using independent t-tests. A Kaplan-Meier survival analysis with the log-rank test was utilized to compare the cumulative probability of survival without significant vision loss over time between the groups. Patient matching and linear mixed-effect models were calculated using R software version 3.5.3 (R Project for Statistical Computing, Vienna, Austria). Analyses other than patient matching and the linear mixed-effect models were performed using SPSS software version 25.0.K (IBM Corporation, Chicago, IL, USA). A p-value of less than 0.05 was considered statistically significant.

Results
Study participants. The study's baseline demographic and clinical characteristics are presented in Table 1, which were generally similar and balanced between the RBZ and AFL treatment groups. Baseline mean VA and the proportion of eyes with VA ≥ 70 letters and VA ≤ 35 letters were similar between groups. There was a significant difference in the proportion of subretinal hemorrhage (SRH). SRH was observed in 22.9% of the RBZ group and 11.9% in the AFL group (p = 0.043), and both groups showed a mean early onset of SRH after initial treatment (2.57 ± 7.04 months [RBZ] vs. 3.00 ± 9.00 months [AFL], p = 0.880). The results of the subgroup comparison for typical nAMD and PCV were well-balanced as shown in , p = 0.013), but this difference was not maintained. The Kaplan-Meier survival curves for the cumulative probability of survival without significant vision loss (losing 10 letters) are presented in Fig. 4. The log-rank test for total eyes and the typical nAMD and PCV subgroups showed no difference between the two treatment groups (p > 0.05). The survival analysis showed that more than half of the typical nAMD eyes lost 10  www.nature.com/scientificreports/ 3.0 ± 1.5 for the AFL group, which did not differ between the treatment groups (p = 0.692). The proportion of eyes without a yearly injection was similar between the groups during the entire study period (p > 0.05).
Treatment switching. Treatment switches were only reported in the RBZ treatment group. Thus, switching from RBZ to AFL was significantly more frequent (13.7% [RBZ] vs. 0% [AFL], p = 0.000). The mean follow-up period prior to switching was 2.3 ± 0.6 years. The mean VA when the treatment was switched and mean VA at 1 year after switching were 69.3 ± 9.6 and 66.9 ± 9.2 letters, respectively, and did not differ significantly (p = 0.425, paired t-test). The subgroup analysis for typical nAMD and PCV showed similar results to the total eyes (see Supplemental Table S2).  Table 1). We compared the variables between the completers and noncompleters at each time point and found that baseline VA, VA at the last follow-up, and whether they were diagnosed with typical nAMD or PCV did not differ, but non-completers after 1 year of follow-up were significantly older (p < 0.05; see Supplemental Table S3).

Discussion
In this 4-year long-term multicenter retrospective study, the visual outcomes in the form of adjusted predictions, mean VA change from baseline, proportion of eyes stratified by VA, and survival analysis without significant vision loss were not different between the two treatment groups. The number of injections and the proportion of eyes without yearly injection also did not differ between the groups. The subgroup analysis for typical nAMD and PCV showed comparable results between the treatment groups.   www.nature.com/scientificreports/ The majority of current real-world studies have reported short-term 1-or 2-year outcomes [19][20][21] , and results of a single anti-VEGF agent [22][23][24][25][26][27] or a mixture of anti-VEGFs without classification [28][29][30][31][32][33][34] . Few studies have compared two anti-VEGF drugs [10][11][12][13] , and long-term head-to-head comparative outcomes to date were limited to 3 years or shorter (Table 4) 12 . We now report the long-term, 4-year outcomes between RBZ and AFL. Furthermore, to the best of our knowledge, we report the first comparative study between treatments for PCV in an Asian population.
The visual outcomes of the present study did not differ between the two treatment groups.  20 , the study of UK AMD EMR Users group with RBZ (+ 2 letters at 1 year, + 1 letter at 2 years) 23 , and the comparative   Although the exact impact on the injection counts remains unclear, this could have substantially impacted the results of the present study. The authors assert that differences in insurance systems must be considered when interpreting the results of real-world studies, as the AURA study discovered that the number of visits and injections and visual outcomes varied between countries 20 . However, in this study, comparable visual outcomes were achieved with substantially fewer injections, showing the characteristics of nAMD patients in South Korea, with substantial differences in demographics including younger age (69. 8  ) and worse baseline VA, as mentioned above.
We could not evaluate disease activity with the present study's data. Thus, we calculated the proportion of eyes without a yearly injection. The proportion did not differ between the two treatment groups as well as in the subgroup analyses. However, eyes without a yearly injection may include stable and inactive conditions, poor response, or advanced lesions with geographic atrophy or disciform scar change. Further studies that evaluate lesion activity are needed to confirm the results of the present study.
It remains contentious as to whether PCV is a subtype of nAMD or a distinct disease entity 36 . Two large pivotal trials were conducted for the treatment of PCV. The EVEREST II study evaluated RBZ monotherapy vs. RBZ combined with PDT and found higher visual gains in the combination group at 12 months (5.1 vs. 8.3 letters) 37 . However, the results of the PLANET study reported that AFL monotherapy was non-inferior to AFL with rescue PDT up to 96 weeks (10.7 vs. 9.1 letters), and the proportion of patients requiring rescue PDT was small (17%) 38 . The real-world outcomes of the Asian population, with a higher rate of PCV occurrence, have been underrepresented 14 . Matsumiya et al. reported 2-year visual gains of + 5.7 letters with RBZ in the PCV group 15 , and Nishikawa et al. showed that long-term, 4-year results with aflibercept and vision were retained above baseline after the 4-year treatment 17 . In the present study, 48.4% of the total eyes (104 of 215 eyes) had PCV, and the mean VA change from baseline showed that VA was maintained for the entire 4 years in the PCV subgroup, on the contrary, it was below the initial values after 1 year in the typical nAMD subgroup. The survival analysis for significant vision loss also showed that half of the typical nAMD eyes lost 10 letters during the 4-year follow-up period. In contrast, only one-fourth of the PCV subgroup experienced vision loss. Recent real-world outcomes with the FRB! registry compared anti-VEGF monotherapy with a combination of anti-VEGF and PDT and found that the combination group showed larger vision gains with fewer injections 16 . Only two patients in the PCV subgroup received PDT in this study, and we were, therefore, unable to compare the treatment modalities. Our data shows that anti-VEGF monotherapy is the mainstream treatment for PCV in South Korea. Further studies should be conducted to find the best treatment option for Asian people with PCV.
In this study, treatment switches were only reported in the RBZ treatment group, from RBZ to AFL (18 eyes, 13.7%). The rate of switching treatment is comparable to the results of previous studies (12.5, 15%) [10][11][12] , and eyes that switched treatment did not show a VA difference after 12 months, as previously reported by Barthelmes et al. and Chakravarthy et al. 39,40 . We contemplate that this one-way result was due to the effect of newly introduced drugs and RBZ-refractory cases. However, the possible effect of practitioners preferring AFL for poor response eyes could not be ruled out. The results of the report that AFL further inhibits VEGF B and placental growth factor, as well as VEGF A, might have affected the drug choice 41 .
The LTFU rate in our study was comparable to the results of observational reports 42 . The LTFU results were similar between the two treatments and in the typical nAMD and PCV subgroups. Non-completers after 1 year were significantly older than completers (p < 0.05), and we believe that the inability to visit clinics and high mortality and comorbidity rates in older patients may contribute to LTFU. A previous study by Lotery et al. also reported similar results that discontinuing eyes were older, although they also found low baseline VA in non-completers 11 . Bhandari et al. reported that reasons for discontinuation were not due to poor outcomes in most cases 12 . Many previous studies used the last observation carried forward (LOCF) method, which carries the latest observed value of non-completers to the end, to deal with LTFU 10,12,43,44 . We concluded that the LOCF method is not applicable in the current study and could over-or under-estimate the outcomes because of the high proportion of LTFU after the completion of 1 year. Instead, we adopted a mixed-effects regression model to make full use of the data of non-completers. www.nature.com/scientificreports/ This study has several limitations. The present study was retrospective and non-randomized in design, which can lead to selection bias. This study has lower internal validity than randomized controlled trials by nature, and practitioners' personal preferences could affect the initial drug choice and treatment regimen. Additionally, the results of the present study could be affected by the domestic insurance policy. However, our study appears to reflect the long-term, real-world clinical practice in South Korea and may be used as a clinical management resource. Further limitations include a high proportion of LTFU after 1 year; however, this is inevitable in realworld studies. In addition, the number of visits, lesion size and activity, reasons for SRH occurrence, initial drug choice, treatment discontinuation, and switching treatment could not be evaluated using the data collected as part of this study. Future well-designed studies with larger cohorts are warranted to validate the results of the present study.
In conclusion, the visual outcomes did not differ between RBZ and AFL in the treatment of treatment-naïve eyes with nAMD and PCV over a 4-year period. The number of injections and the proportion of eyes without a yearly injection were also not different between the groups. The subgroup analysis for typical nAMD and PCV showed comparable results between the treatment groups. Our study likely reflects the long-term, real-world clinical practice and treatment patterns in South Korea and compares the outcomes of two treatments for typical nAMD and PCV.

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.