Baby Growth, and What Crossing Centiles Actually Means

Prescriptive and descriptive are not the same job

Almost every growth chart in history has been descriptive: measure a lot of children, plot where they landed, and call the middle line the median. That tells you what was typical, which is not the same as what is healthy. The WHO Multicentre Growth Reference Study set out to do something different. It recruited infants from six countries whose mothers did not smoke, who had access to adequate nutrition and health care, and who were breastfed under recommended conditions, and it excluded measurements from children whose growth was clearly compromised or excessive. The resulting tables are a standard rather than a reference: they describe how infants grow when the conditions are right, which makes them a target rather than a mirror.

The CDC 2000 charts were built the other way, from US national survey data collected across decades in which most infants were formula-fed from early on. That is not a criticism of them — a descriptive reference is exactly the right instrument for many questions. But it means the two charts answer different questions, and an infant plotted on one cannot be compared with an infant plotted on the other. Health services made a choice here: the WHO standard is used for infants and young children in the United Kingdom, most of Europe and, since 2010, in the United States for children under two.

How big is the gap, exactly

You can measure it directly, because both agencies publish their tables. Take a boy who sits precisely on the WHO median weight at every age and ask what percentile that weight is on the CDC chart. At birth the WHO median is 6.5 ounces lighter than the CDC median, so he reads the 37th centile. By two months he has crossed over: the WHO median is now heavier, and he reads the 66th. He peaks at the 67th at three months, comes back through the 52nd at six months as the curves cross again, and then falls steadily — the 36th at nine months, the 27th at twelve, the 26th at eighteen. At twelve months the WHO median is 1 pound 7 ounces below the CDC median.

Read that sequence again, because it is the whole point. Between three and twelve months, a baby who has never once deviated from a single chart line appears to fall forty percentile points on the other chart. Nothing happened. The two curves simply have different shapes across the first year, and the difference is largest exactly where parents look hardest. The consequence for cut-offs is even sharper: the WHO lower cut-off, two standard deviations below the median, corresponds to about the 4th CDC centile at birth, the 8th at three months, the 3rd at six months and the 0.4th at twelve. In the second half of the first year a baby has to be far lighter to be flagged by the WHO standard than by the CDC reference — which is the arithmetic behind the published finding that the WHO standard flags fewer children for poor weight gain and more for excess weight.

Why formula-fed babies drift the other way

The reason the WHO curve bends downward relative to a formula-era reference is that breastfed and formula-fed infants have genuinely different weight trajectories, and the difference appears in a specific window. Dewey and colleagues followed both groups from birth to eighteen months in the DARLING study, published in Pediatrics in 1992. Weight and lean-mass gains were lower in the breastfed group from three to nine months; the mean weight of breastfed infants dropped below the reference median from around six to eight months and stayed significantly below the formula-fed group from six to eighteen months. Length and head circumference were similar in the two groups. What differs is weight, and mostly in the second half of the first year.

Put the two facts together and the practical consequence follows. A breastfed baby plotted on a descriptive, formula-era reference will look as though they are slowing down from around six months, when they are following exactly the pattern the WHO standard was built to describe. A formula-fed baby plotted on the WHO standard will tend to drift upward across the same window, for the mirror-image reason. Neither drift is a verdict on the feeding method, and neither is a reason for a parent to change anything. The point is narrower and more useful: the direction of an apparent drift in the second half of the first year depends on which chart is under the pencil, so the chart has to be named before the drift can be interpreted.

The first days: how much weight loss is expected

Newborns lose weight in the first days, and this is not faltering growth. It is mostly a fluid adjustment as the body sheds the extra water it was born with, and NICE's guidance on faltering growth is explicit that weight loss in the early days is a separate phenomenon from the growth question. The published expectations are that the loss usually stops after about three or four days, and that most infants have returned to their birth weight by three weeks of age. The threshold that triggers action is a loss of more than 10% of birth weight: at that point the guidance asks for a clinical assessment looking for dehydration or illness, a detailed feeding history, and a direct observation of a feed by someone trained to do it.

It is worth knowing how common that 10% is, because the number sounds alarming and the distribution is not. Flaherman and colleagues built weight-loss nomograms from 161 471 term singleton newborns in Pediatrics in 2015, and found that almost 5% of vaginally delivered exclusively breastfed newborns, and more than 10% of those delivered by caesarean, had lost 10% or more of their birth weight by 48 hours. By 72 hours, more than 25% of caesarean-delivered newborns had. So crossing the line is not rare, particularly after a caesarean, and crossing it triggers a look rather than a diagnosis. In practical terms, on a 7 lb 12 oz baby, 10% is 12.4 ounces.

Crossing centiles: what the thresholds actually say

"Crossing centiles" needs a unit before it means anything, and the unit is the centile space — the gap between two adjacent printed lines on the chart. On the UK-WHO charts there are nine of them (the 0.4th, 2nd, 9th, 25th, 50th, 75th, 91st, 98th and 99.6th), and they are spaced two-thirds of a standard deviation apart, which is why the numbers look irregular. One centile space is therefore a fixed statistical distance, not a fixed number of percentile points: near the median it is about twenty-five percentile points wide, and out at the edges it is a fraction of one.

With that unit in hand, the NICE thresholds for concern about faltering growth are readable. A fall across one or more weight centile spaces matters if birth weight was below the 9th centile; two or more if birth weight was between the 9th and the 91st; three or more if it was above the 91st. A current weight below the 2nd centile is a concern whatever the birth weight was. The logic is that a baby who started small has less room to fall before the fall means something, and a baby who started very large is expected to drift down somewhat as they regress toward their own genetic channel. Note also what these thresholds are not: they are triggers for assessment, phrased by the guideline as reasons to look, not as diagnoses.

Two practical cautions about the data going into all of this. A single weight is noisier than it looks — scales differ, a full nappy is worth a surprising amount, and a baby weighed dressed on one visit and undressed on the next has apparently gained or lost for no reason at all. And a fall judged over two weighing intervals is worth more than one judged between two points, which is why guidance talks about a sustained drop rather than a single reading. If you are reading a chart at home, the honest summary is that the shape over months is informative and any single dot is not.

What to do with a number that worries you

Take it to the person who weighs your baby. That is not a polite deflection; it is where the missing information is. They have the whole series rather than one point, they know whether the last two weights were taken on the same scales, they can watch a feed, and they can examine a baby who a chart cannot see. Nothing on this page — and nothing any calculator returns — can distinguish a healthy baby settling into their own channel from one who needs help, because that distinction is made from the shape of the curve plus the child in front of you.

And if you take one thing away, let it be the thing this whole article is built on: check which chart you are looking at before you read anything into a movement. A WHO plot and a national-reference plot of the same baby will not agree in the first year, and the disagreement can look exactly like a problem when it is a change of paper.

Frequently asked questions

My baby dropped from the 50th to the 25th centile. Is that bad?

That is one centile space, and on the published thresholds one space is not by itself a concern for a baby whose birth weight was between the 9th and 91st centile — the trigger there is two or more. It is also exactly the kind of movement produced by settling into a family pattern, by a change of chart, or by two weights taken on different scales. That said, the thresholds are written for clinicians with the whole record in front of them. If the fall bothers you, ask for the baby to be weighed again in a couple of weeks and looked at, which is what the guidance itself would do.

Should I use the WHO chart or my country's chart?

The one your health service uses, consistently, and never a mixture. For infants and young children that is the WHO standard across the United Kingdom, most of Europe and the United States under age two. The wrong move is to plot a baby on one chart at one visit and a different chart at the next, because the difference between the two — up to forty percentile points across the first year — will look exactly like a change in the child. If you want to see both, plot both from birth and compare the shapes, not the individual numbers.

Does a low centile mean my baby is not getting enough milk?

No. A centile is a position, not a diagnosis, and some perfectly well-fed babies belong on the 9th centile in the same way some belong on the 91st. What is informative is the pattern over time, together with the things a chart cannot show: how the baby feeds, how many wet and dirty nappies there are, how alert they are, whether they are content after a feed. Those are the observations a midwife, health visitor or doctor uses, and they are why an assessment involves watching a feed rather than reading a chart.

How much weight loss after birth is normal?

Some loss is expected and usually stops after about three or four days, with most infants back to birth weight by three weeks. The published action threshold is a loss of more than 10% of birth weight, which prompts a clinical assessment rather than a diagnosis. On a 7 lb 12 oz baby that is 12.4 ounces. It is not rare to reach it: in a study of 161 471 newborns, almost 5% of vaginally delivered exclusively breastfed babies and more than 10% of those born by caesarean had lost 10% or more by 48 hours. Weighing and judgement belong to your midwife or doctor, and any concern about feeding, alertness or nappies should go to them straight away.

Why do the printed centile lines have such odd numbers?

Because they are laid out in standard deviations, not in round percentages. The UK-WHO charts print nine lines — the 0.4th, 2nd, 9th, 25th, 50th, 75th, 91st, 98th and 99.6th — spaced exactly two-thirds of a standard deviation apart. That spacing makes each gap the same statistical distance, which is what allows the guidance to talk about a fall of one, two or three centile spaces and have it mean the same thing anywhere on the chart. Round percentile values would look tidier and would not have that property.

Are length and head circumference affected the same way?

Much less. The divergence between the WHO standard and formula-era references is mostly a weight phenomenon: the DARLING study found weight and lean-mass gains differing between breastfed and formula-fed infants, while length and head circumference were similar in the two groups. That is a useful thing to know when a chart shows a baby drifting down in weight but holding their line in length, which is a common and usually reassuring pattern — but it is still a pattern for the clinician following your child to interpret, not for a web page.