BACKGROUND: Tamoxifen, a drug with antioestrogenic effects, is predicted to prevent the occurrence of breast cancer. We have undertaken a trial of tamoxifen in healthy women who are at increased risk of breast cancer because of family history. We report a planned interim analysis. METHODS: Between October, 1986, and April, 1996, we accrued 2494 healthy women aged between 30 and 70 with a family history of breast cancer. They have been randomised (double blind) to receive tamoxifen 20 mg per day orally or placebo for up to 8 years. Follow-up included clinical assessment, annual mammography, and assessment of toxicity and compliance. The primary endpoint was the occurrence of breast cancer, which was analysed on an intention-to-treat basis with a survival curve. FINDINGS: With a median follow-up of 70 months, 2471 women (tamoxifen 1238, placebo 1233) were suitable for analysis. The groups were evenly matched at baseline, and compliance was good. The overall frequency of breast cancer is the same for women on tamoxifen or placebo (tamoxifen 34, placebo 36, relative risk 1.06 [95% CI 0.7-1.7], p=0.8). Participants who were already on hormone-replacement therapy when they entered the study had an increased risk of breast cancer compared with non-users. Those participants who started such therapy while on trial had a significantly reduced risk. The safety profile of tamoxifen was as expected. INTERPRETATION: We have been unable to show any effect of tamoxifen on breast-cancer incidence in healthy women, contrary to the report from the NSABP-P1 study showing a 45% reduction in healthy women given tamoxifen versus placebo. Differences in the study populations for the two trials may underlie these conflicting findings: eligibility in our trial was based predominantly on a strong family history of breast cancer whereas in the NSABP trial was mostly based on non-genetic risk factors. The importance of oestrogen promotion may vary between such populations.