Positive (in brown) and negative (in blue) Spearman correlation coefficients of visual function and retinal measures. FAZ indicates foveal avascular zone; GCC, ganglion cell complex; RNFL, retinal nerve fiber layer; SCP, superficial capillary plexus; VA, visual acuity, VD, vessel density.
Linear splines with knots at the lowest 10% quantile for distance visual acuity (VA) and contrast sensitivity in the distribution of RNFL (A and B, respectively) and GCC thickness (C and D, respectively).
eFigure. Flowchart Showing Sample Selection for Analytic Sample
eTable 1. Demographic, Medical History, Lifestyle Information, OCT- and OCTA-Based Measures and Ocular Characteristics of All Participants
eTable 2. Association of Visual Function Measures With OCT Structural and Angiographic Measures of All Participants
Customize your JAMA Network experience by selecting one or more topics from the list below.
Dong Y, Guo X, Arsiwala-Scheppach LT, et al. Association of Optical Coherence Tomography and Optical Coherence Tomography Angiography Retinal Features With Visual Function in Older Adults. JAMA Ophthalmol. 2022;140(8):809–817. doi:10.1001/jamaophthalmol.2022.2099
How do the distributions of optical coherence tomography (OCT) and OCT angiography (OCTA) retinal features differ across distinct populations of older adults, and how are they associated with metrics of vision function?
In this cross-sectional study of 759 participants, distribution of retinal features differed by race/community. Contrast sensitivity, but not visual acuity, was associated with retinal nerve fiber layer thickness whereas neither measure of vision was associated with angiographic features.
These results suggest that OCT and OCTA reference values should take into account certain demographic characteristics and further suggest that contrast sensitivity is a more sensitive marker of retinal neuronal differences than visual acuity.
Although there is abundant evidence relating neuronal and vascular optical coherence tomography (OCT) and OCT angiography (OCTA) measures to retinal disease, data on the normative distribution of retinal features and their associations with visual function in a healthy, older, community-based population are sparse.
To characterize the normative OCT and OCTA measures in older adults and describe their associations with visual function.
Design, Setting, and Participants
This was a cross-sectional, observational study conducted from May 17, 2017, to May 31, 2019. The study included a community-based sample. Participants in the Atherosclerosis Risk in Communities study from Jackson, Mississippi (all self-reported Black participants), and Washington County, Maryland (all self-reported White participants), were recruited in the Eye Determinants of Cognition study (EyeDOC). Data analyses were conducted from June 14, 2020, to May 31, 2021.
Main Outcomes and Measures
Retinal measurements, including retinal nerve fiber layer (RNFL) thickness, macular ganglion cell complex (GCC) thickness, macular vessel density (VD) in the superficial capillary plexus (SCP) and deep capillary plexus (DCP), and foveal avascular zone (FAZ) area, were captured with spectral-domain OCT and OCTA. Visual function, including presenting distance vision, corrected distance vision, near visual acuity (VA), and contrast sensitivity (CS), was assessed.
A total of 759 participants (mean [SD] age, 80 [4.2] years; 480 female participants [63%]; 352 Black participants [46%]) were included in the study. Mean (SD) GCC thickness (89.2 [9.3] μm vs 92.3 [8.5] μm) and mean (SD) FAZ (0.36 [0.16] mm2 vs 0.26 [0.12] mm2) differed between Jackson and Washington County participants, respectively. Mean (SD) RNFL thickness and mean (SD) VD in SCP and DCP were greater for participants 80 years or younger than for participants older than 80 years (RNFL: ≤80 years, 93.2 [10.5] μm; >80 years, 91.1 [11.6] μm; VD SCP, ≤80 years, 44.3% [3.5%]; >80 years, 43.5% [3.8%]; VD DCP, ≤80 years, 44.7% [4.9%]; >80 years, 43.7% [4.8%]). Linear regression showed each 10-μm increment in RNFL thickness and GCC thickness was positively associated with 0.016 higher logCS among all participants (RNFL: 95% CI, 0.005-0.027; P = .004; GCC: 95% CI, 0.003-0.029; P = .02), with stronger associations among Jackson participants. The associations of VA and structural measures were found only in Jackson participants, with coefficients per 10-μm increment of 0.012 logMAR VA (RNFL: 95% CI, 0.000-0.023; P = .049) and 0.020 logMAR VA (GCC: 95% CI, 0.004-0.034; P = .04).
Conclusions and Relevance
In this cross-sectional study, better CS was associated with greater RNFL thickness and GCC thickness, but no visual measures were associated with angiographic features overall. These findings suggest that clinical application of normative references for OCT- and OCTA-based measures should consider demographic and community features.
Optical coherence tomography (OCT) has transformed clinical practice, allowing clinicians to diagnose and monitor the progression of ophthalmic diseases such as glaucoma, age-related macular degeneration, and diabetic retinopathy.1-7 Structural measures, including retinal nerve fiber layer (RNFL) thickness and ganglion cell complex (GCC) thickness, have served as diagnostic markers for the presence and severity of diseases such as glaucoma, and changes in these measures are used clinically as indicators of disease progression.2,8-12 More recently, OCT angiography (OCTA) has provided additional insight into retinal microvascular health through measures such as vessel density (VD) and foveal avascular zone (FAZ) area size.13-15 These measures of retinal vascular health hold important information about retinal changes related to the risk for vision loss in populations with eye disease.5,13-15
The value of OCT in the clinical setting among specific disease groups is well established.16-23 However, there are questions as to whether there are differences in the distribution of OCT- and OCTA-based measures across populations, and whether demographic factors should be considered when judging meaningful thresholds for distinguishing healthy retinal tissue from possible pathology.24 Moreover, although OCT- and OCTA-based measures in clinical populations can be used to classify the presence or absence of ocular pathologies,25-27 very little is known about the degree to which these measures reflect differences in visual function in the general older adult population without eye disease.
Here, we present data from the Eye Determinants of Cognition (EyeDOC) study, a large bicommunity study evaluating OCT and OCTA retinal measures with older Black and White individuals. In this study, we aimed to (1) describe and compare the normative distribution of OCT- and OCTA-based retinal measures in a healthy older population across race/community and age, and (2) understand the associations between retinal measures and visual function.
The EyeDOC study recruited 2 community-based samples of older adult participants from the Atherosclerosis Risk in Communities (ARIC) study, including self-reported Black participants from Jackson, Mississippi, and self-reported White participants from Washington County, Maryland. Participants were recruited from May 17, 2017, to May 31, 2019, and underwent a single study visit nested within the ARIC study, at which visual function and retinal imaging were captured. Participants with evidence of overt eye disease (including glaucoma, diabetic retinopathy, age-related macular degeneration, retinal vessel occlusion, proliferative retinopathy, and macular edema) were identified during a retinal pathology review conducted by ophthalmologists at the Wilmer Eye Institute in accordance with the Early Treatment Diabetic Retinopathy Study (ETDRS) Retinal Grading Protocol.28,29 Participants with quality OCT and/or OCTA images and visual function data were included in this cross-sectional analysis. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines were also followed.30
Both the ARIC and EyeDOC studies were approved by the institutional review boards at relevant ARIC sites and all participants provided written informed consent at the EyeDOC study visit. Participants were compensated for their time and travel to the study visit.
OCT structural scans across the macular and optic disc areas and OCTA images of both 3 × 3-mm2 and 6 × 6-mm2 areas around the macula were captured with the AngioVue imaging system RTVue-XR Avanti system spectral OCT machine (OptoVue) in both eyes in a random 10% study participant sample and in 1 eye in the remaining 90% study participants after dilation. RNFL and GCC thickness were derived from OCT structural scans. Retinal VD, defined as the percentage of the imaging region occupied by blood vessels, was estimated in 2 distinct vascular plexuses (superficial capillary plexus [SCP] and deep capillary plexus [DCP]) from the macular 6 × 6-mm2 OCTA images. FAZ area, defined as the region in the fovea devoid of retinal blood vessels, was estimated from the macular 3 × 3-mm2 OCTA images. Both signal strength index (SSI) and a composite quality index (QI) were used to assess image quality. Quality images were included in the analysis using the following cutoffs: OCT structural scans with SSI of 37 or greater (for RNFL thickness) or SSI of 44 or greater (for GCC thickness),31 OCTA images with QI of 6 or greater, and SSI of 55 or greater. Axial length (AL) was derived from ocular biometry using the IOL Master V.5.4 (Carl Zeiss Meditec).
Distance visual acuity (VA) and contrast sensitivity (CS) were measured monocularly in each eye of each participant. Presenting-distance VA with habitual corrective lenses was measured using the retroilluminated ETDRS chart (Precision Vision). Corrected-distance VA was measured using an autorefractor, the Topcon KR-8000 (Topcon) for Jackson and Nidek ARK 560A (Marco Technologies) for Washington County. Binocular reading acuity derived from MNRead chart testing (Precision Vision) was used to assess near VA. CS was assessed using the MARS chart (The Mars Precentrix Corporation).
Participants’ age was available from the ARIC sixth visit (2016-2018) and then adjusted to account for the time between the ARIC sixth visit and the EyeDOC study visit. Participants’ age was categorized as 80 years or younger and older than 80 years. Sex, self-reported race, and community information (ie, Jackson vs Washington County), medical history, and lifestyle information were obtained from the closest ARIC visits. Body mass index was calculated as weight in kilograms divided by square of height in meters. Hypertension was defined as a systolic blood pressure greater than 140 mm Hg, a diastolic blood pressure greater than 90 mm Hg, or the use of antihypertensive medication. Present diabetes was defined if a fasting glucose level was greater than or equal to 126 mg/dL (to convert to millimoles per liter, multiply by 0.0555) or the nonfasting glucose level was greater than or equal to 200 mg/dL, or if using medication for diabetes or self-report diagnosis of diabetes. Self-reported alcohol use and cigarette smoking were categorized into current drinker/smoker and nondrinker/nonsmoker.
Two-tailed t tests and χ2 test were used to compare demographic, medical history, lifestyle information, retinal parameters, and ocular characteristics among all participants between race/community and age strata. Pairwise Spearman correlations along with graphical analysis were used to assess the strength of associations between retinal parameters and visual function measures. All P values were 2-sided and were not adjusted for multiple comparisons. Regression models were adjusted for demographic variables (eg, age, sex, race/community), AL, and OCT- and OCTA-based retinal parameters (SSI and QI for OCTA parameters; SSI for OCT parameters) to examine the independent associations of retinal measures with visual function. To better understand race/community differences in the associations between vision and retinal measures (OCT- and OCTA-based measures), we used both stratified analyses and race/community interaction terms to evaluate the presence of notable effect size difference and statistical interaction. Linear splines with knots at the lowest 10% quantile in the distribution of RNFL and GCC thickness were used to explore the potential nonlinear effects at the lower tail (participants with the thinnest RNFL and GCC) of the distribution. Nonlinearity in regression models with polynomial terms was explored, and models were compared based on Akaike information criterion. Cutoff for statistical significance was set at .05. Data analyses were conducted from June 14, 2020, to May 31, 2021, using R software, version 4.1.3 (R Project for Statistical Computing).
A total of 1073 participants were recruited in the EyeDOC study; of these, 1007 participants provided quality OCT and/or OCTA images. After retinal image review, 248 participants with evidence of overt eye disease were excluded from the study, leaving a final sample of 759 participants without ocular pathology for analysis (eFigure in the Supplement). Mean (SD) age of the analytic population was 80 (4.2) years. There were 480 female participants (63%) and 279 male participants (37%). Race categories included 352 Black (46%) and 407 White (54%). A total of 608 participants (80.6%) reported hypertension, and 295 (39.1%) self-reported as being a current drinker.
Several OCT and OCTA measures differed across race/community and age groups (Table 1). Mean RNFL (SD) thickness was 92.3 (11.0) μm among all participants, and no significant difference was observed across race/communities. Mean (SD) RNFL thickness was greater in participants 80 years or younger compared with participants older than 80 years (93.2 [10.5] μm vs 91.1 [11.6] μm). Mean (SD) GCC thickness was 90.9 (9.0) μm overall but was lower in Jackson vs Washington County participants (89.2 [9.3] μm vs 92.3 [8.5] μm).
For OCTA measures, difference by race/community in mean VD was noted only in the DCP layer. Participants from Jackson had a greater mean (SD) VD DCP than Washington County participants (45.1% [5.1%] vs 43.5% [4.6%]). Mean (SD) VD DCP was also greater among participants 80 years or younger than in those older than 80 years (44.7% [4.9%] vs 43.7% [4.8%]), as was mean VD SCP (44.3% [3.5%] vs 43.5% [3.8%]). Jackson participants had a significantly larger mean (SD) FAZ area than Washington County participants (0.36 [0.16] mm2 vs 0.26 [0.12] mm2), although no differences were noted across age groups.
Measures of visual function (ie, CS, corrected/presenting distance VA, and near VA) were generally positively correlated (ie, better visual function in 1 measure corresponded to better visual function in all other measures), with all correlation coefficients (ρ) exceeding 0.21 in magnitude (Figure 1). Correlations of visual function with retinal measures were weak, although the direction suggested that better visual function was correlated with both thicker neural tissue and greater vascular density.
For neuronal measures, better CS was associated with both thicker RNFL and thicker GCC overall, with stronger associations noted among Jackson participants, although not to a significant degree as judged by interaction terms (Table 2). Each 10-μm-thicker RNFL was associated with 0.016 better logCS (95% CI, 0.005-0.027; P = .004) for all participants studied, with a larger association for Jackson participants. Similarly, each 10-μm-thicker GCC was associated with 0.016 log better CS (95% CI, 0.003-0.029; P = .02), again, with a larger effect (0.022 log better CS) among Jackson participants (95% CI, 0.006-0.039; P < .001). Associations of both thicker RNFL and thicker GCC with better corrected distance VA were found only in Jackson participants, with each 10-μm-thicker GCC associated with 0.020 logMAR VA (95% CI, 0.004-0.034; P = .04) (which corresponds to approximately 1 letter change on the ETDRS chart) and each 10-μm-thicker RNFL associated with 0.012 logMAR VA (95% CI, 0-0.023; P = .049) for RNFL thickness. The association between corrected distance VA and RNFL thickness differed by race/community (Jackson, 0.012; 95% CI, 0.000-0.023; Washington, −0.015; 95% CI, −0.032 to 0.003; P = .03 for interaction term).
For vascular measures, no visual function measures were significantly associated with either VD or FAZ in the overall sample. In community-stratified models, there was an association of corrected distance VA with SCP VD among Washington County participants only.
The scatterplots suggest that associations between visual function and structural measures (RNFL and GCC thickness) may be driven by the thinnest 10% of RNFL and GCC values (Figure 2). That is, corrected distance VA and CS are correlated with RNFL and GCC thickness only among those in the lowest 10% of the distribution. However, regression models incorporating polynomial terms to fit nonlinearity were not a better fit to the data and linear terms were used for the final models.
Our cross-sectional study used the large EyeDOC sample and found that in this healthy older adult population, differences in the normative OCT- and OCTA-based retinal measures were noted across race/community and age (ie, older old vs younger old), though differences were generally small. Regarding the second study aim, our results demonstrated that thicker RNFL and GCC were modestly associated with better CS. Thicker RNFL and GCC were modestly associated with better corrected distance VA, but in only 1 of the 2 communities. However, we did not find any consistent associations of visual function with OCTA measures of retinal VD. These findings support the need to explore a clinically meaningful threshold for distinguishing relatively healthy retinal tissue from possible pathology under different race/community contexts if OCT is to be used as a tool for screening the general older adult population for risk of vision loss.
There are some notable differences in the distributions of OCT and OCTA measures reported in our study compared with prior studies. These differences could be the result of a number of factors: (1) the EyeDOC study population is an older adult sample and represents the ages at which eye disease become increasingly prevalent whereas prior studies involved younger participants, most commonly aged 20 to 40, and15,32 (2) different OCT devices and/or different calculation algorithms can introduce discrepancies, resulting in low reproducibility between studies.32,33
For neuronal measures obtained from OCT, mean RNFL thickness in our full sample was slightly lower than that of existing data, as expected.17,24,32,34,35 White participants have been shown to have thinner RNFL than other ethnic groups, such as Hispanic and Asian, but comparable RNFL with Black participants.24,35 In our analysis, RNFL thickness also did not differ between the Black and White communities. Mean GCC thickness in our data was thinner than the range of 95.1 to 107.9 μm that has been reported previously on a variety of instruments.9,36-38 Notably, mean GCC thickness was roughly 3 μm thinner in Jackson participants than in those from Washington County, a fairly large difference corresponding to roughly 6% of the dynamic range of values.
With regard to retinal vascular measures from OCTA, earlier studies reported generally greater mean VD DCP than data in our study.7,39-41 Prior estimates for mean VD SCP have not been consistent across populations, especially among people older than 60 years.42,43 A population-based study in Hong Kong with participants aged approximately 50 to 60 years showed greater mean VD DCP (55.1%), yet similar mean VD SCP (47.3%) compared with the EyeDOC study.41 The estimated mean FAZ in the EyeDOC study was similar to that of existing studies, which could be a mixed effect of shrinkage owing to eye disease and enlargement attributable to aging.5,44-46 As with GCC measures, a rather substantial difference was noted in mean FAZ area between communities. Jackson participants had a mean FAZ nearly 40% greater than Washington County participants. Differences in the distribution of both neuronal and vascular measures across race/community groups suggest that the normative range of these measures is race/community specific.
Understanding the extent to which vision function is a function of measurable retinal neuronal and vascular features could also inform thresholds of normative values for retinal parameters. Most associations of visual function and OCT and OCTA measures have been reported in people with eye conditions (eg, myopic maculopathy, glaucoma).38,47,48 For example, 1 previous study showed greater GCC thickness associated with better CS in a population with prevalent eye diseases.47 The EyeDOC study is unique in assessing these associations among community-dwelling older adults without overt ocular pathology. In our analysis, correlations were notably weak, though correlations of CS with RNFL and GCC thickness were detected in both communities, with stronger correlations among Jackson participants (differences between communities were nonsignificant). From the regression analysis, the magnitude of these associations was modest, with a significant increase of GCC/RNFL thickness (10 μm) corresponding to only a small improvement in CS (0.016 log units, roughly half a letter in a MARS chart). The univariate associations between visual function and OCT neuronal measures seemed to be strongest among the 10% of participants with the thinnest RNFL and GCC values, which suggests a potential threshold for screening purposes.
We did not find any definitive associations of VA or CS with VD or FAZ. The association of corrected distance VA with VD SCP among Washington County participants may be attributed to chance, as the other associations were seen among Jackson participants, and there is no reason to hypothesize such a large degree of community heterogeneity. A significant positive association of VA and FAZ area was previously reported only in persons with macular edema or branch retinal vein occlusion, whereas only a minimal correlation of VA and FAZ has been found in normal healthy individuals.5,49 Coefficient estimates of the current study tended to be weaker compared with the few reports in the literature; discrepancies may be the result of inclusion of individuals with eye disease in existing reports,50 demographic characteristics of the study sample, or sampling variability in smaller studies.49
Our study had several strengths. First, it was one of the largest community-based studies to date of OCT and OCTA retinal measures among Black and White participants. Second, our results were derived from a less-studied older adult population and included 2 different communities. Third, the visual and retinal parameters were measured in a systematic way following strict, prespecified protocols and rigorous image quality control procedures. Finally, we examined the associations among participants without overt ocular pathology to elucidate associations in the general population. Such results are more applicable to community screening programs and general ophthalmic preventive care. Additional analysis on participants including older subjects with eye disease was included for comparison (eTable 1 and eTable 2 in the Supplement).
However, our findings are subject to some limitations. The cross-sectional study design was appropriate for assessing the degree to which OCT findings reflect abnormalities in vision function, but the study was not designed to assess causal associations or directions of effect. In this analysis, only SSI and QI were used as the image quality assessment metric in the regression models without a manual image-grading process, which is well-matched with clinical practice but may affect precision. This strategy was adopted because of the challenges in obtaining optimal high-quality retinal images among an older population.28 The P values for associations and 95% CIs around association estimates were not adjusted for multiple comparisons as all associations evaluated were specified a priori and represented separate hypothesis about potential mechanisms linking OCT findings to various aspects of vision function. Lastly, only overall OCT- and OCTA-based measures were examined without further exploring these measures by imaging sector or quadrant.
In conclusion, results from this cross-sectional study including an aging, bicommunity sample of healthy older adults suggest age- and community-related differences in retinal feature distributions and associations of greater RNFL thickness and GCC thickness with better CS. Further work is warranted to determine the extent to which different reference/normative values need to be used for different age and community groups when evaluating structural and functional retinal measures.
Accepted for Publication: May 14, 2022.
Published Online: July 14, 2022. doi:10.1001/jamaophthalmol.2022.2099
Corresponding Author: Alison G. Abraham, MHS, MS, PhD, Department of Epidemiology, University of Colorado, Anschutz Medical Campus, 13001 E 17th Pl, 3rd Floor, Aurora, CO 80045 (email@example.com).
Author Contributions: Dr Abraham and Ms Dong had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Dong, Arsiwala-Scheppach, Sharrett, Ramulu, Abraham.
Acquisition, analysis, or interpretation of data: Dong, Guo, Arsiwala-Scheppach, Ramulu, Mihailovic, Pan-Doh, Mosley, Coresh, Abraham.
Drafting of the manuscript: Dong, Abraham.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Dong, Arsiwala-Scheppach, Abraham.
Obtained funding: Ramulu, Coresh, Abraham.
Administrative, technical, or material support: Dong, Guo, Ramulu.
Supervision: Ramulu, Mosley, Abraham.
Conflict of Interest Disclosures: Ms Dong reported receiving grants from the National Institutes of Health and the US National Institute on Aging. Ms Arsiwala-Scheppach reported receiving grants from the US National Institute on Aging and from the National Heart, Lung, and Blood Institute. Dr Ramulu reported receiving grants from the National Institutes of Health during the conduct of the study. Ms Mihailovic reported receiving grants from the US National Institute on Aging during the conduct of the study. Dr Mosely reported receiving grants from the National Institutes of Health and the National Heart, Lung, and Blood Institute during the conduct of the study. Dr Coresh reported receiving grants from the National Institutes of Health and the National Heart, Lung, and Blood Institute; serving as a consultant for and receiving stock options from Healthy.io; and serving as a scientific advisor for SomaLogic outside the submitted work. Dr Abraham reported receiving grants from the National Institutes of Health, the National Heart, Lung, and Blood Institute, and the US National Institute on Aging and consulting fees from Implementation Group Inc during the conduct of the study. No other disclosures were reported.
Funding/Support: This study was supported in part by grant 1R01AG052412 from the US National Institute on Aging; HHSN268201700001I, HHSN268201700002I, HHSN268201700003I, HHSN268201700005I, HHSN268201700004I from the National Heart, Lung, and Blood Institute; U01 2U01HL096812, 2U01HL096814, 2U01HL096899, 2U01HL096902, 2U01HL096917 from the National Institutes of Health; and R01-HL70825 from the National Heart, Lung, and Blood Institute.
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.