A retrospective study of maternity records from 1977-1985 (38,000 deliveries) was conducted to determine the number of stillbirths that might have been prevented by a new method of antenatal assessment. During this period there were 240 stillbirths, 154 of which involved a normally formed fetus who died prior to the onset of labour. After a review of the literature, a set of risk factors were selected relevant to stillbirth. Such factors were found in 60% of the study group, compared with 38% of control patients. Seventy-five of the normally formed 'antepartum stillbirths' occurred in the high-risk group after 31 weeks gestation, so that a perfect method of prediction and treatment, applied from 31 weeks onwards, would potentially have prevented half of the 154 deaths in this study. Since intensive monitoring and the subsequent intervention cannot attain such perfection, an assessment was made, using realistic sensitivity and specificity values and other reasonable assumptions, to show that approximately one third of antepartum stillbirths might have been prevented by a new method for monitoring of all high-risk pregnancies. A sensitivity analysis was used to test this conclusion over a range of possible test performance values. The implications of these findings for clinical trials and cost-utility analysis are discussed.