Abstract
Background:
Vision impairment is a risk factor for mild cognitive impairment (MCI) among stroke survivors, but it is unclear if this association is driven by vision impairment present before or due to the stroke, and if similar associations exist with dementia.
Objective:
To (1) characterize the prevalence of pre-stroke and stroke-related vision impairment(s) among stroke survivors, and (2) quantify associations of vision impairment with dementia and cognitive impairment (MCI/dementia).
Methods:
Using participants from the Atherosclerosis Risk in Communities (ARIC) dataset with adjudicated incident strokes, we gathered descriptive statistics on the cohort, assessed if vision impairment was present at the time of incident stroke, and classified the impairments as pre-stroke or stroke-related. Multivariable logistic regression was used to estimate the association between these types of vision impairment and post-stroke cognitive impairment.
Results:
Among 233 incident stroke survivors (mean=69 years old and 50.2% female sex), 23.2% with pre-stroke vision impairment and 18.9% with stroke-related vision impairment, there were 124 (53%) cases of cognitive impairment (n=76 MCI, n=48 dementia). Stroke-related vision impairment was significantly associated with higher odds of dementia (ref =normal/MCI) (adjusted odds ratio (aOR)=2.32, 95% confidence interval (CI)=1.08–4.92, p=0.029), but not any cognitive impairment (ref=normal) (aOR= 1.33 95% CI= 0.67–2.70, p=0.425). Further adjusting for stroke severity score attenuated the association of stroke-related vision impairment with dementia (aOR= 2.0, 95% CI= 0.90, 4.32, p=0.08).
Conclusions:
Stroke-related vision impairment, but not pre-stroke vision impairment, was associated with higher odds of dementia. There is evidence that stroke severity could, at least partially, explain the observed association.
Keywords: vision impairment, mild cognitive impairment, dementia, stroke, ARIC
Introduction
Strokes are extremely common diseases in the United States.1 Stroke survivors can experience many medical conditions as a result of the stroke, one of which is stroke-related vision impairment, which has been reported to be a common sequela that affects as many as 60% of survivors.2 Stroke-related vision impairment alone can be extremely debilitating,3,4 and can contribute to increased likelihood of additional symptoms and overall disability, as measured by the National Institutes of Health Stroke Scale (NIHSS) and the modified Rankin Scale, when compared to patients without vision impairments.4,5
Another type of vision impairment that is common in stroke survivors is vision impairment due to age-related eye conditions such as macular degeneration, glaucoma, and cataracts, for which etiology is unrelated to the stroke. We and others have reported that stroke-related vision impairment and age-related vision conditions can co-exist6–8 and can negatively impact functional activities such as reading.9 Many times, these age-related vision impairments develop prior to the stroke especially if the stroke survivor is older in age.8
In addition to vision impairment, stroke survivors can experience cognitive impairment due to the stroke itself.7,10 To better understand the contribution vision impairment has on post-stroke cognitive changes, we and others have used extant data and showed that there is a strong association between self-reported vision impairment and cognitive impairment in stroke survivors.10,11 Specifically we found that stroke survivors with worse self-reported vision had lower cognitive functioning than those with excellent-to-very good vision; but we did not find that they had greater cognitive decline over time when compared to stroke survivors with excellent-to-very good vision.10 These findings were limited because vision impairment was defined using a self-reported rating of their vision, which lacks granularity about the visual symptoms the person is experiencing.10 This makes it difficult to determine which types of vision impairment (either or both pre-stroke or stroke-related) confer the greatest risk for post-stroke cognitive impairment. This is necessary to know in order to develop informed and personalized interventions. Others reported that visual field defects are a marker of having a larger more severe stroke5, which could therefore provide some insight into explaining these relationships and risks.
In addition, it is unclear exactly how specific types of vision impairments and dementia may be interrelated in stroke survivors. There is one report on people with vision impairment due to retinopathy being more likely to have a history of stroke as well as be at increased risk of vascular contributions to cognitive impairment and dementia.12 The epidemiologic link may be due to shared risk factors rather than a causal relationship between vision loss and cognitive decline.13 In other words, people with stroke could be at higher risk of vision impairment2, or some of the risk factors for having a non-stroke-related vision impairment (e.g., diabetic retinopathy) may also be a risk factor for dementia (e.g., diabetes).14
To address the lack of research on the associations between specific vision impairments and cognitive impairment (mild cognitive impairment [MCI] or dementia) in a cohort of stroke survivors, we first assessed the prevalence of well-characterized vision impairments in a group of stroke survivors. Next, we evaluated whether the presence of pre-stroke vision impairment or stroke-related vision impairment is associated with post-stroke cognitive impairment (MCI or dementia) or dementia alone. We hypothesized that stroke survivors with stroke-related vision impairments will have higher odds of cognitive impairment and also dementia compared to those with normal vision. We also hypothesized that adjusting for stroke severity via the NIHSS score would reduce variability in the model and strengthen the associations with cognitive impairment and dementia.
Materials and Methods
Study Population
The Atherosclerosis Risk in Communities (ARIC) study is an ongoing prospective community-based study, that began data collection in 1987 with adults between the ages of 45 and 64 years of age who self-identify as Black or White.15 While the primary objective of the ARIC study was to investigate the etiology of clinical atherosclerotic risk factors and diseases in middle age, the diversity of data collected on the ARIC cohort for over 35 years has allowed a much wider array of questions to be investigated.16 The ARIC cohort of almost 16,000 participants was sampled from four communities including Forsyth County, North Carolina, Jackson, Mississippi, Minneapolis, Minnesota, and Washington County, Maryland. The study involves a home visit, several clinic exams, annual (prior to 2012) or semi-annual (since 2012) telephone follow ups, and determination and adjudication of clinical events through surveillance.
Cohort Selection
Using ARIC study cohort data, we conducted a secondary analysis examining the association between vision impairment at stroke and subsequent cognitive status among stroke survivors. We selected the incident ischemic stroke event as our exposure timepoint and the ARIC visit with the most recent available cognitive status as our outcome timepoint. Stroke events were determined from stroke hospitalizations captured from the cohort surveillance from 1987–2022 (except the Jackson center stroke, which were only captured through 2020, because of difficulties accessing hospital records). We considered only strokes that had an adjudicated final diagnosis of a probable ischemic stroke or a definite ischemic stroke, which was defined as either cardioembolic or thrombotic brain infarctions.17 We then subset the cohort to the 1069 participants that had stroke events with a calculated stroke severity score (NIHSS), as described in Koton et al, 2022.18 This limited our dataset to those with incident strokes prior to 2019. We retained participants who had at least one available cognitive diagnosis from the ARIC Neurocognitive Study (NCS) after their incident stroke – at visit 5 (2011–2013), visit 6 (2016–2017), visit 7 (2018–2019), or visit 8 (2020). We retained the latest, or most recent, visit with cognitive status for analysis. If there was no stroke severity score corresponding to a participant’s incident ischemic stroke, they were excluded (n=2). For analyses including pre-stroke vision impairment, we additionally excluded 28 participants without a retinal form prior to their incident ischemic stroke event. See Figure 1 for the CONSORT diagram describing the cohort selection and Figure 2 for the schematic of the cohort timeline.
Figure 1:
CONSORT Diagram describing cohort selection
*Stroke-related vision impairment analyses use the final cohort
** The pre-existing vision impairment and any vision impairment analyses use the sub-cohort
Figure 2:
Timeline of Data Collection
Definition of Primary Exposures
Pre-stroke Vision Impairment.
Retinal photography, along with a retinal examination form, was conducted at ARIC visits 3, 4, and 5. For this study, we used the retinal examination form questions, which concerns the ophthalmic history of each participant. Specifically, participants were asked to self-report on whether they have ever been told by a doctor that they have a variety of eye conditions. We gathered information on glaucoma, macular degeneration, cataracts, and blindness in one/both eyes (not asked at visit 5) for this analysis. Participants who reported at least one of these conditions from their most recent retinal form before their incident ischemic stroke event were given a status of pre-stroke vision impairment (yes/no).
Stroke-Related Vision Impairment.
Along with stroke diagnosis, the ARIC cohort surveillance also captured the presence of hemianopia (visual field loss), diplopia (double vision), and cranial nerve palsy (dysfunctional nerves) either present on or leading to hospital admission or occurring during hospitalization. Participants with at least one of these conditions at their incident ischemic stroke event are given a status of stroke-related vision impairment (binary yes/no).
Any Vision Impairment.
Participants with pre-stroke vision impairment and/or stroke-related vision impairment were considered to have any vision impairment.
Definition of Potential Confounders and Descriptive Covariates
Descriptive data was captured for each participant; namely age, sex, race, diabetes, hypertension, education, and stroke severity. Race-center, sex, and education were self-reported at visit 1 (ARIC baseline). Sex was coded as male/ female; race-center with categories Black Forsyth NC, Black Jackson MS, White Forsyth NC, White Minneapolis MN, and White Washington County MD; and education was coded as ‘Did Not Complete High School’/ ‘High School, GED, or Vocational School’/ ‘At least some college’. Age, hypertension, and diabetes were taken from the outcome timepoint, or the most recent visit with a cognitive measurement. Hypertension was defined as systolic blood pressure (mean of second and third of three measures) ≥ 140, diastolic blood pressure (mean of second and third of three measures) ≥ 90, or use of medication for high blood pressure. Diabetes was defined as fasting glucose values ≥ 126, non-fasting glucose greater than or equal to 200 mg/dL, taking medication for diabetes, or a diagnosis of diabetes by a physician (self-reported). For participants whose cognitive measurements were taken from Visit 8, their diabetes and hypertension measures were taken from Visit 7 if available, since these measurements were not available for Visit 8 (converted to a virtual, phone-based visit due to the COVID-19 pandemic). Stroke severity, which was calculated by Koton et al 2022, who trained clinicians to use an algorithm for retrospective scoring of the NIHSS, was taken from the incident ischemic stroke event.18 They classified the continuous scores into 4 levels: minor (NIHSS ≤5), mild (NIHSS 6–10), moderate (NIHSS 11–15) and severe (NIHSS ≥16). The calculation of the NIHSS involves an item related to visual fields, requiring the administrator to test by confrontation to determine if there is no visual field loss, partial, complete, or bilateral vision loss. Therefore, we expect NIHSS to be associated to stroke-related vision impairment.
Definition of Outcomes
Cognitive Impairment.
The ARIC Neurocognitive Study (NCS) was conducted at visits 5, 6, 7, and 8 and resulted in a cognitive diagnosis for each participant.19 These diagnoses were adjudicated based on in-person cognitive testing (with the exception of Visit 8, which was administered over the phone) and informant interviews, using an algorithm devised as a diagnosis guide, as described further in Knopman et al, 2016.19 The possible diagnoses include normal, mild cognitive impairment (MCI), and dementia. We operationalized two outcome definitions for our models defined as: (1) dementia alone and (2) any cognitive impairment (MCI or dementia).
Statistical Analysis
Demographic characteristics are presented for the overall cohort and stratified by cognitive status; we reported mean and standard deviation for continuous variables and frequency and percentages for categorical variables. Further, we compared the mean continuous stroke severity scores between (1) those with stroke-related vision impairment and those without and (2) those included in the analytic cohort and those excluded (as described in Figure 1), using a Welch two-sample t-test. We also tested for independence between stroke-related vision impairment and inclusion in the analytic dataset using a chi-square test. Six logistic regression models were run with each combination of the following outcomes: (1) dementia vs. no dementia (i.e., MCI or normal) and (2) any cognitive impairment vs. no cognitive impairment (i.e., normal), with the following primary predictors: (1) pre-stroke vision impairment vs. no pre-stroke vision impairment, (2) stroke-related vision impairment vs. no stroke-related vision impairment, and (3) a composite measure of any vision impairment (including pre-stroke vision impairment and/or stroke-related vision impairment) vs. no vision impairment. All six models were run unadjusted and adjusted for age, sex, education, and number of years since the incident ischemic stroke. These potential confounders were chosen as they are expected to be strongly related to the outcome of interest and may also be related to the predictor of interest. It is known that older adults have different trajectories to cognitive impairment than someone who is younger.20 There are known sex differences in stroke outcomes21, and education is often a proxy for cognitive reserve which can also impact stroke recovery.22 Finally, the number of years post stroke is included since some people have had longer post-stroke periods before their cognitive status, and this may impact dementia development. The models with any vision impairment or pre-stroke vision impairment as the exposure also included the number of years from retinal form to stroke in the adjustment set. The models with stroke-related vision impairment as the primary exposure were further adjusted for stroke severity, using the continuous NIHSS score.
Sensitivity Analyses:
We ran a sensitivity analysis using categorical coding of the NIHSS score instead of the continuous score. We used the Akaike Information Criterion (AIC) to determine which categorical coding --- (1) ≤ 5 (minor), 6–10 (mild), 11–15 (moderate), and 16+ (severe), (2) ≤ 5 (minor), 6+ (mild to severe), and (3) ≤ 10 (minor to mild), 11+ (moderate to severe) --- resulted in the best model. We also ran sensitivity analyses for all adjusted models, further adjusting for the race-center variable. For all regression models, we report the estimated odds ratios, the 95% Confidence Intervals, and the associated 2-sided p-value.
Post-Hoc Analyses:
We ran a post hoc analysis for our regression models with dementia as the outcome and stroke-related vision impairment as the primary predictor, recoding the stroke-related vision impairment predictor into a three-level categorical variable. The levels were (1) vision impairment with hemianopia, (2) vision impairment without hemianopia, and (3) no vision impairment (reference variable). We were interested in determining if the association between stroke-related vision impairment and dementia is driven by hemianopia, which is often permanent and more disabling, as compared to diplopia or cranial nerve palsies, which often improve and are correctable with intervention.
All hypothesis tests used a critical value of α=0.05. Given the exploratory nature of this study with a restricted sample size, there was no correction for multiple testing as that would severely reduce power to detect a significant effect and cloud potential discovery. Future confirmatory studies are needed for formal hypothesis testing.23–25 All analyses were completed using R 4.1.3.26,27
Results
Our final analytic cohort contained 233 individuals who sustained a probable or definite incident ischemic stroke. We describe their characteristics in Table 1, overall and by post-stroke cognitive status (normal, MCI, and dementia). The mean (standard deviation) age at time of cognitive diagnosis was 81.7 (5.6) years, and at time of incident stroke was 71.8 (8.8) years. Half identified as female sex (50.2% or 117/233), and 34.8% (81/223) as Black race. We report the percentages of diabetes and hypertension at the time of cognitive assessment for descriptive purposes; 42.5% (99/233) of the sample had diabetes, and 74.2% (173/233) had hypertension. In total, 76.8% (179/233) of participants had a thrombotic incident stroke, and 23.2% (54/233) had a non-carotid embolic incident stroke. Overall, 44 participants (18.9% of 233) had stroke-related vision impairment symptoms at the incident stroke, and 54 participants (26.3% of 205) had a pre-stroke vision impairment. Among the 205 stroke survivors with a retinal form prior to incident stroke, eight participants had both a stroke-related vision impairment and a pre-stroke vision impairment.
Table 1:
Clinical Characteristics of the cohort at time of cognitive assessment, stratified by 3 levels of cognitive impairment
| Level of Cognitive Impairment | ||||
|---|---|---|---|---|
| Overall (N=233) | Normal (N=109) | MCI (N=76) | Dementia (N=48) | |
| Age | ||||
| Mean (SD) | 81.7 (5.59) | 81.6 (5.53) | 81.1 (5.88) | 82.7 (5.21) |
| Median [Min, Max] | 82.0 [67.0, 94.0] | 81.0 [68.0, 92.0] | 81.0 [67.0, 94.0] | 83.0 [74.0, 93.0] |
| Sex | ||||
| Female | 117 (50.2%) | 65 (59.6%) | 31 (40.8%) | 21 (43.8%) |
| Male | 116 (49.8%) | 44 (40.4%) | 45 (59.2%) | 27 (56.3%) |
| Race-Center | ||||
| Black, Forsyth, NC | 6 (2.6%) | 2 (1.8%) | 4 (5.3%) | 0 (0%) |
| Black, Jackson, MS | 74 (31.8%) | 35 (32.1%) | 19 (25.0%) | 20 (41.7%) |
| White, Forsyth, NC | 30 (12.9%) | 17 (15.6%) | 11 (14.5%) | 2 (4.2%) |
| White, Minneapolis, MN | 64 (27.5%) | 27 (24.8%) | 25 (32.9%) | 12 (25.0%) |
| White, Washington County, MD | 58 (24.9%) | 28 (25.7%) | 16 (21.1%) | 14 (29.2%) |
| Missing | 1 (0.4%) | 0 (0%) | 1 (1.3%) | 0 (0%) |
| Education | ||||
| Did Not Complete HS | 51 (21.9%) | 20 (18.3%) | 15 (19.7%) | 16 (33.3%) |
| HS, GED, or Vocational School | 97 (41.6%) | 55 (50.5%) | 29 (38.2%) | 13 (27.1%) |
| At least some college | 85 (36.5%) | 34 (31.2%) | 32 (42.1%) | 19 (39.6%) |
| Diabetes | ||||
| No | 108 (46.4%) | 56 (51.4%) | 32 (42.1%) | 20 (41.7%) |
| Yes | 99 (42.5%) | 39 (35.8%) | 32 (42.1%) | 28 (58.3%) |
| Missing | 26 (11.2%) | 14 (12.8%) | 12 (15.8%) | 0 (0%) |
| Hypertension | ||||
| No | 34 (14.6%) | 15 (13.8%) | 9 (11.8%) | 10 (20.8%) |
| Yes | 173 (74.2%) | 80 (73.4%) | 55 (72.4%) | 38 (79.2%) |
| Missing | 26 (11.2%) | 14 (12.8%) | 12 (15.8%) | 0 (0%) |
| # Years since Stroke | ||||
| Mean (SD) | 9.91 (7.10) | 9.85 (7.24) | 10.1 (6.96) | 9.81 (7.13) |
| Median [Min, Max] | 9.00 [0, 30.0] | 8.00 [1.00, 26.0] | 8.50 [0, 30.0] | 10.0 [0, 28.0] |
| Stroke Type | ||||
| Probable/Definite Thrombotic Brain Infarction (TIB) | 179 (76.8%) | 85 (78.0%) | 58 (76.3%) | 36 (75.0%) |
| Probable/Definite Non-carotid Embolic Brain Infarction (EIB) | 54 (23.2%) | 24 (22.0%) | 18 (23.7%) | 12 (25.0%) |
| Stroke Severity | ||||
| NIHSS <=5 | 184 (79.0%) | 86 (78.9%) | 66 (86.8%) | 32 (66.7%) |
| NIHSS 6–10 | 32 (13.7%) | 15 (13.8%) | 7 (9.2%) | 10 (20.8%) |
| NIHSS 11+ | 17 (7.3%) | 8 (7.3%) | 3 (3.9%) | 6 (12.5%) |
| Stroke-Related Vision Impairment | ||||
| No | 189 (81.1%) | 92 (84.4%) | 64 (84.2%) | 33 (68.8%) |
| Yes | 44 (18.9%) | 17 (15.6%) | 12 (15.8%) | 15 (31.3%) |
| Hemianopia | ||||
| No | 201 (86.3%) | 98 (89.9%) | 68 (89.5%) | 35 (72.9%) |
| Yes | 32 (13.7%) | 11 (10.1%) | 8 (10.5%) | 13 (27.1%) |
| Diplopia | ||||
| No | 221 (94.8%) | 102 (93.6%) | 73 (96.1%) | 46 (95.8%) |
| Yes | 12 (5.2%) | 7 (6.4%) | 3 (3.9%) | 2 (4.2%) |
| Cranial Nerve Palsy | ||||
| No | 228 (97.9%) | 108 (99.1%) | 74 (97.4%) | 46 (95.8%) |
| Yes | 5 (2.1%) | 1 (0.9%) | 2 (2.6%) | 2 (4.2%) |
| # Years between Retinal Exam and Stroke | ||||
| Mean (SD) | 9.92 (6.42) | 9.85 (6.81) | 10.3 (5.89) | 9.57 (6.39) |
| Median [Min, Max] | 9.00 [0, 25.0] | 8.00 [0, 25.0] | 10.0 [0, 23.0] | 8.50 [0, 23.0] |
| Missing | 27 (11.6%) | 11 (10.1%) | 10 (13.2%) | 6 (12.5%) |
| Pre-Existing Vision Impairment | ||||
| No | 151 (64.8%) | 70 (64.2%) | 51(67.1%) | 30 (62.5%) |
| Yes | 54 (23.2%) | 28 (25.7%) | 15 (19.7%) | 11 (22.9%) |
| Missing | 28 (12.0%) | 11 (10.1%) | 10 (13.2%) | 7 (14.6%) |
| Glaucoma | ||||
| No | 195 (83.7%) | 92 (84.4%) | 62 (81.6%) | 41 (85.4%) |
| Yes | 11 (4.7%) | 6 (5.5%) | 4 (5.3%) | 1 (2.1%) |
| Missing | 27 (11.6%) | 11 (10.1%) | 10 (13.2%) | 6 (12.5%) |
| Macular Degeneration | ||||
| No | 197 (84.5%) | 93 (85.3%) | 64 (84.2%) | 40 (83.3%) |
| Yes | 9 (3.9%) | 5 (4.6%) | 2 (2.6%) | 2 (4.2%) |
| Missing | 27 (11.6%) | 11 (10.1%) | 10 (13.2%) | 6 (12.5%) |
| Cataracts | ||||
| No | 163 (70.0%) | 76 (69.7%) | 55 (72.4%) | 32 (66.7%) |
| Yes | 42 (18.0%) | 22 (20.2%) | 11 (14.5%) | 9 (18.8%) |
| Missing | 28 (12.0%) | 11 (10.1%) | 10 (13.2%) | 7 (14.6%) |
| Blind in one/both eyes | ||||
| No | 174 (74.7%) | 80 (73.4%) | 58 (76.3%) | 36 (75.0%) |
| Yes | 1 (0.4%) | 1 (0.9%) | 0 (0%) | 0 (0%) |
| Missing | 58 (24.9%) | 28 (25.7%) | 18 (23.7%) | 12 (25.0%) |
Notes:
Diabetes and hypertension were not collected from visit 8, so missing data from those fields are from participants whose cognitive status was taken from visit 8.
Blind in one/both eyes was not collected from the visit 5 retinal form, so the extra missing data from that field is from participants whose retinal exam was taken from visit 5.
Based on the frequencies in Table 1, it appears that stroke severity score is higher in those with dementia (33.3% with the two highest stroke severity categories in the dementia group vs 21.0% overall), and stroke-related vision impairment is also more common in those with dementia (31.3% in the dementia group vs 18.9% overall). The mean continuous stroke severity for those with stroke-related vision impairment was a score of 5.43 and for those without was a score of 3.54; the difference in stroke severity score for these two groups is statistically significant at a critical value of 0.05 (p = 0.02).
When we compared the mean stroke severity scores in the included and excluded groups, we found a significant difference (p < 0.001). Namely, the included stroke survivors had lower stroke severity scores, indicating that a lower score increases the chance of inclusion in the analysis. The test for independence between stroke-related vision impairment and inclusion was not statistically significant (p =0.61), indicating no evidence of dependence between these variables. These comparisons give us insight into the importance of adjusting for stroke severity to control its confounding effect in our regression model.
Regression results are displayed in Table 2. Vision impairment (any versus none) was not statistically significantly associated with prevalent dementia after adjustment for demographics (dementia versus normal or MCI; aOR = 1.47, 95% CI = (0.67, 3.17), p= 0.33) or any cognitive impairment (MCI or dementia versus normal; aOR = 1.05; 95% CI = (0.56, 1.99), p= 0.88). The same holds true when pre-stroke vision impairment was considered as the primary exposure (dementia: aOR = 0.69; 95% CI = (0.25, 1.76), p= 0.94 and cognitive impairment: aOR = 0.67, 95% CI = (0.31, 1.45), p= 0.31). When assessing stroke-related vision impairment as the primary exposure, there was a statistically significant association with prevalent dementia after adjusting for demographics (aOR = 2.32, 95% CI = (1.08, 4.92), p= 0.03), but upon further adjusting for stroke severity, we see an attenuated odds ratio (aOR =2.00, 95% CI = (0.90, 4.32), p=0.08).
Table 2:
Regression output for each combination of primary exposure (i.e., type of vision impairment) and outcome. Each cell displays (1) the estimated odds ratio (OR) (2) the 95% Confidence Interval (CI), and (3) the p-value for the unadjusted and adjusted models.
| Outcome | Any Vision Impairment (n=205) | Pre-stroke Vision Impairment (n=205) | Stroke-related Vision Impairment (n=233) | |||||
|---|---|---|---|---|---|---|---|---|
| Unadjusted | Adjusted* | Unadjusted | Adjusted* | Unadjusted | Adjusted* | Adjusted SS† | ||
| Dementia | OR | 1.64 | 1.47 | 1.03 | 0.69 | 2.45 | 2.32 | 2.00 |
| (ref= MCI or normal) | 95% CI | (0.82, 3.28) | (0.67, 3.17) | (0.46, 2.19) | (0.25, 1.76) | (1.16, 5.03) | (1.08, 4.92) | (0.9, 4.32) |
| p-value | 0.16 | 0.33 | 0.94 | 0.45 | 0.02 | 0.03 | 0.08 | |
| Cognitive Impairment | OR | 1.14 | 1.05 | 0.80 | 0.67 | 1.51 | 1.33 | 1.38 |
| (ref = normal) | 95% CI | (0.65, 2.00) | (0.56, 1.99) | (0.43, 1.50) | (0.31, 1.45) | (0.78, 2.99) | (0.67, 2.70) | (0.68, 2.84) |
| p-value | 0.64 | 0.88 | 0.49 | 0.31 | 0.23 | 0.43 | 0.38 | |
Adjusted for age, sex, education, and number of years since stroke for all models. The models with any vision impairment or pre-stroke vision impairment as the exposure also included number of years from retinal form to stroke in the adjustment set.
Additionally, adjusted for stroke severity (SS), using the continuous National Institutes of Health Stroke Severity (NIHSS) score
Sensitivity Analyses
When we ran a sensitivity analysis using categorical coding of the NIHSS instead of the continuous score, the binary coding with minor strokes versus mild to severe strokes resulted in the best model. The adjusted odds ratio for stroke-related vision impairment was 2.02 (95% CI: 0.92– 4.34; p = 0.08). This is consistent with our results obtained from using the continuous NIHSS score. The results from all categorical classifications are reported in Supplemental Table 1.
Further, we conducted a sensitivity analysis with all adjusted models including the race-center variable as a potential confounder. These analyses had one less participant included due to their inclusion in a non-defined race-center category. All analyses with race-center resulted in very similar results to the primary analyses. These results are quantified in Supplemental Table 2.
Post-Hoc Analyses
For the post-hoc analyses, we split the stroke-related vision impairment variable into a three-level categorical variable to explore if hemianopia might drive the association between stroke-related vision impairment and dementia. 13.7% of participants had vision impairment with hemianopia, with a mean stroke severity score of 6.41. Additionally, 5.2% of participants had vision impairment without hemianopia with a mean stroke severity score of 2.83. The no vision impairment group (81.1%) had a mean stroke severity score of 3.54. Supplemental Table 3 displays the results from the post-hoc regression analyses. There is statistically significant evidence that vision impairment with hemianopia is associated with odds of dementia, even after adjustment for stroke severity score (aOR = 2.43, 95% CI = (1.01,5.70), p = 0.04), while vision impairment without hemianopia is not (aOR = 1.06, 95% CI = (0.15, 4.46), p =0.95).
Discussion
This study is the first to investigate the association between stroke-related vision impairment symptoms and subsequent dementia separately from pre-stroke vision impairments and dementia. This work is key because the literature lacks knowledge about clinical factors that predict dementia post-stroke. We found that only the stroke-related vision impairment was associated with higher adjusted odds for dementia. Not only was there an association, but the estimated 2-fold increase is clinically relevant. This suggests that post-stroke vision impairment is not just a disabling comorbidity, but it may contribute to dementia risk. We cannot make any claims about causality or the exact mechanisms linking vision impairment to dementia. Although further adjusting for stroke severity led to a non-significant p-value, the slightly attenuated estimated odds ratio of 2.0, is still clinically meaningful. This implies that stroke severity may partially confound the relationship between stroke-related vision impairment symptoms and dementia, and there may also still be an independent impact of stroke-related vision impairment on dementia. The relationship between these variables is further complicated by the fact that stroke-severity and stroke-related vision impairment symptoms are measured simultaneously as a result of the same cerebrovascular event, making it challenging to disentangle their individual contributions to the odds of developing dementia. It would be valuable to know whether participants in ARIC had access to occupational therapy services—particularly those targeting stroke-related vision impairment28-- as such services have been shown to support stroke survivors in learning strategies to engage in meaningful activities including their activities of daily living.29
This work also investigated the prevalence of pre-stroke and stroke-related vision impairments among stroke survivors in the ARIC database, and we compared vision impairment prevalence across groups defined by subsequent cognitive outcomes (normal, MCI, dementia). This descriptive information helped us gather valuable data about the health status of these individuals. It was not surprising that the most common stroke-related vision impairment in our cohort was hemianopia or a visual field defect (13.7%-- 32 out of 233) because it has been reported to affect 20–57% of people post-stroke.30 The pre-stroke retinal form revealed the sample as having some different types of vision conditions, most of which were aging types of visual impairments such as cataracts (18%). Considering the age of our participants, at time of cognitive exam was a mean of 81 years, it is not surprising that age-related vision conditions were present.
Finally, we ran a post-hoc investigation to explore whether the association between stroke related vision impairment and dementia is driven by hemianopia. This was a sensible way to further categorize stroke-related vision impairment, since we know that hemianopia, an afferent visual pathway disorder, is a debilitating, typically permanent, post-stroke symptom.31 In contrast, diplopia, for example, which is an efferent ocular motor dysfunction is often treatable or self-resolving.32 The results determined that hemianopia contributes to explaining the link between stroke-related vision impairment and dementia, when adjusting for presence of diplopia and/or cranial nerve palsy. It’s been shown that hemianopia is associated with reduced independence in activities such reading and driving33, as well as a negative impact on quality of life34, but to our knowledge this specific condition has never been linked to dementia before now. However, given the post-hoc nature of the analysis, these results should be viewed as exploratory and warrant further prospective investigation.
Limitations
There are some limitations to this work. First, the sample is not generalizable to other U.S. areas outside of the four ARIC field center locations and may not be generalizable to races other than Black or White. We only included race-center as a descriptive variable rather than a confounder in our primary model in recognition of race as a social construct which would not accurately reflect the mechanisms driving disparities.35,36 However, we ran a sensitivity analyses to check and these results showed that the race-center variable did not confound the relationships in any of our models.
Additionally, some of the data we used was self-report. Also, the sample size for our analytic cohort resulted in limited power. We could adjust for only a small number of variables in the adjusted models to maintain as much parsimony and precision as possible. Given this, we did not adjust for possible comorbidities, such as hypertension and diabetes. These variables are not confounders as measured, since they were taken from the outcome timepoint and therefore have association with vision impairment at stroke. However, being that we expect them to be associated with the outcome of interest, they could have assisted in the precision of our models, if our sample size allowed for inclusion. Another limitation is that we did not have continuous data on pre-stroke vision impairment up until the incident stroke for all participants because the collection of these data only took place at some of the earlier visits. The length of time between the most recent vision assessment and incident stroke occurrence was long, with a median of 9 years. This leads us to believe that there probably are more people with an age-related condition then what we have to report on because age-related conditions are acquired over time. We also have no information on any age-related vision conditions acquired after the stroke, which may add another complication to the relationship between vision impairment and dementia. The number of stroke-related vision impairments in this stroke cohort may also be underrepresented because of the way these data were captured. The data were abstracted from the medical record by a member of the ARIC team. Therefore, we don’t know what assessments were used to assess for the stroke- related vision impairments at the time of stroke and therefore don’t have details such as which type of cranial nerve palsy or more about the diplopia being experienced. This is a problem with secondary data analyses type papers. We also are aware that identifying vision impairment post-stroke in hospitals can still vary among professionals and patients with stroke are not always referred for assessments by vision, therefore underrepresenting impairments.37 Another limitation is that some cognitive tests in ARIC, which were used to determine the mild cognitive impairment status, rely on vision to complete the test. For example, the digit symbol substitution test requires the person match symbols to numbers according to a key located on the top of the page. If the person administering the test did not recognize or determine the stroke survivor had a visual impairment such as a visual field cut, then the results of the assessment could be biased due to the person being unable to see the assessment, not because they had impaired cognition. We recognize that additional steps were taken in attempt to ensure data accuracy. For example, participants who had either a low score on the Mini Mental Status Exam or any of the 5 cognitive domains had either (1) additional in person evaluation, which included the clinical dementia rating scale, (2) telephone interview to get follow up data, were determined to have any hearing loss, were checked to see if there was an ICD-9 dementia diagnosis, and/or had their proxy share more information with the study investigators.19 Finally, even though the stroke-related vision impairment was only captured at the time of stroke, based on clinical experience we expect this to be persistent. However, we realize that we only have one time point of data and therefore this is a limitation.
To address potential selection bias in our study, we conducted several analyses discussed in the Results section. While we confirmed that there was no evidence of dependence between stroke-related vision impairment and inclusion in the analytic cohort, there was a significant difference in the mean stroke severity score when comparing among the included and excluded participants. This indicates that individuals with more severe strokes were more likely to be excluded from our analysis, either due to death or loss to follow-up, potentially introducing selection bias because those participants may have been more likely to have dementia. As such, our results are only generalizable to the population of stroke survivors who lived long enough to receive the cognitive assessment. Future studies may explore strategies to expand this research to broader populations. Ultimately, we included stroke severity as a covariate in our regression model to control for the confounding effect of stroke severity on the relationship between vision impairment and dementia. While this is a crucial step in addressing bias, we acknowledge that some residual bias may remain due to the complexity of the dropout mechanism and potential unmeasured confounders.
Despite these limitations, this dataset is the only existing dataset available that has the necessary data collected (i.e., stroke, dementia, and vision impairment) to begin exploring our study hypotheses.
Conclusions
This study informs the prevalence of many types of vision impairments in a group of stroke survivors and shows the importance of differentiating pre-existing from stroke-related vision impairments. We conclude that it is crucial to assess for both types of vision impairment in this population, given the estimated 2-fold increase in dementia odds among stroke survivors with stroke-related vision impairment. In addition, there is some evidence that the stroke severity score complicates the relationship and may possibly be due to vision impairment indicating more severe strokes.
This work supports the need for additional research, such as databases that use measures such as the Blind MOCA and or non-visually-dependent tests, to ensure any tests for cognition in a population that may also have vision impairment are accurately capturing the impairment. In addition, data collection should be completed in larger, more diverse populations to more confidently determine how specific types of vision impairments in stroke survivors (pre-stroke and stroke-related) are associated with cognitive impairment or dementia. And finally, we recognize this work would benefit from follow up analyses that explore the relationship between vision impairment and time to dementia diagnosis, in addition to this analysis that sticks to a single outcome timepoint, as well as a prospective study to evaluate if the association between stroke-related vision impairment and dementia is driven by hemianopia.
Supplementary Material
Acknowledgements:
Medical editor Katharine O’Moore-Klopf, ELS (East Setauket, NY, USA) provided professional medical editing of this article.
Funding:
This work was supported by National Institutes of Health under Grant K01HD106010 (PI: Kimberly Hreha), the Duke Pepper Older Americans Independence Center (NIA P30AG028716) and the Biostatistics, Epidemiology and Research Design (BERD) Methods Core funded through Grant Award Number UL1TR002553 from the National Center for Advancing Translational Sciences (NCATS), a component of the National Institutes of Health (NIH). We also want to acknowledge the Atherosclerosis Risk in Communities study, which has been funded in whole or in part with Federal funds from the National Heart, Lung, and Blood Institute, National Institutes of Health, Department of Health and Human Services, under Contract nos. (75N92022D00001, 75N92022D00002, 75N92022D00003, 75N92022D00004, 75N92022D00005). The ARIC Neurocognitive Study is supported by U01HL096812, U01HL096814, U01HL096899, U01HL096902, and U01HL096917 from the NIH (NHLBI, NINDS, NIA and NIDCD). The authors thank the staff and participants of the ARIC study for their important contributions.
Footnotes
Ethical Considerations: All ethical considerations were considered and followed.
Consent to Participate: All ARIC participants were consented at the start of their study enrollment.
Consent for Publication: These data were deidentified and participants were informed that publications may result after their participation in this study.
Declaration of conflicting interests: The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.
Data Availability Statement:
Most of the ARIC data is publicly available, but some require formal submission to ARIC with a proposal form and acceptance in order to obtain.
References
- 1.Tsao CW, Aday AW, Almarzooq ZI, et al. Heart Disease and Stroke Statistics-2023 Update: A Report From the American Heart Association. Circulation 2023; 147: e93–e621. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Rowe FJ, Hepworth LR, Howard C, et al. High incidence and prevalence of visual problems after acute stroke: An epidemiology study with implications for service delivery. PLoS ONE 2019; 14: e0213035. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Khan S, Leung E, Jay WM. Stroke and visual rehabilitation. Top Stroke Rehabil 2008; 15: 27–36. [DOI] [PubMed] [Google Scholar]
- 4.Johnson N, Nisar T, Criswell A, et al. Long-Term Disability Outcomes for Patients With Ischemic Stroke Presenting With Visual Deficits. J Neuro-Ophthalmol Off J North Am Neuro-Ophthalmol Soc 2022; 42: 518–523. [Google Scholar]
- 5.Sand KM, Thomassen L, Næss H, et al. Diagnosis and Rehabilitation of Visual Field Defects in Stroke Patients: A Retrospective Audit. Cerebrovasc Dis Extra 2012; 2: 17–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Hreha KP, Fisher SR, Reistetter TA, et al. Use of the ICD-10 vision codes to study ocular conditions in Medicare beneficiaries with stroke. BMC Health Serv Res 2020; 20: 628. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Hreha K, Gupta P, Reistetter T. A case of comorbidities highlighting cerebral stroke, vision impairment, and dementia. SAGE Open Med Case Rep 2020; 8: 2050313X20975246. [Google Scholar]
- 8.Lotery A Correctable visual impairment in stroke rehabilitation patients. Age Ageing 2000; 29: 221–222. [DOI] [PubMed] [Google Scholar]
- 9.Rowe F, Wright D, Brand D, et al. Reading difficulty after stroke: ocular and non ocular causes. Int J Stroke Off J Int Stroke Soc 2011; 6: 404–411. [Google Scholar]
- 10.Hreha KP, Downer B, Ehrlich JR, et al. Association between vision impairment and cognitive decline in older adults with stroke: Health and Retirement Study. Aging Clin Exp Res 2021; 33: 2605–2610. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Patel MD, Coshall C, Rudd AG, et al. Cognitive impairment after stroke: clinical determinants and its associations with long-term stroke outcomes. J Am Geriatr Soc 2002; 50: 700–706. [DOI] [PubMed] [Google Scholar]
- 12.Deal JA, Sharrett AR, Albert M, et al. Retinal signs and risk of incident dementia in the Atherosclerosis Risk in Communities study. Alzheimers Dement J Alzheimers Assoc 2019; 15: 477–486. [Google Scholar]
- 13.Varadaraj V, Munoz B, Deal JA, et al. Association of Vision Impairment With Cognitive Decline Across Multiple Domains in Older Adults. JAMA Netw Open 2021; 4: e2117416. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Lee CS, Krakauer C, Su Y-R, et al. Diabetic Retinopathy and Dementia Association, Beyond Diabetes Severity. Am J Ophthalmol 2023; 249: 90–98. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.The ARIC Investigators. The Atherosclerosis Risk In Community (ARIC) Study: Design and Objectives. Am J Epidemiol 1989; 129: 687–702. [PubMed] [Google Scholar]
- 16.Wright JD, Folsom AR, Coresh J, et al. The ARIC (Atherosclerosis Risk In Communities) Study: JACC Focus Seminar 3/8. J Am Coll Cardiol 2021; 77: 2939–2959. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Rosamond WD, Folsom AR, Chambless LE, et al. Stroke incidence and survival among middle-aged adults: 9-year follow-up of the Atherosclerosis Risk in Communities (ARIC) cohort. Stroke 1999; 30: 736–743. [DOI] [PubMed] [Google Scholar]
- 18.Koton S, Pike JR, Johansen M, et al. Association of Ischemic Stroke Incidence, Severity, and Recurrence With Dementia in the Atherosclerosis Risk in Communities Cohort Study. JAMA Neurol 2022; 79: 271–280. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Knopman DS, Gottesman RF, Sharrett AR, et al. Mild cognitive impairment and dementia prevalence: The Atherosclerosis Risk in Communities Neurocognitive Study. Alzheimers Dement Diagn Assess Dis Monit 2016; 2: 1–11. [Google Scholar]
- 20.Pang J, Xu Y, Liu Q, et al. Trajectories of cognitive function development and predictive factors in disabled middle-aged and older adults. Front Public Health 2024; 12: 1436722. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Reeves MJ, Bushnell CD, Howard G, et al. Sex differences in stroke: epidemiology, clinical presentation, medical care, and outcomes. Lancet Neurol 2008; 7: 915–926. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Mungas D, Gavett B, Fletcher E, et al. Education amplifies brain atrophy effect on cognitive decline: Implications for cognitive reserve. Neurobiol Aging 2018; 68: 142–150. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Hooper R To adjust, or not to adjust, for multiple comparisons. J Clin Epidemiol 2025; 180: 111688. [DOI] [PubMed] [Google Scholar]
- 24.Perneger TV. What’s wrong with Bonferroni adjustments. BMJ 1998; 316: 1236–1238. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Rothman KJ. No adjustments are needed for multiple comparisons. Epidemiol Camb Mass 1990; 1: 43–46. [Google Scholar]
- 26.R: The R Project for Statistical Computing, https://www.r-project.org/ (accessed 3 May 2022).
- 27.Green P, MacLeod CJ. SIMR: an R package for power analysis of generalized linear mixed models by simulation. Methods Ecol Evol 2016; 7: 493–498. [Google Scholar]
- 28.Hreha K, Boisvert CJ, Tagg N, et al. Multidisciplinary Outpatient Vision Care: Evidence from an Eye Stroke Clinic. J Ophthalmol Vis Sci. [Google Scholar]
- 29.Hsu Y-P, Babaeva I Acquired Monocular Vision : Its Effects on Occupational Performance and Occupational Therapy Treatment Interventions. Top Geriatr Rehabil 2022; 38: 231–235. [Google Scholar]
- 30.Alex Pollock, Christine Hazelton, Henderson Clair A., et al. Interventions for Visual Field Defects in Patients With Stroke. Stroke 2012; 43: e37–e38. [DOI] [PubMed] [Google Scholar]
- 31.Rowe FJ. Vision In Stroke cohort: Profile overview of visual impairment. Brain Behav; 7. Epub ahead of print 6 October 2017. DOI: 10.1002/brb3.771. [DOI] [Google Scholar]
- 32.Hanna KL, Hepworth LR, Rowe FJ. The treatment methods for post-stroke visual impairment: A systematic review. Brain Behav 2017; 7: e00682. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Wolter M, Preda S. Visual deficits following stroke: maximizing participation in rehabilitation. Top Stroke Rehabil 2006; 13: 12–21. [DOI] [PubMed] [Google Scholar]
- 34.Choi H-Y, Kim S-J, Kim SY, et al. Comparison of vision-related quality of life in patients with homonymous hemianopia and monocular blindness. Sci Rep 2022; 12: 6558. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Ioannidis JPA, Powe NR, Yancy C. Recalibrating the Use of Race in Medical Research. JAMA 2021; 325: 623–624. [DOI] [PubMed] [Google Scholar]
- 36.Vyas DA, Eisenstein LG, Jones DS. Hidden in Plain Sight — Reconsidering the Use of Race Correction in Clinical Algorithms. N Engl J Med 2020; 383: 874–882. [DOI] [PubMed] [Google Scholar]
- 37.Roberts PS, Wertheimer J, Ouellette D, et al. Feasibility and Clinician Perspectives of the Visual Symptoms and Signs Screen: A Multisite Pilot Study. Top Geriatr Rehabil 2024; 40: 69. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Most of the ARIC data is publicly available, but some require formal submission to ARIC with a proposal form and acceptance in order to obtain.