Article Text

Download PDFPDF

Compensation of retinal nerve fibre layer thickness as assessed using optical coherence tomography based on anatomical confounders
  1. Jacqueline Chua1,
  2. Florian Schwarzhans2,
  3. Duc Quang Nguyen3,
  4. Yih Chung Tham1,
  5. Josh Tjunrong Sia1,
  6. Claire Lim1,
  7. Shivani Mathijia3,
  8. Carol Cheung4,
  9. Aung Tin5,
  10. Georg Fischer2,
  11. Ching-Yu Cheng3,
  12. Clemens Vass6,
  13. Leopold Schmetterer7
  1. 1 Singapore Eye Research Institute, Singapore National Eye Centre, Singapore, Singapore
  2. 2 Section for Medical Information Management and Imaging, Center for Medical Statistics, Informatics and Intelligent Systems, Medical University of Vienna, Vienna, Austria
  3. 3 Ocular Epidemiology Research Group, Singapore Eye Research Institute, Singapore, Singapore
  4. 4 Ophthalmology and Visual Sciences, The Chinese University of Hong Kong, Ma Liu Shui, Hong Kong
  5. 5 Singapore National Eye Centre, Singapore, Singapore
  6. 6 Ophthalmology, Medical University of Vienna, Vienna, Austria
  7. 7 Department of Clinical Pharmacology, Medical University of Vienna, Vienna, Austria
  1. Correspondence to Professor Leopold Schmetterer, Department of Clinical Pharmacology, Medical University of Vienna, Vienna 1090, Austria; leopold.schmetterer{at}seri.com.sg

Abstract

Background/Aims To compensate the retinal nerve fibre layer (RNFL) thickness assessed by spectral-domain optical coherence tomography (SD-OCT) for anatomical confounders.

Methods The Singapore Epidemiology of Eye Diseases is a population-based study, where 2698 eyes (1076 Chinese, 704 Malays and 918 Indians) with high-quality SD-OCT images from individuals without eye diseases were identified. Optic disc and macular cube scans were registered to determine the distance between fovea and optic disc centres (fovea distance) and their respective angle (fovea angle). Retinal vessels were segmented in the projection images and used to calculate the circumpapillary retinal vessel density profile. Compensated RNFL thickness was generated based on optic disc (ratio, orientation and area), fovea (distance and angle), retinal vessel density, refractive error and age. Linear regression models were used to investigate the effects of clinical factors on RNFL thickness.

Results Retinal vessel density reduced significantly with increasing age (1487±214 µm in 40–49, 1458±208 µm in 50–59, 1429±223 µm in 60–69 and 1415±233 µm in ≥70). Compensation reduced the variability of RNFL thickness, where the effect was greatest for Chinese (10.9%; p<0.001), followed by Malays (6.6%; p=0.075) and then Indians (4.3%; p=0.192). Compensation reduced the age-related RNFL decline by 55% in all participants (β=−3.32 µm vs β=−1.50 µm/10 years; p<0.001). Nearly 62% of the individuals who were initially classified as having abnormally thin RNFL (outside the 99% normal limits) were later reclassified as having normal RNFL.

Conclusions RNFL thickness compensated for anatomical parameters reduced the variability of measurements and may improve glaucoma detection, which needs to be confirmed in future studies.

  • imaging
http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

View Full Text

Statistics from Altmetric.com

Introduction

Glaucoma is the leading cause of blindness worldwide1 and reduction of retinal nerve fibre layer (RNFL) thickness imaged with optical coherence tomography (OCT) is an early hallmark of glaucoma.2 Despite significant improvements with OCT, there remain limitations in relation to its use for glaucoma diagnosis.3 First, the OCT modality does not adjust for ocular biometry and morphology such as optic disc (size and area) and disc-fovea angle,4–6 retinal vessel position,6–8 refractive error7 9 and axial length,10 which can reduce the precision of RNFL thickness measurement. Second, retinal vessels reduce with age, systemic comorbidities11 and also in glaucoma.12 This further complicates matter when RNFL thickness measured from OCT devices also includes retinal vessels. Hence, the thinning of measured RNFL may also represent the narrowing of retinal vessels. Overall, an imprecise RNFL thickness may result in an inaccurate clinical assessment of glaucoma.

We developed a comprehensive multivariate model which compensates the influence of anatomical parameters as mentioned above and showed it improved the limits of RNFL variability in healthy eyes.13 However, this model was tested on a small sample of healthy predominantly young Caucasians. Population-based data on the normal RNFL variability of a multiethnic population are vital, because these are more generalisable and less susceptible to selection biases. Singapore consists of people of Chinese, Malaysian and Indian ancestries, with interethnic difference in central corneal thickness,14 corneal biomechanics,15 16 RNFL thickness,17 macular thickness,18 retinal vessels,19 refractive error20 and axial length.20 Hence, this multiethnic Asian population presents as an attractive sample to examine the effectiveness of this model, and may potentially impact the utility of RNFL normative databases.

In this study, we applied a model which compensates the RNFL thickness for anatomical parameters,13 on a multiethnic Asian population-based sample, free of ocular diseases and also investigated the associations with demographic and clinical factors. We hypothesised that after compensating for anatomical parameters, the interindividual variability of compensated RNFL will be reduced and anatomical factors that were initially correlated with measured RNFL will be attenuated.

Methods

Study participants

Participants were enrolled from the Singapore Epidemiology of Eye Diseases (SEED) programme, comprising Chinese, Malays and Indians aged 40–80 years. Study methodology was identical and has been described elsewhere.21–23 Data were derived from 2698 participants, of which 1076 were Chinese (year 2009–2011), 704 Malays (year 2010–2014) and 918 Indians (year 2013–2015). All study participants were provided with written informed consent in adherence to the Declaration of Helsinki.

Ocular examinations

Participants underwent an ocular examination including visual acuity, subjective refraction, slit-lamp biomicroscopy, gonioscopy, intraocular pressure (IOP) measurement using Goldmann applanation tonometry, measurement of central corneal thickness using an ultrasound pachymeter (CCT Advent; Mentor O & O, Norwell, USA), corneal curvature and refractive error using an autorefractor (Canon RK-5 Autorefractor Keratometer; Canon, Japan), axial length using non-contact partial coherence interferometry (IOL Master V3.01, Carl Zeiss Meditec, Germany) and posterior segment examination at the slit-lamp using a 78 dioptre lens.14

Other measurements

Detailed interviewer-administered questionnaire was used to collect demographic data, medication and ocular surgery histories. Blood pressure was measured using a digital automatic blood pressure monitor (Dinamap model Pro Series DP110X-RW, GE Medical Systems Information Technologies, Milwaukee).24 Non-fasting venous blood samples were collected for biochemistry analysis. Diabetes mellitus was defined as random glucose of ≥11.1 mmol/L, diabetic medication usage or a physician diagnosis of diabetes. Hypertension was defined as systolic blood pressures ≥140 mm Hg or diastolic blood pressures ≥90 mm Hg or physician-diagnosed hypertension or self-reported history of hypertension. Hyperlipidaemia was defined as total cholesterol ≥6.2 mmol/L or self-reported use of lipid-lowering drugs.

OCT imaging

Participants underwent Cirrus SD-OCT (Carl Zeiss Meditec, Dublin, CA) imaging after pupil dilation. We acquired optic disc 200×200 cube scan25 and macular 200×200 cube scan.18 Trained graders masked to the participant characteristics reviewed the quality of OCT scans. Poor-quality images (signal strength less than 6 and/or movement artefacts within the RNFL measurement circle) were excluded from the analysis.

Automated parameter extraction

Automated parameter extraction was performed, as previously published.13 Briefly, we registered the optic disc and macular cube scans of each eye, using an automatic OCT layer segmentation algorithm (Retinal Image Analysis Lab, Iowa Institute for Biomedical Imaging, Iowa City, IA).26–28 We then extracted optic disc parameters including its area, orientation (angle between the horizontal axis and the major axis of the optic disc) and ratio (quotient between major and minor axes) from the spectral-domain OCT (SD-OCT). We also segmented the retinal vessel tree from the optic disc and macular images.13 We considered all vessels that are within a band of diameter around the centre of the optic disc, extending from 3.28 to 3.64 mm to integrate a 256-sector vessel profile. The thickness values of all vessels within the same sector were summed up. After which, we generated a semicontinuous profile, similar to RNFL distribution, that expresses vessel density corresponding to the RNFL measurement area, where each individual discrete profile of vessel distribution was convoluted with a Gaussian window, as described previously.29 30 From the registered image and considering fovea centre as automatically determined in SD-OCT, we obtained the fovea parameters: first, the fovea distance, which corresponds to the distance between optic disc and fovea centres; second, the fovea angle, which corresponds to the angle between a line connecting fovea and optic disc centres and a horizontal line passing through the optic disc centre.

Inclusion/exclusion criteria

Of the 14 908 eyes of 7454 SEED participants, 9735 eyes of 5221 had available OCT data (figure 1). We excluded 2902 eyes due to presence of glaucoma/glaucoma suspect/self-reported glaucoma,31 retinopathies32 and age-related macular degeneration,33 2575 eyes with poor-quality OCT scans and 299 eyes due to missing clinical variables. In total, 3959 eyes of 2698 participants who were without eye diseases had high-quality SD-OCT images. If bilateral OCT data were available for each participant, one eye was randomly selected for subsequent analyses. This left a total of 2698 eyes (2698 participants), including 1076 Chinese eyes (1076 participants), 704 Malay eyes (704 participants) and 918 Indian eyes (918 participants) for analysis.

Figure 1

Flow chart of the inclusion and exclusion criteria of study eyes. Systematic selection process identified 2698 eyes, comprising 1076 Chinese, 704 Malays and 918 Indians with high-quality spectral-domain optical coherence tomography (SD-OCT) images from individuals without eye diseases for analysis. SCES, Singapore Chinese Eye Study; SEED, Singapore Epidemiology of Eye Diseases; SIMES, Singapore Malay Eye Study; SINDI, Singapore Indian Eye Study.

Statistical analyses

Primary outcomes were measured and compensated RNFL thickness measurements. Shapiro-Wilk test was used to assess the normality of the distribution of the continuous variables. To compare the variables between groups, one-way analysis of variance and χ2 tests were performed. P trend was assessed by modelling the median value of the various age groups in the linear regression analysis (non-normally distributed continuous) or with the use of the χ2 test for linear trends (categorical). To compare the variation of the compensated RNFL with the variation when using the conventional approach without compensation, base asymptotic test for the equality of coefficients of variation was used. Linear regression models were used to investigate the effects of demographic and clinical factors (independent variables) on RNFL thickness (dependent variable).

We further classified the 2698 individuals into three categories based on the thickness of their measured RNFL measurements: (1) normal individuals, where 95% of RNFL measurements fall in the green category (within 95% of normal limits; n=2563); (2) suspect, the thinnest 5% of RNFL measurements fall in the yellow category (within 99% of normal limits; n=108); and (3) abnormal individuals, the thinnest 1% of RNFL measurements fall in the red category (outside of 99%; n=27). We then determined the range (minimum to maximum) of measured RNFL measurements within each category. After which, we applied the same cut-off to the entire distribution of compensated RNFL measurements. We then took note of how many persons were reclassified, that is, ‘moved’ or stayed within category after compensation. We then examined the concordance of these two classifications. As such, we examined the number of subjects who were erroneously categorised into the risk categories in the measured data set. To examine the impact that compensation had on RNFL thickness classification, a contingency table was used to display the frequency distribution of RNFL thickness reclassification. Data were analysed with statistical software (STATA V13.1; StataCorp).

Results

The mean±SD age of the 2698 participants was 57±8 years for Chinese, 59±8 years for Malays and 60±8 years for Indians (p<0.001) and 51% were women (p=0.279; table 1). Differences between the ethnic groups were found for age, diabetes, hypertension, hyperlipidaemia, lens status, IOP, central corneal thickness (CCT), corneal curvature, clinical vertical cup-to-disc ratio, optic disc and fovea parameters, retinal vessel density, spherical refractive error and axial length (p<0.05; table 1). Compensated RNFL was thicker than measured RNFL for the overall population (95.0±9.7 µm vs 93.5±10.6 µm; p<0.001), Chinese (97.8±8.7 µm vs 96.6±9.6 µm; p=0.002) and Indians (90.7±9.8 µm vs 87.8±9.9 µm; p<0.001) whereas there were no differences between the compensated and measured RNFL for Malays (96.3±9.1 µm vs 96.0±9.7 µm; p=0.517).

Table 1

Characteristics of included participants among the three ethnic groups

Characteristics of the participants were further summarised according to age groups (table 2). Older participants tended to have diabetes, hypertension, hyperlipidaemia, pseudophakia, thinner CCT, reduced density of retinal vessels, long-sightedness, shorter axial length, and thinner measured and corrected RNFL (p trends <0.05). Figure 2 further showed the distribution of retinal vessel density among the varying age groups. For every 1 year increase in age, there was a −28.09 µm reduction in retinal vessel density in the overall population (95% CI −38.23 to 17.96; p<0.001).

Figure 2

Distribution of retinal vessel density of varying age groups. Retinal vessel density reduced with age. For every 1 year increase in age, there was a −28.09 µm reduction in retinal vessel density in the overall population (95% CI −38.23 to 17.96; p<0.001).

Table 2

Characteristics of included participants among the varying age groups

Coefficients of variation of measured and compensated RNFL thicknesses were compared in each of four quadrants, stratified by ethnicity (table 3). Compensation reduced the variability of global RNFL in the overall Asian population (9.6%; p<0.001), where the greatest improvement was found for the temporal quadrant (11.8%), followed by the superior quadrant (9.3%), inferior quadrant (8.9%) and, lastly, nasal quadrant (8.1%; all p<0.001). In terms of ethnicity, the effect was greatest among the Chinese (10.9%; p<0.001), followed by Malays (6.6%; p=0.075) and then Indians (4.3%; p=0.192).

Table 3

Coefficient of variations for measured and compensated RNFL thickness values among the three ethnic groups

Online supplementary table shows the univariate linear regression of demographic and clinical factors on measured and compensated RNFL, stratified by ethnicity. Age was negatively correlated with measured and compensated RNFL thicknesses in the overall population (β=−3.32 µm; p<0.001 vs β=−1.50 µm; p=0.004) and Malays (β=−2.83 µm; p<0.001 vs β=−1.17 µm; p=0.004) whereas age was only correlated with measured RNFL and not with compensated RNFL for Chinese (β=−2.26 µm; p<0.001 vs β=−0.16 µm; p=0.666) and Indians (β=−2.33 µm; p<0.001 vs β=−0.49 µm; p=0.258). Compensation reduced the age-related RNFL decline by 55% in the overall population (β=−3.32 µm vs β=−1.50 µm/10 years; p<0.001), where the greatest reduction was among the Chinese (93%; β=−2.26 µm vs β=−0.16 µm/10 years; p<0.001), followed by Indians (79%; β=−2.33 µm vs β=−0.49 µm/10 years; p=0.005) and then Malays (59%; β=−2.83 µm vs β=−1.17 µm/10 years; p=0.002). Figure 3 further showed the distribution of measured and compensated RNFL thicknesses among the varying age groups, stratified by ethnicity.

Figure 3

Distribution of measured and compensated RNFL thicknesses among participants of varying age groups and ethnicity. Compensation reduced the age-related RNFL decline by 55% in (A) all participants, where the greatest reduction was among the (B) Chinese (93%), followed by (D) Indians (79%) and then (C) Malays (59%). RNFL, retinal nerve fibre layer.

In Chinese (online Supplementary table), refractive error, optic disc (area and orientation), fovea distance and retinal vessel density were correlated with measured RNFL (p<0.001). However, optic disc ratio and fovea angle were not associated with measured RNFL. Optic disc (area and orientation), fovea distance and retinal vessel density were no longer correlated with compensated RNFL. Although refractive error remained correlated with compensated RNFL, compensation attenuated the effect of refractive error by 69% (β=0.98 µm vs β=0.30 µm; p<0.001).

In Malays (online supplementary table), corneal curvature, refractive error, fovea distance, optic disc area and retinal vessel density were correlated with measured RNFL (p<0.05). Optic disc (ratio and orientation) and fovea angle were not associated with measured RNFL. Refractive error and retinal vessel density were no longer correlated with compensated RNFL. Corneal curvature, optic disc area and fovea distance remained correlated with compensated RNFL.

In Indians (online supplementary table), refractive error, optic disc area and retinal vessel density were correlated with measured RNFL (p<0.05), but not optic disc (ratio and orientation) and fovea (distance and angle). Refractive error and retinal vessel density were no longer correlated with compensated RNFL. Optic disc area remained correlated with compensated RNFL.

Figure 4 shows the impact of compensation on the reclassification of individuals into varying categories of RNFL thickness measurements, stratified by RNFL quadrants. Out of 2563 subjects who were initially classified as normal when using OCT, 0.6% (global RNFL; n=17), 0.7% (inferior RNFL; n=17), 1.0% (superior RNFL; n=24), 1.5% (nasal RNFL; n=39) and 1.5% (temporal RNFL; n=39) were reclassified as having RNFL falling outside of 95% normal limits after compensation. Out of 27 subjects who were classified as abnormal when using OCT, 55.6% (global RNFL; n=15), 59.3% (inferior RNFL; n=16), 40.7% (superior RNFL; n=11), 81.5% (nasal RNFL; n=22) and 74.1% (temporal RNFL; n=20) were reclassified as having RNFL falling within 99% of normal limits after compensation. Figure 5 shows examples of three individuals where compensation of ocular anatomical parameters altered RNFL thickness measurements, from having ‘an abnormally thinned RNFL’ to ‘normal RNFL’.

Figure 4

Classification of individuals based on their measured and compensated RNFL thicknesses into three categories, stratified by varying quadrants. First category, those within the 95% normal limits, considered as normal (green category); second category, those outside the 95% normal limits but within the 99% normal limits, considered as suspect (yellow category); and third category, those outside of 99%, considered as abnormal (red category). RNFL, retinal nerve fibre layer.

Figure 5

Three examples of participants who went from having an abnormally thinned RNFL at the superior (patient 1), nasal (patient 2) and temporal (patient 3) quadrants to a normal RNFL. Patient 1 is a 61-year-old woman, where compensation of anatomical parameters increased the RNFL thickness measurement in the superior quadrant of the right eye by 19 µm. (A–C) Fundus photo, optical coherence tomography (OCT) projection image and binarised image of the vessel tree showing a small optic disc area (1.44 mm2) and slightly sparse retinal vessels at superior quadrant (306 µm). (D, E) OCT RNFL thickness and deviation map showing a thinned RNFL thickness at superior quadrant. (F) RNFL circumpapillary thickness line graph showing a thicker compensated RNFL than measured RNFL, especially the superior quadrant (red arrows). (G) RNFL quadrant hours showing a thinned measured RNFL at superior quadrant, falling in the red zone, indicating outside the 99% distribution. After compensation, compensated RNFL of the superior quadrant now falls in the green zone, indicating it is within the 95% distribution. Patient 2 is a 50-year-old woman, where compensation of anatomical parameters increased the RNFL thickness measurement in the nasal quadrant of the left eye by 10 µm. (A–C) Fundus photo, OCT projection image and binarised image of the vessel tree showing a great optic disc-fovea angle (−15.6°) (red arrow) and sparse retinal vessels at nasal quadrant (110 µm). (D, E) OCT RNFL thickness and deviation map showing a thinned RNFL thickness at nasal quadrant. (F) RNFL circumpapillary thickness line graph showing a thicker compensated RNFL than measured RNFL, especially the nasal quadrant (red arrow). (G) RNFL quadrant hours showing a thinned measured RNFL at nasal quadrant, falling in the red zone, indicating outside the 99% distribution. After compensation, compensated RNFL of the nasal quadrant now falls in the green zone, indicating it is within the 95% distribution. Patient 3 is a 67-year-old man, where compensation of anatomical parameters increased the RNFL thickness measurement in the temporal quadrant of the left eye by 18 µm. (A–C) Fundus photo, OCT projection image and binarised image of the vessel tree showing sparser retinal vessels at temporal quadrant (0 µm). (D, E) OCT RNFL thickness and deviation map showing a thinned RNFL thickness at temporal quadrant. (F) RNFL circumpapillary thickness line graph showing a thicker compensated RNFL than measured RNFL, especially the temporal quadrant (red arrows). (G) RNFL quadrant hours showing a thinned measured RNFL at temporal quadrant, falling in the red zone, indicating outside the 99% distribution. After compensation, compensated RNFL of the temporal quadrant now falls in the green zone, indicating it is within the 95% distribution. RNFL, retinal nerve fibre layer.

Discussion

In this multiethnic, population-based study of adult Chinese, Malays and Indians in Singapore, we demonstrated an overall reduction in the interindividual variability of RNFL thickness after compensating for various anatomical parameters. The effect was greatest for Chinese, followed by Malays and then Indians. This highlights the potential advantages of using compensated RNFL values for clinical use such as early diagnosis of glaucomatous damage, which may be gained by compensating away measurement variability influenced by anatomical parameters. Our study indicated that compensation of RNFL thickness measurements for anatomical parameters may improve diagnostic performance in different ethnicities. Importantly, RNFL thickness measurement is not limited to glaucoma, but has also been useful in diseases such as non-arteritic ischaemic optic neuropathy,34 Parkinson’s disease,35 Alzheimer’s disease36 and multiple sclerosis.37

We report several findings that are important in advancing our understanding of the measurement of RNFL thickness with SD-OCT imaging. Even though the current model reduced the interindividual variability in Asians, several observations indicate that a more refined model will be required. First, our model improved the RNFL measurement variability most effectively in Chinese and least effectively in Indians. Given that the age and refraction distribution in Chinese, Malays and Indians were different, it is plausible that compensation of RNFL thickness had a higher impact in Chinese because they were younger and had higher degrees of myopia. Second, certain anatomical features known to affect RNFL measurements remained associated with compensated RNFL. One such example is refractive error, where it remained associated with compensated RNFL in Chinese. This may be because the initial model was developed in Caucasians13 with lower incidence of myopia than the Chinese.20 Therefore, accounting for a wider range of refractive error in the new model would be important, in particular of the rapidly increasing prevalence of myopia globally,38 and the diagnostic challenges with glaucoma presented by highly myopic eyes.39 Third, new anatomical features such as corneal curvature may be needed as it was associated with RNFL in Malays. Studies have shown that corneal curvature can have a magnification effect on RNFL measurements40 41 and should be considered in the new model. Overall, future models will need to account for a wider range of existing parameters and include new anatomical features to better compensate away RNFL measurement variability influenced by anatomical parameters, which would provide better sensitivity and specificity for glaucoma detection.

Age-related decline in measured RNFL was greatly attenuated by ~55% for global RNFL in the overall population after accounting for anatomical factors. One of the most likely explanations is that OCT is not able to differentiate retinal vessels from neuronal axons. Hence, the RNFL thickness measurement will include retinal vessels. Age-related decline in RNFL thickness has been reported in studies42 43 and may therefore represent to a large degree a change in retinal vasculature. Using Cirrus SD-OCT, Leung et al showed a significant negative correlation between age and measured RNFL thickness (−0.33 µm/year; p=0.011),43 which was comparable to our findings in the overall population (−0.33 µm/year; p<0.001). However, compensating for anatomical factors reduced this age-related decline in RNFL thickness as much as 50% (−0.15 µm/year; p<0.001). This suggests that age-related ocular parameters such as retinal vessels should be accounted for when using RNFL measurements for diagnostic purpose as well as in OCT-related longitudinal glaucoma studies. Our results are compatible with previous reports on the age-related reduction of retinal vessels44 and retinal blood flow with age.45 Moreover, glaucomatous optic neuropathy is also associated with a loss of retinal microvessels and retinal perfusion deficits.46–52 Hence, compensation of RNFL thickness measurements for anatomical parameters may be important for glaucoma eyes.

Our model reclassified ~62% of abnormal individuals out of the ‘at risk category’ and ~1% of normal individuals into the ‘at risk category’. This meant that at least 62% of those who were initially considered as having thinned RNFL by conventional OCT in fact only appeared to have thinned RNFL as a result of anatomical variations (figure 4). This is clinically relevant as well as impactful in the management of patients with glaucoma. Differentiating between normal and early glaucoma is often difficult and clinicians often rely on structural OCT imaging to give an objective assessment of the RNFL before determining if treatment or closer monitoring for glaucoma should be carried out. Our comprehensive model, which comprises precise alignments, vessel removal and input of patient’s anatomical features to OCT scans before sectorial analysis, would therefore allow the accurate identification of glaucoma suspects.

Strengths and limitations

The strengths of this study included the large, multiethnic Asian population-based sample size of eyes without ocular diseases. Thus, we believe that the findings represent population variations of RNFL thickness in different ethnicities. In addition, our study adopted standardised OCT imaging, thus allowing relatively direct and objective comparisons across the ethnic groups. In this study, the compensation was done with one specific OCT device, it can in principle be applied to OCT machines from other companies. Our study had a few limitations. First, OCT imaging was introduced halfway into the study, hence fewer Chinese underwent OCT imaging compared with Malays or Indians. Second, owing to the cross-sectional nature of our study, we could identify an attenuation in the age-related decline in RNFL thickness after compensation, but is not able to address the temporal relationship. Longitudinal data in normal eyes are needed to clarify this relationship. Third, the current model was employed for RNFL measurements only, but not for macular measures of glaucoma damage such as retinal ganglion cell complex.53 Based on our results it is likely that the interindividual variability of such measures will also be reduced by anatomical compensation, although additional studies are required to prove this hypothesis. Most importantly, our compensation model of RNFL thickness is not perfect; it provided a simplified correction of potential anatomical parameters that are related to RNFL thickness. We have only considered for optic disc (ratio, orientation and area), fovea (distance and angle), retinal vessel density, refractive error and age. Other underlying causes of interindividual variability of RNFL might be due to age-related optic nerve head changes, vessel geometry or posterior sclera changes.

In conclusion, we have demonstrated that RNFL thickness compensated for ocular biometry and morphology had major impact on healthy subjects. The greatest effect was observed in Chinese, and least effective in Indians. Anatomical compensation should be considered when refining the RNFL normative database, which may improve glaucoma diagnosis.

References

View Abstract

Footnotes

  • JC and FS are joint first authors.

  • Contributors All the authors met the four criteria of contributorship. Acquisition, analysis or interpretation of data: all authors. Drafting of the manuscript: JC, LS. Critical revision of the manuscript for important intellectual content: all authors. Statistical analysis: DQN, JC. Obtained funding: LS, CYC. Administrative, technical or material support: all authors. Study supervision: LS, JC, FS and CV prepared and reviewed the main manuscript, tables and figures.

  • Funding This work was funded by grants from the National Medical Research Council, Singapore (CG/C010A/2017; CG/SERI/2010; STaR/0003/2008; CIRG/1417/2015; CIRG/1488/2018; OFLCG/004a/2018).

  • Disclaimer The sponsor or funding organisation had no role in the design or conduct of this research.

  • Competing interests None declared.

  • Patient consent for publication Not required.

  • Ethics approval SingHealth Centralized Institutional Review Board.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data availability statement All data relevant to the study are included in the article or uploaded as supplementary information.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles

  • At a glance
    Keith Barton Jost B Jonas James Chodosh