Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 Jun 1.
Published in final edited form as: Am J Clin Nutr. 2026 Apr 30;124(1):101337. doi: 10.1016/j.ajcnut.2026.101337

Optimal Measures of Carbohydrate Quality to Lower the Risk of Type 2 Diabetes: Findings from Three Prospective Cohort Studies

Hala B AlEssa 1, Tong Xia 1, Seyed Mohammad Mousavi 1, Caleigh Sawicki 1, Walter C Willett 1, Shilpa N Bhupathiraju 1,*, Frank B Hu 1,*
PMCID: PMC13222467  NIHMSID: NIHMS2179665  PMID: 42069223

Abstract

Background:

Numerous carbohydrate quality metrics (CQMs) have been suggested, yet the optimal one(s) associated with the lowest type 2 diabetes (T2D) risk remains unknown.

Objective:

We aimed to systematically compare 23 CQMs with T2D risk, identify the five strongest associations, propose an alternate Carbohydrate Quality Index (aCQI), and compare it with the existing CQI regarding T2D risk and cardiometabolic biomarkers.

Methods:

We included participants of three prospective cohort studies (Nurses’ Health Study I (1984–2020) and II (1991–2019), and Health Professionals Follow-up Study (1986–2020), free of cancer, diabetes, and cardiovascular disease. Our primary outcome was incident T2D. We examined thirteen plasma biomarkers in relation to CQIs among a subset.

Results:

During 5,628,955 person-years of follow-up among 213,704 adults, 22,351 cases of incident T2D were ascertained. In multivariable-adjusted models, comparing Q5 to Q1, intakes of cereal fiber (relative risk (RR):0.77 (0.74 – 0.81)), whole-fruit carbohydrates (RR:0.80 (0.76 – 0.84)), glycemic index (RR:1.20 (1.14 –1.26)), sugar from sugar-sweetened beverages (RR:1.22 (1.17–1.28)), and whole grain carbohydrates (RR:0.86 (0.82 – 0.91)) had the strongest associations with T2D risk. The aCQI (RR:0.71 (0.68 – 0.75)), comprising these variables, had a larger magnitude of association with T2D risk than the original CQI (RR:0.82 (0.79 – 0.87)), which included total fiber intake, glycemic index, the ratios of whole to total grains and solid to total carbohydrates. The aCQI had significant associations with larger percentage of differences in cardiometabolic biomarker concentrations, such as C-peptide, leptin, and LDL-cholesterol, than the CQI (all P-trend ≤ 0.001).

Conclusion:

The novel aCQI, comprised of carbohydrates from whole fruits, whole grains, sugar-sweetened beverages, cereal fiber, and glycemic index, was more strongly associated with risk of T2D and cardiometabolic biomarkers than its individual components or the existing CQI, necessitating further research.

Keywords: carbohydrate quality, type 2 diabetes, dietary fiber, whole grains, glycemic index, sugar-sweetened beverages, fruit juice, alternate carbohydrate quality index

INTRODUCTION

Globally, 11% of adults have diabetes, and 12% have pre-diabetes in 2024 (1). In 184 countries, over 70% of incident T2D was attributable to suboptimal diet in 2018 (2). Of all food groups, carbohydrates have some of the strongest associations with T2D risk. A high intake of refined grains, potatoes – especially French fries, sugar-sweetened beverages (SSBs), and juice, coupled with a low intake of whole grains and fruit, elevates the risk of T2D (2–5).

The quality of carbohydrates may be more consequential in developing T2D than the quantity of carbohydrates (3, 6). This is due to the wide variety of carbohydrate foods, ranging from whole grains and whole fruits to starchy vegetables, legumes, and potatoes to processed foods high in added sugar and sugary beverages. They differ in source, form, components, degree of processing, postprandial glycemic effects, and other dimensions. Thus, carbohydrate quality metrics (CQMs) capture different dimensions of this heterogeneity. Existing CQMs vary substantially: some are single variables like dietary fiber and glycemic index (GI), while others are composite measures of carbohydrate quality, such as the carbohydrate to fiber ratio and the Carbohydrate Quality Index (CQI), which includes total fiber, GI, whole to total grain ratio, and solid to total carbohydrate ratio. These differences have important consequences for assessing T2D risk. Single-variable CQMs may misclassify the overall quality of carbohydrate in the diet because they reflect only one dimension without accounting for other important attributes like food matrix, degree of processing, or food sources (e.g., whole grains, whole fruits). Thus, variation in what each CQM measures can lead to divergent associations with T2D, highlighting the need for comparative evaluation.

The CQI, first proposed by Zazpe et al. (2014) within the SUN Project, was developed to assess carbohydrate quality through multiple dietary components. Since its introduction, it has been applied in diverse populations and health contexts, including metabolic, cardiovascular, and cancer research (7–11). Because CQI integrates several components relevant to T2D, including dietary fiber intake, glycemic response, whole-grain consumption, and degree of processing, it is increasingly recognized as one of the most comprehensive CQMs currently available. For these reasons, it serves as an appropriate and widely used benchmark against which to compare alternative CQMs.

Many CQMs have been proposed to assess carbohydrate quality, yet the optimal metric associated with the lowest risk of T2D remains unknown. This study aims to systematically compare 23 CQMs with risk of T2D. It will identify the five individual CQMs most strongly associated with T2D and capturing distinct, complementary dimensions of carbohydrate quality, forming the basis for a new alternative CQI (aCQI). Finally, the new aCQI will be compared to the existing CQI in relation to T2D and cardiometabolic biomarkers.

METHODS

Study population

The study population included three prospective cohort studies: the Nurses’ Health Study (NHS), NHSII, and the Health Professionals Follow-up Study (HPFS). The NHS, initiated in 1976, included 121,701 female nurses (30–55y). The HPFS started in 1986 with 51,529 male health professionals (40–75y). NHSII, initiated in 1989, included 116,671 female nurses (25–42y). Participants were followed up biennially using questionnaires to update their medical history, lifestyle, and incidence of chronic disease. At study baseline (NHS: 1984, HPFS: 1986, NHSII: 1991), we included participants who completed the baseline study questionnaire and the validated food frequency questionnaire (FFQ). We excluded participants who had implausible or missing caloric intake (<500 or >3500 kcal/d for women; <800 or >4200 for men), missing carbohydrate intake information, missing baseline age, or had T2D, cardiovascular disease, or cancer.

A single blood sample was collected from each participant in nested case-control studies of T2D, cardiovascular disease, and several cancers (including breast, colorectal, endometrial, ovarian, and prostate cancers) within each cohort. Samples were obtained from 32,826 women in the NHS (1989 – 1990), 18,159 men in the HPFS (1993–1995), and 29,611 women in NHSII (1996–1999). The same inclusion and exclusion criteria described above were applied. The Institutional Review Boards of the Harvard T.H. Chan School of Public Health and the Brigham and Women’s Hospital have approved the study protocol.

Dietary assessment

Dietary data was collected at baseline and every 4y using a validated semi-quantitative FFQ, where participants were asked on average how often they consumed a specific portion or serving size of each food over the past year, with responses ranging from “never” to “six or more times per week” (12–14). The frequency of consumption was multiplied by the nutrient content of the specific portion sizes of each dietary item (obtained from the USDA food composition tables and supplemented by manufacturer data) to calculate the nutrient intakes (15).

We first evaluated 23 individual or composite (ratio) carbohydrate variables, including intakes of total carbohydrate, starch, total fiber, cereal fiber (from grains, bran, oats, etc.), total glucose equivalents (TGE), total fructose equivalents (TFE), whole and refined grain carbohydrate, whole-fruit carbohydrate, fruit juice sugar, SSB sugar, added sugar, vegetable carbohydrate, starchy vegetables (peas, corn, sweet potatoes/yams(16)), non-starchy vegetables (all vegetables except starchy vegetables), potato carbohydrate, legume carbohydrate, GI, glycemic load (GL) and the ratios of carbohydrate or starch to total or cereal fiber. TGE was calculated by summing the glucose monosaccharide and glucose from sucrose, maltose, lactose, and starch, and TFE was calculated by summing fructose monosaccharide with fructose from sucrose (17). The methods used to calculate GI and GL have been described in detail previously (18). The CQI, described earlier, is a carbohydrate quality score that includes four factors. After systematically evaluating the associations of all single-variable CQMs with T2D, we identified five components that (1) showed the strongest and most consistent associations with T2D risk and (2) represented distinct and complementary dimensions of carbohydrate quality, including factors related to fiber content, food sources, processing, and glycemic impact. Similar to the original four-component CQI the aCQI includes five components: cereal fiber, GI (reverse scored), whole grain carbohydrate, SSB carbohydrate (reverse scored), and whole-fruit carbohydrate (Supplemental table 1). All dietary intakes were adjusted for energy intake using the residual method to control for confounding and reduce extraneous variation by body size, metabolic efficiency, and physical activity.

Ascertainment of incident T2D

Participants self-reported incident T2D on the biennial questionnaires, and if so, they are sent a validated supplementary questionnaire asking about symptoms, diagnostic tests, and T2D medication. The criteria of the National Diabetes Data Group were used to confirm cases before 1998 (19). From 1998, criteria of the American Diabetes Association were used to confirm cases (19). Only cases of participants who completed the supplementary questionnaires were included in the analysis.

Assessment of biomarkers

This study included thirteen cardiometabolic plasma biomarkers previously associated with the etiology of cardiometabolic disease such as T2D for which measurements were available in our cohorts. They include insulinemic/glycemic markers (C-peptide, HbA1c), inflammatory markers (C-reactive protein (CRP), interleukin-6 (IL-6), tumor necrosis factor receptors (TNF-R1, TNF-R2), intracellular adhesion molecule 1 (ICAM-1), adiponectin, leptin), and blood lipids (HDL-cholesterol (HDL-C), LDL-cholesterol (LDL-C), triglyceride (TG), and TG/HDL-C ratio). Upon receipt, mailed blood samples were centrifuged, aliquoted, and stored in liquid nitrogen (−150°C) until analysis. The protocol and procedures for blood collection in the three cohorts have been reported previously in detail (20). Biomarkers were collected using standard methods, and quality control samples were routinely frozen along with study samples to monitor for plasma changes due to long-term storage and to monitor assay variability. All intra-assay coefficients of variation for the biomarkers were less than 20%.

Fasting plasma concentrations of C-peptide, interleukin-6 (IL-6), intracellular adhesion molecule 1 (ICAM-1), and tumor necrosis factor receptors 1 and 2 (TNF-R1 and TNF-R2) were measured using enzyme-linked immunosorbent assay (ELISA) or radioimmunoassay methods, as previously described (21–23). Glycated hemoglobin (HbA1c) was quantified by temperature controlled high-performance liquid chromatography (HPLC) (24). Total, high-density lipoprotein (HDL), and directly measured low-density lipoprotein (LDL) cholesterol and triglycerides were determined by standardized enzymatic procedures using reagents from Roche Diagnostics (Indianapolis, IN) and Genzyme (Cambridge, MA) (25). C-reactive protein (CRP) was measured using a high-sensitivity immunoturbidimetric assay (Denka Seiken, Niigata, Japan).

Assessment of covariates

Through biennial questionnaires, we collected updated information on age (years), race/ethnicity (white, black, native American, Asian, Hawaiian), smoking status (never, past, current 1–14 cigs/day, current 15 – 24 cigs/day, or current ≥ 25 cigs/day), physical activity level (<3, 3 - <9, 9 - <18, 18 - <27, 27 - <42, or ≥ 42 MET-hrs/wk), multivitamin use (yes/no), and family history of diabetes (yes/no), and among women information on menopausal status and postmenopausal hormone use (premenopause, never user, current user, past user), and oral contraceptive use (current, past, never; NHSII only) was collected. Height was reported at cohort baseline, and weight was collected biennially to calculate BMI (<23, 23-<25, 25-<27.5, 27.5-<30, or ≥ 30 kg/m2). Intakes of alcohol intake (0, 0.1 – 4.9, 5.0 – 14.9. or 15 g/d), red meat (quintiles of servings/day), coffee (quintiles of servings/day), and polyunsaturated, saturated, and trans fats (all in quintiles of daily % energy intake) were ascertained every 4y from FFQs. All covariates, apart from race/ethnicity and family history of diabetes, were time-varying.

Statistical analysis

We calculated participants’ person-time from the date of return of the baseline questionnaire until the date of T2D diagnosis, death, or end of follow-up (NHS: June 30, 2020, HPFS: January 31, 2020, NHSII: June 30, 2019), whichever was earlier. Pearson correlation coefficients were used to examine the correlations between the twenty-five CQMs at baseline. We used age-stratified time-varying Cox proportional hazards models to calculate the relative risk (RR) and 95% confidence intervals (95% CI) between each CQM, in quintiles, and the risk of T2D. To test for linear trend, we assigned the median values of each quintile to a continuous variable for each main exposure. The first model was age-adjusted. The second model was further adjusted for family history of diabetes, smoking status, alcohol intake, physical activity level, multivitamin use, race, menopausal status and postmenopausal hormone use (NHS and NHSII), oral contraceptive use (NHSII), total energy, and intakes of red meat, polyunsaturated fat, saturated fat, trans fat, and coffee. Models with cereal fiber as the main exposure were adjusted for fruit and vegetable fiber. Whole and refined grain models were mutually adjusted. A third model further adjusted for BMI. We used cumulative averages of all dietary variables to better reflect long-term diet and reduce measurement errors. We did not update diet after the development of major outcomes such as cancer or cardiovascular disease, as such diagnoses could alter diet (26). We also conducted a range of sensitivity analyses to evaluate the robustness of our findings (described in the supplementary appendix). Analyses were conducted in each cohort individually and then pooled using the fixed-effects models.

For the biomarker analysis, we identified and excluded outliers using the Studentized Deviate Many-Outlier Procedure (Generalized Extreme Studentized Deviate [ESD] test) (27). We adjusted the biomarkers for the batch effect within each cohort to account for inter-assay variability (28). For this cross-sectional analysis, we used the average dietary intake of the three FFQs available prior to blood collection, and lifestyle variables from the questionnaire closest to the blood collection date. We used multivariable generalized linear models to evaluate the associations between the CQI and aCQI and biomarker concentrations. To improve normality, biomarkers were natural log-transformed and results are presented as percentage differences calculated using the equation: [exp (β-coefficient)-1] × 100%, as in a previous analysis (29).

We also conducted additional sensitivity analyses. We conducted an a priori test for effect modification of the association between cereal fiber and incident T2D by age (<60, ≥ 60y), BMI (<25, ≥ 25 kg/m2), physical activity (< median, ≥ median), smoking (ever, never), family history of diabetes (yes, no), and AHEI score (< median, ≥ median) by including a cross-product term in the models, and we conducted a stratified analysis by these variables. We also evaluated the joint effects of cereal fiber intake and GI on T2D incidence. We also performed a stratified analysis and tested for potential effect modification by BMI (<25, ≥ 25 kg/m2) for all CQMs. To test for potential reverse causation, we did a lagged analysis using cumulative average exposure until 4- and 8 years before T2D diagnosis. We also repeated the main analysis with continuous updating of diet even after intermediate endpoint development. Statistical tests were two-sided, and a P-value < 0.05 was considered statistically significant. We conducted all analyses using SAS version 9.4 for UNIX (SAS Institute).

RESULTS

After applying all inclusion and exclusion criteria, the main analysis included 213,704 adults (NHS: 74,325, HPFS: 44,232, and NHSII: 90,403), and the secondary biomarker analysis included 41,687 adults (NHS: 15,815, HPFS: 6,966, and NHSII: 18,906; Supplemental Figure 1). During up to 36y of follow-up, we documented 22,351 cases (NHS: 9,447; HPFS: 4,404; NHSII: 8,500) of incident T2D over 5,628,955 person-years.

Descriptive characteristics

The baseline characteristics of the study population by total carbohydrate intake are presented in Table 1. Participants with higher carbohydrate intake had lower BMI, alcohol intake, and smoking, and higher physical activity and multivitamin use. In the three cohorts, on average, the sources of carbohydrates in the diet at baseline were from refined grains (21%E– 25%E), total added sugar (20%E– 23%E), whole fruits (11%E– 15%E), potatoes (7%E– 8%E), vegetables (6%E– 7%E), whole grains (5%E– 6%E), fruit juice (4%E– 5%E), legumes (3%E), and some were from other sources such as dairy, alcoholic beverages, nuts, etc. (14–15%E) (Figure 1). The Pearson correlation coefficients of all exposure variables are presented in Supplemental Table 2, and those of the CQI, aCQI, and their components are in Supplemental Table 3.

Table 1.

Age-adjusted baseline characters of the NHS study population by quintiles of total carbohydrate intake in the Nurses Health Study (NHS, n=74,325, baseline 1984), Health Professional Follow=Up Study (HPFS, n=44,232, baseline 1986), and Nurses Health Study II (NHSII, n=90,403, baseline 1991).1

NHS HPFS NHSII
Q1 Q3 Q5 Q1 Q3 Q5 Q1 Q3 Q5
Median (g) 146 186 224 182 234 288 184 223 268
N 14,865 14,290 14,595 8,844 8,926 8,789 17,896 18,310 17,759
Age (years)* 50 ± 7 50 ± 7 51 ± 7 54 ± 9 53 ± 10 54 ± 10 36 ± 5 36 ± 5 36 ± 5
BMI (kg/m2) 25.2 ± 4.6 25.1 ± 4.6 24.4 ± 4.4 26.1 ± 3.4 25.5 ± 3.1 24.8 ± 3.2 25.8 ± 5.9 24.6 ± 5 23.5 ± 4.6
Caucasian, % 99 98 96 96 96 93 97 97 94
Postmenopausal, % 46 46 47 - - - 3 3 3
PMH use, past and current, % 49 48 48 - - - 94 95 94
Family history of diabetes, % 29 29 28 18 19 18 35 33 32
High blood pressure at baseline, % 13 11 12 23 19 20 4 3 3
Elevated cholesterol at baseline, % 4 5 6 10 10 14 10 9 10
Physical activity (met-hrs/wk) 13 ± 19 14 ± 20 16 ± 25 17 ± 22 21 ± 25 25 ± 30 18 ± 24 20 ± 25 26 ± 33
Current smoker, % 34 22 20 17 8 5 18 11 10
Alcohol intake (g/day) 15 ± 17 5 ± 8 3 ± 5 22 ± 22 9 ± 12 5 ± 7 5 ± 10 3 ± 5 2 ± 3
Multivitamin supplement user, % 35 37 40 40 41 45 38 44 48
Total energy (kcal/day) 1698 ± 526 1796 ± 531 1687 ± 532 1961 ± 632 2029 ± 622 1951 ± 625 1759 ± 560 1823 ± 540 1748 ± 555
Carbohydrate: total fiber 11 ± 3 12 ± 3 13 ± 6 12 ± 3 12 ± 3 12 ± 5 12 ± 3 13 ± 3 15 ± 6
Carbohydrate: cereal fiber 61 ± 43 54 ± 37 63 ± 57 62 ± 55 51 ± 39 52 ± 59 49 ± 24 47 ± 22 53 ± 39
Starch: total fiber 4 ± 1 4 ± 1 4 ± 1 4 ± 1 4 ± 1 3 ± 1 5 ± 1 5 ± 1 4 ± 2
Starch: cereal fiber 19 ± 8 17 ± 6 16 ± 8 18 ± 11 14 ± 7 12 ± 7 18 ± 6 16 ± 5 15 ± 6
Carbohydrate Quality Index 11 ± 3 12 ± 3 12 ± 4 11 ± 3 12 ± 3 13 ± 3 11 ± 3 12 ± 3 13 ± 3
Total fiber (g/day) 13 ± 3 16 ± 4 19 ± 6 16 ± 4 21 ± 5 27 ± 9 15 ± 4 18 ± 4 21 ± 8
Cereal fiber (g/day) 3 ± 1 4 ± 2 5 ± 3 4 ± 2 6 ± 3 8 ± 6 4 ± 2 6 ± 2 7 ± 5
Whole grains (g/day) 6 ± 6 9 ± 8 14 ± 14 9 ± 9 14 ± 12 24 ± 22 10 ± 7 13 ± 9 19 ± 16
Refined grains (g/day) 35 ± 13 43 ± 14 44 ± 18 43 ± 16 54 ± 18 59 ± 25 49 ± 15 57 ± 17 60 ± 24
Red meat intake (servings/day) 1.5 ± 0.9 1.2 ± 0.6 0.7 ± 0.5 1.6 ± 1 1.2 ± 0.7 0.6 ± 0.5 1 ± 0.6 0.7 ± 0.4 0.4 ± 0.3
Coffee (cups/day) 2.7 ± 1.9 2.5 ± 1.8 2 ± 1.8 2.4 ± 1.9 1.9 ± 1.8 1.4 ± 1.6 1.8 ± 1.8 1.6 ± 1.7 1.3 ± 1.5
Polyunsaturated fat (% energy) 7.3 ± 2.1 6.7 ± 1.5 5.7 ± 1.4 6.5 ± 1.9 6 ± 1.4 5.2 ± 1.3 6.5 ± 1.6 5.7 ± 1.1 4.7 ± 1.1
Saturated fat (% energy) 14.4 ± 2.8 12.7 ± 1.9 10.1 ± 1.9 12.9 ± 2.9 11.2 ± 2 8.4 ± 2.2 13.5 ± 2.3 11.3 ± 1.6 8.7 ± 1.8
Trans fat (% energy) 2.1 ± 0.6 2.0 ± 0.6 1.6 ± 0.6 1.4 ± 0.5 1.3 ± 0.5 1 ± 0.5 1.9 ± 0.7 1.7 ± 0.6 1.3 ± 0.5
Open in a new tab

Values are means (SD) or medians (Q25, Q75) for continuous variables; percentages or ns or both for categorical variables,and are standardized to the age distribution of the study population.

*

Value is not age adjusted

Figure 1.

Figure 1.

Open in a new tab

Average carbohydrate intake (g/day, %) by source at baseline in the Nurses Health Study (NHS, n=74,325, baseline 1984), Health Professional Follow=Up Study (HPFS, n=44,232, baseline 1986), and Nurses Health Study II (NHSII, n=90,403, baseline 1991).

Main analyses

The pooled relative risk (RR) and 95% confidence intervals (CI) of the associations between single-variable and composite indices of carbohydrate quality and risk of T2D are shown in Table 2. For individual-variable indices, comparing Q5 to Q1, total carbohydrate, starch, and TGE intake were not associated with the risk of T2D in the BMI-adjusted multivariable model (RR: 1.04, 95% CI: 0.97, 1.11, P-trend=0.52, RR: 1.04, 95% CI: 0.99, 1.09, P-trend=0.22, and RR: 1.07, 95% CI: 1.01, 1.13, P-trend=0.06 respectively) Q5 to Q1. As for fiber, both total and cereal fiber were associated with a 16% and 23% lower risk, respectively (RR: 0.84, 95% CI: 0.79, 0.88, P-trend<0.001 and RR: 0.77, 95% CI: 0.74, 0.81, P-trend<0.001). TFE intake was associated 7% higher risk of T2D (RR: 1.07, 95% CI: 1.02, 1.12, P-trend=0.008).

Table 2.

Pooled relative risks and 95% confidence intervals of type 2 diabetes by individual-variable and composite indices of carbohydrate quality among 74,325 women in the NHS, 44,232 men in the HPFS, and 90,403 women in the NHSII.1

Quintiles of energy-adjusted intake
Variable 1 2 3 4 5 P-trend2
Individual-variable indices of carbohydrate quality
 Total carbohydrates (g/d)
  Median
NHS/HPFS/NHS2 161 / 192 / 189 184 / 222 / 212 198 / 242 / 227 211 / 261 / 242 231 / 290 / 266
  Cases/Person-years 4,975 / 1,075,478 5,083 / 1,157,147 4,699 / 1,179,741 4,324 / 1,155,373 3,270 / 1,061,218
  Model 1 1.00 (Ref) 0.95 (0.91, 0.99) 0.86 (0.83, 0.89) 0.81 (0.77, 0.84) 0.66 (0.63, 0.69) <0.001
  Model 2 1.00 (Ref) 0.91 (0.87, 0.95) 0.82 (0.79, 0.87) 0.78 (0.74, 0.83) 0.70 (0.65, 0.75) <0.001
  Model 3 1.00 (Ref) 0.99 (0.95, 1.03) 0.98 (0.93, 1.03) 1.03 (0.97, 1.09) 1.04 (0.97, 1.11) 0.52
 Starch (g/d)
  Median
NHS/HPFS/NHS2 46 / 52 / 60 57 / 67 / 73 64 / 78 / 81 71 / 87 / 90 81 / 103 / 104
  Cases/Person-years 4,131 / 1,014,708 4,889 / 1,148,891 4,678 / 1,190,816 4,721 / 1,185,548 3,932 / 1,088,994
  Model 1 1.00 (Ref) 1.04 (1.00, 1.09) 0.96 (0.93, 1.01) 0.98 (0.94, 1.03) 0.91 (0.87, 0.95) <0.001
  Model 2 1.00 (Ref) 1.06 (1.01, 1.10) 0.99 (0.94, 1.03) 1.02 (0.97, 1.06) 1.00 (0.95, 1.04) 0.36
  Model 3 1.00 (Ref) 1.07 (1.02, 1.11) 1.00 (0.96, 1.05) 1.04 (1.00, 1.09) 1.04 (0.99, 1.09) 0.22
 Total fiber (g/d)
  Median
NHS/HPFS/NHS2 12 / 15 / 13 15 / 18 / 16 17 / 21 / 18 20 / 25 / 21 24 / 30 / 26
  Cases/Person-years 4,995 / 1,040,512 5,016 / 1,142,948 4,650 / 1,170,295 4,323 / 1,163,787 3,367 / 1,111,415
  Model 1 1.00 (Ref) 0.90 (0.87, 0.94) 0.81 (0.78, 0.85) 0.75 (0.72, 0.78) 0.60 (0.58, 0.63) <0.001
  Model 2 1.00 (Ref) 0.98 (0.94, 1.02) 0.95 (0.91, 0.99) 0.95 (0.90, 0.99) 0.89 (0.84, 0.94) <0.001
  Model 3 1.00 (Ref) 0.96 (0.92, 1.00) 0.91 (0.87, 0.95) 0.89 (0.85, 0.93) 0.84 (0.79, 0.88) <0.001
 Cereal fiber (g/d)
  Median
NHS/HPFS/NHS2 2.5 / 3.0 / 3.4 3.6 / 4.6 / 4.6 4.5 / 5.9 / 5.6 5.6 / 7.5 / 6.8 7.6 / 10.3 / 8.9
  Cases/Person-years 5,024 / 1,021,319 5,174 / 1,141,970 4,816 / 1,183,475 4,317 / 1,173,329 3,020 / 1,108,864
  Model 1 1.00 (Ref) 0.89 (0.86, 0.93) 0.80 (0.76, 0.83) 0.71 (0.68, 0.74) 0.52 (0.49, 0.54) <0.001
  Model 2 1.00 (Ref) 0.95 (0.91, 0.99) 0.89 (0.86, 0.93) 0.85 (0.81, 0.89) 0.69 (0.65, 0.72) <0.001
  Model 3 1.00 (Ref) 0.94 (0.90, 0.98) 0.90 (0.86, 0.94) 0.87 (0.83, 0.91) 0.77 (0.74, 0.81) <0.001
 Total glucose equivalents (g/d)
  Median
NHS/HPFS/NHS2 89 / 101 / 112 104 / 121 / 126 112 / 134 / 135 120 / 145 / 144 131 / 162 / 159
  Cases/Person-years 4,414 / 1,051,413 5,026 / 1,150,019 4,971 / 1,176,618 4,446 / 1,172,918 3,494 / 1,077,989
  Model 1 1.00 (Ref) 1.04 (1.00, 1.09) 1.01 (0.97, 1.06) 0.91 (0.88, 0.95) 0.79 (0.76, 0.83) <0.001
  Model 2 1.00 (Ref) 0.99 (0.95, 1.03) 0.94 (0.90, 0.99) 0.86 (0.82, 0.91) 0.80 (0.76, 0.85) <0.001
  Model 3 1.00 (Ref) 1.04 (1.00, 1.08) 1.05 (1.00, 1.10) 1.04 (0.99, 1.09) 1.07 (1.01, 1.13) 0.06
 Total fructose equivalents (g/d)
  Median
NHS/HPFS/NHS2 26 / 30 / 28 33 / 39 / 35 39 / 46 / 41 44 / 54 / 48 53 / 67 / 62
  Cases/Person-years 4,737 / 1,062,177 4,749 / 1,155,394 4,484 / 1,174,878 4,219 / 1,167,659 4,162 / 1,068,851
  Model 1 1.00 (Ref) 0.91 (0.87, 0.95) 0.84 (0.80, 0.87) 0.79 (0.76, 0.82) 0.86 (0.83, 0.90) <0.001
  Model 2 1.00 (Ref) 0.92 (0.89, 0.96) 0.86 (0.83, 0.90) 0.81 (0.78, 0.85) 0.85 (0.81, 0.89) <0.001
  Model 3 1.00 (Ref) 0.97 (0.93, 1.01) 0.97 (0.93, 1.01) 0.98 (0.94, 1.03) 1.07 (1.02, 1.12) 0.008
 Total added sugar (g/d)
  Median
NHS/HPFS/NHS2 21 / 22 / 26 30 / 33 / 36 37 / 43 / 45 45 / 54 / 57 60 / 76 / 83
  Cases/Person-years 4,095 / 1,027,402 4,564 / 1,156,866 4,591 / 1,177,629 4,443 / 1,171,244 4,658 / 1,095,816
  Model 1 1.00 (Ref) 0.99 (0.95, 1.03) 0.98 (0.93, 1.02) 0.95 (0.91, 0.99) 1.09 (1.04, 1.14) <0.001
  Model 2 1.00 (Ref) 0.92 (0.88, 0.96) 0.87 (0.83, 0.91) 0.81 (0.77, 0.85) 0.84 (0.80, 0.88) <0.001
  Model 3 1.00 (Ref) 0.96 (0.92, 1.01) 0.97 (0.92, 1.01) 0.96 (0.92, 1.00) 1.05 (1.01, 1.10) 0.003
 Sugar-sweetened beverage sugar (g/d)
  Median
NHS/HPFS/NHS2 0.1 / 0.1 / 0.1 0.6 / 0.7 / 0.6 2.7 / 3.0 / 2.7 6.9 / 8.0 / 6.9 20.2 / 21.0 / 20.2
  Cases/Person-years 3,437 / 924,931 3,859 / 1,036,406 4,468 / 1,219,994 4,286 / 1,056,593 4,825 / 1,003,315
  Model 1 1.00 (Ref) 1.00 (0.96, 1.05) 1.05 (1.00, 1.10) 1.12 (1.07, 1.17) 1.35 (1.29, 1.41) <0.001
  Model 2 1.00 (Ref) 0.99 (0.95, 1.04) 0.96 (0.92, 1.01) 0.99 (0.95, 1.03) 1.07 (1.02, 1.12) <0.001
  Model 3 1.00 (Ref) 1.03 (0.98, 1.07) 1.06 (1.01, 1.10) 1.12 (1.07, 1.17) 1.22 (1.17, 1.28) <0.001
 Fruit juice sugar (g/d)
  Median
NHS/HPFS/NHS2 1.3 / 1.6 / 1.3 4.2 / 5.3 / 4.2 7.8 / 10.0 / 7.8 12.3 / 15.7 / 12.3 20.3 / 25.2 / 20.3
  Cases/Person-years 4,356 / 1,058,021 4,919 / 1,158,609 4,852 / 1,168,992 4,367 / 1,155,081 3,857 / 1,088,254
  Model 1 1.00 (Ref) 1.03 (0.99, 1.08) 1.00 (0.96, 1.04) 0.91 (0.87, 0.95) 0.85 (0.81, 0.88) <0.001
  Model 2 1.00 (Ref) 1.07 (1.02, 1.11) 1.06 (1.02, 1.11) 1.02 (0.97, 1.06) 1.03 (0.98, 1.08) 0.96
  Model 3 1.00 (Ref) 1.05 (1.00, 1.09) 1.07 (1.03, 1.12) 1.09 (1.04, 1.13) 1.14 (1.09, 1.19) <0.001
 Whole fruit carbohydrates (g/d)
  Median
NHS/HPFS/NHS2 13 / 13 / 10 21 / 24 / 17 28 / 33 / 24 36 / 44 / 31 49 / 63 / 44
  Cases/Person-years 5,239 / 1,071,888 5,024 / 1,154,213 4,566 / 1,166,190 4,237 / 1,154,967 3,285 / 1,081,698
  Model 1 1.00 (Ref) 0.87 (0.83, 0.90) 0.76 (0.73, 0.80) 0.70 (0.67, 0.73) 0.57 (0.55, 0.60) <0.001
  Model 2 1.00 (Ref) 0.97 (0.93, 1.01) 0.91 (0.87, 0.95) 0.90 (0.86, 0.94) 0.82 (0.78, 0.86) <0.001
  Model 3 1.00 (Ref) 0.94 (0.90, 0.98) 0.88 (0.84, 0.92) 0.87 (0.84, 0.91) 0.80 (0.76, 0.85) <0.001
 Vegetable carbohydrates (g/d)
  Median
NHS/HPFS/NHS2 8 / 9 / 8 11 / 12 / 11 14 / 15 / 14 16 / 19 / 18 21 / 25 / 23
  Cases/Person-years 4,130 / 1,079,629 4,545 / 1,164,692 4,696 / 1,170,876 4,561 / 1,154,490 4,419 / 1,059,268
  Model 1 1.00 (Ref) 1.00 (0.96, 1.04) 1.02 (0.98, 1.06) 1.00 (0.95, 1.04) 1.04 (1.00, 1.09) 0.05
  Model 2 1.00 (Ref) 1.09 (1.05, 1.14) 1.17 (1.12, 1.22) 1.21 (1.16, 1.27) 1.39 (1.32, 1.45) <0.001
  Model 3 1.00 (Ref) 1.04 (1.00, 1.08) 1.07 (1.02, 1.11) 1.05 (1.00, 1.09) 1.12 (1.06, 1.17) <0.001
 Starchy vegetables (servings/d)
  Median
NHS/HPFS/NHS2 0.10 / 0.10 / 0.07 0.18 / 0.19 / 0.17 0.25 / 0.28 / 0.25 0.34 / 0.39 / 0.36 0.55 / 0.65 / 0.60
  Cases/Person-years 3,689 / 930,734 4,418 / 1,188,969 4,596 / 1,172,354 4,693 / 1,207,559 4,955 / 1,129,109
  Model 1 1.00 (Ref) 0.95 (0.91, 0.99) 0.98 (0.94, 1.02) 1.00 (0.96, 1.04) 1.11 (1.07, 1.16) <0.001
  Model 2 1.00 (Ref) 0.99 (0.95, 1.04) 1.00 (0.96, 1.04) 0.98 (0.94, 1.03) 1.04 (0.99, 1.09) 0.96
  Model 3 1.00 (Ref) 1.00 (0.96, 1.05) 1.02 (0.98, 1.07) 1.02 (0.98, 1.07) 1.08 (1.03, 1.13) <0.001
 Non-starchy vegetables (servings/d)
  Median
NHS/HPFS/NHS2 1.2 / 1.1 / 0.90 1.8 / 1.8 / 1.5 2.4 / 2.4 / 2.1 3.1 / 3.2 / 2.8 4.3 / 4.6 / 4.2
  Cases/Person-years 4,155 / 977,600 4,539 / 1,144,588 4,546 / 1,170,590 4,618 / 1,174,784 4,493 / 1,161,201
  Model 1 1.00 (Ref) 0.89 (0.85, 0.92) 0.85 (0.82, 0.89) 0.86 (0.82, 0.89) 0.83 (0.80, 0.87) <0.001
  Model 2 1.00 (Ref) 0.97 (0.93, 1.01) 1.00 (0.96, 1.05) 1.07 (1.02, 1.12) 1.14 (1.09, 1.19) <0.001
  Model 3 1.00 (Ref) 0.96 (0.92, 1.00) 0.96 (0.92, 1.01) 0.97 (0.93, 1.02) 0.96 (0.92, 1.01) 0.32
 Potato carbohydrates (g/d)
  Median
NHS/HPFS/NHS2 5 / 7 / 6 9 / 11 / 10 13 / 16 / 14 16 / 21 / 18 23 / 29 / 26
  Cases/Person-years 3,363 / 1,064,134 4,204 / 1,169,677 4,812 / 1,175,628 5,025 / 1,150,731 4,947 / 1,068,786
  Model 1 1.00 (Ref) 1.13 (1.08, 1.19) 1.29 (1.24, 1.35) 1.38 (1.32, 1.44) 1.48 (1.41, 1.54) <0.001
  Model 2 1.00 (Ref) 1.06 (1.01, 1.11) 1.12 (1.07, 1.17) 1.14 (1.09, 1.19) 1.14 (1.09, 1.20) <0.001
  Model 3 1.00 (Ref) 1.04 (0.99, 1.09) 1.09 (1.05, 1.15) 1.11 (1.06, 1.17) 1.15 (1.09, 1.20) <0.001
 Legume carbohydrates (g/d)
  Median
NHS/HPFS/NHS2 2.0 / 2.8 / 1.7 3.4 / 4.8 / 3.5 4.5 / 6.5 / 5.0 / 5.8 / 8.6 / 6.8 8.4 / 13.4 / 10.9
  Cases/Person-years 4,348 / 1,049,677 4,796 / 1,154,475 4,710 / 1,169,020 4,526 / 1,162,569 3,971 / 1,093,216
  Model 1 1.00 (Ref) 0.98 (0.94, 1.02) 0.94 (0.91, 0.98) 0.91 (0.87, 0.94) 0.84 (0.81, 0.88) <0.001
  Model 2 1.00 (Ref) 1.00 (0.96, 1.04) 0.99 (0.95, 1.03) 0.98 (0.93, 1.02) 0.99 (0.95, 1.03) 0.36
  Model 3 1.00 (Ref) 1.02 (0.98, 1.06) 1.02 (0.98, 1.07) 1.00 (0.96, 1.04) 1.02 (0.98, 1.07) 0.53
 Whole grains (g/d)
  Median
NHS/HPFS/NHS2 3 / 4 / 5 6 / 10 / 10 10 / 16 / 14 15 / 23 / 20 24 / 36 / 30
  Cases/Person-years 4,686 / 960,503 5,127 / 1,140,107 4,938 / 1,195,078 4,242 / 1,206,962 3,358 / 1,126,305
  Model 1 1.00 (Ref) 0.91 (0.88, 0.95) 0.83 (0.80, 0.87) 0.70 (0.67, 0.73) 0.59 (0.56, 0.62) <0.001
  Model 2 1.00 (Ref) 0.99 (0.95, 1.03) 0.97 (0.93, 1.01) 0.88 (0.84, 0.92) 0.83 (0.79, 0.87) <0.001
  Model 3 1.00 (Ref) 0.95 (0.92, 0.99) 0.94 (0.90, 0.98) 0.87 (0.83, 0.91) 0.86 (0.82, 0.91) <0.001
 Refined grains (g/d)
  Median
NHS/HPFS/NHS2 28 / 33 / 38 37 / 44 / 48 43 / 52 / 55 49 / 61 / 64 60 / 77 / 79
  Cases/Person-years 4,049 / 1,043,011 4,745 / 1,159,541 4,827 / 1,181,766 4,614 / 1,171,313 4,116 / 1,073,326
  Model 1 1.00 (Ref) 1.06 (1.02, 1.11) 1.07 (1.03, 1.12) 1.04 (1.00, 1.09) 1.03 (0.99, 1.08) 0.95
  Model 2 1.00 (Ref) 1.07 (1.02, 1.11) 1.08 (1.03, 1.13) 1.04 (1.00, 1.09) 1.04 (0.99, 1.09) 0.59
  Model 3 1.00 (Ref) 1.06 (1.01, 1.10) 1.06 (1.02, 1.11) 1.04 (0.99, 1.08) 1.05 (1.00, 1.10) 0.24
 Glycemic Index
  Median
NHS/HPFS/NHS2 49 / 49 / 50 51 / 52 / 52 53 / 53 / 53 54 / 55 / 55 56 / 57 / 57
  Cases/Person-years 3,811 / 1,094,187 4,428 / 1,165,607 4,559 / 1,168,323 4,755 / 1,148,769 4,798 / 1,052,070
  Model 1 1.00 (Ref) 1.10 (1.05, 1.14) 1.14 (1.09, 1.19) 1.22 (1.17, 1.28) 1.38 (1.32, 1.44) <0.001
  Model 2 1.00 (Ref) 1.01 (0.96, 1.05) 0.96 (0.92, 1.01) 0.94 (0.90, 0.99) 0.96 (0.92, 1.01) 0.011
  Model 3 1.00 (Ref) 1.07 (1.03, 1.12) 1.08 (1.04, 1.13) 1.11 (1.06, 1.16) 1.20 (1.14, 1.26) <0.001
 Glycemic load
  Median
NHS/HPFS/NHS2 82 / 99 / 98 95 / 116 / 112 104 / 128 / 121 112 / 140 / 130 125 / 157 / 146
  Cases/Person-years 4,698 / 1,078,943 4,995 / 1,153,152 4,676 / 1,181,251 4,393 / 1,153,556 3,589 / 1,062,053
  Model 1 1.00 (Ref) 1.00 (0.96, 1.04) 0.91 (0.88, 0.95) 0.88 (0.84, 0.92) 0.79 (0.76, 0.82) <0.001
  Model 2 1.00 (Ref) 0.94 (0.90, 0.98) 0.85 (0.81, 0.89) 0.81 (0.77, 0.86) 0.75 (0.71, 0.80) <0.001
  Model 3 1.00 (Ref) 1.02 (0.98, 1.06) 1.01 (0.97, 1.06) 1.07 (1.02, 1.13) 1.12 (1.06, 1.19) <0.001
Composite indices of carbohydrate quality
 Carbohydrate Quality Index
  Median
NHS/HPFS/NHS2 8 / 8 / 8 11 / 10 / 10 12 / 12 / 12 14 / 14 / 14 17 / 16 / 16
  Cases/Person-years 5,153 / 1,084,985 5,059 / 1,178,203 4,528 / 1,114,785 4,176 / 1,154,186 3,458 / 1,096,796
  Model 1 1.00 (Ref) 0.88 (0.85, 0.92) 0.80 (0.77, 0.84) 0.73 (0.70, 0.76) 0.63 (0.61, 0.66) <0.001
  Model 2 1.00 (Ref) 0.97 (0.94, 1.01) 0.96 (0.92, 1.00) 0.95 (0.91, 1.00) 0.94 (0.90, 0.99) 0.007
  Model 3 1.00 (Ref) 0.93 (0.89, 0.96) 0.89 (0.85, 0.92) 0.86 (0.82, 0.90) 0.82 (0.79, 0.87) <0.001
 alternate Carbohydrate Quality Index
  Median
NHS/HPFS/NHS2 9 / 9 / 10 12 / 13 / 13 15 / 15 / 15 18 / 17 / 17 21 / 20 / 20
  Cases/Person-years 5,380 / 1,046,864 5,061 / 1,084,720 4,696 / 1,209,384 4,069 / 1,135,594 3,145 / 1,152,396
  Model 1 1.00 (Ref) 0.85 (0.82, 0.88) 0.73 (0.71, 0.76) 0.64 (0.61, 0.66) 0.49 (0.47, 0.51) <0.001
  Model 2 1.00 (Ref) 0.96 (0.92, 1.00) 0.92 (0.88, 0.95) 0.88 (0.84, 0.91) 0.78 (0.74, 0.82) <0.001
  Model 3 1.00 (Ref) 0.90 (0.86, 0.93) 0.84 (0.81, 0.88) 0.80 (0.76, 0.83) 0.71 (0.68, 0.75) <0.001
 Carbohydrate : total fiber
  Median
NHS/HPFS/NHS2 8.8 / 8.6 / 9.3 10.2 / 10.1 / 10.9 11.3 / 11.3 / 12.2 12.7 / 12.7 / 13.8 15.3 / 15.5 / 17.0
  Cases/Person-years 3,882 / 1,079,974 4,397 / 1,148,976 4,534 / 1,160,940 4,673 / 1,154,903 4,865 / 1,084,135
  Model 1 1.00 (Ref) 1.08 (1.04, 1.13) 1.12 (1.07, 1.17) 1.18 (1.13, 1.23) 1.34 (1.29, 1.40) <0.001
  Model 2 1.00 (Ref) 0.98 (0.94, 1.02) 0.95 (0.91, 0.99) 0.94 (0.90, 0.99) 0.97 (0.93, 1.02) 0.19
  Model 3 1.00 (Ref) 1.02 (0.97, 1.06) 1.02 (0.98, 1.07) 1.07 (1.02, 1.12) 1.15 (1.10, 1.21) <0.001
 Carbohydrate : cereal fiber
  Median
NHS/HPFS/NHS2 30 / 26 / 27 39 / 35 / 35 47 / 43 / 41 57 / 53 / 50 79 / 76 / 68
  Cases/Person-years 3,142 / 1,092,053 4,271 / 1,176,570 4,796 / 1,178,471 5,092 / 1,142,855 5,048 / 1,038,783
  Model 1 1.00 (Ref) 1.26 (1.20, 1.32) 1.42 (1.36, 1.48) 1.56 (1.49, 1.63) 1.72 (1.64, 1.80) <0.001
  Model 2 1.00 (Ref) 1.17 (1.12, 1.22) 1.25 (1.19, 1.31) 1.31 (1.25, 1.37) 1.36 (1.30, 1.43) <0.001
  Model 3 1.00 (Ref) 1.09 (1.04, 1.14) 1.15 (1.09, 1.20) 1.21 (1.15, 1.26) 1.27 (1.21, 1.33) <0.001
 Starch : total fiber
  Median
NHS/HPFS/NHS2 2.5 / 2.3 / 3.1 3.2 / 3.1 / 3.9 3.7 / 3.7 / 4.4 4.3 / 4.3 / 5.1 5.2 / 5.3 / 6.1
  Cases/Person-years 3,489 / 1,048,224 4,341 / 1,146,383 4,708 / 1,169,452 4,841 / 1,165,548 4,972 / 1,099,320
  Model 1 1.00 (Ref) 1.16 (1.11, 1.21) 1.26 (1.20, 1.31) 1.33 (1.27, 1.39) 1.49 (1.43, 1.56) <0.001
  Model 2 1.00 (Ref) 1.06 (1.02, 1.11) 1.07 (1.03, 1.13) 1.06 (1.01, 1.11) 1.09 (1.03, 1.14) 0.011
  Model 3 1.00 (Ref) 1.09 (1.04, 1.14) 1.13 (1.08, 1.18) 1.13 (1.08, 1.19) 1.18 (1.13, 1.24) <0.001
 Starch : cereal fiber
  Median
NHS/HPFS/NHS2 9.8 / 8.2 / 10.6 12.8 / 11.3 / 13.1 15.2 / 13. 8 / 15.0 17.9 / 16.8 / 17.1 22.4 / 21.8 / 21.1
  Cases/Person-years 2,896 / 1,092,276 3,978 / 1,172,591 4,786 / 1,176,020 5,252 / 1,148,437 5,437 / 1,039,406
  Model 1 1.00 (Ref) 1.30 (1.24, 1.36) 1.58 (1.51, 1.65) 1.79 (1.71, 1.88) 2.07 (1.98, 2.17) <0.001
  Model 2 1.00 (Ref) 1.20 (1.14, 1.26) 1.36 (1.30, 1.42) 1.44 (1.37, 1.51) 1.52 (1.45, 1.60) <0.001
  Model 3 1.00 (Ref) 1.10 (1.05, 1.15) 1.21 (1.15, 1.26) 1.26 (1.20, 1.32) 1.36 (1.29, 1.42) <0.001
Open in a new tab
1:

RRs and 95% CIs were calculated with the use of the Cox proportional hazards regression model. Model 1: Age-adjusted. Model 2: Adjusted for age, family history of diabetes, smoking status, alcohol intake, physical activity level, multivitamin use, race, menopausal status and postmenopausal hormone use (NHS and NHSII), oral contraceptive use (NHSII), total energy, and intakes of red meat, polyunsaturated fat, saturated fat, trans fat, and coffee. Cereal fiber models were adjusted for fruit and vegetable fiber. Whole and refined grain models were mutually adjusted for each other. Model 3: Additionally adjusted for BMI.

2:

Test for trend based on variable containing median value for each quintile.

When comparing Q5 to Q1 carbohydrate intakes from different food sources, in the BMI-adjusted multivariable models, carbohydrate from whole grains, but not from refined grains, was inversely associated with T2D risk intake (RR: 0.86, 95% CI: 0.82, 0.91, P-trend<0.001 and RR: 1.05, 95% CI: 1.00, 1.10, P-trend=0.24, respectively). Whole-fruit carbohydrate was associated with a 20% lower risk of T2D (RR: 0.80, 95% CI: 0.76, 0.85), but fruit juice sugar was associated with a 14% higher risk (RR: 1.14, 95% CI: 1.09, 1.19; both P-trend<0.001). Similarly, SSB sugar and total added sugar were associated with a 22% and 5% higher risk of T2D, respectively (RR: 1.22, 95% CI: 1.17, 1.28, and RR: 1.05, 95% CI: 1.01, 1.10, both P-trend<0.001). In addition, potato carbohydrate was associated with a 15% higher risk of T2D (RR: 1.15, 95% CI: 1.09, 1.20, P-trend<0.001), while legume carbohydrate was not (RR: 1.02, 95% CI: 0.98, 1.07, P-trend=0.54). Vegetable carbohydrates were significantly associated with a 12% higher risk of T2D (RR: 1.12, 95% CI: 1.06, 1.17), and starchy vegetable consumption was associated with an 8% higher risk (RR: 1.08, 95% CI: 1.03, 1.13, both P-trend<0.001). In contrast, non-starchy vegetable consumption was not associated with T2D (RR: 0.96, 95% CI: 0.92, 1.01, P-trend=0.32).

When comparing Q5 to Q1of metrics that measure the physiological effect of carbohydrates, the GI and GL were associated with a 20% and 12% higher risk of T2D (RR: 1.20, 95% CI: 1.14, 1.26 and RR:1.12, 95% CI: 1.06, 1.19, both P-trend<0.001). For composite CQMs (Table 2), the CQI was associated with an 18% lower risk of T2D, whereas the aCQI was associated with a 29% lower risk (RR: 0.82, 95% CI: 0.79, 0.87 and RR:0.71, 95% CI: 0.68, 0.75, both P-trend<0.001). Ratios of carbohydrate or starch to total or cereal fiber were all positively associated with T2D risk, with the with the largest magnitude of association observed for the starch to cereal fiber ratio (RR: 1.36, 95% CI: 1.29, 1.42, P-trend <0.001).

Figure 2 summarizes the pooled associations comparing Q5 to Q1 of intake of all twenty-five CQMs and risk of T2D from the BMI-adjusted multivariable model. The largest magnitude inverse associations were for aCQI, cereal fiber, and whole-fruit carbohydrate, which were linked to a 29%, 23%, and 20% lower T2D risk, respectively. Conversely, the strongest positive associations were between the starch to cereal fiber ratio, SSB sugar, and GI, which were associated with a 36%, 22%, and 20% higher risk of T2D, respectively.

Figure 2.

Figure 2.

Open in a new tab

Pooled relative risks and 95% confidence intervals of type 2 diabetes by individual measures of carbohydrate quality, comparing Q5 to Q1, among 74,325 women in the NHS, 44,232 men in the HPFS, and 90,403 women in the NHSII.1

• Significant inverse association • Significant positive association • No significant association

1: RRs and 95% CIs were calculated with the use of the Cox proportional hazards regression model, comparing extreme quintiles of intake. Models were adjusted for age, family history of diabetes, smoking status, alcohol intake, physical activity level, multivitamin use, race, menopausal status and postmenopausal hormone use (NHS and NHSII), oral contraceptive use (NHSII), total energy, and intakes of red meat, polyunsaturated fat, saturated fat, trans fat, and coffee, and BMI. Cereal fiber models were adjusted for fruit and vegetable fiber. Whole and refined grain models were mutually adjusted for each other.

Figure 3 displays the percentage difference in cardiometabolic biomarker concentrations, comparing Q5 to Q1 of the CQI and aCQI, in the multivariable and BMI-adjusted models. The aCQI was associated with a greater number of biomarkers than the CQI (Supplemental table 4). The CQI and aCQI were associated with 5% and 10% lower concentrations of the glycemic biomarker C-peptide, respectively (both P-trend <0.001). For inflammatory biomarkers, the largest percent differences were with CRP, where the CQI and aCQI were associated with 13.9% and 16.1% lower concentrations, respectively (both P-trend <0.001). Additionally, the aCQI was linked to 6.10% lower leptin and 4.51% lower LDL-C (both P-trend <0.001); the CQI showed no association.

Figure 3.

Figure 3.

Open in a new tab

Pooled percentage of differences (95% CI) in biomarker concentrations comparing Q5 to Q1 of the CQI and aCQI among a subset of the study population in a cross-sectional analysis.1,2

1: Percent differences and 95% CIs were calculated with the use of the multivariable generalized linear model, comparing extreme quintiles of intake. Associations were adjusted for age at blood draw (continuous), sex (M/F), case-control status (case/control), fasting status (fasting/nonfasting), family history of diabetes (yes/no), smoking status (never, past, current 1–14 cigs/day, current 15 – 24 cigs/day, or current ≥ 25 cigs/day), alcohol intake (0, 0.1 – 4.9, 5.0 – 14.9. or 15 g/d), physical activity level (<3, 3 - <9, 9 - <18, 18 - <27, 27 - <42, or ≥ 42 MET-hrs/wk), multivitamin use (yes/no), race (white/non-white), menopausal status (pre/post/unknown status) and postmenopausal hormone use (never/past/current) (NHS and NHSII), and intakes of total energy red meat, polyunsaturated fat, saturated fat, trans fat, coffee (all in quintiles), and BMI (<23, 23-<25, 25-<27.5, 27.5-<30, or ≥ 30 kg/m2).

2: The four components of the CQI are total fiber, glycemic index, solid to total carbohydrates, and whole to total grains. The five components of the aCQI are cereal fiber, glycemic index, SSBs, whole grains, and whole fruit consumption.

*: P-value for trend < 0.05. Test for trend based on variable containing median value for each quintile.

Additional analyses

The pooled joint effects analysis of cereal fiber intake and glycemic index with risk of T2D showed that participants with either low or medium glycemic index with highest cereal fiber intake had a 34 – 37% lower risk of T2D compared to those with the highest glycemic index and lowest cereal fiber intake (p-interaction=0.017) (Supplemental figure 2). There was a significant effect modification of the association between cereal fiber and risk of T2D by age (p-value < 0.001) and AHEI (p-value = 0.035; Supplemental figure 3). Among participants younger than 60y, cereal fiber was associated with a 26% lower risk of T2D (RR: 0.74, 95% CI: 0.68, 0.79, P-trend <0.001), compared to a 19% lower risk of T2D among those aged 60 or above (RR: 0.81, 95% CI: 0.76, 0.87, P-trend <0.001). In addition, among participants with AHEI levels below the median, cereal fiber intake was associated with a 27% lower risk of T2D (RR: 0.73, 95% CI: 0.68, 0.78, P-trend <0.001), while it was associated with a 17% lower risk of T2D among those who had an aHEI score equal to or greater than the median (RR: 0.83, 95% CI: 0.77, 0.90, P-trend <0.001).

There was no effect modification by BMI (<25, ≥ 25 kg/m2) for most CQMs (Supplemental figure 4). Results were similar when we examined associations with a 4- and 8year lag (Supplemental figure 5). Furthermore, we repeated the primary analysis without updating the diet after intermediate endpoint development, and many associations were slightly stronger than the primary analysis findings (Supplemental table 5).

DISCUSSION

In these three large prospective cohorts, higher intakes of cereal fiber, whole-fruit carbohydrate, and whole-grain carbohydrate, along with lower intakes of SSB sugar and GI were associated with lower T2D risk. Higher total fiber intake and lower consumption of fruit juice sugar, added sugar, potato carbohydrate, starchy vegetables, TFE, and GL were also associated with lower T2D risk, albeit with smaller effect sizes. Additionally, the aCQI, a new composite CQM, was inversely associated with T2D risk.

Our results reaffirm previous studies showing that higher intakes of cereal fiber (6, 30), whole grains (31, 32), whole fruit (33, 34), along with lower SSB consumption (35), and diets with low GI (36), are associated with lower T2D risk. In line with the meta-analysis by Jenkins et al., we observed comparable robust inverse associations for cereal fiber and low GI (36). We also identified a significant GI–cereal fiber interaction: participants in the highest cereal fiber quintile and low-to-medium GI quintiles had a 34%–37% lower T2D risk than those in the highest GI and lowest cereal fiber quintiles, highlighting the complementary value of both metrics in assessing carbohydrate quality.

Our results further underscore the importance of distinguishing cereal from total fiber. Whereas the CQI uses total fiber, the aCQI’s cereal fiber component demonstrated a larger effect size for T2D risk, suggesting fiber source rather than quantity alone may be more informative (37, 38). Additionally, the effect modification of cereal fiber by overall diet quality (aHEI) suggests a more pronounced association among participants with lower diet quality, whereas those with higher aHEI scores may already achieve similar metabolic benefits from healthier diets.

Total carbohydrate intake was not associated with T2D risk, consistent with prior cohort studies (39). In contrast to our results, earlier analyses in these cohorts and in Japanese women reported positive associations for starch, likely because they adjusted for cereal or total fiber intake (6, 40). The modest effect sizes for sugars may reflect the heterogeneous food sources. For example, whole-grain cereals and dark bread are associated with lower T2D risk while SSBs are associated with higher risk (31, 41). Other added sugar sources show mixed associations: a recent study in our cohorts reported no association for confectionery but inverse associations for packaged sweet snacks, fruit-based products, and yogurt or dairy-based desserts (42).

Research consistently links SSB, and to a lesser extent fruit juice, consumption with a higher T2D risk (41). Our analysis of sugars from these beverages aligns with findings of the beverages themselves: a meta-analysis of prospective cohort studies found 13% (95% CI: 6 – 21%) higher T2D risk per serving of SSBs and 7% (95% CI: 1 – 14%) for fruit juices, independent of adiposity, suggesting fruit juices may not be a healthier alternative to SSBs for T2D risk (41). Future research should consider the role of food form (liquid or solid) and degree of processing (e.g., whole fruit or ultra-processed beverages), to better elucidate how carbohydrate quality in different matrices influences T2D risk.

The observed higher T2D risk with higher consumption of starchy vegetables and potato carbohydrate is likely due to their higher carbohydrate and lower fiber content. Further adjustment of the potato carbohydrate, starchy, and non-starchy vegetable models for cereal fiber did not substantially alter the associations (Supplemental table 6). Our findings on the associations of carbohydrate/starch to total/cereal fiber ratios reinforce the importance of carbohydrate quality by demonstrating that diets higher in carbohydrates or starch relative to total or cereal fiber are associated with substantially greater T2D risk. This pattern highlights that insufficient fiber, especially cereal fiber, may amplify the adverse metabolic effects of higher-glycemic carbohydrate sources, underscoring the value of emphasizing fiber-rich foods when evaluating dietary carbohydrate intake.

Finding a universally accepted golden standard for measuring carbohydrate quality has been challenging due to differences in food sources (natural or added; fruit, vegetable, or grain), degree of processing (refined or whole), form (liquid or solid), composition (starch, fiber, or sugar), physiological response (postprandial glycemic), as well as interactions among these factors. The top five CQMs used to construct the aCQI reflect these dimensions: cereal fiber (composition), GI (physiological response), whole grains (processing and source), SSB (source and form), and whole fruit (source and form).

In this study, the aCQI showed a stronger association with T2D risk and a broader range of cardiometabolic biomarkers than the CQI. This may be due to the smaller effect sizes of the components of the CQI, where not only is cereal fiber associated with a larger effect size than total fiber, the CQI also includes two ratios, and in our study, the whole to total grain ratio was associated with a 17% lower T2D risk, while the solid to total carbohydrates ratio was not (RR: 1.03, 95% CI: 0.99, 1.08, P-trend=0.21; Supplemental table 7). In prospective cohort studies, CQI was not associated with gestational diabetes among Spanish women, and only marginally inversely associated with waist circumference among US adults (43, 44). Cross-sectional study findings have been inconsistent: among Iranian adults, CQI was largely not associated with cardiometabolic risk factors and insulin resistance, whereas among Ghanian women, it was inversely associated with general and abdominal obesity (7, 8).

Due to the diversity and complexity of carbohydrates, numerous mechanisms underlie their association with T2D. First, carbohydrate intake raises postprandial glycemia and insulin response, but longer-chain carbohydrates in legumes and whole grains slow glucose absorption and delay gastric emptying due to organic acids (45). Second, insoluble cereal fiber moderately affects satiety and weight loss, improves insulin resistance, and influences protein absorption, digestion, and amino acid metabolism (30). Soluble fiber from fruits and vegetables delays gastric emptying, reduces postprandial glucose spikes, and improves blood lipids. Fermented by gut microbiota, it modestly affects satiety and weight loss (30). Third, added sugar may increase T2D risk, independently and dependently on body weight, by increasing energy intake and contributing to adverse cardiometabolic profiles, including higher triglycerides, cholesterol, and blood pressure (45). Lastly, less-understood pathways, such as the role of fructose, warrant further investigation.

Our study has notable strengths, including the most comprehensive systematic assessment of CQMs and T2D to date, detailed dietary data enabling development of the aCQI, and biomarker validation across thirteen cardiometabolic biomarkers. The large prospective design, many cases, and long follow-up at high rates further strengthen our findings. Limitations include the observational design and potential residual confounding, although repeated dietary measures help reduce measurement error and better represent long-term diet. Diet was self-reported via FFQs, which can lead to misclassification, but FFQs were validated (12–14). Generalizability may be limited due to the predominantly white US adult study population, though global dietary patterns suggest the findings may be relevant in other contexts.

In conclusion, our study demonstrates that carbohydrate quality, rather than quantity, is a stronger determinant for T2D risk. Diets high in cereal fiber, whole-fruit and whole-grain carbohydrates, and low in SSB sugars and GI had the lowest risk of T2D. The novel aCQI, which incorporates these five components, offers a more comprehensive measure of carbohydrate quality for cardiometabolic health than single-variable indicators and merits further evaluation in other populations. Dietary guidelines should emphasize improving carbohydrate quality while also maintaining appropriate portion control, particularly for lower-quality carbohydrate foods and for individuals with prediabetes or obesity.

Supplementary Material

1

Acknowledgements:

We thank the participants and staff of the Nurses’ Health Study, Nurses’ Health Study II and the Health Professionals Follow-up Study for their valuable contributions.

Funding:

The Nurses’ Health Studies and Health Professional Follow-up Studies are supported by National Institutes of Health grants UM1 CA186107, R01 CA49449, R01 HL034594, U01 HL145386, R01 HL088521, U01 CA176726, R01 CA49449, U01CA167552, R01 HL60712, and R01 HL35464. This work was also supported by the Cardiovascular Health Research Grant to HA from the Kuwait Heart Foundation. This study was also partially supported by the Novo Nordisk Foundation (NNF24SA0092022). The sponsors had no role in the study design; the collection, analysis, or interpretation of data; the writing of the report; or in the decision to submit the article for publication. SB and FH are guarantors of the work, where they had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Declaration of interests: The authors report no other relationships or activities that could appear to have influenced the submitted work.

Ethical approval: The Nurses’ Health Study, Nurses’ Health Study II, and the Health Professionals Follow-up Study were approved by the institutional review boards at Brigham and Women’s Hospital and Harvard TH Chan School of Public Health. The return of the completed self-administered questionnaires was considered to imply informed consent.

Data sharing:

Because of participant confidentiality and privacy concerns, data cannot be shared publicly and requests to access the Nurses’ Health Study, the Nurses’ Health Study II, and the Health Professionals Follow-up Study data must be submitted in writing. According to standard controlled access procedures, applications to use resources from the three studies will be reviewed by our External Collaborations Committee to verify that the proposed use maintains the protection of the privacy of participants and the confidentiality of the data. Investigators wishing to use data from the three studies are asked to submit a brief description of the proposed project (go to https://www.nurseshealthstudy.org/researchers [contact email nhsaccess@channing.harvard.edu] and https://sites.sph.harvard.edu/hpfs/for-collaborators/ for details).

The lead author (HA) affirms that the manuscript is an honest, accurate, and transparent account of the study being reported. No important aspects of the study have been omitted, and any discrepancies from the study as planned have been explained.

References:

  • 1.Federation ID. Diabetes Atlas, 11th edn. Brussels, Belgium: International Diabetes Federation; 2025. [Google Scholar]
  • 2.O’Hearn M, Lara-Castor L, Cudhea F, Miller V, Reedy J, Shi P, et al. Incident type 2 diabetes attributable to suboptimal diet in 184 countries. Nature Medicine. 2023;29(4):982–95. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Ley SH, Hamdy O, Mohan V, Hu FB. Prevention and management of type 2 diabetes: dietary components and nutritional strategies. Lancet. 2014;383(9933):1999–2007. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Neuenschwander M, Ballon A, Weber KS, Norat T, Aune D, Schwingshackl L, et al. Role of diet in type 2 diabetes incidence: umbrella review of meta-analyses of prospective observational studies. Bmj. 2019;366:l2368. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Mousavi SM, Gu X, Imamura F, AlEssa HB, Devinsky O, Sun Q, et al. Total and specific potato intake and risk of type 2 diabetes: results from three US cohort studies and a substitution meta-analysis of prospective cohorts. BMJ. 2025;390:e082121. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.AlEssa HB, Bhupathiraju SN, Malik VS, Wedick NM, Campos H, Rosner B, et al. Carbohydrate quality and quantity and risk of type 2 diabetes in US women. Am J Clin Nutr. 2015;102(6):1543–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Suara SB, Siassi F, Saaka M, Rahimi Foroshani A, Sotoudeh G. Association between Carbohydrate Quality Index and general and abdominal obesity in women: a cross-sectional study from Ghana. BMJ Open. 2019;9(12):e033038. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Nikrad N, Hosseini B, Pakmehr A, Tousi AZ, Ardekani AM, Farhangi MA, et al. Dietary carbohydrate quality index (CQI), cardio-metabolic risk factors and insulin resistance among adults with obesity. BMC Endocrine Disorders. 2023;23(1):171. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.ElİBol E, Eski S, Gez E, Çamdeviren G. The impact of carbohydrate quality index on menopausal symptoms and quality of life in postmenopausal women. BMC Women’s Health. 2025;25(1):262. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Lin F-T, Tu K-X, Ou Q-J, Deng X-Q, Fang Y-J, Zhang C-X. Association of low-carbohydrate diet score and carbohydrate quality index with colorectal cancer risk: a large-scale case-control study. European Journal of Nutrition. 2024;64(1):15. [DOI] [PubMed] [Google Scholar]
  • 11.Bulló M, Papandreou C, Ruiz-Canela M, Guasch-Ferré M, Li J, Hernández-Alonso P, et al. Plasma Metabolomic Profiles of Glycemic Index, Glycemic Load, and Carbohydrate Quality Index in the PREDIMED Study. The Journal of Nutrition. 2021;151(1):50–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Willett WC, Sampson L, Stampfer MJ, Rosner B, Bain C, Witschi J, et al. Reproducibility and validity of a semiquantitative food frequency questionnaire. American journal of epidemiology. 1985;122(1):51–65. [DOI] [PubMed] [Google Scholar]
  • 13.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. American journal of epidemiology. 1992;135(10):1114–26; discussion 27–36. [DOI] [PubMed] [Google Scholar]
  • 14.Yuan C, Spiegelman D, Rimm EB, Rosner BA, Stampfer MJ, Barnett JB, et al. Relative Validity of Nutrient Intakes Assessed by Questionnaire, 24-Hour Recalls, and Diet Records as Compared With Urinary Recovery and Plasma Concentration Biomarkers: Findings s for Women. American journal of epidemiology. 2018;187(5):1051–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Nutrient Data Laboratory (U.S.), Consumer and Food Economics Institute (U.S.). USDA nutrient database for standard reference. Riverdale, Md.: USDA, Nutrient Data Laboratory, Agricultural Research Service,; 1999. [Google Scholar]
  • 16.Association AD. Understanding Carbs: American Diabetes Association; 2025. [updated October 28. Available from: https://diabetes.org/food-nutrition/understanding-carbs.
  • 17.Dennis KK, Wang F, Li Y, Manson JE, Rimm EB, Hu FB, et al. Associations of dietary sugar types with coronary heart disease risk: a prospective cohort study. The American Journal of Clinical Nutrition. 2023;118(5):1000–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Bhupathiraju SN, Tobias DK, Malik VS, Pan A, Hruby A, Manson JE, et al. Glycemicindex, glycemic load, and risk of type 2 diabetes: results from 3 large US cohorts and an updated meta-analysis. Am J Clin Nutr. 2014;100(1):218–32. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Classification and diagnosis of diabetes mellitus and other categories of glucose intolerance. National Diabetes Data Group. Diabetes. 1979;28(12):1039–57. [DOI] [PubMed] [Google Scholar]
  • 20.Nimptsch K, Brand-Miller JC, Franz M, Sampson L, Willett WC, Giovannucci E. Dietary insulin index and insulin load in relation to biomarkers of glycemic control, plasma lipids, and inflammation markers. Am J Clin Nutr. 2011;94(1):182–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Wu T, Giovannucci E, Pischon T, Hankinson SE, Ma J, Rifai N, et al. Fructose, glycemic load, and quantity and quality of carbohydrate in relation to plasma C-peptide concentrations in US women123. The American Journal of Clinical Nutrition. 2004;80(4):1043–9. [DOI] [PubMed] [Google Scholar]
  • 22.Chan AT, Ogino S, Giovannucci EL, Fuchs CS. Inflammatory Markers Are Associated With Risk of Colorectal Cancer and Chemopreventive Response to Anti-Inflammatory Drugs. Gastroenterology. 2011;140(3):799–808.e2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Kang JH, Wiggs JL, Pasquale LR. A Nested Case Control Study of Plasma ICAM-1, E-Selectin and TNF Receptor 2 Levels, and Incident Primary Open-Angle Glaucoma. Investigative Ophthalmology & Visual Science. 2013;54(3):1797–804. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Meyer KA, Conigrave KM, Chu N-F, Rifai N, Spiegelman D, Stampfer MJ, et al. Alcohol Consumption Patterns and HbA1c, C-Peptide and Insulin Concentrations in Men. Journal of the American College of Nutrition. 2003;22(3):185–94. [DOI] [PubMed] [Google Scholar]
  • 25.Willett W, Stampfer M, Chu NF, Spiegelman D, Holmes M, Rimm E. Assessment of questionnaire validity for measuring total fat intake using plasma lipid levels as criteria. American journal of epidemiology. 2001;154(12):1107–12. [DOI] [PubMed] [Google Scholar]
  • 26.Willett WC. Nutritional Epidemiology. New York: Oxford University Press; 1998. [Google Scholar]
  • 27.Rosner B Percentage Points for a Generalized ESD Many-Outlier Procedure. Technometrics. 1983;25(2):165–72. [Google Scholar]
  • 28.Rosner B, Cook N, Portman R, Daniels S, Falkner B. Determination of blood pressure percentiles in normal-weight children: some methodological issues. American journal of epidemiology. 2008;167(6):653–66. [DOI] [PubMed] [Google Scholar]
  • 29.Li X, Hur J, Cao Y, Song M, Smith-Warner SA, Liang L, et al. Moderate alcohol consumption, types of beverages and drinking pattern with cardiometabolic biomarkers in three cohorts of US men and women. Eur J Epidemiol. 2023;38(11):1185–96. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Weickert MO, Pfeiffer AFH. Impact of Dietary Fiber Consumption on Insulin Resistance and the Prevention of Type 2 Diabetes. J Nutr. 2018;148(1):7–12. [DOI] [PubMed] [Google Scholar]
  • 31.Hu Y, Ding M, Sampson L, Willett WC, Manson JE, Wang M, et al. Intake of whole grain foods and risk of type 2 diabetes: results from three prospective cohort studies. Bmj. 2020;370:m2206. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Aune D, Norat T, Romundstad P, Vatten LJ. Whole grain and refined grain consumption and the risk of type 2 diabetes: a systematic review and dose-response meta-analysis of cohort studies. Eur J Epidemiol. 2013;28(11):845–58. [DOI] [PubMed] [Google Scholar]
  • 33.Muraki I, Imamura F, Manson JE, Hu FB, Willett WC, van Dam RM, et al. Fruit consumption and risk of type 2 diabetes: results from three prospective longitudinal cohort studies. Bmj. 2013;347:f5001. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Li M, Fan Y, Zhang X, Hou W, Tang Z. Fruit and vegetable intake and risk of type 2 diabetes mellitus: meta-analysis of prospective cohort studies. BMJ Open. 2014;4(11):e005497. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Imamura F, O’Connor L, Ye Z, Mursu J, Hayashino Y, Bhupathiraju SN, et al. Consumption of sugar sweetened beverages, artificially sweetened beverages, and fruit juice and incidence of type 2 diabetes: systematic review, meta-analysis, and estimation of population attributable fraction. Bmj. 2015;351:h3576. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Jenkins DJA, Willett WC, Yusuf S, Hu FB, Glenn AJ, Liu S, et al. Association of glycaemic index and glycaemic load with type 2 diabetes, cardiovascular disease, cancer, and all-cause mortality: a meta-analysis of mega cohorts of more than 100 000 participants. Lancet Diabetes Endocrinol. 2024;12(2):107–18. [DOI] [PubMed] [Google Scholar]
  • 37.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(7):1394–408. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Kabthymer RH, Karim MN, Hodge AM, de Courten B. High cereal fibre but not total fibre is associated with a lower risk of type 2 diabetes: Evidence from the Melbourne Collaborative Cohort Study. Diabetes Obes Metab. 2023;25(7):1911–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Hosseini F, Jayedi A, Khan TA, Shab-Bidar S. Dietary carbohydrate and the risk of type 2 diabetes: an updated systematic review and dose–response meta-analysis of prospective cohort studies. Scientific Reports. 2022;12(1):2491. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Kanehara R, Goto A, Sawada N, Mizoue T, Noda M, Hida A, et al. Association between sugar and starch intakes and type 2 diabetes risk in middle-aged adults in a prospective cohort study. Eur J Clin Nutr. 2022;76(5):746–55. [DOI] [PubMed] [Google Scholar]
  • 41.Imamura F, O’Connor L, Ye Z, Mursu J, Hayashino Y, Bhupathiraju SN, et al. Consumption of sugar sweetened beverages, artificially sweetened beverages, and fruit juice and incidence of type 2 diabetes: systematic review, meta-analysis, and estimation of population attributable fraction. BMJ : British Medical Journal. 2015;351:h3576. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Chen Z, Khandpur N, Desjardins C, Wang L, Monteiro CA, Rossato SL, et al. UltraProcessed Food Consumption and Risk of Type 2 Diabetes: Three Large Prospective U.S. Cohort Studies. Diabetes care. 2023;46(7):1335–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Fernández-González E, Martínez-González M, Bes-Rastrollo M, Suescun-Elizalde D, Basterra-Gortari FJ, Santiago S, et al. Association between pre-conceptional carbohydrate quality index and the incidence of gestational diabetes: the SUN cohort study. Br J Nutr. 2022;129(4):1–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Sawicki CM, Lichtenstein AH, Rogers GT, Jacques PF, Ma J, Saltzman E, et al. Comparison of Indices of Carbohydrate Quality and Food Sources of Dietary Fiber on Longitudinal Changes in Waist Circumference in the Framingham Offspring Cohort. Nutrients. 2021;13(3). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Ludwig DS, Hu FB, Tappy L, Brand-Miller J. Dietary carbohydrates: role of quality and quantity in chronic disease. BMJ. 2018;361:k2340. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS2179665-supplement-1.docx (8.5MB, docx)

Data Availability Statement

Because of participant confidentiality and privacy concerns, data cannot be shared publicly and requests to access the Nurses’ Health Study, the Nurses’ Health Study II, and the Health Professionals Follow-up Study data must be submitted in writing. According to standard controlled access procedures, applications to use resources from the three studies will be reviewed by our External Collaborations Committee to verify that the proposed use maintains the protection of the privacy of participants and the confidentiality of the data. Investigators wishing to use data from the three studies are asked to submit a brief description of the proposed project (go to https://www.nurseshealthstudy.org/researchers [contact email nhsaccess@channing.harvard.edu] and https://sites.sph.harvard.edu/hpfs/for-collaborators/ for details).

The lead author (HA) affirms that the manuscript is an honest, accurate, and transparent account of the study being reported. No important aspects of the study have been omitted, and any discrepancies from the study as planned have been explained.

RESOURCES