Tackling the Obesity Crisis
Analyses of the “Growing up in Scotland (GUS)” child cohort to inform the design of obesity/overweight surveillance systems internationally
The theme of this year’s World Obesity Day is #EveryBodyNeedsEverybody.
If 2020 has highlighted anything, it is that many aspects of our health and wellbeing are not only determined by our own actions, but are dramatically shaped by the society we live in. That is why it is critically important that we continue to work to increase our knowledge of obesity and its root causes in order to drive improved support, policies and actions that will improve our collective health.
Childhood overweight and obesity estimates have shown steady growth over the last three decades, with figures from UNICEF’s State of the World’s Children report showing that the obesity epidemic is on the rise almost everywhere. The proportion of children 5-19 years old who are overweight or obese has nearly doubled in the past 15 years, to 18% globally.
Note: Throughout this blog we have used the standard terminology -- “overweight” -- for children whose weight, for their height, is at a worrisomely high level, but who are not yet “obese” – the level at which significant health, psychological and social effects can be expected. The cutoffs for defining these levels of weight-for-health vary somewhat from country to country, as explained in detail in the UNICEF report linked above.
In Scotland, almost a quarter of children are starting school overweight or obese, with the poorest children at over twice the risk (13.3% versus 6.4%) of the wealthiest of being obese. With the influx of cheap and convenient (but very caloric – i.e. energy-dense) processed foods in our supermarkets, what was once thought of as a problem for the wealthy has steadily become more of an issue for more disadvantaged families, in that all Scottish children had a similar rate of obesity two decades ago.
Studies suggest that children who are obese are five times more likely to become obese as an adult. In addition, it has been shown that once obesity is established, very few children and their families will complete the intensive treatment programmes required to reverse it, even when these are readily available at no cost (which is generally not true in much of the UK). It is therefore arguable that it would be beneficial to predict child obesity prior to it becoming well-established, enabling health professionals to offer earlier and less-intensive treatment that may be more likely to succeed.
Despite the gravity of the obesity epidemic, many countries have yet to develop a successful method by which to regularly monitor the height and weight of a large national sample of children at any age. Currently statistics are based on a combination of national health surveys, such as the Demographic Health Survey (DHS) and Multiple Indicator Cluster Survey (MICS) in low-income countries, although the former only examines children under five. To help tackle the obesity crisis there is a need to develop an effective, yet inexpensive, routine measurement system that can flag indicators required to prevent obesity, protect children from malnutrition and promote healthy lifestyles. A pre-requisite for such screening is a validated predictive algorithm based on universally available (routinely collected), machine-readable and linkable predictor variables, which has been derived from prospective cohort studies with high-quality predictor and outcome data.
Article 24 of the UN Convention on the Rights of the Child states that every child has the right to the best standard of health, including nutritious food. If a child is overweight or obese, it can lead to numerous long-term and short-term health issues, such as early onset type-2 diabetes and coronary heart disease. In Scotland, being overweight or obese is the norm in adults (five-eighths are affected), resulting in health and socio-economic impacts on a significant scale. If we can find better ways to identify predictive indicators of future obesity in children and enable early or preventative interventions, we can encourage and support a new generation to make healthier choices and inform policy more effectively. Scotland’s target is to halve cases of childhood obesity by 2030, so better methods of regular monitoring, assessment, and early intervention will be vital in helping to achieve this goal.
Our project:
Our project team, led by Professor John Frank from the Usher Institute at the University of Edinburgh, has been working with the Growing Up in Scotland (GUS) cohort data to develop predictive algorithms for identifying children at around age 5-6 – when all students entering first grade (P1) have their heights and weights measured -- who are at high risk of obesity at age 12, utilising combinations of published life-course predictors of obesity, for which data are routinely-collected and machine-readable at the population level, at least in Scotland.
The GUS dataset is a long-term monitoring project, which has interviewed a population-based sample of over 3,000 parents and their children on nine occasions. A wide range of information has been collected about the children involved in the study since they were 10 months old in 2004/5, as well as their families, including physical and mental health, home and family life and education.
The aim has been to examine the feasibility of accurate prediction of obesity at age 12 through population-wide screening of all children at half that age, for referral to intensive child and/or family preventative treatment, well before obesity becomes firmly established and thus harder to treat successfully.
In addition, the team have been working to identify novel risk factors, related to Adverse and Protective Childhood Experiences (ACEs/PCEs), as well as markers of family deprivation, that also independently predict obesity at age 12. A secondary study objective has been to test whether any subset of these novel risk factors for/predictors of obesity at age 12 can also identify children whose parents’ perception of their excess weight-for-height, is at odds with reality.
In effect, this study is examining the feasibility for using routinely collected, linkable data in universal screening of Scottish children at age 5-6 (school entry) to predict their risk of obesity at age 12, so that earlier and less-intensive intervention can be offered to those at risk, to prevent the development of fully established obesity.
The full results of this study will be available via journal publication in the near future.
For those wanting more details on the pros and cons of screening, we recommend the excellent book by Angela Raffle and Muir Grey: Screening: Evidence and Practice. 2019: Oxford University Press, USA
Some considerations about the benefits and risks of screening for future obesity risk:
There is room in the public health battle against obesity for both targeting individuals (selecting candidates for intervention based on particular risk factors) and universal policies (aiming to reduce obesity risks in all individuals within a given population, such as taxing high-energy foods). Whilst interventions targeted towards individuals may be more appropriate to each person’s specific risk, they also run the risk of increasing stigma and harmful labelling – especially if the screening test used to detect risk gives a false negative or false positive result. However, children are not a homogenous group, and their social environment (especially the family) plays an integral part in their health and lifestyle. Universal interventions, therefore, may not effectively identify those most at need of help and support. They also tend to be politically challenging for policy makers, as public support for measures such as taxing high-energy foods is far from universal.
All screening programmes have a “cost” and the potential to do net harm to some participants (so-called “false positives” and “false negatives,” arising from inevitable errors in the risk-prediction process – see below.)
In this case, the main cost of such screening is mostly made up of the “referral burden” – the proportion of children/families who are found at age 5-6 screening to be at high risk, who would then need to be referred to intensive dietary and physical activity intervention, in an effort to prevent obesity at age 12. The potential benefits of such a screening programme, if those benefits are robustly established by further evaluative studies, would be the detection of obesity risk at an early stage, before obesity fully manifests, leading to improved treatment success with less-intensive interventions. This would not only benefit the individual; it could also have positive effects for onward economic growth and prosperity through the enhancement of a child’s future productivity, earnings and life expectancy.
The challenge would be to balance the benefits of providing such intensive treatment to the significant proportion of children predicted to become obese a half-decade later, against two sorts of potential harms, arising from the fact that no predictive models are ever 100% accurate:
Those affecting false negatives (children not identified as at risk at age 5-6) who then actually go on to develop obesity at age 12: such children and their families could experience “false reassurance,” rendering them less likely to pay attention to the child’s weight, and take appropriate action, in future years;
Those affecting false positives (children identified as at risk, who would not have gone on to develop obesity by age 12): such children and their families would experience the stress, costs and disruption of being referred for intensive treatment for a risk the child does not have – leading not only to waste of scarce NHS treatment resources, but also unnecessary medicalisation of normal children’s lives.
In other words, it is important to account for the potential errors in predictive models in a screening programme, including the consequences of both false-positive and false-negative predictions to an individual, the family, the NHS, and society as a whole. The ideal model would have 100% sensitivity (no false negatives), the ability to correctly detect all those destined to experience obesity at age 12, and 100% specificity (no false positives), the ability to correctly identify all those who will not develop obesity at age 12. However, this is considered to be unrealistic – all risk-prediction models with many predictors have multiple possible risk-cutoffs, each of which has a different sensitivity and specificity.
A model with high sensitivity, but low specificity, would identify nearly all children who will actually develop obesity (i.e. few false negatives). However, this would mean that many children would be mistakenly identified as being at-risk when they were not (false positives). A model with low sensitivity and high specificity will have a lower false-positive rate but would fail to predict obesity in a larger number of children who will in fact develop it. [Somewhat mitigating this potential for harm is the fact that virtually no children are currently receiving preventive obesity treatment in the UK now – treatment resources are far too stretched by the need to attend to children with established obesity.]
In sum, then, the feasibility of using routinely collected data at age 5-6 to predict obesity at age 12, in the interests of early referral with better outcomes and less intensive treatment, is entirely dependent on the accuracy (sensitivity and specificity) of the optimal multi-predictor model which can be created using data from a large number of children, followed up for the half-decade between these two ages, with weight-for-height measurements taken again at age 12. The GUS Cohort in Scotland is one of the largest public datasets in the UK capable of supporting such an analysis, for children born since 2000.
We are incredibly excited by the work that Professor John Frank and his team have completed on the Growing Up in Scotland cohort of data and hope to be able to share the official publications and findings with you in due course.