This prospective, double-blind, randomized, multicentre study compared the efficacy and safety of recombinant human follicle stimulating hormone (r-hFSH; Gonal-F((R))) versus highly purified urinary FSH (u-hFSH HP; Metrodin((R)) HP) in women undergoing ovarian stimulation for in-vitro fertilization, including intracytoplasmic sperm injection. A total of 278 patients began a long gonadotrophin-releasing hormone agonist protocol, then 139 received r-hFSH and 139 u-hFSH HP, 150 IU/day administered s.c., for the first 6 days of treatment. On day 7, the dose was adjusted, if necessary, according to ovarian response. Human chorionic gonadotrophin (HCG, 10 000 IU, s.c.) was administered once there was more than one follicle 18 mm in diameter and two others >/=16 mm. Oocyte retrieval was performed 36-38 h after HCG injection: 128 patients (92%) receiving r-hFSH and 113 (81%) receiving u-hFSH HP had at least one oocyte retrieved. Among patients receiving r-hFSH, there was a significantly higher mean (+/- SD) number of oocytes retrieved (11.0 +/- 5.9 versus 8.8 +/- 4.8 with u-hFSH HP; P = 0. 002) and mean number of embryos obtained (5.1 +/- 3.7 versus 3.5 +/- 2.9 with u-hFSH HP; P = 0.0001). With r-hFSH, significantly fewer FSH treatment days (11.7 +/- 1.9 versus 14.5 +/- 3.3) and 75 IU ampoules (27.6 +/- 10.2 versus 40.7 +/- 13.6) were required than with u-hFSH HP (P = 0.0001). Embryo replacement on day 2-3 after oocyte retrieval resulted in 36 liveborn children in the Gonal-F((R)) group and 33 in the Metrodin HP((R)) group (not significant). There were seven cases (5.0%) of ovarian hyperstimulation syndrome in the r-hFSH group and three (2.2%), in the u-hFSH HP group (not significant). It is concluded that r-hFSH is more effective than u-hFSH in inducing multiple follicular development.