Weight centile growth charts: why they can’t predict your child’s recovery weight

Are you wanting a target weight, a prediction of the healthy weight your child needs for recovery from their eating disorder?

I'm going to show you how one of the better methods — the use of a growth chart — is still inaccurate — to the point it could mislead. I'll explain how these percentile charts for weight, height or BMI, are statistical compilations. They don't show the considerable variability from one young person to another.

In real life, it's normal for a healthy child or teen to deviate from those smooth percentile curves. They'll have periods of being above, or below. By a bit, or by a lot. They may shoot up and away from a centile curve and never return to it.

The only certainty we have, for almost every youngster, is that if a child lost weight, they must, as a minimum, regain it. How much do they need above that? This is where we hope growth charts will help, but they give statistical rather than individual predictions.

I plan to end with this message you probably don't want to hear: you'll only know your child is weight-recovered when you've got there. Any prediction may be seriously wrong.

Don't let your child get attached to a weight goal that may well turn out to be insufficient. For you, as a parent, there's always a risk you will underestimate quite how much weight your child needs to gain. If you are told a target weight, use that as a minimum ballpark to get you fired up for the refeeding work.

Why are percentile growth charts used?

First, why are we even talking about weight recovery? I explain this on ‘Weight-restoration: why and how much weight gain?’

Next, I explain how a 'one-size-fits-all' weight target is not fit for purpose, Just like shoe size, any estimation of healthy weight must be individualised. If you've been given a goal weight based on 'one hundred percent weight for height' or "median BMI', read my post 'Is your child's target weight a gift to the eating disorder?' or see my YouTube: 'What is a BMI or '% Weight-for-Height' target, and how wrong it could be?'

There's a common message (which I have helped to spread) that percentile growth charts give you that individualized goal weight prediction. Now I realise that's not true. Let me explain.

'A child should follow their percentile growth curve': no

What most of us hear is that each healthy young person person will naturally track along their percentile curves, for height and for weight (some track BMI), throughout their growth years. We hear that percentile charts are therefore a handy tool for predicting healthy weight if a child drops off their curve.

I've been in this field a long time and I've heard this so often that I took it as medical gospel. How nice that growth can be so ordered and predictable! Now I realise that is not so, and I'm annoyed that I repeated the message in my resources, with just a brief warning about variability.

Those percentile curves do not track the speed of growth of a child. They show a statistical compilation of heights and weights, taken from a bunch of kids of different ages. They are tools for whole populations, and were not intended to be used for an individual who might have periods of slower growth, or a growth spurt.

How growth charts are commonly used

I'll illustrate a typical use of a growth chart with a girl — I'll call her Robogirl because she's too good to be true. Year after year, Robogirl tracks ever so nicely on the 75th percentile for weight and the 50th percentile for height. (I explain percentiles here, but briefly: for any age, if you line up 100 girls by order of weight, the weight of the 75th girl along is the 75th percentile).

Now here's the theory, and in a minute I'll show you how it's flawed. The theory is that if Robogirl deviated from her usual percentile curves, that would be abnormal. We would aim to get her back to her usual curve for weight and for height.

So, reading off the chart, at age 16 we'd expect Robogirl to weight 61 kg. (Note that with the awful '100 percent weight-for-height' method, where we expect Robogirl's BMI to be exactly average, that figure would be 7 kg less. How's that for accuracy!)

So if Robogirl had anorexia age 15 we'd be aiming to get her up to 61kg by her 16th birthday.

But… in real life, healthy youngsters do not track so very nicely. Here's an example where a growth spurt introduces variability.

Alert: growth charts don't show inevitable growth spurts

Below is an example I made up, using some data from a growth velocity (growth speed) study tracking individuals, rather than groups.

Height-wise, Amanda was always very average: she tracked on the 50th percentile for height (I made this artificially neat and tidy on my plot). Around age 10, though, many of her peers were taller. Then around age 12 she started shooting up. By age 15 she was back on the 50th for height.

This kind of growth spurt is not only normal, it's to be expected. For more on growth spurts, read my post here.

How come the usual growth charts don't show growth spurts?

Why are those red percentile lines not all shaped like Amanda's plot? Why don't they show growth spurts? Because other kids have their growth spurts at other ages. So while, at age 12, Amanda is shooting up the league tables, other kids are having a slow spell. It all averages out.

The red or blue (for boys) percentile chart represents the data from a snapshot in time of groups of boys and girls of all ages. It's 'cross-sectional' data.

Imagine collecting the height and weight of all the 12 year old girls in a school. Then you do the same for the 13-year olds. And so on. You plot the percentiles for each age group and draw nice smooth curves to link all your dots. That's a 'cross-sectional' survey, and it's what you see in those percentile charts.

What you've not done is follow each 12-year old girl individually to see where each one will be age 13, then age 14, and so on. That would be a 'longitudinal' survey. That's where you see how real growth goes in wiggly ups and downs — sometimes in quite big bumps.

Some youngsters deviate massively from those percentile charts

Some healthy youngsters have times of being a bit below — or a bit above — their usual percentile curve.

And some deviate hugely. Here's how it might look:

This is almost a real example. The girl did have a childhood of tracking pretty nicely along centile curves. And she did end up way higher, as in the chart above. For simplicity, I made up the bit in between, because in real life she shot down while she had anorexia.

A shoe size analogy

In case you need to get your head round this, here's an analogy. It's handy for a shoe manufacturer to know typical shoe sizes for various age groups. Shoes aimed at 11 year old girls are full of pink hearts. Shoes aimed at 16-year olds definitely don't.

Imagine your local shop has a note of the shoes your child has bought over the last few years. Can they predict what size will fit your child today? No. Indeed when I used to take my daughter in for new shoes, I remember at times being surprised how much the size had or had not changed.

A weather prediction analogy

What will predict tomorrow's weather better: data from snapshots of the weather over the last 50 years, or observations of what's happening with weather fronts right now?

Why are we even looking at percentile charts?

Growth charts with those nice percentile curves are presented on many reputable websites as tools for clinical work. There, you'll see underlying message that these are good tools to review the growth of an individual child. It takes time to reflect on what is behind those charts, and how they are designed for population studies, not for individuals.

"The longitudinal-type standards are those which should be used in following the individual child, whether in the paediatric or adolescent clinic or as a routine monitoring procedure in healthy children. The cross-sectional values are the appropriate ones to use in making comparisons between population groups each studied only in cross-sectional surveys."

Tanner et al 1976

Assess the whole person

Let's not throw the baby out with the bathwater. A severe drop off a percentile curve, or a low BMI-for-age, or tracking along a very low weight centile — these things should alert parents and clinicians. Is this normal for this child? What else is going on? What is this child's physical and mental state, and how has that been changing?

Perhaps we should think of BMIs and growth charts as canaries in a coal mine. They make you take notice. But they don't provide an accurate picture:

"For adolescents and young adults, setting the individualized target weight should include assessment of the patient’s premorbid height, weight, and BMI percentiles; menstrual history (in adolescents with secondary amenorrhea); and current pubertal stage ."

(Golden et al. 2015a)

The mention of 'pubertal stage' above is because a young person's growth potential, and expectations for their weight, depends on how far they are along their puberty. There's a good article explaining this by Dr Julie O'Toole in 'Determining ideal body weight'.

The American Psychiatric Association (APA) guideline recommends that a lot more than a growth chart is used to estimate a target weight as part of the initial treatment plan. They also point out that when you're using a growth chart you need some historical data for your child ('longitudinal') because it's misleading to try and extrapolate from one height and weight ('cross-sectional'):

"Growth curves should be followed and are most useful when longitudinal data are available, given that extrapolations from cross-sectional data at one point in time can be misleading. Bone age may be accurately estimated from wrist X-rays and nomograms. In conjunction with bone measurements, mid-parental heights, assessments of skeletal frame, and Centers for Disease Control and Prevention growth charts (available at www.cdc.gov/growthcharts) may be used to accurately estimate individually appropriate ranges for “expected” weights for current age."

The American Psychiatric Association (APA) Practice Guideline 2023 page 33

I find it interesting that the guideline uses the expression "accurately estimate". The two words kind of clash. The same section does indicate that initial estimate will need regular reassessment:

"In adolescents, target weight will be adjusted upward to correspond to increases in the patient’s height, and it can be helpful to discuss this
with them from the initiation of treatment. During a period of growth, the target weight should be reassessed every 3–6 months."

So where does that leave us?

It's frustrating to not have a precise, reliable target weight for our child. We would like to know that at X kilos, they'll be in a good state physically and mentally. We'd like to know their brain is not starved any more, and their body doesn't behave as if in a famine. We hope that after those X kilos, we can start making progress with normal behaviours and normal thoughts and beliefs.

That's why so many of us parents, and so many clinicians, cling on to target weights, to predictions. The truth is, for your individual child, we just don't know. Human beings are just too variable.

So the message is: keep doing the work of recovery, keep assessing your child's physical and mental state with your treatment team, and try to make peace with uncertainty.

---

For more on related topics:

• My main post on all aspects of weight-restoration, with many quotes from professionals: ‘Weight-restoration: why and how much weight gain?’
• More from me on growth spurts: Weight gain in growth spurts, and why it matters in eating disorder treatment
• I gather numerous quotes from esteemed researchers, clinicians and professional organisations, all stating that goal weight should be based on your individual's child's history as well as their current needs, in Experts say, "Recovery weight should be individualized" Yet middle-size-fits-all targets are still common
• My YouTube: ‘Growth charts and goal weights made simple’ shows you how to use your child's historical height or weight data… but now you've read this page you know how inaccurate that still is.
• More on how very inaccurate a ‘one-size-fits-all’ target can be, in my post 'Is your child's target weight a gift to the eating disorder?' Includes an infographic showing huge discrepancies between two methods. You'll learn about the 'one hundred weight-for-height' / Median BMI method, and I show you how growth charts are plotted.
• If you are confused about BMI and weight-for-height measures, you're not alone. I discovered that quite a few clinicians don't quite know what's what either. I make it simple in a YouTube: 'What is a BMI or '% Weight-for-Height' target, and how wrong it could be?' and a corresponding article ‘What do BMI and Weight-For-Height mean?'
• If your child has failed to gain weight with age, or has lost weight, that's weight suppression. When you suspect your child has an eating disorder, make sure your clinicians aren't only looking at their current weight. Here I've gathered evidence that the amount and speed of weight loss matter at least as much: How much weight did your child lose? Weight suppression is critical in eating disorder diagnosis and treatment
• Atypical anorexia diagnosis? Handle with care! What's 'atypical anorexia', the issue with weight being declared 'normal' or high, the risk from malnourishment and the need to regain lost weight.