Diabetes, Fasting Glucose Levels, and Risk of Ischemic Stroke and Vascular Events

Findings from the Northern Manhattan Study (NOMAS)

  1. Bernadette Boden-Albala, MPH, DRPH12,
  2. Sam Cammack, MS1,
  3. Ji Chong, MD1,
  4. Culing Wang, PHD3,
  5. Clinton Wright, MD, MS1,
  6. Tatjana Rundek, MD4,
  7. Mitchell S.V. Elkind, MD, MS1,
  8. Myunghee C. Paik, PHD3 and
  9. Ralph L. Sacco, MD, MS4
  1. 1Department of Neurology, Columbia University College of Physicians and Surgeons, New York, New York
  2. 2Department of Sociomedical Science, Columbia University Mailman School of Public Health, New York, New York
  3. 3Department of Biostatistics, Columbia University Mailman School of Public Health, New York, New York
  4. 4Department of Neurology, University of Miami, Miami, Florida
  1. Corresponding author: Bernadette Boden-Albala, DrPH, Neurological Institute, 710 W. 168 St., New York, NY 10032. E-mail: bb87{at}columbia.edu

Abstract

OBJECTIVE—There is insufficient randomized trial data to support evidence-based recommendations for tight control of fasting blood glucose (FBG) among diabetic subjects in primary stroke prevention. We explored the relationship between FBG among diabetic subjects and risk of ischemic stroke in a multiethnic prospective cohort.

RESEARCH DESIGN AND METHODS—Medical and social data and FBG values were collected for 3,298 stroke-free community residents: mean age ± SD was 69 ±10 years; 63% were women, 21% were white, 24% were black, and 53% were Hispanic; and follow-up was 6.5 years. Baseline FBG levels were categorized: 1) elevated FBG: history of diabetes and FBG ≥126 mg/dl (7.0 mmol/l); 2) target FBG: history of diabetes and FBG <126 mg/dl (7.0 mmol/l); or 3) no diabetes/reference group. Cox models were used to calculate hazard ratios (HRs) and 95% CI for ischemic stroke and vascular events.

RESULTS—In the Northern Manhattan Study, 572 participants reported a history of diabetes and 59% (n = 338) had elevated FBG. Elevated FBG among diabetic subjects was associated with female sex (P < 0.04), Medicaid (P = 0.01), or no insurance (P = 0.03). We detected 190 ischemic strokes and 585 vascular events. Diabetic subjects with elevated FBG (HR 2.7 [95% CI 2.0–3.8]) were at increased risk of stroke, but those with target FBG levels (1.2 [0.7–2.1]) were not, even after adjustment. A similar relationship existed for vascular events: elevated FBG (2.0 [1.6–2.5]) and target FBG (1.3 [0.9–1.8].

CONCLUSIONS—This prospective cohort study provides evidence for the benefits of tighter glucose control for primary stroke prevention.

Diabetes is a major public health problem. In the U.S. today, more than 18.2 million people have diabetes. The prevalence of diabetes increases with age, and certain populations, including minority groups, may be more vulnerable. According to the American Diabetes Association, one of every four African-American and Hispanic individuals >65 years of age has diabetes. In addition, it is estimated that another 10–15% of the U.S. population has elevated blood glucose levels and may be considered to have “pre-diabetes.” Diabetes is a major risk factor for cardiovascular disease and is associated with a >2-fold risk of cardiovascular outcomes, including incident myocardial infarction and mortality (1,2). Similarly, diabetes has been associated with an increased risk of stroke with relative risks (RRs) ranging from 1.8 to 6 (35).

Despite the public health significance of diabetes, there is a paucity of data emphasizing tight control of fasting blood glucose (FBG) as a stroke prevention measure. According to the American Heart Association “Guidelines for the Primary Prevention of Stroke,” there is insufficient randomized trial data to support glycemic control among diabetic subjects as a stroke prevention measure (6). American Heart Association recommendations based on the UK Prospective Diabetes Study (UKPDS) support glycemic control among individuals with diabetes to “reduce microvascular complications, nephropathy, and retinopathy, as well as peripheral neuropathy” (6,7). In the UKPDS, tight glycemic control of a prospective cohort of individuals with newly diagnosed diabetes did not significantly reduce stroke risk (7).

The current obesity epidemic and associated increasing prevalence of diabetes, especially among minority populations, warrants further investigation of whether tight control of FBG is associated with decreased stroke risk. The aim of this study was to explore the relationship between FBG levels among diabetic subjects and the risk of ischemic stroke as well as other vascular events in a multiethnic prospective cohort

RESEARCH DESIGN AND METHODS

The Northern Manhattan Study (NOMAS) is a prospective population-based cohort study designed to study incidence, risk factors, and prognosis of stroke in a multiethnic urban community. Based in Northern Manhattan, an area of ∼260,000 people, with 104,000 individuals aged >39 years of age, this study has a unique race-ethnicity distribution of ∼63% Hispanic, ∼20% black, and ∼15% white. The methodology for NOMAS has been described previously and is summarized briefly below (8).

Selection of prospective cohort

A total of 3,298 subjects were recruited and enrolled between 1993 and 2001. Participants were eligible if they 1) had never had a stroke, 2) were aged ≥40 years, and 3) had resided for at least 3 months in a household with a telephone in Northern Manhattan. Subjects were identified by random digit dialing by using dual frame sampling, and bilingual interviews were conducted (9). The telephone response rate was 91% (9% refused to be screened), and 87% of those eligible indicated willingness to participate. This study was approved by the local governing institutional review board, and written consent was obtained.

Baseline evaluation

Subjects were recruited from the telephone sample to have an in-person baseline interview and assessment. The enrollment response rate was 75% with an overall response rate of 68% (telephone response × enrollment response). Standardized questions regarding medical history were adapted from the Centers for Disease Control and Prevention Behavioral Risk Factor Surveillance System (10). Race-ethnicity was based upon self-identification.

Glucose levels

To examine the impact of FBG on stroke risk among diabetic subjects, we divided baseline FBG levels into three categories: 1) elevated FBG: a history of diabetes and FBG ≥126 mg/dl (7.0 mmol/l); 2) target FBG: a history of diabetes and FBG <126 mg/dl (7.0 mmol/l); or 3) no diabetes/reference group, including those with no history of diabetes. Fasting serum glucose was measured according to standard procedures using a glucose dehydrogenase method (11).

Other baseline assessments

Hypertension was defined as either systolic blood pressure levels ≥140 mmHg, diastolic blood pressure levels ≥90 mmHg, or a history of hypertension. Cardiac disease included history of angina, myocardial infarction, coronary artery disease, or valvular heart disease. Obesity was defined using BMI. Smoking was categorized as never, former, and current smoker. Moderate alcohol use was defined as current drinking of >1 drink per month and ≤2 drinks per day. Physical activity was defined as engaging in leisure activity versus not over the 10 days before baseline enrollment. Social resources were defined by educational level and health insurance status. Education was dichotomized into those who had completed high school versus those who had not. Health insurance was separated into three mutually exclusive groups: 1) individuals who had Medicaid or Medicaid/Medicare or no insurance; 2) individuals who had private insurance or private/Medicare; and 3) individuals with Medicare only (reference group). We combined the no insurance group with the Medicaid group on the basis of a similar risk ratio in these groups, as well as the very low prevalence (7%) of noninsured individuals in this cohort.

Annual prospective follow-up

Subjects were screened annually by telephone interview to determine any change in vital status, detect neurological and cardiac symptoms and events, review any interval hospitalizations, review risk factor status, review changes in medication, and determine changes in functional status. Phone assessment served as a screen for events. The telephone interview simple stroke question (“Since your last visit have you been diagnosed with a stroke?”) had a sensitivity of 92% and specificity of 95%. Moreover, a 10% random sample of the cohort was followed annually in person for 5 years to evaluate for any telephone false-negative results and evaluate for serial changes in baseline measures. In between follow-up interviews, subjects and family members were reminded to notify us in the event of stroke, myocardial infarction, or death.

Subjects who screened positive by telephone were scheduled for an in-person assessment. All affirmative responses to neurological symptoms and conditions required a review and examination by the study neurologists. In addition, ongoing hospital surveillance of admission and discharge ICD-9 codes provided data on mortality and morbidity that may not have been captured during the annual telephone follow-up.

Outcome classifications (ischemic stroke, myocardial infarction, or vascular death)

For these analyses, two outcome measures were 1) first ischemic stroke and 2) first vascular event, defined as either first ischemic stroke or first myocardial infarction or vascular death. Stroke was defined as “rapidly developing clinical signs of focal (at times global) disturbance of cerebral function, lasting >24 h or leading to death with no apparent cause other than that of vascular origin” (12). Subjects who are hospitalized for stroke at Columbia-Presbyterian Medical Center (CPMC) have systematic evaluations during their hospitalization, which include standard diagnostic tests: admission serological studies, head computed tomography, carotid duplex Doppler, transcranial Doppler, echocardiography, electrocardiography, magnetic resonance imaging, and angiography. The diagnostic evaluation by neurologists at non-Northern Manhattan hospitals is difficult to control, but any participant who needs vascular noninvasive tests can have these done as part of their in-person follow-up visit at CPMC. Subjects who are suspected of having a stroke and have not had a prior diagnostic evaluation have one completed either through their primary care physician, the CPMC neurology clinic, or the Clinical Research Center. If a patient has been hospitalized, then medical records will be reviewed to verify details.

More than 98% of these patients with stroke outcomes had at least one form of imaging, predominately a computed tomographic scan. Although our surveillance system captured both hemorrhagic and ischemic stroke, for the purposes of this study we are reporting only on ischemic stroke outcomes. Hemorrhagic stroke accounted for 9% of all stroke, and we did not have the ability to examine this stroke subtype separately. More than 70% of the patients with stroke were hospitalized at the Columbia University Medical Center.

Myocardial infarction was defined by criteria adapted from the Lipid Research Clinics Coronary Primary Prevention Trial (13) and required at least two of the following three criteria: 1) ischemic cardiac pain determined to be typical angina; 2) cardiac enzyme abnormalities defined as abnormal creatinine phosphokinase MB fraction or troponin values; and 3) electrocardiogram abnormalities.

For subjects who died, the date of death was recorded along with cause of death. Deaths were classified as vascular or nonvascular on the basis of information obtained from the family and physicians and validated with medical records and death certificates. Causes of vascular death included ischemic stroke, myocardial infarction, heart failure, pulmonary embolus, cardiac arrhythmia, and other vascular causes. Nonvascular causes of death included accident, cancer, pulmonary disorders (pneumonia, chronic obstructive pulmonary disease, and others) and other nonvascular causes. All vascular outcomes were reviewed and validated in a manner similar to that for stroke outcomes by our team of three study cardiologists.

Statistical analyses

The prevalence of sociodemographic, conventional vascular risk factors, and other baseline variables was calculated. Kaplan-Meier curves of survival free of ischemic stroke and survival free of vascular events (ischemic stroke, myocardial infarction, or vascular death) were calculated (Fig. 1 of the online appendix available at http://dx.doi.org/10.2337/dc07-0797). Cox proportional regression models were used to calculate hazard ratios (HRs) and 95% CI for ischemic stroke and all vascular events (stroke, myocardial infarction, and vascular death). Times to first event among stroke and vascular outcomes (ischemic stroke, myocardial infarction, or vascular death) were analyzed as outcomes with censoring at the time to either nonvascular event or last follow-up.

In addition, we performed a separate cross-sectional analysis among those with diabetes to identify factors associated with elevated FBG versus target FBG levels (Fig. 2 of the online appendix). Age, sex, race/ethnicity, education, and insurance status were considered as sociodemographic factors, and models were adjusted for hypertension, obesity, dyslipidemia, cardiac disease, smoking, physical inactivity, and alcohol consumption.

RESULTS—

A cohort of 3,298 community residents were enrolled. At baseline, mean ± SD age was 69 ±10 years, 63% were women, 21% were white, 24% were black, and 53% were Hispanic. More than 18% (n = 572) of the cohort had reported a diagnosis of diabetes (Table 1). Among those with diabetes, 19% achieved control with diet, 55% were taking oral hypoglycemic agents, and 26% were taking insulin at the baseline interview.

Control of FBG in diabetic subjects was poor, with >59% (338 subjects) of individuals with diabetes having FBG levels ≥126 mg/dl. Elevated FGB levels were similar among white, black, and Hispanic diabetic subjects. The mean duration of diabetes in the cohort was 12 years with a median of 8 years. There was no difference in the duration of diabetes among those with target FBG levels (mean 11.3 years and median 8 years) versus elevated FBG levels (mean 12 years and median 8 years). Likewise, the use of oral hypoglycemic agents and insulin among diabetic subjects was not statistically different between diabetic subjects with poor control of fasting glucose and those with better fasting glucose levels.

During a mean of 6.5 years of follow-up, 190 ischemic strokes and 585 vascular events were detected. Among the diabetic group, 62 ischemic stroke events were documented (14 in the controlled diabetes group and 48 in the uncontrolled diabetes group) In a multivariable Cox model with no diabetes as the reference, elevated FBG levels (HR 2.7 [95% CI 1.9–3.8]) were significantly associated with increased risk of stroke, whereas target FBG levels (1.2 [0.7–2.1]) did not significantly increase stroke risk after adjustment for age, race-ethnicity, sex, insurance status, education, hypertension, coronary artery disease, lipid levels, obesity, physical inactivity, alcohol intake, and smoking. A similar relationship existed for risk of any vascular events: elevated FBG (2.0 [1.6–2.5]) and target FBG (1.3 [0.9–1.8]). In a separate analysis to determine whether the effect of elevated FBG was significantly different from target FBG, we found that they were significantly different (P = 0.03).

We further explored the dose-response model for levels of FBG for ischemic stroke, vascular death, and any vascular event. We found that FBG levels ≥126 but <150 were associated with a HR of 4.1 [95% CI 2.3–7.2] and FBG levels ≥150 were associated with a HR of 2.7 [1.8–3.9], after controlling for age, race-ethnicity, sex, education, and physical activity.

In a separate analysis, we explored factors associated with elevated versus target FBG levels. Three dimensions of factors were examined: demographics, risk factors, and social resources. In multivariable analyses, elevated FBG was significantly greater among women (P < 0.04), those with Medicaid (P = 0.01), or those with no insurance (P = 0.03). Vascular risk burden or prevalence of risk factors was not significantly associated with elevated versus target FBG levels among those with diabetes.

CONCLUSIONS—

In our multiethnic prospective community-based stroke-free cohort, we report a strong relationship between elevated FBG levels among diabetic subjects (FBG ≥126 mg/dl [7.0 mmol/l]) and increased risk of incident ischemic stroke as well as vascular events after adjustment for sociodemographic and vascular risk factors. The increased risk associated with elevated FBG was similar whether fasting glucose was 126–150 mg/dl or greater. In our cohort, risk of ischemic stroke among a diabetic population with FBG levels within the target range was associated with a lower and nonsignificant risk of stroke and vascular events. Furthermore, there was a significantly different relationship in stroke risk between those with elevated FBG levels and target FBG levels. Few studies have examined the relationship between control of FBG levels and ischemic stroke risk, and most include stroke as a combined outcome with cardiovascular disease. In a meta-analysis, chronic hyperglycemia defined by a 1 point increase in A1C was found to be associated with a pooled RR of 1.18 [95% CI 1.10–1.26] for all vascular end points, and a pooled RR of 1.15 [1.08–1.23] for stroke (14). Similarly, in a nested sample from the Atherosclerosis Risk in Communities (ARIC) cohort, elevated baseline A1C was reported to be an independent predictor of cardiovascular outcomes (1.1 [1.1–1.20]) per 1 percentage point increase in A1C (15) In a registry cohort of Finnish diabetic patients, both high plasma glucose levels (>13.4 mmol/l) and elevated A1C (>10.7%) were strongly associated with increased risk of prevalent stroke (16). In the large prospective ARIC study in which the relationship between elevated FBG levels and risk of ischemic stroke was examined, RRs were compared among two definitions of diabetes. Poor glycemic control defined as fasting glucose ≥140 mg/dl or a history of diabetes conferred a 2.2-fold risk [95% CI 1.6–3.2] of ischemic stroke, and poor glycemic control defined by a history of diabetes or fasting glucose ≥126 mg/dl was associated with a HR of 1.9 [1.3–2.7] for ischemic stoke. However, the ARIC study did not examine the impact of elevated glucose levels among diabetic subjects. Although there appears to be a dose-response relationship between the glucose levels and outcomes in the lower glucose group compared with the mid-range and high-range group, the relative risks in the mid-range group and the high-range group are similar. Indeed the CIs are almost identical. In addition, the mid-range group (n = 73) is small. Nevertheless, the lack of dose response between the two upper glucose groups may suggest that damage to organ systems begins in the mid-level group. Other studies have reported that FBG level is an independent risk factor for coronary heart disease among nondiabetic populations (3,5). In the ARIC cohort, the relationship between FBG levels and ischemic stroke among nondiabetic subjects demonstrated a slightly increased risk but failed to reach significance (3). We found similar results (Table 2).

Only a few randomized trials have been conducted to examine whether tight FBG control leads to better vascular outcomes, and these trials have been criticized for small numbers of subjects and insufficient follow-up. The University Group Diabetes Program (UGDP) found no significant benefit of tight glycemic control among 400 diabetic subjects (17). This study, however, suggested that two agents (tolbutamide and phenformin) used for glycemic control might actually be associated with increased cardiovascular mortality. Another trial of 153 type 2 diabetic men assigned to intensive versus conventional therapy demonstrated no difference in cardiovascular events during 22 months of follow-up (18).

The largest randomized clinical trial examining the relationship between vascular outcomes and tight glycemic control among type 2 diabetic subjects was the UKPDS. In the UKPDS, intensive glycemic control reduced the relative risk of any diabetes-related end point by 12% and microvascular complications by 25% (7). However, tight glycemic control was not significantly associated with decreased risk of stroke (HR 1.1 [95% CI 0.8–1.5]) among 5,102 diabetic subjects randomly assigned to conventional (diet-controlled only) versus intensive treatment regiments (chlorpropamide, glyburide, insulin, or metformin among obese subjects). A strength of the UKPDS was its meticulous, serial measurement of glycemic control using A1C techniques. One limitation of the UKPDS was its definition of stroke as “a major stroke with symptoms or signs lasting 1 month or longer.” This outcome criteria may have led to an underidentification of cases of stroke, especially those of small vessel etiology. In addition, the short follow-up period associated with clinical trials may underestimate the “control effect,” whereas a FBG value in an observational study may be an effective indicator of control over many years. Further, because of the “intent to treat” nature of the trial, >80% of the control group required one or more of the therapies reserved for the intensive intervention regiment. The NOMAS cutoff for glycemic control was 126 mg/dl (7.0 mmol/l), which was actually higher than target glucose levels in the UKPDS. Finally, in the UKPDS, none of the monotherapies were capable of maintaining target “control” glucose levels of <108 mg/dl (<6.0 mmol/l) (7).

In a separate analysis, we also examined associations between FBG levels and vascular and social risk factors. We report that health insurance type is an independent predictor of elevated FBG levels among our diabetic patients. Indeed, FBG targets may be difficult to achieve in clinical practice settings in which continuity of care and adherence to treatment are not emphasized. Other studies have shown similar results. In a retrospective design of subjects with newly treated diabetes captured through an administrative claims database, female sex, preferred provider organization insurance plans, and use of insulin were associated with early nonadherence as well as treatment discontinuation (19). In a small cross-sectional study, lower family socioeconomic status and lower rate of health insurance, but not education or ethnicity, were associated with higher A1C levels (20). The UKPDS showed little difference by race/ethnicity in glycemic control (21). In our study, despite differences in the prevalence of diabetes by race/ethnicity, there were no significant differences in elevated versus target glucose levels between whites, blacks, and Hispanics. The inclusion of a large multiethnic, older, heterogeneous cohort with similar geographical access to the medical center is generalizable to other multiethnic urban populations and allows for more valid comparisons across race/ethnicity categories.

Other strengths include a prospective population-based design for which both baseline exposures and outcomes were well documented. Our aggressive follow-up strategies have resulted in <1% loss to follow-up. Study participants were seen in person at both study entry and follow-up to document outcome events. One potential limitation in our results is that ascertainment of outcomes might differ by diabetes control status. It is possible that patients with poor control may be identified earlier because of other related comorbidities. Overall, 54% of our outcome information was obtained through CPMC daily admissions information, and there was no difference in the proportion of subjects with controlled and uncontrolled diabetes identified this way. Similarly, the proportion of outcome events reported during follow-up of patient and family report (43%) did not differ by diabetes control status. Finally, 3% of outcomes were picked up through screening and examination with no difference by diabetes control status.

An important limitation was our use of a single sample of fasting glucose to define baseline glycemic control. However, given its use in previous epidemiological studies and our lack of A1C data, we felt this measurement of FBG has reasonable utility (4). Most studies have used A1C as the indicator of better versus poor long-term regulation of plasma glucose. Indeed, A1C is thought to reflect long-term glycemic control and may be a more accurate and stable measure than fasting glucose levels (22). Theoretically, A1C reflects the average fasting glucose level throughout the 3-month cycle of the erythrocyte. In reality, the measure of A1C is actually a weighted average with more recent glucose levels contributing more than earlier levels. Studies also indicate that fasting plasma glucose levels may underestimate A1C levels, especially at higher A1C levels (23). Indeed, there may be a systematic bias toward incorrectly assigning persons with a higher A1C as having a lower FBG (better control), which would bias results toward the null hypothesis. Hence, the effect of FBG on risk of incident stroke may be greater than what we observed. Other limitations of this study include our inability to capture microvascular end points as well as our nonspecific identification of therapies for glycemic control.

A number of issues remain unresolved regarding elevated versus targeted FBG among diabetic subjects. Data from this and other prospective studies provide evidence that targeted FBG levels among diabetic subjects are associated with a reduction of macrovascular risk including ischemic stroke and other vascular events. Ongoing rigorous clinical trials such as Action to Control Cardiovascular Risk in Diabetes (ACCORD), a study of >10,000 adults with diabetes, will ultimately provide conclusive evidence regarding the importance of glycemic control in preventing macrovascular disease including ischemic stroke.

Table 1—

Baseline sociodemographic characteristics of the Northern Manhattan Study cohort

Table 2—

Adjusted HRs of ischemic stroke and vascular events by level of glucose among diabetic subjects

Acknowledgments

This work has been funded by National Institutes of Health National Institute of Neurological Disorders and Stroke NS 29993.

The authors thank Janet DeRosa, MPH, NOMAS project manager, and the entire NOMAS research staff for their continued hard work and dedication. We also thank Lin Huang for her statistical help.

Footnotes

  • Published ahead of print at http://care.diabetesjournals.org on 13 March 2008. DOI: 10.2337/dc07-0797.

    Additional information for this article can be found in an online appendix at http://dx.doi.org/10.2337/dc07-0797.

    The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C Section 1734 solely to indicate this fact.

    • Accepted March 4, 2008.
    • Received April 24, 2007.

References

| Table of Contents

This Article

  1. Diabetes Care vol. 31 no. 6 1132-1137
  1. Online-Only Appendix
  2. All Versions of this Article:
    1. dc07-0797v1
    2. 31/6/1132 most recent