Ganglion Cell Complex Thickness and Visual Function in Chronic Leber Hereditary Optic Neuropathy

Abstract

Purpose: To evaluate the correlation between the macular ganglion cell complex (GCC) thickness measured with manually corrected segmentation and visual function in individuals with chronic Leber hereditary optic neuropathy (LHON).

Methods: Twenty-six chronic LHON subjects (60% treated with idebenone or Q₁₀) from the Swedish LHON registry were enrolled. Best-corrected visual acuity (BCVA), visual field tests, and optical coherence tomography (OCT) were performed. Visual field was evaluated with the Haag-Streit Octopus 900 with the Esterman test and a custom 30° test. Canon OCT-HS100 scans were exported to the Iowa Reference Algorithm. GCC thickness was obtained after the segmentation was corrected manually in nine macular sectors.

Results: The GCC thickness was overestimated by 16 to 30 µm in different macular sectors with the automated segmentation compared with the corrected (P < 0.001). GCC thickness in all sectors showed significant correlation with all functional parameters. The strongest correlation was seen for the external temporal sector (BCVA: r = 0.604, P < 0.001; mean defect: r = 0.457, P = 0.001; Esterman score: r = 0.421, P = 0.003). No differences were seen between treated and untreated subjects with regard to GCC and visual field scores (P > 0.05), but BCVA was better among treated subjects (P = 0.017).

Conclusions: The corrected GCC thickness showed correlation with visual function in chronic LHON subjects. The frequently occurring segmentation errors in OCT measurements related to chronic LHON can potentially be misleading in monitoring of disease progression and in evaluating the treatment effects. Precise measurements of GCC could serve as a sensitive tool to monitor structural changes in LHON. We therefore emphasize the importance of careful evaluation of the accuracy of OCT segmentation.

Leber hereditary optic neuropathy (LHON) is a maternally inherited optic neuropathy driven by mitochondrial DNA (mtDNA) mutations^1–3 leading to mitochondrial complex I dysfunction and resulting in neuronal cell death in the retina. Approximately 1 in 30,000 to 50,000 people are affected with LHON, making it the most common inherited mitochondrial disease.^4,5 

LHON outbreaks typically start as unilateral impairment of central vision, and within a year this is typically bilateral.⁶ The visual loss results from a selective degeneration of retinal ganglion cells (RGCs), the output neurons of the retina, the axons of which make up the optic nerve. RGCs are highly metabolically active and have been well documented as vulnerable to metabolic insults and bioenergetic insufficiency.^7,8 Most individuals with LHON progress to severe visual impairment with central scotomas and a visual acuity of 1.0 logMAR or worse.⁹ In the acute phase of the disease, the characteristic findings include swollen optic disc, peripapillary telangiectatic blood vessels, and peripapillary retinal nerve fiber layer swelling. In the early phase, the smaller caliber fibers of the papillomacular bundle and ganglion cell layer (GCL) are primarily affected.^10–12 The GCL and RNFL thickness decreases during the months following the onset of vision loss.^13,14 LHON is currently an incurable metabolic optic neuropathy, with the only current treatment option being idebenone, a synthetic analog of coenzyme Q₁₀, which received marketing authorization from the European Commission for LHON treatment in September 2015.^15,16 Q₁₀ carries electrons from complexes I and II to complex III in the electron transport chain. Several therapeutic strategies are undergoing clinical trials, including gene therapies, which are an attractive option due to the accessibility of the RGCs in the retina and the success of gene therapies for other monogenetic Mendelian eye diseases.¹⁶ 

GCL thinning quantified with optical coherence tomography (OCT) is suggested to be a better structural predictor for visual function in LHON compared to other structural measurements.¹⁷ Another study showed that in LHON subjects a significant difference in the thickness of the macular GCL and inner plexiform layer (IPL) can be identified before seeing conclusive changes in peripapillary RNFL in the months following the onset of symptoms.¹⁸ In the context of analyzing the chronic stages of LHON, the structure of the RNFL and GCL layer (together with the IPL, making up the ganglion cell complex [GCC]) might be severely thinned. This severe thinning might lead to the floor effect—that is, the point at which no further thinning can be measured accurately with OCT.^19,20 The accuracy of the automated segmentation tool in OCTs can be affected due to morphological variations such as altered configuration of the individual layers or absence of certain layers.²¹ In LHON patients, the visual field defects, especially the central scotomas, are commonly assessed and monitored with the Humphrey Field Analyzer (Carl Zeiss Meditec, Jena, Germany). Previous studies have used a Swedish Interactive Thresholding Algorithm (SITA) strategy and Goldmann size III stimulus to assess the visual field defects in LHON patients and have demonstrated increased mean deviation that reaches the bottom end of the scale.^9,22,23 In order to increase the range and decrease the floor effect, the use of a Goldmann size IV stimulus has been proposed to enable a more sensitive and precise evaluation of the visual field.²⁴ In chronic LHON, due to severe thinning of the GCC neuronal layers and the floor effect, previous studies have shown opposing results regarding the structure function correlation.^25,26 Precise evaluation of the structural and functional parameters is crucial to monitor disease progression and estimate the effect of current and future treatment effects. 

To understand the pathophysiology of LHON better, precise measurements of the structural and functional deviations are needed. In this cross-sectional study, we aimed to assess macular GCC using a semi-automated segmentation and its correlation with visual function in individuals with chronic LHON from the Swedish LHON registry. 

Methods

Subjects

This study was conducted at the Unit of Optometry, Karolinska Institutet, Sweden. Subjects in the chronic stage of LHON were recruited through the Swedish LHON registry, which was founded in 2018 and contains 98 affected individuals and 144 asymptomatic carriers (as of May 16, 2024). At the date of last inclusion to this study (February 19, 2024), the Swedish LHON registry had 93 affected individuals, of whom 75 had given consent to be contacted. All of the subjects who had given consent to be contacted were invited to participate in the study. The type of LHON mutation, disease duration, and treatment history were recorded from the registry. All included subjects underwent a comprehensive visual examination following the Swedish optometry quality standard. The study protocol adhered to the tenets of the Declaration of Helsinki and was approved by the Swedish Ethical Review Authority (Dnr 2021-06402-01 and Dnr 2023-01105-02). Written informed consent was obtained from participants after they received an explanation of the purpose, nature, and possible consequences of the study. 

OCT Measurement and GCC Segmentation

OCT measurements of the macula were performed with the OCT-HS100 (version 4.7.1.0; Canon, Tokyo, Japan) through undilated pupils. This spectral-domain OCT can perform a scan with an axial resolution of 3 µm. Macular scans were performed with the following specifications: 10 × 10-mm scan window size, 128 B-scans each containing 1024 A-scans (interscan distance, 77 µm; B-scan orientation, vertical). OCT scans with poor centration, blurred images, presence of blink, or poor fixation were excluded. Automated segmentation by the OCT device was used to obtain GCC thickness in nine sectors using the Early Treatment Diabetic Retinopathy Study (ETDRS) grid (Fig. 1). The B-scans were exported to the Iowa Reference Algorithm (IRA) (Retinal Image Analysis Lab, Iowa Institute for Biomedical Imaging, Iowa City, IA, USA), and the GCC automated segmentation was performed. The segmentation was manually rechecked and corrected by the same experienced examiner for each B-scan. The GCC values obtained with the corrected IRA segmentation were noted in the ETDRS sectors. The average values were calculated for the internal ring (internal superior, internal inferior, internal temporal, internal nasal) and external ring (external superior, external inferior, external temporal, external nasal). 

Figure 1.

 

Schematic representation of the macula sectors overlapped on scanning laser ophthalmoscopy images from the OCT. Macula scan, 10 × 10 mm; 1024 × 128 B-scans. The yellow grid represents the ETDRS grid defined by three concentric rings with diameters of 1, 3, and 6 mm. CC, central circle; IR, internal ring; IS, internal superior; IN, internal nasal; II, internal inferior; IT, internal temporal; ER, external ring; ES, external superior; EN, external nasal; EI, external inferior; ET, external temporal.

Visual Acuity

Monocular and binocular distance visual acuity was measured using a ETDRS chart at a distance of 4 meters with best correction. If the participant was unable to read any letter at the 4-meter distance, then visual acuity was measured by changing the viewing distance. 

Visual Field

Monocular Esterman visual field tests and custom 30° visual field tests were performed with the Octopus 900 device (Haag-Streit, Koeniz, Switzerland). In order to evaluate the central visual field sensitivity in chronic LHON subjects with severe visual impairment, we created a custom visual field pattern with 40 prepositioned testing points, presented at 10°, 15°, 20°, 25°, and 30° (Fig. 2). The stimuli size was Goldman V, and the fixation object was a circle with maximal light intensity. The stimuli size was chosen based on a previous study showing increased dynamic range and decreased intertest variability.²⁴ The test was designed as follows: white-on-white stimulus, Goldmann size V stimulus, stimulus duration of 200 ms, background luminance of 31.4 apostilb (asb), stimulus maximum luminance of 4000 asb, and a dynamic strategy for step size. All tests were performed by a trained examiner in a dark room without windows. Before each test, a light calibration was performed. During each visual field test, the examiner monitored the patient's gaze; further instructions were given in case of incorrect fixation, and the test was repeated. Only exams with false-positive rates of <30% were included. From the Esterman test, the Esterman score (points seen in percentage) was noted. From the custom test, the pattern of the visual field defect and the mean defect (arithmetic mean of the sensitivity loss displayed in the comparison representation) values were noted in decibels (dB) and with antilog (nonlogarithmic scales) before averaging the sensitivity loss and conversion back to decibels.^27,28 

Figure 2.

 

Points of stimuli at 10°, 15°, 20°, 25°, and 30°. The test was designed with the following protocol: custom, dynamic, low vision, white-on-white, Goldmann size V stimulus.

Statistical Analysis

The demographics of the participants are summarized with descriptive statistics. The thickness of the GCC in each ETDRS sector was calculated separately. The internal ring and external ring were calculated by averaging the values of corresponding sectors. A paired-samples t-test (Student’s or Wilcoxon signed-rank based on the assumption checks of normality, or Shapiro–Wilk) was used to examine the differences between automated and re-evaluated segmentation for each of the nine sectors, as well as differences between first and second affected eyes for both structure and function. Independent samples t-test was used to investigate the differences between treated and untreated eyes. Linear regression was used to examine the relationships among disease duration, onset, functional tests, and structural measurements. For all statistical tests, the significance level was set to P < 0.05. JASP 0.16.3 was used for statistical analysis.²⁹ 

Results

Subject Demographic Data

Thirty-four subjects (45% of consented subjects from the registry) were examined; of these, 26 had reliable OCT measurements and were included in the analysis. The detailed demographic data are presented in Table 1. Among these chronic LHON individuals, 18 subjects had 11778G>A (MT-ND4) mutation, four subjects had 3460G>A (MT-ND6) mutation, and one subject had 14484T>C (MT-ND1) mutation; two subjects had other mutations, and one subject had no registered mutation in the Swedish LHON registry. Of the participants 54% had received treatment with oral idebenone, 6% had received Q₁₀, and 40% of the subjects had received no treatment for LHON. 

Table 1.

 

Demographic Data

 

Structural Measurements

The thicknesses of the GCC from the automated segmentation and the corrected segmentation are displayed in Table 2. All nine ETDRS sectors showed a statistically significant difference between the automated and corrected segmentation (P < 0.001). The average overestimation of the automated segmentation ranged from about 16 µm to 30 µm for the different sectors. In subjects with unilateral and subsequent bilateral involvement, the corrected segmentation GCC values did not differ between the first and second affected eyes in any of the sectors (P > 0.05). Internal ring, external ring, and average macula GCC thicknesses exhibited weak to moderate correlation with age at disease onset (r = 0.379, P = 0.007; r = 0.411, P = 0.003; r = 0.398, P = 0.004, respectively). Internal ring GCC thickness exhibited weak correlation with disease duration (r = 0.282, P = 0.048). When comparing GCC thickness among treated and untreated subjects, there were no significant differences (P > 0.05). 

Table 2.

 

Automated Segmentation and Re-Evaluated Segmentation of GCC Thickness

 

Functional Measurements

The results from the functional tests (best-corrected visual acuity [BCVA], Esterman test, and custom central 30° visual field test) are displayed in Table 3. Except for 11 eyes, the best corrected visual acuity was >1.0 logMAR. In seven eyes, the BCVA was better than 0.5 logMAR. Similar to BCVA, the mean defect values from the custom visual field measurements and the Esterman score also showed large variations among the individuals. The average inter-eye difference in the mean defect was 2.23 ± 1.35 dB. Four subjects had differences in the visual field defect patterns between the eyes and had an inter-eye mean defect difference of more than 5 dB. The Esterman score displayed a moderate correlation with age at disease onset (r = 0.447, P = 0.001). Treated eyes had better BCVA compared to untreated eyes (difference, 0.38 logMAR; P = 0.017). However, no differences were seen for the Esterman score and mean defect scores in treated and untreated subjects (P > 0.05). In subjects with unilateral and subsequent bilateral involvement, the visual function tests did not differ between the first and second affected eyes (P > 0.05). 

Table 3.

 

Visual Function Metrics

 

Structure–Function Correlation

There was a significant correlation with GCC thickness in all sectors for all functional parameters measured (Table 4). BCVA exhibited moderate correlation to GCC thickness in all sectors, whereas the mean defect and Esterman score showed weak to moderate correlation. The weakest correlation was seen for GCC thickness in the central circle and visual field scores (Esterman score: r = 0.290, P = 0.046; mean defect dB values: r = 0.353, P = 0.014). The strongest correlation was seen for GCC thickness in the external temporal sector and all functional parameters (BCVA: r = 0.604, P < 0.001; mean defect: r = 0.457, P = 0.001; Esterman score: r = 0.421, P = 0.003). For the mean defect values calculated with averaged nonlogarithmic values converted to decibels, the weakest correlation was seen for external nasal sector, and the strongest correlation was seen for the internal inferior and external temporal sectors (r = 0.341, P = 0.016, r = 0.424, P = 0.002; r = 0.420, P = 0.002, respectively). 

Table 4.

 

Linear Regression Analysis for GCC Thickness and Visual Function

 

Discussion

The aim of this study was to evaluate the retinal structural changes and visual function in individuals with chronic LHON from the Swedish LHON registry. We first investigated the correctness of the automated segmentation of the GCC thickness in the macula by comparing it with the manually rechecked and corrected segmentation. Second, we evaluated the visual function by examining visual acuity and the visual field. In general, the automated segmentation misplaced the retinal layer borders and displayed an overestimation of GCC thickness. Thickness values from the corrected segmentation showed significant correlation with functional parameters. OCT imaging and precise GCC thickness evaluations in the macula can provide useful information about the structural damage and level of visual impairment in chronic LHON subjects that might be missed using automated segmentation methods. 

In this cross-sectional study, 26 subjects in the chronic stage of LHON participated. Among these subjects, 40% were affected before any treatment was available (i.e., before the approval of idebenone for LHON). Misplacement of the retinal layer border was evident in these patients’ OCT scans with automatic segmentation, and this effect was seen across different commercially available OCTs (Canon OCT-HS100, ZEISS CIRRUS 5000) (Fig. 3). This is not surprising, as in chronic LHON the morphology of the neuroretina is severely altered, and the segmentation algorithms are known to be inaccurate in such conditions.^21,30 In the current study, OCT B-scans from the macular region were exported and segmentation was manually adjusted using IRA software to ensure accurate GCC thickness measurements. The OCT B-scans from the disc region were not evaluated, as it was difficult to check and correct for the segmentation errors, as the RNFL and GCL boundaries were indistinguishable, particularly in the nasal side due to severe thinning. After correcting the segmentation errors in the macular B-scans, it can be noted that, on average, all of the eyes had overestimated GCC values for the automated segmentation compared to the corrected segmentation. It should be taken into consideration that the inaccuracies in thickness estimation could increase in subjects with thinner GCCs, as the border between RNFL and GCL will be more indistinguishable. 

Figure 3.

 

Automatic segmentation of the GCL–IPL by the OCT instrument built-in automated segmentation algorithm of a foveal B-scan of a chronic LHON subject. (A) Foveal B-scan from Canon OCT-HS100 (purple and yellow lines represent GCL–IPL segmentation). (B) Foveal B-scan from the ZEISS CIRRUS 5000 (orange and cyan lines represent GCL–IPL segmentation). In A, segmentation error can be seen on the temporal side, and in B on the nasal side.

Comparing GCC thicknesses from the current study with previously reported data on chronic LHON subjects^31,32 demonstrates differences and similarities (Table 5). Pajic et al.³¹ used another commercially available OCT to evaluate chronic LHON subjects and manually corrected the segmentation when the automated segmentation was incorrect. The value reported in the central circle was similar (36.9 µm) compared to our automated values, but was thicker than our corrected value. In the internal and external rings, the values were similar to our corrected values but thinner than our automated values. Majander et al.³² reported that automatically segmented GCC thicknesses in the internal ring also showed greater values (66.3 µm) compared to out corrected values but thinner than our automated values. Other studies^17,33–37 have used a different grid pattern (4 × 4.8-mm ellipse) to report GCL–IPL thicknesses and reported values ranging from 43 to 57 µm. The average GCC thickness in the internal and external rings in the present cohort is 56 µm, including the RNFL. A previous study³⁶ reported that the average macular RNFL value in chronic LHON subjects was 13 µm. Although the above-mentioned studies used a different macular area (4 × 4.8-mm ellipse compared to 6-mm ring), some previous studies have reported thicker values, perhaps a result of misplaced segmentation borders and eventual overestimation. Incorrect measurements can potentially be misleading in the monitoring of disease progression and evaluation of treatment effects. It is therefore crucial to be aware of the frequently occurring segmentation errors in OCT measurements related to chronic LHON and correct them. Correcting the segmentation error will ensure accurate measurements and reduce variability in repeated measurements at different disease stages, during follow-up, and in the evaluation treatment effect. 

Table 5.

 

Mean GCC Thickness Using ETDRS Grid Pattern

 

Although several studies^17,31–36 have investigated the macular ganglion cell thickness in LHON subjects, most studies have included only smaller sample sizes in the chronic stage. To our knowledge, only two studies have had a larger sample size than the current study, with about 70 to 80 eyes. One study specified GCL-IPL thickness,³⁴ and the other study specified GCL thickness.¹⁷ However, the thickness values presented in both studies could be an overestimation, as they are similar to our GCC values (RNFL–GCL–IPL). The potential inaccuracies that could affect the GCC measurements should be taken into account when evaluating LHON subjects in different disease phases and when evaluating treatment efficacy. 

The visual function parameters evaluated in the current study showed large inter-individual variations but displayed small or no interocular differences. Several previous studies have also reported large inter-individual variations in BCVA^17,33,38 but similar BCVA between eyes.^33,38 Although there was a significant difference in BCVA between treated and untreated subjects, the visual field tests scores did not differ in these groups. Only four subjects showed differences in the visual field defect patterns between the eyes. Of those four, three had received treatment (one received treatment after the first eye was affected, and two received treatment after both eyes were affected), which could indicate that the treatment has been more effective for these individuals’ visual function in one of their eyes. We cannot draw any conclusions on why there was a difference in visual function between the eyes in some of the individuals, but we can speculate that it could be due to the timing of the treatment or other confounding factors. In previous studies examining the effect of idebenone,^39,40 the largest treatment effect was seen among patients with the 11778G>A (MT-ND4) and 3460G>A (MT-ND6) mutations. These are also the two most common mutations in our cohort. 

There were no significant differences in the structural and functional parameters evaluated in this study between the first and second affected eyes. The structural parameters did not show any differences between idebenone-treated and untreated subjects in the current study. Previous studies evaluating the effect of idebenone have shown that visual function can be maintained or restored, but the structural damage cannot be reversed.^40–42 Both GCC thickness and Esterman score correlated to age at disease onset, with relatively thicker GCCs and higher Esterman scores in subjects affected at younger ages. There is also evidence that points toward better recovery of vision during the dynamic phase for those affected at younger ages.^6,43 In the acute and dynamic stages of LHON, it is likely that the ganglion cells are a heterogeneous mix of alive, stressed, sick, dying, and dead cells. Idebenone can probably recover ganglion cells that are in the stressed to dying stages and support dendritic regrowth but cannot regrow their axons. 

Thinning of the GCL and RNFL is known to be strongly associated with a decline in visual function in LHON patient eyes. It has been suggested that it is important to take the anti-log before averaging the measures of the visual field sensitivity and then convert it back to decibels to evaluate structure–function correlations.²⁸ In the current study, mean defect in decibels and the mean defect calculated with non-logarithmic sensitivity values showed similar regression values with GCC thickness. In the chronic stage of LHON due to severe thinning of these neuronal layers, we might not see a strong structure–function correlation. Previous studies have shown contradictory findings on the structure–function correlations in chronic LHON. One study showed a correlation between thinning of the GCL–IPL and worsening of visual acuity,²⁵ whereas another study did not find any correlation between retinal thickness and visual function.²⁶ In these previous studies, the number of late chronic stage subjects ranged from 9 to 13 individuals, and there is no mention about the correctness of the segmentation or manual correction of the segmentation. In the current study, GCC thickness measured in 26 subjects in late chronic stage showed correlation with functional parameters, where thicker GCC presented with better function, and the strongest correlation was found for the external temporal sector. Supplementary Table S1 shows the correlation between the automated segmentation values of GCC thickness and visual function. It can be noted that, although a significant correlation was seen for some of the sectors for the automated segmentation values (Supplementary Table S1), the corrected GCC thickness shows correlation in all sectors (Table 4). 

Correlation between macular GCL and visual function has been well demonstrated in various other optic neuropathies, as well.^44–46 Understanding the structure–function correlations in LHON is of interest due to the differences in pathophysiology and natural history. With the development of new treatment strategies, it is becoming necessary to accurately measure the structural changes in order to evaluate treatment efficacy. In the present study, we only evaluated macular GCC thickness as a structural parameter, as it is shown to be a better predictor of visual function.^17,18 Even in the chronic stage of LHON, we observed a correlation between GCC thinning and visual function, which can be more evident with accurate measurements of GCC. Hence, we suggest that confirming the accuracy of segmentation to obtain precise GCC measurements is necessary in chronic LHON. 

This is the first study on the Swedish LHON population to include 45% of consented LHON subjects from the Swedish LHON registry. The limitation of sample size in rare-disease studies makes analyses regarding structure–function correlations during different disease stages and evaluating treatment effect more difficult. LHON is caused by mainly three different mutations that by themselves can impact disease outcomes. There are differences in the prevalence of different mutations in Nordic countries compared to other populations. In Denmark, Finland, and our Swedish cohort, the second most common mutation is 3460G>A (MT-ND6) rather than 14484T>C (MT-ND1).^2–4,43,47 Therefore, this present study adds information to the existing knowledge on the Nordic LHON population. In the current study, only limited information on the natural history of LHON for the subjects could be included, as it was a cross-sectional study. Our recommendation for future studies is to include more information regarding the actual disease course. We are currently investigating lifestyles and quality of life among LHON affected and asymptomatic carriers from the Swedish LHON registry and are performing comprehensive ophthalmological examinations on asymptomatic carriers to evaluate the baseline parameters in order to identify clinical precursors related to disease outbreak. 

In conclusion, the macular GCC thickness obtained using corrected segmentation showed correlation with visual acuity and visual field scores in chronic LHON subjects. The frequently occurring segmentation errors in OCT measurements related to chronic LHON can potentially be misleading in monitoring of disease progression and in evaluating the treatment effects. We therefore emphasize the importance of careful evaluation of the accuracy of OCT segmentation. 

Acknowledgments

Supported by grants from LHON Eye Society, Ögonfonden, St. Erik Eye Hospital philanthropic donations, and Vetenskapsrådet (2022-00799). 

Disclosure: J. Hedström, None; M. Nilsson, None; M. Engvall, None; P.A. Williams, None; A.P. Venkataraman, None