Evaluation of adolescent statural growth in health and disease: reliability of assessment from height measurement series and development of an automated algorithm.

Objectives: The precise evaluation of adolescent growth spurt is necessary for numerous clinical research studies of growth disorders and treatments. The objectives of our study were: (1) to evaluate the reliability of clinicians' 'manual' evaluation of the adolescent growth spurt from a collected series of height data, and (2) to construct an automated algorithm to determine the duration of the two phases of growth in health and disease (spurt and final slow growth) independent of clinical pubertal stages.

Methods: One hundred and seventy-four growth curves of normally growing, GH-deficient and Turner's syndrome subjects were presented twice to 2 experienced clinicians. Disagreement between evaluations and clinicians were settled to obtain a 'consensual gold standard' evaluation versus which the algorithm was assessed. Kappa statistics and Bland-Altman analyses were used to evaluate the reliability and agreement of the evaluations.

Results: The reliability of 'manual' evaluation of adolescent growth spurt from collected series of height data appeared to be poor. Conversely, the developed algorithm is perfectly reliable and satisfactorily valid. Discrepancies with the clinical consensual gold standard were always fewer than the discrepancies between the expert clinicians, and were observed in similarly difficult curves.

Conclusion: The developed algorithm may be useful for diverse clinical and biological research applications in children with growth disorders. This study also confirms the value of a comprehensive investigation of growth during adolescence independent of clinical staging.