Effects of physical activity on reduction of adiposity across the entire obesity distribution in primary school-aged children: findings from the Millennium Cohort Study

Abstract

BackgroundAround one in five 4–5-year-olds living in England are overweight or obese. Inactive children are at risk of obesity and of continued inactivity and its attendant risks in later life. Observational studies have reported an inverse relation between physical activity (PA) and bodyweight, body-mass index (BMI), and adiposity in childhood. If obese children are less active as a consequence rather than as a cause of their weight status then we hypothesise that these associations would not be present across the entire weight distribution. We examined whether the conditional distributions of fat mass and fat-free mass indices vary according to objectively measured levels of PA and sedentary behaviour (SB) in a nationally representative population of primary-school-aged children. MethodsAccelerometer data were obtained between May, 2008, and August, 2009, from 6497 7-year-olds (3321 girls) enrolled in the UK-wide prospective Millennium Cohort Study (MCS). PA and SB were measured objectively with the Actigraph GT1M accelerometer, which children wore for 7 continuous days during waking hours. For each child, we derived the mean daily minutes spent in moderate and vigorous PA (MVPA) and in SB. Anthropometric measures were obtained at home interview between January and December, 2008. Height was measured with Leicester height measure stadiometers, recorded to the nearest 0·1 cm, and weight (to the nearest 0·1 kg) and percentage body fat with the Tanita BF-522W scales. Fat mass (kg) was derived by multiplying the percentage body fat by weight (kg), and fat-free mass (kg) by subtracting fat mass from weight. Fat mass index (fat mass [kg]/height [m]2) and fat-free mass index (fat-free mass [kg]/height [m]2) and the corresponding sex-specific and age-specific Z scores were calculated, which were the primary outcomes. We fitted quantile regression models to assess the association between MVPA, SB, and centiles of fat mass index and fat-free mass index. These models do not assume linearity or any distributional form. The models were adjusted for weekend and season of measurement, child's ethnic origin, maternal BMI, mother's age at the birth of the child, maternal socioeconomic circumstances, maternal education, household annual income, regular car use, lone parent status, number of children in the household, country of residence, electoral ward, and urban and rural indicators. Data were analysed with R version 2.14.2. Findings85% (95% CI 82–88) children in the sample were white, 51% (49–53) were boys, 88% (87–89) were living in households including at least one other child, and 62% (63–65) lived in England. The mothers of 40% (95% CI 38–42) of children were not employed during the data collection period, and 23% (21–24) were lone parents; 14% (13–15) of children were overweight and 6% (5–7) obese. The reported proportions are weighted estimates considering the MCS sampling design and attrition. We confirmed an inverse association between PA and fat mass index at the three quartiles of the distribution: the variation in fat mass index Z score for a 15-min increase of MVPA was −0·032 (95% CI −0·051 to −0·013) at the 25th percentile, −0·060 (−0·079 to −0·041) at the 50th centile, and −0·096 (−0·125 to −0·068) at the 75th percentile. We also noted an inverse association between SB and fat-free mass index: the variation in Z scores was −0·093 (95% CI −0·156 to −0·039) at the 50th centile for a 60-min increase in SB. InterpretationThe conditional distributions of adiposity indices in children at lower or higher levels of PA do not only differ for obese children. From a causative perspective, this finding suggests that reverse causation does not entirely explain the association between PA and adiposity. From a public health perspective, this finding supports a strategy of improving activity across the whole childhood population. FundingThe Wellcome Trust (grant 084686/Z/08/A) funded the accelerometry data collection within the MCS.