1. Introduction
Childhood obesity is a major public health issue and its prevalence is increasing both in the UK and internationally. The World Obesity Federation Atlas published in 2019 forecast that childhood obesity will rise markedly by 2030 [
1]. To develop strategies for the prevention of obesity and thereby avoid long-term health effects [
2,
3,
4,
5,
6], it is important to identify modifiable predisposing factors.
Obesity research has focused on nutrition in infancy and early childhood and associations between early life influences and adult cardiovascular and metabolic diseases have become apparent [
7]. Studies investigated the influence of early nutrition on later obesity risk but there is no agreement about the roles of specific macronutrients [
8]. In early childhood a high protein intake has been linked to later body composition as it enhances secretion of insulin-like growth factor-1, which may increase growth and adiposity [
9]. Dietary fat intake is reduced during complementary feeding with the partial replacement of milk by carbohydrate-rich foods [
10,
11]. During this transition period many important hormonal and enzymatic changes occur that affect carbohydrate and lipid metabolism [
12]. A high carbohydrate, low fat diet at this early age has been shown to increase insulin and decrease plasma glucagon concentrations in infants of different species, including humans [
12]. Animal studies have shown that a high carbohydrate, low fat diet during both the suckling [
13] and weaning period [
14] programs susceptibility to obesity later in life. The programming potential of the fat content of early diet on the development of adiposity in humans is not known.
A few longitudinal studies have investigated macronutrient intake and body composition using a range of outcome measures including BMI, skinfold thicknesses, and fat and lean mass, but with differing results. [
8,
15]. The ELANCE study followed a small group of children longitudinally for 20 years with outcomes at 8 years (
n = 112) [
16] and 20 years of age (
n = 73) [
17]. At 8 years a positive association between protein intake at age 2 years and body fatness was found but no relationship with fat intake [
16]. At 20 years the study suggested that low fat intake in early childhood increased the susceptibility to the development of obesity and leptin resistance in adulthood, and that early programming of leptin resistance might be one way in which nutrition in infancy could affect adiposity development [
17].
The ELANCE participants had low intakes of dietary fat for this age: the mean fat %E at 10 months was 27.7% (SD 4.8%) for boys and 28.2% (SD 4.4%) in girls; in comparison at 8 months of age the ALSPAC participants had a mean fat %E intake of 35.2% (SD 5.0%) for boys and 35.6% (SD 5.0%) for girls. WHO recommends 30–40% fat %E during weaning until 3 years [
18]. There is a strong case for repeating these analyses in another setting where fat intakes are more varied, with a larger cohort size to increase the investigative power and additional confounders that have associations with childhood obesity [
19,
20,
21,
22]. Furthermore it is relevant to analyse boys and girls separately since the
Atlas of Childhood Obesity reported a large difference in the prevalence of obesity by sex, with a greater prevalence among boys [
1]. These sex differences may be due to biological differences in body composition or hormonal differences [
23].
The Avon Longitudinal Study of Parents and Children (ALSPAC) collected diet data at similar timepoints to the ELANCE cohort (8 and 18 months of age) and includes body composition data at similar ages (9 and 17 years) and serum leptin (9 years). The aims were therefore: (1) to investigate the generalisability of the findings in ELANCE by replicating the investigation in ALSPAC and (2) to extend the investigation with additional analyses to take advantage of the larger sample size and range of confounders available.
4. Discussion
Our results suggest that a diet high in fat and low in carbohydrate in early childhood is associated with increased adiposity later in life in boys in this cohort. There was strong evidence among boys of a positive association between fat %E at 18 months and FM at age 9 years (0.2 (95% CI 0.0, 0.3) kg,
p = 0.008) and the association was present but attenuated at age 17 years (0.3 (95% CI 0.0, 0.6) kg,
p = 0.045). This equated to a difference of 2.3 kg (at 9 years,
p = 0.003) and 5.1 kg (at 17 years,
p = 0.009) in FM between boys with fat intakes <35% compared with intakes >40% of energy (see
Table 6), a substantial effect. This was strongly associated with the MUFA %Fat. There was also evidence among boys, particularly in plausible reporters, of a positive association between energy intake at 18 months and BMI at 9 years with the increase being in FFM rather than FM, evident at both 9 (0.3 (95% CI 0.2, 0.5) kg
p ≤ 0.001) and 17 (0.7 (95% CI 0.2, 1.3) kg,
p = 0.015) years. In girls there were few associations which were inconsistent.
In ELANCE at 8 years of age they found a positive relationship between energy intake in early childhood and BMI [
16]. In our replication we found similar associations with BMI at 9 years, although the association was with FFM not FM so is a partial replication. In contrast to ELANCE, our study did not support the hypothesis that early protein intake is associated with obesity in mid childhood, although protein intakes were lower in ALSPAC. In ALSPAC there was evidence of a positive association between fat intake at 18 months and fat mass at 9 years, this was not evident in the ELANCE study.
In the second ELANCE study at 20 years of age [
17] the authors found no relationship with protein intake, but there were negative relationships between total fat %E at 2 years and FM and subscapular skinfold thickness, which was related to SFA, leading to the conclusion that low fat intakes in the first 2 years of life were associated with increased body fat in later years [
16]. Our study found fewer associations at 17 years but did confirm a positive association between energy intake and FFM at least in boys. In contrast to ELANCE we found the opposite association between total fat, SFA and MUFA %E and FM: as fat intake increased so did FM. Our findings do not support the conclusions of the ELANCE study regarding fat intake.
There are several differences between ELANCE and ALSPAC that could account for the conflicting results. ELANCE had a much smaller study sample: 112 at age 8 years and 73 at age 20 years compared with >700 at 8 years and >400 at 17 years in ALSPAC. Fat intakes in ELANCE were particularly low: at 10 months mean fat %E was 28% and at 2 years 32%. These were below the mean intakes reported in ALSPAC of 35% at 8 months and 37% at 18 months and below international reference intakes of 40–60% during infancy gradually reducing to 30–40% by 24–36 months [
18,
31]. In ELANCE instead of fat %E intake reducing as the participants got older, intake increased to 38% at 8 years of age. It was proposed that this change from an early low to a later high fat intake may promote the development of adiposity [
32]. In ALSPAC fat intake stayed close to the reference intakes in early years [
26,
27] and remained stable throughout childhood [
33,
34]. The low fat intakes in ELANCE could be explained by their milk intake: at 10 months no children were breastfed, 44% consumed formula milk and the remainder were having cows’ milk mostly with reduced fat levels.
In our study there was a relative increase in fat intake between 8 months and 18 months and this was all in saturated fat fraction and reflected the move from infant foods to family foods. The types of foods which were eaten more frequently by children at 18 months than at 8 months were those containing fat and sugar; for example, the proportion of children eating biscuits rose from 35% to 87% and those eating cakes/puddings rose from 38% to 58% [
26,
27]. This may account for the fact that we only found associations with the diet at 18 months. Our results support recommendations for feeding young children that suggest limiting these fat and sugar containing foods [
35].
Other studies have examined early diet in relation to body composition pre- and post-adolescence. These studies have used different measures of body fatness both direct and indirect but have reported similar results to our study. Fat intakes recorded between 2 and 8 years were positive predictors of BMI at 8 years [
36] with an inverse association for carbohydrate, while in 9–11-year-olds [
37] percentage of body fat correlated positively with fat intakes. Across the age range 2–15 years energy—adjusted fat intakes were positively, and carbohydrate intakes inversely related to greater skinfold thicknesses [
10]. In our study adiposity was measured directly by DXA, so we were able to explore the positive relationship between fat intake and BMI and show that it is driving a greater change in FM than FFM in boys as reported previously [
38]. Our findings also showed that higher energy intake in boys at 18 months was associated with a greater increase in BMI specifically FFM at 9 years but in girls increased energy intake resulted in higher FM. Other studies have corroborated that as BMI increases throughout childhood boys lay down more FFM whereas girls accumulate more FM [
39,
40].
We were able to study boys and girls separately which is important as there are biological differences in body composition between the sexes. We found many associations between diet and body composition in boys but less in girls. Few studies have looked at boys and girls separately, but two found an association between dietary fat and FM in boys but not girls similar to our study [
41,
42]. There were also differences observed for serum leptin levels between boys and girls. As fat intake increased for boys there was a concurrent increase in FM and leptin. This is unremarkable as leptin is a hormone that is produced by fat cells and is positively correlated with the amount of body fat present [
39,
43]. The variations between the sexes may be due to biological differences: girls naturally have higher FM and circulating serum leptin concentrations than boys [
44], so it is possible that this masks any association between dietary fat %E and body composition and leptin. In addition, the variation in fat %E in this study may not be large enough to show an effect in girls.
In our study we found a strong positive association between the fatty acid types MUFA %E and SFA %E and FM corroborated in another study in which percentage of body fat in children aged 9–11 years correlated positively with intakes of all types of fat and negatively with carbohydrate [
37]. This complementary relationship between carbohydrate and fat (as fat %E increases it displaces carbohydrate %E) was also present in our study. We did not find evidence that a diet low in fat and high in carbohydrate in early childhood programmes adiposity in later life, and our findings in boys were the reverse of this. However, a study that investigated children’s food and nutrient intake including MUFA in pre-school years and body composition in mid childhood found that higher longitudinal intakes of dairy products and MUFA were associated with lower body fat determined by DXA [
45] in contrast to our findings.
In boys but not girls, intakes of fat <35%E in early childhood were associated with lower BMI and FM in later childhood and young adulthood compared with intakes of fat >40%E, thus validating, at least in boys, the recommendations for a fat intake not to exceed 40%E. Previously in ALSPAC we have shown that at 18 months full fat dairy products are a primary source of fat and fatty acids in the diet [
46]. Children with higher fat diets consumed more whole milk, and cheese. Some children with the highest fat intakes were consuming in excess of 750mL milk per day. To reduce fat intake children’s whole milk intake could be limited to 500mL or changed to semi-skimmed milk from the second year of life provided they were eating a well-balanced diet [
46].
Our study has many strengths including: (1) the long follow up period from early childhood to early adulthood; (2) a large sample size; (3) repeated measures of diet and the ability to identify plausible reporters; (4) anthropometric measures using DXA rather than skinfold thicknesses or bioimpedance; (5) a wide range of confounding variables available for analyses, including maternal smoking in pregnancy which has been associated with childhood obesity [
19,
20], maternal education which has been identified as a good proxy for social class and is associated with childhood obesity in ALSPAC [
21], and paternal overweight\obesity as well as maternal overweight\obesity which are associated with obesity in offspring [
22]. As we did not have complete data for fathers available, to retain full power in our main model we ran a second model to include the fathers’ data.
There are also limitations. (1) The associations could be different among participants and non-participants, and the study was undertaken in one geographical area of the UK, limiting its generalisability. (2) Participants were lost to follow up, reducing the amount of data available at the later time points; multiple imputation showed that this was unlikely to have biased our findings. (3) It is difficult to assess dietary intakes accurately, but this was addressed by measuring the level of misreporting. Restricting to plausible reporters attenuated some results but strengthened others: the main findings for the association between fat %E and FM and energy and FFM in boys remained. Reporting over a relatively short period of 3 days may also not be fully representative of habitual diet in children beyond the preschool years [
47]. (4) Attrition in a longitudinal cohort is inevitable and is likely to result in the cohort over-representing participants of greater socioeconomic status. This can potentially cause bias and limit generalisability [
48]. (5) There may be other confounders that we were unable to account for in the extended analysis. We did not include adjustment for adult nutrition in the replication of the ELANCE study, to mirror their adjustment. We were unable to include adjustment for adult nutrition in the extended analyses as these data were not available in ALSPAC. The samples for leptin were obtained during the day, but we do not have any data on the exact time of day for each sample, so that we were unable to account for diurnal variation. (6) The results of the goodness of fit tests showed greater adjusted R
2 values in some cases than we would expect for similar studies. Prediction intervals calculated from SEE were generally acceptable (>90%) especially at 7 years old. Lower values in some instances may reflect relatively low case numbers in each category and highlights the importance of replication of this study in other cohorts.
In conclusion our results did not corroborate the findings from the ELANCE study. We found that as fat intake in early childhood increased so did FM in late childhood and early adulthood. Energy intake at 18 months was positively associated with FFM in later life. However, these findings were only apparent in boys. Dietary composition in early life may have implications for later childhood and adolescent obesity: this needs further investigation in other cohorts with longitudinal data.