Associations of childcare type, age at start, and intensity with body mass index trajectories from 10 to 42 years of age in the 1970 British Cohort Study

School of Sport, Exercise and Health Sciences, Loughborough University, Loughborough, UK Centre for Longitudinal Studies, University College London (UCL) Institute of Education, London, UK UKCRC Centre for Diet and Activity Research (CEDAR), MRC Epidemiology Unit, School of Clinical Medicine, University of Cambridge, Cambridge, UK Department of Health, Behavior and Society, Johns Hopkins Bloomberg School of Public Health, Johns Hopkins University, Baltimore, Maryland


| INTRODUCTION
Obesity is a major public health problem, accounting for four million deaths globally and 120 million disability-adjusted life-years in 2015. 1 In the United Kingdom (UK), approximately 25% of the adulthood population are currently affected by obesity, and 40% of men and 30% of women by overweight. 2 Even more worrying is that more recent generations are developing obesity at increasingly younger ages. 3The latest available estimates from the UK's National Child Measurement Programme report that 22.3% of children in Reception year (aged 5 years) and 34.3% of children in Year Six (aged 10-11 years) were already classified as having overweight or obesity. 4cause obesity tracks from childhood to adulthood, 5 this means that future rates of obesity in adulthood are projected to increase. 6ildcare settings (eg, nurseries and playgroups) may offer an ideal opportunity for population-level interventions and targeted policy changes.8][19][20] All these publications found that available evidence is largely limited to cross-sectional and short-term follow-up studies.The two reviews focusing on obesity risk during early childhood 19,20 found mostly mixed or harmful effects, while the two reviews that included outcomes at later ages 17,18 generally found no association or a harmful effect.There is some evidence that earlier age of starting childcare (indicating greater duration of exposure) and higher intensity of attendance are particularly associated with greater obesity.While certain aspects of childcare may be related to increased risk for obesity in childhood, we know almost nothing about how exposure to childcare might be associated with body mass index (BMI) into and across adulthood.Such knowledge would provide evidence that targeting childcare policies and practices might ultimately benefit adulthood health and wellbeing.
We aimed to examine the associations of childcare attendance, and type, age at start, and intensity among those children who did attend, with BMI trajectories from 10 to 42 years of age in the 1970 British Birth Cohort.

| Outcome
Weight and height were measured by community medical officers, health visitors, or school nurses at ages 10 and 16 years according to standard protocols.At the adulthood sweeps, weight and height were self-reported in questionnaires (age 26 years) or face-to-face interviews (ages 30, 34, and 42 years).BMI was computed as weight (kg)/ height (m 2 ).In total there were 32 563 observations of BMI.Approximately 65% of the sample had four or more observations (out of a maximum of six) and approximately 72% of the sample had serial BMI measurements spanning more than 20 years (out of a maximum of 32 years).

| Exposures
At age 5 years, mothers reported type of childcare setting, age when the participant started, and how many morning and afternoon sessions per week the participant attended.Because these data were collected only for the most recent and previous main childcare placements that lasted for 3 months or longer, our exposure variable is limited to the main childcare setting attended by the participant.The first exposure we computed was attendance: "no" if all placements were coded as does not attend or "yes" if any placement was coded as does attend.The second exposure was type: "formal" for nursery school, nursery class attached to a school, or day nursery or "informal" for playgroup.The "informal" classification is restricted to playgroup because no detailed information on the number of sessions per week and age at start and finish of other informal care, such as friends or grandparents.Nursery schools, nursery classes, and day nurseries were formal settings run by professionals, many linked or attached to primary schools, which could be privately owned or public. 23Local authorities were responsible for providing this public provision, but it was not mandatory for local authorities to secure such provision. 23Playgroups were a non-for-profit provision mainly set up by mothers during the 1960's to counter the lack of nursery school places available, 23 which were typically open for two or three sessions per week and each session lasting up to 3 hours. 24Children reported as attending special education settings, a crèche, or other settings were removed from the sample due to low numbers.The third exposure was age at start: "4 to 5," "3 to 3.99," or "0 to 2.99" years old when started.The fourth exposure was intensity: "1 day/week" if attending 1 to 2 sessions, "2 days/week" for 3 to 4 sessions, "3 days/week" for 5 to 6 sessions, or "4 to 5 days/week" for 7 to 10 sessions.When the current placement started after 5 years of age, or a mother reported that their child did not attend childcare, or data were missing, we used the information from the most recent previous placement (with complete data) to derive exposures two to four.

| Potential confounders
In addition to sex (female vs male), Father's occupational class and mother's age of leaving full time education were considered as proxies of socio-economic position (SEP).Father's occupation, measured when the participant was 10 years old, was classified according to the Registrar General's Social Class (I professional, II managerial and technical, IIIN skilled non-manual, IIIM skilled manual, IV partly-skilled, and V unskilled). 25Mother's age at leaving full time education was ascertained at the birth sweep, and was categorized as ≥17 years old, 16 years old, 15 years old, or ≤14 years old.

| Statistical analyses
Descriptive statistics were produced.Further, in order to understand the interrelatedness of childcare type, age at start, and intensity, these three categorical variables were tabulated against each other.χ 2 test were used to quantify the strength of association between any given two variables.

| Trajectory modelling
BMI trajectories were modelled in a multilevel general linear regression framework (measurement occasion at level one and individuals at level two), 26,27 incorporating systematic differences in the sampleaverage trajectory according to the childcare exposures and adjustment for potential confounders.
The time scale was decimal years of age modelled as a linear spline with a single knot at 26 years of age, thereby producing three easily interpretable parameters: (a) BMI (kg/m 2 ) at age 10 years, (b) BMI change (kg/m 2 /year) between ages 10 to 26 years, and (c) BMI change (kg/m 2 /year) between ages 26 to 42 years.The constant and two spline terms were allowed to have random effects at level two, with an unstructured variance-covariance matrix.Further, the level one variance (ie, error) was allowed to differ according to whether the data were measured or self-reported.Placing the knot at the other logical point (ie, age 30 years) or using more complex functions, including fractional polynomials, did not result in noticeably better fitting models.

| Individual childcare exposures
In the first set of models, each of the four childcare exposures was considered separately: (a) childcare attendance (yes vs no), (b) type of childcare (none, formal or informal childcare), (c) age at start of childcare ("4-5," "3-3.99," or "0-2.99"years old when started), and (d) intensity ("1 day/week," "2 days/week," "3 days/week," or "4-5 days/week").Each exposure was included as a main effect and as interactions with the two spline terms, thereby producing estimates capturing the association of the exposure with BMI at age 10 years, BMI change between ages 10 to 26 years, and BMI change between ages 26 to 42 years.Sex, father's occupational class, and mother's age at leaving full time education were included in the same way to provide robust adjustment for these potential confounders.For parsimony, the two SEP variables were converted and entered into models as ridit scores; associated estimates capture the difference in BMI between the lowest and highest SEP.The models were also used to estimate associations of the exposures with BMI at ages 26 and 42 years.No strong evidence of effect modification by SEP was found in stratified models or in models incorporating exposure-by-SEP interaction terms.Similarly, sensitivity analysis further controlling for two additional SEP variables -childhood household income and housing tenure (as an indicator of wealth) -did not modify the results.

| Different childcare age at start and intensity combinations
Because childcare type, age at start, and intensity were related to each other, it was not prudent to fit a single model that included mutual adjustment for these three variables.Instead, in the second set of models, the relationships of different age at start and intensity of childcare combinations with BMI trajectories were investigated in all children who were reported to attend childcare.The childcare intensity variable was collapsed and combined with the age at start of childcare variable to create a new exposure with six responses: (a) 4 to 5 years old when started and attended 1 to 2 days/week, (b) 4 to 5 years old when started and attended 3 to 5 days/week, (c) 3 to 3.99 years old when started and attended 1 to 2 days/week, (d) 3 to 3.99 years old when started and attended 3 to 5 days/week, (e) 0 to 2.99 years old when started and attended 1 to 2 days/week, and (f) 0 to 2.99 years old when started and attended 3 to 5 days/week.This exposure was entered into a single model in the same way that each individual exposure was investigated in the first set of models, with the same set of adjustments for potential confounders.Again, no strong evidence of effect modification, this time by childcare type (ie, informal vs formal), was observed in exploratory analyses.

All procedures were performed in Stata 15 (StataCorp LP, College
Station, TX).The command runmlwin was used for the multilevel models. 28

| RESULTS
Approximately 85% of the sample reported using childcare, with a preference for informal over formal childcare (62% vs 38%) (Table 1).
Among those who attended any childcare, the majority (61%) started between 3 to 4 years of age and the largest proportion (36%) attended for just 1 day a week.As shown in Table 2, however, age at start and intensity differed according to type of childcare, with children in formal childcare tending to start at a later age and attend more times per week than children who attended informal childcare.Conversely, those children who started at a later age tended to go more times per week.

| Individual childcare exposures
Table 3 shows the estimated associations of each childcare exposure with BMI trajectories from 10 to 42 years of age.Children who attended childcare did not have higher BMI at age 10 years than those who did not attend childcare (β 0.039 kg/m 2 ; 95% CI −0.103, 0.180), and there was no evidence that this association changed over followup.When childcare was separated into "formal" and "informal" we similarly found no evidence that these groups had different BMI trajectories compared to children who did not attend childcare.Results for the age at start exposure did show that the BMI of participants who started childcare at 0 to 2.99 years of age was 0.266 kg/m 2 (95% CI 0.095, 0.437) higher at 10 years of age than participants who started childcare at 4 to 5 years of age.But they subsequently gained less BMI between ages 10 to 26 years (β −0.013 kg/m 2 /year; 95% CI −0.032, 0.006) such that effect sizes were attenuated and null at ages 26 and 42 years.For the intensity exposure, however, attending childcare 4 to 5 days a week (compared to 1 day a week) resulted in higher BMI at age 10 years (β 0.170 kg/m 2 ; 95% CI −0.006, 0.058) and this effect persisted to age 26 years (β 0.349 kg/m 2 ; 95% CI 0.024, 0.673) and age 42 years (β 0.380 kg/m 2 ; 95% CI −0.066, 0.826).

| Different age at start and intensity of childcare combinations
Table 4 shows the estimated associations of different age at start and intensity of childcare combinations with BMI trajectories.At 10 years of age, there was a clear dose-response relationship between greater exposure to childcare and increased BMI.For example, among participants who attended childcare 1 to 2 days a week, those who started when 3 to 3.99 years old had a 0.197 kg/m 2 (95%CI −0.004, 0.399) higher BMI than those who started when 4 to 5 years old, and those who started when 0 to 2.99 years old had a 0.289 kg/m 2 (95%CI 0.049, 0.529) higher BMI.Similarly, when holding age at start of childcare constant, estimated effects sizes were always larger for participants who attended childcare 3 to 5 days a week compared to those who attended 1 to 2 days a week.These results clearly demonstrate that age at start and intensity of childcare had additive effects on BMI in late childhood.Subsequently, at ages 26 and 42 years, the effect sizes were not as well ranked according to the level of exposure to childcare.Nonetheless, even at 42 years of age, individuals with the highest level of exposure (ie, 0-2.99 years old when started and attended 3-5 days/week) had a 1.356 kg/m 2 (95% CI 0.637, 2.075) higher BMI than individuals with the lowest level of exposure (ie, 4-5 years old when started and attended 1-2 days/week).

T A B L E 1
Description of sample of 8234 participants Using data from a large UK birth cohort study, our results demonstrate a clear dose-response relationship between starting childcare earlier and attending more frequently with higher BMI at age 10 years, in line with the findings of the Black et al systematic review. 18While this finding is important in and of itself, the real gap in the literature is that, after childhood, we know almost nothing about how exposure to childcare might be related to BMI trajectories.
The key finding of the present paper is that greater exposure to childcare resulted in higher estimated BMI up until age 42 years, perhaps with intensity being more deleterious than age at start of childcare attendance.
Individuals with the highest level of childcare exposure were estimated to have between a 0.5 to 1.5 kg/m 2 higher BMI than individuals with the lowest level of exposure across the studied age range.
This effect size corresponds to approximately 0.1 to 0.3 standard deviations (SDs), which is not negligible given that a BMI reduction of more than 0.20 to 0.25 SDs has been proposed to be clinically important. 29We similarly found no evidence of an association between childcare attendance or type with BMI at ages 42 years but did find evidence for childcare intensity.Batty et al also reported no association of nursery attendance with a range of cardiometabolic disease risk factor (eg, blood pressure and cholesterol) and mortality.Further work, however, needs to investigate the relationships of age at start and intensity of childcare with these and other outcomes.key role in the establishment of healthy behaviors and weight across the life course, instead of being associated with a more adverse BMI trajectory.
, and 42 (2012) years.The study has received ethical approval and obtained informed parental and/or participant consent for all data collection; detailed information is available from the cohort study website and Sheperd & Gilbert's 22 review of ethical procedures used for all sweeps of data collection.At the most recent sweep, the 9841 individuals still participating in the study remained broadly representative of the national population of men and women of the same age.Starting with the 14 874 cohort members who were still participating at age 10 years (ie, when BMI was first assessed), 4843 were excluded because of missing childcare data, 608 because they did not have a single BMI measurement, and 1189 because of missing data on potential confounders.The resulting sample comprised 8234 individuals, representing 55% of the eligible cohort (ie, N = 14 874).

T A B L E 2
Tabulations between childcare type, age at start, and intensity among 6969 participants 32,33 studies have also reported higher BMI in children cared for by grandparents.32,33Unfortunately,dataon care by grandparents was not collected in the 1970 British Cohort Study.There are only two studies that we are aware of that have investigated the association of childcare with BMI in adulthood.In a sample of 783 women attending Cracow or Opole universities in Poland in 2005, Wronka and Pawlinnska-Chmara found no association of childcare type with BMI at 20 to 24 years of age. 34In another of the UK nationallyrepresentative studies, the 1958 National Child Development Study, Batty et al found that the BMI at age 44/45 years of individuals who attended nursery was only 0.01 (−0.05, 0.07) SDs higher compared to individuals who did not attend nursery.
31evious research in another of the UK nationally-representative studies, the 2000 Millennium Cohort Study, reported that children in informal childcare were more likely to be affected by overweight at age 3 years than those cared for by a parent, but this may have been driven by 75% of informal care in that sample being primarily by grandparents.31