A cross-sectional study of ninth grade students (aged 13–14) from lower secondary schools in Vest-Agder County, Norway, was performed from September to November 2012. All secondary schools (grades 8–10) in five communities were invited to participate in the study. Of the 15 invited schools, nine (60%) agreed to participate in the study. All schools found the project interesting, and the main reason for non-participation was lack of time. In total, 742 children were invited to participate in the study. Information about the study was given through several information channels; oral information was given to the students on an information day at school, two information letters were given to the students, one for themselves and one that for their parents. The schools also informed the parents through students’ weekly plans and via SMS. Parents gave written consent on the projects website. In total, 531 students were given permission to participate, which represented 71.6% of the invited ninth-grade students. Of these, 517 completed the food frequency questionnaire, giving a participation rate of 69.7%. Non-responses among students were due to absence on the day the survey was administered. All classes who participated in the study received a gift of 1,000 NOK. An ID number, given to both parents and students, was used to link the questionnaires. The number of parents who responded to the questionnaire was 335, a response rate of 45.1%. The total number of dyads with registered information on both parents and students was 308, giving a response rate of 41.5% of the invited participants and 59.6% of the students who filled out the questionnaire. These 308 dyads, including 131 boys, 177 girls and their parents, represent the total sample in this study. This study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human subjects were approved by The Norwegian Social Science Data Services. Written consent was obtained from the parents of all participating adolescents and the adolescents themselves.

The students’ main characteristics are shown in Table 1. The students’ mean age was 13.9 years and 11.6% of the students were classified as overweight according to Cole et al.'s classification (24). A sociodemographic description of the parents is shown in Table 2. Of the participating parents in the study 12.3% were single parents, 23.1% had a lower family income than 600,000 NOK of those 86.8% single parents, and 76.9% had a higher family income than 600,000 NOK. Lower education (≤12 years of education) was reported by 41.6% of the mothers and 48.7% of the fathers, and higher education was reported by 58.4% of the mothers and 51.3% of the fathers. More girls than boys (63.4% vs. 51.5%, p=0.037) had mothers with high education. There were no such differences for fathers’ education. Ethnic origin was not used in the analysis as the vast majority of respondents were of Norwegian origin (92.8%).

Both questionnaires, for parents and students, were pre-tested for clarity and length among a small sample of parents and students in the same age group and in the same part of Norway prior to the main study. These were not participants in the main study.

The food and drink intake of the students was reported by the student themselves using a validated web-based retrospective food frequency questionnaire (25, 26). The questionnaire was completed at school, in the presence of two project workers. It took about 45 min to complete the questionnaire. Respondents indicated how frequently they had consumed 112 food items and 20 types of drinks during the last 4 weeks. The FFQ measured one item for each question. The present analyses are based on a subset of 30 items from the FFQ, which were categorized into six food groups: fruit (f), vegetables (v), total fruits and vegetables, sugar-sweetened soft drinks, fish and fast food. These were selected based on current dietary recommendations and due to their importance to health. Examined fruit variables contained apples, pears, banana, orange/tangerine/grapefruit, nectarine/peach/plum, dried fruit, melon, kiwi, pineapple, fresh/frozen berries, grapes, and raisins. Vegetables overall contained broccoli, cauliflower, onion/garlic/leek, avocado, corn, mushrooms, peas, mixed lettuce, peppers, carrots, and cucumbers. Fish for dinner contained the variables fatty fish, lean fish, and other fish products such as fish balls, fishcakes, fish sticks, and puddings. Fast food contained the variables sausages/frankfurters, hamburgers, pizza and French fries. The response categories for fruit and vegetables were ‘never’, ‘1–3 times a month’, ‘once a week’, ‘2–3 times a week’, ‘4–6 times a week’, and ‘1 or more times a day’. Before merging into new variables all fruit and vegetables items were re-coded to reflect consumption in times per week (0, 0.5, 1, 2.5, 5, 10). After merging the variables, it was divided by seven to reflect consumption of times per day. The seven response categories for consuming a bottle (0.5 l; standard portion) of sugar-sweetened soft drinks were ‘never’, ‘1–3 bottles a month’, ‘1 bottle a week’, ‘2–6 bottles a week’, ‘1 bottle a day’ and ‘2–3 bottles a day’, ‘more than 3 bottles a day’. The response categories were re-coded to reflect consumption in times per week (0, 0.5, 1, 4, 7, 17.5, 25). To reflect consumption in liters per week the variable was divided by two. Fish and fast food variables were given the alternative frequencies ‘never’, ‘1–3 times a month’, ‘once a week’, ‘2–4 times a week’, ‘more than three times a week’. To reflect consumption per week, these variables were re-coded into 0, 0.5, 1, 3, 5. Furthermore, the questionnaire also included questions about background information such as gender, age, ethnicity, residential area, height, and weight. To classify the students’ weight status, Coles’ standard definition of sex- and age-specific cutoff points similar to the widely used 25 kg/m² for overweight in adults (24), for children aged 13 and 14, were used. These cutoffs are 21.91 and 22.58 kg/m² for 13-year-old boys and girls, respectively, and 22.62 and 23.34 kg/m² for 14-year-old boys and girls, respectively.

To assess socioeconomic information, the parents filled in a web-based questionnaire at the website of the project. SES was determined using family income and parental education. Income is measured as the total income of the family before taxes, not adjusted for number of adults or children living in the household. Income was given five response categories: <NOK 300,000; NOK 300–600,000; NOK 600,000–800,000; NOK 800,000–1,000,000; >NOK 1,000,000. Family income was dichotomized using NOK 600,000 as a cutoff for lower respective higher income. Parental education, both mother and father, is measured as respondents’ total years of education and was grouped as a four-category classification: ‘primary school’ (9 years or less); ‘upper secondary education or vocational school’ (3 years of secondary education); ‘university or university college’ (4 years or less); and ‘university or university college’ (more than 4 years). In analysis, the four parental levels of education were dichotomized into ‘equal or less’ than 12 years of education’ and ‘completed college or university education’.

Parents reported their knowledge of dietary guidelines by the question ‘Do you know the dietary guidelines from the Norwegian Directorate of Health?’ This question had three response categories: ‘yes, to a large extent’, ‘yes, little/some extent’ and ‘no’. In analysis, these categories were presented as two categories where ‘good’ includes ‘yes, to a large extent’, and ‘less or no’ knowledge includes the two last categories.

All statistical analyses were performed with SPSS statistical software package version 19.0 (IBM Corporation, Armonk, NY, USA). The description of participants and their diet are presented as median with interquartile range or percentages using descriptive statistics. Given the skewed nature of the data, analyses of between group differences in dietary habits were conducted using non-parametric statistics, Mann Whitney U-tests. Pearson's Chi-square tests, χ², were used when analyzing the differences between categorical variables, knowledge of dietary guidelines and SES. A p-value of ≤0.05 was considered to be statistically significant. All respondents were included in the analysis. Only a few missing were registered in the questionnaires, and these were treated as missing in the analysis using exclude cases listwise.

Among the strengths of the present study is that we have reports from two different sources, that is, from both students and parents. That the parents report information of SES gives us more reliable data and less missing data, than if the students themselves had reported this (44). At baseline, the study has a relative high response rates and large sample sizes of schools and students, which minimized the risk of selection bias. Another strength is that the FFQ used in this study has been recognized as a valid instrument in epidemiological studies ranking adolescents according to their usual food intake (25, 26). Another advantage of this study is few missing data on diet and SES, partly due to design of questionnaires. A possible strength of our study is that socioeconomic status is presented as two parameters, that is, parental education and income. Research on diet and socioeconomic inequalities most often include education as an indicator of SES (45). However, education and income cover different areas of social stratification (45) and both are important to include. The relation between education and dietary and health habits (45) are most often explained by the person's general knowledge which may influence dietary choices. Income is partly related to education level however also include the person's social position (46). Marmot argue that there are two ways income may be related to health, through a direct effect on the material conditions necessary for biological survival (e.g. food), and through an effect on social participation and opportunity to control life circumstances (47). Our study only presents relations between diet and income; however, future studies could further explore if there are causal relations between income and dietary habits.

Several methodological limitations of this study must be considered. Despite a high response rate among students, selection bias may have been introduced as a consequence of lower parental response rates. The reason for lower participation of parents is uncertain and may be due to many factors. Because we were unable to determine parental socioeconomic status among the non-respondents, the absence of many parents may influence the study's ability to generalize to the population. A disadvantage of this study is that information and questionnaires only were available in Norwegian, which may have influenced the inclusion of all social groups, especially those with foreign backgrounds. Although the study population represents different socioeconomic groups, it may not be representative of the population. According to findings from a study by Turrell et al. (48), there is reason to expect that parents not participating in studies like this represent lower socioeconomic groups. In this study, the proportion of total parents with higher education is 54.7%, which is significantly higher than in Vest-Agder County in total (26.1%) and nationwide (29.1%). It must be taken into account that this statistics are for all adults over 16 years, while the majority (68%) of the parents in this study are aged between 40 and 50. The proportion of higher education in the same age group nationwide is 35.1% (49). The fact that the study is non-representative may affect the validity of the study. However, because of the medium size study population and diverse study population, the tendency of the results can be seen.

Although the FFQ has been validated, the validity of reported food intake is of concern. The dietary data are based on self-reports and may be over- or underestimated depending on social desirability bias and recall bias. We cannot exclude the possibility of misreporting in dietary intake due to under- and overestimating of self-reported data. A well-known phenomenon in studies like this is that obviously unhealthy foods are underestimated, while healthy foods are overestimated (23). Similarly, some parents may have overreported personal information such as income. To reduce social desirability bias among the students, we introduced the survey by telling them that it was not an exam and that there is no right or wrong answers. Students were encouraged to answer the questions as honestly as possible. Furthermore, adolescents may have had difficulties in recalling past dietary intake for the last 4 weeks (50). It must also be considered that students may have been unfocused when filling out the questionnaire (51), which may have resulted in misreporting of intake. However, this was not a major problem in the study, as most students in the study were focused and positive to answer the survey.

The parental knowledge about dietary guidelines was assessed with only one question. We did not do any performance test to assess actual knowledge, the results only reflects the parents’ perceived knowledge. This is a limitation with our study. It may also be that there is a social desirability influencing the results of this question and it should be investigated further.

These are factors that must be taken into consideration in all studies of this kind.