Introduction

Overweight and obesity are major health concerns worldwide since both are associated with higher incidences of chronic diseases, including of several types of cancer, type 2 diabetes and cardiovascular disease.1, 2

Whole grains have been suggested to have a role in body weight management, in part because whole grain foods may be more satiating than refined foods and due to lower digestibility of nutrients such as fat and protein.3, 4 Furthermore, dietary fiber in whole grains may have beneficial effects on postprandial blood glucose and insulin,5 and may delay gastric emptying and increases bowel transit time.6 It is possible that different whole grain sources, for example, wheat, rye and oats, have different associations with body weight management because of different contents and types of dietary fibers and bioactive compounds.7 Cohort studies have rather consistently found that whole grain and cereal fiber intake are related to lower body weight gain, but the whole grain intake in these studies are almost entirely wheat-dominated.8 In randomized controlled studies, both mixed whole grain diets and exclusively whole grain wheat, rye, oats and barley have been investigated,9 but no effect on body weight whole grain intake has been observed.9 One reason for the conflicting results of cohort studies and randomized controlled studies might be that controlled studies are too short to detect an effect of whole grains on body weight, and that intervention studies have typically not been designed to study weight loss. Furthermore, most of the studies are conducted in overweight persons and it might be that whole grain intake has greater effect in the primary prevention of overweight and obesity rather than in the treatment of already overweight or obese persons. The discrepancies might also be due to confounding in cohort studies as whole grain consumption is associated with a healthy lifestyle pattern.10

Assessment of whole grain intake in observational studies is challenging because it is difficult to quantify the amounts of whole grains in the products consumed,11 and no universal definition of whole grains has been established.7 Dietary biomarkers could therefore be useful as a complementary measure of the whole grain intake. Alkylresorcinols are phenolic lipids found almost exclusively in the bran of wheat and rye (and not in oats), and are considered suitable biomarkers of whole grain wheat and rye intake in populations with frequent and regular intake.12 Five homologs are usually measured and their sum are indicative of the intake of whole grain wheat and rye, and the ratio between two of the homologs (C17:0/C21:0) is indicating the ratio of rye to wheat intake,13 with the ratios <0.2 indicating a wheat-dominated whole grain diet and higher indicating that rye also is consumed.14, 15, 16

The aim of the present study was to investigate the association between whole grain intake and subsequent changes in anthropometry, including body weight and waist circumference (WC) in a Danish cohort with large variation in whole grain intake, both regarding quantity and sources. Whole grain intake was estimated from food frequency questionnaires (FFQ) and plasma concentrations of alkylresorcinols were measured in a subset. A further aim was to investigate the associations for different whole grain types, mainly wheat, rye and oats, as well as different whole grain products.

Subjects and methods

Study population

The Danish Diet, Cancer and Health prospective cohort study was initiated in 1993. During 1993-1997, 57 053 participants were included (35% of those invited). The study was approved by the relevant scientific committees and the Danish Data Protection Agency, and all the participants gave written consent. All participants filled in a FFQ and a lifestyle questionnaire.17

Participants were invited to take part in follow-up study in 1999–2002.17 Of the 57 053 original participants, 54 355 participants were eligible and were invited. Of these, 9468 did not participate (17%). Hence, 44 887 participants agreed to participate in the follow-up study.

Anthropometric measures

Anthropometry was measured at baseline and at follow-up.17, 18

At baseline, body weight, height and WC was measured by trained personnel. The participants were wearing light clothes or underwear. Body weight was measured to the nearest 0.1 kg using a digital scale and height was measured to the nearest 0.5 cm with subjects standing without shoes. WC and hip circumference (HC) was measured with a rigid measuring tape and recorded to nearest 0.5 cm. WC was measured at the smallest horizontal circumference between the ribs and iliac crest, or, in case of an indeterminable waist narrowing, halfway between the lower rib and the iliac crest. HC was measured at the largest horizontal expansion of the buttocks.18, 19

Follow-up anthropometric measures were self-reported and measured at home by the participants. They were instructed to weigh themselves (to the nearest 1 kg), and were provided with a measuring tape and were told to measure WC at the level of the umbilicus (to the nearest 1 cm).18

Assessment of dietary intake and lifestyle factors

Baseline dietary intakes are used in the present study. Prior to the visit to the study center, participants were asked to fill in a 192-item validated FFQ.20, 21 The intakes of specific foods were calculated using the software program FoodCalc.22

Whole grains were defined according to the following definition from the American Association of Cereal Chemists: ‘Whole grains shall consist of the intact, ground, cracked or flaked caryopsis, whose principal anatomical components—the starchy endosperm, germ and bran—are present in the same relative proportion as they exist in the intact caryopsis'.23

Estimations on the whole grain product intake was based only on information from the FFQ. The whole grain products rye bread, whole grain bread, rolled oats and muesli, or crispbread were investigated alone, and the sum of them (whole grain products). The majority of the products included were whole grain products, but a few non-whole grain products were also included, such as non-whole grain crispbread.

Total whole grain intake (g/day) was estimated using the FFQ and the 24 h dietary recall and the methods have previously been reported.24 Total whole grain intake (g/day) was the calculated sum of the absolute intake of the specific whole grain type (wheat, rye, oats and other grains).

Information about other lifestyle factors, such as smoking habits and physical activity was collected from a separate lifestyle questionnaire.17

Alkylresorcinol concentrations in plasma

Blood samples were drawn at baseline, and the participants were not fasting, but time since last meal was recorded. The samples were drawn and processed, and aliquoted at the study centers in Copenhagen or Aarhus, respectively, and thereafter stored under −150 °C under nitrogen vapor in a biobank until analysis.17

The concentrations of alkylresorcinol homologues (C17:0, C19:0, C21:0, C23:0 and C25:0) were determined in plasma (200 μl) from each participant by gas chromatography-mass spectrometry.25 The sum of the homologs as well as the C17:0/C21:0 ratio were used as exposures in the present study.

These laboratory analyses were done as part of a previous published study on colorectal cancer, of which the present study included the Danish participants.26 Thus, 50% of the participants were diagnosed with colorectal cancer during follow-up.

Definition of the study population analyzed

Of the 44 887 participants who participated in the follow-up study, 838 were excluded because of missing information on whole grain intake or covariates. Further, 156 participants were excluded due to extreme anthropometric measures (height <130 cm, BMI<16 kg/m2, WC<40 cm, WC>160 cm, bodyweight change>5 kg/year, change in WC >7 cm per year). Consequently, 43 893 participants (23 201 women and 20 692 men) were included in the present study, of which alkylresorcinol concentrations were previously measured in 516 participants (234 women and 282 men).

Statistical methods

The associations between whole grain intake or plasma concentrations of alkylresorcinols and changes in WC and body weight were estimated using multiple linear regression analyses. Changes in WC or body weight between baseline and follow-up (expressed as 5-year changes), respectively, were the dependent variable and whole grain intake or plasma concentrations of alkylresorcinols, respectively, were the independent variables. Logistic regression was used with 5-year increments in WC ⩾15 cm or 5-year gains⩾5 kg in body weight as outcomes (dichotomous variables), or with attained BMI at follow-up >25 kg/m2 as dichotomous outcome (sensitivity analysis among those with BMI<25 at baseline only, nwomen=12 257, nmen=7406).

When the association was expressed as linear increments, increments were 25 g/day or 50 g/day for total whole grain intake (g/day) and whole grain products (gram product/day), respectively and 25 g/day for whole grain product sub-groups and 10 g/day for whole grain subgroups.

Plasma concentrations of alkylresorcinols were investigated per increments of 50 nmol/l and the ratio of alkylresorcinol homologs C17:0 and C21:0 as increments of 0.1 unit.

For the multiple linear regression analyses, the dependent variables were tested for normal distribution, and no deviations were found. Furthermore, all continuous independent variables were tested for linearity using linear splines. The assumption of linearity was made by visual inspection of the plots, and F-tests. No significant departures from linearity were found.

All models, except those with alkylresorcinols or the alkylresorcinol C17:0/C21:0 ratio as exposure, were adjusted for energy intake using the residual method.27 To ease interpretability, sex-specific means were added to the residuals to recuperate the original scale.28 In model 1, only age at baseline and total energy intake were added as independent variables. In model 2, the following variables were further added: smoking status (current, former, never), participation in sports activities (yes/no) and schooling (short <8y, medium 8–10 y, long >10y). An additional model (model 3) was made where the following dietary variables were added to the model, intakes of processed meat (g/day), dairy products (g/day), fruits (g/day), vegetables (g/day), protein (g/day) and saturated fat (g/day). In model 4, further adjustment was made for baseline BMI (kg/m2), and for the analyses of WC also baseline WC (cm). When the alkylresorcinol homolog C17:0/C21:0 ratio was the independent variable, the concentration of total alkylresorcinols was further included as independent variable in the model. In analyses of subgroups of total whole grain intake (wheat, rye, and oats) and whole grain product intake (rye bread, whole grain bread, rolled oats and muesli and crispbread), mutual adjustments were made.

Correlations (Spearman’s) between plasma concentrations of alkylresorcinols and the alkylresorciniol C17/C21:0 ratio and whole grain variables were estimated. We conducted a sensitivity analysis, restricting our analyses to women who were postmenopausal and never users of HRT at baseline.

Possible interaction by BMI of the association between whole grain intake and risk of gaining ⩾5 kg were investigated by making a composite exposure variable by combining whole grain intake (in quartiles) and BMI categories (BMI<25, BMI=25–30, BMI>30). Furthermore, an interaction between whole grain as linear variable and BMI groups was estimated, and difference in slopes was tested using the Wald’s test.

Alkylresorcinol measurement were available on a subset from a previous study on colorectal cancer, and thus ~50% of them were diagnosed with colorectal cancer during follow-up. A t-test was used to test if there was difference in whole grain intake or plasma concentrations between cases and controls, but no statistical significant differences were found (all P>0.62). Thus, all participants of the subset (n=516) were included in the present study.

The statistical analyses were performed using SAS software version 9.3 (SAS Institute, Cary, NC, USA). The univariate and freq procedures were used for the descriptive statistics and the proc ttest was used for the t-test. The glm and logistic procedures were used for the multiple linear regression and logistic regression analyses, respectively. The proc corr was used for the Spearman’s correlations.

Results

Of the 57 053 cohort participants of the Diet, Cancer and Health cohort, information on follow-up body weight and WC as well as covariates were available for 43 893 of the participants (77%) who were included in the present study. The median time from baseline to follow-up was 5.3 years. During follow-up, participants marginally increased WC, whereas a small reduction in body weight was observed (Table 1). Whole grain intake was moderately correlated (correlation coefficients 0.23–0.32) with plasma concentrations of alkylresorcinols (Supplementary Table 1).

Table 1 Characteristics of the study population: Diet, Cancer and Health cohort, Denmark
Full size table

Waist circumference (WC)

For women and men, total whole grain intake was associated with a higher 5-year WC increase, but the association was no longer significant when adjusting for baseline BMI and WC (Table 2). Total whole grain product intake was not associated with 5-year WC change for women. For men it was related to 0.24 cm higher 5-year WC change (ΔWCmodel3 per 50 g increment: 0.24 cm, 95%CI: 0.18 cm; 0.31 cm), but it no longer was significant after adjustment of the baseline BMI and WC. Of the subgroups of whole grain intake, rye (women) and crispbread (both sexes) were related to gains in WC.

Table 2 Difference and 95% confidence interval of 5-year changes in waist circumference (cm) according to intakes of whole grain products and whole grains in grams per day as well as subgroups using energy residualsa—the Danish Diet, Cancer and Health Cohort
Full size table

Odds of a 5-year ⩾15 cm increment in WC was also investigated and a significant association was found for crispbread only (Table 3). A 25 g increase in crispbread was related to an 18 and 51% higher risk for women and men, respectively (ORwomen, model4: 1.18, 95% CI: 1.07, 1.31; ORmen, model4: 1.51, 95% CI: 1.11, 2.03). Quartiles estimates are shown in Supplementary Table 2.

Table 3 Odds ratios (OR) of 5-year increments in waist circumference ⩾15 cm according to intakes of whole grain products and whole grains in grams per day as well as subgroups by using energy residualsa—the Danish Diet, Cancer and Health cohort
Full size table

Body weight

For women, total whole grain intake and whole grain product intake was associated with a small 5-year body weight increase, but was no longer significant after adjustment for baseline BMI (Table 4). However, with the exception of crispbread and rye, which were consistently associated with higher 5-year body weight change. For men, total whole grain intake was associated with gains in 5-year body weight change (ΔWCmodel3 per 25 g increment: 150 g, 95% CI: 78 g; 222 g) but no association was found with adding baseline BMI to the model. Whole grain product intake was also associated with higher 5-year body weight change (ΔWCmodel3 per 50 g increment: 54 g, 95% CI: 8 g; 100 g) but adjusting for baseline BMI changed the association in the opposite direction (ΔWCmodel4 per 50 g increment: −59 g, 95% CI:−104 g; −13 g). All subgroups except crispbread changed association depending on whether or not adjustment was made for baseline BMI. Quartiles estimates are shown in Supplementary Table 3.

Table 4 Difference and 95% confidence intervals of 5-year changes in body weight (g) according to intakes of whole grain products and whole grains in grams per day as well as subgroups by using energy residualsa—the Danish Diet, Cancer and Health cohort
Full size table

Odds of a 5-year ⩾5 kg body weight gain was also investigated (Table 5). For women, total whole grain intake was associated with higher body weight gain (ORmodel4 per 50 g increment: 1.08, 95% CI: 1.01; 1.16) as was intake of rye and crispbread. For men, total whole grain intake and whole grain product intake was related to an 8 and 7% lower odds of body weight gain, respectively. All subgroups, except rye and crispbread, were also related to lower odds of weight gain. A sensitivity analyses were also made among those with a BMI <25 kg/m2 at baseline where odds of attained BMI>25 kg/m2 at follow-up was estimated, but similar results were found (Supplementary Table 4). Results of sensitivity analyses restricted to include only women who were postmenopausal and never users of HRT at baseline were similar to those where all women were included (data not shown).

Table 5 Odds ratios (OR) of 5-year body weight gain ⩾ 5 kg according to intakes of whole grain products and whole grains in grams per day as well as subgroups by using energy residualsa—the Danish Diet, Cancer and Health Cohort
Full size table

For men, no effect modification by baseline BMI was found. For women, total whole grain intake and whole grain products intake was associated with higher body weight gain among women who were obese or overweight at baseline only (Supplementary Table 5).

Alkylresorcinols

On a subset, plasma concentrations of the whole grain biomarkers alkylresorcinols were available (Table 6). Total plasma alkylresorcinol concentrations were not related to changes in body weight but they were inversely related to increments in WC in women (ΔWCmodel4 per 50 mol/l increment: −0.69 cm, 95% CI: −1.24 cm; −0.14 cm). For men, the alkylresorcinols C17.0/C21:0 ratio (marker of whole grain diet dominated by rye compared with wheat) was related to increments in WC (ΔWCmodel4 per 0.1 unit increment: 0.29 cm, 95% CI: 0.01 cm; 0.57 cm). The estimates did not change appreciably depending on whether adjustment for baseline WC and BMI was included in the model.

Table 6 Difference and 95% confidence interval of 5-year changes in waist circumference (cm) and 5-year changes in body weight (g) according to plasma concentrations of total alkylresorcinols or alkylresorcinol C17:0/C21:0 ratio — on a subset of the Danish Diet, Cancer and Health Cohort
Full size table

Discussion

In this Danish cohort study consisting of men and women with a high and varied whole grain intake, we found no strong association between whole grain intake and subsequent change in body weight or WC. For women, whole grain intake was related to higher gains in body weight, but the association was primarily found among overweight and obese women. Crispbread was related to gains in body weight for both sexes, and for women only, rye was related to higher gains in WC and body weight.

The Danish population is suitable for conducting research on whole grains and health due to the high and varied intake. Furthermore, we had detailed information on whole grain intake from both FFQ and 24HDR as well as from biomarkers, and thus measurement errors of the two measurements would be expected to be independent, and thus it is a strength that both measures were available (alkylresorcinols only available at subset). The plasma alkylresorcinol concentrations have been found to be moderately correlated with whole grain and cereal fiber intake, which is estimated from FFQ in this population (correlation coefficient~0.30).29, 30 The prospective nature of the study is also a strength as the whole grain intake was investigated in relation to subsequent changes in WC and body weight as opposed to baseline measures only (cross-sectional design).

The study is, however, not without limitations. Not all participants from the original Diet, Cancer and Health cohort participated in the follow-up study. We found that those participating in the follow-up study reported slightly higher whole grain intake at baseline than nonrespondents (mean=43 g/day versus 40 g/day, P for difference <0.0001), so selection bias cannot be ruled out. Intake of whole grain, estimated from FFQ/24HDR and alkylresorcinols (on a subset), was only available at baseline and changes in whole grain intake could therefore not be investigated. This is a weakness since the participants might have changed their diet during follow-up. Even though we had two independent measures of whole grain intake, each of these are prone to measurement errors. Furthermore, whole grain intake has previously been shown to be associated with a healthy lifestyle in this cohort,10 so even though potential confounders were carefully considered, residual confounding cannot be ruled out. Lastly, the anthropometric measurements at follow-up were self-reported, and the self-reported measures at follow-up were likely more prone to measurement errors.19 We tried to restrict our analyses to women who were postmenopausal and never users of HRT at baseline in order to account for any differences in body composition due to change in menopausal status, but no association was found for this restricted group either (data not shown).

For men, baseline adjustment seemed to change the direction of the associations in some cases. Previous studies of whole grains and body weight management in general adjust for baseline anthropometry,31, 32 but also typically present crude models with no large difference between crude and baseline anthropometry-adjusted models. However, whether baseline adjustment is the most suitable approach or not remains a controversial issue.33, 34 For comparability with results from previous studies, we showed both baseline anthropometry-adjusted and unadjusted analyses. When using the alkylresorcinols as measure of whole grain intake, no difference was found depending on whether or not adjustment was done. Furthermore, when investigating odds of major increment in WC (⩾15 cm) or major weight gain (⩾5 kg), baseline adjustment did not seem to change the results much. This suggests that the associations we find with changes in WC and weight are in fact just chance findings.

A recent meta-analysis of randomized controlled studies found no effect of whole grains on body weight or WC but found a small beneficial effect on body fat.9 This may imply that body fat measurements are more sensitive to assess small changes in anthropometry than our anthropometry measures. Moreover, whole grain intervention studies have typically not been designed to study weight loss. Previous cohort studies have, however, been able to detect associations based on the same type of anthropometric measurements as in the present study. Two large American cohort studies, that is, Nurses’ Health Study and Health Professionals Follow-up Study, both found that an increase in whole grain intake was inversely associated with body weight gain when adjusting for baseline anthropometry;31, 32 however, the magnitude of effects (<0.5 kg difference over 8–12 years follow-up) were relatively small. Previous Scandinavian cohort studies, of which one is the same as the present study, have investigated several foods in relation to changes in WC adjusted for baseline anthropometry and found no association for whole grains.35, 36 Rye intake (women), alkylresorcinols C17:0/C21:0 ratio for men (a measure of a rye-dominated diet) and crispbread (both sexes) were related to gains in both WC and body weight. This finding is surprising as rye has a high content of dietary fiber, and might increase satiety.37 The association for crispbread might be due to that this category included both whole grain and nonwhole grain types. Both rye bread (the major source of rye) and crispbread are often eaten with toppings such as cold cuts, cheese, jam, butter or fish. We adjusted the analyses for processed meat and dairy products to take into account that for example, meat cold cuts might be directly associated with weight gain, but we cannot rule out residual confounding.

The cohort is overrepresented by people with higher socioeconomic status,17 and is only representing middle-aged people. However, we do not expect that these factors may affect the generalizability of the findings.

In conclusion, whole grain intake was not strongly associated with weight or WC change over an average of 5 years. Furthermore, the very modest association found for men with weight change was largely dependent on whether or not adjustment was made for baseline BMI.