To evaluate non-invasively the role of levonorgestrel releasing devices in direct contact with the endometrium on menstrual spotting and endometrium inactivation, we inserted levonorgestrel releasing devices (20 micrograms/24 h) either into the cervical canal or the uterine cavity of 30 fertile women. Both before insertion and over the following 3 months, we used transvaginal sonography to measure the endometrial thickness in 20 of the women and Doppler flow to measure the uterine blood flow in the remaining 10 women. The women were asked to keep records of menstrual bleeding and they gave blood samples for the measurement of serum oestradiol, progesterone and levonorgestrel. By 10 weeks after insertion there was a significant decrease in endometrial thickness in both groups. Intracervical levonorgestrel release allowed the endometrium to maintain cyclic changes, whereas direct intrauterine levonorgestrel release eliminated the cyclical changes. The total number of spotting days was significantly less (P = 0.0249) in the intracervical release group at 3 months; 1.2 +/- 0.6 versus 8.1 +/- 1.8 (mean +/- SE). There were no significant differences in hormone concentrations between the groups. The pulsatility index did not change significantly during the study. We concluded that the inactivation process of the endometrium can be monitored by transvaginal sonography and that locally administered levonorgestrel does not change circulatory conditions detectable by Doppler flow. Our results also suggest that the inactivation process of the endometrium is different between intracervical and intrauterine levonorgestrel administration and may explain the difference in the number of spotting days.