Development of a clinical decision support tool for monitoring and treatment of children with Congenital Adrenal Hyperplasia

Introduction

Congenital adrenal hyperplasia (CAH) is an autosomal recessive condition caused in over 90% of cases by 21-hydroxylase deficiency (21OHD), that leads to cortisol deficiency. This thesis applies longitudinal statistical modelling to gain insights into clinical trial and international registry data from patients with CAH, to inform and develop a prototype clinical decision support tool that supports clinicians managing children and young people with CAH.

Methods

Contemporary models that predict height or weight were systematically reviewed. A nationwide UK service evaluation quantified the satisfaction of services provided for children with CAH. Longitudinal multilevel modelling was used to assess biomarkers, blood pressure and growth in patients with CAH. A web application was developed in a browser interface to exhibit a prototype clinical decision support tool. All analysis was carried out in R (https://www.R-project.org/).

Results

Models developed (n=180) or validated (n=61) to predict height or weight in 2022 were at significant risk of bias. Questionnaires from 229 respondents showed satisfaction with services for CAH in the UK is high, although families would appreciate further education about the condition. Spline modelling of 477 24-hour profiles of 17-Hydroxyprogesterone from 122 patients show they follow a similar pattern to those of androstenedione, with adults taking modified-release hydrocortisone exhibiting reduced variability in these markers throughout the day. Registry data from 554 children with 21OHD showed higher BP was common at younger ages, but the clinical significance of this remains uncertain. Analysis of the growth of 573 children with CAH showed an early adiposity rebound and blunted latest peak height velocity.

Conclusion

The prototype clinical decision support tool (https://endocrinology.shinyapps.io/cah_data_visualiser/) demonstrates how data visualisation and automated calculations of clinical relevance could be incorporated into registry data entry platforms, to make the process of data entry more attractive and useful for frontline clinicians.