Adiposity and cardiovascular outcomes in three-year-old children of participants in UPBEAT, an RCT of a complex intervention in pregnant women with obesity

Objective: We have investigated whether the UPBEAT intervention influenced childhood cardiometabolic outcomes or was associated with sustained improvements in maternal lifestyle 3-years after delivery. Methods: In UPBEAT mother-child dyads at the 3-year follow-up, we assessed childhood blood pressure, resting pulse rate, and adiposity (body mass index, skinfold thicknesses, body fat, waist and arm circumferences) and maternal diet, physical activity, and anthropometry. Results: 514 three-year-old children attended the appointment (49% intervention, 51% standard care). There was no difference in the main outcome of interest, subscapular skinfold thickness, between the trial arms ( − 0.30 mm, 95% confidence interval: − 0.92, 0.31). However, the intervention was associated with a lower resting pulse rate ( − 5 bpm [ − 8.41, − 1.07]). There was also a non-significant lower odds of overweight/obesity (OR 0.73; 0.50, 1.08). Maternal dietary improvements observed in the UPBEAT trial, including glycaemic load and saturated fat were maintained 3-years postpartum. Conclusion: This study has demonstrated that an antenatal dietary and physical activity intervention in women with obesity is associated with lower offspring pulse rate and sustained improvement in maternal diet. Whilst larger than previous cohorts, there remains potential for bias from attrition and these findings require validation in future cohorts.


| INTRODUCTION
The World Health Organization (WHO) estimates that the global prevalence of childhood overweight and obesity will reach 70 million by 2025. 1 The causal pathways, widely explored in observational studies 2,3 suggest that maternal obesity may contribute to the development of childhood obesity through exposures during in utero development, 4,5 which remain following adjustment for confounders. 6,7 These relationships have been observed in animal studies, in which environmental and genetic contributions can be tightly controlled. 8 In contrast, observational studies using Mendelian randomisation, in which maternal genetic variants are used as instrumental variables to test the effect of maternal obesity on offspring adiposity, have not supported a causal intrauterine effect of greater maternal BMI on offspring adiposity, 9,10 inferring that the relationship is explained by shared genetic traits.
Epidemiological observations from mother-child cohorts have also reported associations between maternal obesity and cardiovascular morbidity and mortality rates in their children. 11, 12 The inference of in utero effects of maternal obesity on offspring cardiovascular function is supported by animal models; numerous studies in experimental animals, reported by ourselves 13,14 and others, 15 have described a relationship between pre-pregnancy maternal obesity and offspring cardiovascular dysfunction, including heart rate variability, enhanced cardiovascular response to stress, hypertension and higher circulating atherogenic lipids, [16][17][18] observations which have been reported consistently across species.
Many antenatal randomised controlled trials (RCTs) have attempted to reduce gestational weight gain (GWG) or improve obesity related pregnancy outcomes, especially gestational diabetes through antenatal diet and/or physical activity interventions. 19,20 Whilst improvement in gestational diabetes and other antenatal clinical outcomes has seldom been achieved, the majority of interventions have shown some benefit in limiting GWG and improving selfreported diet. 20 These RCTs provide an important opportunity to explore the causal relationship between maternal obesity and subsequent obesity and cardiovascular risk in the offspring, by studying children born to women who participated in these trials. However, few studies have progressed to childhood follow-up and, in those, that have, the sample size has frequently been inadequate to detect any effects with certainty. 21 The UK Pregnancies Better Eating and Activity Trial (UPBEAT), was a multi-centre RCT of a dietary and physical activity intervention in 1555 pregnant women with obesity. 22 Women were randomised to an intensive 8-week behavioural intervention or to standard antenatal care. The intervention had no effect on the primary outcomes, the incidence of gestational diabetes and large for gestational age (LGA) infants. However, there were improvements in several secondary maternal outcomes; specifically, lower total GWG, sum of skinfold thicknesses, dietary glycaemic load (GL) and saturated fat intake (SFA), and a modest increase in self-reported physical activity. The intervention also contributed to a healthier metabolic profile across pregnancy. 23 At six months postpartum we found the maternal dietary benefits of the intervention were sustained and also observed a lower infant subscapular skinfold thicknesses in the offspring of women randomised to the intervention. 24 The aim of the present study was to assess whether the UPBEAT intervention influenced childhood adiposity and cardiovascular function at three years of age and if improvements in maternal lifestyle behaviours were sustained three years after delivery.

| Study design and setting
This was a secondary analysis of the UPBEAT RCT. 22 We undertook a three-year postpartum follow-up study in eight trial centres. In the original trial, 1 555 women with obesity (≥16 years of age; prepregnancy BMI≥30 kg/m 2 ) were recruited in early pregnancy; exclusion criteria included pre-existing disease and multiple pregnancy. The participants were randomised to the intervention or standard antenatal care at 15 +0 -18 +6 weeks' gestation as reported previously. 25 In brief, the intention of the intervention was to prevent GDM through the promotion of healthy dietary intake and incremental increases in daily physical activity, over the 8-week intervention period. The dietary recommendations focused on reducing GL and SFA intake and were tailored to the woman's habitual diet and cultural preferences.
With respect to daily exercise, all women were provided with a pedometer and a DVD of suitable exercises. Further details are available in the protocol. 25

| Participants and consent
All participants provided written informed consent. Consent to the trial included agreement to contact the participants at a later date participants in the trial were invited to attend a three year postdelivery visit with their child. 22 Research midwives/research assistants completed the data collection. Continued training and regular contact between the sites was sustained throughout the data collection period. Women were excluded from the analysis if they were pregnant or had given birth in the previous four months at the time of follow-up. Children were excluded if they were suffering from severe illness (n = 4) (chronic lung disease, developmental delay, down's syndrome and Spina bifida) as these could affect growth or development or if they were born before 34 weeks' gestation (n = 5).

| Childhood outcomes
Since we had previously reported lower subscapular skinfold thickness in six month old children in the intervention, compared with the control arm, 24 subscapular skinfold thickness was the pre-specified childhood outcome of interest for the present study. Additional offspring outcomes included triceps, bicep, suprailiac, and abdomen skinfold thicknesses, and sum of skinfold thicknesses (calculated by addition of the five measures). All skinfold thicknesses were evaluated in triplicate using Holtain children skinfold callipers. Mid-upper arm and waist circumferences, estimated total body fat percentage (by bioelectrical impedance analysis; BIA, ImpediMed SFB7), weight (using a calibrated scale), WHO growth standard BMI z-score, 26 and age adjusted International Obesity Task Force (IOTF) BMI centiles were also determined. 27 The WHO reference standards are adjusted for age and sex and applicable irrespective of ethnicity and mode of early infant feeding. Childhood overweight and obesity were defined by IOTF sexspecific centiles (90.5th and 98.8th centiles for boys and 89.3th and 98.6th centiles for girls). 27 For BIA estimation of body fat percentage, the child was asked to lie on a couch for five minutes during data collection, after which blood pressure was measured in duplicate when feasible and a single resting pulse rate measurement was recorded by a WelchAllyn 53S00-E4 device, with an appropriately sized arm cuff. This order ensured measurement of resting pulse rate. Blood pressure was converted to age and height appropriate centiles. 28

| Maternal outcomes
Maternal diet and physical activity were assessed with the same questionnaires used in the original UPBEAT study 25 . These included a semi-quantitative food frequency questionnaire to estimate dietary GL, macronutrient, and energy intake. Women were excluded from this analysis if calorie intake was calculated to be under-reported at the baseline visit (15 +0 -18 +6 weeks' gestation). Physical activity was assessed using the International Physical Activity Questionnaire (IPAQ) and summarised as metabolic equivalents (METs) of energy expenditure. 29 Maternal anthropometric measurements at the three year follow-up included mid-upper arm, waist and thigh circumferences, subscapular, triceps, bicep, and suprailiac skinfold thicknesses (measured in triplicate using skinfold callipers). BMI was calculated from weight and height data using standardised methods.

| Statistical analyses
For summary statistics, binary and categorical variables are presented using counts and percentages. The distribution of continuous variables was assessed using coefficients of skewness and then summarised by mean and SD or median and interquartile range, where appropriate. Comparison of demographic details were made between the intervention and control groups; if the outcome of interest was binary, an odds ratio was calculated, when categorical, chi-squared test was used. Mann-Whitney U tests or t tests were used for continuous data, depending on the distribution of the data.

| Effect of the intervention on maternal and offspring outcomes three years postpartum
To analyse the effect of the intervention a complete case analysis (including only those with complete data on all variables used in any analyses) was undertaken for all participating mothers and children.
Treatment effects for continuous outcomes were expressed as differences in means obtained from multivariable linear or quantile regression. Linear regression was used for most outcomes, with quantile regression employed for sum of skinfolds and maternal physical activity as the data were positively skewed. Binary endpoints were expressed as odds ratios with 95% confidence interval using logistic regression. Analyses were adjusted for minimisation variables (maternal BMI at trial enrolment, parity and ethnicity) and child sex and age at follow-up.

| Sensitivity analyses to explore selection bias due to attrition
Although attrition was similar in each trial arm, we explored potential selection bias due to loss to follow-up by comparing maternal baseline characteristics and neonatal outcomes by randomisation arm for those included in this analysis (n = 514) with those lost to follow-up (n = 1 006).
We undertook additional analyses, imputing missing childhood outcome data due to loss to follow-up. For the offspring outcomes we used multivariate imputation chained equations to impute missing data for childhood adiposity and cardiovascular outcomes, to provide a total sample size of n = 1 520. Data were imputed to create 50 datasets using 10 burn-in iterations for live-born infants using the multivariate imputation model including: maternal early pregnancy BMI, age, ethnicity, parity, early pregnancy smoking status, randomisation arm, measures of maternal anthropometry including GWG, maternal diet (glycaemic load, saturated fat, carbohydrate, protein, energy intake) and physical activity at 27 +0 -28 +6 , 34 +0 -36 +0 weeks' gestation, gestation at delivery, birthweight, mode of feeding on hospital discharge and 3-year maternal diet (glycaemic load, saturated fat, carbohydrate, protein, energy intake) and physical activity, and child sex and age at follow up. Analyses were performed using Stata version 15.0 (StataCorp, College Station, TX, USA).

| Participants
One thousand, five hundred and fifty-five participants were randomised to the UPBEAT trial and 1233 were approached between three to four years after delivery with 1018 of these responding to contact. The predominant reason for the reduction in numbers contacted compared to the original study population was the child being outside the prescribed age range. Of the women originally randomised, n = 514 (33%) mother-child dyads took part (n = 250 intervention; 264 standard antenatal care), Figure 1. For the 514 mothers and children, 495 had complete outcome data, with 10 providing only questionnaire data completed at home. Nine children were excluded on the basis of severe illness or delivery <34 weeks' gestation. For those who completed the follow-up there was no difference in the majority of maternal baseline (trial entry) characteristics (Table 1) or neonatal characteristics (Table 2) between the intervention and standard care arm, except for a significantly higher odds ratio of LGA and a higher birthweight for infants in the intervention arm. Mothers who attended the three-year follow-up were on average, compared to those who did not attend, older (1.1 years), had a lower early pregnancy BMI, more likely to be White European and nulliparous, and less likely to smoke (Supplementary,   Table S1). There was a higher proportion of breastfeeding on hospital discharge amongst infants who completed the three-year follow-up (Supplementary, Table S2).
In this sub-population (n = 514), and in common with the original trial, sum of maternal skinfold thicknesses at 26 +0 -28 +6 weeks' gestation were lower in the intervention arm compared to the standard care arm, as were GL and reported SFA intake (Supplementary, Table S3).
In common with the main trial population, physical activity was higher in those in the intervention arm (Supplementary, Table S3). In contrast to the main trial, there was no significant difference in total GWG between the intervention and control groups (difference in mean in main trial −0.55 kg (95% CI: −1.08 to −0.02) vs −0.38 kg (−1.17 to 0.42), or in the metabolic profile in pregnancy (Supplementary, Figure S1). 22,23 3.2 | Intervention effects on childhood adiposity outcomes 34% of all children with adiposity measurements were classified as having a BMI equivalent to the adult BMI classification of ≥25.0 kg/m 2 , with 8% having obesity. The mean (SD) BMI z-score was 0.88 (1.0). There were no differences in the adjusted coefficients for BMI z-score between the intervention and standard care arms.
Despite a trend for lower odds of overweight/obesity in the intervention arm (OR 0.73; 95CI: 0.50, 1.08) and for lower adiposity as measured by skinfold thicknesses there was no statistical evidence for a difference between arms (Table 3) There were also no differences in the adjusted coefficients for waist circumference and mid-upper arm circumference between trial arms (Tables 3).

| Intervention effects on child cardiovascular outcomes
Resting pulse rate was −5 beats per minute (bpm) (−8.6 to −1.07) lower in the intervention arm (P < 0.01), compared with standard care (Table 3). Further analysis identified a bimodal distribution of pulse

| Effect of the intervention on maternal diet and body composition three years postpartum
Compared to women who received standard antenatal care, women in the intervention arm who provided complete dietary data reported lower glycaemic load, maternal energy and SFA intake, and higher protein intake three years after delivery (Supplementary,   Table S5).

| DISCUSSION
To our knowledge this investigation of 514 pre-school children born to mothers with obesity randomised to a lifestyle intervention in pregnancy is the most comprehensive reported to date. In a previous study in UPBEAT infants at six months of age (n = 698), we reported a reduction in subscapular skinfold thickness, a measure which, in adults, is associated with risk of metabolic disease. 30 At three years of age this effect was not sustained, despite trends towards lower adiposity, a lower incidence of overweight/obesity and diastolic blood pressure in the intervention group. The reduction in resting pulse rate in three- year-old children of mothers randomised to a lifestyle intervention, is an entirely novel observation. Notably, we found that the improvement in maternal diet during pregnancy in response to the UPBEAT intervention is still evident three years after delivery.
The reduction in the resting pulse rate of the three-year-old children could imply reduced cardiovascular risk. In adult populations, increased resting pulse rate is associated with hypertension and cardiovascular dysfunction. 31 Of the few reports in children, a higher resting pulse rate has, as might be anticipated, been related to higher blood pressure. 32 Resting pulse rate in children has also been reported to be inversely related to physical activity, 33 but this is an unlikely explanation for the difference in pulse rates observed between intervention arms in F I G U R E 3 Maternal dietary intake across pregnancy and to 3-years postpartum, by randomisation arm this study, as there was no association with parent-reported child activity and sedentary time and resting pulse rate. An association between maternal obesity and offspring cardiometabolic dysfunction is widely reported in experimental animals. [13][14][15][16]34 Rodent maternal obesity has been related to a sustained increase in offspring central sympathetic activity at the level of the hypothalamic neuronal pathways involved in peripheral autonomic regulation. In turn, this central pathway has been implicated in the sustained increase in blood pressure and altered heart rate variability observed. 16 Mechanistically this may occur through permanently changed hypothalamic function through epigenetic processes. 35,36 Our data could support a similar pathway to sympathetic activation in the children of women with obesity, and prompts more detailed investigation of sympathetic pathways and the epigenome in UPBEAT children at an older age. Assessment of the heart rate variability using ECG recordings, for example, would provide a read out of efferent parasympathetic and sympathetic autonomic activity 37 and the peripheral blood epigenome may provide insight into sustained epigenetic signals originating in utero. A recently published study of 184 women provides added support for in utero origins of altered autonomic nervous system activity. 38 Using the method of magnetoencephalography, the authors reported that maternal overweight and obesity was associated with lower fetal heart rate variability and a higher heart rate, in comparison to normal weight mothers. Furthermore, in a preliminary analysis of a study focusing on MRI assessment of newborn cardiovascular function, we have recently shown a significantly higher heart rate associated with abnormal cardiac structure and function in neonates born to mothers with obesity compared to those born to lean mothers (A. Groves, personal communication).
We could find no obvious explanation for the bimodal distribution of heart rate observed in the children as it was not influenced by maternal dietary intake or resting pulse rate, child's diet, weekly activity and sedentary time, time of day, seasonality or centre of measurement (heart rate monitoring device). To our knowledge this clear bimodal distribution has not been previously reported and the origin remains unknown.
The observation that the effect of the intervention on maternal diet was maintained to three years, having previously been demonstrated at six months 24 is important and has potential implications for longer term maternal and family health. It supports the theory that pregnancy is a "teachable" moment for initiating longer-term improvements in dietary intake. 39 In a planned follow up of the older children, we shall explore long term effects of the intervention on the mothers' behaviours and her health outcomes including obesity, Type 2 diabetes, hypertension and cardiovascular disease. Healthier behaviours in mothers may also impact upon the health of her offspring as they age, and of other family members, although we have previously reported no differences between dietary patterns in the 3 year old children from mothers in the standard care and intervention arms 40 . As the children grow older, we shall nonetheless assess relationships between in utero and contemporary family exposures with child health outcomes. As persistently healthier maternal behaviours may also impact on the next pregnancy, maternal and infant health in subsequent pregnancies will also be of interest.
Although there are many reports of the consequences of maternal obesity on offspring health from animal 15 and cohort studies 7 very few have addressed the influence of an antenatal intervention beyond infancy. 21 The only study of an antenatal diet and physical activity intervention in women with obesity, which was effective at reducing GWG, reported no effect on offspring body composition at 2.8 years of age (n = 254). 41 In conjunction with our six month 24 43 The repeatedly demonstrated modest effect of lifestyle interventions on pregnancy outcomes, has contributed to a new focus on optimising BMI before pregnancy. Pre-conception interventions, not limited to the narrow gestational window of nine months, are now seen by many as a as a potentially more effective strategy for improving pregnancy and longer-term outcomes for mother and child. 44 Ultimately, effective interventions which together improve health behaviours in the preconception period and achieve substantive improvements in pregnancy outcomes, will inevitably have greater reach and benefit.

| Strengths and limitations
Strengths of the study include the prospective collection of in-depth data of pregnancy demographic, health, metabolic and lifestyle variables and individual determinants of childhood body composition and health outcomes, allowing for adjustment of potential confounders. The principal limitation is the follow-up of 33% of those eligible from the original RCT, 22 and some minor differences in baseline characteristics between the two trial arms, may have resulted in selection bias. The main outcomes were however consistent when comparing complete case analyses with those from analyses using multivariable regression to impute data from those lost to follow-up.
The consistency between the two methods suggests that selection bias may not have importantly influenced our findings. 45 The lack of a significant reduction in GWG and pregnancy meta- In conclusion, this study provides some evidence to suggest that improving health behaviours in women with obesity may have a positive effect on cardiovascular health of the offspring but is not associated with a reduction in childhood adiposity at the age of three years.
The intervention in pregnancy, in common with many other studies, had only modest effects on maternal diet, weight gain, and adiposity.
It remains important therefore to develop better interventions in pregnant women with obesity before or during pregnancy which substantially improve maternal health, and to determine if these are associated with greater health benefits for the child. Importantly, we found that the improved dietary behaviours arising from the UPBEAT intervention were maintained three years after delivery. Further follow-up of the UPBEAT participants will be valuable in ascertaining the longer-term effects of the UPBEAT intervention and the mother's diet on her and her offspring's health, including detailed assessment of cardiovascular function and adiposity.