Abstract
Background: Carbohydrate quality may be an important determinant of type 2 diabetes (T2D); however, relations between various carbohydrate quality metrics and T2D risk have not been systematically investigated.
Objective: The purpose of this study was to prospectively examine the association between carbohydrates, starch, fibers, and different combinations of these nutrients and risk of T2D in women.
Design: We prospectively followed 70,025 women free of cardiovascular disease, cancer, and diabetes at baseline from the Nurses’ Health Study (1984–2008). Diet information was collected with the use of a validated questionnaire every 4 y. Cox regression was used to evaluate associations with incident T2D.
Results: During 1,484,213 person-years of follow-up, we ascertained 6934 incident T2D cases. In multivariable analyses, when extreme quintiles were compared, higher carbohydrate intake was not associated with T2D (RR = 0.98; 95% CI: 0.89, 1.08; P-trend = 0.84), whereas starch was associated with a higher risk (RR = 1.23; 95% CI: 1.12, 1.35; P-trend <0.0001). Total fiber (RR = 0.80; 95% CI: 0.72, 0.89; P-trend < 0.0001), cereal fiber (RR = 0.71, 95% CI: 0.65, 0.78; P-trend < 0.0001), and fruit fiber (RR = 0.79; 95% CI: 0.72, 0.85; P-trend < 0.0001) were associated with a lower T2D risk. The ratio of carbohydrate to total fiber intake was marginally associated with a higher risk of T2D (RR = 1.09; 95% CI: 1.00, 1.20; P-trend = 0.04). On the other hand, we found positive associations between the ratios of carbohydrate to cereal fiber (RR = 1.28; 95% CI: 1.17, 1.39; P-trend < 0.0001), starch to total fiber (RR = 1.12; 95% CI: 1.02, 1.23; P-trend = 0.03), and starch to cereal fiber (RR = 1.39; 95% CI: 1.27, 1.53; P-trend < 0.0001) and T2D.
Conclusions: Diets with high starch, low fiber, and a high starch-to–cereal fiber ratio were associated with a higher risk of T2D. The starch-to–cereal fiber ratio of the diet may be a novel metric for assessing carbohydrate quality in relation to T2D.
Keywords: carbohydrate quality, carbohydrate-to-fiber ratio, fiber, starch, type 2 diabetes
INTRODUCTION
An estimated 366 million people in the world suffered from type 2 diabetes (T2D)8 in 2011, and by the year 2030 that number is estimated to rise to 552 million (1). In the United States, as of 2011, an estimated 23.7 million people had T2D, and by 2030 this number is estimated to rise to 29.6 million (1). It has been established that poor carbohydrate quality is a significant risk factor of T2D (2, 3). Meta-analyses of observational studies have shown that glycemic index (GI), a carbohydrate quality metric, is positively associated with the risk of T2D (4), whereas other carbohydrate quality metrics, such as whole grains and fiber—specifically total and cereal fiber—have consistently been inversely associated with T2D (5, 6). However, most studies found no association between fruit and vegetable fiber intake and risk of T2D (6). Mixed results have been reported for the association between total carbohydrate intake and T2D, with most demonstrating no association (7–12) and one showing a positive association (13). Likewise, some studies do show that starch is associated with an increased risk of T2D (11, 13), whereas others have found no association (7, 9, 12, 14).
The American Heart Association recommends choosing foods with a carbohydrate-to-fiber ratio of no more than 10:1 (15). This criterion identified more healthful whole grain products than did other carbohydrate selection criteria, such as the whole grains stamp and whole grains listed as the first ingredient (16). Other variations of this ratio include ratios of carbohydrate to cereal fiber, starch to total fiber, and starch to cereal fiber. Although these ratios have been recommended to assess the quality of individual foods, their potential use in determining risk of diseases such as T2D has not been explored. These ratios can differentiate between different carbohydrate quality diets, in which diets rich in refined carbohydrates and sugar-sweetened beverages could have a high carbohydrate-to-fiber ratio, whereas diets rich in whole grains could have a low ratio. Therefore, the objectives of this study were to examine the association between 1) carbohydrates, starch, and total, cereal, fruit, and vegetable fiber consumption and the risk of T2D; 2) carbohydrate- and starch-to-fiber ratios as a measure of carbohydrate quality and the risk of T2D; and 3) the joint effects of carbohydrate and total fiber, carbohydrate and cereal fiber, starch and total fiber, and starch and cereal fiber on the risk of T2D in a large cohort of US women.
METHODS
Study population
The Nurses’ Health Study (NHS) was initiated in 1976 as a prospective cohort study in which 121,701 female registered nurses between the ages of 30 and 55 y were recruited from 11 US states. Every 2 y, nurses updated their information on medical history, lifestyle, and incidence of chronic diseases with the use of validated questionnaires. Follow-up rates are kept at a 90% minimum and mortality follow-up is more than 98%.
For the current investigation, we excluded participants with a baseline history of diabetes, cardiovascular disease, or cancer because these diagnoses may result in changes in diet (17). We also excluded women who left 10 or more items blank on the food-frequency questionnaire (FFQ) or who had an implausible energy intake (<500 or >3500 kcal/d). The final analyses include 70,025 women with complete information. The study was approved by the Human Research Committee of Brigham and Women’s Hospital in Boston.
Assessment of diet
In 1984, diet was assessed with the use of a 116-item semiquantitative FFQ. In 1986, the FFQ was expanded to 133 items and was mailed to subjects every 4 y to update diet information. Participants were asked how often on average (“never” to “6 or more times per day”) they consumed a specified common portion size or serving size of specific foods. The validity and reproducibility of the FFQ in measuring food intake has been demonstrated previously (18–21). In a previous validation study in a subsample of 173 nurses in the Boston area, FFQ assessment of total carbohydrate and total fiber were moderately correlated with the average of four 1-wk diet records (total carbohydrate, r = 0.64; total fiber, r = 0.56) (18, 22). Carbohydrate-rich food items had similar correlation coefficients (cold breakfast cereal, r = 0.79; white bread, r = 0.71; dark bread, r = 0.77; pasta/rice, r = 0.35; and potatoes, r = 0.66) (19).
The main exposure variables are carbohydrates, starch, total fiber, cereal fiber, fruit fiber, vegetable fiber, and ratios of carbohydrate to total fiber, carbohydrate to cereal fiber, starch to total fiber, and starch to cereal fiber. Nutrient intake was calculated by multiplying the frequency of consumption by the nutrient content of the specified portion sizes of each food. Then, the nutrient content of all food items in a participant’s diet was summed up to estimate the individual nutrient intake. The nutrient contents were determined with the use of the USDA food composition tables and complemented with information from manufacturers (23). All dietary variables were adjusted for total energy intake to control for confounding and remove extraneous variation from differences in body size, metabolic efficiency, and physical activity (24).
Assessment of T2D
Our primary outcome of interest was incident T2D assessed by self-report in the biennial questionnaires sent between 1984 and 2008. On each questionnaire sent out to the participants, from baseline and every 2 y after that, participants were asked if and when they were ever diagnosed with diabetes, either type 1 or 2. Participants who reported a diagnosis of diabetes were mailed a supplementary questionnaire that asked about symptoms, diagnostic tests, and hypoglycemic therapy. The National Diabetes Data Group criteria were used to define cases that were diagnosed before 1998 (25). To confirm that a subject is diagnosed with diabetes, as per the guidelines of the National Diabetes Data Group, the subject must report at least one of the following criteria on the supplementary questionnaire: 1) one or more classic symptoms (excessive thirst, polyuria, weight loss, hunger, pruritus, or coma) plus fasting plasma glucose ≥7.8 mmol/L or random plasma glucose ≥11.1 mmol/L; or 2) ≥2 elevated plasma glucose concentrations on different occasions (fasting plasma glucose 7.8 mmol/L and/or random plasma glucose ≥11.1 mmol/L and/or plasma glucose ≥11.1 mmol/L at 2 h on oral glucose tolerance testing) in the absence of symptoms; or 3) treatment with hypoglycemic medication (insulin or oral hypoglycemic agent). From 1998, the criteria of the American Diabetic Association were used to define cases, in which a fasting plasma glucose ≥7 mmol/L was the threshold for diagnosing diabetes instead of 7.8 mmol/L (26). The supplementary questionnaire used has been validated in 2 studies in which a substudy of the NHS showed that 98% of self-reported cases in the supplementary questionnaires were confirmed by reviews of medical records (27, 28).
Assessment of covariates
Participants were sent questionnaires biennially, in which they provided updated information on their age, weight, menopausal status, postmenopausal hormone use, smoking status, multivitamin use, and disease status, including hypertension and hypercholesterolemia. Height was reported in the 1976 questionnaire when the NHS was initiated. Family history of diabetes in first-degree relatives was reported in 1982 and 1988. Physical activity was assessed as metabolic equivalent task (MET) hours per week. Correlations between the physical activity reported on the questionnaires and that reported on recalls and diaries in 2 validation studies were high (0.79 and 0.62) in the NHS II (29).
Statistical analysis
Person-time for each subject was calculated from the date of return of the 1984 questionnaire, baseline, to the date of diagnosis of T2D, loss to follow-up, death, or the end of follow-up (1 June 1 2008), whichever was earliest. Pearson correlation coefficients were used to evaluate associations between carbohydrates, starch, total fiber, cereal fiber, ratio of starch to total fiber, ratio of starch to cereal fiber, GI, and glycemic load (GL) intake in the study population at baseline. Time-varying Cox proportional hazards regression was used to estimate RR of T2D by quintiles of energy-adjusted intake of exposure variables, stratified by 5-y age categories. A test for linear trend was conducted with the use of quintiles of the dietary exposure variable as a continuous variable by assigning the median values of the quintiles to the variable.
We used cumulative averages of all available dietary data to reduce within-person variation and to best represent long-term diet (17). However, we stopped updating dietary exposures of participants that developed disease outcomes such as cancer, heart disease, and stroke at the beginning of each 2-y follow-up cycle. This was done to eliminate the potential confounding of the diet–disease association from change in diet due to the diagnosis of these diseases (17). To reduce the influence of outliers, participants were grouped into quintiles of energy-adjusted dietary exposure variables, with the lowest quintile being the reference group.
We adjusted for several confounders in our Cox models. Model 1 was age-adjusted. In model 2, we further adjusted for various lifestyle covariates and diabetes risk factors, including BMI (in kg/m2; <21, 21–<23, 23–<25, 25–<27, 27–<30, 30–<33, 33–<35, 35–<40, and ≥40), race (Caucasian, African American, Native American, Asian, or Hawaiian), total energy intake (kcal/d, in quintiles), smoking status (never, past, current 1–15, current 16–25, or current ≥26 cigarettes/d), alcohol consumption (0, 0.1–4.9, 5.0–9.9, 10.0–14.9, or ≥15 g/d), physical activity (<3, 3–< 9, 9–< 18, 18–< 27, or ≥27 MET-h/wk), postmenopausal hormone use (premenopause, postmenopausal never user, postmenopausal current user, or postmenopausal past user), family history of diabetes (yes or no), and multivitamin use (yes or no). In model 3, we further adjusted for dietary covariates such as red meat (servings/d), coffee (cups/d), ratio of polyunsaturated fat to saturated fat, trans fat (percentage of total energy), sugar-sweetened beverage intake (cups/d), fruits and vegetables (servings/d), and magnesium (mg/d), and they were all assessed in quintiles. Nondietary variables are updated every 2 y, whereas dietary variables are updated every 4 y, starting in 1984.
We tested for potential effect modification of the association between the 4 ratios and risk of T2D by age (<60 and ≥60 y) and physical activity (<10 MET-h/wk and ≥10 MET-h/wk) by including a crossproduct term in our fully adjusted models (Wald test, 1 df). We also performed an a priori–determined stratified analysis of starch and starch-to–cereal fiber intake on the risk of T2D by age (<60 and ≥60 y), BMI (<28 and ≥28), family history of diabetes (yes or no) and physical activity (<10 MET-h/wk and ≥10 MET-h/wk). We also tested for the joint effects of carbohydrate and total fiber, carbohydrate and cereal fiber, starch and total fiber, and starch and cereal fiber intake on risk of T2D. In sensitivity analyses, we repeated our main analysis after diet was continuously updated throughout follow-up, even after participants developed outcomes such as cardiovascular disease and cancer. We tested for potential nonlinearity in the association between starch and starch-to–cereal fiber intake and risk of T2D by using restricted cubic splines with 5 knots at the 5th, 27.5th, 50th, 72.5th, and 95th percentiles of exposure (30), in which we gave all values <quartile 1 − 3 × IQR the value of quartile 1 − 3 × IQR and all values >quartile 3 + 3 × IQR the value of quartile 3 + 3 × IQR. All statistical tests were 2-sided and a P value < 0.05 was considered statistically significant. SAS version 9.3 for UNIX was used for all statistical analysis.
RESULTS
During 24 y of follow-up (1984–2008), 6934 of the 70,025 participants were diagnosed with T2D. The age-adjusted baseline characteristics of the study participants according to their carbohydrate, starch, total, and cereal fiber intake are presented in Table 1. Women who had a diet higher in carbohydrates and starch were, on average, less likely to be current smokers and hypertensive at baseline and had higher GI, GL, and total fiber and cereal fiber intake and lower red meat and coffee intake. Women who had a diet higher in total and cereal fiber were, on average, more physically active, more likely to use multivitamin supplements, less likely to be current smokers, and less likely to be hypertensive at baseline; had a lower intake of sugar-sweetened beverages, red meat, and coffee; and had a higher intake of fruits and vegetables, carbohydrates, starch, and magnesium.
TABLE 1.
Age-adjusted baseline (1984) characteristics of the NHS study population by nutrient intake1
| Quintiles of intake of energy-adjusted nutrient |
||||||||||||
| Carbohydrates |
Starch |
Total fiber |
Cereal fiber |
|||||||||
| Variable | Q1 | Q3 | Q5 | Q1 | Q3 | Q5 | Q1 | Q3 | Q5 | Q1 | Q3 | Q5 |
| n | 13,970 | 13,516 | 13,661 | 13,999 | 12,715 | 14,057 | 14,230 | 14,511 | 13,601 | 14,364 | 13,089 | 13,689 |
| Median, g/d | 146 | 185 | 224 | 39.3 | 58.2 | 78.6 | 11.0 | 15.7 | 22.6 | 2.00 | 3.60 | 7.10 |
| Anthropometric/lifestyle factors | ||||||||||||
| Age,2 y | 49.7 ± 6.9 | 49.8 ± 7.2 | 51.1 ± 7.3 | 51.3 ± 6.9 | 49.9 ± 7.2 | 49.4 ± 7.3 | 48.4 ± 6.9 | 49.8 ± 7.1 | 52.4 ± 7.0 | 49.9 ± 7.0 | 49.4 ± 7.1 | 51.7 ± 7.2 |
| BMI, kg/m2 | 25.1 ± 4.6 | 24.9 ± 4.5 | 24.3 ± 4.3 | 24.8 ± 4.5 | 24.9 ± 4.5 | 24.7 ± 4.5 | 24.8 ± 4.7 | 25.0 ± 4.5 | 24.5 ± 4.1 | 25.0 ± 4.6 | 25.0 ± 4.5 | 24.3 ± 4.1 |
| Caucasian | 99 | 98 | 96 | 98 | 98 | 97 | 97 | 98 | 97 | 97 | 98 | 98 |
| Postmenopausal | 45 | 45 | 46 | 46 | 45 | 44 | 45 | 44 | 47 | 46 | 44 | 46 |
| PMH use (past and current) | 49 | 48 | 48 | 48 | 48 | 47 | 46 | 48 | 50 | 46 | 47 | 51 |
| Family history of diabetes | 18 | 19 | 19 | 19 | 19 | 20 | 18 | 19 | 19 | 19 | 19 | 19 |
| Hypertension | 22 | 19 | 20 | 22 | 20 | 19 | 21 | 19 | 19 | 23 | 19 | 18 |
| Hypercholesterolemia | 7 | 7 | 9 | 7 | 7 | 8 | 7 | 7 | 8 | 7 | 7 | 8 |
| Physical activity, MET-h/wk | 13.0 ± 19.1 | 13.9 ± 19.7 | 16.0 ± 25.5 | 15.8 ± 23.2 | 14.1 ± 21.5 | 13.3 ± 21.1 | 10.7 ± 17.9 | 13.5 ± 17.4 | 19.1 ± 26.6 | 13.6 ± 21.8 | 13.8 ± 20.2 | 15.9 ± 23.1 |
| Current smoker | 34 | 22 | 19 | 30 | 23 | 20 | 38 | 22 | 14 | 35 | 23 | 14 |
| Multivitamin supplement user | 35 | 37 | 40 | 40 | 37 | 35 | 33 | 36 | 44 | 34 | 36 | 42 |
| Dietary factors | ||||||||||||
| Total energy, kcal/d | 1698 ± 524 | 1797 ± 530 | 1687 ± 533 | 1720 ± 559 | 1790 ± 524 | 1656 ± 497 | 1707 ± 538 | 1777 ± 530 | 1715 ± 523 | 1703 ± 550 | 1775 ± 538 | 1693 ± 481 |
| Alcohol, g/d | 15.0 ± 17.5 | 5.6 ± 8.1 | 2.7 ± 5.0 | 11.6 ± 16.5 | 6.4 ± 9.5 | 3.9 ± 6.5 | 11.5 ± 16.5 | 6.4 ± 9.6 | 4.3 ± 7.0 | 11.6 ± 16.3 | 6.1 ± 9.4 | 4.6 ± 7.6 |
| Fruit and vegetables, servings/d | 4.3 ± 2.0 | 5.2 ± 2.3 | 6.0 ± 3.0 | 5.6 ± 2.9 | 5.3 ± 2.4 | 4.6 ± 2.2 | 3.1 ± 1.4 | 5.1 ± 1.7 | 7.6 ± 2.9 | 5.1 ± 2.7 | 5.2 ± 2.5 | 5.3 ± 2.4 |
| Red meat, servings/d | 1.5 ± 0.8 | 1.2 ± 0.6 | 0.7 ± 0.5 | 1.2 ± 0.8 | 1.2 ± 0.7 | 0.9 ± 0.6 | 1.4 ± 0.8 | 1.2 ± 0.6 | 0.8 ± 0.6 | 1.3 ± 0.8 | 1.2 ± 0.7 | 0.9 ± 0.6 |
| Protein, g/d | 78 ± 14 | 72 ± 11 | 63 ± 11 | 75 ± 17 | 61 ± 12 | 68 ± 11 | 69 ± 14 | 71 ± 12 | 74 ± 13 | 73 ± 16 | 70 ± 12 | 71 ± 12 |
| Polyunsaturated:saturated fat ratio | 0.5 ± 0.2 | 0.5 ± 0.2 | 0.6 ± 0.2 | 0.5 ± 0.2 | 0.6 ± 0.2 | 0.6 ± 0.2 | 0.5 ± 0.2 | 0.6 ± 0.2 | 0.6 ± 0.2 | 0.5 ± 0.2 | 0.6 ± 0.2 | 0.6 ± 0.2 |
| trans fat, % of total energy | 2.1 ± 0.6 | 2.0 ± 0.6 | 1.6 ± 0.6 | 1.7 ± 0.6 | 2.0 ± 0.6 | 1.9 ± 0.6 | 2.1 ± 0.6 | 2.0 ± 0.6 | 1.6 ± 0.5 | 1.8 ± 0.6 | 2.0 ± 0.6 | 1.8 ± 0.6 |
| Sugar-sweetened beverages,3 cups/d | 0.9 ± 1.2 | 0.8 ± 1.0 | 1.0 ± 1.2 | 1.1 ± 1.3 | 0.9 ± 1.0 | 0.8 ± 1.0 | 1.1 ± 1.3 | 0.9 ± 1.0 | 0.7 ± 1.0 | 1.1 ± 1.3 | 0.9 ± 1.1 | 0.7 ± 0.9 |
| Coffee (including decaffeinated),3 cups/d | 2.8 ± 1.9 | 2.5 ± 1.8 | 2.0 ± 1.8 | 2.6 ± 1.9 | 2.4 ± 1.8 | 2.3 ± 1.8 | 2.6 ± 1.9 | 2.5 ± 1.8 | 2.2 ± 1.8 | 2.5 ± 1.9 | 2.5 ± 1.8 | 2.3 ± 1.8 |
| Glycemic load | 74 ± 11.5 | 99 ± 6.8 | 125 ± 13.4 | 85 ± 23 | 99 ± 15 | 113 ± 16 | 90 ± 22 | 99 ± 17 | 109 ± 19 | 88 ± 23 | 100 ± 16 | 110 ± 17 |
| Glycemic index | 51.8 ± 4.0 | 53.4 ± 3.4 | 54.7 ± 3.8 | 50.2 ± 4.4 | 53.5 ± 2.7 | 56.0 ± 2.7 | 53.8 ± 4.3 | 53.6 ± 3.4 | 52.3 ± 3.7 | 51.8 ± 4.6 | 53.7 ± 3.3 | 54.0 ± 3.3 |
| Carbohydrates, g/d | — | — | — | 170± 38 | 185 ± 27 | 202 ± 26 | 168 ± 36 | 183 ± 25 | 207 ± 28 | 170 ± 37 | 185 ± 26 | 202 ± 27 |
| Starch, g/d | 48.6 ± 13.4 | 60.3 ± 13.8 | 65.5 ± 19.5 | — | — | — | 51.5 ± 15.5 | 60.2 ± 14.7 | 63.2 ± 18.4 | 43.6 ± 12.5 | 61.1 ± 12.2 | 70.0 ± 16.8 |
| Total fiber, g/d | 13.4 ± 3.4 | 16.3 ± 3.9 | 19.2 ± 6.2 | 14.9 ± 5.4 | 16.2 ± 4.3 | 18.0 ± 5.0 | — | — | — | 14.0 ± 4.7 | 15.7 ± 3.9 | 20.3 ± 4.9 |
| Cereal fiber, g/d | 2.9 ± 1.5 | 4.2 ± 1.9 | 5.2 ± 3.2 | 2.6 ± 1.7 | 4.1 ± 1.9 | 5.8 ± 2.8 | 2.8 ± 1.1 | 3.9 ± 1.6 | 6.0 ± 3.4 | — | — | — |
| Fruit fiber, g/d | 2.3 ± 1.6 | 3.4 ± 2.0 | 4.9 ± 3.2 | 3.9 ± 3.0 | 3.5 ± 2.2 | 3.2 ± 2.1 | 1.6 ± 1.0 | 3.3 ± 1.5 | 6.0 ± 3.0 | 3.3 ± 2.7 | 3.4 ± 2.3 | 4.0 ± 2.4 |
| Vegetable fiber, g/d | 5.7 ± 2.4 | 6.2 ± 2.6 | 6.7 ± 3.5 | 6.3 ± 3.1 | 6.2 ± 2.6 | 6.2 ± 2.9 | 4.0 ± 1.3 | 6.0 ± 1.7 | 9.2 ± 3.7 | 6.3 ± 3.1 | 6.1 ± 2.6 | 6.3 ± 2.9 |
| Magnesium, mg/d | 276 ± 64 | 288 ± 73 | 299 ± 93 | 299 ± 92 | 285 ± 69 | 286 ± 76 | 244 ± 70 | 280 ± 57 | 355 ± 79 | 271 ± 79 | 274 ± 63 | 337 ± 81 |
| Cold cereal, servings/wk | 0.9 ± 1.6 | 2.0 ± 2.2 | 2.7 ± 3.2 | 1.1 ± 1.8 | 1.9 ± 2.2 | 2.6 ± 3.2 | 1.0 ± 1.8 | 1.8 ± 2.2 | 3.0 ± 3.1 | 0.6 ± 1.4 | 1.5 ± 1.9 | 4.2 ± 3.0 |
| White bread, servings/wk | 4.0 ± 5.5 | 4.8 ± 6.2 | 4.0 ± 6.0 | 2.3 ± 3.4 | 4.4 ± 5.6 | 6.7 ± 8.0 | 5.9 ± 6.8 | 4.7 ± 6.1 | 2.5 ± 4.3 | 3.4 ± 4.4 | 5.5 ± 6.8 | 3.4 ± 5.5 |
| Dark bread, servings/wk | 3.4 ± 4.6 | 4.5 ± 5.5 | 4.6 ± 5.8 | 2.9 ± 3.6 | 4.4 ± 5.3 | 5.5 ± 6.8 | 2.4 ± 3.9 | 4.4 ± 5.4 | 5.8 ± 6.2 | 1.7 ± 2.2 | 4.2 ± 4.8 | 7.0 ± 7.1 |
| English muffins/bagels, servings/wk | 1.0 ± 1.5 | 1.4 ± 1.8 | 1.3 ± 1.9 | 0.7 ± 0.9 | 1.2 ± 1.5 | 1.9 ± 2.5 | 1.1 ± 1.6 | 1.4 ± 1.8 | 1.2 ± 1.7 | 0.8 ± 1.1 | 1.5 ± 1.9 | 1.3 ± 1.9 |
| Pasta, servings/wk | 0.9 ± 0.8 | 1.1 ± 0.9 | 1.0 ± 1.0 | 0.7 ± 0.6 | 1.0 ± 0.8 | 1.3 ± 1.2 | 0.9 ± 0.8 | 1.1 ± 0.9 | 1.0 ± 1.0 | 0.8 ± 0.7 | 1.1 ± 0.9 | 1.1 ± 1.1 |
| White rice, servings/wk | 0.6 ± 0.7 | 0.8 ± 0.9 | 0.9 ± 1.4 | 0.5 ± 0.5 | 0.7 ± 0.8 | 1.1 ± 1.7 | 0.7 ± 1.1 | 0.8 ± 1.0 | 0.8 ± 1.1 | 0.7 ± 1.1 | 0.8 ± 1.1 | 0.7 ± 0.9 |
| Crackers, servings/wk | 3.3 ± 6.7 | 2.9 ± 5.9 | 2.3 ± 5.1 | 2.5 ± 5.4 | 3.0 ± 6.0 | 3.0 ± 6.3 | 2.6 ± 5.9 | 3.1 ± 6.1 | 2.7 ± 5.7 | 2.2 ± 4.8 | 2.9 ± 5.8 | 3.2 ± 6.7 |
| Potatoes, servings/wk | 1.9 ± 1.7 | 2.4 ± 1.8 | 2.4 ± 2.0 | 1.4 ± 1.4 | 2.4 ± 1.7 | 2.8 ± 1.3 | 1.8 ± 1.5 | 2.4 ± 1.8 | 2.4 ± 2.1 | 2.3 ± 2.0 | 2.3 ± 1.8 | 2.1 ± 1.7 |
Values are means ± SDs or percentages and are standardized to the age distribution of the study population. MET, metabolic equivalent task; NHS, Nurses’ Health Study; PMH, postmenopausal hormone; Q, quintile.
Value is not age-adjusted
1 cup = 237 mL.
Carbohydrates, starch, total fiber, cereal fiber, starch-to–total fiber ratio, starch-to–cereal fiber ratio, GI, and GL intake at baseline were all significantly correlated with each other (Supplemental Table 1), with correlation coefficients (r) ranging from −0.66 to 0.94. The weakest correlation was between the starch-to–total fiber ratio and cereal fiber (r = −0.03), whereas the strongest correlation was between carbohydrates and GL (r = 0.94). The correlation between starch-to–total fiber ratio and carbohydrates and starch was −0.05 and 0.58, respectively, whereas the association between the starch-to–cereal fiber ratio and carbohydrates and starch was −0.18 and −0.06, respectively.
The association between carbohydrate, starch, and total, cereal, and fruit and vegetable fiber and risk of T2D is presented in Table 2. Higher carbohydrate intake was associated with a significantly lower risk of T2D when adjusting for age (RR = 0.77; 95% CI: 0.71, 0.83; P-trend < 0.0001) and after further adjustment for BMI, race, total energy, family history of diabetes, postmenopausal status, smoking status, alcohol intake, physical activity level, and multivitamin use (RR = 0.89; 95% CI: 0.82, 0.97; P-trend = 0.007) in the highest quintile of intake compared with the lowest. However, further adjusting for dietary variables (red meat, coffee, magnesium, ratio of polyunsaturated fat to saturated fat, and trans fat) attenuated this association (RR = 0.98; 95% CI: 0.89, 1.08; P-trend = 0.84). On the other hand, starch was not significantly associated with T2D in the age-adjusted model or in the age and lifestyle–adjusted models but was positively associated with T2D after further adjustment of dietary factors (RR for highest quintile relative to lowest: 1.23; 95% CI: 1.12, 1.35; P-trend < 0.0001).
TABLE 2.
RRs (and 95% CIs) of T2D by quintiles of carbohydrate, starch, and total, cereal, fruit, and vegetable fiber intake1
| Quintile of energy-adjusted intake |
||||||
| Variable | 1 | 2 | 3 | 4 | 5 | P-trend2 |
| Carbohydrates | ||||||
| Median, g/d | 159 | 181.0 | 195 | 208.3 | 228.4 | |
| Cases/person-years | 1390/283,174 | 1508/299,048 | 1465/299,859 | 1381/303,398 | 1190/298,735 | |
| Model 1 | 1.00 (Ref) | 1.02 (0.95, 1.10) | 0.97 (0.90, 1.04) | 0.90 (0.84, 0.97) | 0.77 (0.71, 0.83) | <0.0001 |
| Model 2 | 1.00 (Ref) | 0.94 (0.87, 1.01) | 0.92 (0.85, 0.99) | 0.93 (0.86, 1.00) | 0.89 (0.82, 0.97) | 0.007 |
| Model 3 | 1.00 (Ref) | 0.96 (0.89, 1.04) | 0.96 (0.89, 1.04) | 0.99 (0.91, 1.08) | 0.98 (0.89, 1.08) | 0.84 |
| Starch | ||||||
| Median, g/d | 45.4 | 55.6 | 62.2 | 69.3 | 79.6 | |
| Cases/person-years | 1241/272,607 | 1464/298,842 | 1445/303,906 | 1456/310,195 | 1328/298,664 | |
| Model 1 | 1.00 (Ref) | 1.07 (0.99, 1.15) | 1.04 (0.96, 1.12) | 1.03 (0.95, 1.11) | 0.98 (0.91, 1.06) | 0.37 |
| Model 2 | 1.00 (Ref) | 1.04 (0.96, 1.12) | 1.00 (0.93, 1.08) | 0.99 (0.92, 1.07) | 0.98 (0.90, 1.06) | 0.28 |
| Model 3 | 1.00 (Ref) | 1.11 (1.03, 1.20) | 1.12 (1.03, 1.22) | 1.17 (1.07, 1.27) | 1.23 (1.12, 1.35) | <0.0001 |
| Total fiber | ||||||
| Median, g/d | 12.2 | 14.9 | 17.0 | 19.3 | 23.3 | |
| Cases/person-years | 1488/288,459 | 1551/302,435 | 1446/304,094 | 1357/293,442 | 1092/295,783 | |
| Model 1 | 1.00 (Ref) | 0.97 (0.90, 1.04) | 0.88 (0.82, 0.94) | 0.83 (0.77, 0.90) | 0.65 (0.60, 0.70) | <0.0001 |
| Model 2 | 1.00 (Ref) | 0.95 (0.88, 1.02) | 0.89 (0.82, 0.95) | 0.89 (0.82, 0.96) | 0.76 (0.70, 0.82) | <0.0001 |
| Model 3 | 1.00 (Ref) | 0.96 (0.89, 1.03) | 0.90 (0.83, 0.98) | 0.91 (0.84, 1.00) | 0.80 (0.72, 0.89) | <0.0001 |
| Cereal fiber | ||||||
| Median, g/d | 2.40 | 3.43 | 4.28 | 5.36 | 7.40 | |
| Cases/person-years | 1537/275,155 | 1621/304,468 | 1430/299,750 | 1360/303,460 | 986/301,381 | |
| Model 1 | 1.00 (Ref) | 0.93 (0.87, 1.00) | 0.82 (0.77, 0.89) | 0.76 (0.71, 0.82) | 0.54 (0.50, 0.59) | <0.0001 |
| Model 2 | 1.00 (Ref) | 0.90 (0.83, 0.96) | 0.83 (0.77, 0.89) | 0.81 (0.75, 0.87) | 0.69 (0.63, 0.74) | <0.0001 |
| Model 3 | 1.00 (Ref) | 0.90 (0.84, 0.97) | 0.84 (0.78, 0.91) | 0.82 (0.76, 0.89) | 0.71 (0.65, 0.78) | <0.0001 |
| Fruit fiber | ||||||
| Median, g/d | 1.45 | 2.55 | 3.55 | 4.69 | 6.68 | |
| Cases/person-years | 1504/296,218 | 1446/299,737 | 1485/304,047 | 1324/296,519 | 1175/287,692 | |
| Model 1 | 1.00 (Ref) | 0.93 (0.86, 1.00) | 0.91 (0.85, 0.98) | 0.82 (0.76, 0.88) | 0.73 (0.68, 0.79) | <0.0001 |
| Model 2 | 1.00 (Ref) | 0.90 (0.84, 0.97) | 0.91 (0.85, 0.98) | 0.83 (0.77, 0.90) | 0.79 (0.72, 0.85) | <0.0001 |
| Model 3 | 1.00 (Ref) | 0.90 (0.84, 0.97) | 0.92 (0.85, 0.99) | 0.85 (0.78, 0.92) | 0.80 (0.73, 0.88) | <0.0001 |
| Vegetable fiber | ||||||
| Median, g/d | 3.69 | 4.92 | 5.94 | 7.10 | 9.24 | |
| Cases/person-years | 1299/288,921 | 1409/304,811 | 1437/302,276 | 1440/291,878 | 1349/296,326 | |
| Model 1 | 1.00 (Ref) | 1.01 (0.93, 1.09) | 1.04 (0.97, 1.12) | 1.06 (0.98, 1.14) | 0.98 (0.91, 1.06) | 0.78 |
| Model 2 | 1.00 (Ref) | 1.01 (0.94, 1.09) | 1.07 (0.99, 1.15) | 1.08 (1.00, 1.16) | 1.00 (0.92, 1.08) | 0.86 |
| Model 3 | 1.00 (Ref) | 1.03 (0.96, 1.12) | 1.12 (1.03, 1.21) | 1.15 (1.06, 1.24) | 1.09 (1.00, 1.19) | 0.02 |
RRs and 95% CIs were calculated with the use of the Cox proportional hazards regression model. Model 1 was age-adjusted. Model 2 was adjusted for age, BMI, family history of diabetes, postmenopausal status, smoking status, alcohol intake, physical activity level, multivitamin use, race, and total energy intake. Model 3 additionally was adjusted for red meat, coffee, magnesium, ratio of polyunsaturated fat to saturated fat, and trans fat (dietary variables all in quintiles). Model for starch additionally was adjusted for cereal fiber, sugar-sweetened beverages, and fruit and vegetables. Model for total fiber additionally was adjusted for glycemic load. Models for cereal, fruit, and vegetable fibers additionally were adjusted for glycemic load and the other 2 subtypes of fiber. Ref, reference; T2D, type 2 diabetes.
Test for trend based on variable containing median value for each quintile.
We observed a significant inverse association between total, cereal, and fruit fiber and the risk of T2D. This association was significant between each of these fiber variables and the risk of T2D in the age-adjusted model and the age and lifestyle–adjusted model, and after further adjustment for dietary factors, although the effect was slightly attenuated after further adjustment for lifestyle and dietary factors. The age, lifestyle, and dietary factor–adjusted RRs of T2D for the highest quintile of intake compared with the lowest were 0.80 (95% CI: 0.72, 0.89; P-trend < 0.0001) for total fiber, 0.71 (95% CI: 0.65, 0.78; P-trend < 0.0001) for cereal fiber, and 0.80 (95% CI: 0.73, 0.88; P-trend < 0.0001) for fruit fiber. Vegetable fiber was not associated with the risk of T2D in the age-adjusted model, but after further adjustment for potential confounders, there was a marginally significant positive association in which the RR of T2D for the highest quintiles of intake compared with the lowest was 1.09 (95% CI: 1.00, 1.19; P-trend = 0.02).
The ratios of carbohydrate to total fiber and carbohydrate to cereal fiber were both positively associated with the risk of T2D (Table 3). The age-adjusted RRs of T2D of the highest quintile of intake compared with the lowest were 1.32 (95% CI: 1.23, 1.43; P-trend < 0.0001) for carbohydrate to total fiber and 1.68 (95% CI: 1.55, 1.82; P-trend < 0.0001) for the carbohydrate-to–cereal fiber ratio. However, the RRs were attenuated after further adjustment for lifestyle and dietary factors, in which the highest quintile of the ratio compared with the lowest had an RR of 1.09 (95% CI: 1.00, 1.20; P-trend = 0.04) for carbohydrate to total fiber, and 1.28 (95% CI: 1.17, 1.39; P-trend < 0.0001) for carbohydrate to cereal fiber.
TABLE 3.
RRs (and 95% CIs) of T2D by quintiles of the ratios of carbohydrate to total fiber, carbohydrate to cereal fiber, starch to total fiber, and starch to cereal fiber intake1
| Quintile of energy-adjusted intake |
||||||
| Variable | 1 | 2 | 3 | 4 | 5 | P-trend2 |
| Carbohydrate:total fiber | ||||||
| Median | 8.79 | 10.24 | 11.43 | 12.87 | 15.66 | |
| Cases/person-years | 1202/289,016 | 1321/291,930 | 1404/301,881 | 1499/307,674 | 1508/293,712 | |
| Model 1 | 1.00 (Ref) | 1.11 (1.03, 1.20) | 1.15 (1.06, 1.24) | 1.23 (1.14, 1.33) | 1.32 (1.23, 1.43) | <0.0001 |
| Model 2 | 1.00 (Ref) | 1.06 (0.98, 1.15) | 1.06 (0.98, 1.14) | 1.13 (1.05, 1.22) | 1.18 (1.09, 1.27) | <0.0001 |
| Model 3 | 1.00 (Ref) | 1.03 (0.95, 1.11) | 1.01 (0.93, 1.10) | 1.06 (0.98, 1.16) | 1.09 (1.00, 1.20) | 0.04 |
| Carbohydrate:cereal fiber | ||||||
| Median | 30.15 | 39.79 | 48.10 | 58.65 | 80.47 | |
| Cases/person-years | 1021/293,981 | 1282/300,811 | 1463/306,371 | 1616/304,833 | 1552/278,218 | |
| Model 1 | 1.00 (Ref) | 1.25 (1.15, 1.36) | 1.41 (1.30, 1.53) | 1.59 (1.47, 1.72) | 1.68 (1.55, 1.82) | <0.0001 |
| Model 2 | 1.00 (Ref) | 1.08 (0.99, 1.17) | 1.18 (1.09, 1.28) | 1.29 (1.19, 1.39) | 1.36 (1.25, 1.47) | <0.0001 |
| Model 3 | 1.00 (Ref) | 1.05 (0.96, 1.14) | 1.13 (1.04, 1.23) | 1.21 (1.12, 1.32) | 1.28 (1.17, 1.39) | <0.0001 |
| Starch:total fiber | ||||||
| Median | 2.53 | 3.21 | 3.72 | 4.27 | 5.20 | |
| Cases/person-years | 1111/283,449 | 1343/297,138 | 1444/300,078 | 1479/299,713 | 1557/303,835 | |
| Model 1 | 1.00 (Ref) | 1.17 (1.08, 1.27) | 1.28 (1.19, 1.39) | 1.35 (1.24, 1.46) | 1.44 (1.33, 1.56) | <0.0001 |
| Model 2 | 1.00 (Ref) | 1.12 (1.03, 1.21) | 1.18 (1.09, 1.27) | 1.20 (1.11, 1.30) | 1.21 (1.12, 1.31) | <0.0001 |
| Model 3 | 1.00 (Ref) | 1.08 (1.00, 1.18) | 1.12 (1.03, 1.22) | 1.13 (1.03, 1.23) | 1.12 (1.02, 1.23) | 0.03 |
| Starch:cereal fiber | ||||||
| Median | 9.82 | 13.00 | 15.50 | 18.20 | 22.70 | |
| Cases/person-years | 926/294,806 | 1250/301,681 | 1490/302,318 | 1618/299,065 | 1650/286,343 | |
| Model 1 | 1.00 (Ref) | 1.36 (1.25, 1.48) | 1.66 (1.53, 1.80) | 1.86 (1.72, 2.02) | 2.02 (1.86, 2.19) | <0.0001 |
| Model 2 | 1.00 (Ref) | 1.15 (1.05, 1.25) | 1.32 (1.22, 1.44) | 1.42 (1.31, 1.54) | 1.52 (1.40, 1.66) | <0.0001 |
| Model 3 | 1.00 (Ref) | 1.12 (1.03, 1.22) | 1.26 (1.16, 1.38) | 1.33 (1.22, 1.46) | 1.39 (1.27, 1.53) | <0.0001 |
RRs and 95% CIs were calculated with the use of the Cox proportional hazards regression model. Model 1 was age-adjusted. Model 2 was adjusted for age, BMI, family history of diabetes, postmenopausal status, smoking status, alcohol intake, physical activity level, multivitamin use, race, and total energy intake. Model 3 was adjusted additionally for red meat, coffee, magnesium, ratio of polyunsaturated fat to saturated fat, and trans fat. The starch:total fiber model additionally was adjusted for sugar-sweetened beverages. The starch:cereal fiber model additionally was adjusted for sugar-sweetened beverages and fruit and vegetable fiber. T2D, type 2 diabetes.
Test for trend based on variable containing median value for each quintile.
The ratios of starch to total fiber and starch to cereal fiber were also positively associated with the risk of T2D. The multivariate-adjusted RRs of T2D for the highest quintile of intake compared with the lowest were 1.12 (95% CI: 1.02, 1.23; P-trend = 0.03) for starch to total fiber, and 1.39 (95% CI: 1.27, 1.53; P-trend < 0.0001) for starch to cereal fiber. After further adjustment for GI and GL, separately, the association of starch-to–cereal fiber ratio with T2D was not attenuated. Continuously updating diet throughout follow-up, even after participants developed outcomes such as cardiovascular disease and cancer, did not alter our results (Supplemental Table 2). There was no effect modification of the associations between each of the 4 ratios and the risk of T2D by age, BMI, and physical activity. In addition, the association of starch-to–cereal fiber ratio is independent of diabetes risk factors, and the magnitude of association is comparable to that for physical activity and smoking (Supplemental Table 3).
In the a priori–determined stratified analysis, we found that there were significant positive associations between intake of starch and starch-to–cereal fiber ratio and risk of T2D in all 8 different subgroups (age <60 or ≥60 y; BMI <28 or ≥28; family history of diabetes, yes or no; and physical activity <10 MET-h/wk or ≥10 MET-h/wk) (Table 4). In addition, there was a significant linear trend among quintiles of intake of both starch and starch-to–cereal fiber ratio and risk of T2D among all these subgroups, except among those with family history of diabetes in which participants at the highest intake of starch had a 22% higher RR of T2D than those at the lowest quintile of intake (95% CI: 1.04, 1.44), but the P value for a linear trend was 0.07. However, there was no significant effect modification by any of these 4 variables on the associations between starch and starch-to–cereal fiber intake on risk of T2D.
TABLE 4.
RRs of T2D by quintiles of starch and starch-to–cereal fiber intake stratified by age, BMI, family history of diabetes, and physical activity1
| Quintiles of energy-adjusted intake |
|||||||
| Cases | 1 | 2 | 3 | 4 | 5 | P-trend2 | |
| Starch | |||||||
| Age <60 y | 2361 | 1.00 (Ref) | 1.13 (0.99, 1.30) | 1.08 (0.93, 1.25) | 1.16 (0.99, 1.35) | 1.27 (1.08, 1.50) | 0.007 |
| Age ≥60 y | 4573 | 1.00 (Ref) | 1.09 (0.99, 1.20) | 1.13 (1.02, 1.25) | 1.17 (1.05, 1.30) | 1.20 (1.07, 1.35) | 0.001 |
| BMI <28 kg/m2 | 2201 | 1.00 (Ref) | 1.06 (0.92, 1.22) | 1.25 (1.08, 1.45) | 1.18 (1.01, 1.38) | 1.31 (1.11, 1.55) | 0.0009 |
| BMI ≥28 kg/m2 | 4733 | 1.00 (Ref) | 1.14 (1.03, 1.25) | 1.06 (0.95, 1.17) | 1.15 (1.03, 1.28) | 1.19 (1.06, 1.34) | 0.007 |
| No family history of diabetes | 4661 | 1.00 (Ref) | 1.07 (0.97, 1.18) | 1.14 (1.03, 1.26) | 1.19 (1.07, 1.32) | 1.23 (1.09, 1.38) | 0.0002 |
| Family history of diabetes | 2273 | 1.00 (Ref) | 1.19 (1.04, 1.37) | 1.08 (0.93, 1.26) | 1.12 (0.96, 1.31) | 1.22 (1.04, 1.44) | 0.07 |
| Physical activity <10 MET-h/wk | 4064 | 1.00 (Ref) | 1.13 (1.02, 1.25) | 1.10 (0.98, 1.22) | 1.17 (1.04, 1.31) | 1.21 (1.07, 1.37) | 0.004 |
| Physical activity ≥10 MET-h/wk | 2664 | 1.00 (Ref) | 1.08 (0.95, 1.23) | 1.15 (1.00, 1.31) | 1.14 (0.99, 1.32) | 1.26 (1.08, 1.47) | 0.004 |
| Starch:cereal fiber | |||||||
| Age <60 y | 2361 | 1.00 (Ref) | 1.08 (0.91, 1.29) | 1.22 (1.03, 1.45) | 1.31 (1.10, 1.55) | 1.42 (1.19, 1.70) | <0.0001 |
| Age ≥60 y | 4573 | 1.00 (Ref) | 1.15 (1.04, 1.27) | 1.30 (1.17, 1.44) | 1.36 (1.22, 1.51) | 1.39 (1.24, 1.55) | <0.0001 |
| BMI <28 kg/m2 | 2201 | 1.00 (Ref) | 1.26 (1.09, 1.46) | 1.47 (1.26, 1.70) | 1.56 (1.34, 1.82) | 1.70 (1.45, 2.00) | <0.0001 |
| BMI ≥28 kg/m2 | 4733 | 1.00 (Ref) | 1.06 (0.95, 1.18) | 1.17 (1.05, 1.31) | 1.22 (1.09, 1.36) | 1.25 (1.11, 1.40) | <0.0001 |
| No family history of diabetes | 4661 | 1.00 (Ref) | 1.07 (0.96, 1.19) | 1.26 (1.13, 1.40) | 1.31 (1.17, 1.46) | 1.35 (1.20, 1.51) | <0.0001 |
| Family history of diabetes | 2273 | 1.00 (Ref) | 1.21 (1.05, 1.41) | 1.26 (1.08, 1.46) | 1.36 (1.17, 1.59) | 1.47 (1.25, 1.73) | <0.0001 |
| Physical activity <10 MET-h/wk | 4064 | 1.00 (Ref) | 1.18 (1.04, 1.33) | 1.31 (1.16, 1.47) | 1.37 (1.22, 1.55) | 1.36 (1.20, 1.54) | <0.0001 |
| Physical activity ≥10 MET-h/wk | 2664 | 1.00 (Ref) | 1.04 (0.91, 1.19) | 1.20 (1.05, 1.38) | 1.28 (1.11, 1.47) | 1.52 (1.31, 1.76) | <0.0001 |
RRs and 95% CIs were calculated with the use of the Cox proportional hazards regression model. The analysis was adjusted for age, BMI, family history of diabetes, postmenopausal status, smoking status, alcohol intake, physical activity level, multivitamin use, race, total energy intake, red meat, coffee, magnesium, ratio of polyunsaturated fat to saturated fat, and trans fat. Models for starch were additionally adjusted for cereal fiber, sugar-sweetened beverages, and fruits and vegetables. The starch:cereal fiber models were additionally adjusted for sugar-sweetened beverages and fruit and vegetable fiber. MET, metabolic equivalent task; Ref, reference; T2D, type 2 diabetes.
Test for trend based on variable containing median value for each quintile.
We looked at the joint effects of carbohydrate and total fiber, carbohydrate and cereal fiber, starch and total fiber, and starch and cereal fiber on the risk of T2D by classifying subjects by both variables (Figure 1). The multivariate-adjusted RRs of a diet high in carbohydrates and low in total fiber compared with the extreme opposite of that was 1.26 (95% CI: 1.07, 1.48). For a diet high in carbohydrates and low in cereal fiber the RR was 1.23 (95% CI: 1.06, 1.42); for a diet high in starch and low in total fiber the RR was 1.08 (95% CI: 0.95, 1.25); and for a diet high in starch and low in cereal fiber the RR was 1.43 (95% CI: 1.20, 1.75). However, there was no significant interaction between any of these variables, jointly, and risk of T2D (P ≥ 0.09).
FIGURE 1.
RR of T2D by joint effect analysis of carbohydrate or starch and total or cereal fiber intake (cases = 6934, person-years = 1,484,213). Adjusted RR of T2D according to joint classifications of carbohydrates and total fiber (A), carbohydrates and cereal fiber (B), starch and total fiber (C), and starch and cereal fiber (D). The analysis was adjusted for age, BMI, family history of diabetes, postmenopausal status, smoking status, alcohol intake, physical activity level, multivitamin use, race, total energy intake, red meat, coffee, magnesium, ratio of polyunsaturated fat to saturated fat, and trans fat. Panel C is additionally adjusted for sugar-sweetened beverages. Panel D is additionally adjusted for sugar-sweetened beverages and fruit and vegetable fiber. RRs were calculated with the use of the Cox proportional hazards regression model. P values for interactions are presented below. T2D, type 2 diabetes.
We also examined the association between starch and starch-to–cereal fiber ratio and the risk of T2D with the use of spline regression (Figure 2). Starch had a significant positive association with the risk of T2D (P for nonlinearity = 0.44). However, there seemed to be a nonlinear association between starch-to–cereal fiber ratio and risk of T2D (P for nonlinearity = 0.01), in which the association reaches a plateau after a starch-to–cereal fiber ratio of 15.
FIGURE 2.
RR of T2D by starch (A) and starch-to–cereal fiber (B) intake with the use of restricted cubic spline regression. (A) P-nonlinear relation = 0.44, and P-linear relation < 0.0001. (B) P-nonlinear relation = 0.01, and P-overall significance of the curve < 0.0001. The analysis was adjusted for age, BMI, family history of diabetes, postmenopausal status, smoking status, alcohol intake, physical activity level, multivitamin use, race, total energy intake, red meat, coffee, magnesium, ratio of polyunsaturated fat to saturated fat, and trans fat. The model for starch was additionally adjusted for cereal fiber, sugar-sweetened beverages, and fruits and vegetables. The starch-to–cereal fiber model was additionally adjusted for sugar-sweetened beverages and fruit and vegetable fiber. T2D, type 2 diabetes.
DISCUSSION
In this large prospective cohort of US women, we found that a higher intake of starch and a lower intake of total, cereal, and fruit fiber were associated with a higher risk of T2D. In addition, all 4 ratios of carbohydrates- or starch-to–total or –cereal fiber were positively associated with risk of T2D, and the strongest association was between starch-to–cereal fiber and risk of T2D, which reflects a diet rich in highly refined and processed grains.
We did not find an association between total carbohydrate intake and the risk of T2D, consistent with some previous findings (7–12). A plausible explanation for this is confounding by healthier subtypes of carbohydrates such as whole grains, fruits, vegetables, and fiber. A marginally significant positive association was observed in the European Prospective into Cancer and Nutrition–Netherlands cohort study after adjusting for potential confounders, including fiber, indicating that carbohydrate quality rather than quantity is more important in developing T2D (13). Furthermore, consistent with our findings, starch intake was associated with a 25% (95% CI: 1.07, 1.46) higher risk of T2D when comparing extreme quintiles in the European Prospective Investigation into Cancer and Nutrition–Netherlands cohort (13). Additionally, a study of Australian adults found that for every 100 g of starch consumed, the likelihood of T2D was 1.52 (95% CI: 1.09, 2.11) (11). However, 4 other cohort studies found no association between starch and T2D (7, 9, 12, 14), but none adjusted for cereal fiber intake, which is an important confounder, and adjusting for it made the association significant in our study. It should also be noted that starch is an entity that does not allow us to distinguish between amylose and amylopectin, which have differential effects on glycemic responses and may have a different impact on T2D risk.
Our findings are consistent with those of a meta-analysis of 9 prospective cohort studies analyzing fiber by source and risk of T2D, showing that cereal fiber is associated with a 33% (95% CI: 28, 38) lower risk of T2D when comparing extreme quintiles (6). Yet, unlike our findings, there was no association between fruit or vegetable fiber and T2D risk, with a RR of 0.89 (95% CI: 0.70, 1.13) and 0.93 (95% CI: 0.74, 1.17), respectively, comparing extreme quintiles (6). This discrepancy could be due to the fact that these studies, including an earlier analysis of the NHS, did not adjust for dietary factors (8, 9, 11, 31, 32), or for other subtypes of fiber (6, 33), or were conducted in a younger-age cohort (10). The lack of an inverse association between vegetable fiber and T2D risk in our study could be due to the food sources of vegetable fiber, including potatoes, white bread, and white rice. These foods, which have been associated with increased glycemia, may have counterbalanced the effects of vegetable fiber.
In the multivariable analysis, the associations between the ratios of carbohydrate or starch to total fiber and risk of T2D were borderline significant. However, the ratios of carbohydrate or starch to cereal fiber were associated with a higher risk of T2D, in which nurses at extreme quintiles of intake had a 27% and 39% higher risk of T2D, respectively, because cereal fiber has a stronger association with T2D risk than total fiber. Individually, none of these nutrients provide an association as strong as the starch-to–cereal fiber ratio, indicating a potential biological interaction between these nutrients that could explain these findings. A high starch-to–cereal fiber ratio identifies highly processed and refined grain products. Several mechanisms could explain the association between the lower ratios and the reduced risk of T2D. Fiber improves postprandial glucose and insulin response by both slowing sugar absorption and causing a bulking effect in the stomach, and both mechanisms delay the rate of hunger return and/or increase duration of satiety, hence reducing energy intake (34, 35). The benefits of cereal fiber on T2D risk partially could be due to the micronutrients available in cereal fiber–rich foods such as vitamins E and B and magnesium (36, 37).
Nearly 85% of our study population had a carbohydrate-to-fiber ratio >10:1 at baseline (Supplemental Figure 1). This ratio has been proposed for selecting individual foods; therefore, a higher ratio is expected in our cohort, because it measures the carbohydrate content of all foods. Nonetheless, this ratio illustrates the low intake of fiber-rich foods and high intake of refined carbohydrates. Participants consuming diets high in carbohydrates or starch and low in cereal fiber were at higher risk of T2D than participants consuming diets low in carbohydrates or starch and high in cereal fiber. In previous analyses in the NHS, NHS II, and Health Professionals Follow-Up Study, subjects who consumed diets high in GI and low in cereal fiber were also found to be at higher risk of T2D than those who consumed diets low in GI and high in cereal fiber (8, 10, 31). The carbohydrate- or starch-to-fiber ratios not only reflect the carbohydrate and fiber content of the food but may also reflect the GI, GL, whole-grain, and micronutrient content of diets. Therefore, eating whole grains and having a low GI is associated with lower carbohydrate- or starch-to-fiber ratios and a lower risk of T2D.
Our results remained robust in several sensitivity analyses. Updating diet until end of follow-up only mildly attenuated the results, suggesting that even though participants might have changed their diet after such diagnoses, it mildly affected their T2D risk. Also, further adjusting for hypertension and hypercholesterolemia, 2 independent risk factors of T2D that can be in the causal pathway for T2D development (38–41), did not substantially change our results.
The strengths of this study include large sample size, long duration of follow-up, high follow-up rates, large number of T2D cases, and availability of fiber subtypes. In addition, because of the prospective design, any measurement errors in exposure variables were independent of the outcome, and therefore attenuated the association toward the null. To reduce within-person variability and better represent long-term diet, the cumulative average from all available FFQs was used as the exposure variable (17). To reduce the degree of recall bias from the presence of other relevant chronic disease, all participants with diet-altering chronic diseases were excluded at baseline, and diet intake was not updated when a participant developed a relevant chronic disease during follow-up. There are several limitations to our study. First, all participants were nurses, mainly Caucasian, which increases internal validity but may decrease generalizability to other populations, although it is unlikely the association will be largely different in the general population. Second, this was an observational study, so residual and unmeasured confounding are possible, although we adjusted for all known and potential risk factors of T2D. Third, although exposure variables were measured with the use of validated FFQs, there is some degree of nondifferential misclassification (18). In addition, FFQs also do not have detailed information regarding the degree of food processing. Finally, the accuracy of the dietary fiber calculations via the USDA food composition tables is uncertain because of the natural variation of foods (42). The use of biomarkers in nutritional studies should be considered in future research.
In conclusion, starch was positively associated with T2D risk, whereas total, cereal, and fruit fiber were inversely associated with it. Ratios of carbohydrate to cereal fiber, starch to total fiber, and starch to cereal fiber were all positively associated with T2D risk, suggesting the possibility of a biological interaction between these individual variables. The ratios provide a novel way of evaluating the carbohydrate quality of the entire diet, warranting further research with other health endpoints.
Acknowledgments
The authors’ responsibilities were as follows—HBA, WCW, and FBH: designed the analysis; HBA: conducted the analysis, interpreted the data, and wrote the manuscript; SNB, VSM, NMW, HC, BR, WCW, and FBH: assisted in interpreting the data and edited the manuscript; WCW and FBH: obtained funding and managed and conducted the Nurses’ Health Study; HBA and FBH: had primary responsibility for final content; and all authors: critically reviewed the manuscript for important intellectual content and read and approved the final manuscript. None of the authors reported a conflict of interest related to the study.
Footnotes
Abbreviations used: FFQ, food-frequency questionnaire; GI, glycemic index; GL, glycemic load; MET, metabolic equivalent task; NHS, Nurses’ Health Study; T2D, type 2 diabetes.
REFERENCES
- 1.International Diabetes Federation. Diabetes Atlas. 5th ed. Brussels (Belgium): International Diabetes Federation; 2011.
- 2.Hu FB, van Dam RM, Liu S. Diet and risk of Type II diabetes: The role of types of fat and carbohydrate. Diabetologia 2001;44:805–17. [DOI] [PubMed] [Google Scholar]
- 3.Willett W, Manson J, Liu S. Glycemic index, glycemic load, and risk of type 2 diabetes. Am J Clin Nutr 2002;76:274S–80S. [DOI] [PubMed] [Google Scholar]
- 4.Barclay AW, Petocz P, McMillan-Price J, Flood VM, Prvan T, Mitchell P, Brand-Miller JC. Glycemic index, glycemic load, and chronic disease risk–a meta-analysis of observational studies. Am J Clin Nutr 2008;87:627–37. [DOI] [PubMed] [Google Scholar]
- 5.de Munter JS, Hu FB, Spiegelman D, Franz M, van Dam RM. Whole grain, bran, and germ intake and risk of type 2 diabetes: a prospective cohort study and systematic review. PLoS Med 2007;4:e261. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Schulze MB, Schulz M, Heidemann C, Schienkiewitz A, Hoffmann K, Boeing H. Fiber and magnesium intake and incidence of type 2 diabetes: a prospective study and meta-analysis. Arch Intern Med 2007;167:956–65. [DOI] [PubMed] [Google Scholar]
- 7.Barclay AW, Flood VM, Rochtchina E, Mitchell P, Brand-Miller JC. Glycemic index, dietary fiber, and risk of type 2 diabetes in a cohort of older Australians. Diabetes Care 2007;30:2811–3. [DOI] [PubMed] [Google Scholar]
- 8.Salmerón J, Manson JE, Stampfer MJ, Colditz GA, Wing AL, Willett WC. Dietary fiber, glycemic load, and risk of non-insulin-dependent diabetes mellitus in women. JAMA 1997;277:472–7. [DOI] [PubMed] [Google Scholar]
- 9.Meyer KA, Kushi LH, Jacobs DR Jr., Slavin J, Sellers TA, Folsom AR. Carbohydrates, dietary fiber, and incident type 2 diabetes in older women. Am J Clin Nutr 2000;71:921–30. [DOI] [PubMed] [Google Scholar]
- 10.Schulze MB, Liu S, Rimm EB, Manson JE, Willett WC, Hu FB. Glycemic index, glycemic load, and dietary fiber intake and incidence of type 2 diabetes in younger and middle-aged women. Am J Clin Nutr 2004;80:348–56. [DOI] [PubMed] [Google Scholar]
- 11.Hodge AM, English DR, O’Dea K, Giles GG. Glycemic index and dietary fiber and the risk of type 2 diabetes. Diabetes Care 2004;27:2701–6. [DOI] [PubMed] [Google Scholar]
- 12.Schulze MB, Schulz M, Heidemann C, Schienkiewitz A, Hoffmann K, Boeing H. Carbohydrate intake and incidence of type 2 diabetes in the European Prospective Investigation into Cancer and Nutrition (EPIC)-Potsdam Study. Br J Nutr 2008;99:1107–16. [DOI] [PubMed] [Google Scholar]
- 13.Sluijs I, van der Schouw YT. van der AD, Spijkerman AM, Hu FB, Grobbee DE, Beulens JW. Carbohydrate quantity and quality and risk of type 2 diabetes in the European Prospective Investigation into Cancer and Nutrition-Netherlands (EPIC-NL) study. Am J Clin Nutr 2010;92:905–11. [DOI] [PubMed] [Google Scholar]
- 14.Janket SJ, Manson JE, Sesso H, Buring JE, Liu S. A prospective study of sugar intake and risk of type 2 diabetes in women. Diabetes Care 2003;26:1008–15. [DOI] [PubMed] [Google Scholar]
- 15.Lloyd-Jones DM, Hong Y, Labarthe D, Mozaffarian D, Appel LJ, Van Horn L, Greenlund K, Daniels S, Nichol G, Tomaselli GF, et al. Defining and setting national goals for cardiovascular health promotion and disease reduction: the American Heart Association’s strategic Impact Goal through 2020 and beyond. Circulation 2010;121:586–613. [DOI] [PubMed] [Google Scholar]
- 16.Mozaffarian RS, Lee RM, Kennedy MA, Ludwig DS, Mozaffarian D, Gortmaker SL. Identifying whole grain foods: a comparison of different approaches for selecting more healthful whole grain products. Public Health Nutr 2013;16:2255–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Hu FB, Stampfer MJ, Rimm E, Ascherio A, Rosner BA, Spiegelman D, Willett WC. Dietary fat and coronary heart disease: a comparison of approaches for adjusting for total energy intake and modeling repeated dietary measurements. Am J Epidemiol 1999;149:531–40. [DOI] [PubMed] [Google Scholar]
- 18.Willett WC, Sampson L, Stampfer MJ, Rosner B, Bain C, Witschi J, Hennekens CH, Speizer FE. Reproducibility and validity of a semiquantitative food frequency questionnaire. Am J Epidemiol 1985;122:51–65. [DOI] [PubMed] [Google Scholar]
- 19.Salvini S, Hunter DJ, Sampson L, Stampfer MJ, Colditz GA, Rosner B, Willett WC. Food-based validation of a dietary questionnaire: the effects of week-to-week variation in food consumption. Int J Epidemiol 1989;18:858–67. [DOI] [PubMed] [Google Scholar]
- 20.Rimm EB, Giovannucci EL, Stampfer MJ, Colditz GA, Litin LB, Willett WC. Reproducibility and validity of an expanded self-administered semiquantitative food frequency questionnaire among male health professionals. Am J Epidemiol 1992;135:1114–26, discussion 27–36. [DOI] [PubMed] [Google Scholar]
- 21.Feskanich D, Rimm EB, Giovannucci EL, Colditz GA, Stampfer MJ, Litin LB, Willett WC. Reproducibility and validity of food intake measurements from a semiquantitative food frequency questionnaire. J Am Diet Assoc 1993;93:790–6. [DOI] [PubMed] [Google Scholar]
- 22.Willett WC. Nutritional epidemiology. New York: Oxford University Press; 1998. [Google Scholar]
- 23.Nutrient Data Laboratory (US), Consumer and Food Economics Institute. USDA nutrient database for standard reference [CD-ROM]. Riverdale (MD): USDA, Nutrient Data Laboratory, Agricultural Research Service; 1999. [Google Scholar]
- 24.Willett WC, Howe GR, Kushi LH. Adjustment for total energy intake in epidemiologic studies. Am J Clin Nutr 1997;65:1220S–8S; discussion 9S–31S. [DOI] [PubMed] [Google Scholar]
- 25.Classification and diagnosis of diabetes mellitus and other categories of glucose intolerance. National Diabetes Data Group. Diabetes 1979;28:1039–57. [DOI] [PubMed] [Google Scholar]
- 26.Report of the Expert Committee on the Diagnosis and Classification of Diabetes Mellitus. Diabetes Care 1997;20:1183–97. [DOI] [PubMed] [Google Scholar]
- 27.Manson JE, Rimm EB, Stampfer MJ, Colditz GA, Willett WC, Krolewski AS, Rosner B, Hennekens CH, Speizer FE. Physical activity and incidence of non-insulin-dependent diabetes mellitus in women. Lancet 1991;338:774–8. [DOI] [PubMed] [Google Scholar]
- 28.Hu FB, Leitzmann MF, Stampfer MJ, Colditz GA, Willett WC, Rimm EB. Physical activity and television watching in relation to risk for type 2 diabetes mellitus in men. Arch Intern Med 2001;161:1542–8. [DOI] [PubMed] [Google Scholar]
- 29.Wolf AM, Hunter DJ, Colditz GA, Manson JE, Stampfer MJ, Corsano KA, Rosner B, Kriska A, Willett WC. Reproducibility and validity of a self-administered physical activity questionnaire. Int J Epidemiol 1994;23:991–9. [DOI] [PubMed] [Google Scholar]
- 30.Harrell FE., Jr Regression modeling strategies: with applications to linear models, logistic regression, and survival analysis. New York: Springer-Verlag; 2001. [Google Scholar]
- 31.Salmerón J, Ascherio A, Rimm EB, Colditz GA, Spiegelman D, Jenkins DJ, Stampfer MJ, Wing AL, Willett WC. Dietary fiber, glycemic load, and risk of NIDDM in men. Diabetes Care 1997;20:545–50. [DOI] [PubMed] [Google Scholar]
- 32.Stevens J, Ahn K, Juhaeri, Houston D, Steffan L, Couper D. Dietary fiber intake and glycemic index and incidence of diabetes in African-American and white adults: the ARIC study. Diabetes Care 2002;25:1715–21. [DOI] [PubMed] [Google Scholar]
- 33.Montonen J, Knekt P, Jarvinen R, Aromaa A, Reunanen A. Whole-grain and fiber intake and the incidence of type 2 diabetes. Am J Clin Nutr 2003;77:622–9. [DOI] [PubMed] [Google Scholar]
- 34.Jenkins DJ, Wolever TM, Leeds AR, Gassull MA, Haisman P, Dilawari J, Goff DV, Metz GL, Alberti KG. Dietary fibres, fibre analogues, and glucose tolerance: importance of viscosity. BMJ 1978;1:1392–4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Howarth NC, Saltzman E, Roberts SB. Dietary fiber and weight regulation. Nutr Rev 2001;59:129–39. [DOI] [PubMed] [Google Scholar]
- 36.Bakker SJ, Hoogeveen EK, Nijpels G, Kostense PJ, Dekker JM, Gans RO, Heine RJ. The association of dietary fibres with glucose tolerance is partly explained by concomitant intake of thiamine: the Hoorn Study. Diabetologia 1998;41:1168–75. [DOI] [PubMed] [Google Scholar]
- 37.InterAct Consortium. Dietary fibre and incidence of type 2 diabetes in eight European countries: the EPIC-InterAct Study and a meta-analysis of prospective studies. Diabetologia 2015;58:1394–408. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.McPhillips JB, Barrett-Connor E, Wingard DL. Cardiovascular disease risk factors prior to the diagnosis of impaired glucose tolerance and non-insulin-dependent diabetes mellitus in a community of older adults. Am J Epidemiol 1990;131:443–53. [DOI] [PubMed] [Google Scholar]
- 39.Conen D, Ridker PM, Mora S, Buring JE, Glynn RJ. Blood pressure and risk of developing type 2 diabetes mellitus: the Women’s Health Study. Eur Heart J 2007;28:2937–43. [DOI] [PubMed] [Google Scholar]
- 40.Wei GS, Coady SA, Goff DC Jr, Brancati FL, Levy D, Selvin E, Vasan RS, Fox CS. Blood pressure and the risk of developing diabetes in african americans and whites: ARIC, CARDIA, and the framingham heart study. Diabetes Care 2011;34:873–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Wilson PW, Meigs JB, Sullivan L, Fox CS, Nathan DM, D’Agostino RB Sr. Prediction of incident diabetes mellitus in middle-aged adults: the Framingham Offspring Study. Arch Intern Med 2007;167:1068–74. [DOI] [PubMed] [Google Scholar]
- 42.Greenfield H, Southgate DAT, Food and Agriculture Organization of the United Nations. Food composition data: production, management, and use. Rome: FAO; 2003. [Google Scholar]