Possible indication of breast cancer risk through discriminant functions

In the Ferrara province of Italy, a screening programme for the detection of breast cancer was initiated in 1965; by 1976, 21,120 women had been examined. The biologic history of the women participating in the program is routinely collected at registration. Several biological variables can be extracted from the history. We used 15 biologic variables in order to construct discriminant functions between women affected by breast cancer and a group of control women. It was found that four variables suffice to separate the group of the women with breast cancer from the control group, although with some degree of uncertainty. The theory was advanced that the individual values of a discriminant function may be associated with the risk of healthy women for developing breast cancer. Data collected after the construction of the discriminant function hint at such an association. We have attempted to calculate a discriminant value for healthy women participating in screening programs, so that those having a discriminant value below a given threshold may be considered at higher risk than those above the threshold. The women in the high-risk group may then be submitted to an appropriate examination schedule.