Artillery Row

The myth of childhood obesity

The statistics are unfit for purpose

It is not particularly surprising that childhood obesity rose during the first year of the pandemic. Comfort eating and a lack of physical activity during lockdown are the most likely culprits. Nevertheless, the scale of the increase is striking. According to figures released this week, rates of obesity among 4-5 year olds shot up from 9.9 per cent to 14.4 per cent between 2019/20 and 2020/21 and from 21 per cent to 25.5 per cent among 11 year olds. 

The statistics are so wide of the mark that it is incredible that anyone takes them seriously

This is the first significant rise in child obesity since the start of the millennium. Using the broader measure of “overweight”, 41 per cent of kids in their final year of primary school are now officially too fat. Almost unbelievably, 45 per cent of 11 year old boys are overweight.

It is almost unbelievable because it isn’t true. It might be what the statistics say, but the statistics have always been wrong. There are not just a little bit wrong. They are not merely flawed. They are so wide of the mark that it is incredible that anyone has ever taken them seriously. For years, the mantra of organisations such as the late Public Health England has been that one in five children start primary school overweight and this rises to one in three by the time they leave. As any parent who picks their child up from school knows, this is a lie.

Every year, countless parents across Britain receive a letter from school informing them that their perfectly slim child is “very overweight”. Most of them assume there has been a mistake and throw the letter in the bin. A few of them get so angry that they go to the media with photographic evidence to prove that their offspring are of a “healthy weight”. But nothing ever changes. The letters continue to be sent. The statistics continue to be parroted.   

Listen carefully, however, and you can hear the first stirrings of discontent. When the National Food Strategy was published in July, its author Henry Dimbleby added an appendix to explain why it did not include any child obesity statistics. It was, he wrote, because the way child obesity is measured in the UK is “problematic” and “troublesome enough to require rethinking”. When the BBC reported the new figures this week, its head of statistics, Robert Cuffe, said “the number of obese kids may be lower in truth than estimated here” and “it’s probably not as bad as it looks”. 

These are relatively modest criticisms, but they are a start. Let me be blunter.

Obesity in adults is easy enough to measure. Body Mass Index (BMI) is weight in kilograms divided by the square of height in metres. A BMI of 30 or more makes you obese. In theory, the cut-off of 30 is used because this is the point at which being fat increases your risk of premature death. A BMI of 25 or more makes you overweight, but this doesn’t have any clinical significance. It is just a round number.

Kids are not shaped like adults, they do not have the same fat/muscle ratio and they are constantly growing

There are well known problems with BMI, not least the fact that it does not distinguish between muscle and fat. It is excess body fat that we are interested in and this is best diagnosed by clinical examination, but when that is not possible, the BMI system correctly identifies obesity around 80 per cent of time. It’s not perfect, but it’s good enough for a national prevalence survey.

But it doesn’t work with children. Kids are not shaped like adults, they do not have the same fat/muscle ratio and they are constantly growing. Pre-teen children hardly ever have have a BMI over 25 and almost never have a BMI over 30. Conversely, an obese child can easily have a BMI of less than 25.

To make up for this, clinicians use a chart like the one below which gives bespoke, age-specific and gender-specific BMI cut-offs for children. For example, at the age of six and a half, boys are considered obese if their BMI exceeds 20.2. By the time the boy is eleven, the cut-off has risen to 25.1.


To see how these thresholds were derived, we need to look at the work of Professor Tim Cole and his colleagues. In 1995, they published the study upon which the chart above is based. They looked at the BMIs of children at different ages and divided them into percentiles. The data used was taken from between 1978 and 1990, before the rise in childhood obesity got underway, and it provided a reference point from which future changes in obesity could be measured. If two per cent of 11 year old boys were obese in the 1980s and they had a BMI of at least 26, then the threshold for obesity for an 11 year old boy was 26. To measure child obesity going forward, all you had to do was work out how many 11 year old boys would have been in the top two per cent in the 1980s. You then do the same for girls and then for 12 year olds, 13 year olds and so on.

It sounds simple and it had the virtue of being convenient. The system made it possible to estimate child obesity rates nationwide without clinicians having to examine anybody. Detailed figures based on this system have been collected by the UK government since 1995, all using Cole’s reference curves based on what we think we know about the distribution of BMI between 1978 and 1990, some of which comes from tailors’ records of school uniform sizes.

Cole et al. assumed similar obesity rates among both twenty year olds and four year olds

The problem is that we don’t actually know how many children were obese in the 1980s because nobody was paying the issue much attention. Cole assumed that the rate of child obesity was two per cent but this was only inferred from the rate of obesity among young adults. Common sense dictates that the child obesity and adult obesity figures should “link up”, which is to say that both systems should produce similar estimates for young adults. It would be odd if 17 year olds had a rate of obesity that was much different to that of 18 year olds, especially since BMI tends to rise with age.

Cole et al. noticed that 20 years olds in 1990 who had a BMI of 29 (and were therefore nearly obese) appeared at the 98th percentile, which meant the rate of obesity among this age group was just under two per cent. They also noticed that 20 year olds at the 99.6th percentile (the top 0.4 per cent) had BMIs of at least 32.8. They therefore concluded that “These centiles seem to be reasonable definitions of child obesity and superobesity respectively.”

This assumption was highly debatable. There is a big difference between twenty year olds and four year olds and yet Cole et al. assumed similar obesity rates among both. Nevertheless, they came to a similar conclusion when they published further research in 2000. Looking at BMIs in Britain between 1978 and 1993 (“predating the recent increase in prevalence of obesity”), they found that obese 18 year olds were at the 99th percentile. In other words, only one per cent of those who had recently become adults were obese by the usual definition.

They found similarly low rates of obesity for 18 year olds in Brazil and Singapore over the same time period. The Netherlands had an even lower rate of 0.3 per cent while the USA had a much higher rate of 3-4 per cent.

If the rate of obesity among 18 year olds in the 1980s was one per cent, it was most unlikely that the rate among children would have been higher. In a subsequent study, Cole and Lobstein concluded that “the obesity cut-off [for children] is well above the 98th centile”. It was nevertheless decided that the 98th percentile would be used as the threshold for child obesity in the growth charts. Perhaps they saw this as erring on the side of caution. Whatever the reason, it was always going to exaggerate the scale of child obesity. 

They used sketchy data from the 1980s and tried to reverse engineer statistics

If this strikes you as an unnecessarily tortuous way to work out the BMI of a fat child, I do not disagree. Remarkably, no one seems to have proposed the more obvious method of getting a bunch of clinicians to physically examine a group of children and take a note of the BMIs of those they considered obese. Instead they used sketchy data from the 1980s and tried to reverse engineer statistics derived from young adults onto children as young as two.

Whatever its merits, the 98th percentile is widely used by clinicians to diagnose childhood obesity, using the chart above. The 98th percentile is also the internationally recognised threshold for child obesity. To be clear, that is the 98th percentile of each age group in the 1980s. Because the child obesity rate was almost certainly less than two per cent in the 1980s, using this threshold creates many false positives. We can see this in those strange letters received by angry and bemused parents. They are automatically sent out based on nothing more than a record of the height and weight of the child. If they would have been in the top two per cent in the 1980s, the parent is told that they are “very overweight”.  

But it gets worse. When we measure childhood obesity in the UK at the national level, we drop the threshold to the 95th percentile. Why? Nobody seems to know. There is no justification for it in the scientific literature other than that it is “the convention”. Cole himself has said that the methodology is “all built on sand”.

The most likely explanation for dropping the cut-off to the 95th percentile – if we exclude the possibility that it was deliberately intended to exaggerate the scale of the problem – is that the USA did it first and we copied them. By the end of the 1990s, the USA had started using the 95th percentile as the cut-off for “overweight” among children, with the 85th percentile used to define “at risk of overweight” – terms that would later be changed to “obese” and “overweight”. This was not wholly unreasonable. The rise of obesity in America began earlier than it did in Britain and rates of obesity have always been higher. It is likely that around five per cent of American children were at least overweight, if not obese, by the end of the 1980s and would therefore have been above the 95th percentile.

But Britain is not America. Cole’s figures showed that the obesity rate among 18 year olds in Britain was much lower than it was in the USA – at around one per cent – and while he recognised the need for a cut-off, he asked the obvious question:

… why base it on data from the United States, and why use the 85th or 95th centile? Other countries are unlikely to base a cut off point solely on American data, and the 85th or 95th centile is intrinsically no more valid than the 90th, 91st, 97th, or 98th centile.

Whatever the reason for using the 95th percentile, it has had the effect of massively inflating childhood obesity figures in Britain for as long as they have been recorded. It implicitly assumes that five per cent of 18 year olds were obese in the 1980s when we know that the real figure was less than half that and perhaps less than a quarter of it. Starting from this false premise, everything that follows from it is wrong. We are classifying huge numbers of children as obese who would not have been diagnosed as such by a doctor in the 1980s and would not be diagnosed as such today. If you look at the chart above, you will see that the 95th percentile is not even marked on it. It is of no clinical relevance whatsoever.

As a result, we get figures which defy credibility. Amongst the absurdities is the “fact” that obesity among older children is much higher than among young adults. Since rates of obesity rise with age, we should see a gradual increase from one age group to the next. Instead we see a strange drop in prevalence as soon as kids leave school which implies a near-miraculous loss of weight. This is particularly remarkable since we are told that “four out of five obese schoolchildren will remain dangerously overweight for the rest of their lives”.

If you take Britain’s obesity figures at face value, we now have a higher rate of childhood obesity than the USA. This is almost certainly untrue. The USA has a much higher rate of adult obesity than Britain (42 per cent vs. 28 per cent). It is almost inconceivable that its rate of child obesity would be lower – you only have to visit the place to see that – but this is what happens when you apply the same relative measure to countries which start from very different places.

A few years ago I gave evidence to an Australian select committee on childhood obesity. I was amazed to hear that their rate of child obesity was just seven per cent. On paper, this is two-thirds lower than Britain’s, despite Australia having a slightly higher adult obesity rate. But once you understand that the method of counting obese children comes from the think-of-a-number school of statistics, it starts to make sense.

As these examples illustrate, using the 95th percentile to define childhood obesity makes it impossible to make meaningful comparisons between different countries. Campaigners who are unaware of this, or who simply don’t care about the facts, breathlessly claim that London has a higher rate of child obesity than New York. It probably doesn’t, but we don’t have the data to prove it either way.

This is not to deny that the average BMI of a child, like the average BMI of an adult, has risen over the years. Whatever the childhood “obesity” statistics are measuring has gone up, but they are not measuring obesity by any reasonable definition. The statistics for “overweight” children are even more risible. They are based on the 85th percentile for no reason whatsoever. It is beyond arbitrary. It is not even guesswork. 

If you met the children, you would often agree with their parents that they are not fat

When parents have been asked about their supposedly obese children, only 52 per cent of them say that they are “too heavy”. When it comes to children who are “overweight”, only 11 per cent of parents think they are “too heavy”. The children themselves agree. Only 51 per cent of “obese” children and 17 per cent of “overweight” children think that they are too heavy. This is further evidence that the statistics are not aligned with reality, but rather than reassess their methodology, those who collate the data have simply dropped these questions from the Health Survey for England. 

Meanwhile, public health campaigners who perpetuate the myth of a childhood obesity “epidemic” accuse families of deluding themselves. Parents who reject the diagnosis of an arbitrary and patently flawed mathematical assumption are accused of being “no longer able to tell whether their children are overweight”. When a study found that 94 per cent of parents considered the weight of their technically overweight children to be “about right”, the researchers accused them of suffering from “Goldilocks Syndrome”.

Never mind Goldilocks. The Emperor’s New Clothes is the more appropriate fairy tale analogy here. We are told that there is a childhood obesity epidemic. We are told that more than 40 per cent of 11 year olds are overweight and one in four are clinically obese. We go along with this story despite seeing very few obese children with our own eyes. A doctor who examined the “obese” children would reject the diagnosis in most cases. If you met the children, you would often agree with their parents that they are not fat.

There was never any justification for the British government measuring childhood obesity in the way it does. The system used by our statistics agencies is not fit for purpose – unless the purpose is to vastly inflate the scale of the problem. To say that it is “troublesome enough to require rethinking”, as Mr Dimbleby does, is an understatement. It is scandalous that it has survived so long.

Enjoying The Critic online? It's even better in print

Try five issues of Britain’s newest magazine for £10

Critic magazine cover