OBJECTIVES: To test the hypothesis that most excess weight gain occurs by school entry in a large sample of English children, and to determine when the greatest gain in excess weight occurred between birth and 15 years.
METHODS: Longitudinal data were collected annually from birth to 15 years in 625 children. Weight and BMI at each time point were expressed relative to UK 1990 growth reference as z scores. Excess weight gain was calculated as the group increase in weight and BMI z scores between specific time periods.
RESULTS: Weight z score did not increase from birth to 5 years (mean difference: 0.04 [95% confidence interval (CI): −0.03–0.12] P = .30) but increased from 5 to 9 years (mean difference: 0.19 [95% CI: 0.14–0.23] P < .001). BMI z score increased from 7 to 9 years (mean difference: 0.22 [95% CI: 0.18–0.26] P < .001), with no evidence of a large increase before 7 years and after 9 years.
CONCLUSIONS: Our results do not support the hypothesis that most excess weight gain occurs in early childhood in contemporary English children. Excess weight gain was substantial in mid-childhood, with more gradual increases in early childhood and adolescence, which indicates that interventions to prevent excess weight should focus on school-aged children and adolescents as well as the preschool years.
WHAT'S KNOWN ON THIS SUBJECT:
In a previous study, it was found that most excess weight gain in English children had taken place by age 5. The authors suggested that interventions to prevent obesity should focus only on children younger than 5 years of age.
WHAT THIS STUDY ADDS:
From birth to 5 years of age was not characterized by the greatest excess weight gain in a larger sample of English children and adolescents. Greatest gains in weight and BMI z scores occurred during mid-childhood. Efforts to prevent obesity should not only focus on preschool children, but also include school-aged children and adolescents.
There has been a rapid rise in the prevalence of childhood obesity in recent years in the United Kingdom1 and more widely,2,3 although in some nations pediatric obesity prevalence seems to be stabilizing.4,5 Obesity is now 1 of the most common pediatric health problems, and it has a range of adverse effects on physical and psychosocial health in childhood and adulthood.6,7
There is increasing evidence that growth in early life may be important to later risk of obesity, and excessive weight gain in early childhood in particular seems to be strongly predictive of later obesity risk.8,9 Timing of the “adiposity rebound” during the preschool years has long been considered to be a potential critical period for programming of later obesity, but evidence for programming is inconclusive,10 and early childhood may be important to later obesity risk simply because that is when “obesogenic” lifestyles and growth trajectories are often established.11 An improved understanding of the natural history of obesity development (how and when obesity typically develops) would improve our understanding of the etiology of pediatric obesity, and so inform the development of future interventions aimed at obesity prevention.12,13 However, it is unclear at present when excess weight gain typically occurs in childhood (eg, early, mid-, or late childhood) and therefore when best to target obesity prevention interventions.
In a recent study, Gardner et al14 examined excess weight gain (defined as an increase in weight z score) between birth, 5 years, and 9 years in small cohort of 233 English children. Gardner et al14 found that weight gain was substantial from birth to 5 years, whereas subsequent weight gain from 5 to 9 years was much smaller. Therefore, the study indicates that the preschool period is critical to the development of excess weight gain in English children. This study has received a great deal of attention in the United Kingdom because its findings have important implications for public health policy, indicating that future pediatric obesity prevention strategies should target preschool children if obesity is set by school age.14 However, the study by Gardner et al14 included only 134 boys and 99 girls who were followed to 9 years and used change in weight z score as the sole measure of weight status. Given the importance of these findings, the hypothesis that excess weight gain develops largely before the age of school entry needs to be further explored, ideally in a larger sample of children followed for a longer period, and extended by the inclusion of a more specific measure of weight status such as BMI z score. Therefore, the primary aim of the present study was to test the hypothesis (suggested by Gardner et al14) that most excess weight gain occurs by 5 years in a larger sample of English children, by comparing changes in weight and BMI z score between birth and 15 years. A secondary aim was to determine when the greatest gain in excess weight occurred.
The Avon Longitudinal Study of Parents and Children (ALSPAC) is a large prospective cohort study of children born in the Southwest of England in 1991 and 1992; study design and methods are described in detail elsewhere.15,16 Briefly, 14 541 pregnant women with an expected date of delivery between April 1991 and December 1992 were enrolled, resulting in 13 988 children alive at 1 year. Detailed information has been collected by using questionnaires, data extraction from medical notes, linkage to routine information systems, and at research clinics for children.
A 10% sample of the ALSPAC cohort, known as the Children in Focus (CiF) group, attended research clinics at 4, 8, 12, 18, 25, 31, 37, 43, 49, and 61 months, when detailed physical examinations were undertaken. The CiF group was chosen at random from the last 6 months of ALSPAC births (1432 families attended at least 1 clinic). The CiF group was broadly socioeconomically representative of both the entire ALSPAC cohort15 and the United Kingdom. From age 7 onward, the entire ALSPAC cohort was invited to attend regular research clinics. Ethical approval for the study was obtained from the ALSPAC Law and Ethics Committee and the local health service research ethics committees. In the present study, data from the CiF group are used as a means of replicating the findings of Gardner et al14 and meeting our primary aim. Data from the CiF sample were available for children up to 15 years old, and so were also used to meet our secondary aim of exploring when the greatest increases in weight status occurred during childhood and adolescence.
Gestational age and birth weight as recorded in the delivery room was obtained from medical charts, and birth length was measured by trained ALSPAC staff within 24 hours of birth where possible. At each measurement occasion, length/height was measured to 0.1 cm, and weight was measured to 0.1 kg in underwear. BMI was calculated as weight (kg)/length (m)2 at each time point. Weight and BMI at each time point were converted to z scores by comparison with British 1990 reference data17 by using age at measurement for each individual (or gestational age for birth weight z score).
Testing the ‘Gardner Hypothesis’
The present study aimed first to replicate the study by Gardner et al14 by quantifying gain in weight z score in the CiF cohort before and after school entry, ie, from birth to age 9 (in England almost all children have started school by 5 years old): Gardner et al14 considered weight z score changes before 5 years, and from 5 to 9 years. Although a degree of “regression to the mean” in weight status is to be expected in individuals,17 the pattern of group changes in weight status in United Kingdom cohorts studied after the onset of the obesity epidemic in the late 1980s has been 1 of increases in weight status (eg, BMI z score) with age and over time.14,18,19 Support for the Gardner hypothesis would be provided in the present study by evidence of increases in mean or median weight z score (or BMI z score), which were most marked up to around the age of school entry (age 5 in the United Kingdom), and which were relatively stable or declining thereafter. The present study would be inconsistent with the Gardner hypothesis if the increases in weight z scores (and/or BMI z scores) were more substantial after school entry (eg, between ages 5 and 9 years) than before age of school entry.
To meet our secondary aim (to identify periods of greatest gains in weight status across a wider age range than studied by Gardner et al14) we conducted a formal comparison of gains in weight and BMI z scores across the period of birth to 15 years.
Increases in weight and BMI z score with age are to be expected in United Kingdom cohorts of children and adolescents born in the 1990s,14,18,19 and these are considered as excess weight gain because they represent greater gains in weight status than expected from UK 1990 reference data.14,17
In the present study, changes in weight status during specific time periods was calculated (using the same method as Gardner et al14) as the difference in weight z score and BMI z score between ages (eg, weight gain at age 3 to 5 years equals weight z score at 5 years minus weight z score at 3 years). An increase in mean weight or BMI z score would reflect excess weight gain, as defined by Gardner et al14
In the present study, we used the CiF group in the analyses because research clinic weight and height measurements were available in early childhood (ie, up to 5 years) for this sample, and so this provided the best opportunity to replicate the findings of Gardner et al, and to examine our second aim of assessing when excess weight was greatest between infancy and age 15. Children born preterm (ie, <37 weeks' gestation; n = 40) were excluded from the analysis.
Research clinic-assessed weight and height data for the entire ALSPAC cohort were only available at birth and from 7 years onward. Although the CiF sample was selected randomly from the entire cohort, and would not be expected to be biased by selection, we checked this by comparing patterns of excess weight gain in the entire cohort and the CiF sample from 7 years old.
Statistical Analysis and Power
Power was fixed by the size of the CiF group, and so no formal power calculation was conducted. However, the sample size in the present study was much larger than the sample used by Gardner et al,14 by which the hypothesis being tested was suggested. Statistical analyses were performed using SPSS 16.0 (SPSS Inc, Chicago, IL). Weight and BMI z scores were normally distributed, and data are presented as means (SD).
To meet our primary aim of testing the Gardner hypothesis, within-child changes in weight z score and BMI z score during specific time periods (eg, birth to 5 years and 1 to 5 years versus 5 to 9 years) were calculated. The results obtained were then compared with those predicted by the Gardner hypothesis.14 The statistical significance of within-child changes in weight z score and BMI z score was determined using paired t tests. Changes between specified time points and 95% confidence intervals for the changes are presented along with P values. The analytical approach was identical to that taken by Gardner et al,14 although in the present study, changes in both BMI z score and weight z score were available, whereas Gardner et al reported only weight z scores.
To meet our secondary aim, to extend our exploration of the timing of increases in weight and BMI z scores beyond the age of 9, when Gardner et al14 stopped reporting data, within-child changes were calculated up to 15 years old, and the significance of changes between specified periods again was assessed by use of paired t tests and 95% confidence intervals for the changes.
We reran all analyses, taking the following considerations into account: including children born at <37 weeks' gestation; restricting analysis to participants with all data at all time periods (n = 447); and restricting analysis to participants with a difference of 1.5 to 2.5 years between measurement periods. Similar results were obtained for the rerun analyses (available from author on request). We compared study participants who were included in the analyses of change in weight z score from 1 to 3 years and from 11 to 13 years with those who were included in the 1- to 3-year analysis but were lost to follow-up by the 11 to 13 years analysis for a number of characteristics using t tests.
Characteristics of Study Participants
There were 1358 children who attended at least 1 CiF clinic during infancy, and had a gestational age of ≤37 weeks. Of the 1358 (730 boys, 628 girls), 96.3% were from the majority (white) ethnic group (1297); 773 (59.6%) mothers had been educated to 16 years old, and 525 (40.4%) had been educated to 18 years or beyond; 7.0% (85) of mothers and 87 (9.3%) of their partners were obese (defined as BMI > 30.0, with self-reported weights at 12 weeks' gestation).
We compared study participants who were included in the analyses of change in weight z score from 1 to 3 years and from 11 to 13 years with those who were included in the 1- to 3-year analysis but were lost to follow-up by the 11- to 13-year analysis. Compared with those who were included in both analyses, slightly more boys were lost to follow-up and had mothers who were slightly less educated and were more likely to be obese (Supplemental Table 4 includes comparisons of those followed up versus lost to follow-up; Supplemental Table 5 includes anthropometric characteristics from birth to 15 years). Weight and BMI z scores were similar between those followed up and lost to follow-up.
Primary Aim: Test of the ‘Gardner Hypothesis’; Changes in Weight z Score to 5 Years and 5 to 9 Years Described Longitudinally
Weight and BMI z scores for the CiF sample are included in Table 1 from birth to 15 years old (n = 625 at age 15). In Table 1 it is shown that, in contrast to the predictions from Gardner et al, mean BMI and weight z scores increased steadily over the study period, and changes seem most marked after the age of school entry.
The changes in weight z scores from birth to 5 years and 5 to 9 years are shown in Table 2, in the most direct comparison with the Gardner et al study.14 In contrast to the findings of Gardner et al, weight z score did not increase significantly from birth to 5 years, whereas increase in weight z score was statistically significant from 5 to 9 years in the present study (Table 2). Findings were similar in both boys and girls; weight z score did not change significantly from birth to 5 years (z score for boys was 0.06, P = .21; for girls it was 0.02, P = .78) and increased significantly from 5 to 9 years among boys (z score: 0.26, P < .001) and to a lesser extent in girls (z score: 0.10, P < .001).
Primary Aim: Test of the ‘Gardner Hypothesis’; Changes in BMI z Score From 1 to 5 Years and 5 to 9 Years Described Longitudinally
There were small but statistically significant increases in BMI z score from 1 to 5 years and 5 to 9 years (Table 2), although-in contrast to the prediction of Gardner et al- there were no marked differences in BMI z score gain between the 2 time periods. Among boys, BMI z score increased from 1 to 5 years (0.10 z score, P = .01) and 5 to 9 years (0.14 z score, P < .001). Among girls, there was an increase in BMI z score from 1 to 5 years (0.09 z score, P = .03) but not from 5 to 9 years (-0.03 z score, P = .46).
Secondary Aim: Timing of Excess Weight and BMI Gain to 15 Years; Changes in Weight and BMI z Scores Described Longitudinally
The change in weight and BMI z scores to 15 years old are shown in Table 3. Weight z score did not increase significantly in early childhood (ie, 1–7 years). In contrast, there were statistically significant gains in weight z score from 7 to 9 years and 9 to 11 years, with no marked increases thereafter. BMI z score increased significantly from 7 to 9 years, with no evidence of a large increase before 7 or after 9 years of age.
Results obtained from the entire ALSPAC cohort (from age 7 to 15) did not differ substantively from those obtained in the CiF sample (Supplemental Table 6 and Supplemental Table 7).
Modest gains in weight and BMI z scores were evident throughout the period of the present study, although these were generally most marked, and most likely to be statistically significant in the periods after the age of school entry, in particular between 7 to 9 years. Moreover, within-child correlations between BMI z scores across childhood in the present study were significant, but declined with age: for example, correlation coefficients from age 3 were 0.80 at age 5, and fell to 0.49 at age 11, which suggests more movement in weight status than might be predicted from the findings of Gardner et al.14 Thus, the present study does not support the hypothesis suggested in the study by Gardner et al14 that most excess weight gain occurs in early childhood (before school entry) in English children, or that weight status is set by the time of school entry.
Gardner et al14 quantified changes in weight z score only, and from birth to 9 years. In the present study we investigated changes in both weight and BMI z scores in a larger sample of children over a longer period of time, from birth to 15 years. The reasons for the differences in findings of the present study and those of Gardner et al are unclear. The sample in Gardner's study was born in 1995, and the CiF subsample in the present study born in 1992. This 3-year difference means that a period effect might have been operating: the extent to which children of any age were susceptible to excess weight gain might have differed between periods. Geographical or socioeconomic differences between the 2 populations might also contribute to differences between our study findings and those of Gardner et al14
In the present study we used the CiF subgroup of the ALSPAC cohort for the analyses because weight and height measurements were available in early childhood (ie, up to 5 years) for this sample, whereas data were only available at birth and from 7 years onward in the entire ALSPAC cohort. However, to examine the consistency of our findings across the entire cohort, we also examined excess weight gain from birth to 7 years and from 7 to 15 years in the entire ALSPAC cohort, and the results confirmed the findings from the CiF group (see Supplemental Table 6 and Supplemental Table 7).
Study Implications and Comparisons With Other Studies
One interpretation of the study by Gardner et al,14 which has gained widespread currency in the United Kingdom,20 is that if most excess weight gain occurs before school entry, then obesity prevention interventions should focus on preschool children, with a reduced emphasis on such interventions during later childhood and adolescence.20 Because the results of the present study are not consistent with those of Gardner et al14 the present study provides no support for the suggestion that obesity prevention and treatment strategies should be focused exclusively on young children.
Evidence from other populations should be considered when examining the Gardner hypothesis. In some studies the age-related pattern of obesity prevalence does not suggest that the years before school entry are especially obesogenic relative to later periods. It is suggested by recent data from the United States, for example, that obesity risk increases with age from the preschool years into the middle and elementary school years.21 In their review, Whitlock et al22 also suggested that screening for child and adolescent obesity would not identify those likely to be obese as adults very effectively, implying that weight status is not “set” by school entry.
Identifying the optimal time to target pediatric obesity preventive and treatment interventions will depend on a number of factors, including studies of the natural history of excess weight gain,12,13 but also the efficacy and cost-effectiveness of interventions applied at different ages.23,24
Direct comparisons of the findings of the present study with previous studies is difficult because so few previous studies have focused on quantifying the extent and timing of excess weight gain in contemporary children and adolescents, with the notable exception of the study by Gardner et al. However, it was reported in a recent study of a large sample of English adolescents25 that persistent obesity was generally established before 11 years, and marked increases in obesity risk were not present in ages 11 to 15 in England, broadly consistent with the findings of the present study.
Strengths and Limitations of the Present Study
A degree of attrition in longitudinal studies is inevitable, and although some characteristics of participants lost to follow-up differed slightly from those retained in later analyses, changes in both weight and BMI z scores were similar between groups in the present study; it seems likely that loss to follow-up did not have a marked impact on study findings. Analyses restricted only to those children with complete follow-up data were essentially the same as those with the entire sample.
One major issue arising from the present study is generalizability. The characteristics of the ALSPAC study participants will have a bearing on generalizability. The sample was broadly representative of the United Kingdom in terms of socioeconomic status, but estimated maternal obesity prevalence seems to be low because it is based on self-reported weights obtained at 12 weeks' gestation in 1991 and 1992. Both the present study and the study by Gardner et al14 that we attempted to replicate were very dependent on the UK 1990 reference data for weight and BMI. Apparent patterns in timing of excess weight gain may have arisen because of features inherent to the reference population, and both the present study and that of Gardner et al are limited by any weaknesses in the reference data. Extrapolation of the present study findings to other populations and other sets of population reference data should also be considered with caution: differences between populations and between reference data sets mean that the magnitude and pattern of change in weight and BMI z score might be quite different in other cohorts in other settings. The present study is best considered as an exploration of the pattern and timing of excess weight gain within the English population of children and adolescents.
Our results do not support the hypothesis that most excess weight gain occurs in early childhood in contemporary English children. In this cohort born in the early 1990s and followed to 2008, excess weight gain was substantial in mid-childhood, with more gradual increases in early childhood and adolescence. Population-based interventions to prevent excess weight should therefore focus on school-aged children and adolescents as well as the preschool years.
The United Kingdom Medical Research Council (Grant 74882), the Wellcome Trust (Grant 076467), and the University of Bristol provide core support for ALSPAC. The United Kingdom Medical Research Council (G0600705) and the University of Bristol provide core funding for the MRC Centre of Causal Analyses in Translational Epidemiology.
We thank the families who took part in this study, the midwives for their help in recruiting them, and the whole ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists, and nurses.
- Accepted November 22, 2010.
- Address correspondence to John J. Reilly, PhD, School of Psychological Sciences and Health, University of Strathclyde, Jordanhill Campus, 76 Southbrae Dr, Glasgow G13 1PP, Scotland. E-mail:
This publication is the work of the authors. Dr Hughes and Prof Reilly will serve as guarantors for the contents of this article.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
- ALSPAC =
- Avon Longitudinal Study of Parents and Children •
- CiF =
- Children in Focus
- Reilly JJ,
- Dorosty AR,
- Emmett PM
- Reilly JJ,
- Methven E,
- McDowell ZC,
- et al
- Reilly JJ,
- Kelly J
- Reilly JJ,
- Armstrong J,
- Dorosty AR,
- et al
- Baird J,
- Fisher D,
- Lucas P,
- Kleijnen J,
- Roberts H,
- Law C
- Gardner DSL,
- Hosking J,
- Metcalf BS,
- Jeffery AN,
- Voss LD,
- Wilkin TJ
- Ness A
- Golding J
- Cole TJ,
- Freeman JV,
- Preece MA
- Rudolf MC,
- Greenwood DC,
- Cole TJ,
- Levine R,
- Sahota P,
- Walker J,
- Holland P,
- Cade J,
- Truscott J
- 20.↵British Broadcasting Corporation. Obesity set before age of five. Available at: http://news.bbc.co.uk/1/hi/health/7786229.stm. Accessed January 18, 2011
- Whitlock EP,
- Williams SB,
- Gold R,
- Smith P,
- Slipman S
- Wardle J,
- Brodersen NH,
- Cole TJ,
- Jarvis MJ,
- Boniface DR
- Copyright © 2011 by the American Academy of Pediatrics