Study population
Growth data of MWS patients previously published by our workgroup were incorporated into the study [7]. The invitation to participate in the study, explaining its purpose and design, was sent to all the colleagues who previously published on MWS. Furthermore, the Italian Mowat-Wilson Association (AIMW) and the MWS Foundation forwarded the invitation to their members. To those who agreed to participate, we sent a specially designed questionnaire where they could insert the length/height, weight and head circumference measurments at different ages. The collected data were reviewed on a password-protected database and if inconsistencies, probable errors or missing data were detected, the family/referring doctor was contacted for further information and correction of data, if needed. Only individuals with both clinical and molecular confirmation of the disease were included in the study. Two patients, in which we did not find the causal ZEB2 variant, were excluded from the study.
A total of 20 nationalities were represented in the study: 38 Italian patients, 16 from the USA, 7 from Poland, 6 from Japan, 5 from Denmark, 4 from Spain, 4 from the UK, 3 from Portugal, 3 from Finland, 2 from Croatia, 2 from Serbia, and 1 each from Albania, Brazil, Chile, Cyprus, France, Germany, Mexico, Turkey and Ukraine. None of the affected individuals originated from a consanguineous family. In one Italian family we had 2 affected siblings. None of the patients were treated with growth hormone.
A total of 99 patients were included in the study, 53 females and 46 males, born between 1986 and 2016. A total of 2,865 measurements were available: 1,220 data points from 46 males, and 1,645 data points from 53 females. These patients had a mean age of 12 years (range: 4-20 years) in males and 13 years (range: 3-23 years) in females at the time of assessment.
A total of 1,013 height measurements (573 for females, 440 for males), 1,110 weight measurements (631 for females, 479 for males), and 742 head circumference measurements (441 for females, 301 for males) were available. For 97 patients we had at least 2 measurements, and for 89 we gathered more than 4 measurements. We had an average of 10.39 (SD 6.52) height measurements, 11.37 (SD 6.77) weight measurements, and 7.64 (SD 5.76) head circumference measurements per patient (Additional file 1- Table 1).
Due to limited data, no growth curves could be generated for individuals with MWS past the age of 16 years.
Data from 1986 until 2019 were included. Height, weight and head circumference parameters were analyzed from birth until 23 years of age, both for males and females. BMI was calculated from these data, using the formula BMI=mass(kg)/(height(m))2.
Distribution centiles for birth, weight, length and head circumference were computed including preterm and at term newborns, i.e. from 32 to 42 gestational weeks (mean 36.5), and presented separately for males and females.
The data after birth were divided into different age and sex groups, with 1-month intervals during the first year of life, 3-months intervals during the 2nd and 3rd years of life, and 6-months intervals between the 4th and 23rd years of life.
Our data were compared with the CDC reference percentiles for anthropometric measurements for all the growth charts, with a few exceptions: for those concerning the neonatal anthropometric assessment we referred to the INeS growth charts, those concerning the head circumference of males and females aged between 3 and 16 years were compared with the Tanner reference percentiles for anthropometric measurements, and those concerning the standard deviations of the head circumference of males and females aged between birth and 18 years were compared with the Nellhaus reference standard deviations [21-28].
Statistical analysis
The growth charts were developed using the Generalised Additive Model for Location Scale and Shape (GAMLSS) [29,30] package for the R statistical program [31].
Centile estimation of anthropometric data Y (weight, length, BMI and head circumference) for any given age was:
Y ⁓ D (µ, δ, ν, τ)
g1(µ)=h1(x)
g2(δ)=h2(x)
g3(ν)=h3(x)
g4(τ)=h4(x)
x=ageξ
where the distribution D represents the best general probability (density) function according to Akaike Infromation Criterion (AIC) and µ the first parameter of the distribution (usually location), δ the second parameter of the distribution (usually scale, coefficient of variation), ν the third parameter of the distribution (usually shape, e.g. skewness), τ the fourth parameter of the distribution (usually shape, e.g. kurtosis).
The g() functions represent appropriate link functions, the h() are non-parametric smoothing functions and ξ is a power transformation of age (Additional file 2. Table 2).
The more traditional LMS method developed by Cole and Green [32] can be fitted within this framework by assuming that the response variable has a Box-Cox Cole and Green distribution. [33,34] Table 2 (Additional file 2) describes the model specifications by anthropometric measure and sex. The Q test was conducted to evaluate the model fit [35].
The estimated centiles by age and sex were converted back to Excel to create charts for each anthropometric variable, including the CDC growth charts for healthy children.