Sherman, M. E.

Abstract Uptake of tamoxifen is limited to 5-15% of all eligible women due to concerns regarding medication side effects combined with the lack of ability to predict which women will benefit from the therapies. We used a unique retrospective sample of n=92 high-risk women with paired benign biopsies obtained before (PreTam) and during tamoxifen (OnTam) to discover tissue gene expression biomarkers of biologic response, defined as change (Δ) in lobular Ki67. Ki67 was digitally quantified in normal lobules from PreTam and OnTam benign tissue samples and samples with PreTam Ki67 = 0 were excluded. Data were analyzed from 51 women with both PreTam and OnTam Ki67 (median age = 49 years and median BMI = 25.5 kg/m2 at PreTam biopsy). OnTam biopsies were collected a median of 11.5 months after PreTam biopsy (median 6.5 months after starting tamoxifen; dose 20 mg/day in 86%). Median Ki67 significantly decreased overall (Δ Ki67 = -0.29%, p=0.003), but the effect was restricted to women with baseline Ki67 ≥ 0.5% (n=38, Δ Ki67 = -0.78%, p=0.0002); no significant change occurred with baseline Ki67 < 0.5 (Δ Ki67 = 0.19%, p=0.23). Δ Ki67 was inversely proportional to baseline Ki67. RNA-seq was successful for 171 benign tissue samples (including 53 PreTam-OnTam pairs). After edgeR normalization (>14,600 genes), PreTam vs OnTam comparison across 33 pairs with baseline Ki67 ≥ 0.5% yielded 66 differentially expressed genes (fold-change > 2 and p<0.01). At p<0.01, ranked gene set enrichment analysis revealed 27 gene sets enriched for reduced expression with Tam (with top-ranked gene set related to steroid hormone biosynthesis) and 59 gene sets enriched for increased expression with Tam (with top-ranking gene sets related to B cell and T cell receptor signaling, regulation of cell cycle, and other immune-related pathways). To identify predictors of response, we correlated baseline PreTam gene expression with Δ Ki67 in 37 women with available baseline PreTam RNAseq and Δ Ki67 data. Partial correlation analysis (via ‘ppcor’ package, adjusted for library qubit) revealed 73 genes with significant positive correlation (rho > 0.50) and 14 genes with significant negative correlation (rho < -0.50) with Δ Ki67 (p<0.01). At p<0.01, ranked gene set enrichment analysis yielded 67 gene sets enriched with negative correlation (i.e. increased expression associated with a reduction in Ki67), including pathways related to ribosome biogenesis, DNA recombination, and nucleic acid metabolic processes; there were 62 gene sets enriched for positive correlation (i.e. decreased expression associated with a reduction in Ki67), including pathways such as VEGF and ERBB signaling. Our data support (i) baseline Ki67 ≥ 0.5 % and (ii) an 87-gene baseline signature as complementary biomarkers of tamoxifen response. Prospective validation of this combined assay could target tamoxifen to the women most likely to benefit, potentially overcoming a major barrier to chemoprevention uptake. Citation Format: S. J. Winham, C. Wang, Y. Liu, B. M. McCauley, D. J. Tabrizi, D. L. Gehling, J. L. Fischer, L. R. Seymour, M. H. Solanki, M. E. Sherman, K. R. Lohani, T. L. Hoskin, D. C. Radisky, A. C. Degnim. Gene expression signatures of beneficial response to tamoxifen for breast cancer prevention [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS3-01-18.

Abstract Background: Artificial intelligence (AI) algorithms based on deep learning approaches show promise in improving breast cancer (BC) detection on mammography and may also improve prediction of future BC risk compared with clinical risk prediction models. Historical clinical risk prediction models underperform among women with BBD; however, AI models for BC detection on mammography have not been tested among women at elevated BC risk due to benign breast disease (BBD). We evaluated if an AI cancer detection algorithm can aid BC risk prediction among women with BBD. Methods: We examined women without a prior BC who had an initial BBD biopsy at Mayo Clinic, Rochester, between 2002 and 2013. Incident BC occurring after BBD was identified using the Mayo Tumor Registry and supplemented by follow-up questionnaires. Only women with a full field digital mammogram at least 2 years prior to BC diagnosis were eligible. A breast pathologist evaluated BBD according to increasing severity, as non-proliferative (NP), proliferative disease without atypia (PDWA) or atypical hyperplasia (AH). AI malignancy scores derived from the Transpara detection algorithm (1-10) and volumetric percent density (VPD) from Volpara (per one standard deviation, SD) were assessed from mammograms close to BBD diagnosis. Cox proportional hazards regression models were applied to estimate hazard ratios (HRs) with 95% Confidence Intervals (CIs), adjusted for age and BMI. C-statistics were estimated to assess the contribution of each factor to BC risk prediction. Likelihood ratio tests (LRTs) and bootstrapping methods were used to compare model performance with the addition of AI malignancy score. Results: The BBD cohort with mammography consisted of 3,125 women followed for a median 12.9 years, with 250 incident BC. Of these, 221 BC occurred at least two years after the mammogram and were used in analyses. As expected, increased BC risk was associated with BBD severity [for AH HR=3.18 (95%CI: 2.04, 4.95) and for PDWA HR=1.59 (95%CI: 1.17, 2.14) compared to NP] and with higher VPD [HR=1.24 per SD (95% CI: 1.05, 1.45)] (Table ). The AI-malignancy score alone was also associated with BC risk [HR=1.16 per 1 unit score (95%CI: 1.09,1.23)] and showed similar discriminatory accuracy [C-statistic=0.626 (95%CI: 0.586, 0.665)] as the model with BBD severity and VPD combined [C-statistic=0.627, (95%CI: 0.585, 0.669)] (Table ). In models with BBD severity and VPD, the AI-malignancy score was an independent risk factor for BC [HR=1.14 (95%CI: 1.07, 1.21); PLRT<0.001] but only achieved a marginally significant improvement in discriminatory accuracy [C-statistic=0.651 (0.609, 0.693), ΔC-statistic=0.024 (95% CI: 0.000, 0.047)]. Conclusion: In this preliminary study, an AI BC detection algorithm achieved similar performance in predicting future BC risk compared to established risk factors. Citation Format: C. M. Vachon, S. Winham, M. Jensen, D. Hursh, L. Pacheco-Spann, A. Norman, J. Fischer, S. Schrup, L. Seymour, D. Gehling, S. Nyante, M. Troester, N. Karssemeijer, R. Vierkant, D. Radisky, C. Scott, S. I. Maimone, A. Degnim, M. Sherman. An AI algorithm for breast cancer detection improves future BC risk prediction among women with benign breast disease [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PD4-02.