MDCS is a prospective cohort study that was conducted in the city of Malmö in Sweden (23). All men born between 1923 and 1945 and all women born between 1923 and 1950 in Malmö were invited via personal letters and advertisements in the local newspaper and public places during the baseline examination conducted in 1991–1996. Their age at baseline was between 44 and 74 years of age. The only exclusion criteria were mental incapacity and inadequate Swedish language skills. In total, 28,098 individuals (17,035 women and 11,063 men) provided complete dietary information and anthropometric measures, representing about 40% of the eligible individuals. All participants visited the research center on two occasions about 10 days apart. On the first visit, they received detailed instructions and information on the study, and the study material was given to them. They were all asked to fill in a menu booklet and two questionnaires: one with questions about their food habits and one with questions concerning lifestyle and socioeconomic aspects. At the first visit, nurses drew blood and anthropometrical measures were taken. On the second visit, the participants were interviewed by trained staff to complete the dietary history and to check the accuracy of the complete questionnaire. All individuals provided a written informed consent, and the ethics committee of Lund University approved the MDCS protocols.

In the present study, we have included 22,799 individuals (8,797 men and 14,002 women) who had DNA sample available and were genotyped for rs9939609, but who did not have a history of cardiovascular disease (n=954), cancer (n=1747), or diabetes (n=897) at the time of dietary data collection.

MDCS used a modified interview-based diet history method that was specially designed for the study (24) which included: 1) a 7-day menu book for registration of cooked meals and cold beverages; 2) a 168-item questionnaire for assessment of intake frequencies and portion sizes of regularly consumed foods not covered by the menu book; and 3) a 45-min interview including more questions about the cooking methods and the product food choices. The interviewer also very carefully controlled that the answers in the questionnaire and in the 7-day menu book did not overlap. The average daily food intake (g/day) was calculated based on the questionnaire and the 7-day menu book. To convert the food intake into energy and nutrient information the MDCS database was used. The majority of the nutrient information in the MDCS database was from the PCKOST2-93 from the National Food Administration in Uppsala, Sweden. A variable was created for the season of data collection; winter (December–February), spring (March–May), summer (June–August), and fall (September–November). The processing of data was slightly altered in September 1994, and the method version variable indicates if the data was collected before or after the 1 September 1994.

By taking the reported total energy intake and energy expenditure into account, we can identify participants that may have over- or underreported their energy intake. The individually estimated physical activity level (PAL) was expressed as total energy expenditure divided with the basal metabolic rate (BMR). The total energy expenditure was calculated for each individual from the self-reported amount of physical activity at work, leisure-time physical activity, hours of household work, estimated sleeping hours, self-care, and passive time. Non-adequate energy reporters were defined as those with a ratio of reported energy intake to BMR outside 95% confidence intervals (CI) of the calculated PAL (25).

The analyzed macronutrient variables were total energy (MJ), fiber density, and percentage energy (E%) from carbohydrates, protein, fat, sucrose, saturated fatty acids (SFAs), monounsaturated fatty acids (MUFA), polyunsaturated fatty acids (PUFAs), and alcohol. The 27 food groups (g/MJ) were defined mainly depending on fat and sugar content (Appendix 1).

Weight was measured using a balance-beam scale. The participants wore light clothes and no shoes. Height was measured with a fixed stadiometer calibrated in centimeters. BMI was calculated from the weight in kilograms divided by the height in square meters (kg/m²). For categorization of BMI, we used the WHO guidelines BMI ≤25 as normal weight and >25 as overweight or obese.

Genotyping of the rs9939609 was made either by matrix-assisted laser desorption ionization-time of flight mass spectrometry on the Sequenom MassARRAY platform (Sequenom, San Diego, CA, USA), or by Taqman (Applied Biosystems, Foster City, CA, USA). Genotyping was successful for 97.2% of the participants. The distribution of the rs9939609 was in Hardy–Weinberg equilibrium (p=0.99). The concordance rate was 0.9926.

For all the analyses we used Statistical Package of the Social Science (version 20.0; SPSS Inc., Chicago, IL, USA). The analyses were made for all individuals and for men and women separately because of a possible difference in food choices and meal patterns (26). Analyses were also performed in two BMI subgroups based on WHO guidelines, normal weight, and overweight (BMI ≤25 or >25 kg/m²). Two BMI groups where used to keep as much statistical power as possible, and because the median BMI in MDCS was approximately 25. We used the general linear model for continuous variables to test the differences across FTO genotypes. The 27 food variables were logarithmically transformed to get them normally distributed. A small value (0.001) was added before the transformation to handle zero-consumers. We also tested these variables with a non-parametric, Jonckeheere Terpsta test (27). No major differences were found for the p-values; therefore, the results from the parametric test were presented. In sensitivity analysis we excluded individuals classified as non-adequate energy reporters. To correct for multiple testing in the main analyses of food groups we used Bonferroni to calculate the cut-off value for significance level, 0.05/27=0.002. All analyses with dietary factors were adjusted for age, sex, season, and dietary method version. We also performed additional adjustment for BMI to exclude the possibility that the associations with food groups are secondary to association between FTO genotype and BMI. The interactions between the FTO genotype and gender and the FTO genotype and BMI status on intake of the food groups were assessed by introducing a multiplicative factor with the continuous variables.

BMI and reported intake of total energy and macronutrients by the FTO genotype (rs9939609) are shown as background characteristics in Table 1 for all 22,799 individuals in MDCS. As reported earlier in MDCS, FTO associated strongly with BMI (22). We found a significant difference for protein intake (p for trend=0.008), sucrose (p for trend=0.001), and total energy intake (p for trend <0.001) across the genotypes. Carriers of the A-allele had a lower reported total energy intake and consumption of sucrose, but similar to our previous observations a significantly larger part of the energy came from protein, as compared with TT carriers (6). Significant differences were also observed for non-adequate reporting of energy across the FTO genotypes (P for trend <0.001), the highest 17% frequency of under-reporting was observed among AA-genotype carriers as compared to 15.4% among TT genotype carriers. In addition, under-reporting of energy was more common in the overweight group than in the normal weight group (19.0 vs. 11.0%, p<0.0001).

After correcting for multiple testing, we observed significant differences by the FTO genotype (i.e. p<0.002) for reported intake of biscuits and pastry and soft drinks, where the A-allele carriers consumed more biscuits and pastry (0.07g/MJ/allele), but less soft drinks (−0.17g/MJ/allele) compared to TT-genotype carriers (Table 2). In addition, A-allele carriers reported nominally higher intakes of fruits, cereals, high fat meat, ice cream, and cheese, and a nominally lower intake of salty snacks (Table 2). The results were similar in men and women except for a significant interaction (p=0.003) for soft drinks; the trend was in the same direction, but stronger in women (Supplementary Table 2).

When we stratified for BMI, the trends across the FTO genotypes were similar except for the consumption of juice (p-interaction=0.04), which indicated a nominally significant increased intake for the A-allele carriers among overweight individuals but not among the normal weight individuals (Table 3).

In the sensitivity analysis only including individuals classified as adequate energy reporters, our results for food intakes across FTO genotypes remained virtually unchanged, except for intakes of fruits and salty snacks, which did not remain nominally significant. An additional adjustment for BMI did not change any of the reported results.