Mary R. Sammon, Dan Doyle, Julie Hoover-Fong, and Karen W. Gripp contributed equally to this work.
How to Cite this Article: Sammon MR, Doyle D, Hopkins E, Sol-Church K, Stabley DL, McGready J, Schulze K, Alade Y, Hoover-Fong J, Gripp KW. 2012. Normative growth charts for individuals with Costello syndrome. Am J Med Genet Part A 158A: 2692–2699.
Costello syndrome is a very rare condition caused by a germline mutation in the proto-oncogene HRAS [Aoki et al., 2005]. Individuals typically have failure-to-thrive with macrocephaly, short stature, characteristic coarse facial features, hypertrophic cardiomyopathy (HCM), papillomata, malignant tumors, and cognitive impairment [for overview, see Gripp and Lin, 2009]. Infants with Costello syndrome often have mild fetal overgrowth. Feeding difficulties become a major problem in early infancy and many individuals develop severe failure-to-thrive, often requiring nasogastric or gastrostomy tube feedings. Linear growth is slow and usually results in short stature. Growth hormone (GH) deficiency occurs in about 40% [Gripp et al., 2006], and GH replacement therapy is often provided. Despite a concern that GH therapy may promote HCM, our data show no conclusive relationship between GH replacement therapy and HCM [Rauen et al., 2008]. Macrocephaly is a common finding in individuals with Costello syndrome, but is more often relative, with an OFC centile significantly above the length-for-age centile. We recently showed that this relative macrocephaly is likely due to progressive postnatal cerebellar overgrowth, resulting in cerebellar herniation or Chiari 1 malformation, and hydrocephalus or syringomyelia [Gripp et al., 2010]. No growth curves exist for individuals with Costello syndrome. With growth issues common to most individuals with Costello syndrome, it would be valuable to have normative curves specific to this condition. Here, we present the first set of normative curves for length, weight, and head circumference in individuals with Costello syndrome.
MATERIALS AND METHODS
This study was approved by the Institutional Review Board of A.I. duPont Hospital (A. I. duPont Hospital IRB #2003-006 and #2005-051) and the Johns Hopkins Medical Institution in Baltimore, Maryland (NA_00044029). A dataset was created from mixed cross-sectional and longitudinal anthropometric data collected by the authors (K.W.G., D.D., and M.S.) during routine clinic visits and national Costello Support Group meetings, and extracted retrospectively from medical records and growth curves as provided by the families enrolled in this study. Data represent children born 1976–2011. All data were compiled anonymously for analysis.
All diagnoses of Costello syndrome were agreed upon by a single clinical geneticist with particular expertise with this condition (K.W.G.) and confirmed at a molecular level. Only subjects in whom the diagnosis of Costello syndrome was indisputable were included for analysis, excluding individuals in whom no HRAS mutation could be identified; or with a mutation in a gene other than HRAS; or with somatic mosaicism for an HRAS mutation. The majority of measurements were obtained at Costello syndrome family meetings held biannually during an 8-year period from 2003 to 2011. The height, weight, and head circumference measurements were obtained by the same pediatric endocrinologist (D.D.) at these five seminars. All anonymized anthropometric data were entered into Excel and analyzed with R for Windows [R Development Core Team, 2010] using the gamlss macro [Stasinopoulos et al., 2008] to estimate penalized cubic smoothing splines. All statistical analyses were performed at Johns Hopkins Medical Institutions and Bloomberg School of Public Health.
The following demographic and anthropometric data were included in this database: date of birth, date when measurements were obtained, and age at measurement calculated as the difference between these dates, gender, height, weight, and head circumference. Several individuals in this cohort were reported to use GH treatment, and data obtained after first exposure to GH was excluded from our primary analysis. Height and weight measurements obtained after the first exposure to GH were subsequently compared to the curves generated from measurements in GH naïve individuals. For age 36–120 months, curves were generated comparing GH exposed to GH naïve individuals, but no data existed on GH exposed individuals younger than 36 months to allow such a comparison. Data from male and female individuals was combined because the total number of patients with Costello syndrome is small, resulting in limited data points. Further, subjectively there is no appreciable difference in the length or weight between age-matched male and female individuals with Costello syndrome. Scatterplots of anthropometric data by age were generated, and centiles (5th, 50th, and 95th) were estimated across the age continuum (0–10 years) for each growth parameter for males and females combined. A 1-month window (1 month around each time point of interest) for the period of 0–36 months, and a 3-month window for the period of 36 months (3 years) to 120 months (10 years) was used for each estimate and the centiles smoothed using quadratic, penalized-smoothing splines. The differing window choices were applied based on the relatively high density of data points in the former time frame compared to the latter time frame.
Reference data for height, weight, and head circumference for average stature children from birth through 10 years was obtained from 2000 CDC/NCHS growth curves [Kuczmarski et al., 2000, 2002]. These normative curves were superimposed on the male and female Costello anthropometry to compare growth patterns of individuals with Costello syndrome to average stature peers of the same age. Anthropometric values from male and female Costello patients were combined, but compared to gender-specific normative CDC curves. Smoothed percentile curve estimates for height and weight were created over two intervals; birth to 36 months, and birth to 10 years. A fixed cubic polynomial with a subject-specific random effect was fit to test for differing average growth trajectories for both height and weight comparing subjects with a history of GH usage to those without a history of GH usage. Based on the finding of a statistically significant interaction, separate smoothed percentile curves were estimated for subjects with a history of GH usage and subjects without this history for the period of 36 months (3 years) to 120 months (10 years). The smoothing splines for weight and height were fit using an effective degrees of freedom of 6; the smoothing spline for OFC was fit using an effective degrees of freedom of 3. The degrees of freedom was selected via a sensitivity analysis procedure comparing model fit across a range of options for the effective degrees of freedom parameter.
Data were obtained from 94 individuals with Costello syndrome (Table I), 77.7% of whom carried the HRAS p.G12S mutation, followed by 11.7% with p.G13C and 4.3% with p.G12A. This dataset of height, weight, and OFC measurements garnered 1,397 growth points from individuals age 0–10 years (Table II) for inclusion in this analysis, excluding data obtained after use of GH and past the age range included in the curves presented here. We collected 417 weight data points from 80 individuals aged 0–36 months. Weight data from 82 individuals aged 0–10 years garnered 585 points. Height measurements from 78 individuals aged 0–36 months and from 90 individuals aged 0–10 years totaled 391 and 591 data points, respectively. There was limited data in regard to head circumference with measurements from 55 individuals aged 0–36 months garnering 221 data points.
Table I. Dataset Characteristics
Males (N = 45)
Females (N = 49)
Average number of measurements per patient (range) used to generate curves
Males and females
Patients with GH treatment
Number of individuals
Table II. Dataset Characteristics Used to Generate Centile Curves
Number of data points obtained
Weight curves 0–36 months
Weight curves 0–10 years
Height curves 0–36 months
Height curves 0–10 years
OFC curves 0–36 months
Weight for age charts for Costello syndrome for clinical use for 0–36 months (Fig. 1) and 0–10 years (Fig. 2) are presented with 5th, 50th, and 95th centile isopleths.
Figure 1 indicates a lack of the rapid weight gain from 0 to 3 months in patients with Costello syndrome compared to reference data. The 50th centile for patients 0–36 months is at about the 5th centile for CDC/NCHS growth curves. Figure 2 shows overlap of the highest weight-for-age range in Costello syndrome with the lowest normal range in the general population.
Subsequent comparison of weight centiles derived from 72 observations in 15 unique individuals after their first exposure to GH suggests increased weight gain compared to GH naïve individuals (Supplementary Fig. 1). Additional individuals in our cohort used GH, however, because their first exposure occurred past age 10 years, this information is not reported here.
Height for age charts for Costello syndrome for clinical use for 0–36 months (Fig. 3) and 0–10 years (Fig. 4) are presented with the 5th, 50th, and 95th centile isopleths. The Costello syndrome curves revealed a lack of rapid growth from 0 to 6 months compared to the CDC/NCHS curves. The 95th centile of children with Costello syndrome is about equivalent to the 5th centile of the CDC/NCHS curves. Comparison of height centiles derived from 96 observations in 15 unique individuals after their first exposure to GH suggests no significant change in height compared to GH naïve individuals (Supplementary Fig. 2). Start of GH therapy after age 10 years is not reported here.
Head circumference for age charts for clinical use are presented with the 5th, 50th, and 95th centile isopleths (Fig. 5). Due to limited data, no curves could be generated for individuals with Costello syndrome past age 36 months.
The charts provided here (Figs. 1–5) represent the first set of normative growth curves for individuals with Costello syndrome. Weight and linear growth of these patients are below average for typical individuals of comparable age, consistent with the previously reported severe failure-to-thrive and short stature in individuals with Costello syndrome. Neonates with Costello syndrome are often large for gestational age, likely due to the hydrops noted in many [Lin et al., 2009], reflected by the higher weight at the 95th centile at birth compared to the typical individuals (Fig. 1). The severe failure-to-thrive seen in infancy in individuals with Costello syndrome results in the centiles falling below those for average stature individuals, with a low point at around age 12 months, when the 95th centile for individuals with Costello syndrome is just above the 5th centile of weight for age in typical individuals. Notably, the patients included in this work received nutritional support, in most cases including placement of a feeding tube. Thus, the data presented here cannot be considered “natural history,” but rather represent the development of individuals receiving medical care. Similarly, the height centiles reflect the growth of patients with Costello syndrome receiving medical care. We excluded any data points obtained after exposure to GH in our analysis, except as outlined in the comparison of these data points to the curves generated as described above (Supplementary Figs. 1 and 2). Due to the small size of the cohort overall, it was not possible to create separate charts for males and females.
The novel OFC for age curves were consistent with prior publications which have found absolute or relative macrocephaly. This research has shown that the macrocephaly is due to evolving megalencephaly, and cerebellar overgrowth which often results in postnatal cerebellar herniation and Chiari I malformation [Gripp et al., 2010]. Central nervous system abnormalities, including hydrocephalus, Chiari 1 malformation, and syrinx, require surgical intervention in 46% of patients [Gripp et al., 2010], making knowledge of normal head circumference for Costello syndrome individuals a valuable tool.
Genotype–phenotype correlation in Costello syndrome, based on the specific predicted HRAS amino acid change, is an emerging field. The HRAS p.G12S mutation is the most common change, seen in about 80% of individuals [Sol-Church and Gripp, 2009] and it is present in 77.7% of patients in the cohort analyzed here. Rarer HRAS germline changes, including p.G12C, may be associated with a more severe phenotype [van der Burgt et al., 2007; Lo et al., 2008]; whereas others, such as p.G13C, may result in a milder phenotype [Gripp et al., 2008, 2011]. Because the growth parameters of individuals with HRAS p.G13C may be closer to the range for typical individuals than those of patients with p.G12S [Gripp et al., 2011], it is important to note that data from Costello syndrome patients with rare mutations are included in our analysis (Table I).
Limitations of our study are largely due to the small number of patients with Costello syndrome, and are compounded by the lack of reliable growth data as exemplified by our inability to generate OFC centiles past age 36 months. Separate charts for male and female individuals may be achieved in the future. Alternatively, growth parameters may be analyzed based on the specific HRAS mutation, which may have a larger impact than gender. Our data obtained after first exposure to GH (Supplementary Figs. 1 and 2) suggest a positive effect on weight, but a lesser effect on height gain. However, these data need to be interpreted very cautiously as they are derived from only 15 individuals and reflect all measurements obtained after GH exposure, rather than while consistently using GH.
In conclusion, we present the first normative growth centiles for individuals with Costello syndrome. These charts are consistent with prior findings of failure-to-thrive, short stature, and relative macrocephaly.
We thank the patients and their families for allowing us to share this information.