IUDs releasing 20 mcg/day of levonorgestrel (LNg20) were in randomized trial together with the Copper T, model TCu 380Ag, in seven centers involving 2244 women. Two-year (25 months) gross cumulative pregnancy rates were 0.2 +/- 0.2 and 0.9 +/- 0.3 for the levonorgestrel and copper releasing devices, respectively (P greater than 0.05). There were no ectopic pregnancies in more than 1600 woman-years of use of each device. Removal rates for bleeding and/or pain or for medical reasons other than menstrual problems did not differ significantly between devices. Oligomenorrhea or amenorrhea prompted 10.7 per hundred (gross rate, 8.4 net rate) women using the LNg 20 IUD to request removal in the two-year period, significantly above the 0.2 per hundred rate among women with the Copper IUD (P less than 0.001). At the end of two years an estimated 59.4 per 100 women were continuing use of the LNg 20 IUD, and 67.5 per 100 (P less than 0.001) with the TCu 380Ag. This difference is almost wholly ascribable to a marked reduction in bleeding episodes and days among women using the LNg 20 device with concomitant removal of device. Hemoglobin rose an average of 0.5 g/dl (P less than 0.001) for this group whereas women using the TCu 380Ag experienced a decline of 0.2 g/dl compared with baseline values (P less than 0.001).