The main finding of this study was that no single "healthy lifestyle" orientation could be identified among the women who reported frequent use of organic food. The frequent use of organic food was associated with lower and higher age groups, lower BMI, a vegetarian diet, cigarette smoking and use of alcohol during pregnancy, regular exercise, lower and higher levels of education, the participant and/or her partner being a student, low household income, urban living area, and participation in MoBa between 2005 and 2007. The associations between socio-demographic and lifestyle variables and eating organic food reflect complexity and indicate that no quick label like "young and idealistic" or "well educated and wealthy" can be applied to describe women who report frequent intake of organic food during pregnancy. Some previous studies also reported that use of organic food is quite widely distributed across socio-economic groups and associated with various types of motivation [10, 29], while others, particularly within the marketing tradition, have identified consumer segments in the market with a high likelihood to buy organic food such as "the engaged", "the eco-healthy", or "the practical green" etc. .
The strength of this study is the large sample of pregnant women with participants from both urban and rural regions, representing all age groups and all socioeconomic groups. The participation rate in MoBa is 43% and the prevalence of organic consumption may not be representative for all pregnant women in Norway . However, this is not likely to influence the associations between reported use of organic food and characteristics of the respondents. The potential bias due to self-selection in MoBa was recently evaluated by Nilsen et al., 2009. No statistically relative differences in association measures were found between participants and the total population regarding eight exposure-outcome associations evaluated .
In the present study we examined the associations between participant characteristics and frequent organic consumption with and without women having missing data on participant characteristics. Covariates with missing data were BMI (2.6%), exercise (8.4%), education (2.2%), income (8.4%), and year of participation (2.0%). The numbers of missing values were higher in year 2002, in the youngest age group, among women with low education, among smokers, and among frequent organic food consumers. Including missing as separate categories in a regression is a simple method of dealing with missing values compared to the more correct but also more complex method of multiple imputations. Including missing in a regression will, contrary to popular belief, increase bias from confounding, but will reduce bias from possible heterogeneity of effects (interaction) between responders and non-responders/missing. It will also increase sample size and thereby power. The results from models with and without missing included were similar in our data, indicating that neither confounding nor heterogeneous effects played a strong role here.
The sum index provides a robust indicator of the consumption of a variety of the main organic food groups in the diet, appropriate for the explorative aim in this study. The sum index attributes equal importance to each food category and thus some detail may have been lost. The categories are dissimilar with regard to number of items within the category (e.g. 'eggs' containing only one item while 'fruit' and 'vegetables' contain numerous items). Eating 'mostly organic' vegetables is a more extensive practice than eating 'mostly organic' eggs. A related, but different challenge is that we do not know a respondents total variety of consumption within each category, e.g. whether 'mostly organic fruit' refers to only apples - or a whole range of different fruits. However, the high correlation between organic consumption within the six food groups supports the viability of using a sum index in this study (Table 2).
Eating a vegetarian diet was the characteristic which was most strongly associated with frequent consumption of organic food during pregnancy. Even though the total number of vegetarians among the MoBa-participants was low (0.2%), this was the single most predictive factor among all variables in the analysis, with 23.1 percentage points higher prevalence of frequent organic use among vegetarians than among non-vegetarians (Table 3). This reinforces earlier findings of an association between eating organic food and a vegetarian diet [32, 33]. Frequent use of organic food and eating a vegetarian diet may well be part of a healthy lifestyle, as a vegetarian diet has been associated with many health benefits such as lower risk of heart disease and type 2 diabetes [34–37]. Further associations between eating a vegetarian diet and the general dietary quality among the women with a frequent consumption of organic food will be published in a separate paper. Well-planned vegetarian diets are considered appropriate for individuals in all life-phases, including pregnancy . Lower levels of BMI have been reported among vegetarians in various populations [35, 37], and also among consumers of organic food in a study of soy consumers and non-soy consumers in Minnesota, USA .
Participation in regular exercise and being underweight or normal weight were also associated with being a frequent organic consumer. Being physically active is an important contributor to a healthy lifestyle in the general population as well as among pregnant women, and is strongly inversely related to excessive body mass [39–41].
Entering pregnancy with a normal weight is beneficial with regard to pregnancy complications and health outcomes for both the mother and the child. Maternal obesity is a risk factor for all major pregnancy complications, which have increased in prevalence in later years , including gestational diabetes, pre-eclampsia, foetal overgrowth, preterm births, and cesarean delivery [42–45]. Being underweight, on the other hand, is also associated with unfavorable birth outcomes such as preterm birth and low birth weight, while overall, the outcome is favorable and several adverse outcomes are less common in this group of women . We plan to further investigate dietary habits among organic food consumers and to examine potential diet-health relationships related to organic food consumption during pregnancy. It would also be interesting to further investigate subgroups within the population, as the characteristics associated with frequent organic consumption in the present study have also been related to a higher prevalence of eating disorders, such as adhering to a vegetarian diet , exercising more than three times per week, older age and being a student .
It is well established that both cigarette smoking and alcohol consumption during pregnancy is associated with increased risk of adverse health outcomes for the fetus, and consequently health authorities in many countries, including Norway recommend that pregnant women and those trying for a baby should totally avoid alcohol and smoking . In this study we found a higher prevalence of smokers among women with frequent consumption of organic food. If the use of organic food is motivated by perceived health benefits, this finding may appear surprising. However, sociological studies of health behavior have indicated that the associations between them are complex. In a Finish study, an attempt to construct 'health indices' based on all relevant factors associated with good health proved difficult, and even though there were clear associations between health behaviors, their distribution into different combinations were quite diverse [50, 51]. Smoking has been shown to be central in the interplay between health behaviors, and the majority of smokers had either only smoked or had one additional unhealthy habit .
Our finding of a higher prevalence of cigarette smoking among women with a frequent consumption of organic food is contrary to a European multi-country study that reported less maternal smoking during pregnancy and current smoking in families with anthroposophic lifestyle (having children at Waldorf schools, eating organic/bio-dynamic food and/or living at farms practicing Bio-dynamic farming) compared to reference families [52, 53], while a Swedish study reported equal prevalence of parental smoking in families with anthroposophic lifestyle vs. reference families . In the present study, however, consumers of all types of organic food are included, and we have not looked at respondents adhering to an anthroposophic lifestyle in particular.
We found a two-sided trend between frequent use of organic food both in regard to age and education, with the highest and lowest age and education groups being more likely to be frequent organic consumers than the middle groups. Next to eating a vegetarian diet, being in the lowest age group (<20 years) represented the highest likelihood for being a frequent organic consumer, with 15.6 percentage points higher prevalence than in the reference group (30-34 years). The present study has a smaller age range than studies in the general population as only pregnant women were included. Studies in the general population have also reported diverging results related to age, some finding that younger age (see for example [55–58]) and others that older age  was associated with a higher propensity to buy organic food, while some reported no difference with regard to age . It has been suggested that there may be a pattern whereby there are higher shares of younger consumers among early adopters in developing markets, while older consumers are in higher numbers in more mature markets . In Denmark and Great Britain, both being mature markets for organic food, the highest likelihood of buying organic food was found in households with middle-aged. In Great Britain the likelihood was lower in both the younger and older groups, while in Denmark it generally increased with age, but with a peak for the age group 40-49 years . The Norwegian market for organic food is not by far as mature as the British or Danish, even though it has developed quite rapidly during the last decade. It might be that we see a combination of young new-comers and women who have longer experience with eating organic food.
Household economy could be expected to be crucial for the level of organic consumption given the fact that these products generally are more expensive. However, our finding of an inverse relation between household income and frequency of organic consumption indicate a complexity beyond economic ability alone, and that other factors are more decisive for the likelihood of eating organic food. Higher levels of education, on the other hand, were associated with a higher likelihood of eating organic food compared with middle-levels of education, and so was being a student - even in the older age groups. Both education and income are strongly associated with better health in population studies. Our finding of lower household income may therefore be interpreted as an indicator of vulnerability with regard to health, while higher levels of education - particularly with increasing age, provides for robustness. The combination of higher education and lower income among the frequent consumers of organic food may point to a different value-orientation in this group.
Previous Norwegian surveys have reported higher levels of education among consumers with more frequent use of organic food and food produced without use of pesticides [29, 59, 62], while a regional survey did not find any association between length of education and likelihood of buying organic food . Studies from other countries have also reported differing results regarding respondents' level of education and use of organic food [30, 57, 64, 65]. Higher income was associated with a higher likelihood of buying organic food in one Norwegian study , while no difference with regard to income was reported in another . A study of consumption of organic food in Denmark and Great Britain reported an increased propensity to buy organic food with higher 'social group', a composite indicator of educational level and income. However, in both countries, the highest propensity to buy organic food was found for the middle class households, while it was actually lower for the upper middle class . Another Danish study also found that income explained very little of the purchasing behavior related to organic food .
Taken together, the lifestyle- and socio-economic characteristics of the frequent organic food consumers in the present study point to a complex phenomenon, involving diverse groups of women which go beyond narrowly defined 'consumer segments'. Some of the characteristics, such as regular exercise, lower BMI, and - in part - higher education indicate robustness with regard to a healthy lifestyle, while other characteristics, such as cigarette smoking and use of alcohol, and - in part - lower levels of education, may indicate vulnerability with regard to health of mother and child.
Our finding of a higher prevalence of frequent organic users in urban areas, is contrary to previous Norwegian studies that reported no difference between urban and rural areas [29, 63], but in line with results from Denmark and Great Britain [60, 66]. Further, our more nuanced finding that the highest level of organic consumption in the urban areas was followed by the most rural areas, with the lowest consumption in the two middle categories, is supported by other findings , and may be related to the possible 'idealistic' nature of organic consumption. Codron et al (2006) describe that a typical development for radical movements, such as those associated with organic food, is that when these products first enter the market, buyers are typically either local rural consumers, or urban consumers with higher income levels, including members of consumer associations and politically active movements sharing these values. In later phases, the constellations may change. These suggested lines of development may be relevant for Norway, where the market situation may still be described as not having reached maturity with regard to organic food - and we do find the highest levels of consumption among the contrasting groups: either urban or rural.