Modifiable early life exposures associated with adiposity and obesity in 3‐year old children born to mothers with obesity

Children born to mothers with obesity are at increased risk of obesity. Influences underlying this predisposition include in‐utero exposures, genetic predisposition and a shared family environment. Effective intervention strategies are needed to prevent obesity in these high‐risk children; this requires evaluation of modifiable pregnancy and early‐life risk factors.


| INTRODUCTION
In parallel with the global obesity epidemic, childhood obesity is increasing worldwide. 1 Between 2000 and 2013 the number of children with overweight and obesity rose from 32 to 42 million, 2 with global prevalence expected to reach 70 million by 2025. 3 The immediate effects of childhood obesity include health complications, such as behavioural disorders, fatty liver disease and asthma. 4,5 Childhood or early-life obesity is known to track across the lifecourse 6 increasing the risk of cardiovascular disease 7 and type 2 diabetes 8 in the longer term. Prevention of childhood obesity is a worldwide public health priority. 3 In the UK a quarter of children enter primary school with overweight or obesity, 9 the highest prevalence being amongst ethnic minorities and those living in disadvantaged areas. 10 Environmental and lifestyle factors are widely implicated in the rising prevalence of childhood obesity, including interactions with hereditary predisposition. A substantial body of evidence suggests a key role for the environment in the earliest stages of life. Experimental animal studies, 11 observational cohorts 12 and some randomized controlled trials 13 suggest that adverse in-utero exposures, including maternal obesity 14 or excessive gestational weight gain (GWG), 15 may contribute to offspring obesity, which can persist into later life. 16 Nutritional exposures and behaviours in infancy and early childhood are also increasingly recognized to be influential. 17 These include, short or no breastfeeding duration, 18,19 and the development of eating habits and behaviours, such as responsiveness to food, 20 high intake of energy dense foods and a higher rate of food consumption. 21 Longitudinal analyses suggest that once established, these eating habits and behaviours persist into adulthood. 22 Therefore, effective strategies implemented during these windows of vulnerability are needed to stem the rising trend of childhood obesity.
Previous studies addressing the early-life determinants of childhood obesity have generally focused on children born to women of heterogeneous BMI 17,23,24 in cohorts, which frequently have a small proportion of women with clinical obesity (BMI≥30 kg/m 2 ) yet children of women with obesity are those at greatest risk. In accord with the WHO ECHO report 3 recommending that effective public health strategies to prevent childhood obesity be tailored to high-risk women and their families, we have attempted to identify modifiable risk factors in a cohort confined to children born to women with obesity.
Maternal BMI was distributed across the WHO categories (I, II and III) and the women were ethnically diverse and of a predominantly social deprived backgrounds. In a follow-up study of the UK Pregnancies Better Eating and Activity Trial (UPBEAT) 25 we examined the association between six modifiable early-life exposures and their cumulative contribution on nine measures of offspring adiposity and obesity outcomes at 3 years of age. These exposures comprised early pregnancy BMI (WHO obesity categories I, II and III), GWG, mode of infant feeding and childhood eating habits (food responsiveness, slowness in eating and a processed/snacking dietary pattern score) at 3-years of age.

| Setting
UPBEAT, a multi-centre randomized controlled trial, investigated the effect of an intense 8-week diet and physical activity intervention in 1555 pregnant women with obesity (BMI≥30 kg/m 2 ). 25 Participants were randomized to the intervention or to standard antenatal care and were from UK inner-city settings of ethnic diversity and from predominantly socioeconomically deprived backgrounds. Details of the intervention inclusion and exclusion criteria have been published previously. 26  All participants provided written informed consent. The intervention had no effect on the primary outcomes, the incidence of maternal gestational diabetes and large-for-gestational-age infants. However, improvements were observed in several secondary maternal outcomes, including a reduction in total GWG. 25   Ltd.). Weight was measured to the nearest 0.1 kg with calibrated electronic scales (Seca), after removal of shoes and heavy clothing or jewellery; (2) GWG categorized using the National Academy of Medicine (NAM) guidelines 28 (inadequate: <5 kg vs adequate: 5-9 kg vs excessive >9 kg). GWG was calculated using estimated weight before pregnancy by the difference in the mother's weight measured at baseline minus 1.25 kg and weight recorded at 34 +0 -36 +6 weeks gestation, and (3) mode of infant feeding recorded on hospital discharge as exclusively breastfeeding, exclusively formula feeding or partial breastfeeding (defined as any breast feeding). The nutritional exposures were recorded at 3-years of age and included (1) a "processed/ snacking" dietary pattern score, (2) child's food responsiveness and (3) slowness in eating. The data collection and methodology of these dietary variables have been published previously. 29 In brief, dietary patterns were derived using factor analysis of a culturally appropriate 85-item food frequency questionnaire. Eating behaviours were assessed using the validated Childhood Eating Behaviour Questionnaire (CEBQ), 30 consisting of 35 items, divided into eight eating behaviours. Slowness in eating and food responsiveness, were selected for analysis; as being most amenable to intervention, 31 Furthermore, we reported strong associations between these two eating behaviours and adiposity and obesity in the 3-year old UPBEAT children. 29 2.4 | Child Outcomes: Body composition and measures of obesity and height (using the Leicester height measurer) to the nearest 0.1 cm was used to derive the WHO z-scores 32 and to define childhood overweight by International Obesity Task Force (IOTF) sexspecific centiles (boys overweight = 90.5th centile and girls overweight = 89.3rd centile). 33

| Statistical analysis
We have previously reported that the UPBEAT intervention did not affect any measure of adiposity/obesity in the 3-year old children, 34 therefore, the data was treated as a cohort. Demographic results were expressed as mean ± SD (SD), median and interquartile range or percent and number as appropriate. Children's sum of skinfolds was positively skewed and log-transformed for analysis. Depending on the outcome of interest, adjusted linear or logistic regression was used.
Poisson regression with robust variance was used to estimate the relative risk of child overweight. All outcomes were adjusted for maternal age, parity, ethnicity, smoking status at baseline, years spent in full time education, randomization arm and gestational age at delivery.
Additional adjustments were made for child age at follow-up (months) and infant sex when indicated. Using regression analyses, the first objective was to assess the relationship between the individual maternal exposures (early pregnancy BMI, excessive GWG and mode of infant feeding on hospital discharge) and measures of childhood adiposity and obesity outcomes. The second objective was to address the incremental impact of the exposure variables on childhood outcomes, using three composite models were created: (i) maternal exposures (BMI, GWG and mode of feeding on hospital discharge), (ii) child nutritional exposures (processed/snacking dietary pattern score, food responsiveness and slowness in eating) and (iii) the combined F I G U R E 1 Consort diagram of participants enrolled in the UPBEAT trial at 3 years after delivery contribution of all six exposures. To create the composite models, binary variables were derived for each exposure (BMI: 30.0-34.9 kg/ m 2 = 0, ≥35.0 kg/m 2 = 1; inadequate/adequate GWG = 0, excessive GWG = 1 and exclusively breastfeeding = 0, partial breastfeeding or formula feeding = 1). As the nutritional exposures are continuous variables, with no published reference guidelines to dichotomize the variables, we categorized a high (=1) association as mean ± 1 SD, with the remainder categorized as normal/low (=0); food responsiveness and the dietary pattern scores are positively associated with measures of adiposity, therefore, the high categories were defined as mean + 1 SD, as slowness in eating in negatively associated with measures of adiposity the high category was defined as mean -1 SD. Each child was assigned a score for the three models, the maternal model ranged from 0 to 3, the nutritional model ranged from 0 to 2, as categorizes 2 and 3 were combined. For the combined model and overall score was calculated ranging from 0 to 6 (0 was the reference group for all models). Using adjusted regression, the association with childhood outcomes was examined on a continuous and categorical scale. All data were analysed using Stata software, version 15.0 (StataCorp, College Station, Texas).

| RESULTS
Five hundred and fourteen (33%) of the 1555 women from UPBEAT agreed to participate and attended the follow-up appointments 3 years after delivery. Of those who attended the follow-up visit five children were born <34 weeks gestation, four were suffering from major ill health and 10 provided no outcome data for this analysis.
The study population, therefore, comprised of 495 children ( Figure 1).
The average age of the mothers at trial baseline was 31.2 years, 49% were nulliparous, 68% were White and the median early-pregnancy BMI was 34.7 kg/m 2 (32.5-37.9). The average GWG was 7.5 (4.3) kg and using the NAM guidelines for GWG, 28%, 35% and 37% of women were categorized as having inadequate, adequate and excessive GWG, respectively. 63% of mothers were exclusively breast feeding on hospital discharge. For the children, the average age at follow-up was 42 (3.4) months, the BMI z-score was 0.88 (1.0) and 35% were classified as having overweight/obesity according to the IOTF criteria (Table 1) Table 1).

| Combined contribution of maternal exposures
A score, from 0 to 3, based on three maternal exposures was generated for each child. 23% of the children had no exposures, 41% had one exposure, 30% had two exposures and 6% had three ( Table 2).
Child BMI z-score, weight-for-height (WH) z-score, arm circumference and overweight/obesity were included in the combined model analyses as these were most frequently associated with maternal risk factors. On a continuous scale, for each additional maternal exposure, child WH and BMI z-scores increased by 0.25 SD (0.13-0.36, P < 0.001) and 0.23 SD (0.12-0.36, P < 0.001), respectively. For overweight/obesity, the relative risk increased by 1.24 (1.07-1.44, P = 0.004), and arm circumference by 0.35 cm (0.16-0.55, P < 0.001).

| Combined contribution of childhood dietary exposures analysis
For the dietary exposures (range 0-3), categorizes 2 and 3 were combined. 58% of children had no exposures, 31% had one exposure, 11% had two or more exposures (Table 3) Table 4).

| Sensitivity analyses
As the risk factors are associated and may be a result of maternal obesity, we completed an additional analysis to examine the relationship between the offspring outcomes and the combined exposures with maternal BMI as a covariate, rather than an exposure. These results show that the significant incremental increase in childhood adiposity and obesity remained with the five risk factors. Furthermore, our decision to exclude infants born <34 weeks gestation may have resulted in collider bias as prematurity may be a mediator between GWG and childhood obesity. However, inclusion of all infants did not change the final results. also observed positive, additive associations between each of the maternal and nutritional exposures and childhood adiposity and obesity, suggesting that each of these modifiable factors are potential targets for intervention.

| DISCUSSION
In agreement with a meta-analysis of children aged 1-18 years, 12 and our previous study in 6-year old children from the Screening for Pregnancy Endpoints (SCOPE) cohort, 14 we confirmed that maternal early-pregnancy BMI was strongly associated with childhood adiposity and obesity in 3-year olds; 35% of the children studied were overweight or had obesity. Similar relationships have been widely described in animal models of maternal obesity in which maternal metabolic disturbances have been implicated in offspring obesity through persistent developmental changes in the foetus. 36 In common with others, we showed excessive maternal GWG to be a risk factor for offspring obesity. 37,38 A recent meta-analysis of 37 cohorts has assessed the separate and combined association of maternal BMI and GWG with the risk of overweight/obesity throughout childhood, and in common with this study, pre-pregnancy BMI was found to be a greater determinant of childhood overweight/obesity than GWG. This meta-analysis also identified the strongest association with maternal BMI and GWG to be in late childhood (10-18 years). Translating this to the UPBEAT children might suggest that the current 35% will be overweight or have obesity in later childhood and adolescence.
Formula feeding on hospital discharge (within 72 hours of delivery) was independently associated with increased rates of obesity and adiposity compared to breastfed children. We have already shown in this cohort that adiposity in 6-month infants was associated with formula feeding 39 and now report that this persists until early childhood.
This aligns with previous studies in children of weight heterogeneous women, including the CHOP RCT, which attributed higher adiposity rates in formula-fed children compared to breast fed children to the higher protein content of formula milk. 40 Our data also concord with the recent WHO European Childhood Obesity Surveillance Initiative from 22 European countries, which concluded that formula feeding from birth was associated with the highest rates of obesity in older children (6-9 years) born to women of heterogeneous BMI. 41 While the benefits of any breastfeeding on prevention of childhood obesity remain equivocal 42,43 our findings support initiation of breastfeeding at birth to reduce obesity in pre-school children born to women with obesity. Importantly, we report for the first time that maternal BMI, GWG and neonatal feeding combine independently to increase the risk of obesity in these children.
Several reports suggest independent relationships between eating behaviours 44,20 or dietary intake 45 29 In the present study, analysing the exposures on a continuous scale, we have found positive independent relationships between the high "processed/snacking" dietary pattern, food responsiveness and overweight/obesity. Conversely, we found a negative relationship between slowness in eating and childhood adiposity and obesity. This added information confirms a robust association between these exposures and childhood adiposity and obesity. While eating behaviours have been repeatedly causally associated with obesity, we are aware of one report, which infers bi-directionality of effect. 44 However, in that study the association was stronger between eating behaviours and obesity compared with the reverse.

| Strengths and limitations
Strengths of the study include the rich UPBEAT dataset, which provided comprehensive information including multiple indicators of childhood body composition and adiposity. The mother-child dyads were ethnically diverse and predominantly from low socioeconomic backgrounds, a population with a high-risk of obesity. 47 To our knowledge, the UPBEAT cohort is unique in the size of the cohort of women with pre-pregnancy obesity and with detail of in-utero, early postnatal and dietary exposures and multiple health outcomes, with a focus on measures of childhood adiposity, enabling adjustment for recognized confounding factors. Very few previous studies have focused on children of this age, yet pre-school adiposity tracks into adulthood. 6 The main limitation is the observational study design, which is subject to residual confounding and potential overestimation of reported effects. The measures of adiposity and obesity, although detailed, have limitations; BMI is an indirect measure of fat mass, and the BIA method has not been validated against dual-energy X-ray absorptiometry, the gold standard for adiposity measurement. 48 Attrition of the study population may have resulted in selection bias. However, the UPBEAT population studied at the 3-year visit was a representative sample of the main UPBEAT cohort. 34   The funders had no role in study design, data collection, data analysis, data interpretation or writing of the final report. The corresponding author had access to all the data in the study and had final responsibility for the decision to submit for publication.

CONFLICT OF INTEREST
In relation to the current work, none of the authors report any conflicts of interest. Prof Godfrey has been reimbursed for presenting at Nestlé Nutrition Institute conferences and has received grants from Abbott Nutrition and Nestec, outside the submitted work.