Dietary Patterns in a High-Risk Population for Glucose Intolerance

We evaluated dietary habits as risk factor for glucose intolerance in a high risk population of Japanese-Brazilians enrolled in a study on the prevalence of diabetes (DM). Based on oral glucose tolerance test and WHO criteria, 331 had normal tolerance (NGT), 88 impaired tolerance (IGT) and 83 had type 2 DM (51 self-reported, 32 newly diagnosed diabetics). Clinical, laboratory and dietary data, assessed by food frequency questionnaire (FFQ), were compared between the NGT group and another composed of IGT and newly diagnosed DM (disturbed glucose tolerance or DGT group). Associations of total energy intake and nutrient intakes with glucose intolerance were analyzed by logistic regression. Also, subjects with NGT and DGT entered into separate models of multiple linear regression including BMI as the dependent variable, and total energy intake or each nutrient as independent variables. DGT group showed higher waist-to-hip ratio, blood pressure, plasma glucose and insulin levels and worse lipid profile. Total energy intake, macronutrients, fibers, alcohol and saturated fat intakes did not differ between groups; DGT was not associated with any nutrient intake in multivariate analyses. BMI of the subjects with DGT but not with NGT was associated with protein and cholesterol intakes in linear regression analysis. Our findings did not support an association between nutritional factors and glucose intolerance even in subjects who are unaware of their DGT, using FFQ to reflect current habits. However, we suggest that protein and cholesterol intakes may be markers of increased BMI. Despite assuming that obesity and insulin resistance precedes DM, FFQ may not be useful in the assessment of unfavorable dietary patterns among subjects at risk for glucose intolerance, such as JapaneseBrazilians with elevated BMI. J Epidemiol, 2000 ; 10 : 111-117


INTRODUCTION
An important environmental factor implicated in the genesis of glucose intolerance is diet.Migrant populations who markedly changed their dietary habits have been the target of investigations on the role of diet for type 2 diabetes mellitus (DM).Brazil has the largest Japanese population living outside Japan which constitutes a socially homogeneous community despite their growing integration into political, economic, and cultural development of the country.We found increased prevalence of DM among Japaneses and their descendants living in Brazil as compared to Japan 1) in agreement with findings in the U.S. 2,3), reinforcing a role for environment in the etiopathogenesis of the disease.Fat deposition, particularly of abdominal location, can lead to insulin resistance that is associated with DM and other manifestations of the metabolic syndrome."Western" diet is an attractive hypothesis to explain in part an unfavorable impact of environment for fat deposition and glucose intolerance seen in Asian migrants in America.
The diet of Japanese-Americans with DM was shown to resemble more the American diet than that consumed in Japan 4).Associations of DM with animal fat and protein intakes were reported in a Japanese community living in Washington a and with animal fat, simple carbohydrate and low complex carbohydrate intakes in those living in Hawaii 3).Similar results were found in other ethnic groups [6][7][8] while other studies failed to demonstrate associations between specific nutrient intakes and risk for DM [9][10][11].Methodological limitations in the assessment of dietary patterns may partially contribute to controversial results and the role of specific nutritional factors are still inconclusive.
The aim of this study was to examine the association between specific nutritional factors and glucose intolerance and obesity among first and second-generation Japanese-Brazilians, a community previously demonstrated to be at high risk for DM, after adjustments for confounders.

MATERIALS AND METHODS
A cross-sectional study was conducted in Bauru, a developed city of the state of Sao Paulo, Brazil, to estimate the prevalence of DM in a representative sample of the Japanese population.A census of this Japanese community conducted in 1993 showed a total of 2,954 individuals, including first to fourth-generation Japanese-Brazilians and mixed people.The study sample was comprised of all first-generation and a random sample of the second-generation, in the age group 40-79 years.The entire population of second-generation was listed in alphabetical order and every third subject was selected.A description of the study population and the sample selection process was detailed in a previous paper 1).Participants enrolled in the study were 238 first-and 292 second-generation Japanese-Brazilians (male-female ratios of 1.14 and 0.87, respectively).The percentage of refusal to participate was 11.8% for first-and 11.1% for the second-generation.Residents in the households were informed about the purposes of the survey; selected subjects were interviewed using standardized questionnaires and scheduled to clinical examination.
Prior to the study, members of this Japanese community were asked about their food intake and preferences, including typical Japanese and Brazilian foods and recipes, and a full fist of food was developed that would best estimate the usual intake of this sample.The food frequency questionnaire (FFQ) developed followed a format adapted from Tsunehara et al 5) study in Japanese-Americans.FFQ was used to quantify nutrient intake.All subjects were asked about their usual frequency of consumption of 177 food items for the 2-month period before the participation in the study.Twenty subjects answered the questionnaire twice within a 2-month interval to check its reproducibility.According to the kappa statistics, the reproducibility of total energy intake and macronutrients ranged from 0.52 to 0.70.Average portion sizes were estimated by the subject using food models and household measuring utensils.These portion sizes were then converted to grams using the frequency estimate to reflect intake for one day.Nutrient analysis of the computerized database was performed by a local software, complemented by international 12), Japanese 13) and Brazilian tables 14).Intake of macronutrients was expressed in grams and also as a percentage of total energy intake.Fat was further broken into saturated and unsaturated fat.An unsaturated-to-saturated ratio was calculated.Absolute amounts of cholesterol, fiber, alcohol and refined sugar were also obtained.Physical activity was estimated using questionnaires which included questions concerning type, duration and frequency of activities performed during working and leisure times.A score was attributed to each subject who was classified as level 1when physical activity was light or they were sedentary -or level 2 if it was moderate or heavy.
Subjects were scheduled for clinical and laboratory procedures after an overnight fast.Anthropometric data (weight, height and circumferences) and blood pressure measurements by random-zero sphygmomanometry were obtained.BMI was calculated as weight (kilograms) divided by height (meters) squared.Waist-to-hip ratio (WHR) was calculated as the ratio of waist (measured at the umbilicus) to the hip (at the level of the trochanter major) circumferences.Hypertension was defined by systolic or diastolic blood pressure >_ 140/90 mmHg or by the use of antihypertensive medications 15).A 75-g oral glucose tolerance test (except to those taking insulin or fasting glycemia >_ 200 mg/dl) were performed and diagnostic categories were based on WHO criteria 16).A fasting blood sample was taken for glucose, cholesterol, triglyceride and insulin; plasma glucose and insulin were also determined 2-hour postglucose load.Newly diagnosed diabetic subjects plus those classified as having impaired glucose tolerance (IGT), who were not self-reported, were grouped for comparison purposes.This procedure was adopted because the sample size would be too small if composed only of the newly diagnosed diabetic subjects.Such group of subjects with disturbed glucose tolerance (DGT), who was unaware of their diabetic or IGT condition and who has not changed diet pattern, was compared to those with normal glucose tolerance (NGT).Plasma glucose was determined by the glucose-oxidase method.Cholesterol contents of lipoprotein fractions and serum triglycerides were measured enzymatically.Insulin was determined by monoclonal antibody-based immunofluorimetric assay 17).
Statistical analysis included unpaired Student's t test to compare means of clinical and laboratory data between NGT and DGT groups and chi-square test to compare frequencies.Data were given as means and standard deviations.The effects of variables of main interest -total energy intake, macronutrients, fibers, alcohol, saturated fat intakes and unsaturated-to-saturated fat ratio -on glucose tolerance, adjusted for confounders (age, sex, generation, BMI and physical activity level), was analyzed by unconditional logistic regression.The analytic approach suggested by Willett and Stampfer 18) was employed.
Calorie-adjusted nutrient intakes were computed as the residuals from the regression model with total energy intake as the independent variable and absolute nutrient intake as the dependent variable.Point and interval estimates of the odds ratio were presented.In a second analysis, subjects with NGT and DGT entered into separate models of multivariate linear regression including BMI as the dependent variable, and total energy intake or each nutrient as independent variables, adjusted for age, sex, generation and physical activity.Level of significance was set at p < 0.05.Data analysis was performed by Stata 5.0 software 19).

RESULTS
Out of 530 subjects enrolled on the prevalence study, nutritional data were not available for 28 subjects.No difference was observed between people who participate or not in the study considering variables such as age, sex, generation, BMI and self-reported DM.Among 502 subjects, 331 had NGT, 88 IGT and 83 had type 2 DM.Fifty-one subjects had prior knowledge of their diabetic status and 32 were newly diagnosed based on the results of the oral glucose tolerance test.The DGT group was composed of 120 subjects.Subjects with DGT were slightly older than the NGT ones (57.9± 10.6 vs 59.7 ± 10.6 years, p = 0.05).Male/female ratios (0.91 for NGT and 1.00 for DGT) and generation distributions were not different between groups.DGT group showed higher BMI (25.3±3.5 vs 23.8±3.4Kg/m2, p<0.001) and WHR (0.94± 0.06 vs 0.91±0.07,p < 0.001) than NGT group.Mean systolic and diastolic blood pressures were higher among DGT subjects as compared with NGT (139±23 / 84± 14 vs 127± 19 / 78± 11 mmHg, p < 0.001).A worse lipid profile was observed among the subjects with glucose intolerance.In addition to the expected higher mean fasting and 2-hr plasma glucose, they showed higher total cholesterol (223±41 vs 213±42 mg/dl, p < 0.05) and triglycerides levels (206± 160 vs 155± 117 mg/dl, p < 0.01) when compared with the NGT ones.LDL-cholesterol did not differ between groups (136±45 vs 134±53 mg/dl, respectively), while HDL-cholesterol was lower in the DGT group (39±13 vs 43±11 mg/dl, p < 0.01).Fasting (47.6±69.6 vs 30.0±46.3 pmol/1, p < 0.01) and 2-hr insulin levels (353.7±380.0vs 189.7±253.1 pmol/1, p < 0.001) were higher in DGT as compared with NGT group.
Total energy intake and crude macronutrients intakes were quite similar between groups as well as the percentages of protein, fat and carbohydrate in relation to total energy intake of their diets (table 1).Saturated fat, cholesterol, fiber, alcohol, and refined sugar intakes and unsaturated-to-saturated fat ratio did not differ.Proportions of drinkers in each group were similar.Logistic regression analysis models indicated no association of glucose tolerance status with total energy intake, macronutrients, saturated fat, cholesterol, fiber, alcohol, and refined sugar intakes, whose odds ratios are shown in table 2. Adjustments for confounders showed that BMI was persistently associated with the diagnostic status in all the models (p < 0.001) and also a borderline significance was detected concerning the inverse association with physical activity (p = 0.05).
Considering that high BMI represents an important risk fac-Table 1.Total energy and nutrient intakes (means±SD) of groups according to glucose tolerance status.
NGT, normal glucose tolerance DGT, disturbed glucose tolerance for for insulin resistance and glucose intolerance, specific nutrients' contributions for BMI were evaluated.Multivariate linear regression analysis showed that protein and cholesterol intakes were associated with the BMI of the subjects with DGT, but not with BMI of the NGT group (table 3), adjusting for sex, age, generation, and physical activity level.On the other hand, association of refined sugar intake was found only in the latter (table 3).A borderline significance (p =0.05) was detected concerning an inverse association between BMI and physical activity in the models including protein, fat, carbohydrate, unsaturated fat and cholesterol as the independent variables of main interest.Confounding variables such as sex, age and generation were not associated with BMI in any group of subjects.

DISCUSSION
Changes in dietary patterns should be considered in the explanations of the increased prevalence of type 2 DM seen in Japanese migrants.However, the establishment of a nutrientdisease relationship is complex, requiring careful collection and interpretation of dietary data.It is known that factors such as age, sex, body size and physical activity which account for marked interindividual differences in energy intake.Consideration of the biological implications of analytic approaches is also needed to avoid misleading conclusions.Willett and Stampfer 18) emphasized the importance of employing nutrient intakes adjusted for the effect of total caloric intake.Multivariate regression models and the use of calorieadjusted nutrient intakes, procedures employed in the present study, have been recommended in order to minimize these aspects.
FFQ have been accepted as an accurate method to assess nutrient intake in epidemiological studies [20][21][22].Using such method, the comparison of crude intakes and as percentages of TEI between NGT and DGT Japanese-Brazilians did not support a relationship between any specific nutrient and glucose intolerance.The lack of associations of macronutrients, saturated fat, cholesterol, fiber, alcohol and refined sugar intakes with diagnostic categories was confirmed after adjustments for a number of confounding variables.Contrasting results were reported in literature but, in agreement to ours, some studies also failed to detect any relationship between diet and glucose intolerance [9][10][11].A link of high-fat Western diet to DM has been suggested by several investigators 5.23.24).In fact, in vitro findings and studies in animals and humans demonstrated reduction in tissue sensitivity to insulin and in pancreatic responsiveness to glucose consequent to high-fat diet 25-28).Similarly, elevated basal plasma amino acid concentrations were associated with a worsening of insulin sensitivity 29,30).Considering that type 2 DM represents a insulin-resistant state, a high-fat and protein diet could be an underlying factor for the glucose metabolism deterioration.Some reasons for our discordant results when compared particularly to those observed in Japanese migrants living in Seattle could be raised 5) .First of all, our option by the use of the DGT group which included a large proportion of people with IGT rather than newly diagnosed DM probably weakened any association between dietary factors and glucose intolerance.In spite of being an at high risk condition to develop DM, IGT represents a heterogeneous group of subjects who on retest could be classified in another diagnostic category.Secondly, the relevance of the diet consumed in the last 2 months for the genesis of glucose intolerance, which progresses over a number of years, may be questioned.Actually, the diet of the Japanese community in Brazil changed soon after the arrival of the first Japanese migrants at the beginning of this century due to the scarcity of oriental food, becoming similar to the typical Western diet since .
Considering that the whole community should be already acculturated with respect to its dietary habits, a FFQ focusing on consumption only of the last 2 months, could be inappropriate to demonstrate different nutritional patterns between the groups occurred many years ago.In fact, we found that the diet composition of the Japanese community living in Brazil is more similar to western than Oriental habits.Japanese-Brazilians consume approximately 15% of energy as protein, 32% as fat and 53% as carbohydrate, which is nearer to the usual diet of the U.S. population (16%, 37% and 42%, respectively) 31) than the people living in Japan, who consume 15% of energy as protein, 17% as fat and 61% as carbohydrate 32).Data shown in table 4 particularly compare dietary data obtained in Japanese living in Japan and migrants of similar age groups living in Brazil and the U.S.An intermediate position between the Japanese and Japanese-American diet is suggested since a less contrasting diet composition of the Japanese-Brazilians compared with people living in Japan was observed.In spite of that, a relatively high amount of fat in the diet of the Japanese-Brazilians is still noted when compared with Japan which may be contributing to elevate their BMI.
Higher BMI and WHR among the Japanese-Brazilians with DGT were found and also the known associations of obesity and/or central adiposity 2, 33-36) with disturbances of lipid and glucose metabolism and blood pressure were confirmed, in accordance to other studies in migrant populations 37) .Considering that elevated BMI is a risk factor for insulin resistance and glucose intolerance and the finding of higher BMI and insulin levels in DGT group, we investigated if different diet compositions could account for different outcomes concerning glucose tolerance.In fact, when specific nutrients' contributions for BMI were evaluated, significant and independent associations suggested that a diet with high content of protein and cholesterol may predispose to elevated BMI and worsening of insulin sensitivity.Considering the relatively high amount of fat and low amount of carbohydrate in the diet of the Japanese-Brazilians as compared to Japan (table 4) , we expected to find associations with BMI as well as with glucose intolerance as previously described 5).However , only a tendency of people with DGT to consume less carbohydrate and more fat was observed as the BMI increases , although the former did not reach statistical significance.The latter associations were detected in Japanese-Americans whose diet composition was even more contrasting than ours when compared with the low-fat and high-carbohydrate diet consumed in Japan.Despite our FFQ have focused only the last 2 months prior the study, it seemed to be able to detect a relationship between diet and BMI whose elevation frequently precedes type 2 DM in its natural history.
Since factors such as physical activity level and BMI should determine or cause modifications in food intake , adjustments for these variables are necessary when analyzing dietary patterns between subgroups.Physical activity represents a major risk factor for type 2 DM.Our data pointed to an independent association of low level of physical activity with the risk to glucose intolerance, which is in agreement with a number of studies 3 8 ,33,38).However, the cross-sectional nature of this study does not allow to explore cause-effect relationship between these variables.
Although our findings did not definitively confirm a role of nutritional habits for the risk of type 2 DM, the importance of this environmental factor is reinforced by the association with BMI which represents a major risk factor to develop glucose intolerance.The inverse association of activity level with BMI and glucose intolerance showed that other environmental factors should be also involved.It is possible that both, a highprotein and high-fat diet and low physical activity, interacting with an underlying genetic susceptibility to DM, may contribute to the elevated prevalence of disturbances of glucose metabolism seen in the first-and second-generation Japanese-Brazilians 1).Therefore, it is reasonable to propose that Japanese migrants should maintain their original dietary habits and physical activity level in order to avoid the development of obesity and insulin resistance in the Western environment.Despite availability, Japanese foods are still expensive to be consumed routinely, being more feasible to recommend a decrease in fat and protein content of their usual diet.Further studies in the migrant populations are desirable to clarify the contributory roles of environmental and genetic factors for the genesis of DM.

Table 2 .
Odds ratios of calorie-adjusted nutrient intakes obtained by the respective logistic regression models for the association with glucose intolerance.
Models adjusted for sex, age, generation, physical activity and BMI

Table 3 .
Coefficients of calorie-adjusted nutrient intakes obtained by the respective multiple linear regression models with BMI as dependent variable in Japanese-Brazilians with disturbed and normal glucose tolerance.C.I., confidence interval.Models adjusted for sex, age, generation and physical activity

Table 4 .
Mean macronutrient intakes of Japanese and Japanese migrants in some populations.