RESUMEN
Twenty two view mammograms without magnification, with microcalcifications corresponding to non palpable breast lesions, were submitted to 23 experienced observers, either gynaecologists or radiologists. These mammograms consisted of 10 malignant lesions and 10 benign lesions; all of them underwent surgical procedures. The microcalcifications were graded and categorized into one of the four groups: malignant; suspect; slightly suspect; benign. One of the four options was proposed: biopsy; mammogram within 3 months; mammogram within 6 months; mammogram within a year. Twelve observers out of 23 have referred to a classification. The mammograms were assessed according to a consensus. The validity of that diagnostic test was studied in various clinical situations and the lack of homogeneity of the responses was quantified. In the trade-off situation between sensitivity and specificity (suspect or malignant considered as positive), sensitivity is 50% and specificity is 70%. Moreover, the responses are not homogeneous at all and this lack of homogeneity is found to be statistically significant, greater than by chance alone. The diagnosis of cancer is more frequently put forward by the radiologists. This study explains the low predictive value of non palpable microcalcifications undergoing surgical procedures: from 11.5 to 44% in the literature. Similarly, there is no agreement on the frequency of the follow-up for these microcalcifications when they do not lead to surgical procedures.