Hyperreflective Choroidal Foci Characterization Supports the Existence of Different Clinical Phenotypes of Geographic Atrophy

Abstract

Purpose

To characterize and quantify optical coherence tomography (OCT) hyperreflective choroidal foci (HCF) in geographic atrophy (GA) of age-related macular degeneration (AMD) and to assess their distribution across different GA phenotypes.

Methods

Eyes with bilateral GA (B-GA) or unilateral GA (U-GA) with macular neovascularization in the fellow eye, as well as controls, were included. HCF were classified in different subtypes according to size (small, ≤30 µm; large, >30 µm) and reflectivity (high or intermediate). Each subtype was also identified according to location (choriocapillaris vs. choroid). HCF counts were quantified within complete retinal pigment epithelium and outer retinal atrophy (cRORA), incomplete retinal pigment epithelium and outer retinal atrophy (iRORA), and non-atrophic outer retinal alteration (noRORA). The comparison of HCF count between the GA groups and the controls was performed only in noRORA segments.

Results

Seventy-seven eyes were enrolled (40 B-GA, 18 U-GA, and 19 controls). The extent of cRORA, iRORA, and noRORA regions did not differ between the two GA groups (P > 0.05 for all). In noRORA regions, both GA groups exhibited significantly higher HCF number compared to controls (P < 0.0001 for both). Total HCF count did not differ between the GA groups across all RORA regions (P > 0.05 for all). However, the analysis of HCF subtypes revealed significant differences. Small intermediate-reflectivity choriocapillaris HCF were significantly more numerous in U-GA eyes than in B-GA in cRORA and noRORA regions (P = 0.0006 and P = 0.006, respectively). Large high-reflectivity choriocapillaris HCF were significantly more numerous in B-GA eyes than in U-GA within cRORA regions (P = 0.025). In choroid, HCF subtypes did not differ between the two GA groups in the different RORA regions.

Conclusions

Although overall HCF burden was similar between the two GA phenotypes, significant distinct distribution of HCF subtypes was documented. Detailed HCF characterization provides new insights into GA heterogeneity, supporting the value of novel biomarkers in further characterizing different GA phenotypes.

Geographic atrophy (GA), the atrophic late-stage form of age-related macular degeneration (AMD), is one of the leading causes of blindness in individuals older than 50 years in developed countries.¹ GA prevalence increases exponentially with age, making it an escalating public health concern.² Optical coherence tomography (OCT) has become an essential retinal imaging modality for the evaluation of GA. Advanced OCT technologies, such as spectral-domain OCT (SD-OCT) and swept-source OCT, enable layered visualization of the retina and choroid, providing a precise analysis of tissue loss and changes in retinal and choroidal layers related to GA. To provide a more precise framework for studying GA, an international panel of experts, the Classification of Atrophy Meeting (CAM) Group, developed a consensus classification system aimed at standardizing terminology and anatomical endpoints.³ The CAM Group defines GA as a progressive condition evolving from an incomplete retinal pigment epithelium (RPE) and outer retinal atrophy (iRORA) to a complete one (cRORA) as detected by OCT.^3,4 OCT facilitates the identification of retinal and choroidal structural features that can enhance the ability to monitor GA progression, and, by introducing earlier and more robust anatomical endpoints, it can improve the efficiency and precision of clinical trials. Among the new OCT features, hyperreflective retinal foci (HRF) and hyperreflective choroidal foci (HCF) have been identified.^5,6 Although the characterization of the HRF inside the retina is better defined and understood,⁵ the significance of HCF within the choriocapillaris and choroid has received comparatively less attention and is still under investigation. HCF have been described in different conditions such as diabetic retinopathy,⁶ Stargardt disease,⁷ and central serous chorioretinopathy,⁸ and even in healthy subjects.⁹ Several hypotheses have been proposed to explain their origin, including migration of neuroglia or RPE cells, activated macrophages, or accumulation of melanolipofuscin deposits.^7,10–13 However, the precise biological implications of these findings remain largely unclear. In GA, HCF have been recently described and analyzed in vivo using B-scan OCT. They were specifically distinguished within the choriocapillaris and Sattler's and Haller's layers, and their number across the OCT volume was correlated with the extent of the atrophic area.¹⁴ However, in the absence of histological confirmation, their significance in this condition remains poorly understood. HCF exhibit significant variability in their structural characteristics (size and reflectivity) and spatial distribution, which is likely indicative of distinct underlying pathophysiological processes. The present study aims to systematically categorize and analyze HCF in GA eyes. Specifically, we classify HCF based on their size, reflectivity, and spatial distribution (choriocapillaris vs. choroid), in relation to regions of atrophic areas (cRORA and iRORA) and non-atrophic areas (noRORA), as well as across different GA phenotypes, bilateral GA (B-GA) and unilateral GA (U-GA) with macular neovascularization (MNV) in the fellow eye.^13,15

Methods

Population

This retrospective observational study included subjects with GA secondary to AMD and healthy controls followed at our institution between January 2022 and December 2024 with available retinal and choroidal multimodal imaging. This study adhered to the tenets of the Declaration of Helsinki, and written informed consent approved by the local ethics committee was obtained from all participants. Inclusion criteria were subjects more than 50 years of age with unifocal or multifocal GA in at least one eye, with a total GA lesion area in the study eye > 1.25 mm² as assessed with blue light fundus autofluorescence (FAF).¹⁶ Patients were divided into two groups: GA in both eyes (B-GA), and GA in one eye and MNV in the fellow-eye (U-GA) as confirmed by multimodal imaging, including fluorescein and indocyanine green angiography and/or OCT angiography. We also considered a group of healthy subjects (age-matched to the affected population), who had no history of relevant eye diseases. Exclusion criteria included diabetes mellitus, evidence or history of MNV in the study eye (confirmed by OCT angiography or fluorescein and indocyanine green angiography), media opacities affecting imaging quality, glaucoma, hereditary retinal dystrophies, epiretinal membranes or macular pucker, previous or concomitant inflammatory or retinal vascular diseases, anterior segment abnormalities, refractive errors > 6 diopters (D), and previous macular laser treatment and intraocular surgery within the last 6 months. All subjects underwent a complete ophthalmologic evaluation, including best-corrected visual acuity (BCVA) quantified using the Early Treatment Diabetic Retinopathy Study (ETDRS) protocol, anterior segment examination, intraocular pressure measurement, and biomicroscopic examination under mydriasis using a 90-D lens.

Imaging and OCT Analysis

All subjects underwent multimodal retinal and choroidal imaging, including near-infrared (NIR) and FAF images and OCT, which was performed using a SPECTRALIS HRA+OCT (Heidelberg Engineering, Heidelberg, Germany), under mydriatic condition after pupil dilation with 1% tropicamide, in a dark room and with the in-built eye tracker always activated. NIR and FAF images were obtained by a Heidelberg confocal scanning laser ophthalmoscope (cSLO). The cSLO device uses an optically pumped solid-state laser source to generate a blue-light excitation wavelength of 488 nm for FAF imaging. Based on the analysis of the FAF images, the following GA features were obtained: total area of GA (in mm²), semi-automatically measured by means of RegionFinder 2.6.6.0 within the Heidelberg platform, as previously described,¹⁷ and the distinction in unifocal or multifocal GA according to the presence of a single or multiple circular areas of reduced FAF signal.¹⁸ The FAF pattern in the junctional zone surrounding the atrophic area was classified as focal, banded, patchy, or diffuse.¹⁶ The presence of foveal sparing¹⁹ was identified in the NIR images when the atrophic lesions did not include the foveal center and confirmed with subfoveal preservation of the outer retinal layers on structural SD-OCT. The presence of subretinal drusenoid deposits (SDDs) was evaluated in each eye on structural SD-OCT scans, as accumulations of granular hyperreflective material located in the subretinal space above the RPE, and was confirmed on NIR images. The structural OCT imaging protocol included a 6 × 6-mm volumetric map composed of 49 horizontal B-scans, with the central scan passing through the foveola. Scans were acquired in high-speed (HS) mode with ≥12 automatic real-time tracking (ART) and a quality index > 28. All OCT examinations were performed late in the morning to avoid possible choroidal diurnal variations. Central subfield thickness (CST) was automatically calculated in the central 1 mm of the ETDRS grid by the device. Subfoveal choroidal thickness (CT) was measured manually by a single physician (FP) in the single horizontal OCT line scan acquired in high-resolution (HR) mode passing through the foveola at 180°, 9-mm length, with ART set at 100 frames. Therefore, the number of HCF was calculated in the central 3 mm of this single horizontal HR OCT linear scan. As previously proposed,¹⁴ the identification and quantification of HCF were performed on high-magnification, white-over-black OCT scans to enhance HCF visualization. Although HCF have previously been defined as round hyperreflective dots with a diameter between 10 and 50 µm in size,⁷ in this study we classified four different types of HCF according to size and reflectivity. HCF were identified and categorized using the same parameters already used to classify HRF^12,20: (1) dimension, classified as small (≤30 µm) or large (>30 µm) in size, and (2) reflectivity, based on OCT signal intensity and categorized as high reflective (similar to that of RPE layer) or intermediate reflective (similar to that of the nerve fiber layer). The HCF subtypes were also identified according to their presence in the choriocapillaris or in the deeper choroidal layers (named choroid). HCF were assigned to the choriocapillaris compartment when located within a slab extending from 4 to 20 µm beneath the outer boundary of Bruch's membrane, corresponding to the thin hyperreflective layer immediately deep to the RPE–Bruch's membrane complex and superficial to Sattler's and Haller's layers; foci located deeper than this band were classified within the choroidal compartment.^21,22 To provide a more detailed characterization of HCF, we analyzed their distribution relative to regions of atrophy as defined by the CAM consensus,^3,4 segmenting linear B-scans into regions of cRORA, iRORA, and noRORA. As shown in Figure 1, for each of these regions, HCF were quantified based on their location (choriocapillaris vs. choroid), size (≤30 µm vs. >30 µm), and reflectivity (high vs. intermediate). Two independent graders (GM and FP) identified and classified HCF in each layer (choriocapillaris or choroid). HCF were counted separately within each region, and the linear extension (LE; in µm) of each region was recorded. Because cRORA, iRORA, and noRORA regions differ in extension both within and between eyes, regional lesion length was not used to derive a raw density ratio but was entered as a continuous covariate in the statistical models so HCF burden could be compared across regions and groups independently of regional extension. This model-based approach also mitigates the limitations posed by varying choroidal reflectivity between atrophic and non-atrophic areas, a challenge previously highlighted.¹⁴ NIR and FAF images were used to verify the areas with cRORA, as previously described.³ All data were collected and analyzed for each of the three study groups: B-GA, U-GA, and healthy controls.

Figure 1.

(a) A horizontal OCT scan acquired in high-resolution (HR) mode passing through the foveola of an eye affected by GA. (b) A cRORA region is highlighted in red, iRORA region in yellow, and noRORA region in green. (c) Small HCF with intermediate reflectivity are included within a red triangle, large HCF with intermediate reflectivity within a purple circle, small-HCF with high reflectivity within a rhombus, and large HCF with high reflectivity within a green square. All HCF have also been magnified for clarity.

Statistical Analysis

Descriptive statistics were used to summarize demographic and baseline characteristics. For quantitative variables, a repeated-measures (linear mixed-effects) model was used, with patient included as a random effect, as both eyes were evaluated in patients with bilateral GA, to account for within-subject inter-eye correlation. Quantitative variables, such as CST, CT, visual acuity, GA area, the number of total HCF in the whole choroid, and the number of the subtypes of HCF in the choriocapillaris and choroid, were summarized using the mean and standard deviation (SD). Comparisons among the study groups (B-GA, U-GA, and controls) were performed using analysis of variance (ANOVA) or Kruskal–Wallis tests for continuous variables, and χ² tests for categorical variables. Post hoc pairwise comparisons were conducted using the Tukey–Kramer test when applicable. HCF counts within each region were analyzed with a generalized linear mixed model, with group as a fixed effect, the LE of the corresponding region as a continuous covariate, and patient as a random effect to account for the inclusion of both eyes in patients with B-GA. Only regions with measurable LE were included; observations in which the corresponding region was absent were excluded from the model for that region, as region-specific quantification is not defined in the absence of the region itself and would otherwise introduce structural zeros. This approach allowed comparisons between groups independently from lesion size. Estimates are reported as least squares (LS) means ± SE at the mean regional LE. To evaluate grading reliability, intragrader and intergrader agreement was assessed using intraclass correlation coefficients (ICCs) and Bland–Altman plots. Intragrader agreement was determined from two separate evaluations by each grader, and intergrader agreement was based on the first evaluation by each grader. Pairwise comparisons of LS means within each region-specific model were adjusted using the Bonferroni correction for the number of comparisons performed in that model; significance was set at P < 0.05. All analyses were conducted using SAS 9.4 (SAS Institute, Cary, NC, USA).

Results

Population and Geographic Atrophy Features

Seventy-seven eyes of 61 Caucasian subjects were enrolled in this study, including 40 eyes with B-GA (24 subjects), 18 eyes with U-GA (18 subjects), and 19 eyes of age-matched healthy controls (19 subjects). Gender distribution significantly differed among groups (P = 0.011), with B-GA and U-GA patients being predominantly female (92% and 78%, respectively), whereas controls were almost balanced for sex (47% males and 53% females). No significant difference in sex distribution was observed between B-GA and U-GA (P = 0.375). Age distribution did not differ among groups (P = 0.062) and was similar between B-GA and U-GA. Demographic data are reported in Table 1.

Table 1.

Study Population Demographics

Parameter	B-GA (n = 24)	U-GA (n = 18)	Controls (n = 19)	P
Sex (males; females), n (%)	2 (8); 22 (92)	4 (22); 14 (78)	9 (47); 10 (53)	0.375*
Age (y), mean ± SD (range)	80.4 ± 8.4 (66–93)	83.3 ± 6.1 (72–92)	81.1 ± 7.3 (65–86)	0.062*

^*B-GA group versus U-GA.

Mean BCVA was higher in controls than in GA eyes (P = 0.002 in controls vs. B-GA; P = 0.008 in controls vs. U-GA), but it was not significantly different between the two GA groups (P = 0.871). All subjects met the prespecified image-quality criteria and were considered gradeable; no eyes were excluded because of insufficient OCT image quality. GA clinical features, including mean GA area, foveal sparing, presence of SDDs, multifocal versus unifocal GA distribution, FAF pattern, and mean cRORA, iRORA, and noRORA extension did not significantly differ between U-GA and B-GA eyes (P > 0.05 for all). Data are summarized in Table 2.

Table 2.

BCVA and GA Clinical Features

Parameter	B-GA (n = 40 Eyes)	U-GA (n =18 Eyes)	Controls (n = 19 Eyes)	P
BCVA (ETDRS letters), mean ± SD (range)	61.5 ± 23.5 (1–85)	64.4 ± 22.1 (3–85)	85.0 ± 0.0 (85–85)	0.871*0.002†
				0.008 ‡
GA area in FAF (mm2), mean ± SD (range)	10.4 ± 8.1 (1.25–32.9)	8.2 ± 9.3 (1.25–30.0)	NA	0.385*
FAF pattern, n (%)
Banded	14 (35)	3 (17)	NA	0.111*
Diffuse	16 (40)	5 (28)
Focal	10 (25)	8 (44)
Patchy	0 (0)	2 (11)
Multifocal; unifocal, n (%)	18 (45); 22 (55)	10 (56); 8 (44)	NA	0.572*
Foveal sparing (yes; no), n (%)	20 (50); 20 (50)	12 (67); 6 (33)	NA	0.390*
SDDs (yes; no), n (%)	28 (70); 12 (30)	8 (44); 10 (56)	NA	0.147*
cRORA extension (µm), mean ± SD (range)	1390.2 ± 977.74 (250.0–3000.0)	1089.72 ± 726.88 (250.0–2440.0)	NA	0.378*
iRORA extension (µm), mean ± SD (range)	73.63 ± 138.92 (0.0–248.0)	60.33 ± 90.95 (0.0–249.0)	NA	0.904*
noRORA extension (µm), mean ± SD (range)	1536.18 ± 963.2 (0.0–3000.0)	1849.94 ± 732.51 (560.0–3000.0)	3000.0 ± 0 (3000.0–3000.0)	0.339*

NA, not applicable.

Significant P values are indicated in bold.

^*B-GA versus U-GA.

^†Controls versus B-GA.

^‡Controls versus U-GA.

Additionally, the number of eyes available for each region-specific analysis varied according to the presence of the corresponding lesion along the analyzed B-scan. Specifically, the cRORA analysis included 36 B-GA and 16 U-GA eyes, the iRORA analysis included 10 B-GA and six U-GA eyes, and the noRORA analysis included 36 B-GA and 18 U-GA eyes. All 19 control eyes were included in the noRORA analysis.

OCT Features and Quantitative Analysis of Total HCF

Mean CT measured 174.7 ± 100.1 µm in the B-GA group, 151.0 ± 57.1 µm in the U-GA group and 263.9 ± 85.2 µm in controls. CT was significantly thinner in the B-GA and U-GA groups compared to controls (P = 0.001 and P = 0.0006, respectively), and it was thinner in the U-GA group versus the B-GA group (P = 0.037). CST was reduced in GA eyes compared to controls. Mean CST measured 243.7 ± 45.4 µm in the B-GA group, 250.3 ± 37.0 µm in the U-GA group, and 280.2 ± 26.7 µm in controls. The reduction reached statistical significance in B-GA eyes versus controls (P = 0.004). No significant difference in CST was observed between the two GA groups ( P = 0.830).

The comparisons of HCF count between the GA groups and controls were performed just in noRORA segments because cRORA and iRORA segments were absent in healthy subjects. In noRORA, both GA groups showed significantly higher HCF count compared to controls, with mean values of 42.9 ± 2.6 in B-GA eyes, 48.6 ± 3.5 in U-GA eyes, and 5.6 ± 4.0 in controls (P < 0.0001), without significant differences between B-GA and U-GA eyes (P = 0.370). In cRORA, mean total HCF count was 64.7 ± 5.2 in B-GA eyes and 75.2 ± 7.6 in U-GA eyes, with no significant intergroup difference (P = 0.258). In iRORA, the mean total HCF count was 15.1 ± 2.8 in B-GA eyes and 17.3 ± 3.8 in U-GA eyes, without significant differences between the two groups (P = 0.683). Data are summarized in Table 3.

Table 3.

Adjusted Total HCF Counts According to RORA Region

	Total HCF, LS Mean ± SE (n)
RORA Region	B-GA	U-GA	Controls	P
cRORA	64.7 ± 5.2 (36 eyes)	75.2 ± 7.6 (16 eyes)	NA	0.258*
iRORA	15.1 ± 2.8 (10 eyes)	17.3 ± 3.8 (6 eyes)	NA	0.683*
noRORA	42.9 ± 2.6 (36 eyes)	48.6 ± 3.5 (18 eyes)	5.6 ± 4.0 (19 eyes)	<0.0001 † ,<0.001‡, 0.370*

Significant P values are indicated in bold. Values are adjusted for the linear extension of the corresponding region. Estimates represent the expected absolute HCF count at the mean regional extension. Only eyes with measurable extension for the corresponding region were included in each analysis.

^*B-GA versus U-GA.

^†Controls versus B-GA.

^‡Controls versus U-GA.

Quantitative Analysis of HCF Subtypes

HCF subtype counts (by size and reflectivity) in the choriocapillaris and choroid, stratified by RORA-defined regions, are reported in Table 4. Control data were analyzed versus noRORA regions only. The HCF subtype analysis detected significant differences between the two GA groups and controls. In cRORA regions, U-GA eyes exhibited a significantly higher number of small intermediate-reflectivity choriocapillaris HCF compared to B-GA eyes (18.5 ± 2.65 vs. 9.0 ± 1.88; P = 0.0006), whereas they had a lower number of large high-reflectivity choriocapillaris HCF (2.4 ± 0.53 vs. 5.1 ± 0.85; P = 0.025). In iRORA regions, no significant differences were observed between U-GA and B-GA eyes across all evaluated HCF subtypes (all P > 0.05), although a nonsignificant trend toward higher number of small intermediate-reflectivity choriocapillaris HCF was observed in U-GA eyes (P = 0.087).

Table 4.

Adjusted HCF Subtype Counts Across RORA Regions

	cRORA, HCF Subtype Count (LS Mean ± SE)			iRORA, HCF Subtype Count (LS Mean ± SE)			noRORA, HCF Subtype Count (LS Mean ± SE)
	B-GA (n = 36)	U-GA (n = 16)	P	B-GA (n = 10)	U-GA (n = 6)	P	B-GA (n = 36)	U-GA (n = 18)	Controls (n = 19)	P
Choriocapillaris
Small
H-ref	7.4 ± 1.07	7.3 ± 1.73	0.977	0.6 ± 0.38	1.4 ± 0.98	0.442	1.9 ± 0.58	1.1 ± 0.45	0.1 ± 0.46	0.113
I-ref	9.0 ± 1.88	18.5 ± 2.65	0.0006 ∗	2.2 ± 1.2	4.7 ± 0.82	0.087	16.6 ± 1.70	23.1 ± 2.31	0.3 ± 1.33	<0.001 †
										0.006 ‡
Large
H-ref	5.1 ± 0.85	2.4 ± 0.53	0.025 ∗	1.0 ± 0.35	0.2 ± 0.16	0.162	1.0 ± 0.35	0.2 ± 0.07	0.1 ± 0.11	0.052
I-ref	3.2 ± 0.85	1.8 ± 0.45	0.328	1.7 ± 0.63	0.3 ± 0.2	0.214	3.9 ± 0.65	3.8 ± 0.61	1.0 ± 0.60	0.028 †
										0.061‡
Choroid
Small
H-ref	8.4 ± 1.55	8.9 ± 2.55	0.839	1.7 ± 0.89	0.2 ± 1.2	0.315	0.5 ± 0.28	0.4 ± 0.33	0.2 ± 0.01	0.801
I-ref	20.4 ± 3.13	24.6 ± 5.45	0.453	4.5 ± 1.99	8.7 ± 1.80	0.168	11.6 ± 1.95	11.7 ± 3.06	1.8 ± 1.95	0.032 †
										0.044 ‡
Large
H-ref	5.2 ± 1.15	3.9 ± 1.08	0.454	0.3 ± 0.19	0.5 ± 0.2	0.722	0.7 ± 0.28	2.1 ± 1.56	0.1 ± 0.01	0.133
I-ref	6.0 ± 1.02	7.6 ± 1.65	0.374	3.2 ± 0.89	1.3 ± 0.61	0.235	6.5 ± 1.17	6.2 ± 1.25	2.0 ± 0.62	0.065

Significant P values are indicated in bold. Values are adjusted for the linear extension of the corresponding region. Estimates represent the expected absolute HCF count at the mean regional extension. Only eyes with measurable extension for the corresponding region were included in each analysis.

^∗B-GA versus U-GA.

^†Controls vs. B-GA.

^‡Controls vs. U-GA.

Within noRORA regions, statistically significant differences were mainly observed as small intermediate-reflectivity HCF. Small intermediate-reflectivity choriocapillaris HCF were significantly increased in both GA groups compared to controls (U-GA vs. controls, P < 0.001; B-GA vs. controls, P < 0.001), and they were significantly higher in U-GA compared to B-GA eyes (23.1 ± 2.31 vs. 16.6 ± 1.70; P = 0.006). Figure 2 summarizes the statistically significant differences in choriocapillaris HCF subtype distribution between U-GA and B-GA in cRORA and noRORA regions. At the choroidal level, only small intermediate-reflectivity HCF were significantly increased in GA eyes compared to controls, in both B-GA (P = 0.032) and U-GA eyes (P = 0.044), without differences between the two GA groups.

Figure 2.

Choriocapillaris (CC) HCF counts across RORA-defined regions. Statistically significant differences between U-GA and B-GA eyes are shown. In cRORA, eyes with U-GA exhibited a higher number of small intermediate-reflectivity (I-ref) HCF compared with B-GA eyes (P = 0.0006), whereas large high-reflectivity (H-ref) HCF were more frequent in B-GA eyes (P = 0.025). In noRORA regions, small I-ref choriocapillaris HCF were also significantly increased in U-GA compared with B-GA eyes (P = 0.006).

Reproducibility

Grading repeatability was evaluated with ICCs (two-way model, absolute agreement, single measures) and Bland–Altman analysis, as shown in Supplementary Figure S1, in the whole choroid, the choriocapillaris, and the choroid layer. Intergrader agreement was good to excellent (whole choroid ICC = 0.904; CC ICC = 0.837; choroid layer ICC = 0.917), and intragrader agreement was high for both readers (GM, ICC = 0.874–0.906; FP, ICC = 0.997–0.999). All analyses reported here were based on the gradings of a single prespecified reader (FP), the more experienced grader, with the independent gradings of the second reader used solely for the assessment of inter- and intragrader reproducibility; given the degree of agreement observed, this choice would not be expected to affect the results.

Discussion

This study supports HCF as a new choroidal OCT biomarker in eyes with GA demonstrating a significant increase of this parameter compared to healthy age-matched controls. Overall choroidal changes represent a well-established hallmark of GA as confirmed by clinical and histopathologic studies.^23–26 Choroidal thinning observed in our GA cohort is consistent with extensive structural and histopathologic evidence indicating early and progressive involvement of the choriocapillaris–choroid complex in AMD.^23,27 Previous studies have shown that alterations of overall choroid may precede the development of overt retinal atrophy, suggesting an active role of these changes in the pathogenesis of GA rather than a secondary epiphenomenon.^28,29 Notably, the significant more pronounced choroidal thinning in U-GA eyes compared to B-GA eyes further supports the existence of biologically distinct GA phenotypes, based on differences in choroidal thickness, retinal sensitivity, and disease progression patterns.³⁰ Growing evidence from OCTA studies further supports a primary role of choriocapillaris dysfunction in GA pathogenesis.^31–33 Choriocapillaris flow deficits have been shown to correlate with GA enlargement rates and to be detectable in surrounding regions, suggesting that choroidal vascular impairment may precede or actively contribute to outer retinal degeneration.^21,31–34 Specifically, several studies have demonstrated that reduced choriocapillaris perfusion outside GA margins is associated with subsequent lesion progression.^32,33 More recent longitudinal OCTA studies further indicate that widespread choriocapillaris impairment may already be present in earlier AMD stages and precede the development of GA, supporting a dynamic involvement of the choriocapillaris across disease progression,³² as previously described in histopathologic studies^23,34 and more recently confirmed by a clinicopathologic correlation study by McLeod et al.³⁵

In the scenario of these structural and vascular choroidal changes, HCF may represent a potential imaging biomarker of the underlying choroidal disease. Specifically, we observed a significant increase of the number of HCF in both GA phenotypes compared to controls. The increased number of HCF within noRORA regions confirms that HCF are specific OCT features and not OCT findings merely related to an increase of choroidal visualization due to RPE atrophy.¹⁴ Moreover, our results reinforce previous findings that choroidal changes are widespread because HCF are detectable beyond GA margins, even in non-atrophic regions (noRORA). This observation is consistent with the hypothesis that subclinical choroidal alterations may represent early biomarkers of the disease, even before the development of overt atrophy.^28,29,36 Although previous studies have emphasized a close relationship between HCF burden and atrophic extent,³⁷ our findings suggest that the spatial distribution of HCF, particularly their presence in noRORA regions, may provide complementary information beyond the atrophic areas.

Although total HCF count was comparable between B-GA and U-GA eyes across all RORA regions, subtype HCF analysis revealed marked differences at the choriocapillaris level. In our cohort, the comparable extent of cRORA, iRORA, and noRORA regions between the two GA phenotypes reduces the likelihood that these differences in HCF distribution merely reflect disparities in atrophic burden.^14,37 U-GA eyes were characterized by a predominance of small intermediate-reflectivity choriocapillaris foci, in both cRORA and noRORA regions. Previous experimental and clinical studies have shown that inflammatory HRF (I-HRF), as defined by Frizziero et al.,⁵ which represent activated inflammatory cells, display a distinct phenotypic distribution across GA subtypes.³⁸ I-HRF are significantly more numerous in U-GA compared to B-GA.^5,39 Therefore, the predominance of small intermediate-reflectivity choriocapillaris HCF in U-GA eyes may reflect an inflammatory milieu, rather than merely representing a consequence of greater atrophic burden.⁴⁰

This interpretation is also supported by histopathologic evidence showing increased numbers and activated choroidal mast cells in GA, particularly in close proximity to choriocapillaris and Bruch's membrane.^35,41,42 When activated, mast cells release a wide array of inflammatory mediators, which activate matrix metalloproteinases and promote alterations of extracellular matrix, basement membranes, and stromal tissue.^41,42 These mechanisms have been proposed to contribute to choroidal thinning, choriocapillaris dropout, and RPE damage consistent with GA.⁴¹ This hypothesis is further supported by experimental evidence showing that mast cells promote choroidal neovascularization in AMD models,⁴³ potentially accounting for the higher inflammatory activity observed in U-GA eyes with MNV in the fellow eye. Moreover, the choroid is characterized by a dense network of resident macrophages and dendritic cells, along with occasional trafficking naïve lymphocytes that can penetrate the permeable choroidal vasculature.⁴⁴

Macrophages may play a key role in AMD pathogenesis contributing to Bruch's membrane disruption. Notably, the macular choriocapillaris exhibits higher intercellular adhesion molecule 1 (ICAM-1) expression, which may facilitate leukocyte adhesion and transmigration, promoting endothelial injury and reduced perfusion.⁴⁵ Furthermore, macrophage-derived cytokines and growth factors may further exacerbate choriocapillaris dysfunction and RPE degeneration.⁴⁶ This hypothesis is also supported by broader literature describing differential immune regulation and macrophage-driven inflammation in AMD pathogenesis.^10,40,47,48 An increased inflammatory activity might explain the higher rate of MNV developing in the GA study eye of patients with MNV in the fellow eye compared to patients with bilateral GA, as previously reported.¹⁵ In this context, the increased number of small intermediate-reflectivity choriocapillaris HCF observed in U-GA eyes may therefore reflect a more pronounced inflammatory and tissue-remodeling activity at the Bruch's membrane–choriocapillaris interface. Therefore, the increased number of hyperreflective foci showing similar imaging features (size and reflectivity) at both retinal (HRF) and choriocapillaris (HCF) levels in U-GA eyes further strengthens the hypothesis that inflammatory pathways contribute to the disease expression in this GA phenotype.

In contrast, B-GA eyes demonstrated a higher prevalence of large high-reflectivity choriocapillaris HCF, especially within cRORA regions. Larger and highly reflective choriocapillaris foci have been histologically associated with chronic extracellular deposits, calcific material, advanced RPE-derived remodeling processes, and transdifferentiated or degenerating RPE remnants and pigmented melanocytes.^{12,25,27,35,44} In addition, these features may reflect aggregates of activated, debris- or lipid-laden macrophage-like cells (foam cell–like), together with extracellular material, further supporting their association with chronic tissue remodeling processes.⁴⁵ Both OCT and histologic studies have shown that thickening of the RPE– Bruch's membrane complex and accumulation of deposits are associated with GA enlargement.^49,50 These structural changes are thought to reflect long-standing tissue remodeling and impaired metabolic exchange across Bruch's membrane which contribute to irreversible outer retinal and choriocapillaris degeneration. Overall, the predominance of large high-reflectivity choriocapillaris foci in eyes with GA, particularly in B-GA, likely represents a more advanced and structurally consolidated disease phenotype, in which chronic degenerative and remodeling processes prevail over active inflammatory mechanisms.

No statistically significant difference between the two GA phenotypes was observed within iRORA regions. However, U-GA eyes showed a consistent trend toward higher expression of small intermediate-reflectivity choriocapillaris HCF, exhibiting the same imaging features as those observed in noRORA and cRORA regions. This pattern supports the interpretation of iRORA as a biologically transitional stage, characterized by ongoing degeneration in which phenotype-specific features begin to emerge but are not yet fully expressed. Such heterogeneity is consistent with the CAM framework, which recognizes iRORA as an intermediate and variably progressive condition.^3,4,29 Longitudinal imaging studies further support this view, showing that structural and functional biomarkers in iRORA display variable trajectories before converging toward complete atrophy.^51,52

The cross-sectional design and the absence of longitudinal follow-up represent the principal limitation of this study, as they preclude determining whether HCF burden or distribution precedes, accompanies, or follows GA progression; our findings should therefore be interpreted as imaging-based associations rather than as evidence that HCF drive disease progression. Prospective longitudinal studies with repeated OCT imaging are required to assess whether HCF characteristics predict GA development since early stages of the disease, as intermediate AMD, growth, new satellite lesion development, or phenotypic transition. Moreover, because the study was cross-sectional, we cannot exclude that some U-GA patients will develop GA in the fellow eye over time; however, U-GA patients were not younger and their atrophic area was not smaller than in B-GA patients, so the different HCF findings are unlikely to depend on U-GA eyes being at an earlier disease stage. A further limitation is the manual count of HCF on OCT scans which is time consuming and requires expert and well-trained operators. However, to the best of our knowledge, to date no automatic method has been validated to clearly identify and reproducibly quantify choroidal foci.

An additional limitation is the sample size. However, all subjects with a diagnosis of GA were consecutively enrolled from the patient population of our institution, and multimodal imaging analysis (OCT angiography and/or fluorescein and indocyanine green angiography) to evaluate eventual MNV was required to be included. A further limitation is the use of SD-OCT without enhanced depth imaging, rather than swept-source OCT, which offers less depth penetration and a less favorable sensitivity roll-off at the level of the choroid. However, two considerations restrict the impact of this constraint. First, all B-scans were acquired with 100-frame ART averaging, which substantially increased the signal-to-noise ratio at the level of the RPE–Bruch's membrane complex and of the choriocapillaris and choroid. Second, intergrader agreement for foci in the choroid layer was excellent (ICC = 0.917), with no systematic or proportional bias on Bland–Altman analysis. Moreover, as this limitation applied equally to both groups, any residual under-detection would be non-differential and would be expected to bias intergroup comparisons toward the null rather than to generate the differences observed.

Finally, HCF analysis limited to a single high-resolution foveal B-scan may have underestimated the spatial variability of HCF across the macula. However, the use of model-based adjustment for regional linear extension and the region-specific analyses helped reduce the potential influence of differences in regional extent and the impact of enhanced OCT signal transmission in atrophic regions, enabling meaningful intergroup comparisons.^9,14

Conclusions

This study provided a detailed in vivo characterization of HCF in GA, highlighting phenotype-specific differences in HCF subtypes despite a comparable overall HCF burden between bilateral and unilateral GA. These findings suggest that HCF location and specific features, rather than absolute quantity, may reflect distinct pathogenetic mechanisms underlying GA heterogeneity.^14,30,38,39 The distinct choriocapillaris HCF patterns observed between the two GA phenotypes are consistent with divergent pathogenetic pathways and highlight the potential role of choroidal biomarkers in refining early GA phenotyping, risk stratification, and disease monitoring. Dedicated clinicopathologic correlation studies integrating imaging with molecular and histologic analyses are warranted to elucidate the cellular origin, biological significance, and potential prognostic role of HCF in GA progression.