Abstract
Purpose
Retinal vascular changes could serve as an early, easily visible indicator of cerebrovascular health. Prior research on retinal vessel traits and stroke has predominantly focused on diameter, while vessel area and tortuosity remain underinvestigated and have yielded inconsistent results. Our goal was to examine longitudinal associations between retinal vessel traits and incident stroke.
Design
Prospective cohort study.
Participants
Baseline and 6-year follow-up data from the Canadian Longitudinal Study on Aging Comprehensive Cohort (n = 30 097) were used.
Methods
A deep learning algorithm called Quantitative Analysis of Retinal Vessel Topology and Size extracted data from retinal images including arteriolar and venular diameter, area, and tortuosity. Participants were asked if a doctor had previously diagnosed them with stroke. Logistic regression was used to adjust for demographic, lifestyle, and clinical factors. Interactions with sex were also assessed.
Main Outcome Measures
Stroke.
Results
Having wider arterioles at baseline was associated with lower odds of stroke development over 6 years (odds ratio [OR] = 0.84, 95% confidence interval [CI]: 0.74‑0.95). Having larger venular area was associated with a higher odds of developing stroke (OR = 1.71, 95% CI: 1.11‑2.63), whereas larger arteriolar area showed the inverse association (OR = 0.62, 95% CI: 0.44‑0.86). Statistically significant interactions were found between sex and both arteriolar and venular tortuosity (P < 0.05).
Conclusions
This study confirms that retinal vessel area and stroke are associated. Vessel area and arteriolar width are associated with the development of stroke, which may help to develop stroke prediction models in the future.
Financial Disclosure(s)
Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.
Keywords: CLSA, Stroke, Arteriole, Venule, Sex.
Stroke occurs when a cerebral infarction or hemorrhage disrupts blood flow and subsequently causes acute damage to the brain tissue.1 It is the second leading cause of death globally and the third leading contributor to disability, often resulting in cognitive or motor impairments.2 Most strokes stem from systemic vascular etiologies such as small vessel disease, large artery disease, or cardio-embolism and well-established risk factors including diabetes and hypertension.1
The retina shares anatomical, physiological, and embryological characteristics with the brain. Therefore, retinal vascular changes could serve as an early, easily visible indicator of cerebrovascular health.3,4 Systematic reviews have shown that a variety of quantitative retinal vessel measures have been associated with clinical and subclinical stroke, its recurrence, and stroke-related mortality.4,5 Specifically, narrower retinal arteriolar diameter and wider venular diameter have been linked to an increased risk of stroke across several prospective studies.6, 7, 8, 9 A meta-analysis of 6 cohort studies comprised of 20 798 subjects also confirmed that wider venular diameter was associated with incident stroke.10 Overall, these findings highlight the potential utility of retinal vessel diameter in assessing cerebrovascular risk.
While prior research on retinal vessel traits and stroke has predominantly focused on diameter, other parameters that reflect the overall retinal vessel pattern such as vessel area and tortuosity (curvature) remain underinvestigated and have yielded inconsistent results. For example, Cheung et al found null associations between arteriolar and venular tortuosity and incident stroke.9 However, in a study of individuals with type 2 diabetes, Sandoval-Garcia et al reported a positive association between arteriolar tortuosity and incident stroke.11 To the best of our knowledge, only one published study has investigated retinal vessel area and stroke reporting an inverse association.12
Given the few longitudinal studies that have examined retinal vessel traits like area and tortuosity, we aimed to examine the longitudinal associations between retinal vessel diameter, area, and tortuosity with incident stroke in a large cohort of participants from the Canadian Longitudinal Study on Aging (CLSA). We will also investigate differences by sex because they have previously been reported.6,10
Methods
Study Population and Design
Longitudinal analyses were conducted using data from the CLSA Comprehensive Cohort.13 This study includes 30 097 Canadian adults aged 45 to 85 years, recruited through random sampling of provincial health care databases and random digit dialing of landline phones. Baseline data were gathered between 2012 and 2015, with follow-up 2 data collected 6 years later, from 2018 to 2021. Data collection involved in-home interviews and physical assessments at CLSA data collection sites in cities across Canada, including Victoria, Vancouver, Surrey, Calgary, Winnipeg, Hamilton, Ottawa, Montreal, Sherbrooke, Halifax, and St. John's.
Participants aged 45 to 85 years were eligible if they lived in the community at recruitment, spoke English or French, and resided within 25 to 50 km of a data collection site. Exclusion criteria included cognitive impairment at recruitment, full-time membership in the Canadian Armed Forces, living on a federal First Nations reserve or settlement, residing in a long-term care facility at recruitment, or not being a permanent resident or citizen. The research adhered to the Declaration of Helsinki. All participants provided written informed consent. Research Ethics Board approval was acquired for all CLSA sites in July 2010. The University of Ottawa Office of Research Ethics and Integrity gave approval for the present analysis in January 2023 (H-01-23-8331).
Data Collection
Retinal Images
Nonmydriatic images of the retina were taken in each eye at baseline using a Topcon TRC-NW8 fundus camera with an attached Nikon D90 camera. Images were centered on the macula and had a 45° field-of-view. Most images were saved in JPEG format with a resolution of 4288 × 2848 pixels. Images from one site were saved with a resolution of 4928 × 3264 pixels and were resized to 4288 × 2848 pixels to be consistent with the other images.
Retinal Vessel Traits
Retinal images were processed using a deep learning algorithm called Quantitative Analysis of Retinal Vessel Topology and Size (QUARTZ).14,15 Quantitative Analysis of Retinal Vessel Topology and Size analyzes vessels in fundus images across the inner retinal layer. These vessels, supplied by the ophthalmic artery via the central retinal artery, include arterioles, precapillaries and postcapillaries, and venules. Vessel width was measured using zero-crossings of the second derivative and tortuosity (curvature) was measured using a subdivided chord length method. These methods have been previously well documented and validated.15 In the CLSA data, QUARTZ can determine inadequate image quality with a sensitivity of 91% and a specificity of 100%. Quantitative Analysis of Retinal Vessel Topology and Size can also distinguish between an arteriole and a venule with 96% accuracy using a probability cutoff of 0.8.14 Thousands of measures of diameter, area, and tortuosity from across the entire retinal image were obtained. The mean diameter and tortuosity of each arteriole and venule were then summarized weighted by the vessel segment length for each image. Area was calculated from the average vessel width multiplied by length, summed across vessel segments. Diameter is measured in pixels, while area is presented in kilopixels.2 Tortuosity is a unitless measure.
Stroke, Demographic, Health, and Lifestyle Data
During the interviewer-led survey at baseline and follow-up, participants were asked “if a doctor had ever told them they had a stroke or CVA (cerebrovascular accident).” They were also asked “if a doctor had ever told them they had experienced a ministroke or TIA (transient ischemic attack)?” Mortality data were collected by proxy interviews. Primary and contributing causes of death were obtained. We defined a person as dying from stroke if either their primary or contributing cause of death was reported as stroke.
Participants were also asked about physician diagnoses of diabetes (none, type 1, type 2, or neither), heart disease, or high blood pressure (hypertension). Blood pressure was also measured using the VSM BpTRU monitor (Medaval). Six readings were acquired, and the average of the second through sixth measurements was used for analysis. High blood pressure was defined as either reporting a diagnosis of it or currently having elevated blood pressure (systolic blood pressure ≥130 mmHg or diastolic blood pressure ≥80 mmHg). Individuals without high blood pressure did not report a doctor diagnosis and had normal blood pressure (systolic blood pressure <130 mmHg and diastolic blood pressure <80 mmHg).
Weight was measured using a 140-10 Healthweigh Digital Physician Scale, while height was measured using a Seca 213 stadiometer. Body mass index was calculated as weight (kg)/height squared (m2) and classified according to the World Health Organization categories (underweight <18.5 kg/m2, normal weight 18.5–24.9 kg/m2, overweight 25.0–29.9 kg/m2, and obese ≥30.0 kg/m2).
Demographic information, including age, sex, income, and race/ethnicity, was collected at baseline through an interviewer-administered questionnaire. Household income was assessed by asking participants “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 grouped into 5 categories: >$100K, $100 to $50K, $50K to $20K, <$20K, and Refused.
Smoking status was determined using answers from 2 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)?” Current smokers were defined as those who had smoked ≥100 cigarettes and currently smoke daily or occasionally. Former smokers were those who reported smoking ≥100 cigarettes in life but had not smoked in the last 30 days. Nonsmokers were defined as individuals who had smoked <100 cigarettes in their lifetime. Similarly, frequency of alcohol consumption was assessed by asking participants “About how often during the past 12 months did you drink alcohol?” Responses were categorized into never, occasionally (0–3 times a month), weekly (between 1 and 5 times a week), or daily (≥6 times a week).
Blood Tests
Nonfasting blood samples were collected via venipuncture during the data collection site visit and analyzed at the respective site laboratory. Several soluble and cellular biomarkers were measured in serum including total cholesterol and the concentration of high-sensitivity C-reactive protein using the Cobas 8000 modular analyzer (Roche Diagnostics).
Statistical Analysis
Our primary outcome was nonfatal stroke. Incident nonfatal stroke was defined as the presence of a nonfatal stroke or cerebrovascular accident (CVA) at follow-up 2 in those without a stroke or CVA at baseline. Our secondary outcome was fatal or nonfatal stroke. Incident fatal or nonfatal stroke was defined as the report of a nonfatal stroke or CVA at follow-up 2 in those without a stroke or CVA at baseline or the presence of death by stroke as either a primary or contributing cause by the date of their follow-up 2 examination. Those with and without incident stroke were compared. Logistic regression models were used to assess the associations between each retinal vessel trait (diameter, tortuosity, and area) and the 6-year development of stroke. Each outcome’s model was run separately, containing both the arteriolar and venular trait as there is evidence to suggest a confounding effect between the vessels.3,16 Models were adjusted for age, sex, race/ethnicity, income, hypertension, body mass index, history of heart disease, total cholesterol, high-sensitivity C-reactive protein, diabetes, alcohol, smoking, and province. Effect modification by sex was assessed to determine whether the relationship between vessel traits and outcomes varied across subgroups. This was done by including interaction terms (e.g., sex × vessel diameter) in the statistical models and conducting stratified analyses. The complex survey design (analytic weights and strata) of the CLSA was accounted for in all analyses. Regression models were run using right eye data only because it was not possible to include both eyes and account for both weights and strata in the analyses. A P value of 0.05 was considered statistically significant. Multiple comparisons were not taken into account.17 Stata Version 16 was used for all analyses.
Results
Of the 30 097 CLSA participants, 26 076 (87%) had ≥1 acceptable QUARTZ image quality score, as presented in the flow chart in Figure. Those with unacceptable QUARTZ scores were older, had lower incomes, were more likely to have high blood pressure, were more likely to smoke, were more likely to have heart disease, and were more likely to have diabetes (Table S1, available at www.ophthalmologyscience.org). Of those with acceptable QUARTZ image quality scores who returned to follow-up and provided stroke data (n = 21 547), characteristics of CLSA participants by 6-year incident stroke status are compared in Table 1. There were 264 people who reported a nonfatal physician diagnosis of stroke at the 6-year follow-up examination (1.2%). Those who developed stroke were older, more likely to have a lower income, and be current smokers (P < 0.05). Those with incident stroke were also more likely to report health conditions such as high blood pressure, heart disease, and diabetes and to have higher C-reactive protein levels (P < 0.05). They also had wider venular diameter and smaller arteriolar area (P < 0.05).
Figure 1.
Flowchart of CLSA participants in the analysis. CLSA = Canadian Longitudinal Study on Aging.
Table 1.
Characteristics of CLSA Participants by 6-Yr Incidence of Nonfatal Stroke
| Develop Stroke (n = 264) Mean (SD) or % | Do Not Develop Stroke (n = 21 283) Mean (SD) or % | P Value | |
|---|---|---|---|
| Age, yrs | 65.4 (10.8) | 58.1 (9.2) | <0.001 |
| Female sex | 47.1% | 53% | 0.174 |
| Race/ethnicity | |||
| White | 94.8% | 94.3% | 0.809 |
| Non-White | 5.2% | 5.7% | |
| Income | |||
| >$100 000 | 22.6% | 39.3% | <0.001 |
| $50 000-$100 000 | 36.2% | 33.0% | |
| $20 000-$50 000 | 28.8% | 18.0% | |
| <$20 000 | 8.6% | 4.4% | |
| Missing | 3.7% | 5.3% | |
| BMI | |||
| Underweight | 0.9% | 0.5% | 0.101 |
| Normal | 20.7% | 28.9% | |
| Overweight | 47.8% | 39.2% | |
| Obese | 30.6% | 31.49% | |
| High blood pressure | |||
| No | 25.5% | 49.8% | <0.001 |
| Yes | 74.5% | 50.2% | |
| Smoking | |||
| Never | 36.5% | 46.0% | 0.017 |
| Former | 47.2% | 44.4% | |
| Current | 16.3% | 9.6% | |
| Alcohol | |||
| Never | 2.2% | 2.2% | 0.150 |
| Occasional | 48.2% | 40.6% | |
| Weekly | 34.6% | 43.8% | |
| Daily | 15.0% | 13.4% | |
| Heart disease | |||
| No | 74.5% | 91.9% | <0.001 |
| Yes | 25.5% | 8.1% | |
| Diabetes | |||
| None | 75.9% | 84.2% | 0.011 |
| Any type | 24.1% | 15.8% | |
| hs-CRP, mg/L | 3.0 (4.4) | 2.5 (4.8) | 0.039 |
| Cholesterol, mmol/L | 5.0 (1.4) | 5.2 (1.1) | 0.169 |
| Arteriolar diameter, pixels | 15.5 (1.6) | 15.6 (1.6) | 0.594 |
| Venular diameter, pixels | 17.8 (2.0) | 17.5 (1.8) | 0.035 |
| Arteriolar tortuosity∗ | 1.3 (0.4) | 1.2 (0.4) | 0.343 |
| Venular tortuosity∗ | 1.2 (0.3) | 1.2 (0.2) | 0.178 |
| Arteriolar area, kilopixels2 | 217.1 (44.3) | 233.9 (48.4) | <0.001 |
| Venular area, kilopixels2 | 254.8 (57.5) | 260.5 (49.0) | 0.215 |
BMI = body mass index; CLSA = Canadian Longitudinal Study on Aging; hs-CRP = high-sensitivity C-reactive protein; SD = standard deviation.
Tortuosity measures were multiplied by 1000 and the natural log taken to achieve normality.
The overall 6-year incidence of nonfatal stroke was 1.2% with the highest incidence of 3.8% in the oldest age group from 75 to 85 years at baseline. The pairwise correlation coefficients for arteriolar diameter, area, and tortuosity are all <0.2. The same is true for venular diameter and tortuosity. The longitudinal associations between retinal vessel traits and the 6-year development of nonfatal stroke are shown in Table 2. Those with wider arterioles at baseline had a lower odds of developing stroke (odds ratio [OR] = 0.84, 95% confidence interval [CI]: 0.74‑0.95 per 1 pixel change in diameter). Venular width was not associated with the development of stroke (OR = 1.07, 95% CI: 0.98‑1.18, P = 0.151). Both arteriolar and venular areas were associated with the development of nonfatal stroke but in opposite directions. Having a larger arteriolar area was associated with a reduced odds of incident stroke (OR = 0.62, 95% CI: 0.44‑0.86), while having larger venular area was associated with a higher odds of developing a stroke (OR = 1.71, 95% CI: 1.11‑2.63). Arteriolar tortuosity was not associated with 6-year development of nonfatal stroke in the overall sample (P > 0.05). However, sex differences were found such that arteriolar tortuosity was not associated with the development of nonfatal stroke in men (OR = 0.74, 95% CI: 0.44‑1.25), but it was in women (OR = 2.17, 95% CI: 1.02‑4.61) (interaction P value < 0.05). The same was true of venular tortuosity such that it was not associated with the 6-year development of nonfatal stroke in the overall sample (OR = 0.98, 95% CI: 0.39‑2.44). However, statistically significant interaction by sex was found such that the ORs were in opposite directions in men (OR = 1.99, 95% CI: 0.61‑6.47) versus women (OR = 0.38, 95% CI: 0.11‑1.35) (interaction P value < 0.05).
Table 2.
Longitudinal Associations between Retinal Vessel Traits and the 6-Yr Development of Nonfatal Stroke
| Model∗ | Vessel Trait† | 6-Yr Odds of Stroke, Men and Women, n = 18 473, OR (95% CI) | 6-Yr Odds of Stroke, Men Only, n = 9048, OR (95% CI) | 6-Yr Odds of Stroke, Women Only, n = 9425, OR (95% CI) |
|---|---|---|---|---|
| 1 | Arteriolar diameter | 0.84 (0.74–0.95) | 0.91 (0.77–1.08) | 0.78 (0.65–0.93) |
| Venular diameter | 1.07 (0.98–1.18) | 0.99 (0.86–1,13) | 1.14 (1.00–1.29) | |
| 2 | Arteriolar tortuosity | 1.20 (0.74–1.93) | 0.74 (0.44–1.25)‡ | 2.17 (1.02–4.61)‡ |
| Venular tortuosity | 0.98 (0.39–2.44) | 1.99 (0.61–6.47)‡ | 0.38 (0.11–1.35)‡ | |
| 3 | Arteriolar area | 0.62 (0.44–0.86) | 0.58 (0.37–0.92) | 0.62 (0.39–0.98) |
| Venular area | 1.71 (1.11–2.63) | 1.72 (0.88–3.37) | 1.75 (1.08–2.86) |
CI = confidence interval; OR = odds ratio.
Each model runs separately containing both the arteriolar and venular trait; models also adjusted for age, sex, income, ethnicity, hypertension, body mass index, diabetes, heart disease, smoking, alcohol, total cholesterol, high-sensitivity c-reactive protein, and province.
Per 1 pixel increase in diameter, 1 unit increase in tortuosity, or 100 kilopixel2 increase for area.
Statistically significant interaction by sex (P < 0.05).
The overall 6-year incidence of fatal or nonfatal stroke was 1.3%. The longitudinal associations between retinal vessel traits and the 6-year development of fatal or nonfatal stroke are shown in Table 3. Arteriolar diameter remained associated with the 6-year odds of developing a fatal or nonfatal stroke (OR = 0.84, 95% CI: 0.74‑0.94). Arteriolar and venular area also remained associated with the 6-year development of fatal or nonfatal stroke (OR = 0.66, 95% CI: 0.47‑0.92 and OR = 1.64, 95% CI: 1.07‑2.51, respectively). However, the interactions between sex and tortuosity were no longer statistically significant for either arteriolar tortuosity (P = 0.280) or venular tortuosity (P = 0.083), although the ORs remained very heterogeneous.
Table 3.
Longitudinal Associations between Retinal Vessel Traits and the 6-Yr Development of Nonfatal or Fatal Stroke
| Model∗ | Vessel Trait† | 6-Yr Odds of Stroke Men and Women, n = 18 490, OR (95% CI) |
6-Yr Odds of Stroke Men Only, n = 9060, OR (95% CI) |
6-Yr Odds of Stroke Women Only, n = 9430, OR (95% CI) |
|---|---|---|---|---|
| 1 | Arteriolar diameter | 0.84 (0.74–0.94) | 0.89 (0.76–1.05) | 0.79 (0.66–0.93) |
| Venular diameter | 1.07 (0.98–1.17) | 1.00 (0.88–1,14) | 1.12 (0.99–1.27) | |
| 2 | Arteriolar tortuosity | 1.36 (0.86–2.16) | 0.95 (0.56–1.62) | 2.21 (1.06–4.63) |
| Venular tortuosity | 1.09 (0.45–2.60) | 2.20 (0.72–6.73) | 0.42 (0.13–1.43) | |
| 3 | Arteriolar area | 0.66 (0.47–0.92) | 0.63 (0.40–1.01) | 0.65 (0.41–1.01) |
| Venular area | 1.64 (1.07–2.51) | 1.61 (0.84–3.08) | 1.75 (1.09–2.83) |
CI = confidence interval; OR = odds ratio.
Each model runs separately containing both the arteriolar and venular trait; models also adjusted for age, sex, income, ethnicity, hypertension, body mass index, diabetes, heart disease, smoking, alcohol, total cholesterol, high-sensitivity c-reactive protein, and province.
Per 1 pixel increase in diameter, 1 unit increase in tortuosity, or 100 kilopixel2 increase for area.
In supplemental analyses, we investigated whether retinal vessel traits were associated with the 6-year incidence of a transient ischemic attack, reported by 1.9% of the sample. There were no statistically significant associations between retinal vessel diameter, area, or tortuosity with incident transient ischemic attack (Table S2, available at www.ophthalmologyscience.org). In addition, to assure that area provides new information apart from diameter, we adjusted area models for diameter. All inferences remained the same.
Discussion
Our work examines the relationship between retinal vessel diameter, area, and tortuosity with the incidence of stroke. We observed that the venular area was associated with a higher odds of developing stroke, and both arteriolar area and diameter were inversely associated with the development of stroke. We also found strong sex differences in the association between arteriolar and venular tortuosity and stroke incidence.
Retinal vessel area has not often been examined.12,18, 19, 20 Area combines information about retinal vessel width and length into one measure. We previously reported that blood pressure, a strong risk factor for stroke,2 was associated with retinal vessel area, diameter, and tortuosity.20 Only one study, to our knowledge, has previously examined retinal vessel area and stroke.12 Yusufu et al reported that greater arteriolar vessel area density, as measured with the Retina-Based Microvascular Health Assessment System, was associated with a lower adjusted risk of incident stroke (hazard ratio = 0.86, 95% CI: 0.80‑0.93). Our findings are consistent with theirs. We found that greater arteriolar area was associated with a lower 6-year odds of developing stroke, while a greater venular area was associated with a higher 6-year odds of developing stroke. These associations were very strong and persisted after adjustment for a variety of potential confounding factors. Vessel area and vessel diameter are related to one another but they are not very correlated with correlation coefficients <0.2. Although arteriolar diameter and arteriolar area share similar results with stroke incidence in our data, the same cannot be said for venular diameter and venular area. Indeed, when we adjust area for diameter, it remains statistically significant. More research is needed on retinal vessel area to understand its biological meaning.
We also observed that wider arteriolar diameter was inversely associated with 6-year development of stroke such that a 1-pixel increase in diameter was associated with a 16% lower odds of developing a stroke. This is a moderately sized OR similar in size to what is seen for eating a diet high in fruit and vegetables and its association with stroke.21 Our results are consistent with longitudinal studies6, 7, 8,22 and 3 meta-analyses.23, 24, 25 Previous studies have established links between retinal arteriolar narrowing and known risk factors for stroke including hypertension, reduced large-artery compliance, increased carotid artery stiffness, and reduced aortic distensibility.3,20,26,27 Narrower arterioles may also reflect atherosclerotic changes in the vasculature;25 thromboembolism from atherosclerotic lesions can cause stroke.1 Furthermore, narrowing of retinal blood vessels may indicate cerebral artery stenosis (narrowing), which causes cerebral ischemia and ultimately stroke.28
We were unable to confirm the association between wider venular diameter and the higher risk of stroke reported in meta-analyses and several population-based studies.10,25 The OR was elevated at 1.07, but it was not statistically significant (P = 0.151). Quantitative Analysis of Retinal Vessel Topology and Size takes measures across the entire retinal image, whereas some other analysis software only takes measures near the optic disc. If venular widening is more pronounced near the optic disc in people at a risk of stroke, this may explain the discrepancy in our results. Wider retinal venules have been linked to hypoxia, endothelial dysfunction, and inflammation,29 which play an essential role in the development of stroke.30 We did find that having a larger venular area was associated with the 6-year development of stroke. Also, we previously reported using CLSA data that higher C-reactive protein levels, a marker of inflammation, were associated with venular area.20
There were sex differences in the association between tortuosity and the incidence of stroke. Having more tortuous (curved) arterioles was associated with the incidence of stroke but only in women (OR = 2.17, 95% CI: 1.02‑4.61). This was a statistically significant interaction. By contrast, the ORs for venular tortuosity were in opposite directions for men and women such that there was a higher risk of stroke in men (OR = 1.99, 95% CI: 0.61‑6.47) and a lower risk of stroke in women (OR = 0.38, 95% CI: 011‑1.35). Owen et al also observed using QUARTZ that more tortuous arterioles were associated with prevalent stroke in the EPIC Norfolk study but they did not report sex differences.31 Sandoval-Garcia et al also reported that more tortuous arterioles were associated with long-term risk of stroke in those with type 2 diabetes11 but they did not find sex differences. In contrast, in the Singapore Malay Eye Study, arteriolar tortuosity was not associated with the risk of stroke.9 Although no studies have reported sex differences between retinal vessel tortuosity and stroke, studies examining vessel diameter and stroke have reported sex differences. McGeechan et al10 observed that wider central retinal vein equivalent was associated with incident stroke in men only, but not women, after stratification. Conversely, Seidelmann et al6 assessed stroke as part of a composite outcome among other incident cardiovascular events and showed that narrower arterioles and wider venules were associated with cardiovascular events in women but not in men. Other studies have not reported statistically significant interactions between sex and retinal vessel diameter.10,32,33 Sex differences could indicate different biological mechanisms of stroke by sex34 or these could be chance findings.
The sex differences that we found are aligned with evidence suggesting a more widespread contribution of microvascular dysfunction in relation to cardiovascular disease in women compared to that in men.35 Women generally have more diffuse atherosclerosis and arteriolar dysfunction35 and stroke risk factors often have stronger associations with stroke in women than in men.34 Women also generally tend to have more tortuous arteries, potentially due to hormonal factors.36 Therefore, sex-specific differences in retinal microvascular geometric patterns may contribute to stroke risk. Biologically, our results are supported by evidence showing that more tortuous vessels are common in those with major risk factors for stroke, including hypertension and atherosclerosis, potentially due to increased mechanical stress.36,37 Highly tortuous vessels can restrict or occlude blood flow and may also encourage platelet activation and the formation of blood clots, contributing to stroke development.37,38 Thus, as a reflection of the cerebral and systemic microvasculature, more tortuous arterioles in the retina may represent underlying pathogenic processes related to stroke. Further research to clarify the role of retinal vessel tortuosity on the development of stroke and whether there are any sex differences is warranted.
The strengths of this study are the large sample size, the population-based sample, and the comprehensive data in the form of survey responses and physical assessments including retinal images. A limitation of this work is that stroke was assessed by self-report of a doctor’s diagnosis and we did not have any information on the type of stroke. The prevalence of stroke in the CLSA is 1.5%, which is almost identical to what was self-reported in the Canadian Community Health Survey (1.3%).39 The self-report of stroke has been reported to have a sensitivity of 80% and a specificity of 99%.40 Therefore, some people who actually had a stroke are likely to be misclassified as having reported not having a stroke. However, we would not expect the misclassification of stroke to vary by the level of retinal vessel characteristics. Another limitation is that our population is predominantly of a White, European ancestry background. Results may differ in populations with substantially different ethnic backgrounds. We did test 3 retinal vessel traits in their relationship with stroke and did not adjust for multiple comparisons. Therefore, there is a possibility that our findings are due to chance. Finally, we did not have data on axial length so were unable to convert pixels to microns. However, a rough estimate can be provided of how many microns a pixel equals in the CLSA. The average vertical optic disc diameter in the CLSA was 418 pixels, while the average vertical optic disc diameter in White people in a paper by Quigley et al was 1820 microns.41 Therefore, 1 pixel equals approximately 4.35 microns.
Conclusions
This study provides longitudinal evidence into the relationship between retinal vessel characteristics, including understudied traits such as tortuosity and area, with stroke incidence. More research on the timing of retinal vessel changes and their relationship to the development of stroke would be helpful for use in risk prediction models.12,42,43
Acknowledgments
This research was made possible using the data/biospecimens collected by the Canadian Longitudinal Study on Aging (CLSA). This research has been conducted using the CLSA Baseline Comprehensive Dataset version 7.0, Follow-up 2 Comprehensive Dataset version 1.0, and Baseline Retinal Scans under Application Number 2209017. The CLSA is led by Drs Parminder Raina, Christina Wolfson, and Susan Kirkland. The time and commitment of the participants to the CLSA study platform is gratefully acknowledged, without whom this research would not be possible.
Manuscript no. XOPS-D-25-00725.
Footnotes
Supplemental material available at www.ophthalmologyscience.org.
Disclosure(s):
All authors have completed and submitted the ICMJE disclosures form.
The authors made the following disclosures:
E.F.: All support for the present manuscript ‑ Canadian Institutes of Health Research.
D.M.: Grants or contracts from any entity ‑ Regenxbio; Participation on a Data Safety Monitoring Board or Advisory Board ‑ Bayer.
M.H.R.G.: All support for the present manuscript ‑ Canadian Institutes of Health Research.
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, as well as the following provinces, Newfoundland, Nova Scotia, Quebec, Ontario, Manitoba, Alberta, and British Columbia. The funders had no role in the design, analysis, or the interpretation of results. The opinions expressed in this manuscript are the author’s own and do not reflect the views of the Canadian Longitudinal Study on Aging. Data are available from the Canadian Longitudinal Study on Aging (www.clsa-elcv.ca) for researchers who meet the criteria for access to deidentified CLSA data. This research project was further funded by CIHR operating grant PJT-183690 led by Dr Ellen Freeman.
Support for Open Access publication was provided by the Canadian Institutes of Health Research (CIHR).
HUMAN SUBJECTS: Human subjects were included in this study. The research adhered to the Declaration of Helsinki. All participants provided written informed consent. Research Ethics Board approval was acquired for all CLSA sites in July 2010. The University of Ottawa Office of Research Ethics and Integrity gave approval for the present analysis in January 2023 (H-01-23-8331).
No animal subjects were included in this study.
Author Contributions:
Conception and design: Freeman.
Analysis and interpretation: O’Neil, Welikala, Barman, Owen, Rudnicka, Rakesh, Roy-Gagnon, Maberley, Kergoat, Paquet, Freeman.
Data collection: Welikala, Barman, Rakesh, Roy-Gagnon, Freeman.
Overall responsibility: O’Neil, Welikala, Barman, Owen, Rudnicka, Rakesh, Roy-Gagnon, Maberley, Kergoat, Paquet, Freeman.
Obtained funding: Freeman.
Supplementary Data
References
- 1.Hilkens N.A., Casolla B., Leung T.W., de Leeuw F.E. Stroke. Lancet. 2024;403:2820–2836. doi: 10.1016/S0140-6736(24)00642-1. [DOI] [PubMed] [Google Scholar]
- 2.Collaborators GBDS Global, regional, and national burden of stroke and its risk factors, 1990-2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet Neurol. 2021;20:795–820. doi: 10.1016/S1474-4422(21)00252-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Sun C., Wang J.J., Mackey D.A., Wong T.Y. Retinal vascular caliber: systemic, environmental, and genetic associations. Surv Ophthalmol. 2009;54:74–95. doi: 10.1016/j.survophthal.2008.10.003. [DOI] [PubMed] [Google Scholar]
- 4.Rim T.H., Teo A.W.J., Yang H.H.S., et al. Retinal vascular signs and cerebrovascular diseases. J Neuroophthalmol. 2020;40:44–59. doi: 10.1097/WNO.0000000000000888. [DOI] [PubMed] [Google Scholar]
- 5.Dumitrascu O.M., Demaerschalk B.M., Valencia Sanchez C., et al. Retinal microvascular abnormalities as surrogate markers of cerebrovascular ischemic disease: a meta-analysis. J Stroke Cerebrovasc Dis. 2018;27:1960–1968. doi: 10.1016/j.jstrokecerebrovasdis.2018.02.041. [DOI] [PubMed] [Google Scholar]
- 6.Seidelmann S.B., Claggett B., Bravo P.E., et al. Retinal vessel calibers in predicting long-term cardiovascular outcomes: the atherosclerosis risk in communities Study. Circulation. 2016;134:1328–1338. doi: 10.1161/CIRCULATIONAHA.116.023425. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Kawasaki R., Xie J., Cheung N., et al. Retinal microvascular signs and risk of stroke: the Multi-Ethnic Study of Atherosclerosis (MESA) Stroke. 2012;43:3245–3251. doi: 10.1161/STROKEAHA.112.673335. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Wieberdink R.G., Ikram M.K., Koudstaal P.J., et al. Retinal vascular calibers and the risk of intracerebral hemorrhage and cerebral infarction: the Rotterdam Study. Stroke. 2010;41:2757–2761. doi: 10.1161/STROKEAHA.110.599084. [DOI] [PubMed] [Google Scholar]
- 9.Cheung C.Y., Tay W.T., Ikram M.K., et al. Retinal microvascular changes and risk of stroke: the Singapore Malay Eye Study. Stroke. 2013;44:2402–2408. doi: 10.1161/STROKEAHA.113.001738. [DOI] [PubMed] [Google Scholar]
- 10.McGeechan K., Liew G., Macaskill P., et al. Prediction of incident stroke events based on retinal vessel caliber: a systematic review and individual-participant meta-analysis. Am J Epidemiol. 2009;170:1323–1332. doi: 10.1093/aje/kwp306. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Sandoval-Garcia E., McLachlan S., Price A.H., et al. Retinal arteriolar tortuosity and fractal dimension are associated with long-term cardiovascular outcomes in people with type 2 diabetes. Diabetologia. 2021;64:2215–2227. doi: 10.1007/s00125-021-05499-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Yusufu M., Friedman D.S., Kang M., et al. Retinal vascular fingerprints predict incident stroke: findings from the UK Biobank cohort study. Heart. 2025;111:306–313. doi: 10.1136/heartjnl-2024-324705. [DOI] [PubMed] [Google Scholar]
- 13.Raina P., Wolfson C., Kirkland S., et al. Cohort profile: the Canadian longitudinal Study on aging (CLSA) Int J Epidemiol. 2019;48:1752–1753j. doi: 10.1093/ije/dyz173. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Welikala R., Fajtl J., Johnson G., et al. Institute of Electrical and Electronics Engineers (IEEE)-EMBS International Conference on Biomedical and Health Informatics (BHI’24); Nov 10-13, Houston, USA. 2024. Artificial intelligence-enabled retinal vasculometry at Scale utilizing the UK Biobank, CLSA, and NEL DESP datasets. [Google Scholar]
- 15.Welikala R.A., Fraz M.M., Habib M., et al. Institute of Electrical and Electronics Engineers (IEEE) Seventh International Conference on Image Processing Theory, Tools and Applications, Montreal, Canada. 2017. Automated quantification of retinal vessel morphometry in the UK biobank cohort. [Google Scholar]
- 16.Liew G., Wong T.Y., Mitchell P., Wang J.J. Are narrower or wider retinal venules associated with incident hypertension? Hypertension. 2006;48:e10. doi: 10.1161/01.HYP.0000231652.97173.4c. author reply e11. [DOI] [PubMed] [Google Scholar]
- 17.Rothman K.J. No adjustments are needed for multiple comparisons. Epidemiology. 1990;1:43–46. [PubMed] [Google Scholar]
- 18.Fukutsu K., Saito M., Noda K., et al. A deep learning Architecture for vascular area measurement in fundus images. Ophthalmol Sci. 2021;1 doi: 10.1016/j.xops.2021.100004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Maloca P.M., Feu-Basilio S., Schottenhamml J., et al. Reference database of total retinal vessel surface area derived from volume-rendered optical coherence tomography angiography. Sci Rep. 2022;12:3695. doi: 10.1038/s41598-022-07439-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.O'Neil A., Welikala R.A., Barman S., et al. Factors associated with retinal vessel traits in the Canadian longitudinal Study on aging. Invest Ophthalmol Vis Sci. 2025;66:13. doi: 10.1167/iovs.66.3.13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Oude Griep L.M., Verschuren W.M., Kromhout D., et al. Colors of fruit and vegetables and 10-year incidence of stroke. Stroke. 2011;42:3190–3195. doi: 10.1161/STROKEAHA.110.611152. [DOI] [PubMed] [Google Scholar]
- 22.Yatsuya H., Folsom A.R., Wong T.Y., et al. Retinal microvascular abnormalities and risk of lacunar stroke: atherosclerosis Risk in Communities Study. Stroke. 2010;41:1349–1355. doi: 10.1161/STROKEAHA.110.580837. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Girach Z., Sarian A., Maldonado-Garcia C., et al. Retinal imaging for the assessment of stroke risk: a systematic review. J Neurol. 2024;271:2285–2297. doi: 10.1007/s00415-023-12171-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Wu H.Q., Wu H., Shi L.L., et al. The association between retinal vasculature changes and stroke: a literature review and meta-analysis. Int J Ophthalmol. 2017;10:109–114. doi: 10.18240/ijo.2017.01.18. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Guo S., Yin S., Tse G., et al. Association between caliber of retinal vessels and cardiovascular disease: a systematic review and meta-analysis. Curr Atheroscler Rep. 2020;22:16. doi: 10.1007/s11883-020-0834-2. [DOI] [PubMed] [Google Scholar]
- 26.Cheung N., Islam F.M., Jacobs D.R., Jr., et al. Arterial compliance and retinal vascular caliber in cerebrovascular disease. Ann Neurol. 2007;62:618–624. doi: 10.1002/ana.21236. [DOI] [PubMed] [Google Scholar]
- 27.De Silva D.A., Woon F.P., Manzano J.J., et al. The relationship between aortic stiffness and changes in retinal microvessels among Asian ischemic stroke patients. J Hum Hypertens. 2012;26:716–722. doi: 10.1038/jhh.2011.88. [DOI] [PubMed] [Google Scholar]
- 28.Gao Y., Xu L., He N., et al. A narrative review of retinal vascular parameters and the applications (Part II): diagnosis in stroke. Brain Circ. 2023;9:129–134. doi: 10.4103/bc.bc_9_23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Liu M., Lovern C., Lycett K., et al. The association between markers of inflammation and retinal microvascular parameters: a systematic review and meta-analysis. Atherosclerosis. 2021;336:12–22. doi: 10.1016/j.atherosclerosis.2021.09.025. [DOI] [PubMed] [Google Scholar]
- 30.Kelly P.J., Lemmens R., Tsivgoulis G. Inflammation and stroke risk: a new target for prevention. Stroke. 2021;52:2697–2706. doi: 10.1161/STROKEAHA.121.034388. [DOI] [PubMed] [Google Scholar]
- 31.Owen C.G., Rudnicka A.R., Welikala R.A., et al. Retinal vasculometry associations with cardiometabolic risk factors in the European prospective investigation of cancer-norfolk Study. Ophthalmology. 2019;126:96–106. doi: 10.1016/j.ophtha.2018.07.022. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Wong T.Y., Kamineni A., Klein R., et al. Quantitative retinal venular caliber and risk of cardiovascular disease in older persons: the cardiovascular health study. Arch Intern Med. 2006;166:2388–2394. doi: 10.1001/archinte.166.21.2388. [DOI] [PubMed] [Google Scholar]
- 33.Longstreth W., Jr., Larsen E.K., Klein R., et al. Associations between findings on cranial magnetic resonance imaging and retinal photography in the elderly: the Cardiovascular Health Study. Am J Epidemiol. 2007;165:78–84. doi: 10.1093/aje/kwj350. [DOI] [PubMed] [Google Scholar]
- 34.Rexrode K.M., Madsen T.E., Yu A.Y.X., et al. The impact of sex and gender on stroke. Circ Res. 2022;130(4):512–528. doi: 10.1161/CIRCRESAHA.121.319915. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Hanssen H., Streese L., Vilser W. Retinal vessel diameters and function in cardiovascular risk and disease. Prog Retin Eye Res. 2022;91 doi: 10.1016/j.preteyeres.2022.101095. [DOI] [PubMed] [Google Scholar]
- 36.Ciurica S., Lopez-Sublet M., Loeys B.L., et al. Arterial tortuosity. Hypertension. 2019;73:951–960. doi: 10.1161/HYPERTENSIONAHA.118.11647. [DOI] [PubMed] [Google Scholar]
- 37.Han H.C. Twisted blood vessels: symptoms, etiology and biomechanical mechanisms. J Vasc Res. 2012;49:185–197. doi: 10.1159/000335123. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Chesnutt J.K., Han H.C. Tortuosity triggers platelet activation and thrombus formation in microvessels. J Biomech Eng. 2011;133 doi: 10.1115/1.4005478. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Singh S.S., Zhong S., Rogers K.A., et al. Prevalence of dementia, heart disease and stroke in community-dwelling adults in Canada, 2016-2021: opportunities for joint prevention. Arch Public Health. 2023;81:158. doi: 10.1186/s13690-023-01171-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Engstad T., Bonaa K.H., Viitanen M. Validity of self-reported stroke: the Tromso Study. Stroke. 2000;31:1602–1607. doi: 10.1161/01.str.31.7.1602. [DOI] [PubMed] [Google Scholar]
- 41.Quigley H.A., Brown A.E., Morrison J.D., Drance S.M. The size and shape of the optic disc in normal human eyes. Arch Ophthalmol. 1990;108:51–57. doi: 10.1001/archopht.1990.01070030057028. [DOI] [PubMed] [Google Scholar]
- 42.Hong C., Pencina M.J., Wojdyla D.M., et al. Predictive accuracy of stroke risk prediction models across Black and White race, sex, and age groups. JAMA. 2023;329:306–317. doi: 10.1001/jama.2022.24683. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Rudnicka A.R., Welikala R., Barman S., et al. Artificial intelligence-enabled retinal vasculometry for prediction of circulatory mortality, myocardial infarction and stroke. Br J Ophthalmol. 2022;106:1722–1729. doi: 10.1136/bjo-2022-321842. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.