Abstract Study question In a cohort of idiopathic and unexplained infertile men we aimed to identify subgroups with similar characteristics, and therewith underlying etiologic factors, by clustering approach. Summary answer We identified two distinct patient clusters. Across all diverse phenotypes of infertility, the strongest segregation markers were FSHB c.-211G>T, FSH, and bi-testicular volume. What is known already In about 30-75% of infertile men no major causative factors can be identified; leading to the diagnose of unexplained (normozoospermia) or idiopathic (abnormal semen parameters) male infertility. This cohort of men remains very heterogenous, albeit the detailed andrological characterization that is currently applied in infertility workup. New analysis tools such as machine learning and cluster analysis can provide a more in-depth approach. Such explorative analyses have the potential to uncover hitherto hidden patterns in data that might be difficult to spot for andrologists but become visible by these tools. Study design, size, duration A Cluster analysis was retrospectively performed in a clinically well characterized cohort of 2742 men with unexplained or idiopathic male infertility. These men had visited our Centre within a 10-year period (2008-2018) for infertility workup. Due to the well curated database (Androbase®) we were able to include up to 37 andrologic parameters in the unbiased cluster analysis. Participants/materials, setting, methods After applying strict selection criteria 2742, of initially 7627, infertile men remained for cluster analysis (exclusion: obstructive -, genetic -, other causative factors, female factor; inclusion: azoo- to normozoospermia, FSH ≥ 1IU/l, Testosterone ≥ 8nmol/l). For subsequent analyses the following parameters were included: somatic/semen/hormone parameters, testicular sonography and testis volume, genotyping of the FSHB c.-211G>T (rs10835638) single nucleotide polymorphism. For cluster analysis, partitioning around medoids method was employed based on Gower distance between patients. Main results and the role of chance The applied cluster approach for the study population yielded two separate clusters (average silhouette width ∼0.12). These clusters showed significantly different distributions in bi-testicular volume, FSH and FSHB genotype. Cluster 1 contained all men homozygous for G (wildtype) in FSHB c.-211G>T (100%), while Cluster 2 contained most patients carrying a T allele (>96.6%). Even in subgroup analysis (Total sperm count (TSC) <1Mill and TSC 1³Mill) two clusters each were formed too. Again, the strongest segregation markers between the respective clusters were FSHB c.-211G>T, bitesticular volume, and FSH, supporting the notion of a contributing genetic factor. Surprisingly, sperm parameters like TSC, motility and morphology played a minor role in cluster formation; as well as testicular maldescent, varicocele, smoking, and microlithiasis testes. The genetic parameter of FSHB c.-211G>T in combination with the established parameters FSH and testicular volume should attract more attention in future clinical workups of infertile men with unknown etiologic factors. Limitations, reasons for caution Categorical and numeric features contribute diversely to the calculation of patient dissimilarity. Potentially, categorical features can have a higher impact because patients are rated as completely different if they fall in different categories; for numeric features, the dissimilarity depends on the range of values. Wider implications of the findings The FSHB SNP was identified as an informative segregation marker; we therefore suggest introducing diagnostic genotyping into clinical routine in men with so far idiopathic or unexplained male infertility. This may reduce the high number of infertile men with so far unknown origin by nearly one-third. Trial registration number DFG Clinical Research Unit 326 Male Germ Cells
Read full abstract