Long-term outcome of intravitreal anti-vascular endothelial growth factor treatment for pachychoroid neovasculopathy

To compare the long-term effectiveness of intravitreal anti-vascular endothelial growth factor (VEGF) treatment for pachychoroid neovasculopathy (PNV), polypoidal choroidal vasculopathy/aneurysmal type 1 neovascularization (PCV/AT1), and typical neovascular age-related macular degeneration (nAMD). Forty-one eyes with PNV, 68 eyes with PCV/AT1, and 56 eyes with typical nAMD were retrospectively included for analysis. All patients were treatment-naïve and received a three-monthly loading injection of anti-VEGF, followed by further injections, as required. The visual and anatomical outcomes after treatment were evaluated up to 36 months from baseline. No significant intergroup difference was found in terms of best-corrected visual acuity (BCVA) and changes in central foveal thickness at 12, 24, and 36 months after the baseline. In addition, no significant difference was found between the groups regarding the proportions of improved or worsened (increased or decreased more than 3-lines) visual acuity. However, the PNV group participants received significantly fewer anti-VEGF injections (11.7 ± 6.9) than those in the PCV/AT1 (12.4 ± 7.0; P = 0.031) and typical nAMD groups (13.2 ± 7.4; P = 0.016). The incidence of macular atrophy (MA) development was also significantly lower for the PNV (4/41 eyes, 9.8%) than the typical nAMD (15/56 eyes, 26.8%; P = 0.033) eyes. There was no significant difference between PNV, PCV/AT1, and typical nAMD regarding visual acuity improvement after anti-VEGF treatment over 36 months. However, the number of injections for PNV was significantly lower compared to that for PCV/AT1 and typical nAMD, and the incidence of MA development was significantly lower than in typical nAMD.

www.nature.com/scientificreports/ of anti-VEGF treatment for PNV [10][11][12] . However, to date, long-term evaluation of the treatment efficacy of anti-VEGF for PNV has not been sufficiently reported. The purpose of the current study was to evaluate the longterm therapeutic efficacy of anti-VEGF treatment for PNV and to compare the therapeutic results with those of other types of nAMD.

Methods
The patients diagnosed with nAMD and treated with anti-VEGF (ranibizumab or aflibercept) between January 2016 and August 2017 were identified by computerized searching of the electronic medical charts. All patients were diagnosed and treated at the Retina Center of Kim's Eye Hospital, Konyang University College of Medicine. The current research followed the tenets of the Declaration of Helsinki, and was approved by the institutional review board at Kim's Eye Hospital. The need for informed consent was waived by the Institutional Review Board.
Subjects. The inclusion criteria were as follows: (1) over 50 years old; (2) confirmation of n AMD with spectral-domain optical coherence tomography (SD-OCT, Spectralis®; Heidelberg Engineering, Dossenheim, Germany), fluorescein angiography (FA), and indocyanine angiography (ICGA; HRA-2, Heidelberg Engineering) at the first visit; (3) no previous anti-VEGF treatment; and (5) a minimum follow-up period of 36 months. When a case of bilateral nAMD was encountered, only the eye earlier diagnosed was included.
All the eyes were classified as PNV, PCV/AT1, and typical nAMD after reviewing the medical records of the patients with nAMD. The definition of PNV was based on the latest reports 2,10,13,14 , since it is still evolving. PNV was diagnosed using the following criteria: (1) definite type 1 CNV on OCT and ancillary images, including FA/ICGA images; (2) no drusen in both eyes (no AMD, AREDS category; (3) presence of dilated choroidal vessels below the macular neovascularization; and (4) PPE or CSC characteristics including choroidal vascular hyperpermeability (CVH) on ICGA 15 , or a history of CSC. When it was difficult to ascertain whether a shallow irregular RPE elevation was a true type 1 macular neovascularization or not, OCT angiography was used for confirmation. The diagnosis of PCV/AT1 was based on the definite presence of polypoidal lesions with or without branching vascular networks in the ICGA images. When the polypoidal lesion could not be evaluated due to severe subretinal hemorrhage at baseline, the lesion was re-evaluated after the loading injection, following repeated ICGA. The representative cases are shown in Fig. 1.
The exclusion criteria were as follows: (1) type 3 macular neovascularization (also known as retinal angiomatous proliferation [RAP]); (2) end-stage nAMD with macular fibrosis or geographic atrophy (GA) at baseline; (3) treated with concomitant photodynamic therapy; and (4) presence of other concomitant ocular diseases, such as diabetic retinopathy or retinal vein occlusion.  www.nature.com/scientificreports/ As a routine for nAMD treatment in our institution, all the patients were administered three loading injections of anti-VEGF (ranibizumab [0.5 mg/0.05 mL] or aflibercept [2 mg/0.05 mL]) at monthly intervals. After the initial loading injections, the patients were followed up during the study period at 4 to 8-week intervals, and they received further injections as required. All the patients underwent standardized examinations at every visit, including best-corrected VA (BCVA), fundus examination, SD-OCT (consisting of 19 or 31 horizontal lines [6 mm × 6 mm area]), and additional FA/ICGA, OCT angiography, or autofluorescence (AF) at the discretion of the physician.
Patient assessment and outcome measures. The main visual outcome was the mean change in bestcorrected visual acuity (BCVA; logarithm of the minimal angle of resolution [logMAR] converted from Snellen BCVA) from baseline to 3, 12, 24, and 36 months. The proportion of patients who gained or lost more than three lines of BCVA compared to the baseline was also recorded.
The main anatomical outcome was the mean change in central foveal thickness from baseline to 3, 12, 24, and 36 months. Central foveal thickness was measured between the vertical distance from the hyperreflective line of Bruch's membrane to the inner limiting membrane on the fovea-centered SD-OCT images. In addition, the development of macular atrophy (MA) in the patients was evaluated during the 36-month follow-up period. Presence of MA was detected via color fundus photography (CFP), SD-OCT, infrared reflectance image, and autofluorescence images (AF). MA was defined as in our previous investigations 16,17 : (1) well-demarcated, hypopigmented area with a minimum linear dimension of 250 µm; (2) occurred within the macular vascular arcades; (3) uniformly reduced autofluorescence signals on AF or increased visibility of the underlying choroidal vessels; and (4) confirmation of increased signal transmission on the SD-OCT findings due to the absence of RPE.
All measurements were estimated using the Heidelberg Eye Explorer software (v. 5.6.4.0; Heidelberg Engineering) by two retinal specialists (J.L. and S.K.N.) who were blinded to the patient information. When the evaluation was inconsistent for the diagnosis of PNV between the graders, a senior investigator (H.J.C.) made the final decision after an open discussion.
Statistical analysis. SPSS for Windows, Version 18.0 (SPSS Inc., Chicago, IL, USA) was used for all the statistical analyses. Frequencies were compared between the groups by using the chi-square or Fischer's exact test. One-way analysis of variance (ANOVA) was used to compare continuous variables between the groups. The Bonferroni correction was used to make statistical adjustments for multiple comparisons. Inter-grader agreement was measured using Cohen's Kappa coefficient. A P-value of less than 0.05 was considered statistically significant.

Results
Baseline characteristics. In total, 165 eyes (165 patients) met the inclusion criteria and were enrolled in the study for analysis. After reviewing the patients' medical records, 41 eyes were categorized into the PNV group, 68 eyes into the PCV/AT1 group, and 56 eyes into the typical nAMD group. The intra-grader agreement between the graders for the diagnosis of PNV showed an almost perfect agreement (the Cohen's Kappa was 0.88). All patients were South Koreans, and the clinical details of the enrolled patients are presented in Table 1.
No significant intergroup difference was observed regarding sex, baseline BCVA, baseline central foveal thickness, lesion location, and lesion size (Table 1). However, significant differences between the groups were identified with respect to age, subfoveal choroidal thickness, presence of choroidal vascular hyperpermeability, and features of fluids at baseline ( Table 1).
The feature of exudative fluid at baseline was significantly different between the groups. No significant difference was found regarding the incidence of subretinal fluid (SRF) between the groups, However, the PNV group showed a significantly lower incidence of intraretinal fluid (IRF, 5/41 eyes, 12.2%) and retinal hemorrhage (5/ Visual and anatomical outcomes. The mean BCVA change from the baseline to 36 months was similar between the PNV, PCV/AT1, and typical nAMD groups during the anti-VEGF treatment (Fig. 2). The www.nature.com/scientificreports/ The BCVA at 3 months was significantly different between the groups (ANOVA, P = 0.026); Bonferroni correction showed that the PNV group showed better BCVA compared to typical nAMD (P = 0.040), and no difference compared to PCV/AT1 (P = 0.220). However, no difference was found between PCV/AT1 and typical nAMD (P = 0.057). Moreover, no significant difference of BCVA between groups was found at 12 months (ANOVA, P = 0.327), 24 months (P = 0.118), and 36 months (P = 0.059).
The injection number during the 36 months showed a significant difference between the groups (ANOVA, P = 0.023, Table 2). Bonferroni correction showed that the injection number of PNV (11.7 ± 6.9) was significantly Table 1. Baseline characteristics of patients treated with anti-VEGF for macular neovascularization. a Based on one-way analysis of variance. b Based on chi-square test. c Patients for whom the anti-VEGF treatment was switched during the study period; there was a switch from ranibizumab to aflibercept for 14 eyes (66.7%) and a switch from aflibercept to ranibizumab for the others. BCVA best-corrected visual acuity, IRF intraretinal fluid, logMAR logarithm of the minimum angle of resolution, nAMD neovascular age-related macular degeneration, PCV/AT1 polypoidal choroidal vasculopathy/aneurysmal type 1 neovascularization, PNV pachychoroid neovasculopathy, SD standard deviation, SRF subretinal fluid, VEGF vascular endothelial growth factor.

Discussion
In the current study, anti-VEGF treatment for PNV showed similar long-term visual and anatomical outcomes compared to PCV/AT1 and typical nAMD. Although the BCVA of the PNV group at three months tended to be better than that of the PCV/AT1 and typical nAMD groups, with statistical difference, thereafter, the BCVA at 12, 24, and 36 months showed no difference between the groups. No significant difference was found regarding Figure 2. Changes in mean best-corrected visual acuity (BCVA), expressed as logarithm of the minimal angle of resolution, during the 36-month anti-vascular endothelial growth factor treatment for macular neovascularization. The BCVA at 3 months was significantly different between the pachychoroid neovasculopathy (PNV), polypoidal choroidal vasculopathy/aneurysmal type 1 neovascularization (PCV/AT1), and typical neovascular age-related macular degeneration (nAMD) (ANOVA, P = 0.031). However, there was no significant difference between the groups at 12, 24, and 36 months. Table 2. Three-year results of the treatment with anti-VEGF for macular neovascularization. a Based on oneway analysis of variance. b Based on chi-square test. BCVA best-corrected visual acuity, logMAR logarithm of the minimum angle of resolution, nAMD neovascular age-related macular degeneration, PCV/AT1 polypoidal choroidal vasculopathy/aneurysmal type 1 neovascularization, PNV pachychoroid neovasculopathy, SD standard deviation, VEGF vascular endothelial growth factor. www.nature.com/scientificreports/ the proportion of improved and worsened (increased or decreased more than 3-lines) visual acuity, at 36 months, between the groups. Notably, the visual outcome was achieved in the PNV group with fewer injections compared to the PCV/AT1 and typical nAMD groups.

PNV (41 eyes) PCV/AT1 (68 eyes) Typical nAMD (56 eyes) P
Recently, our study group reported a low recurrence rate of PNV with fewer injections during 12 months of follow-up 10 . Another investigation has reported that the PNV group received fewer aflibercept injections than the type 1 nAMD group with the treat-and-extend (TNE) regimen 18 . The current study also showed that the PNV group received fewer anti-VEGF injections during the 36 months of long-term follow-up. Our results suggest that the response to anti-VEGF treatment may differ in PNV and other types of nAMD.
Recent evidence suggests that there are distinct profiles of VEGF and cytokines between PNV and other types of nAMD. Hata et al. reported that the mean VEGF concentration of PNV was significantly lower than that of nAMD 19 . Kato et al. reported elevated VEGF levels without elevated complement component 3a (C3a) and macrophage chemoattractant protein 1 (MCP-1) for PNV, compared to typical nAMD 14 . Furthermore, in PNV patients, a lack of correlation between cytokine levels and anti-VEGF treatment response has been noted, whereas a negative association is typical in nAMD patients 13 . These differences in VEGF and cytokine profiles between PNV and typical nAMD could have led to the different therapeutic efficacies of anti-VEGF, or the mechanism by which VEGF is involved in angiogenesis.
The feature of baseline fluid also might affect the results. IRF was significantly less frequent for PNV than for PCV/AT1 and typical nAMD in this study. Given that IRF is associated with lower baseline visual acuity 20 , delayed response to anti-VEGF treatment 21 , more frequent injections than SRF 22 , and poor visual outcome 23 , the significant infrequency of IRF in the PNV group could have affected the injection frequency.
Interestingly, although it has been proposed that PNV and PCV/AT1 share clinical phenotypes in many aspects except the presence of polypoidal lesions 24 , the injection number in PNV was also lower compared to PCV/AT1. Presence of polypoidal lesions is a hallmark of PCV/AT1, and a major cause of retinal hemorrhage, exudation, and pigment epithelial detachment (PED) 25 . Often, retinal hemorrhage from the rupture of the polypoidal lesion might need an intensive anti-VEGF treatment 25 . Thus, the paucity of retinal hemorrhage in the PNV than the PCV/AT1 groups could be associated with the fewer injections in the PNV group. Further investigation is needed regarding the reason for the different frequency of anti-VEGF injection, including the difference in pathophysiology between PNV and other types of nAMD.
The incidence of MA was significantly lower in the PNV group than in the typical nAMD group, being similar to that of PCV/AT1 in this study. The development of MA during anti-VEGF treatment for nAMD is an unwanted side effect of the therapy. Once RPE atrophy has developed, it could influence the visual outcome over a long-term follow-up period because the lesion tends to increase in size over time 16 . Furthermore, the risk www.nature.com/scientificreports/ of developing MA raises the possibility that the visual benefit achieved from anti-VEGF treatment may not be maintained in the long term. It is well known that the incidence of MA is different for the subtypes of nAMD; the incidence of type 3 neovascularization (RAP) is higher 16 , whereas the incidence of PCV/AT1 is lower than that of typical nAMD 17 . Since the choroidal perfusion status is a causative factor for MA development 26 , and low choroidal thickness is a risk factor for GA development after anti-VEGF treatment 16 , it could be speculated that the relatively high choroidal thickness of PNV might be associated with the lower incidence of MA development. Another reason for the low incidence of MA development in PNV could be the low number of anti-VEGF injections in our study. However, in our study, there was no significant difference in terms of MA development between PNV and PCV/AT1, despite the difference in injection number. Indeed, there is still a controversy as to whether frequent anti-VEGF injections could affect MA development. Previously, the CATT study showed that monthly anti-VEGF injection had a higher risk of MA development than as-needed injection 27 . On the contrary, Recent investigations reported that MA development was not associated with the number of anti-VEGF injections between PRN and TNE regimens 28 . The effect of number of injections administered on the development of RPE atrophy should be investigated in the future.
Although the current study is the first to investigate the long-term therapeutic efficacy of anti-VEGF treatment for PNV during more than 36 months of follow-up, the study has several limitations, including its retrospective nature. First, the diagnostic criteria for PNV may not be ideal because no definite diagnostic criteria have been established so far. We defined PNV according to the most recent studies; nevertheless, a standard definition of PNV should be investigated in the future. Second, aflibercept and ranibizumab were not strictly differentiated. However, the two anti-VEGF agents have comparable visual outcomes for the treatment of neovascular AMD 29 . A recent study reported that ranibizumab and aflibercept had comparable visual outcomes for the treatment of PNV 12 . Third, we excluded the patients who underwent PDT because adjuvant PDT during anti-VEGF treatment might affect the number of injections 30 . Despite PDT, also reported as one of effective treatment modalities for PNV 31 , the long-term effect of PDT on PNV could be investigated in a future study.
In conclusion, there was no significant difference between PNV, PCV/AT1, and typical nAMD regarding visual acuity improvement after anti-VEGF treatment during 36 months. However, the injection number for PNV was significantly lower compared to PCV/AT1 and typical nAMD, and incidence of MA development was significantly lower than for typical nAMD. Further investigations regarding the differences in the pathophysiology between PNV and other types of nAMD are warranted in the future.