A prospective, randomized, comparative, assessor-blind study was carried out in two centres to compare the efficacy and safety of recombinant human follicle stimulating hormone (r-hFSH; Gonal-F) versus highly purified urinary FSH (u-hFSH HP; Metrodin HP), both administered s.c. in women undergoing ovarian stimulation for in-vitro fertilization including intracytoplasmic sperm injection (ICSI). A total of 235 patients started a long gonadotrophin-releasing hormone agonist protocol: 119 received r-hFSH and 114 received u-hFSH HP (150 IU/day) for the first 6 days. Two patients were excluded from the study because they mistakenly received the incorrect treatment combination. Human chorionic gonadotrophin (HCG; 10000 IU, s.c.) was administered once there was at least one follicle 18 mm in diameter and two others > or = 16 mm. In all, 119 (100%) and 102 (89%) of the patients respectively in the r-hFSH and u-hFSH HP groups achieved the criteria for HCG. The mean numbers (+/- SD) of oocytes recovered (the primary endpoint) were 12.2 +/- 5.5 and 7.6 +/- 4.4 in the r-hFSH and u-hFSH HP groups respectively (P < 0.0001). However, the number of FSH treatment days (11.0 +/- 1.6 versus 13.5 +/- 3.7) and the number of 75 IU ampoules (21.9 +/- 5.1 versus 31.9 +/- 13.4) used were significantly less (P < 0.0001) in the r-hFSH group than in the u-hFSH HP group. In patients treated using ICSI (63 patients in each group), no difference in oocyte maturation was observed. The mean numbers of embryos obtained were 8.1 +/- 4.2 and 4.7 +/- 3.5 (P < 0.0001), in favour of the r-hFSH group. In the majority of patients (96 and 99% respectively) only one or two embryos were replaced (mean 2.0 +/- 0.2 and 1.9 +/- 0.1 respectively) in the r-hFSH and u-hFSH HP groups. The clinical pregnancy rates per started cycle and per embryo transfer were 45 and 36%, and 48 and 47%, respectively in the r-hFSH and u-hFSH HP groups (not significant). There were six (5.1%) and two (1.7%) cases of ovarian hyperstimulation syndrome respectively. In conclusion, it was found that r-hFSH was more effective than u-hFSH at inducing multiple follicular development. However, the high rate of low ovarian response in the u-hFSH group compared with our general experience was unexpected. The availability of a gonadotrophin with less inter-batch variation would be beneficial for clinicians. r-hFSH seems to fulfil such a requirement.