Objectives: To determine the IVF-ET pregnancy potential of women with variably elevated day 3 FSH. Design: Cohort evaluation of 1,868 consecutive IVF-ET cycles January 1991 to December 1994. Setting: University hospital infertility unit. Patients: Four cohorts of couples were defined based on day 3 FSH determinations with an arbitrary threshold of 20 mIU/mL, only greater than or equal to 20 mIU/mL, always <20 mIU/mL, current <20 mIU/mL but one previous greater than or equal to 20 mIU/mL, and current <20 mIU/mL but two or more previous greater than or equal to 20 mIU/mL (conversion factor to SI unit, 1.00). Intervention: In vitro fertilization-embryo transfer. Main Outcome Measure: Fetal heart activity on luteal day 40 transvaginal ultrasound. Results: No pregnancies occurred in 53 cycles with day 3 FSH only greater than or equal to 20 mIU/mL. In 1,750 women whose day 3 FSH levels were always <20 mIU/mL, the pregnancy rate (PR) per cycle was 16.5%. In 54 cycles in which day 3 FSH was greater than or equal to 20 one time only, but <20 mIU/mL during the treatment cycle, the PR was 5.6%. In 11 cycles where two or more previous FSH determinations were greater than or equal to 20 mIU/mL but with a current day 3 FSH <20 mIU/mL, no pregnancies occurred. Conclusion: Our data leads us to the conclusion that day 3 FSH determination precede every IVF cycle and that cycles with FSH greater than or equal to 20 mIU/mL be canceled. It also suggests that women with two previous elevations of day 3 FSH be discouraged from future IVF cycles. The 5.6% pregnancy per cycle with one previously elevated FSH warrants extreme pessimism in discussion of further cycles.