AJCN. First published ahead of print April 1, 2015 as doi: 10.3945/ajcn.114.103010.

Food sources of fat may clarify the inconsistent role of dietary fat intake for incidence of type 2 diabetes1–4 Ulrika Ericson, Sophie Hellstrand, Louise Brunkwall, Christina-Alexandra Schulz, Emily Sonestedt, Peter Wallstro¨m, Bo Gullberg, Elisabet Wirfa¨lt, and Marju Orho-Melander ABSTRACT Background: Dietary fats could affect glucose metabolism and obesity development and, thereby, may have a crucial role in the cause of type 2 diabetes (T2D). Studies indicated that replacing saturated with unsaturated fats might be favorable, and plant foods might be a better choice than animal foods. Nevertheless, epidemiologic studies suggested that dairy foods are protective. Objective: We hypothesized that, by examining dietary fat and its food sources classified according to fat type and fat content, some clarification regarding the role of dietary fat in T2D incidence could be provided. Design: A total of 26,930 individuals (61% women), aged 45–74 y, from the Malmo¨ Diet and Cancer cohort were included in the study. Dietary data were collected by using a modified diet-history method. During 14 y of follow-up, 2860 incident T2D cases were identified. Results: Total intake of high-fat dairy products (regular-fat alternatives) was inversely associated with incident T2D (HR for highest compared with lowest quintiles: 0.77; 95% CI: 0.68, 0.87; P-trend , 0.001). Most robust inverse associations were seen for intakes of cream and high-fat fermented milk (P-trend , 0.01) and for cheese in women (P-trend = 0.02). High intake of low-fat dairy products was associated with increased risk, but this association disappeared when low- and high-fat dairy were mutually adjusted (P-trend = 0.18). Intakes of both high-fat meat (P-trend = 0.04) and low-fat meat (P-trend , 0.001) were associated with increased risk. Finally, we did not observe significant association between total dietary fat content and T2D (P-trend = 0.24), but intakes of saturated fatty acids with 4–10 carbons, lauric acid (12:0), and myristic acid (14:0) were associated with decreased risk (P-trend , 0.01). Conclusions: Decreased T2D risk at high intake of high- but not of low-fat dairy products suggests that dairy fat partly could have contributed to previously observed protective associations between dairy intake and T2D. Meat intake was associated with increased risk independently of the fat content. Am J Clin Nutr doi: 10. 3945/ajcn.114.103010. Keywords: cohort study, diet, dietary fats, food intake, type 2 diabetes

INTRODUCTION

The worldwide adaption of westernized energy-rich diets is considered an important contributor to the increasing prevalence of obesity and type 2 diabetes (T2D).5 These diets tend to be high in animal foods and low in unrefined plant foods, which generally

result in high intakes of fat, SFAs, linoleic acid (LA; 18:2n–6), and protein but lower in dietary fiber and several micronutrients. Because fat is energy dense, and fatty acids affect glucose metabolism, fat intake may have a crucial role in the development of T2D. Potential effects on gene expression, cell membrane function, lipid metabolism, and gut microbiota may also explain associations with T2D (1–4). Evidence from randomized lifestyle interventions indicated that reduced intakes of total and saturated fats, in combination with increased fiber intake and physical activity, prevent the development of T2D in individuals with impaired glucose tolerance (5, 6). However, associations between dietary fat and T2D from epidemiologic studies have been inconsistent (7), and the importance of dietary fat content and food sources of fat with regard to risk of T2D remains to be clarified. The replacement of dietary intakes of SFAs with PUFAs may, via various mechanisms, lead to improved insulin sensitivity (8), and epidemiologic studies have indicated that the replacement of foods high in SFAs with food sources of MUFAs and PUFAs could be favorable in the prevention of diabetes development (8). In addition, high blood concentrations of LA may counteract the development of hyperglycemia and T2D (9). In line with those findings, plant sources of fat were suggested to be a better choice than animal sources (10). Indeed, high intakes of red meat and meat products show positive associations with risk of T2D (11). Nevertheless, several epidemiologic studies indicated that high intake of dairy products may be protective (12). Effects of different dairy products or dairy components, including possible 1 From the Department of Clinical Sciences, Malmo¨, Diabetes and Cardiovascular Disease, Genetic Epidemiology (UE, SH, LB, C-AS, ES, and MO-M) and the Department of Clinical Sciences, Malmo¨, Nutritional Epidemiology, Lund University, Lund, Sweden (PW, BG, and EW). 2 Supported by the Swedish Research Council, the Region Ska˚ne, the Ska˚ne University Hospital, the Novo Nordic Foundation, and the Albert Pa˚hlsson Research Foundation. 3 Supplemental Tables 1 and 2 are available from the “Supplemental data” link in the online posting of the article and from the same link in the online table of contents at http://ajcn.nutrition.org. 4 Address correspondence to U Ericson, Clinical Research Centre, Building 60, Floor 13, Ska˚nes Universitetssjukhus in Malmo¨, Entrance 72, Jan Waldenstro¨ms gata 35, SE-205 02 Malmo¨, Sweden. E-mail: ulrika.ericson@ med.lu.se. 5 Abbreviations used: LA, linoleic acid; MDC, Malmo¨ Diet and Cancer; T2D, type 2 diabetes. Received November 12, 2014. Accepted for publication March 6, 2015. doi: 10.3945/ajcn.114.103010.

Am J Clin Nutr doi: 10.3945/ajcn.114.103010. Printed in USA. Ó 2015 American Society for Nutrition

Copyright (C) 2015 by the American Society for Nutrition

1 of 16

2 of 16

ERICSON ET AL.

beneficial effects of yogurt, cheese, and specific fatty acids, were proposed to lie behind these observations (13–15). In addition, intake of fatty fish (16) as well as intakes and blood concentrations of total n–3 PUFAs (17), a-linolenic acid (18), and longchain fish n–3 PUFA from foods (19) were, in some studies, inversely associated with T2D risk, whereas results from other studies did not indicate that fatty fish or n–3 PUFA have an important protective role in the cause of T2D (8, 18, 20). In this population-based prospective study of men and women from the MDC (Malmo¨ Diet and Cancer) cohort, we examined if dietary fat intake and, in particular, different types of fatty acids and food sources of fat classified according to fat type and fat content were associated with incidence of T2D. METHODS

Study population and data collection The MDC study is a population-based prospective cohort study in Malmo¨, which is a city in the south of Sweden. Baseline examinations were conducted between 1991 and 1996. All women born between 1923 and 1950 and all men born between 1923 and 1945 who were living in the city of Malmo¨ were invited to participate (n = 74,138). Details of the cohort and the recruitment procedures are described elsewhere (21). The only exclusion criteria were mental incapacity and inadequate Swedish language skills (eligible persons: n = 68,905). Participants filled out questionnaires that covered socioeconomic, lifestyle, and dietary factors, recorded meals, and underwent a diet-history interview. Anthropometric measures were conducted by nurses. Weight was measured by using a balance-beam scale with subjects wearing light clothing and no shoes. Standing height was measured by using a fixed stadiometer calibrated in centimeters. Waist circumference was measured midway between the lowest rib margin and iliac crest. Body composition was estimated by using a bioelectrical impedance analyzer (BIA 103,single-frequency analyzer; RJL Systems). The percentage of body fat was calculated by using an algorithm provided by the manufacturer. During the screening period, 28,098 participants (40% of eligible persons) completed all baseline examinations. Of nonparticipants, 49% did not reply to the invitation letter, 39% answered that they were not willing to take part, 7% died or moved before they had received an invitation, and 5% failed to complete all baseline examinations (21). MDC participants have been compared with participants in a mailed health survey in Malmo¨ with a higher participation rate (75%) with regard to subjective health, sociodemographic characteristics, and lifestyle (21). In the current study, we included 26,930 participants without diabetes at baseline. We excluded 1168 participants on the basis of self-reported diabetes diagnosis, self-reported diabetes medication, or information from medical data registries that indicated a date of diagnosis preceding the baseline examination date. The ethical committee at Lund University approved the study (LU 51–90), and participants gave their written informed consent. Dietary data Dietary data were collected once during the baseline period. The MDC study used an interview-based modified diet-history method that combined 1) a 7-d menu book for the recording of intakes from meals that varied from day to day (usually lunch and dinner meals), cold beverages, and nutrient supplements and

2) a 168-item questionnaire for the assessment of consumption frequencies and portion sizes of regularly eaten foods that were not covered by the menu book. Finally, 3) a 45-min interview completed the dietary assessment. The MDC method has been described in detail elsewhere (22, 23). Diet analyses were adjusted for a variable called the diet-method version because slightly altered coding routines of dietary data were introduced in September 1994 to shorten the interview time (from 60 to 45 min). This adjustment resulted in 2 slightly different method versions (before or after September 1994) without any major influence on the ranking of individuals (23). The relative validity of the MDC method was evaluated in the Malmo¨ Food study 1984–1985 by comparing the method with 18-d weighed-food records (24, 25). Pearson correlation coefficients, which were adjusted for total energy, between the reference method and MDC method were, in women and men, respectively, 0.69 and 0.64 for total fat, 0.68 and 0.56 for SFA, 0.66 and 0.59 for MUFA, 0.64 and 0.26 for PUFA, 0.68 and 0.23 for LA, 0.58 and 0.22 for a-linolenic acid (18:3n–3), 0.38 and 0.24 for EPA (20:5n–3), 0.40 and 0.37 for docosapentaenoic acid (22:5n–3), 0.27 and 0.20 for DHA (22:6n–3), 0.51 and 0.43 for low-fat meat, 0.80 and 0.40 for high-fat meat, 0.92 and 0.92 for low-fat milk, 0.75 and 0.76 for high-fat milk, and 0.59 and 0.47 for cheese (24, 25). The mean daily intake of foods was calculated on the basis of the frequency and portion-size estimates from the questionnaire and menu book. Food intake was converted to energy and nutrient intakes by using the MDC nutrient database whereby the majority of the nutrient information comes from PC-KOST2-93 from the National Food Agency in Uppsala, Sweden. Nutrient intakes from supplements were calculated on the basis of supplement consumption recorded in the menu book. Supplement consumption was converted into nutrient intakes by using the MDC supplement database (26). Dietary variables examined in this study are listed and described in Supplemental Table 1. Examined nutrient intakes were the sum from foods and supplements. Main food sources of fat were identified in the MDC cohort (27) and primarily grouped according to fat type and fat content. Some less-important fat sources were also examined to facilitate the interpretation of results regarding high-fat alternatives of the same types of foods. Total intake of high-fat dairy products was defined as the sum of portions of butter; regular-fat alternatives ($2.5% fat) of milk, yogurt, and sour milk; cream (.12% fat); and regular-fat cheese (.20% fat). Portions (instead of grams) were used to analyze the sum of dairy products with different water contents and usually consumed in different weights (e.g., cheese and milk). Standard portion sizes from the MDC study or National Food Agency in Sweden were used (28) as follows: milk and yogurt (200 g/portion), cheese (20 g/portion), cream (25 g/portion), ice cream (75 g/portion), and butter (7 g/portion). Energy-adjusted dietary intakes were obtained by regressing intakes on nonalcohol energy intake. Quintiles of nutrient and food residuals were used as exposure categories. If .20% of the individuals were zero consumers, they constituted the lowest intake category, and the higher categories were defined as quartiles in consumers. Diabetes case ascertainment We identified 2860 incident cases of T2D during 377,642 person-years of follow-up via at least one of 7 registries (90%) or

FOOD SOURCES OF FAT AND INCIDENT TYPE 2 DIABETES

at new screenings or examinations during follow-up (10%). The mean follow-up time was 14 y (63.9 SD). Subjects contributed person-time from date of enrollment until date of diabetes diagnosis, death, migration from Sweden, or end of follow-up (December 2009), whichever occurred first. During follow-up, 0.5% of subjects had migrated from Sweden. If available, we used information on the date of diagnosis from 2 registries prioritized in the following order: 1) the regional Diabetes 2000 registry of Scania (29) and 2) the Swedish National Diabetes Registry (30). These registries required a physician diagnosis according to established diagnosis criteria (fasting plasma glucose concentration $7.0 mmol/L or fasting whole blood concentration $6.1 mmol/L, measured at 2 different occasions). Individuals with at $2 glycated hemoglobin values .6.0% with the Swedish Mono-S standardization system (corresponding to 6.9% in the US National Glycohemoglobin Standardization Program and 52 mmol/mol with International Federation of Clinical Chemistry and Laboratory Medicine units) (31, 32) were categorized as diabetes cases in the Malmo¨ HbA1c Registry. In addition, cases were identified via 4 registries from the National Board of Health and Welfare in Sweden as follows: the Swedish National Inpatient Registry, the Swedish Hospitalbased outpatient care, the Cause-of-death Registry, and the Swedish Prescribed Drug Registry. Other variables Information on age was obtained from the personal identification number. Age was divided into 5-y categories. BMI (in kg/m2) was calculated from the direct measurement of weight and height. Leisure-time physical activity was assessed by asking participants to estimate the number of minutes per week they spent on 17 different activities. The duration was multiplied by an activity specific intensity coefficient, and an overall leisure-time physical activity score was created. The score was divided into sex-specific quintiles. The smoking status of participants was defined as current smokers (including irregular smokers), ex-smokers, and never smokers. The total consumption of alcohol was defined by a 4-category variable. Participants who reported zero consumption in the menu book and indicating no consumption of any type of alcohol during the previous year were categorized as zero reporters. Other category ranges were ,15 g alcohol/d for women and ,20 g/d for men (low), 15–30 g/d for women and 20–40 g/d for men (medium), and .30 g/d for women and .40 g/d for men (high). Participants were divided into 4 categories according to their highest level of education (#8, 9–10, or 11–13 y or university degree). Season was defined as the season of dietdata collection (winter, spring, summer, and fall). Dietary change in the past (yes or no) was based on the question “Have you substantially changed your eating habits because of illness or some other reasons?” Statistical analysis The SPSS statistical computer package (version 20.0; IBM Corp.) was used for all statistical analyses. All food variables were log transformed (e-log) to normalize the distribution before analysis. To handle log transformation for zero intakes, we added a very small amount (0.0001 g).

3 of 16

We examined baseline characteristics in cases and noncases of T2D and across intake quintiles of fat and its food sources by using the general linear model for continuous variables (adjusted for age and sex) and with the chi-square test for categorical variables. In a post hoc analysis, we used the general linear model to examine intakes of nondairy foods (meat, fish, potatoes, fruit, vegetables, cereal products, margarine, pastry, chocolate, and sugar-sweetened beverages) across intake quintiles of cream and high-fat fermented milk. We used Cox proportional hazards regression model to estimate HRs of diabetes incidence associated with quintiles of dietary intakes adjusted for energy intake by using the residual method. The first quintile was used as the reference. Years of follow-up was used as the underlying time variable. We used covariates obtained from baseline examinations. The basic model included adjustments for age (continuous), sex (when applicable), method version, season (categorical), and total energy intake (continuous). Our full multivariate model further included adjustments for the following categorical variables: leisure-time physical activity, smoking, alcohol intake, and education, and, finally, BMI as a continuous variable. Because associations between dietary fat and T2D may partly be mediated via BMI, we also performed analyses with an intermediate multivariate model without the inclusion of BMI. Covariates were identified from the literature and indicated potential confounding in the MDC cohort because of associations with incident T2D and dietary intakes. Missing values for variables were treated as separate categories. Analyses with additional adjustments for waist circumference or dietary change in the past were also performed. Finally, additional adjustments were made for possible dietary confounders, which previously showed associations with T2D are found, in the examined foods or central in the same dietary pattern (intake quintiles of protein, fiber, sucrose, calcium, vitamin D, magnesium, meat, fruit and vegetables, sugar-sweetened beverages, or high-fat dairy products). We also performed all analyses for men and women separately. Tests for interactions between sex and nutrient and food intakes with regard to diabetes incidence were performed [sex 3 quintile of nutrients and foods (treated as continuous variables)]. Tests for interactions between BMI (#25 or .25) and dietary variables were also performed. In a sensitivity analysis, we excluded individuals with a reported dietary change in the past (24% of the individuals). In a second sensitivity analysis, we excluded individuals with prevalent cardiovascular disease (coronary event or stroke) at baseline (3%). All statistical tests were 2 sided, and significance was assumed at P , 0.05.

RESULTS

Baseline characteristics At baseline, several established risk factors for T2D, as well as potential confounders of dietary associations, differed between cases and noncases of incident T2D (Table 1). Cases were older and had higher BMI, a more sedentary lifestyle, lower alcohol intake, higher protein intake, and lower intakes of carbohydrates and dietary fiber. In addition, there were more individuals who reported a dietary change in the past, more ever smokers, and fewer individuals with a high level of education in cases. Baseline characteristics differed also between low and high consumers of several fat sources (Table 2). Subjects who reported a high total

4 of 16

ERICSON ET AL. TABLE 1 Baseline characteristics for cases and noncases of incident T2D in the MDC cohort after the exclusion of individuals with prevalent diabetes at baseline (1991–1996)1 Baseline variable Sex, F, % Age, y BMI, kg/m2 Waist, cm Body fat, % Systolic blood pressure, mm Hg Diastolic blood pressure, mm Hg Hb A1c, % Fasting blood glucose, mmol/L Triglycerides, mmol/L HDL cholesterol, mmol/L LDL cholesterol, mmol/L Fasting plasma insulin, mU/L HOMA-IR Leisure-time physical activity score5 Alcohol intake,6 g/d Smoking, previous and current, % Education .10 y, % Dietary change in the past, % Energy, kcal/d Protein, E% Carbohydrates, E% Fat, E% Saturated fat, E% Fiber, g/1000 kcal Calcium, mg/d Dairy products, portions/d Dairy products, low-fat, portions/d Dairy products, high-fat, portions/d Margarine, g/d Egg, g/d Meat and meat products, g/d Fish, high-fat, g/d Pastry and biscuits, g/d Chocolate, g/d

n 26,930 26,930 26,894 26,885 26,772 26,892 26,890 5104 5104 5110 5062 4999 4931 4692 26,754 25,286 26,920 26,865 26,893 26,930 26,930 26,930 26,930 26,930 26,930 26,930 26,930 26,930 26,930 26,930 26,930 26,930 26,930 26,930 26,930

Cases (n = 2860) 58.7 28.4 93.0 28.3 147 89.1 5.20 5.91 1.74 1.23 4.29 12.9 2.42 7510 11.8

2320 15.9 45.9 38.3 16.3 9.0 1170 6.3 2.0 4.1 31 25 120 16.6 37 8.0

48.5 (58.5, 59.0)3 (28.2, 28.5) (92.6, 93.4) (28.1, 28.5) (146, 147) (88.7, 89.4) (5.16, 5.23) (5.86, 5.96) (1.68, 1.79) (1.21, 1.26) (4.22, 4.37) (12.3, 13.5) (2.31, 2.53) (7270, 7760) (11.4, 12.3) 64.8 25.5 27.4 (2300, 2340) (15.8, 16.0) (45.7, 46.2) (38.0, 38.5) (16.1, 16.4) (8.9, 9.1) (1150, 1180) (6.2, 6.5) (2.0, 2.1) (4.0, 4.3) (30, 32) (24, 26) (119, 122) (15.9, 17.4) (36, 39) (7.6, 8.5)

Noncases (n = 24,070)

P2

62.7 (58.0, 58.2) (25.4, 25.4) (84.4, 84.7) (25.3, 25.4) (140, 141) (85.4, 85.7) (4.74, 4.77) (4.86, 4.90) (1.29, 1.33) (1.37, 1.39) (4.11, 4.17) (6.89, 7.35) (1.50, 1.57) (8130, 8310) (12.3, 12.6) 61.6 32.9 21.8 (2340, 2360) (15.4, 15.5) (46.2, 46.4) (38.4, 38.6) (16.5, 16.6) (9.1, 9.2) (1140, 1160) (6.6, 6.7) (1.8, 1.8) (4.6, 4.7) (30, 31) (23, 24) (113, 114) (15.8, 16.4) (39, 40) (7.8, 8.2)

,0.001 ,0.001 ,0.001 ,0.001 ,0.001 ,0.001 ,0.001 ,0.001 ,0.001 ,0.001 ,0.001 ,0.001 ,0.001 ,0.0014 ,0.001 0.02 0.001 ,0.001 ,0.001 0.01 ,0.001 0.01 0.36 ,0.001 0.04 0.10 ,0.001 ,0.001 ,0.001 0.10 ,0.001 ,0.001 0.25 0.004 0.88

58.1 25.4 84.6 25.3 140 85.5 4.75 4.88 1.31 1.38 4.14 7.1 1.54 8220 12.4

2350 15.4 46.3 38.5 16.6 9.1 1150 6.6 1.8 4.6 31 23 114 16.1 39 8.0

1

E%, percentage of energy; Hb A1c, glycated hemoglobin; MDC, Malmo¨ Diet and Cancer; T2D, type 2 diabetes. A general linear model was used for continuous variables and adjusted for age and sex. The examination of diet was also adjusted for the diet-method version, season, and energy intake. The chi-square test was used for categorical variables. 3 Mean; 95% CI in parentheses (all such values). 4 P value for ln-transformed values. 5 A high score indicates a high level of leisure-time physical activity. 6 In subjects who reported that they consumed alcohol during the year before baseline examinations. 2

dietary fat content were younger and had lower BMI, but apart from these variables, they were characterized by a rather unhealthy lifestyle pattern; they had a more-sedentary lifestyle and higher alcohol intake, and there were also more ever smokers and fewer individuals with a high level of education in subjects who reported a high dietary fat content. Finally, fewer of these individuals reported a dietary change in the past. Except for the observation regarding education, a similar pattern was seen for individuals with a diet rich in high-fat dairy products. Dietary content of total fat and fatty acids in relation to incidence of T2D We did not observe any significant associations between the dietary content of total fat and incidence of T2D (P-trend = 0.24) (Table 3). In the full multivariate analysis, we observed a significant

inverse association between intake of SFA and T2D (P-trend = 0.01). However, the association disappeared after adjustment for intake of high-fat dairy products (P-trend = 0.61). Moreover, in analyses of SFAs with different chain lengths, we only observed significant decreased risk of T2D at high aggregated intakes of short- to medium-chain SFAs with 4–10 carbons (P-trend , 0.001) as well as at high intakes of lauric acid (12:0) (P-trend = 0.003) and myristic acid (14:0) (P-trend , 0.001). In contrast, high intakes of SFAs with a longer chain length, palmitic acid (16:0) (P-trend = 0.10) and stearic acid (18:0) (P-trend = 0.36), were not associated with T2D. Intakes of MUFAs and PUFAs were not significantly associated with T2D in the full multivariate analysis. Except for an interaction between n–3 PUFA intake and sex (P = 0.046), we did not detect any significant interactions between fat intakes and sex. Men in the highest intake quintile of n–3 PUFAs tended to be at decreased risk (HR: 0.87; 95% CI: 0.74, 1.02; P = 0.08), whereas

26,894 25.8 (25.7, 25.9) 25.4 (25.3, 25.5) ,0.001 25.3 (25.2, 25.4) 26.3 (26.2, 26.4) ,0.001 26.1 (26.0, 26.2) 25.3 (25.2, 25.4) ,0.001 25.4 (25.3, 25.5) 25.8 (25.7, 25.9) ,0.001 25.4 (25.3, 25.5) 26.2 (26.1, 26.3) ,0.001 25.9 (24.9, 25.0) 26.4 (26.2, 26.5) ,0.001 25.8 (25.7, 25.8) 25.9 (25.8, 25.9) 0.001 25.8 (25.7, 25.9) 25.6 (25.5, 25.7) 0.01 25.9 (25.8, 26.0) 25.6 (25.5, 25.8) 0.04

58.5 (58.3, 58.7)3 58.1 (57.9, 58.3) ,0.001

57.7 (57.5, 57.9) 58.5 (58.3, 58.7) ,0.001

59.2 (59.0, 59.4) 57.1 (56.9, 57.3) ,0.001

58.1 (57.9, 58.3) 58.9 (58.7, 59.1) ,0.001

57.5 (57.3, 57.7) 58.5 (58.3, 58.7) ,0.001

58.9 (58.7, 59.1) 56.7 (56.5, 56.9) ,0.001

56.7 (56.6, 57.0) 59.9 (59.7, 60.1) ,0.001

56.0 (55.8, 56.2) 61.4 (61.1, 61.6) ,0.001

58.4 (58.2, 58.6) 58.3 (58.0, 58.5) 0.72

BMI, kg/m2

26,930

Age, y

8000 (7800, 8200) 8000 (7800, 8100) 0.67

8100 (7900, 8300) 8100 (7900, 8300) 0.48

7800 (7600, 8000) 8200 (8000, 8400) ,0.001

8900 (8700, 9100) 7500 (7300, 7700) ,0.001

8100 (7900, 8300) 8100 (7900, 8300) 0.90

8200 (8000, 8500) 7600 (7400, 7800) ,0.001

8200 (8100, 8400) 8000 (7800, 8200) 0.04

7800 (7600, 8000) 8600 (8400, 8700) ,0.001

9000 (8800, 9200) 7400 (7300, 7600) ,0.001

26,754

Leisure-time physical activity score

13 (13, 13) 13 (12, 13) 0.37

15 (15,16) 10 (10, 11) ,0.001

10 (10, 11) 15 (14, 15) ,0.001

11 (11, 11) 15 (14, 15) ,0.001

11 (11, 12) 14 (14, 14) ,0.001

13 (13, 14) 11 (11, 11) ,0.001

11 (10, 11) 14 (13, 14) ,0.001

14 (13, 14) 11 (10, 11) ,0.001

10 (10, 11) 15 (14, 15) ,0.001

25,286

Alcohol intake,2 g/d

52.7 76.6 ,0.001

48.4 73.0 ,0.001

56.9 64.7 0.05

74.8 45.2 ,0.001

57.9 65.2 ,0.001

59.3 50.2 ,0.001

51.1 68.6 ,0.001

52.8 64.5 ,0.001

61.3 60.5 0.41

26,930

Sex, F, %

68.3 62.8 ,0.001

75.1 50.7 ,0.001

64.9 61.5 ,0.001

57.0 70.4 ,0.001

60.7 65.2 ,0.001

64.8 65.2 0.70

61.3 65.9 ,0.001

67.7 60.8 ,0.001

55.3 70.7 ,0.001

26,920

Smoking, ex/current, %

29.9 31.9 0.02

37.4 24.7 ,0.001

33.2 32.2 0.85

40.5 28.0 ,0.001

34.4 32.0 0.01

35.2 24.4 ,0.001

24.6 34.1 ,0.001

31.2 31.8 0.56

34.1 30.6 ,0.001

26,865

Education, .10 y, %

27.9 20.5 ,0.001

25.1 21.9 ,0.001

22.3 26.9 ,0.001

28.9 20.7 ,0.001

25.1 23.7 0.11

22.2 22.0 0.77

34.1 16.3 ,0.001

17.7 32.6 ,0.001

36.1 14.6 ,0.001

26,893

Dietary change in the past, %

A general linear model was used for continuous variables and adjusted for age, sex, diet-method version, and season. The chi-square test was used for categorical variables. Quintiles for dietary intakes were adjusted for energy by using the residual method. P-trend values were calculated across quintiles for continuous variables. In addition, P values were calculated for the comparison of percentages in highest and lowest quintiles for categorical variables. MDC, Malmo¨ Diet and Cancer; E%, percentage of energy. 2 In subjects who reported that they consumed alcohol during the year before baseline examinations. 3 Mean; 95% CI in parentheses (all such values). 4 Zero consumers; higher categories are quartiles in consumers.

1

n Fat (E%) Quintile 1 (31) Quintile 5 (46) P-trend Dairy products, low-fat (portions) Quintile 1 (0.1) Quintile 5 (4) P-trend Dairy products, high-fat (portions) Quintile 1 (0.9) Quintile 5 (8.3) P-trend Margarine (g) Quintile 1 (5) Quintile 5 (59) P-trend Eggs (g) Quintile 1 (4) Quintile 5 (45) P-trend Meat and meat products (g) Quintile 1 (55) Quintile 5 (163) P-trend Fish, high-fat (g) 04 (0) 4 (46) P-trend Pastry and biscuits (g) Quintile 1 (6) Quintile 5 (72) P-trend Chocolate (g) Quintile 1 (0) Quintile 5 (16) P-trend

Dietary intake quintile (median intake/d)

TABLE 2 Baseline characteristics in quintiles of energy-adjusted dietary intakes of fat and food sources of fat in individuals without prevalent diabetes from the MDC cohort (1991–1996)1

FOOD SOURCES OF FAT AND INCIDENT TYPE 2 DIABETES

5 of 16

6 of 16

ERICSON ET AL.

TABLE 3 HRs (95% CIs) of incident T2D associated with intakes of total fat and different fatty acids in the MDC cohort1 Nutrient quintile (median intake) Fat (E%) 1 (31) 2 (35) 3 (38) 4 (41) 5 (46) P-trend Saturated fat (E%) 1 (12) 2 (14) 3 (16) 4 (18) 5 (22) P-trend Fatty acids 4:0–10:0 (E%) 1 (0.7) 2 (1.0) 3 (1.3) 4 (1.7) 5 (2.6) P-trend Lauric acid (12:0) (E%) 1 (0.6) 2 (0.8) 3 (1.0) 4 (1.1) 5 (1.4) P-trend Myristic acid (14:0) (E%) 1 (1.1) 2 (1.4) 3 (1.7) 4 (2.0) 5 (2.7) P-trend Palmitic acid (16:0) (E%) 1 (6) 2 (7) 3 (8) 4 (9) 5 (10) P-trend Stearic acid (18:0) (E%) 1 (2.7) 2 (3.3) 3 (3.6) 4 (4.0) 5 (4.5) P-trend MUFAs (E%) 1 (11) 2 (12) 3 (13) 4 (14) 5 (16) P-trend PUFAs (E%) 1 (4) 2 (5) 3 (6)

n cases/ person-years

Basic model2

Multivariate model without BMI3

Full multivariate model with BMI4

P-interaction with sex 0.59

590/76,508 598/75,642 565/75,519 550/75,703 557/74,269 —

1.03 0.98 0.97 0.99

1.00 (0.92, 1.16) (0.87, 1.10) (0.86, 1.09) (0.88, 1.11) 0.52

1.00 1.02 (0.91, 1.14) 0.97(0.86, 1.09) 0.93 (0.83, 1.05) 0.93 (0.82, 1.04) 0.08

1.00 0.95 0.93 0.96

1.00 (0.88, 1.12) (0.85, 1.07) (0.83, 1.05) (0.85, 1.08) 0.24

626/76,561 663/75,374 542/76,008 534/75,327 495/74,372 —

1.11 0.92 0.94 0.89

1.00 (1.00, 1.24) (0.82, 1.04) (0.83, 1.05) (0.79, 1.00) 0.002

1.00 1.11 (0.99, 1.24) 0.92 (0.82, 1.03) 0.91 (0.81, 1.03) 0.85 (0.75, 0.96) ,0.001

1.07 0.91 0.93 0.91

1.00 (0.96, 1.19) (0.81, 1.02) (0.82, 1.04) (0.81, 1.02) 0.01

731/75,372 628/76,036 542/76,029 493/75,646 466/74,558 —

1.00 0.90 (0.81, 1.00) 0.82 (0.73, 0.91) 0.76 (0.68, 0.85) 0.72 (0.64, 0.82) ,0.001

1.00 0.93 (0.83, 1.03) 0.86 (0.76, 0.96) 0.79 (0.71, 0.89) 0.74 (0.66, 0.83) ,0.001

1.00 0.92 (0.83, 1.03) 0.88 (0.78, 0.98) 0.84 (0.75, 0.95) 0.83 (0.74, 0.93) ,0.001

650/75,604 608/75,082 557/75,907 560/75,699 485/75,349 —

1.00 0.97 (0.86, 1.08) 0.88 (0.78, 0.98) 0.89 (0.80, 1.00) 0.80 (0.71, 0.90) ,0.001

1.00 0.95 (0.84, 1.06) 0.87 (0.78, 0.97) 0.88 (0.78, 0.98) 0.76 (0.67, 0.86) ,0.001

0.98 0.92 0.92 0.84

1.00 (0.87, 1.09) (0.82, 1.03) (0.82, 1.03) (0.75, 0.95) 0.003

701/75,680 644/75,851 540/75,639 501/75,649 474/74,823 —

1.00 0.96 (0.86, 1.07) 0.84 (0.75, 0.94) 0.79 (0.70, 0.89) 0.76 (0.68, 0.86) ,0.001

1.00 0.98 (0.88, 1.09) 0.86 (0.77, 0.96) 0.81 (0.72, 0.91) 0.76 (0.68, 0.86) ,0.001

1.00 0.98 (0.88, 1.08) 0.86 (0.76, 0.96) 0.87 (0.77, 0.98) 0.83 (0.74, 0.94) ,0.001

614/76,720 619/75,884 556/75,597 559/75,104 512/74,336 —

1.05 0.97 1.00 0.92

1.00 (0.94, 1.18) (0.86, 1.09) (0.89, 1.12) (0.82, 1.03) 0.09

1.06 0.96 0.97 0.87

1.00 (0.95, 1.18) (0.85, 1.08) (0.87, 1.09) (0.78, 0.98) 0.01

1.03 0.93 0.97 0.92

1.00 (0.92, 1.15) (0.82, 1.04) (0.87, 1.09) (0.81, 1.03) 0.10

0.95 1.01 1.00 1.01

1.00 (0.85, 1.07) (0.90, 1.14) (0.89, 1.12) (0.90, 1.14) 0.60

0.94 0.99 0.95 0.92

1.00 (0.84, 1.06) (0.88, 1.11) (0.85, 1.07) (0.82, 1.04) 0.28

0.93 0.96 0.94 0.94

1.00 (0.83, 1.05) (0.86, 1.08) (0.84, 1.06) (0.83, 1.05) 0.36

545/76,597 561/76,057 555/75,419 595/75,276 604/74,292 —

1.04 1.03 1.10 1.10

1.00 (0.92, 1.17) (0.91, 1.16) (0.98, 1.23) (0.98, 1.24) 0.06

1.03 1.01 1.05 1.02

1.00 (0.91, 1.16) (0.89, 1.13) (0.94, 1.18) (0.91, 1.15) 0.62

0.99 0.98 1.03 1.01

1.00 (0.88, 1.12) (0.87, 1.10) (0.91, 1.16) (0.89, 1.13) 0.72

496/74,893 570/74,850 573/76,267

1.00 1.13 (1.00, 1.28) 1.10 (0.97, 1.24)

0.36

0.37

0.41

0.33

0.92

0.31 593/76,505 556/76,074 580/75,377 564/75,284 567/74,401 —

0.75

0.62 1.00 1.12 (0.99, 1.26) 1.09 (0.96, 1.23)

1.00 1.08 (0.96, 1.22) 1.04 (0.92, 1.17) (Continued)

7 of 16

FOOD SOURCES OF FAT AND INCIDENT TYPE 2 DIABETES TABLE 3 (Continued ) Nutrient quintile (median intake)

n cases/ person-years

Basic model2

Multivariate model without BMI3

Full multivariate model with BMI4

4 (7) 5 (8) P-trend Total n–3 PUFAs (E%) 1 (0.7) 2 (0.8) 3 (0.9) 4 (1.1) 5 (1.4) P-trend ALA (E%) 1 (0.5) 2 (0.6) 3 (0.7) 4 (0.8) 5 (1.0) P-trend Long-chain n–3 PUFAs (E%) 1 (0.07) 2 (0.12) 3 (0.19) 4 (0.29) 5 (0.52) P-trend Total n–6 PUFAs (E%) 1 (3.2) 2 (4.0) 3 (4.7) 4 (5.5) 5 (6.8) P-trend Ratio n–3:n–6 1 (0.14) 2 (0.17) 3 (0.19) 4 (0.23) 5 (0.30) P-trend Ratio ALA:LA 1 (0.11) 2 (0.14) 3 (0.15) 4 (0.17) 5 (0.21) P-trend

600/75,952 621/75,679 —

1.14 (1.01, 1.29) 1.17 (1.04, 1.32) 0.02

1.13 (1.00, 1.28) 1.13 (1.00, 1.28) 0.07

1.08 (0.96, 1.22) 1.07 (0.95, 1.20) 0.37

570/75,798 533/76,093 550/76,008 575/75,111 632/74,633 —

0.92 0.93 0.95 1.02

1.00 (0.81, (0.82, (0.85, (0.91, 0.47

0.92 0.92 0.95 1.03

1.00 (0.82, (0.81, (0.84, (0.92, 0.43

0.90 0.91 0.93 1.00

1.00 (0.80, (0.81, (0.83, (0.89, 0.80

0.89 0.96 0.88 0.96

1.00 (0.79, (0.85, (0.79, (0.86, 0.49

0.88 0.93 0.86 0.92

1.00 (0.78, (0.83, (0.76, (0.82, 0.12

0.85 0.94 0.85 0.94

1.00 (0.76, (0.84, (0.76, (0.83, 0.31

1.05 1.05 0.96 1.12

1.00 (0.93, (0.93, (0.85, (0.99, 0.29

1.06 1.07 1.01 1.18

1.00 (0.94, (0.95, (0.90, (1.05, 0.03

1.01 0.99 0.92 1.07

1.00 (0.90, (0.88, (0.81, (0.94, 0.72

1.18 1.14 1.17 1.18

1.00 (1.05, (1.01, (1.04, (1.05, 0.02

1.17 1.13 1.16 1.15

1.00 (1.04, (1.00, (1.03, (1.02, 0.07

1.13 1.07 1.11 1.09

1.00 (1.00, (0.95, (0.98, (0.97, 0.28

0.95 1.00 0.96 0.91

1.00 (0.84, (0.89, (0.86, (0.81, 0.22

0.95 1.02 0.99 0.94

1.00 (0.85, (0.91, (0.88, (0.84, 0.56

0.90 1.00 0.98 0.91

1.00 (0.80, (0.90, (0.87, (0.81, 0.46

0.93 0.95 1.02 0.81

1.00 (0.83, (0.84, (0.91, (0.72, 0.02

0.93 0.94 1.03 0.80

1.00 (0.93, (0.84, (0.92, (0.71, 0.02

0.91 0.93 1.03 0.86

1.00 (0.81, (0.83, (0.93, (0.76, 0.26

P-interaction with sex

0.046 1.03) 1.04) 1.07) 1.15)

1.03) 1.03) 1.07) 1.16)

1.02) 1.02) 1.05) 1.12) 0.84

606/75,539 548/76,483 581/75,467 542/75,365 583/74,788 —

1.00) 1.07) 0.99) 1.08)

0.99) 1.04) 0.96) 1.03)

0.95) 1.05) 0.95) 1.05) 0.10

519/76,234 565/75,565 577/75,121 550/75,920 649/74,802 —

1.18) 1.18) 1.09) 1.26)

1.19) 1.20) 1.14) 1.33)

1.14) 1.12) 1.04) 1.20) 0.93

488/74,326 582/74,845 577/75,958 600/76,514 613/75,998 —

1.33) 1.28) 1.32) 1.34)

1.32) 1.28) 1.31) 1.29)

1.28) 1.21) 1.25) 1.23) 0.41

595/76,912 565/76,253 587/75,424 569/75,140 544/73,913 —

1.06) 1.12) 1.08) 1.03)

1.07) 1.14) 1.12) 1.06)

1.01) 1.13) 1.10) 1.03) 0.65

631/77,233 576/76,389 577/75,353 595/74,562 481/74,105 —

1.04) 1.06) 1.14) 0.92)

1.04) 1.06) 1.16) 0.91)

1.02) 1.04) 1.16) 0.97)

HRs were calculated by using a Cox proportional hazards model. ALA, a-linolenic acid; E%, percentage of energy; LA, linoleic acid; MDC, Malmo¨ Diet and Cancer; T2D, type 2 diabetes. 2 Adjusted for age (continuous), sex (when applicable), method version (categorical), season (categorical) and total energy intake (continuous). 3 Adjusted as for the basic model and for the following categorical variables: leisure-time physical activity, smoking, alcohol intake, and education. 4 Adjusted as for the basic model and for the following categorical variables: leisure-time physical activity, smoking, alcohol intake, education, and BMI (continuous). 1

the findings for women with highest intakes were rather in the opposite direction (HR: 1.14; 95% CI: 0.97, 1.35; P = 0.12). However, no significant trends across quintiles were seen in men or women (P-trend $ 0.12). Food sources of fat and incidence of T2D We did not observe any significant association between total intake of dairy products (i.e., high fat and low fat) but a lower

incidence of T2D with higher total intake of high-fat dairy products (P-trend , 0.001) (Table 4), and similar protective associations were seen for both fermented (P-trend = 0.01) and nonfermented (P-trend , 0.001) high-fat dairy products. Decreased risk of T2D was seen with higher intakes of cream (Ptrend = 0.001), butter (P-trend = 0.001), and high-fat fermented milk (P-trend = 0.007) as well as higher intake of cheese in women (P-trend = 0.02, P-interaction with sex = 0.01).

8 of 16

ERICSON ET AL.

TABLE 4 HRs (95% CIs) of incident T2D associated with intake of food sources of fat in the MDC cohort1 Nutrient quintile (median intake/d) Dairy products, 1 (3) 2 (4) 3 (5) 4 (7) 5 (10) P-trend Dairy products, 1 (0.1) 2 (0.7) 3 (1.5) 4 (2.4) 5 (4.0) P-trend Dairy products, (portions) 1 (0.02) 2 (0.3) 3 (0.9) 4 (1.6) 5 (3.0) P-trend Dairy products, (portions) 05 (0) 1 (0.2) 2 (0.5) 3 (1.1) 4 (2.4) P-trend Dairy products, 1 (0.9) 2 (2.3) 3 (3.3) 4 (5.0) 5 (8.3) P-trend Dairy products, (portions) 1 (0.1) 2 (0.4) 3 (0.9) 4 (2.4) 5 (5.8) P-trend Dairy products, (portions) 1 (0.3) 2 (1.1) 3 (1.7) 4 (2.4) 5 (3.6) P-trend Milk, total (g) 1 (71) 2 (221) 3 (331) 4 (450) 5 (633) P-trend

n cases/ person-years

Basic model2

Multivariate model without BMI3

Full multivariate model with BMI4

total (portions)

P-interaction with sex 0.09

641/4745 601/4785 596/4790 539/4847 483/4903 —

1.00 (0.90, 1.12) (0.94, 1.17) (0.87, 1.09) (0.77, 0.98) 0.03

1.00 1.05 0.97 0.87

541/4845 536/4850 494/4892 606/4780 683/4703 —

1.00 1.00 (0.89, 1.13) 0.92 (0.92, 1.04) 1.14 (1.02, 1.28) 1.29 (1.15, 1.45) ,0.001

1.04 1.10 1.03 0.90

1.00 (0.93, 1.16) (0.97, 1.23) (0.92, 1.16) (0.79, 1.01) 0.13

1.00 1.04 0.99 0.90

1.00 (0.89, 1.12) (0.93, 1.16) (0.88, 1.11) (0.80, 1.02) 0.14

1.00 0.93 1.08 1.14

1.00 (0.89, 1.13) (0.82, 1.05) (0.96, 1.22) (1.01, 1.28) 0.01

low-fat (portions)

0.44 1.00 1.04 (0.92, 1.17) 0.97 (0.86, 1.10) 1.19 (1.06, 1.34) 1.34 (1.20, 1.51) ,0.001

low-fat, nonfermented

0.11 557/4829 508/4878 500/4886 579/4807 716/4670 —

0.92 0.90 1.04 1.32

1.00 (0.81, 1.03) (0.79, 1.01) (0.92, 1.17) (1.18, 1.48) ,0.001

0.94 0.93 1.07 1.31

1.00 (0.84, 1.06) (0.82, 1.05) (0.95, 1.20) (1.17, 1.46) ,0.001

0.93 0.92 0.98 1.12

1.00 (0.82, 1.05) (0.81, 1.04) (0.87, 1.10) (1.00, 1.25) 0.02

low-fat, fermented

0.78 1249/10,478 398/3402 403/3398 385/3416 425/3376 —

1.08 1.06 0.98 1.06

1.00 (0.96, 1.21) (0.94, 1.19) (0.87, 1.10) (0.94, 1.18) 0.56

1.13 1.13 1.05 1.15

1.00 (1.00, 1.26) (1.01, 1.27) (0.94, 1.18) (1.03, 1.29) 0.02

1.07 1.09 0.99 1.06

1.00 (0.96, 1.20) (0.98, 1.22) (0.88, 1.12) (0.95, 1.18) 0.42

high-fat (portions)

0.15 739/4657 611/4775 552/4834 514/4872 444/4942 —

1.00 0.86 (0.77, 0.95) 0.82 (0.74, 0.92) 0.78 (0.70, 0.88) 0.69 (0.61, 0.77) ,0.001

1.00 0.88 (0.79, 0.98) 0.86 (0.77, 0.96) 0.82 (0.73, 0.92) 0.69 (0.62, 0.78) ,0.001

1.00 0.88 (0.79, 0.98) 0.89 (0.79, 0.99) 0.89 (0.79, 1.00) 0.77 (0.68, 0.87) ,0.001

high-fat, nonfermented

0.60 716/4670 622/4764 528/4858 534/4852 460/4926 —

0.91 0.78 0.78 0.71

1.00 (0.82, 1.02) (0.69, 0.87) (0.70, 0.88) (0.63, 0.79) ,0.001

0.95 0.82 0.80 0.70

1.00 (0.85, 1.06) (0.73, 0.91) (0.72, 0.90) (0.62, 0.79) ,0.001

0.99 0.88 0.88 0.80

1.00 (0.89, 1.10) (0.78, 0.98) (0.79, 0.99) (0.71, 0.90) ,0.001

high-fat, fermented

0.01 709/4677 634/4752 535/4851 516/4870 466/4920 —

0.92 0.78 0.78 0.76

1.00 (0.82, 1.02) (0.70, 0.88) (0.70, 0.88) (0.68, 0.86) ,0.001

1.00 0.94 (0.85, 1.05) 0.82 (0.74, 0.92) 0.85 (0.75, 0.95) 0.83 (074, 0.94) ,0.001

570/4816 513/4873 535/4851 586/4800 656/4730 —

1.00 0.91 (0.81, 1.02) 0.96 (0.86, 1.08) 1.10 (0.98, 1.24) 1.29 (1.15, 1.44) ,0.001

1.00 0.93 (0.83, 1.05) 0.99 (0.88, 1.12) 1.12 (1.00, 1.26) 1.27 (1.14, 1.43) ,0.001

0.98 0.85 0.88 0.89

1.00 (0.88, 1.09) (0.76, 0.95) (0.79, 0.99) (0.79, 1.01) 0.01

0.92 0.95 1.07 1.09

1.00 (0.82, 1.04) (0.84, 1.07) (0.95, 1.20) (0.98, 1.23) 0.02

0.23

(Continued)

9 of 16

FOOD SOURCES OF FAT AND INCIDENT TYPE 2 DIABETES TABLE 4 (Continued )

Nutrient quintile (median intake/d) Milk, low-fat (g) 05 (0) 1 (57) 2 (182) 3 (322) 4 (546) P-trend Milk, low-fat, nonfermented (g) 05 (0) 1 (43) 2 (157) 3 (289) 4 (503) P-trend Milk, low-fat, fermented (g) 05(0) 1 (29) 2 (71) 3 (140) 4 (250) P-trend Milk, high-fat (g) 1 (6) 2 (29) 3 (68) 4 (161) 5 (330) P-trend Milk, high-fat, nonfermented (g) 1 (3) 2 (16) 3 (33) 4 (63) 5 (271) P-trend Milk, high-fat, fermented (g) 05(0) 1 (25) 2 (61) 3 (107) 4 (179) P-trend Milk, nonfermented (g) 1 (24) 2 (119) 3 (244) 4 (357) 5 (515) P-trend Milk, fermented (g) 05 (0) 2 (36) 3 (75) 4 (143) 5 (250) P-trend Cheese (g) 1 (11) 2 (27) 3 (40)

n cases/ person-years

Basic model2

Multivariate model without BMI3

666/5911 479/4609 463/4625 570/4519 682/4406 —

1.00 0.96 (0.86, 1.08) 0.94 (0.84, 1.06) 1.14 (1.01, 1.27) 1.34 (1.20, 1.49) ,0.001

1.00 1.01 (0.90, 1.14) 1.00 (0.88, 1.12) 1.20 (1.07, 1.34) 1.37 (1.24, 1.54) ,0.001

833/7702 440/4158 448/4151 497/4102 642/3957 —

1.00 0.98 (0.87, 1.10) 1.03 (0.92, 1.16) 1.13 (1.01, 1.26) 1.48 (1.33, 1.64) ,0.001

1.00 1.03 (0.92, 1.16) 1.08 (0.96, 1.22) 1.17 (1.04, 1.30) 1.46 (1.32, 1.62) ,0.001

1938/16,011 224/2021 239/2006 216/2030 243/2002 —

1.00 (0.90, (0.93, (0.81, (0.82, 0.38

Full multivariate model with BMI4

P-interaction with sex 0.28

0.99 0.93 1.10 1.15

1.00 (0.88, 1.12) (0.83, 1.05) (0.98, 1.23) (1.04, 1.29) 0.003

1.00 1.02 1.05 1.21

1.00 (0.90, 1.13) (0.91, 1.14) (0.94, 1.18) (1.09, 1.34) 0.001

1.04 1.04 0.96 1.04

1.00 (0.90, (0.91, (0.83, (0.91, 0.73

1.00 0.93 0.89 0.91

1.00 (0.90, (0.84, (0.79, (0.81, 0.02

0.95 1.04 0.98 0.94

1.00 (0.84, (0.93, (0.87, (0.84, 0.52

0.99 0.97 0.94 0.80

1.00 (0.87, 1.13) (0.86, 1.10) (0.83, 1.07) (0.69, 0.92) 0.007

1.05 1.05 1.07 1.24

1.00 (0.93, 1.19) (0.92, 1.18) (0.95, 1.21) (1.10, 1.39) 0.001

1.04 0.93 0.97 0.91

1.00 (0.94, (0.83, (0.86, (0.81, 0.08

0.12

0.46 1.04 1.06 0.93 0.94

1.20) 1.22) 1.08) 1.08)

1.08 1.12 1.00 1.02

1.00 (0.94, 1.25) (0.98, 1.28) (0.87, 1.15) (0.89, 1.17) 0.48

1.20) 1.20) 1.11) 1.19) 0.83

647/4739 627/4759 568/4818 513/4873 505/4881 —

1.00 1.01 (0.90, 1.12) 0.93 (0.83, 1.04) 0.83 (0.74, 0.94) 0.86 (0.76, 0.96) ,0.001

601/4785 577/4809 595/4791 555/4831 532/4854 —

1.00 (0.86, (0.92, (0.87, (0.83, 0.40

1.00 1.01 (0.90, 1.13) 0.93 (0.83, 1.04) 0.84 (0.75, 0.95) 0.84 (0.75, 0.94) ,0.001

1.12) 1.05) 1.00) 1.03) 0.40

0.98 1.03 0.98 0.94

1.08) 1.16) 1.10) 1.06)

0.98 1.04 0.97 0.89

1.00 (0.87, 1.10) (0.93, 1.17) (0.86, 1.09) (0.79, 1.00) 0.09

1.06) 1.17) 1.10) 1.06) 0.77

1812/13,917 287/2513 283/2517 270/2531 208/2592 —

1.00 0.94 (0.83, 1.07) 0.92 (0.81, 1.04) 0.87 (0.79, 0.99) 0.69 (0.59, 0.79) ,0.001

1.00 0.97 (0.86, 1.10) 0.95 (0.84, 1.08) 0.91 (0.80, 1.03) 0.73 (0.63, 0.84) ,0.001

509/4877 532/4854 540/4846 579/4807 700/4686 —

1.00 1.05 (0.93, 1.19) 1.07 (0.95, 1.21) 1.19 (1.05, 1.34) 1.55 (1.38, 1.74) ,0.001

1.00 1.06 (0.94, 1.20) 1.08 (0.95, 1.22) 1.16 (1.03, 1.31) 1.47 (1.31, 1.65) ,0.001

1132/8292 469/3907 440/3937 431/3946 388/3988 —

1.00 0.97 (0.87, 1.08) 0.89 (0.80, 1.00) 0.86 (0.77, 0.96) 0.77 (0.68, 0.86) ,0.001

1.03 0.95 0.94 0.85

1.00 (0.92, 1.14) (0.85, 1.06) (0.84, 1.05) (0.76, 0.96) 0.01

666/4720 589/4797 571/4851

1.00 0.89 (0.80, 0.99) 0.89 (0.79, 0.99)

1.00 0.93 (0.83, 1.04) 0.94 (0.84, 1.06)

0.15

0.40 1.16) 1.04) 1.08) 1.02) 0.01 1.00 0.94 (0.84, 1.05) 0.92 (0.82, 1.03) (Continued)

10 of 16

ERICSON ET AL.

TABLE 4 (Continued )

Nutrient quintile (median intake/d) 4 (53) 5 (82) P-trend Cream (g) 1 (0.3) 2 (5) 3 (11) 4 (18) 5 (32) P-trend Ice cream (g) 1 (0) 2 (3) 3 (6) 4 (13) 5 (29) P-trend Butter/butter blends (g) 1 (0) 2 (3) 3 (16) 4 (28) 5 (33) P-trend Margarine total (g) 1 (5) 2 (13) 3 (25) 4 (38) 5 (59) P-trend Margarine, low-fat (g) 05(0) 1 (8) 2 (19) 3 (30) 4 (52) P-trend Margarine, high-fat (g) 1 (3) 2 (6) 3 (8) 4 (12) 5 (26) P-trend Oils and dressing (g) 1 (0) 2 (1) 3 (4) 4 (7) 5 (14) P-trend Eggs (g) 1 (4) 2 (12) 3 (19) 4 (28) 5 (45) P-trend

n cases/ person-years

Basic model2

Multivariate model without BMI3

Full multivariate model with BMI4

532/4854 502/4884 —

0.86 (0.77, 0.96) 0.87 (0.77, 0.98) 0.01

0.94 (0.83, 1.05) 0.96 (0.85, 1.07) 0.50

0.93 (0.83, 1.05) 0.92 (0.81, 1.04) 0.21

671/4715 623/4763 590/4796 501/4885 475/4911 —

1.00 0.95 (0.85, 1.06) 0.91 (0.82, 1.02) 0.76 (0.68, 0.86) 0.71 (0.63, 0.80) ,0.001

1.00 0.99 (0.88, 1.10) 0.96 (0.86, 1.07) 0.81 (0.72, 0.91) 0.75 (0.67, 0.85) ,0.001

1.01 1.00 0.88 0.85

1.00 (0.90, 1.13) (0.89, 1.12) (0.78, 0.99) (0.76, 0.96) 0.001

632/4754 539/4847 575/4811 530/4856 584/4802 —

0.86 0.94 0.86 0.94

1.00 (0.76, 0.96) (0.84, 1.05) (0.77, 0.96) (0.84, 1.05) 0.33

0.89 0.97 0.91 1.00

1.00 (0.79, 1.00) (0.86, 1.08) (0.81, 1.02) (0.89, 1.12) 0.93

0.89 0.93 0.87 0.93

1.00 (0.79, (0.83, (0.77, (0.83, 0.20

1781/13,548 281/2619 290/2610 255/2646 253/2647 —

1.00 (0.73, 0.94) (0.77, 0.99) (0.67, 0.87) (0.72, 0.93) ,0.001

0.88 0.90 0.77 0.79

1.00 (0.77, 1.00) (0.79, 1.02) (0.67, 0.88) (0.69, 0.91) ,0.001

0.89 0.94 0.83 0.86

1.00 (0.78, 1.01) (0.83, 1.06) (0.73, 0.95) (0.75, 0.98) 0.001

0.60

0.83 0.87 0.76 0.72

541/4845 499/4887 578/4808 601/4785 641/4745 —

1.00 (0.84, 1.07) (0.95, 1.21) (0.96, 1.22) (0.97, 1.22) 0.03

0.97 1.10 1.08 1.04

1.00 (0.86, 1.10) (0.97, 1.23) (0.96, 1.21) (0.93, 1.17) 0.19

0.94 1.04 1.03 0.99

1.00 (0.83, (0.93, (0.91, (0.88, 0.69

0.28

0.95 1.07 1.08 1.09

1.05 1.09 1.07 1.13

1.00 (0.93, 1.18) (0.97, 1.22) (0.96, 1.20) (1.02, 1.26) 0.02

1.09 1.12 1.08 1.09

1.00 (0.96, 1.23) (1.00, 1.26) (0.96, 1.20) (0.98, 1.21) 0.06

1.01 1.08 1.02 1.02

1.00 (0.89, (0.96, (0.92, (0.91, 0.55

0.92 0.91 0.91 0.95

1.00 (0.82, 1.03) (0.81, 1.02) (0.81, 1.02) (0.85, 1.06) 0.36

0.92 0.90 0.92 0.94

1.00 (0.82, 1.03) (0.80, 1.01) (0.82, 1.03) (0.84, 1.05) 0.32

0.91 0.88 0.94 0.97

1.00 (0.81, (0.78, (0.84, (0.86, 0.77

566/4820 632/4754 560/4826 561/4825 541/4845 —

1.16 1.04 1.04 1.04

1.00 (1.03, 1.30) (0.92, 1.17) (0.93, 1.18) (0.92, 1.17) 1.00

1.21 1.11 1.13 1.14

1.00 (1.08, 1.35) (0.98, 1.25) (1.00, 1.27) (1.01, 1.28) 0.17

1.16 1.06 1.09 1.09

1.00 (1.04, (0.94, (0.97, (0.96, 0.49

528/4858 565/4821 538/4848 592/4794 637/4749 —

1.00 1.08 (0.96, 1.21) 1.02 (0.90, 1.15) 1.16 (1.03, 1.30) 1.27 (1.13, 1.42) ,0.001

1.07 0.99 1.10 1.14

1.00 (0.95, (0.88, (0.98, (1.02, 0.03

P-interaction with sex

0.28

0.27 1.00) 1.04) 0.98) 1.04)

1.07) 1.18) 1.15) 1.11) 0.16

1102/9962 382/3584 421/3546 440/3527 515/3451 —

1.14) 1.21) 1.14) 1.13) 0.42

617/4769 565/4821 557/4829 543/4843 578/4808 —

1.02) 0.98) 1.06) 1.08) 0.38 1.31) 1.20) 1.22) 1.23) 0.89

1.00 1.08 (0.96, 1.21) 1.02 (0.90, 1.15) 1.16 (1.03, 1.31) 1.27 (1.13, 1.42) ,0.001

1.20) 1.12) 1.24) 1.28) (Continued)

11 of 16

FOOD SOURCES OF FAT AND INCIDENT TYPE 2 DIABETES TABLE 4 (Continued )

Nutrient quintile (median intake/d) Meat and meat products, total (g) 1 (55) 2 (84) 3 (102) 4 (123) 5 (163) P-trend Meat and meat products, low-fat (g) 1 (9) 2 (24) 3 (35) 4 (49) 5 (75) P-trend Meat, red, low-fat, nonprocessed (g) 1 (1) 2 (15) 3 (24) 4 (36) 5 (57) P-trend Meat products, low-fat, processed (g) 05(0) 1 (3) 2 (8) 3 (14) 4 (27) P-trend Meat and meat products, high-fat (g) 1 (16) 2 (36) 3 (51) 4 (68) 5 (93) P-trend Meat, red, high-fat, nonprocessed (g) 1 (4) 2 (16) 3 (25) 4 (36) 5 (55) P-trend Meat products, high-fat, processed (g) 1 (2) 2 (16) 3 (29) 4 (38) 5 (50) P-trend Poultry (g) 1 (0) 2 (1) 3 (13) 4 (22) 5 (36) P-trend Fish and shellfish, low-fat (g) 1 (0) 2 (9) 3 (21)

n cases/ person-years

Basic model2

Multivariate model without BMI3

Full multivariate model with BMI4

394/4992 522/4864 581/4805 499/4787 764/4622 —

1.00 1.28 (1.12, 1.46) 1.41 (1.24, 1.60) 1.44 (1.27, 1.64) 1.82 (1.61, 2.06) ,0.001

1.00 1.24 (1.09, 1.41) 1.36 (1.19, 1.54) 1.36 (1.19, 1.55) 1.68 (1.48, 1.91) ,0.001

1.00 1.13 (1.00, 1.29) 1.20 (1.06, 1.37) 1.15 (1.01, 1.31) 1.36 (1.20, 1.55) ,0.001

504/4926 553/4892 573/4810 592/4784 658/4658 —

1.06 1.15 1.20 1.34

1.00 (0.94, 1.20) (1.03, 1.31) (1.08, 1.37) (1.19, 1.51) ,0.001

1.00 1.05 (0.93, 1.19) 1.14 (1.01, 1.29) 1.17 (1.04, 1.32) 1.25 (1.11, 1.41) ,0.001

509/4877 545/4841 542/4844 602/4784 662/4724 —

1.00 1.08 (0.96, 1.22) 1.08 (0.96, 1.22) 1.18 (1.05, 1.33) 1.30 (1.16, 1.46) ,0.001

1.00 1.09 (0.99, 1.22) 1.08 (0.96, 1.22) 1.19 (1.06, 1.34) 1.28 (1.15, 1.46) ,0.001

1.00 1.11 (0.99, 1.26) 1.07 (0.95, 1.21) 1.17 (1.04, 1.32) 1.24 (1.10, 1.39) ,0.001

628/5628 582/4586 520/4649 542/4627 588/4580 —

1.17 1.05 1.10 1.20

1.00 (1.04, 1.31) (0.94, 1.18) (0.98, 1.23) (1.08, 1.35) 0.01

1.20 1.07 1.12 1.23

1.00 (1.07, 1.34) (0.96, 1.21) (1.00, 1.26) (1.10, 1.37) 0.01

444/5032 531/4913 566/4851 655/4676 664/4598 —

1.15 1.21 1.42 1.44

1.00 (1.01, 1.30) (1.07, 1.37) (1.25, 1.60) (1.28, 1.63) ,0.001

1.10 1.13 1.30 1.27

1.00 (0.97, 1.25) (1.00, 1.28) (1.15, 1.47) (1.12, 1.44) ,0.001

504/4882 527/4859 587/4799 610/4776 632/4754 —

1.00 1.02 (0.90, 1.15) 1.14 (1.01, 1.28) 1.17 (1.04, 1.32) 1.22 (1.08, 1.37) ,0.001

1.00 1.10 1.12 1.13

1.00 (0.88, 1.13) (0.97, 1.24) (0.99, 1.26) (1.00, 1.27) 0.01

433/4953 552/4834 578/4808 667/4719 630/4756 —

1.00 1.20 (1.05, 1.36) 1.26 (1.11, 1.42) 1.47 (1.30, 1.66) 1.43 (1.26, 1.62) ,0.001

1.00 1.16 (1.02, 1.32) 1.20 (1.16, 1.36) 1.37 (1.21, 1.55) 1.29 (1.14, 1.46) ,0.001

114/9498 412/3667 396/3684 458/3622 480/3599 —

1.03 0.99 1.11 1.12

1.00 (0.92, 1.16) (0.88, 1.11) (1.00, 1.24) (1.00, 1.24) 0.02

1.00 (0.93, 1.17) (0.91, 1.14) (1.03, 1.28) (1.03, 1.28) 0.004

620/4766 579/4807 551/4835

1.00 0.94 (0.84, 1.05) 0.90 (0.81, 1.02)

P-interaction with sex 0.80

1.00 (0.94, 1.19) (1.02, 1.29) (1.06, 1.35) (1.19, 1.50) ,0.001

1.06 1.16 1.21 1.34

0.56

0.46

0.69 1.16 1.07 1.09 1.16

1.00 (1.04, (0.95, (0.97, (1.04, 0.06

1.30) 1.20) 1.22) 1.30)

1.04 1.06 1.17 1.09

1.00 (0.92, (0.94, (1.04, (0.97, 0.04

1.18) 1.20) 1.32) 1.24)

0.96 1.04 1.04 1.01

1.00 (0.85, (0.93, (0.92, (0.90, 0.48

1.12 1.18 1.29 1.15

1.00 (0.99, (1.04, (1.14, (1.01, 0.01

1.02 1.00 1.10 1.06

1.00 (0.91, (0.90, (0.99, (0.96, 0.11

0.17

0.51 1.08) 1.18) 1.17) 1.14) 0.97 1.28) 1.33) 1.46) 1.30) 0.33 1.04 1.02 1.15 1.14

1.14) 1.13) 1.23) 1.18) 0.46

1.00 0.98 (0.87, 1.10) 0.95 (0.84, 1.06)

1.00 0.95 (0.85, 1.07) 0.95 (0.84, 1.06) (Continued)

12 of 16

ERICSON ET AL.

TABLE 4 (Continued )

Nutrient quintile (median intake/d) 4 (33) 5 (55) P-trend Fish, high-fat (g) 05 (0) 1 (3) 2 (9) 3 (23) 4 (46) P-trend Pastry and biscuits (g) 1 (6) 2 (20) 3 (33) 4 (48) 5 (72) P-trend Chocolate (g) 1 (0) 2 (2) 3 (4) 4 (8) 5 (16) P-trend

n cases/ person-years

Basic model2

Multivariate model without BMI3

Full multivariate model with BMI4

553/4833 557/4829 —

0.89 (0.80, 1.00) 0.90 (0.80, 1.00) 0.04

0.94 (0.84, 1.06) 0.98 (0.87, 1.10) 0.52

0.95 (0.85, 1.06) 0.97 (0.86, 1.09) 0.60

691/5978 549/4516 534/4531 507/4559 579/4486 —

1.01 0.98 0.90 1.04

1.00 (0.90, 1.14) (0.87, 1.10) (0.80, 1.02) (0.93, 1.16) 0.78

1.05 1.03 0.96 1.11

1.00 (0.93, (0.92, (0.86, (0.99, 0.31

1.07 1.02 0.93 1.05

1.00 (0.95, (0.91, (0.82, (0.94, 0.86

660/4726 573/4813 542/4844 554/4832 531/4855 —

1.00 0.87 (0.77, 0.97) 0.81 (0.72, 0.91) 0.83 (0.74, 0.93) 0.80 (0.71, 0.90) ,0.001

0.90 0.84 0.86 0.82

1.00 (0.80, 1.00) (0.75, 0.94) (0.77, 0.97) (0.72, 0.92) 0.001

0.92 0.87 0.90 0.89

1.00 (0.82, (0.78, (0.80, (0.79, 0.06

656/4730 564/4822 561/4825 542/4844 537/4849 —

1.00 (0.74, 0.93) (0.76, 0.96) (0.77, 0.97) (0.80, 1.01) 0.14

0.87 0.91 0.92 0.94

1.00 (0.78, (0.81, (0.82, (0.84, 0.54

0.88 0.94 0.93 1.01

1.00 (0.79, (0.84, (0.83, (0.90, 0.66

P-interaction with sex

0.67 1.17) 1.15) 1.08) 1.25)

1.20) 1.14) 1.04) 1.18) 0.70 1.03) 0.98) 1.01) 1.01) 0.08

0.83 0.86 0.86 0.90

0.97) 1.02) 1.03) 1.06)

0.99) 1.06) 1.05) 1.14)

1 HRs were calculated by using a Cox proportional hazards model. ALA, a-linolenic acid; LA, linoleic acid; MDC, Malmo¨ Diet and Cancer; T2D, type 2 diabetes. 2 Adjusted for age (continuous), sex (when applicable), method version (categorical), season (categorical), and total energy intake (continuous). 3 Adjusted as for the basic model and for the following categorical variables: leisure-time physical activity, smoking, alcohol intake, and education. 4 Adjusted as for the basic model and for the following categorical variables: leisure-time physical activity, smoking, alcohol intake, education, and BMI (continuous). 5 Zero consumers; higher categories are quartiles in consumers.

Although high intake of total low-fat dairy products was associated with increased risk (P-trend = 0.01), this association was NS when intakes of low- and high-fat dairy products were mutually adjusted (P-trend = 0.18), whereas the protective association with high-fat dairy products remained significant (P-trend = 0.003). Furthermore, the association with low-fat dairy products also disappeared after adjustment for protein intake (P-trend = 0.37); similar observations were made for lowfat nonfermented milk. Results regarding high-fat dairy products remained unchanged. High intakes of meats, both low-fat (Ptrend , 0.001) and high-fat (P-trend = 0.04) meat and meat products, were associated with increased risk of T2D. Increased risk seemed mainly driven by intakes of low-fat nonprocessed red meat (P-trend , 0.001) and high-fat processed meat products (P-trend = 0.01). Finally and similarly to what has previously been reported after analyses with shorter follow-up time in the MDC cohort (33), high egg intake was associated with increased risk. All observed associations remained virtually unchanged after additional adjustments for dietary change in the past. A post hoc analysis indicated that intakes of several nondairy foods tended to differ significantly across intake quintiles of cream and high-fat fermented milk; decreased intakes of both sugar-sweetened beverages and fiber-rich bread and cereals were, for example, seen across quintiles (Supplemental Table 2). However, adjustment for dietary intakes (fiber, sucrose, calcium, vitamin D, magnesium, meat, fruit and vegetables, or

sugar-sweetened beverages) did not substantially affect any of our observed associations. Except for the interaction between cheese intake and sex [also reflected in the interaction between intake of high-fat fermented dairy products and sex (P = 0.01)], we did not observe any significant interactions between any other examined food intakes and sex. Statistical models without BMI Overall, statistical models without BMI did not substantially change our observations. However, inverse associations between several of the high-fat dairy foods and T2D were somewhat stronger before adjustment for BMI (i.e., for cream, high-fat fermented milk, and butter). In addition, individuals in the highest quintile of high-fat nonfermented milk tended to be at decreased risk before adjustment for BMI (HR: 0.89; CI: 0.79, 1.00). Moreover, high-fat nonprocessed red meat was significantly associated with increased risk of T2D only before adjustment for BMI (P-trend = 0.01). The inclusion of waist circumference in our statistical models did not substantially affect any results. Sensitivity analysis In an analysis excluding individuals who reported less-stable food habits (24% of participants), we did not observe an inverse association between total intake of SFAs and T2D (P-trend = 0.69 in the full multivariate model including BMI). However,

FOOD SOURCES OF FAT AND INCIDENT TYPE 2 DIABETES

although decreased risks of T2D at high intakes of SFAs with 4– 10 carbons, lauric acid, and myristic acid were no longer significant, tendencies of protective associations were still seen, especially in women (P-trend = 0.09, 0.06, and 0.06, respectively). As concerns specific food sources of fat, the consumption of cream and high-fat fermented milk remained significantly and inversely associated with T2D (P-trend = 0.01), whereas inverse associations with butter became nonsignificant. Observed associations with intakes of different types of meat and meat products remained virtually unchanged. Moreover, an increased incidence of T2D was now seen in individuals in the highest compared with lowest quintiles of n–6 PUFA intake (HR: 1.17; 95% CI: 1.02, 1.35), and in line with this result, we observed decreased risk in the highest quintile of the ratio between intakes of n–3 and n–6 PUFAs (HR: 0.86; 95% CI: 0.75, 0.99), but no significant trends were seen across quintiles (P-trend = 0.10 and 0.16, respectively). After the exclusion of individuals who reported less-stable food habits and additional adjustment for protein intake to the full multivariate model, a significant trend of increased risk across intake quintiles of n–6 PUFAs was seen (P-trend = 0.046) as well as a tendency of decreased risk with a higher ratio between intakes of n–3 and n–6 PUFAs (P-trend = 0.07). Finally, after adjustment for protein, we also observed an inverse association between long-chain n–3 PUFAs and T2D (P-trend = 0.04) in men. The results remained virtually unchanged in analysis that excluded individuals with prevalent cardiovascular disease at baseline. Finally, our observations were similar in normal-weight and overweight individuals. DISCUSSION

Our main findings were that, in contrast to low-fat dairy products, high intake of high-fat dairy products was associated with decreased incidence of T2D, whereas the consumption of both low- and high-fat meats was associated with increased incidence. Intakes of palmitic and stearic acids did not show significant associations, whereas SFAs with shorter chain lengths were inversely associated with T2D. Although associations with dairy intakes were weaker after the exclusion of individuals with potentially unstable dietary habits, high intakes of cream and high-fat fermented milk, together with cheese in women, were still associated with significantly decreased risk of T2D. In a sensitivity analysis, we also observed increased risk at high n–6 PUFA intake. SFAs 12–16 are known to have adverse effects on LDLcholesterol concentrations, and even though there is evidence that shows that these SFAs can have beneficial effects on HDL cholesterol, triglycerides, or apolipoprotein A-I (34), a systematic review indicated similar improvements by unsaturated fats (7), and SFAs do not seem to change the ratio between total cholesterol and HDL cholesterol (34). Nevertheless, findings from observational studies provided a more balanced picture of the role of SFAs in the development of obesity and cardiometabolic diseases, and researchers recognized the value of examining food sources of fat and specific SFAs separately (14, 35). In a recent report from the European Prospective Investigation into Cancer and Nutrition-InterAct, odd-chained SFAs in plasma phospholipids were positively correlated with

13 of 16

intake of dairy products and showed inverse associations with T2D (15). Similar observations were made in other studies (36, 37). Also, it was recently concluded, in a meta-analysis, that total dairy intake was inversely associated with risk of T2D (12). However, in contrast to our findings, the meta-analysis showed protective associations with low-fat dairy. Only one of the included studies reported an inverse association with high-fat dairy products (38), and high-fat dairy was inversely associated with the metabolic syndrome in another study (39). Still, all studies could not distinguish between low- and high-fat dairy (13), studies included different dairy foods (40, 41), and classifications of low- compared with high-fat products differed. Finally, dairy intake in Sweden is relatively high compared with that in other populations. Dairy is the most important fat source in the MDC cohort and contributes, on average, 30% of total fat intake and 35% of SFA intake in Sweden (27, 42); corresponding figures in the United States are 12% and 24% (43), whereas meat contributes less to fat intake in Sweden (42, 44). In line with previous findings, in the MDC cohort after a shorter follow-up and in other cohorts (33, 45), we observed higher risk of T2D for individuals with high-meat diets. Because our observations were independent of the fat content of the meat, compounds such as nitrite, heterocyclic amines, and iron may explain our findings. Dietary SFAs with shorter chain lengths (4:0–10:0) that indicated protective associations with T2D in this study were mainly found in dairy products. Dairy products are also better sources of lauric acid (12:0) and myristic acid (14:0) than are other important food sources of fat in Sweden (46). In contrast, palmitic acid (16:0) and stearic acid (18:0), which are abundant in both dairy foods and meat, fish, and eggs, showed null associations with T2D (46). The importance of dairy fats in the development of T2D is not well understood. Fatty acids reflected in blood and tissue concentrations such as odd-chained SFAs pentadecanoic acid (15:0) and heptadecanoic acid (17:0) (15, 36, 37), which are also present in fish, and trans palmitoleic acid (trans-16:1n–7), which is a biomarker of dairy fat, showed inverse associations with T2D (47). Likewise, conjugated LA may have beneficial effects in metabolic disease, but studies have been inconclusive and even suggested adverse effects (48). A study in mice indicated that butyrate (4:0) may prevent dietinduced insulin resistance (49). Although SFAs with shorter chains may also promote insulin resistance via mechanisms that lead to inflammatory processes (8), some experimental studies indicated that SFAs with $16 carbons were more prone to cause insulin resistance (50, 51). In addition, biomarkers of dairy fat and a food pattern characterized by dairy fat have indicated cross-sectional protective associations with hyperinsulinemia (52, 53). Future studies need to distinguish whether any SFAs with chain lengths from 4 to 14 carbons may account for any diabetes protective properties or if these properties can be referred to other dairy fats or food-specific characteristics. Health effects of dairy are most likely the result of a complex interplay between many components. For instance, SFAs in cheese may be less detrimental for serum cholesterol than SFAs in butter (54). Besides, fermented dairy may affect the gut microbiota composition, which was shown to be altered in individuals with T2D (4), and high intake has been associated with lower risk of T2D (13). However, we observed inverse associations with high-fat dairy products regardless of fermentation.

14 of 16

ERICSON ET AL.

Although sucrose, calcium, vitamin D, and magnesium in dairy may affect insulin secretion, insulin sensitivity, and risk of T2D (55, 56), adjustment for those intakes did not change our findings. Also, whey proteins showed favorable effects on glucose metabolism (57), but they are not likely to explain differing associations with high- compared with low-fat dairy. Our observation of increased risk at high n–6 PUFA intake in sensitivity analysis needs cautious interpretation. Some reports suggested that diets with a high ratio between n–3 and n–6 PUFAs may prevent insulin resistance and T2D (58), which supports our findings, but other reports were inconclusive (18, 59). High n–6 PUFAs may also improve insulin sensitivity (60), and blood concentration of LA has been inversely associated with T2D (7). Even if fat intake may promote a positive energy balance, and there has been probable evidence for a positive association with body weight from randomized controlled trials, the magnitude of this association is most likely modest (7), and results from longterm prospective studies have been less convincing (61). Furthermore, there has been evidence that linked high-fat dairy intake to satiety (62). Nevertheless, BMI adjustments are crucial to minimize confounding by differing food preferences in lean and obese individuals. Besides, misreporting may be of special concern in obese individuals (63). We observed inverse associations between high-fat dairy intake and T2D regardless of BMI adjustment, and our findings were similar in normal-weight and overweight persons. Because adjustment for waist circumference had an even smaller influence on our results, it is possible that overweight in general (independently of fat distribution) is a more important confounder because of potentially stronger links to food choices and dietary reporting. A loss of power may have partly explained the weakened associations between high-fat dairy products and T2D and that associations with specific SFAs became nonsignificant when participants who reported unstable food habits were excluded (almost one-fourth of subjects were excluded). Inverse associations with cream and high-fat fermented milk remained significant. Because dietary change was more common in subjects who developed T2D and inversely related to high-fat dairy intake, we also treated the variable as a confounder [i.e., a potential marker of unhealthy dietary habits earlier in life because health reasons were the major cause of dietary change (64)]. This adjustment did not significantly affect any results. This study had several strengths. It was a large study with a long follow-up time. Because it was a population-based prospective study, selection bias and reverse causation were minor issues. A main objective of the MDC study was to examine fat intake (65), and the relative validity for dietary intakes of importance to this study (e.g., total milk correlation coefficient was 0.8) has been well documented (24, 25). The intake range was wide for most foods (e.g., median high-fat dairy intakes were 1 and 8 portions/d in extreme quintiles). We had extensive information on potential confounders and the possibility to exclude individuals with reported dietary changes in the past. Moreover, we were able to distinguish between low- and high-fat products. High-fat dairy foods could be part of a healthy-lifestyle pattern, and individuals who developed diabetes may have been more prone to adapt a healthy lifestyle, which could have led to observations that reflected reverse causation. However, a counter argument is that our analyses indicated that high-fat dairy intake

was associated with unhealthy lifestyle characteristics, and other studies also showed that high-fat dairy is more common in individuals with a lower socioeconomic status (66). Besides, meat intake was associated with higher rather than lower risk, as would have been expected on the basis of a similar potential for reverse causation. The relatively low validity for some PUFA intakes was a limitation (25). A lower relative validity in men may have explained that the association with cheese intake was restricted to women and that associations with intakes of specific SFAs were more robust in women. Moreover, an analysis of some plant sources of fat, such as nuts and seeds, was not meaningful because of low intakes. Finally, we could not exclude the occurrence of residual confounding. In conclusion, our results indicate that analyses of food sources of fat may partially clarify the inconsistent role of dietary fat for risk of T2D. We observed a decreased incidence of T2D at high intake of high-fat dairy products but not of low-fat dairy products. Meat intake was associated with increased risk independently of fat content. Although intake of palmitic acid, which is the mostabundant SFA in both dairy and meat, was not significantly associated with T2D, intakes of SFAs with 4–14 carbons, which are more abundant in dairy than in meat, showed inverse associations with T2D. Our study indicates a protective role of fat from dairy and suggests that dairy fat may also have contributed to previously observed protective associations between dairy intake and T2D. Of 28,098 participants in the MDC cohort, 1758 incident diabetes cases and 1758 controls are included in the European Prospective Investigation into Cancer and Nutrition InterAct Consortium for the study of genetic factors and gene-lifestyle interactions in regard to incident diabetes. As a large cohort study, the MDC represents a different study design than the case-control study design of the European Prospective Investigation into Cancer and Nutrition InterAct. Dietary data used within the European Prospective Investigation into Cancer and Nutrition InterAct are harmonized between several study centers, and many details in the MDC dietary data used in the current study were lacking in these harmonized data. That is, a different study design, different study size, extensive information on confounding variables, the possibility to exclude individuals with reported dietary change, and uniform dietary data of high relative validity ensured the uniqueness of the current study compared with the pooled analyses that may be performed within the European Prospective Investigation into Cancer and Nutrition InterAct. The authors’ responsibilities were as follows—UE: designed the research, performed the statistical analysis, wrote the manuscript, and had primary responsibility for the final content of the manuscript; BG: gave statistical advice; and all authors: contributed to the interpretation of results and revision of the manuscript and read and approved the final version of the manuscript. None of the funders had any role in the study design, data collection and analysis, interpretation of data, decision to publish, or preparation of the manuscript. None of the authors reported a conflict of interest related to the study.

REFERENCES 1. Pégorier JP, Le May C, Girard J. Control of gene expression by fatty acids. J Nutr 2004;134:2444S–9S. 2. Storlien LH, Pan DA, Kriketos AD, O’Connor J, Caterson ID, Cooney GJ, Jenkins AB, Baur LA. Skeletal muscle membrane lipids and insulin resistance. Lipids 1996;31(Suppl):S261–5. 3. Lottenberg AM, Afonso Mda S, Lavrador MS, Machado RM, Nakandakare ER. The role of dietary fatty acids in the pathology of metabolic syndrome. J Nutr Biochem 2012;23:1027–40. 4. Qin J, Li Y, Cai Z, Li S, Zhu J, Zhang F, Liang S, Zhang W, Guan Y, Shen D, et al. A metagenome-wide association study of gut microbiota in type 2 diabetes. Nature 2012;490:55–60.

FOOD SOURCES OF FAT AND INCIDENT TYPE 2 DIABETES 5. Penn L, White M, Lindstrom J, den Boer AT, Blaak E, Eriksson JG, Feskens E, Ilanne-Parikka P, Keinanen-Kiukaanniemi SM, Walker M, et al. Importance of weight loss maintenance and risk prediction in the prevention of type 2 diabetes: analysis of European Diabetes Prevention Study RCT. PLoS ONE 2013;8:e57143. 6. Knowler WC, Barrett-Connor E, Fowler SE, Hamman RF, Lachin JM, Walker EA, Nathan DM. Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. N Engl J Med 2002;346:393–403. 7. Schwab U, Lauritzen L, Tholstrup T, Haldorssoni T, Riserus U, Uusitupa M, Becker W. Effect of the amount and type of dietary fat on cardiometabolic risk factors and risk of developing type 2 diabetes, cardiovascular diseases, and cancer: a systematic review. Food Nutr Res 2014;58. 8. Risérus U, Willett WC, Hu FB. Dietary fats and prevention of type 2 diabetes. Prog Lipid Res 2009;48:44–51. 9. Mahendran Y, Cederberg H, Vangipurapu J, Kangas AJ, Soininen P, Kuusisto J, Uusitupa M, Ala-Korpela M, Laakso M. Glycerol and fatty acids in serum predict the development of hyperglycemia and type 2 diabetes in Finnish men. Diabetes Care 2013;36:3732–8. 10. Parillo M, Riccardi G. Diet composition and the risk of type 2 diabetes: epidemiological and clinical evidence. Br J Nutr 2004;92:7–19. 11. Micha R, Wallace SK, Mozaffarian D. Red and processed meat consumption and risk of incident coronary heart disease, stroke, and diabetes mellitus: a systematic review and meta-analysis. Circulation 2010;121:2271–83. 12. Aune D, Norat T, Romundstad P, Vatten LJ. Dairy products and the risk of type 2 diabetes: a systematic review and dose-response meta-analysis of cohort studies. Am J Clin Nutr 2013;98:1066–83. 13. Sluijs I, Forouhi NG, Beulens JW, van der Schouw YT, Agnoli C, Arriola L, Balkau B, Barricarte A, Boeing H, Bueno-de-Mesquita HB, et al. The amount and type of dairy product intake and incident type 2 diabetes: results from the EPIC-InterAct Study. Am J Clin Nutr 2012; 96:382–90. 14. Warensjo E, Nolan D, Tapsell L. Dairy food consumption and obesityrelated chronic disease. Adv Food Nutr Res 2010;59:1–41. 15. Forouhi NG, Koulman A, Sharp SJ, Imamura F, Kroger J, Schulze MB, Crowe FL, Huerta JM, Guevara M, Beulens JW, et al. Differences in the prospective association between individual plasma phospholipid saturated fatty acids and incident type 2 diabetes: the EPIC-InterAct case-cohort study. Lancet Diabetes Endocrinol 2014;2:810–8. 16. Patel PS, Forouhi NG, Kuijsten A, Schulze MB, van Woudenbergh GJ, Ardanaz E, Amiano P, Arriola L, Balkau B, Barricarte A, et al. The prospective association between total and type of fish intake and type 2 diabetes in 8 European countries: EPIC-InterAct Study. Am J Clin Nutr 2012;95:1445–53. 17. Alhazmi A, Stojanovski E, McEvoy M, Garg ML. Macronutrient intake and type 2 diabetes risk in middle-aged Australian women. Results from the Australian Longitudinal Study on Women’s Health. Public Health Nutr 2014;17:1587–94. 18. Wu JH, Micha R, Imamura F, Pan A, Biggs ML, Ajaz O, Djousse L, Hu FB, Mozaffarian D. Omega-3 fatty acids and incident type 2 diabetes: a systematic review and meta-analysis. Br J Nutr 2012;107(Suppl 2): S214–27. 19. Villegas R, Xiang YB, Elasy T, Li HL, Yang G, Cai H, Ye F, Gao YT, Shyr Y, Zheng W, et al. Fish, shellfish, and long-chain n–3 fatty acid consumption and risk of incident type 2 diabetes in middle-aged Chinese men and women. Am J Clin Nutr 2011;94:543–51. 20. Djoussé L, Gaziano JM, Buring JE, Lee IM. Dietary omega-3 fatty acids and fish consumption and risk of type 2 diabetes. Am J Clin Nutr 2011;93:143–50. 21. Manjer J, Carlsson S, Elmstahl S, Gullberg B, Janzon L, Lindstrom M, Mattisson I, Berglund G. The Malmo Diet and Cancer Study: representativity, cancer incidence and mortality in participants and nonparticipants. Eur J Cancer Prev 2001;10:489–99. 22. Callmer E, Riboli E, Saracci R, Akesson B, Lindgarde F. Dietary assessment methods evaluated in the Malmo food study. J Intern Med 1993;233:53–7. 23. Wirfa¨lt E, Mattisson I, Johansson U, Gullberg B, Wallstrom P, Berglund G. A methodological report from the Malmo Diet and Cancer study: development and evaluation of altered routines in dietary data processing. Nutr J 2002;1:3. 24. Elmsta˚h S, Riboli E, Lindgarde F, Gullberg B, Saracci R. The Malmo Food Study: the relative validity of a modified diet history method and an extensive food frequency questionnaire for measuring food intake. Eur J Clin Nutr 1996;50:143–51.

15 of 16

25. Riboli E, Elmstahl S, Saracci R, Gullberg B, Lindgarde F. The Malmo Food Study: validity of two dietary assessment methods for measuring nutrient intake. Int J Epidemiol 1997;26(Suppl 1):S161–73. 26. Elmsta˚hl S, Wallstro¨m P, Berglund G, Janzon L, Johansson U, Larsson S, Mattisson I. The use of dietary supplements in relation to dietary habits in a Swedish middle-aged population. Scand J Nutr 1994;38: 94–7. 27. Mattisson I, Wirfalt E, Andren C, Gullberg B, Berglund G. Dietary fat intake–food sources and dietary correlates in the Malmo Diet and Cancer cohort. Public Health Nutr 2003;6:559–69. 28. The Swedish National Food Administration SLV: Vikttabeller (weight tables). Uppsala: Swedish National Food Agency. 1989. p.1-49. 29. Lindholm E, Agardh E, Tuomi T, Groop L, Agardh CD. Classifying diabetes according to the new WHO clinical stages. Eur J Epidemiol 2001;17:983–9. 30. Cederholm J, Eeg-Olofsson K, Eliasson B, Zethelius B, Nilsson PM, Gudbjornsdottir S. Risk prediction of cardiovascular disease in type 2 diabetes: a risk equation from the Swedish National Diabetes Register. Diabetes Care 2008;31:2038–43. 31. Hanas R, John G. 2010 consensus statement on the worldwide standardization of the hemoglobin A1C measurement. Diabetes Care 2010; 33:1903–4. 32. Hoelzel W, Weykamp C, Jeppsson JO, Miedema K, Barr JR, Goodall I, Hoshino T, John WG, Kobold U, Little R, et al. IFCC reference system for measurement of hemoglobin A1c in human blood and the national standardization schemes in the United States, Japan, and Sweden: a method-comparison study. Clin Chem 2004;50:166–74. 33. 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: 1143–53. 34. Mensink RP, Zock PL, Kester AD, Katan MB. 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 2003;77:1146–55. 35. Astrup A. Yogurt and dairy product consumption to prevent cardiometabolic diseases: epidemiologic and experimental studies. Am J Clin Nutr 2014;99:1235S–42S. 36. Krachler B, Norberg M, Eriksson JW, Hallmans G, Johansson I, Vessby B, Weinehall L, Lindahl B. Fatty acid profile of the erythrocyte membrane preceding development of type 2 diabetes mellitus. Nutr Metab Cardiovasc Dis 2008;18:503–10. 37. Hodge AM, English DR, O’Dea K, Sinclair AJ, Makrides M, Gibson RA, Giles GG. Plasma phospholipid and dietary fatty acids as predictors of type 2 diabetes: interpreting the role of linoleic acid. Am J Clin Nutr 2007;86:189–97. 38. Malik VS, Sun Q, van Dam RM, Rimm EB, Willett WC, Rosner B, Hu FB. Adolescent dairy product consumption and risk of type 2 diabetes in middle-aged women. Am J Clin Nutr 2011;94:854–61. 39. Louie JC, Flood VM, Rangan AM, Burlutsky G, Gill TP, Gopinath B, Mitchell P. Higher regular fat dairy consumption is associated with lower incidence of metabolic syndrome but not type 2 diabetes. Nutr Metab Cardiovasc Dis 2013;23:816–21. 40. Liu S, Choi HK, Ford E, Song Y, Klevak A, Buring JE, Manson JE. A prospective study of dairy intake and the risk of type 2 diabetes in women. Diabetes Care 2006;29:1579–84. 41. 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 EPICPotsdam study. Eur J Clin Nutr 2013;67:412–9. 42. NationalFoodAgency. In: Riksmaten-vuxna (National food habit surveyadults) 2011-2012: Byra˚ 4; 2012. 43. Huth PJ, DiRienzo DB, Miller GD. Major scientific advances with dairy foods in nutrition and health. J Dairy Sci 2006;89:1207–21. 44. Schmid A. The role of meat fat in the human diet. Crit Rev Food Sci Nutr 2011;51:50–66. 45. Aune D, Ursin G, Veierod MB. Meat consumption and the risk of type 2 diabetes: a systematic review and meta-analysis of cohort studies. Diabetologia 2009;52:2277–87. 46. National Food Agency in Uppsala: Sweden www.slv.se/. The food database 2014-01-28. 47. Mozaffarian D, de Oliveira Otto MC, Lemaitre RN, Fretts AM, Hotamisligil G, Tsai MY, Siscovick DS, Nettleton JA. trans-Palmitoleic acid, other dairy fat biomarkers, and incident diabetes: the Multi-Ethnic Study of Atherosclerosis (MESA). Am J Clin Nutr 2013;97:854–61.

16 of 16

ERICSON ET AL.

48. Kratz M, Baars T, Guyenet S. The relationship between high-fat dairy consumption and obesity, cardiovascular, and metabolic disease. Eur J Nutr 2013;52:1–24. 49. Gao Z, Yin J, Zhang J, Ward RE, Martin RJ, Lefevre M, Cefalu WT, Ye J. Butyrate improves insulin sensitivity and increases energy expenditure in mice. Diabetes 2009;58:1509–17. 50. Holzer RG, Park EJ, Li N, Tran H, Chen M, Choi C, Solinas G, Karin M. Saturated fatty acids induce c-Src clustering within membrane subdomains, leading to JNK activation. Cell 2011;147:173–84. 51. Solinas G, Naugler W, Galimi F, Lee MS, Karin M. Saturated fatty acids inhibit induction of insulin gene transcription by JNK-mediated phosphorylation of insulin-receptor substrates. Proc Natl Acad Sci USA 2006;103:16454–9. 52. Warensjo¨ E, Jansson JH, Berglund L, Boman K, Ahren B, Weinehall L, Lindahl B, Hallmans G, Vessby B. Estimated intake of milk fat is negatively associated with cardiovascular risk factors and does not increase the risk of a first acute myocardial infarction. A prospective case-control study. Br J Nutr 2004;91:635–42. 53. Wirfa¨lt E, Hedblad B, Gullberg B, Mattisson I, Andren C, Rosander U, Janzon L, Berglund G. Food patterns and components of the metabolic syndrome in men and women: a cross-sectional study within the Malmo Diet and Cancer cohort. Am J Epidemiol 2001;154: 1150–9. 54. Hjerpsted J, Leedo E, Tholstrup T. Cheese intake in large amounts lowers LDL-cholesterol concentrations compared with butter intake of equal fat content. Am J Clin Nutr 2011;94:1479–84. 55. Dong JY, Xun P, He K, Qin LQ. Magnesium intake and risk of type 2 diabetes: meta-analysis of prospective cohort studies. Diabetes Care 2011;34:2116–22. 56. Pittas AG, Lau J, Hu FB, Dawson-Hughes B. The role of vitamin D and calcium in type 2 diabetes. A systematic review and meta-analysis. J Clin Endocrinol Metab 2007;92:2017–29.

57. Gunnerud UJ, Ostman EM, Bjorck IM. Effects of whey proteins on glycaemia and insulinaemia to an oral glucose load in healthy adults; a dose-response study. Eur J Clin Nutr 2013;67:749–53. 58. Simopoulos AP. Essential fatty acids in health and chronic disease. Am J Clin Nutr 1999;70:560S–9S. 59. Melanson EL, Astrup A, Donahoo WT. The relationship between dietary fat and fatty acid intake and body weight, diabetes, and the metabolic syndrome. Ann Nutr Metab 2009;55:229–43. 60. Bjermo H, Iggman D, Kullberg J, Dahlman I, Johansson L, Persson L, Berglund J, Pulkki K, Basu S, Uusitupa M, et al. Effects of n–6 PUFAs compared with SFAs on liver fat, lipoproteins, and inflammation in abdominal obesity: a randomized controlled trial. Am J Clin Nutr 2012; 95:1003–12. 61. Hooper L, Abdelhamid A, Moore HJ, Douthwaite W, Skeaff CM, Summerbell CD. Effect of reducing total fat intake on body weight: systematic review and meta-analysis of randomised controlled trials and cohort studies. BMJ 2012;345:e7666. 62. Schneeman BO, Burton-Freeman B, Davis P. Incorporating dairy foods into low and high fat diets increases the postprandial cholecystokinin response in men and women. J Nutr 2003;133:4124–8. 63. Lissner L, Troiano RP, Midthune D, Heitmann BL, Kipnis V, Subar AF, Potischman N. OPEN about obesity: recovery biomarkers, dietary reporting errors and BMI. Int J Obes (Lond) 2007;31:956–61. 64. Sonestedt E, Wirfalt E, Gullberg B, Berglund G. Past food habit change is related to obesity, lifestyle and socio-economic factors in the Malmo Diet and Cancer Cohort. Public Health Nutr 2005;8:876–85. 65. Berglund G, Elmstahl S, Janzon L, Larsson SA. The Malmo Diet and Cancer Study. Design and feasibility. J Intern Med 1993;233:45–51. 66. Méjean C, Droomers M, van der Schouw YT, Sluijs I, Czernichow S, Grobbee DE, Bueno-de-Mesquita HB, Beulens JW. The contribution of diet and lifestyle to socioeconomic inequalities in cardiovascular morbidity and mortality. Int J Cardiol 2013;168:5190–5.