Incidence of Diabetes in American Indians of Three Geographic Areas: The Strong Heart Study

The design and methods of the SHS have been described elsewhere (2). At the time of the baseline examination (1989–1991),∼1,500 tribal members, aged 45–74 years, were recruited from each of the three centers: Arizona (AZ), Oklahoma (OK), and South/North Dakota (SD/ND). A personal interview was conducted to obtain demo-graphic data, personal medical history, health habits, and family history of CVD and diabetes. A physical examination was performed, a fasting blood sample was collected for measurement of lipids, lipoproteins, apolipoproteins, insulin, glucose, and glycated hemoglobin, and a morning urine sample was collected for measurements of creatinine and albumin. The 2-h glucose tolerance test was performed only if the participant was not taking insulin or any oral hypoglycemic medication for known diabetes and if the participant had a fasting glucose value <225 mg/dl as determined using Acucek II (Baxter Healthcare, Grand Prairie, TX).

Participants were classified into five diabetes status groups at the baseline examination: known diabetes, new diabetes, impaired glucose tolerance (IGT), normal glucose tolerance (NGT), and diabetes status undetermined. Detailed criteria for each group have been described elsewhere (1). Criteria used for newly identified diabetes, IGT, and NGT followed the 1985 World Health Organization recommendations (3,4).

In 1993–1995, the surviving participants were invited to a second personal interview and physical examination, which were similar to those of the baseline examination. Several additional procedures were added to the second examination, including an echocardiogram, a pulmonary function test, and a 24-h-recall dietary survey. The average follow-up time between the baseline and second examinations was ∼4 years.

Participants were classified as an incident cases of diabetes if they were classified at baseline examination as NGT or IGT and if they satisfied the criteria for diabetes at the second examination. The 4-year cumulative incidence was obtained by dividing the number of incident diabetes cases by the number of participants at risk for diabetes at baseline. Baseline characteristics considered as possible risk factors include age, sex, family history of diabetes, degree of Indian blood, BMI, waist-to-hip ratio, percentage of body fat, physical activity level, hypertension status, fasting plasma glucose, 2-h glucose, HbA_1c, fasting insulin, fibrinogen, total cholesterol, HDL cholesterol, LDL cholesterol, total triglyceride, albumin-creatinine ratio, apoprotein A1, and apoprotein B. Statistical methods used included the χ² test, the Student’s t test, analysis of variance with Tukey’ s method of multiple comparison (5), and stepwise linear logistic regression methods. When a continuous variable had a skewed distribution (e.g., triglyceride), logarithmic transformation was used to stabilize the variance (6,7). A statistically significance level of 0.05 was used unless otherwise indicated.

Of the 4,549 participants examined at baseline, 413 died during the follow-up period, 498 were lost to follow-up, and 3,638 (1,206 in AZ, 1,255 in OK, and 1,177 in SD/ND) returned for the second examination (1,384 men and 2,254 women). Participants who died had significantly higher mean age (59 vs. 56 years), higher average degree of Indian blood (93.3 vs. 87.7%), higher average fasting plasma glucose (164.7 vs. 151.3 mg/dl), and higher rate of albuminuria (51 vs. 29%) but lower BMI (29.9 vs. 31.1 kg/m²) at baseline than those who participated in the second examination. Compared with the follow-up participants, those lost to follow-up had significantly lower mean log(insulin) (2.66 vs. 2.78) and BMI (30.3 vs. 31.1 kg/m²).

Glucose tolerance status was determined in 3,401 (93%) of the 3,638 participants who underwent the second examination. The outcome of study participants at the second examination is summarized in Table 1. Overall, of the 1,664 participants who were free of diabetes at baseline examination and were reexamined, 326 (19.6%) developed diabetes during the 4-year follow-up period (average annual incidence rate 4.9%). Specifically, 128 (11.3%) of the 1,132 participants with NGT developed diabetes (average annual incidence rate 2.8%). Among those who had IGT at baseline, 37% (198/532, average annual incidence rate 9.3%) progressed to diabetes and 27.6% (147/532, average annual rate 6.9%) reverted to NGT. During the 4 years, a small fraction of the participants who were initially diagnosed as having diabetes but who were not taking medication reverted to IGT (2.9%) and NGT (2.0%). Among the 46 participants who reverted to IGT, 43 were newly diagnosed diabetes cases, and 3 were known to have diabetes at the baseline examination. Of the 33 diabetic patients who reverted to NGT, 26 were new diabetes cases diagnosed by the SHS at baseline examination, and 7 were known diabetics at baseline. No substantial changes were found in BMI, percentage of body fat, or waist-to-hip ratio in these participants.

The incidence rates of diabetes by center, gender, and baseline diabetes status are shown in Fig. 1. The overall 4-year incidence rate was 19.7%. In men, ∼9% (average annual rate 2.3%) of those who had NGT developed diabetes during the 4-year period. Incidence rates were higher among women than men. The differences between centers were not statistically significant in either men or women. Among the participants who had IGT at baseline, the overall rates of conversion to diabetes were 38.6% (average annual rate 9.7%) and 36.6% (average annual rate 9.2%) in men and women, respectively. The AZ men had a higher conversion rate (52.8%) than OK and SD/ND men, but the difference was significant only between AZ and SD/ND. In women, the conversion rate from IGT to diabetes was not significantly different among the three centers. When the participants with NGT and IGT at baseline were combined, diabetes developed in 16.2% (average annual rate 4.1%) of the men and 22.2% (average annual rate 5.6%) of the women during the 4-year follow-up period. Among the three centers, AZ men had significantly higher 4-year incidence rates of diabetes than OK and SD/ND men. In women, AZ had the highest 4-year incidence rate of diabetes, which was followed by SD/ND and OK. However, only the difference between AZ and OK was statistically significant.

The cumulative incidence rates of diabetes among participants with NGT at baseline were significantly higher in those with higher baseline values of BMI, waist-to-hip ratio, fasting glucose, and fasting insulin as well as in those with albuminuria in both men and women (Table 2). Degree of Indian blood was significantly associated with the development of diabetes only in women. In men only, percentage of body fat and total triglyceride levels were significantly positively related to the development of diabetes, whereas HDL cholesterol and apoprotein A1 were inversely related to risk. Among participants who had IGT at baseline, those with 100% Indian blood and higher 2-h glucose, fasting insulin, and HbA_1c had significantly higher rates of conversion to diabetes in both men and women. Fasting glucose and albuminuria were significantly associated with conversion to diabetes in women only, and BMI and percentage of body fat were significantly associated with conversion to diabetes in men only.

Variables included in the logistic regression analysis were age, BMI, percentage of Indian blood, fasting plasma glucose, 2-h glucose, fasting insulin, total triglyceride, and albuminuria (present or absent). These variables were chosen for the multivariate analysis because of their significance in the univariate analysis. Stepwise selection procedure was used to select those variables that were significantly associated with incident diabetes after adjusting for the other variables. The variables that were identified as significantly associated with the development of diabetes are shown in Table 3 by sex and baseline glucose tolerance status.

In men, the most predictive variables for development of diabetes in participants with NGT were BMI and triglycerides. The odds ratios indicated that for every 0.6-mg/dl increase in log(triglyceride) and every 5-unit increase in BMI, the risk of developing diabetes in men increased 40 and 38%, respectively. In women with NGT, fasting insulin was the only significant variable related to the risk of developing diabetes. For every 0.7-μU/dl increase in ln(insulin), the risk for women to develop diabetes increased 91%. Similar results were obtained when 2-h glucose was excluded in the analysis.

For men with IGT at baseline, the significant predictive variables were BMI, 2-h glucose, and percentage of Indian blood. For every 5-unit increase in BMI and every 27% increase in Indian blood, the risk of converting to diabetes increased 90 and 74%, respectively. However, for every 34-mg/dl increase in 2-h glucose, the risk increased nearly fourfold. BMI and percentage of Indian blood were the only significant variables identified by the regression method when 2-h glucose was excluded. For women with IGT, fasting glucose, 2-h glucose, and albuminuria were the most significant independent predictors. A 12-mg/dl increase in fasting glucose increased the risk of diabetes by 54%. However, the presence of albuminuria or a 34-mg/dl increase in 2-h glucose increased the risk of diabetes by 2.4-fold. When 2-h glucose was excluded in the regression analysis, fasting insulin was identified as a significant risk factor, along with fasting glucose and the presence of albuminuria.

When all of the participants with NGT and IGT at baseline were combined, IGT, percentage of Indian blood, fasting glucose, 2-h glucose, fasting insulin, and presence of albuminuria were found to be positively related to the development of diabetes, and age was negatively related. Sex was not a significant variable (Table 3).

The prevalence of diabetes has increased in the U.S., particularly in some minority populations, such as Mexican-Americans and American Indians (8,9). The high incidence rates found in the SHS further confirmed that diabetes is one of the most severe health problems in this population. Nearly all of the diabetic patients had type 2 diabetes. The incidence rates in the SHS seem to be reasonably comparable to those reported earlier in the Pima Indians (10,11) but much higher than those reported in other populations, for example, middle-aged Swedish men, Mexican-American and non-Hispanic whites, and Korean men and women (12,13,14).

Similar to previous reports (15,16,17,18,19,20), individuals with IGT had a higher risk of developing diabetes than individuals with NGT. In reviewing results from several population based studies, Harris (16) found that rates of conversion from IGT to type 2 diabetes ranged from 1.5% per year in Bedford, U.K. to 7% in both U.S. Mexican-Americans in Colorado and Japanese-American men (17,18,19,20,21,22). The overall average annual rate of conversion from IGT to diabetes found in the SHS was an alarming 8.4% (almost 10% in men), which exceeds all of these reported conversion rates. However, compared with a South-African Indian cohort (annual rate 12.6%), our conversion rates were lower (23).

Results from several longitudinal studies have shown that fasting glucose or glucose tolerance is a significant predictor of the development of diabetes (12,14,22,23,24). In the SHS, baseline fasting glucose was an independent predictor of diabetes after adjusting for other variables only for women with IGT. Two-hour glucose levels were an independent predictor of conversion to diabetes from IGT in both men and women. Hyperinsulinemia was common among individuals in whom diabetes developed. In our study, fasting insulin, which is frequently used as a measure of insulin sensitivity, was found to be an independent predictor of diabetes risk in women with NGT. Moreover, fasting triglyceride, a reflection of insulin resistance, was predictive in men with NGT.

The Whitehall Study (24) showed that high triglyceride levels were predictive of diabetes in individuals with impaired glucose tolerance. In the SHS, we found triglyceride to be an independent predictor of diabetes in men with NGT. Obesity, measured by BMI or waist-to-hip ratio, has been found to have significant predictive power in several studies (10,12,13,25). In our study, BMI was a significant independent predictor of diabetes among men with NGT or IGT at baseline. Age was found to have an inverse association with risk of diabetes. This may be because many of the older participants had died during the follow-up period and diagnosis of type 2 diabetes occurred most often in the fourth and fifth decades of life in this population (1). We also found that women with IGT and albuminuria had a nearly threefold increased risk of developing diabetes than those without albuminuria.

In conclusion, this study, although limited to adults aged 45–74 years, demonstrates that diabetes, already an epidemic in the Native-American population, continues to increase. The incidence rates of diabetes in this population were several times higher than in other ethnic groups. Recent data also showed that type 2 diabetes is increasingly diagnosed among children and adolescents in this population (26). Regular screening of plasma glucose, insulin, and albuminuria is necessary to ensure early detection and timely treatment. The high progression rates from IGT to diabetes were particularly alarming. A recent Finnish study (27) of individuals with IGT concluded that reducing weight and increasing physical activity could prevent type 2 diabetes. Therefore, patients with IGT should be put on a rigid weight-reduction program including diet and exercise or should be treated with diabetes medication to reduce the risk of progression to diabetes.

This study was supported by Grants U01 HL-41642, U01 HL-41652, and U01 HL-41654 from the National Heart, Lung and Blood Institute.

We thank the 13 participating American-Indian tribes/communities, Indian Health Service, and participants for their support and assistance; the SHS field center coordinators Beverly Blake and Alan Crawford of SD/ND, Betty Jarvis of AZ, and Dr. Tauqeer Ali of OK; and the SHS staff for collecting the data.