Profiles of Ganglion Cell-Inner Plexiform Layer Thickness in a Multi-Ethnic Asian Population: The Singapore Epidemiology of Eye Diseases Study

Profiles of Ganglion Cell-Inner Plexiform Layer Thickness in a Multi-Ethnic Asian Population: The Singapore Epidemiology of Eye Diseases Study

Journal Pre-proof Profiles of Ganglion Cell-Inner Plexiform Layer Thickness in a Multi-ethnic Asian Population: The Singapore Epidemiology of Eye Dise...

1MB Sizes 0 Downloads 26 Views

Journal Pre-proof Profiles of Ganglion Cell-Inner Plexiform Layer Thickness in a Multi-ethnic Asian Population: The Singapore Epidemiology of Eye Diseases Study Yih-Chung Tham, PhD, Miao Li Chee, BSc, Wei Dai, MPH, Zhi Wei Lim, BSc, Shivani Majithia, OD, Rosalynn Siantar, MBBS, Sahil Thakur, MBBS, Tyler Rim., MD, PhD, Carol Y. Cheung, PhD, Charumathi Sabanayagam, MD, PhD, Tin Aung, MD, PhD, Tien Yin Wong, MD, PhD, Ching-Yu Cheng, MD, PhD PII:

S0161-6420(20)30140-8

DOI:

https://doi.org/10.1016/j.ophtha.2020.01.055

Reference:

OPHTHA 11115

To appear in:

Ophthalmology

Received Date: 25 September 2019 Revised Date:

31 December 2019

Accepted Date: 30 January 2020

Please cite this article as: Tham Y-C, Chee ML, Dai W, Lim ZW, Majithia S, Siantar R, Thakur S, Rim. T, Cheung CY, Sabanayagam C, Aung T, Wong TY, Cheng C-Y, Profiles of Ganglion Cell-Inner Plexiform Layer Thickness in a Multi-ethnic Asian Population: The Singapore Epidemiology of Eye Diseases Study, Ophthalmology (2020), doi: https://doi.org/10.1016/j.ophtha.2020.01.055. This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2020 Published by Elsevier Inc. on behalf of the American Academy of Ophthalmology

1 Profiles of Ganglion Cell-Inner Plexiform Layer Thickness in a Multi-ethnic Asian 2

Population: The Singapore Epidemiology of Eye Diseases Study

3 4

Yih-Chung Tham, PhD1,2, Miao Li Chee, BSc1, Wei Dai, MPH1, Zhi Wei Lim, BSc1,

5 Shivani Majithia, OD1, Rosalynn Siantar, MBBS3, Sahil Thakur, MBBS1, Tyler Rim. MD, 6 PhD1,2, Carol Y. Cheung, PhD4, Charumathi Sabanayagam, MD, PhD1,2,, Tin Aung, MD, 7

PhD1,2,5, Tien Yin Wong, MD, PhD1,2,5, Ching-Yu Cheng, MD, PhD1,2,5

8 91Singapore Eye Research Institute, Singapore National Eye Centre, Singapore; 102Ophthalmology & Visual Sciences Academic Clinical Program (Eye ACP), Duke-NUS Medical 11School, Singapore; 123Department of Ophthalmology, Tan Tock Seng Hospital, Singapore. 134Department of Ophthalmology and Visual Sciences, The Chinese University of Hong Kong, 14Hong Kong 155Department of Ophthalmology, Yong Loo Lin School of Medicine, National University of 16Singapore, Singapore 17 18Correspondence: Professor Ching-Yu Cheng 19The Academia, 20 College Road 20Discovery Tower, Level 6 21Singapore 169856

22Tel: +65 6322 4500 23Fax: +65 6322 4598 24Email: [email protected] 25Financial Disclosures: None 26Running head: Profiles of Ganglion Cell-Inner Plexiform Layer Thickness in a Multi-ethnic 27Asian Population: The Singapore Epidemiology of Eye Diseases Study. 28

29Abstract 30Purpose: To examine the normative profile and determinants of macular ganglion cell-inner 31plexiform layer (GCIPL) thickness based on spectral-domain OCT (SD-OCT) in a non32glaucoma, multi-ethnic Asian population. 33Design: Population-based, cross-sectional study. 34Participants: Ethnic Chinese, Malay and Indian adults, aged ≥40 years, recruited from the 35Singapore Epidemiology of Eye Diseases Study. 36Methods: All participants underwent standardized examinations. GCIPL thickness was 37measured using Cirrus HD-OCT (Carl Zeiss Meditec). Participants with glaucoma or poor quality 38scans were excluded. Eye-specific data were used. Associations of ocular and systemic factors 39with GCIPL thickness parameters were investigated using multivariable linear regression with 40generalised estimating equation models to account for correlation between both eyes. 41Main Outcome Measure: Average, superior, and inferior hemisphere GCIPL thicknesses. 42Results: A total of 4,464 participants (7,520 eyes) consisting of 1,625 Chinese, 1,212 Malay, 43and 1,627 Indian adults contributed to this analysis. Average GCIPL thickness was 82.6± 6.1μm 44in Chinese, 81.5± 6.8μm in Malays and 78.0± 6.9μm in Indians (P<0.001 by ANOVA). The 5th 45percentile limit of average GCIPL thickness was 72μm in Chinese, 70μm in Malays, and 67μm 46in Indians. The 1st percentile value for average GCIPL was 68µm in Chinese, 64µm in Malays, 47and 60 µm in Indians. In multivariable analysis adjusting for age, gender, axial length, presence 48of cataract, OCT signal strength, disc area, hypertension, diabetes, and hyperlipidaemia, Indian 49eyes were observed to have 3.43µm thinner GCIPL on average, compared to Chinese 50(P<0.001), and 3.36 µm thinner GCIPL, compared to Malays (P<0.001). In addition, older age 51(per decade; β=-2.51), female gender (β=-1.57), longer axial length (per mm; β=-1.54), and

52presence of chronic kidney disease (β=-1.49) were significantly associated with thinner average 53GCIPL (all P≤0.008). Larger optic disc area (per mm2; β=0.78; P<0.001) was associated with 54thicker GCIPL. These factors were consistently observed to be significant for superior and 55inferior hemisphere GCIPL thickness. 56Conclusions: Average and subsector GCIPL thickness profiles were significantly thinner in 57Indians, compared to Chinese and Malays. Our findings further highlight the need of a more 58refined, ethnic-specific normative database for GCIPL thickness, which may in turn improve the 59detection and diagnosis of glaucoma in Asians in clinics. 60

61Precis 62GCIPL thickness profile in Indians is significantly thinner compared with Chinese and Malays, 63suggesting that ethnic-specific OCT normative database is needed, even among Asians. 64 65

66INTRODUCTION 67Glaucoma is a leading cause of irreversible blindness,1 characterised by degeneration of retinal 68ganglion cells.2 As visual field loss only becomes clinically apparent after 25-35% loss of retinal 69ganglion cells,3 early identification of retinal ganglion cell loss is important for early detection of 70glaucoma. With the advent of spectral-domain optical coherence tomography (SD-OCT), the 71ganglion cell layer and inner plexiform layer at the macula can now be quantified objectively and 72non-invasively. In this regard, the ganglion cell-inner plexiform layer (GCIPL) thickness 73parameters by commercial SD-OCTs had been shown as useful markers in detecting 74glaucoma.4-8 In addition, GCIPL parameter was also reported to be less affected by the 75presence of ‘floor effect’ in cases of severe glaucoma, compared to peripapillary retinal nerve 76fibre layer (RNFL) thickness parameters.9 77

Despite the increasingly wide usage of GCIPL evaluation in clinical practice, some gaps

78still remain. First, although Asians account for 60% of glaucoma cases globally,1 Asians are 79poorly represented in current commercial OCT normative databases. For example, in the Cirrus 80HD-OCT’s Ganglion Cell Normative Database, there were only 24% of Asian eyes.10 81Furthermore, given that RNFL thickness profiles vary between different ethnic populations and 82among Asians,11, 12 it is also plausible to suggest that GCIPL thickness profile may differ across 83different ethnic groups. Previous studies mainly demonstrated that thinner GCIPL was 84associated with older age, female gender, longer axial length and thinner RNFL thickness.13-16 85However, ethnic variations in GCIPL thickness had not been comprehensively evaluated in 86these studies, an aspect which may affect the accuracy of glaucoma detection. Taken together, 87a detailed evaluation on the normative profile of GCIPL thickness among Asians is warranted. 88

In addition, GCIPL asymmetry parameters such as inter-eye GCIPL thickness asymmetry

89and intra-eye GCIPL hemisphere asymmetry were recently explored, and were suggested as

90potential markers for early glaucoma detection.17-19 However, the translation of these newly 91proposed GCIPL asymmetry parameters into clinical adoption is currently hindered, partly due 92to the reason that normative profiles of these newly proposed asymmetry parameters have yet 93been comprehensively evaluated, in particular in population-based studies.20 94

Hence, in this study, we aimed to examine and describe the profiles of GCIPL thickness,

95among non-glaucoma eyes in a multi-ethnic Asian population. Second, we aimed to identify 96systemic and ocular factors associated with GCIPL thickness. These findings would be useful in 97establishing a more refined, Asian-specific normative reference for GCIPL thickness 98parameters, which may in turn help to improve glaucoma detection and diagnosis among 99Asians. 100 101METHODS 102Study Population 103

The Singapore Epidemiology of Eye Diseases (SEED) Study is a population-based

104cross-sectional study for three major ethnic groups in Singapore: Malays, Indians and Chinese. 105The detailed methodology of SEED has been described in detail previously.21-24 Briefly, the 106study sampled adults aged ≥40 years residing in the South-western part of Singapore, and used 107a standardized study protocol across the three ethnic groups. Our study population is made up 108of 3,353 Chinese from the baseline visit in year 2009-2011 (response rate 72.8%), 1,901 Malays 109from the 6-year follow up visit in 2011-2013 (response rate 72.1%), and 2,200 Indians from the 1106-year follow up visit in year 2013-2015 (response rate 75.5%). The baseline visits in the Malay 111and Indian population did not undergo OCT assessments and hence were not included in this 112study. This study follows the principles of the Declaration of Helsinki with ethical approval

113obtained from SingHealth Centralized Institutional Review Board. Written informed consent was 114obtained from all participants. 115 116Ocular Examinations 117

All subjects underwent standardized systemic and ophthalmic examinations at the

118Singapore Eye Research Institute. Auto-refraction was performed with auto-refractor (Canon 119RK-5 Auto Ref-Keratometer, Canon Inc. Ltd., Japan) and axial length was measured using non120contact partial coherence laser interferometer (IOL Master V3.01; Carl Zeiss Meditec AG, Jena, 121Germany). Intraocular pressure (IOP) was measured using a Goldmann applanation tonometer 122(Haag-Streit, Konig, Switzerland). Central corneal thickness (CCT) was measured using an 123ultrasound pachymeter (Advent; Mentor O & O Inc., Norwell, MA). Subjective refraction and 124best-corrected visual acuity (VA) were measured by trained optometrists.25 After pupil dilation 125with tropicamide 1% and phenylephrine 2.5%, fundus examinations were performed. Lens 126opacities were graded using Lens Opacity Classification System (LOCS III) during slit-lamp 127examinations.26 Presence of cataract was defined as nuclear opalescence ≥grade 4, nuclear 128colour ≥grade 4, cortical cataract ≥grade 2 or posterior subcapsular cataract ≥grade 2, based on 129LOCS III grading.26 In addition, gonioscopy and 24-2 SITA Fast Humphrey visual field 130(Humphrey Field Analyzer II; Humphrey Instruments, San Leandro, CA) test were performed for 131glaucoma suspects and participants with known glaucoma cases prior to dilation. 132Systemic Examinations 133

Non-fasting venous blood samples were collected and analyzed for biochemical testing

134of serum glycated haemoglobin (HbA1c), glucose, total cholesterol, high-density lipoprotein 135(HDL) cholesterol, low-density lipoprotein (LDL) cholesterol and creatinine. Diabetes was 136defined as random glucose ≥11.1mmol/L, HbA1c ≥6.5%, use of diabetic medication, or self-

137reported history. Hypertension was defined as systolic blood pressure (BP) ≥140 mmHg, 138diastolic BP ≥90 mmHg, antihypertensive drugs usage, or self-reported history of hypertension. 139Hyperlipidaemia was defined as total cholesterol ≥6.2 mmol/L, use of lipid lowering drugs, or 140self-reported history of hyperlipidaemia. Kidney function was assessed using estimated 141glomerular filtration rate (eGFR) from serum creatinine using the chronic kidney disease 142epidemiology collaboration (CKD-EPI) equation.27 Subjects were defined to have chronic kidney 143disease (CKD) if GFR <60 mL/min/1.73m2. Body mass index (BMI) was calculated as weight in 144kilograms divided by height in meters squared. 145Other Measurements 146

A detailed interviewer-administered questionnaire was used to collect information

147including sociodemographic information, medication use, history of neurodegenerative, systemic 148and ocular diseases, retinal surgery and laser treatment, as well as current smoking status. 149Spectral-Domain OCT Imaging 150

After pupil dilation, image acquisition of the macula was performed in each study eye

151using Cirrus high definition-optical coherence tomography (HD-OCT, Carl Zeiss Meditec, Dublin, 152CA.). Macular cube 512 ×128 scan protocol was performed, where a 6 × 6mm2 area centred at 153the fovea was scanned with 128 horizontal B-scans, each consisting of 512 A-scans per B-scan. 154An automated ganglion cell analysis (GCA) algorithm incorporated in the Cirrus HD-OCT 155software version 6.5, was used to segment and measure GC-IPL thickness. Detailed description 156of the GCA algorithm of Cirrus OCT had been reported elsewhere. 14, 28 In brief, the algorithm 157measured the GCIPL thickness inside a 14.13 mm2 elliptical annulus area centred on the fovea 158(Figure 1A). Within this area, the posterior boundary of the RNFL and posterior boundary of the 159IPL, were automatically delineated by the GCA algorithm (Figure 1B). This segmentation 160method further led to generation of average GCIPL thickness parameter, and the 6 sectoral

161GCIPL parameters (supero-temporal, superior, supero-nasal, infero-nasal, inferior, infero162temporal, Figure 1A). Superior hemisphere GCIPL thickness was averaged from 163superiotemporal, superior and superionasal subfields, while inferior hemisphere GC-IPL 164thickness was averaged from inferionasal, inferior and inferiotemporal subfields. Intra-eye 165GCIPL hemifield asymmetry was calculated as the absolute difference between the superior 166and inferior hemisphere GCIPL thickness (Figure 1E) On the other hand, inter-eye GCIPL 167asymmetry was calculated as the absolute difference between the right and left eye’s average 168GCIPL thickness (Figure 1D). Quality control was performed to isolate OCT scans with errors or 169artefacts that might affect GCIPL measurements, for example, poor OCT signal strength (<6), 170misaligned scans, motion artefacts, and segmentation errors. 171Inclusion and Exclusion Criteria 172

We only included subjects that underwent OCT imaging. Subjects were excluded from

173the analysis if signal strength was <6; GCIPL demarcation errors, including motion or blinking 174artefacts; best-corrected VA worse than 20/40; retinal diseases observed from fundus photos 175and OCT scans, including epiretinal membrane, age-related macular degeneration, geographic 176atrophy, diabetic retinopathy, clinically significant macular oedema, retinal detachment and 177other retinopathy cases which affect measurement of GCIPL thickness. Neurodegenerative 178disease and glaucoma cases (diagnosed based on the International Society for Geographical 179and Epidemiological Ophthalmology guidelines) were also excluded. 180Statistical Analysis 181

All statistical analyses were performed using Stata 14.0 (StataCorp LP, College Station,

182TX). Characteristics comparisons between the three ethnic groups were performed using 183analysis of variance (ANOVA) for continuous variables, and chi-square tests for categorical 184variables.

185

Univariate linear regression analyses were performed to examine the associations

186between demographic, systemic and ocular factors with GCIPL thickness. Age, gender, and 187factors which were significant in the univariate analyses (P<0.05, as shown in Supplementary 188Table 1), were then further included in the multivariable linear regression model. Eye-specific 189data was used for this study, and generalized estimating equation (GEE) with exchangeable 190correlation structure and Gaussian link was used in the regression models to account for the 191correlation between pairs of eyes for each individual. P value for significance was set at <0.05. 192

For evaluation on the distribution of GCIPL thickness, a Gaussian model was assumed

193for defining the 1st, 5th percentile cutoff values for GCIPL thickness, and the 95th and 99th 194percentile values for intra-, inter-eye GCIPL asymmetry values. 195

Lastly, as Chinese participants at the time of clinical examination were generally younger

196than Malay and Indian population by approximately 5 to 6 years, we further performed a 197sensitivity analysis by including only participants aged 48 to 82 in our study sample to further 198account for, and minimise the impact of residual confounding by age. 199 200RESULTS 201Of the total 7,454 study subjects, 5,333 subjects (10,049 eyes) had Cirrus HD-OCT scans done. 202Of them, 2,529 eyes were excluded due to poor OCT signal strength (N =70), segmentation in 203error in scan (N =196), presence of retinal diseases (N =1,878), glaucoma or neurodegenerative 204diseases (N =285), and missing relevant covariates data (N=100), thus leaving 7,520 eyes from 2054,464 participants (1,625 Chinese, 1,212 Malay, and 1,627 Indian adults) included for the final 206analysis (Supplementary Figure 1). 207

Table 1 summarises the characteristics of included subjects.The mean age for each ethnic

208group was 60.4±8.8 years for Malays, 60.5±7.8 years for Indians, and 55.2±7.5 years for

209Chinese participants. Among the three ethnic groups, Malay participants had higher systolic BP 210level, higher BMI and a greater proportion individuals with hypertension, CKD, and of current 211smokers (all P<0.001). Indian participants had higher proportions of individuals with diabetes, 212and hyperlipidemia (all P<0.001, Table 1). Across the three ethnic groups, Chinese had longer 213axial length, thickest CCT, lowest IOP (all P<0.001, Table 1). 214

Normative distribution of average GCIPL thickness by ethnic and age groups are described

215in Table 2. Overall, we observed that Indian eyes have thinner average GCIPL thickness, 216compared to Malays and Chinese (all P<0.001). The mean (± standard deviation) value of 217average GCIPL thickness was 80.6 ±6.9 µm in all participants, 82.6 ±6.1 µm in Chinese, 21881.5±6.8 µm in Malays, and 78.0±6.9 µm in Indians. The 5th percentile value for average GCIPL 219thickness was 72µm in Chinese, 70µm in Malays, and 67µm in Indians; whereas the 1st 220percentile value for average GCIPL thickness was 68µm in Chinese, 64µm in Malays, and 60µm 221in Indians. Thinner GCIPL profiles in Indian eyes compared to Malays and Chinese (i.e. lower 222mean, 5th, and 1st percentile values) were also consistently observed across the superior, 223inferior hemisphere, and the 6 sub-sectors of GCIPL thickness parameters (P<0.001 across all 224parameters, Figure 2; Table 3 & 4; Supplementary Table 2 & 3). In further sensitivity analysis, 225including only those aged 48 to 82 years old, we still observed thinner GCIPL profiles in Indian 226eyes compared to Malays and Chinese (all P<0.001, Supplementary Table 4 & 5). 227

The normative distribution of GCIPL thickness for each ethnic group is further illustrated in

228Figure 3. Green area represents GCIPL thickness measurements which fall within the 90% of 229this sample (denoted as the 5th to 95th percentile range). Yellow represents the sample with the 230thinnest 5% of measurements (5th percentile and below), and red represents the thinnest 1% of 231measurements (1st percentile and below). As consistently shown in average, superior, and 232inferior hemisphere GCIPL thickness, the 5th percentile and 1st percentile limits were also 233smallest in Indians, compared to Chinese and Malays (Figure 3A-C).

234

On the other hand, in terms of intra-eye GCIPL hemifield asymmetry, the mean absolute

235difference was 2.3 ±2.1 µm in all participants, and was largely similar across Chinese (2.5 ±2.0 236µm), Malays (2.4 ±2.1 µm), and Indians (2.2 ±2.2 µm) (Table 5). The 95th percentile limit (i.e. 237representing the largest hemifield asymmetry in 5% of sample) for absolute intra-eye hemifield 238difference was 6.0 µm in all eyes and was similar across the three ethnic groups. The 99th 239percentile limit was 10.3μm in Indians, 9.3μm in Malays, and 8.7μm for Chinese. 240

In terms of inter-eye GCIPL asymmetry, the mean absolute difference was 1.7 ±2.3 µm in

241all participants, and was also largely similar across Chinese (1.5 ±1.7 µm), Malays (1.7±2.0 242µm), and Indians (1.8 ±2.9 µm) (Table 6). The 95th percentile limit (i.e. representing the largest 243inter-eye GCIPL thickness difference in 5% of sample) was 5 µm in all participants, and was 244similar across the three ethnic groups. The 99th percentile limit was 11μm in Indians, 9μm in 245Malays, and 8μm for Chinese. 246

In multivariable analysis adjusting for age, gender, axial length, presence of cataract, OCT

247signal strength, disc area, presence of diabetes, CKD, hyperlipidaemia, hypertension, BMI, 248current smoking status, IOP, and CCT, Indians were found to have thinner average GCIPL, 249compared to Chinese (β= -3.43; 95% CI, -3.93 to -2.92; P<0.001; Table 7) and Malays (β= 250-3.36; 95% CI, -3.84 to -2.88; P<0.001; data not shown in table). For superior and inferior 251hemisphere GCIPL parameters, we also observed thinner GCIPL in Indian eyes (Table 7). In a 252further sensitivity analysis on subgroup aged 48-82 years, Indians were still observed to have 253thinner GCIPL (Supplementary Table 6), compared to Chinese eyes. 254

On the other hand, after adjusting for the same covariates, older age (per decade, β=

255-2.51), female gender (β= -1.57), longer axial length (per mm, β= -1.54), presence of any 256cataract (β= -0.50), and presence of CKD (β= -1.49) were significantly associated with thinner 257average GCIPL thickness (all P ≤ 0.008, Table 7). Larger optic disc area (per mm2; β= 0.78;

258P<0.001) was associated with thicker GCIPL. These factors were consistently observed to be 259significant for superior and inferior hemisphere GCIPL thickness. Among these significant 260factors (other than ethnicity), age and axial length were the 2 most important determinants (i.e. 261largest standardized beta coefficient values for age and axial length compared to other factors) 262for average GCIPL thickness. The associations of older age and longer axial length with thinner 263average GCIPL, were consistently observed across the 3 ethnic groups with P trend of <0.001 264(Figure 4 & 5). 265 266DISCUSSION 267

In this study, we evaluated the profile of GCIPL thickness in a non-glaucomatous multi-

268ethnic Asian population. To our knowledge, this is the first large, multi-ethnic population-based 269study which demonstrated differences in GCIPL thickness among Asians. Specifically, we 270observed that Indians have thinner GCIPL profiles compared to Chinese and Malays. 271Correspondingly, the 5th percentile limit of GCIPL thickness which is typically used as the 272reference cut-off for glaucoma screening, varies between the three ethnic groups, with Indian 273eyes having a lower 5th percentile limit, further emphasizing that a new and ethnic-specific cut274off is warranted for GCIPL thickness. In addition, we also presented the normative profiles of 275intra-, and inter-eye GCIPL asymmetry from this population study. Altogether, these 276comprehensive findings on normative GCIPL profiles will be useful in establishing a more 277refined, Asian-specific normative database for GCIPL thickness measured from SD-OCT, which 278may in turn help to improve the detection and diagnosis of glaucoma in clinics. 279

Furthermore, it should also be noted that that the magnitude of observed inter-ethnic

280difference between Indian, Chinese and Malay eyes, was also greater than previously reported 281test-retest variability of GCIPL parameters, which range from 0.61 to 2.64 μm. 29, 30 This further

282indicates that the observed inter-ethnic difference is clinically substantial. The observed inter283ethnic difference in GCIPL profiles among Asians was consistent with RNFL inter-ethnic 284difference observed in our earlier study, where Indian eyes were also reported to have thinner 285RNFL profiles compared to Chinese and Malays.11 Indian eyes correspondingly have lower 286values of 5th percentile and 1st percentile GCIPL profiles (average, superior, and inferior 287hemisphere), compared to Chinese and Malays (Figure 3). Particularly, with regards to the 5th 288percentile limit of average GCIPL, there was a considerable difference of 5μm between Chinese 289(72μm) and Indian eyes (67μm); for the 1st percentile limit, there was a greater difference of 2908μm between Chinese (68μm) and Indian eyes (60μm). Putting this into context, if assuming 291that a higher single GCIPL cut-off of 72μm was used for screening across Asians as a whole, a 292proportion of false positives amongst Indians might be resulted (for Indian eyes with GCIPL 293between 67 to 72μm); and vice versa for potential false negative classification in a proportion of 294Chinese eyes if a lower limit of 67μm was deployed as an ‘across the board’ cut-off. As shown 295in Table 4, given the different 5th percentile limits across ethnic groups for other GCIPL 296subsector parameters, similar scenarios of misclassifications may also occur in these subsector 297parameters which are more sensitive in detecting early localised glaucomatous damage.5 298Hence, a more refined, ethnic-specific Asian normative database for all GCIPL parameters (i.e. 299average and subsector) may help to minimize such misclassifications and improve screening 300performance among Asian patients. 301

Several previous studies reported normative distributions or range for GC-IPL thickness

302in healthy adults.13, 14, 31-34 However, these studies were not population-based studies and did not 303perform inter-ethnic valuation among Asians. It is worthy to further compare our observations 304with the two GCIPL normative databases (the Diversified version and Asian version) currently 305deployed in the commercially available Cirrus HD OCT. First, the Cirrus Diversified GCIPL 306normative database consist of 282 healthy subject of multiple ethnicities (122 European

307descents, 62 Asians [inclusive of Chinese, Korean and Japanese], 51 Africans, 33 Hispanics, 308and 14 Indians). Based on this dataset, Mwanza et al. reported that European descents on 309average, had thinner GCIPL profiles (84.1µm) compared to Asians (89.4µm; P=0.003) and 310Hispanics (88.8µm; P=0.007).34 Following adjustments for age, axial length and RNFL, this inter311ethnic difference became non-significant. However, further evaluation of ethnic differences 312among Asians was not reported, possibly due to the small number of Asians in the sample.34 It 313should also be noted that Indians were not categorised as part of Asians in this Cirrus 314Diversified normative dataset. The mean and 5th percentile value of average GCIPL reported in 315the Cirrus Diversified normative dataset were 84.7 ±7.1µm, and 71µm, respectively. On the 316other hand, the Cirrus Asian normative database was established from 315 health subjects (139 317Japanese, 139 Chinese, and 37 Indians), with a mean and 5th percentile value of the average 318GCIPL to be 83.2 ±5.3µm, and 73µm, respectively. Comparing with these two Cirrus normative 319databases, our overall Asian sample and each ethnic group had smaller mean and 5th percentile 320limits of average GCIPL (Table 2). Taken together, this indicates that the current GCIPL cut-off 321(whether the Diversified or Asian version) deployed in commercial OCT may potentially still 322result in higher false positive misclassifications among Asians. 323

With regards to the regional distribution of GCIPL, we observed that the superior

324hemisphere to be thicker than inferior hemisphere (Table 3). This was also similarly reported by 325Mwanza et al.,34 and concurred with previous histology studies which showed superior area in 326the central macula has more ganglion cells than the inferior area.35, 36 Nevertheless, the regional 327distribution of GCIPL is contrary to the ‘ISNT’ rule of the neuroretinal rim (NRR) in which inferior 328rim is thicker than superior.37 Our previous population study on RNFL also revealed thicker 329RNFL inferiorly than superiorly.11 This discrepancy in regional distribution between GCIPL with 330circumpapillary RNFL and NRR may be due to the foveal position relative to the optic disc. The 331fovea is typically positioned 6.3±3.0° inferior to the horizontal plane of the optic disc center,38

332and the relative lower position of the fovea had been reported as an important determinant of 333the circumpapillary RNFL distribution in normal eyes.39, 40 In this regard, Hood et al. suggested 334that given the lower relative position of the fovea, the inferior temporal side of the 335circumpapillary region is more crowded, with a higher density of retinal ganglion cell axons 336converging into the optic disc than other regions.38 This postulation was further supported by 337previous studies which found that eyes with a more inferior foveal position relative to the optic 338disc indeed had thicker inferior RNFL.39, 40 Taken together, it is plausible to observe thicker 339GCIPL in the superior hemifield but yet thicker inferior circumpapillary RNFL and inferior NRR. 340

Structural glaucomatous damage may appear to be asymmetric across the vertical

341hemifields of retina and between eyes, especially in early stage of glaucoma.18, 41 In this regard, 342recent studies also demonstrated intra-eye GCIPL hemifield asymmetry, and inter-eye GCIPL 343asymmetry measurements to have good performance in detecting early glaucoma.18, 20, 41-44 344Despite these reports, it should be noted that normative profile of intra- and inter-eye GCIPL 345asymmetry had yet been comprehensively evaluated in population-based setting. Such 346information is important to further determine and position the clinical utility of these new 347asymmetry parameters. In this aspect, we observed that the 95th percentile limit of hemifield 348asymmetry to be 6.0 µm in our study eyes, and this was almost identical across the three ethnic 349groups (Table 5). On the other hand, in our subsample with both eyes’ data available (N=3,056; 350Table 6), the 95th percentile limit of inter-eye asymmetry was 5.0 µm, and was also almost 351identical across the three ethnic groups. This is the first study which comprehensively 352established the normative ranges and potential screening cut-offs (based on 95th percentile 353limits) for GCIPL symmetry-related parameters, based on Asian population data. These 354information provide greater clarity on the potential clinical utility of GCIPL symmetry parameters 355for detection, and potentially monitoring of glaucoma in the Asian population. Nevertheless, the 356diagnostic performance (i.e. sensitivity and specificity) of these symmetry-related parameters,

357coupled with the suggested Asian-specific cut-offs, still need to be further evaluated and 358validated in future clinical studies. 359

Among the significant determinants observed, older age and longer axial length were

360most strongly associated with thinner GCIPL profiles. These trends of associations were also 361consistently observed across the three ethnic groups (Figure 4 & 5), further suggesting that age 362and axial length are important factors that need to be taken into account when interpreting 363GCIPL measurements. Based on our cross-sectional findings, it is estimated that the rate of 364GCIPL thinning per 10 years older age band would be approximately 2.5 μm (i.e. 0.25 μm/year). 365Similarly, in a 3-year longitudinal follow-up study of a selected sample of 60 healthy eyes, Lee et 366al. reported the mean rates of GCIPL thinning to be approximately −0.31 μm/year.45 In another 367study of 72 healthy eyes of Chinese adults, Leung et al. similarly reported age-related GCIPL 368thinning to be −0.318 μm/year.46 Collectively, these findings provide fundamental understanding 369on the normative patterns of age-related GCIPL thinning, which may in turn aid in differentiating 370between true glaucoma progression, and age-related GCIPL changes over time. 371

The associations between longer axial length as well as more myopic refraction with

372thinner GCIPL, may be due to the overall thinning of retina in eyes with longer axial length as a 373result of mechanical elongation and stretching of the sclera in more myopic eyes.47 This 374observation was also consistent with previous reports which demonstrated longer axial length to 375be associated with thinner peripapillary RNFL.11 Interestingly, Chinese eyes had longer axial 376length in our study sample, however, after further taking into account axial length in our 377multivariable model (Table 7), the GCIPL profiles in Chinese were still thicker compared to 378Indian eyes. 379

In addition, our study found that female gender was associated with thinner GCIPL. This

380is consistent with our earlier study on Chinese adults.14 In concordance, previous histology 381studies on animal models reported that females had thinner retina than males, due to higher

382ratio of parvocellular retina ganglion cells (the ‘midget’ type of RGC) in females than males, 383further supporting this observation.48 In addition, Mwanza et al. also reported that females had 384thinner GCIPL in the supero- and infero-temporal subsectors.34 On the other hand, Bloch et al. 385did not observe significant association between gender and GCIPL thickness.49 Thus, further 386studies are still needed to ascertain the influence of gender on GCIPL thickness profile. 387

We also observed that CKD was associated with thinner GCIPL in our sample. This may

388be partly explained by the developmental and vasculature similarities between retina and 389kidney.50 A previous study in a western sample similarly reported that lower levels of eGFR was 390associated with thinner ganglion cell complex thickness (GCIPL plus macular RNFL).49 391Nevertheless, as this association has yet been widely reported, future studies are still required 392to confirm its validity. 393

The strengths of our study include a large multi-ethnic Asian population-based sample,

394as well as standardized assessment protocol which allow relatively objective comparison across 395the 3 ethnic groups. Furthermore, evaluation of a comprehensive range of systemic and ocular 396factors allowed us to control for multiple potential confounders in our analyses. However, our 397study also has a few limitations. First, there is a difference in age distributions across the three 398ethnic groups. Despite adjustment of age in our models, there might still be residual 399confounding effect caused by age. To further account for this, we performed relevant sensitivity 400analyses which removed participants younger than 48 years old, and those older than 82 years 401old. Significant ethnic differences were still observed in these sensitivity analyses 402(Supplementary Table 5 & 6). Second, amongst the original 10,049 study eyes with macular 403OCT scans done, 2,528 eyes were excluded due to poor scan quality, retinal pathologies, or 404missing relevant data. Among these excluded eyes, a greater proportion were Indians 405(n=1,126), followed by Malays (n=890) and Chinese (n=512). Thus, the presence of bias cannot 406be entirely ruled out in our final evaluation.

407

In conclusion, in this multi-ethnic Asian population, we demonstrated that Indians have

408thinner GCIPL profiles compared to Chinese and Malay eyes; an observation that is consistent 409with the ethnic difference observed in peripapillary RNFL thickness. Our findings further 410highlight the need of an ethnic-specific normative database, even among Asians. A more refined 411Asian normative database may help to further improve the detection and diagnosis of glaucoma 412in Asians. 413

414REFERENCES 415 Tham YC, Li X, Wong TY, et al. Global prevalence of glaucoma and projections of 4161. 417glaucoma burden through 2040: a systematic review and meta-analysis. Ophthalmology 4182014;121(11):2081-90. Almasieh M, Wilson AM, Morquette B, et al. The molecular basis of retinal ganglion cell 4192. 420death in glaucoma. Prog Retin Eye Res 2012;31(2):152-81. Kuang TM, Zhang C, Zangwill LM, et al. Estimating Lead Time Gained by Optical 4213. 422Coherence Tomography in Detecting Glaucoma before Development of Visual Field Defects. 423Ophthalmology 2015;122(10):2002-9. Mwanza JC, Oakley JD, Budenz DL, et al. Macular ganglion cell-inner plexiform layer: 4244. 425automated detection and thickness reproducibility with spectral domain-optical coherence 426tomography in glaucoma. Invest Ophthalmol Vis Sci 2011;52(11):8323-9. Mwanza JC, Budenz DL, Godfrey DG, et al. Diagnostic performance of optical 4275. 428coherence tomography ganglion cell--inner plexiform layer thickness measurements in early 429glaucoma. Ophthalmology 2014;121(4):849-54. Koh V, Tham YC, Cheung CY, et al. Diagnostic accuracy of macular ganglion cell-inner 4306. 431plexiform layer thickness for glaucoma detection in a population-based study: Comparison with 432optic nerve head imaging parameters. PLoS One 2018;13(6):e0199134. Oddone F, Lucenteforte E, Michelessi M, et al. Macular versus Retinal Nerve Fiber 4337. 434Layer Parameters for Diagnosing Manifest Glaucoma: A Systematic Review of Diagnostic 435Accuracy Studies. Ophthalmology 2016;123(5):939-49. Shin JW, Sung KR, Park SW. Patterns of Progressive Ganglion Cell-Inner Plexiform 4368. 437Layer Thinning in Glaucoma Detected by OCT. Ophthalmology 2018;125(10):1515-25. Bowd C, Zangwill LM, Weinreb RN, et al. Estimating Optical Coherence Tomography 4389. 439Structural Measurement Floors to Improve Detection of Progression in Advanced Glaucoma. 440Am J Ophthalmol 2017;175:37-44. Cirrus HD-OCT with Retinal Nerve Fiber Layer and Macular Normative Databases: 44110. 442United States Food & Drug Administration (FDA). 2009; v. 2018. Ho H, Tham YC, Chee ML, et al. Retinal Nerve Fiber Layer Thickness in a Multiethnic 44311. 444Normal Asian Population: The Singapore Epidemiology of Eye Diseases Study. Ophthalmology 4452018. Knight OJ, Girkin CA, Budenz DL, et al. Effect of race, age, and axial length on optic 44612. 447nerve head parameters and retinal nerve fiber layer thickness measured by Cirrus HD-OCT. 448Arch Ophthalmol 2012;130(3):312-8. Huo YJ, Guo Y, Li L, et al. Age-related changes in and determinants of macular ganglion 44913. 450cell-inner plexiform layer thickness in normal Chinese adults. Clin Exp Ophthalmol 4512018;46(4):400-6.

Koh VT, Tham YC, Cheung CY, et al. Determinants of ganglion cell-inner plexiform layer 45214. 453thickness measured by high-definition optical coherence tomography. Invest Ophthalmol Vis Sci 4542012;53(9):5853-9. Lee YP, Ju YS, Choi DG. Ganglion cell-inner plexiform layer thickness by swept-source 45515. 456optical coherence tomography in healthy Korean children: Normative data and biometric 457correlations. Sci Rep 2018;8(1):10605. Tham YC, Cheung CY, Koh VT, et al. Relationship between ganglion cell-inner plexiform 45816. 459layer and optic disc/retinal nerve fibre layer parameters in non-glaucomatous eyes. Br J 460Ophthalmol 2013;97(12):1592-7. Ghassabi Z, Nguyen AH, Amini N, et al. The Fovea-BMO Axis Angle and Macular 46117. 462Thickness Vertical Asymmetry Across The Temporal Raphe. J Glaucoma 2018;27(11):993-8. Kim YK, Yoo BW, Kim HC, Park KH. Automated Detection of Hemifield Difference 46318. 464across Horizontal Raphe on Ganglion Cell--Inner Plexiform Layer Thickness Map. 465Ophthalmology 2015;122(11):2252-60. Kim YK, Yoo BW, Jeoung JW, et al. Glaucoma-Diagnostic Ability of Ganglion Cell-Inner 46619. 467Plexiform Layer Thickness Difference Across Temporal Raphe in Highly Myopic Eyes. Invest 468Ophthalmol Vis Sci 2016;57(14):5856-63. Chen TC, Hoguet A, Junk AK, et al. Spectral-Domain OCT: Helping the Clinician 46920. 470Diagnose Glaucoma: A Report by the American Academy of Ophthalmology. Ophthalmology 4712018;125(11):1817-27. Foong AW, Saw SM, Loo JL, et al. Rationale and methodology for a population-based 47221. 473study of eye diseases in Malay people: The Singapore Malay eye study (SiMES). Ophthalmic 474Epidemiol 2007;14(1):25-35. Rosman M, Zheng Y, Wong W, et al. Singapore Malay Eye Study: rationale and 47522. 476methodology of 6-year follow-up study (SiMES-2). Clin Exp Ophthalmol 2012;40(6):557-68. Sabanayagam C, Yip W, Gupta P, et al. Singapore Indian Eye Study-2: methodology 47723. 478and impact of migration on systemic and eye outcomes. Clin Exp Ophthalmol 2017;45(8):77947989. Lavanya R, Jeganathan VS, Zheng Y, et al. Methodology of the Singapore Indian 48024. 481Chinese Cohort (SICC) eye study: quantifying ethnic variations in the epidemiology of eye 482diseases in Asians. Ophthalmic Epidemiol 2009;16(6):325-36. Huang OS, Tay WT, Ong PG, et al. Prevalence and determinants of undiagnosed 48325. 484diabetic retinopathy and vision-threatening retinopathy in a multiethnic Asian cohort: the 485Singapore Epidemiology of Eye Diseases (SEED) study. Br J Ophthalmol 2015;99(12):1614-21. Chylack LT, Jr., Wolfe JK, Singer DM, et al. The Lens Opacities Classification System 48626. 487III. The Longitudinal Study of Cataract Study Group. Arch Ophthalmol 1993;111(6):831-6. Sabanayagam C, Yip W, Gupta P, et al. Singapore Indian Eye Study-2: methodology 48827. 489and impact of migration on systemic and eye outcomes. Clin Exp Ophthalmol 2017.

Mwanza JC, Oakley JD, Budenz DL, et al. Macular ganglion cell-inner plexiform layer: 49028. 491automated detection and thickness reproducibility with spectral domain-optical coherence 492tomography in glaucoma. Invest Ophthalmol Vis Sci;52(11):8323-9. Lee HJ, Kim MS, Jo YJ, Kim JY. Ganglion Cell-Inner Plexiform Layer Thickness in 49329. 494Retinal Diseases: Repeatability Study of Spectral-Domain Optical Coherence Tomography. Am 495J Ophthalmol 2015;160(2):283-9 e1. Francoz M, Fenolland JR, Giraud JM, et al. Reproducibility of macular ganglion cell-inner 49630. 497plexiform layer thickness measurement with cirrus HD-OCT in normal, hypertensive and 498glaucomatous eyes. Br J Ophthalmol 2014;98(3):322-8. Chaglasian M, Fingeret M, Davey PG, et al. The development of a reference database 49931. 500with the Topcon 3D OCT-1 Maestro. Clin Ophthalmol 2018;12:849-57. Perez CI, Chansangpetch S, Thai A, et al. Normative Database and Color-code 50132. 502Agreement of Peripapillary Retinal Nerve Fiber Layer and Macular Ganglion Cell-inner Plexiform 503Layer Thickness in a Vietnamese Population. J Glaucoma 2018;27(8):665-73. Xu X, Xiao H, Guo X, et al. Diagnostic ability of macular ganglion cell-inner plexiform 50433. 505layer thickness in glaucoma suspects. Medicine (Baltimore) 2017;96(51):e9182. Mwanza JC, Durbin MK, Budenz DL, et al. Profile and predictors of normal ganglion cell50634. 507inner plexiform layer thickness measured with frequency-domain optical coherence tomography. 508Invest Ophthalmol Vis Sci 2011;52(11):7872-9. Curcio CA, Allen KA. Topography of ganglion cells in human retina. J Comp Neurol 50935. 5101990;300(1):5-25. Curcio CA, Messinger JD, Sloan KR, et al. Human chorioretinal layer thicknesses 51136. 512measured in macula-wide, high-resolution histologic sections. Invest Ophthalmol Vis Sci 5132011;52(7):3943-54. Jonas JB, Fernandez MC, Sturmer J. Pattern of glaucomatous neuroretinal rim loss. 51437. 515Ophthalmology 1993;100(1):63-8. Hood DC, Raza AS, de Moraes CG, et al. Glaucomatous damage of the macula. Prog 51638. 517Retin Eye Res 2013;32:1-21. Choi JA, Kim JS, Park HY, et al. The foveal position relative to the optic disc and the 51839. 519retinal nerve fiber layer thickness profile in myopia. Invest Ophthalmol Vis Sci 2014;55(3):141952026. Jonas RA, Wang YX, Yang H, et al. Optic Disc - Fovea Angle: The Beijing Eye Study 52140. 5222011. PLoS One 2015;10(11):e0141771. Yamada H, Hangai M, Nakano N, et al. Asymmetry analysis of macular inner retinal 52341. 524layers for glaucoma diagnosis. Am J Ophthalmol 2014;158(6):1318-29.e3. Hwang YH, Ahn SI, Ko SJ. Diagnostic ability of macular ganglion cell asymmetry for 52542. 526glaucoma. Clin Exp Ophthalmol 2015;43(8):720-6.

Chen MJ, Yang HY, Chang YF, et al. Diagnostic ability of macular ganglion cell 52743. 528asymmetry in Preperimetric Glaucoma. BMC Ophthalmol 2019;19(1):12. Hou H, Moghimi S, Zangwill LM, et al. Inter-eye Asymmetry of Optical Coherence 52944. 530Tomography Angiography Vessel Density in Bilateral Glaucoma, Glaucoma Suspect, and 531Healthy Eyes. Am J Ophthalmol 2018;190:69-77. Lee WJ, Baek SU, Kim YK, et al. Rates of Ganglion Cell-Inner Plexiform Layer Thinning 53245. 533in Normal, Open-Angle Glaucoma and Pseudoexfoliation Glaucoma Eyes: A Trend-Based 534Analysis. Invest Ophthalmol Vis Sci 2019;60(2):599-604. Leung CKS, Ye C, Weinreb RN, et al. Impact of age-related change of retinal nerve fiber 53546. 536layer and macular thicknesses on evaluation of glaucoma progression. Ophthalmology 5372013;120(12):2485-92. Wu PC, Chen YJ, Chen CH, et al. Assessment of macular retinal thickness and volume 53847. 539in normal eyes and highly myopic eyes with third-generation optical coherence tomography. Eye 540(Lond) 2008;22(4):551-5. Salyer DL, Lund TD, Fleming DE, et al. Sexual dimorphism and aromatase in the rat 54148. 542retina. Brain Res Dev Brain Res 2001;126(1):131-6. Bloch E, Yonova-Doing E, Jones-Odeh E, et al. Genetic and Environmental Factors 54349. 544Associated With the Ganglion Cell Complex in a Healthy Aging British Cohort. JAMA 545Ophthalmol 2017;135(1):31-8. Savige J, Ratnaike S, Colville D. Retinal abnormalities characteristic of inherited renal 54650. 547disease. J Am Soc Nephrol 2011;22(8):1403-15. 548 549

550Figure Legends: 551Figure 1. A) Macular scan of the right eye in Cirrus HD-OCT showing a colour-coded GCIPL 552thickness map within the 14.13 mm2 elliptical annulus area, centred on the fovea; B) single 553horizontal B scan of the macula showing segmentation of the GCIPL (boundaries of layer 554demarcated by the purple and yellow lines); C) the 6 subsectors of the GCIPL parameters within 555the elliptical annulus area; D) illustration of inter-eye average GCIPL symmetry evaluation 556between right and left eye (as denoted by the yellow lines); E) illustration of intra-eye GCIPL 557hemifield symmetry evaluation (as denoted by the dark blue lines). 558 559Figure 2. Comparison of ganglion cell-inner plexiform layer thickness parameters by ethnic 560groups. Graph shows mean and 95% confidence intervals. 561† denotes P<0.001 for pairwise comparisons between Malays and Chinese. 562‡ denotes P=0.006 for pairwise comparisons between Malays and Chinese. 563* denotes P<0.001 for pairwise comparisons between Indians with Malays and Chinese. 564 565Figure 3. Normative distribution of average (a), superior hemisphere (b), and inferior 566hemisphere (c) ganglion cell-inner plexiform layer thickness by ethnic groups. White region 567represents the thickest 5% of measurements (≥95 th percentile), green region represents the 56890% of measurements ranging between 5th and 95th percentiles, yellow region represents the 569thinnest 5% of measurements (≤5th percentile), and red region represents the thinnest 1% of 570measurements (≤1st percentile). Dotted vertical yellow line indicates the 5 th percentile value, and 571dotted vertical black line indicates the mean value. 572 573Figure 4. Association between age and average ganglion cell-inner plexiform layer thickness by 574ethnic groups. Data presented are in mean ± standard error. 575 576Figure 5. Association between axial length and average ganglion cell-inner plexiform layer 577thickness. Data presented are in mean ± standard error.

578ACKNOWLEDGMENTS 579Financial Support: The author(s) have no proprietary or commercial interest in any materials 580discussed in this article. This study was supported by the National Medical Research Council, 581Republic of Singapore grant nos.: NMRC/1249/2010, NMRC/CIRG/1371/2013, 582NMRC/CIRG/1417/2015, and NMRC/CIRG/1488/2018; NMRC/CSASI/0012/2017 [C.-Y.C.] and 583NMRC/MOH-TA18nov-0002 [Y.-C.T.]; and the Biomedical Research Council, Singapore, 584Republic of Singapore (grant no.: 08/1/35/19/550). 585 586Conflict of Interest: No conflicting relationship exists for any author. 587 588Author Contributions: 589Conception and design: Y.C.Tham, T Aung, T.Y.Wong, C.Y.Cheng. 590Analysis and interpretation: Y.C.Tham, M.L.Chee, W.Dai, Z.W.Lim, S.Majithia, C.Y.Cheung, and 591C.Y.Cheng. 592Data collection: Y.C.Tham, M.L Chee, W.Dai, Z.W.Lim, S.Majithia, R. Siantar, S.Thakur, T.Rim, 593C.Y.Cheung, C.Sabanayagam, and C.Y.Cheng. 594Obtained funding: T Aung, T.Y.Wong, and C.Y.Cheng. 595Manuscript preparation and overall responsibility: Y.C.Tham, M.L.Chee, W.Dai, Z.W.Lim, 596S.Majithia, R. Siantar, S.Thakur, T.Rim, C.Y.Cheung, C.Sabanayagam, T. Aung, T.Y.Wong, and 597C.Y.Cheng. 598

Precis GCIPL thickness profile in Indians is significantly thinner compared with Chinese and Malays, suggesting that ethnic-specific OCT normative database is needed, even among Asians.

Table 1: Characteristics of Study Participants. Demographic and systemic characteristics

Chinese (N= 1,625)

Malay (N = 1,212)

Indian (N = 1,627)

P-value†

Age, years

55.2 (7.5)

60.4 (8.8)

60.5 (7.8)

<0.001

Gender (Female)

842 (51.8)

647 (53.4)

822 (50.5)

0.32

Body mass index, kg/m2

23.6 (3.5)

27.0 (5.2)

26.6 (4.6)

<0.001

Current smoker, n

219 (13.5)

233 (19.2)

209 (12.9)

<0.001

Diabetes, n

176 (10.8)

325 (26.8)

578 (35.5)

<0.001

Hyperlipidaemia, n

657 (41.4)

657 (56.2)

956 (61.4)

<0.001

40 (2.5)

150 (12.5)

89 (5.7)

<0.001

Hypertension, n

789 (48.6)

812 (67.1)

1020 (62.7)

<0.001

Ocular Characteristics

Chinese (n= 2,635)

Malay (n= 2,084)

Indian (n= 2,801)

P-value‡

Spherical Equivalent, dioptre

-0.82 (2.38)

0.16 (1.82)

0.29 (1.81)

<0.001

24.0 (1.3)

23.6 (1.0)

23.5 (1.0)

<0.001

553.3 (32.6)

542.4 (33.8)

541.3 (33.0)

<0.001

IOP, mmHg

14.3 (2.9)

14.5 (3.0)

15.3 (2.8)

<0.001

OCT signal strength

9.3 (0.9)

9.1 (1.0)

8.9 (1.1)

<0.001

Disc area, mm2

1.97 (0.38)

2.05 (0.42)

1.97 (0.38)

<0.001

Vertical cup-to-disc ratio

0.50 (0.14)

0.51 (0.14)

0.55 (0.13)

<0.001

Presence of any cataract*, n

398 (15.6)

1012 (51.5)

1476 (59.5)

<0.001

Chronic kidney disease, n

Axial Length, mm CCT, µm

N = number of subjects; n = number of eyes; CCT = central corneal thickness; IOP = intraocular pressure; OCT = optical coherence tomography.

Data presented are mean (SD) for continuous variables, and number (%) for categorical variables.

†Comparison across three ethnicities was evaluated by ANOVA for continuous variables, and chi-square tests for categorical variables.

‡Eye-specific data. Comparison across three ethnicities was evaluated by linear regression models in generalized estimating equations (GEE) for continuous variables, and by logistic regression in GEE for categorical variables.

*Defined based on Lens Opacity Classification System (LOCS III).

Table 2. Normative Distribution of Average Ganglion Cell-Inner Plexiform Layer (GCIPL) Thickness by Ethnic and Age groups.

Ethnicity

Overall

Chinese

Malay

Indian

Age (year)

Number (n)

40 - 49

Average GCIPL thickness, µm Mean (SD)

50th Percentile

5th Percentile

1st Percentile

Range

1,280

83.7 (5.9)

84

74

71

50 - 104

50 - 59

3,494

81.3 (6.5)

81

70

66

34 - 106

60 - 69

2,039

79.4 (6.6)

80

69

62

28 - 106

70+

707

75.1 (7.4)

76

62

57

28 - 97

Total

7,520

80.6 (6.9)

81

69

63

28 - 106

40 - 49

892

84.0 (5.8)

84

75

72

67 - 104

50 - 59

1,121

82.7 (6.0)

83

72

67

58 - 103

60 - 69

510

80.9 (6.0)

81

71

68

60 - 98

70+

112

79.0 (6.2)

80

68

61

58 - 92

Total

2,635

82.6 (6.1)

83

72

68

58 - 104

40 - 49

271

84.1 (5.9)

84

74

69

63 - 101

50 - 59

878

83.1 (6.1)

83

73

68

64 - 106

60 - 69

652

80.7 (6.6)

81

70

60

28 - 98

70+

283

76.1 (6.7)

77

65

61

57 - 95

Total

2,084

81.5 (6.8)

82

70

64

28 - 106

40 - 49

117

80.7 (6.2)

80

72

68

50 - 94

50 - 59

1,495

79.2 (6.4)

79

68

64

34 - 100

60 - 69

877

77.5 (6.5)

78

67

60

43 - 106

70+

312

72.9 (7.6)

73.5

60

54

28 - 97

Total

2,801

78.0 (6.9)

78

67

60

28 - 106

Table 3. Comparison of Ganglion Cell-Inner Plexiform Layer (GCIPL) Thickness Parameters between Ethnic Groups. Chinese (n= 2,635)

Malay (n= 2,084)

Indian (n= 2,801)

P-value (Chinese vs Malay)

P-value (Chinese vs Indian)

P-value (Malay vs Indian)

Average

82.6 (6.1)

81.5 (6.8)

78.0 (6.9)

<0.001

<0.001

<0.001

Superior hemisphere

83.5 (6.2)

82.3 (6.9)

78.4 (6.9)

<0.001

<0.001

<0.001

Supero-temporal

81.5 (6.2)

80.3 (6.8)

76.4 (6.7)

<0.001

<0.001

<0.001

Superior

83.3 (6.7)

82.2 (7.4)

78.3 (7.4)

<0.001

<0.001

<0.001

Supero-nasal

85.5 (6.8)

84.2 (7.5)

80.6 (7.6)

<0.001

<0.001

<0.001

Inferior hemisphere

81.7 (6.2)

80.8 (6.9)

77.7 (7.1)

<0.001

<0.001

<0.001

Infero-nasal

83.0 (6.7)

82.0 (7.5)

78.9 (7.6)

<0.001

<0.001

<0.001

Inferior

80.0 (6.6)

79.4 (7.3)

76.4 (7.5)

0.006

<0.001

<0.001

Infero-temporal

82.1 (6.4)

81.0 (7.0)

77.7 (7.2)

<0.001

<0.001

<0.001

GCIPL thickness parameters, μm

Table 4: Normative Distribution of Ganglion Cell-Inner Plexiform Layer (GCIPL) Thickness Parameters, by Ethnic Groups. GCIPL Thickness (μm) Mean (SD)

10th Percentile

5th Percentile

1st Percentile

Range

Average

80.6 (6.9)

72

69

63

28 - 106

Superior hemisphere

81.3 (7.1)

72

70

63

27 - 115

Supero-temporal

79.3 (6.9)

71

68

62

22 - 104

Superior

81.2 (7.5)

72

69

62

25 - 110

Supero-nasal

83.3 (7.6)

74

71

63

29 - 140

Inferior hemisphere

80.0 (7.0)

71

68

62

27 - 106

Infero-nasal

81.2 (7.5)

72

69

61

31 - 110

Inferior

78.5 (7.3)

69

66

59

28 - 104

Infero-temporal

80.1 (7.2)

71

68

61

23 - 115

Average

82.6 (6.1)

75

72

68

58 - 104

Superior hemisphere

83.5 (6.2)

76

73

68

58 - 106

Superiotemporal

81.5 (6.2)

74

71

65

57 - 104

Superior

83.3 (6.7)

75

72

67

55 - 108

Superionasal

85.5 (6.8)

77

74

69

56 - 110

Inferior hemisphere

81.7 (6.2)

74

71

66

57 - 104

Inferionasal

83.0 (6.7)

75

72

67

58 - 108

Inferior

80.0 (6.6)

72

69

63

54 - 102

Inferiotemporal

82.1 (6.4)

74

72

65

54 - 105

Average

81.5 (6.8)

73

70

64

28 - 106

Superior hemisphere

82.3 (6.9)

73

70

65

28 - 105

Supero-temporal

80.3 (6.8)

72

69

63

23 - 103

Superior

82.2 (7.4)

73

70

64

25 - 105

Supero-nasal

84.2 (7.5)

75

72

66

29 - 127

Inferior hemisphere

80.8 (6.9)

72

69

62

27 - 106

Infero-nasal

82.0 (7.5)

73

69

62

31 - 110

Inferior

79.4 (7.3)

70

67

60

28 - 104

Infero-temporal

81.0 (7.0)

72

69

61

23 - 109

Average*

78.0 (6.9)

70

67

60

28 - 106

Superior hemisphere

78.4 (6.9)

70

67

61

27 - 115

Superiotemporal

76.4 (6.7)

68

66

59

22 - 97

Superior

78.3 (7.4)

69

66

59

26 - 110

Superionasal

80.6 (7.6)

71

68

60

34 - 140

Inferior hemisphere

77.7 (7.1)

69

66

59

28 - 101

Inferionasal

78.9 (7.6)

69

66

59

32 - 103

Inferior

76.4 (7.5)

67

64

57

28 - 101

Inferiotemporal

77.7 (7.2)

69

66

58

23 - 115

Overall (N = 7,520)

Chinese (n = 2,635)

Malay (n = 2,084)

Indian (n = 2,801)

Table 5. Normative Distribution of Intra-Eye Ganglion Cell-Inner Plexiform Layer (GCIPL) Hemifield Asymmetry, by Ethnic and Age groups.

Ethnicity

Overall

Chinese

Malay

Indian

Age (year)

Number (n)

40 - 49

Absolute Difference between superior and inferior hemisphere GCIPL thickness, µm Mean (SD)

50th Percentile

95th Percentile

99th Percentile

Range

1,280

2.3 (1.9)

2.0

5.7

8.7

0.0 - 17.7

50 - 59

3,494

2.2 (2.0)

1.7

5.7

9.3

0.0 - 23.3

60 - 69

2,039

2.4 (2.3)

2.0

6.0

10.0

0.0 - 30.7

70+

707

2.7 (2.5)

2.0

7.7

12.3

0.0 - 18.3

Total

7,520

2.3 (2.1)

2.0

6.0

9.7

0.0 - 30.7

40 - 49

892

2.4 (2.0)

2.0

6.0

8.3

0.0 - 17.7

50 - 59

1,121

2.4 (2.0)

2.0

5.7

8.3

0.0 - 23.0

60 - 69

510

2.6 (2.2)

2.3

6.3

10.3

0.0 - 19.0

70+

112

3.0 (2.2)

2.7

7.7

8.0

0.0 - 10.0

Total

2,635

2.5 (2.0)

2.0

6.0

8.7

0.0 - 23.0

40 - 49

271

2.2 (1.9)

1.7

5.3

10.7

0.0 - 12.3

50 - 59

878

2.3 (1.9)

2.0

5.7

8.7

0.0 - 16.3

60 - 69

652

2.5 (2.2)

2.0

6.0

9.3

0.0 - 23.3

70+

283

2.9 (2.5)

2.0

7.7

12.7

0.0 - 17.0

Total

2,084

2.4 (2.1)

2.0

6.0

9.3

0.0 - 23.3

40 - 49

117

2.1 (1.9)

1.7

5.7

9.7

0.0 - 11.7

50 - 59

1,495

2.1 (2.0)

1.7

5.3

9.7

0.0 - 23.3

60 - 69

877

2.3 (2.4)

1.7

6.0

10.7

0.0 - 30.7

70+

312

2.5 (2.6)

1.7

8.3

12.7

0.0 - 18.3

Total

2,801

2.2 (2.2)

1.7

5.7

10.3

0.0 - 30.7

Table 6. Normative Distribution of Inter-Eye Ganglion Cell-Inner Plexiform Layer (GCIPL) Asymmetry, by Ethnic and Age Groups.

Ethnicity

Overall

Chinese

Malay

Indian

Age (year)

Number (n)*

40 - 49

Absolute Inter-eye Difference on average GCIPL thickness, µm Mean (SD)

50th Percentile

95th Percentile

99th Percentile

Range

560

1.5 (1.7)

1

4

9

0 - 21

50 - 59

1,487

1.6 (2.2)

1

4

8

0 - 57

60 - 69

781

1.8 (2.4)

1

5

9

0 - 33

70+

228

2.5 (3.5)

2

7

13

0 - 43

Total

3,056

1.7 (2.3)

1

5

9

0 - 57

40 - 49

378

1.4 (1.5)

1

4

8

0 - 11

50 - 59

433

1.5 (1.6)

1

4

8

0 - 11

60 - 69

168

1.8 (2.4)

2

4

8

0 - 28

70+

31

2.1 (1.4)

2

4

7

0-7

Total

1,010

1.5 (1.7)

1

4

8

0 - 28

40 - 49

128

1.6 (2.4)

1

5

10

0 - 21

50 - 59

385

1.6 (1.6)

1

4

9

0 - 12

60 - 69

268

1.8 (2.2)

1

6

12

0 - 18

70+

91

2.2 (2.0)

2

6

9

0-9

Total*

872

1.7 (2.0)

1

5

9

0 - 21

40 - 49

54

1.4 (1.2)

1

3

6

0-6

50 - 59

669

1.6 (2.8)

1

4

11

0 - 57

60 - 69

345

1.8 (2.5)

1

5

8

0 - 33

70+

106

2.9 (4.8)

2

8

16

0 - 43

Total*

1,174

1.8 (2.9)

1

5

11

0 - 57

*Denote number of individuals. Only subjects with GCIPL data available in both eyes were included for this analysis.

Table 7: Multivariable Regression Analysis on the Associations between Demographic, Systemic and Ocular factors with Average Ganglion Cell-Inner Plexiform Layer (GCIPL) thickness. Average GCIPL (N = 7,520)

Superior hemisphere*(N = 7,520)

Inferior hemisphere^(N = 7,520)

β

95% CI

β'

P Value

β

95%CI

β'

P Value

β

95%CI

β'

P Value

Age, per decade

-2.51

(-2.81, -2.21)

-2.23

<0.001

-2.42

(-2.73, -2.11)

-2.16

<0.001

-2.58

(-2.89, -2.27)

-2.30

<0.001

Female gender

-1.57

(-1.99, -1.16)

-0.79

<0.001

-1.56

(-1.99, -1.13)

-0.78

<0.001

-1.57

(-1.99, -1.15)

-0.79

<0.001

Ethnicity Chinese

Reference

Reference

Reference

Malay

-0.27

(-0.78, 0.24)

-0.12

0.296

-0.41

(-0.95, 0.12)

-0.19

0.127

-0.11

(-0.62, 0.41)

-0.05

0.685

Indian

-3.43

(-3.93, -2.92)

-1.64

<0.001

-3.88

(-4.40, -3.35)

-1.85

<0.001

-2.93

(-3.44, -2.42)

-1.40

<0.001

Axial Length, mm

-1.54

(-1.71, -1.37)

-1.71

<0.001

-1.42

(-1.60, -1.24)

-1.57

<0.001

-1.62

(-1.80, -1.45)

-1.80

<0.001

Spherical equivalent, per negative dioptre†

-0.68

(-0.76, -0.60)

-1.58

<0.001

-0.65

(-0.74, -0.56)

-1.51

<0.001

-0.72

(-0.80, -0.63)

-1.66

<0.001

Any cataract, yes

-0.50

(-0.83, -0.17)

-0.24

0.003

-0.61

(-0.98, -0.25)

-0.30

0.001

-0.43

(-0.78, -0.08)

-0.21

0.017

OCT signal strength

0.04

(-0.06, 0.13)

0.04

0.418

0.04

(-0.07, 0.15)

0.04

0.462

0.03

(-0.08, 0.13)

0.03

0.641

Optic Disc area, mm2

0.78

(0.43, 1.14)

0.31

<0.001

0.80

(0.43, 1.18)

0.31

<0.001

0.92

(0.53, 1.31)

0.36

<0.001

Diabetes, yes

-0.41

(-0.92, 0.10)

-0.17

0.114

-0.38

(-0.91, 0.14)

-0.16

0.150

-0.43

(-0.95, 0.08)

-0.18

0.099

CKD, yes

-1.49

(-2.58, -0.39)

-0.30

0.008

-1.36

(-2.50, -0.22)

-0.28

0.020

-1.60

(-2.69, -0.50)

-0.33

0.004

Hyperlipidemia, yes

-0.08

(-0.48, 0.31)

-0.04

0.678

-0.11

(-0.52, 0.29)

-0.06

0.588

-0.04

(-0.45, 0.36)

-0.02

0.830

Hypertension, yes

-0.18

(-0.60, 0.24)

-0.09

0.412

-0.20

(-0.62, 0.23)

-0.10

0.372

-0.17

(-0.59, 0.26)

-0.08

0.438

Body mass index, kg/m2

0.02

(-0.02, 0.07)

0.11

0.295

0.03

(-0.02, 0.08)

0.14

0.218

0.02

(-0.03, 0.06)

0.07

0.486

Current smoking, yes

-0.07

(-0.66, 0.53)

-0.02

0.829

-0.20

(-0.81, 0.41)

-0.07

0.512

0.07

(-0.54, 0.68)

0.02

0.825

IOP, mmHg

-0.02

(-0.10, 0.06)

-0.06

0.613

0.00

(-0.08, 0.07)

0.00

0.971

-0.03

(-0.11, 0.04)

-0.09

0.413

CCT, µm

0.00

(-0.01, 0.00)

-0.04

0.671

0.00

(-0.01, 0.00)

-0.05

0.621

0.00

(-0.01, 0.00)

-0.04

0.692

N=number of eyes; CI=confidence interval; OCT= optical coherence tomography; CKD= chronic kidney disease; IOP= intraocular pressure; CCT= central corneal thickness. β denotes the change in GC-IPL thickness (µm) per unit change in predictor variables. β' represents standardized coefficient. *Superior hemisphere parameter was averaged from measurements of superiotemporal, superior and superionasal subfields. ^Inferior hemisphere parameter was averaged from measurements of inferiotemporal, inferior and inferionasal subfields. †Results for spherical equivalent was based on multivariable model excluding those with history of cataract surgery. Spherical equivalent and axial length were evaluated in separate models.

- . . - Chinese ~ Malay

----+--- Indian

g L,--------,---------,--------.---------.--------,-~ Superiortemporal

Superior

Superiornasal

lnferiortemporal

Inferior

lnferiomasal

0

'"

-----...--- Malay

-----+----

India n

-----+----- Ch inese

~

1. "'

. c

~

0

"'

0

~ ~

~

"-

"

"'

0

.

u

"'~ "

~

~


0

w

P trend <0.001 across all ethnic groups.

L,-----,-----,-----,-----,-----,-----,----~

S:45

50

55

60

65

Age (years)

70

75


0

'"

-----...--- Malay

-----+----

Indian

-----+----- Ch inese

~

1. "'

. c

~

0

"'

0

i'

>-~

"-

u

~

"

"'

~


0

w

P trend <0.001 across all ethnic groups.

L,------,------,------,------,------,-----~

:52 1

22

23

24 Axial Length (mm )

25

26

!?.27