Background and purpose: The 21-gene assay has been recommended by guidelines for helping clinicians decide whether adjuvant chemotherapy is needed in estrogen receptor-positive

lymph node-negative breast cancer patients. However

as an invasive and expensive test

the Oncotype DX 21-gene test can’t fully reflect the gene expression of the overall heterogeneous tumors. The 21-gene test has not been routinely used clinically. Mammography has been widely used to detect

diagnose and evaluate the treatment effect of breast diseases

but it has been unclear whether mammography has prognostic value. This study aimed to explore the association between mammographic imaging features and recurrence risk quantified with 21-gene recurrence score (RS) in breast cancer patients. Methods: A total of 529 patients who underwent a 21-gene expression assay from Apr. 2017 to Mar. 2019 were identified. Mammographic imaging features were reviewed and recorded according to the 2013 version of the Breast Imaging Reporting and Data System (BI-RADS) lexicon. One-way analysis of variance and Spearman’s correlation test were used to evaluate the correlation between mammographic imaging features and the RS. Multivariate linear regression was used to assess the added effect of significant mammographic features. Results: Mammographic calcification distribution and mass margin were significantly associated with the RS (P0.001 and P=0.005). The average RS for regional calcification distribution was 39.9

which was obviously higher than that for segmental and grouped calcification distributions

corresponding to 24.4 and 31.8

respectively. Mass with circumscribed margins had a lower average RS of 17.4

whereas the RS for all other mass margins was 27.6. When these two significant features were added to a multivariable regression model including pathological factors

the mass margin remained statistically significant

and the R-squared value increased from 0.170 to 0.193. Conclusion: Calcification distribution and mass margin upon mammography have the potential to predict a breast cancer patient’s recurrence risk. Further prospective studies are needed to verify this preliminary finding.