Age-related macular degeneration (AMD) is the leading cause of legal blindness in patients aged 65 years and older. It is estimated that 196 million people will be affected by AMD by 2020, and this number will increase to 288 million by 2040 [1]. Intermediate AMD is defined by the presence of large drusen > 125 μm and/or pigmentary abnormalities [2].

Another hallmark feature is pigmentary epithelial detachment (PED). Optical coherence tomography (OCT) can be used to identify various aspects of PEDs owing to the presence of drusen, sub-retinal pigment epithelium (RPE) fluid, and fibrovascular tissue in AMD [3]. Roughly, PEDs are divided into 3 types including a drusenoid PED, a serous PED, and a shallow irregular PED in intermediate AMD.

In contrast, late AMD, including exudative AMD, is distinguished from intermediate AMD by the presence of choroidal neovascularization (CNV). Traditionally, exudative AMD and neovascular AMD were interchangeable since they had identical characteristics on angiography and OCT [4,5]. However, non-exudative neovascularization has recently been described in patients with dry AMD [5,6]. In 2013, Querques et al. [6] first described subclinical, non-exudative CNV using multimodal imaging. The diagnosis requires the presence of a moderately reflective material between an elevated RPE and Bruch’s membrane on OCT, the absence of fluid on OCT, staining with fluorescein angiography (FA), and a plaque identified with indocyanine green angiography (ICGA). This study confirmed the existence of non-exudative neovascular AMD and showed that exudation was not required in the presence of CNV [6,7].

Recently, with the advent of optical coherence tomography angiography (OCTA), several studies have demonstrated the presence of non-exudative CNV as a blood flow signal under the RPE using OCTA in eyes with dry AMD [5,8-11]. OCTA, which can detect CNV by identifying moving blood flow under the RPE rather than relying on the presence of leakage on FA and ICVA or fluid on OCT, is an ideal imaging modality to further study non-exudative CNV. In 2015, Palejwala et al. [5] first described the use of OCTA to detect non-exudative CNV. In their study, 32 fellow eyes in patients with neovascular AMD were scanned using OCTA. Two cases of subclinical non-exudative CNV were identified using OCTA. In 2016, Carnevali et al. [10] detected CNV in 18/22 study eyes using OCTA, without false-positive results. This group concluded that OCTA reliably detected CNV with a sensitivity of 81.8% and specificity of 100%. Thus, we believe that OCTA may effectively detect subclinical non-exudative CNV in eyes with intermediate AMD.

Moreover, there are limitations to performing FA and ICGA tests to identify subclinical non-exudative CNV in real-world clinics in patients not suspected of having exudative AMD. FA and ICGA methods are expensive, time-consuming, uncomfortable for patients, and pose a risk of allergic reactions. Because of these limitations, angiographic monitoring of eyes with intermediate AMD has not become routine [9]. For these reasons, the exact incidence of subclinical, non-exudative CNV is unclear. However, it is clinically important to understand the frequency of non-exudative CNV because one-quarter of such eyes develop exudative AMD over 4 years [12-14].

In this study, we aimed to investigate the frequency of subclinical non-exudative CNV in eyes with intermediate AMD using OCTA. To our knowledge, this is the first report in Korea identifying non-exudative CNV in patients with intermediate AMD.

We retrospectively reviewed the medical records of 97 patients (140 eyes) with intermediate AMD who visited the retina center of Hangil Eye Hospital between January 1, 2023, and December 31, 2023. The study was conducted in accordance with the tenets of the Declaration of Helsinki and approved by the Institutional Review Board of Hangil Eye Hospital (IRB approval number: 24002). Informed consent was waived because of the retrospective nature of the study and the analysis used anonymous clinical data.

The definition of intermediate AMD in this study follows the Ferris classification [15], which suggests that patients with large drusen and/or pigmentary abnormalities associated with at least medium drusen should be considered to have intermediate AMD. Additionally, we only enrolled patients in whom neither subretinal fluid nor intraretinal fluid was observed on OCT examination.

All patients had a standardized history and clinical examination, including measurement of best corrected visual acuity (BCVA), slit-lamp biomicroscopy, and fundus examination. Additionally, they underwent Spectral-domain OCT (Spectralis; Heidelberg Engineering) and Spectral-domain OCTA (Spectralis; Heidelberg Engineering).

The inclusion criteria were as follows: age > 50 years; ophthalmoscopic signs of intermediate AMD; PEDs on OCT; and diagnosis of non-exudative CNV, defined as a CNV showing no intraretinal or subretinal exudation on OCT but detectable on OCTA.

The exclusion criteria were any prior treatment in either eye, such as laser photocoagulation, photodynamic therapy, intravitreal injections of steroid or anti-vascular endothelial growth factor, or vitrectomy; and non-exudative CNV due to causes other than AMD, such as pathologic myopia, CSC, and inherited macular dystrophy.

We identified PEDs on OCT B-scans and divided them into 4 types: drusenoid, serous, shallow irregular, and combined. A drusenoid PED is defined as a dome-shaped RPE elevation overlying deposits with homogeneous internal reflectivity. A serous PED is defined as a dome-shaped elevation of the RPE overlying a homogeneously hyporeflective space. A shallow irregular PED manifests as flat, irregular RPE elevations, referred to as the double-layer sign on OCT. Combined PEDs are defined as a mixed appearance of these PEDs [3].

When the PED was identified, the OCTA image was analyzed to evaluate the presence of CNV. Spectral-domain (SD)-OCTA was performed based on full-spectrum amplitude decorrelation angiography using a probabilistic approach to generate high-contrast, almost binarized OCTA images. This instrument had an axial resolution of approximately 3 μm and a transverse resolution of approximately 6 μm. OCTA was performed in all patients, with a scanning area of 3 × 3 mm centered on the foveal area. The process of removing segmentation and projection artifacts involved two steps. First, segmentation and projection artifacts were automatically removed using a program provided by the OCTA manufacturer. Second, all OCT and OCTA images were reviewed by 2 trained retinal specialists (H.Y.C and A.R.C) who were blinded to each other’s imaging findings, and the remaining artifacts were manually corrected.

Using the OCTA software, the retinal layers and generated en-face images of the superficial vascular complex, deep vascular complex, avascular retina, and choriocapillaris were automatically segmented. Next, the en-face images of the avascular retina were analyzed to evaluate the presence of CNV in PEDs.

In the cross-sectional OCTA, we analyzed the image that included a flow signal in false color over a structural OCT B-scan in standard gray color. Identification of the flow signal in the PED provides a high level of confidence for the presence of CNV. It was possible to differentiate projection artifacts from true flow.

Statistical analyses were performed using IBM SPSS software version 21 (IBM Corp.). Statistical significance was set at p < 0.05.

We included 140 eyes from 97 patients with unilateral or bilateral intermediate AMD. The mean age was 68.9 ± 10.4 years (range, 54–88 years). We identified 17 eyes (13.8%) in 16 patients with subclinical non-exudative CNV among the eyes with intermediate AMD.

Among the 17 eyes with non-exudative CNV, 10 (9 patients) were in males (p = 0.51). One male patient had bilateral non-exudative CNV.

There was no significant age difference between patients with non-exudative CNV and those without CNV (with CNV: 69.1 ± 12.9 years; without CNV: 68.9 ± 9.6 years; p = 0.94).

The mean BCVA of the eyes with non-exudative CNV was 0.68 ± 0.24 (0.15–1.0), and that of the non-affected eyes was 0.77 ± 0.21 (0.2–1.0). There was no significant difference in BCVA between the eyes with non-exudative CNV and those without CNV (p = 0.20).

To evaluate choroidal features between eyes with non-exudative CNV and those without CNV, subfoveal choroidal thickness was measured. There was no significant difference in choroidal thickness between eyes with non-exudative CNV and those without CNV (with CNV: 268 ± 107 μm; without CNV: 234 ± 126 μm; p = 0.16).

The typical fundoscopic findings in patients with intermediate AMD are macular drusen and pigmentary abnormalities that appear on OCT as RPE increase with no evidence of intraretinal or subretinal fluid.

The PEDs were classified into 4 types (drusenoid, shallow irregular, serous, and combined) using OCT. Drusenoid PEDs accounted for 47.1% (66 eyes), shallow irregular PEDs for 20.0% (28 eyes), serous PEDs for 18.6% (26 eyes), and combined PEDs for 14.3% (20 eyes). Non-exudative CNV was only observed in eyes with shallow irregular PEDs (8 eyes, 28.6%) and combined PEDs (9 eyes, 45.0%). There were 16 shallow irregular PEDs among the combined PEDs, and the frequency of non-exudative CNV in shallow irregular PEDs was 38.6% (17 eyes). Non-exudative CNV was not detected in eyes with drusenoid or serous PEDs (Fig. 1).

Fig. 1. Frequencies of pigment epithelial detachment (PED) subtypes and of non-exudative choroidal neovascularization (CNV) according to PED subtype in 140 eyes from 97 patients.

Among the 44 eyes with shallow irregular PEDs, CNV was observed in 19 on en-face OCTA and in 17 on cross-sectional OCTA. There were 4 discrepant cases in which the results of the en-face OCTA did not match those of the cross-sectional OCTA. Among them, 3 exhibited CNV on en-face OCTA but not on cross-sectional OCTA. However, in the remaining case, cross-sectional OCTA detected CNV, while en-face OCTA did not (Fig. 2).

Fig. 2. Frequency of subclinical non-exudative choroidal neovascularization based on en-face and cross-sectional optical coherence tomography angiography.

Figs. 3–5 show several cases of correspondence and discrepancy. Fig. 3 shows OCTA images of the 3 corresponding cases in the results of en-face and cross-sectional OCTA images. Both en-face OCTA and cross-sectional OCTA detected non-exudative CNV in shallow irregular PEDs. Fig. 4 shows 3 cases of non-exudative CNV on en-face OCTA images (Fig. 4A, C, and E); however, CNV could not be detected by cross-sectional OCTA (Fig. 4B, D, and F). All 3 cases in Fig. 4 were classified into the group without CNV based on manual artifact removal analysis because they were considered projection artifacts missed by the removal algorithms in the OCTA device program. Fig. 5 shows the opposite situation, in which a clear flow signal was observed in the cross-sectional OCTA images but no CNV lesion was observed in the en-face OCTA images. In this case, the CNV of very small size was likely misidentified as a projection artifact on en-face OCTA and was removed. Thus, the case in Figure 5 was classified into the group with CNV in the analysis. Overall, among the 140 eyes, 17 cases of CNV in shallow irregular PED were identified.

Fig. 3. Three corresponding cases in the results of the en-face and cross-sectional OCTA images. (A, D, G) Optical coherence tomography (OCT) B-scan images show shallow irregular pigment epithelial detachments (PEDs). (B, E, H) An en-face optical coherence tomography angiography (OCTA) image shows choroidal neovascularization (CNV) (arrows). (C, F, I) In the same lesions, a cross-sectional OCTA image shows the flow signal (arrow heads) in shallow irregular PEDs. Both en-face OCTA and cross-sectional OCTA detected non-exudative CNV in shallow irregular PEDs.

Fig. 4. Three cases of discrepancy between the en-face image and the cross-sectional image. Non-exudative choroidal neovascularization (CNV) in shown on en-face optical coherence tomography angiography (OCTA) (arrows in A, C, and E). However, there is no flow signal detected in the cross-sectional OCTA image (B, D, and F). All three cases are classified into the group without CNV based on manual artifact removal analysis because they are considered projection artifacts missed by the removal algorithms in the OCTA device program.

Fig. 5. A case of flow signal masking during calibration of projection artifacts. (A) Fundus photography shows pigmentary change on the fovea. (B) Optical coherence tomography shows a focal pigment epithelial detachment and hyper-reflective material in the outer retina. The enface optical coherence tomography angiography (OCTA) image does not show any choroidal neovascularization (CNV) (C), but a clear flow signal can be seen in the cross-sectional OCTA image (D, arrowhead). In this case, the CNV size appears to be very small; it is likely that the automatic artifact removal algorithm mistook it for a projection artifact in en-face OCTA and excluded the flow signal, resulting in classification into the group with CNV in the analysis.

In the present study, the frequency of non-exudative CNV in patients with intermediate AMD was 13.8% (17 eyes), well in the previously reported range from 6 to 27% [9,16-20].

Our findings suggest that non-exudative CNV may be heterogeneous in clinical presentation and ocular history. The presence of non-exudative CNV must also be suspected in eyes with low- and intermediate-risk of exudative AMD and not only in the fellow eyes of patients with exudative AMD, as reported by Roisman et al. [9]. Indeed, the incidence rate of non-exudative CNV in patients with intermediate AMD has been reported to be 6–27%, which is not low [9,16-20].

In 1976, Sarks [16] showed that CNV could exist in eyes with intermediate AMD by performing histopathological analyses of postmortem eyes. In 1997, Schneider et al. [21] used ICGA to detect subclinical non-exudative CNV in situ in patients with dry AMD. Although ICGA can identify these subclinical lesions, its potential usefulness has been ignored over the years because of the cost and risk of discomfort associated with its use as a screening tool in dry AMD. However, recent developments in OCTA have greatly facilitated the detection of subclinical, non-exudative CNV [5]. Currently, because several studies have reported the incidence of non-exudative CNV using OCTA, researchers have proposed that eyes with intermediate AMD require a new classification system that distinguishes between neovascular and non-neovascular intermediate AMD [5,9].

It is important to detect CNV in intermediate AMD as subclinical non-exudative CNV has been shown to be a harbinger of exudative disease [18]. Yanagi et al. [22] reported that non-exudative CNV carries a significant risk of exudation (estimated annual incidence, 18.1%). Eyes with non-exudative CNV have an almost 10 times higher risk of developing exudative changes than those without these lesions (estimated annual incidence, 2.0%). De Oliveira Dias et al. [19] found a cumulative incidence of exudation of 24% within 1 year in patients with non-exudative macular neovascularization; after 2 years of follow-up, the cumulative incidence of exudation was 34.5%. Yanagi et al. [22] reported an exudation incidence of 22.2% by 6 months. All authors recommend close follow-up without treatment to detect early signs of exudation [18-20,22].

Thus, our findings suggest that 13.8% of patients with intermediate AMD may have a high risk of exudative AMD, and close monitoring is needed. Indeed, OCTA may be necessary in patients with intermediate AMD to identify the group at a higher risk of exudative AMD.

We defined intermediate AMD according to the Ferris classification without angiographic evaluation [15]. It would have been better to perform angiographic tests to define intermediate AMD precisely. However, conducting FA and ICGA tests in real-world clinics in patients not suspected of having exudative AMD is associated with difficulties. In addition, some previous studies defined intermediate AMD using only fundus and OCT images without angiographic evaluations and then confirmed subclinical CNV through OCTA, similar to our study [23,24]. Shi et al. [23] also pointed out that performing ICGA in patients with non-exudative intermediate AMD is riskier, more time-consuming, and more expensive.

In addition, shallow irregular PEDs and CNV were observed in eyes with intermediate AMD, bringing into question the classification as ‘intermediate.’ A prospective diagnostic accuracy cohort study conducted in the United Kingdom over 3 years mentioned that shallow irregular elevation of the RPE without subretinal fluid was not considered neovascular AMD [25]. Moreover, Roisman et al. [9] reported irregular PEDs and CNV in eyes diagnosed with intermediate AMD after FA and ICGA. Accordingly, Roisman et al. [9] proposed a new terminology called non-exudative, neovascular intermediate AMD. Therefore, even when irregular PEDs and subclinical CNV are detected on OCTA, intermediate AMD may be diagnosed.

In this study, non-exudative CNV was only observed in eyes with shallow irregular PEDs, at a frequency of 38.6% (17/44 eyes). The double-layer sign is a typical feature of shallow irregular PED and represents flat, irregular elevation of the RPE. This OCT configuration is likely indicative of non-exudative CNV rather than simply coalescent drusen. The specificity and sensitivity of the double-layer sign for identifying non-exudative CNV are as high as 88% [23]. Dansingani et al. [8] reported a 95% prevalence of non-exudative CNV in patients with a pachychoroid spectrum and shallow irregular PEDs on OCTA. In this study, non-exudative CNV was only detected in shallow irregular PED, at 38.8%, when using OCTA. This frequency is lower than that reported in previous studies [8,9,23].

This result may be because we used SD-OCTA to detect non-exudative CNV. Previous reports have compared the performance of SD-OCTA and swept source-OCTA (SS-OCTA) for detecting CNV and concluded that SS-OCTA yielded larger CNV areas than did SD-OCTA. Additionally, SS-OCTA had the ability to produce higher-quality angiograms than those by SD-OCTA. Those studies suggested that SS-OCTA may provide a more accurate representation of CNV than SD-OCTA [26,27]. Thus, we might have missed some cases of non-exudative CNV due to the relatively low image quality of SD-OCTA.

The sensitivity of OCTA to detect CNV using FA and ICGA as the “gold standard” ranged from 50–100%. Yanagi et al. [11] reported an OCTA sensitivity of 50% compared with the 71% sensitivity of ICGA for detection of non-exudative CNV. OCTA identified flow signals in only 3 of 10 eyes with ICGA features suggestive of non-exudative CNV. Some CNV, especially polypoidal choroidal vasculopathy, is thought to originate from the inner choroid and penetrate Bruch’s membrane. Therefore, it is possible that CNV visualized using ICGA but not OCTA represents abnormal blood vessels located in the inner choroid that are difficult to visualize using commercially available OCTA instruments. In the present study, non-exudative CNV might have been more frequently detected if we had used ICGA.

Finally, distinguishing between shallow irregular PED and drusenoid PED derived from adjacent confluent soft drusen can be difficult because both appear as collections of moderately reflective material in the sub-RPE space on OCT [10]. Thus, some of the cases that we identified as shallow irregular PEDs actually could be drusenoid PEDs.

For these reasons, we detected a relatively low frequency of non-exudative CNV in shallow irregular PED, such as low image quality of SD-OCTA, wide range of sensitivity of OCTA to detect CNV, and possibility of misclassifying drusenoid PED as shallow irregular PED. Further evaluation using SS-OCTA, FA, and ICGA is required to confirm the presence of non-exudative CNV in shallow irregular PED.

In this study, non-exudative CNV was identified in PEDs using both en-face and cross-sectional OCTA. The CNV in shallow irregular PEDs was observed in 19 eyes using enface OCTA and in 17 using cross-sectional OCTA. Four patients exhibited contradictory results. Among them, 3 cases were presumed to be projection artifacts because there was no flow signal on cross-sectional OCTA. While OCTA is a powerful tool, it can include confounding projection artifacts that result in false-positive CNV. Such artifacts can be created with a light beam that encounters the superficial retinal plexus passes through the moving blood cells, projecting the scanned vessel during reconstruction of the deeper retinal plexus [28-32]. Zheng et al. [32] suggested that projection artifacts may be interpreted as vascularized drusen. Our results also suggest that projection artifacts may limit accurate evaluation. Thus, both en-face OCTA and cross-sectional OCTA were performed to compensate for these errors [28].

There was no significant difference in BCVA between eyes with non-exudative CNV and those without CNV (p = 0.20). There is a growing concern that subclinical, non-exudative CNV can be treated with anti-vascular endothelial growth factor therapy [9]. While no controlled clinical trials have been performed to provide definitive recommendations, the reviewed studies recommend that non-exudative CNV be closely monitored for evidence of exudative disease [18-20,22]. Non-exudative CNV may persist for 2 years or longer without exudation or affecting vision [10]. Therefore, we plan to monitor patients with intermediate AMD and non-exudative CNV more frequently than those without CNV.

The limitations of this study include its retrospective design and the relatively small number of patients from a single institution. In addition, this study did not perform FA or ICGA and could not confirm the presence of non-exudative CNV. Moreover, longitudinal studies are needed to confirm changes in CNV lesions. However, this study is valuable in that it is the first to analyze subclinical non-exudative CNV in intermediate AMD using both en-face and cross-sectional OCTA in Korea.

In conclusion, our study demonstrated that OCTA could identify subclinical non-exudative CNV in up to 13.8% of eyes with intermediate AMD. Non-exudative CNV was observed only in eyes with shallow irregular PEDs. Despite some artifacts, OCTA may allow clinicians to noninvasively identify non-exudative CNV and may be considered a useful tool to guide the frequency of return visits and make treatment decisions. We recommend the use of OCTA to evaluate patients with intermediate AMD, especially in eyes with shallow irregular PEDs.

The authors declare no conflicts of interest relevant to this article.

Conception (H.C.); Design (H.C.); Data acquisition (H.C., A.C.); Analysis (H.C., A.C.); interpretation (H.C., A.C., G.H.); writing (H.C., G.H.); review (H.C., G.H., J.S.); Final approval of the article (All authors)