The Application of Principal Component Analysis on Clinical and Biochemical Parameters Exemplified in Children With Congenital Adrenal Hyperplasia

Purpose: Principal component analysis (PCA) is a mathematical model which simplifies data into new, combined variables. Optimal treatment of pediatric congenital adrenal hyperplasia (CAH) remains a challenge and requires evaluation of all biochemical and clinical markers. The aim of this study was to introduce PCA methodology as a tool to optimize management in a cohort of pediatric and adolescent patients with CAH by including adrenal steroid measurements and clinical parameters.

Methods: This retrospective, longitudinal cohort of 33 children and adolescents with CAH due to 21-hydroxylase deficiency included 406 follow-up observations. PCAs were applied to serum hormone concentrations and compared to treatment efficacy evaluated by clinical parameters.

Results: We provide and describe the first PCA models with hormone parameters denoted in sex- and age-adjusted standard deviation (SD) scores to comprehensibly describe the combined 'endocrine profiles' of patients with classical and non-classical CAH, respectively. Endocrine profile scores were predictive markers of treatment efficacy for classical (AUC=92%; accuracy 95%; p=1.8e-06) and non-classical CAH (AUC=80%; accuracy 91%; p=0.004). A combined PCA demonstrated clustering of patients with classical and non-classical CAH by serum 17-hydroxyprogesterone (17-OHP) and dehydroepiandrosterone-sulphate (DHEAS) concentrations.

Conclusion: As an example of the possibilities of PCA, endocrine profiles were successfully able to distinguish between patients with CAH according to treatment efficacy and to elucidate biochemical differences between classical and non-classical CAH.