Statistics from Altmetric.com
What is the problem?
Recent surveys have found almost a quarter of children at school entry to be overweight or obese, leading to suggestions that this is where preventive interventions should be focused. However, others dispute this, as tracking of body mass index (BMI) from preschool age is weak, with 63% of obese toddlers reverting to overweight or healthy weight at later ages,1 while BMI z scores relate weakly to adiposity in younger compared with older overweight children.2
This paper explores the possibility that this lack of persistence of early obesity may be the result of a classification artefact. This issue arose when an obese toddler, seen by a specialist, was reported to have improved, despite no change in BMI. The explanation for this was evident on the BMI chart. This is illustrated with a fictitious example in figure 1, where a child’s centile normalises over time, despite no reduction in BMI. We are used to the idea of children catching down towards normality, as other children grow taller or heavier around them, as the whole growth curve rises over time. However, here the median BMI is the same at both ages. Can it be right that extreme BMI centiles are crossed at so much lower levels in younger children than in mid-childhood, when the average BMI is not increasing?
How is obesity in childhood currently defined?
We still have no properly validated upper threshold for unhealthy BMI in childhood. The upper thresholds for healthy adult BMI were established using their association with adverse health outcomes and mortality,3 but it was (and is still) not possible to do this in childhood. This is because we lack the large-scale, long-term data required, as until recently overweight in childhood was rare, while most adverse adult health outcomes do not occur until middle age or beyond. Therefore, the upper centiles on BMI charts were first introduced simply ‘to identify children who were unusually fat or thin’.4 However, using a fixed centile as an upper threshold defined the same proportion of children at all ages as being at risk of obesity. As time has gone on, the arbitrary nature of those thresholds has been forgotten, and they are now generally considered to represent actual overweight and obesity. We currently have no way of determining the true prevalence of obesity at different ages, but it seems unlikely that the prevalence of overweight and obesity would actually be constant across childhood, as few newborn infants are overweight, yet two-thirds of adults are overweight. When these centile thresholds are applied to contemporary BMI data5 (figure 2), the proportion with overweight and obesity is much higher than when the thresholds were set, reflecting the marked overall increase in obesity over time, but the proportions are still broadly similar at all ages, although rather higher in the teens.
What do the BMI centile curves tell us?
Median BMI is largely constant up to the age of 9 and rises gently thereafter (figure 1). In contrast, the upper limit of ‘normal’ BMI increases more steeply with age, giving the BMI chart a distinctive wedge shape. An interpretation of this is that obesity is actually still relatively rare in the first few years, so that the range of BMI is narrow. Then, as increasing numbers of obese-prone individual children become overweight and then obese, the upper limit increases, reflecting an increasing proportion of children with unhealthy BMIs. Meanwhile, the median continues to reflect the BMI of mainly non-obese individuals. This upper limit also seems to increase more steeply with age in populations with higher rates of obesity. The UK 1990 (UK90) reference4 used data collected early in the UK obesity epidemic in the 1970s and 1980s, the WHO 2007 reference used US data collected before 1977,6 while the US Centers for Disease Control and Prevention (CDC) reference includes data collected up to 1994.7 The median curves for all three charts are very similar at most ages, and there is little difference between the +2 SD lines before age 5. However, by age 12 the recent CDC +2 SD line is 3.5 kg/m2 higher than both UK90 and WHO, and by age 15–19 years the WHO is more than 1 kg/m2 higher than the UK90. This variability has also been described in the data sets used to construct the International Obesity Task Force (IOTF) cut-offs, where the coefficient of variation, a measure of the width of the distribution, rose sharply for all data sets in the first few years, but differed substantially between countries in their slope.8
An alternative approach to assessing BMI?
Given that the median BMI is much less variable, both over age and between populations, we should consider whether the per cent above (or below) median BMI (%median) would be a useful alternative. This approach, which has been used widely for growth monitoring in the past, adjusts for the change in the average value with age, but not for any change in the amount of variability. If 30 years ago those defining childhood obesity had opted for %median instead of BMI centiles, what effect would that have on the prevalence of obesity at different ages now? There are no standard thresholds for %median BMI, but an adult BMI of 25 kg/m2 is roughly 20% above and BMI of 30 kg/m2 roughly 40% above the median at ages 18–20 years. It could be argued that it is not appropriate to apply adult obesity thresholds to children, but in the absence of any directly validated thresholds, an explicit link to validated adult levels seems sensible; this is analogous to the approach used to set the IOTF centile-based thresholds. Using these %median thresholds in the same cohort5 (figure 2) gives a very different picture, with very low rates of overweight and obesity in the youngest children, rising progressively to similar proportions as for the centile thresholds in the teens.
The %median approach has its own limitations, as it assumes that, in health, there is no increase in variability of BMI with age, in contrast to centile-based thresholds, which assume that all the increase in variability with age (at the time the data were collected) represented healthy normality. It seems likely that as children go into their teens there may be more variability in BMI, reflecting changes in lean mass, but it also seems highly likely that progressively more individuals become unhealthily obese as they progress into their teens. We have no gold standard measure of childhood obesity, but the true prevalence must lie somewhere between these two estimates. Further work is needed, ideally to explore the extent to which these different approaches predict worse adult health, but this is only possible if suitable data can be found. In the mean time, an exploration of how reliably each detects children who are overfat would be helpful.
Centile charts are vital tools in child health, but when applied to BMI it is likely that they greatly inflate the prevalence of preschool obesity, with the risk that attention is deflected from older children who are at higher risk. We must also recognise the inherent danger of defining healthy limits for teenagers, based on upper BMI thresholds which rise as the obesity epidemic increases. Further attention needs to be given to defining how the true prevalence of obesity varies with age, but it seems likely that more stringent thresholds are required for younger children.
I am grateful to Jill Morrison for first pointing out the clinical implication of this anomaly, to Jonathan Wells and Tim Cole for many discussions about how to investigate it and for Tim’s helpful comments on an earlier draft of this manuscript, as well as the three anonymous reviewers. I am grateful to all those involved with the Gateshead Millennium Study. I am indebted to the families and children who took part, to Ashley Adamson who now leads the GMS cohort, to Kathryn Parkinson for her help in setting the cohort up and for her work on it over the years, to John Reilly and Ann Le Couteur who led relevant survey waves, to Angela Jones and Laura Basterfield for their comments on the drafts, as well as to the External Reference Group, Gateshead Health NHS Foundation Trust, Gateshead Education Authority and local schools.
Contributors CMW set up the GMS cohort, conceived, designed and undertook the analysis, and drafted the manuscript. She is its guarantor.
Funding There was no specific funding for this analysis. The GMS cohort has been supported by a wide range of grants over time, notably the Henry Smith Charity and Sports People Aiding Research for Kids (SPARKS), the Gateshead NHS Trust R&D, Northern and Yorkshire NHS R&D, Northumberland, Tyne and Wear NHS Trust, the National Prevention Research Initiative (incorporating funding from British Heart Foundation; Cancer Research UK; Department of Health; Diabetes UK; Economic and Social Research Council; Food Standards Agency; Medical Research Council; Research and Development Office for the Northern Ireland Health and Social Services; Chief Scientist Office, Scottish Government Health Directorates; Welsh Assembly Government and World Cancer Research Fund), the Children’s Foundation, and the Scottish Government Health Directorates Chief Scientist Office.
Competing interests None declared.
Patient consent for publication Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.