Online Poster Portal

  • Author
    Johnny Xu
  • PI

    Lucy Chow MD, Bo Li MD

  • Co-Author

    Bill Zhou MD

  • Title

    Discordance in Breast Biopsies: An Analysis of Contributing Variables

  • Program


  • Other Program (if not listed above)

  • Abstract

    Introduction: Image-guided core needle biopsy is less invasive than excisional biopsy. However, sampling error can occur yielding benign results that do not match with what was identified in previous breast imaging. These are discordant cases and they highlight a concern for missed malignancy. Thus, additional evaluation such as imaging, repeat biopsy, or excisional biopsy is performed. The purpose of this study is to determine the factors that influence whether the discordant results, upon additional biopsy, change to malignant or remain benign. We define true discordance as a final malignant pathology and false discordance as benign pathology.

    Methods: In UCLA's institutional database, 7635 biopsies were performed between 2016-2021 with 122 yielding discordant results with adequate follow up. In these cases, patient demographics, imaging features, biopsy method, and pathology findings were recorded. These variables were analyzed using R statistical software to assess correlation and to develop a logistic regression model.

    Results: Our institutional discordance rate was low at 2% with a true discordance rate of 24.6%. Among the variables tested, lesion size (mm), BI-RADS, LAD, lesion type, history of atypia, and biopsy method are each significantly correlated with a true discordant lesion. The variables were narrowed down to lesion size (mm), BI-RADS, LAD, and history of atypia when developing the logistic regression mode because biopsy method and lesion type negatively affected the predictive accuracy of the model. The model had a sensitivity of 37% and a specificity of 96%, with a predictive accuracy of 81% (95% CI: (0.73, 0.88)).  An ROC curve has been constructed from the output probability of the model with an AUROC = 77.4%. Removing any variable reduced the AUROC.

    Conclusion: Six variables tested were individually found to be significantly correlated against true and false discordance. Although not all six were useful in predictive modeling, they should still be carefully reviewed in the setting of discordance. Other conclusions that can be made are that US-guided biopsies are more prone to false discordance likely due to the technical challenges and smaller needle size. Additional MR breast evaluation may also prevent patients from needing an invasive excisional biopsy.

  • PDF


  • Zoom