Child BMI percentiles, explained

Why children get percentiles rather than fixed cut-offs

The adult BMI categories work (to the limited extent they do work) because adult body composition is roughly stable. A healthy 30-year-old and a healthy 50-year-old carry broadly similar amounts of fat for their size, so one threshold can serve both. Childhood is nothing like that. Body fatness is high in infancy, falls through the preschool years to a minimum somewhere around age five or six, and then rises steadily through the school years and adolescence. That minimum has a name, the adiposity rebound, and it is visible directly in the growth reference data.

The table below is computed from the CDC BMI-for-age parameters this site ships, so you can see the effect rather than take it on faith. The median BMI for girls drops from 16.4 at age two to 15.2 at age six, then climbs to 21.3 by eighteen. A fixed overweight threshold of 25 would classify essentially no six-year-old as overweight and a great many eighteen-year-olds, for reasons that have nothing whatever to do with health.

Notice the last row. By eighteen, the 85th percentile has arrived at 25.7 for both sexes and the 95th at 30.3 for girls and 29.0 for boys, which is to say, almost exactly the adult overweight and obesity thresholds. That convergence is deliberate. The charts were built so that a child tracking at the 95th percentile arrives at adulthood at roughly the adult obesity line, which lets paediatric and adult classification hand over to one another without a discontinuity.

What the CDC LMS method does

A percentile is a statement about where a measurement sits in a distribution, and to make that statement you need to know the shape of the distribution at that exact age and sex. BMI is not distributed symmetrically in children: the upper tail is much longer than the lower one, because there is a floor on how small a healthy child can be and no comparable ceiling. A method that assumed a normal distribution would place children at the top of the range badly wrong.

The LMS method, set out in the CDC's 2002 methods report by Kuczmarski and colleagues, solves this with three parameters published for every month of age and for each sex. M is the median BMI at that age. S is a measure of relative spread. L is a Box-Cox power that describes and removes the skew. Together they transform an individual BMI onto a standard normal scale, after which a percentile is just the area under the normal curve up to that point.

Worked through for a girl aged exactly eight years, 130 cm and 30 kg: BMI is 30 ÷ 1.30², which is 17.75. The CDC parameters for girls at 96.5 months are L = −2.617, M = 15.827 and S = 0.1172. Applying the transform gives a z-score of 0.85, and the normal distribution puts a z of 0.85 at the 80th percentile. So she sits at the 80th, inside the healthy-weight band, which runs from the 5th to below the 85th, and close enough to its upper boundary to be worth watching at the next visit rather than acting on today.

It is worth understanding what those three parameters mean for you as a reader. Because M is the median, the percentile is always relative to children of the same age and sex in the reference population. Because L and S vary by age, the same BMI difference means different things at different ages: half a BMI point separates the 50th from the 60th percentile at age four, but a much smaller slice of the distribution at fourteen, when the spread is wider.

• z = ((BMI ÷ M)^L − 1) ÷ (L × S) when L is not zero; z = ln(BMI ÷ M) ÷ S when L is zero.
• Percentile = the proportion of the standard normal distribution below that z, multiplied by 100.
• L, M and S are published for every half-month of age from 24 to 240 months, separately for boys and girls.
• Between published ages the parameters are interpolated, which is why a child's exact age in months matters and a rounded age introduces error.
• Above the 95th percentile the z-score becomes an unreliable way to express position, which is why a separate metric exists for severe obesity.

A reference is not a standard

This distinction does more work than any other idea on this page, and it is almost never explained. A growth reference describes how a particular population of children actually grew. A growth standard describes how children grow when the conditions for healthy growth are met. One is descriptive and one is prescriptive, and they answer different questions.

The CDC growth charts are a reference. They were built from United States national survey data collected between the 1960s and 1994, describing how American children did grow over that period. The CDC deliberately excluded the more recent weight data for children over six when constructing the charts, precisely because including the sharp rise in childhood obesity through the 1980s and 1990s would have shifted the reference upward and made a worsening population look normal. Even with that correction, the charts describe a real population with real problems rather than an ideal.

The WHO Child Growth Standards for children under five are a standard. They came from the Multicentre Growth Reference Study, which followed children in six countries who were raised under conditions considered favourable for growth: breastfed, with good nutrition and health care, and with non-smoking mothers. The resulting curves describe how children grow when nothing is holding them back, which is why WHO presents them as a target rather than a description.

The WHO growth reference for children aged five to nineteen is a third thing again, and its name is honest: it is a reference, not a standard. De Onis and colleagues reconstructed it in 2007 by merging the 1977 NCHS/WHO data with the under-fives standard sample and applying the same modern smoothing methods, specifically so that the curves would join the under-five standards at age five and arrive at the adult cut-offs at nineteen. They succeeded: at nineteen the +1 standard deviation value is 25.4 for boys and 25.0 for girls, against an adult overweight threshold of 25.0, and the +2 standard deviation value is 29.7 for both sexes against an adult obesity threshold of 30.0.

The consequence for anyone reading a number is simple and important. A child can sit at the 90th percentile on one chart and the 80th on another, and neither chart is wrong. Always know which chart produced the number you are looking at, and never compare a percentile from one chart against a percentile from another.

The recognised thresholds and what they are called

The CDC publishes four bands for children and teenagers aged two to nineteen, and they are worth learning in their exact form because the boundaries are inclusive at one end and exclusive at the other. Underweight is below the 5th percentile. Healthy weight runs from the 5th percentile to below the 85th. Overweight runs from the 85th to below the 95th. Obesity is the 95th percentile or above.

A fifth category sits above those. Severe obesity is defined as a BMI at or above 120% of the 95th percentile for that age and sex, or a BMI of 35 or more, whichever is lower. That odd-looking definition exists because of a genuine statistical problem: above the 95th percentile the reference distribution has very few observations, so percentiles compress and z-scores become unstable. Two children with meaningfully different BMIs can both be reported at the 99th percentile, which makes it impossible to tell whether treatment is working.

Kelly and colleagues, in a 2013 American Heart Association scientific statement, set out the percent-of-the-95th-percentile approach and the class 2 and class 3 thresholds at 120% and 140% of the 95th percentile. Freedman and colleagues later showed directly why the older approach fails, reporting that BMI z-scores are a poor indicator of adiposity among children with very high BMIs. For our eight-year-old girl above, the 95th percentile BMI is 20.70, so 120% of it is 24.8 and 140% is 29.0.

The WHO system uses different language for the same idea. For children aged five to nineteen it defines overweight as BMI-for-age above +1 standard deviation and obesity as above +2 standard deviations, which correspond to roughly the 84th and 98th percentiles. That is why prevalence figures quoted from WHO sources and from CDC sources for the same population do not match, and why any published rate of childhood obesity should always name its chart.

The extended percentiles, and the problem they solve

The 2000 CDC growth charts stop at the 97th percentile. That was a defensible choice at the time, because there were too few children above it in the survey data to model the curve reliably. It stopped being defensible as severe childhood obesity became more common: by the late 2010s, United States survey data indicated several million children whose BMI sat above the top line on the chart their clinician was using.

The CDC's response was the extended BMI-for-age growth charts, which add percentiles above the original ceiling by drawing on a larger pool of data for children with obesity collected between 1988 and 2016. The statistical work behind them is set out by Wei and colleagues, who published a method for calculating BMI z-scores and percentiles above the 95th percentile of the CDC growth charts. The extended charts do not change the 5th, 85th or 95th percentile lines, so nothing about the standard categories moved.

For a parent or a clinician, the practical gain is that a child well above the 95th percentile now has a position on a chart and a plottable trajectory, instead of simply sitting off the top of the page. That matters for the same reason any measurement matters: without a scale you cannot tell improvement from deterioration. Our child BMI calculator reads the extended parameter files directly, which is why it can report both a percentile and a percent-of-the-95th figure.

Why the trajectory across visits beats any single reading

A single percentile tells you where a child sits today relative to a reference population. It does not tell you whether that position is where they have always sat, where they are heading, or an artefact of the morning they were measured. Children grow in height in uneven spurts, and because BMI divides weight by height squared, a growth spurt can drop a percentile sharply for several months before weight catches up. A child who has simply grown four centimetres since the last visit can look as though something has changed when nothing has.

Measurement error adds its own layer. Height measured with shoes on, or with a child not standing straight, easily moves BMI by half a point. Weight measured after breakfast rather than before does the same. At the steep part of the distribution, half a BMI point can be five or ten percentile points, which is larger than most of the changes anyone is trying to detect.

What carries real information is the line rather than the point. A child who has tracked along the 70th percentile from age four to age nine is growing consistently, and that consistency is itself reassuring regardless of which line it is. A child who has moved from the 50th to the 85th over two years is crossing centiles, and the crossing is the finding, not the 85th itself. The expert committee recommendations that shaped current paediatric practice are built around this idea: BMI-for-age is a screening tool that identifies children for further assessment, and the assessment looks at growth over time, family history, diet, activity, sleep and any clinical signs.

Practically, this means three things. Ask for the plotted chart, not just the number. Use the same setting and method each time, because a school screening and a clinic measurement are not directly comparable. And resist reacting to a single reading: the honest interpretation of one percentile is almost always to measure again in six to twelve months.

How to talk about this with a child without doing harm

This is the part of the topic where getting it wrong has real costs, and the evidence on that is reasonably clear. The American Academy of Pediatrics' 2016 clinical report on preventing obesity and eating disorders in adolescents identifies several practices that raise the risk of disordered eating: dieting, family conversations focused on weight, and weight-based teasing. It recommends the opposite emphasis: on healthy behaviours rather than on weight, on family meals and sleep and activity, and on never making a child's body a topic of household commentary.

The concrete advice that follows is to talk about what the family does, not about what the child is. A conversation about adding a vegetable to dinner, walking to school, or putting screens away an hour before bed is a conversation a child can act on without learning that their body is a problem. A conversation about a percentile is not, and a child cannot do anything useful with the number anyway.

If a percentile has to be discussed (because a clinician has raised it, or because a child has seen it on a school letter), the honest framing is the accurate one. It is a comparison with other children of the same age, not a verdict and not a measure of health. It moves with height as well as weight. Plenty of perfectly healthy children sit high or low on it. And the reason anyone measured it is to decide whether to look more closely, not to decide anything about the child.

Two things are worth avoiding outright. The first is putting a child on a restrictive diet without clinical supervision; paediatric weight management exists as a specialty precisely because growing bodies have nutritional requirements that adult dieting advice ignores. The second is weighing a child at home as a routine. If a measurement is needed, it belongs in a clinical setting where it comes with context, and where the person reporting it can answer the question that always follows.

Frequently asked questions

Put it into practice

Run your own numbers through the child BMI calculator, the BMI calculator and the healthy weight range calculator. Related reading: BMI explained, Pregnancy weight gain and 2025 obesity definition.

Sources

1. Kuczmarski RJ, Ogden CL, Guo SS, et al. 2000 CDC growth charts for the United States: methods and development. Vital Health Stat 11 2002;246:1–190. The LMS method set out in full. www.cdc.gov/nchs/data/series/sr_11/sr11_246.pdf
2. CDC. About child and teen BMI: the weight status categories for ages 2 to 19, including severe obesity at 120% of the 95th percentile. www.cdc.gov/bmi/child-teen-calculator/
3. CDC. Extended BMI-for-age growth charts: the percentiles added above the original 97th percentile ceiling. www.cdc.gov/growthcharts/extended-bmi.htm
4. CDC. BMI-for-age growth chart percentile data files: the L, M and S parameters themselves. www.cdc.gov/growthcharts/percentile_data_files.htm
5. Wei R, Ogden CL, Parsons VL, Freedman DS, Hales CM. A method for calculating BMI z-scores and percentiles above the 95th percentile of the CDC growth charts. Ann Hum Biol 2020;47:514–21. doi.org/10.1080/03014460.2020.1808065
6. de Onis M, Onyango AW, Borghi E, Siyam A, Nishida C, Siekmann J. Development of a WHO growth reference for school-aged children and adolescents. Bull World Health Organ 2007;85:660–7. doi.org/10.2471/BLT.07.043497
7. WHO. Growth reference data for 5–19 years: BMI-for-age. www.who.int/tools/growth-reference-data-for-5to19-years/indicators/bmi-for-age
8. Kelly AS, Barlow SE, Rao G, et al. Severe obesity in children and adolescents: identification, associated health risks, and treatment approaches. An American Heart Association scientific statement. Circulation 2013;128:1689–1712. doi.org/10.1161/CIR.0b013e3182a5cfb3
9. Freedman DS, Butte NF, Taveras EM, et al. BMI z-scores are a poor indicator of adiposity among 2- to 19-year-olds with very high BMIs, NHANES 1999-2000 to 2013-2014. Obesity 2017;25:739–46. doi.org/10.1002/oby.21782
10. Barlow SE. Expert committee recommendations regarding the prevention, assessment, and treatment of child and adolescent overweight and obesity: summary report. Pediatrics 2007;120:S164–92. doi.org/10.1542/peds.2007-2329C
11. Golden NH, Schneider M, Wood C; AAP Committee on Nutrition. Preventing obesity and eating disorders in adolescents. Pediatrics 2016;138:e20161649. doi.org/10.1542/peds.2016-1649