Data Driven Endocrine Metabolic Phenotypes in Young Women With Polycystic Ovary Syndrome and Associations With Cardiometabolic Risk Markers

ContextPolycystic ovary syndrome (PCOS) is a heterogeneous endocrine disorder associated with reproductive dysfunction and long-term cardiometabolic risk. Traditional phenotype classifications based on diagnostic criteria may not fully capture the multidimensional biological variability underlying endocrine and metabolic risk profiles, particularly in young women. ObjectiveTo identify data-driven endocrine-metabolic phenotypes in young women with PCOS and evaluate their association with established cardiometabolic risk markers. Design and SettingCross-sectional study conducted at a tertiary Gynecological Endocrinology Clinic in Poland between January 2018 and May 2025. ParticipantsA total of 1300 young women diagnosed with PCOS according to Rotterdam criteria were included. The primary analytic cohort comprised 1032 participants aged 16-25 years with complete endocrine-metabolic biomarker data. Main Outcome MeasuresEndocrine-metabolic phenotypes were derived using principal component analysis followed by Gaussian mixture model clustering. Cardiometabolic risk endpoints included impaired glucose tolerance (2-hour plasma glucose during an oral glucose tolerance test [≥]140 mg/dL), an atherogenic lipid profile (triglycerides (TG)/high-density lipoproteins (HDL-C) ratio >3.50), elevated non-HDL cholesterol ([≥]130 mg/dL), and a composite outcome of any abnormality. ResultsPrincipal component analysis retained 10 components explaining 81.9% of total variance. Unsupervised clustering identified two stable phenotypes (silhouette = 0.392; ARI = 0.842). Cluster 0 (n=954; 92.4%) represented a mixed endocrine-metabolic profile, whereas cluster 1 (n=78; 7.6%) was enriched for thyroid/autoimmune features, with higher anti-thyroid peroxidase antibody levels and higher thyroid-stimulating hormone. Cluster 1 showed a higher prevalence of an atherogenic lipid profile compared with cluster 0, while differences in glucose intolerance and non-HDL cholesterol were modest. Logistic regression analyses suggested phenotype-specific variation in cardiometabolic risk markers. ConclusionsIn a large cohort of young women with PCOS, data-driven analysis identified two reproducible endocrine-metabolic phenotypes, including a distinct thyroid/autoimmune-enriched subgroup. These findings highlight clinically relevant heterogeneity beyond traditional diagnostic phenotypes and support the potential value of integrated endocrine-metabolic profiling for early risk stratification in PCOS.