Parents provide both genes and the home environment for their children, and as such, are central in shaping children’s early experiences with food and eating behavior. Parents and other caregivers have considerable control over the foods that their children eat through the types of foods made available in and out of the home, food preparation methods, food-related parenting style, frequency of family meals, and deciding where the family goes out to eat [1
]. For example, the availability of fruits, vegetables, and dairy foods in the household is an important predictor of children’s intake of these foods [2
Children have a genetic predisposition to prefer foods that are sweet and salty and reject those that are sour and bitter, such as found in some vegetables [4
]. They can, however, learn to accept these tastes through repeated exposure [5
]. During family meals, parents or other familiar adults can serve as important role models for children’s willingness to try novel foods [6
]. Eating meals together has been associated with healthful dietary patterns that track into adulthood, including increased intakes of fruits and vegetables, calcium-rich foods, fiber and micronutrients, and reduced intakes of fried food and soft drinks [7
As children enter adolescence, they become more independent of their parents and have greater autonomy over their food intake. Peers exert an increasing influence on their snack and soft drink consumption, particularly when the availability of these foods at schools is high [9
]. As children progress from childhood to adolescence, they are less likely to participate in family dinners at home [7
], more likely to skip breakfast [10
], and their diet quality and diet variety declines [10
Previous research suggests that the parent–offspring resemblance in dietary intake is weak and that factors other than parental diet and home environment play an important role in influencing dietary intake in children and young adults [12
]. As Wang et al. [13
] previously summarized in a review of the literature and meta-analysis, most previous studies are based on small sample sizes and the resemblance varies by nutrients, foods and parent-child pairs. In a large study of 2692 child-parent pairs from the U.S., there were no consistent differences by age, i.e., the parent-child resemblance for the healthy eating index was stronger for children below the age of 10 year-old than older children, while the parent-child resemblance for energy, most nutrients and soft drinks was stronger for children older than 10 year-old than younger children [14
In order to develop effective family-based interventions and target them appropriately, it is important to know whether familial or non-familial factors are more important in determining children’s food intake and whether the family environment exerts a stronger effect in younger than older children. For example, if familial effects weaken as children get older it may be advisable to introduce individual and peer related interventions instead of familial interventions to achieve dietary behavior change. In the I.Family study, the inclusion of a large number of families and children of a wide age range allowed us to calculate familial correlations and familiality estimates for usual intakes of energy, 4 macronutrients and 13 food groups and determine whether these correlations vary by generation, sex, types of foods or between younger and older children.
The results of this large multi-center European study confirm a substantial familial resemblance in usual intake of energy, macronutrients, and food groups. These may arise from shared genetic factors between biologically-related relatives, shared family environmental factors between family members living in the same household, and assortative mating or social homogamy between spouses. Our parent–child correlations are consistent with the magnitude of correlations reported by Wang et al. in their review and meta-analysis on parent-child resemblance in dietary intake [13
]. The current study extends this knowledge, by showing firstly, that familial correlations—particularly parent–offspring correlations—are stronger for healthy food intake than unhealthy food intake; and secondly, that familial resemblance, particularly parent–offspring correlations is stronger for younger children aged 2–10 than for older children aged 11–19.
In the present study, familial factors explained 61% of the variance in the intake of healthy foods and only half as much (30%) in the intake of unhealthy foods. A possible explanation could be that children have a strong biological preference for sweet and salty foods and this preference decreases with increasing age and as they enter adulthood [31
]. Thus, the age difference between parents and children could contribute to the weak parent–child resemblance in unhealthy food intake in the present study. Another explanation could be that in the European societies in which the present study was conducted, there are more external influences on unhealthy food intake than healthy food intake. For example, television viewing has been shown to influence obesity risk in children and to contribute to overconsumption of higher-fat and higher-sugar diets [33
]. Since advertising is often targeted specifically at children, children can be assumed to be affected by advertising from an early age onwards. Most foods marketed to children on television are for products high in sugar, fat, and sodium such as fast food, sweets, snacks, and unhealthy cereals [34
]. Food advertising could lead to the child requesting the food through what has been termed the “nag factor” or “pester power” and could consequently influence what children eat [35
]. This could further contribute to parent–child dissimilarity in the intake of foods high in sugar and fat. Friends and peers could also contribute to the lower resemblance between parents and their children in unhealthy food intake. Among students, it has been shown that attitudes towards fruit were predicted by those of their parents rather than friends, while preference for sweet was predicted by friends’ rather than parents’ attitudes, suggesting parental influence as regards healthy foods and peer influence as regards junk food [36
]. This could be mediated by social norms, as it has been reported that adolescents tend to associate healthy food with family and junk food with friends, and these associations predict food consumption frequency [37
We further found that parents resemble their younger children’s intake of healthy foods more than their older children’s intake of the same foods, and familial factors are a more important factor in determining variation in healthy food intake among younger than older children. In contrast, for the intake of unhealthy foods, no differences were seen between younger and older children. As younger children consume most of their main meals at home, the home environment is likely to be the main determinant of the intake of healthy foods, including fruit and vegetables. If such foods are not made available to children at home, children are unlikely to consume them outside the home. Family meals have been found to be associated with more healthful eating patterns, and the frequency of family meals declines as children become older and more independent [7
]. Longitudinal research has shown that as children move from elementary to junior high and middle school, their consumption of breakfast, fruits, vegetables, and milk decreases and the popularity of soft drink consumption increases [10
]. One previous study reported the parent–child correlations for the healthy eating index to be stronger for children below the age of 10 than for children above the age of 10, whereas for fiber, calcium and some other nutrients, the resemblance was stronger for older than younger children [14
]. Direct comparison with the current study is difficult as only very few food groups were examined in that study.
Overall, we found few sex differences. For a few nutrients and food groups, including unhealthy milk and dairy products and unhealthy meat and meat products, father–offspring dyads were not correlated. Furthermore, for unhealthy meat and meat products and healthy meat alternatives, heterogeneity was found for the different subtypes, and mother–child correlations tended to be stronger than father–child correlations. Some previous studies have reported stronger correlations between mothers and daughters in dietary intake than between mothers and sons [38
], whereas others did not detect any sex differences [40
]. In a population-based sample of 471 mother–child pairs from Stockholm, there were significant associations between mothers and sons in the intake of many food groups. Overall, however, mothers and daughters were more strongly correlated in their dietary intakes than mothers and sons, and macronutrient and energy intakes were significantly correlated between mothers and daughters, but not between mothers and sons [39
]. Mothers usually devote more time to child care than fathers, whether they are employed or not [42
], and women are more likely to take primary responsibility in meal planning, shopping, and preparation than men [43
]. In addition, long working hours and schedules among fathers were found to be positively associated with take-out meals, missed family meals, prepared entrees, and eating while working [45
]. Thus the greater resemblance of mother–child than father–child pairs in some aspects of dietary intake could be explained by the additional environmental factors that mothers share with their children as compared to fathers, such as the time that mothers devote to food preparation and feeding their children. Furthermore, stronger maternal associations could also suggest maternal in utero programming of offspring appetite by maternal intake during pregnancy [46
]. It is unlikely that non-paternity, which is estimated to be only about 1–3% for Western populations [47
] has significantly contributed to the few correlations that are lower between children and their fathers versus their mothers.
When we compare our sibling correlations to those reported in other studies, we find larger correlations than previously reported. For example, in a prospective cohort of postmenopausal women from the Iowa Women’s Health Study, sibling correlations for nutrient intakes ranged from 0.04 to 0.17 [49
], while sibling correlations for dietary traits in the present analysis ranged from 0.21 to 0.43. It is likely that this difference is due to the fact that siblings in the I.Family study still live in the same home, whereas the adult siblings in other studies generally no longer share these home environmental factors. We found that correlations were about the same for sister–sister, sister–brother, and brother–brother pairs, and little evidence that siblings of the same sex would share a more common social environment than siblings of the opposite sex.
Mothers and fathers were modestly correlated for all dietary factors examined, and for many dietary factors, these were of the same magnitude as parent–offspring correlations, suggesting a shared household effect. Alternatively, assortative mating can contribute to spouses’ resemblance, and this has been well documented for height [50
], obesity [50
], and lifestyle behaviors [52
]. Interestingly, the strength of the spousal correlation did not differ between healthy and unhealthy food groups, suggesting that home environment influences the intake of healthy and unhealthy food groups to an equal extent in adults.
The present analysis has important strengths, including the recently collected large sample of families with information on biological relatedness and in-depth phenotyping of children and their parents which allowed a comprehensive assessment of habitual dietary intake. In order to increase the accuracy of the dietary assessment, we used a computer-assisted 24HDR that included standardized photographs, multiple plausibility checks, and reminding questions that facilitated reporting of accurate portion sizes and complete recalls. We corrected for reporting bias by excluding incomplete recalls and recalls with implausible energy intakes.
However, our study also has limitations that should be considered when interpreting our findings. We acknowledge that the estimate of familiality in this study is not directly comparable to the heritability estimates obtained from previous twin and pedigree studies. Unfortunately, we cannot distinguish between genetic and shared environmental effects. Many twin studies also report an estimate for shared environment effects together with heritability this can be cautiously compared to familiality estimates. Our study of nuclear families does not allow us to conclude whether the familiality estimate is due to genetic or shared environmental factors, and whether the lower familiality estimate in older children is due to stronger genetic influences in younger children or diminished shared environmental influences in older children. Previous twin studies suggest that genetic influences are largest and shared environmental influences lowest for healthy foods including fruits, vegetables, and protein [54
]. This has been shown to vary by age since in young children shared environmental factors have a substantially stronger influence on food preferences than genetic factors, while in adolescence the shared environmental influence appears to disappear completely [55
]. Although our overall sample size was large, it was lower when comparing the correlations between the different dyads by sex or age.
Children may have consumed some foods outside the home and this might have decreased the accuracy of 24HDR in younger children when proxy reports by parents were used. In addition, younger children more often required assistance in completing one or more 24HDR and this may have contributed to stronger similarity between parents and their younger children. However, less than 5% of participating children were below seven years of age. Previous research has shown that by the age of 8–10 years children can reliably report their food intake [56
] and that children as young as 6 years can accurately recall their school lunch intake for one occasion while teachers record with less accuracy [58