Socioeconomic and demographic characteristics
The most notable finding of our study was that the dietary intake patterns varied mostly according to social characteristics, since the bigger coefficients are for SEP. Although we had found associations between the other explanatory variables and some specific dietary intake components, the magnitude of effect for these variables were not as big as seen for SEP.
Higher SEP was associated with lower intake of snacks and treats (such as crisps, soft drinks and chocolates), but also with less fruits and vegetables. Richer children had more milk and less rice and beans (the traditional daily food in Brazil), and they also consumed more cheese and processed meats. Based on this picture, it was not possible to identify a clear health-oriented pattern among richer children, although the trend of poorer children eating more treats was already seen at earlier ages in this cohort [19].
It seems that the nutritional transition, which started a few decades ago in Brazil, is not yet finished in children from Pelotas, since we could not observe a clear picture associating high SEP with a ‘healthier’ pattern, as seen in other studies [27–29]. Studies with children from high-income countries found a clear indication of a ‘healthy’ dietary intake pattern in richer children, with, for example, higher intake of fruits and vegetables [27, 28], while children from lower SEP present high intake of ‘processed’ pattern [28]. Moreover, a recent review including studies with adults from low- and middle-income countries - among them 9 studies from Brazil - concluded that a ‘healthier’ dietary pattern was associated with higher SEP [29]. Lower intake of fruits and vegetables and snacks and treats at same time among richer children may be an indication that children from Pelotas are in the midway of a nutritional transition, going from a dietary intake pattern typical of low-income countries to one typical of high-income countries [30].
Maternal age at birth was also associated with dietary intake patterns at 6 years. Children who were born to teenage mothers consumed more snacks and treats, characterized by a high intake of candies, soft drinks and crisps, and less fruits and vegetables, suggesting unhealthy food consumption. High intake of treats was already associated with younger mothers at 48 months in this cohort [19]. In agreement with our results, two studies conducted with children from the Avon Longitudinal Study of Parents and Children (ALSPAC) found that children from younger mothers (≤20 years) consumed more biscuits, sweets and crisps at 6 and 15 months [10] and more ‘junk’ foods at 3 years [31].
We also observed that children from mothers who were more than 35 years at the time of giving birth consumed more fruits and vegetables and less cheese and processed meats. Previous studies have reported that children from older mothers presented higher intake of healthy (fruits, vegetables and fish) and traditional (meat and vegetables) components [32, 33].
Here, it is clear that higher maternal age at birth is associated with a healthier dietary intake pattern. As young mothers, especially the teenagers, are less educated and do not have full access to health information, it is important to focus on nutritional counselling to this group in health practice, in order to promote healthier diets and lifestyles among these mothers and their children.
Consistent with other studies from different settings [34–36], boys presented higher intake of milk than girls as well as more coffee and bread, more rice and beans, and more daily energy intake. In contrast, girls presented higher consumption of fruits and vegetables than boys. These differences seen in adherence to dietary components may be explained by differences in food preferences according to sex [37]. Furthermore, unpublished results of this cohort showed that boys have higher physical activity than girls at 6 years (measured by accelerometer), which could increase dietary needs in boys, increasing their daily energy intake as well as their adherence to the majority of dietary components. Nevertheless, this is speculative since we did not run any analysis to assess the effect of physical activity on dietary intake in this cohort.
Skin colour was also seen as an important factor associated with dietary intake patterns at 6 years. We observed that non-white children consumed more coffee and bread and rice and beans, and less milk, irrespective of SEP, sex and daily energy intake. Skin colour is an important marker of wealth inequality in Brazil, and the non-white population (mainly the blacks) always present worse results for socioeconomic indicators. This scenario influences food availability at the household level and the number of meals per day, which could explain the disparities in children’s dietary intake patterns according to skin colour in our sample. Recent studies showed that the availability of foods at the household level has an impact on children’s food consumption [38, 39]. Unfortunately, we were not able to measure household food availability and the number of meals per day in this cohort. Further research would be useful to help us to understand the mechanisms underlying the differences in dietary intake by skin colour in Brazil.
Early feeding practices
Breastfeeding duration has been positively associated with healthy dietary patterns in children. A study conducted with 6 year-old children in the US found that breastfeeding duration was associated with higher intake of fruits and lower intake of sugar-sweetened beverages [40]. Similarly, another study conducted in Australia found that breastfeeding duration was associated with a healthy dietary pattern in 2–8 year-old children [41]. We found similar results, since children who were exclusively breastfed for less than 3 months and who started complementary feeding before 4 months of age consumed more snacks and treats (and more coffee and bread) and less fruits and vegetables, suggesting that early weaning along with early introduction of complementary feeding is related to ‘unhealthier’ feeding habits at 6 years. Some studies have raised an hypothesis where breastfeeding can improve children’s acceptability to new foods [42, 43], increasing the variety in children’s dietary intake.
BMI z-score at 6 years
We expected that obese children would have higher consumption in the majority of dietary components, mainly in those indicative of unhealthier diet (such as snacks and treats), and lower consumption of fruits and vegetables. Nonetheless, we observed that overweight and obese children presented lower intake of four out of seven dietary components, including snacks and treats. As children’s consumption was collected based on mothers’ report, a possible explanation for this result is the fact that mothers of obese children, aware about their obesity status, could be underestimating their food consumption. Moreover, both food consumption and BMI status were collected at the same age and this association may be affected by reverse causality.
Extra analyses treating BMI as outcome showed that high intake of coffee and bread was negatively associated with BMI z-score at 6 years, while moderate and high consumption of cheese and processed meats was positively associated with BMI z-score at 6 years. This finding is interesting to note, as recent investigations in Brazil have shown that ultra-processed foods may play a key role in the obesity epidemic, providing around 30 % of daily energy intake in adolescents and adults [44, 45]. In addition, a recent study conducted in the 1982 Pelotas cohort study showed that high intake of processed foods was positively correlated with intake of sodium, cholesterol, and fats [46]. Therefore, our results could indicate that high intake of ultra-processed foods, characterized by ready-to-eat foods with minimal or no preparation [47], may also be important in the development of childhood obesity in this cohort. But again we must be cautious with interpretation due to reverse causality. More studies are needed in order to assess the longitudinal effect of food intake patterns on BMI and adiposity status.
Strengths and limitations
Some strengths of this study are the cohort’s size and longevity, the low losses and refusal rates (only 9.8 %), and regular data collection that helped to minimize biases. Moreover, the FFQ used in our study was validated based on three 24-h dietary recalls. Food frequency questionnaires are very useful to investigate food consumption in large population-based studies as this method allows collection of complex dietary intake information in a simple, cheap and time-effective way [48].
Regarding PCA, the seven independent components identified in our study explained almost 50 % of the variation in children’s food consumption. The variance preserved in other studies which used PCA to analyse dietary intake patterns ranged from 26.8 up to 48.4 % [9, 11, 13, 19], which puts our results along with others with the highest percentage of variation explained. In addition, the seven components that we selected are visibly distinct by the type of food group that loaded on it (Table 2). PCA has been successfully used to describe the feeding practices of children in this cohort (at earlier ages) [19] and in other studies [9, 10, 15].
On the other hand, the 12 months recall period for the FFQ can be considered a limitation of our study, since it is a very long time frame and can result in recall bias. In addition, the FFQ was administered to the child’s mother, representing an indirect measure of the child’s food consumption, which may result in measurement error.