The Effect of Indian or Anglo Dietary Preference on the Incidence of Diabetes in Pima Indians

RESEARCH DESIGN AND METHODS

An age- and sex-stratified sample of 575 Pima Indians aged 18 to 74 years underwent a dietary survey in 1988 (4). The present study included the 165 subjects (77 men and 88 women) who did not have diabetes at the time of the survey, as ascertained from glucose tolerance tests during a research examination within 1 year of the dietary assessment. For the present analysis, this examination was considered as the baseline examination. Women who were pregnant or <6 weeks postpartum at the time of the dietary survey were excluded. The subjects were examined periodically between June 1988 and June 1999. Each subject included in this study had at least two examinations. Examinations included a glucose tolerance test with determination of the glucose concentration in venous plasma drawn after an overnight fast and 2 h after a 75-g oral carbohydrate load. Diabetes was defined according to the criteria of the World Health Organization (8). The date of diagnosis was determined from research examinations or from review of clinical records if diabetes was diagnosed in the course of routine medical care.

Blood pressure was measured to the nearest 2 mmHg with a mercury sphygmomanometer while the subject rested in the supine position. Diastolic blood pressure was measured at the fourth Korotkoff sound. BMI was computed as the weight in kilograms divided by the square of height in meters.

Plasma glucose concentrations were measured with an autoanalyzer by the potassium ferricyanide method until 1991 and by the hexokinase method thereafter (9). HbA₁ was measured by electrophoresis (10) at baseline, and thereafter HbA_1c was measured by high-performance liquid chromatography.

Leisure-time physical activity was measured in a subsample (n = 139) of the study population with a modified activity questionnaire developed for use in this population (11,12,13). The questionnaire recorded the frequency and duration of activities that require energy expenditure greater than the activities of daily living, i.e., bathing, grooming, and feeding. The frequency and duration of these activities were weighted by their relative intensity (metabolic equivalent) (14) to produce an estimate of total leisure-time physical activity for the previous year (12).

Parents of many subjects were examined as part of the longitudinal diabetes survey, and the presence of diabetes in the parent at the time of the dietary interview was determined from these examinations. For each subject, the number of siblings was computed as the number of offspring from the same mother.

Dietary intake was assessed by a 24-h food recall and a quantitative food frequency assessment, as previously described (4). Results of the quantitative food frequency assessment were used in this study because they correlate most closely with the total energy expenditure measured with the doubly labeled water method (15). Interviews were conducted by trained Pima women and supervised by a research dietitian using standard methods recommended by the Nutrition Coordination Center of the University of Minnesota, Minneapolis, Minnesota (16). The quantitative assessment had an open format and inquired about the subject’s usual dietary intake. This method permits a more detailed estimate of the portion size than a self-administered food frequency questionnaire (17). Subjects were also asked to classify their diet as being mainly “Indian,” “Anglo,” or “mixed.” The Indian diet has been described in detail by Reid et al. (18), which reflects the influence of popular Sonoran-style foods common in the Southwest and includes dishes such as tortillas, tacos, tamales, chili stew, menudo, and chorizo sausage, as well as the traditional desert foods, and the melons and squash introduced by the Spanish in the late 17th century (19). Legumes are a staple in this diet, as pinto beans are combined with many food items (4). The Anglo diet, on the other hand, reflects a preference for a Western diet typical of the remaining regions of the U.S. Energy and nutrient intake was calculated using the University of Minnesota Nutrient Calculation System and database (version 1.3, released June 1989) developed by the University of Minnesota Nutrition Coordinating Center (16). The database was expanded to include specific American-Indian foods (19). Carbohydrates were subdivided into simple (sucrose, galactose, glucose, fructose, and lactose) and complex (starch and dietary fiber) carbohydrates.

The studies were approved by the review boards of the National Institute of Diabetes and Digestive and Kidney Diseases and by the Tribal Council of the Gila River Indian Community. Each participant gave informed consent.

Statistical analysis

Data analysis was performed using Stata software (20). Student’s t and χ² tests were used to test differences in means and proportions. Fisher’s exact test (21) was used to test differences in proportions when there were less than five expected observations in any cell. Analysis of variance was used to test differences between the three dietary groups. The effect of dietary pattern on the development of type 2 diabetes was examined using a Cox proportional hazards model to control for the effects of potentially confounding variables. A test for trend (P trend) across the different dietary preference categories was obtained by coding the Indian diet as 0, the mixed diet as 1, and the Anglo diet as 2 in the proportional hazards model. Follow-up time was computed from the date of the baseline examination to the date of the diagnosis of diabetes or the last examination before June 1999. Age at dietary interview was entered as a continuous variable and as a quadratic term. The quadratic term was not significant and was not included in the final model. BMI and total energy intake were included as standardized variables. Leisure-time physical activity was dichotomized at the median for each sex. Product terms among predictor variables were not examined because of the limited number of cases. The adequacy of the proportional hazards model was assessed using the method proposed by Grambsch and Therneau (22). In this method, a generalized linear regression of the scaled Schoenfeld residuals on functions of time is performed, and a nonzero slope (rejection of the null hypothesis) indicates that the log hazard ratio function was not constant over time. All variables satisfied the proportional hazards assumption.

RESULTS

Of the 165 subjects without diabetes at baseline, 49 classified their diet as Indian, 95 as mixed, and 21 as Anglo. The mean duration of follow-up was 6.2, 6.3, and 6.0 years in the Indian, mixed, and Anglo groups, respectively. The clinical and demographic characteristics of the population at baseline are shown in Table 1 according to self-reported dietary pattern. Blood pressure, BMI, waist circumference, and glucose tolerance were similar in all groups. Weight and height were also similar in all groups (data not shown). At baseline, there were no significant group differences in the proportion of subjects living on or off the reservation, the number of siblings in the family, or the parental family history of diabetes. Leisure-time physical activity was highest in the Indian-diet group and lowest in the Anglo-diet group, but the differences in physical activity levels were not statistically significant (Table 1).

Among women, complex carbohydrates (P < 0.05), dietary fiber (P = 0.05), insoluble fiber (P = 0.04), and vegetable proteins (P = 0.04) were significantly higher in the Indian-diet group than in the Anglo-diet group, and the values for the mixed-diet group were intermediate (Table 2). Fat and protein contents of the three diets were similar. Differences in complex carbohydrate and vegetable proteins remained significant after expressing the nutrients as a percentage of total caloric intake (Table 2). However, the fiber content of the three diets did not differ significantly when expressed as a fraction of total energy intake. In men, there were no significant differences in nutrient intake by dietary preference, although there was a tendency for the Anglo diet to be lower in complex carbohydrates and higher in alcohol consumption when both were expressed as a fraction of total energy intake. At the end of the study, the mean fasting, 2-h glucose, and HbA_1c concentrations were higher, but not significantly, in Anglo-diet subjects than in Indian-diet subjects, and concentrations were intermediate in mixed-diet subjects. Systolic and diastolic blood pressures were slightly higher, but not significantly, in Indian-diet subjects compared with those of the two other groups (Table 3).

Diabetes developed in 36 (22%) of the subjects (8 men and 28 women) during a mean follow-up of 6.2 years (range 0.9–10.9). Subjects reporting an Anglo diet were more likely to develop diabetes over the study period (Table 4). The incidence of diabetes in the Indian, mixed, and Anglo dietary groups were 23, 35, and 63 cases per 1,000 person-years, respectively. The risk of developing diabetes in Anglo-diet subjects was 2.9 times (95% CI 1.1–8.1) as high as that of Indian-diet subjects (Table 4). Compared with the Indian diet, the unadjusted sex-specific hazard of diabetes in the Anglo diet was 3.3 (0.9–11.5) in women and 1.6 (0.2–16.1) in men. Additional adjustments for age, baseline BMI, total caloric intake, and sex did not substantially alter the point estimate (2.5; 0.9–7.2), but the CI included 1. After adjustment for age, BMI, and total energy intake, the risks were 3.0 (0.8–10.7) in women and 1.4 (0.1–16.6) in men.

Further adjustment for physical activity in the subset of subjects with physical activity data was associated with a greater risk of diabetes in Anglo-diet subjects (5.0; CI 1.2–20.3) (Table 4), but the effect of the Anglo diet in this subset was greater than in the total sample, even without adjustment for physical activity (5.2; CI 1.3–21.3).

No relationship was observed between any individual nutrient and the risk of diabetes after adjustment for total caloric intake (Table 5).

CONCLUSIONS

This study suggests that the adoption of an Anglo diet may increase the risk of developing diabetes in Pima Indians, but it does not provide definitive evidence for this hypothesis. Fewer men developed diabetes in our study, therefore these results pertain primarily to women. In the women, the nutrient pattern of a self-reported traditional Indian diet was distinct from that of an Anglo diet, and the Indian diet was associated with a lower risk of diabetes over 6 years. This finding is consistent with previous short-term studies that suggest that a traditional lifestyle protects some indigenous people from diabetes. Our results are strengthened by the intermediate values of both the nutrient intakes and the glucose and insulin concentrations in women who rated their diet as consisting of a mixture of the Indian and Western diets. Clearly, more investigation of this relationship is needed.

The pattern of nutrient intake for the Indian diet in women was closer to that expected for the early Pima Indian diet (18), with the exception of fat intake. The increased fat intake may be an indication that the Pima diet has evolved over time toward the Western diet. Societal pressures and the commodity food distribution programs that serve this community undoubtedly influence the nutrient intake, even in subjects who wish to adhere to the early Indian diet (23). The food distribution programs provide staples such as canned meats, vegetables, and fruits, American cheese, dry cereals, fruit juices, canned or dried milk, and egg mix, which are frequently incorporated into the Pima diet (4).

The main differences between the three diets in women were in total caloric intake and in the amount of complex carbohydrate, fiber, and vegetable protein they contained, with the Indian diet containing the highest levels of each of these nutrients. Although the fiber composition of the diet was not significantly different after adjusting for total energy intake, the influence of fiber on the dietary patterns may be a bulk effect, and if so, the absolute amounts would be more important. In the men, the difference in nutrient composition was probably not statistically significant because of the small sample size in the Anglo group. Presumably, the higher consumption of legumes in the Indian diet is responsible for the increased fiber intake, although this cannot be confirmed because quantitative food item data were not available for this analysis. Nevertheless, diets high in complex carbohydrate and fiber have a lower glycemia index (24–26) and are associated with lower postprandial glucose and insulin responses (25–27) than those containing lower levels of these nutrients. Each of these metabolic factors exerts a protective effect on glucose tolerance in prospective studies (28,29).

The CI for the hazard of developing diabetes in the Anglo group included unity after adjusting for age, BMI, sex, and total energy intake. We believe that this may have been caused by the small sample size of the study, but we cannot rule out the possibility of chance or residual confounding in this relationship. However, the magnitude of the hazard ratio suggests that the Anglo group is at an increased risk of developing diabetes in this population. In this study, the dietary preference classification was made by the participants themselves and was subsequently shown to reflect differences in the nutrient contents of the three groups. The women who reported a mixed diet had values intermediate between the Anglo and Indian diet and may identify a group of subjects whose diets are in transition between these two groups. This method of classifying diet is subjective and may not be robust, but it does identify differences in dietary patterns in this population.

The effect of dietary preference on the development of diabetes was more pronounced in women, as the number of incident cases was greater in women and the dietary patterns were more distinct than in men. Alternative explanations for the relationship between dietary pattern and diabetes were examined, with the main consideration being a lifestyle effect that was not only limited to diet. Physical activity levels did not affect the relationship between the different diets and development of diabetes. Our quantitative physical activity measurement only included leisure-time activities because the quantification of occupational activities is difficult, as the population has a high turnover of jobs (12). Therefore, we propose that dietary factors indeed influence the development of diabetes, but that the effects of individual nutrients are too small or too difficult to measure with the food frequency assessment used in this study. The effect may instead be captured by the global question of dietary preference.

In summary, this study suggests that an adherence to an Indian diet may reduce the risk of diabetes among Pima Indians. Although the ascertainment of dietary pattern was subjective, distinct differences in the nutrient composition of the diet were found in women according to this classification. Increased acceptance of Western dietary habits and access to commodity food programs in this community may further increase the risk of diabetes by adversely affecting the nutrient composition of even the most traditional Indian foods and by increasing the total caloric intake well beyond the nutritional requirements of this population. This hypothesis should be tested in a larger sample of the population and in other populations undergoing similar transitions.