Stroke Recurrence in Diabetics
Does Control of Blood Glucose Reduce Risk?
Background and Purpose Patients with diabetes are at increased risk of stroke. Risk might be reduced if blood glucose level were controlled.
Methods In a population-based study, we enrolled 621 patients within a month of an initial ischemic stroke and followed them regularly twice annually; 198 were diabetic. We monitored blood glucose level in 142 (72%) using glycosylated hemoglobin (HbAlc). Recurrent stroke frequency was determined by history, examination, and medical records. Cox proportional hazards models were used to examine the relationship between risk of recurrent stroke and HbAlc level. The models included interaction with time-dependent HbAlc level and history of diabetes, selected medical comorbidities, age, and sex. HbAlc level was analyzed as both a continuous and a dichotomous variable (ie, controlled versus uncontrolled); “controlled” was defined with different cut points.
Results All but 17 patients (12%) whose blood glucose was monitored were well controlled (HbAlc <8%). HbAlc level was not associated with increased risk of stroke recurrence (hazard ratio [HR], 0.87 per 1% increment in HbAlc; 95% confidence interval [CI], 0.623 to 1.219), nor was there a trend toward increased risk of recurrent stroke as the cut point defining “controlled” increased: with HbAlc at <6%, the HR for the uncontrolled group was 0.51 (95% CI, 0.176 to 1.503); at <7%, it was 0.43 (95% CI, 0.089 to 1.923); and at <8%, it was also 0.43 (95% CI, 0.057 to 3.317).
Conclusions Among diabetic patients with an initial stroke, no association between HbAlc level over time and risk of stroke recurrence was found. However, most patients in this cohort were well controlled, and any adverse effect of poor control could not be adequately tested.
Patients with diabetes have an increased risk of stroke compared with nondiabetic patients.1 2 3 4 5 It has been conclusively demonstrated in the Diabetes Control and Complications Trial that intensive treatment regimens in patients with IDDM reduce the risk of development (or progression) of retinopathy, nephropathy, and neuropathy,6 but there have been no randomized clinical trials comparable to the Diabetes Control and Complications Trial studying the effect of good control of blood glucose on these complications or on stroke in patients with NIDDM. Strokes occur with greater frequency in older individuals, among whom NIDDM is more common than IDDM. Thus, it remains uncertain whether the risk of stroke can be reduced by controlling blood glucose. In a previous report based on the prospective study of the Lehigh Valley Stroke Cohort, we found that individuals classified as diabetic by history had an increased mortality but no increased risk of another stroke compared with patients without diabetes even after age, sex, and comorbidities were taken into account.7 8 In that analysis, we did not consider level of glucose control subsequent to the initial stroke. In this report we examine the effect of level of glucose control on stroke recurrence rate in patients with diabetes.
Subjects and Methods
The Lehigh Valley Stroke Cohort consists of individuals admitted during 1987 to 1989 with their first stroke to one of the eight acute care hospitals in the Lehigh Valley, a region in northeastern Pennsylvania. The patients were followed for up to 4 years (mean, 24 months). The region had approximately 600 000 inhabitants, of whom 95% were white (non-Hispanic), 1.7% were black, and 3.3% were Hispanic. The medical facilities in this region are well developed so that inhabitants tended to seek medical help locally. This allowed ascertainment of virtually all strokes that occurred.
Informed consent was obtained for all study participants. We examined the effect on stroke recurrence over time of five medical conditions: diabetes mellitus, hypertension, myocardial infarction, cardiac arrhythmia, and transient ischemic attack. Each of these medical conditions was defined in our earlier report.9 The presence of these conditions was determined at the time of enrollment as well as during each follow-up visit. In this report diabetes is the main focus.
The medical status of patients with respect to diabetes mellitus (as well as the other four medical conditions) was determined systematically through a structured interview at enrollment as well as through review of the medical records of the patient. Trained nurses conducted the interviews with the patient and often with members of the family. One of us (G.F.), a neurologist, also reviewed the medical records and other data collected by the study nurses.
A patient was classified as diabetic if a history of diabetes mellitus was reported by the patient or a relative or diabetes was listed in the medical records. In addition, use of insulin, oral hypoglycemic agents, or diet to reduce blood glucose level constituted evidence for diabetes mellitus in this study. These patients had HbAlc assays during follow-up to monitor glucose control. HbAlc is a widely accepted and clinically useful measure of glucose control over the previous several months.10 The initial blood glucose on admission to the hospital for the stroke was deliberately not used to diagnose diabetes so as to avoid labeling as diabetic those patients with reactive (or “stress”) hyperglycemia. However, patients without a history of or treatment for diabetes, but with elevated blood glucose during hospitalization, had an HbAlc assay on follow-up visits. If the HbAlc was >6.2%, the upper limit of normal in our central laboratory (Damon, Inc, Trenton, NJ) or the patient was being treated for diabetes, the patient was classified as diabetic. Any questions regarding the patient’s medical status were resolved in conference by the two study neurologists (M.A. and G.F.).
The criteria of the Stroke Data Bank with minor modifications9 were used to diagnose stroke type. In addition, all patients had at least one CT or MRI scan of the brain. Differences between the investigators as to stroke type were also resolved in conference by the study neurologists.
Patients were followed regularly after discharge from the hospital by trained nurses who were part of the research team. The visits occurred at home or in a care facility first at 4 months and then approximately every 6 months for up to 4 years (mean follow-up, 24 months). At each visit, the surviving patients were queried as to medical status including treatment, if any, for diabetes and symptoms compatible with diabetes, such as polydipsia, polyuria, distal limb numbness and tingling, and loss of vision. If such symptoms were elicited, an HbAlc assay was performed. If >6.2%, the patients’ physician was notified and the patient was classified as diabetic in the analysis. For previously diagnosed patients with diabetes and for those whose HbAlc assay was >6.2% on a follow-up visit assay, HbAlc level was monitored on each follow-up visit. If new neurological symptoms or signs were found by the nurse, the patient was examined by one of the study neurologists to determine whether a new stroke had occurred. The nurse also inquired about any neurological deficits that had developed in the interim since the previous visit and performed a brief neurological examination. Criteria for our diagnosing stroke recurrence have been detailed previously9 and included a history, neurological examination by the study neurologist, and review of medical records.
Differences in demographic characteristics between patients with and without diabetes were analyzed with the χ2 or the t test. The risk of stroke recurrence in relation to the presence or absence of a history of diabetes mellitus was analyzed by the Kaplan-Meier method, and statistical significance was determined by the log-rank test.
In evaluating the possible association between glucose control and risk of stroke recurrence, we used several time-dependent Cox proportional hazards models that included history of diabetes, age, sex, comorbidities, and level of HbAlc. The Cox analysis in which HbAlc was treated as a continuous variable was based only on patients with diabetes, and therefore history of diabetes was not included in that analysis. In other Cox analyses in which HbAlc was treated as a dichotomous variable (ie, controlled or noncontrolled), nondiabetic patients whose HbAlc had presumably been normal were included among the “controlled” patients. By including nondiabetic patients as controls, the power of these analyses was increased, although the HbAlc in the patients without diabetes was not actually measured during follow-up. The cut points for “controlled” that were tested were 6%, 7%, and 8%.
We also estimated the stability of the HbAlc over time by calculating the median SD of HbAlc over all assays for each of the 142 patients with diabetes on whom data were available. (A large SD, of course, implies fluctuation in blood glucose level over time, while a small SD implies a stable blood glucose level.)
Approximately 4000 individuals diagnosed or suspected of having had an acute stroke who were hospitalized during the accrual period were screened, and 692 were enrolled. Reasons for failure to qualify for enrollment among those screened were determined on a consecutive sample of 200 individuals and included prior stroke (34.5%), nonstroke (21.0%), transient ischemic attack (13.5%), non–Lehigh Valley resident (14.5%), refusal (8.5%), more than 1 month from onset (5.5%), early death (1.5%), and no CT/MRI (1%). Thus, most patients screened and excluded did not meet the primary enrollment criteria of having had an initial stroke and being residents of the Lehigh Valley. Consequently, those excluded introduced minimal selection bias. All patients included were enrolled within 1 month of onset of their first stroke (93% within 2 weeks). Eight patients were subsequently excluded because additional information showed that they were ineligible. Another 22 died in the hospital and were excluded because, obviously, they could not be followed up. Only 5 of those who died in the hospital had a history of diabetes. Of the remaining 662, 41 had a cerebral hemorrhage. Since a previous study had shown no association between diabetes and risk of cerebral hemorrhage,5 we based this report only on the 621 patients with an ischemic stroke.
Among the 621 patients with an initial acute ischemic stroke, a total of 198 were classified as having diabetes mellitus, 189 by history at enrollment and 9 after enrollment, who were newly discovered during follow-up by the patients’ physician or by us based on symptoms of diabetes and an HbAlc >6.2%. Of the 198 diabetics, 142 (72%) had an HbAlc assay on at least one follow-up visit. If a patient with symptoms suggestive of diabetes had an HbAlc ≤6.2% and was not being treated for diabetes, the patient was not considered diabetic.
Age, sex, baseline glucose, stroke risk factors, initial stroke type, and study outcome for patients with and without diabetes are shown in Table 1⇓. The distribution of stroke types was similar in the two groups. However, the frequency of stroke risk factors showed some differences. A history of hypertension and electrocardiographic evidence of myocardial infarction at enrollment were significantly more frequent among patients with diabetes. Atrial fibrillation was significantly more common in patients without diabetes, but other types of cardiac arrhythmia occurred with similar frequency in the two groups. Transient ischemic attacks were more frequent among patients with diabetes, with the difference approaching statistical significance.
Blood glucose levels based on values recorded in the medical records on or about the third day of hospitalization for the initial stroke were significantly higher in the diabetic group, as expected, which supports the accuracy of our classification. However, these assays were not necessarily fasting specimens. During follow-up, 66 (46%) of the 142 patients with diabetes whose HbAlc was assayed had mean levels <6%, 111 (78%) had mean levels <7%, and 125 (88%) had mean levels <8%, leaving only 17 (12%) with mean HbAlc levels >8%, the level at which control of blood glucose is usually considered suboptimal. The number of HbAlc assays available per patient was similar among those with higher levels and lower levels. Mean number of assays ranged from 3.1 to 4.3 per patient (Table 2⇓).
To estimate stability of the HbAlc levels over time, we calculated the SD for each patient using HbAlc values over all available follow-up visits for that patient. The median SD within patients across an incremental range of mean HbAlc levels varied from 0.73 to 1.31 (Table 2⇑), indicating that HbAlc levels tended to be stable over time. There was, however, a trend toward greater fluctuation in individuals with higher mean HbAlc levels.
Since not all patients completed the study, we analyzed the percentage with HbAlc levels >7% among those patients with diabetes with an HbAlc assay who were lost to follow-up before the end of the study. Among those who died (n=29), 6 (21%) had a mean HbAlc >7%; patients who did not complete the study with values >7% included 1 of 5 patients who moved away (20%), 2 of 11 who elected not to continue (18%), and 3 of 17 who developed a recurrent stroke (18%). Among those diabetics with HbAlc assays who remained at the end of the study, 19 of 80 (24%) had a mean HbAlc assay >7%. Thus, when >7% was used as a cut point, there was no disproportionate premature loss from the study of patients with high HbAlc levels.
We have previously reported the frequency of recurrent stroke in patients with and without diabetes in a multifactorial analysis of risk factors for recurrent stroke.9 Here we reiterate these results to provide a more complete picture of the lack of association between diabetes and risk of recurrent stroke. Among the 198 patients with diabetes, 27 (13.6%) had a recurrent stroke by the end of the study compared with 11.8% (50/423) among the patients without diabetes. This difference is not statistically significant (P=.53). The rate of stroke recurrence over time in patients with and without diabetes is illustrated in the Figure⇓ with the use of Kaplan-Meier survival curves. Although the curves appear to diverge slightly toward the end of the study period, there was no significant difference between them (P=.31). The association between control of blood glucose as reflected by level of HbAlc and risk of stroke recurrence among patients with diabetes was not previously reported in our cohort.
To determine whether there was a relationship between level of blood glucose over time and risk of stroke recurrence, we analyzed the data using time-dependent Cox proportional hazards models testing the effect of HbAlc as a continuous variable on risk of recurrent stroke after controlling for age, sex, and four comorbidities. The HR for recurrent stroke, based on the 142 patients with diabetes with at least one HbAlc assay, was 0.87 per 1% increment in HbAlc (95% CI, 0.623 to 1.219). After adjustment for age, sex, and comorbidities in other Cox models in which HbAlc was a dichotomous variable (ie, controlled or not controlled) and with the assumption that HbAlc in patients without diabetes was controlled (n=621), there was no trend toward increased risk of stroke recurrence with increase in the cut point used to define “controlled.” With “controlled” defined as <6%, the HR for a recurrent stroke when we compared controlled and not controlled was 0.51 (95% CI, 0.176 to 1.503); at <7%, the HR was 0.43 (95% CI, 0.089 to 1.923); and at <8%, it was also 0.43 (95% CI, 0.057 to 3.317). Thus, the HR for recurrent stroke was virtually the same regardless of the cut point used to define control of blood glucose, and no association could be demonstrated between level of glucose control and risk of stroke recurrence among patients with diabetes in this cohort. The number of poorly controlled patients with diabetes in our cohort did not permit us to test adequately for an adverse effect on risk of stroke recurrence.
Our study of the association of stroke and diabetes was unusual in that control of blood glucose over time was monitored with the sensitive HbAlc assay rather than blood glucose level. Goldstein10 pointed out rather colorfully that the “relation of glycosylated hemoglobin to a single blood glucose determination is similar to the relation between a baseball player’s batting average for the entire season and the player’s average for just one game.” The test reflects exposure of an individual’s red blood cells to the mean blood glucose level during the preceding 2 to 3 months. It is widely considered to be the best and most sensitive measure of control of blood glucose currently available.10
Several studies have found increased risk of initial stroke among patients with diabetes compared with nondiabetics. For example, Kannel and McGee,1 using the Framingham population, reported a relative risk for initial stroke in patients with diabetes of 2.18 in men and 2.17 in women. In a later study of the same population, the 10-year relative risk of initial stroke in patients with diabetes was 1.40 in men and 1.72 in women.4 Burchfiel et al5 studied Japanese-American men in Honolulu and reported relative risks of initial stroke in patients with diabetes that varied between 1.43 and 2.45, with the higher rate occurring in those whose initial blood glucose was higher. Barrett-Conner and Khaw3 reported a relative risk for initial stroke in patients with diabetes of 1.8 in men and 2.2 in women in the Rancho Bernardo retirement community. Whisnant et al11 reported an OR of 2.0 (95% CI, 1.46 to 2.83) for an initial ischemic stroke in patients with diabetes compared with matched control subjects in a population-based study in Rochester, Minn. In none of these studies was control of blood glucose adequately monitored. In the Framingham Study blood glucose was assayed only every 2 years, which provided a cross-sectional assessment of blood glucose level at rather widely separated intervals rather than measures of glucose control over time. Moreover, in most previous studies the risk reported was for an initial stroke rather than for recurrent stroke, as in the present study.
Several groups have studied recurrent stroke in patients with diabetes. Burn et al12 in Oxfordshire, UK, reported a relative risk of 1.71 (95% CI, 0.90 to 3.26). It was not significantly increased compared with patients without diabetes. Sacco et al13 also investigated recurrent stroke in patients with diabetes but limited the study to the risk of recurrence within 30 days of the initial stroke. They reported that the percentage of early stroke recurrence in patients with diabetes was 4.88% compared with 2.65% in patients without diabetes (P=.05).
Hier et al14 determined the stroke recurrence rate in a Stroke Data Bank cohort that included patients with and without diabetes. Using Kaplan-Meier estimates, they determined that the recurrence rate of stroke at 2 years in patients with diabetes with no prior stroke was 15.2% compared with 11.4% in patients without diabetes. These recurrence rates are similar to ours, which were 13.6% in patients with diabetes and 11.8% in patients without diabetes, also with an average follow-up of 2 years. The difference between patients with and without diabetes was statistically significant in the study of Hier et al (P=.005) but not in ours.
More germane was whether control of the blood glucose in patients with diabetes affected the rate of stroke recurrence, but this was not addressed in the study of Hier et al.14 Our study failed to show such an effect. The fact that so few patients with diabetes in our cohort had poorly controlled blood glucose levels (only 17 patients had average levels of HbAlc >8%) certainly limited our ability to detect any effect on risk of stroke recurrence from substantially and persistently elevated blood glucose levels. Even so, we would have expected to demonstrate a trend toward increased risk of stroke recurrence in association with increasing HbAlc levels, but we did not. The HRs were similar and not significant across the three different incremental cut points that we used to define “controlled glucose.” They were, in fact, <1.0. However, we previously reported9 that when treatment of diabetes in our patients was considered, those who received insulin had a higher cumulative stroke recurrence rate than patients with diabetes who had other noninsulin treatments. Insulin-dependent patients may have a worse prognosis for stroke recurrence than non–insulin-dependent patients.
Toyry et al15 followed a cohort of patients with NIDDM. They found no significant difference in mean HbAlc levels at the 5-year follow-up assessment comparing those who had had an initial stroke during follow-up with those who did not (9.6±2.7% and 9.3±2.6%, respectively; P=.713). These investigators did report that high initial blood glucose was associated with increased risk of stroke (OR, 1.2; 95% CI, 1.04 to 1.4), but high glucose at 5-year follow-up was not. The 95% CI included 1 (OR, 1.2; 95% CI, 1.0 to 1.5). Kuusisto et al,16 in a Finnish study of patients with NIDDM, found that HbAlc (along with fasting and 2-hour glucose) at enrollment was a strong predictor of initial stroke that outweighed other classic risk factors. However, neither Toyry’s nor Kuusisto’s group monitored control of blood glucose over time.
No bias was introduced into our analysis by a disproportionate loss of patients with diabetes in poor control. The proportion of patients with average HbAlc levels >7% was virtually the same for each group of dropouts and lower than in those who completed the study. Moreover, loss of patients with diabetes among those who died early during the initial hospitalization could not have appreciably altered our results because of 22 such patients, only 5 had diabetes.
Yet another variable that could have affected our results is the age of the population studied. While Boysen et al2 observed a decline in risk of stroke with age among patients with diabetes, Whisnant et al11 reported an OR for stroke among such patients of 1.7 (95% CI, 1.2 to 2.4) across all ages studied. Thus, the role of age with regard to risk of stroke in patients with diabetes is unclear.
We adjusted for four major comorbidities in our analyses, while Whisnant et al11 and Barrett-Conner and Khaw,3 for example, adjusted for many more. Risk factor adjustment usually reduces the relative risk estimates. Thus, our negative finding is unlikely to be the result of having failed to adjust for an important variable.
The longer a cohort is followed, the more likely it is that additional strokes will occur. We followed our patients for up to 4 years, with a mean follow-up of 2 years. Our results are comparable to those of Hier et al14 in terms of frequency of stroke recurrence after 2 years. The HR for recurrent stroke in the study of Hier et al was similar to that reported by Burn et al,12 who followed the patients in Oxfordshire, UK, up to 6.5 years. Therefore, it is unlikely that moderately longer follow-up would alter results.
Since 95% of the population in the Lehigh Valley is white, our results may not be generalizable to blacks and Asian-Americans. Our results also need to be confirmed in other white populations.
A more rigorous test than ours of whether control of blood glucose affects risk of stroke recurrence would require more frequent monitoring of HbAlc and prospective follow-up of patients with diabetes with a wider range of HbAlc levels, especially including more patients with poor control. However, were it known that blood glucose was elevated in patients with diabetes, treatments to lower the blood glucose would certainly be attempted. Pulsinelli et al17 and Helgason18 have shown that hyperglycemia is associated with poorer neurological outcomes after stroke so that it is important to strive for optimal control in such patients, ie, an HbAlc level <8%. Accrual of adequate numbers of patients with persistently elevated HbAlc levels would, therefore, not be easy.
Blood glucose level, as managed by the patient’s own physician in our cohort, was generally in good control, with 88% having an HbAlc level on average <8%. While evidence that poor control of blood glucose is associated with an increased frequency of recurrent stroke was not available in our cohort, our data support the conclusion that when blood glucose is well controlled, the risk of recurrent stroke up to 4 years after an initial stroke is no greater than in patients without diabetes. Moreover, no trend toward increased risk of recurrent stroke was observed with incremental levels of HbAlc over the range of available HbAlc values in our cohort.
Selected Abbreviations and Acronyms
|IDDM||=||insulin-dependent diabetes mellitus|
|NIDDM||=||non–insulin-dependent diabetes mellitus|
This study was supported by the National Institutes of Health/National Institute of Neurological Disorders and Stroke (RO1 NS22188-06). Dr Henry Hooker participated in the early analysis of data. The hospital directors, physicians, nurses, patients, and their families in the Lehigh Valley were key to the success of this study.
Reprint requests to Dr M. Alter, Department of Neurology, Medical College of Pennsylvania, Allegheny University, 3300 Henry Ave, Philadelphia, PA 19129.
- Received November 12, 1996.
- Revision received March 10, 1997.
- Accepted March 10, 1997.
- Copyright © 1997 by American Heart Association
Boysen G, Nyboe J, Appleyard M, Sorensen PS, Boas J, Somnier F, Jensen G, Schnohr P. Stroke incidence and risk factors for stroke in Copenhagen, Denmark. Stroke. 1988;19:1345-1352.
Barrett-Connor E, Khaw K-T. Diabetes mellitus: an independent risk factor for stroke? Am J Epidemiol. 1988;128:116-123.
Wolf PA, D’Agostino RB, Belanger AJ, Kannel WB. Probability of stroke: a risk profile from the Framingham Study. Stroke. 1991;22:312-318.
Burchfiel CM, Curb JD, Rodriguez BL, Abbott RD, Chiu D, Yano K. Glucose intolerance and 22-year stroke incidence: the Honolulu Heart Program. Stroke. 1994;25:951-957.
Lai SM, Alter M, Friday G, Sobel E. A multifactorial analysis of risk factors for recurrence of ischemic stroke. Stroke. 1994;25:958-962.
Lai SM, Alter M, Friday G, Sobel E. Prognosis for survival after an initial stroke. Stroke. 1995;26:2011-2015.
Whisnant JP, Wiebers DO, O’Fallon WM, Sicks JD, Frye RL. A population-based model of risk factors for ischemic stroke: Rochester, Minnesota. Neurology. 1996;47:1420-1428.
Burn J, Dennis M, Bamford J, Sandercock P, Wade D, Warlow C. Long-term risk of recurrent stroke after a first-ever stroke: the Oxfordshire Community Stroke Project. Stroke. 1994;25:333-337.
Sacco RL, Foulkes MA, Mohr JP, Wolf PA, Hier DB, Price TR. Determinants of early recurrence of cerebral infarction: the Stroke Data Bank. Stroke. 1989;20:983-989.
Hier DB, Foulkes MA, Swiontoniowski M, Sacco RL, Gorelick PB, Mohr JP, Price TR, Wolf PA. Stroke recurrence within 2 years after ischemic infarction. Stroke. 1991;22:155-161.
Toyry JP, Niskanen LK, Lansimies EA, Partanen KPL, Uusitupa MIJ. Autonomic neuropathy predicts the development of stroke in patients with non–insulin-dependent diabetes mellitus. Stroke. 1996;27:1316-1318.
Kuusisto J, Mykkanen L, Pyorala K, Laakso M. Non–insulin dependent diabetes and its metabolic control are important predictors of stroke in elderly subjects. Stroke. 1994;25:1157-1164.
Helgason CM. Current concepts of cerebrovascular disease and stroke: blood glucose and stroke. Stroke. 1988;19:1049-1053.