Nasal‐Predominant Ganglion Cell‐Inner Plexiform Layer Thinning Distinguishes Early‐Stage Leber Hereditary Optic Neuropathy From Acute Optic Neuritis

ABSTRACT

Background

To compare acute‐phase macular ganglion cell–inner plexiform layer (GCIPL) thickness on optical coherence tomography (OCT) between Leber hereditary optic neuropathy (LHON) and optic neuritis (ON), and to evaluate the diagnostic performance of nasal GCIPL thickness.

Methods

In this retrospective single‐centre study, OCT (CIRRUS HD‐OCT; Carl Zeiss Meditec, Dublin) acquired within 45 days of symptom onset was analysed. Sectoral GCIPL thickness was compared between LHON and ON—including double‐seronegative (DN‐ON), aquaporin‐4 antibody–positive (AQP4‐ON) and myelin oligodendrocyte glycoprotein antibody–positive (MOG‐ON) subtypes—using linear mixed‐effects models adjusted for age, sex and onset‐to‐OCT interval. Receiver operating characteristic (ROC) analyses evaluated nasal GCIPL thickness (average of superonasal and inferonasal sectors) alone and combined with clinical covariates.

Results

LHON showed significantly thinner GCIPL than ON in the superonasal (adjusted mean difference −12.1 μm) and inferonasal (−12.4 μm) sectors in the acute phase (FDR‐adjusted q = 0.02 and 0.01, respectively). The nasal GCIPL difference was evident versus DN‐ON and MOG‐ON, whereas separation from AQP4‐ON was less apparent. Nasal GCIPL thickness alone discriminated LHON from ON with an area under the curve (AUC) of 0.88 (Youden‐optimal cut‐off 72.5 μm; sensitivity 0.93, specificity 0.82); a multivariable model including age, sex and onset‐to‐OCT category increased the AUC to 0.94.

Conclusion

Acute‐phase OCT reveals nasal‐predominant GCIPL thinning in LHON. This pattern differentiates LHON from DN‐ON and MOG‐ON but is less distinct from AQP4‐ON. Nasal GCIPL thickness, alone or combined with clinical covariates, may aid differential diagnosis in acute presentations.

1. Introduction

Leber hereditary optic neuropathy (LHON) is a mitochondrial DNA‐associated inherited optic neuropathy that involves preferential degeneration of retinal ganglion cells (RGCs) [1]. Clinically, LHON often presents with acute, painless visual loss in young adults. In the early phase, it may mimic idiopathic optic neuritis (ON) or antibody‐associated ON, which can complicate the initial differential diagnosis. Management differs between these conditions: high‐dose corticosteroids and plasma exchange are established first‐line treatments for ON, whereas these interventions generally do not improve vision in LHON. Early distinction is therefore important [2].

Optical coherence tomography (OCT) provides quantitative structural measures in optic nerve disorders. Macular ganglion cell–inner plexiform layer (GCIPL) thickness and circumpapillary retinal nerve fibre layer (cpRNFL) thickness capture changes at the level of ganglion cell bodies and axons. In acute LHON, progressive GCIPL thinning (especially on the nasal side) with cpRNFL swelling, sometimes described as pseudoedema, has been reported [3, 4]. This differs from acute ON, in which inflammatory oedema may predominate. However, few studies have directly compared LHON with double‐negative ON (DN‐ON) under a standardised early imaging window [5]. Unlike prior early‐phase OCT studies limited by narrow comparator groups, we evaluated subtype‐specific contrasts (DN‐ON, aquaporin‐4 antibody–associated ON [AQP4‐ON] and myelin oligodendrocyte glycoprotein antibody–associated ON [MOG‐ON]) within the same early imaging window to improve clinical interpretability.

The purpose of this study was to quantitatively compare sectoral GCIPL thickness between LHON and ON subtypes (DN‐ON, AQP4‐ON and MOG‐ON) using OCT obtained within 45 days of onset. We sought to clarify whether early nasal‐sector GCIPL thinning can help distinguish acute LHON from ON in routine clinical practice.

2. Methods

This single‐centre retrospective observational study included consecutive patients evaluated at our institution between January 2016 and March 2025 who were diagnosed with LHON with the m.11778G>A mutation (n = 14), DN‐ON (n = 24), AQP4‐ON (n = 18) or MOG‐ON (n = 18) (Figure 1). All participants underwent OCT within 45 days of symptom onset. Time from onset to OCT was defined as the interval from symptom onset to OCT acquisition. Clinical data, including visual acuity and OCT measurements, were extracted from medical records and institutional examination databases. For ON cases, the analysed eye corresponded to the acute episode under evaluation. Only eyes presenting with a first episode of ON were included.

FIGURE 1.

Flowchart of patient selection. Patients diagnosed with Leber hereditary optic neuropathy (LHON; m.11778G>A mutation), double‐negative optic neuritis (DN‐ON), aquaporin‐4 antibody–positive optic neuritis (AQP4‐ON), or myelin oligodendrocyte glycoprotein antibody–positive optic neuritis (MOG‐ON) who underwent optical coherence tomography (OCT) within 45 days of onset between January 2016 and March 2025 were identified. Only eyes presenting with a first episode of optic neuritis were included. In each group, patients with other ocular comorbidities or high myopia (spherical equivalent ≤ −6.0 dioptres) were excluded. The final analysis included 64 patients (71 eyes): 11 patients (15 eyes) with LHON, 20 patients (20 eyes) with DN‐ON, 18 patients (18 eyes) with AQP4‐ON and 15 patients (18 eyes) with MOG‐ON.

A diagnosis of LHON was made based on characteristic clinical findings and confirmed by detection of the m.11778G>A mutation [6]. AQP4‐ON and MOG‐ON were diagnosed according to the 2015 International Consensus Criteria [7] and the 2023 International MOGAD Panel Criteria [8], respectively, with seropositivity for the corresponding antibody and compatible clinical features. DN‐ON was defined as ON with negative results for both AQP4 and MOG antibodies after excluding alternative aetiologies such as infectious, granulomatous, or systemic autoimmune diseases, including sarcoidosis [9]. Patients were excluded if they had ocular conditions that could affect OCT measurements (e.g., glaucoma or retinal disease) or high myopia, defined as spherical equivalent ≤ −6.0 dioptres. The study was approved by the Institutional Review Board of our hospital (approval number: 202601‐2) and adhered to the tenets of the Declaration of Helsinki. Given the retrospective design, consent was obtained using an opt‐out approach via the hospital website. All data were anonymised before analysis.

OCT was performed using the CIRRUS HD‐OCT system (Carl Zeiss Meditec, Dublin). For GCIPL analysis, automated segmentation was applied to fovea‐centred 6 × 6 mm Macular Cube scans to obtain thickness values for the six sectors of the ETDRS inner ring: superior, superonasal, inferonasal, inferior, inferotemporal and superotemporal (Figure S1). These sectors were prespecified as the primary regions of interest. For cpRNFL assessment, the Optic Disc Cube 200 × 200 protocol was used. The device software calculated average and quadrant cpRNFL thickness (superior, inferior, temporal and nasal) along a 3.46‐mm‐diameter calculation circle centred on the optic disc. All OCT scans were reviewed for image quality and obvious segmentation failure. Scans with signal strength < 7, severe motion artefact, decentration, or clear segmentation error affecting GCIPL or cpRNFL measurements were excluded. Because manual re‐segmentation is not available in the CIRRUS platform, only scans judged qualitatively acceptable on automated segmentation output were included.

The primary outcome was the between‐group difference in sectoral GCIPL thickness on OCT obtained within 45 days of onset, comparing LHON with the overall ON group. To account for inclusion of both eyes, we used linear mixed‐effects models with a patient‐level random intercept. The dependent variable was sectoral GCIPL thickness, and the main exposure was disease group (LHON vs. ON). Models were adjusted for age (continuous), sex and time from onset to OCT, categorised as 0–7, 8–21 and 22–45 days. Regression coefficients are presented as adjusted mean differences (μm) with the ON group as the reference. In key secondary analyses, LHON was compared with each ON subtype (DN‐ON, AQP4‐ON and MOG‐ON) using the same modelling framework. For sensitivity analyses, we restricted the dataset to one eye per patient and fitted general linear models with the same covariates; for bilateral cases, the eye with the shorter onset‐to‐OCT interval was selected, and when intervals were identical, the right eye was used. Multiple testing across the six GCIPL sectors was addressed using false discovery rate control with the Benjamini–Hochberg procedure, and adjusted q values are reported. No missing data were present for the primary OCT variables; analyses were conducted using complete cases.

ROC analyses were performed to evaluate discrimination between LHON and ON using nasal GCIPL thickness. We generated ROC curves for nasal GCIPL thickness (average of superonasal and inferonasal) alone and for a multivariable logistic regression model including nasal GCIPL thickness, age, sex and time from onset to OCT; the model‐predicted probability of LHON was used as the classifier. AUCs with 95% confidence intervals (DeLong method) were calculated, and optimal cut‐offs were defined by the Youden index. We additionally evaluated the diagnostic performance of temporal cpRNFL thickness. ROC analyses were performed for temporal cpRNFL thickness alone and for a multivariable logistic regression model including temporal cpRNFL thickness, age, sex and time from symptom onset to OCT category. As an additional sensitivity analysis, we repeated the nasal GCIPL models after adding average cpRNFL thickness as a covariate to account for acute phase cpRNFL swelling.

Baseline characteristics are reported as median (range) for continuous variables and number (percentage) for categorical variables. BCVA was measured using a standard Japanese decimal visual acuity chart and converted to logMAR units using the formula logMAR = −log10(decimal VA). Counting fingers, hand motion and light perception were assigned logMAR values of 2.0, 2.3 and 2.6, respectively [10]. Statistical significance was defined as two‐sided p < 0.05. Analyses were performed in R (version 4.4.2; R Foundation for Statistical Computing, Vienna, Austria).

3. Results

Baseline characteristics are summarised in Table 1. We analysed 64 patients (71 eyes), including 11 patients (15 eyes) with LHON and 53 patients (56 eyes) with ON. The ON cohort comprised 20 patients (20 eyes) with DN‐ON, 18 patients (18 eyes) with AQP4‐ON and 15 patients (18 eyes) with MOG‐ON. The median age (range) was 43 years (27–56) overall and 27 years (17–44) in the LHON group versus 45 years (34–58) in the ON group. Median age (range) by ON subtype was 41 years (26–55) for DN‐ON, 56 years (42–65) for AQP4‐ON and 44 years (34–51) for MOG‐ON. Overall, 35 patients were male (55%) and 29 were female (45%); subgroup counts are provided in Table 1. The median interval (range) from onset to OCT was 10 days (6–16) overall, 20 days (17–42) for LHON, and 8 days (5–12) for ON. Median baseline BCVA (logMAR) was 1.10 (0.22–2.00) overall, 1.22 (0.40–2.00) for LHON and 1.02 (0.19–2.00) for ON. BCVA at 6 months is summarised in Table 1 for all groups.

TABLE 1.

Baseline patient characteristics.

	Overall, N = 64 (71 eyes)	LHON, N = 11 (15 eyes)	ON, N = 53 (56 eyes)	DN‐ON, N = 20 (20 eyes)	AQP4‐ON, N = 18 (18 eyes)	MOG‐ON, N = 15 (18 eyes)
Age (years)	43 (27–56)	27 (17–44)	45 (34–58)	41 (26–55)	56 (42–65)	44 (34–51)
Male, n (%)	35 (55%)	10 (91%)	25 (47%)	13 (65%)	5 (28%)	7 (47%)
Female, n (%)	29 (45%)	1 (9%)	28 (53%)	7 (35%)	13 (72%)	8 (53%)
Times to OCT measurement (days)	10 (6–16)	20 (17–42)	8 (5–12)	8 (6–11)	8 (6–16)	8 (3–13)
Initial BCVA, logMAR	1.10 (0.22–2.00)	1.22 (0.40–2.00)	1.02 (0.19–2.00)	1.00 (0.05–2.00)	2.00 (1.05–2.30)	0.35 (0.15–1.52)
Six months BCVA, logMAR	0.05 (−0.08–1.46)	2.00 (1.52–2.00)	0.00 (−0.08–0.10)	−0.08 (−0.08–0.22)	0.43 (0.10–1.00)	−0.08 (−0.08–0.00)

Note: Age, sex, days from onset to OCT imaging and best‐corrected visual acuity (BCVA; logMAR) at baseline and at 6 months are presented for the overall cohort and for each disease group (LHON, overall ON, DN‐ON, AQP4‐ON and MOG‐ON). Values are expressed as median (range) for age, days from onset to OCT imaging and BCVA and as number (percentage) for sex.

Abbreviations: AQP4‐ON = aquaporin‐4 antibody–positive optic neuritis, BCVA = best‐corrected visual acuity, DN‐ON = double‐negative optic neuritis (negative for both anti‐AQP4 and anti‐MOG antibodies), LHON = Leber hereditary optic neuropathy, MOG‐ON = myelin oligodendrocyte glycoprotein antibody–positive optic neuritis, OCT = optical coherence tomography, ON = optic neuritis.

Table 2 and Figure 2 shows sectoral differences in GCIPL thickness between LHON and ON. In linear mixed‐effects models with a patient‐level random intercept, LHON had thinner GCIPL than ON in the superonasal and inferonasal sectors. The adjusted mean differences were −12.11 (95% CI, −20.42 to −3.80) and −12.39 μm (95% CI, −19.44 to −5.33), with FDR‐adjusted q values of 0.02 and 0.01, respectively. No other sector differed significantly after FDR correction. In sensitivity analyses restricted to one eye per patient, the superonasal and inferonasal sectors remained significant (q = 0.03 and 0.01), while the other sectors did not.

TABLE 2.

Differences in sector‐specific GCIPL thickness between LHON and ON.

GCIPL sector	LHON vs. ON, mixed‐effects model (both eyes)				LHON vs. ON, sensitivity analysis (one eye per patient)
	Estimate (μm)	95% CI	p	q values (FDR)	Estimate (μm)	95% CI	p	q values (FDR)
Superior	−1.46	−8.36 to 5.44	0.68	0.82	−0.98	−8.13 to 6.17	0.78	0.78
Superonasal	−12.11	−20.42 to −3.80	0.01	0.02	−11.94	−20.71 to −3.17	0.01	0.03
Inferonasal	−12.39	−19.44 to −5.33	0	0.01	−12.49	−19.89 to −5.09	0	0.01
Inferior	−6.21	−12.46 to 0.03	0.06	0.11	−6.14	−12.59 to 0.32	0.06	0.12
Inferotemporal	−3.62	−9.32 to 2.08	0.22	0.33	−3.28	−9.07 to 2.51	0.26	0.39
Superotemporal	0.53	−5.06 to 6.12	0.85	0.85	1.52	−3.86 to 6.89	0.57	0.69

Note: Differences in macular GCIPL thickness between LHON and ON are presented for each sector. The left columns show estimates derived from linear mixed‐effects models including random intercepts for each patient to account for inter‐eye correlation (mixed‐effects model, both eyes). The right columns show results from the sensitivity analysis including only one eye per patient (sensitivity analysis, one eye per patient). Estimates represent the adjusted mean difference (μm) calculated as LHON minus ON; negative values indicate thinner GCIPL in the LHON group.

Abbreviations: 95% CI = 95% confidence interval, p = significance probability; q values (FDR) = values after false discovery rate correction for multiple comparisons.

FIGURE 2.

Forest plots comparing sector‐specific macular GCIPL thickness between LHON and ON. (A) Adjusted mean differences (μm) and 95% confidence intervals (CIs) for each macular GCIPL sector between LHON and ON are presented. These estimates were derived from linear mixed‐effects models including random intercepts for each patient to account for inter‐eye correlation. (B) Corresponding adjusted mean differences and 95% CIs from the sensitivity analysis, which included only one eye per patient, are also shown. The horizontal axis represents the difference calculated as LHON minus ON; negative values indicate thinner GCIPL in the LHON group. All models were adjusted for age, sex and days from onset to OCT imaging. The dashed vertical line denotes a difference of 0 μm.

Figure 3 summarises comparisons between LHON and ON subtypes versus DN‐ON, LHON had thinner GCIPL in the superonasal and inferonasal sectors, with adjusted mean differences of −10.41 (95% CI, −16.34 to −4.47; FDR‐adjusted q = 0.01) and −10.28 μm (95% CI, −16.36 to −4.21; q = 0.01), respectively. The inferior sector showed a smaller difference (−5.61 μm; 95% CI, −11.18 to −0.05), which did not remain significant after FDR correction (q = 0.12). No sector differed significantly between LHON and AQP4‐ON after FDR adjustment versus MOG‐ON, LHON again showed thinner GCIPL in the superonasal and inferonasal sectors, with adjusted mean differences of −16.46 (95% CI, −27.36 to −5.55; q = 0.03) and −14.39 μm (95% CI, −24.39 to −4.38; q = 0.03), respectively; no other sector differed significantly (Table 3). Analyses restricted to one eye per patient showed the same nasal‐predominant pattern versus DN‐ON and MOG‐ON, with no significant differences versus AQP4‐ON (Table 4).

FIGURE 3.

Forest plots comparing sector‐specific macular GCIPL thickness between LHON and optic neuritis subtypes. Adjusted mean differences (μm) and 95% confidence intervals (CIs) for each macular GCIPL sector are presented for comparisons between (A) LHON and DN‐ON, (B) LHON and AQP4‐ON and (C) LHON and MOG‐ON. Estimates were derived from linear mixed‐effects models including random intercepts for each patient to account for inter‐eye correlation, adjusted for age, sex and days from onset to OCT imaging. The horizontal axis represents the difference calculated as LHON minus the respective ON subtype; negative values indicate thinner GCIPL in the LHON group. Dots represent adjusted mean differences, horizontal lines represent 95% CIs and the dashed vertical line denotes a difference of 0 μm.

TABLE 3.

Comparisons of sector‐specific GCIPL thickness between LHON and optic neuritis subtypes (linear mixed‐effects models).

GCIPL sector	LHON vs. DNON				LHON vs. AQP4				LHON vs. MOG
	Estimate	95% CI	p	q value (FDR)	Estimate	95% CI	p	q value (FDR)	Estimate	95% CI	p	q value (FDR)
Superior	−2.82	−8.74 to 3.10	0.34	0.39	3.27	−9.84 to 16.38	0.61	0.85	0.1	−7.23 to 7.43	0.98	0.98
Superonasal	−9.93	−15.92 to −3.95	0	0.01	−2.18	−14.82 to 10.45	0.73	0.85	−13.57	−24.82 to −2.32	0.02	0.07
Inferonasal	−9.85	−15.91 to −3.79	0	0.01	−4.91	−15.22 to 5.40	0.34	0.79	−12.47	−22.43 to −2.52	0.02	0.07
Inferior	−6.1	−11.77 to −0.44	0.04	0.06	−6.89	−18.43 to 4.64	0.23	0.79	−2.48	−9.12 to 4.16	0.45	0.74
Inferotemporal	−3.75	−9.60 to 2.10	0.2	0.28	−5.45	−14.98 to 4.07	0.25	0.79	−1.56	−8.16 to 5.05	0.63	0.74
Superotemporal	0.83	−6.47 to 8.14	0.82	0.82	−0.46	−11.47 to 10.54	0.93	0.93	2.07	−6.53 to 10.66	0.63	0.74
Average	−5.49	−10.60 to −0.38	0.04	0.06	−2.36	−10.94 to 6.23	0.58	0.85	−3.32	−10.86 to 4.21	0.37	0.74

Note: Differences in macular GCIPL thickness between LHON and DN‐ON, AQP4‐ON and MOG‐ON are presented for each sector. Estimates were derived from linear mixed‐effects models including random intercepts for each patient to account for inter‐eye correlation. Estimates represent the adjusted mean difference (μm) calculated as LHON minus the respective ON subtype; negative values indicate thinner GCIPL in the LHON group.

Abbreviations: 95% CI = 95% confidence interval, AQP4‐ON = aquaporin‐4 antibody–positive optic neuritis, DN‐ON = double‐negative optic neuritis (negative for both anti‐AQP4 and anti‐MOG antibodies), GCIPL = ganglion cell–inner plexiform layer, MOG‐ON = myelin oligodendrocyte glycoprotein antibody–positive optic neuritis, p = significance probability, q value (FDR) = value after false discovery rate correction for multiple comparisons.

TABLE 4.

Comparisons of sector‐specific GCIPL thickness between LHON and optic neuritis subtypes including only one eye per patient (sensitivity analysis).

GCIPL sector	LHON vs. DN‐ON, sensitivity analysis (one eye per patient)				LHON vs. AQP4‐ON, sensitivity analysis (one eye per patient)				LHON vs. MOG‐ON, sensitivity analysis (one eye per patient)
	Estimate	95% CI	p	q value (FDR)	Estimate	95% CI	p	q value (FDR)	Estimate	95% CI	p	q value (FDR)
Superior	−2.43	−8.65 to 3.80	0.43	0.52	3.71	−10.49 to 17.92	0.59	0.82	−0.67	−7.98 to 6.63	0.85	0.85
Superonasal	−10.62	−17.10 to −4.15	0	0.01	−2.78	−16.65 to 11.09	0.68	0.82	−15.96	−28.05 to −3.87	0.01	0.05
Inferonasal	−10.35	−16.96 to −3.73	0	0.01	−5.39	−16.70 to 5.92	0.33	0.67	−14.07	−25.28 to −2.86	0.02	0.05
Inferior	−5.96	−11.77 to −0.14	0.05	0.09	−7.08	−19.51 to 5.35	0.25	0.67	−3.27	−10.19 to 3.65	0.34	0.67
Inferotemporal	−3.27	−9.43 to 2.90	0.29	0.43	−5.54	−15.64 to 4.55	0.27	0.67	−1.89	−8.55 to 4.77	0.56	0.78
Superotemporal	1.31	−5.42 to 8.03	0.69	0.69	0.11	−10.09 to 10.30	0.98	0.98	1.62	−5.70 to 8.93	0.65	0.78

Note: Differences in macular GCIPL thickness between LHON and DN‐ON, AQP4‐ON and MOG‐ON are presented for each sector in a monocular cohort including only one eye per patient. Comparisons were performed using general linear models adjusted for age, sex and days from onset to OCT imaging (categorised into three intervals). Estimates represent the adjusted mean difference (μm) calculated as LHON minus the respective ON subtype; negative values indicate thinner GCIPL in the LHON group.

Abbreviations: 95% CI = 95% confidence interval, p = significance probability, q value (FDR) = value after false discovery rate correction using the Benjamini–Hochberg procedure.

ROC analyses were performed to assess the ability of nasal macular GCIPL thickness to differentiate LHON from ON (Figure 4). Using nasal GCIPL thickness alone, the AUC was 0.877 (95% CI, 0.778–0.976). The Youden‐optimal cut‐off was 72.5 μm, providing a sensitivity of 0.933 and specificity of 0.821. We then constructed a multivariable logistic regression model including nasal GCIPL thickness, age, sex and time from onset to OCT category. ROC analysis based on the model‐predicted probability of LHON showed improved discrimination (AUC, 0.943; 95% CI, 0.877–1.000). At the Youden‐optimal probability threshold of 0.518, sensitivity was 0.800 and specificity was 0.982.

FIGURE 4.

Receiver operating characteristic (ROC) curves for differentiating Leber hereditary optic neuropathy (LHON) from optic neuritis (ON). (A) ROC curve using nasal macular GCIPL thickness alone. The area under the curve (AUC) was 0.877 (95% CI, 0.778–0.976). The Youden‐optimal cut‐off was 72.5 μm (LHON defined as nasal GCIPL thickness ≤ 72.5 μm), yielding a sensitivity of 0.933 and specificity of 0.821. (B) ROC curve derived from a multivariable logistic regression model including nasal GCIPL thickness, age, sex and time from onset to OCT category (categorised as 0–7, 8–21 and 22–45 days). The ROC was constructed using the model‐predicted probability of LHON. The AUC was 0.943 (95% CI, 0.877–1.000). The Youden‐optimal probability threshold was 0.518 (LHON defined as predicted probability ≥ 0.518), yielding a sensitivity of 0.800 and specificity of 0.982. Black circles indicate Youden‐optimal thresholds; the diagonal line indicates chance‐level discrimination.

We additionally evaluated temporal cpRNFL thickness (Figure S2). Temporal cpRNFL thickness alone showed limited discrimination between LHON and ON, with an AUC of 0.648 (95% CI, 0.496–0.800). The Youden‐optimal cut‐off was 92.5 μm, yielding a sensitivity of 0.667 and a specificity of 0.732. A multivariable logistic regression model including temporal cpRNFL thickness, age, sex and onset‐to‐OCT category showed improved overall discrimination, with an AUC of 0.923 (95% CI, 0.835–1.000). At the Youden‐optimal probability threshold of 0.368, sensitivity was 0.800 and specificity was 0.929. However, temporal cpRNFL thickness itself was not a significant independent predictor in the multivariable model, suggesting that the improved discrimination reflected the combined contribution of clinical covariates rather than temporal cpRNFL thickness alone.

Unadjusted descriptive OCT measures within 45 days are presented in Table S1. The LHON group showed higher cpRNFL thickness than the overall ON group, most notably in the superior, inferior and temporal quadrants (e.g., average: 125 [107–131] vs. 110 [97–119] μm). In contrast, GCIPL thickness was lower in the nasal sectors in LHON (superonasal 70 [67–72] and inferonasal 65 [63–68] μm) than in ON (82 [77–87] and 79 [76–83] μm), and average GCIPL thickness was also lower (73 [70–75] vs. 80 [74–83] μm). By subtype, GCIPL and cpRNFL were generally preserved in DN‐ON; average GCIPL was slightly reduced in AQP4‐ON; and superior cpRNFL tended to be higher in MOG‐ON. Overall, LHON showed a pattern of early cpRNFL thickening with nasal‐sector GCIPL thinning. In additional analyses with further adjustment for average cpRNFL thickness, nasal GCIPL remained thinner in LHON than in DN‐ON (adjusted mean difference, −9.45 μm; 95% CI, −15.29 to −3.60; p = 0.0026). In contrast, the differences versus AQP4‐ON (−4.35 μm; 95% CI, −15.71 to 7.02; p = 0.439) and versus MOG‐ON (−8.12 μm; 95% CI, −17.99 to 1.74; p = 0.103) were not statistically significant (Figure S3). These findings indicate that the difference versus MOG‐ON was attenuated after additional adjustment for average cpRNFL thickness.

4. Discussion

This study quantitatively compared sectoral GCIPL thickness between LHON and ON subtypes using linear mixed‐effects models applied to OCT obtained within 45 days of onset. LHON showed thinner GCIPL in the nasal sectors (superonasal and inferonasal) early after onset. This nasal‐predominant pattern remained after correction for multiple testing, both in the comparison with the overall ON group and in direct comparisons with DN‐ON and MOG‐ON. In contrast, no sector differed significantly between LHON and AQP4‐ON. The same pattern was observed in sensitivity analyses restricted to one eye per patient. In additional models adjusted for average cpRNFL thickness, the difference remained significant versus DN‐ON but was attenuated and no longer statistically significant versus MOG‐ON and AQP4‐ON. This attenuation may partly reflect differences in acute‐phase cpRNFL swelling, particularly in MOG‐ON, although the small sample size should also be considered. In exploratory ROC analyses, a nasal GCIPL threshold derived from our cohort (72.5 μm) provided good discrimination, while the multivariable model prioritised specificity when higher diagnostic confidence is required.

In acute LHON, cpRNFL swelling can occur while macular RGC degeneration progresses early [4]. Our results are consistent with this dissociation, with nasal‐sector GCIPL thinning despite relative cpRNFL thickening in LHON. The nasal predominance aligns with the anatomical distribution of the papillomacular bundle (PMB), which is preferentially affected in LHON [11, 12]. Because the PMB is composed of small‐calibre fibres that are vulnerable to mitochondrial dysfunction, early damage may be captured in the nasal macular sectors that contain the corresponding ganglion cell bodies. These findings are broadly consistent with the report by Zimmermann et al. [5], who described early inner macular thinning in LHON compared with DN‐ON. However, that study used Heidelberg OCT and evaluated ganglion cell layer thickness rather than GCIPL thickness; therefore, the results are not directly comparable, although they support the same general pattern of early nasal‐predominant inner retinal involvement in LHON. Our data extend this observation by showing that the nasal‐predominant pattern persists in subtype‐based comparisons that include DN‐ON and MOG‐ON.

Subtype‐specific comparisons may inform the differential diagnosis in the acute setting. The nasal‐sector GCIPL thinning characteristic of LHON was present in comparisons with DN‐ON and MOG‐ON, whereas no significant differences were observed versus AQP4‐ON. Prior work has shown that AQP4‐ON and MOG‐ON can be associated with substantial GCIPL and RNFL loss compared with multiple sclerosis–associated ON [13]. Despite a similar degree of structural thinning in some reports, AQP4‐ON is often associated with poorer visual outcomes and is considered a more severe phenotype [14]. As an astrocytopathy with intense inflammation and tissue injury, AQP4‐ON may cause rapid RGC damage early after onset, which could reduce detectable differences from LHON in this time window [15]. These findings should be interpreted cautiously given the limited sample size. Temporal cpRNFL thickening may also be a supportive acute‐phase OCT feature in LHON. However, in our additional analysis, its standalone discriminatory performance was lower than that of nasal GCIPL thickness, supporting nasal GCIPL thinning as the more informative OCT marker for differentiating LHON from ON.

Several limitations should be considered. First, this single‐centre retrospective study included relatively small numbers in each subgroup, which may limit precision, particularly for subtype comparisons. Second, OCT was assessed at a single acute‐phase time point within 45 days of symptom onset, and longitudinal changes in GCIPL and cpRNFL were not evaluated. Although nasal GCIPL thinning may begin early in LHON, the subsequent temporal evolution and stability of this pattern could not be assessed in the present study. Longitudinal OCT studies will be important to clarify how early nasal‐sector GCIPL thinning evolves over time and how it compares with dynamic cpRNFL changes in acute LHON and ON. Third, image‐related bias, including segmentation errors or reduced fixation due to visual loss, remains possible. Fourth, LHON cases were limited to patients with the m.11778G>A mutation. Therefore, whether the same nasal‐predominant GCIPL thinning pattern applies to other primary LHON mutations or rarer mitochondrial variants remains uncertain. Genotype‐specific differences in the susceptibility and topographic distribution of RGC injury may exist, and further studies including a broader spectrum of LHON genotypes will be needed. Larger multicentre cohorts with longitudinal OCT assessments would help address these limitations.

In conclusion, early OCT (within 45 days of onset) showed nasal‐predominant GCIPL thinning in LHON compared with ON. This pattern differentiated LHON from DN‐ON and was directionally similar versus MOG‐ON, while differences versus AQP4‐ON were less clear. Nasal GCIPL thinning with concurrent cpRNFL thickening may support recognition of acute LHON, warranting prospective validation and external assessment of diagnostic thresholds.

Funding

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Example of macular GCIPL measurement using ZEISS CIRRUS HD‐OCT.

Figure S2: ROC curves for additional analysis of temporal cpRNFL thickness in differentiating LHON from ON.

Figure S3: Difference in nasal GCIPL thickness between LHON and optic neuritis groups after additional adjustment for average cpRNFL thickness.

Table S1: Descriptive statistics for cpRNFL and GCIPL thickness measured by OCT within 45 days of onset.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.