The Relationship Between Dysglycemia and Atherosclerosis in South Asian, Chinese, and European Individuals in Canada: A randomly sampled cross-sectional study

SHARE was a population-based study of carotid AS and its determinants in 985 individuals of South Asian, Chinese, and European origin living in Canada for ≥5 years. Detailed descriptions of the protocol, sampling frames, and major results have been published (11,12). Briefly, a database of South Asian and Chinese surnames in the telephone directories of Hamilton, Toronto, and Edmonton, Canada, was randomly sampled. Selected households were mailed an introductory letter and then contacted by phone. Eligible individuals of the appropriate ethnicity (defined by their ancestral origins) aged 35–75 years who had lived in Canada for ≥5 years were invited for assessment. Once a South Asian or Chinese participant agreed to attend, a European household was randomly sampled from all of the other surnames within the same geographic postal code. The results in 979 people in whom glucose tolerance status could be determined are reported here.

After providing written informed consent, participants had a detailed clinical assessment and provided urine and fasting blood samples. Individuals with no history of diabetes also had an oral glucose tolerance test. All assays were performed centrally without any clinical information. Glucose and lipids were measured as previously described (11). HbA_1c was measured by low-pressure cation exchange chromatography on a Ciba Corning 765 Glycomat analyzer and a Glucamal reagent kit (Drew Scientific, London, U.K.). Insulin was measured by a highly specific radioimmunoassay, with cross-reactivity with only proinsulin of ∼40% at mid-curve (Coat A Count; Diagnostic Products, Los Angeles, CA). Free fatty acids were measured by an assay that detects fatty acids of 6–20 carbon lengths (NEFA C; Wako Chemical, Richmond, VA). The arithmetic approximation to the homeostasis model assessment (HOMA) for β-cell secretion (HOMA-β) and for insulin resistance (HOMA-IR) were calculated from the fasting plasma glucose and insulin levels as follows (13): HOMA-β = (20 × fasting insulin [mU/l])/(fasting plasma glucose [mmol/l] − 3.5) and HOMA-IR = (fasting insulin [mU/l]) × (fasting plasma glucose [mmol/l]/22.5).

All participants underwent carotid B-mode ultrasonography as previously described (11). The mean of the maximum intimal media thickness (MMIMT) was calculated by averaging the segment maximum IMT measurements from the far and near walls of the left and right common, bifurcation, and internal carotid artery segments. All scans and off-line measurements were performed by sonographers blinded to the participants’ other data.

All analyses were performed with SAS (Version 6.12) and (unless indicated otherwise) were age- and sex-adjusted to the whole SHARE population. χ² testing was used to compare proportions; ANCOVA was used to analyze continuous variables, and the Tukey-Kramer test was used for pairwise comparisons. Logistic regression and multiple regression models were used to assess independent determinants of abnormal glucose categories and MMIMT, respectively. HbA_1c deciles were used to divide the study groups, and the mean HbA_1c within each decile (for all participants) was plotted against the adjusted MMIMT.

Data were collected from 979 of the 985 South Asian (n = 340, 45% female), Chinese (n = 315, 49% female), and European (n = 324, 52% female) participants who either completed a glucose tolerance test or had a diabetes history. There were no differences in sex. However, European participants (mean age 51.3 years) were significantly older than South Asian (mean age 49.4 years, P for comparison = 0.01) and Chinese participants (mean age 47.8 years, P for comparison <0.0001), and South Asian participants were significantly older than Chinese participants (P = 0.04). Age- and sex-adjusted MMIMT values for these three groups have been reported previously (11).

Individuals with higher degrees of glucose intolerance were older and more likely to be of South Asian ethnicity than individuals with normal glucose tolerance (Table 1). Indeed, the age- and sex-adjusted HbA_1c levels (± SD) in the subjects of South Asian, Chinese, and European ethnicity were 5.90 ± 0.86, 5.63 ± 0.87, and 5.43 ± 0.86%, respectively (F = 29.13 and P < 0.0001 for differences among all three groups; P = 0.0003 for South Asian vs. Chinese; P < 0.0001 for South Asian vs. European; and P = 0.077 for Chinese vs. European). After adjusting for age and sex, individuals with higher degrees of glucose intolerance were also more likely to have a history of treated hypertension (P < 0.0001); a history of hyperlipidemia (P = 0.02); higher blood pressure, BMI, waist-to-hip ratio, fasting insulin, estimated insulin resistance, free fatty acids, and triglycerides; and lower HDL cholesterol and estimated β-cell function (P < 0.0001 for all) (Table 1).

When these risk factors were analyzed in a multivariate model (Table 2), the independent determinants of any abnormal glucose category (i.e., either impaired fasting glucose, impaired glucose tolerance, or diabetes) were ethnicity (specifically the South Asian compared with the Chinese group), age, waist-to-hip ratio, and the natural logarithm of triglycerides and fasting free fatty acids. Insulin secretion and insulin resistance (by HOMA) were highly correlated with each other and with insulin levels; when tested in separate models that included the other variables in Table 2, they were also strong independent determinants of any abnormal glucose category (P < 0.0001) (Table 3).

The relationship between metabolic CV risk factors and AS was explored by analyzing the relationship between ambient glycemia (i.e., HbA_1c), determinants of abnormal glucose tolerance, and MMIMT. After age and sex adjustment (Table 4), significant determinants of MMIMT included HbA_1c and fasting plasma glucose (P < 0.0001 for both), HOMA-β (P = 0.001), BMI (P = 0.03), and systolic blood pressure (P < 0.0001). MMIMT was not associated with postload glucose, HOMA-IR, any lipid abnormality, fasting free fatty acids, fasting insulin, abdominal obesity, or diastolic blood pressure (Table 4).

Further analyses focused on the relationship between HbA_1c and the carotid MMIMT. First, the whole study population was divided into 10 equally sized groups by HbA_1c decile. As shown in Fig. 1A, there was a progressive curvilinear relationship between age- and sex-adjusted MMIMT and HbA_1c. Although this relationship was most apparent as the HbA_1c level rose above 5.7%, no threshold was detectable using a nonlinear statistical analysis. Second, a similar statistically significant relationship was noted when participants of South Asian (P = 0.018), Chinese (P = 0.002), and European (P < 0.0001) descent were analyzed separately using multiple regression techniques (Fig. 1B). Moreover, the strong significant relationship between HbA_1c and MMIMT persisted after adjusting for ethnicity in the multiple regression analysis (P < 0.0001) (Table 5), with no evidence of any heterogeneity between ethnicities (P for interaction between HbA_1c and ethnicity = 0.07). Third, several other regression models were tested to determine whether the HbA_1c relationship could be explained by confounding with related variables. As noted in Table 5, a strong and robust relationship between MMIMT and HbA_1c was noted in different models that adjusted for diabetes status, abdominal obesity, insulin resistance, insulin secretion, fasting free fatty acids and/or dyslipidemia (P < 0.004 for all models).

Finally, to determine the relative importance of any diabetes (i.e., either new or established) versus HbA_1c as a determinant of AS, both of these variables were assessed separately and then together in the same regression model. After adjustment for age, sex, and ethnicity, both diabetes (P = 0.002) and HbA_1c (P < 0.0001) were significant determinants of MMIMT when they were included in separate models. However, when they were included together in the same model, HbA_1c (P < 0.0001) but not diabetes (P = 0.6) was a significant determinant of MMIMT (Fig. 2). This finding was confirmed when the analysis was repeated in the subset of individuals without evidence of diabetes; as in the overall group, HbA_1c remained an independent determinant of MMIMT after adjustment for age, sex, and ethnicity (P = 0.029).

These data show that the degree of glycemia as measured by HbA_1c is a strong independent determinant of AS. They demonstrate that the relationship between HbA_1c and AS is similar in different ethnic groups and cannot be accounted for by differences in abdominal obesity, dyslipidemia, free fatty acids, insulin secretion, or insulin resistance. The fact that the HbA_1c-AS relationship exists after controlling for the presence of diabetes, and the fact that there is no relationship between diabetes and MMIMT after controlling for HbA_1c (Fig. 2), suggests that the relationship between diabetes and AS is accounted for by the relationship between glucose levels and AS, and not by the presence or absence of diabetes per se.

These data also confirm that diabetes and glucose intolerance are strongly associated with other risk factors for CV disease, including age, abdominal obesity, hypertriglyceridemia, high free fatty acids, reduced insulin secretion, and increased insulin resistance. Furthermore, they confirm that 1) the prevalence of diabetes and glucose intolerance clearly differs between ethnic groups, and 2) the relationship between ethnicity and an abnormal glucose category is independent of the presence of other measured CV risk factors.

A few other studies have examined the relationship between diabetes, glucose, insulin resistance, and the anatomic extent of AS. Although people with diabetes were shown to have a higher IMT than nondiabetic people (14), significant associations between glucose levels, insulin levels, and carotid AS (15), or between insulin resistance and carotid AS (16–19), have not been consistently observed. Unfortunately, HbA_1c was not measured in these previous studies. Indeed, in at least two studies in which HbA_1c was measured, it was noted to be an important risk factor for carotid AS (20,21). These latter studies, as well as recent evidence that a period of glucose lowering with intensified insulin therapy reduced progression of carotid AS in people with type 1 diabetes who were in the Diabetes Control and Complications Trial (22), provide further support for the importance of HbA_1c as a powerful (and potentially modifiable) risk factor for AS.

The HbA_1c level reflects the fasting and postprandial glucose levels during a 2–3 month window (23). The fact that it was a strong determinant of AS suggests that it is to be preferred over other markers of dysglycemia (such as postload glucose) and may be the glycemic measure of choice when assessing risk for AS in both clinical practice and epidemiological research. The absence of a significant relationship between 2-h postload glucose and MMIMT, despite a strong relationship between AS and both HbA_1c and fasting glucose, may have been due to at least three possibilities. First, people with known diabetes did not have glucose tolerance testing in this study. Therefore, postload glucose measures were not available for these individuals, who presumably had the highest levels and the highest MMIMT. Second, few people with normal glucose tolerance have elevated fasting plasma glucose levels; for example, in one study of randomly selected patients in the U.S. (10), only 7.9% of 2,142 people with normal glucose tolerance had a fasting glucose level ≥6.1 mmol/l, whereas 33% of 2,932 people with normal fasting glucose levels already had a postload glucose level ≥7.8 mmol/l (10). Therefore, in nondiabetic individuals, elevated fasting plasma glucose is likely to reflect more advanced and perhaps longer-duration dysglycemia than an elevated postload glucose level, and it may therefore be more closely linked to MMIMT. Third, the high intrasubject variability of postload glucose levels may have obscured any relationship with MMIMT.

These data reflect a strong relationship between dysglycemia and AS that is independent of the presence or absence of diabetes. Moreover, the 0.032-mm difference in MMIMT that was noted per 0.9% rise in HbA_1c (after age, sex, ethnicity, and diabetes adjustment) (Table 5) is similar to the 0.036-mm difference in MMIMT observed between individuals taking 10 mg of ramipril and placebo during 4.5 years of follow-up in a substudy of the Heart Outcomes Prevention Evaluation Study (24,25). The fact that this study showed that ramipril reduced the risk of CV events by 22% suggests that modest differences in HbA_1c may reflect clinically relevant differences in CV risk.

Finally, these data do not explain why rising glucose levels are related to AS; nevertheless, they clearly support the inclusion of HbA_1c level in the list of AS risk factors in all people (not just those with diabetes). They also provide support for the hypothesis that therapies that lower glucose levels in people with even modestly elevated levels may reduce the risk of AS.

This study was funded by the Medical Research Council of Canada (MRC Grant MT-13734) and by Merck Frosst Canada. H.C.G. holds the Population Health Institute Chair in Diabetes Research (sponsored by Aventis); S.A. holds the Eli Lilly Chair in Women’s Health and holds a Canadian Institutes of Health Research Clinician Scientist award; and S.Y. holds a Heart and Stroke Foundation of Ontario Chair in Cardiovascular Research and is a Senior Scientist of the Canadian Institutes of Health Research.