We analyzed data collected in 2003-2004 in the LASER study (Lifestyle in Amsterdam: a Study among Ethnic gRoups). The aim of this study was to gain insight into health-related risk factors and its determinants in young Turkish and Moroccan people living in Amsterdam, the Netherlands. For this study, a random sample was drawn from the Amsterdam population registry, which included people between the ages of 10 and 30 years who had either been born in Turkey or Morocco (first generation) or in the Netherlands with at least one parent who had been born in Turkey or Morocco (second generation). During a home visit, trained interviewers with a similar ethnic background and sex to the participant, conducted face-to-face interviews using a structured questionnaire. This questionnaire was then forward-translated and back-translated into Turkish and Moroccan-Arabic by professional translators.
For the current study, we included Turkish and Moroccan participants aged 15 to 30 years. The main reason for excluding the 10-14 years olds is that 90% of this group belongs to the second generation, which would lead to less reliable assessment of generational differences. In addition, we assume that the influence of social and cultural determinants on overweight may be different within this younger age group.
Of the Turkish sample (aged 15-30) in total 997 persons were contacted at their home address. Of these persons 33% refused to participate in the study and 14% of these persons could not be reached after three attempts of visiting their home. In total 52% of the Turkish sample participated in the study, which resulted in 519 participants. Of the Moroccan sample 601 persons were contacted at their home address. Of these persons 30% refused to participate in the study and 23% was not reached after three attempts. In total 46% of the Moroccan sample participated in the study which resulted in 277 participants. Due to many missing or invalid data (15%) on weight and height, the total numbers of participants used in our analyses were 249 Moroccan and 424 Turkish men and women.
The study sample appeared to be representative of the Turkish and Moroccan populations (aged 15-30 years) living in Amsterdam and according to sex, generational status, city district and educational level. With regard to age, among the Moroccan male sample, the 20-30 year-olds were slightly under represented compared to the general population of Moroccans in Amsterdam.
We certify that all applicable institutional and governmental regulations concerning the ethical use of human volunteers were followed during this research. The study was approved by the Medical Ethical Committee of the Academic Medical Centre in Amsterdam.
Body mass index (BMI) was calculated as weight (kg) divided by height (m2). For people of 18 years and older, overweight was defined as a BMI of 25 or higher. For people of 15 up to 17 years, we used the recommended sex and age-adjusted cut-off points . The weight and height of the participants were measured during the home visits. Weight was measured on an electronic scale to the nearest 0.1 kg after removal of shoes, jackets, heavier clothing and pocket contents. Height was measured twice, standing in an upright position without shoes, with a measuring tape to the nearest 0.1 cm. Due to some logistical problems, not all participants were measured during the home visit. In these cases, which were completely random, weight and height were self-reported (43% of the cases, n = 289). Chi-square tests showed that the self-reported and the measured group did not differ on any of the dependent and independent variables used in this study. The consequences of this measurement issue for the results of the study are discussed in the limitations section (discussion) of this paper.
Age (years) and ethnicity (Turkish, Moroccan) were treated as confounders in the relationship between generational status (first or second generation) and overweight (yes, no), and the potential determinants. Marital status (married or cohabiting, not married/not cohabiting) and having children (yes, no) were considered as potential explanatory factors in the association between overweight and generational status.
Firstly, Educational level was indicated by the highest level of education attained for those people who had finished school. For those people who were still following a course of study, the current level of education was used. Educational level was categorized as "low" when people had no education or only a primary school education and "moderate" when people had had lower to intermediate level vocational training. Participants were considered to have a "high" level of education when they had completed higher professional education or university.
To justify this categorisation for students as well as non-students, we would like to refer to the fact that, unlike other European countries (such as the UK), the educational system in the Netherlands is characterized by the process of streaming. This implies that pupils from the age of 12 years start an educational program at a certain level. This starting position is a good predictor for the level of education they will finally achieve and is also associated with their future position on the labour market .
Secondly, position on the labour market was measured by the current daily 'main activity' of the participants. We divided the participants into four categories: 1) unemployed, 2) homemakers, 3) paid employment, and 4) students still studying. Among men, there was only one homemaker, therefore this category was left out within the logistic regression analyses in men.
Thirdly, we measured occupational status in which persons were categorized according to the highest occupational status within the family. For adolescents who were still living with their parents, we used the highest occupational position of the father or mother. In cases where one or both parents were retired, we used their level of occupation prior to retirement. For young adults with their own household, we used the highest occupation of the participant and his/her partner. The following categories were distinguished, using a standard classification of occupations : 1) manual occupation (i.e. cleaning jobs), 2) non-manual occupation (i.e. administrative work), 3) unemployed, and 4) students.
Migration related factors
Region of origin was measured by asking participants if they (themselves or their family) originally came from a small village/town or from a big city in their country of origin.
In addition, participants were asked what had been the main reason for migration to the Netherlands. First generation participants answered this question for themselves and second generation participants were asked to indicate the main reason for migration of their parents. The reasons were categorized as follows: 1) came with parents, 2) family reunion/marriage or 3) economic reasons, such as education or employment.
The indicators of acculturation were based on Berry's approach whereby this position is considered in terms of orientation towards the majority culture versus culture of origin, and social contacts with the host population versus contacts with people from the culture of origin . Thus the following components were derived.
Firstly, cultural orientation was measured by 10 items about language use with family members and friends, use of media, difficulties with reading Dutch, shopping preferences and emancipation as an example of Western norms and values [26, 27]. For example, one item on the use of media was as follows: "How often do you watch Dutch television programmes?", but also, "How often do you watch Turkish television programmes?" This enabled participants to answer positive to both items which would indicate their bi-cultural orientation. The cultural orientation scale was constructed using principal component analysis and reliability analysis (alpha = .64). Secondly, social contacts were measured by three questions about contacts with native Dutch people during leisure time (i.e., How many of your best friends are ethnic Dutch?) (alpha = .84). For both scales, the scores on the items in each scale were summed up. A mean substitution was made for cases where one item was missing. In total there were 7 male participants and 6 female participants with missing items on one of both acculturation scales.
The final scales were categorized in tertiles in order to denote an individual's cultural position, with subjects in the first tertile being the least oriented towards the majority culture (having the least contacts with ethnic Dutch) and those in the third tertile being the most oriented towards the majority culture (having the most contacts with ethnic Dutch).
In addition, we measured the perceived importance of religion which was scored on a 4-point scale ranging from not important at all (1) to very important (4), which was dichotomized into very important (score 4) versus not, to moderately important (score 1-3).
First we performed all analyses separately for Turks and Moroccans to check that the pattern of associations was similar by ethnicity. It appeared that all associations between social and cultural factors and overweight were in the same direction in both groups. Also, the associations between generational status and overweight were similar in both groups and for women and men. Subsequently, we decided to combine the Turkish and Moroccan groups as ethnicity would not affect the associations.
To assess the extent to which the potential determinants differed between the generations, we calculated percentages by gender and generational status using cross tabulations with Chi-square tests. To investigate which factors could account for the generational differences in overweight, we conducted logistic regression analyses. First we assessed the generational differences in a crude model (adjusted for age and ethnicity). Secondly, we added the separate social and cultural factors to the crude model with generational status as an independent variable and overweight as a dependent variable. In Model 1 we added socio-demographic factors, in Model 2 we added socioeconomic position, in Model 3 we added migration-related factors and in Model 4 acculturation and religion were added to the crude model. All logistic regression analyses were adjusted for age and ethnicity. The results are presented as Odds Ratios (OR) with 95% Confidence Intervals (CI).