Macular Pigment and Retinal Reflectance in a Feasibility Study of Multiple Sclerosis: Toward Novel Biomarkers of Hidden Disease Processes

Abstract

Purpose

To evaluate the feasibility and characteristics of macular pigment optical density (MPOD) and retinal reflectance (RR) measurements using two-wavelength autofluorescence (AF) and multicolor imaging, respectively, in multiple sclerosis (MS), and explore differences compared with healthy controls. Sixteen MS participants and 25 age-matched healthy controls were included in this prospective, observational, single-center, cross-sectional, feasibility study,

Methods

After pupil dilation, blue- and green-light fundus autofluorescence and multicolor imaging were performed using the Heidelberg Spectralis. Custom-built ImageJ plugins extracted MPOD, spatial profiles, macular pigment optical volume (MPOV), and hyperspectral indexes (HSI) from RR. Linear mixed-effects models compared outcomes between MS eyes with optic neuritis (MSON), without optic neuritis (MSnON), and controls. A post hoc power analysis characterized detectable effect sizes given the available sample size.

Results

MPOD and RR were successfully measured in all participants. Macular pigment spatial profiles differed between MS and controls (P = 0.010). Ring-like MPOD patterns were observed in 43% of MS eyes and were most pronounced in MSON (67%). HSI varied regionally (P < 0.001) but was not associated with MS status. MPOV did not differ significantly between groups. MPOD demonstrated inter-individual variability within both groups. Exploratory analyses showed that higher peripheral MPOD (3°–6°) was associated with greater inner/full retinal thickness, whereas higher HSI was associated with thinner outer retina and smaller foveal avascular zone volume (all P ≤ 0.05).

Conclusions

AF-based MPOD measurement is feasible in MS, providing spatially resolved, objective macular pigment quantification. AF-derived MPOD showed limited sensitivity to MS-related changes in isolation. Larger cohort studies with multimodal retinal imaging are required to determine MPOD's role as a biomarker in MS.

Multiple sclerosis (MS) is a chronic inflammatory disease of the central nervous system (CNS), characterized by demyelination, axonal loss, and widespread structural damage. In addition to relapse-associated worsening, more subtle background processes drive disability accumulation independent of relapse activity, often referred to as smoldering MS.¹ Both oxidative stress and low-grade chronic inflammation are thought to play key roles in these smoldering disease mechanisms; however, they are challenging to detect using conventional tools.

The retina offers an accessible window into CNS pathology.² Retinal measures reflect both inflammatory and neurodegenerative injury in MS, with inner retinal thinning typically linked to optic neuritis (ON)³ and emerging evidence of outer retinal involvement, including photoreceptor loss, occurring independently of ON.⁴ These observations motivate exploration of retinal signatures that may capture smoldering disease biology.

Macular pigment, composed primarily of the dietary carotenoids lutein and zeaxanthin, can be quantified in vivo as macular pigment optical density (MPOD). MPOD is an optical surrogate, not a direct biochemical measure, and reflects macular carotenoid distribution rather than directly indexing oxidative stress. Macular pigment optical volume (MPOV), calculated by integrating MPOD over the mapped macular area weighted by the pixel count of each sampling zone, provides a complementary estimate of total pigment content. Prior studies in MS reported lower MPOD and associations with cognitive processing speed, suggesting that macular carotenoids may relate to antioxidant capacity relevant to neuroprotection.^5–7 However, all three studies employed heterochromatic flicker photometry (HFP), a subjective psychophysical technique that relies on flicker nulling and may be influenced by attention, fixation stability, and visual pathway dysfunction—limitations of particular concern in MS. Two-wavelength autofluorescence imaging, an objective, imaging-based approach to MPOD quantification, has not previously been applied in this population.^8,9

Retinal reflectance (RR) provides a complementary optical signature. Hyperspectral retinal imaging has revealed disease-related reflectance changes in neurodegenerative disorders such as Alzheimer's and Parkinson's disease, potentially reflecting protein aggregation and inflammatory states.^16,21,22 Whether analogous spectral patterns occur in MS, potentially reflecting subtle inflammatory processes, remains unknown.

Given the limited and heterogeneous evidence on MPOD alterations in multiple sclerosis, particularly from objective imaging, the present study was designed as an exploratory and methodological investigation. Specifically, we aimed to assess the feasibility and characteristics of two-wavelength autofluorescence-based MPOD measurement and RR in MS, rather than to establish MPOD and RR as a definitive biomarker of disease severity.

Methods

Study Design and Recruitment

This research was conducted as part of the Belfast Eye and MS study (BEAMS), a cross-sectional feasibility study approved by the United Kingdom Research Ethics Committee (REC 18/NW/0334). All participants provided informed written consent, and the study was conducted in accordance with the tenets of the Declaration of Helsinki. Participants with MS were recruited via the MS Research Network of Northern Ireland, whereas age- and sex-matched control participants were recruited from the Queen's University Belfast community. The primary exclusion criterion was a history of optic neuritis (ON) within six months before the study visit, including any recorded relapse activity.

All participants underwent clinical ocular screening before enrolment. Participants with known retinal pathology (including age-related macular degeneration, diabetic retinopathy, and glaucoma) were excluded.

For the MS cohort, clinical information, including disease duration, Expanded Disability Status Scale (EDSS) score, comorbidities, and ON history, was obtained from medical records. For the control cohort, this information was collected via a questionnaire. A total of 16 people with MS (pwMS) and 25 controls were enrolled. The MS cohort included 13 individuals with Relapsing-Remitting MS (RRMS) and three with Secondary Progressive MS (SPMS). Eyes from pwMS were classified as either having no history of optic neuritis (MSnON) or having a prior history of optic neuritis (MSON).

Imaging and Data Processing

Following pupil dilation, macular-centered multicolor (MC) images were acquired using a Heidelberg Spectralis camera (camera model: S3610-CIFMP; Heidelberg Engineering GmbH, Heidelberg, Germany), with a 30° scan angle and 15 ART (Automatic Real-Time averaging). MC is a confocal scanning laser method that uses blue, green, and infrared light, each penetrating to different depths of the eye. These images were used to extract retinal reflectance values.

Subsequently, macula‑centered blue and green light (two-wavelength) autofluorescence (BAF and GAF) video recordings of 30 seconds’ duration were acquired using the same imaging system, with a 30° scan angle and ART mode disabled. These autofluorescence recordings were used to extract MPOD.

MPOD

Macular pigment optical density was measured using two-wavelength fundus autofluorescence imaging. Previous validation studies in health and other conditions have shown that this technique provides highly repeatable and reproducible measurements of MPOD, supporting its use for quantitative assessment of macular pigment distribution.^10–12 Following image acquisition, individual frames from the BAF and GAF video recordings were reviewed using Heidelberg Eye Explorer software (Heyex, version 1.10.4.0) with the Expand function. Frames that were blank due to blinking or were off-center were removed from the image stack prior to further processing. MPOD images were generated from the BAF and GAF video recordings using the Calc MP Image function. The spatial distribution of MPOD was subsequently evaluated using radial intensity plots, which depict MPOD values as a function of retinal eccentricity and are automatically generated using the Radial Intensity function in Heyex. Based on these radial profiles, three distinct MPOD spatial distribution patterns were identified—non-ring-like, intermediate, and ring-like—according to the classification previously described (Fig. 1).¹³ Classification was performed masked to clinical subgroup (JM) and 40% of the images were reviewed by senior investigators (LC) and no discrepancies were found between the two graders. To extract MPOD values, the BAF and GAF MPOD images were exported using the Export as XML function, which generated an XML file containing the associated metadata, along with the corresponding blue and green reflectance images in TIFF format.

Figure 1.

MPOD spatial distribution profile Figure is showing the different spatial distribution profiles we observed in our cohorts (A, no-ring-like; B, intermediate; C, ring-like).

Mean MPOD values were extracted using a novel image-processing approach implemented in ImageJ (National Institutes of Health, Bethesda, Maryland, USA) with custom-built plugins developed by KS and distributed via the ImageJ Update Site (https://sites.imagej.net/CreativeComputation/). BAF and GAF images were digitally subtracted to isolate the wavelength-dependent absorption attributable to macular pigment. The resulting images were spatially registered, after which the plugin automatically identified the MPOD center, defined as the location of peak MPOD signal intensity; this position did not necessarily coincide with the anatomical center of the fovea. MPOD data were then sampled at predefined retinal eccentricities relative to this center to generate MPOD profiles (0°, 0.5°, 1°, 2°, 3°, 4°, 5°, and 6°). Schematic representation of MPOD extraction is shown in Supplementary Figure S1. The pixel count (Nᵢ) of the sampling area at each eccentricity was recorded alongside the mean MPOD value (MPODᵢ).

MPOV was calculated as MPOV = Σ(MPODᵢ × pixelsᵢ), where MPODᵢ is the mean MPOD at eccentricity i and pixelsᵢ is the corresponding pixel count of the sampled area. This pixel-weighted formula accounts for the increasing area sampled at greater eccentricities and uses the actual image-derived pixel counts.

Following quality control procedures, 21 eyes from 12 individuals in the MS cohort (16 MSnON eyes of 10 participants and five MSON eyes of four participants) and 36 eyes from 19 individuals in the control cohort were included in the final analysis. To minimize variability, all images were acquired using a standardized protocol on the same Heidelberg Spectralis platform.

RR

MC images were reviewed using the same Heyex software, and images of insufficient quality were excluded from further analysis. Exclusion criteria included images that were underexposed or overexposed, blurred or out of focus, or affected by shadowing, for example, due to media opacities. The remaining images were processed using two custom macro scripts. The first macro generated a TIFF image stack comprising the three reflectance channels and superimposed an Early Treatment Diabetic Retinopathy Study (ETDRS) grid, subdividing the macular region into four concentric zones. When necessary, the ETDRS grid placement was manually adjusted to ensure accurate centration on the foveola. The second macro produced ETDRS-based masks for each of the four regions, enabling region-specific data sampling (central subfield [C, 0.5 mm diameter circle], central circle [CC, 0.5–1 mm], inner ring [IR, 1–3 mm], and outer ring [OR, 3–6 mm]) across all reflectance wavelengths. For each participant, a total of twelve reflectance measurements were obtained, corresponding to the three wavelengths within each of the four retinal regions. Schematic representation of RR extraction is shown in Supplementary Figure S1. Hyperspectral Index (HSI) ratios were calculated by normalizing blue and green reflectance values to red reflectance, yielding the blue-to-red ratio (HSIb = B/R) and the green-to-red ratio (HSIg = G/R). Following quality control procedures, 27 eyes from 15 individuals in the MS cohort (18 MSnON of 12 participants and nine MSON of six participants) and 46 eyes from 25 individuals in the control cohort were included in the final analysis.

Statistical Analysis

Demographic and clinical data were compared between MS and control groups using the χ² test for categorical variables and the Wilcoxon rank sum test for continuous variables. The distribution of macular pigment spatial profiles (no-ring, intermediate, ring-like) between groups was assessed using Fisher's exact test (Ctrl vs. MS; Ctrl vs. MSnON vs. MSON). Effect sizes were quantified using Cramér's V. To further characterise group differences, standardized residuals from the corresponding contingency tables were examined to identify the specific categories contributing to the overall association. MPOV comparisons across groups were performed using the Kruskal–Wallis test, with post-hoc pairwise Mann–Whitney U tests.

MPOD and HSI in the blue (HSIb) and green (HSIg) spectral ranges were analyzed using linear mixed-effects modeling to account for repeated measurements across retinal eccentricities or zones within the same eye, and for the correlation between eyes within the same participant. For MPOD, retinal eccentricity (0° to 6°) and clinical subgroup (Control, MSnON, MSON) were included as fixed effects, along with their interaction. For HSI data, ETDRS zone (C, CC, IR, OR) and clinical subdiagnosis (Control, MSnON, MSON) were included as fixed effects, with their interaction. Age and sex were included as covariates in all models. The central measurement (0° for MPOD, C for HSI) was used as the reference level; control eyes and male sex served as reference categories. Random intercepts were included to account for within-subject correlation. Model estimates are reported as regression coefficients with 95% confidence intervals and corresponding P-values. Degrees of freedom were estimated using the Kenward–Roger approximation. Statistical significance was set at P < 0.05, and all analyses were performed in R (version 4.5.1).

Analyses were conducted at both the subject and eye levels. Table 1 summarizes subject-level demographic and clinical characteristics for all enrolled participants. In contrast, Table 2 presents eye-level analyses for eyes that met predefined inclusion criteria for MPOD quantification. Not all eyes were eligible for inclusion in the final analysis due to factors such as image quality, availability of complete MPOD measurements, and, where applicable, exclusion of eyes with prior optic neuritis. As a result, the number of eyes in Table 2 differs from the total number of eyes reported in Table 1. This approach ensured that statistical analyses were performed only on high-quality datasets suitable for reliable MPOD evaluation. Given the exploratory nature of the study and the available dataset, analyses were focused on primary MPOD measures rather than extensive multimodal correlations.

Table 1.

Characteristics of Study Participants

	MS (n = 16)	Ctrl (n = 25)	P Value*
Age			0.462
Mean (SD)	51 (11)	48 (12)
Median (IQR)	50 (43, 59)	52 (38, 57)
Range	34, 73	28, 71
Sex			0.249
Male	8 (50%)	8 (32%)
Female	8 (50%)	17 (68%)
IOPg			0.020
Mean (SD)	12.9 (3.4)	15.8 (3.2)
Median (IQR)	11.8 (11.3, 15.5)	16.0 (14.0, 17.5)
Range	7.3, 18.8	10.1, 22.6
Spherical error			0.177
Mean (SD)	0.87 (1.84)	−0.23 (2.03)
Median (IQR)	0.38 (−0.28, 1.97)	0.00 (−0.50, 0.62)
Range	−2.75, 4.75	−5.87, 4.62
Hypertension			0.007
No	8 (50%)	23 (92.0%)
Yes	8 (50%)	2 (8.0%)
ON history
MSnON	21 (65.6%)	NA	NA
MSON	11 (34.4%)	NA
MS sub-type
RRMS	13 (%)	NA	NA
SPMS	3 (%)	NA

IQR, interquartile range; IOPg, intra-ocular pressure (Goldmann); ON history, optic neuritis history.

Summary of demographic and clinical characteristics of study participants, including healthy controls, individuals with MS.

Bold values indicate statistically significant differences (P < 0.05).

^*The χ² test and Wilcoxon rank sum test.

Table 2.

Distribution of MP Spatial Profiles Across Groups

	Ctrl (N = 36)	MS (N = 21)	MSnON (N = 16)	MSON (N = 5)	P Value*	P Value†
MPOD Spatial Profile					0.012	0.015
No-Ring	15 (42%)	5 (24%)	5 (31%)	0 (0%)
Intermediate	18 (50%)	7 (33%)	5 (31%)	2 (40%)
Ring-Like	3 (8.3%)	9 (43%)	6 (38%)	3 (60%)

Values are reported at the eye level. The number of included eyes differs from that in Table 1 because of image quality criteria, availability of complete MPOD measurements, and eye-level exclusion criteria. Values represent the number and percentage of eyes with no-ring, intermediate, or ring-like MPOD patterns for controls (Ctrl) and MS participants, including subgroups without optic neuritis (MSnON) and with a history of optic neuritis (MSON). Group differences were evaluated using Fisher's exact test:

^*Ctrl versus MS.

^†Ctrl vs. MSnON versus MSON.

To explore associations between MPOD/RR and disease burden and structural retinal measures (OCT and OCTA metrics), additional exploratory correlation analyses were performed. The association of MPOD and HSI with disease duration (years since diagnosis) was estimated using LMM with a subject-level random intercept in MS participants only (n = 12 for MPOD, n = 15 for HSI). Retinal eccentricity (for MPOD) or ETDRS zone (for HSI) was included as a fixed-effect covariate alongside age and sex, yielding a single pooled coefficient for the overall association across all eccentricities or zones simultaneously. For associations with OCT-derived retinal thickness, OCTA SVC capillary density, and foveal avascular zone (FAZ) metrics, a single representative MPOD eccentricity was matched to each ETDRS sector based on published MPOD-to-ETDRS zone correspondences^14,15: 1° ∼ central circle (CC, 0–1 mm), 3° ∼ inner ring (IR, 1–3 mm), and 6° ∼ outer ring (OR, 3–6 mm) for OCT and FAZ; 1° ∼ CC and 3° ∼ IR for OCTA SVC capillary density; for HSI versus OCTA, the HSI CC zone (0–1 mm) was paired with OCTA CC (0–1 mm), and the HSI IR zone (1–3 mm) was paired with the mean of OCTA IR (1–2 mm) and OCTA OR (2–3 mm), because these two rings together span the 1–3 mm range. Using single eccentricities avoids the spatial dilution that can occur when averaging MPOD across a zone where values change sharply with eccentricity.¹⁶ GEE analyses used a Gaussian family with exchangeable working correlation, clustering by participant, pooled across controls and MS participants. A pooled approach was adopted because the MS subgroup alone is too small for reliable separate analyses; MS subdiagnosis (MSnON and MSON as indicator variables, Ctrl as reference) was included as a covariate alongside age and sex, so that the estimated MPOD/HSI associations reflect the relationship with structural outcomes after accounting for mean differences between diagnostic groups, assuming a common slope across groups. For GEE analyses, MPOD was z-score standardized within each eccentricity (i.e., expressed in SD units) prior to analysis. This was necessary because raw MPOD at 6° has an SD of only 0.016 OD, making raw per-unit-OD regression coefficients implausibly large and clinically uninterpretable; standardized coefficients represent the change in the outcome per 1 SD increase in MPOD. HSI was retained on its raw scale because per-SD HSI coefficients would be too small to be meaningful, whereas raw per-unit coefficients are already clinically interpretable. FAZ metrics included area and volume. OCT-derived retinal thickness, OCTA SVC capillary density, and FAZ metrics used in these exploratory correlation analyses were acquired as part of the BEAMS study and have been reported separately.^4,17

Results

Participant Characteristics

The study included 16 participants with MS (mean age 51 ± 11 years) and 25 age- and sex-matched controls (mean age 48 ± 12 years). Although there was no significant difference in spherical error (P = 0.177), intraocular pressure was significantly lower in the MS group (P = 0.020). Hypertension was also significantly more prevalent in the MS cohort compared to controls (P = 0.007). Of the 32 eyes from participants with MS, 11 had a history of optic neuritis (MSON), and 21 did not (MSnON). Cohort characteristics are presented in Table 1, and further MS characteristics are shown in Supplementary Table S1.

Macular Pigment Optical Density and Volume

The distribution of MPOD spatial profiles differed across diagnostic groups (Fisher's exact test, P = 0.012), with a large effect size (Cramér's V = 0.41) (Table 2). Representative MPOD maps illustrating each profile type are shown in Figure 1. Standardized residuals demonstrated that this difference was predominantly driven by a marked excess of ring-like profiles in the MS group and a relative absence in controls, while no meaningful differences were observed for no-ring or intermediate patterns.

In the eye level analysis the spatial profiles in MSnON eyes were nearly evenly distributed across the three categories (31%, 31%, and 38%) (Table 2). However, MSON eyes showed a preference for the ring-like profile (60%), with no eyes in this subgroup exhibiting a no-ring pattern. When comparing across all three groups simultaneously, the difference remained significant (P = 0.015).

We observed a strong dependence on retinal eccentricity, with MPOD values decreasing progressively with increasing distance from the MPOD center (Fig. 2). As shown in Table 3, MPOD was significantly lower at all eccentricities from 0.5° to 6° compared with the central reference location (all P < 0.001).

Figure 2.

Distribution of MPOD across retinal eccentricities in Ctrl, MSnON, and MSON. Boxplots show MPOD measured at increasing eccentricities from the foveal center (0°–6°). Boxes represent the interquartile range, the central line denotes the median, whiskers indicate 1.5 × the interquartile range, and points represent outliers.

Table 3.

Linear Mixed-Effects Model Examining the Association Between MS Subdiagnosis and MPOD

MPOD	Beta	95% CI	P Value
Subdiagnosis
MSnON	−0.012	−0.095, 0.071	0.772
MSON	−0.093	−0.208, 0.023	0.114
Eccentricity
0.5	−0.084	−0.136, −0.033	0.001
1	−0.206	−0.257, −0.155	<0.001
2	−0.393	−0.444, −0.341	<0.001
3	−0.425	−0.475, −0.374	<0.001
4	−0.443	−0.493, −0.393	<0.001
5	−0.465	−0.516, −0.415	<0.001
6	−0.484	−0.535, −0.433	<0.001
Age	0.001	−0.002, 0.004	0.527
Female sex	0.022	−0.049, 0.093	0.532
Subdiagnosis * Eccentricity
MSnON * 0.5	−0.003	−0.094, 0.087	0.942
MSON * 0.5	−0.062	−0.203, 0.080	0.391
MSnON * 1	0.010	−0.081, 0.101	0.824
MSON * 1	0.013	−0.129, 0.154	0.860
MSnON * 2	0.012	−0.076, 0.101	0.784
MSON * 2	0.085	−0.076, 0.247	0.298
MSnON * 3	0.010	−0.080, 0.100	0.826
MSON * 3	0.084	−0.057, 0.225	0.242
MSnON * 4	0.012	−0.081, 0.105	0.794
MSON * 4	0.083	−0.048, 0.213	0.213
MSnON * 5	0.013	−0.075, 0.101	0.773
MSON * 5	0.104	−0.057, 0.265	0.205
MSnON * 6	0.016	−0.075, 0.107	0.732
MSON * 6	0.105	−0.037, 0.246	0.146

Control participants served as the reference group for subdiagnosis.

Bold values indicate statistically significant differences (P < 0.05).

Within clinical subgroups, MPOD did not differ significantly between MS participants’ eyes and control eyes (Table 3). MSON eyes showed lower MPOD values than controls, although the difference was not statistically significant (P = 0.114), suggesting an overall reduction in macular pigment associated with prior optic neuritis (Table 3). Individual MPOD eccentricity profiles for all analyzed eyes are shown together in Supplementary Figure S2, grouped by diagnostic subgroup. Despite the significant differences in spatial profile (Table 2), MPOD was not significantly different at any eccentricities measured across groups (Table 3—interaction term). Additionally, neither age nor sex was significantly associated with MPOD in our model (Table 3).

MPOV results are presented in Supplementary Table S2. The Kruskal–Wallis test did not reach statistical significance (P = 0.351). Post-hoc pairwise comparisons suggested lower MPOV values in MSON eyes compared with MSnON and controls, although these differences were not statistically significant.

Hyperspectral Analysis of Retinal Reflectance

In both the blue (HSIb) and green (HSIg) spectral bands, HSI values showed a strong dependence on the ETDRS zone but did not differ significantly between clinical subgroups (Fig. 3). As shown in Table 4, HSIb values were significantly higher in the IR (P < 0.001) and OR (P < 0.001) compared to the central subfield (C). Similarly, for HSIg, values were significantly higher at CC (P = 0.002), IR (P < 0.001), and OR (P < 0.001) than at C (Table 5). No significant differences were detected in any ETDRS fields between clinical subgroups, and neither age nor sex was significantly associated with HSI values in the model.

Figure 3.

Distribution of HSIb and HSIg across ETDRS zones in Ctrl, MSnON, and MSON. Panels show mean HSIb (left) and HSIg (right) across the C, CC, IR, and OR ETDRS regions. Boxes represent the interquartile range, the central line denotes the median, whiskers indicate 1.5 × the interquartile range, and points represent outliers. C, center.

Table 4.

Linear Mixed-Effects Model Examining the Association Between MS Subdiagnosis and HSIb

HSIb	Beta	95% CI	P Value
Subdiagnosis
MSnON	−0.034	−0.094, 0.025	0.256
MSON	0.039	−0.037, 0.115	0.311
ETDRS zone
CC	−0.014	−0.044, 0.017	0.370
IR	0.060	0.030, 0.091	<0.001
OR	0.055	0.025, 0.086	<0.001
Age	0.001	−0.001, 0.003	0.160
Female sex	0.000	−0.046, 0.045	0.997
Subdiagnosis * ETDRS zone
MSnON * CC	0.022	−0.035, 0.079	0.447
MSON * CC	−0.010	−0.085, 0.065	0.801
MSnON * IR	0.022	−0.035, 0.079	0.453
MSON * IR	−0.028	−0.103, 0.047	0.468
MSnON * OR	0.022	−0.035, 0.080	0.441
MSON * OR	−0.008	−0.083, 0.067	0.834

Control participants served as the reference group for subdiagnosis.

Bold values indicate statistically significant differences (P < 0.05).

Table 5.

Linear Mixed-Effects Model Examining the Association Between MS Subdiagnosis and HSIg

HSIg	Beta	95% CI	P Value
Subdiagnosis
MSnON	−0.037	−0.127, 0.053	0.419
MSON	0.056	−0.058, 0.171	0.331
ETDRS zone
CC	0.056	0.021, 0.091	0.002
IR	0.163	0.128, 0.198	<0.001
OR	0.153	0.119, 0.188	<0.001
Age	0.000	−0.003, 0.004	0.758
Female sex	−0.001	−0.076, 0.074	0.987
Subdiagnosis * ETDRS zone
MSnON * CC	0.007	−0.058, 0.073	0.827
MSON * CC	−0.014	−0.100, 0.072	0.746
MSnON * IR	0.003	−0.062, 0.069	0.917
MSON * IR	−0.029	−0.115, 0.057	0.504
MSnON * OR	0.004	−0.062, 0.070	0.903
MSON * OR	0.008	−0.078, 0.094	0.854

Control participants served as the reference group for subdiagnosis.

Bold values indicate statistically significant differences (P < 0.05).

Sample Size and Power Considerations for Future Studies

As this study was designed as a feasibility investigation rather than a formally powered hypothesis-testing study, a post-hoc power analysis was performed to characterize the magnitude of effects that could have been detected with the available sample sizes (MPOD analysis: Ctrl 36 eyes, MSnON 15 eyes, MSON 6 eyes; α = 0.05, power = 0.80, two-sided). The minimum detectable Cohen's d was approximately 1.24 for the Ctrl–MSON comparison and 0.86 for Ctrl–MSnON. The observed Cohen's d for MPOD at the foveal center (0°) was 0.64 for Ctrl versus MSON, indicating the present sample was underpowered to detect this effect size with confidence.

For the HSI analysis (Ctrl 46 eyes, MSnON 18 eyes, MSON 9 eyes; α = 0.05, power = 0.80, two-sided), the minimum detectable Cohen's d was approximately 1.02 for the Ctrl–MSON comparison and 0.78 for Ctrl–MSnON. The observed Cohen's d for HSIb at the central zone was 0.23 for Ctrl versus MSON and 0.26 for Ctrl versus MSnON; for HSIg the corresponding values were 0.22 and 0.21, respectively. These small observed effect sizes fall well below the minimum detectable thresholds, indicating that the HSI analysis was substantially underpowered to detect differences of the magnitude observed between clinical subgroups. The null HSI findings should therefore be interpreted with caution and do not exclude the possibility of clinically meaningful reflectance differences detectable in larger, adequately powered samples. These findings are presented as hypothesis-generating.

Exploratory Correlation Analyses

To explore whether MPOD and retinal reflectance are associated with disease duration and structural retinal changes (OCT and OCTA metrics), exploratory analyses were performed (Supplementary Tables S3–S9). Neither MPOD nor HSI was significantly associated with disease duration in MS participants (Supplementary Table S3; all P > 0.30). Using single representative eccentricities matched to each ETDRS sector, with MPOD z-score standardized within each eccentricity, MPOD at the outer ring (6°, SD = 0.016 OD) was positively associated with full retinal thickness (β = +9.40 µm/SD, 95% CI [3.87, 14.93], P < 0.001) and inner retinal thickness (β = +9.76 µm/SD, 95% CI [4.42, 15.11], p < 0.001) at the corresponding outer ring zone; MPOD at the inner ring (3°, SD = 0.046 OD) was also positively associated with inner retinal thickness (β = +5.80 µm/SD, 95% CI [0.36, 11.23], P = 0.037) (Supplementary Table S4). No significant associations were observed between MPOD and outer retinal thickness, or between HSI and full or inner retinal thickness. HSIb and HSIg were negatively associated with outer retinal thickness at the inner and outer ETDRS rings (β = −3.8 to −6.0 µm/RR unit, P = 0.012–0.050; Supplementary Table S5). Neither MPOD (at 1° for CC; at 3° for IR) nor HSI (CC paired with OCTA CC; IR paired with the mean of OCTA IR and OCTA OR) was significantly associated with SVC capillary density (Supplementary Tables S6–S8). MPOD was not significantly associated with FAZ area or volume at any zone after using single eccentricities (Supplementary Table S6). Higher HSI at the central circle was associated with smaller FAZ volume for both spectral bands (HSIb: β = −0.026 mm³/RR unit, P = 0.004; HSIg: β = −0.011 mm³/RR unit, P = 0.015; Supplementary Table S9). All correlation analyses were exploratory and results should be interpreted with caution given the small sample size.

Discussion

This feasibility study evaluated two complementary retinal signatures, MPOD and RR, as candidate non-invasive biomarkers of smoldering disease activity in MS. The principal signal was a change in the spatial distribution of MPOD, with ring-like profiles enriched in MS and most prominent in eyes with prior optic neuritis (MSON). In contrast, neither mean MPOD across eccentricities nor RR derived from three-wavelength multicolor imaging differentiated MS from controls. MPOV showed nosignificant differences in MSON eyes, consistent with the MPOD findings. Collectively, these findings indicate that the organization of macular pigment may be more sensitive to MS-related macular change than the absolute amount of pigment, and that coarse spectral sampling is unlikely to capture subtle reflectance signatures. The present findings should be interpreted in the context of the exploratory nature of this study. Although prior studies have suggested alterations in MPOD in MS, our results using an objective autofluorescence-based approach did not demonstrate significant group differences. This may reflect methodological differences, but also highlights that AF-based MPOD measurement may capture different aspects of macular pigment biology compared with psychophysical techniques.

Notably, this is the first study to assess MPOD and RR in MS. Using two-wavelength autofluorescence imaging we identify a distinct MPOD spatial profile associated with MS. This profile is evident in both MSnON and MSON eyes. The findings should be considered hypothesis-generating and interpreted with appropriate caution.

Macular Pigment Optical Density and Volume

The enrichment of ring-like profiles suggests redistribution of macular pigment rather than uniform depletion. Because macular carotenoids localize predominantly to the Henle fiber layer and adjacent inner plexiform layers, alterations in foveal architecture and microvasculature—including a larger FAZ—are plausible drivers of the observed pattern.^18,19 These observations align with MS literature reporting inner retinal injury related to ON³ and outer retinal/photoreceptor involvement independent of ON, consistent with the concept that the retina can mirror smouldering neurodegenerative processes in MS. Spatial MPOD profiles are considered relatively stable within individuals and less diet-dependent than absolute MPOD, supporting their potential utility as robust readouts in small or heterogeneous cohorts.

MPOV reinforces the MPOD magnitude trend, with MSON eyes showing the lowest median total pigment. Although the Kruskal–Wallis test did not reach significance, primarily due to the small MSON group, the direction of the effect is consistent across both MPOD and MPOV analyses. The dissociation between altered spatial distribution and unchanged mean MPOD is biologically coherent: spatial shifts in carotenoids between the foveal center and parafoveal annuli can occur with minimal change in total pigment.²⁰ MPOD is known to be influenced by dietary intake of carotenoids, particularly lutein and zeaxanthin. Variations in dietary habits and supplementation can therefore contribute to interindividual variability in MPOD measurements. However, because dietary intake was not specifically controlled or assessed in the present study, its potential impact cannot be ruled out. Importantly, such effects are expected to introduce variability rather than systematic bias between groups, although this remains a limitation to consider when interpreting the findings.

It is important to note that MPOD is an optical surrogate, not a direct biochemical measure of oxidative stress. It reflects macular carotenoid distribution, which may relate indirectly to antioxidant capacity. Direct measurement of oxidative stress markers (e.g., fundus flavoprotein fluorescence or QUEST OCT) would be required to confirm any link between MPOD changes and oxidative pathology in MS and is proposed as a direction for future work.

Methodological Considerations for MPOD Quantification

Our study derived MPOD from two-wavelength autofluorescence imaging (BAF and GAF) using the Heidelberg Spectralis platform, whereas prior MS studies.^5–7 employed HFP. These methods assess related but not identical constructs. HFP is a psychophysical technique that relies on subjective flicker nulling and thus depends on participant performance, and may be influenced by factors such as attention, fixation stability, and dysfunction along the visual pathway, including neurological impairment. In contrast, two-wavelength autofluorescence is an objective imaging-based method that quantifies macular pigment. Differences in response dependence, spatial sampling across retinal eccentricities, and susceptibility to visual pathway dysfunction may plausibly contribute to discrepancies between our findings and those of prior studies.

Formal test-retest repeatability was not assessed in this single-visit study, and this is acknowledged as a limitation. However, AF-based MPOD measurement demonstrated good repeatability and reproducibility, as established in prior validation studies.^10–12 To minimize variability, a standardized acquisition protocol on the same Heidelberg Spectralis platform was used, with rigorous quality control. Spatial MPOD profiles are reported to be relatively stable within individuals and reproducible across measurements.²⁰ Profile classification was performed masked to clinical subgroup and independently reviewed by two investigators.

It is important to note that HFP and two‑wavelength autofluorescence (AF) differ fundamentally in how MPOD is measured. HFP is a psychophysical technique that relies on a subject's perception to detect flicker nulling, making it dependent on attention, fixation stability, and cognitive performance, factors that may be altered in neurological conditions such as multiple sclerosis. In contrast, AF is an objective, imaging-based method that quantifies MPOD by comparing fundus autofluorescence at two excitation wavelengths, yielding spatially resolved maps that are less influenced by subject cooperation. Although HFP has been widely used and may be sensitive to functional changes, AF offers improved reproducibility and the ability to assess spatial distribution, though methodological differences between the techniques may lead to discrepancies in absolute values and disease-related findings. The absence of significant differences is unlikely to be explained solely by measurement variability, because two-wavelength autofluorescence has demonstrated good repeatability and reproducibility in previous validation studies.

Interpretation of Retinal Reflectance Imaging Findings

RR analysis yielded no group differences. Our approach estimated “hyperspectral” behavior from three discrete wavelengths (blue, green, and infrared) rather than true hyperspectral sampling. In neurodegenerative disease, dedicated hyperspectral cameras sampling narrowly across the visible spectrum have detected reflectance alterations that may index protein aggregation and inflammation.^16,21,22 The absence of differences here does not exclude RR-based biomarkers in MS; rather, it underscores the need for higher spectral resolution or complementary high-resolution modalities.^23,24 Retinal reflectance is retained in this report to provide feasibility data informing future hyperspectral imaging studies in MS.

Correlation Analysis

The exploratory correlation analyses suggest a complex spatial relationship between MPOD and retinal structure. MPOD at peripheral eccentricities (3° and 6°) was positively associated with inner and full retinal thickness at the corresponding ETDRS zones, consistent with the notion that eyes with higher overall macular carotenoid status retain more structural integrity in the peripheral macula. HSI at the central circle was inversely associated with foveal avascular zone volume, a finding that may reflect co-localization of macular carotenoids with regions of healthy vascular perfusion, though directionality cannot be established cross-sectionally. These findings are preliminary and require replication in adequately powered studies. This study provides initial correlation data linking MPOD and HSI with OCT, OCTA, and FAZ metrics; the associations identified here require confirmation in larger, prospective cohorts.

Study Limitations and Future Directions

This study has several limitations that should be considered. This single-center, cross-sectional feasibility study was modest in size, which limits statistical power and cannot establish causality. The post hoc power analysis confirmed that the study was powered to detect only moderate-to-large effects. Group imbalance, particularly the small MSON group, limits comparisons involving this subgroup and should be addressed in future studies. Formal test-retest repeatability was not assessed in this single-visit study. In addition, dietary carotenoid intake, which is known to influence macular pigment optical density, was not controlled for or assessed and may have contributed to inter-individual variability. Because of the limited sample size, definitive, well-powered multimodal correlation analyses were beyond the scope of this feasibility study. Accordingly, the study should be considered exploratory, focusing on the feasibility and characteristics of AF-based MPOD measurement in MS rather than definitive biomarker validation. Finally, methodological differences between autofluorescence-based and psychophysical MPOD measurements may complicate direct comparisons with prior studies.

Future work should integrate: dietary questionnaires and serum carotenoid measurements; phenotypic stratification (RRMS, SPMS; disease duration; EDSS; relapse history)³; multimodal imaging including OCT layer metrics, OCTA (FAZ and perfusion), and adaptive optics where relevant^18,23–25; and direct measures of oxidative stress such as fundus flavoprotein fluorescence (FPF) or QUEST OCT. For reflectance, studies employing true hyperspectral or snapshot multispectral cameras are warranted.^16,21,22 Longitudinal designs should test whether MPOD spatial profiles predict smouldering-associated worsening or structural progression.^1,3,4 Confirmation of these findings will require larger, prospectively powered, multicenter studies with balanced cohorts and longitudinal follow-up.

Conclusions

This hypothesis-generating feasibility study demonstrates that MPOD spatial distribution, more than density or volume, emerges as a promising retinal marker candidate for MS. The distinct MPOD spatial profile associated with MS is evident in both MSnON and MSON eyes. MPOD spatial architecture is rapid to acquire, noninvasive, and may index outer retinal/Henle-layer organization and neurovascular status not readily captured by standard OCT alone. As such, it could complement inner retinal thickness metrics³ and add value to retinal oculomics frameworks in neurodegeneration. These findings are intended as a foundation for adequately powered, hypothesis-driven future investigations.