Angiographic Subtypes of Neovascular Age-related Macular Degeneration in Korean: A New Diagnostic Challenge

Neovascular age-related macular degeneration (AMD) is the leading cause of irreversible blindness in elderly population. Several classifications schemes have been developed to provide subtypes of neovascular AMD, which are known to be associated with visual prognosis. However, there is still a large proportion of patient with ambiguous findings according to current classification criteria. In this study, we classified treatment-naïve neovascular AMD patients using novel angiographic classification system and investigated the incidence and clinical characteristics of AMD subtypes. Among 339 eyes, five AMD subtypes were identified: 41 (12.1%) with classic choroidal neovascularization (CNV), 30 (8.8%) with occult CNV, 91 (26.8%) with microaneurysmal choroidal vasculopathy (MCV), 123 (36.3%) with polypoidal choroidal vasculopathy (PCV), and 54 (15.9%) with retinal angiomatous proliferation (RAP). MCV was younger than RAP (P < 0.001). Classic CNV presented with worse visual acuity compared with MCV at baseline (P < 0.001). Central macular subfield thickness was highest in RAP, and lowest in MCV (P = 0.036). Subfoveal choroidal thickness was highest in MCV, and lowest in RAP (P < 0.001). There was a significant difference in visual acuity at 12 months among five subtypes (P = 0.046). Our results highlight the importance of angiography for identifying AMD subtypes, particularly the novel MCV group being distinct from other subtypes.


Results
We identified 353 eyes from 304 patients with neovascular AMD based on their electronic records. Fourteen eyes were excluded because they had vitreous hemorrhages (n = 6), or because of an absence of ICGA photographs at baseline (n = 8). Ultimately, 339 eyes from 290 patients with neovascular AMD were included in this study. The mean age of the patients was 72.4 ± 8.0 years (range, 50~92), and 59.0% were male ( Table 1). The most common types of neovascular AMD were PCV (36.3%) and MCV (26.8%). Classic CNV, occult CNV, and RAP were identified in 12.1%, 8.8%, and 15.9% of eyes, respectively. There was excellent interobserver reproducibility of the angiography-based classification (ICC = 0.804, 95% confidence interval 0.719-0.883). The patients in the MCV and PCV groups were younger than were those in the classic CNV or RAP groups. The proportion of men was significantly higher in the MCV and PCV groups than it was in the other groups (73.6% and 67.5%, respectively). At baseline, the BCVA in classic CNV, occult CNV, MCV, PCV, and RAP were 20/400, 20/100, 20/80, 20/100, and 20/142, respectively (P < 0.001). There were no significant differences across the five groups in the prevalence of subretinal hemorrhage at baseline. Indocyanine green angiography findings. Eyes in the MCV, PCV, and RAP groups had a higher proportion of hot spots than did those in the other groups (75.8%, 95.9%, 75.9%, respectively, P < 0.001). Polyps were identified after there was pneumatic displacement of submacular hemorrhages in 5 eyes in the PCV group. In these cases, the polyps were not visible at baseline. BVN were detected in 14.6% of classic CNV, 6.7% of occult CNV, 100% of MCV, 84.6% of PCV, and 5.6% of RAP patients (P < 0.001). Plaques were detected in the late phase of ICGA in 40.0% of occult CNV and 20.4% of RAP cases, which differed significantly from those in the other groups (P < 0.001).

Discussion
The use of FA and ICGA in neovascular AMD has been a clinically relevant issue under investigation since its introduction 14 . Using ICGA has allowed providers to define subtypes of neovascular AMD according to the angiographic characteristics of CNV. However, we still see patients with unclear diagnoses regardless of advanced imaging technology. In this study, neovascular AMD was classified into five subtypes, including the novel MCV group, which was not present in the previous classification system.
This study provides several important issues in investigating the neovascular AMD. We can estimate the proportions of each subtype of neovascular AMD among the Asian population. When classified using FA only, the proportions of classic CNV and occult CNV were 31.0% and 69.0%, respectively. This incidence is similar to previous study, in which classic CNV was found in 23% of eyes and occult CNV in 71% 15 . In contrast, when using ICGA, the most common types of neovascular AMD were PCV (36.3%) and MCV (26.8%). Classic CNV was identified in 12.1% of eyes, occult CNV in 8.8%, and RAP in 15.9%. The sum of eyes with MCV and PCV exceeded 60% of all eyes with neovascular AMD. However, the proportion of RAP was similar to that of previous studies 16,17 . According to the criteria of the EVEREST study 18 , MCV should belong to the category of PCV. However, it could have been classified as a part of occult CNV or chronic CSC in the preceding studies 12,19 , as there is no definite 'visible' polyp observed on ICGA. This finding may explain the discrepancy in the proportion of PCV across studies, even those focusing on the same races 20 . In this regard, this novel classification system may contribute to improved categorization of neovascular AMD patients. This method is strengthened by its comparatively high ratio of inter-observer correlation.
Another advantage of this study is that we can compare the proportion of subtypes classified using FA and ICGA, respectively. Of the classic CNV patients diagnosed using FA, only 39.0% were proper classic CNV, compared to just 12.8% in the occult CNV. While FA was useful to assess leakage from CNV lesions, it was insufficient for demarcating occult or poorly defined CNV, which is necessary for identifying PCV and MCV. Yannuzzi et al. reported that the prevalence of RAP is between 10% and 15%, with a female and Caucasian predominance 16 . The frequency of RAP in this study (15.9%) is consistent with findings reported in similar studies on Caucasians, but is higher than that reported in Asian populations (4.5%) 21,22 . Lee et al. 23 also reported that the rate of RAP was high in Koreans (25%). Interracial differences in the proportion of neovascular AMD can be expected, because

Characteristics
Classic CNV n = 41 Occult CNV n = 30 MCV n = 91 PCV n = 123 RAP n = 54 P-value    www.nature.com/scientificreports www.nature.com/scientificreports/ the Y402H variant plays a major role in the etiology of AMD and is obviously different between Caucasians and Asians 24 . Most Korean patients do not have the Y402H variant. Therefore, there are likely other, yet unidentified, genetic variants that also promote disease progression in RAP.
Sometimes there are ambiguous features on angiography, and retinal manifestations of PCV resemble those of chronic CSC or occult CNV 12,19 . Moreover, Uzawa et al. 25 and Tan et al. 26 advocated that the PCV is not a homogenous disease entity, but a spectrum of disease with discriminative characteristics. Previous study also revealed a lower anti-VEGF treatment response associated with thicker choroids in PCV patients 27 . These results are likely to be caused by the coexistence of different subtypes within the category we classify as PCV. In this regard, the refined classification system including MCV may reduce such debate and clarify the subtype of neovascular AMD. MCV is distinct from PCV by the absence of definite polyps. It is distinct from occult CNV by the presence of microaneurysms, which are connected with a lacy choroidal vasculature on ICGA (Fig. 3). The choroid of the MCV group patients was most thickened, and the baseline BCVA was the best among the subtypes. However, the MCV group only had one line of visual gain after 12 months of anti-VEGF treatment, which was inferior to that of the PCV group. These results suggest that the MCV group has its own characteristics that distinguish it from other groups.
The characteristics of the patients in MCV group, which includes male predominance, thickened choroid, alterations in RPE layer, and relatively good visual acuity at baseline, is quite interesting because it is comparable to CSC. Freund and associates have proposed the existence of a pachychoroid-driven spectrum of disease comprising "CSC, pachychoroid pigment epitheliopathy, and pachychoroid neovasculopathy, " which may support the hypothesis that CSC develops Type 1 neovascularization 28 . They demonstrated that it is a thickened choroid that drives the RPE changes seen in these eyes. In this regard, the features of MCV may correspond to the pachychoroid spectrum disease and its relevance seems even stronger than PCV. Given that choroid was most thickened in MCV group patients, it is possible that multiple microaneurysms along with the BVN is dominated by changes in the RPE, which, in tum, may be secondarily related to the chronicity of choroidal vascular hyperpermeability. However, since this study is a cross-sectional study, it is difficult to identify the causal relationship between MCV and pachychoroid spectrum diseases. Further longitudinal research is needed in this regard.
There were several limitations in the present study. First, because of its retrospective nature, using data from a single tertiary hospital, our results may not be generalizable to the overall Korean population and microaneurysms of MCV subtype could not be identified histologically. A second limitation is that the classification was applied at the time of the review process; therefore, it did not influence the treatment regimens. This may be the reason why there was no difference in treatment modality. Another limitation of this study is its relatively short follow-up period of 12 months. Our results, therefore, may not reflect the final outcomes of neovascular AMD. Despite these limitations, our data provide evidence that are compared to results from other American, Asian, and even previous Korean studies. We observed five subtypes of neovascular AMD in this study, including classic CNV, occult CNV, MCV, PCV, and RAP. The demographic, angiographic and OCT findings varied depending based on these subtypes. The MCV group has distinctive features from other groups. Further investigation is needed to clarify the pathogenesis and prognosis of this subtype of neovascular AMD.

Methods
Design and setting. We conducted a review of patients at a single center according to the tenets of the Declaration of Helsinki. The study protocol was approved by the institutional review board and ethics committees at Kyung Hee University Hospital. All patients provided written informed consent before the treatment.
The study population consisted of consecutive treatment naïve patients aged 50 years or older who were diagnosed with neovascular AMD between January 2011 and June 2017. The exclusion criteria were as follows: eyes with myopic CNV, central serous chorioretinopathy (CSC), diabetic retinopathy, retinal vasculopathy, inflammatory diseases, intraocular tumors, history of intraocular surgery (other than cataract surgery), and other ocular diseases that could influence visual function.
Our baseline investigations were as follows: an examination for the Snellen best-corrected visual acuity (BCVA) and anterior segment by slit lamp; dilated fundus examination; and confocal scanning laser ophthalmoscopy (SLO) FA and ICGA using eye-tracked Spectralis Heidelberg Retina Angiograph (HRA) + OCT (Heidelberg Engineering, Heidelberg, Germany). Angiograms were obtained using a standardized imaging protocol. The early phases (up to 60 s) were imaged using dynamic ICGA and FA. Consecutive photographs were then taken for up to 1 min, and again at approximately 2.5, 5, 10, 15 and 20 minutes. These images were reviewed using Heidelberg Eye Explorer software (V.1.7.0.0). We also obtained simultaneous SD-OCT cross-sectional images, which correspond to either the site of the hot spot, plaque, or the network vessels on SLO. In addition, horizontal and vertical cross-sectional enhanced depth imaging (EDI)-OCT was performed at the fovea.
ICGA and FA-based classification. We devised a classification system of neovascular AMD based on prior observation of patients from our clinic. Two independent retinal specialists (K.B. & E.S.K.) who were blinded to the treatments classified the angiographic findings into one of five types (Fig. 4). In cases in which there was disagreement between the two observers, a third independent blinded observer (S.Y.Y.) analyzed the images. When it was difficult to specify one of these categories, it was classified as miscellaneous. Angiographic classifications were defined as below: • Classic CNV was defined by discrete, well-demarcated focal areas of hyperfluorescence that can be discerned in the early phases of the FA. Hyperflourescence increases in intensity and extends beyond the boundaries of the hyperfluorescent area, which is identified in earlier phases of the angiogram of mid-and late-phase frames.
www.nature.com/scientificreports www.nature.com/scientificreports/ • Occult CNV was defined by late choroidal-based leakage in the early or mid-phase FA to account for an area of leakage in the late phase. There was no clearly identifiable classic CNV in FA or the presence of a polypoidal lesion, vascular network, or anastomosis in ICGA.  www.nature.com/scientificreports www.nature.com/scientificreports/ treatment protocols. Patients were treated using anti-VEGF agents (Bevacizumab, Ranibizumab, and Aflibercept) by two retinal specialists (E.S.K. and S.Y.Y.) using a standardized treat-and-extent protocol 29 . We defined patients in whom subretinal or intraretinal fluid was present or increased, even in the presence of monthly anti-VEGF injections, as non-response. Active polyps, as well as lesions that were non-responsive to 3 consecutive anti-VEGF injections, were treated using photodynamic therapy (PDT) with verteporfin (Visudyne, Novartis International AG, Basel, Switzerland) guided by ICGA. Full-dose, full-duration PDT was performed. The treatment zone only included the visible polyps in PCV and active CNV lesions in other categories. The treatments were initiated or repeated when there were clinical signs of disease activity, such as a drop in BCVA of ≥1 line, subretinal fluid or macular edema, or evidence of active CNV on angiography. study measurements. We recorded the patient age, sex, BCVA, and the lesion type at the time of the first injection by ICGA. During the follow-up visits at 3, 6, 9, and 12 months after the initial treatment, all patients underwent BCVA measurement, dilated fundus examination, and SD-OCT. At each visit, all of the treatments were recorded, along with the BCVA and central subfield thickness of the macula.
The following parameters were used to evaluate the SD-OCT images: the presence and type of RPE detachment, subretinal fibrosis, atrophy of the ellipsoid zone, and subfoveal choroidal thickness. www.nature.com/scientificreports www.nature.com/scientificreports/ study outcomes. The primary outcome was the difference in the baseline characteristics of the subtypes of neovascular AMD based on novel classification. The secondary outcomes included the mean BCVA and central subfield thickness 12 months after initiating treatment, and the number of intravitreal injections. statistical analysis. Statistical analyses were performed using SPSS version 24 (SPSS, Chicago, Illinois, USA). Interobserver agreements were evaluated using the intraclass correlation coefficient (ICC) value. The Snellen BCVA was converted into logarithm of the minimal angle of resolution (LogMAR) units prior to the analysis. The chi-square test was used to compare categorical variables, while the one-way analysis of variance (ANOVA) was used for continuous variables. Appropriate parametric analyses were performed when the data had a normal distribution. P-values less than 0.05 were considered statistically significant.