Identification of Metabolic Phenotypes in Polycystic Ovary Syndrome Using Unsupervised Clustering: A Case-Control Study: Metabolic Phenotypes in Polycystic Ovary Syndrome Using Unsupervised Clustering

Aqsa Haroon; Irzah Farooq; Manahl Imran

doi:10.54393/pbmj.v9i7.1385

Authors

Aqsa Haroon Department of Medicine, Sahiwal Medical College, Sahiwal, Pakistan
Irzah Farooq School of Health Sciences, University of Management and Technology, Lahore, Pakistan
Manahl Imran College of Statistical Sciences, University of the Punjab, Lahore, Pakistan

DOI:

https://doi.org/10.54393/pbmj.v9i7.1385

Keywords:

Polycystic Ovary Syndrome, Insulin Resistance, Sex Hormone-Binding Globulin, Cluster Analysis, Hirsutism, Body Mass Index, ROC Curve, Phenotype, Hyperandrogenism

Abstract

PCOS is an endocrine disorder of women. It remains unclear whether adding biomarkers to clinical assessment improves diagnosis, or if androgens explain clinical hirsutism severity. Objectives: To identify metabolic phenotypes via clustering; test if a biomarker panel predicts PCOS better than clinical assessment; and explore the serum androgen-hirsutism correlation. Methods: This retrospective case-control study analyzed Kaggle data (208 PCOS cases and controls). Clinical, biochemical (total/free testosterone, DHEAS, LH/FSH, HOMA-IR, SHBG), and ultrasound (ovarian volume, follicle count) variables were recorded. Logistic regression (clinical vs. full panel) was compared using AUC, DeLong's test, LRT, bootstrap correction, and Hosmer-Lemeshow. K-means clustering (k=2) applied to standardized BMI, HOMA-IR, and SHBG. Pearson correlation assessed FG score. Results: The full model (test AUC 0.596) did not outperform the clinical model (AUC 0.592; DeLong p=0.886; LRT p=0.835). Optimism correction reduced the full AUC to 0.507 versus 0.526 for clinical, with poor calibration (p=0.005). No biomarker independently predicted PCOS. No serum hormone correlated with FG score (|r|≤0.111). Clustering revealed obese/low-SHBG (n=98) and lean/insulin-resistant (n=110) phenotypes, differing in BMI, SHBG, and HOMA-IR (p<0.050), but not testosterone or LH/FSH. Conclusions: Adding biomarkers does not improve diagnosis and is overfitted. Serum androgens do not explain hirsutism. Clustering identifies distinct metabolic subtypes, supporting phenotype-based management.

References

Salari N, Nankali A, Ghanbari A, Jafarpour S, Ghasemi H, Dokaneheifard S et al. Global Prevalence of Polycystic Ovary Syndrome in Women Worldwide: A Comprehensive Systematic Review and Meta-Analysis. Archives of Gynecology and Obstetrics. 2024 Sep; 310(3): 1303-14. doi: 10.1007/s00404-024-07607-x. DOI: https://doi.org/10.1007/s00404-024-07607-x

Teede HJ, Tay CT, Laven JJ, Dokras A, Moran LJ, Piltonen TT et al. Recommendations from the 2023 International Evidence-Based Guideline for the Assessment and Management of Polycystic Ovary Syndrome. European Journal of Endocrinology. 2023 Aug; 189(2): G43-64.

Fauser BC, Tarlatzis BC, Rebar RW, Legro RS, Balen AH, Lobo R et al. Consensus on Women’s Health Aspects of Polycystic Ovary Syndrome (PCOS): the Amsterdam ESHRE/ASRM-Sponsored 3rd PCOS Consensus Workshop Group. Fertility and Sterility. 2012 Jan; 97(1): 28-38. doi: 10.1016/j.fertnstert.2011.09.024. DOI: https://doi.org/10.1016/j.fertnstert.2011.09.024

van der Ham K, Moolhuijsen LM, Brewer K, Sisk R, Dunaif A, Laven JS et al. Clustering Identifies Subtypes with Different Phenotypic Characteristics in Women with Polycystic Ovary Syndrome. The Journal of Clinical Endocrinology and Metabolism. 2024 Dec; 109(12): 3096-107. doi: 10.1210/clinem/dgae298. DOI: https://doi.org/10.1210/clinem/dgae298

Dapas M, Lin FT, Nadkarni GN, Sisk R, Legro RS, Urbanek M et al. Distinct Subtypes of Polycystic Ovary Syndrome with Novel Genetic Associations: An Unsupervised, Phenotypic Clustering Analysis. PLOS Medicine. 2020 Jun; 17(6): e1003132. doi: 10.1371/journal.pmed.1003132. DOI: https://doi.org/10.1371/journal.pmed.1003132

Burns KA, Stuckey AW, Wilson SG, Watts GF, Stuckey BG. Characterization of Distinct Polycystic Ovary Syndrome Subtypes by Cluster and Principal Component Analyses. Frontiers in Endocrinology. 2025 Oct; 16: 1572427. doi: 10.3389/fendo.2025.1572427. DOI: https://doi.org/10.3389/fendo.2025.1572427

Melson E, Rocha TP, Veen RJ, Abdi L, McDonnell T, Tandl V et al. Unsupervised Steroid Metabolome Cluster Analysis to Dissect Androgen Excess and Metabolic Dysfunction In 488 Women with Polycystic Ovary Syndrome-Results from the Prospective DAISy-PCOS Study. In Endocrine Abstracts. 2023 Oct; 94. doi: 10.1530/endoabs.94.OC4.1. DOI: https://doi.org/10.1530/endoabs.94.OC4.1

Zhang Y, Movva VC, Williams MS, Lee MT. Polycystic Ovary Syndrome Susceptibility Loci Inform Disease Etiological Heterogeneity. Journal of Clinical Medicine. 2021 Jun; 10(12): 2688. doi: 10.3390/jcm10122688. DOI: https://doi.org/10.3390/jcm10122688

Zeng X, Xie YJ, Liu YT, Long SL, Mo ZC. Polycystic Ovarian Syndrome: Correlation Between Hyperandrogenism, Insulin Resistance and Obesity. Clinica Chimica Acta. 2020 Mar; 502: 214-21. doi: 10.1016/j.cca.2019.11.003. DOI: https://doi.org/10.1016/j.cca.2019.11.003

Cassar S, Misso ML, Hopkins WG, Shaw CS, Teede HJ, Stepto NK. Insulin Resistance in Polycystic Ovary Syndrome: A Systematic Review and Meta-Analysis of Euglycaemic–Hyperinsulinaemic Clamp Studies. Human Reproduction. 2016 Nov; 31(11): 2619-31. doi: 10.1093/humrep/dew243. DOI: https://doi.org/10.1093/humrep/dew243

Coskun A, Ercan O, Arikan DC, Özer A, Kilinc M, Kiran G et al. Modified Ferriman–Gallwey Hirsutism Score and Androgen Levels in Turkish women. European Journal of Obstetrics and Gynecology and Reproductive Biology. 2011 Feb; 154(2): 167-71. doi: 10.1016/j.ejogrb.2010.10.001. DOI: https://doi.org/10.1016/j.ejogrb.2010.10.001

Hestiantoro A, Karimah PD, Shadrina A, Wiweko B, Muharam R, Astuti BP. Triglycerides, Independent of Ferriman-Gallwey Score, are a Main Determinant of Free Testosterone Index in PCOS. F1000Research. 2019 Jan; 8: 94. doi: 10.12688/f1000research.16815.1. DOI: https://doi.org/10.12688/f1000research.16815.1

Khan A, Karim N, Ainuddin JA, Fahim MF. Polycystic Ovarian Syndrome: Correlation between Clinical Hyperandrogenism, Anthropometric, Metabolic and Endocrine Parameters. Pakistan Journal of Medical Sciences. 2019 Sep; 35(5): 1227. doi: 10.12669/pjms.35.5.742. DOI: https://doi.org/10.12669/pjms.35.5.742

Ilagan MK, Paz-Pacheco E, Totesora DZ, Clemente-Chua LR, Jalique JR. The Modified Ferriman-Gallwey Score and Hirsutism among Filipino women. Endocrinology and Metabolism. 2019 Dec; 34(4): 347. doi: 10.3803/EnM.2019.34.4.374. DOI: https://doi.org/10.3803/EnM.2019.34.4.374

Tarroza ML, Pacquing-Songco DF, Zamora BB. Correlation of Visual (Modified Ferriman–Gallwey Scoring) and Biochemical Evaluation of Hirsutism in Polycystic Ovary Syndrome Patients in a Tertiary Hospital: A Cross-Sectional Study. Philippine Journal of Obstetrics and Gynecology. 2025 Jan; 49(1): 3-9. doi: 10.4103/pjog.pjog_36_24. DOI: https://doi.org/10.4103/pjog.pjog_36_24

Ali B, Hadadi I, Adam M, Faqihi WA, Sayed M, Alotaibi SA et al. Assessment of Sonographic Ovarian Volume as a Potential Predictor of Polycystic Ovarian Syndrome (PCOS): A Hospital-Based Cross-Sectional Study. Endocrine, Metabolic & Immune Disorders-Drug Targets. 2026 Jan; 26(1): E18715303396709. doi: 10.2174/0118715303396709250721152449. DOI: https://doi.org/10.2174/0118715303396709250721152449

Matthews DR, Hosker JP, Rudenski AS, Naylor BA, Treacher DF, Turner R. Homeostasis Model Assessment: Insulin Resistance and Β-Cell Function from Fasting Plasma Glucose and Insulin Concentrations in Man. Diabetologia. 1985 Jul; 28(7): 412-9. doi: 10.1007/BF00280883. DOI: https://doi.org/10.1007/BF00280883

Zad Z, Jiang VS, Wolf AT, Wang T, Cheng JJ, Paschalidis IC et al. Predicting Polycystic Ovary Syndrome with Machine Learning Algorithms from Electronic Health Records. Frontiers in Endocrinology. 2024 Jan; 15: 1298628. doi: 10.3389/fendo.2024.1298628. DOI: https://doi.org/10.3389/fendo.2024.1298628

Agirsoy M and Oehlschlaeger MA. A Machine Learning Approach for Non-Invasive PCOS Diagnosis from Ultrasound and Clinical Features. Scientific Reports. 2025 Sep; 15(1): 33638. doi: 10.1038/s41598-025-10453-9. DOI: https://doi.org/10.1038/s41598-025-10453-9

Deng J, Elghobashy ME, Zang K, Patel SK, Guo E, Heybati K. So You’ve Got a High AUC, Now What? An Overview of Important Considerations When Bringing Machine-Learning Models from Computer to Bedside. Medical Decision Making. 2025 Aug; 45(6): 640-53. doi: 10.1177/0272989X251343082. DOI: https://doi.org/10.1177/0272989X251343082

Van Calster B, McLernon DJ, Van Smeden M, Wynants L, Steyerberg EW, Moons Karel GM et al. Calibration: the Achilles Heel of Predictive Analytics. BioMed Central Medicine. 2019 Dec; 17(1): 230. doi: 10.1186/s12916-019-1466-7. DOI: https://doi.org/10.1186/s12916-019-1466-7