ARRS 2022 Abstracts


E1103. Discordance in Breast Biopsies: An Analysis of Contributing Variables
  1. Johnny Xu; University of California - Los Angeles
  2. Bill Zhou; University of California - Los Angeles
  3. Lucy Chow; University of California - Los Angeles
  4. Bo Li; University of California - Los Angeles
Image-guided core needle biopsy is a less invasive method than excisional biopsy. However, sampling error can occur. In discordant cases where there is concern for missed malignancy, additional evaluation such as imaging, repeat biopsy, or excisional biopsy is performed. These additional steps cause patient anxiety and financial burden. Our objective is to determine factors that contribute to discordant results.

Materials and Methods:
Our institutional database was searched for discordant biopsies performed between 2016-2021. Patient demographics, imaging features, biopsy method, and pathology findings were recorded. Discordant rates were calculated. True discordance was defined as additional evaluation that resulted in a final malignant pathology. False discordance was defined as additional evaluation that resulted in a final benign pathology. Means between the true and false discordance groups were assessed for statistical significance using IBM SPSS, with Pearson’s chi-squared test and Fischer’s exact test used for categorical variables and two-sided Student’s t-test for numerical variables. Linear regression was used to generate a receiver operating characteristic curve for modeling likelihood of true discordance.

Out of a total of 7635 biopsies performed during the 5-year period, 155 biopsies were discordant (0.02%), of which 32 patients were lost to follow up. Of the 123 patients who underwent additional evaluation, 30 were true discordant (24.4%) while 93 were false discordant (75.6%). Statistically significant variables between the two groups include history of prior or concurrent breast atypia (p=0.02), size of finding (p = 0.04), presence of axillary lymphadenopathy (p=0.03), BI-RADS score (p=0.01), and biopsy method (p=0.04). Non-significant variables include age, IBIS score, history of non-breast malignancies, history of prior breast procedures, family history of breast malignancies, breast density, lesion type, and attending experience. A ROC curve using BI-RADS score 4C or greater and history of prior or concurrent atypia has an area under curve value of 0.734. US-guided biopsies had a higher rate of true discordance compared to stereotactic-guided and MR-guided biopsies (52.6%, 14.8%, 25.0%, respectively). This is likely due to technical challenges and smaller needle size with US-guided biopsies, which limit accurate sampling on the initial biopsy. Of the different lesion types, calcifications had the highest rate of true discordance (50.0%), with 15 out of the 21 (71.4%) discordant cases caused by lack of calcifications seen on pathology. The most common final pathologies in false discordance cases were stromal fibrosis, fat necrosis, and radial scar. Of the 93 benign cases, 33 patients had high-risk lesions on final pathology (35.5%). 42 had patients additional breast MRI as part of their workup, and 7 (16.6%) patients had their BI-RADS score downgraded as a result, obviating the need for additional biopsy.

This study emphasizes the need for careful radiologic-pathologic review to help guide the radiologist in the setting of discordance.