Macular Optical Coherence Tomography Parameters and Incident Glaucoma Among Myopic Eyes

Abstract

Purpose

The purpose of this study was to estimate the correlations between macular optical coherence tomography (OCT)-derived metrics and incident glaucoma risk in myopic eyes.

Methods

This longitudinal observational study included 24,181 individuals with myopia (spherical equivalence [SE] ≤ −0.5 diopters [D]) from the UK Biobank study.

Results

Participants (mean age = 55.5 ± 8.0 years, 54.5% women) were followed up for 13.2 ± 1.2 years and incident glaucoma was diagnosed in 582 eyes. Those who developed glaucoma were significantly older (P < 0.001), more likely to be male participants (P < 0.001), and had more pronounced myopic refractive error (P < 0.001). Cox regression analyses indicated that participants with thinner retinal nerve fiber layer (RNFL; hazard ratio [HR] = 0.97, 95% confidence interval [CI] = 0.96–0.99, P < 0.001), thinner ganglion cell-inner plexiform layer (GCIPL; HR = 0.97, 95% CI = 0.96–0.98, P < 0.001), and thinner ganglion cell complex (GCC; HR = 0.98, 95% CI = 0.97–0.99, P < 0.001) had an increasing risk of incident glaucoma after adjustment for age, sex, ethnic group, and SE. Meanwhile, the thicker inner nuclear layer (INL; HR = 1.03, 95% CI = 1.01–1.05, P = 0.002) and photoreceptor segments (PS; HR = 1.03, 95% CI = 1.01–1.06, P = 0.024) were positive for the incidence of glaucoma.

Conclusions

This longitudinal study suggested that baseline RNFL, GCIPL, GCC, INL, and PS thickness were significant predictors for the incidence of glaucoma among myopic participants, which indicated a pattern of internal layer thinning (except INL) and outer layer thickening in these pre-glaucoma participants.

Translational Relevance

Our study highlighted the potential of OCT-derived indicators for early glaucoma risk assessment and clinical monitoring.

Introduction

Individuals with myopic eyes exhibit an elevated risk for the development of glaucoma.^1–5 Thus, recognizing the risk factors linked to glaucoma onset among myopic participants are crucial for healthcare management as glaucoma is often asymptomatic in early stages.⁶ Optical coherence tomography (OCT) imaging has significantly improved clinical capabilities for identifying neurodegenerative eye disorders at initial stages, while enabling precise monitoring of progressive optic nerve fiber deterioration.⁷ Substantial clinical evidence has demonstrated the diagnostic utility of specific retinal layer metrics including retinal nerve fiber layer (RNFL), ganglion cell-inner plexiform layer (GCIPL), and ganglion cell complex (GCC) thickness profiles in glaucoma detection and monitoring.^8–12 However, it is yet unclear whether these parameters can help predict those who are at high risk of glaucoma onset among myopic participants.

To our knowledge, no prior research efforts have assessed the relationship of macular OCT parameters and glaucoma onset among myopic participants using longitudinal data. Thus, this current study included myopic participants with over a decade of follow-up to examine the association between baseline OCT parameters and glaucoma onset. Bidirectional Mendelian randomization (MR) analysis was subsequently implemented to investigate putative causal associations between OCT-derived structural biomarkers and glaucoma.

Methods

The UK Biobank is a comprehensive and ongoing research project that involves more than 500,000 individuals, with baseline data collection spanning from 2006 to 2010. The study collected comprehensive systemic and ocular data, all of whom provided informed consent. Participants provided thorough information about their lifestyle, surroundings, and medical background using questionnaires. Health-related outcomes were monitored through linkage to Hospital Episode Statistics (HES) and national mortality registers. The methodologies used for the UK Biobank’s dataset and research have been presented in earlier publications in detail.¹³

The investigation followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) framework. Ethical clearance was obtained from the Northwest Regional Ethics Board (approval code: 11/NW/0382). All procedures adhered to the principles of the Declaration of Helsinki, with documented informed consent secured through UK Biobank protocols (access codes: 99029 and 55836).

Participants

Figure 1 presents the flowchart detailing participant inclusion and exclusion criteria. The UK Biobank cohort initially comprised 502,499 participants. Exclusion criteria were sequentially applied as follows: 387,730 participants lacking refractive error measurements, 2341 with unreliable refraction (a score large than 4),¹⁴ 44,624 without OCT imaging records, 1586 with low OCT image quality (image quality < 45),¹⁵ and 41,471 non-myopic individuals (spherical equivalent [SE] > −0.50 diopters [D]).¹⁶ Additionally, 563 participants were excluded due to a baseline diagnosis of glaucoma. During the follow-up, three participants were lost. Thus, a total of 24,181 myopic individuals were included in this study.

Figure 1.

Flow diagram illustrating participant inclusion and exclusion criteria for the cohort study.

Glaucoma Ascertainment

Glaucoma status at the time of baseline and follow-up assessments was determined using International Classification of Diseases, Ninth Revision (ICD-9) or Tenth Revision (ICD-10) codes (ICD-9 = H40.0–H40.9, and H42, and ICD-10 = 365.0–365.2, 365.5, 365.6, 365.9, and V80.1). The incidence of glaucoma was defined as the initial diagnosis of glaucoma occurring during the period of follow-up, which spanned from the baseline assessment to September 1, 2023.

OCT and Retinal Layer Segmentation

Spectral domain OCT (SD-OCT) was performed at baseline screening implemented during the 2009 to 2010 period. OCT examinations were conducted with the Topcon 3D OCT-1000 MkII system under low ambient illumination, with pharmacological pupil dilation deemed unnecessary. The scanning protocol incorporated three-dimensional macular volumetric acquisition (6 mm × 6 mm acquisition grid comprising 512 axial scans per B-scan and 128 cross-sectional scans arranged in raster pattern) for binocular retinal assessments.

Automated computation of retinal layer thicknesses (RNFL, GCIPL, GCC, inner nuclear layer [INL], photoreceptor segments [PS], retinal pigment epithelium [RPE]; µm scale) was achieved through Topcon TABS (version 1.6.2.6) using dual-scale gradient analysis for precise sublayer quantification, with all measurements derived from six-grid macular protocols and quality-controlled images (≥ 45 rating).¹⁷

Statistical Analyses

Participant characteristics were stratified by data type: continuous variables reported as mean ± SD, whereas categorical variables are expressed as n (%). Demographic variables were categorized based on measurement scales: quantitative measures summarized as mean with standard deviation (SD), whereas qualitative data were presented through frequency counts (percentage distribution). Statistical comparisons were conducted using parametric t-tests or nonparametric χ² analyses contingent on variable characteristics. Multivariable-adjusted Cox models incorporating age, gender, ethnicity, and SE as covariates assessed longitudinal associations between individual OCT biomarkers and glaucoma incidence. Statistical significance thresholds were established at P < 0.05 (two-tailed) using R statistical computing platform version 4.4.1.

To explore whether the associations between baseline OCT parameters and incident glaucoma differ by the severity of myopia, we categorized participants into three subgroups: high myopia (SE ≤ −6.00 D), moderate myopia (SE > −6.00 D and ≤ −3.00 D), and low myopia (SE > −3.00 D and ≤ −0.50 D). Subgroup analysis of myopia severity was conducted for the relationship between baseline OCT parameters and glaucoma. We adjusted for age, sex, and ethnicity group but it was not adjusted for the subgroup analysis variables themselves.

MR Analysis

Genetic studies have found risk genes for both myopia and glaucoma, some genes may increase the risk for both conditions.^18–20 However, the relationship between genetic factors underlying myopia and glaucoma remains unclear, prompting growing interest in the use of genetic tools to improve risk assessment of glaucoma in myopic individuals. To address this, we performed a Mendelian randomization (MR) analysis to strengthen causal inference from a genetic standpoint.²¹ Analytical strategies encompassed five complementary methods: inverse-variance weighted (IVW) serving as the principal estimator, augmented by weighted median estimator, pleiotropy-adjusted MR-Egger, basic mode, and weighted mode techniques.

The MR analyses were deemed valid by considering three key assumptions for causal inferences: (1) instrument relevance (genome-wide significant exposure-linked SNPs satisfying P < 5 × 10⁻⁸ threshold); (2) independence (linkage disequilibrium [LD]-pruned variants: r² < 0.001 within 10,000, kilobases [Kb] windows); and (3) exclusion restriction (no pleiotropic pathways).²² To prevent sample overlap, European cohorts were strictly drawn from two discrete data sources. Genome-wide association studies (GWAS) on OCT parameters featuring UK Biobank participants (44,823 participants) provided the summary statistics used in this analysis.²³ The GWAS data for glaucoma were from FinnGen study (412,181 participants).²⁴ We calculated MR estimates by using IVW for the main analysis, conducting two-sample MR assessing six associations: (1) RNFL thickness and glaucoma, (2) GCIPL thickness and glaucoma, (3) GCC thickness and glaucoma, (4) INL thickness and glaucoma, (5) PS thickness and glaucoma, and (6) RPE thickness and glaucoma. To determine the strength of the IVW estimates, further MR analyses, like weighted median, MR Egger regression, simple mode, and weighted mode, were conducted, due to the fact that these techniques can offer more reliable predictions across an expanded set of situations.²⁵ Sensitivity analyses included MR-PRESSO global testing for horizontal pleiotropy, Cochran’s Q statistics for heterogeneity assessment, and leave-one-out validation of IVW estimates.^26,27

Results

Baseline Characteristics

The study cohort comprised 24,181 individuals with a mean age of 55.5 ± 8.0 years, of whom 54.5% were women. Over a median follow-up duration of 13.2 ± 1.2 years, glaucoma incidence was observed in 582 eyes. Baseline demographic and clinical characteristics of myopic participants, stratified by glaucoma status, are detailed in Table 1. Participants with incident glaucoma demonstrated significantly higher baseline age (P < 0.001), male predominance (P < 0.001), and greater myopic refractive error (P < 0.001) compared with non-glaucoma controls. Ethnic distribution showed no statistically significant intergroup differences (P = 0.579).

Table 1.

Baseline Characteristic of the Myopic Participants by Incidence of Glaucoma

Characteristics	Glaucoma Group (n = 582)	Non-Glaucoma Group (n = 23,599)	P Value
Age, y	59.90 ± 6.91	55.43 ± 8.02	<0.001
Sex, n (%)			<0.001
Female	267 (45.88)	12,921 (54.75)
Male	315 (54.12)	10,678 (45.25)
Ethnic, n (%)			0.579
White	517 (88.83)	21,340 (90.43)
Mixed	13 (2.23)	573 (2.43)
Asian or Asian British	26 (4.47)	871 (3.69)
Black or Black British	21 (3.61)	667 (2.82)
Missing	5 (0.86)	148 (0.63)
SE, diopter	−3.39 ± 3.21	−2.74 ± 2.35	<0.001
RNFL, µm	28.50 ± 8.96	30.22 ± 7.59	<0.001
GCIPL, µm	69.11 ± 7.93	72.10 ± 7.54	<0.001
GCC, µm	97.61 ± 15.53	102.32 ± 12.93	<0.001
INL, µm	31.98 ± 3.34	31.74 ± 3.15	0.099
PS, µm	25.28 ± 3.61	24.69 ± 3.15	<0.001
RPE, µm	27.55 ± 20.23	26.95 ± 20.47	0.493

GCC, ganglion cell complex; GCIPL, ganglion cell-inner plexiform layer; INL, inner nuclear layer; PS, photoreceptor segments; RNFL, retinal nerve fiber layer; RPE, retinal pigment epithelium; SE, spherical equivalent value.

Continuous variables were compared between groups using the Student’s t-test and categorical variables were using the Chi-square test.

Baseline OCT Parameters and Incident Glaucoma

Comparative analysis of retinal layer metrics revealed significantly reduced thickness in the glaucoma group for RNFL, GCIPL, and GCC (all P < 0.001; see Table 1). Conversely, PS thickness demonstrated a significant increase among glaucoma cases (P < 0.001).

Multivariable-adjusted Cox proportional hazards models, controlling for age, sex, ethnicity group, and SE indicated that myopic participants with thinner RNFL (hazard ratio [HR] = 0.97, 95% confidence interval [CI] = 0.96–0.99, P < 0.001), thinner GCIPL (HR = 0.97, 95% CI = 0.96–0.98, P < 0.001), thinner GCC (HR = 0.98, 95% CI = 0.97–0.99, P < 0.001), thicker INL (HR = 1.03, 95% CI = 1.01–1.05, P = 0.002), and thicker PS (HR = 1.03, 95% CI = 1.01–1.06, P = 0.024) had increasing risk of glaucoma onset (Table 2).

Table 2.

Association Between Baseline OCT Parameters and the Incidence of Glaucoma Among Myopic Participants

	Unadjusted		Adjusted
Variables	Hazard Ratio (95% CI)	P Value	Hazard Ratio (95% CI)	P Value
RNFL	0.97 (0.96∼0.98)	<0.001	0.97 (0.96∼0.99)	<0.001
GCIPL	0.96 (0.95∼0.96)	<0.001	0.97 (0.96∼0.98)	<0.001
GCC	0.97 (0.97∼0.98)	<0.001	0.98 (0.97∼0.99)	<0.001
INL	1.02 (1.00∼1.03)	0.110	1.03 (1.01∼1.05)	0.002
PS	1.05 (1.03∼1.07)	<0.001	1.03 (1.01∼1.06)	0.024
RPE	1.00 (1.00∼1.00)	0.790	1.00 (1.00∼1.00)	0.747

CI, confidence interval.

HRs derived from Cox proportional hazards regression models of incidence of glaucoma against OCT parameters. Adjusted models control for age, sex, ethnicity group, and SE.

The cumulative glaucoma rates over the baseline OCT parameters are shown in Figure 2. Participants in the bottom 10% of RNFL thickness (HR = 0.48, 95% CI = 0.35–0.67), GCIPL thickness (HR = 0.23, 95% CI = 0.17–0.32), and GCC thickness (HR = 0.24, 95% CI = 0.17–0.34), had event rates exceeding 2.06, 4.19, and 4.06 times those of the top 10% for glaucoma onset, respectively.

Figure 2.

Cumulative probability of glaucoma development by baseline OCT parameter groups: highest decile, central decile, and lowest decile. (A) Cumulative glaucoma rate by deciles of RNFL thickness. (B) Cumulative glaucoma rate by deciles of GCIPL thickness. (C) Cumulative glaucoma rate by deciles of GCC thickness. (D) Cumulative glaucoma rate by deciles of INL thickness. (E) Cumulative glaucoma rate by deciles of PS thickness. (F) Cumulative glaucoma rate by deciles of RPE thickness.

In stratified Cox regression analyses using low myopia as the reference group, RNFL, GCC, INL, and RPE thickness was significantly associated with incident glaucoma risk in both moderate and high myopia (all P < 0.05). GCIPL and PS thickness showed significant associations only in the moderate myopia group when compared to the low myopia group (P = 0.006 and P = 0.001, respectively; Fig. 3).

Figure 3.

Associations between baseline OCT parameters and glaucoma risk stratified by myopia severity.

MR Analyses

The MR analyses showed that OCT parameters were not casually associated with glaucoma based on IVW estimates (RNFL = odds ratio [OR] = 0.91, 95% CI = 0.81–1.03, P = 0.139; GCIPL = OR = 1.00, 95% CI = 0.98–1.03, P = 0.731; GCC = OR = 0.92, 95% CI = 0.79–1.08, P = 0.320; INL = OR = 0.96, 95% CI = 0.87–1.06, P = 0.456; PS = OR = 0.98, 95% CI = 0.88–1.09, P = 0.726; and RPE = OR = 1.02, 95% CI = 0.80–1.30, P = 0.850). Complementary MR approaches yielded consistent effect estimates across analyses. However, RPE thickness showed divergent associations in simple mode (P = 0.009) and weighted mode (P = 0.030) analyses, as detailed in Table 3. Using OCT parameters as the outcome and glaucoma as the exposure, the study revealed no evidence of glaucoma’s impact on OCT parameters except borderline causal association between glaucoma and INL (Table 4). The sensitivity analysis revealed no substantial change in the overall effect estimate upon the exclusion of any one SNP. See Supplementary Figure S1 for more detail.

Table 3.

MR Results When OCT Parameters Were Considered as Exposure

		IVW		Weighted Median		MR Egger		Simple Mode		Weighted Mode
Outcome	Exposure	OR (95% CI)	P Value	OR (95% CI)	P Value	OR (95% CI)	P Value	OR (95% CI)	P Value	OR (95% CI)	P Value
Glaucoma	RNFL	0.91 (0.81–1.03)	0.139	0.92 (0.84–1.01)	0.089	0.89 (0.63–1.25)	0.498	0.93 (0.80–1.09)	0.373	0.94 (0.84–1.05)	0.272
	GCIPL	1.00 (0.98–1.03)	0.731	1.02 (1.00–1.04)	0.088	1.01 (0.94–1.07)	0.842	1.03 (0.99–1.06)	0.148	1.02 (1.00–1.05)	0.113
	GCC	0.92 (0.79–1.08)	0.320	0.90 (0.81–1.01)	0.090	1.07 (0.65–1.78)	0.790	0.89 (0.74–1.08)	0.298	0.84 (0.69–1.01)	0.103
	INL	0.96 (0.87–1.06)	0.456	1.04 (0.96–1.13)	0.332	1.00 (0.83–1.21)	0.966	1.07 (0.92–1.24)	0.415	1.05 (0.97–1.13)	0.241
	PS	0.98 (0.88–1.09)	0.726	0.93 (0.86–1.02)	0.101	0.96 (0.67–1.36)	0.805	0.86 (0.74–1.01)	0.112	0.92 (0.81–1.04)	0.173
	RPE	1.02 (0.80–1.30)	0.850	0.99 (0.86–1.14)	0.937	1.17 (0.63–2.18)	0.626	0.75 (0.62–0.91)	0.009	0.85 (0.74–0.97)	0.030

IVW, inverse variance weighted; OR, odds ratio.

The P values in bold represent statistical significance.

Table 4.

MR Results When Glaucoma was Considered as Exposure

		IVW		Weighted Median		MR Egger		Simple Mode		Weighted Mode
Outcome	Exposure	OR (95% CI)	P Value	OR (95% CI)	P Value	OR (95% CI)	P Value	OR (95% CI)	P Value	OR (95% CI)	P Value
RNFL	Glaucoma	0.98 (0.89–1.07)	0.626	0.99 (0.91–1.08)	0.806	0.90 (0.64–1.27)	0.567	1.04 (0.87–1.25)	0.659	1.04 (0.92–1.17)	0.571
GCIPL		1.01 (0.93–1.09)	0.877	1.01 (0.93–1.08)	0.890	0.83 (0.63–1.10)	0.205	1.01 (0.88–1.15)	0.899	1.00 (0.90–1.10)	0.926
GCC		1.02 (0.96–1.09)	0.455	1.00 (0.94–1.07)	0.955	0.84 (0.68–1.04)	0.124	1.00 (0.89–1.12)	0.983	1.00 (0.90–1.09)	0.849
INL		1.05 (1.00–1.11)	0.048	1.07 (1.00–1.14)	0.054	0.98 (0.81–1.18)	0.828	1.09 (0.97–1.23)	0.145	1.09 (0.99–1.20)	0.087
PS		1.01 (0.96–1.06)	0.695	0.99 (0.93–1.06)	0.835	1.00 (0.85–1.18)	0.976	0.95 (0.85–1.07)	0.419	1.00 (0.92–1.09)	0.943
RPE		1.00 (0.96–1.05)	0.920	1.03 (0.96–1.10)	0.437	1.19 (1.01–1.40)	0.056	0.90 (0.86–1.13)	0.876	1.05 (0.95–1.16)	0.381

The P values in bold represent statistical significance.

Discussion

This prospective cohort study enrolled 24,181 myopic participants undergoing baseline OCT measurement with longitudinal follow-up (mean = 13.2 years) and found that thinner RNFL, GCIPL, and GCC, as well as a thicker INL and PS, were significantly correlated with increasing risk of developing glaucoma among myopic participants. These observations suggested a potential pre-glaucoma biomarker profile marked by inner retinal structural loss (excluding INL) and outer retinal hypertrophic changes.

There is a paucity of studies that have investigated the predictive value of baseline OCT parameters in assessing the risk of glaucoma development in myopic individuals. Our findings suggested that baseline macular OCT parameters, including macular RNFL, GCIPL, and GCC thickness, were important factors for future glaucoma onset among myopic participants. This suggests that early macular involvement may occur prior to the development of the condition, which is consistent with manifestations observed in the early stages of glaucoma. It is biologically plausible that the thinning of the inner layer in pre-glaucoma is associated with ganglion cell death.²⁸

The correlations of INL thickness and outer retinal layer thickness with glaucoma were contradictory. Tong et al. investigated the thickness of the INL and the outer retinal layer in 271 participants with glaucoma and 548 healthy eyes. They observed a notable decrease in INL thickness in the glaucoma group, whereas no consistent pattern was identified in the outer retinal complex group.²⁹ However, Garcia-Medina et al. found that INL and outer retinal layers were thickening in early-stage glaucoma.³⁰

Our study showed that INL and PS were thicker in eyes with glaucoma onset. The underlying reason for the observed thickening in the INL and outer retinal layers remains unclear, but it may be attributed to the infiltration of glial or inflammatory cells, or the buildup of extracellular matrix components. Previous studies also identified activation of neuroinflammatory and resident glia in glaucomatous retinas and optic nerve heads, particularly in INL, consistent with results observed in experimental glaucoma models.^31–36

Many investigations have conclusively shown that glaucoma and myopia are causally related.^37–40 Therefore, in our research, a Mendelian mediation analysis of myopia and glaucoma utilizing OCT parameters as mediators was not conducted. Instead, we directly performed MR analysis on OCT parameters and glaucoma. Our findings indicated that there was no relationship between genetic liability to retinal layer thickness and glaucoma, and the consistency was also found in reverse analyses except INL. The observed discrepancy between the significant associations in observational analyses and the null findings in MR analyses likely stems from residual confounding, tissue-specific genetic regulation, and instrument strength. These methodological constraints underscore the persistent challenges in differentiating between correlational and causal relationships in large-scale epidemiological research.

Furthermore, several limitations need to be recognized. First, the demographic composition of the UK Biobank cohort, predominantly comprising younger, healthier individuals, may introduce selection bias, potentially compromising the generalizability of our results. Furthermore, the substantial exclusion of participants necessitated by our study design may further compound these limitations. Second, a methodological limitation stems from the unavailability of axial length (AL) measurements in the UK Biobank. This precluded comprehensive assessment of AL-associated structural changes and their mechanistic relationships with glaucoma pathogenesis. Third, the diagnostic accuracy of glaucoma cases in the UK Biobank requires rigorous validation. Case identification relied solely on ICD codes extracted from hospital admission and primary care records, without confirmation through standardized ophthalmic examinations. Whereas ICD-coded diagnoses provide operational utility for large-scale epidemiological research, this methodology may lead to potential overestimation or underestimation of glaucoma prevalence in population-based registries. Additionally, although changes in peripapillary OCT parameters might hold value for assessing the risk of pre-glaucoma, the UK Biobank does not provide relevant data. Our capacity to identify causal relationships that are more directly related to the pathophysiology of glaucoma may have been hampered, in particular, by the lack of optic nerve head RNFL. Finally, residual confounding effects cannot be fully ruled out. For instance, in our sample, over 88% of participants were of White ethnicity, with no significant difference in racial distribution between glaucoma and non-glaucoma groups (P = 0.579). Although this reduces the likelihood of ethnic confounding, it limits the generalizability of our findings to more diverse populations.

Conclusions

In this longitudinal study, we observed that a thinner baseline RNFL, GCIPL, and GCC, alongside a thicker INL and PS, were significantly linked to a higher likelihood of developing glaucoma in myopic individuals. These findings suggest a distinctive macular change pattern in participants at the pre-glaucoma stage, which could offer insights into the underlying mechanisms of glaucoma. Moreover, these macular OCT parameters could be useful indicators for assessing glaucoma risk, facilitating early diagnosis, and tracking disease progression. Additional studies are necessary to validate these findings across different populations.