Excerpt from my new paper "Throwing a Curve at Growth Charts" to be
published in the proceedings of the NMAA conference to be held this October.
Not to be cited, just FYI and MESHOAA (My ever so humble opinion as always).

>The current growth standards used by most health care professionals around
the world for assessing children's growth are known variously as the World
Health Organization charts (WHO), the Centers for Disease Control charts
(CDC), the National Center for Health Statistics (NCHS) charts, or the Ross
Labs charts (Hamill et al., 1979).  This latter designation comes from the
fact that Ross Labs prints up pink and blue versions of these growth
standards and distributes them free to doctors' offices, clinics, and other
venues.  It is commonly stated that these charts are based on "formula-fed
children" and are therefore inappropriate for measuring the growth of
breastfed children.
>
>In fact, the data used to construct the WHO/CDC/NCHS charts came from two
sources.  One was the Fels Research Institute Data, based on studies of
growth in young children from mostly middle-class, mostly Caucasian families
in Ohio back in the 1950s.  Some of these children were breastfed, though no
one can say what percentage, or for how long.  Some of these children were
bottle-fed, though not on modern commercial formulas.  Some were fed on
early versions of commercial formulas, while many were raised on various
homemade concoctions based on evaporated milk, water, and Karo syrup.  Many
had solids introduced as early as three weeks of age, and most were
undoubtedly getting solids by three months of age.  The second source of
growth data for the NCHS standards was the National Health and Nutrition
Examination Survey (NHANES) carried out across the United States in the
1970s.  This data set included thousands of normal, healthy children from a
variety of socioeconomic and ethnic backgrounds.  Some of these children
were breastfed, and some were bottle-fed.  Most of the bottle-fed children
were given commercial formulas, but again, these were of different
compositions than the formulas available 20 years later (today).  Solids
were still being started early in the United States in the 1970s.
>
>The data from the Fels Research Institute and from the NHANES surveys were
combined, and then statistically smoothed to create the beautifully even
growth curves seen today.  The growth standards are designed to reflect
percentiles of growth.  The 50th percentile represents the line dividing the
top 50% of the sample from the bottom 50%.  That is to say, half of all
normal children are expected to grow above the 50th percentile, and half of
all normal children are expected to grow below the 50th percentile.
Likewise, the 10th percentile represents the line below which 10% of all
normal children are expected to fall.  Although this may be elementary to
most readers, I feel compelled to explain it in some detail, as I have heard
the opinion expressed that any child who is not growing "above the 70th
percentile" is faltering.  Normal healthy children are expected to be found
at every percentile, including above the 99th and below the 1st.
>
>Growth charts were originally developed for researchers to use to compare
the average growth of groups of children.  A sample of, say, 350 Boston
school girls, aged 10, could be measured and their average weight-for-age,
height-for-age, weight-for-height, and head circumferences could be compared
with the standards.  Or a sample of 115 Malian children from birth to three
years of age could be measured and their average for each age and sex
category could be compared with the standards.  The standards were designed
to be used to compare the average values for one group of children to the
standards for children of the same age and sex.  The standards were not
developed for evaluating a single child's growth, as individual children
grow erratically, not in smooth curves like the average for a group of
children, or like the standards.  In addition, the standards were designed
to be used as a yardstick, to allow people easily to compare the growth of
different groups of children, not as a goal to be achieved nor as an
indicator of maximal or optimal growth.  If I measure my sample children in
inches, and another researcher measures her sample in centimeters, and a
third researcher measures his children in Coca-Cola can lengths, or in
"hands" like a horse, then we cannot easily compare our data.  If we all
agree to take measurements in centimeters, and we further all agree to
express our data as z-scores — standard deviation units away from the mean
of the standards — then we can easily compare our results.  Based as they
are on a hodgepodge of children, most of whom were not raised according to
current WHO infant feeding recommendations, the current standards should not
be considered to represent maximal growth, nor optimal growth, nor even
adequate growth.  They are a yardstick to ease comparisons between different
samples from around the world, nothing more, nothing less. 
 
>When misunderstanding about the meaning of percentiles is added to
misunderstanding of the purpose of the standards, the potential for
misinterpretation of the data is high, especially the data on weight for
age, is high, as we will discuss below.

I go on to explain reasons why breastfed children in some recent might not
grow as well as the standards, from catch-down growth to over-inflated birth
weights due to IV fluids mom got during labor, to infrequent feeding and
solitary sleeping cutting down on breast milk quantity and quality (fat
content).  
Katherine A. Dettwyler, Ph.D.
Associate Professor of Anthropology and Nutrition
Texas A&M University