Eur J Epidemiol (2013) 28:845–858 DOI 10.1007/s10654-013-9852-5

REVIEW

Whole grain and refined grain consumption and the risk of type 2 diabetes: a systematic review and dose–response meta-analysis of cohort studies Dagfinn Aune • Teresa Norat • Pa˚l Romundstad Lars J. Vatten

•

Received: 6 February 2013 / Accepted: 16 September 2013 / Published online: 25 October 2013 Ó Springer Science+Business Media Dordrecht 2013

Abstract Several studies have suggested a protective effect of intake of whole grains, but not refined grains on type 2 diabetes risk, but the dose–response relationship between different types of grains and type 2 diabetes has not been established. We conducted a systematic review and meta-analysis of prospective studies of grain intake and type 2 diabetes. We searched the PubMed database for studies of grain intake and risk of type 2 diabetes, up to June 5th, 2013. Summary relative risks were calculated using a random effects model. Sixteen cohort studies were included in the analyses. The summary relative risk per 3 servings per day was 0.68 (95 % CI 0.58–0.81, I2 = 82 %, n = 10) for whole grains and 0.95 (95 % CI 0.88–1.04, I2 = 53 %, n = 6) for refined grains. A nonlinear association was observed for whole grains, pnonlinearity \ 0.0001, but not for refined grains, pnonlinearity = 0.10. Inverse associations were observed for subtypes of whole grains including whole grain bread, whole grain cereals, wheat bran and brown rice, but these results were based on few studies, while white rice was associated with increased risk. Our meta-analysis suggests that a high whole grain intake, but not refined grains, is associated with reduced

Electronic supplementary material The online version of this article (doi:10.1007/s10654-013-9852-5) contains supplementary material, which is available to authorized users. D. Aune
P. Romundstad
L. J. Vatten Department of Public Health and General Practice, Faculty of Medicine, Norwegian University of Science and Technology, Trondheim, Norway D. Aune (&)
T. Norat Department of Epidemiology and Biostatistics, School of Public Health, Imperial College London, St. Mary’s Campus, Norfolk Place, Paddington, London W2 1PG, UK e-mail: [email protected]

type 2 diabetes risk. However, a positive association with intake of white rice and inverse associations between several specific types of whole grains and type 2 diabetes warrant further investigations. Our results support public health recommendations to replace refined grains with whole grains and suggest that at least two servings of whole grains per day should be consumed to reduce type 2 diabetes risk. Keywords Whole grains
Refined grains
Cereals
Type 2 diabetes
Meta-analysis

Introduction The prevalence of diabetes type 2 is rapidly increasing worldwide, with an estimated 311 million persons living with diabetes in 2011 and this number is expected to increase to 552 million by 2030 [1]. Diabetes patients have increased risk cardiovascular disease, some cancers, eye and kidney disease [2]. Total medical costs of diabetes were estimated at US$245 billion in 2012 in the US [3]. Changes in body weight and physical activity are likely to contribute to these increased rates [4], but diet may also influence diabetes risk, directly and indirectly through an effect on obesity. Whole grains contain endosperm, germ, and bran, in contrast to refined grains which have the germ and bran removed during the milling process. Whole grains have been hypothesized to reduce the risk of type 2 diabetes based on their content of fiber, vitamins and minerals and phytochemicals which may improve insulin sensitivity and glucose metabolism, and by reducing overweight and obesity [5]. In contrast, refined grains may increase risk because of their high glycemic index or glycemic load and reduced fiber and nutrient content. Several studies of whole

123

846

grain intake in relation to type 2 diabetes risk have reported inverse associations with higher intake [5–10], but some found no significant association [11, 12]. Inverse associations have been reported with intake of specific whole grain products as well, including brown bread [13–15], whole grain breakfast cereals [13, 16] and brown rice [7], although the results are not entirely consistent [17, 18]. In contrast, most studies of refined grain intake have shown no association overall [5, 12, 13, 19], although two suggested inverse associations [8, 10], while high intake of white bread [17] or white rice [7, 20, 21] has been associated with increased risk, although not consistently [17, 22]. Although two previous meta-analyses have been conducted on whole grains and type 2 diabetes [23, 24], the optimal intake of whole grains for prevention of type 2 diabetes is not established because the shape of the dose– response relationship has not been investigated. In addition, there is increasing evidence suggesting that whole grains reduces the risk of overweight and obesity and weight gain [24–30], thus it is possible that body mass index may be an intermediate factor more than a confounder, but it is not known how much of the association that may be explained by reduced body fatness. We conducted a systematic review and meta-analysis of the evidence from prospective studies with the aim of clarifying (1) the association between the intake of grains and different types of grains and type 2 diabetes risk, (2) the dose– response relationship between intake of grains and specific types of grains and type 2 diabetes risk, and (3) how much of the association that may be explained by reduced body fatness.

Methods Search strategy We conducted a comprehensive search in the PubMed database up to June 5th, 2013 for studies of various food groups and type 2 diabetes risk. The search terms relevant to this analysis included ‘‘cereal OR breakfast cereal OR grain OR whole grain OR rice OR bread’’ AND ‘‘diabetes’’. The full search is provided in the Supplementary Appendix. We also searched the reference lists of all the studies that were included in the analysis and the reference lists of published meta-analyses [23, 24]. Study selection To be included, the study had to have a prospective design and to investigate the association between the intake of grains and type 2 diabetes risk. Estimates of the relative risk (hazard ratio, risk ratio) had to be available with the

123

D. Aune et al.

95 % confidence intervals in the publication and for the dose–response analysis, a quantitative measure of intake and the total number of cases and person-years had to be available in the publication. We identified 28 publications that reported on intake of grains in relation to diabetes [5–23, 31–39]. Three publications were excluded because no risk estimates were provided [35, 36, 39], two publications were excluded because they were cross-sectional studies [37, 38] and four because they were duplicates [31–34]. One publication [23] was included only in the sensitivity analysis with and without adjustment for BMI because the most recent publication [7] from these two studies did not provide results both adjusted and unadjusted for BMI. In addition several publications from the same studies reported on different grain items and all were included in the analyses, but each study was only included once in the analysis of the relevant grain variable. Data extraction We extracted the following data from each study: The first author’s last name, publication year, country where the study was conducted, the study name, follow-up period, sample size, gender, age, number of cases, dietary assessment method (type, number of food items and whether it had been validated), exposure, quantity of intake, RRs and 95 % CIs for the highest versus the lowest grain intake and variables adjusted for in the analysis. Statistical methods To take into account within and between studies heterogeneity we used random effects models to estimate summary RRs and 95 % CIs for the highest versus the lowest level of grain intake and for the dose–response analysis [40]. The average of the natural logarithm of the RRs was estimated and the RR from each study was weighted by the inverse of its variance. A two-tailed p \ 0.05 was considered statistically significant. We used the method described by Greenland and Longnecker [41] for the dose–response analysis and computed study-specific slopes (linear trends) and 95 % CIs from the natural logs of the RRs and CIs across categories of grain intake. The method requires that the distribution of cases and person-years or non-cases and the RRs with the variance estimates for at least three quantitative exposure categories are known. We estimated the distribution of cases or person-years in studies that did not report these, but reported the total number of cases/person-years [42]. The median or mean level of grain intake in each category of intake was assigned to the corresponding relative risk for each study. For studies that reported grain intake by ranges of intake we estimated the midpoint for each category by

Whole grain and refined grain consumption

calculating the average of the lower and upper bound. When the highest or lowest category was open-ended we assumed the open-ended interval length to be the same as the adjacent interval. In studies that reported the intakes in grams per day we used 30 g as a serving size for recalculation of the intakes to a common scale (servings per day) [43]. We used 158 g as a serving size for intake of white rice and brown rice consistent with a recent study [44]. The dose–response results in the forest plots are presented for a 3 serving per day increment [43]. We examined a potential nonlinear dose–response relationship between grain intake and type 2 diabetes by using fractional polynomial models [45]. We determined the best fitting second order fractional polynomial regression model, defined as the one with the lowest deviance. A likelihood ratio test was used to assess the difference between the nonlinear and linear models to test for nonlinearity [46]. The intake in the reference category was subtracted from the intake in each category for the linear dose–response analysis, but not for the nonlinear dose– response analysis. Heterogeneity between studies was assessed by the Q test and I2 [47]. I2 is the amount of total variation that is explained by between study variation. I2 values of approximately 25, 50 and 75 % are considered to indicate low, moderate and high heterogeneity, respectively. Publication bias was assessed with Egger’s test [48] and Begg’s test [49] with the results considered to indicate publication bias when p \ 0.10. We conducted sensitivity analyses excluding one study at a time to ensure that the results were not simply due to one large study or a study with an extreme result, when there were at least 5 studies in the analysis. The statistical analyses were conducted using Stata, version 10.1 software (StataCorp, College Station, TX, USA).

Results We identified sixteen cohort studies (nineteen publications) that were included in the analyses of grain intake and type 2 diabetes risk [5–23] (Table 1; Fig. 1). Seven studies were from the US, six were from Europe, two from Asia and one was from Australia (Table 1). Whole grains Ten cohort studies (8 publications) [5–12] were included in the analysis of total whole grain intake and type 2 diabetes risk and included 19,829 cases among 385,868 participants. One of the studies only reported a continuous result and was not included in the high versus low analysis [11]. The summary RR for high versus low intake was 0.74 (95 % CI

847

Fig. 1 Flow-chart of study selection

0.71–0.78, I2 = 0 %, pheterogeneity = 0.43) (Supplementary Figure 1). The summary RR per 3 servings per day was 0.68 (95 % CI 0.58–0.81, I2 = 82 %, pheterogeneity \ 0.0001) (Fig. 2a). The summary RR ranged from 0.65 (95 % CI 0.56–0.77) when excluding the EPIC-Potsdam study to 0.72 (95 % CI 0.63–0.83) when excluding the Nurses’ Health Study 1. There was no evidence of small study bias with Egger’s test, p = 0.49 or with Begg’s test, p = 0.37. There was evidence of a nonlinear association between whole grain intake and type 2 diabetes risk, pnonlinearity \ 0.0001, with a steeper reduction in risk when increasing intake from low levels and most of the benefit was observed up to an intake of two servings per day (Fig. 2b, Supplementary Table 1). Refined grains Six studies [5, 8, 12, 13, 19] reported on refined grain intake and type 2 diabetes and included 9,545 cases among 258,078 participants. The summary RR for high versus low intake of refined grains was 0.94 (95 % CI 0.82–1.09, I2 = 64 %, pheterogeneity = 0.02) (Supplementary Figure 2). The summary RR per 3 servings per day was 0.95 (95 % CI 0.88–1.04, I2 = 53 %, pheterogeneity = 0.06) (Fig. 3a). The summary RR ranged from 0.93 (95 % CI 0.86–1.00) when the Nurses’ Health Study 1 was excluded to 0.98 (95 % CI 0.90–1.08) when the Women’s Health Initiative was excluded. There was no evidence of small study bias with

123

Study name

Malmo Diet and Cancer Cohort

Women’s Health Initiative Observational Study

European Prospective Investigation into Cancer and Nutrition– Potsdam study

NA

The Pizarra Study

Health Professionals Follow-up Study

Author, publication year [Ref. no.], country

Ericson et al 2013 [12], Sweden

123

Parker et al 2013 [10], USA

Von Ruesten et al 2013 [18], Germany

Wirstro¨m et al 2013 [9], Sweden

Soriguer et al 2013 [22], Spain

Sun et al 2010 [7], USA

1986–2006, 20 years followup

1997/ 1998–2003/ 2004, 6 years follow-up

1992/1998–NA, 8–10 years follow-up

1994/1998–NA, 8 years follow-up

1993/ 1998–2005, 7.9 years follow-up

1991/ 1996–2006, 12 years follow-up

Follow-up period

39,765 m, age 32–87 years: 2,648 cases

605 m & w, age 18–65 years: 54 cases

5,477 m & w, age 35–56 years: 165 cases

23,531 m & w, age 35–65 years: 837 cases

72,215 w, age 50–79 years: 3,465 cases

27,140 m & w, age 45–74 years: 1,709 cases

Study size, gender, age, number of cases

Table 1 Cohort studies of grain intake and type 2 diabetes risk

Validated FFQ, 131 food items

Validated FFQ, NA

Validated FFQ, NA

Validated FFQ, 148 food items

Validated FFQ, 122 food items

Validated diet history, FFQ 168 food items, interview

Dietary assessment

0.88 (0.74–1.04)

Per 30 g/day

C5/week versus \1/month C2/week versus \1/month 47.1 versus 5.1 g/day 14.3 versus 0.6 g/day 2.3 versus 0.2 g/day

Brown rice Whole grain Bran Germ

1.04 (0.89–1.21)

0.69 (0.60–0.81)

0.72 (0.63–0.83)

0.96 (0.82–1.12)

1.02 (0.77–1.34)

0.43 (0.19–0.95)

0.71 (0.48–1.04)

[59.1 versus \30.6 g/day

2–3/week versus B1/week

0.92 (0.82–1.03)

Per 50 g/day

White rice

White rice

Whole grains

Whole grain bread

0.73 (0.58–0.93)

C6.0 versus \1.0 serv/day

1.02 (0.82–1.26)

Refined grains

4.6 versus 1.1 portions/day

Refined cereals, m

0.84 (0.68–1.04)

0.79 (0.66–0.94)

2.3 versus 0.01 portions/day

Fibre-rich bread and cereals, m

1.07 (0.87–1.32)

C2.0 versus 0 serv/day

2.9 versus 0.7 portions/day

Refined cereals, w

0.85 (0.68–1.06)

RR (95% CI)

Whole grains

2.0 versus 0.1 portions/day

Quantity

Fibre-rich bread and cereals, w

Exposure

Age, ethnicity, BMI, FH–DM, smoking status, cigarettes per day, alcohol, multivitamins, physical activity, total energy, red meat, fruits and vegetables, white rice or brown rice in the respective analyses

Age, sex, BMI, abnormal glucose regulation

Age, sex, FH–DM, BMI, leisure-time physical activity, smoking, education, blood pressure

Age, sex, smoking status, pack-years of smoking, alcohol, leisure-time physical activity, BMI, WHR, prevalent hypertension, high blood lipid levels, education, vitamin supplementation, nonconsumption of the food group, total energy, other food groups

Age, energy intake, race/ethnicity, physical activity, smoking status, pack-years of cigarettes, alcohol, HRT, education, income, FH–DM, BMI, dairy, fruit, vegetables

Age, dietary method, season, total energy, education, smoking, alcohol, leisure time physical activity, BMI

Adjustment for confounders

848 D. Aune et al.

Study name

Nurses’ Health Study 1

Nurses’ Health Study 2

Japan Public Health Center-Based Prospective Study

European Prospective Investigation into Cancer and Nutrition– Potsdam study

Author, publication year [Ref. no.], country

Sun et al 2010 [7], USA

Sun et al 2010 [7], USA

Nanri et al 2010 [21], Japan

Fisher et al 2009 [11], Germany

Table 1 continued

1994/ 1998–2005, 7.1 years follow-up

Cohort 2: 1998–2003, 5 years followup

1995–2000

Cohort 1:

1991–2005, 14 years followup

1984–2006, 22 years followup

Follow-up period

2,318 m & w, age 35–65 years: 724 cases

25,666 m & 33,622 w, age 45–75 years: 1,103 cases

88,343 w, age 26–45 years: 2,359 cases

69,120 w, age 37–65 years: 5,500 cases

Study size, gender, age, number of cases

Validated FFQ, 148 food items

Validated FFQ, 147 food items

Validated FFQ, 131 food items

Validated FFQ, 116 food items

Dietary assessment

176.9 versus 29.0 g/day

Noodles

Per 50 g/day

60 versus 4 g/day

Bread

Whole grains, rs7903146 CT ? TT genotype

560 versus 165 g/day

Rice, w

Per 50 g/day

225 versus 41.3 g/day

Noodles

Whole grains, rs7903146 CC genotype

47.1 versus 0 g/day

Bread

2.0 versus 0.3 g/ day

Germ

700 versus 280 g/day

12.1 versus 1.0 g/ day

Bran

Rice, m

40.0 versus 6.2 g/ day

Whole grain

1.5 versus 0.2 g/day

Germ

C2/week versus \1/month

9.5 versus 0.6 g/day

Bran

Brown rice

31.3 versus 3.6 g/day

Whole grain

C5/week versus \1/month

C2/week versus \1/month

Brown rice

White rice

C5/week versus \1/month

Quantity

White rice

Exposure

1.08 (0.96–1.23)

0.86 (0.75–0.99)

1.15 (0.83–1.58)

0.99 (0.73–1.34)

1.65 (1.06–2.57)

0.89 (0.68–1.17)

0.85 (0.64–1.14)

1.19 (0.85–1.68)

1.04 (0.90–1.21)

0.83 (0.71–0.97)

0.81 (0.70–0.94)

0.89 (0.75–1.07)

1.40 (1.09–1.80)

0.88 (0.79–0.97)

0.77 (0.69–0.86)

0.70 (0.64–0.77)

0.83 (0.72–0.96)

1.11 (0.87–1.43)

RR (95% CI)

Age, sex, BMI, waist circumference, education, occupational activity, sports, smoking, alcohol, red meat, processed meat, low-fat dairy, butter, margarine, vegetable fat, total energy

Age, study area, smoking status and cigarettes per day, alcohol, FH–DM, total physical activity, hypertension, occupation, total energy intake, coffee, calcium, magnesium, fruit, vegetables, fish, BMI

Age, ethnicity, BMI, FH–DM, smoking status, cigarettes per day, alcohol, multivitamins, physical activity, menopausal status, hormone use, OC use, total energy, red meat, fruits and vegetables, white rice or brown rice in the respective analyses

Age, ethnicity, BMI, FH–DM, smoking status, cigarettes per day, alcohol, multivitamins, physical activity, menopausal status, hormone use, OC use, total energy, red meat, fruits and vegetables, white rice or brown rice in the respective analyses

Adjustment for confounders

Whole grain and refined grain consumption 849

123

123 9,702 m &15,365 w, age 35–65 years: 844 cases

1996/ 2000–2004, 5 years followup

1994/ 1998–2005, 7 years followup

Shanghai Women’s Health Study

European Prospective Investigation into Cancer and Nutrition– Potsdam study

European Prospective Investigation into Cancer and Nutrition– Norfolk study

Physicians’ Health Study 1

Black Women’s Health Study

Villegas et al 2007 [20], China

Schulze et al 2007 [14], Germany

Simmons et al 2007 [15], UK

Kochar et al 2007 [16], USA

Van Dam et al 2006 [6], USA

1995–2003, 8 years followup

1981/ 1983–2002, 19.1 years follow-up

1993/ 1998–2000, 4.6 years follow-up

64,117 w, age 40–70 years: 1,608 cases

1991–2003, 12 years followup

Nurses’ Health Study 2

de Munter et al 2007 [23], USA

41186 w, age 21–69 years: 1,964 cases

21,152 m, mean age 53 years: 1,958 cases

25,633 m & w, age 40–79 years: 417 cases

88,410 w age 26–46 years: 2,739 cases

73,327 w, age 37–65 years: 4,747 cases

1984–2002, 18 years followup

Nurses’ Health Study 1

Study size, gender, age, number of cases

de Munter et al 2007 [23], USA

Follow-up period

Study name

Author, publication year [Ref. no.], country

Table 1 continued

Validated FFQ, 68 food items

FFQ, NA

Validated FFQ,

Validated FFQ, 146 food items

Validated FFQ, 77 food items

Validated FFQ, 131 food items

Validated FFQ, 116 food items

Dietary assessment

1.78 (1.48–2.15)

300 versus \200 g/day Quintile 5 versus 1

Rice Staple food items (rice, noodles, steamed bread, bread)

1.29 versus 0.03 serv/day

C7 versus 0 serv/ week

Refined cereals Whole grains

C7 versus 0 serv/ week

Whole grains cereals

0.69 (0.60–0.79)

0.95 (0.73–1.30)

0.60 (0.50–0.71)

0.69 (0.60–0.79)

0.72 (0.53–0.97)

C1 versus \1 portion/day

C7 versus 0 serv/ week

0.78 (0.62–0.97)

1.37 (1.11–1.69)

80.2 versus 4.4 g/day

Breakfast cereals

Wholemeal/brown bread

Whole grain bread

1.00 (0.85–1.17)

1.9 versus 0.3 g/day

Germ

0.84 (0.71–1.00)

12.0 versus 1.1 g/day

0.86 (0.72–1.02)

0.83 (0.75–0.92)

Bran

1.5 versus 0.2 g/day

Germ

0.72 (0.65–0.80)

39.9 versus 6.2 g/day

9.6 versus 0.6 g/day

Bran

0.75 (0.68–0.83)

RR (95% CI)

Whole grains

31.2 versus 3.7 g/day

Quantity

Whole grains

Exposure

Age, total energy, BMI, smoking status, strenous physical activity, alcohol, parental history of DM, education, coffee, sugarsweetened soft drink, processed meat, red meat, low-fat dairy

Age, smoking, vitamin intake, alcohol, vegetables, physical activity BMI

Unadjusted

Age, sex, BMI, sports activities, education, cycling, occupational activity, smoking, alcohol, total energy intake, waist circumference, PUFA:SFA ratio, MUFA:SFA ratio, carbohydrate, magnesium

Age, energy intake, BMI, WHR, smoking status, alcohol, physical activity, income level, education level, occupation, hypertension

Age, smoking status, physical activity, alcohol, HRT, OC use, FH–T2DM, coffee, sugar-sweetened soft drinks, fruit punch, total energy, processed meat, PUFA/SFA ratio, BMI

Age, smoking status, physical activity, alcohol, HRT, OC use, FH–T2DM, coffee, sugar-sweetened soft drinks, fruit punch, total energy, processed meat, PUFA/SFA ratio, BMI

Adjustment for confounders

850 D. Aune et al.

Study name

Melbourne Collaborative Cohort Study

Finnish Mobile Clinic Health Examination Survey

Health Professionals Follow-up Study

Author, publication year [Ref. no.], country

Hodge et al 2004 [17], Australia

Montonen et al 2003 [8], Finland

Fung et al 2002 [19], USA

Table 1 continued

1986–1998, 12 years followup

1966/ 1972–1995, 23 years follow-up

1990/1994–NA, 4 years follow-up

Follow-up period

42,898 m, age 40–75 years: 1,197 cases

2,286 m & 2,030 w, age 40–69 years: 52/102 cases

31,641 m & w, age 40–69 years: 365 cases

Study size, gender, age, number of cases

Validated FFQ, 131 food items

Dietary history interview, [100 food items

FFQ, 121 food items

Dietary assessment

3.2 versus 0.4 serv/day 4.1 versus 0.8

Refined grains

91–389 versus 0–33 g/day

Refined grain from wheat Whole grains

111–567 versus 0–45 g/day

Refined grain

0.79 (0.56–1.10)

C11.0 versus \2.0 times/week

Other cereal

76–632 versus 0–5 g/day

0.86 (0.60–1.23)

C3.0 versus \0.5 times/week

Pasta

Other whole grain

1.22 (0.89–1.69)

C1.5 versus \0.5 times/week

Savory cereal products

182–1026 versus 0–58 g/day

0.86 (0.63–1.18)

C17.5 versus \0.5 times/week

Whole-meal bread

Rye

1.13 (0.86–1.50)

C7.0 versus \0.5 times/week

White bread

238–1321 versus 0–109 g/day

1.12 (0.79–1.58)

C18.0 versus \6.0 times/week

Bread

Whole grain

0.93 (0.68–1.27)

C2.5 versus \1.0 times/week

Rice

1.08 (0.87–1.33)

0.70 (0.57–0.85)

0.69 (0.41–1.17)

0.62 (0.36–1.06)

1.14 (0.69–1.87)

0.65 (0.36–1.18)

0.65 (0.36–1.18)

0.38 (0.19–0.77)

1.01 (0.75–1.35)

C7.0 versus \0.01 times/week

Breakfast cereal

340–1535 versus 10–181 g/day

1.05 (0.73–1.52)

C41 versus \20 times/week

Cereal

Total grain

RR (95% CI)

Quantity

Exposure

Age, period, physical activity, energy intake, missing FFQ, smoking, FH–DM, alcohol intake, fruit intake, vegetable intake, BMI

Age, sex, geographic area, smoking, BMI, intake of energy, fruit, berries and vegetables

Age, sex, country of birth, physical activity, FH–DM, alcohol intake, education, weight change in the last 5 years, energy intake, BMI, WHR

Adjustment for confounders

Whole grain and refined grain consumption 851

123

123

Nurses’ Health Study 1

Iowa Women’s Health Study

Liu et al 2000 [13], USA

Meyer et al 2000 [5], USA

1986–1992, 6 years followup

1984–1994, 10 years followup

Follow-up period

35,988 w, age 55–69 years: 1,141 cases

75,521 w, age 38–63 years: 1,879 cases

Study size, gender, age, number of cases

Validated FFQ, 127 food items

FFQ, 126 food items

Dietary assessment

5–6/week versus almost never \1/week versus almost never

Bran Other grains

29.5 versus 3.5 serv/week

5–6/week versus almost never

Wheat germ

Refined grains

5–6/week versus almost never

Brown rice

20.5 versus 1.0 serv/week

C1/day versus almost never

Cooked oatmeal

Whole grains

C1/day versus almost never

Popcorn

41.5 versus 9.5 serv/week

0.77 (0.63–0.94)

C1/day versus almost never

Whole-grain breakfast cereal

Total grains

0.54 (0.41–0.72)

C1/day versus almost never

Dark bread

1.11 (0.94–1.30)

0.87 (0.70–1.08)

0.79 (0.65–0.96)

0.68 (0.54–0.87)

0.85 (0.52–1.37)

0.47 (0.15–1.45)

0.73 (0.35–1.54)

0.88 (0.59–1.31)

0.66 (0.55–0.80)

0.77 (0.66–0.90)

1.26 (1.08–1.46)

Quintile 5 versus 1 Quintile 5 versus 1

Refined/whole grain ratio

0.73 (0.63–0.85)

0.75 (0.63–0.89)

RR (95% CI)

Refined grain

Quintile 5 versus 1 2.70 versus 0.13 serv/dayay

Whole grain

Quantity

Total grain

Exposure

adj. adjustment, BMI body mass index, DM diabetes mellitus, FFQ food frequency questionnaire, FH family history, m men, NA not available, WHR waist-to-hip ratio, w women

Study name

Author, publication year [Ref. no.], country

Table 1 continued

Age, total energy intake, BMI, WHR, education, pack-years of smoking, alcohol intake, physical activity

Age, BMI, physical activity, cigarette smoking, alcohol intake, FH–DM 2 in a 1st degree relative, use of multivitamins or vitamin E supplements, total energy intake

Adjustment for confounders

852 D. Aune et al.

Whole grain and refined grain consumption

853

A

A Relative Risk

Relative Risk (95% CI)

Study

Study

(95% CI)

Ericson, 2013

0.77 ( 0.63, 0.94)

Parker, 2013

0.83 ( 0.69, 0.99)

Ericson, 2013

0.98 ( 0.85, 1.13)

Wirström, 2013

0.68 ( 0.41, 1.12)

Parker, 2013

0.89 ( 0.82, 0.96)

Sun, 2010, HPFS

0.66 ( 0.55, 0.79)

Sun, 2010, NHS1

0.46 ( 0.39, 0.56)

Montonen, 2003

0.66 ( 0.43, 1.00)

Fung, 2002

1.03 ( 0.86, 1.22)

Liu, 2000

1.07 ( 0.95, 1.20)

Meyer, 2000

0.93 ( 0.79, 1.08)

Sun, 2010, NHS2

0.69 ( 0.54, 0.88)

Fisher, 2009

0.96 ( 0.81, 1.13)

van Dam, 2006

0.41 ( 0.30, 0.56)

Montonen, 2003

0.75 ( 0.48, 1.17)

Meyer, 2000

0.77 ( 0.63, 0.93)

Overall

0.95 ( 0.88, 1.04)

Overall

0.68 ( 0.58, 0.81)

.25

.5

.75

1

.25

1.5

.75

.5

1

1.5

Relative Risk

Relative Risk

B B

1.2

1.2

1.0

1.0

0.8

0.8

RR

RR

0.6

0.6

0.4

0.4 0

1

2

3

4

5

Whole grains (serving/day) Best fitting fractional polynomial 95% confidence interval

0

1

2

3

4

5

6

7

Refined grains (servings/day) Best fitting fractional polynomial 95% confidence interval

Fig. 2 Whole grains and type 2 diabetes. Summary estimates were calculated using a random-effects model

Fig. 3 Refined grains and type 2 diabetes. Summary estimates were calculated using a random-effects model

Egger’s test, p = 1.00 or with Begg’s test, p = 1.00. There was no evidence of a nonlinear association between refined grain intake and type 2 diabetes risk, pnonlinearity = 0.10 (Fig. 3b, Supplementary Table 2).

summary RR for high versus low intake was 0.82 (95 % CI 0.72–0.94, I2 = 50 %, pheterogeneity = 0.11, n = 4) for whole grain bread [5, 13, 14, 17], 0.66 (95 % CI 0.57–0.77, I2 = 35 %, pheterogeneity = 0.21, n = 3) for whole grain cereals [5, 13, 16], 0.76 (95 % CI 0.69–0.84, I2 = 30 %, pheterogeneity = 0.24, n = 3) for wheat bran [7], 0.97 (95 % CI 0.86–1.10, I2 = 59 %, pheterogeneity = 0.09, n = 3) for wheat germ [7], 0.89 (95 % CI: 0.81–0.97, I2 = 0 %, pheterogeneity = 0.40, n = 3) for brown rice [7], 1.17 (95 % CI: 0.93–1.47, I2 = 78 %, pheterogeneity \ 0.0001, n = 7) for white rice [7, 17, 20–22], and 0.82 (95 % CI 0.56–1.18, n = 2) for total cereals [16, 17] (Table 2). Nonlinear associations were observed for whole grain bread, pnonlinearity = 0.01, whole grain cereals, pnonlinearity \ 0.0001, wheat bran, pnonlinearity = 0.007, and brown rice, pnonlinearity = 0.02, and consistent with the analysis of overall whole grain intake, the reduction in risk was steepest when increasing the intake from low levels (Supplementary Figure 5a-d). We were not able to fit a nonlinear curve for

Total grains and subtypes of grains Fewer studies had reported on total grains and subtypes of grains. The summary RR for high versus low total grain intake was 0.74 (95 % CI 0.58–0.93) [5, 8, 13, 17] with moderate heterogeneity, I2 = 60 %, pheterogeneity = 0.06 (Supplementary Figure 3). The summary RR per 3 servings per day was 0.83 (95 % CI 0.75–0.91, I2 = 36 %, pheterogeneity = 0.19) (Supplementary Figure 4a). There was evidence of a nonlinear association between total grain intake and type 2 diabetes, pnonlinearity = 0.001, and the reduction in risk was steeper at the lower and higher end of the intake, with a slight flattening at intermediate intakes (Supplementary Figure 4b, Supplementary Table 3). The

123

854

D. Aune et al.

Table 2 Subtypes of grains and type 2 diabetes risk Type of grain

High versus low comparison N

RR (95 % CI)

I

2

Dose-response analysis Pheterogeneity

Dose

N

RR (95 % CI)

I2

Pheterogeneity

Whole grain bread

4

0.81 (0.74–0.89)

0

0.60

Per 3 serv/day

3

0.74 (0.56–0.98)

44.1

0.17

Whole grain breakfast cereal

3

0.72 (0.55–0.93)

77.8

0.01

Per 1 serv/day

3

0.73 (0.59–0.91)

80.3

0.006

Brown rice

3

0.89 (0.81–0.97)

50

0.11

Per 0.5 serv/day

3

0.87 (0.78–0.97)

26.1

0.26

Wheat bran

3

0.76 (0.69–0.84)

30

0.24

Per 10 g/day

3

0.79 (0.72–0.87)

49.1

0.14

Wheat germ

3

0.97 (0.86–1.10)

59

0.09

Per 2 g/day

3

0.98 (0.87–1.11)

50.1

0.14

White rice

7

1.17 (0.93–1.47)

78.1

Per 1 serv/day

6

1.23 (1.15–1.31)

21.4

0.27

\0.0001

white rice, possibly due to large differences in the intake between studies. Subgroup and sensitivity analyses There was no significant heterogeneity between subgroups in analyses of whole grains and type 2 diabetes stratified by gender, duration of follow-up, geographic area, number of cases and adjustment for confounding factors and inverse associations were apparent in most subgroups, although they were not always statistically significant (Table 3). Although the test for heterogeneity was not significant, pheterogeneity = 0.15, the association appeared to be slightly stronger in the American studies than among the European studies. Because BMI may be an intermediate variable we also restricted the analysis to the five studies (four publications) that had presented risk estimates both adjusted and not adjusted for BMI [10, 12, 19, 23]. The summary RR per 3 servings per day increase in whole grain intake was 0.69 (0.60–0.80, I2 = 58 %, pheterogeneity = 0.05) with BMI adjustment (and this was similar to the result from the main analysis) and 0.53 (95 % CI 0.41–0.69, I2 = 88 %, pheterogeneity \ 0.001) without BMI adjustment (Fig. 4a) and there were similar differences in the results by BMI adjustment in the nonlinear analysis (Fig. 4b).

Discussion Our meta-analysis supports the hypothesis that a high whole grain and total grain intake protects against type 2 diabetes with a 32 and 17 % reduction in the relative risk per 3 servings per day, but we found no association between overall refined grain intake and type 2 diabetes risk. There was evidence of a nonlinear inverse association between whole grains and total grains and type 2 diabetes with most of the reduction observed when increasing the intake up to 2 servings per day for whole grain intake, while for total grains there was also a steep reduction in relative risk when increasing intake from low levels,

123

followed by a slight flattening of the curve with intermediate intakes and a steeper reduction at higher intakes. However, the inverse association with high total grain intake should be interpreted with caution as it was based on relatively few studies, and is likely to be driven by higher whole grain intake since there was no association with overall refined grain intake. A positive association was observed with intake of white rice. In addition, we found that several subtypes of whole grains including whole grain cereals, brown bread and brown rice were associated with reduced risk, but these analyses were based on few studies and need further confirmation. Our meta-analysis has limitations which affect the interpretation of the results. The main limitation is the low number of cohort studies available apart from the total whole grain analysis. Further studies are therefore needed before firm conclusions can be made for the remaining exposures. Although it is possible that the inverse association between whole grain intake and type 2 diabetes could be due to unmeasured or residual confounding by other lifestyle factors we found that the association persisted in several subgroup analyses where such factors had been adjusted for. There was high heterogeneity in the dose– response analysis of whole grains and type 2 diabetes, although not in the comparison of the highest versus the lowest intake. There was less heterogeneity in studies conducted among men than among women, but there was no significant heterogeneity between these subgroups, or when stratified by number of cases, duration of follow-up or adjustment for confounding factors. A slightly stronger association was observed in the American studies than among the European studies, but there was also no significant heterogeneity by geographic location, suggesting that this finding could be due to chance. Because of the low number of studies our ability to test for publication bias may have been limited, however, there was no indication of asymmetry in the funnel plots. In addition, because of the low number of studies with very high intakes of whole grains and total grains, the results in the high ranges ([3 servings for whole grains, and [7 servings for total grains) were based on relatively few datapoints and should be

Whole grain and refined grain consumption

855

Table 3 Subgroup analyses of whole intake and type 2 diabetes, dose–response Whole grains, 3 servings per day n

RR (95 % CI)

I2 (%)

Pah

10

0.68 (0.58–0.81)

81.9

\0.0001

\10 years follow-up

5

0.72 (0.56–0.93)

82.3

\0.0001

C10 years follow-up

5

0.65 (0.53–0.79)

75.3

0.003

Men

3

0.70 (0.61–0.81)

0

0.53

Women Men and women

7 2

0.64 (0.51–0.80) 0.93 (0.79–1.09)

82.0 1.5

\0.0001 0.31

Europe

4

0.84 (0.72–0.97)

23.8

0.27

America

6

0.62 (0.51–0.77)

84.0

\0.0001

3

0.88 (0.73–1.06)

13.6

0.31

All studies

Pbh

Duration of follow-up 0.26

Sex 0.43/0.723

Geographic location 0.15

Number of cases Cases \1,000 Cases 1,000–\2,000

3

0.64 (0.46–0.89)

84.5

0.002

Cases C2.000

4

0.65 (0.50–0.83)

85.1

\0.0001

0.68 (0.58–0.81)

81.9

\0.0001

0.68 (0.57–0.81)

83.9

\0.0001

0.32

Adjustment for confounders Body mass index

Yes

10

No

0

Physical activity

Yes

9

No

1

0.75 (0.48–1.17)

Smoking

Yes

10

0.68 (0.58–0.81)

81.9

\0.0001

Alcohol

No Yes

0 8

0.68 (0.56–0.82)

85.9

\0.0001

No

2

0.72 (0.52–1.01)

0

0.78

Yes

1

0.41 (0.30–0.56)

No

9

0.72 (0.61–0.84)

78.7

\0.0001

Coffee Red and/or processed meat Dairy products Fruits and/or vegetables Energy intake

Yes

5

0.61 (0.46–0.83)

90.5

\0.0001

5

0.78 (0.71–0.87)

0

0.94

Yes

3

0.70 (0.47–1.05)

90.9

\0.0001

No

7

0.67 (0.57–0.78)

69.0

0.004

Yes

5

0.66 (0.53–0.82)

80.6

\0.0001

No

5

0.71 (0.55–0.91)

82.0

\0.0001

Yes

9

0.68 (0.57–0.82)

83.9

\0.0001

No

1

0.68 (0.41–1.12)

P for heterogeneity within each subgroup

2

P for heterogeneity between subgroups with meta-regression analysis

3

P for heterogeneity between men and women (excluding studies with both genders)

0.78 NC 0.84 0.07

No

a

NC

0.23 0.94 0.74 0.99

NC not calculable

interpreted with caution. Measurement errors in the exposure assessment are known to bias effect estimates, but because we only included prospective cohort studies such measurement errors are most likely to have resulted in attenuation of the association between whole grain intake and type 2 diabetes risk. None of the studies published to date have corrected their results for measurement error. The definition of whole grains differed in some of the

studies (Supplementary Table 4) with several American studies considering breakfast cereals to be made of whole grains if the product contained C25 % whole grain or bran by weight [5, 7, 13, 19, 23], while one Swedish study used C50 % as a cut-off point [9]. Several other studies did not state how whole grains were defined, thus it is difficult to assess whether the differing definitions might have influenced the results. Further studies using biomarkers of

123

856

D. Aune et al.

A Study

Relative Risk (95% CI)

with BMI adjustment Ericson, 2013

0.77 ( 0.63, 0.94)

Parker, 2013 de Munter, 2007, NHS1

0.76 ( 0.64, 0.91) 0.53 ( 0.43, 0.64)

de Munter, 2007, NHS2 Fung, 2000

0.73 ( 0.55, 0.97) 0.70 ( 0.57, 0.85) 0.69 ( 0.60, 0.80)

Subtotal no BMI adjustment Ericson, 2013 Parker, 2013

0.75 ( 0.62, 0.91) 0.60 ( 0.50, 0.72) 0.35 ( 0.29, 0.42) 0.48 ( 0.36, 0.64)

de Munter, 2007, NHS1 de Munter, 2007, NHS2

0.56 ( 0.46, 0.68) 0.53 ( 0.41, 0.69)

Fung, 2000 Subtotal

.1

.25

.5

.75

1

1.5

Relative Risk

B 1.2 1.0

0.8

RR 0.6

0.4 0

1

2

3

4

Whole grains (serv/day) without BMI adjustment with BMI adjustment

95% CI 95% CI

Fig. 4 Whole grains and type 2 diabetes, with and without adjustment for BMI. Summary estimates were calculated using a randomeffects model

whole grain intake could be useful to assess the impact of measurement errors in the dietary assessment [50] and any further studies on dietary whole grain intake should report the definition of whole grain foods used in the analysis for comparison between studies. A protective effect of whole grain consumption against type 2 diabetes is biologically plausible and several mechanisms may operate to reduce the risk. Several studies have reported inverse associations between whole grain intake and prospective weight gain [25–30] and we found that the size of the association between whole grains and type 2 diabetes was about 1/3 stronger when the analyses were not adjusted for BMI compared with adjustment for BMI (RR = 0.53 vs. 0.69, respectively) [10, 12, 19, 23]. Thus, reduced body fatness may explain part, but not all of the protective effect of whole grains against type 2 diabetes risk.

123

The results of the nonlinear analysis stratified by adjustment for BMI suggest that reduced body fatness may explain a larger part of the association at higher levels compared with lower levels of whole grain intake as the association appeared to have a more linear shape in analyses without adjustment for BMI than when adjusted for BMI. Whole grains are an important source of cereal fiber, phytochemicals, vitamins and minerals. High whole grain intake has been associated with greater insulin sensitivity and lower fasting insulin concentration and this was observed for dark breads, and in particular high-fiber cereals [51]. Intake of cereal fiber, but not fruit or vegetable fiber, has been associated with reduced type 2 diabetes risk in a meta-analysis of prospective studies [14]. Greater intake of soluble fiber reduces the rate of gastric emptying and leads to a slower blood glucose and insulin response [52–54]. However, whole grains contain more insoluble fiber, thus other mechanisms are probably involved than just the latter. Intake of rye bread has been shown to result in a lower postprandial insulin response and this was found to be independent of its fiber content [55]. In addition, high intake of whole grains may reduce risk of type 2 diabetes by reducing concentrations of inflammatory markers including plasminogen activator inhibitor type 1 and C-reactive protein [56–60] and liver enzymes including gamma-glutamyltransferase and aspartate aminotransferase [56], as higher concentrations of these proteins may increase type 2 diabetes risk [61–63]. In addition, a high intake of whole grains and cereal fiber has been associated with greater blood concentrations of adiponectin [57, 64], a cytokine that increases insulin sensitivity and reduces inflammation [65]. Further studies are needed to explore potential mechanisms that could explain the nonlinear associations observed. Our meta-analysis also has several strengths. Because we based our analysis on prospective cohort studies recall bias is not likely to explain our findings, and the possibility for selection bias is reduced. Although the number of studies was moderate they included up to 19,800 cases and 385,000 participants and we therefore had adequate statistical power to detect moderate associations. We conducted several subgroup analyses and observed that the inverse association persisted in most subgroup analyses, and the findings were also robust in sensitivity analyses where each study was excluded one at a time. We quantified the association between grain intake and type 2 diabetes by conducting linear and nonlinear dose–response analyses and found that most of the benefit of whole grains on type 2 diabetes risk is observed with an intake of at least 2 servings per day (60 g/day). However, if whole grains reduce body fatness and body mass index is a mediating factor, further reductions in the risk may be observed with higher intakes. Increasing whole grain intakes is also likely to reduce the risk of cardiovascular disease [66],

Whole grain and refined grain consumption

overweight and obesity [24–30] and colorectal cancer [43], and it is possible that there are greater benefits for these outcomes with even higher intakes. In summary, our meta-analysis suggests that a high intake of whole grains, but not refined grains, is associated with reduced type 2 diabetes risk. However, a positive association with intake of white rice and inverse associations between several specific types of whole grains and type 2 diabetes warrant further investigations. Our results support public health recommendations to replace refined grains with whole grains and suggest that at least two servings of whole grains per day should be consumed to reduce type 2 diabetes risk. Acknowledgement DA designed the project, conducted the literature search and analyses and wrote the first draft of the paper. DA, TN, PR, LJV interpreted the data and revised the subsequent drafts for important intellectual content and approved the final version of the paper to be published. The authors declare that there is no duality of interest associated with this manuscript. This project has been funded by Liaison Committee between the Central Norway Regional Health Authority (RHA) and the Norwegian University of Science and Technology (NTNU). We thank Ulrika Ericson for clarifying the definition of highfibre cereals and breads in the Malmo¨ Diet and Cancer cohort.

References 1. Whiting DR, Guariguata L, Weil C, Shaw J. IDF diabetes atlas: global estimates of the prevalence of diabetes for 2011 and 2030. Diabetes Res Clin Pract. 2011;94(3):311–21. 2. Standards of medical care in diabetes–2006. Diabetes Care. 2006;29 Suppl 1:S4–42. 3. Economic costs of diabetes in the u.s. In 2012. Diabetes Care 2013;36(4):1033–46. 4. Rana JS, Li TY, Manson JE, Hu FB. Adiposity compared with physical inactivity and risk of type 2 diabetes in women. Diabetes Care. 2007;30(1):53–8. 5. Meyer KA, Kushi LH, Jacobs DR Jr, Slavin J, Sellers TA, Folsom AR. Carbohydrates, dietary fiber, and incident type 2 diabetes in older women. Am J Clin Nutr. 2000;71(4):921–30. 6. van Dam RM, Hu FB, Rosenberg L, Krishnan S, Palmer JR. Dietary calcium and magnesium, major food sources, and risk of type 2 diabetes in US black women. Diabetes Care. 2006;29(10):2238–43. 7. Sun Q, Spiegelman D, van Dam RM, Holmes MD, Malik VS, Willett WC, et al. White rice, brown rice, and risk of type 2 diabetes in US men and women. Arch Intern Med. 2010;170(11):961–9. 8. Montonen J, Knekt P, Jarvinen R, Aromaa A, Reunanen A. Whole-grain and fiber intake and the incidence of type 2 diabetes. Am J Clin Nutr. 2003;77(3):622–9. 9. Wirstro¨m T, Hilding A, Gu HF, Ostenson CG, Bjorklund A. Consumption of whole grain reduces risk of deteriorating glucose tolerance, including progression to prediabetes. Am J Clin Nutr. 2013;97(1):179–87. 10. Parker ED, Liu S, Van Horn L, Tinker LF, Shikany JM, Eaton CB, et al. The association of whole grain consumption with incident type 2 diabetes: the Women’s Health Initiative Observational Study. Ann Epidemiol. 2013;23(6):321–7. 11. Fisher E, Boeing H, Fritsche A, Doering F, Joost HG, Schulze MB. Whole-grain consumption and transcription factor-7-like 2 (TCF7L2) rs7903146: gene-diet interaction in modulating type 2 diabetes risk. Br J Nutr. 2009;101(4):478–81.

857 12. Ericson U, Sonestedt E, Gullberg B, Hellstrand S, Hindy G, Wirfalt E, Orho-Melander M. High intakes of protein and processed meat associate with increased incidence of type 2 diabetes. Br J Nutr. 2013;109(6):1143–53. 13. Liu S, Manson JE, Stampfer MJ, Hu FB, Giovannucci E, Colditz GA, et al. A prospective study of whole-grain intake and risk of type 2 diabetes mellitus in US women. Am J Public Health. 2000;90(9):1409–15. 14. 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(9):956–65. 15. Simmons RK, Harding AH, Wareham NJ, Griffin SJ. Do simple questions about diet and physical activity help to identify those at risk of Type 2 diabetes? Diabet Med. 2007;24(8):830–5. 16. Kochar J, Djousse L, Gaziano JM. Breakfast cereals and risk of type 2 diabetes in the Physicians’ Health Study I. Obesity (Silver Spring). 2007;15(12):3039–44. 17. Hodge AM, English DR, O’Dea K, Giles GG. Glycemic index and dietary fiber and the risk of type 2 diabetes. Diabetes Care. 2004;27(11):2701–6. 18. von Ruesten A, Feller S, Bergmann MM, Boeing H. Diet and risk of chronic diseases: results from the first 8 years of follow-up in the EPIC-Potsdam study. Eur J Clin Nutr. 2013;67(4):412–9. 19. Fung TT, Hu FB, Pereira MA, Liu S, Stampfer MJ, Colditz GA, et al. Whole-grain intake and the risk of type 2 diabetes: a prospective study in men. Am J Clin Nutr. 2002;76(3):535–40. 20. Villegas R, Liu S, Gao YT, Yang G, Li H, Zheng W, 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(21):2310–6. 21. Nanri A, Mizoue T, Noda M, Takahashi Y, Kato M, Inoue M, et al. Rice intake and type 2 diabetes in Japanese men and women: the Japan Public Health Center-based Prospective Study. Am J Clin Nutr. 2010;92(6):1468–77. 22. Soriguer F, Colomo N, Olveira G, Garcia-Fuentes E, Esteva I, Ruiz de Adana MS, et al. White rice consumption and risk of type 2 diabetes. Clin Nutr. 2013;32(3):481–4. 23. de Munter JS, Hu FB, Spiegelman D, Franz M, van Dam RM. Whole grain, bran, and germ intake and risk of type 2 diabetes: a prospective cohort study and systematic review. PLoS Med. 2007;4(8):e261. 24. Ye EQ, Chacko SA, Chou EL, Kugizaki M, Liu S. Greater wholegrain intake is associated with lower risk of type 2 diabetes, cardiovascular disease, and weight gain. J Nutr. 2012;142(7): 1304–13. 25. Liu S, Willett WC, Manson JE, Hu FB, Rosner B, Colditz G. Relation between changes in intakes of dietary fiber and grain products and changes in weight and development of obesity among middle-aged women. Am J Clin Nutr. 2003;78(5):920–7. 26. Koh-Banerjee P, Franz M, Sampson L, Liu S, Jacobs DR Jr, Spiegelman D, et al. Changes in whole-grain, bran, and cereal fiber consumption in relation to 8-y weight gain among men. Am J Clin Nutr. 2004;80(5):1237–45. 27. Bazzano LA, Song Y, Bubes V, Good CK, Manson JE, Liu S. Dietary intake of whole and refined grain breakfast cereals and weight gain in men. Obes Res. 2005;13(11):1952–60. 28. van de Vijver LP, van den Bosch LM, van den Brandt PA, Goldbohm RA. Whole-grain consumption, dietary fibre intake and body mass index in the Netherlands cohort study. Eur J Clin Nutr. 2009;63(1):31–8. 29. Mozaffarian D, Hao T, Rimm EB, Willett WC, Hu FB. Changes in diet and lifestyle and long-term weight gain in women and men. N Engl J Med. 2011;364(25):2392–404. 30. Quick V, Wall M, Larson N, Haines J, Neumark-Sztainer D. Personal, behavioral and socio-environmental predictors of

123

858

31.

32.

33.

34.

35.

36.

37.

38.

39.

40. 41.

42.

43.

44.

45.

46.

47. 48.

49.

D. Aune et al. overweight incidence in young adults: 10-yr longitudinal findings. Int J Behav Nutr Phys Act. 2013;10:37. Danaei G, Pan A, Hu FB, Hernan MA. Hypothetical Midlife Interventions in Women and Risk of Type 2 Diabetes. Epidemiology. 2013;24(1):122–8. Schulze MB, Hoffmann K, Boeing H, Linseisen J, Rohrmann S, Mohlig M, et al. An accurate risk score based on anthropometric, dietary, and lifestyle factors to predict the development of type 2 diabetes. Diabetes Care. 2007;30(3):510–5. Montonen J, Jarvinen R, Heliovaara M, Reunanen A, Aromaa A, Knekt P. Food consumption and the incidence of type II diabetes mellitus. Eur J Clin Nutr. 2005;59(3):441–8. Hodge AM, English DR, O’Dea K, Giles GG. Dietary patterns and diabetes incidence in the Melbourne Collaborative Cohort Study. Am J Epidemiol. 2007;165(6):603–10. Colditz GA, Manson JE, Stampfer MJ, Rosner B, Willett WC, Speizer FE. Diet and risk of clinical diabetes in women. Am J Clin Nutr. 1992;55(5):1018–23. Feskens EJ, Virtanen SM, Rasanen L, Tuomilehto J, Stengard J, Pekkanen J, et al. Dietary factors determining diabetes and impaired glucose tolerance. A 20-year follow-up of the Finnish and Dutch cohorts of the Seven Countries Study. Diabetes Care. 1995;18(8):1104–12. Williams DE, Wareham NJ, Cox BD, Byrne CD, Hales CN, Day NE. Frequent salad vegetable consumption is associated with a reduction in the risk of diabetes mellitus. J Clin Epidemiol. 1999;52(4):329–35. Mohan V, Radhika G, Sathya RM, Tamil SR, Ganesan A, Sudha V. Dietary carbohydrates, glycaemic load, food groups and newly detected type 2 diabetes among urban Asian Indian population in Chennai, India (Chennai Urban Rural Epidemiology Study 59). Br J Nutr. 2009;102(10):1498–506. Lecomte P, Vol S, Caces E, Born C, Chabrolle C, Lasfargues G, et al. Five-year predictive factors of type 2 diabetes in men with impaired fasting glucose. Diabetes Metab. 2007;33(2):140–7. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials. 1986;7(3):177–88. Greenland S, Longnecker MP. Methods for trend estimation from summarized dose-response data, with applications to meta-analysis. Am J Epidemiol. 1992;135(11):1301–9. Aune D, Greenwood DC, Chan DS, Vieira R, Vieira AR, Navarro Rosenblatt DA, et al. Body mass index, abdominal fatness and pancreatic cancer risk: a systematic review and non-linear doseresponse meta-analysis of prospective studies. Ann Oncol. 2012;23(4):843–52. Aune D, Chan DS, Lau R, Vieira R, Greenwood DC, Kampman E, et al. Dietary fibre, whole grains, and risk of colorectal cancer: systematic review and dose-response meta-analysis of prospective studies. BMJ. 2011;343:d6617. Hu EA, Pan A, Malik V, Sun Q. White rice consumption and risk of type 2 diabetes: meta-analysis and systematic review. BMJ. 2012;344:e1454. Royston P. A strategy for modelling the effect of a continuous covariate in medicine and epidemiology. Stat Med. 2000;19(14): 1831–47. Bagnardi V, Zambon A, Quatto P, Corrao G. Flexible metaregression functions for modeling aggregate dose-response data, with an application to alcohol and mortality. Am J Epidemiol. 2004;159(11):1077–86. Higgins JP, Thompson SG. Quantifying heterogeneity in a metaanalysis. Stat Med. 2002;21(11):1539–58. Egger M, Davey SG, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. BMJ. 1997;315(7109): 629–34. Begg CB, Mazumdar M. Operating characteristics of a rank correlation test for publication bias. Biometrics. 1994;50(4):1088–101.

123

50. Ross AB, Bourgeois A, Macharia HN, Kochhar S, Jebb SA, Brownlee IA, et al. Plasma alkylresorcinols as a biomarker of whole-grain food consumption in a large population: results from the WHOLEheart Intervention Study. Am J Clin Nutr. 2012;95(1):204–11. 51. Liese AD, Roach AK, Sparks KC, Marquart L, D’Agostino RB Jr, Mayer-Davis EJ. Whole-grain intake and insulin sensitivity: the Insulin Resistance Atherosclerosis Study. Am J Clin Nutr. 2003;78(5):965–71. 52. Torsdottir I, Alpsten M, Holm G, Sandberg AS, Tolli J. A small dose of soluble alginate-fiber affects postprandial glycemia and gastric emptying in humans with diabetes. J Nutr. 1991; 121(6):795–9. 53. Holt S, Heading RC, Carter DC, Prescott LF, Tothill P. Effect of gel fibre on gastric emptying and absorption of glucose and paracetamol. Lancet. 1979;1(8117):636–9. 54. Jenkins DJ, Wolever TM, Leeds AR, Gassull MA, Haisman P, Dilawari J, et al. Dietary fibres, fibre analogues, and glucose tolerance: importance of viscosity. Br Med J. 1978;1(6124): 1392–4. 55. Juntunen KS, Laaksonen DE, Autio K, Niskanen LK, Holst JJ, Savolainen KE, et al. Structural differences between rye and wheat breads but not total fiber content may explain the lower postprandial insulin response to rye bread. Am J Clin Nutr. 2003;78(5):957–64. 56. Montonen J, Boeing H, Fritsche A, Schleicher E, Joost HG, Schulze MB, et al. Consumption of red meat and whole-grain bread in relation to biomarkers of obesity, inflammation, glucose metabolism and oxidative stress. Eur J Nutr. 2013;52(1):337–45. 57. Qi L, van Dam RM, Liu S, Franz M, Mantzoros C, Hu FB. Whole-grain, bran, and cereal fiber intakes and markers of systemic inflammation in diabetic women. Diabetes Care. 2006;29(2):207–11. 58. Gaskins AJ, Mumford SL, Rovner AJ, Zhang C, Chen L, Wactawski-Wende J, et al. Whole grains are associated with serum concentrations of high sensitivity C-reactive protein among premenopausal women. J Nutr. 2010;140(9):1669–76. 59. Masters RC, Liese AD, Haffner SM, Wagenknecht LE, Hanley AJ. Whole and refined grain intakes are related to inflammatory protein concentrations in human plasma. J Nutr. 2010;140(3): 587–94. 60. Katcher HI, Legro RS, Kunselman AR, Gillies PJ, Demers LM, Bagshaw DM, et al. The effects of a whole grain-enriched hypocaloric diet on cardiovascular disease risk factors in men and women with metabolic syndrome. Am J Clin Nutr. 2008;87(1):79–90. 61. Hu FB, Meigs JB, Li TY, Rifai N, Manson JE. Inflammatory markers and risk of developing type 2 diabetes in women. Diabetes. 2004;53(3):693–700. 62. Qi L, Rifai N, Hu FB. Interleukin-6 receptor gene, plasma C-reactive protein, and diabetes risk in women. Diabetes. 2009;58(1):275–8. 63. Fraser A, Harris R, Sattar N, Ebrahim S, Davey SG, Lawlor DA. Alanine aminotransferase, gamma-glutamyltransferase, and incident diabetes: the British Women’s Heart and Health Study and meta-analysis. Diabetes Care. 2009;32(4):741–50. 64. Qi L, Rimm E, Liu S, Rifai N, Hu FB. Dietary glycemic index, glycemic load, cereal fiber, and plasma adiponectin concentration in diabetic men. Diabetes Care. 2005;28(5):1022–8. 65. Li S, Shin HJ, Ding EL, van Dam RM. Adiponectin levels and risk of type 2 diabetes: a systematic review and meta-analysis. JAMA. 2009;302(2):179–88. 66. Mellen PB, Walsh TF, Herrington DM. Whole grain intake and cardiovascular disease: a meta-analysis. Nutr Metab Cardiovasc Dis. 2008;18(4):283–90.