This work presents a data mining effort to discover pure ‚or almost-pure clusters with respect to atherosclerosis risk factors, from a medical database used by the STULONG project. One originality of this work is to produce overlapping clusters with two recents algorithms: ECCLAT and PoBOC. Such clusters, described by social characteristics and physical and biochemical examinations on patients, allow to characterize patients affected by disease due to atherosclerosis, and may lead to relevant factors. We compare the two algorithms, and we observe if the results point out the role of some examinations.