Hostname: page-component-78c5997874-j824f Total loading time: 0 Render date: 2024-10-31T16:38:52.182Z Has data issue: false hasContentIssue false

Carbohydrate intake and incidence of type 2 diabetes in the European Prospective Investigation into Cancer and Nutrition (EPIC)-Potsdam Study

Published online by Cambridge University Press:  01 May 2008

Matthias B. Schulze*
Affiliation:
Department of Epidemiology, German Institute of Human Nutrition Potsdam-Rehbruecke, Arthur-Scheunert-Allee 114-116, 14558Nuthetal, Germany
Mandy Schulz
Affiliation:
Department of Epidemiology, German Institute of Human Nutrition Potsdam-Rehbruecke, Arthur-Scheunert-Allee 114-116, 14558Nuthetal, Germany
Christin Heidemann
Affiliation:
Department of Epidemiology, German Institute of Human Nutrition Potsdam-Rehbruecke, Arthur-Scheunert-Allee 114-116, 14558Nuthetal, Germany
Anja Schienkiewitz
Affiliation:
Department of Epidemiology, German Institute of Human Nutrition Potsdam-Rehbruecke, Arthur-Scheunert-Allee 114-116, 14558Nuthetal, Germany
Kurt Hoffmann
Affiliation:
Department of Epidemiology, German Institute of Human Nutrition Potsdam-Rehbruecke, Arthur-Scheunert-Allee 114-116, 14558Nuthetal, Germany
Heiner Boeing
Affiliation:
Department of Epidemiology, German Institute of Human Nutrition Potsdam-Rehbruecke, Arthur-Scheunert-Allee 114-116, 14558Nuthetal, Germany
*
*Corresponding author: Dr Matthias B. Schulze, fax +49 33200 88 721, email mschulze@dife.de
Rights & Permissions [Opens in a new window]

Abstract

It remains unclear what long-term effects of substituting carbohydrates at the expense of protein or fat may have with regard to diabetes risk. Our objective was to evaluate carbohydrate intake in predicting type 2 diabetes using substitution models for fat and protein. We conducted a prospective cohort study of 9702 men and 15 365 women aged 35–65 years and free of diabetes at baseline (1994–8) who were followed for incident type 2 diabetes until 2005. Dietary intake of macronutrients was estimated with a validated FFQ. We estimated the relative risk (RR) using Cox proportional hazards analysis. During 176 117 person-years of follow-up we observed 844 incident cases of physician-confirmed type 2 diabetes. After adjustment for age, BMI, waist circumference, potential lifestyle and dietary confounders, substituting 5 % of energy intake from total, saturated, or monounsaturated fat with carbohydrates was not associated with diabetes risk. In contrast, substituting carbohydrates for protein or PUFA was inversely related to diabetes risk (RR for 5 % energy substitution of protein 0·77 (95 % CI 0·64, 0·91); RR for PUFA 0·83 (95 % CI 0·70, 0·98)). These associations appeared to be similar for men and women, but gained statistical significance only among men for protein (RR 0·78 (95 % CI 0·61, 0·99)). Restricted cubic spline regression did not indicate non-linearity of these associations (P for non-linearity in full cohort was 0·353 and 0·349). In conclusion, a higher carbohydrate intake at the expense of protein and PUFA might be associated with decreased diabetes risk.

Type
Full Papers
Copyright
Copyright © The Authors 2007

While obesity and lifestyle characteristics such as physical activity are established risk factors for type 2 diabetes(Reference Hu, Manson, Stampfer, Colditz, Liu, Solomon and Willett1, Reference Schulze and Hu2), less is known about dietary factors. The quantity of carbohydrates has received particular interest, as high-carbohydrate diets generally produce high postprandial glucose and insulin responses. However, the total percentage of energy derived from carbohydrates in the diet has generally not been found to predict diabetes risk in prospective studies(Reference Salmeron, Ascherio, Rimm, Colditz, Spiegelman, Jenkins, Stampfer, Wing and Willett3Reference Schulz, Liese, Fang, Gilliard and Karter8). Previous studies largely ignored that differences in carbohydrate intake in isoenergetic settings reflect substitutions for other macronutrients(Reference Salmeron, Ascherio, Rimm, Colditz, Spiegelman, Jenkins, Stampfer, Wing and Willett3Reference Hodge, English, O'Dea and Giles6, Reference Schulz, Liese, Fang, Gilliard and Karter8). Substituting carbohydrates for protein or fat or different fatty acids may have very different metabolic consequences beyond glucose responses. For example, HDL-cholesterol blood levels, which are a potent predictor of diabetes risk(Reference Fagot-Campagna, Narayan, Hanson, Imperatore, Howard, Nelson, Pettitt and Knowler9, Reference Fagot-Campagna, Knowler, Narayan, Hanson, Saaddine and Howard11), depend on the macronutrient composition with distinct effects of different fatty acids(Reference Mensink, Zock, Kester and Katan12). The intake of n-3 PUFA may regulate adiponectin secretion(Reference Neschen, Morino, Rossbacher, Pongratz, Cline, Sono, Gillum and Shulman13), which is strongly related to diabetes risk(Reference Hara, Yamauchi and Kadowaki14). In addition, protein-rich foods are known to increase postprandial insulin secretion without augmenting glucose concentrations(Reference Nuttall, Mooradian, Gannon, Billington and Krezowski15, Reference Simpson, McDonald, Wahlqvist, Atley and Outch16). Analysing the association between carbohydrate intake and diabetes risk in statistical models, which take into consideration the other nutrients that carbohydrates are substituted with(Reference Hu, Stampfer, Manson, Rimm, Colditz, Rosner, Hennekens and Willett17, Reference Faerch, Lau, Tetens, Pedersen, Jorgensen, Borch-Johnsen and Glumer18), may help clarify this complex relationship.

We aimed to evaluate the association between carbohydrate intake and risk of type 2 diabetes, using macronutrient substitution models in a large prospective cohort study of men and women.

Materials and methods

Study population

The European Prospective Investigation into Cancer and Nutrition (EPIC)-Potsdam study involves 27 548 subjects, 16 644 women aged mainly 35–65 years and 10 904 men aged mainly 40–65 years, from the general population of Potsdam, Germany, recruited between 1994 and 1998(Reference Boeing, Wahrendorf and Becker19, Reference Boeing, Korfmann and Bergmann20). The baseline examination included anthropometric measurements as well as a personal interview and a questionnaire on prevalent diseases and socio-demographic and lifestyle characteristics. Follow-up questionnaires have been administered every 2–3 years. Response rates for follow-up rounds 1, 2, 3 and 4 were 96, 95, 91 and 90 % (by 31 August 2005).

The prevalence of diabetes mellitus at baseline was evaluated by a physician using information on self-reported medical diagnoses, medication records and dieting behaviour. Uncertainties regarding a proper diagnosis were clarified with the participant or treating physician. After exclusion of participants with prevalent diabetes at baseline or with self-reported diabetes during follow-up but without physician confirmation (n 1567), with missing follow-up time (n 589), with missing diet and confounder information at baseline (n 226), or with implausible energy intake below 3350 kJ (800 kcal) or above 25 100 kJ (6000 kcal) per d (n 99), a total of 9702 men and 15 365 women remained for analyses.

Ascertainment of type 2 diabetes

Potentially incident cases of diabetes were identified in each follow-up questionnaire via self-reports of a diabetes diagnosis, diabetes-relevant medication or dietary treatment due to diabetes. All potentially incident cases of diabetes were verified by questionnaires mailed to the diagnosing physician asking about the date and type of diagnosis, diagnostic tests, and treatment of diabetes. Only cases with a physician diagnosis of type 2 diabetes (International Statistical Classification of Diseases (ICD10) code E11) and a diagnosis date after the baseline examination were considered as confirmed incident cases of type 2 diabetes.

Dietary assessment

All participants were asked to complete a semi-quantitative FFQ which assessed the average frequency of intake and the portion size of 148 foods consumed during the 12 months before examination. Frequency of intake was measured using ten categories, ranging from ‘never’ to ‘five times per day or more’. Portion sizes were estimated using photographs of standard portion sizes. Information on frequency of intake and portion size was used to calculate the amount of each food item in g consumed on average per d. Nutrient intake was calculated according to the German Food Code and Nutrient Data Base(Reference Dehne, Klemm, Henseler and Hermann-Kunz21) version II.3. Intakes estimated from the FFQ at baseline were calculated in g/d. These intakes were then calibrated to account for systematic over- or underestimation. Here, the single 24 h recalls of the EPIC calibration study with 2297 participants were used as the reference instrument(Reference Slimani, Ferrari and Ocke22, Reference Kynast-Wolf, Becker, Kroke, Brandstetter, Wahrendorf and Boeing23). Before calibration, intake from the single 24 h recall was shrunken to the sex- and age-group-specific mean using the external within-person variance estimate from another calibration study with repeated 24 h recalls. Shrinkage excludes the intra-individual variance component and the shrunken intake values can be considered as estimates of habitual dietary intake. Then, a linear calibration method was applied ensuring that the mean and the variance of the calibrated FFQ data are equal to the mean and variance of estimated habitual dietary intake from 24 h recalls. Intakes of macronutrients were expressed as absolute intake (g/d) and nutrient density (% of total energy intake); fibre and Mg intake were adjusted for total energy intake using the residual method(Reference Willett, Howe and Kushi24). The deattenuated correlation coefficients between non-calibrated FFQ and twelve 24 h dietary recalls for energy-adjusted intake of carbohydrates was 0·58 for total carbohydrates, 0·66 for monosaccharides, 0·65 for disaccharides and 0·65 for polysaccharides(Reference Kroke, Klipstein-Grobusch, Voss, Moseneder, Thielecke, Noack and Boeing25, Reference Bohlscheid-Thomas, Hoting, Boeing and Wahrendorf26).

Assessment of lifestyle exposures

Information on educational attainment, smoking, occupational activity level and leisure-time physical activity were assessed with a self-administered questionnaire and a personal interview. We considered sport activities and biking as leisure-time activities, both calculated as the average time spent per week during the 12 months before baseline recruitment. Anthropometric measurement procedures followed standard protocols under strict quality control(Reference Kroke, Bergmann, Lotze, Jeckel, Klipstein-Grobusch and Boeing27, Reference Klipstein-Grobusch, Georg and Boeing28).

Statistical analyses

We estimated the relative risk (RR) for each quintile of carbohydrate intake compared with the lowest quintile using Cox proportional hazards analysis stratified by age. Age was used as the primary time-dependent variable in all models, with entry time defined as the subject's age at recruitment and exit time as the date of diagnosis of diabetes, death, or return of the last follow-up questionnaire. We used information on covariates obtained from the baseline examination in multivariate analyses, including sex, education, occupational activity, sport activity, biking, smoking, total energy intake and alcohol intake. Additional adjustments were made for BMI and waist circumference as well as fibre intake, Mg intake, and the PUFA:SFA and MUFA:SFA ratios. In multivariate nutrient-density models(Reference Willett, Lenart and Willett29), we simultaneously included energy intake, the percentages of energy derived from carbohydrates and alcohol and other potentially confounding variables. We also considered energy densities of protein, total fat and fatty acids. The coefficients from these models can be interpreted as isoenergetic substitution of 5 % of energy from carbohydrates for the same percentage of energy from protein, fat or specific fatty acids. Restricted cubic spline regression was used to examine non-linearity of the RR function for nutrient-density data(Reference Harrell, Lee and Pollock30) using the SAS Macro RCS(Reference Heinzl and Kaider31). Four knots were selected separately for men and women according to the 5th, 25th, 75th and 95th percentiles of carbohydrate intake. The reference value was fixed at 45 % energy contribution from carbohydrates. Analyses were stratified by sex and were performed with SAS release 9.1 (SAS Institute, Cary, NC, USA).

Results

At baseline, subjects with higher carbohydrate intake were older, cycled more frequently, had a lower prevalence of smoking but a lower educational level (Table 1). Men with high carbohydrate intake had lower BMI and waist circumferences, while anthropometry was not related to carbohydrate intake among women. With regard to diet, participants with higher carbohydrate intake had higher intakes of fibre and Mg and lower intake of fat, protein and alcohol.

Table 1 Baseline characteristics by quintiles of carbohydrate intake (% energy) among men and women in the European Prospective Investigation into Cancer and Nutrition-Potsdam Study, 2005 (Mean values or percentages)

During 176 117 person-years of follow-up we observed 844 incident cases of type 2 diabetes (491 men and 353 women). The crude incidence of diabetes increased with increasing age and was higher among men than women (Fig. 1). To evaluate the association between carbohydrate intake and diabetes risk, we first used multivariate nutrient-density models expressing carbohydrate intake as percentage of total energy intake. A higher carbohydrate intake was associated with a lower risk of diabetes in age-adjusted models among men (Table 2). The RR for the highest quintile compared with the lowest was 0·71 (95 % CI 0·54, 0·94). This association was attenuated controlling for lifestyle confounders as well as BMI and waist circumference (RR 0·83 (95 % CI 0·62, 1·12)). This association was further attenuated after controlling for the intake of Mg, fibre, the PUFA:SFA ratio and the MUFA:SFA ratio (RR 0·91 (95 % CI 0·66, 1·26); P = 0·78 for trend). Associations among women were very similar, although they did not gain statistical significance in any model. In the full adjusted model, the RR for extreme quintiles was 0·89 (95 % CI 0·62, 1·29; P = 0·64 for trend).

Fig. 1 Incidence rate (per 10 000 person-years) of type 2 diabetes mellitus among men () and women (■) for different age groups in the European Prospective Investigation into Cancer and Nutrition-Potsdam Study, 2005.

Table 2 Relative risk (RR) of type 2 diabetes by quintiles of total carbohydrate intake (% energy) among men and women in the European Prospective Investigation into Cancer and Nutrition-Potsdam Study, 2005

* Adjusted for age, education (in or no training, vocational training, technical school, or technical college or university degree), occupational activity (light, moderate, heavy), sport activity (0, 0·1–4·0, > 4·0 h/week), cycling (0, 0·1–2·4, 2·5–4·9, ≥  5 h/week), smoking (never, past, current <  20 cigarettes/d, current ≥  20 cigarettes/d), alcohol intake, and total energy intake.

Additionally adjusted for fibre intake, Mg intake, PUFA:SFA ratio, and MUFA:SFA ratio.

We further used different multivariate nutrient-density models to model specific energy substitution. Exchanging carbohydrates for total fat was not associated with diabetes risk (Fig. 2). Similarly, exchanging carbohydrates for SFA or MUFA was not significantly related to diabetes risk. However, a 5 % higher energy contribution by carbohydrates at the expense of PUFA and protein was significantly associated with a lower risk of developing type 2 diabetes. The RR among all participants was 0·83 (95 % CI 0·70, 0·98) for 5 % energy at the expense of PUFA and 0·77 (95 % CI 0·64, 0·91) for 5 % energy at the expense of protein. These associations appeared to be similar for men and women, but reached statistical significance only among men for protein (RR 0·78 (95 % CI 0·61, 0·99)). When we restricted the analysis to participants with moderate to high carbohydrate intakes ( ≥  45 % energy) the inverse association for a carbohydrate-for-protein substitution remained statistically significant (RR 0·64 (95 % CI 0·43, 0·96)). While the association for the carbohydrate-for-PUFA substitution remained similar, it was not statistically significant anymore (RR 0·67 (95 % CI 0·43, 1·04)). Restricted cubic spline regression was used to examine non-linearity of these RR functions (Fig. 3). There was no indication for an association between a carbohydrate-for-fat substitution at any carbohydrate intake level (Fig. 3 (A)). In contrast, carbohydrate-for-protein (Fig. 3 (B)) and carbohydrate-for-PUFA (Fig. 3 (C)) substitutions were inversely related to diabetes risk with no indication for non-linearity (P = 0·353 and 0·349). The inverse associations between a carbohydrate-for-protein and a carbohydrate-for-PUFA substitution appeared to be slightly stronger at low carbohydrate intake levels among men (data not shown). However, tests for non-linearity were non-significant (P = 0·127 and 0·127). There was also no indication of non-linearity among women (P = 0·565 and 0·561). We further examined whether these associations remained similar in subgroup analyses based on BMI and the reported energy intake:BMR ratio. Associations appeared to be stronger among non-obese participants (data not shown). However, tests for interaction were non-significant. We also repeated the analyses using models without adjustment for total energy intake, BMI and waist circumference, but this had minimal impact on our observations. The RR was 0·82 (95 % CI 0·70, 0·97) for 5 % energy from carbohydrates at the expense of PUFA and 0·77 (95 % CI 0·65, 0·91) for 5 % energy from carbohydrates at the expense of protein in multivariate models not controlling for total energy intake. Using models not controlling for total energy intake, BMI, and waist circumference, the RR were 0·81 (95 % CI 0·69, 0·95) for 5 % energy carbohydrate-for-PUFA substitution and 0·61 (95 % CI 0·51, 0·72) for 5 % energy carbohydrate-for-protein substitution.

Fig. 2 Relative risk of type 2 diabetes for isoenergetic substitution of 5 % energy with carbohydrates for other macronutrients among all participants, men and women in the European Prospective Investigation into Cancer and Nutrition-Potsdam Study, 2005. The models were adjusted for age, education (in or no training, vocational training, technical school, or technical college or university degree), occupational activity (light, moderate, heavy), sport activity (0, 0·1–4·0, > 4·0 h/week), cycling (0, 0·1–2·4, 2·5–4·9, ≥  5 h/week), smoking (never, past, current <  20 cigarettes/d, current ≥  20 cigarettes/d), alcohol intake, total energy intake, fibre intake, and Mg intake. Models for total fat and protein intake were also adjusted for PUFA:SFA ratio, and MUFA:SFA ratio. Models for all study participants were also adjusted for sex. Vertical bars represent 95 % CI.

Fig. 3 Relative risk of type 2 diabetes for isoenergetic substitution of energy with carbohydrates for total fat (A), protein (B) and PUFA (C) among participants in the European Prospective Investigation into Cancer and Nutrition-Potsdam Study, 2005. Lines are restricted cubic splines showing the shape of dose–response curve according to carbohydrate intake on a continuous basis: relative risks (———); 95 % CI (–––). The reference point is 45 % energy from carbohydrates with knots placed at the 5th (31·1 % energy), 25th (37·6 % energy), 75th (46·3 % energy) and 95th percentile (52·5 % energy) of carbohydrate intake distribution. All models were adjusted for age, sex, education (in or no training, vocational training, technical school, or technical college or university degree), occupational activity (light, moderate, heavy), sport activity (0, 0·1–4·0, > 4·0 h/week), cycling (0, 0·1–2·4, 2·5–4·9, ≥  5 h/week), smoking (never, past, current <  20 cigarettes/d, current ≥  20 cigarettes/d), alcohol intake, total energy intake, fibre intake, Mg intake, PUFA:SFA ratio, and MUFA:SFA ratio.

The age-adjusted RR for the highest quintile compared with the lowest quintile of absolute intake (g/d) of carbohydrates among men was 0·75 (95 % CI 0·56, 1·00; P = 0·026 for trend; Table 3). This association was stronger controlling for lifestyle confounders (RR 0·53 (95 % CI 0·34, 0·85); P = 0·005 for trend), but was not significant after adjustment for anthropometric and diet characteristics (RR 0·67 (95 % CI 0·41, 1·08); P = 0·104 for trend). Similar associations were observed among women, but were not statistically significant (Table 3). We further evaluated whether different types of carbohydrates are related to diabetes risk. After adjustment for lifestyle confounders, anthropometry and diet characteristics, starch, sucrose, glucose and fructose were not significantly associated with diabetes risk in men or women (Table 4).

Table 3 Relative risk (RR) of type 2 diabetes by quintiles of total carbohydrate (g/d) intake among men and women in the European Prospective Investigation into Cancer and Nutrition-Potsdam Study, 2005

* Adjusted for age, education (in or no training, vocational training, technical school, or technical college or university degree), occupational activity (light, moderate, heavy), sport activity (0, 0·1–4·0, > 4·0 h/week), cycling (0, 0·1–2·4, 2·5–4·9, ≥  5 h/week), smoking (never, past, current <  20 cigarettes/d, current ≥  20 cigarettes/d), alcohol intake, and total energy intake.

Additionally adjusted for fibre intake, Mg intake, PUFA:SFA ratio, and MUFA:SFA ratio.

Table 4 Relative risk (RR) of type 2 diabetes by quintiles of carbohydrate intake among men and women in the European Prospective Investigation into Cancer and Nutrition-Potsdam Study, 2005

* Adjusted for age, sex, education (in or no training, vocational training, technical school, or technical college or university degree), occupational activity (light, moderate, heavy), sport activity (0, 0·1–4·0, > 4·0 h/week), cycling (0, 0·1–2·4, 2·5–4·9, ≥  5 h/week), smoking (never, past, current <  20 cigarettes/d, current ≥  20 cigarettes/d), alcohol intake, total energy intake, BMI, waist circumference, fibre intake, Mg intake, PUFA:SFA ratio, and MUFA:SFA ratio.

Discussion

We found that among men and women, a substitution of protein with carbohydrates was associated with a reduced risk of type 2 diabetes independently of age, anthropometry, alcohol consumption, physical activity and other lifestyle risk factors. Higher carbohydrate intake at the expense of total fat was not related to risk; however, substituting carbohydrates for PUFA was also associated with a lower diabetes risk.

Several previous cohort studies have evaluated whether carbohydrate intake is associated with the incidence of type 2 diabetes(Reference Salmeron, Ascherio, Rimm, Colditz, Spiegelman, Jenkins, Stampfer, Wing and Willett3Reference Schulz, Liese, Fang, Gilliard and Karter8). After multivariate adjustment, none of the previous studies has observed a significant association. However, in contrast to the present study, most previous studies did not evaluate specific macronutrient substitutions. In the Nurses' Health Study II, an isoenergetic substitution of carbohydrates for protein was evaluated, but in contrast to the present study, no significant association was observed(Reference Schulze, Liu, Rimm, Manson, Willett and Hu7). Our finding, that substituting carbohydrates for PUFA reduces diabetes risk is in contrast to findings from the Nurses' Health Study, where women in the highest quintile of PUFA intake were at lower risk of developing diabetes controlling for protein, alcohol, and other fatty acids, thus modelling a substitution of PUFA for carbohydrates(Reference Salmeron, Hu, Manson, Stampfer, Colditz, Rimm and Willett32). Although we also evaluated absolute carbohydrate intake, multivariate nutrient-density models may be particularly valuable if similar effects of an increment in intake for subjects with high and low energy intakes may not be plausible(Reference Willett, Lenart and Willett29). In addition, under isoenergetic settings assuming a steady state of energy balance, thus without changing the amount of energy consumed, it is impossible to change intake of one macronutrient without changing at least one other macronutrient. Macronutrient intake is therefore generally characterised by substitutions. As a consequence, observed associations cannot be interpreted as the effect of one single nutrient, but rather as a combination of two or more nutrients. It should be noted that energy excess and subsequent weight gain are important causes of type 2 diabetes. It could therefore be argued that adjustment for energy intake would represent over-control of a variable in the causal pathway. However, the multivariate nutrient density allows us to evaluate whether carbohydrate intake is related to diabetes risk independent of its contribution to energy intake per se. In this sense, the model is conceptually similar to metabolic studies, where, for example, changes in lipoproteins were evaluated when carbohydrates constituting a particular amount of dietary energy are replaced isoenergetically with fat or specific fatty acids(Reference Mensink, Zock, Kester and Katan12).

We did not observe a significant association between the intake of starch and diabetes risk. In the Women's Health Study(Reference Janket, Manson, Sesso, Buring and Liu33) and the Iowa Women's Study(Reference Meyer, Kushi, Jacobs, Slavin, Sellers and Folsom5), non-significant inverse associations between starch intake and diabetes risk were observed(Reference Janket, Manson, Sesso, Buring and Liu33), while starch intake was positively associated with risk in the Melbourne Collaborative Cohort Study after adjustment for BMI and waist:hip ratio(Reference Hodge, English, O'Dea and Giles6). Cohort studies on the role of starch intake in the development of diabetes are therefore inconclusive so far. We also found no significant association between sucrose intake and diabetes risk. Although previous studies are inconsistent, there is some suggestion that sucrose intake might be inversely associated with diabetes risk. No significant association between sucrose consumption and diabetes risk was observed in the Nurses' Health Study(Reference Colditz, Manson, Stampfer, Rosner, Willett and Speizer34). However, higher sucrose intake was inversely associated with risk in the Women's Health Study after excluding participants with a history of hypertension and hypercholesterolaemia(Reference Janket, Manson, Sesso, Buring and Liu33). Similarly, higher sucrose intake was significantly inversely associated with risk in the Iowa Women's Study(Reference Meyer, Kushi, Jacobs, Slavin, Sellers and Folsom5). Sugars were also inversely associated with diabetes risk in the Melbourne Collaborative Cohort Study after adjustment for obesity(Reference Hodge, English, O'Dea and Giles6), although the role of specific sugars remains unclear from this study. The finding of an inverse association between higher sucrose consumption and diabetes risk by some cohort studies is in contrast to experiments in human subjects, which have produced very conflicting results, and experimental research on animals, which has shown a clear and consistent effect of high-sucrose diets in decreasing insulin sensitivity(Reference Daly35). Glucose and fructose were not associated with risk in our cohort, similar to findings from the Women's Health Study(Reference Janket, Manson, Sesso, Buring and Liu33). However, significant positive associations with diabetes risk were observed for glucose and fructose in the Iowa Women's Study(Reference Meyer, Kushi, Jacobs, Slavin, Sellers and Folsom5). Similarly, frequent consumption of sugar-sweetened soft drinks containing high-fructose maize syrup and thus providing large amounts of fructose and glucose has been related to higher diabetes risk(Reference Schulze, Manson, Ludwig, Colditz, Stampfer, Willett and Hu36). Still, the inconsistency across studies may result from different effects from sugars consumed as added sugar v. those naturally occurring in foods.

Although protein acts as an insulin secretagogue(Reference Nuttall, Mooradian, Gannon, Billington and Krezowski15, Reference Simpson, McDonald, Wahlqvist, Atley and Outch16), and stimulation of insulin secretion counterbalances the increased gluconeogenesis due to postprandial amino acid elevations, the gluconeogenic effect of amino acids might be substantial in subjects with impaired insulin secretion(Reference Linn, Santosa, Gronemeyer, Aygen, Scholz, Busch and Bretzel37, Reference Krebs, Brehm and Krssak38). In addition, short-term amino acid infusion induces peripheral insulin resistance in healthy humans by inhibition of glucose transport and phosphorylation and subsequent reduction in glycogen synthesis(Reference Krebs, Krssak, Bernroider, Anderwald, Brehm, Meyerspeer, Nowotny, Roth, Waldhausl and Roden39). This has led to the notion that protein-rich diets contribute to the disturbance of carbohydrate metabolism(Reference Krebs40), although trials on low-carbohydrate high-protein diets have generally not shown such effects(Reference Bravata, Sanders, Huang, Krumholz, Olkin and Gardner41); regarding the effects of protein on HDL-cholesterol concentrations, data are not conclusive so far(Reference Katan42). Still, the observation that red meat consumption is related to higher risk of diabetes independent of dietary fat(Reference Schulze, Manson, Willett and Hu43, Reference Song, Manson, Buring and Liu44) highlights the possibility that protein-rich foods, particularly red meat, may adversely affect carbohydrate metabolism. It remains, however, unclear why the substitution of carbohydrates for PUFA was related to lower diabetes risk in the present study. Such substitution is expected to result in a decrease in HDL-cholesterol(Reference Mensink, Zock, Kester and Katan12) which should lead to higher diabetes risk(Reference Fagot-Campagna, Narayan, Hanson, Imperatore, Howard, Nelson, Pettitt and Knowler9Reference Fagot-Campagna, Knowler, Narayan, Hanson, Saaddine and Howard11). In addition, higher carbohydrate intake is inversely related to adiponectin concentrations(Reference Pischon, Girman, Rifai, Hotamisligil and Rimm45), and, in contrast, intake of n-3 PUFA may up regulate adiponectin secretion(Reference Neschen, Morino, Rossbacher, Pongratz, Cline, Sono, Gillum and Shulman13). Thus, substituting carbohydrates for PUFA would be expected to result in lower adiponectin concentrations, increased insulin resistance, and increased risk of diabetes(Reference Hara, Yamauchi and Kadowaki14). However, PUFA intake has not been consistently shown to be related to lower diabetes risk(Reference Salmeron, Hu, Manson, Stampfer, Colditz, Rimm and Willett32, Reference Meyer, Kushi, Jacobs and Folsom46, Reference van Dam, Willett, Rimm, Stampfer and Hu47) and long-chain n-3 PUFA even with an increased risk(Reference Meyer, Kushi, Jacobs and Folsom46). We were not able to distinguish between cis- and trans-PUFA which have very different physiological consequences(Reference Mozaffarian, Katan, Ascherio, Stampfer and Willett48). Thus, interpretation of our observations should be made cautiously.

All potential cases in the present study were verified through the treating physician. Given the resulting high positive predictive value of the disease classification, the remaining misclassification (non-identified cases) should not have biased the estimated risk(Reference Greenland, Rothman and Greenland49). However, we considered only clinically apparent type 2 diabetes and did not screen our study population for diabetes at baseline, thus it is possible that prevalent but undiagnosed cases of diabetes remained in our analyses. The glycaemic index, a measure of carbohydrate quality in terms of quantified glucose response to foods, is not available in the German Food Code and Nutrient Data Base version II.3(Reference Dehne, Klemm, Henseler and Hermann-Kunz21). Therefore, we were not able to address whether the association between carbohydrate intake and diabetes risk is independent or may be modified by the carbohydrate quality in terms of quantified glucose response to particular foods, for example, the glycaemic index and load. While we controlled in our analyses for fibre, Mg and fat quality, other nutrients related to carbohydrate intake may account for the associations. The potential of residual confounding due to unknown factors or incomplete statistical control applies to the present study as to observational studies in general. It is therefore possible that other factors may explain the observed associations or that they may mask potential associations. Although we calibrated FFQ data to the mean and variance of estimated habitual dietary intake from 24 h recalls, thus allowing quantitative analyses of the FFQ data, there is a significant degree of error in measuring dietary carbohydrate intake with our FFQ. In addition, it is possible that participants in the present study changed their dietary intake after the baseline measurement. Random measurement error and the lack of repeated measurement of diet may have led to an underestimation of association(Reference Hu, Stampfer, Rimm, Ascherio, Rosner, Spiegelman and Willett50). Furthermore, adjustment for energy intake, body weight and waist circumference may reflect adjustment for covariates being on the causal pathway between carbohydrate intake and diabetes risk, thus representing an over-adjustment. However, we are not able to clarify whether body fatness is a mediator or confounder in the present study. Adjustment allows us not only to control for a potential confounding effect of body fatness but also to control for the effect of selective underreporting of fat and energy intake which has been related to obesity in our study(Reference Voss, Kroke, Klipstein-Grobusch and Boeing51) and in other studies(Reference Livingstone and Black52). Still, our observations remained similar when we did not control for energy intake, BMI and waist circumference, supporting that substitutions of carbohydrates for PUFA or protein might have physiological consequences beyond excessive energy intake and weight gain. Although the present study involved a relatively large number of incident cases, it might have been underpowered to detect associations within specific subgroups. For example, although inverse associations were observed for carbohydrate-for-protein and carbohydrate-for-PUFA substitutions in the full cohort, only the carbohydrate-for-protein substitution among men gained significance in sex-specific analyses. Because spline regression did not indicate non-linear associations and the inverse association was observable across subgroups of sex, BMI, carbohydrate intake, and the reported energy intake:BMR ratio, the non-significance within subgroups might reflect the limited power of such analyses rather than the absence of these associations. Although we had considerable variation in carbohydrate intake in our cohort (5th percentile, 31·1 % energy; median, 42·1 % energy; 95th percentile, 52·5 % energy), we were not able to reliably estimate associations for macronutrient substitutions at very low or very high carbohydrate intakes. We did not evaluate foods or food groups that are rich in carbohydrates or rich in protein or PUFA and that might widely vary in their nutrient composition. Thus, translation of our findings into food-based dietary guidelines is limited.

In conclusion, a higher carbohydrate intake at the expense of protein and PUFA might be associated with decreased diabetes risk.

Acknowledgements

The recruitment phase of the EPIC-Potsdam Study was supported by the Federal Ministry of Science, Germany (01 EA 9401) and the European Union (SOC 95201408 05F02). The follow-up of the EPIC-Potsdam Study was supported by German Cancer Aid (70-2488-Ha I) and the European Community (SOC 98200769 05F02). M. B. S. was partly supported by the European Union (FP6-2005-513946). M. B. S. contributed to the development of the analysis plan, conducted the statistical analyses, collaborated on the interpretation of the results and wrote the manuscript. K. H. and H. B. provided significant consultation on the statistical analysis plan, interpretation of results and writing of the manuscript. M. S., C. H. and A. S. provided significant consultation on the interpretation of results and writing of the manuscript. None of the authors had any financial or personal interest in any company or organisation sponsoring this research, including advisory board affiliations.

Kurt Hoffmann died during the review process of the manuscript and the authors would like to dedicate this paper to our colleague and friend.

References

1Hu, FB, Manson, JE, Stampfer, MJ, Colditz, G, Liu, S, Solomon, CG & Willett, WC (2001) Diet, lifestyle, and the risk of type 2 diabetes mellitus in women. N Engl J Med 345, 790797.CrossRefGoogle ScholarPubMed
2Schulze, MB & Hu, FB (2005) Primary prevention of diabetes: what can be done and how much can be prevented? Annu Rev Public Health 26, 445467.CrossRefGoogle ScholarPubMed
3Salmeron, J, Ascherio, A, Rimm, EB, Colditz, GA, Spiegelman, D, Jenkins, DJ, Stampfer, MJ, Wing, AL & Willett, WC (1997) Dietary fiber, glycemic load, and risk of NIDDM in men. Diabetes Care 20, 545550.CrossRefGoogle ScholarPubMed
4Salmeron, J, Manson, JE, Stampfer, MJ, Colditz, GA, Wing, AL & Willett, WC (1997) Dietary fiber, glycemic load, and risk of non-insulin-dependent diabetes mellitus in women. JAMA 277, 472477.CrossRefGoogle ScholarPubMed
5Meyer, KA, Kushi, LH, Jacobs, DR Jr, Slavin, J, Sellers, TA & Folsom, AR (2000) Carbohydrates, dietary fiber, and incident type 2 diabetes in older women. Am J Clin Nutr 71, 921930.CrossRefGoogle ScholarPubMed
6Hodge, AM, English, DR, O'Dea, K & Giles, GG (2004) Glycemic index and dietary fiber and the risk of type 2 diabetes. Diabetes Care 27, 27012706.CrossRefGoogle ScholarPubMed
7Schulze, MB, Liu, S, Rimm, EB, Manson, JE, Willett, WC & Hu, FB (2004) Glycemic index, glycemic load, and dietary fiber intake and incidence of type 2 diabetes in younger and middle-aged women. Am J Clin Nutr 80, 348356.CrossRefGoogle ScholarPubMed
8Schulz, M, Liese, AD, Fang, F, Gilliard, TS & Karter, AJ (2006) Is the association between dietary glycemic index and type 2 diabetes modified by waist circumference? Diabetes Care 29, 11021104.CrossRefGoogle ScholarPubMed
9Fagot-Campagna, A, Narayan, KM, Hanson, RL, Imperatore, G, Howard, BV, Nelson, RG, Pettitt, DJ & Knowler, WC (1997) Plasma lipoproteins and incidence of non-insulin-dependent diabetes mellitus in Pima Indians: protective effect of HDL cholesterol in women. Atherosclerosis 128, 113119.CrossRefGoogle ScholarPubMed
10Njolstad, I, Arnesen, E & Lund-Larsen, PG (1998) Sex differences in risk factors for clinical diabetes mellitus in a general population: a 12-year follow-up of the Finnmark Study. Am J Epidemiol 147, 4958.CrossRefGoogle Scholar
11Fagot-Campagna, A, Knowler, WC, Narayan, KM, Hanson, RL, Saaddine, J & Howard, BV (1999) HDL cholesterol subfractions and risk of developing type 2 diabetes among Pima Indians. Diabetes Care 22, 271274.CrossRefGoogle ScholarPubMed
12Mensink, RP, Zock, PL, Kester, AD & Katan, MB (2003) Effects of dietary fatty acids and carbohydrates on the ratio of serum total to HDL cholesterol and on serum lipids and apolipoproteins: a meta-analysis of 60 controlled trials. Am J Clin Nutr 77, 11461155.CrossRefGoogle ScholarPubMed
13Neschen, S, Morino, K, Rossbacher, JC, Pongratz, RL, Cline, GW, Sono, S, Gillum, M & Shulman, GI (2006) Fish oil regulates adiponectin secretion by a peroxisome proliferator-activated receptor-{γ}-dependent mechanism in mice. Diabetes 55, 924928.CrossRefGoogle ScholarPubMed
14Hara, K, Yamauchi, T & Kadowaki, T (2005) Adiponectin: an adipokine linking adipocytes and type 2 diabetes in humans. Curr Diab Rep 5, 136140.CrossRefGoogle ScholarPubMed
15Nuttall, FQ, Mooradian, AD, Gannon, MC, Billington, C & Krezowski, P (1984) Effect of protein ingestion on the glucose and insulin response to a standardized oral glucose load. Diabetes Care 7, 465470.CrossRefGoogle ScholarPubMed
16Simpson, RW, McDonald, J, Wahlqvist, ML, Atley, L & Outch, K (1985) Macronutrients have different metabolic effects in nondiabetics and diabetics. Am J Clin Nutr 42, 449453.CrossRefGoogle ScholarPubMed
17Hu, FB, Stampfer, MJ, Manson, JE, Rimm, E, Colditz, GA, Rosner, BA, Hennekens, CH & Willett, WC (1997) Dietary fat intake and the risk of coronary heart disease in women. N Engl J Med 337, 14911499.CrossRefGoogle ScholarPubMed
18Faerch, K, Lau, C, Tetens, I, Pedersen, OB, Jorgensen, T, Borch-Johnsen, K & Glumer, C (2005) A statistical approach based on substitution of macronutrients provides additional information to models analyzing single dietary factors in relation to type 2 diabetes in Danish adults: the Inter99 study. J Nutr 135, 11771182.CrossRefGoogle ScholarPubMed
19Boeing, H, Wahrendorf, J & Becker, N (1999) EPIC-Germany - a source for studies into diet and risk of chronic diseases. Ann Nutr Metab 43, 195204.CrossRefGoogle ScholarPubMed
20Boeing, H, Korfmann, A & Bergmann, MM (1999) Recruitment procedures of EPIC-Germany. European Investigation into Cancer and Nutrition. Ann Nutr Metab 43, 205215.CrossRefGoogle ScholarPubMed
21Dehne, LI, Klemm, C, Henseler, G & Hermann-Kunz, E (1999) The German Food Code and Nutrient Data Base (BLS II.2). Eur J Epidemiol 15, 355359.CrossRefGoogle ScholarPubMed
22Slimani, N, Ferrari, P, Ocke, M, et al. (2000) Standardization of the 24-hour diet recall calibration method used in the European Prospective Investigation into Cancer and Nutrition (EPIC): general concepts and preliminary results. Eur J Clin Nutr 54, 900917.CrossRefGoogle ScholarPubMed
23Kynast-Wolf, G, Becker, N, Kroke, A, Brandstetter, BR, Wahrendorf, J & Boeing, H (2002) Linear regression calibration: theoretical framework and empirical results in EPIC, Germany. Ann Nutr Metab 46, 28.CrossRefGoogle ScholarPubMed
24Willett, WC, Howe, GR & Kushi, LH (1997) Adjustment for total energy intake in epidemiologic studies. Am J Clin Nutr 65, Suppl. 4, 1220S1228S.CrossRefGoogle ScholarPubMed
25Kroke, A, Klipstein-Grobusch, K, Voss, S, Moseneder, J, Thielecke, F, Noack, R & Boeing, H (1999) Validation of a self-administered food-frequency questionnaire administered in the European Prospective Investigation into Cancer and Nutrition (EPIC) Study: comparison of energy, protein, and macronutrient intakes estimated with the doubly labeled water, urinary nitrogen, and repeated 24-h dietary recall methods. Am J Clin Nutr 70, 439447.CrossRefGoogle Scholar
26Bohlscheid-Thomas, S, Hoting, I, Boeing, H & Wahrendorf, J (1997) Reproducibility and relative validity of energy and macronutrient intake of a food frequency questionnaire developed for the German part of the EPIC project. European Prospective Investigation into Cancer and Nutrition. Int J Epidemiol 26, Suppl. 1, S71S81.CrossRefGoogle ScholarPubMed
27Kroke, A, Bergmann, MM, Lotze, G, Jeckel, A, Klipstein-Grobusch, K & Boeing, H (1999) Measures of quality control in the German component of the EPIC study. European Prospective Investigation into Cancer and Nutrition. Ann Nutr Metab 43, 216224.CrossRefGoogle Scholar
28Klipstein-Grobusch, K, Georg, T & Boeing, H (1997) Interviewer variability in anthropometric measurements and estimates of body composition. Int J Epidemiol 26, Suppl. 1, S174S180.CrossRefGoogle ScholarPubMed
29Willett, W & Lenart, E (1998) Implications of total energy intake for epidemiologic analyses. In Nutritional Epidemiology, pp. 273301 [Willett, W, editor]. New York: Oxford University Press.CrossRefGoogle Scholar
30Harrell, FE Jr, Lee, KL & Pollock, BG (1988) Regression models in clinical studies: determining relationships between predictors and response. J Natl Cancer Inst 80, 11981202.CrossRefGoogle ScholarPubMed
31Heinzl, H & Kaider, A (1997) Gaining more flexibility in Cox proportional hazards regression models with cubic spline functions. Comput Methods Programs Biomed 54, 201208.CrossRefGoogle ScholarPubMed
32Salmeron, J, Hu, FB, Manson, JE, Stampfer, MJ, Colditz, GA, Rimm, EB & Willett, WC (2001) Dietary fat intake and risk of type 2 diabetes in women. Am J Clin Nutr 73, 10191026.CrossRefGoogle ScholarPubMed
33Janket, SJ, Manson, JE, Sesso, H, Buring, JE & Liu, S (2003) A prospective study of sugar intake and risk of type 2 diabetes in women. Diabetes Care 26, 10081015.CrossRefGoogle ScholarPubMed
34Colditz, GA, Manson, JE, Stampfer, MJ, Rosner, B, Willett, WC & Speizer, FE (1992) Diet and risk of clinical diabetes in women. Am J Clin Nutr 55, 10181023.CrossRefGoogle ScholarPubMed
35Daly, M (2003) Sugars, insulin sensitivity, and the postprandial state. Am J Clin Nutr 78, 865S872S.CrossRefGoogle ScholarPubMed
36Schulze, MB, Manson, JE, Ludwig, DS, Colditz, GA, Stampfer, MJ, Willett, WC & Hu, FB (2004) Sugar-sweetened beverages, weight gain, and incidence of type 2 diabetes in young and middle-aged women. JAMA 292, 927934.CrossRefGoogle Scholar
37Linn, T, Santosa, B, Gronemeyer, D, Aygen, S, Scholz, N, Busch, M & Bretzel, RG (2000) Effect of long-term dietary protein intake on glucose metabolism in humans. Diabetologia 43, 12571265.CrossRefGoogle ScholarPubMed
38Krebs, M, Brehm, A, Krssak, M, et al. (2003) Direct and indirect effects of amino acids on hepatic glucose metabolism in humans. Diabetologia 46, 917925.CrossRefGoogle ScholarPubMed
39Krebs, M, Krssak, M, Bernroider, E, Anderwald, C, Brehm, A, Meyerspeer, M, Nowotny, P, Roth, E, Waldhausl, W & Roden, M (2002) Mechanism of amino acid-induced skeletal muscle insulin resistance in humans. Diabetes 51, 599605.CrossRefGoogle ScholarPubMed
40Krebs, M (2005) Amino acid-dependent modulation of glucose metabolism in humans. Eur J Clin Invest 35, 351354.CrossRefGoogle ScholarPubMed
41Bravata, DM, Sanders, L, Huang, J, Krumholz, HM, Olkin, I & Gardner, CD (2003) Efficacy and safety of low-carbohydrate diets: a systematic review. JAMA 289, 18371850.CrossRefGoogle ScholarPubMed
42Katan, MB (2006) Alternatives to low-fat diets. Am J Clin Nutr 83, 989990.CrossRefGoogle ScholarPubMed
43Schulze, MB, Manson, JE, Willett, WC & Hu, FB (2003) Processed meat intake and incidence of type 2 diabetes in younger and middle-aged women. Diabetologia 46, 14651473.CrossRefGoogle ScholarPubMed
44Song, Y, Manson, JE, Buring, JE & Liu, S (2004) A prospective study of red meat consumption and type 2 diabetes in middle-aged and elderly women: The Women's Health Study. Diabetes Care 27, 21082115.CrossRefGoogle ScholarPubMed
45Pischon, T, Girman, CJ, Rifai, N, Hotamisligil, GS & Rimm, EB (2005) Association between dietary factors and plasma adiponectin concentrations in men. Am J Clin Nutr 81, 780786.CrossRefGoogle ScholarPubMed
46Meyer, KA, Kushi, LH, Jacobs, DR Jr & Folsom, AR (2001) Dietary fat and incidence of type 2 diabetes in older Iowa women. Diabetes Care 24, 15281535.CrossRefGoogle ScholarPubMed
47van Dam, RM, Willett, WC, Rimm, EB, Stampfer, MJ & Hu, FB (2002) Dietary fat and meat intake in relation to risk of type 2 diabetes in men. Diabetes Care 25, 417424.CrossRefGoogle ScholarPubMed
48Mozaffarian, D, Katan, MB, Ascherio, A, Stampfer, MJ & Willett, WC (2006) Trans fatty acids and cardiovascular disease. N Engl J Med 354, 16011613.CrossRefGoogle ScholarPubMed
49Greenland, S (1998) Basic methods for sensitivity analysis and external adjustment. In Modern Epidemiology, 2nd ed. pp. 343358 [Rothman, JR and Greenland, S, editors]. Philadelphia, PA: Lippincott-Raven.Google Scholar
50Hu, FB, Stampfer, MJ, Rimm, E, Ascherio, A, Rosner, BA, Spiegelman, D & Willett, WC (1999) Dietary fat and coronary heart disease: a comparison of approaches for adjusting for total energy intake and modeling repeated dietary measurements. Am J Epidemiol 149, 531540.CrossRefGoogle ScholarPubMed
51Voss, S, Kroke, A, Klipstein-Grobusch, K & Boeing, H (1997) Obesity as a major determinant of underreporting in a self-administered food frequency questionnaire: results from the EPIC-Potsdam Study. Z Ernahrungswiss 36, 229236.CrossRefGoogle Scholar
52Livingstone, MB & Black, AE (2003) Markers of the validity of reported energy intake. J Nutr 133, Suppl. 3, 895S920S.CrossRefGoogle ScholarPubMed
Figure 0

Table 1 Baseline characteristics by quintiles of carbohydrate intake (% energy) among men and women in the European Prospective Investigation into Cancer and Nutrition-Potsdam Study, 2005 (Mean values or percentages)

Figure 1

Fig. 1 Incidence rate (per 10 000 person-years) of type 2 diabetes mellitus among men () and women (■) for different age groups in the European Prospective Investigation into Cancer and Nutrition-Potsdam Study, 2005.

Figure 2

Table 2 Relative risk (RR) of type 2 diabetes by quintiles of total carbohydrate intake (% energy) among men and women in the European Prospective Investigation into Cancer and Nutrition-Potsdam Study, 2005

Figure 3

Fig. 2 Relative risk of type 2 diabetes for isoenergetic substitution of 5 % energy with carbohydrates for other macronutrients among all participants, men and women in the European Prospective Investigation into Cancer and Nutrition-Potsdam Study, 2005. The models were adjusted for age, education (in or no training, vocational training, technical school, or technical college or university degree), occupational activity (light, moderate, heavy), sport activity (0, 0·1–4·0, > 4·0 h/week), cycling (0, 0·1–2·4, 2·5–4·9, ≥  5 h/week), smoking (never, past, current <  20 cigarettes/d, current ≥  20 cigarettes/d), alcohol intake, total energy intake, fibre intake, and Mg intake. Models for total fat and protein intake were also adjusted for PUFA:SFA ratio, and MUFA:SFA ratio. Models for all study participants were also adjusted for sex. Vertical bars represent 95 % CI.

Figure 4

Fig. 3 Relative risk of type 2 diabetes for isoenergetic substitution of energy with carbohydrates for total fat (A), protein (B) and PUFA (C) among participants in the European Prospective Investigation into Cancer and Nutrition-Potsdam Study, 2005. Lines are restricted cubic splines showing the shape of dose–response curve according to carbohydrate intake on a continuous basis: relative risks (———); 95 % CI (–––). The reference point is 45 % energy from carbohydrates with knots placed at the 5th (31·1 % energy), 25th (37·6 % energy), 75th (46·3 % energy) and 95th percentile (52·5 % energy) of carbohydrate intake distribution. All models were adjusted for age, sex, education (in or no training, vocational training, technical school, or technical college or university degree), occupational activity (light, moderate, heavy), sport activity (0, 0·1–4·0, > 4·0 h/week), cycling (0, 0·1–2·4, 2·5–4·9, ≥  5 h/week), smoking (never, past, current <  20 cigarettes/d, current ≥  20 cigarettes/d), alcohol intake, total energy intake, fibre intake, Mg intake, PUFA:SFA ratio, and MUFA:SFA ratio.

Figure 5

Table 3 Relative risk (RR) of type 2 diabetes by quintiles of total carbohydrate (g/d) intake among men and women in the European Prospective Investigation into Cancer and Nutrition-Potsdam Study, 2005

Figure 6

Table 4 Relative risk (RR) of type 2 diabetes by quintiles of carbohydrate intake among men and women in the European Prospective Investigation into Cancer and Nutrition-Potsdam Study, 2005