OBJECTIVE

To examine longitudinal and dose-dependent associations of dietary glycemic index (GI), glycemic load (GL), and fiber with body weight and glycemic status during 3-year weight loss maintenance (WLM) in adults at high risk of type 2 diabetes.

RESEARCH DESIGN AND METHODS

In this secondary analysis we used pooled data from the PREVention of diabetes through lifestyle Intervention and population studies in Europe and around the World (PREVIEW) randomized controlled trial, which was designed to test the effects of four diet and physical activity interventions. A total of 1,279 participants with overweight or obesity (age 25–70 years and BMI ≥25 kg ⋅ m−2) and prediabetes at baseline were included. We used multiadjusted linear mixed models with repeated measurements to assess longitudinal and dose-dependent associations by merging the participants into one group and dividing them into GI, GL, and fiber tertiles, respectively.

RESULTS

In the available-case analysis, each 10-unit increment in GI was associated with a greater regain of weight (0.46 kg ⋅ year−1; 95% CI 0.23, 0.68; P < 0.001) and increase in HbA1c. Each 20-unit increment in GL was associated with a greater regain of weight (0.49 kg ⋅ year−1; 0.24, 0.75; P < 0.001) and increase in HbA1c. The associations of GI and GL with HbA1c were independent of weight change. Compared with those in the lowest tertiles, participants in the highest GI and GL tertiles had significantly greater weight regain and increases in HbA1c. Fiber was inversely associated with increases in waist circumference, but the associations with weight regain and glycemic status did not remain robust in different analyses.

CONCLUSIONS

Dietary GI and GL were positively associated with weight regain and deteriorating glycemic status. Stronger evidence on the role of fiber is needed.

Type 2 diabetes is a global health problem and is related to multiple comorbidities such as cardiovascular disease (CVD) (1). Substantial evidence supports that type 2 diabetes may be largely prevented by management of body weight (BW) and improvement in glucose homeostasis with lifestyle modification (2). Studies have shown effectiveness of various dietary interventions on weight loss (WL) and diabetes prevention (3,4). In particular, low-energy diets (LED) based on total or partial meal replacements have been found to result in rapid WL, exemplified by the findings in PREVention of diabetes through lifestyle Intervention and population studies in Europe and around the World (PREVIEW) (5,6). However, weight regain is a common problem after rapid WL, and maintaining WL is a considerable challenge (7).

Diet composition, including the relative contribution of the different macronutrients to total energy intake, may play a role in weight loss maintenance (WLM) and diabetes prevention (8,9). In addition to carbohydrate quantity, carbohydrate quality is of interest (10), but the effect of glycemic index (GI), a marker of carbohydrate glycemic effect (11), and glycemic load (GL), a marker of carbohydrate quality and quantity combined (12), remains controversial. A recent meta-analysis of four randomized controlled trials (RCTs) suggested that there was no difference between low- and high-GI diets in prevention of weight regain, but this result had high heterogeneity (13). For glucose regulation and diabetes incidence, a recent meta-analysis of prospective cohort studies indicated that dietary GI and GL were important predictors of type 2 diabetes development worldwide (14). Nonetheless, most previous clinical trials reporting the effect of GI and GL on WLM and glycemic status did not exceed 1 year (1517) and, hence, were potentially too short given the longer time frame over which the disease manifests itself. Moreover, few observational studies focused on long-term WLM and glycemic status, especially after diet-induced WL.

A recent meta-analysis of prospective and clinical studies suggested that dietary fiber could be a better marker than GI and GL of potential weight control efficacy and risk of noncommunicable diseases including diabetes (10). Unfortunately, evidence based on large-scale observational studies regarding long-term effects of fiber on WLM is scarce. Only one secondary analysis of a long-term RCT explored the association of fiber with 30-month WLM (18), but as with most previous studies on GI and GL, it was conducted in individuals who had excessive BW but were otherwise healthy. It is unclear whether an association would be observed in adults with higher risk of developing type 2 diabetes.

Therefore, the aim of the current study was to investigate longitudinal and dose-dependent associations of GI, GL, and fiber with 3-year WLM and glycemic status after rapid diet-induced WL in adults at high risk of type 2 diabetes in PREVIEW, a randomized controlled trial designed to examine the effects of four diet and physical activity (PA) interventions on diabetes prevention.

Study Design

PREVIEW was a long-term, large-scale RCT conducted at eight intervention centers in Denmark, Finland, the Netherlands, the U.K., Spain, Bulgaria, Australia, and New Zealand (19). The study was initially performed to examine the effects and interactions of two diets and two PA programs on the prevention of type 2 diabetes. There were two phases in this study: an 8-week WL phase with a formula LED containing 810 kcal ⋅ day−1 consumed by all participants (6) and a 148-week WLM intervention phase. The four intervention groups were high-protein, low-GI (HP-LGI) diet (25% of energy [25 E%] from protein, GI <50) or moderate protein, moderate GI diet (15 E% protein, GI >56) combined with either high- or moderate-intensity PA. The study was approved by the human ethics committees at all intervention centers and was conducted in accord with the latest revision of the Declaration of Helsinki.

The current study is a secondary observational analysis based on the data on WLM (8–156 weeks) of PREVIEW, irrespective of original randomization. The start of WLM (at 8 weeks) was considered the baseline for this analysis. Main outcomes were BW and glycosylated hemoglobin A1c (HbA1c). Other outcomes of interest were fat mass (FM), waist circumference (WC), fasting plasma glucose (FPG), fasting insulin, HOMA of insulin resistance (HOMA-IR), and type 2 diabetes events, information regarding which was collected at 8, 26, 52, 104, and 156 weeks relative to the pre-WL baseline and cardiovascular events self-reported by participants over the course of the study. Self-reported dietary intakes, 24-h urinary nitrogen or urea level, and information regarding PA were collected at 26, 52, 104, and 156 weeks.

Study Population

Participants were recruited from June 2013 to April 2015. The main inclusion criteria were age 25–70 years, overweight or obesity, and prediabetes (19). Overweight and obesity were defined as BMI 25–29.9 kg ⋅ m−2 and ≥30 kg ⋅ m−2, respectively. Prediabetes was evaluated in accordance with American Diabetes Association criteria (20). Eligible participants provided written informed consent and were enrolled into the study. Those who lost ≥8% of initial BW and were not diagnosed with diabetes after WL were allowed to enter the WLM phase. Both complete-case and available-case analyses were conducted in this study. The complete-case and available-case analyses involved participants who finished all phases and those who commenced the WLM phase, respectively. Participants with unavailable dietary GI and fiber data at 26 weeks and/or implausible energy intake data (<600 or >3,500 kcal ⋅ day−1 for women and <800 or >4,200 kcal ⋅ day−1 for men) (21) were excluded.

Dietary Assessment

Dietary intake was assessed by food diaries on four consecutive days including one weekend day. Participants were instructed how to use scales and conventional household measurements and to record the foods consumed in detail. Food diaries were collected and in the case of lack of clarity were discussed and, when possible, checked with the participant to clarify any ambiguities. All data from food diaries were entered into national nutrient analysis software (Supplementary Material). The GI of each food was obtained with use of GI databases (Supplementary Table 1). Regarding mixed meals and some recipes, the weighted mean GI of the components was used (22). Total GI and GL were calculated according to the formula of van Woudenbergh et al. (23):

Table 1

Characteristics of participants at the start of WLM (8 weeks) or at 26 weeks (diet and lifestyle outcomes)

All participantsCompletersNoncompletersP
N 1,279 847 432 — 
Sociodemographics*     
 Women, n (%) 851 (66.5) 550 (64.9) 301 (69.7) 0.089 
 Age (years) 56 (45, 63) 57 (48, 63) 53 (43, 61) <0.001 
 Height (m) 1.68 ± 0.09 1.68 ± 0.09 1.67 ± 0.09 0.057 
 Ethnicity, n (%)    0.016 
  Caucasian 1,158 (90.5) 780 (92.1) 378 (87.5) — 
  Asian 29 (2.3) 19 (2.2) 10 (2.3) — 
  Black 19 (1.5) 13 (1.5) 6 (1.4) — 
  Arabic 4 (0.3) 3 (0.4) 1 (0.2) — 
  Hispanic 22 (1.7) 12 (1.4) 10 (2.3) — 
  Other 47 (3.7) 20 (2.4) 27 (6.3) — 
Anthropometric outcomes and body composition*     
 BW (kg) 83.0 (74.1, 94.4) 81.3 (72.8, 90.7) 86.3 (76.5, 99.7) <0.001 
 BMI (kg ⋅ m−229.2 (26.6, 32.8) 28.4 (26.0, 31.7) 30.8 (28.2, 34.6) <0.001 
 FM (kg) 31.1 (24.5, 39.8) 29.6 (23.1, 37.0) 35.2 (28.0, 42.5) <0.001 
 WC (cm) 98.7 ± 12.4 97.4 ± 11.9 101.2 ± 13.1 <0.001 
Glucose tolerance and blood biochemistry*     
 FPG (mmol ⋅ L−15.7 ± 0.5 5.6 ± 0.5 5.7 ± 0.6 <0.014 
 HbA1c (%) 5.3 ± 0.3 5.3 ± 0.3 5.3 ± 0.3 0.825 
 HbA1c (mmol ⋅ mol−134.6 ± 3.1 34.5 ± 3.1 34.8 ± 3.2 0.893 
 Fasting insulin (mU ⋅ L−16.8 (5.1, 9.5) 6.5 (4.9, 8.9) 7.6 (5.6, 10.4) <0.001 
 HOMA-IR 1.7 (1.3, 2.4) 1.6 (1.2, 2.3) 1.9 (1.4, 2.8) <0.001 
Diet and lifestyle outcomes     
 Energy intake from food diary (kcal) 1,648.9 ± 447.6 1,674.1 ± 431.9 1,599.7 ± 473.4 0.005 
 GI from food diary 53.5 ± 8.2 53.9 ± 8.0 52.8 ± 8.5 0.028 
 GL from food diary 90.5 ± 35.6 92.5 ± 35.6 86.5 ± 35.5 0.005 
 Dietary fiber from food diary (g ⋅ day−122.6 ± 8.2 23.4 ± 8.1 21.2 ± 8.1 <0.001 
 Dietary fiber from food diary (E%) 2.7 ± 0.8 2.7 ± 0.8 2.6 ± 0.8 0.001 
 Protein intake from food diary (E%) 21.0 ± 4.6 21.0 ± 4.6 21.1 ± 4.5 0.688 
 Protein intake from urinary nitrogen (g ⋅ day−188.9 ± 35.6 89.4 ± 34.8 88.0 ± 37.3 0.518 
 Carbohydrate intake from food diary (E%) 41.0 ± 8.2 41.1 ± 8.4 40.8 ± 7.9 0.527 
 Fat intake from food diary (E%) 32.8 ± 6.9 32.5 ± 7.0 33.4 ± 6.7 0.022 
 PA (counts ⋅ min−1311.6 (248.2, 396.9) 315.1 (255.2, 404.1) 305.2 (228.0, 382.0) 0.022 
All participantsCompletersNoncompletersP
N 1,279 847 432 — 
Sociodemographics*     
 Women, n (%) 851 (66.5) 550 (64.9) 301 (69.7) 0.089 
 Age (years) 56 (45, 63) 57 (48, 63) 53 (43, 61) <0.001 
 Height (m) 1.68 ± 0.09 1.68 ± 0.09 1.67 ± 0.09 0.057 
 Ethnicity, n (%)    0.016 
  Caucasian 1,158 (90.5) 780 (92.1) 378 (87.5) — 
  Asian 29 (2.3) 19 (2.2) 10 (2.3) — 
  Black 19 (1.5) 13 (1.5) 6 (1.4) — 
  Arabic 4 (0.3) 3 (0.4) 1 (0.2) — 
  Hispanic 22 (1.7) 12 (1.4) 10 (2.3) — 
  Other 47 (3.7) 20 (2.4) 27 (6.3) — 
Anthropometric outcomes and body composition*     
 BW (kg) 83.0 (74.1, 94.4) 81.3 (72.8, 90.7) 86.3 (76.5, 99.7) <0.001 
 BMI (kg ⋅ m−229.2 (26.6, 32.8) 28.4 (26.0, 31.7) 30.8 (28.2, 34.6) <0.001 
 FM (kg) 31.1 (24.5, 39.8) 29.6 (23.1, 37.0) 35.2 (28.0, 42.5) <0.001 
 WC (cm) 98.7 ± 12.4 97.4 ± 11.9 101.2 ± 13.1 <0.001 
Glucose tolerance and blood biochemistry*     
 FPG (mmol ⋅ L−15.7 ± 0.5 5.6 ± 0.5 5.7 ± 0.6 <0.014 
 HbA1c (%) 5.3 ± 0.3 5.3 ± 0.3 5.3 ± 0.3 0.825 
 HbA1c (mmol ⋅ mol−134.6 ± 3.1 34.5 ± 3.1 34.8 ± 3.2 0.893 
 Fasting insulin (mU ⋅ L−16.8 (5.1, 9.5) 6.5 (4.9, 8.9) 7.6 (5.6, 10.4) <0.001 
 HOMA-IR 1.7 (1.3, 2.4) 1.6 (1.2, 2.3) 1.9 (1.4, 2.8) <0.001 
Diet and lifestyle outcomes     
 Energy intake from food diary (kcal) 1,648.9 ± 447.6 1,674.1 ± 431.9 1,599.7 ± 473.4 0.005 
 GI from food diary 53.5 ± 8.2 53.9 ± 8.0 52.8 ± 8.5 0.028 
 GL from food diary 90.5 ± 35.6 92.5 ± 35.6 86.5 ± 35.5 0.005 
 Dietary fiber from food diary (g ⋅ day−122.6 ± 8.2 23.4 ± 8.1 21.2 ± 8.1 <0.001 
 Dietary fiber from food diary (E%) 2.7 ± 0.8 2.7 ± 0.8 2.6 ± 0.8 0.001 
 Protein intake from food diary (E%) 21.0 ± 4.6 21.0 ± 4.6 21.1 ± 4.5 0.688 
 Protein intake from urinary nitrogen (g ⋅ day−188.9 ± 35.6 89.4 ± 34.8 88.0 ± 37.3 0.518 
 Carbohydrate intake from food diary (E%) 41.0 ± 8.2 41.1 ± 8.4 40.8 ± 7.9 0.527 
 Fat intake from food diary (E%) 32.8 ± 6.9 32.5 ± 7.0 33.4 ± 6.7 0.022 
 PA (counts ⋅ min−1311.6 (248.2, 396.9) 315.1 (255.2, 404.1) 305.2 (228.0, 382.0) 0.022 

Data are mean ± SD, median (25th, 75th percentiles), or the number of participants (%).

*

Data were collected at 8 weeks.

Data were collected at 26 weeks.

Participants who entered the WLM phase. Difference between completers and noncompleters in characteristics was examined by t test, Wilcoxon nonparametric test, and χ2 test. t test was used for approximately, normally distributed variables; Wilcoxon nonparametric test was used for non–normally distributed variables; and χ2 test was used for categorical variables.

Protein intake was also objectively assessed by 24-h urine collection (nitrogen or urea) with the following formula: dietary protein (g ⋅ day−1) = [24-h urinary nitrogen (g) / 0.81] × 6.25, and conversion factor urea × 0.4664 = nitrogen (24,25).

Assessment of Anthropometric Outcomes and Body Composition

BW was measured when participants were in a fasting state with an empty bladder and wearing light clothing or underwear. FM was measured by DEXA, bioelectrical impedance, or BOD POD at different intervention centers. WC was measured when participants were at the end of breath expiration, at the midway point between the bottom of the rib cage and the top of the iliac crest.

Assessment of Markers of Glycemic Status

Fasting (>10 h) blood samples were drawn from an antecubital vein and, after processing, were frozen at −80°C and transported to the Finnish Institute for Health and Welfare for determination of FPG, HbA1c, and fasting insulin. HOMA-IR was calculated with the following formula: HOMA-IR = (fasting insulin in mU ⋅ L−1 × FPG in mmol ⋅ L−1)/ 22.5. The diagnosis of type 2 diabetes and CVD incidence is described in Supplementary Material.

Covariates Assessment

Information on age, sex, and ethnicity was collected with self-administered questionnaires at week 0. Stature was measured at week 0. BMI was calculated as BW in kilograms divided by the square of height in meters. PA was assessed by 7-day accelerometry (ActiSleep+, ActiGraph, LLC, Pensacola, FL) in order to obtain mean activity counts, expressed in counts ⋅ min−1 over valid wear time.

Statistical Analysis

Descriptive statistics were used to summarize characteristics for participants. Further information is described in Supplementary Material.

Participants were merged into one group for assessment of longitudinal associations among GI, GL, fiber, and yearly changes in BW, body composition, and markers of glycemic status during WLM. Yearly changes were calculated as changes in outcomes from 8 to 26, 52, 104, and 156 weeks divided by corresponding changes in years. For best representation of the long-term dietary and PA patterns of participants during WLM, a cumulative average method (21) based on all available measurements of self-reported dietary intake, protein intake from urinary nitrogen, and accelerometry-measured PA was used in all analyses. Cumulative average GI, GL, fiber, and other dietary components and PA from 8 to 26, 52, 104, and 156 weeks were calculated. PA and 26-week diet were used to estimate the average PA and dietary intake from 8 to 26 weeks (Supplementary Material and Supplementary Table 2).

Linear mixed models with repeated measurements were used, assuming that missing data occurred at random. Model 1 was adjusted for age (continuous), sex (categorical), ethnicity (categorical [Caucasian, Asian, Black, Arabic, Hispanic, or other]), weight-related or glycemic outcomes at 8 weeks (continuous), BMI at 8 weeks (continuous), and time (categorical) as fixed effects and intervention center (categorical) and participant identifier as random effects. Model 2 was additionally adjusted for time-varying accelerometry-measured PA (continuous) and self-reported energy intake (kcal ⋅ day−1) and dietary components (continuous) including percentage of energy from fat, protein, fiber or carbohydrate, and alcohol (continuous, all in E%) as fixed effects. Model 3 was additionally adjusted for time-varying yearly weight change. A stratified analysis was conducted to examine potential effect modification by sex, age, ethnicity, BMI at 8 weeks, PA, and dietary components.

All participants with available dietary data, irrespective of whether they completed the study, were divided into tertiles by cumulative average GI, GL, and fiber at each interval from 26 weeks (Supplementary Table 2). Dose-dependent associations of GI, GL, and fiber with BW and HbA1c were assessed with use of linear mixed models with repeated measurements adjusted for the covariates in model 2 or 3. For markers of glycemic status, the models were additionally adjusted for time-varying weight change. Differences among three tertiles or between the highest and lowest tertiles in changes in BW and HbA1c were examined. Post hoc analyses with multiple comparisons with Bonferroni adjustment or pairwise comparisons were performed to compare tertiles at each time point where appropriate.

Assessment of the association of Gl, GL, and fiber with type 2 diabetes or CVD incidence is described in Supplementary Material.

We performed sensitivity analyses by 1) replacing self-reported protein intake (E%) with protein intake from urinary nitrogen (g ⋅ day−1), 2) further adjusting for intervention group as fixed effect, and 3) assuming that data were not missing at random (26) (results were reported if they were modified). All data analyses were performed by IBM SPSS, version 26.0 (SPSS, Chicago, IL). Statistical significance was set at a two-tailed P value of <0.05.

The flow of participants is shown in Supplementary Fig. 1. A total of 1,857 participants entered the WLM phase. Of these, 1,279 (4,033–4,130 observations of main outcomes) had available dietary GI and fiber data and plausible energy intake data and were included in the available-case analysis. Of these, 43 and 22 developed type 2 diabetes and CVD, respectively. A total of 847 (3,268–3,344 observations) were included in the complete-case analysis. The median age of the 1,279 participants (66.5% women) was 56 years (range 25–70) (Table 1) and the median values for BW and BMI were 83.0 kg (25th, 75th percentiles 74.1, 94.4) and 29.2 kg ⋅ m−2 (26.6, 32.8), respectively, at 8 weeks. Mean ± SD HbA1c was 5.3% ± 0.3% (34.6 ± 3.1 mmol ⋅ mol−1) at 8 weeks and GI 53.5 ± 8.2, GL 90.5 ± 35.6, and fiber intake 22.6 ± 8.2 g ⋅ day−1 at 26 weeks. On average, compared with noncompleters, completers were older and had lower BW, BMI, FM, WC, FPG, fasting insulin, and HOMA-IR and higher energy intake, GI, GL, and fiber intake at 8 or 26 weeks.

Figure 1

Longitudinal associations of cumulative average GI (each 10 unit) with yearly weight regain and changes in markers of glycemic status during WLM. Model 1 was adjusted for age, sex, ethnicity, anthropometric outcomes or body composition, and markers of glycemic status at 8 weeks, BMI at the start of WLM (8 weeks), and time as fixed effects and intervention center and participant identifier as random effects. Model 2 was additionally adjusted for time-varying accelerometry-measured PA and time-varying self-reported energy intake (kcal ⋅ day−1) and dietary components including percentage of energy from fat, protein, fiber, and alcohol (all in E%). Model 3 was additionally adjusted for time-varying yearly changes in BW. *Yearly mean change and 95% CI of main effects indicating the increase in anthropometric outcomes or body composition or markers of glycemic status increased per year by 10-unit increment in GI. †P values for main effects.

Figure 1

Longitudinal associations of cumulative average GI (each 10 unit) with yearly weight regain and changes in markers of glycemic status during WLM. Model 1 was adjusted for age, sex, ethnicity, anthropometric outcomes or body composition, and markers of glycemic status at 8 weeks, BMI at the start of WLM (8 weeks), and time as fixed effects and intervention center and participant identifier as random effects. Model 2 was additionally adjusted for time-varying accelerometry-measured PA and time-varying self-reported energy intake (kcal ⋅ day−1) and dietary components including percentage of energy from fat, protein, fiber, and alcohol (all in E%). Model 3 was additionally adjusted for time-varying yearly changes in BW. *Yearly mean change and 95% CI of main effects indicating the increase in anthropometric outcomes or body composition or markers of glycemic status increased per year by 10-unit increment in GI. †P values for main effects.

Close modal

Figure 1 shows the associations of cumulative average GI with yearly weight regain and changes in markers of glycemic status during WLM. In model 2, GI was positively associated with regains in BW and FM and increases in HbA1c, fasting insulin, and HOMA-IR in both the complete-case and available-case analyses. Only the association with HbA1c remained significant after adjustment for weight change. Some of the associations were weaker in older participants and those with greater PA volume and alcohol intake, whereas there were no differences in sex (Supplementary Fig. 2). In the available-case analysis, there were significant changes in BW and HbA1c at multiple time points among the GI tertiles (Fig. 2A and B). Specifically, compared with those in the lowest tertile (GI ∼45 [∼52% of participants from the HP-LGI group]), participants in the highest GI tertile (GI ∼61 [∼48% from HP-LGI]) had greater weight regain and increases in HbA1c (Supplementary Fig. 3A and B).

Figure 2

Changes in BW and markers of glycemic status over time during WLM by tertiles of cumulative average GI, GL, and fiber. Values are estimated marginal mean and 95% CI in changes in BW (kg) (A) and HbA1c (%) (B) by GI tertiles, changes in BW (kg) (C) and HbA1c (%) (D) by GL tertiles, and changes in BW (kg) (E) and HbA1c (%) (F) by fiber tertiles. We performed analyses using a linear mixed model with repeated measurements with adjustment for age, sex, ethnicity, anthropometric outcomes or body composition, and markers of glycemic status at the start of WLM (8 weeks); BMI at 8 weeks; and time, time-varying accelerometry-measured PA, and self-reported energy intake (kcal ⋅ day−1) and dietary components including percentage of energy from fat, protein, fiber, or carbohydrate, and alcohol (all in E%) as fixed effects and with participant identifier and intervention center as random effects. For markers of glycemic status, the models were additionally adjusted for time-varying weight change. Time-by-tertile group interaction terms were added. Main effects, time effects, and time-by-tertile group interaction were reported. Post hoc analyses with multiple comparisons with Bonferroni adjustment were performed to compare the tertiles at each time point where appropriate. Values with different lowercase letters (a, b, and c) are significantly different (P < 0.05).

Figure 2

Changes in BW and markers of glycemic status over time during WLM by tertiles of cumulative average GI, GL, and fiber. Values are estimated marginal mean and 95% CI in changes in BW (kg) (A) and HbA1c (%) (B) by GI tertiles, changes in BW (kg) (C) and HbA1c (%) (D) by GL tertiles, and changes in BW (kg) (E) and HbA1c (%) (F) by fiber tertiles. We performed analyses using a linear mixed model with repeated measurements with adjustment for age, sex, ethnicity, anthropometric outcomes or body composition, and markers of glycemic status at the start of WLM (8 weeks); BMI at 8 weeks; and time, time-varying accelerometry-measured PA, and self-reported energy intake (kcal ⋅ day−1) and dietary components including percentage of energy from fat, protein, fiber, or carbohydrate, and alcohol (all in E%) as fixed effects and with participant identifier and intervention center as random effects. For markers of glycemic status, the models were additionally adjusted for time-varying weight change. Time-by-tertile group interaction terms were added. Main effects, time effects, and time-by-tertile group interaction were reported. Post hoc analyses with multiple comparisons with Bonferroni adjustment were performed to compare the tertiles at each time point where appropriate. Values with different lowercase letters (a, b, and c) are significantly different (P < 0.05).

Close modal
Figure 3

Longitudinal associations of cumulative average GL (each 20 unit) with yearly weight regain and changes in markers of glycemic status during WLM. Analyses were performed with use of a linear mixed model with repeated measurements. Model 1 was adjusted for age, sex, ethnicity, weight- or glycemic status–related outcomes at the start of WLM (8 weeks), BMI at 8 weeks, and time as fixed effects and intervention center and participant identifier as random effects. Model 2 was additionally adjusted for time-varying accelerometry-measured PA and time-varying self-reported energy intake (kcal ⋅ day−1) and dietary components including intake of fat, protein, fiber, and alcohol (all in E%). Model 3 was additionally adjusted for time-varying yearly changes in BW. *Yearly mean change and 95% CI of main effects indicating the amount of increase in anthropometric outcomes or body composition or markers of glycemic status increased per year by 20-unit increment in GL. †P values for main effects.

Figure 3

Longitudinal associations of cumulative average GL (each 20 unit) with yearly weight regain and changes in markers of glycemic status during WLM. Analyses were performed with use of a linear mixed model with repeated measurements. Model 1 was adjusted for age, sex, ethnicity, weight- or glycemic status–related outcomes at the start of WLM (8 weeks), BMI at 8 weeks, and time as fixed effects and intervention center and participant identifier as random effects. Model 2 was additionally adjusted for time-varying accelerometry-measured PA and time-varying self-reported energy intake (kcal ⋅ day−1) and dietary components including intake of fat, protein, fiber, and alcohol (all in E%). Model 3 was additionally adjusted for time-varying yearly changes in BW. *Yearly mean change and 95% CI of main effects indicating the amount of increase in anthropometric outcomes or body composition or markers of glycemic status increased per year by 20-unit increment in GL. †P values for main effects.

Close modal

Figure 3 shows the associations of cumulative average GL with yearly weight regain and changes in markers of glycemic status during WLM. In model 2, GL was positively associated with regains in BW and FM and increase in HbA1c and fasting insulin in the complete-case and available-case analyses. Only the association with HbA1c remained significant to adjustment for weight change. After addition of protein intake from urinary nitrogen as a covariate, WC showed significant association with GL, and the associations of fasting insulin and HbA1c with GL were independent of weight change (data not shown). Some of the associations were modified by age, ethnicity, BMI at 8 weeks, PA, and fat and protein intake, whereas there was no difference in sex (Supplementary Fig. 2). In the available-case analysis, there were significant differences in changes in BW and HbA1c at multiple time points among the GL tertiles (Fig. 2C and D). Specifically, compared with those in the lowest tertile (GL ∼58), participants in the highest GL tertile (GL ∼125) had greater weight regain and increases in HbA1c (Supplementary Fig. 3A and B).

Supplementary Fig. 4 shows the associations of cumulative average fiber intake with yearly weight regain and changes in markers of glycemic status during WLM. In model 2, fiber was inversely associated with regains of FM and increases in WC, HbA1c, and fasting insulin in the available-case analysis. Only the association with HbA1c remained significant to adjustment for weight change. After adjustment for protein intake from urinary nitrogen, the association between FM and fiber was lost (data not shown). The associations were not modified by age and sex (Supplementary Fig. 2). There were no differences among the fiber tertiles (Fig. 2F and Supplementary Fig. 3F) or quartiles or quintiles or sextiles (data not shown) in HbA1c in the available-case analysis.

After multivariable adjustment, there were no associations of GI, GL, or fiber with type 2 diabetes or CVD incidence (Supplementary Table 3).

In this secondary analysis of individuals with a high risk of type 2 diabetes from a large international, multiethnic cohort, we show that higher cumulative average GI and GL were associated with increases in BW and markers of glycemic status. Specifically, the associations of GI and GL with HbA1c were independent of weight change. Participants in the highest GI and GL tertiles had significantly greater weight regain and increases in HbA1c than those in the lowest tertiles. Greater fiber intake was associated with decreases in WC, whereas the association of fiber with weight regain, FM, and glycemic status did not remain robust in different analyses.

To date, there are few large-scale clinical trials and observational studies focusing on long-term WLM, especially after rapid diet-induced WL. Regarding clinical trials, in the intention-to-treat analysis of the Diet, Obesity, and Genes (Diogenes) study, there was a difference between the high-GI (GI = 63, 51 E% carbohydrate and 20 E% protein) and low-GI (GI = 58, 51 E% carbohydrate and 18 E% protein) groups in weight regain, FM, and WC at 6 months (16) but not the 1-year follow-up (7). The difference in GI may not have been sufficient for detection of an effect on outcomes. By contrast, in our tertile analysis, we observed larger differences between the highest and lowest tertiles of GI (45 vs. 61) and GL (58 vs. 125). Other clinical trials have reported mixed findings on GI, GL, and WL (27). The lack of reliable data on GI of local foods is another limitation. The relative postprandial glycemic response to mixed meals or diets will be affected by many factors in addition to the GI of each meal component. This includes the fat and protein, as well as carbohydrate content (28), meal preparation methods, and serving temperature (29). In the current study and other studies (7,16), GI and GL were calculated based on food composition and GI databases and the outcomes were adjusted for other macronutrients, which may cause bias.

Considering observational studies, previous cohort studies have simply evaluated baseline GI or GL intake and subsequent changes in BW or body composition (3032). Causal inference is therefore more limited. For instance, Salari-Moghaddam et al. (31) reported that dietary GI was positively related to abdominal obesity in women in a cross-sectional study. Hare-Bruun et al. (30) found positive associations between baseline GI and subsequent 6-year changes in BW, percentage body fat, and WC in women. Unlike these studies, we analyzed long-term, updated, cumulative average GI and GL with concurrent BW changes, which may provide new insights into the causally relevant associations.

Our findings are in line with previous observational studies on GI or GL and glucose metabolism, especially HbA1c. Cheng et al. (33) reported that GI and GL were positively associated with HbA1c in 3,918 Chinese without diabetes. In addition, Wang et al. (34) found a positive longitudinal association between change in GI and change in HbA1c in a 1-year intervention trial without a WL phase. That secondary analysis was based on those classified as having Latino ethnicity and having diabetes, whereas PREVIEW participants were mainly of Caucasian (90%) ethnicity with overweight or obesity and prediabetes. In the present analysis, unlike HbA1c, fasting insulin and HOMA-IR did not show significant associations after adjustment for BW change. This may be because HbA1c, a marker reflecting longer-term glycemic status, is more influenced by postprandial glycemia in individuals with prediabetes (as opposed to individuals with type 2 diabetes) and because fasting insulin and HOMA-IR, markers of shorter-term glycemic status, are less disturbed.

In the available-case analysis, we did not find a link between fiber and BW but we found inverse associations between self-reported fiber and WC, which implies that fiber may be more relevant to central obesity. In a secondary, observational analysis in an RCT with 30-month WLM (18) investigators also found no effect of greater fiber intake on weight. Regarding glucose metabolism, the association of fiber intake with glycemic outcomes in the present analysis failed to remain robust after adjustment for weight change or in the tertile analysis. In contrast, in a cross-sectional study it was found that, after adjustment for age, sex, BMI, and other confounders, the odds ratios for poor glycemic control were reduced with increasing tertiles of fiber intake (35). The casual inference of that study was, however, limited because of its cross-sectional nature. Type of fiber may also be relevant. Different types or sources of fiber, e.g., soluble and insoluble fiber, have different physiological effects. In the primary PREVIEW RCT, participants in the two diet groups were advised not only to select foods from the different food groups in different proportions but also to select different foods from certain food groups for achievement of the differences in percentage of total energy from each macronutrient and GI. This may have resulted in a greater divergence in the types of fiber consumed (and their subsequent functional effects) than would otherwise be expected across the continuum of total fiber intake. Results from large prospective cohort studies have suggested that high insoluble cereal fiber intake may reduce the type 2 diabetes risk, whereas the association with soluble fiber intake is either weak or absent (36). The present analysis focused on total fiber only, so this distinction could not be made. With these factors taken together, the present findings on fiber, WLM, and glucose metabolism should be interpreted with caution.

There are many strengths in the current study. First, we provide new evidence of the associations in question during longer-term WLM, which is more likely to address a lifelong problem, particularly for individuals with obesity. Second, unlike some studies providing standardized meals with fixed caloric content, we determined the associations in a “free-living” context with ad libitum diets. Further, we found that most results, especially regarding GI and GL, from participants who started the WLM were also applicable to completers (who were older and relatively healthier than the available-case population), which implies that selection bias should not be a concern and that the results may be generalizable. Finally, in the main PREVIEW RCT (37), no differences were observed in primary or secondary outcomes between the two diet randomization arms or among the four diet-PA intervention groups. This null result may have occurred because some participants adhered less strictly to their dietary prescriptions and PA protocols, resulting in overlap between arms, especially toward the end. The present analysis merged participants into one group and explored longitudinal associations. In addition, we divided participants into tertiles of GI, GL, and fiber and determined dose-dependent associations and robustness of longitudinal associations.

The current study has several limitations. The GI, GL, and fiber were calculated from self-reported 4-day food diaries. Although food diaries outperform food-frequency questionnaires on estimates of dietary intakes, misreporting is inevitable (38). It is possible that weight regain and dietary misreporting are correlated, which may create bias. Moreover, as food diaries were not collected at the end of the WL phase, we used food intake at 26 weeks to estimate the average food intake from 8 to 26 weeks, which may not be accurate. Finally, GI is a proxy for a certain type of diet, including fruit, vegetables, legumes, berries, and dairy. Although we have correctly adjusted for dietary macronutrient composition, there are several dietary components (e.g., vitamins, minerals, and antioxidants) that we could not adjust for, and, hence, residual and unmeasured confounders may exist. Cigarette smoking has been found to be related to BW (39), and the evidence from epidemiologic studies has demonstrated a clear association between cigarette smoking and increased diabetes risk (40). It is possible that smoking could have affected the results, but this was not measured during WLM and therefore not adjusted for.

In conclusion, this secondary analysis addresses the prevention of weight regain after a period of rapid WL in a large international population with a high risk of diabetes. It may have implications for the lifelong problem of incremental weight gain creep and may provide new evidence that the quality and quantity of carbohydrates are linked to longer-term WLM and glycemic status. Due to the observational nature of the study and residual and unmeasured confounders, the findings should be interpreted with caution. Future research based on large-scale, long-term RCTs should investigate whether diets with lower GI or GL can be recommended to individuals with overweight or obesity and higher risk of diabetes.

Clinical trial reg. no. NCT01777893, clinicaltrials.gov

This article contains supplementary material online at https://doi.org/10.2337/figshare.14364191.

Acknowledgments. The PREVIEW consortium thanks all study participants at every intervention center for their time and commitment and all scientists, advisors, and students for their dedication and contributions to the study. A full list of investigators of the PREVIEW study is available in Supplementary Materials, in addition to the list below. We thank Louise Dye (chairman of the Scientific Advisory Board [SAB]), University of Leeds, Leeds, U.K.; Richard Atkinson (Ethical Officer of the SAB), Virginia Commonwealth University, Richmond, VA; and medical expert and consultant Stephen Colagiuri (University of Sydney). Meyers Madhus A/S is acknowledged for providing training and producing the two cooking books. Furthermore, we gratefully thank the research staff from each center: University of Copenhagen, Laura Pastor-Sanz, Grith Møller, Lone Vestergaard Nielsen, Kasper Nowak, Arne Astrup, Finn Sandø-Pedersen, Morten Bo Johansen, Ulla Skovbæch Pedersen, Maria Roed Andersen, Marianne Juhl Hansen, Jane Jørgensen, Sofie Skov Frost, Lene Stevner; University of Helsinki, Heikki Tikkanen, Saara Kettunen, Tiia Kunnas, Sanna Ritola, Laura Korpipää, Heini Hyvärinen, Karoliina Himanen, Tiina Pellinen, Elina Malkamäki, Heidi Jokinen, Pauliina Kokkonen, Liisi Korhonen, Jaana Valkeapää, Heli Pikkarainen, Martta Nieminen, Tuulia Ingman, Pihla Mäkinen, and Sonja Toijonen; University of Nottingham, Clare Randa ll, Nicky Gilbert, Shelley Archer, Sally Maitland, Melanie Marshall, Cheryl Percival, Jakki Pritchard, Laura Helm, and Peter Mansell; University of Navarra, Blanca Martinez de Morentin, Maria Hernandez Ruiz de Eguilaz, Salome Perez Diez, Veronica Ciaurriz, Angels Batlle, and Maria Jose Cobo; Medical University of Sofia, Georgi Bogdanov, Pavlina Gateva, Rossica Metodieva, and Galia Dobrevska; Swansea University, Kelly Mackintosh, Melitta McNarry, Jeff Stephens, Gareth Dunseath, Steve Luzio, and Masoumeh Minou; Finnish Institute for Health and Welfare, Merja Tukiainen, Ira Greinert, Laura Karjalainen, Jukka Lauronen; University of Sydney, Fiona Atkinson, Michele Whittle, Jessica Burke, Kylie Simpson, Kimberley Way, Sally McClintock, Radhika Seimon, Shelly Keating, Kirsten Bell, Tania Markovic, Cathy Corry, Evalyn Eldering, and Ian Caterson; University of Auckland, Amy Liu, Anne-Thea McGill, Katya Volkova, Madhavi Bollineni, Clarence Vivar, Kelly Storey, Niamh Brennan, Audrey Tay, Lindsay Plank, Nicholas Gant, and Jon Woodhead; Wageningen University, Edith Feskens; and University of Stuttgart, Wolfgang Schlicht.

Funding. This study received funding from EU Framework Programme 7 (FP7/2007-2013) grant 312057; National Health and Medical Research Council - EU Collaborative Grant, AUS 8, identifier 1067711; The Glycemic Index Foundation Australia through royalties to the University of Sydney; the New Zealand Health Research Council (grant 14/191); and University of Auckland Faculty Research Development Fund. The Cambridge Weight Plan donated all products for the 8-week LED period. This study also received funding from the Danish Agriculture & Food Council, the Danish Meat and Research Institute. National Institute for Health Research Biomedical Research Centre (NIHR BRC) (U.K.), Biotechnology and Biological Sciences Research Council (BBSRC) (U.K.), and Engineering and Physical Sciences Research Council (EPSRC) (U.K.). Nutritics (Dublin) donated all dietary analyses software used by University of Nottingham. Juho Vainio Foundation (Finland), Academy of Finland (grants 272376, 314383, 266286, 314135), Finnish Medical Foundation, Gyllenberg Foundation, Novo Nordisk Foundation, Finnish Diabetes Research Foundation, University of Helsinki, Government Research Funds for Helsinki University Hospital (FIN), Jenny and Antti Wihuri Foundation (FIN), Emil Aaltonen Foundation (FIN), and China Scholarship Council also provided funding. I.A.M. is a member of the U.K. government Scientific Advisory Committee on Nutrition, Treasurer of the Federation of European Nutrition Societies, and Treasurer of the World Obesity Federation. J.B.-M. is a member of the Scientific Advisory Board of the Novo Nordisk Foundation.

The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

Duality of Interest. T.M.L. is an advisor for “Sense” diet program. J.B.-M. is President and Director of the Glycemic Index Foundation, oversees a GI testing service at the University of Sydney, and is a co-author of books about diet and diabetes. She is also a member of the Scientific Advisory Board of ZOE Global. S.D.P. was the Fonterra Chair in Human Nutrition during the PREVIEW intervention. I.A.M. is a member of the Mars Scientific Advisory Council, member of the Mars Europe Nutrition Advisory Board, and Scientific Adviser to the Waltham Centre for Pet Nutrition. He was also a member of the Nestle Research Scientific Advisory Board and of the Novozymes Scientific Advisory Board. He is now Scientific Director of the Nestle Institute of Health Sciences. A.R. has received honorariums from Unilever and the International Sweeteners Association. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. The PREVIEW project was designed by M.F., M.S.W.-P., J.B-.M., A.R., Wolfgang Schlicht (W.S.), and Edith Feskens. The protocol for the PREVIEW adult intervention study was written by T.M.L., M.F., and A.R. W.S., S.D.P., G.S., S.H., J.A.M., I.A.M., and M.S.W.-P. were involved in developing the study design. L.M.V. designed the eight postprandial studies to obtain the GI and GL values and led the compilation work for the estimated GI and GL databases of the Fineli database used in the study. R.Z., T.M.L., J.B-.M., and A.R. conceived the research question of this secondary analysis. R.Z., T.M.L., J.B-.M., and A.R. designed the analysis plan. R.Z. performed the data analysis. C.R. provided statistical supervision. R.Z., J.B-.M., and A.R. formed the writing group. R.Z. drafted the manuscript with supervision from A.R. and J.B-.M. All authors contributed to the implementation of the experimental trial and contributed to analysis and interpretation of the data. All authors contributed to critical revision of the manuscript for important intellectual content. All authors agreed that the accuracy and integrity of the work have been appropriately investigated and resolved, and all approved the final version of the manuscript. A.R. attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted. A.R. and R.Z. are the guarantors of this work and, as such, 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.

Prior Presentation. Parts of this study were presented in abstract and ePoster form at the DNSG 2021 - 38th International Symposium on Diabetes and Nutrition of the Diabetes and Nutrition Study Group of the European Association for the study of Diabetes, online, 21–24 June 2021.

1.
Chatterjee
S
,
Khunti
K
,
Davies
MJ
.
Type 2 diabetes
.
Lancet
2017
;
389
:
2239
2251
2.
Merlotti
C
,
Morabito
A
,
Pontiroli
AE
.
Prevention of type 2 diabetes; a systematic review and meta-analysis of different intervention strategies
.
Diabetes Obes Metab
2014
;
16
:
719
727
3.
Johnston
BC
,
Kanters
S
,
Bandayrel
K
, et al
.
Comparison of weight loss among named diet programs in overweight and obese adults: a meta-analysis
.
JAMA
2014
;
312
:
923
933
4.
Ley
SH
,
Hamdy
O
,
Mohan
V
,
Hu
FB
.
Prevention and management of type 2 diabetes: dietary components and nutritional strategies
.
Lancet
2014
;
383
:
1999
2007
5.
Gudzune
KA
,
Doshi
RS
,
Mehta
AK
, et al
.
Efficacy of commercial weight-loss programs: an updated systematic review
.
Ann Intern Med
2015
;
162
:
501
512
6.
Christensen
P
,
Meinert Larsen
T
,
Westerterp-Plantenga
M
, et al
.
Men and women respond differently to rapid weight loss: Metabolic outcomes of a multi-centre intervention study after a low-energy diet in 2500 overweight, individuals with pre-diabetes (PREVIEW)
.
Diabetes Obes Metab
2018
;
20
:
2840
2851
7.
Aller
EE
,
Larsen
TM
,
Claus
H
, et al
.
Weight loss maintenance in overweight subjects on ad libitum diets with high or low protein content and glycemic index: the DIOGENES trial 12-month results
.
Int J Obes
2014
;
38
:
1511
1517
8.
Ebbeling
CB
,
Feldman
HA
,
Klein
GL
, et al
.
Effects of a low carbohydrate diet on energy expenditure during weight loss maintenance: randomized trial
.
BMJ
2018
;
363
:
k4583
9.
Huntriss
R
,
Campbell
M
,
Bedwell
C
.
The interpretation and effect of a low-carbohydrate diet in the management of type 2 diabetes: a systematic review and meta-analysis of randomised controlled trials
.
Eur J Clin Nutr
2018
;
72
:
311
325
10.
Reynolds
A
,
Mann
J
,
Cummings
J
,
Winter
N
,
Mete
E
,
Te Morenga
L
.
Carbohydrate quality and human health: a series of systematic reviews and meta-analyses
.
Lancet
2019
;
393
:
434
445
11.
Wolever
TM
.
Is glycaemic index (GI) a valid measure of carbohydrate quality?
Eur J Clin Nutr
2013
;
67
:
522
531
12.
Monro
JA
,
Shaw
M
.
Glycemic impact, glycemic glucose equivalents, glycemic index, and glycemic load: definitions, distinctions, and implications
.
Am J Clin Nutr
2008
;
87
:
237S
243S
13.
van Baak
MA
,
Mariman
ECM
.
Dietary strategies for weight loss maintenance
.
Nutrients
2019
;
11
:
1916
14.
Livesey
G
,
Taylor
R
,
Livesey
HF
, et al
.
Dietary glycemic index and load and the risk of type 2 diabetes: a systematic review and updated meta-analyses of prospective cohort studies
.
Nutrients
2019
;
11
:
E1280
15.
Sloth
B
,
Due
A
,
Larsen
TM
,
Holst
JJ
,
Heding
A
,
Astrup
A
.
The effect of a high-MUFA, low-glycaemic index diet and a low-fat diet on appetite and glucose metabolism during a 6-month weight maintenance period
.
Br J Nutr
2009
;
101
:
1846
1858
16.
Larsen
TM
,
Dalskov
SM
,
van Baak
M
, et al.;
Diet, Obesity, and Genes (Diogenes) Project
.
Diets with high or low protein content and glycemic index for weight-loss maintenance
.
N Engl J Med
2010
;
363
:
2102
2113
17.
Philippou
E
,
Neary
NM
,
Chaudhri
O
, et al
.
The effect of dietary glycemic index on weight maintenance in overweight subjects: a pilot study
.
Obesity (Silver Spring)
2009
;
17
:
396
401
18.
Champagne
CM
,
Broyles
ST
,
Moran
LD
, et al
.
Dietary intakes associated with successful weight loss and maintenance during the Weight Loss Maintenance trial
.
J Am Diet Assoc
2011
;
111
:
1826
1835
19.
Fogelholm
M
,
Larsen
TM
,
Westerterp-Plantenga
M
, et al
.
PREVIEW: PREVention of diabetes through lifestyle Intervention and population studies in Europe and around the World. Design, methods, and baseline participant description of an adult cohort enrolled into a three-year randomised clinical trial
.
Nutrients
2017
;
9
:
632
20.
American Diabetes Association
.
2. Classification and diagnosis of diabetes: Standards of Care in Diabetes—2017
.
Diabetes Care
2017
;
40
(
Suppl. 1
):
S11
S24
21.
Bhupathiraju
SN
,
Tobias
DK
,
Malik
VS
, et al
.
Glycemic index, 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
:
218
232
22.
Wolever
TM
,
Yang
M
,
Zeng
XY
,
Atkinson
F
,
Brand-Miller
JC
.
Food glycemic index, as given in glycemic index tables, is a significant determinant of glycemic responses elicited by composite breakfast meals
.
Am J Clin Nutr
2006
;
83
:
1306
1312
23.
van Woudenbergh
GJ
,
Kuijsten
A
,
Sijbrands
EJ
,
Hofman
A
,
Witteman
JC
,
Feskens
EJ
.
Glycemic index and glycemic load and their association with C-reactive protein and incident type 2 diabetes
.
J Nutr Metab
2011
;
2011
:
623076
24.
Bingham
SA
.
Urine nitrogen as a biomarker for the validation of dietary protein intake
.
J Nutr
2003
;
133
(
Suppl. 3
):
921S
924S
25.
Bingham
SA
.
The use of 24-h urine samples and energy expenditure to validate dietary assessments
.
Am J Clin Nutr
1994
;
59
(
Suppl
.):
227S
231S
26.
Son
H
,
Friedmann
E
,
Thomas
SA
.
Application of pattern mixture models to address missing data in longitudinal data analysis using SPSS
.
Nurs Res
2012
;
61
:
195
203
27.
Esfahani
A
,
Wong
JM
,
Mirrahimi
A
,
Villa
CR
,
Kendall
CW
.
The application of the glycemic index and glycemic load in weight loss: a review of the clinical evidence
.
IUBMB Life
2011
;
63
:
7
13
28.
Ballance
S
,
Knutsen
SH
,
Fosvold
ØW
,
Fernandez
AS
,
Monro
J
.
Predicting mixed-meal measured glycaemic index in healthy subjects
.
Eur J Nutr
2019
;
58
:
2657
2667
29.
Raben
A
.
Glycemic index and metabolic risks: how strong is the evidence?
Am J Clin Nutr
2014
;
100
:
1
3
30.
Hare-Bruun
H
,
Flint
A
,
Heitmann
BL
.
Glycemic index and glycemic load in relation to changes in body weight, body fat distribution, and body composition in adult Danes
.
Am J Clin Nutr
2006
;
84
:
871
879
31.
Salari-Moghaddam
A
,
Keshteli
AH
,
Haghighatdoost
F
,
Esmaillzadeh
A
,
Adibi
P
.
Dietary glycemic index and glycemic load in relation to general obesity and central adiposity among adults
.
Clin Nutr
2019
;
38
:
2936
2942
32.
Murakami
K
,
McCaffrey
TA
,
Gallagher
AM
,
Neville
CE
,
Boreham
CA
,
Livingstone
MB
.
Dietary glycemic index and glycemic load in relation to changes in body composition measures during adolescence: Northern Ireland Young Hearts Study
.
Int J Obes
2014
;
38
:
252
258
33.
Cheng
G
,
Xue
H
,
Luo
J
, et al
.
Relevance of the dietary glycemic index, glycemic load and genetic predisposition for the glucose homeostasis of Chinese adults without diabetes
.
Sci Rep
2017
;
7
:
400
34.
Wang
ML
,
Gellar
L
,
Nathanson
BH
, et al
.
Decrease in glycemic index associated with improved glycemic control among Latinos with type 2 diabetes
.
J Acad Nutr Diet
2015
;
115
:
898
906
35.
Jiang
J
,
Qiu
H
,
Zhao
G
, et al
.
Dietary fiber intake is associated with HbA1c level among prevalent patients with type 2 diabetes in Pudong New Area of Shanghai, China
.
PLoS One
2012
;
7
:
e46552
36.
Weickert
MO
,
Pfeiffer
AFH
.
Impact of dietary fiber consumption on insulin resistance and the prevention of type 2 diabetes
.
J Nutr
2018
;
148
:
7
12
37.
Raben
A
,
Vestentoft
PS
,
Brand-Miller
J
, et al
.
The PREVIEW intervention study: results from a 3-year randomized 2 x 2 factorial multinational trial investigating the role of protein, glycaemic index and physical activity for prevention of type 2 diabetes
.
Diabetes Obes Metab
2021
;
23
:
324
337
38.
Park
Y
,
Dodd
KW
,
Kipnis
V
, et al
.
Comparison of self-reported dietary intakes from the Automated Self-Administered 24-h recall, 4-d food records, and food-frequency questionnaires against recovery biomarkers
.
Am J Clin Nutr
2018
;
107
:
80
93
39.
Wang
Q
.
Smoking and body weight: evidence from China health and nutrition survey
.
BMC Public Health
2015
;
15
:
1238
40.
Aggarwal
S
,
Khandelwal
D
,
Dutta
D
,
Kalra
S
,
Balhara
YPS
.
Diabetes and smoking: the burden of evidence
. In:
Rodriguez-Saldana
J
, ed.
The Diabetes Textbook
.
Berlin
:
Springer
,
2019
, p.
611
616
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at https://www.diabetesjournals.org/content/license.