In 1950 a trial was set up to evaluate the effects of large doses of stilboestrol and ethisterone on rates of fetal loss in pregnant diabetic women. Eighty women were allocated at random to receive the hormonal treatment and 76 to receive inactive tablets of identical appearance. At follow-up 27 years later, information was obtained about 97% of the women, all but four being traced. All respondents were unaware of who had received hormones. The overall mortality was 4.5 times that of women of comparable age in England and Wales, most deaths being from complications of diabetes. More tumours, mainly benign, of the reproductive tract were reported in the hormone-exposed than the non-exposed group (14 (18%) and two (3%) respectively). Four cases of malignant breast disease were reported in the hormone-exposed women and none in the non-exposed. These findings support other evidence linking oestrogen treatment and breast cancer and suggesting that the latent period before the tumour becomes clinically apparent may be 15 years or longer.