Diabetes: a Risk Factor for Ischemic Stroke in a Large Bi-Racial Population

Abstract

Background and Purpose

We previously reported increased incidence of ischemic stroke among both blacks and whites with diabetes, especially in those <55 years old. With rising prevalence of diabetes in the last decade, we revisit the impact of diabetes on stroke incidence in the same population (~1.3 million) 5 and 10 years later.

Methods

This is a population-based study. First ischemic strokes among African American and white residents of the 5-county Greater Cincinnati/Northern Kentucky region, ≥20 years old, for periods 7/1993 to 6/1994, 1999, and 2005, were included in this analysis. Incidence rates were adjusted for gender, race and age, as appropriate, to the 2000 US population.

Results

History of diabetes among first ischemic strokes was reported for 493/1,709 (28%) in 1993/94, 522/1,778 (29%) in 1999, and 544/1,680 (33%) in 2005. Risk ratios (95% CI) for rates of stroke in those with versus without diabetes for African Americans reduced significantly from 5.6 in 1993/94 to 3.2 in 2005; for whites the risk ratio remained stable at 3.8 in 1993/94 and 2005. However, risk ratios varied with age, with an overall 5- to 14-fold increased risk observed in those 20 to 65 years old.

Conclusions

Those with diabetes remain at greatly increased risk for stroke at all ages, especially <65 years, regardless of race. The rates and risk ratios for 1999 and 2005, while similar to those previously reported for the mid-1990s, take on increased significance, given the epidemic of diabetes and metabolic syndrome throughout the US and the world.

We previously reported an increased incidence of ischemic stroke among both African Americans and whites with diabetes. We found significant variation by age, with high risks for the youngest (especially for those younger than 55).¹ More recently, we also reported on increasing incidence of stroke over time in this relatively younger population.² In addition, George et al, using the Nationwide Inpatient Sample (NIS) database, reported an overall increase in hospitalization rates for ischemic stroke between 1995-1996 and 2007-2008 for age 5-14, 15-34 and 35-44 age groups and a concomitant increase in prevalence rate of diabetes.³ Diabetes, perhaps the second most important risk factor for stroke, is highly associated with other comorbidities, including hypertension.^{4, 5}

The prevalence of diabetes has been increasing during the last decade in all age groups but particularly in the younger population.^5-8 Taken together with our previous findings, one might expect an increasing incidence of stroke, especially in the younger age ranges, related to this higher prevalence of diabetes. In contrast, recent studies, including our own, have shown that the overall incidence of stroke has been decreasing in the white population^9-11 but has remained stable in the African American population.¹⁰ Further examination of our rates by age has revealed decreasing incidence in the older white population.²

Because of the rising epidemic of diabetes in the last decade, we decided to revisit the impact of diabetes on stroke incidence in the same region of 1.3 million people, 5 and 10 years after the initial observation. We sought to examine any potential trends of increasing stroke incidence in those with pre-existing diabetes. The Greater Cincinnati Northern Kentucky Stroke Study (GCNKSS) population reflects the US population with respect to percentage of African Americans, median age and income, and educational level. Thus, any changes observed in our population may be generalizable to the US African American and white population.

Methods

The GCNKSS population is defined as all residents of two counties in southwest Ohio (one of which includes the city of Cincinnati) and three contiguous counties in northern Kentucky, separated by the Ohio River. Only residents of the five study area counties were eligible to be counted as cases for this study. This study was approved by the Institutional Review Board at all participating hospitals.

The methodology and definitions used for our study have been previously reported¹². Study nurses retrospectively reviewed the medical records of all inpatients and hospital emergency department visits with primary or secondary stroke-related ICD-9 discharge diagnoses (430-436) from the acute-care hospitals in the study region. This included 19 hospitals in 1993/94, 18 in 1999, and 17 in 2005; the decreasing numbers are due to consolidations and closings. Strokes were also ascertained by monitoring stroke-related visits to all local public health clinics and hospital-based outpatient clinics. Cases where stroke was listed as the primary or secondary cause of death by one of the five county coroners’ offices were also included. In addition, monitoring was performed by examining the records of potential stroke cases in a random sample of primary care physicians’ offices and nursing homes in the GCNK region. Sampling was necessary given the large number of physician offices and nursing homes in the region; we have previously described this sampling in detail.¹³ Sites were selected randomly, for each study period, by the study statistician from a list generated from a combination of the local yellow pages and the American Medical Association listing of physicians in the region. The random sample consisted of 50 of the 878 primary care physicians’ offices and 25 of the 193 nursing homes in the study region, in 1993/94. The numbers were 37 of the 849 primary care physicians’ offices and 23 of the 171 nursing homes in 1999 and 51 of the 832 primary care physicians’ offices and 26 of the 126 nursing homes in 2005. All events were cross-checked within and between sources to prevent double counting. Thus the total number of first-time strokes in this area is represented; the out of hospital ascertained strokes account for between 10% and 20% of the strokes in our population in the three study periods.¹⁴

To qualify as a case, a patient must have met the criteria for one of the five stroke categories adapted from the Classification for Cerebrovascular Diseases III and epidemiological studies of stroke: cerebral ischemia, intracerebral hemorrhage, subarachnoid hemorrhage, stroke of uncertain cause, or transient ischemic attack (TIA; symptoms lasting < 24 hours).^{1, 2, 13, 15} Once potential cases were identified, a study nurse reviewed and abstracted the medical record. All probable and borderline stroke cases were abstracted for physician review. Data collected included stroke symptoms, past medical and surgical history, and disposition/outcome; also medical and vascular risk factors as recorded in the medical record. Patients were identified as having diabetes based on documentation in the medical record that diabetes had been diagnosed prior to the stroke. It is likely that diabetes was undercounted, as only those patients with a diagnosis of diabetes documented in the medical record were classified as having diabetes (undiagnosed diabetes, i.e., elevated fasting glucose or glycohemoglobin A1C, was not classified as “diabetes” for this study). The classification of race was as self-reported in the medical administrative record. A study physician reviewed each abstract and all available neuroimaging data to determine whether a stroke or TIA occurred. Study physicians also characterized imaging findings and assigned stroke subtype and mechanism to each patient based on all available information. To ensure consistency over time, only those ischemic strokes identified by the strict clinical definition were included in this analysis.^{1, 2, 13, 15} Because we were tracking incident events, recurrent cases of ischemic stroke were excluded.

Statistical analyses

Data management and descriptive and comparative analyses were performed using SAS® versions 8.02 and 9.2 respectively (SAS Institute, Cary, NC). Population estimates were obtained by including the sampling weights in all analyses as dictated by the study design. The sampling weights were one for all strokes except those identified through the physician's office or nursing home monitoring. The sampling weight is calculated, by study period, as the number in the area divided by number sampled, as described above.

Incidence rates of ischemic stroke were estimated separately for patients with and without diabetes. The numerator was the weighted number of first ischemic strokes as determined by physician review, further classified by diabetes status. The denominator by gender, race, and age group was extracted from the US Census Bureau website (www.census.gov) for 1993, 1994, 1999, and 2005 for the 5-county area. The denominators for the populations with and without diabetes were calculated based on the age-, gender-, and race-specific rates of diabetes in the NHANES III, 1999-2000, and 2005-2006 databases as appropriate, for the study year. Diabetes was defined as response to the NHANES question “other than during pregnancy have you ... ever been told by a doctor or other health professional that you have ... diabetes or sugar diabetes?”, as this was thought to best reflect our chart abstraction information. NHANES has been used extensively to report population-based prevalence of diabetes and other chronic diseases.^{6, 16}

Stroke incidence rates were age, gender, and race adjusted to the 2000 US population, as appropriate. Standard errors for incidence rates were estimated assuming a Poisson distribution. Risk ratios for stroke in patients with diabetes were obtained by division of the incidence rates for those with diabetes by the incidence rates for those without diabetes, and standard errors were estimated using the delta method.¹⁷ Generalized estimating equations (GEE)¹⁸ were used to examine the bivariate differences between stroke patients with and without diabetes and the trends over time. This allowed inclusion of a clustering variable to account for the sampling scheme defined above. The working correlation structure giving the best model fit was obtained. Data are reported as raw numbers with weighted percentages or weighted means and the associated standard error.

Results

Among African American and white area residents ≥20 years old identified in the study area, there were 1,709 first-ever ischemic strokes in 1993/94, 1,778 in 1999, and 1,680 in 2005. In 1993/94, 493 (28%) had a documented history of diabetes diagnosed prior to stroke, compared with 522 (29%) in 1999 and 544 (33%) in 2005 (p for trend = 0.01). This trend of increasing diabetes in stroke subjects mirrors the national trend of increasing diabetes in the US population.¹⁹ Data from NHANES III, 1999-2000 and 2005-2006 showed a similar increase in rate of diabetes from the population aged 20 years and older of 5.1%, 5.9% and 7.7%.

The demographics of the stroke patients with and without diabetes are shown in Table 1. In all three study periods, patients with diabetes were somewhat younger on average than those without diabetes, but this was only statistically significant in 1999. Gender distribution was similar over time, but the proportion of African Americans with diabetes was higher than that of whites in all three study periods (18.8% vs 15.5% in 1993/94; 19.0% vs 13.8% in 1999, and 26.7% vs 16.2% in 2005, p<0.001 for 1999 and 2005).

Table 1.

Demographics and Stroke Risk Factors of the 1993/94, 1999 and 2005 stroke population by diabetes status

	1993/94		1999		2005
	Diabetes n=493 (27.6%)	No diabetes n=1216 (72.4%)	Diabetes n=522 (29.1%)	No diabetes n=1256 (70.8%)	Diabetes n=544 (32.6%)	No diabetes n=1136 (67.4%)
Age (years)	71.3 (0.50)	72.4 (0.63)	70.8 (0.62)	73.9 (0.72)*	69.3 (0.68)	70.1 (0.51)
Age <65 years†	117 (22.2%)	284 (24.8%)	151 (35.8%)	291 (24.2%)	197 (22.6%)	387 (31.7%)
Female	287 (57.7%)	687 (55.6%)	302 (60.3%)	703 (55.3%)	297 (55.1%)	627 (52.1%)
African American	99 (18.8%)	197 (15.5%)	116 (19.0%)	183 (13.8%)*	143 (26.7%)	206 (16.2%)*
Hypertension†	379 (75.2%)	649 (56.5%)*	422 (83.6%)	790 (63.7%)*	481 (88.7%)	796 (71.7%)*
High cholesterol†	72 (13.7%)	92 (9.1%)	155 (32.6%)	213 (17.3%)*	288 (56.2%)	356 (35.3%)*
Myocardial infarction	104 (22.9%)	142 (13.2%)#	100 (16.4%)	144 (11.6%)	89 (17.6%)	121 (10.7%)#
Atrial fibrillation	82 (15.6%)	169 (12.8%)	79 (13.0%)	183 (15.8%)	92 (15.5%)	172 (14.2%)
Current smoking†	91 (17.3%)	270 (20.8%)	101 (16.6%)	289 (19.5%)	109 (19.0%)	317 (27.4%)#

Data presented as weighted mean (weighted standard error) or raw n (weighted %)

^#p≤0.05 for within year comparison (diabetes vs. no diabetes)

^*p≤0.001 for within year comparison (diabetes vs. no diabetes)

^†change over years (p<0.05)

Distributions of stroke risk factors by study period and diabetes status are also shown in Table 1. The proportions of stroke risk factors were higher in stroke patients with diabetes, compared with those without diabetes, except for atrial fibrillation and current smoking. The diagnoses of hypertension and high cholesterol increased over time for both those with and without diabetes, and remained significantly increased in each period for patients with diabetes, compared with those without diabetes. There was an overall decrease in history of myocardial infarction over time, it remained increased in those with diabetes compared with those without. Rates of atrial fibrillation did not change over time and were not different by diabetes status. Current smoking rates increased over time for both those with and without diabetes, and were consistently higher in those without diabetes, although only statistically significant in 2005.

Gender- and age- adjusted incidence rates of first-ever ischemic stroke for populations both with and without diabetes, are shown in Table 2. Incidence rates are shown for the study years 1993/94, 1999, and 2005. Diabetes confers a consistent, significantly higher incidence of stroke in all study periods.

Table 2.

Age and Race Specific Incidence Rates per 100,000 (95% CI) for First-Ever Ischemic Stroke in those with and without diabetes

Age	African American			White
	1993/94	1999	2005	1993/94	1999	2005
Diabetes
<65 years*	602† (382, 823) [n=29]	759 (549, 969) [n=56]	529 (402, 656) [n=67]	411 (325, 497) [n=88]	470 (374, 567) [n=95]	454 (375, 532) [n=130]
≥ 65 years*	4383 (3339, 5427) [n=70]	948 (707, 1189) [n=60]	1437 (1114, 1760) [n=76]	1777 (1578, 1977) [n=306]	1726 (1533, 1918) [n=311]	1264 (1113, 1414) [n=271]
No Diabetes
<65 years*	72 (56, 89) [n=74]	77 (60, 94) [n=82]	101 (82, 120) [n=108]	33 (29, 38) [n=210]	33 (28, 37) [n=209]	44 (39, 49) [n=279]
≥ 65 years*	745 (612, 878) [n=123]	795 (640, 951) [n=101]	681 (545, 817) [n=98]	628 (585, 672) [n=809]	680 (634, 725) [n=864]	529 (488, 569) [n=651]

^*Sex adjusted to the 2000 US population

^†Rates based on samples of <30 should be treated with caution

In the population with diabetes, overall incidence rates for stroke for African Americans decreased significantly over the study periods, from 1331/100,000 (95%CI 1037, 1624) in 1993/94 to 617/100,000 (496, 737) in 2005, whereas the incidence rates for the white population remained unchanged, 549/100,000 (484, 614) in 1993/94 and 504/100,000 (443,565) in 2005. This is in contrast to the overall incidence rates in the population without diabetes, where the African American rate did not change significantly; 241/100,000 (208, 275) in 1993/94 and 216/100,000 (185, 246) in 2005. However, the rate in the white population deceased significantly from 169/100,000 (159, 246) in 1993/94 to 145/100,000 (136, 154) in 2005.

Stroke incidence rates by age group (Table 2); for both the African American and white populations with diabetes remain unchanged over time for those less than 65 years of age, but decrease significantly in the older age group. Stroke incidence rates for those without diabetes for the African American population show no significant change over the 15 year period, but there is a significant increase in incidence of stroke seen in the white population under 65 years of age, in contrasted to a significant decrease in incidence in those 65 years and older.

The risk ratios for incidence rate of stroke for those with diabetes compared to those without diabetes are presented in Table 3. The excess risk for those with diabetes is decreasing in the African American population, but remains stable in the white population over time. The overall risk ratio for African Americans decreased significantly from 5.6 in 1993/94 to 3.2 in 2005; in whites the risk ratio was 3.8 in both 1993/94 and 2005. Age-specific risk ratios by race also show a significant decrease over time in the African American population aged 65 years and above, but not for those less than 65 years. In the white population those less than 65 years old have a 12- to 14-fold increase in stroke incidence over the study periods when diagnosed with diabetes prior to stroke; the excess risk is under 3 in the older age group but does not change significantly over time in either age group.

Table 3.

Risk ratios (95% confidence interval (CI)) for ischemic stroke in those with diabetes vs. without diabetes

	African American			White
Age	1993/94	1999	2005	1993/94	1999	2005
<65 years*	8.4 (4.8, 11.9)	9.8 (6.6, 13.0)	5.2 (3.6, 6.8)	12.4 (9.4, 15.5)	14.2 (10.7, 17.7)	12.0 (8.8, 15.2)
≥ 65 years*	5.9 (4.1, 7.6)	1.2 (0.8, 1.6)	2.1 (1.5, 2.7)	2.8 (2.4, 3.2)	2.5 (2.2, 2.9)	2.7 (2.1, 3.4)
All ages†	5.6 (4.2, 7.1)	2.8 (2.0, 3.5)	3.2 (2.4, 3.9)	3.8 (3.2, 4.3)	4.0 (3.5, 4.6)	3.8 (3.3, 4.3)

^*Sex adjusted to the 2000 US population

^†Sex and age adjusted to the 2000 US population

Data were collapsed over the years to examine the effect of age more closely. The risk ratios reiterate the increased risk for stroke conferred by diabetes in the white population compared to African Americans in the age groups less than 65 years, these are statistically significant in the groups 45-54 years and 55-64 years. In the oldest age group 75 years and above a significantly increased risk ratio for the white population compared to the African American population is again observed. The overall risk ratio conferred by diabetes for Whites is significantly greater than that for African Americans. Data are presented in Supplemental Table S1, please see http:/stroke.ahajournals.org.

Discussion

As we had shown before,¹ the population with diabetes is at substantially increased risk for stroke. Our data highlight the increasing prevalence of diabetes among stroke patients over time in line with the “epidemic” of diabetes seen in the general population. We show higher prevalence of risk factors in patients with diabetes, and increasing hypertension and high cholesterol in all stroke patients over time. Furthermore, our data demonstrate a continued 3- to 4-fold increased incidence rate of ischemic stroke for patients with diabetes as compared with those without diabetes. The excess risk is seen across all ages, but it is particularly striking in the age group less than 65 years. We are also seeing a shift over time with evidence of higher ischemic stroke risk in whites compared with blacks conferred by diabetes at virtually all ages.

Our data are of great public health importance given the well-publicized epidemic of diabetes and obesity, or alternatively of the metabolic syndrome.^{19, 20} Diabetes is clearly an important risk factor for stroke especially in those under 65 years of age.^{2, 3} This finding is consistent with those reported from studies in the UK and Europe.^{21, 22} In addition, it has been shown that diabetes is an important risk factor associated with poor post-stroke outcomes and disability.^23-26 Thus, higher rates of stroke in patients with diabetes, especially in younger patients, may be associated with a higher burden of disability, productive life-years lost, and higher cost to society. Given the increasing prevalence of diabetes in 2005 compared to 1993/94, we estimate that there were an additional 25,000 strokes in 2005 associated with this increase.

Our study has several limitations. We have the potential for increased awareness and increase in physician diagnosis of diabetes and other risk factors over time, which may skew the risk ratios. In addition, we do not know duration or time since diagnosis of diabetes, nor can we estimate the rate of undiagnosed diabetes at initial presentation with stroke for all of the periods studied. Finally, because our methodology is retrospective, there is the risk of under-documentation bias, which would result in our diabetes-specific rates being underestimates. Nevertheless, weaknesses are balanced by strengths, including the population-based nature of our study and the ability to assess changes over time with little change in methodology for both stroke and diabetes ascertainment.

Summary

In conclusion, diabetes is not only a highly important risk factor for ischemic stroke at all ages, but is especially so in those less than 65 years of age with risk ratios showing more than a 5-fold risk. Our data support diabetes as a very important preventable ischemic stroke risk factor. Thus, physicians caring for patients at risk for stroke should be vigilant for diabetes, as well as other concurrent stroke risk factors that tend to cluster with diabetes. We see a significant risk difference where diabetes may confer more risk to whites than blacks for stroke, which requires further study.