Methods

The design of the CRIC study has been reported previously. Participants for our RCRIC study were recruited at 6 of the 7 CRIC clinical centers. All 2,605 CRIC participants from these 6 sites were offered participation in our RCRIC investigation. From June 2006 to May 2008, 1936 RCRIC participants were photographed. The institutional review boards of the participating institutions approved the study, and all participants provided written informed consent.

Trained nonophthalmic personnel took the photographs. The participants were seated in a darkened room for 5 minutes to induce physiologic, nonpharmacologic dilation of the pupils. A Canon CR-DGI, non-mydriatic retinal camera (Canon, Tokyo Japan) was used to obtain 45°, digital, color fundus photographs. A set of 2 images, 1 centered on the macula and 1 on the optic disk, was obtained for each eye. A participant was considered eligible for analysis if either the disk or the macula photograph of 1 eye could be evaluated.

The digital fundus photographs were assessed by trained graders and a retinal specialist at the RCRIC Fundus Photograph Reading Center at the University of Pennsylvania. The readers were masked to the participants’ clinical and demographic information. Fundus pathology, including retinopathy (diabetic, hypertensive, or other), and measurement of the diameter of the major retinal vessels were assessed. Because the readers were unaware of the diabetic or hypertensive status of the participants, retinopathy was evaluated without an assumption of cause.

The Atherosclerosis Risk In Communities (ARIC) fundus photographic and the Early Treatment of Diabetic Retinopathy Study (ETDRS) grading protocol were used to grade retinopathy due to diabetes mellitus, systemic hypertension, and other conditions. The Multi-Ethnic Study of Atherosclerosis (MESA) protocol was used for evaluation of macular edema. These grading protocols have been previously validated in diabetic and nondiabetic populations. Evaluations of the digital photographs were done on color-calibrated monitors by a single masked reader using standard protocols with standardized photographic field definitions.

Retinal abnormalities were graded by referring to standard photographs and included microaneurysms, retinal hemorrhages, hemorrhages and/or microaneurysms, retinal hemorrhage type (flame or blot), drusen, hard exudates, cotton wool patches or soft exudates, intraretinal microvascular abnormalities, new vessels on or within 1 disk diameter of the disk or elsewhere, fibrous proliferation, and scars from previous pan retinal photocoagulation.

An Early Treatment of Diabetic Retinopathy Study severity score was assigned for each eye. The score is on an ordinal scale and is not a continuous variable. The scores were classified as normal (<14), very mild, nonproliferative retinopathy (14 to 20), nonproliferative retinopathy (35 to 53), and proliferative retinopathy (>60). The score of the eye with more advanced retinopathy was used as the score of the participant; when only 1 eye was available, that score was used. A total of 116 participants (6%) had photographs that could not be graded for both eyes. Of these, 38 participants had photographs on which no features could be detected. The remaining 78 participants had photographs that were blurry or dark and, although some mild retinopathy features were present, an accurate grade could not be assigned because more advanced and subtle retinopathy features were not discernible.

Grade–regrade reliability was assessed in 200 participants. The weighted κ for participant’s Early Treatment of Diabetic Retinopathy Study score was 0.77 (95% confidence interval 0.67 to 0.88), a value consistent with the reproducibility assessment reported by the Early Treatment of Diabetic Retinopathy Study.

Image processor measurements of the vascular arteriolar and venular calibers were performed according to the Atherosclerosis Risk In Communities protocol, using Interactive Vessel ANalysis software developed at the University of Wisconsin (Madison, Wisconsin). Graders overlaid a grid centered on the disk to establish the distance from the optic nerve. The vessels were measured within an annulus spanning 0.5 to 1 disk diameter from the edge of the disk. The graders identified major arterioles and venules and chose segments most suitable for measurement according to the vessel’s sharpness and straightness. The diameter of ≤6 arterioles and ≤6 venules was averaged.

Identification of clinical cardiovascular outcomes was performed by the CRIC clinical centers. At baseline and then every 6 months, the patients were queried through a questionnaire about possible cardiovascular hospitalizations and outpatient cardiovascular tests and interventions. Cardiovascular events such as myocardial infarction and/or revascularization procedures, stroke, new-onset or worsening congestive heart failure, abnormal heart rhythm, and peripheral arterial disease were identified.

We compared the baseline characteristics of the participants with and without gradable photographs. We used t tests for continuous variables and Fisher’s exact tests for categorical variables. Participant with ungradable photographs were included in a separate retinopathy category, and the analysis did not assume ordering among the categories. We performed 2 separate analyses of the relation between retinopathy and CVD, 1 that included and 1 that excluded the ungradable category.

The relation between fundus features and CVD was assessed using odds ratios and their 95% confidence intervals obtained from univariate and multivariate logistic regression models. The multivariate models were adjusted by the traditional risk factors for CVD, including age, gender, systolic blood pressure, smoking status (never/former/current), diabetes, hypertension, low-density lipoprotein, high-density lipoprotein, triglycerides, hemoglobin A1c, estimated glomerular filtration rate (eGFR), and 24-hour urine protein. Data from the annual visit closest to the fundus photography were used.

For the association between vascular diameter and CVD, the average of the vascular diameters from both eyes was calculated for each participant. When measurements were available for only 1 eye, that measurement was used for that subject. The diameter measurements were divided into quartiles because of the nonmonotonic relations between vascular diameters and CVD. The comparison of CVD among 4 quartiles was assessed by univariate logistic regression models and multivariate logistic regression models adjusted by the traditional risk factors listed.

These analyses were performed for each type-specific CVD, including myocardial infarction and/or previous revascularization, atrial fibrillation or heart arrhythmia, congestive heart failure, peripheral arterial disease, and stroke. The same analysis was performed for any CVD, defined as the presence of any of these type-specific CVDs, excluding atrial fibrillation or heart arrhythmia. For the association between retinopathy or vessel calibers and any CVD, a separate analysis was also performed for diabetic and nondiabetic participants. All data analyses were performed using SAS, version 9.2 (SAS Institute, Cary, North Carolina), and 2-sided p values <0.05 were considered statistically significant.

We used the CRIC definitions of diabetes mellitus and systemic hypertension. The eGFR was calculated using the Modification in Diet in Renal Disease equation.

Results

A total of 1,936 (74%) of 2,605 eligible participants were photographed. Their baseline clinical and demographic characteristics have been described in a previous report. The mean systolic blood pressure and body mass index, prevalence of diabetes, and proportion of women were significantly lower and the mean eGFR was significantly greater in the participants who had photographs, indicating that the photographed group was healthier than the group not photographed.

Of the 1,936 participants with baseline photographs, 1,820 (94.0%) had photographs that were of sufficient quality to allow retinopathy scoring, and 1,599 (82.6%) had photographs that allowed retinal vessel caliber measurements in ≥1 eye. Compared to the 1,820 participants with gradable photographs, the 116 participants with ungradable photographs were older, had a significantly lower average eGFR, and had greater systolic blood pressure, urine protein, hemoglobin A1c, and low-density lipoprotein. They also had a greater prevalence of diabetes and hypertension ( Table 1 ).

Retinopathy was present in 456 (49%) of 925 participants with diabetes and 115 (11%) of 1,011 participants without diabetes (p <0.001). Of the 1,714 participants with hypertension, 548 (32%) had retinopathy, and of the 221 participants without hypertension, 22 (10%) had retinopathy (p <0.001). Also, 182 participants had neither diabetes mellitus nor hypertension, and 4 (2%) had mild retinopathy.

Retinopathy was associated with an increased prevalence of any CVD on univariate analysis (p <0.001; Table 2 ). This significant relation persisted after adjustment for other CVD risk factors (multivariate analysis, p <0.01). However, the risk of any CVD did not increase as the severity of retinopathy increased. The participants with ungradable photographs had the greatest risk of any CVD ( Table 2 ). Of those with diabetes, a similar relation was observed (p <0.01; Table 2 ), and mild retinopathy was associated with the greatest risk of any CVD.

We used a stepwise selection method for logistic regression analysis to identify the retinopathy features independently associated with any CVD. Only fibrous proliferation (p = 0.03), a feature included in the Early Treatment of Diabetic Retinopathy Study grading, and arteriovenous abnormalities (p = 0.02), a feature commonly associated with systemic hypertension, were statistically significantly associated with an increased risk of any CVD.

We investigated whether an interaction was present between eGFR and retinopathy or proteinuria and retinopathy with any CVD. We found no statistically significant modulating effect of eGFR or proteinuria on this relation (data not shown).

We found no associations between the arterial diameter and the presence of any CVD. A significant association was detected, however, between the venous diameter and any CVD (univariate linear trend analysis, p = 0.02; Table 3 ). A larger venous diameter was associated with an increased risk of any CVD, and this association remained after adjustment for CVD risk factors (linear trend, p = 0.01). For the participants with diabetes, this relation was not statistically significant (p = 0.10; Table 3 ).

The strongest associations between retinopathy and specific CVDs were observed for myocardial infarction (MI) and/or revascularization and stroke ( Table 4 ). A significant association was detected between the retinopathy level and a history of MI and/or previous revascularization (univariate analysis, p <0.001; multivariate analysis, p = 0.02). Mild nonproliferative retinopathy showed the strongest relation with MI and/or revascularization (odds ratio 1.93, 95% confidence interval 1.27 to 2.94; Table 4 ). The association between a greater retinal venous diameter and MI was not significant on univariate analysis (p = 0.13) but was on multivariate analysis (p = 0.03, Table 5 ).

Stay updated, free articles. Join our Telegram channel

Join