hMG and recombinant FSH, have both been used successfully for controlled ovarian hyperstimulation in in vitro fertilization and embryo transfer (IVF-ET). To compare the effectiveness of hMG with rFSH in ovarian stimulation protocols in IVF or ICSI treatment cycles. We searched the Cochrane Menstrual Disorders and Subfertility Group trials register (searched 3rd Jan 2002), PubMed, MEDLINE, Web of Science (all searched 1985 to May 15 2002), and reference lists of articles. We also contacted manufacturers and researchers in the field. Randomised trials comparing hMG with rFSH for ovarian stimulation in IVF or ICSI treatment for treatment of infertility in normogonadotrophic women. The main outcome measure was ongoing pregnancy/live birth per woman.Secondary outcomes included total gonadotrophin dose used, cancellation, number of oocytes retrieved, implantation, clinical pregnancy per woman, multiple pregnancy, spontaneous abortion and ovarian hyperstimulation syndrome. Peto odds ratios (OR) for hMG relative to rFSH were calculated after testing for homogeneity of treatment effect across all trials. Analyses were performed separately for the three different GnRHa protocols used: (1) without GnRHa down-regulation, (2) with GnRHa down-regulation using a short protocol and (3) with GnRHa down-regulation using a long protocol. Eight trials that met the inclusion criteria could be identified. One trial did not use down-regulation, one trial used a short protocol and six trials used a long down-regulation protocol. In the one trial with non-down-regulated women and in the one trial that used a short down-regulation protocol there was no evidence of a difference between hMG and rFSH in any clinical outcome. Data of four truly randomised trials in women down-regulated using a long protocol could be pooled. There was no evidence of a difference between hMG and rFSH in ongoing pregnancy/live birth per woman (OR 1.27; 95% CI 0.98 to 1.64). Furthermore there was no clear difference on any of the secondary outcomes, although the clinical pregnancy rate per woman was of borderline significance in favour of hMG (summary OR 1.28; 95% CI 1.00 to 1.64). The other secondary outcomes were comparable for both gonadotrophins. For all three GnRHa protocols analysed there is insufficient evidence of a difference between hMG and rFSH on ongoing pregnancy or live birth. More large randomised trials are needed to estimate the difference between hMG and rFSH more precisely. Such trials should preferably (1) use a consistent long GnRHa protocol and (2) use a fixed dose of gonadotrophin such to prevent potentially subjective decisions of the clinician in dosing and (3) take live birth as primary endpoint. At this moment in time however, in prescribing gonadotrophins for ovarian hyperstimulation in IVF one should use the least expensive medication.