The evolution of mammography has provided possibilities for mass screening programs. Early diagnosis is considered the most important factor in reducing the breast cancer mortality but the mass screening is very expensive if we include all female population. In this paper we show the results of an early diagnosis pilot study with a multivariate data analysis. Data concerning risk factors (age, age at menarche, menopause, parity, age at first childbirth, lactation, abortions and previous benign breast disease) were recorded in 438 patients with breast carcinoma and 1750 patients with benign breast diseases diagnosed in the Early Diagnosis Pilot Program at the National Institute of Oncology and Radiobiology in Cuba. A group of 449 healthy women living in Havana City was also studied. Age and age at first childbirth were the major factor considered. Multivariate data analysis allowed to build stratification trees identifying subgroups with different breast cancer incidence. The usefulness of these stratifications for screening with optimal coverage, sufficiency and efficacy, is discussed.