This study was undertaken to clarify discriminative roles of multiple epidemiologic, hormonal, and biophysical variables for causation of ovarian hyperstimulation syndrome. Three hundred ninety-six patients with anovulatory infertility had ovulation induction with human menopausal gonadotropin throughout 1822 treatment cycles; 54 cycles (3%) were complicated by ovarian hyperstimulation syndrome. Early follicular serum estradiol and prolactin levels were higher in this group than in controls: 75.5 versus 46.2 pg/ml and 18.5 versus 11.7 ng/ml, respectively (p less than 0.01). On the day of human chorionic gonadotropin administration (day 0) the mean serum estradiol level was 1047 +/- 381 in the group with ovarian hyperstimulation syndrome and 719 +/- 339 pg/ml in controls (p less than 0.0001). In all follicular sizes and in all grades of ovarian hyperstimulation syndrome there was a tendency for more recruited follicles, with significantly more small follicles (12 to 14 mm) present on day 0 in all grades of ovarian hyperstimulation syndrome than in controls. Stepwise logistic regression performed on 22 variables identified a high-risk group for this syndrome; the major features are illustrated by young, lean patients who, after relatively few ampules of human menopausal gonadotropin, develop high estradiol levels and multiple small follicles.