Development of Specific Growth Chart for Children with Fanconi Anemia

Description

Fanconi anemia (FA) is often associated with poor growth due to a combination of endocrine and non-endocrine causes. There are currently no disease-specific growth curves. Despite short stature being regarded as a major characteristic of FA (Auerbach, 2009), there is currently no established method of accurately monitoring growth expectations for individuals with this condition. Use of bone age to estimate adult height is viewed as falsely inflating projected growth outcomes in FA as this assumes normal hormone production, nutrition, and puberty onset, which are not seen in FA (Petryk et al., 2015). Hormone response and duration of puberty may also be abnormal in individuals with FA. Midparental target height is frequently not appropriate for setting expected growth in patients with FA (Petryk et al., 2015). As such, we aimed to create disease-specific growth curves using a cohort of individuals with FA seeking care at the University of Minnesota. The Bone Marrow Transplant (BMT) Registry at the University of Minnesota was used to find a large cohort of patients with FA seen at the Fanconi Anemia Comprehensive Care Center at the University of Minnesota since its inception in 1968. Age, sex, height, and weight data from birth until age 20 years old were abstracted from electronic health records (EHR) beginning in spring of 2011 when EPIC was instituted as the EHR at the FA Comprehensive Care Center. Additional data were abstracted including results of FANC gene variant testing and treatment data such as BMT status. Data was collected in a REDCap database housed by the University of Minnesota. Patient growth trajectories were visualized by plotting, longitudinal anthropometric data for height, weight, and BMI serially by clinic visits for age (i.e., spaghetti plots). Prior to statistical modeling, all de-identified patient data were screened for data quality issues such as: same-day anthropometric measurements, duplicate measurements, last measurement carried-forward using a pediatric growth cleaner statistical tool (Daymont et al., 2017). The Lambda-Mu-Sigma (LMS) growth modeling technique was then used to fit height-for-age, weight-for-age, and BMI-for-age growth functions from which reference percentile ranges for children with FA were estimated. The LMS modeling technique is from the Generalized Additive Model for Location Scale and Shape (GAMLSS) family of statistical approaches (Rigby & Stasinopoulos, 2004). It is a modeling methodology based on penalized maximum likelihood estimation with cubic splines to smooth for roughness in age-related growth channels to calculate percentiles (Cole & Green, 1992). Extensive statistical and visual diagnostics GAMLSS tools were used in select our final models. To compare growth differences, the FA-specific percentiles charts were then overlaid on WHO standards for ages 0-2 years old and CDC references for ages 2-20 years old.

Steps to reproduce

In total, we reviewed the charts of 260 patients with FA. 125 patients (60 female) had available data in the Epic EHR. Growth data from 13,110 unique clinic visits were used. Of the 125 patients with FA included in the creation of our disease-specific growth charts, 93 patients (74%) had received at least one HSCT. Majority (56%) of patients with available data are of the FANCA complementation group. Patient growth trajectories were visualized by plotting, longitudinal anthropometric data for height, weight, and BMI serially by clinic visits for age (i.e., spaghetti plots). Prior to statistical modeling, all de-identified patient data were screened for data quality issues such as: same-day anthropometric measurements, duplicate measurements, last measurement carried-forward using a pediatric growth cleaner statistical tool (Daymont et al., 2017). The Lambda-Mu-Sigma (LMS) growth modeling technique was then used to fit height-for-age, weight-for-age, and BMI-for-age growth functions from which reference percentile ranges for children with FA were estimated. The LMS modeling technique is from the Generalized Additive Model for Location Scale and Shape (GAMLSS) family of statistical approaches (Rigby & Stasinopoulos, 2004). It is a modeling methodology based on penalized maximum likelihood estimation with cubic splines to smooth for roughness in age-related growth channels to calculate percentiles (Cole & Green, 1992). Extensive statistical and visual diagnostics GAMLSS tools were used in select our final models. To compare growth differences, the FA-specific percentiles charts were then overlaid on WHO standards for ages 0-2 years old and CDC references for ages 2-20 years old.

Institutions

Categories

Funding

Licence