Associations of Visual Function With Cognitive Performance in Community-Based Older Adults: The Eye Determinants of Cognition Study

Abstract

Background

Given the detrimental impacts of visual and cognitive impairment separately, in older adults, understanding their coexistence could inform strategies against age-related cognitive changes.

Methods

Participants from Washington County (White) and Jackson (Black) sites of the Atherosclerosis Risk in Communities Study were recruited who differed on racial, regional, urbanicity, and community-based factors. Presenting distance visual acuity (DVA; representing vision loss due to disease or lack of eye-care), corrected DVA (representing optimal correction for refractive loss of vision), and contrast sensitivity were measured. Factor scores for global cognition, memory, executive function, and language domains were calculated for 3 visits. We quantified the associations of vision measures with change in cognitive scores, stratified by community/race, using generalized estimating equations.

Results

In 982 participants, mean (standard deviation [SD]) baseline age was 74 (4) years, with 37% males and 45% Jackson/Black participants. As hypothesized, after accounting for potential confounders, in the better-eye, worse presenting DVA was associated with greater 10-year decline rate in global cognition, memory, and executive function in Washington County/White participants (eg, global cognition: −0.08 SD [95% confidence interval: −0.12, −0.04]). Worse corrected DVA was associated with greater 10-year decline rate in executive function in Washington County/White participants (−0.10 SD [−0.15, −0.04]). Better contrast sensitivity was associated with lower 10-year decline rate in global cognition and executive function in Washington County/White participants (eg, global cognition: 0.10 SD [0.06,0.14]). None of these associations were confirmed in Jackson/Black participants.

Conclusions

Our study supports a functional link between some vision measures and cognition in older adults, but in only 1 of the 2 communities studied.

Visual impairment has significant detrimental effects on physical and psychosocial health and affects more than 2.9 million U.S. adults (1–5). The prevalence of blindness and visual impairment increases rapidly with age among all races, particularly among individuals over 75 years (6,7). As the population ages, numbers of individuals with visual impairment will increase as will the number with cognitive decline, and consequently so will the number experiencing the burden of both conditions (3,8,9). Older adults with cognitive impairment experience increased disability and dependency (10,11), increased health care costs (11,12), and early mortality (13). The coexistence of these 2 conditions exacerbates the risk of disability (3).

A growing burden of both visual and cognitive impairment among older adults may, in part, reflect their shared relationship with age (14). However, it has been postulated that visual impairment might cause or exacerbate cognitive decline through a variety of mechanisms including negative impacts on social functioning, emotional well-being, and engagement in cognitively stimulating activities like reading, known as the sensory deprivation theory (3,15). There is also potential for shared effects from systemic disease, as the eye and the brain are linked through common embryonic development (16). The common cause theory suggests that disease processes resulting from inflammation or vascular/neuronal damage may affect both retinal and cerebral tissues, linking visual and cognitive loss (3,17,18). Understanding the relationship between visual and cognitive impairment could help untangle underlying pathologic mechanisms and inform prophylactic strategies for age-related cognitive changes (3), particularly if vision loss is found to be causally related to cognitive impairment.

While numerous studies have found cross-sectional associations between visual and cognitive impairment (19–24), confounders such as education and other socioeconomic factors which strongly influence cognitive test scores (25) complicate interpretation. Only a few studies have examined how visual function relates to cognitive change over time (3,15,26,27). For example, decline in presenting distance visual acuity (DVA) was associated with cognitive function decline in the Salisbury Eye Evaluation study, though cognition was assessed with only a limited screening test (Mini-Mental State Examination [MMSE]) (3). However, some longitudinal studies either examined only a selective cohort (27) or implemented only 2 cognitive measurements (15,26,27). Additional research is warranted using more comprehensive measures of cognitive function (28,29), and a broader assessment of visual function including contrast sensitivity (CS), which has been tied to falls and inability to engage in daily living activities (14,30,31).

Here, we examined the association of visual function with longitudinal change in cognitive function evaluated via an extensive cognitive battery in the biracial, community-based Eye Determinants of Cognition (EyeDOC) Study nested within the Atherosclerosis Risk in Communities (ARIC) Study. Previous work in the ARIC has shown an association between retinal vascular changes in midlife with faster rate of cognitive decline (32). Known racial and ethnic differences in visual (5,33) and cognitive (34) health prompted a further evaluation of how these relationships may differ by race and community in 2 distinct communities. We hypothesize that both poor DVA and CS would be associated with greater decline in cognitive function in both community/race groups, and poor presenting DVA would be more strongly associated with cognitive decline than poor corrected DVA.

Method

Study Population

The EyeDOC Study recruited current participants of the ARIC Study from May 17, 2017 to May 31, 2019 for a single visit. The ARIC Study began as a prospective cohort of 15,792 participants, aged 45–64 years during an initial visit (1987–1989), with participants selected by probability sampling from 4 U.S. communities (35,36), 2 of which were included in the EyeDOC Study: Jackson, Mississippi, where only Black participants were recruited, and Washington County, Maryland, which was predominantly White. These 2 sites were selected because they provided a biracial cohort for the study of visual function. The majority of EyeDOC Study visits (ie, vision measurements) took place between ARIC Study visits 6 and 7.

Eligibility for the EyeDOC Study was restricted to active ARIC participants who at the most recent ARIC visit had a MMSE score of at least 24 in Washington County or at least 22 in Jackson. Different thresholds were used for the 2 study sites based on prior work which effectively excluded dementias in our study populations as diagnosed by cognitive testing and review by the expert panel of the ARIC Study (37).

As the Jackson study site recruited only Black participants and Washington County had only 10 non-White participants, race and community were inextricably connected in our study. We excluded non-White participants from Washington County as there were too few for a meaningful analysis. Participants with missing data on visual, cognitive, demographic, and clinical variables were also excluded. The ARIC and the EyeDOC studies were reviewed and approved by Institutional Review Boards of their respective study sites and the coordinating center. Written informed consent for both studies was obtained from all participants.

Assessment of Vision Measures

The vision assessment (2017−2019) was conducted 6.2 years (median) after the baseline cognitive function evaluation (2011−2013) and included presenting DVA, representing vision loss due to disease, lack of access to care (eg, cataract removal surgery), or inability to obtain proper correction for refractive error (ie, glasses); corrected DVA, representing optimal correction for vision loss due to refractive error (38); and CS, which affects fall risk and daily living activities and is relatively unaffected by uncorrected refractive error (14,30,31). The objective vision measures described below were assessed separately in the right and left eye. Presenting DVA was measured with participants wearing their usual corrective lenses, if any. It was measured using a backlit Early Treatment Diabetic Retinopathy Study (ETDRS) chart at a distance of 4 m. Corrected DVA was measured using an autorefractor (Nidek ARK 560A or Topcon KR 800S) with the Snellen chart on the autorefractor. CS was tested using the MARS chart held at a distance of 50 cm with participants wearing their usual vision correction.

Number of letters correctly read in presenting DVA and CS testing, and the lowest line of the letters read for corrected DVA were used to calculate the 3 vision parameters for each eye in their respective scales and units. Better-eye presenting and corrected DVA were derived as the respective lower (better) logMAR (log₁₀[minimum angle of resolution]) value between both the eyes. Better-eye CS was derived as the higher (better) logarithmic value between both the eyes. Separately, worse-eye values for each parameter were derived as the respective lower (DVA) or higher (CS) value between both eyes.

Age-related macular degeneration (AMD) was identified via retinal imaging following dilation, using a fundus camera (Nidek AFC-330). Pathology grading was based on two 45° retinal photographic images centered, respectively, at the macula and the optic nerve head. Two ophthalmologists (X.G., P.Y.R.) graded these images for presence of AMD. Other conditions like glaucoma and diabetic retinopathy were self-reported.

For presenting and corrected DVA, 0.1 logMAR (the unit increment used for analyses) represented 5 letters or 1 line on the ETDRS chart. For CS, 0.1 logarithm (log) represented 2.5 letters on the MARS chart.

Assessment of Cognitive Measures

A full battery of cognitive testing was implemented for all ARIC participants at visit 5 (2011–2013), visit 6 (2016–2017), and visit 7 (2018–2019). We used visit 5 as the baseline cognitive function evaluation and assessed change in cognition over a median of 6.5 years using 3 ARIC Study visits that crossed the EyeDOC Study visit. To compare neurocognitive function across visits, Gross and colleagues (39) developed a model for calculating factor scores in the ARIC Study for 3 previously identified cognitive domains (39–43): memory, executive function, and language as well as global cognition. Global cognition was represented by the global composite factor score created from the weighted factor analysis using all the test scores (39,44). The tests used in these calculations were selected a priori and were Delayed Word Recall, Digit Symbol Substitution, Incidental Learning, Word Fluency Test (FAS test), Animal Naming, Logical Memory I, Digit Span (backwards), Trails A, Trails B, Boston Naming Test, and Logical Memory II (39).

We used factor scores of global cognition and the aforementioned 3 domains, from visits 5, 6, and 7. The interpretation of factor scores for global cognition and the 3 domains is similar to z-scores because they were scaled to have a mean of 0 and variance of 1 (scaled to ARIC Study visit 5 data, the baseline visit for cognitive testing).

Covariates

Covariate data from visit 5 (2011–2013) were used for analysis, unless specified otherwise. Education level was assessed at visit 1 (1987–1989) and categorized as basic (less than high school), intermediate (high school graduate or vocational school), or advanced (college, graduate school, or professional school). Participants were asked to bring all medications taken in the previous 2 weeks, which were coded by trained personnel. Hypertension was defined as systolic blood pressure ≥140 mmHg, diastolic blood pressure ≥90 mmHg, or taking antihypertensive medications. Diabetes was defined as fasting blood glucose levels ≥126 mg/dL, nonfasting blood glucose level ≥200 mg/dL, taking hypoglycemic medications, or self-report of physician-diagnosed diabetes.

Statistical Analyses

Baseline characteristics at visit 5 (or EyeDOC Study visit, as relevant) were summarized for the entire cohort and by community/race strata, where differences between community/race strata were tested using chi-square tests for categorical variables and 1-way analysis of variance for continuous variables. All analyses were stratified by community/race. We quantified the association of vision measures (per 0.1 unit) with concurrent change in cognitive factor scores from visit 5 to visit 7 (and expressed as 10-year change). Linear regression models were used to regress cognitive factor scores on follow-up time, vision level, and covariates, with generalized estimating equations used to account for the correlation between multiple outcome measurements per participant over time. The follow-up time (years) was calculated for each visit, using visit 5 as the baseline and was included as a predictor variable; the interpretation of the beta estimate was change in cognitive score per year. An interaction term between the exposure (ie, vision) and length of follow-up time was included, whose beta estimate represented the difference in the change in cognitive score per year per unit change in vision measurement. The maximum likelihood estimate for this interaction term was the focus of our analysis for evaluating whether a particular measure of visual function was associated with concurrent change in cognitive score. Models were adjusted for demographic (age, sex, and education) and clinical variables (diabetes and hypertension).

All aforementioned analyses were conducted using measurements from the better-eye and worse-eye separately, to shed light on the mechanisms underlying any observed associations. Visual function of the better-eye may be more closely related to a person’s daily functioning, whereas that of the worse-eye may better reflect the presence (vs absence) of systemic pathology. Hence, associations found with better-eye measurements may be more consistent with the sensory deprivation theory (4), whereas those with worse-eye measurements may be more consistent with the common cause theory.

As a sensitivity analysis, we repeated our analyses after excluding participants with AMD, whose prevalence differed substantially between the 2 communities.

The level of significance was set to p value < .05. All statistical analyses and data management were performed using RStudio (R version 4.0.2).

Results

A total of 1 073 participants were recruited into the EyeDOC Study of whom 982 participants were available for analysis after excluding non-White participants from Washington County (n = 10), those who lacked cognitive function measurement at baseline (visit 5, n = 21), and those with missing information on any vision measures (n = 34), any cognitive measures for a visit (n = 2), demographic measures (age, sex, race, and education, n = 1), or clinical comorbidities (diabetes and hypertension, n = 23; Supplementary Figure 1). When comparing participants who attended the EyeDOC Study to those who were eligible but did not attend the study, differences in age, community/race, education, and cognitive scores were noted, with the former group having better cognitive performance than the latter. The mean age of the included sample at visit 5 was 74 years (standard deviation [SD]: 4), 37% were males, and 45% were from Jackson community, where all participants were African American (Table 1). Compared to participants from Washington County/White group, those from Jackson/Black group were less likely to be male or have AMD, and more likely to have advanced education, hypertension, and generally had slightly better vision. Visual function was worse in older participants in both community/race groups with roughly similar trends across age groups (Figure 1). The Washington County/White group had higher mean cognitive scores than Jackson/Black group (eg, mean [SD] global cognition factor scores were +0.49 (0.67) and −0.32 (0.75), respectively; Table 1). The median follow-up time from ARIC visit 5 to visit 7 was 6.5 years (interquartile interval [IQI]: 6.0, 6.8 years), and from ARIC visit 5 to the EyeDOC visit was 6.2 years (IQI: 5.7, 6.6 years).

Table 1.

Baseline Characteristics of the Entire Study Cohort and Stratified by Community/Race

Characteristics	All	Washington County/White	Jackson/Black	p Value*
N	982	540	442
Age (years)	74 (4)	74 (4)	73 (4)	.001
Male, n (%)	363 (37)	238 (44)	125 (28)	<.001
Education level, n (%)				<.001
Basic education	167 (17)	88 (16)	79 (18)
Intermediate education	392 (40)	265 (49)	127 (29)
Advanced education	423 (43)	187 (35)	236 (53)
Diabetes, n (%)	350 (36)	180 (33)	170 (39)	.11
Hypertension, n (%)	757 (77)	375 (69)	382 (87)	<.001
Presenting distance visual acuity in better-eye (logMAR)	0.16 (0.15)	0.16 (0.15)	0.15 (0.15)	.15
Presenting distance visual acuity in worse-eye (logMAR)	0.34 (0.32)	0.33 (0.28)	0.34 (0.37)	.47
Corrected distance visual acuity in better-eye (logMAR)	0.07 (0.12)	0.08 (0.12)	0.05 (0.11)	<.001
Corrected distance visual acuity in worse-eye (logMAR)	0.19 (0.31)	0.20 (0.24)	0.19 (0.38)	.82
Contrast sensitivity in better-eye (log)	1.43 (0.17)	1.41 (0.18)	1.44 (0.16)	.02
Contrast sensitivity in worse-eye (log)	1.29 (0.28)	1.28 (0.27)	1.31 (0.29)	.04
Age-related macular degeneration, n (%)	73 (8)	63 (12)	10 (2)	<.001
Glaucoma, n (%)	117 (12)	51 (10)	66 (15)	.03
Diabetic retinopathy, n (%)	19 (2)	12 (2)	7 (2)	.51
Global cognition factor score	0.12 (0.81)	0.49 (0.67)	−0.32 (0.75)	<.001
Memory domain factor score	0.19 (0.69)	0.27 (0.72)	0.08 (0.64)	<.001
Executive function domain factor score	0.06 (0.86)	0.46 (0.70)	−0.43 (0.78)	<.001
Language domain factor score	0.06 (0.80)	0.30 (0.72)	−0.24 (0.78)	<.001

Notes: Values for continuous variables (age, presenting distance visual acuity, corrected distance visual acuity, contrast sensitivity, global cognition factor score, memory domain factor score, executive function domain factor score, and language domain factor score) are given as mean (standard deviation). In this table, values of presenting distance visual acuity, corrected distance visual acuity, and contrast sensitivity are presented in the original form, that is, using a scale of 1 unit, where worse presenting and corrected distance visual acuity are represented by higher values and worse contrast sensitivity is represented by lower values. N = number of participants; log = logarithm; logMAR = log₁₀[minimum angle of resolution].

*p Value represents comparison between Washington County/White and Jackson/Black groups.

Figure 1.

Boxplots showing the distribution of visual function measures in better-eye, stratified by age groups and community/race. Notes: In this figure, values of presenting distance visual acuity in better-eye (A), corrected distance visual acuity in better-eye (B), and contrast sensitivity in better-eye (C) are presented in the original scale, that is, using a scale of 1 unit, and where better presenting and corrected distance visual acuity is represented by lower values and better contrast sensitivity is represented by higher values. log = logarithm; logMAR = log₁₀[minimum angle of resolution].

Visual Function Using the Better-Eye and Global Cognitive Change

In stratified regression analyses using better-eye measurements, we estimated that worse presenting DVA (higher value) was associated with a greater estimated rate of decline in global cognition factor score in Washington County/White participants (−0.08 SDs per 10 years per 0.1 logMAR increment [95% confidence interval, CI: −0.12, −0.04]). In contrast, worse presenting DVA was associated with no significant difference in the estimated rate of change per 10 years in global cognition factor score in Jackson/Black group (Table 2).

Table 2.

Generalized Estimating Equation Regression Estimates (95% confidence interval) of Vision Measures in the Better-Eye With Rate of Change in Cognitive Outcomes per 10 Years, Stratified by Community/Race (Washington County/White = 540, Jackson/Black = 442)

		Rate of Change in Global Cognition Factor Score per 10 Years (SD)	Rate of Change in Memory Domain Factor Score per 10 Years (SD)	Rate of Change in Executive Function Domain Factor Score per 10 Years (SD)	Rate of Change in Language Domain Factor Score per 10 Years (SD)
Presenting distance visual acuity (0.1 logMAR worse)	Washington County/White	−0.08 (−0.12, −0.04)	−0.06 (−0.11, −0.02)	−0.08 (−0.12, −0.03)	−0.03 (−0.08, 0.01)
	Jackson/Black	0.007 (−0.03, 0.04)	0.01 (−0.04, 0.06)	0.03 (−0.01, 0.08)	−0.01 (−0.07, 0.04)
Corrected distance visual acuity (0.1 logMAR worse)	Washington County/White	−0.07 (−0.12, −0.01)	−0.06 (−0.13, −0.002)	−0.10 (−0.15, −0.04)	0.02 (−0.02, 0.08)
	Jackson/Black	0.01 (−0.05, 0.07)	0.004 (−0.06, 0.07)	0.04 (−0.01, 0.10)	0.05 (−0.02, 0.14)
Contrast sensitivity (0.1 log better)	Washington County/White	0.10 (0.06, 0.14)	0.08 (0.03, 0.12)	0.10 (0.05, 0.14)	0.04 (−0.003, 0.08)
	Jackson/Black	0.009 (−0.03, 0.05)	0.04 (−0.008, 0.08)	−0.02 (−0.06, 0.01)	0.01 (−0.04, 0.07)

Notes: Models are adjusted for age, sex, education level, diabetes, hypertension, follow-up time, vision:follow-up time interaction. All vision measures are in the scale of 0.1 unit. For presenting and corrected distance visual acuity, 0.1 logMAR represented 5 letters or 1 line on the ETDRS chart. For contrast sensitivity, 0.1 log represented 2.5 letters on the MARS chart. ETDRS = Early Treatment Diabetic Retinopathy Study; log = logarithm; logMAR = log₁₀[minimum angle of resolution]; SD = standard deviation.

We also estimated that better CS (higher value) was associated with a lower estimated rate of decline in global cognition factor score in Washington County/White group (0.10 SDs per 10 years per 0.1 log increment [95% CI: 0.06, 0.14]). In contrast, better CS was associated with no significant difference in the estimated rate of change per 10 years in global cognition factor score in Jackson/Black group.

Visual Function Using the Better-Eye and Domain-Specific Cognitive Change

Worse presenting DVA was associated with a greater estimated rate of decline in memory domain factor score in Washington County/White group (−0.06 SDs per 10 years per 0.1 logMAR increment [95% CI: −0.11, −0.02]). In contrast, it was associated with no significant difference in the estimated rate of decline per 10 years in memory domain factor score in Jackson/Black group, similar to the results for global cognition. Worse presenting DVA was also associated with a greater estimated rate of decline in executive function domain factor score in Washington County/White group (−0.08 SDs per 10 years per 0.1 logMAR increment [95% CI: −0.12, −0.03]), but not in Jackson/Black group.

Better CS was associated with a lower estimated rate of decline in executive function domain factor score among Washington County/White group (0.10 SDs per 10 years per 0.1 log increment [95% CI: 0.05, 0.14]), while no significant 10-year rate of change in executive function domain factor score was noted in Jackson/Black group.

Comparisons of Results Using Corrected Versus Presenting DVA, and Better- Versus Worse-Eye

As individuals with worsening cognitive function may be less likely to access vision care for refractive error, we compared the strength of relationships using corrected versus presenting DVA. Worse presenting DVA was associated with greater rates of decline in global cognition, memory domain, and executive function domain factor scores in Washington County/White group. Replacing presenting with corrected DVA resulted in slightly weakened results in Washington County/White group (Table 2). Relationships consistently remained null among Jackson/Black group.

We also compared our results using measurements from better-eye with those using worse-eye. We observed that most effect estimates were similar in terms of direction but those in the worse-eye were consistently weaker in magnitude (Tables 2 and 3).

Table 3.

Generalized Estimating Equation Regression Estimates (95% confidence interval) of Vision Measures in the Worse-Eye With Rate of Change in Cognitive Outcomes per 10 Years, Stratified by Community/Race (Washington County/White = 540, Jackson/Black = 442)

		Rate of Change in Global Cognition Factor Score per 10 Years (SD)	Rate of Change in Memory Domain Factor Score per 10 Years (SD)	Rate of Change in Executive Function Domain Factor Score per 10 Years (SD)	Rate of Change in Language Domain Factor Score per 10 Years (SD)
Presenting distance visual acuity (0.1 logMAR worse)	Washington County/White	−0.05 (−0.07, −0.02)	−0.05 (−0.08, −0.02)	−0.04 (−0.07, −0.02)	−0.01 (−0.04, 0.006)
	Jackson/Black	−0.004 (−0.02, 0.01)	−0.004 (−0.02, 0.01)	0.002 (−0.01, 0.02)	−0.004 (−0.02, 0.02)
Corrected distance visual acuity (0.1 logMAR worse)	Washington County/White	−0.04 (−0.06, −0.01)	−0.04 (−0.08, −0.01)	−0.04 (−0.06, −0.01)	−0.0006 (−0.03, 0.02)
	Jackson/Black	−0.003 (−0.02, 0.01)	−0.006 (−0.02, 0.01)	0.002 (−0.01, 0.02)	0.001 (−0.02, 0.02)
Contrast sensitivity (0.1 log better)	Washington County/White	0.04 (0.01, 0.06)	0.06 (0.03, 0.08)	0.03 (0.01, 0.06)	0.01 (−0.01, 0.04)
	Jackson/Black	0.007 (−0.01, 0.02)	0.02 (−0.003, 0.04)	−0.008 (−0.03, 0.01)	0.003 (−0.02, 0.03)

Notes: Models are adjusted for age, sex, education level, diabetes, hypertension, follow-up time, vision: follow-up time interaction. All vision measures are in the scale of 0.1 unit. For presenting and corrected distance visual acuity, 0.1 logMAR represented 5 letters or 1 line on the ETDRS chart. For contrast sensitivity, 0.1 log represented 2.5 letters on the MARS chart. ETDRS = Early Treatment Diabetic Retinopathy Study; log = logarithm; logMAR = log₁₀[minimum angle of resolution]; SD = standard deviation.

Sensitivity Analyses

In light of the marked difference in AMD prevalence observed between the community/race groups, which could explain our findings, we repeated our analyses in participants without AMD (n = 889). Results were largely similar to those observed in the main cohort (Supplementary Tables 1 and 2).

Discussion

Our results showed that worse vision was associated with greater rates of concurrent decline in global cognition, as well as memory and executive function domains, but only in Washington County/White group, and not in Jackson/Black group. This pattern of associations was consistent regardless of how visual function was assessed. These associations were present independent of comorbidities which cause cognitive decline, like diabetes and hypertension. The EyeDOC Study includes populations from 2 different contexts: White participants from Washington County, Maryland, and Black participants from Jackson, Mississippi, and differences in visual function between White and Black participants have been previously noted (33). More importantly, context-specific differences (eg, differences in educational access and quality, etc.) in communities are well-known contributors to differences in cognitive function and may contribute to differences in cognitive test scores and perhaps differences in risk for dementia (34,45,46). Additional contributors to the observed differences in the results may be that the 2 communities differ by geographic location and urbanicity; Washington County is a largely rural more northern area, whereas Jackson is a city in the southern United States. These factors may also influence the kind of infrastructure and opportunities available to older adults for outdoor physical activities, educational, and social interactions, etc. which are important for maintaining cognitive function, as postulated by the sensory deprivation theory (4). Our findings suggest only the broad conclusion that community/race has an important influence on the extent to which visual limitations affect the rate of cognitive decline in community-based older adults.

To shed light on the mechanisms underlying the associations found in Washington County/White group, we looked across a broad range of visual function assessments that speak to different mechanistic pathways. Presenting DVA is a function of both underlying local or systemic disease-related vision loss and access to vision care, which if poor, can result in poor vision due to refractive error or unremoved cataract (38,47). We found that, compared with presenting DVA, some associations of corrected DVA (which represents the optimal correction for refractive error (38)) were slightly weakened in Washington County/White group. This could suggest that part of the association of presenting DVA with rate of cognitive decline in Washington County/White group results from suboptimal vision correction. Studies have indicated that older adults with dementia are less likely to visit an ophthalmologist or wear their corrective lenses (48), in which case, cognitive changes may be a cause, not a consequence, of impaired vision, although this is unlikely to occur in our cohort because the EyeDOC Study included only nondemented participants. However, worse CS which cannot be corrected was also associated with worsening cognitive scores among Washington County/White participants. Lack of health care for chronic comorbidities that worsen CS may allow for unchecked progression of disease processes like inflammation or vascular/neuronal damage leading to worse CS. This suggests a common cause mechanism for associations whereby systemic disease processes affect both retinal and cerebral tissues.

We also compared results using the visual function of the better-eye versus that of the worse-eye. Visual function of the better-eye is thought to be the most influential on the daily functioning of an individual (4,49). It is consistent with a causal pathway from poor visual function to cognitive decline through mediators such as a loss of social functioning, physical functioning, and decreased mental well-being (4). In contrast, visual function of the worse-eye may better represent underlying ocular pathology that stems from systemic diseases like diabetes or vascular disease. We found stronger associations with the better-eye, suggesting that vision loss may be linked with rate of cognitive decline through mediators like physical and social functioning consistent with the sensory deprivation theory. However, it is important to note that differences in effect estimates were small and may equally be explained by random noise in the data. Future studies should more rigorously examine the extent to which physical and social functioning mediate relationships between visual function and cognitive changes.

Vision studies often only sample a single community or source population, and our results hint at a broader variation in relationships across populations and contexts than what many studies may reveal. This speaks to a need to expand vision research into diverse geographic and racial/ethnic groups to better understand the impact of visual health in different communities. Insights from more diverse and inclusive vision research could not only inform our understanding of how vision affects aging outcomes like cognitive decline, but also provide the evidence needed to efficiently target specific communities/populations for behavioral or therapeutic interventions, where they would be mostly likely to affect health and quality of life.

The strengths of our study are, first, results were derived from 2 racially and contextually distinct communities with a large cohort of older participants. Second, visual and cognitive function assessments were conducted following standard protocols and were not self-reported, reducing variability in our measures. Third, change in cognitive function was evaluated longitudinally based on repeated measurements over 6.5 years (median). However, our results should be interpreted in light of the following limitations. First, our cohort included White and Black races only, and race and community were inextricably connected. Thus, the differences in results across communities cannot be attributed to race alone but rather represent the combined effect of myriad contextual and demographic factors. Further, community differences may not be generalizable to other contexts. Second, vision assessment was conducted 6.2 years (median) after the baseline cognitive function evaluation, concurrent with the assessment of change in cognition over the 6.5 (median) years of follow-up. Hence, we were unable to assert with certainty that the observed associations do not reflect reverse-causal mechanisms, though we have no reason to believe that cognitive decline is likely to cause damage to the retina or affect refraction in nondemented individuals. Lastly, the participants who attended the EyeDOC Study visit had a slightly higher cognitive performance than the larger eligible population (by a 0.14 SD for global cognitive performance) suggesting that our sample may have a selection bias and a narrower variability in the measures, thus limiting our ability to detect relationships.

Given the high prevalence of visual and cognitive impairment in older adults (1–3,6–9), our findings from the Washington County/White community, if generalizable, have possible public health and clinical implications. Our study supports a functional link between some vision measures and cognitive changes in some, but not all groups. The community/race differences observed in our biracial cohort suggest the need to explore external influences on this link. Further, in some communities, regular visual screening may have a meaningful impact on quality of life, as well as decrease disability and dependency. Correction of refractive error, cataract, and ensuring older adults have access to vision care is a simple and effective strategy for promoting healthy aging including better cognitive health.

Funding

This work was supported by the U.S. National Institute on Aging (NIA; grant number 1R01AG052412). The Atherosclerosis Risk in Communities Study is carried out as a collaborative study supported by National Heart, Lung, and Blood Institute (NHLBI; contract numbers HHSN268201700001I, HHSN268201700002I, HHSN268201700003I, HHSN268201700005I, HHSN268201700004I). Neurocognitive data collection was supported by the National Institutes of Health (NHLBI; National Institute of Neurological Disorders and Stroke; U.S. National Institute on Aging; and National Institute on Deafness and Other Communication; contract numbers U01 2U01HL096812, 2U01HL096814, 2U01HL096899, 2U01HL096902, 2U01HL096917), and with previous brain MRI examinations funded by the National Heart, Lung, and Blood Institute (grant number R01-HL70825).

Conflict of Interest

None declared.

Author Contributions

All authors have read and approved of the submission of this manuscript. A.G.A., P.Y.R., A.R.S., and X.G. were involved in study design, methodology, and data collection. L.T.A., A.G.A., and X.G. were responsible for data management, data quality assurance, and analysis. All authors contributed to the interpretation of results and writing/reviewing of this manuscript. A.G.A. obtained funding and provided study supervision.

Ethics Approvals

The study protocols were approved by the Institutional Review Board (IRB) at the Johns Hopkins Bloomberg School of Public Health. The IRB number is H.34.99.07.02.A1.