Family History of Type 2 Diabetes Is Associated With Increased Carotid Artery Intimal-Medial Thickness in Mexican Americans

Study subjects were participants of the SAFHS, a population-based longitudinal family study designed to investigate the genetics of heart disease and its determinants. Probands (one person from each family who was initially contacted to enroll in the study) for the SAFHS were a random sample of Mexican Americans who resided in a single low-income census track and were recruited without regard to the health status of either themselves or their family members. During the baseline phase of the SAFHS conducted between 1992 and 1996, a total of 1,431 individuals aged ≥16 years were recruited. These included 42 probands, their spouses, and the first-, second-, and third-degree relatives of each. Details of the sampling design and recruitment procedures have been previously described (14). A 4- to 5-year follow-up examination of 860 of the original participants was conducted between 1996 and 1998. The present cross-sectional analyses are based on subjects participating in the follow-up examination and in whom carotid ultrasound assessments were performed.

Examinations were conducted in the morning following an overnight fast. Procedures relevant to this study included an anthropometric assessment, a 2-h glucose tolerance test, and an ultrasound measurement of the CCA to assess thickness of the intimal-medial wall. The systolic (first phase) and diastolic (fifth phase) blood pressures were measured to the nearest even digit using a random-zero sphygmomanometer (Hawksley-Gelman, London, England) on the right arm of the seated participant. Three readings were recorded for each individual, and the subject’s blood pressure was defined as the average of the second and third readings. The glucose tolerance test was based on a 75-g glucose equivalent load, after which glucose levels were determined from a blood sample obtained at fasting (time 0) and 2 h after the glucose challenge. Plasma glucose was measured using an Abbott V/P Analyzer (Abbott Laboratories, Abbott Park, IL). Serum insulin was measured using commercial radioimmunoassay kits (Diagnostic Products, Los Angeles, CA). Diabetes was defined according to the American Diabetes Association criteria as a fasting glucose ≥126 mg/dl (15). Subjects were also considered to have diabetes if they were currently taking antidiabetic medications. Serum insulin concentrations were measured by radioimmunoassay. Lipid and lipoprotein concentrations were measured from fasting blood samples using previously described methods (16).

IMT of the CCA was assessed by high-resolution ultrasound scanning. The scanning and reading protocols were identical to those used in the Cardiovascular Health Study (17). CCA IMT was measured as the mean of two far wall measurements. Tracings were read in three batches by two different readers at the central carotid ultrasound reading center. A batch effect was included as a covariate for all statistical analyses. CCA IMT data were transformed using the inverse function (1/CCA IMT) to achieve normality.

Baseline distributions of potential confounders, such as age and BMI, for the entire study were estimated. Mean levels of insulin and triglyceride concentrations were logarithm-transformed before analysis to remove skewness.

To estimate the heritability of CCA IMT, we used a variance component approach implemented in the SOLAR computer software (18). The heritability was computed by partitioning the phenotypic variance of a trait (V_P) into components that include the additive (polygenic) variance (V_A), the variance due to measured environmental risk factors (V_E), and the residual or nonshared environmental variance (V_R). The genetic heritability (h²) was estimated as the proportion of the total phenotypic variance that can be explained by the additive genetic variance, or h² = V_A/V_P. The estimation of h² follows basic quantitative genetic theory, which models the phenotypic covariances between two individuals in a pedigree as a function of their degree of biologic relatedness (kinship coefficients). Therefore, using the observed phenotypic covariance between relatives and the corresponding expected kinship coefficients, the additive genetic variance (and hence heritability) can be estimated. If a trait is “genetic,” then the degree of phenotypic correlation between relatives should be in proportion to their degree of biological relationship (i.e., phenotypic correlation between siblings should be greater than that between first cousins).

To examine the association between CCA IMT and other cardiovascular risk factors, mean levels of cardiovascular risk factors were compared between subjects in the top and bottom tertile of the CCA IMT distribution (e.g., mean BMI of individuals in the top tertile was compared with mean BMI of individuals in the bottom tertile). To account for the relatedness of study subjects, a likelihood approach was used in the analysis, whereby we computed the likelihood of the data given the pedigree structure (18).

As an initial analysis, we evaluated the impact of a parental history of diabetes on arterial wall thickness by comparing mean CCA IMT levels between subjects with and without a parental history of type 2 diabetes. The parental history was considered to be positive if either parent was found to have the disease at their SAFHS examination and negative if both parents participated in the SAFHS and were found by an oral glucose tolerance test to not have diabetes. If neither parent participated in the SAFHS or one parent had participated and was found to be nondiabetic, then the subject was excluded from this analysis.

We then used a more comprehensive measure of family diabetes burden that included affection status of additional family members. This family history score was designed to summarize the burden of diabetes in all older first-degree relatives of the participants. Scores were calculated by adding +1 for each affected parent and older sibling and −1 for each unaffected parent and older sibling. Unexamined family members did not contribute any information to the family history score. Thus, although excluded from the preliminary “parental history” analysis, individuals with one examined unaffected parent and one unexamined parent could be included in this analysis. This scoring is analogous to that described by Bonney in his Class D regressive models, in which each subject’s phenotype is dependent on the phenotype of his or her parents and all older siblings (19). In the context of our study, a significant effect of this family history score will occur only when the familial correlation between the discrete trait and the quantitative risk factor is not equal to zero.

To determine whether the family history score was correlated with CCA IMT, we modeled CCA IMT as a function of age, age squared, sex, and family history score. To achieve normality of CCA IMT, the inverse of CCA IMT was used as the dependent variable in the regression models. Additional models were constructed that included other known cardiovascular risk factors (e.g., BMI, lipid levels, blood pressure, glucose, and insulin) as additional covariates to determine whether correlations observed between IMT and diabetes family score were independent of these risk factors.

The significance of the family history score was assessed by the likelihood ratio test, which compares the likelihood of a full model (e.g., age, sex, and family history score) with that of a nested model (e.g., age and sex only). The likelihood ratio test compares the difference in the likelihoods between the full and the nested models. Two times the difference between the logarithm of the likelihoods of the two models is distributed asymptotically as a χ² statistic with degrees of freedom equal to the difference in number of parameters in the two models being compared. All analyses were performed using SOLAR computer software (18).

These analyses are based on 620 nondiabetic individuals from 24 large Mexican-American families. The numbers of nondiabetic individuals with CCA IMT measurements in these families ranged from 2 to 57.

Table 1 shows the clinical characteristics of the 620 nondiabetic individuals included in the analysis, stratified by sex. The mean age of both men and women in the present analysis was about 40 years. As shown by previous studies, the SAFHS population is, in general, more obese than the average U.S. white population, with mean BMI of 29.4 and 30.6 kg/m² in male and female participants, respectively.

The distribution of CCA IMT levels is shown for men and women according to age in Fig. 1. Mean values of CCA IMT are higher in men than in women (0.640 vs. 0.607 mm, P < 0.001). The heritability of CCA IMT after adjustment for age and sex was 16% (unadjusted heritability was 10 ± 5.6% [mean ± SE]; P = 0.05). Comparison of model likelihoods with that obtained from a nested model in which the heritability was constrained to be zero revealed this estimate to be statistically greater than 0% (P = 0.009), indicating evidence for a heritable component, albeit small, to CCA IMT in this population of nondiabetic Mexican Americans.

Table 2 shows mean levels of selected cardiovascular risk factors for men and women in the lowest and highest tertiles of the CCA IMT distribution. As expected, individuals with the thickest CCA IMT were older, more likely to be men, and more likely to be smokers. Furthermore, these individuals had significantly higher BMI, diastolic and systolic blood pressures, and fasting levels of total cholesterol, triglycerides, glucose, and insulin. Interestingly, their mean HDL level was not significantly lower than those in the lowest quartile of CCA IMT. Some of the differences in cardiovascular risk profile between those in the lowest and the highest tertiles of CCA IMT were accounted for by age and sex. However, BMI and diastolic and systolic blood pressures were still significantly higher in those within the highest tertile of CCA IMT even after adjustment for age and sex.

As a preliminary analysis, we characterized the nondiabetic subjects according to their parental history of diabetes. The parental history was considered to be positive if either parent was found to have disease at their SAFHS examination and negative if both parents participated in the SAFHS and were found by oral glucose tolerance test to not have diabetes. After adjustment for age and sex, there was a nonsignificant increase in CCA IMT among those with a parental history of diabetes compared with those without (0.592 vs. 0.595 mm, P = 0.84).

Next, we examined the association between cardiovascular risk factors and family history of scores of diabetes, which ranged from −9 to +4. The diabetes family history score was assigned based on nuclear family membership. These analyses included sibships ranging in size from 1 to 10 (130 sibships of size 1, 60 sibships of size 2, 41 sibships of size 3, 17 sibships of size 4, 8 sibships of size 5, 3 sibships of size 6, and 1 sibship each of sizes 8 and 10). Diabetes family history scores decreased modestly but in a linear fashion, with increasing sibship size (mean diabetes family score ranging from −0.65 for sibship size 1 to −4.00 for sibship size 6), a result that might be explained by the fact that small families do not have the opportunity to have large negative scores.

To provide a more clinically relevant aspect of this association, we divided subjects into those with relatively low diabetes family history scores (i.e., score ≤−3, n = 123) and those with relatively high scores (i.e., score ≥0, n = 159) and presented the average cardiovascular profile in these two groups, as shown in Table 3. As expected, these two extreme groups of individuals are quite different with respect to their cardiovascular risk profile. A higher diabetes family history score, reflecting a greater burden of diabetes in the family, is generally associated with adverse cardiovascular risk. To test for statistical associations between the full spectrum of family history score and each cardiovascular risk factor, diabetes family history score was regressed against each risk factor in the full set of 620 individuals. This analysis revealed that diabetes family history scores were significantly associated with BMI (P = 0.002) and systolic blood pressure (P = 0.021), even after adjustment for age and sex.

Figure 2 shows a scatterplot of CCA IMT measurements by diabetes family score. Individuals with higher diabetes family scores had thicker CCA IMT than those with lower scores. When diabetes family history score was treated as an ordinal variable, every unit increase was associated with an increase of 0.026 mm in CCA IMT (P < 0.001). This association was significant even after adjustment for age and sex (0.006 mm increase in CCA IMT for every point increase of diabetes family history score, P = 0.016).

We also examined the relationship between CCA IMT and extremes of diabetes family history score. To determine the extent to which differences in CCA IMT between those with the greatest and those with the least burden of diabetes in the family was explained by concurrent differences in cardiovascular risk profile, we conducted a series of regression analyses, adjusting for groups of potential risk mediators (Table 4). Adjusted for age and sex, the mean CCA IMT of those with the greatest burden was 0.049 mm thicker than the mean IMT of their nondiabetic counterparts with less diabetes burden in the family (model 1 in Table 4, P = 0.016). Further adjustment for smoking status did not significantly change this difference (model 2 in Table 4). In fact, additional adjustment for BMI, waist-to-hip ratio, systolic and diastolic blood pressure, and serum lipid levels did not alter this significant increase in CCA IMT found in nondiabetic Mexican Americans with a greater diabetes burden in the family compared with their counterparts who have less diabetes burden in the family. Lastly, we adjusted for fasting serum insulin, fasting plasma glucose, and 2-h postload plasma glucose. As shown in the results of model 6 (difference of 0.037 mm; P = 0.040), the effects of family history score on CCA IMT persisted and could not be attributed to differences between study subjects in insulin and glucose concentrations. Adjustment for the aforementioned cardiovascular risk factors accounted for only a 24% reduction in the excess thickening of CCA IMT observed in those with high family history score (from 0.049 to 0.037 mm).

Our results lead to the following conclusions. First, heritability of subclinical atherosclerosis as assessed by CCA IMT is low (∼16%) in the nondiabetic Mexican-American participants of the SAFHS. Second, subclinical atherosclerosis as measured by CCA IMT is increased in those with a greater burden of diabetes in the family. Moreover, this is observed even among subjects who themselves do not have diabetes. After adjusting for only age and sex, the mean CCA IMT in individuals with a high family history of diabetes is 0.049 mm higher than that of those with a lower family history score of diabetes. This excess in thickening may be clinically significant as previous studies have shown that a history of myocardial infarction is associated with an increase of 0.07 mm in carotid IMT (20). Third, although some of the excess thickening in the carotid artery in those with greater burden of diabetes in the family was attributed to the adverse cardiovascular risk profile present in these individuals, these “mediating factors” such as BMI, systolic and diastolic blood pressure, and serum lipid levels accounted for only 20% reduction of this excess thickening.

In contrast to our heritability estimate of 16% for CCA IMT in the SAFHS, previous studies have reported heritability estimates ranging from 30 to 64% (21–25). This difference was somewhat surprising; however, ours was a relatively young population (mean age ∼40 years) in whom genetic predispositions may not yet be detectable. Other possible explanations for these differences in heritability estimates include ethnic differences in the population studied, differences in site of IMT measurement, differences in prevalence of existing cardiovascular disease in the population (some families in other studies were selected for presence of existing cardiovascular condition), and measurement error in IMT assessment.

A number of studies have documented a three- to fivefold increase of cardiovascular disease in diabetic individuals compared with nondiabetic individuals (26). Our study suggests that this relationship precedes the clinical manifestations of diabetes because increased thickening of the carotid artery is present even in nondiabetic individuals who are at greater risk of developing type 2 diabetes, as indicated by a greater burden of family history of diabetes, compared with their counterparts with less genetic predisposition to diabetes. The source of this shared etiology is unknown. Although it is possible that a causal relationship could exist between exposure to elevated glucose and/or insulin resistance (as expected in “pre-diabetic” individuals) and atherosclerosis, the association between diabetes family history and subclinical atherosclerosis remained in multivariate analysis, even after adjustment for fasting glucose and insulin levels.

The current analyses, which indicate cross-familial correlations between diabetes in relatives and subclinical atherosclerosis, is consistent with the “common soil” hypothesis that diabetes and cardiovascular disease share common antecedents (13), possibly including specific genetic polymorphisms that influence the cosegregation of atherosclerosis and diabetes risk. One such common antecedent could be the insulin resistance syndrome (27,28). Several natural history studies have previously shown that many cardiovascular risk factors, including dyslipidemia, elevated blood pressure, and obesity, present long before development of overt type 2 diabetes (8–10,29). However, in our analysis, only ∼20% of the excess CCA IMT found in individuals with greater burden of diabetes family history could be attributed to other concurrent cardiovascular risk factors. The remaining increase in atherosclerosis seen in these individuals may have been mediated through impaired fibrinolysis (8), endothelial dysfunction (30,31), or increased inflammatory response, none of which was accounted for in our multivariate model. A recent report indicated that elevation of plasma markers of endothelial cell activation or inflammation and oxidative stress in type 2 diabetes may be distinct from and may not be totally explained simply by differences in the burden of atherosclerosis as assessed by carotid IMT (31).

Atherosclerosis could be a marker for the underlying inflammatory process that is contributing to the risk of type 2 diabetes in these susceptible individuals. One could hypothesize that a certain amount of inflammation that results in increased atherosclerosis has to accumulate before glucose metabolism becomes dysfunctional. This hypothesis is a complete reversal of the classic paradigm of the relationship between diabetes and atherosclerosis; however, it is supported by previous epidemiologic studies showing that markers of inflammation predict both incident diabetes and incident cardiovascular morbidities (32–36) and by the notion that insulin resistance and impaired β-cell function must be present simultaneously in diabetic individuals. In fact, type 2 diabetes has been hypothesized to be a disease of the immune system (37).

One of the limitations of this study is that it assesses only atherosclerosis in the common carotid artery, which represents only one segment of the arterial system; however, IMT measured this way has been validated in previous studies (38). In addition, there was possible misclassification of “burden of diabetes in the family” because diabetes status was assessed only in surviving family members and only at one time point. Finally, as previously mentioned, excess CCA IMT not explained in the adjusted model may also be due to either residual confounding from the existing cardiovascular risk factors included in the model or from other unmeasured cardiovascular risk factors, such as serum C-reactive protein.

Nonetheless, this study has several strengths. Although previous studies have shown increased carotid IMT in individuals with newly diagnosed diabetes (39) and in nondiabetic individuals who are at risk of developing diabetes such as those with impaired glucose tolerance (40), it is not clear if this common antecedent is due to shared genetic factors, shared environmental factors, or both. This study is one of the first to address this issue. Other strengths of this study include its large sample size, population-based design (participant of the SAFHS were not ascertained for any disease, specifically neither type 2 diabetes nor cardiovascular disease), and assessment of various cardiovascular risk factors. Furthermore, this study was able to use a family history score of diabetes, with diabetes status confirmed by glucose tolerance test in all family members, to capture more accurately and more precisely the burden of diabetes in the family compared with just parental history of diabetes. A virtue of this scoring function is that the scores are influenced largely by the “density” of disease burden in the family because information from both affected and unaffected relatives is included. This has an advantage over scores that are based on the number of affected relatives only, as those are heavily dependent on family size because larger families may have more affected individuals than smaller families simply by virtue of their size. Our scoring function places more weight on larger families because extremely high (or extremely low) scores cannot be generated from small and less informative families.

In summary, although the genetic contribution to CCA IMT is modest, increased subclinical atherosclerosis is observed in nondiabetic Mexican Americans with a positive family history of diabetes. Taken together, these results provide new evidence from a family study in support of the hypothesis that type 2 diabetes and atherosclerosis share common genetic risk factors. This finding has implications for both treatment and future candidate gene studies of type 2 diabetes. Treatments for atherosclerosis may very well benefit type 2 diabetes, and nontraditional candidate genes (for example, those not necessarily involved in insulin signaling) may also contribute to risk of type 2 diabetes.

This work was supported by grants PO1-HL45522 and RO1-AR43351 awarded by the National Institutes of Health. Support for the Frederic C. Bartter General Clinical Research Center was made available by Clinical National Institutes of Health Grant MO1-RR-01346.

The authors are deeply grateful for the cooperation of the families participating in the SAFHS.