Pachychoroid neovasculopathy and age-related macular degeneration

Pachychoroid neovasculopathy is a recently proposed clinical entity of choroidal neovascularization (CNV). As it often masquerades as neovascular age-related macular degeneration (AMD), it is currently controversial whether pachychoroid neovasculopathy should be distinguished from neovascular AMD. This is because its characteristics have yet to be well described. To estimate the relative prevalence of pachychoroid neovasculopathy in comparison with neovascular AMD and to investigate the phenotypic/genetic differences of the two diseases, we evaluated 200 consecutive Japanese patients who agreed to participate in the genetic study and diagnosed with pachychoroid neovasculopathy or neovascular AMD. Pachychoroid neovasculopathy was observed in 39 individuals (19.5%), which corresponds to one fourth of neovascular AMD. Patients with pachychoroid neovasculopathy were significantly younger (p = 5.1 × 10−5) and showed a greater subfoveal choroidal thickness (p = 3.4 × 10−14). Their genetic susceptibility to AMD was significantly lower than that of neovascular AMD; ARMS2 rs10490924 (p = 0.029), CFH rs800292 (p = 0.013) and genetic risk score calculated from 11 AMD susceptibility genes (p = 3.8 × 10−3). Current results implicate that the etiologies of the two conditions must be different. Thus, it will be necessary to distinguish these two conditions in future studies.


Results
A total of 200 patients met the criteria for the current study. The demographics of the study population are summarized in Table 1. Mean age was 74.3 yr (SD = 8.8 yr). Characteristics were not significantly different between right and left eyes. Pseudodrusen and calcified drusen were observed in 9.0% and 1.0-1.5% of the study population, respectively. Approximately 30% of the eyes had RPE abnormality independent of the CNV lesion, and 15% of the eyes had choroidal vascular hyperpermeability. Mean subfoveal choroidal thickness was 227 μ m (SD = 101 μ m) in right eyes and 229 μ m (SD = 109 μ m) in left eyes.
Thirty-nine (19.5%) of the participants were diagnosed with pachychoroid neovasculopathy, and 161 (80.5%) patients were diagnosed with neovascular AMD. The characteristics of both groups are shown in Table 2. Patients with pachychoroid neovasculopathy were significantly younger than those with neovascular AMD (68.7 yr, vs. 75.6 yr, respectively; p = 5.1 × 10 −5 ). They also had greater subfoveal choroidal thickness (p = 3.4 × 10 −14 ). Although RPE abnormality and choroidal vascular hyperpermeability were more common in patients with pachychoroid neovasculopathy (p = 7.9 × 10 −12 and 4.5 × 10 −7 , respectively), they were also found in 30.4% and 13.6%, respectively, of patients with neovascular AMD. Although choroidal vascular dilatation and polypoidal lesion tended to exist more in pachychoroid neovasculopathy, the difference were not statistically significant (p = 0.093 and 0.11, respectively). Table 3 and Fig. 1 show the genetic profiles of pachychoroid neovasculopathy and neovascular AMD. The genotype distribution of both ARMS2 A69S and CFH I62V differed significantly between the two conditions ( Table 3). The frequency of the T allele in ARMS2 A69S was 51.3% and 64.8%, respectively, in patients with pachychoroid neovasculopathy and neovascular AMD (p = 0.029). The frequency of the A allele in CFH I62V was 41.0% and 25.5%, respectively, in patients with pachychoroid neovasculopathy and neovascular AMD (p = 0.013). These results showed that patients with pachychoroid neovasculopathy were less genetically susceptible to AMD.
Genetic risk score for AMD susceptibility also differed significantly between the pachychoroid neovasculopathy and neovascular AMD groups (OR = 0.52 and 0.56, respectively; p = 3.8 × 10 −3 and 1.5 × 10 −2 ; per one score increase of genetic risk score weighted by Asian, and per one score increase of genetic risk score weighted by Caucasian, respectively) ( Fig. 1). Again, the pachychoroid neovasculopathy patients showed less genetic susceptibility to AMD.
A total of 108 patients treated with ranibizumab were eligible for the survival analysis. Of them, 28 individuals had pachychoroid neovasculopathy and 80 individuals had neovascular AMD (Table 4). The rate of dry macula after a loading dose did not differ significantly between the two groups (90.9% vs. 83.7%; p = 0.51). However, the Kaplan-Meier curves for the retreatment-free periods were significantly different between the two groups (p = 9.5 × 10 −3 ). Pachychoroid neovasculopathy had a longer retreatment-free period (Fig. 2).

Discussion
Pachychoroid neovasculopathy is a new clinical entity of CNV, which is characterized by its shared features with CSC or PPE 17 . However, pachychoroid neovasculopathy often masquerades as neovascular AMD, and standard diagnostic criteria have not yet been established. Since pachychoroid neovasculopathy may respond differently to photodynamic therapy or anti-VEGF therapy, discriminating it from neovascular AMD is of great importance. We hypothesized that pachychoroid neovasculopathy may commonly have been misdiagnosed as AMD, especially in Asians, and that this may partly explain the heterogeneity seen in neovascular AMD between Asians and Caucasians 8 . To estimate the prevalence of pachychoroid neovasculopathy, we investigated 200 CNV patients aged over 50, classifying them into pachychoroid neovasculopathy and neovascular AMD groups based on their phenotypes. In our cohort, pachychoroid neovasculopathy was seen with high frequency (one-fourth of neovascular AMD-diagnosed patients). In addition, our study revealed significantly different genetic backgrounds between the two groups.
As Table 1 shows, the overall clinical characteristics of the participants are similar to those previously reported in Japanese AMD patients. For example, pseudodrusen were observed in 9.0% of the patients, which is comparable to previous reports showing the prevalence of pseudodrusen in Japanese patients with AMD (10.8-16.8%) 18,19 . Our mean subfoveal choroidal thickness (227 ± 101 μ m in right eyes and 229 ± 109 μ m in left eyes) was also similar to that found in previous reports (subfoveal choroidal thickness; 204-245 μ m in neovascular AMD and 243-293 μ m in PCV) 12,20 . Considering the prevalence of choroidal vascular hyperpermeability in cases of PCV has been reported to be 9.8-34.8% [11][12][13] , the prevalence of choroidal vascular hyperpermeability in the current study (16.5% Table 3. Genotypic differences between the patients with pachychoroid neovasculopathy and neovascular AMD. * Adjusting for age and sex. † Compared to 3,248 normal Japanese. ‡ Cited from the original paper. CSC: central serous chorioretinopathy, AMD: age-related macular degeneration, CNV: choroidal neovascularization.

Figure 1. Distribution of genetic risk score calculated from 11 SNPs of 11 AMD susceptibility genes.
We constructed a multi-locus genetic risk score by summing up the number of risk alleles of each single nucleotide polymorphism (SNP), weighted by its reported effect size (log odds ratio, OR). We evaluated both genetic risk score using effect size in Caucasian (A) and that using effect size in Asian (B). The effect sizes applied in this analysis are summarized in Supplementary  Table 4. Summary of the patients who were eligible for the survival analysis. AMD: age-related macular degeneration.
significantly younger than those with neovascular AMD, which was compatible with the original report 17 .
In the current study, however, we evaluated only patients with age of more than 50 years because the main interest of this study was to contrast pachychoroid neovasculopathy to neovascular AMD. Thus, mean age of the patients with pachychoroid neovasculopathy could be rather younger. Since pachychoroid neovasculopathy, unlike neovascular AMD, is characterized by a lack of drusen, and the genetic backgrounds differed significantly between the two groups ( Table 3, Fig. 1), the etiologies of these two conditions are likely to be different. As shown in Table 3, the ARMS2 rs10490924 effect size for neovascular AMD (OR = 3.17) was higher than the previously reported effect size for Japanese neovascular AMD (OR = 2.41) 21 , while it was similar to the same effect size for Caucasian neovascular AMD (OR = 3.67) 22 . It is possible that pachychoroid neovasculopathy was misdiagnosed and included among neovascular AMD cases in previous reports. This would have led to an underestimation of the effect of AMD susceptibility genes in Japanese patients. Furthermore, a similar underdiagnosis of pachychoroid neovasculopathy may have occurred in other Asian studies, considering the higher prevalence of CSC in Asia 23,24 . Indeed, in The AMD Gene Consortium paper, the effect sizes of ARMS2 rs10490924 and CFH rs10737680 for AMD were higher in Caucasian than in Asian patients, despite a lower prevalence of geographic atrophy, which is associated with a lower risk allele frequency, in Asians ( Supplementary Figures 2A and 2B of The AMD Gene Consortium paper) 25 . Interestingly, the CFH rs800292 A allele frequency in pachychoroid neovasculopathy patients was comparable to that in normal Japanese subjects (41.0% vs. 40.5%, p = 0.92). Furthermore, these genetic findings were quite similar to our previous report that compared consecutive CNVs with choroidal vascular hyperpermeability to those without 14 , CNVs with choroidal vascular hyperpermeability, which would correspond to pachychoroid neovasculopathy at least in part, had normal genotype distribution in terms of ARMS2 rs10490924 and CFH rs800292. Considering above, these CNVs should be distinguished from neovascular AMD.
Freund and colleagues hypothesized that pachychoroid neovasculopathy is associated with PCV 16,17 . Consistent with this, 56.4% of pachychoroid neovasculopathy cases are associated with polypoidal lesion. It follows that PCV figures may have been more augmented by misdiagnosed pachychoroid neovasculopathy cases than the figures of other subtypes. This may explain why PCV have thicker choroid 20,26 , more frequent history of CSC 11,27,28 , and more favorable prognosis 29 , particularly in the Asian population. The results may also explain why the effect of ARMS2 rs10490924 is higher for AMD than for PCV in the Asian population 30,31 . All that being said, in our study 42.9% of eyes with neovascular AMD also had polypoidal lesion. Polypoidal legion is increasingly thought to be a manifestation of long-standing type 1 CNV in AMD as well as a variety of other diseases. In general, PCV has been considered as a subtype of neovascular AMD that accompanies polypoidal legion; thus, we suggest that pachychoroid We selected study participants who had received a loading dose of ranibizumab (3 × one-monthly injections, 0.5 mg), been treated as needed afterwards, and been followed up for more than 12 months. The duration from the third ranibizumab injection to the administration of additional treatment (event), or to the final visit by June 2014, or loss to follow up) was reported on a daily basis. A log-rank test revealed that the curves for pachychoroid neovasculopathy and neovascular AMD differed significantly from one another (p = 0.0095). Pachychoroid neovasculopathy showed longer retreatment-free period. neovasculopathy with polypoidal lesion should be distinguished from neovascular AMD with polypoidal lesion so as not to confuse the concepts of neovascular AMD and pachychoroid neovasculopathy.
To distinguish neovascular AMD and pachychoroid neovasculopathy is important not only because it is related to the etiology of CNV, but also because it can influence disease management. Our survival analysis showed that patients with pachychoroid neovasculopathy had a significantly longer retreatment-free period than those with neovascular AMD (p = 9.5 × 10 −3 ) after a loading dose of anti-VEGF therapy. This indicates that pachychoroid neovasculopathy may be more self-limited or that VEGF secretion in pachychoroid neovasculopathy may be lower than in neovascular AMD. This hypothesis is compatible with the following observations: (1) CNV development takes a long time in pachychoroid neovasculopathy; and (2) the VEGF concentration in the aqueous humor of PCV eyes, whose figures may have been more bloated by misdiagnosed pachychoroid neovasculopathy, was reported to be lower than that of neovascular AMD eyes 32 . On the other hand, the dry macula rate after loading dose was over 80% in both conditions, revealing no statistically significant difference between them (p = 0.51). That said, it must be stated that the loading dose may have been adequate to eliminate VEGF in both conditions except for 10%-20% of patients, who may have been non-responder.
While the current study has significant implications, it does have limitations. Firstly, the current criteria for diagnosing pachychoroid neovasculopathy may not be ideal. Based on previous reports, we set our original diagnostic criteria -choroidal thickness, absence of drusen, choroidal vascular hyperpermeability, and RPE abnormalities -because no standard diagnostic criteria for pachychoroid neovasculopathy have been established so far. The current criteria could be more sophisticated to distinguish pachychoroid neovasculopathy and neovascular AMD with higher sensitivity and specificity. For instance, choroidal thickness varies with age and axial length 33 , so these parameters could be taken into account for diagnosis. As another example, we did not diagnose pachychoroid neovasculopathy when there was only one soft drusen in the fellow eye, in order to increase the specificity of the diagnosis. Ideally, the scoring method could have been employed to balance the sensitivity and specificity of the pachychoroid neovasculopathy diagnosis. Nonetheless, even though our criteria were based solely on phenotype, they successfully demarcated a cluster of CNV that was genetically distinct from neovascular AMD. Secondly, this study only included Japanese individuals. Considering the higher prevalence of CSC in Asians than in Caucasians, we speculate that pachychoroid neovasculopathy may have been misclassified as neovascular AMD in Asians more often. Future studies including other ethnicities are indicated. Thirdly, the significant result in the survival analysis may need to be deducted from the study because it is based on retrospective data.
In conclusion, pachychoroid neovasculopathy was different from neovascular AMD not only phenotypically but also genetically. Pachychoroid neovasculopathy may represent up to one quarter of diagnosed neovascular AMD cases. Although pachychoroid neovasculopathy often masquerades as neovascular AMD, their etiology is likely to be different because pachychoroid neovasculopathy shows lack of drusen and the genotype distribution of AMD susceptibility SNPs differed significantly between the two conditions. Pachychoroid neovasculopathy should be distinguished from neovascular AMD in future epidemiological and genetic studies. Lastly, further research is necessary to manage and prevent this disease.

Methods
The current study was approved by the Institutional Review Board at Kyoto University Graduate School of Medicine and adhered to the tenets of the Declaration of Helsinki. Written informed consent was obtained from each genotyped patient.

Subjects.
We retrospectively reviewed the medical records of consecutive patients (1) who had visited the macular service of Kyoto University Hospital (Kyoto, Japan) between January 2010 and October 2012, agreed to participate in the genetic study (the participation rate ≥ 99%), and were genome-scanned, (2) who were diagnosed with either pachychoroid neovasculopathy or neovascular AMD (diagnostic criteria of these diseases are described in the following section), and (3) who were older than 50 years. Patients with the following conditions were excluded from the study: (1) CNV secondary to high myopia (spherical equivalent, ≤ − 6.00 D or axial length of ≥ 26 mm), trauma, angioid streaks, uveitis, or any other neovascular maculopathy, (2)
Color fundus photographs (field, 40°) were obtained digitally using a Topcon TRC NW6S non-mydriatic retinal camera (Topcon, Tokyo, Japan) after medical dilatation of the pupil (phenylephrine 0.5% and tropicamide 0.5%). IR, FAF, FA, and IA images were acquired using a confocal SLO (Spectralis HRA+ OCT; Heidelberg Engineering, Heidelberg, Germany). IR images were obtained using a light stimulus of 820 nm. FAF images were obtained using an excitation light of 488 nm and a barrier filter beginning at 500 nm. The field of view was set to 30° × 30° centered on the macula. SD-OCT was conducted using a Spectralis HRA+ OCT (Heidelberg Engineering). First, horizontal and vertical line scans through the fovea center were obtained at a 30° angle, followed by serial horizontal scans with an examination field size of 30° × 10°. Inverted OCT images, which enable us to measure choroidal thickness, were routinely obtained in all patients using an enhanced-depth imaging (EDI) technique 35 . At each location of interest on the retina, fifty SD-OCT images were acquired and averaged to reduce speckle noise. Image Analysis and Phenotyping. Soft drusen were graded based on the fundus photographs according to the simplified severity scale for AMD from the Age-Related Eye Disease Study (AREDS) 36 . The occurrence of dot-or ribbon-type pseudodrusen was confirmed using color fundus photography, IR, FAF, IA (late phase), or when there was OCT evidence of definite drusenoid deposits above the RPE.
Subfoveal choroidal thickness and choroidal vascular hyperpermeability were evaluated as previously described 14 . Briefly, subfoveal choroidal thickness was defined as the vertical distance between Bruch's membrane and the chorioscleral interface at the fovea, which was manually measured in the EDI-OCT images by a retinal specialist blinded to study parameters using the built-in caliber. Choroidal vascular hyperpermeability was determined by detecting multifocal hyperfluorescent areas with blurred margin that expanded during the late phase of IA (i.e., 10-15 minutes after dye injection) (Fig. 3A). Choroidal vascular hyperpermeability was confirmed only when independent judgments of two retinal specialists (M.M. and S.O.) were agreed. The number of quadrants with dilated choroidal vessels was also evaluated for each eye, based on wide-field IA images taken five to ten minutes after dye injection. Choroidal vascular hyperpermeability is evidenced by focal areas of hyperfluorescence, which appear during the middle phase of indocyanine green angiography (IA) and expand over time, eventually forming a ring shape (arrowheads). The center of the initially hyperfluorescent area becomes hypofluorescent during the late phase. (B) Retinal pigment epithelium abnormality. Retinal pigment epithelium (RPE) abnormality is seen as patchy areas of granular hypoautofluorescence with occasional discrete hyperautofluorescent specks scattered throughout the fundus autofluorescence (FAF) images 16,17 (arrowheads). However, this was not considered indicative of RPE abnormality when this finding was adjacent to a choroidal neovascularization (CNV; arrows) lesion. RPE abnormality was determined by detecting patchy areas of granular hypoautofluorescence with occasional discrete hyperautofluorescent specks scattered throughout the FAF image, according to the original report of PPE 16 . RPE abnormality is often seen in CNV legion; therefore, only RPE abnormalities occurring independent of CNV lesions were evaluated in this study (Fig. 3B). The presence of RPE abnormality was only confirmed when independent judgments of two retinal specialists (M.M. and S.O.) were agreed.
Choroidal thickness is an important factor that characterizes PPE, though it is not included in the definition of CSC. Although the original report did note "focal, multifocal, or diffuse area of reddish orange background within the arcades with minimal to absent choroidal vascular markings (reduced fundus tessellation), indicative of a thickened choroid, " it was somewhat subjective 16 . For this reason, we employed a choroidal thickness threshold value instead. Because the original report presented a PPE case with a subfoveal choroidal thickness of 231 μ m, we set this threshold value at 200 μ m. In another study, the mean subfoveal choroidal thickness in CSC eyes was 414 μ m (95% range, 200 μ m-628 μ m), while that of the contralateral eyes was 350 μ m (95% range, 123-577 μ m) 37 . This further supports the validity of our threshold value.
As a reference group, we used 3,248 unrelated individuals from the general population, recruited from the Nagahama Prospective Genome Cohort for the Comprehensive Human Bioscience (The Nagahama Study) [38][39][40] . These patients were genotyped using HumanHap610 K quad arrays, HumanOmni2.5 M arrays, and/or HumanExome arrays (Illumina Inc., San Diego, CA). The two SNP genotypes were extracted from the cohort's fixed dataset.
We constructed a multi-locus genetic risk score by summing up the number of risk alleles of each SNP, weighted by their reported effect sizes (log odds ratio, OR). Because of the possibility that the effect size in Asians did not reflect the true AMD susceptibility because of errors in the sample introduced by wrongly diagnosed cases of pachychoroid neovasculopathy, we evaluated genetic risk score using the effect sizes in both Caucasian and Asian populations. The effect sizes applied in this analysis are summarized in Supplementary Table 2.

Survival analysis.
To evaluate the clinical importance of pachychoroid neovasculopathy, we selected study participants who had received a loading dose of ranibizumab (3 monthly injections, 0.5 mg), received treatment as needed afterwards, and been followed up for more than 12 months. After excluding the patients who had had previous ocular treatment (other than cataract surgery), and those with a visual acuity less than 20/200, we retrospectively reviewed medical charts. The durations from their third ranibizumab injection to event (the administration of additional treatment) or censoring (final visit by June 2014, or loss to follow up) was reported on a daily basis. Statistical analysis. Every 2 × 2 table was evaluated using Fisher's exact test. Continuous variables were compared using the unpaired t-test. Genotype distribution was compared using the chi-square test for trends. Survival analysis on the retreatment-free period was conducted using the Kaplan-Meier method. Differences between curves were evaluated using the log-rank test. A p-value < 0.05 was considered statistically significant. These statistical analyses were conducted using R software ver.3.02 (http:// www.r-project.org/). For the survival analysis, the CRAN package "survival" was employed.