PURPOSE:Achieving a pathologic complete response (pCR) following neoadjuvant chemotherapy (NAC) is strongly associated with improved survival. This study investigates whether bilateral asymmetry of quantitative perfusion parameters in normal parenchyma from ultrafast dynamic contrast-enhanced MRI (DCE-MRI), measured using k-means clustering (KMC) before NAC, can predict pCR in breast cancer patients. MATERIALS AND METHODS:Fifty-six breast cancer patients undergoing NAC with pretreatment ultrafast DCE-MRI (3-9 s/image at 3T) were enrolled. KMC was used to classify tumor and normal parenchymal voxels into five clusters based on maximum enhancement rate (A·α). Ipsilateral-to-contralateral (I/C) ratios of background parenchymal enhancement kinetics (kBPE) and tumor kinetics (kT) were compared between pCR and nonpCR groups. Logistic regression models were developed to predict pCR. Statistical tests included bootstrapping, z-test, chi-square, and Wilcoxon rank-sum. RESULTS:Patients with residual disease showed significantly higher kBPE in the normal-appearing parenchyma of the ipsilateral breast compared to the contralateral side. Parameters including enhancement rate α, A·α, area under the enhancement curve for 30 s AUC30, volume transfer constant Ktran s, and rate constant of contrast transfer, Kep, were significantly higher, while extravascular extracellular space fractional volume, ve, was significantly lower in the ipsilateral breast parenchyma versus contralateral breast parenchyma for women who have residual disease (p < 0.05). A prediction model using kBPE asymmetry alone achieved an area under the curve (AUC) of 0.83. Including tumor kinetics improved the AUC to 0.85. CONCLUSIONS:Bilateral asymmetry of kBPE parameters derived from ultrafast DCE-MRI using KMC before NAC initiation can predict pCR with high accuracy, providing a new minimal-invasive biomarker for treatment response.

Purpose:To develop and validate a multimodal recurrence-risk model integrating histology, genomic testing, and clinical variables. Methods:We developed AI-Path, a whole-slide image biomarker for recurrence prediction trained in CALGB 9344, and validated it in three independent cohorts: TAILORx, a multi-site Chicago cohort, and the MDX-BRCA cohort. We then integrated AI-Path with Oncotype DX Recurrence Score (RS), tumor size, and nodal status into a Cox model, PathClinRS, fit using 60% of cases from TAILORx, with the remaining 40% held out for validation. The primary end point was distant recurrence-free interval. Performance was assessed using Harrell's concordance index (C-index) and Kaplan-Meier analyses. Results:A total of 12,418 patients were included. In TAILORx, AI-Path outperformed RS for distant recurrence (C-index, 0.682 vs 0.647; P = .038), driven by superior prediction of late recurrence (0.656 vs 0.567; P < .001). In node-negative disease, PathClinRS outperformed RSClin in the TAILORx fitting (0.72 vs 0.70; P = .016) and validation sets (0.74 vs 0.70; P = .004). In node-positive disease, PathClinRS outperformed RSClinN+ in Chicago (0.94 vs 0.74; P < .001) and MDX-BRCA (0.71 vs 0.66; P = .004) cohorts. Compared with NATALEE eligibility, PathClinRS identified nearly twice as many high-risk node-negative patients while maintaining a comparable 10-year distant recurrence risk (16.7% vs 16.6% per NATALEE eligibility in TAILORx fitting; 21.0% vs 19.4% in TAILORx validation). PathClinRS identified 68% of intermediate risk premenopausal patients as low-risk with no evidence of chemotherapy benefit, compared to only 36% identified as low risk by standard clinicopathologic criteria. Conclusion:Digital histopathology provides prognostic information complementary to genomic assays and has the potential to personalize therapy beyond existing clinicogenomic tools.

Importance:Active surveillance has emerged as a deescalation strategy for low-risk ductal carcinoma in situ (DCIS) to reduce overtreatment while maintaining favorable outcomes. Emerging data in low-risk DCIS, eg, the COMET trial, have highlighted growing interest in surveillance-based management for carefully selected patients. However, recent clinical adoption and national trends in managing low-risk, hormone receptor (HR)-positive DCIS have not been evaluated in the US. Objective:To examine trends and sociodemographic variations in nonsurgical management and other treatment modalities for low-risk, HR-positive DCIS. Design, Setting, and Participants:This cross-sectional study analyzed data from the National Cancer Database from January 1, 2004, to December 31, 2022, and included patients aged 18 years or older with grade 1 to 2, HR-positive DCIS and at least 12 months of follow-up since initial diagnosis. Analyses were performed between January 10 and August 31, 2025. Exposures:Year of diagnosis and sociodemographic characteristics. Main Outcomes and Measures:Nonsurgical management, lumpectomy alone, lumpectomy plus adjuvant radiotherapy, unilateral mastectomy, bilateral mastectomy, and endocrine therapy were measured using descriptive statistics. Results:A total of 316 590 female patients were included (mean [SD] age, 60.8 [12.0] years; 5.8% Asian or Pacific Islander, 13.9% Black, 6.1% Hispanic, 73.3% White, and 0.9% other race and ethnicity). From 2004 to 2022, nonsurgical management increased from 2.1% to 3.5%, bilateral mastectomy increased from 4.1% to 8.7%, and lumpectomy increased from 22.0% to 25.1%, while lumpectomy plus adjuvant radiotherapy decreased from 50.9% to 45.6% and unilateral mastectomy decreased from 20.9% to 17.1%. Nonsurgical management was more common among Black patients and patients with no insurance. Bilateral mastectomy was common in younger, White, and privately insured patients and those who lived in higher-income areas. Endocrine therapy use increased from 2004 to 2020 but declined thereafter. Endocrine therapy was highest after lumpectomy plus adjuvant radiotherapy (69.6%), followed by lumpectomy alone (43.9%), unilateral mastectomy (35.3%), and nonsurgical management (29.2%), with the lowest use in patients younger than 50 years in the no surgery (15.2%) and lumpectomy alone (38.6%) groups. Since 2018, radiotherapy use has increased and become progressively more risk adapted, with increasing use with higher Oncotype DX DCIS scores (low risk, 34.5%; intermediate risk, 63.9%; high risk, 73.1%). Conclusions and Relevance:This cross-sectional study highlights increasing trends and socioeconomic disparities in the nonsurgical management of and the need for precision-based, patient-centered care for low-risk DCIS. Precision prevention may enhance the identification of patients who could benefit most from preventive surgery, prolonged endocrine therapy, or treatment deescalation, paving the way for individualized strategies.

Importance Although adding immune checkpoint inhibitors to neoadjuvant chemotherapy improves outcomes in high-risk early-stage breast cancer, opportunities remain to further enhance response. Dual checkpoint blockade offers a potential strategy to further enhance efficacy. Objective To evaluate the combination of anti−programmed cell death 1 protein (PD-1) cemiplimab and anti−lymphocyte activation gene 3 (LAG-3) added to neoadjuvant therapy in ERBB2 -negative early-stage, high-risk breast cancer. Design, Setting, and Participants The I-SPY2 (Investigation of Serial Studies to Predict Your Therapeutic Response With Imaging and Molecular Analysis 2) is an ongoing randomized clinical platform trial being conducted at multiple US clinical sites including patients with early-stage (II or III) ERBB2 -negative, high-risk breast cancer. Participants, continuously enrolled since 2010, were adaptively randomized from February 2, 2020, to December 9, 2021, to one of several experimental neoadjuvant therapies or control groups based on receptor subtypes defined by hormone receptor (HR), ERBB2 status, and MammaPrint (Agendia Inc) molecular risk, categorized as high (MP1) or ultrahigh (MP2). Data were analyzed from January 1, 2022, to August 5, 2025. Interventions Both groups received weekly paclitaxel for 12 weeks, then doxorubicin and cyclophosphamide followed by surgery; concomitant with paclitaxel, the intervention group also received 4 doses of cemiplimab and fianlimab (PCF) every 3 weeks. Main Outcomes and Measures Pathologic complete response (pCR). Treatments graduated when they achieved 85% bayesian probability of success in a subtype-specific phase 3 trial. Pathway-specific biomarkers were assessed for response prediction. Results A total of 78 participants (mean [SD] age, 47 [39-54] years) were randomized to the intervention group, with 350 participants (mean [SD] age, 48 [39-57] years) randomized to the historical control population. PCF graduated in all clinical signatures, with pCR rates vs control of 44% (95% CI, 34%-53%) vs 21% (95% CI, 17%-25%) in all ERBB2 , 53% (95% CI, 39%-67%) vs 29% (95% CI, 22%-36%) in triple-negative, and 36% (95% CI, 23%-49%) vs 14% (95% CI, 9%-19%) in HR-positive and ERBB2 -negative disease. Among the total participants, 16 (21%) experienced adrenal insufficiency, including hypophysitis (11% grade 3 or 4), mostly occurring after immunotherapy completion. PCF was found to be highly effective in the subset of patients with immune signature positive status (ImPrint positive). Conclusions and Relevance In this randomized clinical trial, the combination of PD-1 and anti−LAG-3 inhibition with standard NAC was effective in early-stage ERBB2 -negative breast cancer, particularly in patients displaying a positive ImPrint immune signature. These results warrant further definitive trials. Trial Registration ClinicalTrials.gov Identifier: NCT01042379

PURPOSE:Despite current standard-of-care endocrine therapy, distant recurrence remains a concern for patients with hormone receptor-positive (HR+)/HER2- early breast cancer (EBC). Understanding individual recurrence risk would aid in clinical decision-making. We used machine learning to identify risk factors and develop recurrence risk prediction models. EXPERIMENTAL DESIGN:Predictor variables were identified by gradient boosting and used to train models on a large, diverse real-world dataset of patients with stage I-III HR+/HER2- EBC obtained from the US-based, electronic health record-derived deidentified Flatiron Health Research Database. An elastic net-penalized Cox proportional hazards model was validated internally with real-world data and externally with data from the NATALEE trial of ribociclib in patients with HR+/HER2- EBC. Prediction and outcome concordance for distant recurrence and treatment effect were analyzed with Harrell's concordance index (C-index) and integrated Brier score; model performance over time was determined by dynamic AUC analysis. RESULTS:The model accurately predicted distant recurrence in the real-world cohort [n = 7,842; C-index: 0.85 (95% confidence interval, 0.8461-0.8598); integrated Brier score: 0.05 (95% confidence interval, 0.0443-0.0495)] over time (AUC >0.7 through 10 years); internal validation and sensitivity analyses confirmed model performance. External validation with the NATALEE nonsteroidal aromatase inhibitor alone arm yielded a lower but still discriminative performance (C-index: 0.66). Training on NATALEE data improved concordance (C-index: 0.70); the NATALEE-trained model predicted a 3.2% reduction in distant recurrence at 48 months with ribociclib treatment in the real-world cohort. CONCLUSIONS:A machine learning model was developed that accurately predicted distant recurrence in HR+/HER2- EBC. The identified predictor variables and developed models may aid in risk-based personalized treatment decision-making.

BACKGROUND:Genomic assays such as Oncotype DX have transformed adjuvant treatment selection for hormone receptor-positive, HER2-negative, early breast cancer but remain inaccessible to many patients because of high cost and logistical barriers. We aimed to develop and validate an artificial intelligence (AI) model that estimates Oncotype DX 21-gene recurrence scores directly from routine histopathology slides and clinicopathological variables. METHODS:In this multicentre, model development and validation study, a multimodal deep-learning model was trained on digital whole-slide images and clinical features using a foundation model pre-trained on 171 189 histopathology slides for predicting Oncotype DX recurrence score. We included slides from patients with hormone receptor-positive, HER2-negative, invasive breast cancers and without scanning artifacts and with at least 100 tissue tiles (1·6 mm2). The model was fine-tuned and validated on the TAILORx randomised trial (8284 patients after quality control). Prognostic and predictive performance was assessed in the TAILORx-test set and externally validated in six independent cohorts (Carmel, Haemek, and Sheba medical centres [Israel], the University of Chicago Medical Center [USA], the Australian Breast Cancer Tissue Bank [Australia], and the Cancer Genome Atlas Breast Invasive Carcinoma project [USA]). FINDINGS:In the TAILORx-test set (n=2407), the AI model classified 1097 (45·6%) patients as low risk, 1021 (42·4%) as intermediate risk, and 289 (12·0%) as high risk. For identifying high genomic-risk disease (recurrence score ≥26), the area under the curve (AUC) was 0·898 (95% CI 0·879-0·913). AI-based risk stratification was prognostic for recurrence-free interval (hazard ratio 2·61 [95% CI 1·68-4·04]), distant recurrence-free interval (2·88 [1·73-4·79]), and disease-free survival (1·32 [0·92-1·89]). Chemotherapy benefit was evident in premenopausal patients classified by AI as being at high risk (0·63 [0·46-0·86]) but absent in postmenopausal patients classified by AI as being at low risk (0·94 [0·78-1·12]). 151 (31·3%) clinically high-risk postmenopausal women (by MINDACT criteria) were reclassified as low AI risk with no chemotherapy benefit. Analysis on external cohorts (5497 patients) showed that the model is transferable to new data with high generalisability (recurrence score ≥26 AUC ranging from 0·858 to 0·903). INTERPRETATION:These findings show that AI applied to routine histopathology can serve as a practical and scalable tool for guiding chemotherapy decisions in hormone receptor-positive, HER2-negative, early breast cancer. This approach has the potential to reduce unnecessary chemotherapy and broaden access to precision oncology, particularly in resource-limited settings where genomic testing remains unavailable or unaffordable. FUNDING:Israel Innovation Authority (Kamin), Zimin Institute for Artificial Intelligence Solutions in Healthcare, Israel Precision Medicine Partnership program, and Israel Cancer Research Fund.

Artificial Intelligence (AI) deployment in healthcare is accelerating, yet governance frameworks remain fragmented and often assume extensive resources. Through a systematic review of 35 frameworks for AI implementation in healthcare (published 2019-2024), we identified seven critical domains of healthcare AI governance. While existing frameworks provide valuable guidance, the resource requirements create barriers for smaller healthcare organizations. To address this gap, we organized key findings from the review to create the Healthcare AI Governance Readiness Assessment (HAIRA), a five-level maturity model that provides actionable governance pathways based on organizational resources. HAIRA spans from Level 1 (Initial/Ad Hoc) to Level 5 (Leading), with specific benchmarks across all seven governance domains. This tiered approach enables healthcare organizations to assess their current AI governance capabilities and establish appropriate advancement targets. Our framework addresses a critical need for adaptive governance strategies that ensure that AI implementation delivers tangible benefits to systems of varying resource levels.

The OncotypeDX 21-gene assay guides adjuvant chemotherapy decisions in early-stage, hormone receptor-positive, HER2-negative breast cancer, but cost and turnaround time limit access. This study presents a deep learning-based approach for predicting OncotypeDX recurrence scores directly from hematoxylin and eosin-stained whole slide images. Our approach leverages a deep learning foundation model pre-trained on 171,189 slides via self-supervised learning, which is fine-tuned for our task. The model was developed and validated using five independent cohorts, out of which three are external. On the two external cohorts that include OncotypeDX scores, the model achieved an AUC of 0.836 and 0.817, and identified 22% and 16.3% of the patients as low-risk with sensitivity of 0.97 and 0.97 and negative predictive value of 0.97 and 0.96, showing strong generalizability despite variations in staining protocols and imaging devices. Kaplan-Meier analysis demonstrated that patients classified as low-risk by the model had a significantly better prognosis than those classified as high-risk, with a hazard ratio of 4.1 (P < 0.001) and 2.0 (P < 0.01) on the two external cohorts that include patient outcomes. This artificial intelligence-driven solution offers a rapid, cost-effective, and scalable alternative to genomic testing, with the potential to enhance personalized treatment planning, especially in resource-constrained settings.

Purpose:To test whether histology-derived gene-expression signatures from routine hematoxylin and eosin slides are prognostic for recurrence and predictive of chemotherapy benefit in early breast cancer. Methods:We conducted a multi-cohort study including CALGB 9344 (anthracycline ± paclitaxel), CALGB 9741 (standard vs dose-dense chemotherapy), a pooled Chicago real-world cohort, and the American Cancer Society (ACS) Cancer Prevention Studies-II and -3. Whole-slide images were processed with a previously described pipeline to generate 61 histology-derived signatures per patient. The primary endpoint was distant recurrence-free interval (DRFI), except in ACS, where breast cancer-specific survival was used. Secondary endpoints include distant recurrence-free survival (DRFS) and overall survival. The most prognostic signature in CALGB 9344, selected by Harrell's C-index, was evaluated in additional cohorts. Signature-treatment interaction was assessed by likelihood-ratio tests. Multivariable Cox models incorporating age, tumor size, nodal status, estrogen/progesterone receptor status, and signature were fit in CALGB 9344 to improve risk stratification. Results:A total of 7,170 patients were included across four cohorts. The top histology-derived signature in CALGB 9344 showed strong prognostic performance for 5-year DRFI (C-index 0.63) and performed well across validation cohorts (C-index 0.60, 0.70, and 0.62 in CALGB 9741, Chicago, and ACS, respectively). The strongest predictive signal for treatment benefit was observed for DRFS. High-risk cases identified by the signature demonstrated greater benefit from taxane in CALGB 9344 (adjusted hazard ratio [aHR] 0.76 for DRFS, 95% CI 0.66-0.88; interaction p=0.028), from dose-dense chemotherapy in CALGB 9741 (aHR 0.69, 95% CI 0.56-0.85; interaction p=0.039), and differential chemotherapy benefit in the Chicago cohort (aHR 0.84, 95% CI 0.59-1.21; interaction p=0.009). Combined clinical-histology models improved risk stratification and identified low-risk groups with a 2%-10% risk of distant recurrence or breast cancer death. Conclusion:Histology-derived signatures from H&E images are broadly prognostic and, unlike clinical factors, may predict chemotherapy benefit.

Importance:Although adding immune checkpoint inhibitors to neoadjuvant chemotherapy improves outcomes in high-risk early-stage breast cancer, opportunities remain to further enhance response. Dual checkpoint blockade offers a potential strategy to further enhance efficacy. Objective:To evaluate the combination of anti-programmed cell death 1 protein (PD-1) cemiplimab and anti-lymphocyte activation gene 3 (LAG-3) added to neoadjuvant therapy in ERBB2-negative early-stage, high-risk breast cancer. Design, Setting, and Participants:The I-SPY2 (Investigation of Serial Studies to Predict Your Therapeutic Response With Imaging and Molecular Analysis 2) is an ongoing randomized clinical platform trial being conducted at multiple US clinical sites including patients with early-stage (II or III) ERBB2-negative, high-risk breast cancer. Participants, continuously enrolled since 2010, were adaptively randomized from February 2, 2020, to December 9, 2021, to one of several experimental neoadjuvant therapies or control groups based on receptor subtypes defined by hormone receptor (HR), ERBB2 status, and MammaPrint (Agendia Inc) molecular risk, categorized as high (MP1) or ultrahigh (MP2). Data were analyzed from January 1, 2022, to August 5, 2025. Interventions:Both groups received weekly paclitaxel for 12 weeks, then doxorubicin and cyclophosphamide followed by surgery; concomitant with paclitaxel, the intervention group also received 4 doses of cemiplimab and fianlimab (PCF) every 3 weeks. Main Outcomes and Measures:Pathologic complete response (pCR). Treatments graduated when they achieved 85% bayesian probability of success in a subtype-specific phase 3 trial. Pathway-specific biomarkers were assessed for response prediction. Results:A total of 78 participants (mean [SD] age, 47 [39-54] years) were randomized to the intervention group, with 350 participants (mean [SD] age, 48 [39-57] years) randomized to the historical control population. PCF graduated in all clinical signatures, with pCR rates vs control of 44% (95% CI, 34%-53%) vs 21% (95% CI, 17%-25%) in all ERBB2, 53% (95% CI, 39%-67%) vs 29% (95% CI, 22%-36%) in triple-negative, and 36% (95% CI, 23%-49%) vs 14% (95% CI, 9%-19%) in HR-positive and ERBB2-negative disease. Among the total participants, 16 (21%) experienced adrenal insufficiency, including hypophysitis (11% grade 3 or 4), mostly occurring after immunotherapy completion. PCF was found to be highly effective in the subset of patients with immune signature positive status (ImPrint positive). Conclusions and Relevance:In this randomized clinical trial, the combination of PD-1 and anti-LAG-3 inhibition with standard NAC was effective in early-stage ERBB2-negative breast cancer, particularly in patients displaying a positive ImPrint immune signature. These results warrant further definitive trials. Trial Registration:ClinicalTrials.gov Identifier: NCT01042379.

Background:Hematoxylin and eosin (H&E) staining is routine in pathology but lacks cellular specificity. Multiplex immunofluorescence (mIF) captures spatial immune relationships in tumors, but cost and complexity limit clinical application. Novel approaches to yield similar information from readily available tumor histology are needed. Objectives:Develop and validate a novel deep learning tool capable of translating standard H&E-stained histopathology images into high-fidelity synthetic mIF images that preserve immune cell information predictive of treatment response in breast cancer. Design:Comparative model evaluation and predictive modeling in a retrospective breast cancer cohort. Methods:Core-needle biopsies from 17 triple-negative breast cancer cases underwent mIF imaging. Hematoxylin and eosin and mIF images for DAPI (nuclei), pan-CK (tumor), CD3/CD4/CD8 (T-cells), and CD20 (B cells) were aligned. A pipeline outperforming standard Pix2Pix and CycleGAN image translation networks was developed, "multiplex Synthetic Immunofluoresence Generated through H&E Translation" (mSIGHT), which integrates a registration network to overcome misalignment between the input and target images. Generated images were evaluated with pixel-level metrics and biological metrics, including cell density and cell-to-cell adjacency. The pipeline was then applied to an external cohort to assess associations between predicted immune features and pathologic response to neoadjuvant chemotherapy. Results:Generated images preserved immune cell distributions and proximity metrics correlated to the ground truth cell counts. In a cohort of 218 breast cancer cases treated with neoadjuvant chemotherapy, predicted density of CD8+ T cells was significantly associated with complete response (adjusted odds ratio 1.89, 95% confidence interval 1.23-2.80, p = 0.002), independent of receptor status, grade, and pathologist TIL annotations. Conclusion:The mSIGHT pipeline enables translation of routine H&E slides into virtual mIF images with interpretable immune biomarkers, offering a scalable and affordable alternative to multiplex imaging. It also identifies immune features predictive of therapeutic response and has the potential to assist in the personalization of neoadjuvant therapy.

Importance:Since 2018, the TAILORx and RxPONDER trials have demonstrated that the 21-gene recurrence score (RS) can be indicative of the benefit of adjuvant chemotherapy in hormone receptor (HR)-positive, ERBB2 (formerly HER2)-negative breast cancer with 3 or fewer positive lymph nodes. However, its applicability to key subgroups with high risk for recurrence, including premenopausal women with positive lymph nodes and racial and ethnic minority individuals, remains unclear. Objective:To assess the temporal patterns of and disparities in adjuvant chemotherapy use in early-stage HR-positive, ERBB2-negative breast cancer by age, genomic risk, and nodal involvement. Design, Setting, and Participants:This retrospective cohort study analyzed clinical data from the 2010 to 2022 National Cancer Database. The cohort included women with stage I to III, HR-positive, ERBB2-negative breast cancer who had undergone a lumpectomy or mastectomy and were eligible for endocrine therapy. Patients were categorized into premenopausal (aged ≤50 years) or postmenopausal (aged >50 years) status. Nodal status (negative or positive) was pathologically confirmed. RS was classified per the TAILORx trial, with RS of 0 to 10 as low genomic risk, RS of 11 to 25 as intermediate genomic risk, and RS of 26 or higher as high genomic risk. Data were analyzed from January 20 to August 11, 2025. Main Outcomes and Measures:Adjuvant systemic therapy, defined as receipt of either endocrine therapy alone or chemoendocrine therapy (chemotherapy plus endocrine therapy), after surgery (lumpectomy or mastectomy). Results:A total of 504 937 women (mean [SD] age, 60.0 [10.7] years; 5.4% Hispanic, 4.3% non-Hispanic Asian or Pacific Islander, 8.1% non-Hispanic Black, 81.3% non-Hispanic White, and 0.9% other race or ethnicity) were included. Among premenopausal patients with node-negative tumors, adjuvant chemotherapy use decreased from 6.5% in 2010 to 0.9% in 2022 for those with low genomic risk and from 29.6% in 2010 to 11.1% in 2022 for those with intermediate genomic risk. However, among premenopausal patients with node-positive disease, chemotherapy use declined from 33.3% in 2010 to 12.7% in 2019 but increased to 25.7% in 2022 for the low genomic risk group. For the intermediate genomic risk group, chemotherapy use declined from 55.8% in 2010 to 38.1% in 2019 but increased to 48.9% in 2022. Among postmenopausal women, chemotherapy use for those with low to intermediate genomic risk continued to decrease from 2010 to 2022 in both node-negative and node-positive disease status. Black women with high genomic risk had lower odds of chemotherapy receipt than White women, regardless of menopausal or nodal status (adjusted odds ratio [AOR], 0.84; 95% CI, 0.78-0.90). Premenopausal Black women with low to intermediate genomic risk also had lower odds of chemotherapy receipt than White women (AOR, 0.85; 95% CI, 0.77-0.94), regardless of nodal status. Conclusions and Relevance:This retrospective cohort study found that adjuvant chemotherapy use almost doubled in premenopausal patients with node-positive tumors and with a low to intermediate genomic risk from 2019 to 2022 but decreased for patients with node-negative disease, coinciding with the publication of the TAILORx and RxPONDER trials. The findings highlight the variability in genomic assay use to facilitate adjuvant therapy recommendations for HR-positive, ERBB2-negative breast cancer.

Rationale and Objectives Pathologic complete response (pCR) following neoadjuvant chemotherapy (NAC) is a key marker of long-term outcomes in patients with triple-negative breast cancer (TNBC). Accurate radiologic assessments of response to NAC using breast MRI and US play an important role in guiding treatment decisions, though the effects of immunotherapy (IO) on radiologic assessments are not well characterized. We examined the discordance between radiologic complete response (rCR) and pCR in patients with TNBC following NAC with or without IO. Materials and Methods This is a retrospective cohort study of patients from a single institution diagnosed between January 2013 and April 2022 with TNBC who underwent dynamic contrast-enhanced MRI or US breast imaging before and after NAC. Radiology reports of post-NAC MRI and US scans were examined for presence of rCR, and these findings were correlated with pathologic stage on surgical resection. Concordance between rCR and pCR was measured using Cohen’s kappa (κ) statistic. Association between the receipt of IO and pCR among patients without rCR on post-NAC MRI/US was measured using Fischer’s exact test. Results We identified 186 patients from the cohort meeting inclusion criteria. 166 (89%) patients had high tumor grade, 174 (94%) patients had ductal histologic subtype, 81 (44%) patients had pCR, and 34 (18%) patients received IO. Among patients not receiving IO, pCR and MRI were moderately concordant (κ = 0.60), and pCR and US were weakly concordant (κ = 0.42). Discordance was pronounced in those receiving IO, with reduced concordance between pCR and MRI (κ = 0.39), and no concordance between pCR and US (κ = 0.05). Among patients with no rCR on MRI, there was a strong association between IO and pCR (p = 0.009) with a poor negative predictive value (NPV) of 50% in patients receiving NAC with IO compared to an NPV of 84% in patients receiving NAC without IO. Conclusion We present data that demonstrate a statistically significant reduction in the NPV of post-NAC MRI in patients with TNBC receiving IO. These findings suggest that MRI following neoadjuvant chemoimmunotherapy may overcall residual tumor in TNBC. Further research is necessary to better characterize any underlying immunotherapy induced inflammatory changes that may explain this discordance.

Deep learning models applied to digital histology can predict gene expression signatures (GES) and offer a low-cost, rapidly available alternative to molecular testing at the time of diagnosis. We optimized transformer-based models to infer GES results and applied this approach to pre-treatment H&E-stained biopsies from 1,940 breast cancer patients treated with neoadjuvant chemotherapy in clinical trial and real-world cohorts. The most predictive histology-derived GES for pathologic complete response (pCR) in the I-SPY2 trial was validated in four external cohorts: CALGB 40601, CALGB 40603, a trial of durvalumab plus CT, and standard-of-care CT-treated patients from the University of Chicago. Among HER2-negative patients, a transformer-based model trained using a signature composed of estrogen-regulated genes, proliferation, apoptosis, and interferon response genes predicted pCR with an AUC of 0.794, outperforming models based on clinical features alone (AUC 0.704, p = 0.001), pathologist TIL assessment, and a model trained directly to predict response from I-SPY2 cases. Tertiles of this signature stratify patients into clinically relevant groups with increasing likelihood of complete response, with pCR rates ≥50% in the top tertile regardless of treatment or hormone receptor status. Additional transformer-based signature models predicted response to specific therapies (but not chemotherapy alone), including a HER2 signaling signature in IO-treated patients, and a claudin-low signature in bevacizumab treated patients. In HER2- cohorts with available gene expression data and histology, models trained on expression data performed similarly to digital histology predictions, but the combination of gene expression and histology outperformed histology alone. These findings suggest that histology-based GES provides additive information to RNA sequencing data and can inform precision treatment selection across breast cancer subtypes.

e13103 Background: Despite recent advances, treatment (tx) of metastatic triple-negative breast cancer (TNBC) remains an important unmet clinical need. Activation of the glucocorticoid receptor (GR) initiates cell survival pathways, leading to chemotherapy resistance. Furthermore, high GR expression is a negative prognostic marker in early-stage, hormone receptor negative breast cancer. We hypothesized that GR antagonism with mifepristone (mif) prior to the administration of cytotoxic chemotherapy would improve efficacy by blocking the potent anti-apoptotic signals mediated by GR activation. Methods: We conducted a phase II, randomized, placebo-controlled, multi-center study of nab-paclitaxel (nab-pac) with or without mif in patients (pts) with locally advanced, unresectable or metastatic TNBC (NCT02788981). Prior tx with nab-pac was not allowed. Two prior chemotherapies allowed in the metastatic setting. Pts were randomized to receive nab-pac 100mg/m 2 on day 1, 8, and 15 of a 28-day cycle with mif 300mg or placebo the day prior and day of each dose of nab-pac. The primary endpoint was progression free survival (PFS). Secondary endpoints included objective response rate (ORR), overall survival (OS), and safety and tolerability. Results: 29 patients were enrolled from September 2017 to July 2021; 16 patients were randomized to nab-pac+placebo and 13 patients to nab-pac+mif. The mean age was 53 years (range 32-73) and 34% of pts were self-reported Black. The median PFS was 3.0 and 3.0 months (mos) in the nab-pac alone and combination arms respectively (hazard ratio [HR] 0.87, 95% CI 0.37 – 2.01, p=0.739]. ORR in the nab-pac and combination arms were 31.5% and 23%, respectively; both arms had 1 complete response (CR). Median OS with nab-pac alone was 6.0 mos and in the combination arm was 9.0 mos (HR=0.67, 95% CI 0.29 – 1.16, p=0.350]. The pt on the mif arm who achieved a CR completed 50 cycles and ultimately died of a non-cancer, non-tx related event. Tx was generally well-tolerated, with a safety profile comparable with nab-pac monotherapy. The most common treatment-related adverse events (TRAEs) were fatigue (50%), neuropathy (42%), and neutropenia (42%); the most common grade 3 TRAE was neutropenia. Grade 3 neutropenia occurred more frequently in pts receiving nab-pac+mif (69% vs 13%). Conclusions: This study did not achieve the desired statistical power due to poor accrual attributed to the COVID-19 pandemic and the approval of checkpoint inhibitors in advanced TNBC. Nonetheless, in this subset of patients, the addition of mif to nab-pac did not significantly improve PFS compared to nab-pac alone. There was a trend towards improvement in OS, primarily driven by one long-term responder. GR activation remains relevant in advanced TNBC. Further investigation of this pathway and other strategies to target chemotherapy resistance are needed. Clinical trial information: NCT02788981 .

Background: Early-stage hormone receptor positive (HR+) human epidermal growth factor receptor 2 negative (HER2-) breast cancer is typically treated with endocrine therapy with or without chemotherapy. While several commercially available genomic assays exist to assess risk of recurrence and guide adjuvant systemic therapy selection, these platforms are time consuming, costly, and may not fully capture disease heterogeneity when performed on a single tissue block. Radiographic imaging allows for a global assessment of tumor heterogeneity, but new approaches are needed to develop effective prognostic imaging tools. Therefore, we developed and validated an AI-based model incorporating pre-treatment DCE-MRI features with clinicopathologic features to predict recurrence risk in HR+/HER2- breast cancer. Methods: The model was trained on 522 women (29.5% neoadjuvant chemotherapy), pooled from a multi-institutional clinical trial and one independent institution, and validated on 496 women (31.2% neoadjuvant chemotherapy), pooled from three independent health systems. All patients had HR+/HER2- breast cancer and pre-treatment DCE-MRI. The model inputs included clinicopathological data (age, race/ethnicity, T stage, N stage, grade) and DCE-MRI features that capture the spatial characteristics of the tumor and surrounding tissues. These image-based features are computed using a validated AI-based segmentation model developed for early-stage breast cancer. Model outputs are thresholded to optimize patient stratification in the training set, resulting in categorical outputs of high or low risk. Finally, we assessed the added prognostic value of our model trained with both clinicopathologic and imaging features against the model trained with clinicopathologic features alone. Results: In the validation cohort, patients had a median follow-up time of 4.6 years, median age of 53 years, and 16% were African American. The majority (90.3%) of women had T1-T2 disease, and 34.9% had 1-3 involved lymph nodes. The 5-year recurrence free survival (RFS) for women predicted as low risk by the model was 93.1% (95% CI: 89.0, 95.6%) compared to 78.0% (95% CI: 67.3, 85.6%) for women predicted as high risk. The unadjusted hazard ratio (HR) for the predicted high vs low recurrence risk groups was 3.8 (95% CI: 2.0, 7.2) at 5 years. After adding age, race/ethnicity, T stage, N stage, and grade into a Cox model, the adjusted HR for high vs low risk groups was 3.2 (95% CI: 1.6, 6.6). The prognostic signature was added to the clinical-only risk score both as a continuous variable and as risk groups, with significant improvement in RFS prediction (p<0.01) over the clinical-only model, as assessed using a likelihood ratio test. Subgroup analyses indicated the model was prognostic regardless of age or nodal status. In women less than 50 years, the 5-year RFS for the low risk group was 89.1% versus 75.9% in the high risk group (HR: 2.8, 95% CI: 1.1, 7.1). In women greater than 50 years, the 5-year RFS was 94.6% in the low risk group vs 80.3% in the high risk group (HR: 4.4, 95% CI: 1.8, 10.6). Effective risk stratification into low risk and high risk was accomplished in both the lymph node negative (HR: 3.8, 95% CI: 1.6, 9.1) and lymph node positive patients (HR: 4.0, 95% CI: 1.5, 11.1). Conclusions: Our AI-based prognostic tool incorporating pre-treatment DCE-MRI allows personalized treatment planning in real time in women with early-stage HR+/HER2- breast cancer. The prognostic benefit exceeded that of clinical features alone and was observed regardless of age and lymph node involvement. Further studies are ongoing to assess the ability of the model to identify patients most suitable for therapy escalation (e.g., chemotherapy, CDK 4/6 inhibitors) or de-escalation (endocrine monotherapy). Citation Format: Poornima Saha, Frederick M. Howard, Erica M. Stringer-Reasor, Yara Abdou, Jennifer McMahon, Yuhan Zhang, Joseph R. Peterson, Bradley Feiger, Michelle Weitz, Judy C Boughey, Matthew P Goetz. Leveraging AI to Predict Recurrence-Free Survival in Breast Cancer Patients through Image-Based assessment of Tumor Characteristics [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2024; 2024 Dec 10-13; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(12 Suppl):Abstract nr PS11-06.

549 Background: Breast cancer (BC) treatment selection is traditionally guided by clinical characteristics. However, as clinical characteristics cannot capture the complexity of a disease, genomic tools have been developed. Recent advances in artificial intelligence (AI) have allowed pathology imaging to be used to build more accurate and comprehensive prognostic/predictive models. In this study, we validated an AI test, powered by a pan-cancer histopathology foundation model, that integrates digital pathology images with clinical variables to predict breast cancer recurrence. Methods: The Ataraxis AI prognostic model (ATX) was developed using 4,659 stage I-III BC patients from 10 distinct cohorts. Ataraxis AI platform first extracts novel morphological features from digitized H&E slides using a pre-trained AI foundation model. These morphological features are then integrated with common clinical characteristics, such as TNM staging, ER/PR/HER2 status, age at diagnosis, or lobular or ductal histology to generate a risk score between 0 and 1. We evaluated ATX on 3,502 patients from 5 external cohorts, including 858 patients with available Oncotype DX (ODX) scores. The primary endpoint of this study was disease-free interval (DFI), defined as the time until first recurrence, with deaths prior to recurrence censored. Results: Across 3,502 patients spanning five validation cohorts, ATX accurately predicted DFI with a C-index of 0.71 [0.68-0.75] and hazard ratio (HR) of 3.63 [3.02-4.37, p < 0.01], computed for every 0.2 unit increase in the test score. Compared to ODX (n = 858), the ATX was more accurate, achieving a C-index of 0.67 [0.61-0.74] versus 0.61 [0.49-0.73]. Additionally, ATX added independent prognostic information to ODX in a multivariate analysis (HR: 3.11 [1.91-5.09, p < 0.01]). ATX demonstrated robust accuracy in TNBC (n = 230, C-index: 0.71 [0.62-0.81], HR: 3.81 [2.35-6.17, p = 0.02]) and HER2+ (n = 353, C-index: 0.67 [0.55-0.80], HR: 2.22 [0.99-5.01, p = 0.05]) groups. Conclusions: (1) ATX is predictive of breast cancer recurrence, (2) ATX improves upon the accuracy of ODX, (3) ATX demonstrates robust performance in all main BC subtypes. ATX evaluated across 5 cohorts individually and pooled, for both Harrell’s C-index and hazard ratio. Cohort N C-index HR Karmanos 168 0.62 [0.49-0.75] 3.82 [1.33-10.98, p=0.01] Basel 269 0.67 [0.58-0.77] 3.98 [1.92-8.25, p<0.01] TCGA 911 0.70 [0.63-0.77] 3.0 [2.1-4.28, p<0.01] Providence 1733 0.74 [0.7-0.79] 4.02 [3.09-5.23, p<0.01] Chicago 421 0.70 [0.60-0.80] 3.25 [1.45-7.31, p<0.01] Pooled 3502 0.71 [0.68-0.75] 3.63 [3.02-4.37, p<0.01]

587 Background: GES predictive of response to therapy across multiple breast cancer subtypes are commercially available or in development. Deep learning models can predict GES from digital histology, and may serve as a lower-cost alternative immediately available at the time of biopsy. Methods: Transformer-based models trained to predict 38 distinct breast cancer signatures from pathology (all with Pearson correlation > 0.5 versus true GES) were previously developed using cases from The Cancer Genome Atlas. These models were applied to digital H&E from pre-treatment biopsies from HER2- cases treated with CT or CT + immunotherapy (IO) from the ISPY2 trial. The histology-derived GES most predictive of pCR in ISPY2 (as per area under the ROC curve [AUROC]) was tested in two external neoadjuvant cohorts - a subset of a trial from Yale of durvalumab + CT (NCT02489448) with TIL annotations, and patients receiving standard of care CT at University of Chicago. AUROC significance was assessed with 1000x bootstrapping, with Benjamini Hochberg correction applied in ISPY2 to account for testing multiple GES models. Tertiles of predicted expression calculated in ISPY2 defined groups with low, medium, and high likelihood of pCR; these cutoffs were tested in the external cohorts. Results: Accuracy for pCR prediction was tested in 578 patients from seven arms of ISPY2 – with breakdown by treatment and hormone receptor (HR) status shown in Table. A histology model for a GES defined by estrogen regulated genes (Oh et al, JCO 2006) – including proliferation, apoptosis, and interferon-response genes – predicted pCR with the highest AUROC (0.794) in ISPY2, and outperformed a logistic regression fit on grade, HR status, and tumor / nodal stage (AUROC 0.705, p for comparison 0.0001). Tertiles of predicted expression for this GES (computed in ISPY2) identified groups with low / high pCR rates which were robust to treatment, HR status, and consistent in validation cohorts (Table). This digital signature (AUROC 0.737) compared favorably to pathologist TIL annotation (AUROC 0.664) from the external Yale cohort. An explainability tool demonstrated that patterns of lymphocytic infiltrate and poor differentiation contributed to high signature predictions from histology. Conclusions: A digital histology-derived GES consistently identifies patients at low / high likelihood of pCR with neoadjuvant CT or CT + IO, and may improve treatment personalization. Subgroup n AUROC p % pCR (low expression) % pCR (mid expression) % pCR (high expression) ISPY2 579 0.794 2 x 10 -28 7.6 26.7 58.6 ISPY2, CT + IO 459 0.810 3 x 10 -26 8.0 28.2 64.1 ISPY2, CT only 120 0.726 0.001 6.4 20.0 36.8 ISPY2, HR- 239 0.704 4 x 10 -7 14.8 29.2 58.4 ISPY2, HR+ 340 0.817 1 x 10 -15 6.5 24.8 59.0 UChicago HR- 151 0.746 3 x 10 -7 11.1 27.3 55.1 UChicago HR+ 63 0.847 1 x 10 -5 5.5 12.0 70.0 Yale HR- 41 0.737 0.005 0.0 50.0 61.5