Models predicting non-sentinel node involvement also predict for regional recurrence in breast cancer patients without axillary treatment

Background: Non-SN prediction models are frequently used in clinical decision making to identify patients that may not need axillary treatment, but these models still need to be validated by follow-up data. Our purpose was the validation of non-sentinel node (SN) prediction models in predicting regional recurrences in patients without axillary treatment. Methods: We followed a cohort of 486 women with favorable primary tumor characteristics and pN0(i+)(sn) or pN1mi(sn) for median 4.5 years. None of the patients underwent axillary treatment. Based on four published non-SN prediction models, the threshold allowing separation into low versus high-risk on non-SN involvement was set at 10%. Results: Overall 5-year regional recurrence rate was 3.0% (SE, +/- 0.1%). Using the Tenon scoring system, 438 low-risk patients had a 5-year regional recurrence rate of 2.3% (+/- 0.8%), and 48 high-risk patients a recurrence rate of 10.1% (+/- 0.4%). The MSKCC nomogram identified 300 low-risk patients with a recurrence rate of 2.8% (+/- 1.1%), Versus 166 high-risk patients with a rate of 3.4% (+/- 0.5%) (20 patients not assessable). The Stanford nomogram identified 21 high-risk patients without recurrence, and 465 low-risk patients with a 3.2% (+/- 0.9%) recurrence rate. A Dutch model discriminated between 384 low-risk patients with a recurrence rate of 2.2% (+/- 0.8%) and 102 high-risk patients with a rate of 6.3% (+/- 2.9%). Conclusion: The Tenon scoring system outperformed the other models as it identified the largest subgroup of patients with low recurrence rate. In patients resembling our cohort we would recommend axillary treatment if they had a Tenon score above 3.5.