Visual Impairment, Eye Disease, and 3-Year Cognitive Decline: The Canadian Longitudinal Study on Aging

ABSTRACT

Purpose

To examine the longitudinal association between vision-related variables and the 3-year change in cognitive test scores in a community-dwelling sample of adults and to explore whether sex, education, or hearing loss act as effect modifiers.

Methods

Data came from two waves of a 3-year population-based prospective cohort study (Canadian Longitudinal Study on Aging), which consisted of 30,097 randomly selected people aged 45–85 years from 7 Canadian provinces. Visual impairment (VI) was defined as binocular presenting visual acuity worse than 20/40. Participants were asked if they had ever had a diagnosis of age-related macular degeneration (AMD), glaucoma, or cataract. Cognitive change over 3 years was examined by calculating the difference between baseline and follow-up scores for the Rey Auditory Verbal Learning Test (RAVLT) and the RAVLT delayed test (memory tests), the Controlled Oral Word Association Test (COWAT) and the Animal Naming Test (ANT) (verbal fluency tests), and the Mental Alternation Test (MAT) (processing speed test). Multiple linear regression was used.

Results

VI, AMD, and cataract were not associated with 3-year changes on any of the 5 cognitive tests after adjusting for age, sex, ethnicity, education, income, smoking, diabetes, stroke, heart disease, and province. A report of glaucoma was associated with greater declines in MAT scores (β = -0.60, 95% CI -1.03, -0.18). No effect modification was detected.

Conclusions

Glaucoma was associated with worsening processing speed. Further research to confirm this finding and to understand the possible reason is necessary.

KEYWORDS:

• Visual impairment
• cognitive decline
• glaucoma
• eye disease
• CLSA

Introduction

The loss of vision is very common in older age as 16% of adults aged 75–85 years old have visual impairment in Canada.¹ A decline in cognitive function is also common in older age, which has led to substantial research to understand if the two are related.^2–6 Theories about why they might be related include that they share a common cause⁷ or that the sensory loss itself causes disuse of the brain leading to atrophy.^8–10

Several longitudinal studies have indicated a relationship between visual impairment (VI) and cognitive function.^11–15 However, it is important to confirm these findings because many of them used cognitive tests that rely on vision, which unfairly penalizes a person with vision loss. For example, the 3 MS cognitive test requires that participants copy intersecting pentagons. A person with even minor vision loss, although able to do the test, may make more mistakes, not because of a cognitive problem but because they cannot see as well. By contrast, very few clinical studies have examined the relationship between eye disease and cognitive function.^16,17 Studies on eye disease, instead, have often used health administrative records and have focused on outcomes like Alzheimer’s disease or dementia reporting positive relationships in some cases^18–21 and null findings in others.²² Prior studies have also neglected to examine potential effect modifiers such as sex, education, or hearing loss. The evaluation of effect modification can tell us whether some people are affected by an exposure but not others. For example, some studies have shown that women have higher rates of Alzheimer’s disease than men²³; therefore, maybe vision loss would affect the cognition of women more than men. Similarly, people with more education may be less likely to develop cognitive impairment,²⁴ which may make them less affected by vision loss. Finally, if sensory loss is thought to cause disuse of the brain leading to atrophy then perhaps people with dual vision and hearing impairments would be at an even higher risk of cognitive decline, as has been indicated in the literature.²⁵

The Canadian Longitudinal Study on Aging (CLSA) is a large, population-based cohort study with data on over 30,000 middle-aged and older adults.²⁶ The CLSA used 5 cognitive tests that did not require vision. We utilized this dataset to examine the association between VI or eye disease and 3-year changes in cognitive test scores and to determine whether sex, education, or hearing loss modify any of the associations.

Methods

Study population

A prospective study of community-dwelling older adults was performed using data from rounds 1 and 2 of the Canadian Longitudinal Study on Aging (CLSA) Comprehensive Cohort consisting of 30,097 individuals.²⁶ Stratified random sampling via the use of provincial healthcare registration databases and random digit dialling of landline telephones was done. When sampling from provincial healthcare databases, people who were temporary visa holders or had transitional health coverage (when the information was available) were excluded. Non-permanent residents and non-Canadian citizens were excluded from all sampling frames. Further inclusion criteria were that participants had to be aged between 45 and 85 years, community-dwelling, cognitively unimpaired at baseline, speak English or French, and provide written informed consent. Full-time members of the Canadian Armed Forces, individuals residing on a federal First Nations reserve or settlement, and individuals living in a long-term care institution were excluded.

Study design

Baseline and follow-up assessments included both a home visit and a data collection site visit. All CLSA Comprehensive Cohort study personnel received standardized training on data collection procedures in order to ensure that participant assessments were standardized across the data collection sites. Baseline assessments were done between December 2011 and July 2015 and follow-up assessments were done between July 2015 and December 2018. The 11 CLSA data collection sites are located in Victoria, Vancouver, Surrey, Calgary, Winnipeg, Hamilton, Ottawa, Montreal, Sherbrooke, Halifax and St. John’s. The follow-up rate was 92%. Research Ethics Board approval was received in July 2010 from all affiliated sites. Ethics approval from the University of Ottawa was received for the present analysis in May 2019 (H-05-19-4466).

Data collection

Visual impairment and eye disease

Baseline visual acuity was measured at the data collection sites using an illuminated Early Treatment of Diabetic Retinopathy Study (ETDRS) chart and its standard protocol²⁷ and scores were converted to logarithm of the minimum angle of resolution (logMAR) units. Binocular visual acuity was evaluated at a 2-meter distance from the letter chart using habitual distance correction (i.e. wearing normal corrective lenses for distance vision). Baseline VI was defined as presenting binocular acuity worse than 20/40 (0.301 logMAR), as is often used in North American research.¹ Participants were asked to report if they have or had a diagnosis of cataract, macular degeneration, or glaucoma.

Cognitive change

Cognitive change was examined by calculating the difference between baseline and follow-up scores on validated non-visual cognitive tests including the Rey Auditory Verbal Learning Test (RAVLT) and the delayed-recall RAVLT trial (RAVLT-delayed),²⁸ the Controlled Oral Word Association Test (COWAT),²⁹ the Animal Naming test (ANT),³⁰ and the Mental Alternation Test (MAT).³¹ Tests were administered in-person at study site visits by trained interviewers. These tests are reliable and widely used in neuropsychology to indicate cognitive decline.^32–34

The RAVLT, a test of memory, involves a 15-item list of words read to participants at a rate of one word per second. After presentation of the 15 words, participants were requested to recall as many words as possible. Thirty minutes later, the participant was again asked to recall as many words as possible (delayed RAVLT). Verbal fluency was measured using the COWAT, which measures letter fluency, and the ANT, which measures category fluency. For the COWAT, participants were given one minute to name as many words as possible beginning with a given letter. The trial was repeated 3 times with 3 different letters (F, A, S) with combined scores used for analyses. The ANT is a measure of category verbal fluency in which participants were asked to name as many animals as they can within one minute. Two trials were conducted with the combined score used for analyses. Finally, the MAT, which measures processing speed and mental flexibility, involves a timed performance in which participants were asked to alternate between number and letter (i.e. 1-A, 2-B, 3-C …) as quickly as possible within 30 seconds.

Demographic, health, and lifestyle data

Demographic data including age, sex, ethnicity, education, and income were collected at baseline during the in-home visit using the interviewer-administered questionnaire. Participant education level was determined by asking, “What is the highest degree, certificate or diploma you have obtained?”. Participants were grouped into 3 categories: more than Bachelor’s degree, Bachelor’s degree, less than Bachelor’s degree. In order to assess household income, participants were asked, “What is your best estimate of the total household income received by all household members, from all sources, before taxes and deductions, in the past 12 months?”. Participants were groups into 5 categories: >$150 K, $150-$100 K, $100-$50 K, $50 K-$20 K, Refused.

Regarding health, hearing was assessed by asking participants, “Is your hearing, using a hearing aid if you use one, excellent, very good, fair, or poor?”. For use in our analyses, participant responses were dichotomized into two categories: good, very good, or excellent versus fair or poor. Participants were asked if they had ever received a physician diagnosis of several comorbid conditions including diabetes (none, Type 1, Type 2, suspect), heart disease (no, yes), or stroke (no, yes).

Smoking status was classified as either current, never, or former based on participant responses to the interview questions “Have you smoked at least 100 cigarettes in your life?” and “At the present time, do you smoke cigarettes daily, occasionally (at least once in last 30 days), or not at all (not in last 30 days)?”. A current smoker was defined as a person who reported smoking at least 100 cigarettes and currently smokes daily or occasionally while a former smoker was someone who reported smoking at least 100 cigarettes in life but had not smoked in the last 30 days.

Statistical analysis

Those who were included and excluded from the analyses and those with and without VI at baseline were compared for demographic, health, and lifestyle factors. Associations between vision-related variables and 3-year changes in cognitive scores were tested using simple linear regression. To adjust for potential confounding factors, multiple linear regression was used. Potential confounding variables included age (as a continuous variable), sex, ethnicity, household income, level of education, smoking status, diabetes, heart disease, stroke, and province. These variables were chosen based on previous research showing their importance to vision or cognitive impairment.^3,6 We did not adjust for baseline cognition as a potential confounder out of concern that it could induce a sizeable bias due to measurement error and regression to the mean.³⁵ The relationship between age and each cognitive outcome was graphically assessed using locally weighted scatterplot smoothing. All relationships were linear and therefore age was entered into the regression models as a continuous, linear term. Effect modification by sex (male, female), education (Bachelor’s degree or more, less than Bachelor’s, or hearing level (≥good, <good) was tested by adding product terms into regression models. Practice effects, or improved performance with repeated testing, are sometimes seen with cognitive tests. We did not adjust for practice effects due to a study by Racine et al reporting that it was not necessary when looking at relative effects.³⁶ Statistical significance was defined as P < .05. Sampling weights and strata variables were incorporated into all analyses using the SVY commands in STATA SE Version 16 (College Station, Texas).

Results

Description of the sample

Our analysis consisted of 27,412 individuals after excluding those missing vision data at baseline (n = 431) and those lost to follow-up (n = 2,254). Of the 2,254 who did not return to follow-up, 553 died (25%) and 938 withdrew (42%) with the remaining 763 (34%) either unable to be contacted or unable to complete data collection. Participants who were lost to follow-up were older, more likely to be female, had lower education, lower household income, and had worse baseline visual acuity and cognitive scores compared to those with who were not lost to follow-up (Table 1).

Table 1. Descriptive statistics of those who were and were not included in the analysis due to missing follow-up or baseline vision data.

Variable	Included n = 27,412 Mean (SD)/Row %	Excluded n = 2,685 Mean (SD)/Row %
Age, Years	59.2 (9.7)	62.3 (12.1)
Sex
Female (n = 13,932)	50.6	60.0
Male (n = 13,480)	49.4	40.0
Ethnic or Racial Background
White (n = 25,897)	93.8	94.5
Black (n = 220)	0.9	1.7
Asian (n = 639)	2.5	2.1
Aboriginal (n = 327)	1.3	0.6
Other (n = 329)	1.5	1.1
Education
More than Bachelor’s (n = 6,106)	22.4	22.6
Bachelor’s (n = 6,602)	25.4	22.2
Less than Bachelor’s (n = 14,660)	52.2	55.2
Annual Household Income
≥ $150,000 (n = 9,748)	43.6	33.3
$50,000 -$100,000 (n = 9,162)	31.6	34.2
$20,000 – $50,000 (n = 5,522)	15.9	22.0
<$20,000 (n = 1,275)	3.5	6.1
Refused/ Don’t know (n = 1,705)	5.3	4.4
Baseline RAVLT	6.1 (1.9)	5.4 (2.0)
Baseline RAVLT Delayed	4.3 (2.2)	3.6 (2.3)
Baseline COWAT	40.3 (12.6)	35.9 (14.0)
Baseline ANT	42.8 (12.0)	37.6 (13.1)
Baseline MAT	27.6 (8.6)	24.4 (9.8)
Baseline Visual Acuity, logMAR	0.01 (0.2)	0.06 (0.2)

SD = standard deviation; RAVLT = Rey Auditory Verbal Learning Test; COWAT = Controlled Oral Word Association Test; ANT = Animal Naming Test; MAT = Mental Alternation Test

Of our analytic sample, the mean age of participants, after correction for the complex survey design, was 59.5 years (SD = 9.9) and 50.8% were female. Overall, at baseline, 5.7% of participants in the cohort had VI, 21.5% reported ever having had a cataract diagnosis, 3.2% reported a diagnosis of macular degeneration, and 3.9% reported a diagnosis of glaucoma.

Baseline characteristics of the cohort by VI status are presented in Table 2. Those with VI were more likely to be older, female, have lower education and annual household income levels, be current smokers, have diabetes, stroke, and heart disease as compared to those without VI.

Table 2. Descriptive statistics of those with and without visual impairment.

Variable	VI n = 1,925 Mean (SD)/Column %	No VI n = 25,487 Mean (SD)/Column %
Age, Years	65.8 (12.1)	58.8 (9.4)
Sex
Female (n = 13,932)	52.7	50.5
Male (n = 13,480)	47.3	49.5
Ethnic or Racial Background
White (n = 25,897)	93.7	93.8
Black (n = 220)	0.5	0.9
Asian (n = 639)	2.7	2.5
Aboriginal (n = 327)	1.5	1.3
Other (n = 329)	1.6	1.5
Education
More than Bachelor’s (n = 6,106)	20.2	22.6
Bachelor’s (n = 6,602)	21.6	25.6
Less than Bachelor’s (n = 14,660)	58.2	51.8
Annual Household Income
≥ $150,000 (n = 9,748)	29.1	44.4
$50,000 -$100,000 (n = 9,162)	33.0	31.5
$20,000 – $50,000 (n = 5,522)	23.6	15.5
<$20,000 (n = 1,275)	6.7	3.4
Refused/ Don’t know (n = 1,705)	7.6	5.2
Smoking
Current (n = 2,315)	46.6	50.7
Former (n = 11,906)	43.9	41.0
Never (n = 13,190)	9.5	8.3
Diabetes
None (n = 22,646)	80.6	85.6
Type 1 (n = 148)	0.9	0.5
Type 2 (n = 2,400)	11.3	7.2
Suspect (n = 1,928)	7.2	6.7
Stroke
No (n = 26,898)	97.2	98.9
Yes (n = 423)	2.8	1.1
Heart Disease
No (n = 24,252)	86.1	91.7
Yes (n = 3,020)	13.9	8.3

VI = visual impairment; SD = standard deviation

Visual impairment and 3-year change in cognitive scores

The baseline, follow-up, and 3-year change scores are shown in Table 3. The RAVLT, RAVLT-delayed, and COWAT test results were better at follow-up compared to baseline. This improvement was expected and likely indicates a practice effect for these tests.³⁷ Over the 3 years, people with VI had smaller improvements on the RAVLT and RAVLT-delayed scores than those without VI (P < .001). People with VI had a greater decline on the ANT than those without VI although this only reached borderline statistical significance (P = .06). Changes in the COWAT (P = .17) and MAT scores (P = .75) did not differ between those with and without VI.

Table 3. Visual impairment and 3-year change in cognitive scores.

Variable	VI n = 1,925 Mean (SD)	No VI n = 25,487 Mean (SD)	P-value
Baseline RAVLT	5.6 (2.2)	6.2 (1.9)	<0.01
Follow-up RAVLT	6.1 (2.5)	6.9 (2.2)	<0.01
3-Year Change RAVLT	0.4 (2.1)	0.8 (1.9)	<0.01
Baseline RAVLT Delayed	3.8 (2.4)	4.4 (2.2)	<0.01
Follow-up RAVLT Delayed	4.2 (2.7)	5.1 (2.4)	<0.01
3-Year Change RAVLT Delayed	0.4 (2.3)	0.7 (2.1)	<0.01
Baseline COWAT	38.3 (14.6)	40.6 (12.4)	<0.01
Follow-up COWAT	38.7 (14.4)	41.3 (12.2)	<0.01
3-Year Change COWAT	0.4 (8.2)	0.8 (7.9)	0.17
Baseline ANT	39.2 (13.8)	43.2 (12.1)	<0.01
Follow-up ANT	38.2 (13.2)	42.8 (11.3)	<0.01
3-Year Change ANT	−0.9 (10.5)	−0.4 (9.7)	0.06
Baseline MAT	25.7 (10.4)	28.0 (8.6)	<0.01
Follow-up MAT	25.1 (9.1)	27.5 (7.5)	<0.01
3-Year Change MAT	−0.6 (7.4)	−0.5 (6.4)	0.75

RAVLT = Rey Auditory Verbal Learning Test; COWAT = Controlled Oral Word Association Test; ANT = Animal Naming Test; MAT = Mental Alternation Test; VI = visual impairment; SD = standard deviation

In Table 4, linear regression was used to adjust for age, sex, ethnicity, education, income, smoking, diabetes, stroke, heart disease, and province. VI was not statistically significantly associated with any of the 5 cognitive change outcomes. VI had a borderline association with the RAVLT (β = −0.11, 95% −0.22, 0.00, P = .061).

Table 4. Relationships between ocular variables and 3-year changes in cognitive score using multiple linear regression models*.

		COGNITIVE	CHANGE	OUTCOMES
Ocular Variable	RAVLT β 95% CI	RAVLT Delayed β 95% CI	COWAT β 95% CI	ANT β 95% CI	MAT β 95% CI
Visual impairment
No	reference	reference	reference	reference	reference
Yes	−0.11 − 0.22, 0.00	−0.05 −0.18, 0.07	−0.00 −0.49, 0.48	−0.28 −0.84, 0.29	0.26 −0.15, 0.66
Cataract
No	reference
Yes	−0.02 −0.09, 0.06	−0.04 −0.12, 0.04	0.06 −0.25, 0.36	−0.21 −0.58, 0.15	−0.10 −0.35, 0.16
AMD
No	reference
Yes	0.03 −0.10, 0.17	−0.01 −0.16, 0.15	−0.02 −0.62, 0.58	−0.36 −1.05, 0.33	−0.04 −0.51, 0.42
Glaucoma
No	reference
Yes	0.04 −0.09, 0.16	−0.08 −0.22, 0.06	0.25 −0.30, 0.80	−0.52 −1.16, 0.12	−0.60 −1.03, −0.18

RAVLT = Reyes Auditory Verbal Learning Test; COWAT = Controlled Oral Word Association Test; ANT = Animal Naming Test; MAT = Mental Alternation Test; SD = standard deviation; CI = confidence interval

*Each ocular variable was in a separate model adjusted for age, sex, ethnicity, education, income, smoking, diabetes, stroke, heart disease, and province.

Age-related eye disease and 3-year change in cognitive scores

In Table 4, those with a report of glaucoma had a greater 3-year decline in MAT scores (β = −0.60, 95% CI = −1.03, −0.18) than those who did not report glaucoma after adjustment. Glaucoma was not significantly associated with the other 3-year changes in cognitive scores. AMD and cataract were not associated with any of the changes in cognitive scores.

Sensitivity of the results to adjustment for baseline cognition

We did not adjust our models for baseline cognition out of a concern that it could induce bias.³⁵ However, our results for VI were very different when we did adjust for baseline cognition. Specifically, VI was associated with 3 of the 5 cognitive outcomes including the RAVLT (β = −0.18, 95% confidence interval (CI) = −0.28, −0.07), the RAVLT-Delayed (β = −0.13, 95% CI = −0.25, −0.02), and ANT scores (β = −0.95, 95% CI = −1.44, −0.45). By contrast, the conclusions were unaffected between age-related eye disease and cognitive change.

Effect modification

No evidence of effect modification was found. None of the interaction terms between visual impairment/eye disease and hearing impairment, low education, or sex were statistically significant.

Supplemental results

To be consistent with other literature, cross-sectional results are also provided. VI was associated with worse scores on all 5 baseline cognitive outcomes, while neither glaucoma, AMD, nor cataract were associated with any baseline cognitive outcomes (Supplemental Table 1).

Results for those 65 and older are also provided to be consistent with other literature. Regression coefficients were not statistically significantly different from those less than 65 years old (Supplemental Table 2).

Discussion

Visual impairment, AMD, and cataract were not associated with 3-year changes in cognitive function for any of the 5 cognitive tests. People who reported having glaucoma had greater declines on the Mental Alternation test, which measures processing speed.

To our knowledge, no other epidemiological studies have examined whether glaucoma is related to information processing speed using a non-visual test like the MAT. Slower processing speed is thought to be a key driver of cognitive aging.^38–40 In support of our finding, small clinical studies have reported that glaucoma patients have structural brain abnormalities and altered anatomical and functional brain connectivity.^41,42 It is possible that age-related conditions like glaucoma and cognitive decline share a common pathology. Sivak et al described how eye diseases like glaucoma share common characteristics with Alzheimer’s disease like glial reactivity, neuroinflammation, and oxidative stress.⁷ Alternatively, the Disuse Hypothesis suggests that cognitive atrophy could develop due to difficulties doing activities that require vision.^8–10 Research has shown that participating in more cognitive activities is associated with a reduced risk of developing dementia.^43,44 We previously reported that people with moderate and severe glaucoma performed fewer activities like reading, playing games, exercising, and socializing than people with normal vision.⁴⁵ Previous studies have shown glaucoma to be related to worse scores on the verbal digit forward and backward tests and the logical memory test,¹⁷ the MMSE-Blind,¹⁶ and with increased Alzheimer’s disease.⁴⁶ Other studies have not shown associations between glaucoma and Alzheimer’s disease.^22,47

We did not find that AMD or cataract were related to cognitive decline. Our measures of eye disease were based on self-report and no information on severity of eye disease was available. One other study found an association between clinically diagnosed AMD and cognition measured using the MMSE-Blind¹⁶ while another study did not find an association between clinically diagnosed AMD and measures of cognitive function like tests of the verbal digit span and verbal fluency.¹⁷ Previous studies using health administrative records have reported associations between glaucoma/AMD and Alzheimer’s disease or dementia.^18–21 Few studies have been conducted on cataract and cognition. One previous study found that cognitive decline was slower after cataract surgery than before it⁴⁸ although a randomized controlled trial found no cognitive benefit to cataract surgery.⁴⁹

Consistent with our null findings for AMD and cataract, we also found that VI was not associated with change in cognitive function, although in a sensitivity analysis, VI was associated with worse change scores on 3 of 5 cognitive tests when we adjusted for baseline cognitive function. A null relationship between VI and cognitive change was also reported in 4 other longitudinal studies.^25,50–52 For example, Hong et al⁵¹ performed a prospective, population-based study of 3,654 participants of the Blue Mountains Eye Study in Australia and reported that VI was not associated with a ≥ 3 point decline in MMSE-Blind scores over 5 years (OR 0.84; 95% CI 0.40–1.79) or 10 years (OR = 1.09; 95% CI 0.52–2.30). This is in contrast to 4 longitudinal studies and one meta-analysis that have found a relationship between VI and cognitive function^{12–14,53,54} and 1 study that found a relationship between visual acuity, used as a continuous variable, and cognitive decline.⁵⁵ Furthermore, several studies have found a relationship between VI and dementia or cognitive impairment^56–60 although not all.⁶¹ Of note, some of these studies used cognitive tests that relied on vision without removing the visual items (MMSE, 3 MS).^12,53 It is unclear why there is so much disagreement in the literature. Future studies should be longitudinal, should use non-visual cognitive testing, and should use measured rather than self-reported visual function.

Future studies should also investigate effect modification, if possible. We did not find evidence of effect modification. The relationships between ocular variables and cognitive decline did not vary by sex, education, hearing loss, or age group. Although there was some heterogeneity by strata, it was not statistically significant. This highlights a methodological challenge in the detection of effect modification. With relatively rare exposures like visual impairment and eye disease, it can be difficult to have sufficient power to detect effect modification unless the study is specifically designed for that purpose or the raw data from studies are pooled together.

One major strength of this research was that the longitudinal nature of the CLSA data allowed us to establish that the vision loss/eye disease was present before the cognitive change occurred. This is a clear advantage over the use of cross-sectional data to address this research question because a measure of association from cross-sectional data can also capture an association due to reverse causality (people with poor cognition delaying their vision care and developing avoidable vision loss). Second, the many variables collected in the CLSA also allowed us to adjust for a large number of potentially confounding variables and to examine potential effect modification. Third, the CLSA included both middle and older-aged adults unlike most other studies. Finally, multiple valid and reliable non-visual cognitive tests were used rather than a single global test of cognition. This allowed us to investigate the specific aspects of cognition that are related to vision or eye disease and to ensure that impaired vision is not interfering with the assessment of cognition.

A limitation of this study is the use of self-reported data on the presence of eye disease. The self-report of eye disease is known to have limited validity, which may result in misclassification.⁶² For example, people might not know that they have glaucoma until late in the disease process because the vision loss begins in the periphery, where it is less noticeable, or people may think they have glaucoma because they take pressure-lowering eye drops, but they have ocular hypertension instead. However, the prevalence of glaucoma that we found in the CLSA (3.9%) is fairly similar to what has been reported in other countries through the use of a comprehensive eye exam (3.0% in the Blue Mountains Eye Study).⁶³ Misclassification of eye disease could be differential in that those with more cognitive decline over the 3 years might have been less likely to accurately report their eye disease at baseline. This could result in regression coefficients that are either biased towards or away from the null. Another limitation is that the follow-up period was only 3 years, which is rather short for detecting cognitive change, especially in middle-aged adults. This may have contributed to some of our null findings. Finally, our results may not be generalizable to those excluded from the CLSA sampling frame.

To conclude, visual impairment, AMD, and cataract were not associated with cognitive decline. However, glaucoma was associated with a reduction in processing speed. Further longitudinal research using a comprehensive eye exam is needed to confirm and explain this finding.

Acknowledgments

This research has been conducted using the CLSA Comprehensive Dataset version 4.0 and Follow-up 1 Comprehensive Dataset version 1.0 under Application Number 190212. The CLSA is led by Drs. Parminder Raina, Christina Wolfson, and Susan Kirkland. The opinions expressed in this manuscript are the author’s own and do not reflect the views of the Canadian Longitudinal Study on Aging. Funding for the Canadian Longitudinal Study on Aging (CLSA) is provided by the Government of Canada through the Canadian Institutes of Health Research (CIHR) under grant reference: LSA 94473 and the Canada Foundation for Innovation and by CIHR operating grant ACD 170303 (EF). The funders had no role in the design, analysis, or the interpretation of results. No potential conflict of interest was reported by the authors.

Data availability

Data are available from the Canadian Longitudinal Study on Aging (www.clsa-elcv.ca) for researchers who meet the criteria for access to de-identified CLSA data.

Additional information

Funding

This work was supported by the Canadian Institutes of Health Research under grants ACD-170303 and LSA 94473 and the Canada Foundation for Innovation.