Longitudinal Behavior of the Hyper-Reflective Ganglion Cell Layer Band and Its Association With Ellipsoid Zone Constriction Rate in RP

Abstract

Purpose

To investigate the longitudinal behavior of the hyper-reflective ganglion cell layer band (HGB) and its association with ellipsoid zone (EZ) constriction over time in molecularly characterized RP, based on the hypothesis that the HGB represents a manifestation of inner retinal remodeling with potential prognostic significance.

Methods

We conducted a retrospective cohort study featuring patients with typical RP caused by variants in photoreceptor-associated genes and ≥6 months follow-up. Optical coherence tomography images were reviewed for the presence of HGB, epiretinal membrane, and cystoid macular edema. Predictors of ganglion cell layer (GCL) thickness and longitudinal changes in best-corrected visual acuity were assessed with linear mixed models. Longitudinal EZ constriction was estimated using exponential mixed models with log-transformed EZ width.

Results

We included 118 eyes from 61 patients (30 females; mean baseline age, 34.7 ± 13.2 years) with a mean follow-up of 4.5 ± 2.6 years. HGB was detected at baseline in 27 eyes (22.9%) from 14 patients. Over follow-up, no eyes developed or lost the HGB. Baseline GCL thickness was significantly greater in eyes with HGB compared with those without (65 ± 10 µm vs. 53 ± 8 µm). Compared with autosomal recessive RP (5.5% per year), EZ constriction was significantly slower in autosomal dominant RP (2.8% per year; P = 0.0005) and faster in X-linked RP (7.7% per year; P = 0.021). Eyes without HGB showed a decline of 6.4% per year, compared with 4.3% per year in eyes with HGB (P = 0.001).

Conclusions

HGB is a relatively stable optical coherence tomography finding in RP and is associated with increased GCL thickness and a slower rate of EZ constriction.

Retinitis pigmentosa (RP) comprises a genetically heterogeneous group of inherited retinal diseases (IRDs), with more than 100 causative genes identified to date and autosomal dominant (ad), autosomal recessive (ar), or X-linked (XL) modes of inheritance (https://retnet.org/).^1,2 Pathogenic variants underlying RP predominantly affect proteins involved in phototransduction, retinal metabolism, ciliary transport, and photoreceptor structural integrity.³ Primary rod photoreceptor degeneration, followed by secondary cone involvement, leads to the characteristic clinical course of RP, which includes early-onset night blindness, progressive visual field constriction, and eventual impairment of central and color vision.^2,4

Optical coherence tomography (OCT) plays a central role in monitoring RP progression by enabling noninvasive, in vivo assessment of retinal structure and the identification of reproducible structural biomarkers.⁵ Among these, ellipsoid zone (EZ) width and EZ area have emerged as some of the most robust and widely used indicators of disease progression, because their progressive constriction closely parallels visual field loss.^6–11

Accumulating evidence indicates that EZ width and EZ area reduction follow a nonlinear, decelerating trajectory, with faster loss observed at younger ages and in eyes with greater baseline EZ extent, followed by a progressively slower decline as the EZ becomes more restricted.^8,9,12 Despite this, most longitudinal studies report EZ constriction rates in microns per year, implicitly assuming linear progression over time.^11,13–15 Conversely, studies adopting nonlinear modeling approaches have largely focused on single genotypes or featured patients lacking genetic confirmation, yielding valuable genotype-specific natural history data but limiting the generalizability of their findings across the broader RP population.^8,12,16,17

Although the longitudinal assessment of RP has primarily focused on outer retinal biomarkers, increasing evidence suggests that inner retinal remodeling also occurs during disease progression. Our group recently described a novel OCT finding, termed the hyper-reflective ganglion cell layer band (HGB), detectable in approximately one-quarter of patients with RP and associated with slightly worse visual acuity.¹⁸ These observations were subsequently replicated by Dias et al.,¹⁹ who reported a prevalence of 19.4% and confirmed the association with poorer visual function. Emerging OCT angiography evidence suggests that HGB may reflect perivascular gliosis surrounding the capillaries of the superficial capillary plexus, supporting a microvascular or neuroglial component to its pathophysiology.²⁰

However, it remains unknown whether HGB represents a stable structural feature or a dynamic finding that may appear or disappear over time, and whether its presence carries prognostic significance for photoreceptor degeneration. Establishing such an association would also justify efforts toward its objective quantification through anatomical surrogate markers. We hypothesized that HGB represents a stable manifestation of inner retinal remodeling and that its presence is associated with the rate of photoreceptor degeneration.

Therefore, in a molecularly characterized, genetically heterogeneous cohort of patients with RP, we investigated the longitudinal behavior of HGB, its relationship with EZ constriction using an exponential progression model, and the potential of ganglion cell layer (GCL) thickness as a surrogate marker for future quantitative characterization.

Methods

Subjects

This retrospective cohort study included patients with molecularly characterized RP or typical rod–cone dystrophy, followed at the Heredodystrophy Unit, Department of Ophthalmology, IRCCS San Raffaele Scientific Institute, Milan, Italy. The study adhered to the Declaration of Helsinki, and the requirement for written informed consent was waived by the local ethics committee (Institutional Review Board approval code: IRDs_OSR). To be included, patients had to carry pathogenic, likely pathogenic, or phenotype-compatible variants of unknown significance in genes well characterized within the classical RP spectrum and supported by sufficient natural history data (Supplementary Table S1). Genotypes associated with atypical phenotypes, limited natural history, or distinct clinical entities were excluded (Supplementary Table S2).

Eligible patients had at least one radial scan (six 30° B-scans, ART > 25) and one 20° × 20° macular volume scan (19 horizontal B-scans, 234 µm spacing, ART > 9) acquired using a Spectralis HRA+OCT (Heidelberg Engineering, Heidelberg, Germany). Eyes were excluded if the presence of the HGB could not be assessed due to poor image quality, extensive cystoid macular edema (CME), or stage 4 epiretinal membrane (ERM).²¹ Additional exclusion criteria included follow-up duration of less than 6 months, absence of follow-up imaging, or presence of other conditions potentially affecting the analysis, such as macular neovascularization, branch retinal vein occlusion, or prior intraocular surgery other than cataract extraction.

Data Collection and Analysis

Best-corrected visual acuity (BCVA) was measured using Snellen chart and converted to Early Treatment for Diabetic Retinopathy Study (ETDRS) letters. The presence of HGB, ERM, or CME was recorded on OCT scans, as previously described.¹⁸ The HGB was defined as a continuous hyper-reflective band located within the GCL (Fig. 1). Its presence at baseline was independently assessed by two graders (RT and GS); in cases of disagreement, a third experienced grader (AA) adjudicated the final decision. The ERM was identified as an irregular, hyper-reflective layer on the inner retinal surface, typically associated with retinal wrinkling and hyporeflective spaces between the ERM and the internal limiting membrane.²¹ Ectopic inner foveal layers, when present, were also distinguished from the HGB. CME was defined as hyporeflective intraretinal cystoid spaces visible on at least two consecutive macular raster scans. The horizontal EZ width was measured on 30° OCT scans and ranged from the full scan length (when the EZ extended beyond scan limits) to 0 µm (when the EZ was no longer visible).¹⁸

Figure 1.

Representative multimodal imaging in RP with and without a HGB. (Top) A 20-year-old woman with MYO7A-related retinopathy showing HGB. Fundus autofluorescence (FAF) demonstrates a relatively preserved posterior pole with a hyperautofluorescent ring. Corresponding OCT B-scans (green lines) show a continuous hyper-reflective band at the GCL level (yellow arrowheads), extending across the macula. (Bottom) A 50-year-old woman with USH2A-related retinopathy without HGB. FAF shows a well-demarcated hyperautofluorescent ring. OCT B-scans do not demonstrate a discernible HGB.

Fovea-centered OCT raster scans were used to quantify GCL thickness, defined as the distance from the outer boundary of the retinal nerve fiber layer to the inner boundary of the inner plexiform layer. Thickness values were extracted from the four inner sectors of the standard ETDRS grid, centered either on the fovea or on the point of maximum convergence of the inner retinal layers, using Heidelberg Eye Explorer software. The average of these four sectors was used for statistical analyses. All B-scans were reviewed for segmentation accuracy and manually corrected by a single experienced grader (GS) using the software's built-in correction tool. GCL thickness was assessed only at baseline and at the last follow-up visit, whereas all other structural OCT parameters, as well as BCVA, were evaluated at all available visits.

Study Outcomes

Coprimary outcomes were to assess whether the HGB can appear or disappear over time and if its detection at baseline visit is associated with faster EZ constriction or BCVA decline. A secondary outcome was to explore the thickness of the GCL as a surrogate measure for HGB thickness.

Statistical Analyses

Continuous variables were summarized as mean ± SD, and categorical variables were reported as frequencies and percentages.

Baseline predictors of GCL thickness were evaluated using linear mixed-effects models with a random intercept for patients to account for the correlation between fellow eyes of the same patient. Each predictor (age, EZ width, HGB, ERM, CME, and inheritance mode) was first tested in separate univariable models. Variables showing statistical significance were subsequently included in a multivariable model.

Mixed-effects models assuming exponential decay were used to estimate the rate of change in EZ width. For EZ width (µm), the outcome was log-transformed and modeled as a function of follow-up time, baseline HGB status, and inheritance mode (with arRP as the reference), with interaction terms for time. In both models, random intercepts were specified for eyes nested within patients, accounting for both the within-eye correlation across repeated visits and the between-eye correlation within the same subject. For in-text reporting, model estimates were back-transformed to yearly percentage change for interpretability. All statistical analyses were performed using R software (R Foundation for Statistical Computing, Vienna, Austria), and P values of ≤0.05 were considered statistically significant.

Results

Study Cohort

An initial cohort of 136 patients with molecularly tested rod–cone dystrophy diagnosed between September 2015 and June 2024 was screened. Forty-eight patients harboring variants in genes listed in Supplementary Table S2 were not considered. Twenty-six additional patients were excluded due to insufficient follow-up (n = 22) or factors preventing the assessment of HGB presence (n = 4). Four eyes from four patients were excluded due to macular neovascularization, branch retinal vein occlusion, suboptimal imaging quality, or prior pars plana vitrectomy for ERM.

The final cohort included 118 eyes from 61 patients (30 females) with a typical RP phenotype and pathogenic variants in photoreceptor-associated genes over a mean follow-up of 4.5 ± 2.6 years. The most frequent mode of inheritance was ar, with 46 cases (75%), followed by 9 (15%) adRP, and 6 (10%) XLRP (Supplementary Table S3). The mean patients age at baseline was 34.7 ± 13.2 years, and this was not statistically different between patients with and without HGB in at least one eye (P = 0.99). Patients’ demographics and baseline clinical characteristics are summarized in Table 1.

Table 1.

Patient Demographics and Baseline Data

	With HGB (n = 14)	Without HGB (n = 47)	P value
Age (years)	34.8 ± 11.0	34.7 ± 13.8	0.99*
Sex (females), n (%)	10 (71)	20 (43)	0.06†
Mode of inheritance			0.46‡
ar, n (%)	12 (86)	34 (72)
ad, n (%)	2 (14)	7 (15)
XL, n (%)	0 (0)	6 (13)
CME	5 (36)	11 (23)	0.30‡
ERM	6 (43)	22 (47)	1.00‡

Values are, mean ± SD or number (%). Data reported at the patient level, with subjects classified as having HGB, CME, or ERM, if the feature was present in at least one eye.

Statistically significant P values in bold.

^*Student t-test.

^†χ² test;

^‡Fisher's exact test.

Baseline Characteristics

At baseline, the HGB was detected in 27 eyes (22.9%) from 14 patients, being more prevalent in arRP (27%) than adRP (19%), and absent in XLRP, although this difference was not statistically significant (Fisher's exact test P = 0.51). Similarly, no differences in CME or ERM prevalence were detected between groups (Fisher's exact test P = 0.30 and P = 1.00, respectively). No eye showed an EZ width exceeding the OCT B-scan length (range, 0–8733 µm). Twelve eyes from the HBG– group had an undetectable EZ. Among the 106 eyes with a measurable EZ, the mean width was 3773 ± 2534 µm in the HGB+ group and 2693 ± 1474 µm in the HGB– group. Among eyes with a preserved EZ, the median BCVA was 0.1 logMAR (interquartile range, 0.0–0.25 logMAR) in HGB+ eyes and 0.0 logMAR (interquartile range, 0.0–0.2 logMAR) in HGB– eyes. Of these, 6 eyes (22%) from the HGB+ group and 12 eyes (13.2%) from the HGB– group had posterior subcapsular cataract (Table 2).

Table 2.

Quantitative Imaging Data

	With HGB (n = 27)		Without HGB (n = 91)
	Baseline	Last Follow-up	Baseline	Last Follow-up
Posterior subcapsular cataract	6 (22.2)	10 (37.0)	12 (13.2)	13 (14.3)
Measurable EZ width	27 (100)	27 (100)	79 (86.8)	76 (83.5)
Horizontal EZ width (µm)	3773 ± 2534	3281 ± 2395	2693 ± 1474	2223 ± 1249
BCVA (logMAR)*	0.1 (0.0–0.25)	0.1 (0.0–0.25)	0.0 (0.0–0.2)	0.0 (0.0–0.2)
GCL thickness (µm)†	64.9 ± 9.6	66.9 ± 11.6	53.3 ± 8.1	56.6 ± 8.5
Superior sector	66.1 ± 9.5	67.4 ± 10.7	54.7 ± 9.9	57.0 ± 8.7
Temporal sector	63.4 ± 12.3	66.1 ± 15.9	52.0 ± 9.5	55.0 ± 9.3
Inferior sector	65.4 ± 10.5	66.1 ± 13.0	51.7 ± 8.8	55.3 ± 8.7
Nasal sector	65 ± 9.4	67.9 ± 10.4	54.7 ± 9.2	58.0 ± 11.1

Values are number (%), mean ± SD, or median (interquartile range). Data are reported at the eye level. Horizontal EZ width and BCVA are showed only for patients with a measurable EZ.

^*Data from eyes with a measurable EZ width.

^†Overall and sectoral GCL thickness refer to the inner ring of the standard ETDRS chart.

Baseline GCL thickness was assessed in the inner ring of the standard ETDRS chart. On average, the GCL was thicker in HGB+ patients, measuring 65 ± 10 µm in this group, and 53 ± 8 µm in eyes without this sign on OCT (Fig. 2); this difference was consistent across all sectors. In a mixed-effects model, the presence of HGB was independently associated with increased GCL thickness (β = 10 µm; P < 0.001), whereas age was associated with thinner GCL (β = −1.7 µm per 10 years, P = 0.029). Conversely, ERM, CME, EZ width, and inheritance mode were not significantly associated with GCL thickness (Table 3).

Figure 2.

Distribution of GCL thickness according to HGB status at baseline and last follow-up. Violin plots show GCL thickness in eyes without HGB and with HGB at baseline (red) and last follow-up (blue). Individual points represent single eyes; the central bar indicates the median and error bars represent the interquartile range. Eyes with HGB show greater GCL thickness at both time points, and a slight increase in GCL thickness is observed over follow-up.

Table 3.

Predictors of GCL Thickness

	Univariable Analysis			Multivariable Analysis
Predictor	Estimate (µm)	95% CI (Lower, Upper)	P Value	Estimate (µm)	95% CI (Lower, Upper)	P Value
Intercept				57.77	54.64 to 66.33	<0.001
Age (per decade)	−1.89	−3.67 to −0.10	0.043	−1.72	−3.25 to −0.21	0.029
EZ width (per 1000 µm)	1.17	−0.09 to 2.25	0.036	0.63	−34.8 to 1.62	0.21
HGB (ref. absent)	10.2	5.82 to 14.65	<0.001	9.81	5.46 to 14.16	<0.001
ERM (ref. absent)	1.90	−0.78 to 4.59	0.168	—	—	—
CME (ref. absent)	1.89	−1.82 to 5.61	0.321	—	—	—
Inheritance mode (vs. arRP)
adRP	0.13	−6.77 to 7.02	0.971	—	—	—
XLRP	−2.56	−10.73 to 5.61	0.541	—	—	—

Dependent variables are GCL thickness (µm). Random intercepts were included at the patient level.

Statistically significant P values in bold.

Longitudinal Changes

Over the follow-up, HGB loss or development was not observed in any eye. EZ width became unmeasurable in three HGB– eyes. Excluding these eyes, the mean change in EZ width was −623 ± 567 µm in HGB– eyes (n = 76) and −492 ± 411 µm in HGB+ eyes (n = 27). Among those with a measurable EZ, the median visual acuity remained stable (Table 2). Conversely, GCL thickness increased slightly over time, with a mean change of +2.63 ± 5.73 µm in HGB– eyes (n = 93) and +1.94 ± 5.40 µm in HGB+ eyes (n = 27).

EZ Constriction Rate

For estimation of the EZ band constriction rate, eyes with an unmeasurable EZ at baseline were excluded, as were follow-up observations after the EZ width reached 0, from the exponential mixed-effects model. Accordingly, the final model included 566 observations from 106 eyes of 56 patients (Fig. 3).

Figure 3.

EZ width decline over time. (A) Scatterplot of EZ width as a function of age, stratified by inheritance mode. (B) Longitudinal change in EZ width expressed as percentage of baseline according to inheritance mode. (C) Longitudinal EZ width decline according to HGB status. Solid lines represent the overall group trend derived from all longitudinal measurements by linear regression, with shaded areas indicating the corresponding 95% CIs. XLRP shows the fastest EZ constriction, whereas autosomal-dominant RP shows the slowest decline. Eyes with HGB exhibit a slower EZ width decline compared with HGB-negative eyes.

In the reference group (HGB–, arRP), EZ width declined at a rate of 6.4% per year (95% confidence interval [CI], 5.6–7.2; P < 0.001). When stratified by inheritance pattern, the rate of EZ constriction differed significantly across genotypes. Compared with arRP (5.5% per year; 95% CI, 4.8–6.1), a significantly slower decline was observed in adRP (2.8% per year; 95% CI, 1.4–4.3; P = 0.0005), whereas a significantly faster decline was observed in XLRP (7.7% per year; 95% CI, 5.8–9.5; P = 0.021). Collectively, HGB– eyes declined at a rate of 6.4% per year (95% CI, 5.6–7.2), compared with 4.3% per year (95% CI, 2.9–5.6) in HGB+ eyes, indicating significantly slower EZ constriction in the presence of HGB (P = 0.001) (Table 4). Restriction of the analysis to patients carrying at least one pathogenic or likely pathogenic variant confirmed HGB presence as an independent predictor of slower EZ constriction (Supplementary Table S4).

Table 4.

Multivariable Mixed-Effect Model of Inheritance Mode and HGB Status as Predictors of EZ Progression Rate

Predictor	Estimate (β)	95% CI (Lower, Upper)	P Value
Baseline effects
Intercept	7.878	7.604 to 8.151	<0.001
HGB at baseline	−0.234	−0.604 to 0.135	0.216
adRP	−0.332	−0.937 to 0.272	0.286
XLRP	−0.038	−0.819 to 0.744	0.925
Longitudinal effects
Follow-up time (years)	−0.0675	−0.0748 to −0.0603	<0.001
Follow-up time × HGB	0.0228	0.0092 to 0.0363	0.001
Follow-up time × adRP	0.0273	0.0119 to 0.0427	<0.001
Follow-up time × XLRP	−0.0235	−0.0434 to −0.0036	0.021

Coefficients (β) are expressed on the log scale (log-µm/year). In-text results were back-transformed to percentage change per year using %Δ = (exp[β] − 1) × 100 for interpretability. A negative slope indicates EZ constriction. Genotype-specific deviations were estimated using arRP as the reference group.

Statistically significant P values are shown in bold.

Discussion

In this study, we investigated the prevalence, longitudinal changes, and prognostic value of HGB in a molecularly confirmed cohort of rod–cone dystrophies. Despite restricting inclusion to RP caused by photoreceptor-associated genes, we observed a baseline prevalence of HGB (22.9%) comparable with that reported in our previous study (25%) and in the study by Dias et al. (19.4%).^18,19 This restriction was essential to improve the internal validity and reproducibility of our findings by excluding poorly characterized genotypes, as well as genes known to produce distinct retinal phenotypes. These included genes associated with early macular atrophy (e.g., PCARE, PROM1),^22,23 abnormal retinal lamination (e.g., CRB1),²⁴ specific disease phenotypes (e.g., NR2E3),²⁵ and extreme refractive errors (e.g., RDH12)²⁶ that could have prevented or confounded measurements of GCL thickness, EZ width, or both.

The presence of an HGB has also been reported in IRDs other than RP, including choroideremia, gyrate atrophy, and NCL3-associated retinopathy, suggesting that shared mechanisms may underlie its development across different genotypes.^27–29 Together with the comparable prevalence observed in our genetically restricted cohort, these findings point to a potential common downstream pathway of the inner retinal changes in response to progressive photoreceptor loss. Nevertheless, the prevalence and clinical significance of this OCT finding across different IRDs remain poorly characterized and warrants further investigation.

Over an average follow-up of approximately 5 years, we did not observe any change in HGB status, suggesting that, once developed, HGB represents a relatively stable OCT feature. However, the absence of detectable changes during the follow-up period does not exclude the possibility that HGB may emerge during earlier phases of retinal remodeling or disappear with advanced retinal degeneration over longer periods of disease evolution. Similarly, although HGB+ eyes showed a slightly poorer median BCVA, central visual function remained stable in both groups throughout follow-up.

Because HGB assessment was qualitative, we cannot exclude that quantitative changes in its magnitude may occur over time. To preliminarily address this, we explored GCL thickness as a potential anatomical surrogate of HGB thickness. At baseline, GCL thickness was approximately 10 µm greater in HGB+ eyes and declined with advancing age, with an estimated reduction of about 2 µm per decade. During follow-up, however, we observed a paradoxical increase in GCL thickness, irrespective of HGB status.

Collectively, these findings suggest that GCL thickness is not a reliable surrogate for quantitative HGB assessment. Although HGB+ eyes consistently showed greater GCL thickness, this observation remains biologically relevant, as it supports the hypothesis that HGB occupies space within the GCL, rather than representing a mere reflectivity change or imaging artifact. Histopathological studies of RP have shown that progressive photoreceptor degeneration is accompanied by substantial inner retinal remodeling, including Müller cell activation and glial proliferation,^30–33 processes that have been proposed as potential contributors to the HGB.²⁰

Because age-related neuronal loss and reactive gliotic remodeling may exert opposing effects on GCL thickness, the net structural changes observed over time likely reflect a balance between degenerative thinning and reactive expansion, rather than HGB status alone. Consequently, future studies aimed at investigating the functional significance of HGB should incorporate dedicated quantitative imaging approaches to directly characterize this OCT feature.

A key outcome of this study was to assess the prognostic relevance of HGB by estimating the rate of EZ constriction. This work is among the few to model EZ constriction exponentially and to do so in a diverse, molecularly confirmed cohort.^8,9,12 In line with previous studies modeling EZ constriction linearly and visual field loss exponentially,^13,16,34,35 we observed genotype-dependent progression rates: compared with arRP, which showed an average EZ constriction rate of about 5.5% per year, XLRP progressed 2.2% per year faster, whereas adRP progressed 2.7% per year more slowly. Unexpectedly, the presence of HGB was associated with a slower rate of outer retinal degeneration. Specifically, EZ constriction in HGB+ eyes was approximately one-third slower than in HGB– eyes (4.3%/year vs. 6.4%/year).

The interpretation of this association is challenging. One possible explanation relates to the complex role of Müller cell–mediated gliosis in degenerating retinas.³⁶ Müller cells are central regulators of retinal homeostasis and respond to photoreceptor degeneration through a gliotic reaction that can exert both protective and detrimental effects.³⁷ Notably, this gliotic response is neither temporally nor topographically restricted to areas of photoreceptor loss. Experimental models of retinal remodeling have shown that Müller cell reactivity and metabolic dysregulation may arise during an early phase I of photoreceptor stress that precedes overt cell death and can be detected in retinal regions where photoreceptors remain structurally preserved.³⁸ This may explain why HGB can be observed across the full spectrum of EZ widths, as documented both in our original description and in the present cohort.¹⁸

Experimental studies have shown that Müller cells may promote photoreceptor survival by releasing neurotrophic factors, removing excitotoxic glutamate, and producing antioxidants, as well as by transcriptionally adapting to support the metabolic needs of surviving photoreceptors.^36,39

At the same time, Müller cells are key regulators of the retinal vascular environment and can influence angiogenic signaling and capillary perfusion.⁴⁰ Gliotic remodeling affecting the superficial capillary plexus has been proposed as a structural substrate of the HGB on OCT and is associated with reduced perfusion density on OCT angiography.²⁰ Within this framework, the presence of HGB could reflect a gliotic response that partially supports photoreceptor survival while simultaneously contributing to inner retinal and vascular alterations.

This dual effect may help to reconcile our findings: enhanced Müller cell–mediated support could contribute to slower outer retinal degeneration and relative preservation of EZ width, while concomitant inner retinal and vascular changes could impair neuronal processing and be associated with the slightly poorer central visual function observed in eyes exhibiting HGB.^18,19

Taken together, our findings suggest that HGB represents an epiphenomenon of inner retinal remodeling, rather than a direct marker of photoreceptor dysfunction. Its presence appears to reflect structural and vascular changes occurring during retinal degeneration, which may coexist with relatively slower outer retinal loss but with subtle impairment of central visual function.

This study has several limitations. First, its retrospective design may introduce inherent selection and measurement biases. Variations in image quality (e.g., due to cataract development or extraction) may have altered HGB visibility, potentially masking or unmasking this sign and thereby affecting its estimated prevalence. Second, the uneven distribution of genotypes limited the statistical power to detect subtle genotype-specific differences. Nonetheless, the stringent inclusion criteria, molecular confirmation of diagnoses, availability of multiple examination time points, and use of exponential modeling of EZ constriction provide robust and reproducible estimates that partially mitigate these limitations.

Future studies are warranted to investigate the prevalence and functional implications of HGB across other IRDs, to determine its potential role as a prognostic biomarker through dedicated quantitative imaging approaches and establish whether HGB may emerge or disappear over longer periods of disease progression.