Retinopathy as a Risk Factor for Nonembolic Stroke in Diabetic Subjects
Background and Purpose Age, hypertension, and the presence of diabetic nephropathy and coronary and peripheral vascular disease are established risk factors for stroke in diabetic subjects, but the association of retinopathy with stroke risk is less clear.
Methods In this nested case-control study, we examined the association of retinopathy with risk of nonembolic ischemic stroke in diabetes. In a cohort of 2124 diabetic persons identified at multiphasic health checkups during the period from 1979 through 1985, 56 suffered a nonembolic ischemic stroke during follow-up, which extended through 1991. For each case subject, one diabetic control subject matched by sex and year of birth was selected from the same cohort of diabetics. Medical records were reviewed to gather information on risk factors for stroke. Information on risk factors was retrieved for 52 of the 56 cases.
Results The estimated relative risk of stroke in diabetic subjects with retinopathy was 2.8 (95% confidence interval, 1.2 to 6.9). After adjustment for age, sex, smoking, use of insulin, average systolic blood pressure, and average random glucose, the estimated relative risk was 4.0 (95% confidence interval, 1.0 to 14.5). The relative risk of stroke in diabetic subjects with retinopathy remained elevated after exclusion of those with complications other than retinopathy.
Conclusions Stroke prevention in diabetic persons is a challenge. Diabetic subjects with retinopathy appear to be a group at particularly high risk of ischemic stroke. Development of preventive interventions might focus on this group.
The risk of ischemic stroke is elevated in diabetic persons.1 2 3 Age, hypertension, and the presence of diabetic nephropathy and coronary and peripheral vascular disease are established risk factors for stroke in those with diabetes.4 5 6 7 The relationship between stroke and other factors, including retinopathy, is less clear.1 2 3
This nested case-control study sought to examine risk factors for nonembolic ischemic stroke in a defined cohort of diabetic subjects who received multiphasic health checkups (MHC) in two Northern California Kaiser Permanente facilities. The original purpose of the study was to determine the relationship between stroke risk and an index of glycemic control. We found that retinopathy was an independent risk factor for nonembolic ischemic stroke. This relationship seems not to have been previously reported, and it is therefore the focus of this report.
This study was reviewed and approved by the institutional review board of Northern California Kaiser Permanente Medical Care Program.
Subjects were identified from the cohort of diabetic persons who had an MHC in either of two Northern California Kaiser Permanente Medical Care Program facilities in the years 1979 through 1985. Persons who stated on a self-administered questionnaire that they were being treated by a doctor for diabetes at the time of the MHC were considered to be diabetic. Potential cases were identified by review of hospitalizations in Kaiser Permanente hospitals in the years 1979 through 1991 for ischemic cerebrovascular disease (International Classification of Diseases, 9th Revision, Clinical Modification codes 433 to 436). For each case subject, one diabetic control subject was selected who was matched on exact year of birth and sex, had an MHC in the same year as the case subject, and was still a member of the Kaiser Permanente Health Plan in the year the case subject was hospitalized. The index date was the date of the hospitalization for the case subject and her or his associated control. The period from the MHC to the hospitalization of the case subject was the index period.
Medical records of case and control subjects were abstracted to verify the diagnosis of diabetes, to establish the diagnosis of first-ever nonembolic ischemic stroke according to predetermined criteria, and to collect information on risk factors for stroke.
Case and control subjects were considered to have verified cases of diabetes if, during the index period, they were treated for diabetes with diet, oral hypoglycemic drugs, or insulin; they had one or more fasting blood glucose measurements >140 mg/dL; or they had two or more postload or random blood glucose measurements >200 mg/dL.
Stroke was defined as the rapid onset of a new neurological deficit lasting more than 24 hours when the deficit was not known to have a non–central nervous system cause. Excluded were case subjects with a prior history of cerebrovascular disease, those with blood seen on computed tomographic or magnetic resonance imaging scan or blood in lumbar puncture, and those with arterial dissection or venous thrombosis. Cases found in persons with atrial fibrillation, rheumatic heart disease, or another possible cardiac source for embolus were considered probably of cardioembolic origin and were also excluded.
A patient was considered to have neuropathy, nephropathy, retinopathy, peripheral vascular disease, or history of myocardial infarction if there was any mention of these diabetic complications in the medical chart during the index period. All blood glucose values from the beginning of the index period up to, but not including, the index date were recorded. For case subjects, measurements made after the onset of neurological symptoms were not recorded. An index of glucose control was constructed by taking the average of the first of these recorded measurements for each quarter year during the index period. This method for creating an index blood glucose control is an attempt to take into account the tendency of physicians to do more frequent glucose measurements in patients whose diabetes is out of control. Separate averages were computed for fasting and for postload or random blood glucose measurements. All blood pressure measurements during the index period were also recorded, and an index of blood pressure was computed by taking the average of the first blood pressure measurement for each quarter.
Thirty-three percent of subjects had missing data for random and/or average blood glucose. Data on smoking status were also missing for two subjects. Because missing data on any variable for either member of a matched pair caused the pair to be dropped, a matched multivariate analysis resulted in the loss of almost half of all pairs. For this reason, the multivariate analysis was performed unmatched. The results of both matched and unmatched univariate analyses are presented.
For categorical variables, the program statxact8 was used to derive exact unmatched maximum likelihood estimates of the odds ratio. The 95% confidence intervals (CIs) presented are mid-probability corrected intervals. Unconditional logistic regression was used in the multivariate analysis and to derive univariate estimates of the odds ratio for continuous variables. Conditional logistic regression was used in the matched analysis. The term “estimated relative risk” (RR) is used as a synonym for odds ratio in the remainder of the article and in the tables.
The diabetic cohort comprised 2124 persons. Among these diabetic subjects, there were 114 hospitalizations with a discharge code suggesting ischemic stroke during the follow-up period. Three records were miscoded, records of two potential cases could not be located, and one subject did not have verified diabetes. Twenty-eight cases of ischemic stroke were found in persons with a history of cerebrovascular disease. Twenty cases were considered embolic stroke, and 4 cases were ischemic neurological events lasting less than 24 hours. This left 56 confirmed cases of nonembolic ischemic stroke. Information on risk factors could be retrieved for 52 of the 56 case/control pairs. The mean age of case and control subjects on the index date was 67 years. Of case and control subjects combined, 52% were male.
Table 1⇓ shows the characteristics of the 52 case and control subjects at the time of the baseline MHC and on the index date and the unadjusted matched and unmatched estimates of the RR of stroke, according to each characteristic. Conclusions based on the matched and unmatched analyses were the same. Smokers at baseline had a lower risk of stroke. Baseline levels of total cholesterol and fasting glucose were associated with a slightly increased risk of stroke. Among variables assessed for the index period, hypertension and insulin treatment were associated with an increased RR of stroke in diabetic subjects. There was also an association of average random and average fasting glucose levels with an increased RR of stroke, although the associations were weak. The RR of stroke was elevated in relation to each of the diabetic complications. Among the diabetic complications, the highest RR was observed for myocardial infarction. However, over 40% of stroke case subjects had retinopathy, and the RR of stroke was significantly increased only in relation to retinopathy.
Table 2⇓ shows the estimated RR of ischemic stroke in diabetic subjects with retinopathy adjusting for various confounding variables and for variables that are potentially a part of the causal pathway. The RR for stroke remains elevated in diabetic subjects with retinopathy after adjusting for age and sex and for other variables associated with increased risk of stroke. These analyses show that retinopathy is an independent risk factor for stroke.
Table 3⇓ shows the estimated RR of ischemic stroke in subgroups of diabetic subjects defined by the absence of other diabetic complications, adjusting first for age and sex and then for these variables as well as average glucose values and average systolic blood pressure. The risk of stroke was increased in diabetic subjects with retinopathy in each of these subgroups, although the CIs are wide in some subgroups, perhaps in part because of the small number of subjects.
We found that retinopathy was a risk factor for nonembolic ischemic stroke in diabetic persons. Other complications of diabetes, including nephropathy, myocardial infarction, and peripheral vascular disease, are also associated with an increased risk of stroke in diabetic subjects,4 5 6 7 and our finding is not very surprising. Diabetes causes both macrovascular and microvascular changes. Microvascular changes occur throughout the body, with the most significant changes occurring in the kidney, eye, and brain. Autopsy studies show that the increased risk of ischemic stroke in diabetes is a result of occlusion of small paramedian-penetrating arteries and not carotid disease.9 10 11 The vascular lesions in the small vessels of the brain are proliferative lesions,9 11 which are also found in retinopathy. It is logical that retinopathy should predict ischemic stroke if both conditions are due to diffuse microvascular changes.
Our data suggest that retinopathy is a risk factor for nonembolic stroke independent of smoking, blood pressure, and other diabetic complications. The poor quality of our measure of glycemic control does not allow a firm conclusion about independence of the association from this factor. The data suggest that glycemic control may be important.
The study has some limitations. Most important, despite the relatively large size of the diabetic population from which cases were derived, the number of cases was small. We had problems with missing data. Missing data have unpredictable effects on multivariate estimates on risk.
The information on presence or absence of retinopathy was based on notations in medical records. These notations are of uncertain accuracy. It was not possible to distinguish diabetic and hypertensive retinopathy. We believe that retinopathy noted in medical records is probably more severe than what might be observed on systematic examination of the retina. Thus, our results may pertain only to severe retinopathy. Studies that distinguish hypertensive from diabetic retinopathy would be needed to confirm our findings.
Information on duration of diabetes was so incomplete that we could not consider it in the analysis. We also could not distinguish type I and type II diabetic subjects. We were not able to reliably separate lacunar and nonlacunar stroke based on chart review. The number of cases of each type would probably not be large enough for subgroup analysis even if we could separate lacunar from nonlacunar stroke.
Prevention of stroke in diabetes is a challenge. In the Early Treatment Diabetic Retinopathy Study,12 650 mg/d aspirin in diabetic subjects with retinopathy decreased the risk of fatal and nonfatal stroke in type I diabetes (RR, 0.60; 95% CI, 0.18 to 2.04) but slightly increased the risk in mixed diabetes (RR, 1.11; 95% CI, 0.59 to 2.07) and type II diabetes (RR, 1.37; 95% CI, 0.77 to 2.43). In the Veterans Administration Cooperative Study,13 which included diabetic patients who had undergone a lower extremity amputation, stroke was significantly less frequent in patients treated with 75 mg/d dipyridamole plus 325 mg/d aspirin than in those who received placebo. In the European Stroke Prevention Study,14 the benefit of 990 mg/d aspirin plus 225 mg/d dipyridamole in preventing stroke in diabetics with cerebrovascular disease was less than in nondiabetics and failed to achieve statistical significance. In the Ticlopidine Aspirin Stroke Study,15 1300 mg/d aspirin was less effective in preventing stroke in diabetic than in nondiabetic subjects, although 500 mg/d ticlopidine was equally effective.
The data are sparse, but they suggest the hypothesis that low-dose aspirin alone is insufficient to prevent stroke in diabetic subjects, especially those who already have manifestations of microvascular disease. Further elucidation of the mechanisms of ischemic stroke in diabetes and determination of the efficacy of antiplatelet and fibrinolytic agents in preventing stroke in diabetic subjects are priority areas for further research. Because this group appears to be at particularly high risk of stroke, preventive interventions might focus on diabetic persons with retinopathy.
The Student Research Program of the University of California at San Francisco School of Medicine provided support for this study. We thank Wansu Chen for help with the statistical analysis.
Reprint requests to Dr Petitti, Southern California Kaiser Permanente Medical Care Program, 393 E Walnut St, Pasadena, CA 91188.
- Received November 4, 1994.
- Revision received January 19, 1995.
- Accepted January 19, 1995.
- Copyright © 1995 by American Heart Association
Bell DS. Stroke in the diabetic patient. Diabetes Care. 1994;17:213-219.
Abbott RD, Donahue RP, MacMahon SW, Reed DM, Yunko K. Diabetes and the risk of stroke: the Honolulu Heart Program. JAMA. 1987;949-952.
STATXACT, Version 2. Cambridge, Mass: Cytel Software; 1991.
Aronson SM. Intercranial vascular lesions in patients with diabetes mellitus. J Neuropathol Exp Neurol. 1973;23:1983-1996.
Bell ET. A postmortem study of vascular disease in diabetes. Arch Pathol. 1952;53:444-455.
Alex M, Baron EK, Goldenberg S, Blumenthal HT. An autopsy study of cerebrovascular accident in diabetes mellitus. Circulation. 1962;25:663-673.
Sivenius J, Laakso M, Riekkinen P, Smets P, Lowenthal A. European Stroke Prevention Study: effectiveness of antiplatelet therapy in diabetic patients in secondary prevention of stroke. Stroke. 1992;23:851-854.