Cross-validation of three predictive tools for non-sentinel node metastases in breast cancer patients with micrometastases or isolated tumor cells in the sentinel node

BackgroundWe cross-validated three existing models for the prediction of non-sentinel node metastases in patients with micrometastases or isolated tumor cells (ITC) in the sentinel node, developed in Danish and Finnish cohorts of breast cancer patients, to find the best model to identify patients who might benefit from further axillary treatment.Material and methodBased on 484 Finnish breast cancer patients with micrometastases or ITC in sentinel node a model has been developed for the prediction of non-sentinel node metastases. Likewise, two separate models have been developed in 1577 Danish patients with micrometastases and 304 Danish patients with ITC, respectively. The models were cross-validated in the opposite cohort.ResultsThe Danish model for micrometatases was accurate when tested in the Finnish cohort, with a slight change in AUC from 0.64 to 0.63. The AUC of the Finnish model decreased from 0.68 to 0.58 when tested in the Danish cohort, and the AUC of the Danish model for ITC decreased from 0.73 to 0.52, when tested in the Finnish cohort. The Danish micrometastatic model identified 14–22% of the patients as high-risk patients with over 30% risk of non-sentinel node metastases while less than 1% was identified by the Finish model. In contrast, the Finish model predicted a much larger proportion of patients being in the low-risk group with less than 10% risk of non-sentinel node metastases.ConclusionThe Danish model for micrometastases worked well in predicting high risk of non-sentinel node metastases and was accurate under external validation.