Background: A Breast Cancer Nomogram (BCN) for predicting nonsentinel lymph node (NSLN) involvement has been developed and prospectively tested in several series. However, its clinical applicability has never been tested among surgeons.
Methods: The BCN was applied to 209 SLN-positive patients. Its performance was assessed by the area under the receiver-operating characteristic (ROC) curve. Surgeons in Quebec were surveyed to determine the predicted NSLN positivity below which they would not dissect the axilla. The accuracy of the BCN was determined in this clinically relevant range.
Results: The predictive accuracy of the BCN had an area under the ROC curve of 0.687. Almost half of interviewed surgeons treat over 20 breast cancer per year. Fourteen out of 82 surgeons questioned would never leave the patient without a completion axillary dissection after a positive SLN, regardless of the BCN result. Seventy one percent of them would not complete axillary dissection if the prediction of a positive NSLN was </=10%. Only 37 of the 209 patients were in this 10% or less category, with a mean observed rate of positive NSLN of 13% (95% confidence interval [CI], 2-24%).
Conclusion: The global performance of the BCN was fair. A majority of surgeons in Quebec would omit an axillary lymph node dissection (ALND) if the predicted probability of positive NSLN is 10% or less. Although useful, the BCN data should be used with caution at the low end of the scale. Because of some limitations in the performance in this category, other clinical factors and judgment must accompany its use.