Metabolic and hormonal heterogeneity in polyendocrine metabolic ovarian syndrome in a South Indian cohort identified by k-means clustering

Context In polyendocrine metabolic ovarian syndrome (PMOS), conventional Rotterdam phenotypes may not adequately capture the multidimensional hormonal and metabolic heterogeneity of the disorder. Aims To identify clinically meaningful PMOS subtypes using a data-driven clustering approach among infertile South Indian women. Methods A cross-sectional study was conducted among 196 infertile women diagnosed with PMOS (Rotterdam criteria) at a fertility centre in South India. Unsupervised k-means clustering was performed and cluster stability was assessed using Jaccard bootstrapping. Key results Three subtypes were identified: gonadotrophin predominant, body mass index (BMI) predominant, and intermediate. The gonadotrophin predominant cluster exhibited intermediate BMI with higher LH and FSH concentrations and the highest LH/FSH ratio. The BMI predominant cluster demonstrated the highest BMI with a relatively low LH/FSH ratio, while the intermediate cluster showed lean-to-normal BMI with the lowest LH, FSH, and LH/FSH ratio. Significant differences were observed between clusters in systolic and diastolic blood pressure and endometrial thickness. No significant between-cluster differences were found for anti-Müllerian hormone, estradiol, prolactin, thyroid-stimulating hormone and haemoglobin A1c. Conclusions Data-driven clustering using BMI, LH, and FSH identified three PMOS subtypes, demonstrating heterogeneity beyond the conventional Rotterdam phenotypic classification. Implications The identified subtypes enhance understanding of clinical heterogeneity of PMOS and provide a basis for potential application in personalized patient management.