Food Patterns Associated with Intakes of Fat, Carbohydrate and Dietary Fibre

Food patterns associated with intakes of fat, carbohydrate and dietary ®bre in a cohort of Danish adults followed for six years

M Osler1,3 and BL Heitmann2,3

1Department of Social Medicine and Psycosocial Health, University of Copenhagen, Blegdamsvej 3, 2200 Copenhagen N, Denmark; 2Danish Epidemiology Science Center at the Institute of Preventive Medicine, Copenhagen Hospital Corporation, Copenhagen Municipal Hospital, 1399 Copenhagen K. Denmark; and 3The Copenhagen County Centre of Preventive Medicine, Department of Medicine C, Glostrup University Hospital, 2600 Glostrup, Denmark

Objective: To examine associations between food consumption patterns, measured by a short food frequency questionnaire (FFQ), and the intakes of fat, carbohydrates and ®bre over time, and in relation to recommended guidelines. Design: The same 329 individuals had their diet intake measured by a short FFQ and a thorough diet history interview, ®rst in 1987/88, and again six years later in 1993/94. Setting: The County of Copenhagen, Denmark. Subjects: Three hundred and twenty-nine men and women, aged 35±65 y selected randomly from a large population sample. Results: At both examinations fat energy displayed the strongest positive associations with the intake of animal fats and negative correlations with the vegetables. These food items explained most of the total explained variation in fat intake. In general the associations between food items and intakes of carbohydrates and ®bre were similar but inverse, to those found for fat. During the study period median fat energy decreased from 41±38%. A less frequent intake of animal fats over time predicted an increase in fat energy both among men and women, while a more frequent intake of fruit and pasta, and a less frequent intake of cakes was associated with an increase in dietary ®bre. Conclusions: Food items like animal fats, vegetables and certain high starch foods can predict compliance to dietary guidelines for fat and carbohydrates. The study also shows that the food pattern of this Danish cohort has changed in the direction of a more healthy diet during the six years of follow-up. Sponsorship: This study was granted by the Danish Agricultural and Veterinary and Danish Medical Councils and the Danish Health Insurance foundation. Descriptors: diet; epidemiology; methods; longitudinal study

Introduction and an increase in the intake of vegetables (PraÈttaÈla et al, 1992; Osler & Schroll, 1995). The high mortality from coronary heart disease and cancer From a public health point of view it is important to in most Western countries has, in part, been attributed to a know if food consumption measured by a short FFQ follow high fat, low ®bre diet (James et al, 1988), and in the last the recommended dietary guidelines. Indeed, ®ndings from two decades health authorities in these countries have earlier studies have suggested associations between certain strongly recommended that the adult populations reduce food patterns and dietary fat intake. For instance, high fat fat intake to between 30 and 35% of daily energy. Also intakes in the general population in US, Australia, UK and an increase in the consumption of carbohydrates to more Finland have been found to be associated with the use of than 55% of energy, and in dietary ®bre intake to more than full-fat milk, pastry or desserts, table fat, pork, and pro- 3 g/MJ, has been emphasized (National Research cessed meats, while low-fat intakes were characterized by a Council, 1989; The Scienti®c Committee for Food, 1992; frequent intake of high-®bre breads, cereals, ®sh, chicken, National Food Agency, 1993). fruit and vegetables (RaÈsaÈnen & Pietinen, 1982; Ursin et al, Today it is generally accepted that attrition to dietary 1993; Barghurst et al, 1994; Subar et al, 1994; Pryer et al, recommendations may be better, and compliance more easy 1995). However, only in a few studies have food consump- to monitor, if recommendations are formulated in terms of tion patterns associated with other nutrients, such as car- foods rather than nutrients (The Scienti®c Committee for bohydrates and dietary ®bre, been investigated. An exception Food, 1992; National Food Agency, 1993). Several studies to this is the study by Randall et al (1991) who, in the control have used short food frequency questionnaires (FFQ) to population of 2255 subjects from a case-control study of diet monitor intake and changes, in food patterns of popula- and cancer in USA, found that dietary ®bre density was tions, and studies from Denmark and Finland have, for positively associated with intake of salads, whole grain food instance, shown a decreasing trend in the intake of butter, and fruits; associations with intake of high-fat foods and snacks were negative. Furthermore, to our knowledge, there arenostudiesdescribinghowchangesinfoodintakemeasured Correspondence: Dr M Osler Received 1 August 1996; revised 30 January 1997; accepted 14 by a short FFQ associate with changes in nutrient intake in the February 1997 general population. Food patterns and associated nutrients M Osler and BL Heitmann 355 Therefore, the present study aims at examining how measured by the diet history interview increased with food intake, estimated from a short FFQ, associate with the increasing frequency category of the FFQ, indicating that intakes of fat, carbohydrates and ®bre, estimated from a the FFQ was able to identify levels of food intake correctly. thorough diet history interview over time, and in relation to Furthermore, those who reported a less frequent intake by the recommended guidelines. the FFQ in 1993/94 compared with 1987/88 also had lower mean daily intakes according to the diet history information. Thus the short FFQ may be used to monitor changes in Materials and methods food patterns at a group level. Subjects The study included 552 Danish citizens aged 35, 45, 55 and Diet history interview 65 y randomly selected from the general population. All In 1987/88, and six years later in 1993/94, the same trained subjects were invited to a general health examination. Of dietician interviewed all the subjects about their habitual the 552 subjects invited, and (435) 79% agreed to partici- diet using the diet history method. The diet was assessed, pate, and to give a diet history interview. Non participation based on information for the previous month, and average has been described earlier (Heitmann, 1993). daily intakes were calculated from this information. Meal The study was a part of the Danish MONICA project (an patterns, dishes and foods were explored by interview using international study conducted under the auspices of the a detailed pre-coded interview form. Quantities were World Health Organization to monitor trends and determi- explored by the use of food models, photo series, cups nants of mortality from cardiovascular disease) and was and measures. This instrument has been validated earlier carried out in collaboration with the Copenhagen Country (Heitmann, 1993). Nutrient calculations were carried out Centre of Preventive Medicine. The study was approved by with the DANKOST programme, which is derived from the the Ethical Comity of the County of Copenhagen and all Danish Food Composition Tables. This database included participants had provided informed consent. The examina- 835 food items in 1988, but in 1991 it was extended and tion (GEN-MONICA) was carried out from December now comprises 1250 food items. The nutrient values were 1987 to November 1988. After six years, 511 of the 552 not up-dated at this occasion. subjects originally sampled were alive, and could be invited The intake of nutrients was estimated from the diet to a follow-up examination, and 365 (71%) attended. Of history reports. As nutrient intakes are strongly associated these, 329 had also participated in the GEN-MONICA with energy intake, density-based measures were calcu- study. The 106 and 36 subjects who did not participate in lated, for example percent of energy from fat and carbohy- the follow-up or base-line examination, respectively, had drates, and amounts of dietary ®bres (g) per MJ consumed. the same distribution on sex, age, body mass and energy Table 1 gives the distribution of the nutrients at both intake as the 329 subjects who attended both examinations examinations. (all P > 0.20). The present analysis include these 329 subjects, only. This sample was considered to be represen- Statistical methods tative of the total sample. The 163 men and the 166 women The intake of nutrients were approximately normally dis- of the sample, were equally distributed on the four age tributed. Differences in mean values were examined using a groups, and had a mean BMI of 25.2 and 24.6 kg/m2 t-test. The associations between intake of each nutrient, as respectively. assessed from the diet history, and the intake of each food group, as obtained from the FFQ, were evaluated using Questionnaire data Pearsons correlation analyses. Stepwise multiple regression The method of data collection was exactly the same at the was used to select the foods in the FFQ that best predicted two examinations. An extensive questionnaire on socio- intakes of each of the nutrients. The dependent variables demographic variables, lifestyle, and health was completed were the relative intakes of fat, carbohydrates and ®bre, before the general health examination, and the dietary calculated from the diet history interview. The 24 foods history interview. This self-administered questionnaire listed in Table 2 were entered into the equations. The included questions about how often 26 food items were criterion applied for adding new variables to the equation consumed: The alternatives used in the frequency scale was that each new variable had to improve the variance were as follows: never, once a month or less, twice a explained by the model with at least 1.5%. All models were month, once a week, 2±3 times a week, once a day, 2±3 screened for collinearity between the independent variables times a day and 4 times or more daily. In the statistical by variance in¯ation factors. This exercise showed that calculations used here, the scale alternatives were con- collinearity was not a problem. The analyses were carried verted into weekly frequencies. The FFQ did not include out for the whole group, for men and women separately, any quantitative assessments of portion sizes. Some food and furthermore for the age groups older and younger than groups were added because they were not mutually exclu- 51 y. sive, for example sausages included with meat for open The FFQ data was measured on an ordinal scale, and sandwiches, and raw vegetables included with boiled vege- differences in the distribution of food intake frequencies in tables, therefore the present analysis is based on 24 foods. groups of subjects complying, or not complying, with The FFQ was developed for the MONICA-project, and recommended intakes of fat (Fat-energy 35%, or less) served two purposes. The ®rst was to give a qualitative were described using a `non-parametric t-test' (Wilcoxon characterization of the individual's usual food habits for test). use in individual-based follow-up studies. The second was Individual changes in food intake, measured by the FFQ to detect differences in the intake of general food items method were estimated by subtracting the 1993/94 scale- such as meat and vegetables between groups of subjects value from the 1987/88 value. Negative values indicated a (Wahrendorf, 1983; Boeing et al, 1989). The FFQ was more frequent food intake, zero values meant no change, validated as a part of the present study (Osler & Heitmann, and positive values indicates a less frequent food intake. 1996). This validation showed that the mean food intake Changes in nutrient intake re¯ected by the diet history Food patterns and associated nutrients M Osler and BL Heitmann 356 interview were assessed in the same manner. The correla- with fat energy percent in men, only (Table 2). Animal fats tions between the two measures of individual changes in (r 0.20 and r 0.28) and vegetables (r 70.22 and food intake were estimated in each food group using the r 70.20) were associated with fat energy in both the Pearson correlation coef®cient. A stepwise regression ana- younger and the older age groups, respectively. Sausages lysis was used to identify changes in intake of individual 24 (r 0.32), white ryebread (r 0.22), and meat (r 0.23) food items that best predicted changes in relative intakes of were associated with fat density in the older age-group fat, carbohydrate and ®bre. only, while ®sh (r 70.19), fruit (r 70.20) and rice (r 70.26) were associated with fat energy percent in the younger age group, only. Results In both men and women, and in both age groups, 20% of subjects complied with the recommended fat intake of 35% Energy from fat or less. Compliers used animal and vegetable fats, sausages, Table 1 shows that fat energy intake decreased during the meat, light bread, cakes and candy less often, and ®sh, fruit, study period in both men and women. Pearson correlation and vegetables more often than non-compliers (Table 3). coef®cients between food groups, as assessed from the A multivariate analysis showed that the variation in FFQ, and fat energy percent, as assessed from the diet percent fat energy intake in 1987/88 was best explained history, are presented in Table 2. A few strong associations by animal fats and vegetables. Sixteen percent of the will be noted in the following. variation in percent fat energy intake was explained by 7 food items (Table 4). In men consumption of dark ryebread, Food intake at baseline sausages, rice and vegetables signi®cantly predicted fat In 1987/88, the strongest positive correlation with fat intake, while animal fats, oatmeal, cakes, pasta, ice cream energy intake percent was seen for the food group animal and vegetables were predictors of fat intake in women. fats, whereas the strongest negative correlation was Vegetables, ice cream, animal fats and rice appeared in the observed for vegetables. Fruit, cakes, oatmeal, and candy model for the younger age groups, while animal fats, were associated with fat density in women, only, while dark sausages, cakes and vegetables were included in the and white ryebread, white bread, and rice were associated model for the older age groups.

Table 1 Median, 25th and 75th percentiles of energy, and absolute and Food intake at follow-up relative intakes of fat, carbohydrate and dietary ®bres in 329 Danish men Bivariate analysis of the data from 1993/94 showed that and women estimated from a diet history interview obtained in 1987/88 compared to the baseline data in 1987/88 most of the and 1993/94. Difference in intakes at baseline and follow-up given (P-values) correlations with animal fats, fruit and vegetables were unchanged (Table 2). From baseline to follow-up, the Percentiles period under study the percent complying with the recom- 25 50 75 P-values mendation had increased from 20±31%. This increase was consistent in all sex and age groups. At follow-up, the Energy MJ/d differences in eating patterns between compliers and non- 1987/88 6.8 8.3 10.4 compliers were nearly the same as those found six years 1993/94 7.2 8.5 10.8 0.07 Men earlier (Table 3). A multivariate analysis showed that 1987/88 8.2 10.2 11.9 animal fat remained a predictor of fat intake. Approxi- 1993/94 8.6 10.5 11.9 0.24 mately 22% of the variation in percent fat energy intake Women was explained by 4 food items. Meat and animal fats were 1987/88 5.9 7.2 8.3 signi®cant predictors of fat intake in men, while white 1993/94 6.4 7.5 8.5 < 0.01 Fat g/day (Fat energy %) ryebread, vegetables, sausages, ®sh and rice predicted fat 1987/88 70(36) 89(41) 109(44) intake in women (Table 4). Vegetables and pasta appears in 1993/94 68(34) 84(38) 110(43) 0.9( < 0.01) the models for the youngest age group. Animal fat, rice, Men and dark ryebread were included in the model for the older 1987/88 88(37) 101(41) 131(44) 1993/94 79(35) 103(39) 127(43) 0.06( < 0.01) age group and explained 45% of the variation of percent fat Women energy intake. 1987/88 59(36) 76(41) 90(43) 1993/94 62(34) 73(38) 89(42) 0.89( < 0.01) Changes in food intake Carbohydrates g/d (carbohydrate energy %) 1987/88 157(35) 190(40) 243(45) During the study period the consumption frequency of dark 1993/94 173(37) 207(42) 259(47) < 0.01( < 0.01) ryebread, oatmeal and pasta increased, while it decreased Men for animal fats, vegetable margarine, cheese and vegeta- 1987/88 177(35) 224(38) 279(44) bles. At the same time 58 subjects (18%), who were non- 1993/94 192(35) 241(41) 296(45) < 0.01( < 0.01) compliers at baseline, complied with fat recommendations Women 1987/88 138(37) 170(42) 206(46) six years later, while 180 subjects (54%) persisted as non- 1993/94 154(38) 189(44) 216(48) < 0.01(0.02) compliers. Those who became compliers had increased Dietary ®bre g/d (®bre density g/MJ) their intake of pasta, and decreased their intake of animal 1987/88 16(2.0) 21(2.5) 26(3.1) fats more than subjects being non-compliers at both exam- 1993/94 16(1.9) 20(2.3) 25(2.8) 0.34( < 0.01) inations. In contrast, those who were non-compliers both at Men 1987/88 18(1.9) 23(2.3) 31(2.9) base-line and follow-up had increased their intakes of cake 1993/94 18(1.8) 23(2.3) 28(2.7) 2.6( < 0.01) and candy. Women Individual changes in food consumption over the 6 year 1987/88 15(2.2) 19(2.7) 24(3.4) period were found to be related to changes in fat energy 1993/94 15(2.0) 18(2.4) 22(3.0) 0.23( < 0.01) intake. For instance, a more frequency intake of animal fats P-values for relative intakes in parenthesis. was associated with an increase in percent fat energy in Food patterns and associated nutrients M Osler and BL Heitmann 357 Table 2 Associations between food intake frequency and fat energy in 329 Danish men and women at baseline and follow-up

Coef®cient of correlation

1987/88 1993/94

All Men Women All Men Women

Fats and dairy foods Animal fats 0.24* 0.19* 0.29* 0.30* 0.28* 0.31* Vegetable m 0.18* 0.16 0.10 0.11 0.11 0.10 Low-fat m 70.08 70.09 70.06 70.06 70.06 70.06 Cheese 0.05 0.05 0.05 0.01 70.03 0.04 Milk, yoghurt 0.03 70.01 0.08 70.07 70.18 70.01 Meats, eggs, ®sh Eggs 0.02 70.03 0.11 0.07 0.05 0.10 Fish 0.02 0.08 70.15 70.14* 70.16 70.13 Meat 0.12 0.05 0.11 70.01 0.12 70.14 Sausages 0.20* 0.19* 0.20* 0.12 0.08 0.19 Fruits and vegetables Fruit 70.17* 70.15 70.18* 70.19* 70.08 70.28* Juice 70.06 0.03 70.16 70.09 70.06 70.13 Vegetables 70.22* 70.20* 70.21* 70.22* 70.18 70.26* Breads, cereals, starches Coarse bread 70.12 70.16 70.08 70.06 70.09 70.07 Dark ryebread 70.12 70.25* 0.02 70.10 70.14 70.07 Light bread 0.13 0.19* 0.06 0.14 0.17 0.12 White ryebread 0.15* 0.23* 0.08 0.15 0.07 0.01 Oatmeal 70.10 70.01 70.20* 70.18* 70.21* 70.14 Pasta 70.05 70.12 70.09 70.10 70.15 70.05 Potatoes 0.06 0.07 0.02 0.04 0.01 0.07 Rice 70.17* 70.23* 0.01 70.15* 70.10 70.19 Baked goods, sweets Cakes, biscuits 0.14* 0.03 0.24* 0.08 70.01 0.14 Candy, chocolate 0.14* 0.03 0.21* 0.13* 0.06 0.18 Jam, honey 70.01 0.01 70.01 0.02 0.03 0.02 Ice cream, soda 70.08 70.03 70.14 0.01 0.06 70.12

Animal fats butter, non-vegetable margarine, lard; m margarine; soda carbonated beverages. *P < 0.01 Pearson correlation analysis. both men and women (r 0.19). From the multivariate Relative intake of dietary ®bres analyses it could be estimated that for each time the weekly In 1987/88, animal fats and vegetables, and in 1993/94, consumption frequency of animal fats increased, fat energy fruit and animal fats explained nearly 25% of the total intake increased with 0.22% (95% CI: 0.12±0.43). variation in relative intake of ®bre (Table 6). In 1993/94, the foods predicting nearly 40% of the variation in the Energy from carbohydrates relative intake of ®bre intake were milk, fruit and animal Associations between foods and percent energy from car- fats in the young age group, animal fats, meat and vege- bohydrates, and the relative intake of dietary ®bres, were tables in the old age group. examined using the same analytic approach as for fat. The Although the intake of energy from carbohydrates correlation and regression analyses pointed at the same increased, the median relative intake of dietary ®bres foods as predictors of carbohydrate intake. Hence, only decreased 0.2 g/MJ during the six years study period. results from the regression analysis will be presented in the From the multiple regression analysis it could be predicted following. In general, associations were similar but inverse, that for each increase in weekly consumption frequency of to those found for fat. cakes, the intake of ®bre decreased with 0.09 g/MJ (95% In both 1987/88 and 1993/94, fruit, animal fats, meat CI: 0.03±0.14), while a more frequent intake of fruit and and jam predicted the intake of percent energy from pasta increased dietary ®bre intake by 0.04 (95% CI: 0.01± carbohydrates in the whole study group. At both examina- 0.09) and 0.20 (95% CI: 0.14±0.27) g/MJ, respectively. tions, there were variables whose explanatory power was related to a speci®c gender (Table 5). In 1987/88, animal Discussion fats, fruit, oatmeal and meat predicted carbohydrate intake in the older age group. Six year later, 2 foods (fruit and The aim of the present analysis was to examine what foods pasta) appeared in the models of the youngest age group. from a short FFQ predict intake and change in fat, carbo- Meat, white and dark ryebread were included in the model hydrate and dietary ®bre. The correlations between food of the older age group. intake measured by the FFQ, and nutrient intakes, assessed Median carbohydrate intake increased 1.8 energy% from by diet history, were rather weak, as might be expected 1987/88 to 1993/94. In bivariate analysis changes in the when comparing data from two different dietary methods. intake frequency of juice (r 0.14), and meat (r 70.14) This may be the reason why only a limited number of foods were associated with an increase in carbohydrate energy in the FFQ accounted for a rather small proportion of the intake, while none of the changes in food consumption over variance in the nutrients. However, when we replicated the the six year period were found to predict changes in the analyses using comparable food groups derived from the multivariate analysis. diet history data nearly the same results were obtained, Food patterns and associated nutrients M Osler and BL Heitmann 358 Table 3 Average food intake frequencies in men and women complying or non-complying, with recommended guidelines for fat energy intake in 329 Danish men and women Mean food intake frequency (times per week)

1987/88 1993/94

Compliers Non-compliers All Compliers Non-compliers All (n=64) (n=265) (n=329) (n=101) (n=228) (n=329)

Fat and dairy foods Animal fats 5.6 9.2** 8.5 5.9 9.3** 8.2* Vegetable m 4.0 7.3** 6.5 4.8 5.5 5.3* Low-fat m 3.3 4.8 3.7 4.5 3.5 3.8 Cheese 6.9 7.2 7.1 6.3 6.6 6.5* Milk, yoghurt 7.1 6.9 7.0 2.1 6.5 6.7 Meats, eggs, ®sh Eggs 2.1 2.1 2.1 1.7 1.8 1.8 Fish 1.1 0.9** 1.0 1.3 1.0** 1.0 Meat 4.6 5.2 5.0 4.6 5.0 4.9 Sausages 4.0 5.4** 5.1 4.5 5.0** 4.8 Fruits and vegetables Fruit 7.7 5.2** 5.7 7.6 4.8** 5.7 Juice 2.5 1.9 1.9 2.5 1.9 2.1 Vegetables 2.0 8.1** 8.8 7.2 4.8** 5.6* Breads, cereals, starches Coarse bread 4.5 4.3 4.3 5.3 4.3 4.6 Dark ryebread 5.9 8.1 8.4 9.4 8.6 8.8* Light bread 4.5 3.8** 3.6 2.9 3.6 3.4 White ryebread 1.9 3.5** 3.1 2.6 4.1** 3.7 Oatmeal 1.5 1.1 1.2 3.0 1.4** 1.9* Pasta 0.6 0.4 0.5 1.0 0.9 0.9* Potatoes 3.7 4.4 4.3 4.0 4.0 4.0 Rice 1.1 0.8 0.9 1.2 1.0 1.0 Baked goods, sweets Cakes, biscuits 1.3 2.3** 2.1 2.3 2.6 2.5 Candy, chocolate 1.0 1.8** 1.6 1.2 2.1** 1.8 Jam, honey 3.2 3.2 3.2 3.0 3.2 3.1 Ice cream, soda 1.9 1.6 1.6 1.5 1.9 1.7

Animal fats butter, non-vegetable margarine, lard; m margarine; soda carbonated beverages. **P < 0.01: Wilcoxon test compliers (Fat energy 35%) vs non-compliers (fat energy > 35%) *P < 0.01: Wilcoxon test 1987/88 vs 1993/94.

indicating that only a smaller part of the variance should be 1994; Pryer et al, 1995). The relative consistency between explained from food intake. It should also be noticed that these studies and the present study suggests that, although many statistical tests were carried out, and hence signi®- the single food items may be of speci®c importance in cance, due to the play of change, is a possible bias. The different countries, healthy eating patterns seem predicted results from the present population based study suggest that by nearly the same groups of food, at least in Westernized certain food patterns associate with dietary guidelines for societies. fat, carbohydrate and ®bre in Danish adults. For instance, The food group including milk products, which is foods like animal fats, vegetables, fruit, coarse bread and among the main contributors to fat-energy in the Danish oatmeal were included in most predictive models. This diet, was not found to be associated with fat intake in the does not indicate that any single food item, oatmeal, for present study. This is in agreement with a study from USA instance, does, in itself, constitute an important source of (Subar et al, 1994), in which total intake of dairy products fat or carbohydrates in the Danish diet. Rather, the results was found to be unassociated with fat energy too. However, indicate that such foods serve as indicators of a certain food after they sub-divided the dairy product categories into pattern. Hence, the use of oatmeal may indicate an attempt low-fat and whole milk products, low-fat milk was nega- to choose foods with a lower fat, and higher ®bre content. tively, and whole milk positively correlated with fat intake. Vegetables, fruit and oatmeal appear to contribute to a diet Unfortunately, the FFQ used in the present study did not both low in fat and high in carbohydrates. Possibly because allow us to sub-divide the milk group further. they are rich in carbohydrates and substitute carbohydrate The different explanatory models for predicting nutrient energy for fat energy, and also possibly because they are densities in men and women, and in people of different part of a dietary pattern that includes low fat food items. In ages, indicate that food patterns are different among people the same way marker foods, such as animal fats, meat and living in the same place, and that they change over a milk predicted a low ®bre intake. Presumably, some of relatively short period of time. These results are in agree- these relationships were due to collinearity and reciprocal ment with a Finnish study of 154 men and women, in which associations among food groups, and such patterns of a 24 item FFQ was related to nutrient intakes estimated consumption should not be used uncritically to advocate from four days dietary records. This analysis also found sex for dietary changes. and age differences in food patterns, and the intake of total In other populations, a high fat intake has been asso- fat was explained by bread and pork in men, but by bread ciated with a frequent use of table fat, and an infrequent use and coffee with cream in women (RaÈsaÈnen & Pietinen, of cereals, fruit and vegetables, too (RaÈsaÈnen & Pietinen, 1982). Furthermore, although the general patterns of diet 1982; Ursin et al, 1993; Barghurst et al, 1994; Subar et al, intake of the different cultures still could be recognized at Food patterns and associated nutrients M Osler and BL Heitmann 359 Table 4 Stepwise multiple regression analysis (Regression coef®cients (b), 95% Con®dence Intervals (95%CI), and Cumulative varians (R2)) of associations between intake frequency of the food items and fat energy percent in 329 Danish men and women at baseline and follow-up

Food items 1987/88 1993/94

b 95% CI R2 b 95% CI R2

All Animal fats 0.2** (0.1±0.3) 0.07 0.3** (0.2±0.4) 0.14 Vegetables 70.2** (70.3±0.0) 0.09 Ð Cakes, biscuits 0.3** (0.1±0.5) 0.11 Ð Pasta 71.1** (72.1±70.6) 0.13 Ð Ryebread, white 0.2** (0.0±0.2) 0.14 Ð Oatmeal 70.3* (70.5±70.0) 0.15 Ð Ice cream, soda 70.1* (70.3±0.0) 0.16 0.3* (0.0±0.5) 0.22 Rice 71.9** (72.9±70.8) 0.16 Meat 0.4** (0.1±0.7) 0.20 Men Ryebread, dark 70.3** (70.4±70.1) 0.11 Ð Sausages 0.3** (0.1±0.6) 0.15 Ð Rice 71.7** (72.7±70.6) 0.17 Ð Vegetables 70.1* (0.0±70.2) 0.19 Ð Animal fat 0.3** (0.1±0.6) 0.13 Meat 0.5** (0.1±0.9) 0.16 Women Animal fat 0.3** (0.2±0.4) 0.10 Ð Oatmeal 70.5** (70.9±70.2) 0.12 Ð Cakes, biscuits 0.4** (0.2±0.6) 0.16 Ð Pasta 71.3* (72.4±70.1) 0.19 Ð Ice cream, soda 70.2* (70.6±0.0) 0.20 Ð Vegetables 70.1* (70.3±0.0) 0.22 70.2** (70.4±70.1) 0.30 Fish 71.3** (72.2±70.4) 0.36 Rice 71.5** (72.7±70.3) 0.41 Ryebread, white 0.7** (0.4±1.1) 0.22 Sausages 0.4* (0.1±0.8) 0.43

* P < 0.05; ** 0.05 < P 0.01. old age, the variation between them had become smaller this context it should be noted, however, that the changes in (Huijbregts et al, 1995). In constrast, in a study of the Danish food patterns in the direction of a more healthy diet, dietary patterns and associated fat intake in 20.243 US are seen both in studies with a stable age structure and in adults, results by demographic sub-groups showed few Danish food supply statistics. Indeed, from 1988±1990 the differences from those found in the total population. yearly use of pasta products doubled from 0.9 kg/capita to These study analyses were based on bivariate analyses of 1.8 kg/capita. In the same period the use of butter and correlations between foods and fat intake (Ursin et al, margarine decreased from 23.9 kg/capita to 20.7 kg/capita 1993), so these results may not be comparable with results (Fagt & Groth, 1992). This is also in agreement with results from our study. from a longitudinal study of ageing cohorts of men in The present study suggests that food patterns and nutri- Finland, Italy and the Netherlands, in which dietary pat- ent intakes in the Danish diet changed over a relatively terns of the Finish and Dutch men also changed towards a short period of time, in the direction of a more healthy diet. more healthy diet (Huijbregts et al, 1995). Furthermore, we During the study period median fat energy intake decreased showed, in another study examining the validity of the FFQ from 41±38%, and the percent of subjects who complied that, over time, similar changes in food intake were with recommended intakes increased from 21±31%. The re¯ected by the FFQ and by the diet history method decline in fat consumption, succeeds a trend of increase in (Osler & Heitmann, 1996). fat intake that has taken place since the 1950s (Fagt & Groth, 1992). The present decline in fat intake was accom- panied by a less frequent use of, in particular, animal fats. Conclusion Furthermore, subjects who became compliers of recom- The present study indicates that intake of fat, carbohydrate mendations had a more frequent intake of pasta. The and ®bre may be monitored by a few questions on con- reported changes in food and nutrient intakes could be an sumption frequency of certain food items. In this context, it artefact or re¯ect real changes. Artefactual trends may was found that foods like animal fats, vegetables, fruit and result from errors due to changes in methodology or certain cereals are predictors of compliance to dietary reporting. We ®nd it unlikely that the changes over time, guideline for fat and carbohydrates. However, the predic- revealed by the present study, are explained by differences tion of nutrient intakes by the present FFQ may be in methodology. The data was collected at the same place, improved by sub-dividing food groups, like milk and by the same dietician, with the same questionnaires and meat, into low and high fat products. using the same food database. On the other hand, we cannot The present study also shows that the food patterns of exclude that an increased awareness of diet intake, acquired this Danish cohort have changed in the direction of a more during participation in a study on diet and health, or healthy diet. It seems that changes in fat intake may be changes in food choice related to aspects of ageing, may estimated by a few foods. However, due to the variation in in part explain the observed changes in dietary patterns. In food consumption in different sex and age groups, and to Food patterns and associated nutrients M Osler and BL Heitmann 360 Table 5 Stepwise multiple regression analysis (Regression coef®cients (b), 95% Con®dence Intervals (95%CI), and Cumulative varians (R2)) of associations between intake frequency of the food items and carbohydrate energy percent in 329 Danish men and women at baseline and follow-up

Food items 1987/88 1993/94

b 95% CI R2 b 95% CI R2

All Fruit 0.4** (0.2±0.5) 0.11 0.3** (0.1±0.5) 0.12 Oatmeal 0.5** (0.2±0.8) 0.15 Ð Animal fats 70.2* (70.3±70.1) 0.19 70.2** (70.3±70.1) 0.22 Meat 70.4* (70.7±70.1) 0.21 70.5** (70.8±70.1) 0.26 Jam, honey 0.2* (0.1±0.4) 0.23 0.5* (0.2±0.9) 0.19 Rice 2.2** (1.0±3.4) 0.15 Ice cream, soda 70.4* (70.6±70.1) 0.29 Men Fruit 0.4** (0.2±0.6) 0.10 0.4** (0.1±0.8) 0.09 Candy, chocolate 0.9** (70.3±71.5) 0.15 Ð Animal fats 70.2** (0.3±0.1) 0.18 Ð Cakes, biscuits 0.4** (0.1±0.8) 0.21 Ð Coarse bread 0.2* (0.0±0.4) 0.24 Ð Jam, honey 0.5** (0.1±1.0) 0.24 Ð Pasta 2.2** (0.1±4.3) 0.19 Women Oatmean 0.8** (0.4±1.1) 0.12 Ð Fruit 0.2** (0.1±0.4) 0.17 Ð Animal fats 70.2** (70.3±70.1) 0.21 Ð Juice 0.3* (0.1±0.6) 0.25 0.3* (0.0±0.7) 0.45 Ryebread, white 70.2** (71.0±70.1) 0.16 Vegetables 0.3* (0.0±0.5) 0.25 Fish 1.5** (0.5±2.7) 0.32 Rice 1.9** (0.5±3.3) 0.36 Sausages 70.5* (70.9±70.1) 0.41

* P < 0.05; ** 0.05 < P < 0.01.

Table 6 Stepwise multiple regression analysis (Regression coef®cients (b), 95% Con®dence Intervals (95%CI), and Cumulative varians (R2)) of associations between intake frequency of the food items and relative intake of ®bre in 329 Danish mean and women at baseline and follow-up

Food items 1987/88 1993/94

b 95% CI R2 b 95% CI R2

All Vegetables 0.5** (0.3±0.6) 0.11 Ð Animal fats 70.2** (70.3±70.1) 0.25 70.2* (70.4±70.6) 0.24 Ryebread, dark 0.2** (0.1±0.4) 0.28 0.2* (0.0±0.4) 0.29 Meat 70.5* (71.0±70.1) 0.30 Ð Oatmeal 0.5** (0.2±0.9) 0.32 0.3* (0.1±0.6) 0.38 Milk 70.1 (70.2±70.0) 0.33 70.3** (70.9±70.1) 0.31 Fruit 0.7** (0.4±0.9) 0.20 Sausages 70.4** (70.7±70.1) 0.27 Candy, chocolate 70.8** (71.5±70.1) 0.33 Coarse bread 0.3* (0.1±0.5) 0.35 Men Fruit 0.6** (0.2±1.0) 0.18 0.5** (0.2±0.9) 0.15 Ryebread, dark 0.4** (0.2±0.8) 0.27 0.3** (0.1±0.5) 0.48 Juice 70.4* (70.8±70.2) 0.30 70.7** (71.1±70.3) 0.43 Animal fats 70.2* (70.4±70.1) 0.35 70.3** (70.5±70.1) 0.29 Coarse bread 0.4** (0.1±0.7) 0.23 Vegetables 0.5** (0.2±0.8) 0.35 Ice cream, soda 70.4** (70.7±70.1) 0.40 Oatmeal 0.4* (0.0±0.8) 0.51 Women Vegetables 0.4** (0.2±0.6) 0.14 Ð Animal fats 70.3** (70.5±70.1) 0.20 Ð Candy chocolate 70.5** (70.7±70.2) 0.25 71.3** (72.0±70.1) 0.32 Fruit 0.6** (0.3±0.9) 0.19 Sausages 70.5** (71.1±70.1) 0.24 Juice 0.5* (0.0±1.0) 0.29 Rice 1.9** (0.1±3.8) 0.36 Milk 70.2* (70.5±0.0) 0.40

* P < 0.05; ** 0.05 < P < 0.01. Food patterns and associated nutrients M Osler and BL Heitmann 361 changes in consumption patterns over time, it may be National Food Agency (1993): Nordic Food Recommendations. necessary to further establish the predictive value of the National Research Council (1989): Committee on diet and health. Food and Nutrition Board. Commission of Life Sciences. Diet and Health. FFQ in assessing nutrient intake for different sub-groups, Implications for Reducing Chronic Disease Risk. Washington, DC: and over time. National Academy Press. Osler M & Heitmann BL (1996) The validity of a short fool frequency AcknowledgementsÐThis work was done in collaboration with the questionnaire and its ability to measure changes in food intake. A Research Department of Human Nutrition of the Royal Veterinary and longitudinal study. Int. J. Epidemiol. 44, 1023±1029. Osler M & Scroll M (1995): Lifestyle and prevention of ischaemic heart Agricultural University, Denmark. The project was supported by grants disease in Denmark. Changes in knowledge and behaviour 1982±1992. from the Danish Agricultural and Veterinary and Danish Medical Councils Eur. J. Public Health 5, 109±112. and the Danish Health Insurance Foundation. We are grateful to Dr. Kim Pryer J, Brunner E, Elliott P, Nichols R, Dimond H & Mammot M (1995): Overvad for his critical and inspiring comments. Who complied with COMA 1994 dietary fat recommendations among a nationally representative sample of British adults in 1986±7 and what did they eat? Eur. J. Clin. Nutr. 49, 718±728. References PraÈttaÈla R. Berg M-A & Puska P (1992): Diminishing or increasing constrasts? Social class variation in Finnish food consumption patterns Barghurst KI, Barghurst PA & Record SJ (1994): Demographic and dietary 1979±1990. Eur. J. Clin. Nutr. 46, 279±287. pro®les of high fat consumers in Australia. J. Epidemiol. Comm. Health Randall DE, Marshall JR, Brasure J & Graham S (1991): Patterns in food 48, 26±32. use and compliance with NCI dietary guidelines. Nutr. Cancer 15, 141± Boeing M, Wahrendorf J, Heinemann L, Kilesza W, Rywik SL, Sznajd J & 158. Theil C (1989): Results from a comparative dietary assessment in RaÈsaÈnen L & Pietinen P (1982): A short questionnaire method for Europe: I Comparison of dietary information derived from concurrently evaluation of diets. Prev. Med. 11, 669±676. applied frequency questionnaires and quantitative measures. Eur. J. Subar AF, Ziegler RG, Patterson BH, Ursin G & Graubard B (1994): US Clin. Nutr. 43, 367±377. dietary patterns association with fat intake: The 1987 National Health Fagt S & Groth MV (1992): Trends in Danish Food Supply 1955±1990. Interview Study. Am. J. Public Health 84, 359±366. National Food Agency. The Scienti®c Committee for Food (1992): Nutrient and Energy Intake for Heirmann BL (1993): The in¯uence of fatness, weight change, slimming the European Community. Brussels: European Union. history and other lifestyle variables on diet reporting in Danish men and Ursin G, Ziegler RG, Subar AF, Graubard BI, Haile RW & Hoover R women aged 35±65 years. Int. J. Obes. 17, 329±336. (1993): Dietary patterns associated with a low-fat diet in the National Huijbregts PPCW, Feskens EJM, RaÈsaÈnen L, Alberti-Fidanza A, Mutanen Health Examination follow-up study: Identi®cation of potential M, Fidanza F & Kromhout D (1995): Dietary intake in ®ve ageing confounders for epidemiological analyses. Am. J. Epidemiol. 137, cohorts of men in Finland, Italy and the Netherlands. Eur. J. Clin. Nutr. 916±927. 49, 852±860. Wahrendorf J (1983): IARC proposal for a cohort study on diet and cancer James WPT, Ferro-Luzzi A, Isaksson B & Szostak WB (1988): Healthy using MONICA sample surveys and cancer registries. In Surveillance of Nutrition. Preventing Nutrition-Related Diseases in Europe. WHO Dietary Habits of the Population with Regard to Cardiovascular Regional Publications, European Series, no 24. Copenhagen: WHO Diseases, eds. De Backer GG, Tunstall-Pedoe H & Ducimetiere P, Regional of®ce for Europe. pp. 97±99, EURO-NUT Report 2.