点击显示 收起
1 From the Departments of Nutrition (TW, EG, TP, and EBR) and Epidemiology (EG, SEH, and EBR), Harvard School of Public Health, Boston; the Department of Laboratory Medicine, Children's Hospital and Harvard Medical School, Boston (TW and NR); and the Channing Laboratory, Department of Medicine, Harvard Medical School and Brigham and Women's Hospital, Boston (EG, SEH, JM, and EBR).
2 Supported by grants CA49449, 5 P01 CA 87969, and 5 R01 DK58845 from the National Institutes of Health and by the Colorectal Cancer Research Fund. 3 Reprints not available. Address correspondence to T Wu, Department of Nutrition, Harvard School of Public Health, 655 Huntington Avenue, Boston, MA 02115. E-mail: tianying{at}hsph.harvard.edu.
ABSTRACT
Background: Circulating C-peptide concentrations are associated with insulin resistance and the development of type 2 diabetes. However, associations between fructose and the quantity and quality of total carbohydrate intake in relation to C-peptide concentrations have not been adequately examined.
Objective: We assessed the association of dietary fructose, glycemic load, and carbohydrate intake with fasting C-peptide concentrations.
Design: Plasma C-peptide concentrations were measured in a cross-sectional setting in 1999 healthy women from the Nurses' Health Study I and II. Dietary fructose, glycemic load, and carbohydrate intake were assessed with the use of semiquantitative food-frequency questionnaires.
Results: After multivariate adjustment, subjects in the highest quintile of energy-adjusted fructose intake had 13.9% higher C-peptide concentrations (P for trend = 0.01) than did subjects in the lowest quintile. Similarly, in the multivariate model, subjects in the highest quintile of glycemic load had 14.1% (P for trend = 0.09) and 16.1% (P for trend = 0.04) higher C-peptide concentrations than did subjects in the lowest quintile after further adjustment for total fat or carbohydrate intake, respectively. In contrast, subjects with high intakes of cereal fiber had 15.6% lower (P for trend = 0.03) C-peptide concentrations after control for other covariates.
Conclusions: Our results suggest that high intakes of fructose and high glycemic foods are associated with higher C-peptide concentrations, whereas consumption of carbohydrates high in fiber, such as whole-grain foods, is associated with lower C-peptide concentrations. Furthermore, our study suggests that these nutrients play divergent roles in the development of insulin resistance and type 2 diabetes.
Key Words: Fructose • glycemic load • carbohydrate • C-peptide • dietary questionnaire • insulin resistance
INTRODUCTION
Plasma C-peptide, a marker of insulin secretion, is associated with insulin resistance and the development of many chronic diseases such as diabetes (1, 2), cardiovascular diseases (3), and colon cancer (4). C-peptide is less actively extracted by the liver and has a longer half-life than insulin. In addition, because of the greater linearity of the kinetics of C-peptide relative to that of insulin, plasma concentrations of C-peptide better reflect the true pancreatic secretion of insulin and might be a better marker of insulin demand and pancreatic stress (5).
Fructose is a naturally occurring sugar in fruit and vegetables. In the United States, the intakes of the natural sources of fructose were relatively stable since the 1970s. However, the consumption of free fructose has dramatically increased over the past 30 y with the increased consumption of soft drinks and other beverages and foods high in fructose. The addition of high-fructose corn syrup sweeteners to foods such as breakfast cereals, baked goods, and prepared desserts account for most of the free fructose (6, 7). Many animal studies showed that intake of fructose increases body weight, plasma glucose, and insulin concentrations and reduces insulin sensitivity and insulin binding activity than other sugars or starch (8-12). However, the long-term effects of fructose intake in relation to insulin action and sensitivity in human studies are less clear. Beck-Nielsen et al (13) and Hallfrisch et al (14) showed that diets with higher proportions of fructose than other carbohydrates lead to reduced insulin action and sensitivity, whereas others found that fructose decreases plasma glucose concentrations (15, 16).
Several studies showed that, when patients with type 2 diabetes are fed isocarbohydrate diets, diets with a low glycemic load (GL) improve glycemic control (17, 18). Some (19, 20) but not all (21, 22) prospective cohort studies showed that a high-GL diet increases risk of type 2 diabetes. However, the use of low-GL diets for the prevention and management of diabetes has still not received conclusive support. The aim of this study was to assess the association of fructose intake, GL, and the quantity and quality of carbohydrate intake in relation to plasma C-peptide concentrations.
SUBJECTS AND METHODS
Study population
The Nurses' Health Study I
The Nurses' Health Study I cohort, established in 1976, consists of 121 700 female registered nurses aged 30-55 y who completed a mailed questionnaire and provided detailed medical history and lifestyle information at baseline and on subsequent biennial mailed questionnaires. Diet was assessed approximately every 4 y with the use of a previously validated (23, 24) semiquantitative food-frequency questionnaire (SFFQ). Between 1989 and 1990, we collected blood samples from 32 826 cohort members who were then aged 43-69 y. As previously published (25), each woman received a kit containing all the supplies needed for blood collection plus a supplemental questionnaire to collect additional information on menopausal status, recent postmenopausal hormone use, time since last meal, and time of day of blood sampling. After blood collection the women mailed the whole-blood sample, cooled with an ice pack, by way of overnight mail. On arrival, samples were centrifuged, and plasma, red blood cells, and buffy coat were stored in the vapor phase of liquid nitrogen freezers. A total of 1534 women were included in this analysis, all of whom were control participants in 3 nested case-control studies of breast cancer, hypertension, and diabetes. None of the women had previously diagnosed cancer, cardiovascular diseases, or diabetes at time of blood drawing. We included only those women who had fasted overnight for 6 h. The Institutional Review Board of the Brigham and Women's Hospital in Boston approved the study.
The Nurses' Health Study II
The Nurses' Health Study II cohort was established in 1989 and comprises 116 671 female registered nurses from 14 states aged 25-44 y at the start of the study. We used methods similar to those described for the Nurses' Health Study I to obtain questionnaire information and blood samples. Blood was collected in 1997-1999 from 29 604 women. Of those women, 473 healthy women who were free of cancer, cardiovascular diseases, and diabetes were previously sampled to study the effects of alcohol consumption patterns on biologic markers. The selection criteria for that subsample were described elsewhere (26). We selected 465 women from that subsample who had a C-peptide measurement, fasted overnight for 6 h, and filled out an SFFQ in 1999 or 1995. The Institutional Review Board of the Harvard School of Public Health approved the study.
Dietary assessment
The reproducibility and validity of the SFFQs were described in detail (23, 24, 27). Participants were asked to report the average frequency of consumption of 130 food items. Standard portion sizes were listed with each food, and 9 frequency choices from "less than once a month to 6 or more times a day" were given. The specific values of each nutrient were obtained from the Harvard University Food Composition Database, which is derived from the US Department of Agriculture [Composition of Foods (28)], information from manufacturers, and published papers. The total nutrient intake was the sum of nutrients derived from each food.
The correlation for energy-adjusted carbohydrate intake between the SFFQ and multiple 7-d diet records was 0.76 (24), and between 2 SFFQs over a 4-y period it was 0.65 (24). The correlation between SFFQ and multiple dietary records for carbohydrate-containing foods was high and included 0.66 for potatoes, 0.71 for white bread, 0.79 for cold cereals, and 0.80 for bananas (29, 30).
Fructose is a monosaccharide, and half of the disaccharide sucrose is fructose, which is split from sucrose in the small intestine. Therefore, total fructose intake is equal to the intake of free fructose plus half the intake of sucrose. The correlations between the intakes measured by SFFQ and diet records for the 4 top contributors of the monosaccharide fructose intake in our data set were 0.78 for orange juice, 0.84 for soft drink (cola) beverages, 0.59 for raisins, and 0.70 for apples (27).
The use of glycemic index is well documented (31). The glycemic index is a method of ranking foods on the basis of the incremental glucose response and insulin demand they produce for a given amount of carbohydrate (31). We used glycemic index to calculate the average GL during the past year for each participant by multiplying the carbohydrate content (grams per serving) for each food by its glycemic index, multiplying the product by the frequency of consumption (serving of that food per day), and summing values for all food items reported:
RESULTS
Among the 1999 women in this analysis, subjects in the lowest quintile consumed 2.8% of energy from free fructose compared with 7.8% of energy for women in the highest quintile. Women who consumed more free fructose were less likely to smoke and to drink alcohol and more likely to be older, to be postmenopausal, and to be physically active. Free fructose consumption was positively associated with intake of total energy, carbohydrate, sucrose, and GL and inversely associated with intake of total fat, protein, and cholesterol (Table 1). BMI was slightly inversely associated with fructose intake. The overall pattern for the associations between GL and other covariates was similar. For example, smoking and alcohol were negatively associated with GL.
View this table:
TABLE 1. Age-standardized characteristics by quintile (Q) of percentage of energy from free fructose intake and quintile of energy-adjusted glycemic load in 1999 women who participated in the Nurses' Health Study I and II1
The age- and BMI-adjusted (also adjusted for fasting hours and laboratory batch) C-peptide concentrations were 8% higher among women in the highest quintile of free fructose intake than in women in the lowest quintile (Table 2). The association was stronger in multivariate analysis (13.9%) and after further adjustment for total carbohydrate intake (20.5%). The associations between total fructose (including the fructose in sucrose) and C-peptide were similar (Table 2). The association between sucrose and C-peptide was not significant in multivariate models (data not shown). GL was positively associated with C-peptide in the multivariate model adjusted for total fat or carbohydrate intake, although P for trend was only marginally significant after adjustment for total fat in the multivariate analysis. Higher intake of carbohydrate was inversely associated with C-peptide concentrations after adjustment for GL. Women in the highest quintile of carbohydrate had 14.2% lower C-peptide concentrations than women in the bottom quintile (P for trend = 0.04). We also examined the association between total protein, total fat, and trans fatty acids in relation to C-peptide. No association was seen between total protein (animal or vegetable protein) and C-peptide in the multivariate-adjusted model, and no relation was seen between total fat or trans fatty acid and C-peptide in the multivariate-adjusted model. BMI was a strong confounder, but the results were similar whether we treated BMI as a continuous variable or as deciles. When BMI was treated as a continuous variable in the multivariate model, we found that every unit increase in BMI was associated with a 0.1 ng/mL increase in C-peptide (P < 0.0001). This finding provides an interval validity, because BMI is strongly associated with insulin resistance.
View this table:
TABLE 2. Changes in fasting plasma C-peptide concentrations from quintile 1 to quintile 5 associated with the change from quintile 1 to quintile 5 in each of the dietary predictors in women1
We further examined the association between dietary predictors and plasma C-peptide concentration within strata of BMI (<25 and 25). P values for interaction were not significant for any of the dietary predictors with BMI strata (<25 and 25). The absolute difference for fructose and GL was slightly greater in the overweight group. In the overweight group, the overall range was above normal; thus, even a small increase might have a substantial effect. We also reexamined these associations by using a different cutoff point for BMI (<30 or 30) and obtained similar results.
To explore the foods that might contribute to these associations we examined the relation between carbohydrate-contributing foods (Table 3), fructose-contributing foods (Table 4), and different sources of fiber in relation to plasma C-peptide concentrations. Among the top fructose-contributing foods in our SFFQ, a positive association with C-peptide was observed only for the caffeine-containing beverages and fruit punch. In contract, concentrations of C-peptide were lower among women who consumed 1 serving of "ready-to-eat" cereal/d compared with women who consumed <1 serving/wk. Among sources of dietary fiber (cereal, vegetable, fruit), only cereal fiber was significantly associated with C-peptide in the multivariate analysis. Women in the highest quintile of cereal fiber (median concentration: 9 g/d) had 15.6% lower C-peptide concentrations than women in the lowest quintile (median concentration: 2 g/d; P for tend = 0.03).
View this table:
TABLE 3. Changes in C-peptide concentrations associated with the change in intake of carbohydrate-contributing foods from 0 servings/wk in women1
View this table:
TABLE 4. Changes in C-peptide concentrations associated with the change in intake of fructose-contributing foods from 0 servings/wk in women1
DISCUSSION
In this large cross-sectional study of diet and C-peptide concentrations among premenopausal and postmenopausal women, dietary fructose intake and GL were positively associated with C-peptide concentrations, whereas cereal fiber was inversely associated with C-peptide. In the early 1960s, fructose was considered beneficial because it was thought to stimulate lower postprandial plasma concentrations of insulin and glucose (7). However, more recently, results from experimental studies in animal models and from short-term feeding trials among humans suggest that higher fructose intake contributes to insulin resistance, impaired glucose tolerance, and hyperinsulinemia (9-12). For example, Thorburn et al (10) fed rats a diet containing 35% of energy as fructose for 4 wk and found reduced insulin sensitivity and whole-body glucose disposal, whereas comparable amounts of starch had no observable effects. Results were similar in rats fed for longer periods (15 mo) with lower amounts of fructose (9). In human studies, Beck-Nielsen et al (13) showed a reduction in insulin binding and insulin activity among healthy subjects fed 1000 extra kcal as fructose for 7 d, whereas intake of 1000 extra kcal glucose had no similar adverse effects. In another study by Hallfrisch et al (14), intake of 15% of total energy as fructose for 5 wk resulted in higher insulin and glucose responses than isocaloric diets with 7.5% of energy from fructose or no fructose. In contrast, among patients with type 2 diabetes, Bantle et al (15) and Osei et al (16) found that fructose reduced blood glucose concentrations in comparison with either isocaloric sucrose or isocaloric diabetic diets in patients with diabetes.
Several mechanisms were proposed to explain the relation between fructose intake and insulin resistance (6). First, an increase in fructose consumption leads to positive energy balance, which might contribute excess body weight (36, 37). Excess adiposity is associated with higher concentration of nonesterified fatty acids (38), which might reduce insulin sensitivity by increasing the intramyocellular lipid content in muscle cells where insulin receptors are located (39). Second, an increased supply of nonesterified fatty acids can lead to an increase in triacylglycerol concentrations, which is associated with reduced insulin sensitivity (40). Furthermore, studies showed that fructose intake increased plasma triacylglycerol concentrations compared with other types of sugar (41-43), and fructose intake resulted in weight gain (36, 37). Taken together, these results suggest high intake of fructose over time might deteriorate insulin sensitivity and promote the development of type 2 diabetes.
We found a positive association between C-peptide concentration with GL and an inverse association with carbohydrates. GL and carbohydrate intake are highly correlated (r = 0.83), but with a large sample size we were able to assess independent associations.
Theoretically, we should observe the effect of GL without controlling for the amount of carbohydrate in the multivariate analysis because GL is a measurement of both the quality and quantity of carbohydrate. However, carbohydrates include components in addition to those contributing to GL. Foods that contribute carbohydrates can be whole grains, refined grains, fruit, and vegetables. If the GL is the same for a woman who consumes mashed potatoes (glycemic index = 102) as for a woman who eats raisin bran cereal (glycemic index = 88) (44), the carbohydrate content of cereal must be 16% higher than that for the mashed potatoes because of the 16% higher glycemic index of mashed potatoes. The food with the additional carbohydrate contains more fiber, vitamins, minerals, and phytochemicals. Therefore, in a model that simultaneously includes GL and carbohydrate, an increase in GL holding carbohydrate constant represents proportionally more carbohydrate from high-GL foods; whereas, carbohydrate holding GL constant represents foods higher in fiber, micronutrients, minerals, and antioxidants. The combination of these components might work synergistically to lower the risk of diabetes (45, 46). The inverse association we observed between intake of cereal fiber and C-peptide concentrations supports our interpretation.
In an additional approach to examine the association between GL and C-peptide, we added total fat (which is not highly correlated with GL) instead of total carbohydrate to the multivariate model. We found a similar positive association between GL and C-peptide. In this particular model, GL represents a true comparison of a high-GL diet with a low-GL diet when holding constant total calories, fat, protein, and implicitly total carbohydrate.
We found that a high consumption of cereals and cereal fiber was inversely associated with C-peptide concentrations, whereas a high consumption of mashed potatoes was positively associated with C-peptide concentrations. Both findings support the possible contribution of a high-GL diet to insulin resistance and of the possible protection by whole-grain foods. Previously, in the Nurses' Health Study and in the Health Professionals Follow-Up Study, we reported a positive association for GL and a negative association for cereal fiber in relation to the incidence of diabetes in both women and men (19, 20). The Iowa Women's Health Study failed to show an association between GL and diabetes (21). The Iowa study might contain more measurement error in the assessment of dietary intake, because the food-frequency questionnaire was completed only once, and the change in diet during the course of follow-up was not assessed. Several other observational and interventional studies reported the dual benefit of low-GL diets and diets high in whole grains on improved glycemic control and decreased plasma insulin concentrations (1, 10, 47). Kiens et al (22) also were unable to show that high-GL diets increase insulin concentration and increase blood glucose compared with low-GL diets after 30 d of intervention; however, their study included only 7 subjects (young, healthy men) and the high-GL diets might exert their adverse effect later in life.
Some limitations and strengths were found in our study. Because of its cross-sectional design, a causal relation cannot be established. The positive association between total fructose (including free fructose and fructose from sucrose, a disaccharide) and C-peptide was not stronger than the positive association between free fructose and C-peptide. This finding suggests that the fructose in disaccharides (sucrose) might not add greater adverse effect. However, the mechanism for this is not clear. Nevertheless, the large sample size of this study enables us to tease out the independent association of highly correlated covariates such as GL and carbohydrate. The suggested adverse effect of fructose could reflect the harmful effect of the excess of caloric intake from soft drinks. However, we have controlled for total caloric intake in our multivariate model and also especially for the amount of calories from protein and carbohydrate.
In conclusion, we found a significant positive association between high-fructose diets and C-peptide concentrations. The association was independent of total carbohydrate quantity and quality. These results suggest potentially adverse metabolic effects of fructose as sweeteners to soft drinks or other foods, and they indicate a strong need for more research in this area. Furthermore, high glycemic foods and the overall GL of the diet also were associated with higher C-peptide concentrations. These results lend further support to dietary guidelines to replace refined starch with whole grains (28).
ACKNOWLEDGMENTS
TW performed the study design, data collection, and data analysis and wrote the first draft of the manuscript. EG, TP, SEH, JM, NR, and EBR participated in the study design, data collection, and data analysis. The authors had no conflicts of interest.
REFERENCES