Identification of andrologic patient groups by cluster analysis

Interrelationship between ultrasonography and biology in the diagnosis of polycystic ovarian syndrome

In order to compare the diagnostic significance of hormonal and ultrasonic criteria of polycystic ovarian syndrome (PCOS), the presence or the absence of ultrasonographic and hormonal features of PCOS were recorded in a heterogeneous population of 90 women presenting with hyperandrogenism and/or menstrual disorders. On clinical and hormonal grounds exclusively, these patients could be separated into five diagnostic subgroups: presumed cases of PCOS (n = 21), idiopathic hirsutism (IH) (n = 26), hypothalamic anovulation (HA) (n = 11), hyperprolactinemia (HPRL) (n = 9), and miscellaneous or undetermined diagnosis (n = 23). By the means of a computed automatic classification of patients (cluster analysis) using five hormonal and ultrasonic criteria of PCOS, four homogeneous clusters of patients were obtained. Cluster #1 (25 patients) had the most characteristic profile of PCOS. It included 15 cases of PCOS and 7 cases of IH. Cluster #4 (47 patients) had the less characteristic profile of PCOS. It included the majority of patients with HA and HPRL and the half of the patients with IH. Cluster #2 included only two hyperandrogenic patients, who were massively obese and in whom ultrasonography may have failed to detect PCOS. Cluster #3 (16 patients) included patients from each diagnostic group, who were gathered together because ultrasonographic and hormonal features were, respectively, present and absent in nearly all of them. With the same analysis, the criteria of PCOS could be graded according to their grouping potential. The presence of an abnormal ovarian stroma by ultrasonography appeared as the most potent criterion. Elevated serum testosterone and androstenedione levels and the polyfollicular pattern of ovaries gave intermediate results, while elevated basal LH level was a much weaker grouping parameter. In conclusion, the automatic classification of patients by cluster analysis using both hormonal and ultrasonographic criteria revealed that the classical diagnostic classification, relying upon hormonal data exclusively, may arbitrarily separate patients having the same disease; and that ultrasonography affords pertinent information that should help provide a better diagnostic definition of PCOS.

Statistical analysis in andrology

The authors began a statistical study on a file group of 9384 records of semen analyses and anamnestic data (P-1). From this file, they selected a group of 5191 patients having undergone analysis at least one time (P-2) and a file of 850 patients having undergone analysis at least three times (P-3). Finally a new group, derived from P-3, of 203 patients followed by their medical team was created (P-4). Two control groups-composed of 20 sperm-bank donors (DON) and 740 prevasectomy subjects (VAS)-were also created. Frequency distributions and cluster analyses were carried out. The multivariate statistical analysis allows a distinction to be made between fertile and infertile subjects regarding a patient's clinical condition. The most discriminating parameters appear to be sperm concentration and, above all, forward motility and sperm morphology.