Multicentre validation of different predictive tools of non-sentinel lymph node involvement in breast cancer.

Sentinel lymph node (SN) biopsy offers the possibility of selective axillary treatment for breast cancer patients, but there are only limited means for the selective treatment of SN-positive patients. Eight predictive models assessing the risk of non-SN involvement in patients with SN metastasis were tested in a multi-institutional setting. Data of 200 consecutive patients with metastatic SNs and axillary lymph node dissection from each of the 5 participating centres were entered into the selected non-SN metastasis predictive tools. There were significant differences between centres in the distribution of most parameters used in the predictive models, including tumour size, type, grade, oestrogen receptor positivity, rate of lymphovascular invasion, proportion of micrometastatic cases and the presence of extracapsular extension of SN metastasis. There were also significant differences in the proportion of cases classified as having low risk of non-SN metastasis. Despite these differences, there were practically no such differences in the sensitivities, specificities and false reassurance rates of the predictive tools. Each predictive tool used in clinical practice for patient and physician decision on further axillary treatment of SN-positive patients may require individual institutional validation; such validation may reveal different predictive tools to be the best in different institutions.