See corresponding editorial on page 675.

Carbohydrate quantity and quality and risk of type 2 diabetes in the European Prospective Investigation into Cancer and NutritionNetherlands (EPIC-NL) study13
Ivonne Sluijs, Yvonne T van der Schouw, Daphne L van der A, Annemieke M Spijkerman, Frank B Hu, Diederick E Grobbee, and Joline W Beulens
ABSTRACT Background: Carbohydrate quantity and quality may play an important role in the development of type 2 diabetes. Objective: We investigated the associations of dietary glycemic load (GL), glycemic index (GI), carbohydrate, and ber intake with the incidence of type 2 diabetes. Design: A prospective cohort study was conducted in 37,846 participants of the EPIC-NL (European Prospective Investigation into Cancer and NutritionNetherlands) study, aged 2170 y at baseline and free of diabetes. Dietary intake was assessed with the use of a validated food-frequency questionnaire. Incident diabetes cases were mainly self-reported and veried against general practitioner records. Results: During a mean follow-up of 10 y, 915 incident diabetes cases were documented. Dietary GL was associated with an increased diabetes risk after adjustment for age, sex, established diabetes risk factors, and dietary factors [hazard ratio (HR) per SD increase: 1.32; 95% CI: 1.14, 1.54; P , 0.001]. GI tended to increase diabetes risk (HR: 1.08; 95% CI: 1.00, 1.17; P = 0.05). Dietary ber was inversely associated with diabetes risk (HR: 0.92; 95% CI: 0.85, 0.99; P , 0.05), whereas carbohydrate intake was associated with increased diabetes risk (HR: 1.15; 95% CI: 1.01, 1.32; P , 0.05). Of the carbohydrate subtypes, only starch was related to increased diabetes risk [HR: 1.25 (1.07, 1.46), P , 0.05]. All associations became slightly stronger after exclusion of energy misreporters. Conclusions: Diets high in GL, GI, and starch and low in ber were associated with an increased diabetes risk. Both carbohydrate quantity and quality seem to be important factors in diabetes prevention. Energy misreporting contributed to a slight attenuation of associations. Am J Clin Nutr 2010;92:90511. INTRODUCTION

and the amount of carbohydrate in a food and reects both the quantity and quality of the carbohydrate (3). Until now, evidence regarding the role of GL and GI in relation to diabetes risk has remained inconclusive. Several studies reported increased risks of GL or GI (39), whereas others did not conrm this (1015). Some studies suggested interactions with the degree of adiposity (4, 6, 9, 14) or with cereal ber intake (3, 7, 8), but this was not always conrmed (4, 8). In Australia, the nutritional guidelines require foods to be labeled with a symbol and their GI value (16); however, the American Diabetes Associations dietary guidelines for diabetes prevention state there is no sufcient, consistent information to conclude that low-GL diets reduce diabetes risk (17). Misreporting of energy intake may be one of the factors explaining the inconsistent ndings regarding GL, GI, and diabetes. To our knowledge, very little is known about the inuence of energy misreporting on GL, GI, and diabetes. Lau et al (18) showed that after correction for energy intake, exclusion of energy underreporters did not affect the associations of GL and GI with body mass index (BMI) (18). To date, only one study has examined the inuence of energy misreporting on the relation of GL and GI with diabetes, and no effect of excluding energy misreporters was reported (12).
From the Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, Utrecht, Netherlands (IS, YTvdS, DEG, and JWB); the Center for Nutrition and Health, National Institute for Public Health and the Environment, Bilthoven, Netherlands (DLvdA); the Center for Prevention and Health Services Research, National Institute for Public Health and the Environment, Bilthoven, Netherlands (AMS); and the Departments of Epidemiology and Nutrition, Harvard School of Public Health, Boston, MA (FBH). 2 Partly supported by the InterAct project, funded by the European Union (Integrated Project LSHM-CT-2006-037197 in the Framework Programme 6 of the European Community). The EPIC-NL study was funded by the European Commission: Public Health and Consumer Protection Directorate 19932004; Research Directorate-General 2005; the Dutch Ministry of Public Health, Welfare, and Sports; the Netherlands Cancer Registry; LK Research Funds; Dutch Prevention Funds; Dutch ZON (Zorg Onderzoek Nederland); and the World Cancer Research Fund. 3 Address correspondence to I Sluijs, Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, PO Box 85500, 3508 GA Utrecht, Netherlands. E-mail: i.sluijs-2@umcutrecht.nl. Received April 1, 2010. Accepted for publication July 9, 2010. First published online August 4, 2010; doi: 10.3945/ajcn.2010.29620.
1

Downloaded from ajcn.nutrition.org by guest on November 12, 2013

Hyperglycemia, a hallmark feature of diabetes, was long viewed as a disorder of carbohydrate metabolism. However, the denition of carbohydrate-containing foods in terms of their capacity to increase blood glucose has been suggested to be a better determinant of diabetes risk (1, 2). Highglycemic index (high-GI) foods contain carbohydrates that break down rapidly and cause high postprandial glucose concentrations, whereas low-GI foods contain carbohydrates that break down slowly and cause lower postprandial glucose concentrations that decline more gradually (1). Glycemic load (GL) is the product of the GI

Am J Clin Nutr 2010;92:90511. Printed in USA. 2010 American Society for Nutrition
Supplemental Material can be found at: http://ajcn.nutrition.org/content/suppl/2010/09/20/ajcn.2010. 29620.DC1.html

905

906

SLUIJS ET AL

In this study, we aimed to prospectively investigate the associations of dietary GL, GI, ber, carbohydrate, and carbohydrate subtypes with the risk of type 2 diabetes. In addition, we investigated the inuence of energy misreporting and explored possible interactions with dietary ber and the degree of adiposity.
SUBJECTS AND METHODS

then divided by the total amount of available carbohydrate consumed (30). Such an expression of dietary GI per gram of carbohydrate reects the overall quality of the daily carbohydrate intake. Daily GL was calculated in the same manner used for the GI but without dividing by the total amount of available carbohydrate consumed (3). The GL represents both the quality and quantity of carbohydrate and the interaction between them. Each unit of dietary GL represents the equivalent of 1 g carbohydrate from glucose. Energy reporting BMR was estimated with the use of the Schoeld equations. Participants with an energy intake compared with BMR of ,1.14 were dened as energy underreporters, whereas those with an energy intake compared with BMR of .2.40 were classied as energy overreporters according to the Goldberg cutoffs (31). Energy misreporters were dened as energy under- plus overreporters. The remaining participants were dened as normal energy reporters. Diabetes The occurrence of diabetes during follow-up was self-reported in 2 follow-up questionnaires at 35 y intervals. Participants were asked whether diabetes was diagnosed, in what year, by whom, and what treatment they received. Diagnoses of diabetes were also obtained from the Dutch Center for Health Care Information, which holds a standardized computerized register of hospital discharge diagnoses. In this register, admission les have been led continuously from all general and university hospitals in the Netherlands from 1990 onward. All diagnoses were coded according to the International Classication of Diseases, Clinical Modication, 9th revision (32). Follow-up was completed 1 January 2006. In the Prospect-EPIC study, incident diabetes cases could also be detected via a urinary glucose strip test, sent out with the rst follow-up questionnaire, for detection of glucosuria. Potential diabetes cases detected by any of these methods were veried against information obtained from the participants general practitioner or pharmacist through mailed questionnaires. Diabetes was dened as present when either of them conrmed the diagnosis. Verication information was available for 89% of the potential diabetes cases, and 72% of these cases were veried as type 2 diabetes and subsequently used for the analysis. Baseline characteristics At baseline, participants completed a general questionnaire containing questions on demographics, presence of chronic diseases, and risk factors for chronic diseases. Smoking was categorized into current, past, or never smoker; and parental history of diabetes was categorized into none, one, or both parents. Physical activity was assessed by means of a questionnaire validated in an elderly population (33), and the Cambridge Physical Activity Score was calculated and used to categorize physical activity (34). Because we could not calculate a total physical activity score for 14% of all participants, we imputed missing scores by single linear regression modeling [Statistical Package for Social Sciences (SPSS), Missing Value Analysis procedure]. Systolic and diastolic blood pressure measurements

Study population The European Prospective Investigation into Cancer and Nutrition (EPIC) was established to investigate the relation between nutrition, various lifestyles, and environmental factors and the incidence of cancer and other chronic diseases (19). EPIC-NL comprises the 2 Dutch contributions to the EPIC study: ProspectEPIC and MORGEN-EPIC. The individual cohorts of EPIC-NL were set up simultaneously in 19931997 within the context of the EPIC study and were merged in 2007 according to standardized processes into one large Dutch EPIC cohort. Its design and rationale are described elsewhere (20). The Prospect-EPIC study includes 17,357 women, aged 4970 y at baseline, who are participating in the national breast cancer screening program and living in the city of Utrecht and its surroundings (21). The MORGEN-EPIC cohort consists of 22,715 men and women aged 2164 y selected from random samples of the Dutch population in 3 towns in the Netherlands (Amsterdam, Doetinchem, and Maastricht) (22, 23). All participants gave informed consent before they were included in the study. The study complied with the Declaration of Helsinki and was approved by the Institutional Board of the University Medical Center Utrecht. After exclusion of prevalent diabetes cases (n = 615), individuals with extremely low or high reported energy intakes (ie, those in the top 0.5% and bottom 0.5% ratio of reported energy intake over estimated energy requirement [estimated with basal metabolic rate (BMR); n = 388]), participants with missing nutritional data (n = 213), and those who did not consent to linkage with disease registries (n = 931), 37,846 participants were left for the analysis. Food intake Daily nutritional intake was obtained from a food-frequency questionnaire (FFQ) containing questions on the usual frequency of consumption of 79 main food items during the year preceding enrollment. This questionnaire allows the estimation of the average daily consumption of 178 foods. The FFQ has been validated against twelve 24-h recalls (2426). Spearmans correlations were 0.60 for GL, 0.62 for GI, 0.74 (men) and 0.76 (women) for carbohydrate, and 0.61 (men) and 0.74 (women) for ber. The GI of the foods was obtained from the international tables compiled by Foster-Powell et al (27) and Atkinson et al (28), which contain all relevant data published between 1981 and 2007. Intakes of nutrients were adjusted for total energy intake by means of the regression residual method (29). Calculation of dietary GI and GL We calculated the daily GI by summing the products of the GI value of a food with its available carbohydrate content, multiplied by the frequency of consumption of that food. These values were

Downloaded from ajcn.nutrition.org by guest on November 12, 2013

GL AND GI AND DIABETES RISK

907

were performed twice in the supine position on the right arm with a Boso Oscillomat (Bosch & Son, Jungingen, Germany; Prospect) or on the left arm with a random zero Sphygmomano-meter (MORGEN), from which the mean was taken. Hypertension was dened as present when one or more of the following criteria were met: diastolic blood pressure 90 mm Hg, systolic blood pressure 140 mm Hg, self-reported antihypertensive medication use, and self-reported presence of hypertension. Waist circumference, height, and weight were measured and the BMI was calculated. All measurements were performed according to standard operating procedures. Weight during follow-up was derived from mailed follow-up questionnaires or physical examination (Doetinchem part). Weight change was dened as the difference between weight at baseline and follow-up. Because the follow-up period varied, we calculated the annual weight change by dividing the weight change by the years of follow-up. Data analysis Characteristics of the study population are presented as the mean (SD) or median (interquartile range) for continuous variables and frequencies (percentages) for categorical variables. For the analyses, GL, GI, and intakes of ber, carbohydrate, sugar, and starch, adjusted for energy intake by the regression residual method (29), were expressed per SD of intake. Cox proportional hazard models were used to calculate crude and adjusted hazard ratios (HRs) and their 95% CIs for the associations between dietary intakes and incidence of type 2 diabetes. Potential confounding factors were selected based on univariable associations of potential confounders with both diabetes and GL or GI, and whether adjustment for the confounder would change the HR by 10%. Variables selected this way were incorporated into a multivariate model by means of a stepwise selection approach. The confounders that were entered into the model were age (continuous), sex (male or female); alcohol consumption (,11, 1125, 2650, .50 g/d); physical activity (not active, moderately inactive, moderately active, active); smoking status (never, past, current); BMI (in kg/ m2; ,20, 2025, 25.130, .30); waist circumference; systolic blood pressure; educational level (high, middle, low); family history of diabetes (none, one parent, both parents); energyadjusted intake of vitamin C, vitamin E, ber, protein, saturated fat, and polyunsaturated fat; and total energy intake (all continuous). All analyses were stratied by cohort (Prospect or MORGEN) by including the cohort in the strata statement. The presence of nonlinear associations of GL, GI, ber, carbohydrate, sugar, or starch was explored by including the quadratic term of these nutrients (per SD increase) in the model with the linear term, and no evidence for nonlinear associations was found (with P values for quadratic terms ranging from 0.14 to 0.70). To examine the inuence of energy under- and misreporting, we repeated the analyses after excluding energy underreporters and energy misreporters, respectively. The inuence of energy overreporting could not be studied because relatively few participants contributed to this group. Interactions with BMI, waist circumference, and ber intake were estimated with the use of a likelihood ratio test. The proportionality assumption was checked visually by means of logminus-log plots, with no deviations detected. Data were analyzed with SPSS (version 15.0; SPSS Inc, Chicago, IL) for Windows.

RESULTS

The mean age of the study population was 51 y, and 25% of the participants were male. Daily mean (6SD) dietary GL and GI and intakes of ber and carbohydrate were 117.9 6 21.3 g, 54.9 6 3.9, 23.4 6 4.8 g, and 222.0 6 30.8 g, respectively. Sugar and starch each contributed 50% to the total carbohydrate. In total, 24.5% of participants were classied as energy underreporters and 1.0% as energy overreporters (Table 1). The main contributors to the GL were bread (34%), potatoes (13%), sweets (11%), and fruit (10%). Milk (products) (20%), bread (17%), fruit (16%), potatoes (15%), and drinks (14%) contributed the most to the GI. Pearsons correlation coefcients between GL and carbohydrate and between GI and carbohydrate were 0.87 and 0.20, respectively. During a mean (6SD) of 10.1 6 1.9 y of follow-up, we documented 915 incident type 2 diabetes cases. In the univariable model, we found a nonsignicantly increased risk of diabetes with higher GL. After adjustment for age, sex, established diabetes risk factors, and nutritional intake, the risk of diabetes was signicantly increased, with an HR of 1.27 (95% CI: 1.11, 1.44) per SD increase in GL. This increase in HR was mainly attributable to correction for dietary factors, and protein intake in particular. For GI, the risk of diabetes was increased in the univariable model. After adjustment for confounders, the association attenuated to borderline signicant (HR per SD increase in GI: 1.08; 95% CI: 1.00, 1.17). Higher dietary ber was associated with a decreased diabetes risk in the multivariable model (HR per SD increase: 0.89; 95% CI: 0.82, 0.98). Higher intakes of carbohydrate and sugar were associated with a lower incidence of diabetes in the univariable analyses, whereas starch was associated with a nonsignicantly increased risk. After adjustment for confounders, the risk of diabetes signicantly increased per SD increase of carbohydrate and starch (HR for carbohydrate: 1.20; 95% CI: 1.01, 1.42; HR for starch: 1.23; 95% CI: 1.07, 1.42) but not for sugar (HR: 1.15; 95% CI: 0.98, 1.35) (Table 2). Energy underreporters tended to be more often female, more often physically inactive, and to have a higher waist circumference and BMI compared with normal energy reporters. Energy overreporters tended to be less often female, less often physically inactive, and to have a lower waist circumference and BMI compared with normal energy reporters. Intakes of potatoes, bread, cakes and cookies, milk (products), and sweets tended to be lower in energy underreporters, whereas these intakes tended to be higher in energy overreporters compared with normal energy reporters (see Table S1 under Supplemental Data in the online issue). Excluding energy underreporters yielded slightly stronger associations with diabetes compared with the associations in the full cohort. We found signicantly increased risks of diabetes with higher dietary GL, GI, carbohydrate, sugar, and starch and a signicantly decreased risk with higher dietary ber. Excluding both energy under- and overreporters yielded comparable results, although the associations of carbohydrate and sugar were attenuated to borderline signicant (Table 3). We observed no interactions with BMI, waist circumference, and ber intake in the multivariable models (P values for interaction for GL: 0.62, 0.99, and 0.15, respectively; for GI: 0.24, 0.89, and 0.47, respectively). There was no interaction between carbohydrate and GI (P value for interaction: 0.67).

Downloaded from ajcn.nutrition.org by guest on November 12, 2013

908

SLUIJS ET AL
TABLE 1 Baseline characteristics and nutritional intake of the study population1 Values Baseline characteristics Male sex [n (%)] Age (y) Higher education [n (%)] Inactive according to Cambridge Physical Activity Index (34) [n (%)] Current smoker [n (%)] BMI (kg/m2) Waist circumference (cm) Systolic blood pressure (mm Hg) Diastolic blood pressure (mm Hg) Hypertension [n (%)] Family history of diabetes [n (%)] Alcohol intake (g/d) Daily nutritional intake Energy (kcal) Glycemic load (g) Glycemic index Carbohydrate (g) Fiber (g) Sugar (g) Starch (g) Saturated fat (g) Polyunsaturated fat (g) Monounsaturated fat (g) Protein (g) Vitamin C (mg) Vitamin E (mg) Energy underreporters [n (%)] Energy overreporters [n (%)]
1 2 3

9684 (25.6) 51 (4258)2 7795 (20.6) 14,379 (38.0) 11,534 (30.5) 25.6 6 4.03 85.1 6 11.4 126.0 6 18.8 77.8 6 10.6 13,569 (35.9) 6817 (18.0) 4.6 (0.714.3) 2053 6 608 117.9 6 21.2 54.9 6 3.9 222.0 6 30.8 23.4 6 4.8 112.4 6 29.5 109.4 6 23.0 32.6 6 5.9 14.9 6 3.9 29.5 6 5.2 75.7 6 11.0 109.4 6 45.3 12.2 6 3.2 9279 (24.5) 380 (1.0)

Downloaded from ajcn.nutrition.org by guest on November 12, 2013

n = 37,846. All dietary variables were adjusted for total energy intake by the regression residual method (29). Median; interquartile range in parentheses (all such values). Mean 6 SD (all such values).

DISCUSSION

In this prospective study, higher dietary GL, GI, and carbohydrate, and lower dietary ber increased the risk of type 2 diabetes in 37,846 Dutch adults aged 2070 y at baseline. Associations became slightly stronger after exclusion of energy misreporters.

The strengths of our study include its prospective design, long follow-up time, large sample size, and large number of incident diabetes cases. Moreover, the use of validated diabetes cases minimized the presence of false-positive diabetes cases. This reduced the dilution of associations. In addition, increments of 1 SD in dietary GL, GI, and carbohydrate were shown to be

TABLE 2 Univariable and adjusted hazard ratios (95% CI) for the association of glycemic load, glycemic index, ber, carbohydrate, sugar, and starch with incident type 2 diabetes1 Model Univariable Model 1 (age, sex) Model 2 (model 1 + diabetes risk factors) Model 3 (model 2 + dietary intake) Glycemic load 1.03 (0.96, 1.10) 1.03 (0.96, 1.10) 1.04 (0.97, 1.13) 1.27 (1.11, 1.44)** Glycemic index 1.15 (1.08, 1.23)** 1.15 (1.08, 1.24)** 1.06 (0.99, 1.14) 1.08 (1.00, 1.17) Fiber 1.04 (0.98, 1.12) 0.98 (0.91, 1.05) 0.81 (0.92, 1.06) 0.89 (0.82, 0.98)* Carbohydrate 0.91 (0.86, 0.97)* 0.92 (0.86, 0.98)* 0.98 (0.91, 1.05) 1.20 (1.01, 1.42)* Sugar 0.87 (0.81, 0.93)** 0.85 (0.79, 0.91)** 0.93 (0.87, 1.00)* 1.15 (0.98, 1.35) Starch 1.06 (0.99, 1.14) 1.11 (1.04, 1.19)* 1.08 (1.01, 1.16)* 1.23 (1.07, 1.42)*

1 Hazard ratios are presented per SD increase. n = 37,846, with 915 incident diabetes cases. Model 1 was adjusted for sex (male or female) and age at recruitment (continuous). Model 2 was adjusted as in model 1 plus energy-adjusted alcohol consumption (,11, 1125, 2650, .50 g/d), physical activity (not active, moderately inactive, moderately active, active), waist circumference (continuous), BMI [in kg/m2; ,20, 2025 (reference group), 25.130, .30], smoking status (never, past, current), mean systolic blood pressure (continuous), educational level (high, middle, low), family history of diabetes (none, one parent, both parents). Model 3 was adjusted as in model 2 plus energy intake and energy-adjusted intake of vitamin C, vitamin E, protein, saturated fat, and polyunsaturated fat (all continuous). Analyses of glycemic load, glycemic index, carbohydrate, sugar, and starch with diabetes risk were additionally adjusted for total ber intake. Analysis of ber with diabetes risk was additionally adjusted for glycemic load. Analysis of sugar with diabetes risk was additionally adjusted for starch intake. Analysis of starch intake with diabetes risk was additionally adjusted for sugar intake. *P , 0.05, **P , 0.001.

GL AND GI AND DIABETES RISK

TABLE 3 Multivariable hazard ratios (95% CI) for the association of glycemic load, glycemic index, ber, carbohydrate, sugar, and starch with incident type 2 diabetes among normal-plus-high or normal energy reporters1 Normal-plus-high energy reporters2 Glycemic load Glycemic index Fiber Carbohydrate Sugar Starch 1.36 (1.15, 1.62)** 1.14 (1.02, 1.27)* 0.86 (0.76, 0.97)* 1.28 (1.03, 1.59)* 1.23 (1.00, 1.51)* 1.31 (1.08, 1.59)* Normal energy reporters3 1.36 1.16 0.85 1.24 1.19 1.28 (1.15, (1.03, (0.76, (1.00, (0.97, (1.06, 1.62)** 1.29)* 0.96)* 1.55) 1.47) 1.56)*

909

1 Hazard ratios are presented per SD increase, adjusted for sex (male or female), age at recruitment (continuous), energy-adjusted alcohol consumption (,11, 1125, 2650, .50 g/d), physical activity (not active, moderately inactive, moderately active, active), waist circumference (continuous), BMI [in kg/m2; ,20, 2025 (reference group), 25.130, .30], smoking status (never, past, current), mean systolic blood pressure (continuous), educational level (high, middle, low), family history of diabetes (none, one parent, both parents), energy intake, and energy-adjusted intake of vitamin C, vitamin E, protein, saturated fat, and polyunsaturated fat (all continuous). Analyses of glycemic load, glycemic index, carbohydrate, sugar, and starch with diabetes risk were additionally adjusted for total ber intake. Analysis of ber with diabetes risk was additionally corrected for glycemic load. Analysis of sugar with diabetes risk was additionally adjusted for starch intake. Analysis of starch intake with diabetes risk was additionally adjusted for sugar intake. *P , 0.05, **P , 0.001. 2 Normal-plus-high energy reporters were dened as energy intake compared with basal metabolic rate of 1.14; n = 28,757, with 563 incident diabetes cases. 3 Normal energy reporters were dened as energy intake compared with basal metabolic rate of 1.14 and 2.40; n = 28,187, with 554 incident diabetes cases.

achievable in practice (35). Some aspects of the study need to be addressed. First, the FFQ was not specically designed to measure GL and GI. However, a validation study showed good agreement of the FFQ with 24-h dietary recalls for both GI and GL (24). Second, we used the Goldberg cutoffs, based on energy intake compared with BMR, to estimate energy misreporting (31). Although these cutoffs are often used in epidemiologic studies, the use of the individual physical activity level to determine energy misreporting would provide more accurate estimations (36). However, no such levels were available in our study. Finally, the presence of diabetes often goes undetected (37). Misclassication of participants with undetected diabetes may have attenuated our ndings. Median dietary GL and GI were relatively low compared with other studies. For example, in the Nurses Health Study I (3), quintile medians for GI ranged from 64 (lowest) to 77 (highest), compared with 50 (lowest) to 60 (highest) in our study. Despite this, we found high dietary GI and GL to be related to increased diabetes. Several previous studies related high-GI diets to increased diabetes risk (3, 4, 69), although this was not always conrmed (1015). GL has also been related to increased diabetes, but there are inconsistencies among studies (39, 1115). There are several possible explanations for these inconsistencies, including variations in study populations (such as age, sex, and ethnicity) and in carbohydrate intake, GL, or GI. Energy misreporting may weaken the associations of GI and GL with diabetes. In this study, 1 in 4 participants misreported their energy intake, and the majority of these were underreporters. This is

comparable to previous ndings (18, 38). We adjusted all nutrients for total energy intake, thereby controlling confounding of energy intake caused by reporting bias (29). However, this method only controls confounding for those who misreport their diet as a whole. Energy underreporters specically tend to misreport foods that largely contribute to the GL and GI, such as sugars, cookies, milk products (relatively low intakes reported), and fruit and vegetables (relatively high intakes reported) (38). Moreover, energy underreporting may be relatively high in incident diabetes cases, as underreporting has been related to obesity (38). One study reported that excluding energy misreporters had no inuence on these relations (12). However, the results of that study were not shown, limiting comparison with our ndings. Another study showed that after adjustment for energy intake, exclusion of energy underreporters did not affect relations of GL and GI with BMI (18). This suggests that energy adjustment sufciently corrected for confounding due to reporting bias. In our study, exclusion of energy misreporters strengthened all associations, even after energy adjustment. However, the inuence of energy misreporting was modest and did not substantially change the conclusions. Energy adjustment thus seems to minimize the problems related to selective misreporting. Consequently, our results, as well as those of previous studies (12, 18), suggest that energy misreporting is not a major contributor to inconsistencies in studies of GL, GI, and diabetes. Several mechanisms may relate high-GI diets to diabetes. High-GI diets can rapidly increase postprandial glucose levels, thereby increasing insulin demand. This may lead to pancreatic exhaustion. In addition, high-GI diets can increase postprandial free fatty acid release, directly increasing insulin resistance (39, 40). Furthermore, high-GI diets may promote weight gain (39, 41), suggesting mediation through weight gain. Additional adjustment for annual weight change did not affect our ndings (HR for GL: 1.26; 95% CI: 1.11, 1.44). However, weight during follow-up was mainly self-reported. This may have limited us in detecting possible mediating effects of weight change. Removal of baseline BMI and waist circumference from the nal model only slightly affected the results (HR for GL: 1.23; 95% CI: 1.08, 1.40). Altogether, this provides little support for a mediation role of body weight (gain). Carbohydrate and sugar were inversely related to diabetes in the univariable, rst, and second models (the latter for sugar only), as also found by others (8, 42, 43). Confounding by healthy lifestyle behaviors may underlie this [eg, high fruit consumption largely contributes to sugar intake (44)]. Indeed, the inverse associations disappeared after further adjustment was made for lifestyle factors. Total carbohydrate was related to increased risk of diabetes in the nal model, in contrast to the majority of prospective studies (3, 4, 811, 42, 43, 45). Carbohydrate intakes were comparable to those reported in the majority of other studies (3, 4, 811, 42, 43, 45), but differences in carbohydrate sources or GL may account for this. The high correlation (0.87) between GL and carbohydrate made it difcult to separate the effects of GL and carbohydrate. However, we found a stronger association for GL than for carbohydrate, suggesting that not only the amount but also the quality of consumed carbohydrate is an important determinant of diabetes. We observed no evidence for different effects of simple and complex carbohydrate subtypes on diabetes risk. The relation of carbohydrate subtypes with diabetes is not yet very clear. Most

Downloaded from ajcn.nutrition.org by guest on November 12, 2013

910

SLUIJS ET AL
of diabetes in the Whitehall II study. Am J Clin Nutr 2007;86: 98894. Sahyoun NR, Anderson AL, Tylavsky FA, Lee JS, Sellmeyer DE, Harris TB. Dietary glycemic index and glycemic load and the risk of type 2 diabetes in older adults. Am J Clin Nutr 2008;87:12631. Schulz M, Liese AD, Fang F, Gilliard TS, Karter AJ. Is the association between dietary glycemic index and type 2 diabetes modied by waist circumference? Diabetes Care 2006;29:11024. Stevens J, Ahn K, Juhaeri J, Houston D, Steffan L, Couper D. Dietary ber intake and glycemic index and incidence of diabetes in AfricanAmerican and white adults: the ARIC study. Diabetes Care 2002;25: 171521. The Glycemic Index (GI) Symbol Program. Available from: http://www. gisymbol.com.au (cited 24 December 2009). American Diabetes Association. Nutrition recommendations and interventions for diabetes 2006. Diabetes Care 2006;29:214057. Lau C, Toft U, Tetens I, et al. Association between dietary glycemic index, glycemic load, and body mass index in the Inter99 study: is underreporting a problem? Am J Clin Nutr 2006;84:6415. Riboli E, Hunt KJ, Slimani N, et al. European Prospective Investigation into Cancer and Nutrition (EPIC): study populations and data collection. Public Health Nutr 2002;5:111324. Beulens JW, Monninkhof EM, Verschuren MW, et al. Cohort prole: the EPIC-NL study. Int J Epidemiol (Epub ahead of print 8 July 2009). Boker LK, van Noord PA, van der Schouw YT, et al. Prospect-EPIC Utrecht: study design and characteristics of the cohort population. European Prospective Investigation into Cancer and Nutrition. Eur J Epidemiol 2001;17:104753. Blokstra A, Smit HA, Bueno de Mesquita HB, Seidell JC, Verschuren WMM. Monitoring van Risicofactoren en Gezondheid in Nederland (MORGEN-project), 19931997. Leefstijl-en risicofactoren: prevalenties en trends. [Monitoring Project on Chronic Disease Risk Factors (MORGEN-project) 19931997: prevalences and trends in lifestyle and risk.] RIVM report 263200008. Bilthoven, Netherlands: National Institute of Public Health and the Environment, 2005 (in Dutch). Verschuren W, Blokstra A, Picavet H, Smit H. Cohort prole: the Doetinchem cohort study. Int J Epidemiol 2008:37:123641. M, Feskens Du H, van der A DL, van Bakel MM, Verberne LD, Ocke EJ. Reproducibility and relative validity of dietary glycaemic index and glycaemic load assessed by the food-frequency questionnaire used in the Dutch cohorts of the European Prospective Investigation into Cancer and Nutrition. Br J Nutr 2009;102:6014. MC, Bueno-De-Mesquita HB, Goddijn HE, et al. The Dutch EPIC Ocke food frequency questionnaire. I. Description of the questionnaire, and relative validity and reproducibility for food groups. Int J Epidemiol 1997;26(suppl 1):S3748. MC, Bueno-De-Mesquita HB, Pols MA, Smit HA, van Staveren Ocke WA, Kromhout D. The Dutch EPIC food frequency questionnaire. II. Relative validity and reproducibility for nutrients. Int J Epidemiol 1997; 26(Suppl 1):S4958. Foster-Powell K, Holt SH, Brand-Miller JC. International table of glycemic index and glycemic load values: 2002. Am J Clin Nutr 2002;76:556. Atkinson FS, Foster-Powell K, Brand-Miller JC. International tables of glycemic index and glycemic load values: 2008. Diabetes Care 2008;31: 22813. Willett WC, Howe GR, Kushi LH. Adjustment for total energy intake in epidemiologic studies. Am J Clin Nutr 1997;65:1220S8S. Wolever TM, Nguyen PM, Chiasson JL, et al. Determinants of diet glycemic index calculated retrospectively from diet records of 342 individuals with non-insulin-dependent diabetes mellitus. Am J Clin Nutr 1994;59:12659. Goldberg GR, Black AE, Jebb SA, et al. Critical evaluation of energy intake data using fundamental principles of energy physiology: 1. Derivation of cut-off limits to identify under-recording. Eur J Clin Nutr 1991;45:56981. World Health Organization. International classication of diseases, 9th revision (ICD-9) Geneva, Switzerland: WHO, 1997. Voorrips LE, Ravelli AC, Dongelmans PC, Deurenberg P, van Staveren WA. A physical activity questionnaire for the elderly. Med Sci Sports Exerc 1991;23:9749. Wareham NJ, Jakes RW, Rennie KL, et al. Validity and repeatability of a simple index derived from the short physical activity questionnaire used in the European Prospective Investigation into Cancer and Nutrition (EPIC) study. Public Health Nutr 2003;6:40713.

prospective studies found no relation between starch and diabetes (10, 11, 42, 43). One reported a positive association (4), in line with our results. Two studies found no relation between total sugar and diabetes (10, 42), whereas one found an inverse relation (4). Possibly, underreporting of sugar intake contributed to these inverse or null ndings. This is supported by our ndings, because the relation of sugar with diabetes strengthened to a signicantly increased risk after exclusion of energy underreporters. Total ber was inversely related with diabetes, as previously reported by others (11, 46, 47) but not all (48). The inverse relation of total ber with diabetes seems mainly attributable to insoluble or cereal ber (48). This may explain the inconsistencies among studies. Unfortunately, our study did not provide information on intake of specic ber subtypes. In conclusion, our ndings support the view that diets high in GL, GI, and carbohydrate, and low in ber increase the risk of diabetes. Both carbohydrate quantity and quality seem to be important factors in diabetes prevention. These ndings were not modied by the degree of adiposity. Energy misreporting may slightly attenuate associations of GL and GI with diabetes, even after correction for total energy intake.
We thank the Central Bureau for Statistics and the PHARMO Institute for providing follow-up data on cardiovascular disease and vital status. The authors responsibilities were as followsIS, YTvdS, DLvdA, AMS, FBH, DEG, and JWB: study concept and design; IS: data analysis; IS, YTvdS, DLvdA, AMS, FBH, DEG, and JWB: interpretation of results; IS, YTvdS, and JWB: drafting of manuscript; and DLvdA, AMS, FBH, and DEG: critical review of manuscript. None of the authors declared a conict of interest.

13.

14.

15.

16. 17. 18.

19.

20. 21.

Downloaded from ajcn.nutrition.org by guest on November 12, 2013

22.

23.

REFERENCES
1. Jenkins DJ, Wolever TM, Taylor RH, et al. Glycemic index of foods: a physiological basis for carbohydrate exchange. Am J Clin Nutr 1981; 34:3626. 2. Sheard NF, Clark NG, Brand-Miller JC, et al. Dietary carbohydrate (amount and type) in the prevention and management of diabetes: a statement by the American diabetes association. Diabetes Care 2004; 27:226671. 3. Salmeron J, Manson JE, Stampfer MJ, Colditz GA, Wing AL, Willett WC. Dietary ber, glycemic load, and risk of non-insulin-dependent diabetes mellitus in women. JAMA 1997;277:4727. 4. Hodge AM, English DR, ODea K, Giles GG. Glycemic index and dietary ber and the risk of type 2 diabetes. Diabetes Care 2004;27:27016. 5. Hu FB, Manson JE, Stampfer MJ, et al. Diet, lifestyle, and the risk of type 2 diabetes mellitus in women. N Engl J Med 2001;345:7907. 6. Krishnan S, Rosenberg L, Singer M, et al. Glycemic index, glycemic load, and cereal ber intake and risk of type 2 diabetes in US black women. Arch Intern Med 2007;167:23049. 7. Salmeron J, Ascherio A, Rimm EB, et al. Dietary ber, glycemic load, and risk of NIDDM in men. Diabetes Care 1997;20:54550. 8. Schulze MB, Liu S, Rimm EB, Manson JE, Willett WC, Hu FB. Glycemic index, glycemic load, and dietary ber intake and incidence of type 2 diabetes in younger and middle-aged women. Am J Clin Nutr 2004;80:34856. 9. Villegas R, Liu S, Gao YT, et al. Prospective study of dietary carbohydrates, glycemic index, glycemic load, and incidence of type 2 diabetes mellitus in middle-aged Chinese women. Arch Intern Med 2007; 167:23106. 10. Barclay AW, Flood VM, Rochtchina E, Mitchell P, Brand-Miller JC. Glycemic index, dietary ber, and risk of type 2 diabetes in a cohort of older Australians. Diabetes Care 2007;30:28113. 11. Meyer KA, Kushi LH, Jacobs DR Jr, Slavin J, Sellers TA, Folsom AR. Carbohydrates, dietary ber, and incident type 2 diabetes in older women. Am J Clin Nutr 2000;71:92130. 12. Mosdol A, Witte DR, Frost G, Marmot MG, Brunner EJ. Dietary glycemic index and glycemic load are associated with high-densitylipoprotein cholesterol at baseline but not with increased risk

24.

25.

26.

27. 28.

29. 30.

31.

32. 33.

34.

GL AND GI AND DIABETES RISK

35. Ebbeling CB, Leidig MM, Feldman HA, Lovesky MM, Ludwig DS. Effects of a low-glycemic load vs low-fat diet in obese young adults: a randomized trial. JAMA 2007;297:2092102. 36. Black AE. The sensitivity and specicity of the Goldberg cut-off for EI: BMR for identifying diet reports of poor validity. Eur J Clin Nutr 2000; 54:395404. 37. Harris MI, Klein R, Welborn TA, Knuiman MW. Onset of NIDDM occurs at least 4-7 yr before clinical diagnosis. Diabetes Care 1992;15:8159. 38. Livingstone MB, Black AE. Markers of the validity of reported energy intake. J Nutr 2003;133(Suppl 3):895S920S. 39. Ludwig DS. The glycemic index: physiological mechanisms relating to obesity, diabetes, and cardiovascular disease. JAMA 2002;287:241423. 40. Willett W, Manson J, Liu S. Glycemic index, glycemic load, and risk of type 2 diabetes. Am J Clin Nutr 2002;76:274S80S. 41. Du H, van der A DL, van Bakel MM, et al. Dietary glycaemic index, glycaemic load and subsequent changes of weight and waist circumference in European men and women. Int J Obes (Lond) 2009;33: 12808. 42. Janket SJ, Manson JE, Sesso H, Buring JE, Liu S. A prospective study of sugar intake and risk of type 2 diabetes in women. Diabetes Care 2003; 26:100815.

911

43. Schulze MB, Schulz M, Heidemann C, Schienkiewitz A, Hoffmann K, Boeing H. Carbohydrate intake and incidence of type 2 diabetes in the European Prospective Investigation into Cancer and Nutrition (EPIC)-Potsdam Study. Br J Nutr 2008;99:110716. 44. Cust AE, Skilton MR, van Bakel MM, et al. Total dietary carbohydrate, sugar, starch and bre intakes in the European Prospective Investigation into Cancer and Nutrition. Eur J Clin Nutr 2009;63(Suppl 4):S3760. 45. Salonen JT, Tuomainen TP, Nyyssonen K, Lakka HM, Punnonen K. Relation between iron stores and non-insulin dependent diabetes in men: case-control study. BMJ 1998;317:727. 46. Montonen J, Knekt P, Jarvinen R, Aromaa A, Reunanen A. Whole-grain and ber intake and the incidence of type 2 diabetes. Am J Clin Nutr 2003;77:6229. 47. Wannamethee SG, Whincup PH, Thomas MC, Sattar N. Associations between dietary ber and inammation, hepatic function, and risk of type 2 diabetes in older men: potential mechanisms for the benets of ber on diabetes risk. Diabetes Care 2009;32:18235. 48. Schulze MB, Schulz M, Heidemann C, Schienkiewitz A, Hoffmann K, Boeing H. Fiber and magnesium intake and incidence of type 2 diabetes: a prospective study and meta-analysis. Arch Intern Med 2007;167: 95665.

Downloaded from ajcn.nutrition.org by guest on November 12, 2013

676

LETTERS TO THE EDITOR

Erratum
Welch AA, Shakya-Shrestha S, Lentjes MAH, Wareham NJ, Khaw K-T. Dietary intake and status of n3 polyunsaturated fatty acids in a population of sh-eating and non-sh-eating meat eaters, vegetarians, and vegans and the precursor-product ratio of a-linolenic acid to long-chain n3 polyunsaturated fatty acids: results from the EPIC-Norfolk cohort. Am J Clin Nutr 2010;92:104051. The term precursor-product ratio used throughout the article would be more correctly called product-precursor ratio. Despite this name change, the ratio and the data and their interpretation remain correct. In addition, inaccurate wording appears in the second sentence of the Results section of the abstract (page 1040). As published, the sentence reads: Total n3 PUFA intakes were 5780% lower in non-sh-eaters than in sh-eaters, but status differences were considerably smaller. Instead, the sentence should read as follows: Total n3 PUFA intakes in non-sh-eaters were 5780% of those in sh-eaters, but status differences were considerably smaller. These gures are referred to correctly in the second sentence of the Discussion section on page 1048. doi: 10.3945/ajcn.110.011346.

Erratum
Sluijs I, van der Schouw YT, van der A DL, et al. Carbohydrate quantity and quality and risk of type 2 diabetes in the European Prospective Investigation into Cancer and NutritionNetherlands (EPIC-NL) study. Am J Clin Nutr 2010;92:90511. In the second sentence of the Results section of the abstract on page 905, the hazard ratio and 95% CI are incorrect. The sentence should read as follows: Dietary GL was associated with an increased diabetes risk after adjustment for age, sex, established diabetes risk factors, and dietary factors [hazard ratio (HR) per SD increase: 1.27; 95% CI: 1.11, 1.44; P , 0.001]. doi: 10.3945/ajcn.110.011361.

Erratum
Freedman DS, Fulton JE, Dietz WH, et al. The identication of children with adverse risk factor levels by body mass index cutoffs from 2 classication systems: the Bogalusa Heart Study. Am J Clin Nutr 2010;92:1298305. After our article (1) was published online, we became aware that the FitnessGram cutoffs for body composition had been revised. (This revision was announced in an e-mail dated November 2010.) According to the FitnessGram website (2), the previous standards for percentage body fat and body mass index (BMI) were based on the best available research at the time they were developed, . . . but some inconsistencies became apparent. The new FitnessGram BMI cutoffs (3) categorize children and adolescents into 4 groups: 1) very lean, 2) within the Healthy Fitness Zone, 3) needs improvementsome risk, and 4) needs improvementhigh risk. The previous and revised cutoffs for the upper categories in FitnessGram, along with the CDC (Centers for Disease Control and Prevention) 85th and 95th percentiles of BMI (4), are shown in Figure 1. As noted in our article (1), the previous FitnessGram BMI cutoffs resulted in marked differences (ranging from 2% to 20%) in the prevalence of children who had a high FitnessGram BMI across ages. The revised FitnessGram BMI cutoffs are fairly close to the CDC 85th percentile, with the cutoffs for some risk varying from the CDC 79th to 83rd percentiles of BMI by sex and age. The high risk cutoffs range from the CDC 87th to 91st percentiles. On the basis of these revised cutoffs, it is likely 1) that the prevalence of children with a high FitnessGram (Continued on next page)