The Effect of Chronic Diseases on Spermogram and Serum Parameters in Male Patients Admitted to the Andrology Laboratory: Unsupervised Machine Learning

This study aimed to evaluate andrological patterns across chronic disease groups in men who presented to the andrology laboratory, by applying an unsupervised machine-learning approach to spermogram and serum laboratory parameters. The retrospective study included 879 patients who presented from January 2021 to January 2025 and who were classified into chronic disease blocks according to ICD codes. Chronic diseases were grouped into allergic diseases, anemia, diabetes, hypertension, endocrine disorders, cardiovascular diseases, neurological diseases, psychiatric diseases, and peptic ulcers. The spermogram model included volume, sperm concentration, total motility, progressive motility, non-progressive motility, and immotility. Ward hierarchical clustering was used as the primary method, and KMeans was applied in a sensitivity analysis. The model based solely on spermogram variables provided the most balanced and clinically interpretable separation. The variables that most strongly distinguished the clusters were total motility, progressive motility, and immotility. Cluster 0 exhibited reduced motility, whereas Cluster 1 exhibited a relatively preserved spermogram. In the clinical reference threshold analysis, values outside the reference range were most frequent for progressive and total motility. In men with chronic disease, the principal andrological differentiation was centered on the spermogram and driven primarily by sperm motility. Serum variables provided complementary information on the systemic biological context of the chronic disease groups but did not strengthen the primary cluster separation.