Exploring the heterogeneity of polycystic ovary syndrome with principal component analysis — ASN Events

Exploring the heterogeneity of polycystic ovary syndrome with principal component analysis (#69)

Bronwyn Stuckey 1 2 3 , Nicole Opie 4 , Andrea Cussons 5 , Gerald Watts 3 , Valerie Burke 3
  1. Keogh Institute for Medical Research, Nedlands, WA, Australia
  2. Department of Endocrinology and Diabetes, Sir Charles Gairdner Hospital, Nedlands, WA, Australia
  3. School of Medicine and Pharmacology, University of Western Australia, Perth, WA, Australia
  4. Third Department of Medicine, Medical School University of Athens, Athens, Greece
  5. Department of Endocrinology and Diabetes, Royal Perth Hospital, Perth, WA, Australia

Context. Polycystic ovary syndrome (PCOS) is a heterogeneous condition associated with variables of cardiometabolic risk. The heterogeneity within the syndrome implies that there is not one aetiological factor for, nor a predictable clinical consequence of, PCOS. Principal component analysis (PCA) is a statistical method which allows the several components of a set of data to be focused into independent, orthogonal, subsets of variables.
Objective. To define orthogonal factors within PCOS that may be of use in delineating subgroups within the syndrome.
Design. We used PCA to examine the endocrine and cardiometabolic variables associated with PCOS as defined by the National Institutes of Health (NIH) criteria – menstrual irregularity and hyperandrogenism.
Patients. Data from 378 unmedicated women with PCOS were retrieved.
Measurements. Data included weight and height, blood pressure, fasting blood for glucose and insulin, lipids, gonadotrophins, ovarian and adrenal androgens. PCA was performed retaining those factors with eigenvalues of at least 1.0. Varimax rotation was used to produce interpretable factors.
Results. We identified three principal components explaining 60% of the variance in PCOS. In component 1 the dominant variables were homeostatic model assessment (HOMA) index, body mass index (BMI), high density lipoprotein (HDL) cholesterol and SHBG; in component 2, systolic blood pressure, low density lipoprotein (LDL) cholesterol and triglycerides; in component 3, total testosterone and high LH:FSH ratio. These three components explained 36%, 13% and 11% of the variance in the PCOS cohort respectively.
Conclusions. These data support three principal components characterized by insulin resistance, dylipidaemia/hypertension and gonadotrophin driven hyperandrogenaemia respectively. These three components are, by definition, distinct and non-collinear and may imply different aetiological factors even though the features of more than one factor may co-exist in the same patient. These findings suggest different pathogenetic pathways within PCOS and/or differing clinical cardiometabolic outcomes.