Normative cut-offs for polycystic ovary syndrome diagnostic features in adolescents using cluster analysis.

Abstract

Listen

Translate article

The diagnosis of polycystic ovary syndrome (PCOS) remains challenging with international guidelines prioritising accurate cut-offs for individual diagnostic features. These diagnostic cut-offs are currently based on arbitrary percentiles, often from poorly characterised cohorts, and are dependent on variable laboratory ranges defined by assay manufacturers, limiting diagnostic accuracy. Cluster analysis is the recommended approach for defining normative cut-offs within populations for clinical syndromes. Few PCOS adult studies have applied cluster analysis, with no studies in adolescents. We aimed to define normative cut-offs for individual PCOS diagnostic features in a community-based population of adolescents using cluster analysis. This analysis utilised data from the Menstruation in Teenagers Study, a subgroup of the Raine Study, which is a population based prospective cohort of 244 adolescents whose mean age at PCOS assessment was 15.2 years. K-means cluster analysis and receiver operating characteristics curves were used to define normative cut-offs for modified Ferriman-Gallwey (mFG) score, free testosterone (free T), free androgen index (FAI), and menstrual cycle length. Normative cut-offs for mFG, free T, FAI, and menstrual cycle lengths were 1.0, 23.4 pmol/L, 3.6, and 29 days, respectively. These corresponded to the 65th, 71st, 70th, and 59th population percentiles, respectively. In this novel study, we define the normative diagnostic criteria cut-offs in this unselected adolescent population and show that these cut-offs correspond to lower percentiles than conventional cut-offs. These findings highlight the pertinent need to re-define PCOS diagnostic cut-offs in adolescents. Validation is required in larger, multi-ethnic, and well-characterised adolescent cohorts.