Glycated Hemoglobin and Risk of Stroke in People Without Known Diabetes in the European Prospective Investigation Into Cancer (EPIC)–Norfolk Prospective Population Study
A Threshold Relationship?
Background and Purpose— Diabetes is a well-recognized risk factor for cardiovascular diseases. Evidence suggests a linear relationship between blood glucose and myocardial infarction, even at blood glucose concentrations below the threshold for diabetes. The relationship between blood glucose concentration and stroke in people without established diabetes has been studied less extensively.
Methods— We examined the prospective relationship between usual blood glucose level measured by glycohemoglobin (HbA1c) concentrations and incident stroke risk in a general population without diabetes and stroke at baseline assessment in the European Prospective Investigation Into Cancer (EPIC)–Norfolk.
Results— A total of 10 489 men and women aged 40 to 79 years at baseline were followed up (mean=8.5 years). Mean age, systolic and diastolic blood pressure, body mass index, total cholesterol, triglycerides, and proportion of current smokers increased and mean high-density lipoprotein cholesterol decreased with increasing HbA1c concentrations. There were 164 incident strokes identified over 88 652 person-years. After adjustment for age, sex, and cardiovascular risk factors, the relative risks (95% CI) for stroke for participants with HbA1c concentrations 5% to 5.4%, 5.5% to 6.9%, and ≥7% were 0.78 (0.50 to 1.22), 0.83 (0.54 to 1.27), and 2.83 (1.40 to 5.74), respectively, compared with those with HbA1c <5%.
Conclusions— In contrast to the continuous linear relationship observed between blood glucose level and coronary heart disease risk, the association between blood glucose level and stroke risk appears to be more consistent with a threshold relationship. These observations may give insights into the differing pathogenesis of different vascular diseases.
The number of adults in the developed world with diabetes is predicted to rise by 42% between 1995 and 2025,1 whereas in the United Kingdom the relative economic burden imposed by type 2 diabetes and its complications has been predicted to increase by 40% to 50%.2 People with diabetes are known to have an increased risk of microvascular complications such as retinopathy as well as myocardial infarction (MI) and stroke.3 There is considerable overlap between blood glucose concentrations in people with and without established diabetes, and the threshold concentrations applied for the diagnosis of diabetes relate to the clear increase in risk of microvascular complications above these concentrations.
The relationship between blood glucose and macrovascular complications is less clear, although increasing evidence suggests a linear relationship between blood glucose and MI, even at blood glucose concentrations well below the threshold for diabetes.4–6 The relationship between stroke and blood glucose concentrations in people without established diabetes has been studied less extensively, particularly in women and older people, and previous work has yielded conflicting results. Glycohemoglobin (HbA1c) concentration, a measure of average blood glucose concentration over the previous 3 months, is widely used in clinical practice to assess control of diabetes and has been shown to predict the development of microvascular complications.7
In the present study we examined the relationship between HbA1c and incident stroke in a large prospective, population-based cohort of men and women.
Subjects and Methods
Study participants came from a general population-based cohort of ≈25 000 men and women aged 40 to 79 years at baseline participating in the Norfolk arm of the multicenter European Prospective Investigation Into Cancer (EPIC-Norfolk), 99.6% of whom are white British individuals. The Norwich Local Research Ethics Committee approved the study, and all participants gave signed informed consent. Methods of recruitment and measurements in the EPIC-Norfolk have been described elsewhere.8
Briefly, between 1993 and 1997, all individuals aged 40 to 79 years registered with 35 participating general practices in the Norfolk area were invited by mail. Participants who consented to the study completed a Health and Lifestyle Questionnaire and attended a health examination at which anthropometric variables and blood pressure were measured and a venous blood sample was taken. From November 1995, when funding became available, an additional blood sample was taken for measurement of HbA1c.
The EPIC Health and Lifestyle Questionnaire included questions on age, smoking, alcohol use, occupation, and previous medical history. Prevalent stroke, heart disease, and diabetes mellitus were determined by a positive response to the question, “Has a doctor ever told you that you have any of the following?” followed by a list of conditions including stroke, heart attack, and diabetes. Participants with diabetes occurring only in pregnancy were considered nondiabetic.
Blood pressure was recorded as the mean of 2 measurements taken in the right arm with the subject seated for 5 minutes, with the use of an Accutorr sphygmomanometer (Datascope, Huntingdon, UK). Height was measured to the nearest millimeter, without shoes, with the use of a stadiometer. Weight was measured with Salter scales to the nearest 0.1 kg without socks and shoes in light clothing. Body mass index (BMI) was derived as weight in kilograms/(height in meters2).
HbA1c was measured by high-performance liquid chromatography with a Bio-Rad Diamat device (Richmond, Calif). The coefficient of variation was 3.6% at the lower end of the range (mean, 4.9%) and 3.0% at the upper end of the range (9.8%). Serum total and high-density lipoprotein cholesterol and triglycerides were measured in nonfasting samples by colorimetry (RA 1000, Bayer Diagnostics, Basingstoke, UK). Plasma low-density lipoprotein (LDL) cholesterol was calculated with the use of the Friedewald formula.9
Incident stroke cases were ascertained by 2 methods: death certificate data and hospital record linkage. All participants are flagged for death at the UK Office of National Statistics. Death certificates are coded by trained nosologists using International Classification of Diseases (ICD), revisions 9 and 10. Participants are also linked to hospital information systems so that admissions anywhere in the United Kingdom are communicated to EPIC-Norfolk. All hospital admissions with coding for cause for each episode anywhere in the United Kingdom for EPIC-Norfolk participants are obtained through routine record linkage and updated annually. The present study is based on follow-up to March 2005. Stroke death was defined as ICD-9 codes 430 to 438 or ICD-10 codes 60 to 69 anywhere on the death certificate. Incident stroke was defined as stroke death or hospital discharge codes ICD-9 430 to 438 or ICD-10 60 to 69.
Statistical analyses were performed with SPSS for Windows version 12.0.1 (SPSS Inc). The study population for this analysis comprised participants aged 40 to 79 years who completed the Health and Lifestyle Questionnaire and attended a baseline health examination, had HbA1c measured, and had no history of stroke and diabetes at baseline. A total of 155 persons who reported previous history of stroke and 248 participants with known diabetes were excluded from the current analyses. Participants with missing values for covariates used in different models were also excluded in these particular analyses.
We divided the population into 4 groups. The first group included those with no known diagnosis of diabetes but HbA1c ≥7%, an accepted threshold for the diagnosis of diabetes; the remainder were divided into approximate thirds of HbA1c with the use of clinically convenient cutoff points. Differences in established cardiovascular risk factors between the 4 groups were tested by the general linear model for means and by the χ2 test for proportions.
Cox proportional hazards regression models were used to estimate the relative risk (RR) of stroke for each group adjusted (1) for age and sex; (2) for age, sex, and other cardiovascular risk factors including BMI (continuous), smoking status (never, current, former), either systolic blood pressure or diastolic blood pressure or pulse pressure (by 10 mm Hg increase), total cholesterol (continuous), triglycerides (continuous), and alcohol consumption (g/d, continuous); and (3) for age, cardiovascular risk factors, and previous history of MI, with the lowest HbA1c group used as the reference category.
Because the resulting data suggested a threshold relationship at the HbA1c cutoff value of ≥7%, ie, those with previously undiagnosed diabetes, we further explored associations statistically by reanalyzing the RRs of stroke in those who had HbA1c of <7% by (1) 1% increase in HbA1c level and (2) HbA1c categories (<5%, 5% to 5.4%, 5.5% to 6.9%) in men and women separately.
In those who attended the first health examination (n=25 639), HbA1c levels were measured in 11 147 people. This was attributable to later introduction of HbA1c measurement in the first health examination with the later availability of funding for the project. After exclusion of participants with prevalent stroke, diabetes, and missing data for all other variables included in the present study, 10 489 men and women aged 40 to 79 years were included in these analyses. Characteristics of the study population are shown in Table 1. The mean (SD) HbA1c was 5.29% (0.71) in men without known diabetes and 5.23% (0.70) in women without known diabetes (P<0.0001).
At baseline, mean age, systolic and diastolic blood pressure, BMI, total cholesterol, LDL cholesterol, and triglycerides increased with increasing HbA1c concentrations. High-density lipoprotein cholesterol and alcohol consumption decreased with increasing HbA1c. The proportion of women and lifelong nonsmokers decreased from the lowest HbA1c group to the highest, whereas the proportion of current smokers rose with increasing HbA1c.
Of the 164 incident strokes identified over ≈88 652 total person-years (mean follow-up, 8.5 years), 52 (32%) were fatal. Stroke incidence was 7.5 times higher among men and women with HbA1c ≥7% than in the lowest HbA1c group, although these men and women were also older. After adjustment for age, sex, and cardiovascular risk factors including BMI, systolic blood pressure, smoking, cholesterol, triglycerides, alcohol consumption, and history of MI, the RR associated with HbA1c ≥7% was attenuated to 2.83 (95% CI, 1.40 to 5.74; Table 2). At concentrations of HbA1c <7%, there was no association between HbA1c and stroke. Exclusion of prevalent MI did not alter the findings materially. Similarly, using diastolic blood pressure or pulse pressure as covariates in the Cox regression model instead of systolic blood pressure did not change the results materially (data not shown).
Repeating the analyses after excluding the participants without known diabetes but who had HbA1c of ≥7% showed no significant relationship between HbA1c value and risk of stroke. The adjusted RR for every increase in HbA1c value of 1% was 1.01 (95% CI, 0.75 to 1.36). With the use of categorical cutoff points, the adjusted RRs for men were 0.90 (95% CI, 0.51 to 1.60) and 0.73 (95% CI, 0.41 to 1.31) for HbA1c categories of 5% to 5.4% and 5.5% to 6.9% compared with HbA1c <5%. The corresponding values for women were 0.61 (95% CI, 0.30 to 1.25) and 0.94 (95% CI, 0.50 to 1.77), respectively.
In contrast to the relationship between blood glucose and coronary heart disease, our results suggest that the relationship between usual blood glucose levels assessed by HbA1c and stroke is not linear in people without known diabetes. Among those without known diabetes at baseline study examination, only those with HbA1c concentrations >7.0%, who are likely to be people with undiagnosed diabetes, had a significantly increased risk of stroke. Our data are consistent with previous European, US, and Australian studies that have shown either no relationship between blood glucose in the nondiabetic range and stroke10–12 or increased stroke risk in people without diabetes only at the upper extreme of the blood glucose distribution.13–15 In the Whitehall study of 18 000 middle-aged men, only a 2-hour postchallenge glucose concentration above the 95th percentile (>5.4 mmol/L) of the nondiabetic distribution was associated with increased stroke mortality.13 In the British Regional Heart Study, nonfasting blood glucose in the upper 2.5% (>8.2 mmol/L) of the nondiabetic distribution was associated with an almost 2-fold increased risk of stroke compared with men in the lowest quintile of blood glucose after adjustment for other cardiovascular risk factors.14 In the Renfrew/Paisley Study, women (but not men) in the top 5% of the nondiabetic distribution of nonfasting glucose (>6.8 mmol/L) were found to be twice as likely to suffer a fatal stroke as the remaining 95%15; however, there was no clear relationship between incident stroke and blood glucose in either sex.16 In a number of studies, the increase in stroke risk at the upper extreme of the blood glucose distribution became nonsignificant after adjustment for multiple cardiovascular risk factors.17,18
The most recently reported data are from the Atherosclerosis Risk in Communities Study. The authors reported that although the adjusted RR of stroke increased with increasing tertile of HbA1c in adults without diabetes (n=10 886 participants, 167 incident strokes), this relationship was not statistically significant.19 Their findings are comparable to those of the present study.
We observed a markedly increased stroke risk in people with HbA1c ≥7% in our cohort. Increasing HbA1c concentrations were positively associated with classic stroke risk factors including blood pressure and cigarette smoking habit (Table 1). However, the excess stroke risk was still apparent after adjustment for all these factors. The group with HbA1c >7% largely comprises people with undiagnosed type 2 diabetes. Undiagnosed type 2 diabetes is common20 and may be present for several years before clinical diagnosis,21 and individuals are at increased risk of cardiovascular events during this time.22 Indeed, people with undiagnosed diabetes may have a higher cardiovascular risk than those with diagnosed diabetes, which could be attributed to increased treatment and lifestyle changes among people with known diabetes.
In the British Regional Heart Study, the excess stroke risk at the upper extreme of the nondiabetic blood glucose distribution became nonsignificant after individuals who developed diabetes during follow-up were excluded. However, people with undiagnosed diabetes who have a stroke are more likely to have their diabetes identified than undiagnosed diabetics who are event free. Perhaps only people with future risk of metabolic syndrome are prone to increased risk of stroke. Excluding people who are diagnosed with diabetes during follow-up could therefore have the effect of attenuating a true association between blood glucose and stroke in the nondiabetic population.
Various mechanisms have been proposed by which blood glucose might increase the risk of macrovascular disease, including impairment of NO-mediated endothelial dilation,23 increased oxidative stress by increased generation of free radicals24 or impaired vitamin C metabolism,25 changes in mechanical stresses on the vessel wall,26 and interleukin-mediated inflammatory mechanisms.27
In contrast to the majority of studies of the association between stroke and blood glucose that have been undertaken in middle-aged men,11,13,14, 17,18 the current cohort comprised large numbers of women and older persons. We have also been able to examine incident stroke rather than stroke mortality, and therefore the results are unlikely to be explained by excess case fatality associated with hyperglycemia. We also believe that the use of HbA1c provides a more reliable measure of usual glycemic status than the random glucose levels used in several previous studies.
In previous UK studies, the association between hyperglycemia and stroke appeared to be stronger in women than in men.16,28 Unfortunately, we did not have sufficient stroke events to examine sex differences, nor were we able to separately examine stroke subtypes because it is possible that the relationship between blood glucose level and stroke may vary according to stroke subtype and by sex. Stroke is a relatively uncommon event compared with coronary heart disease, and it is possible that with the limited number of events, the study lacked the power to detect a continuous relationship. The confidence limits are wide for the RR estimates below HbA1c 7%, and we cannot definitively exclude a continuous association with stroke. Nevertheless, these results are consistent with reports from previous epidemiological studies of no increased risk of stroke at levels of HbA1C below the threshold commonly used for diagnosis of diabetes.
In contrast to the relationship observed with coronary heart disease, which appears to be continuous over the whole distribution of HbA1C, the findings are more consistent with a threshold relationship between HbA1C levels and stroke risk. This threshold relationship is much more similar to the threshold relationships observed between blood glucose levels and microvascular complications of diabetes such as retinopathy and nephropathy.7
There are limitations to our study. Because of the requirement to provide detailed information and to be followed up long term, the initial response rate was modest at 40%. People who were severely ill or confined to bed were not able to participate in the study. Nevertheless, the study population from which the current sample was drawn is comparable to other UK population samples, with a slightly lower prevalence of smokers.8 Moreover, truncation of distribution will only attenuate the study findings, and it would not have an effect on the internal relationship between HbA1c level and stroke risk. We used ICD codes that include all stroke subtypes, including transient ischemic attacks. However, we used only hospital record linkage and death certificate data and excluded self-reported strokes, which are more likely to be transient ischemic attacks. Moreover, persons with transient ischemic attacks are not usually admitted to the hospital in the United Kingdom. Another limitation of our study is that because of the small number of events, we were not able to examine these relationships for each stroke subtype, which have different pathophysiologies. However, we examined the clinical stroke events severe enough to require hospitalization or leading to death, and this is of clinical relevance. We excluded ≈250 participants because of missing data. However, there is no significant difference between those who were included in the study and those with missing data in terms of age, BMI, systolic blood pressure, and cholesterol level. There may be other potential confounders that may have an effect on the relationship but were not adjusted for, such as the length of time of hyperglycemia.
In summary, although diabetes is an established risk factor for both coronary heart disease and stroke, the observed association between blood glucose levels and stroke risk appears to be much more consistent with a threshold relationship in contrast to the continuous linear relationship observed between blood glucose levels and coronary heart disease risk. These observations may provide insights into the differing pathogenesis of different vascular diseases.
We would like to thank participants and general practitioners who took part in the study. We also thank the staff of EPIC-Norfolk and our sources of funding.
Sources of Funding
EPIC-Norfolk is supported by research program grant funding from Cancer Research UK and the Medical Research Council with additional support from the Stroke Association, British Heart Foundation, Department of Health, Research Into Ageing, and Academy of Medical Sciences.
- Received September 1, 2006.
- Accepted September 29, 2006.
King H, Aubert RE, Herman WH. Global burden of diabetes, 1995–2025: prevalence, numerical estimates, and projections. Diabetes Care. 1998; 21: 1414–1431.
Coutinho M, Gerstein HC, Wang Y, Yusuf S. The relationship between glucose and incident cardiovascular events: a metaregression analysis of published data from 20 studies of 95,783 individuals followed for 12.4 years. Diabetes Care. 1999; 22: 233–240.
Khaw KT, Wareham N, Luben R, Bingham S, Oakes S, Welch A, Day N. Glycated haemoglobin, diabetes, and mortality in men in Norfolk cohort of European Prospective Investigation of Cancer and Nutrition (EPIC-Norfolk). BMJ. 2001; 322: 15–18.
Gerstein HC, Pouge J, Mann JF, Lonn E, Dagenias GR, McQueen M, Yusuf S, for the HOPE Investigators. The relationship between dysglycaemia and cardiovascular and renal risk in diabetic and non-diabetic participants in the HOPE study: a prospective epidemiological analysis. Diabetologia. 2005; 48: 1749–1755.
McCance DR, Hanson RL, Charles MA, Jacobsson LT, Pettitt DJ, Bennett PH, Knowler WC. Comparison of tests for glycated haemoglobin and fasting and two hour plasma glucose concentrations as diagnostic methods for diabetes. BMJ. 1994; 308: 1323–1328.
Friedewald WT, Levy RI, Fredrickson DS. Estimation of the concentration of low-density lipoprotein cholesterol in plasma, without use of the preparative ultracentrifuge. Clin Chem. 1972; 18: 499–502.
Folsom AR, Rasmussen ML, Chambless LE, Howard G, Cooper LS, Schmidt MI, Heiss G, for the Atherosclerosis Risk in Communities (ARIC) Study Investigators. Prospective associations of fasting insulin, body fat distribution, and diabetes with risk of ischemic stroke. Diabetes Care. 1999; 22: 1077–1083.
Haheim LL, Holme I, Hjermann I, Leren P. Nonfasting serum glucose and the risk of fatal stroke in diabetic and nondiabetic subjects: 18-year follow-up of the Oslo Study. Stroke. 1995; 26: 774–777.
Wannamethee SG, Perry IJ, Shaper AG. Nonfasting serum glucose and insulin concentrations and the risk of stroke. Stroke. 1999; 30: 1780–1786.
Hart CL, Hole DJ, Smith GD. Risk factors and 20-year stroke mortality in men and women in the Renfrew/Paisley Study in Scotland. Stroke. 1999; 30: 1999–2007.
Hart CL, Hole DJ, Smith GD. Comparison of risk factors for stroke incidence and stroke mortality in 20 years of follow-up in men and women in the Renfrew/Paisley Study in Scotland. Stroke. 2000; 31: 1893–1896.
Balkau B, Shipley M, Jarrett RJ, Pyorala K, Pyorala M, Forhan A, Eschwege E. High blood glucose concentration is a risk factor for mortality in middle-aged nondiabetic men: 20-year follow-up in the Whitehall Study, the Paris Prospective Study, and the Helsinki Policemen Study. Diabetes Care. 1998; 21: 360–367.
Harris MI, Klein R, Welborn TA, Knuiman MW. Onset of NIDDM occurs at least 4–7 yr before clinical diagnosis. Diabetes Care. 1992; 15: 815–819.
Jiang Y, Kohara K, Hiwada K. Association between risk factors for atherosclerosis and mechanical forces in carotid artery. Stroke. 2000; 31: 2319–2324.