New Threshold for Large Optic Discs in Children Using Bruch's Membrane Opening Area

Abstract

Purpose: To compare the peripapillary retinal nerve fiber layer (pRNFL) thickness of large optic discs in children (LOD-C) with normal sized optic discs in children (NOD-C) and large optic discs in adults (LOD-A).

Methods: We included 127 eyes per group (n = 381). Optic discs were considered large if the Bruch's membrane opening (BMO) area was ≥2.5 mm². pRNFL thickness and minimal rim width (BMO-MRW) were compared between the pediatric groups and the large optic disc groups.

Results: The mean global pRNFL thickness (3.5-mm circle) was 100.2 ± 12.1 µm for LOD-C, 95.9 ± 11.7 µm for NOD-C, and 97.7 ± 10.5 µm for LOD-A. It was significantly thicker in LOD-C compared with NOD-C; the difference decreased toward the periphery. The mean global BMO-MRW was 280.0 ± 41.1 µm for LOD-C, 320.7 ± 10.5 µm for NOD-C, and 252.9 ± 45.8 µm for LOD-A. It was significantly thinner in LOD-C compared with NOD-C, and significantly thicker in LOD-C compared with LOD-A. pRNFL and BMO area showed a positive correlation which was higher if BMO area was >2.8 mm², BMO-MRW and BMO are showed a negative correlation which was higher if BMO area was <1.9 mm².

Conclusions: RNFL and BMO-MRW of pediatric optic discs with BMO area <1.9 mm² and >2.8 mm² should be interpreted with caution, because they seem to be strongly influenced by optic disc size.

Translational Relevance: The novel thresholds for small and LOD-C will help to account for the effects of disc status on RNFL and MRW. They will furthermore help clinicians to better diagnose children with suspicious discs for glaucoma.

Introduction

Glaucoma in childhood and adolescence is rare: The incidence of childhood glaucoma in European countries is estimated between 1:10,000 and 1:30,000.^1,2 The prevalence of juvenile glaucoma in the United States was reported to be 1:50,000,³ accounting for about 0.7% of newly diagnosed glaucoma in Caucasians.^4,5 However, the occurrence of glaucoma in childhood and adolescence usually results in irreversible moderate to severe visual impairment.^6–8 High intraocular pressure (IOP) causes retinal nerve fiber loss of the optic nerve, appearing as neuroretinal rim thinning and optic disc cupping, which is the most sensitive and important parameter for glaucoma diagnosis and detection of progression.⁹ 

Although a cup-to-disc ratio of >0.3 is only present in roughly 3% of normal Caucasian newborn eyes,¹⁰ large optic discs frequently appear with a physiological cupping, sharp excavation, and thin rim. These features may lead the ophthalmologist to the diagnose of a glaucoma suspect optic disc. Especially in children, the differentiation between physiological cupping in large optic discs and glaucoma can be challenging; other parameters usually raised in adults, such as visual field analysis or IOP measurement, may not be performed or reliable owing to anxiety or incompliance. 

Optical coherence tomography (OCT) is an objective, fast, and noncontact examination with an axial resolution of 3 µm,¹¹ quantifying the peripapillary retinal nerve fiber layer (pRNFL) thickness. Thinning of pRNFL thickness is an indicator for glaucoma in children and adults.^12–14 Although an OCT examination is simple to perform in most children after the age of 5 to 6 years, the configuration of pRNFL thickness in large optic discs sometimes differs from normal sized optic discs and is categorized suspicious when compared with the normative databases. Thus, the specificity for glaucoma detection might be reduced. In such borderline cases, where visual field diagnostics and IOP measurement are not reliable, only follow-up examinations provide enough information to rule out glaucoma. 

Thus, the aim of this study was to evaluate the pRNFL thickness profile in children with large optic discs (LOD-C) compared with children with normal sized optic discs (NOD-C), and adults with large optic discs (LOD-A). 

Methods

Participants

This retrospective cohort study was performed at the Department of Ophthalmology of the University Medical Center Mainz, Germany. A chart review was conducted searching for LOD-C, NOD-C, and LOD-A who had undergone a pRNFL examination using spectral domain OCT (SPECTRALIS, Heidelberg Engineering, Heidelberg, Germany). Large optic discs were defined as Bruch’s membrane opening (BMO) area of ≥2.5 mm². The searched cohort comprised patients of mixed geographical origins. Exclusion criteria were a history of glaucoma as defined by focal or diffuse thinning of the neuroretinal rim, nerve fiber layer defects, or asymmetric excavation in slit-lamp fundoscopy with corresponding visual field defects and RNFL thinning in the OCT, use of antiglaucomatous medication, or any ocular or cerebral disease possibly influencing the pRNFL thickness. The control group comprised subjects who had undergone comprehensive eye examination because of a suspected disease (e.g., glaucoma suspect optic disc), but had turned out to be healthy (except from minor refractive errors). The diagnoses are listed in Supplemental Material S1. 

Optical Coherence Tomography

The glaucoma module Premium Edition software (version 6.16.7.0) was used to measure 24 high-resolution radial optic nerve head scans and 3 circle RNFL scans with a diameter of 3.5 mm, 4.1 mm, and 4.7 mm, centered at the optic disc. The BMO area and BMO-minimal rim width (MRW) were calculated from the radial scans, and the pRNFL thickness was calculated from circular scans using Heidelberg Engineering Spectralis standard software. A quality review of OCT scans was performed. If correct segmentation of BMO, inner limiting membrane or RNFL by the device software failed, a manual correction was performed by JVS and SW. In cases of significant segmentation mistakes a glaucoma specialist with OCT reading center experience (JR) conducted the correction. Cases were excluded if automated segmentation failed, and a manual correction was not possible in two or more scans. 

Refractive Error, Ocular Biometry, and Visual Field Examination

Apart from OCT, most patients received refractive error measurement using OCULUS/Nidek AR-1s (OCULUS, Wetzlar, Germany) and ocular biometry using IOL Master 700 (Zeiss Meditec, Jena, Germany). Visual field examination using Octopus 900 (Haag-Streit AG, Koenitz, Switzerland) or Humphrey Field Analyzer (Carl Zeiss Meditec). The measurements were not conducted in cases with low compliance. 

Statistics

Both right and left eyes were permitted for evaluation. Eyes were classified in three groups: LOD-C, NOD-C, and LOD-A. Each group contained 167 cases. A two-step 1:1 nearest neighbor matching was applied. In step one, LOD-C and NOD-C were matched by sex and age. Sex was exactly matched and for age calipers of 0.2 standard deviations were used. We found 140 matching pairs out of 167 cases per group. In a second step, the remaining 140 LOD-C from step one were matched with LOD-A by sex and BMO area. Sex was exactly matched, and for BMO area calipers of 0.2 standard deviations were used. 

The sample size calculations were performed using G*Power (Version 3.1.9.6) and based on the mean global RNFL thickness (96 ± 10 µm) in the Gutenberg Health Study, a population-based cohort study from the region of Rhine-Main, Germany.¹⁵ The aim was to detect a difference of 4 µm between the groups. For the comparison of LOD-C with NOD-C and LOD-C with LOD-A, enrollment of 127 patients per group was necessary to generate statistically valid conclusions on an alpha level of 0.025 with a power of 0.8. All other analyses were exploratory. 

Statistical analysis was conducted with RStudio (2022.02.1 Build 461). For normally distributed variables, mean ± standard deviation is presented, and for non-normally distributed variables, median and interquartile ranges (IQRs) are shown. The mean global pRNFL thickness, the mean pRNFL thickness of six segments (temporal superior, superior, nasal superior nasal, nasal inferior, inferior, temporal inferior, temporal), BMO area, and MRW were compared using a mixed model with generalized estimated equations, because both eyes per patient were allowed to be included. The χ2 test was used for categorial comparison of pRNFL classes. The mean for each of the 768 single measurements from the pRNFL thickness circles was computed and a mean pRNFL thickness curve was displayed to compare the pRNFL thickness configurations in each group qualitatively. Linear regression analysis was performed to find relationships with BMO area. Furthermore, locally estimated scatterplot smoothing curves were depicted in the scatterplots and breakpoints with correlation curve slopes were calculated with regression model with segmented relationships. 

Results

We included 127 matching subjects per group in the analysis. Owing to the matching, sex and age did not significantly differ in LOD-C and NOD-C groups (sex, 53% vs. 54% females [P = 0.90]; age, 12.0 ± 3.1 years vs. 11.6 ± 3.1 [P = 0.33]), and sex and BMO area did not significantly differ in LOD-C and LOD-A groups (sex, 54% vs. 55% females [P = 0.80]; BMO area, 2.04 [95% interquartile range (IQR), 1.77–2.26] vs. 2.7 [95% CI, 2.6–2.9] [P = 0.11]). Information on visual acuity, visual field examination, refractive error, axial length and IOP are provided in Table 1. 

Table 1.

 

Baseline Characteristics per Group

 

The mean global pRNFL thickness (3.5-mm circle) was 100.2 ± 12.1 µm for LOD-C, 95.9 ± 11.7 µm for NOD-C, and 97.7 ± 10.5 µm for LOD-A. The mean global BMO-MRW was 280.0 ± 41.1 µm for LOD-C, 320.7 ± 10.5 µm for NOD-C, and 252.9 ± 45.8 µm for LOD-A. Table 2 presents the pRNFL thickness of the 3.5-mm, 4.1-mm, and 4.7-mm circles and the BMO-MRW of all segments per group showing a thicker pRNFL with smaller measurement circle diameter and a decrease of group differences toward the periphery. 

Table 2.

 

pRNFL Thickness and BMO-MRW in the Segments per Group in µm

 

When comparing the 3.5-mm circle of LOD-C with NOD-C, global pRNFL and the nasal and nasal inferior segments were significantly thicker in the large optic discs (Supplemental Material S2). Measuring more peripherally, the pRNFL was significantly thicker in the temporal superior and temporal segments and globally in the 4.1-mm circle, and in the temporal superior and temporal segments in the 4.7-mm circle. The pRNFL thickness of LOD-C did only significantly differ from pRNFL thickness of LOD-A in the temporal superior segment in the 3.5-mm circle and temporal in the 4.1-mm circle. The pRNFL thickness profiles (768 single measurements) of LOD-C vs. NOD-C and LOD-C vs. LOD-A are presented in Figure 1. 

Figure 1.

 

pRNFL thickness profile plots (TSNIT) of large optic discs vs. normal optic discs in children (A–C) and LOD-C vs. adults (D–F).

BMO was significantly (mean difference, 0.93 mm² [95% CI, 0.84–1.02 mm²]; P < 0.001) larger in the LOD-C group than in the NOD-C group, but was not significantly different between LOD-C and LOD-A (mean difference, 0.05 mm² [95% CI −0.04 to 0.15 mm²]; P = 0.25). BMO-MRW revealed to be significantly thinner in all segments of LOD-C compared with NOD-C, and significantly thicker in all segments of LOD-C compared with LOD-A (Supplemental Material S2). 

A comparison of the pRNFL classes (within normal limits, borderline, or outer normal limits in comparison with the device's normative database) between normal and LOD-C revealed that in the pRNFL thickness in the temporal inferior and nasal superior segments was more frequently classified as borderline or outer normal limits in children with large optic discs (P = 0.02 and P = 0.03) (Table 3). 

Table 3.

 

OCT pRNFL Thickness Categorial Output

 

In a visual comparison of the children's pRNFL thickness profiles, the pRNFL seemed thicker in optic discs with BMO area of ≥3.0 mm than in optic discs with a BMO area of <3.0 mm, especially in the nasal inferior and temporal inferior zones (Fig. 2, Supplemental Material S3). 

Figure 2.

 

pRNFL thickness profiles split by BMO area (<2.0 mm², 2.0–2.5 mm², 2.5–3.0 mm², ≥3.0 mm²) for the 3.5-mm measurement circle in children (LOD-C and NOD-C).

There was a positive overall correlation between pRNFL thickness of all three circles and BMO area (r = 0.29; r = 0.28; r = 0.25; P < 0.001 each). Locally estimated scatterplot smoothing curve revealed an increased correlation between pRNFL thickness and BMO area for larger optic discs. Regression model with segmented relationships estimated a breakpoint for BMO area of 2.8 mm² (Fig. 3A, dashed line), with correlation curve slope of 3.7 µm/mm² (95% CI, −0.4 to 7.9 µm/mm²) for optic discs with a BMO area of <2.8 mm² and correlation curve slope of 11.3 µm/mm² (95% CI, 3.1–19.5 µm/mm²) for optic discs with a BMO area of >2.8 mm² (Fig. 3A, red line). The scatterplots, breakpoints and slopes for the 4.1-mm and 4.7-mm circles are presented in the Supplemental Material S4. 

Figure 3.

 

Scatterplots showing (A) global mean pRNFL thickness in relation to BMO area in children (LOD-C and NOD-C groups) for the 3.5-mm circle (4.1-mm circle and 4.7-mm circle are provided as Supplemental Material) and (B) global mean MRW in relation to BMO area in children (LOD-C and NOD-C groups). In (A), locally estimated scatterplot smoothing (LOESS) curve (blue) reveals an increased slope of BMO-pRNFL thickness correlation for BMO area >2.8 mm² (dashed intercept line), calculated with regression model with segmented relationships (red line). In (B), the LOESS curve shows a negative correlation of MRW and BMO area; a breakpoint was calculated at 1.9 mm², below which the negative correlation is stronger than above. The red background highlights the traditional threshold value of 2.5 mm² for large optic discs.

In contrast with pRNFL thickness, BMO-MRW showed a negative correlation with BMO area (r = −0.43; P < 0.001). A breakpoint was estimated at 1.9 mm² using a regression model with segmented relationships (Fig. 3B, dashed line). The correlation curve's slope was −173.2 µm/mm² (95% CI, −246 to −100.7 µm/mm²) for optic discs with a BMO area of <1.9 mm² and −20.3 µm/mm² (95% CI, −36.0 to −4.5 µm/mm²) for optic discs with a BMO area of >1.9 mm². 

To estimate the influence of BMO area and other factors on pRNFL thickness, a regression analysis with generalized estimating equations was performed. Multivariable analysis (Table 4) showed an increase in pRNFL thickness of 3.11 µm per mm² increasing BMO area (95% CI, 0.40–5.82; P = 0.025). Age and gender were not associated with pRNFL thickness. 

Table 4.

 

Linear Regression Analysis

 

Discussion

Optic discs show a large variation in their size and configuration with an inter-individual variability of 0.8 to 5.4 mm² in fundus photographs of Caucasian populations.¹⁶ Moreover, the optic disc size varies depending on the ethnic background, with the largest optic disc sizes in African American populations, followed by Hispanics, Asians, and Caucasians.¹⁷ Owing to their larger cup-to-disc ratio, they are frequently classified as glaucoma suspect, and sometimes OCT yields suspicious results as well. Thus, the aim of this study was to evaluate the pRNFL thickness in LOD-C and to compare them with NOD-C and LOD-A of mixed geographical origins. 

Compared with the NOD-C group, the pRNFL thickness in the LOD-C group was significantly thicker when measured globally and in the temporal superior, nasal and nasal inferior segments when comparing the 3.5-mm circle. This finding is concordant with previous studies reporting a thicker pRNFL in eyes with LOD-C^18,19 and adults.²⁰ The other circles showed a tendency toward decreasing differences between LOD-C and LOD-A toward the periphery. The graphic comparison proved that the differences were mainly located at the pRNFL maxima (Figs. 1 and 2). 

There are two possible explanations for this relation: First, there may be a greater number of retinal nerve fibers in eyes with large optic discs. Some histological studies found a positive correlation between optic disc size and nerve fiber count with an increase of 175.000 nerve fibers per additional square millimeter of optic disc area^21–23; however, some other studies with smaller sample sizes did not find a correlation.^24,25 A previous OCT study reported a slight but in all sectors consistent an positive correlation between optic disc size and macular ganglion cell/inner plexiform layer in children <12 years of age, supporting the theory of a higher number of nerve fibers and ganglion cells in eyes with large optic disc.¹⁹ In contrast, another OCT study in adults found a lower pRNFL and mGCIPL thickness in large optic discs compared with normal sized adult optic discs.²⁶ 

A second explanation might be that the measurement circles are located closer to the optic disc's edge, because they have a defined diameter and are adjusted to the optic disc's center and not its edge, causing an approximation of disc margin and measurement circle. As known from previous studies, the pRNFL thickness increases with smaller circle diameters, that is, with closer distance to the edge.^24,27 This factor could be either explained by increasing cumulation of nerve fibers or owing to the glial support tissue at the optic disc margin. However, the thicker pRNFL in large optic discs is unlikely to be a sole product of nerve fiber cumulation, because the circle does not move closer to the margin in large optic discs, but the margin moves closer to the circle, and the count of nerve fibers should be the same at a given distance from the optic disc center. The more reasonable explanation is the reorganization of nerve fibers with neuroglial cells in the transition zone immediately before entering the optic nerve canal. Astrocytes align both longitudinally in columns and cross-sectionally in honeycomb compartments around nerve fiber bundles for scaffolding, supporting, and metabolic functions.^28,29 This area of pRNFL being interfused with additional supportive tissue moves relatively toward the measurement circles in large optic discs and thus might be responsible for the thick pRNFL in these eyes. 

Compared with LOD-A, none of the three circles in the LOD-C group showed a significant difference in global pRNFL thickness. We expected a difference between children and adults because of the constant nerve fiber loss with age, which was estimated to be 400 to 7000 axons per year in histological studies and 0.16 to 0.44 µm pRNFL thickness per year in OCT morphologic studies depending on age and device type.^21,30–32 The pRNFL thickness does not decrease similarly in all segments over time, which might be an explanation for the inconsistent and insignificant tendency in our study.³² 

BMO-MRW is a sensitive parameter in the detection of early glaucoma. The Hong Kong Children Eye Study found a decrease in BMO-MRW of −47 µm per additional square millimeter BMO area, which is similar to our results (−39.4 µm).³³ Sectoral analysis revealed a significant thinner BMO-MRW in all sectors of LOD-C than of NOD-C, which might be explained by the larger optic disc rim to which the retinal nerve fibers may distribute. Additionally, stereometric OCT-based studies in pediatric large optic discs found a larger cup volume and cup-to-disc ratio, but interestingly also a larger rim area in large optic discs than in normal sized optic discs.^18,34 Comparisons between LOD-C and LOD-A showed a significant thicker BMO-MRW in all sectors, which might be linked with the age-related axonal loss on the one hand, but might also be affected by preperimetric early glaucoma cases that were not yet classified as such. Regarding the pediatric groups, BMO-MRW showed a stronger correlation in eyes with BMO area of <1.9 mm²; thus, it should be interpreted with care in small optic discs when it is used as early glaucoma detection parameter in children. 

The ISNT rule is a widely accepted approach to evaluating the optic disc in fundoscopy regarding glaucomatous damage. The pRNFL thickness measurements in this study adhered to the ISNT rule except from the 4.7-mm circle measurement in children, which showed rather an ISTN configuration. Adult optic discs followed the ISNT rule in all three circles. Previous studies investigated the ISNT rule in nonglaucomatous cupped optic discs in children and found an adherence in 16% to 24% using fundoscopy or OCT.^35,36 Dave at al.³⁶ reported a greater adherence of pediatric optic discs to the IST rule (inferior > superior > temporal) which partly corresponds with our results. A systematic review further found that the ISNT rule is hardly applicable to children, because the reviewed studies showed a high variability in the order of decreasing thicknesses of pRNFL.³⁷ Neither BMO-MRW complied with the ISNT rule in our sample, which showed an inferior > nasal > superior > temporal order in children and an inferior > nasal = superior > temporal order in the LOD-A. 

The classification algorithm indicating the probability of abnormal values compared with a normative database revealed more frequently a borderline or outer normal limits classification in temporal inferior and nasal superior segments in the LOD-C group compared with the NOD-C group. These segments are adjacent to the nasal inferior and temporal superior segments, which presented with a thicker pRNFL in LOD-C than in NOD-C. A possible explanation may be that the retinal nerve fibers take a different course in large optic discs with a greater proportion running in the nasal inferior and temporal superior segments, resulting in a differently configured pRNFL thickness in the adjacent segments temporal inferior and nasal superior and in a false-positive output leading the ophthalmologist to the diagnose of a glaucoma suspect optic disc. 

pRNFL thickness and BMO area showed a slight to moderate positive correlation in all three measurement circles (Fig. 3a, Supplemental Material S4). The correlation was greater in large optic discs. The higher pRNFL thickness could possibly contribute to the false assessment of the OCT classification algorithm. Regression analysis revealed a breakpoint in BMO area of 2.8 mm², indicating a stronger association between pRNFL thickness and BMO area for optic discs sized ≥2.8 mm². We propose to use this value as threshold for LOD-C, because the pRNFL thickness is thicker in such eyes and might present with an unusual distribution pattern: The displaced or broadened pRNFL maxima, and the false positive OCT algorithm classification as “borderline” or “outside normal limits” may lead to a suspicious appearance for glaucoma. OCT in optic discs with BMO area of ≥2.8 mm² or ≤1.9 mm² should be interpreted carefully because their RNFL or MRW shape can appear differently. 

This study has several limitations. First, because of the retrospective method, refractive error was not measured in cyclopegia and thus might be skewed owing to the accommodative abilities of children. Furthermore, the optic discs of highly myopic or hyperopic eyes were not excluded intentionally from the analysis with the aim to capture all possible ocular conditions in which large optic discs occur. The segmentation accuracy was reviewed and corrected manually; however, possible optic artifacts owing to high refractive error cannot be ruled out. Second, the sample size was calculated for the comparison of global pRNFL thickness; a comparison of further segments and further analyses were exploratory. Third, visual field analysis, axial length, refractive error, and IOP were not available in all cases because of the retrospective character of the study and owing to reduced compliance in some cases, which could affect the accuracy of classification as healthy. On the other hand, the setting of this study reflects a real-world population, including cases that need a longitudinal follow-up to make a safe diagnose. Fourth, the control group does not reflect a totally healthy cohort, because referral to our clinic was mainly owing to glaucoma suspect optic discs or nonglaucomatous pathologies such as strabism, which turned out to have healthy OCTs after review by a glaucoma specialist (AKS or EMH). However, minor deviations from totally healthy eyes cannot be fully excluded. Owing to the retrospective character, not all previous diseases might have been considered, and some patients with a suspect optic disc might not have told or do not know their full medical history. 

Conclusions

This study revealed a thicker global pRNFL in LOD-C compared with NOD-C. The positive correlation between pRNFL thickness and BMO area was markedly present in optic discs with a BMO area of >2.8 mm². pRNFL thickness was similar in large optic discs of children and adults. To date, there is no clear evidence if the thicker pRNFL in large optic discs is due to a higher number of nerve fibers, which might be a protective property against glaucoma, or if it is caused by the neuroglial supporting cells at the optic disc margin. Abnormalities in the classification algorithm compared with the normative database might be due to a different distribution of retinal nerve fibers leading to higher peaks in some segments and an altered distribution in adjacent segments. Ophthalmologists should not rely on automated device classification when evaluating LOD-C for glaucoma. Further follow-up of our sample will possibly provide more information on factors associated with glaucomatous damage. 

Acknowledgments

Disclosure: J.V. Stingl, None; C. Braun, None; S. Walter, None; J. Rezapour, None; F.M. Wagner, None; L. Shen, None; A. Strzalkowska, None; I. Schmidtmann, None; A.K. Schuster, Abbvie (F), Apellis (F), Bayer (F), Heidelberg Engineering (F), Novartis (F), PlusOptix (F); E.M. Hoffmann, Abbvie, Heidelberg Engineering (F), Santen (F), Ora (F), Belkin Laser (F), Thea (F)