[Cancer and nipple discharge]

The authors, in a retrospective study, have looked at the clinical and paraclinical parameters that suggest that a cancer may be present when there is a nipple discharge. By using logistic regression on these criteria a model has been able to be constructed with four variables (bloody discharge, and associated palpable tumour, a suspicious mammographic appearance and atypical or proliferating cells in cytology). The probability of a cancer being present varies between three per thousand when none of the four variables are present to 99.1% when all four are present. A first outline of the therapy that can be used is suggested.