Impact of HER2 Status on Risk of Sentinel Node Metastasis: An Independently Validated Multivariable Model.

Abstract

Abstract Purpose The presence of axillary lymph node metastases is dictated by tumor biology and is a strong prognostic indicator in breast cancer. We have previously published a user-friendly nomogram that provides a risk estimate for sentinel lymph node (SLN) metastasis in women with breast cancer (Bevilacqua et al. 2007, J Clin Oncol 25: 3670). At that time, HER2 testing was not uniformly performed, but is now standard of care. The purpose of this study was to determine if the addition of HER2 status or grouping of patients by molecular subtype improves the prediction of SLN metastasis.Patients and Methods: The ability of clinical and pathologic features to predict the presence of SLN metastasis in patients presenting with clinically node negative invasive breast cancer, was assessed with multivariable logistic regression (MVA) for 4723 sequential SLN biopsy procedures with known ER, PR, and HER2 from 1996 to 2004. HER2 status was defined as positive if IHC=3+ and/or FISH≥2. The modeling (n=3297) and validation (n=1426) groups were identified by simple random sampling. Two models were created using the modeling population: one included ER, PR, and HER2 as separate variables and one combined these markers into 4 subtypes defined as: Luminal A-like=ER or PR +, HER2 −; Luminal B-like=ER or PR +, HER2 +; HER2-like=ER and PR −, HER2 +; Basal-like=ER and PR −, HER2 −. The validation group was used to assess the calibration (intercept, slope, Emax, Eavg) and discrimination (AUC, area under the receiver operating curve) of the models.Results: In addition to age, tumor size, tumor type, lymphovascular invasion, tumor location, and multifocality, subtype was found to be an independent predictor of SLN metastasis on MVA (p=0.003). The Basal-like subtype was associated with a significantly lower risk of SLN metastasis (compared with referent Luminal A-like subtype, OR=0.58). HER2 alone was not found to be an independent predictor of SLN metastasis (p=0.56), while ER and PR remained significant (p=.05, p=.02). Compared to our previous model without HER2, the AUC was slightly decreased by the addition of either HER2 or subtype into the model, but the calibration was slightly improved (Table 1). DiscriminationCalibrationModelAUCInterceptSlopeEmaxEavgPrevious Model0.735- 0.0440.8690.0390.020Mew model with HER20.733- 0.0030.8890.0270.019New model with subtype0.731- 0.0010.8870.0270.020Table 1: Discrimination and calibration measures comparing the new model (with HER2 alone or subtype) with the previously published model (without HER2 or subtype).Conclusion: The addition of subtype, defined by combining HER2 status with ER and PR, slightly increases the calibration of the new model but does not increase its discrimination compared to the previous model. Breast tumor subtype is a significant independent predictor of risk of SLN metastasis, with basal-like subtype having a lower risk of SLN metastasis. Citation Information: Cancer Res 2009;69(24 Suppl):Abstract nr 1004.