Results of the multivariable analysis of DR for the image-only model.

Calibration performance of models for 10-year distant recurrence risk in the NSABP B-42 translational cohort.

A, Overview of the M3T model workflow. The model was developed using the NSABP B-42 translational cohort (N = 2,271) with five-fold cross-validation, integrating H&E image–derived features and selected tabular clinicopathologic features through the M3T (multimodal, multitask transformer). An auxiliary task predicting the lowest BMD T-score enhanced prognostic performance and improved stratification of patients with differential ELT benefit but was not required for model inference. The model outputs a continuous risk score that is dichotomized into low- and high-risk groups. Independent external validation was performed in the TAILORx translational cohort (N = 4,300). Prognostic, predictive, and calibration analyses were conducted to evaluate model performance. B, End-to-end image preprocessing, feature extraction, model training, and output architecture. Tissue regions were detected from H&E-stained WSIs, tiled at 10× magnification, and processed using CTransPath to generate tile-level image embeddings. Tile-level image embeddings and clinicopathologic features were projected into a shared representation space and processed through transformer encoder and decoder modules. During training, an auxiliary task predicting the lowest BMD T-score was incorporated to improve prognostic performance and stratification of patients with differential ELT benefit. The primary task generated a continuous risk score for late DR risk stratification. The dashed box shows the detailed architecture of the transformer encoder (left) and transformer decoder (right).

Predictive performance comparison of Clarity models with Breast Cancer Index (BCI).

Triple-negative breast cancer (TNBC) is an aggressive subtype with an activated tumor immune microenvironment. The multicenter, multinational, double-blinded NSABP B-59/GeparDouze trial evaluated the addition of atezolizumab (atezo) (773 patients randomized) or placebo (777 patients) to sequential taxane-carboplatin-anthracycline-based neoadjuvant chemotherapy in stage II-III TNBC. The addition of atezo did not significantly improve the primary endpoint of event-free survival (EFS) (HR, 0.80 (95% CI, 0.062-1.03); stratified log-rank P = 0.083, 4-year EFS rates difference 3.3%). The HR for overall survival was 0.86 (95% CI, 0.62-1.19), with a 4-year benefit of 0.7%. Prespecified subgroup analyses suggested heterogeneity in EFS, with benefit of atezo in patients presenting with clinical lymph node involvement (Pinteraction = 0.039). Treatment-emergent adverse events with grades ≥3 were reported in 75.3% (atezo) versus 73.4% (placebo), and immune-related adverse events were reported in 27.6% (atezo) versus 11.4% (placebo). A total of 196 (25.5%) patients discontinued atezo and 143 (18.8%) patients discontinued placebo in the neoadjuvant phase. In an exploratory mRNA-based subset analysis, patients with basal-like immune-activated tumors may have benefited from atezo. TNBC subtyping to identify basal-like immune-activated tumors and quantification of tumor-infiltrating lymphocytes to identify tumors with high tumor-infiltrating lymphocyte counts could be a promising strategy to identify patients who benefit from the addition of immune checkpoint inhibitors to neoadjuvant chemotherapy. ClinicalTrials.gov registration: NCT03281954 .

Calibration plots showing observed versus predicted 10-year distant recurrence risk.

Predictive performance of the MI Clarity M3T model for ELT benefit across clinical subgroups.

Model interpretation using tile-level embeddings and whole-slide spatial heatmaps. A, UMAP visualization of tile-level histopathology embeddings extracted from H&E-stained WSIs. Points are colored by unsupervised embedding cluster and shaped according to model-defined risk group. B, Distribution of model-predicted risk scores across embedding clusters. Boxplots indicate the median and interquartile range, with pairwise comparison P values shown. C, Representative image tiles from each embedding cluster, illustrating morphologic patterns captured by the image feature extractor.

Results of the multivariable analysis of DR for the M3T model.

Abstract Late distant recurrence (DR) remains a persistent risk in hormone receptor–positive (HR+) early breast cancer after completion of 5 years of endocrine therapy (ET). We developed and validated a multimodal artificial intelligence (AI) model to improve long-term risk stratification and to explore heterogeneity in benefit from extended letrozole therapy (ELT). The deep learning model integrating digitized hematoxylin and eosin whole-slide images with clinicopathologic variables was developed using 2,271 patients from the National Surgical Adjuvant Breast and Bowel Project (NSABP) B-42 trial with five-fold cross-validation and externally validated in 4,300 patients from the TAILORx trial who were disease-free at 5 years from the initial diagnosis. Prognostic performance was evaluated using hazard ratios (HR) and absolute risk differences. Exploratory analyses assessed ELT benefit across model-defined risk groups. In NSABP B-42, the model stratified patients into groups with markedly different outcomes, with a 10-year absolute DR risk difference of 7.95% between high- and low-risk groups [HR, 5.71; 95% confidence interval (CI), 3.5–9.317; P < 0.001]. High-risk patients derived greater absolute benefit from ELT (4.09%) than low-risk patients (0.49%). External validation in the independent TAILORx cohort confirmed prognostic performance, with MI Clarity multimodal–multitask identifying patients with significantly different late DR outcomes (HR, 1.893; 95% CI, 1.413–2.534; P < 0.001). This multimodal AI approach using routine pathology and clinical data enables robust and generalizable stratification of late DR risk in HR+ breast cancer. This scalable strategy may complement existing genomic assays and support more individualized decisions about extended ET. Significance: Late DR is a major cause of death in HR+ breast cancer, and deciding who needs longer hormone therapy is challenging. Current genomic tests are useful but have limitations. This study shows that an AI model using pathology and clinical data identifies patients at high or low risk of late recurrence. High-risk patients may benefit more from extended hormone therapy, offering an accessible way to guide treatment decisions.

Kaplan-Meier analysis of distant recurrence comparing patients treated with ELT versus placebo in subsets of patients from NSABP B-42 translational cohort defined by pathological node status.

Distribution of model predicted risk groups by pathological node status across different models.

Performance comparison of models (image-only, multimodal, M3T model) for predicting benefit from ELT.

558 Background: Accurate prediction of risk of distant recurrence (DR) in HR+, HER2-negative early-stage breast cancer (EBC) is important for optimizing adjuvant therapy decisions, including treatment escalation with CDK4/6 inhibitors (CDK4/6i). Current guidelines recommend consideration of CDK4/6i for node-positive, clinically high-risk EBC, including N1 patients. However, it is not known which patients will benefit, and guideline recommendations suggest that the risks may outweigh benefits for patients with low risk of distant recurrence. Identification of these patients will help prevent over-treatment. RlapsRisk BC (RR) is an AI pathology-based test that integrates features from H&E-stained whole-slide images with clinicopathologic data (age, tumor size, nodal status) to assess risk of distant recurrence in HR+, HER2-negative EBC. The RR model was previously developed and validated using 7 retrospective cohorts totaling 6,039 patients. Here we report validation of RR in the NSABP B-28 cohort of HR+ HER2-negative N+ patients who received post-operative chemoendocrine therapy. Methods: This blinded independent validation study of the pre-specified RR included a subset of NSABP B-28 patients with HR+ and HER2-negative EBC (n=731) and digitized whole slide images from primary tumors. The primary endpoint was distant recurrence-free interval (DRFI). The objective was to validate the prognostic utility of RR score for DR. Univariable and multivariable Cox models were performed. Results: The evaluable cohort had a median follow-up of 11.1 years and included 504 (69%) patients with N1 disease, 87% grade 2/3 tumors, and a median tumor size of 21mm. Across the entire cohort, RR classified 58% of patients as low-risk and 42% as high-risk. RR (high vs low) was significantly associated with DRFI: HR=3.1 (95% CI: 2.3-4.3; p < .001). Estimated 10-year DR-free was 87.5% (95% CI: 83.9-90.4%) for low-risk vs. 65.0% (95% CI: 59.1-70.2%) for high-risk patients. In a multivariable Cox model, the histology-only score remained significant after adjusting for clinicopathologic data: HR=1.6 (95% CI: 1.4-1.9; p < .001). In the N1 subgroup, RR identified 65.7% of patients as low-risk with an estimated 10-year DR-free of 90.6% (95% CI: 86.9-93.4%), compared to 69.8% (95% CI: 62.2-76.2%) for high-risk. Conclusions: RR demonstrates robust prognostic performance in clinically high-risk N+, HR+, HER2-negative EBC patients from the NSABP-28 trial, in which the majority of patients had N1 disease. Our results demonstrate that approximately two-thirds of N1 patients, identified as low-risk by RR, exhibited favorable long-term outcomes with standard chemoendocrine therapies alone. As the clinical landscape shifts toward broader CDK4/6i use, RR could be used to identify N1 patients for whom the benefit of treatment intensification may be minimal.

Abstract PURPOSE: Presence of circulating tumor DNA (ctDNA) is strongly associated with recurrence risk in early triple-negative breast cancer (TNBC). NSABP B-59/GBG-96-GeparDouze evaluated atezolizumab or placebo added to neoadjuvant therapy (NAT) and as adjuvant therapy in 1,550 patients with stage II/III TNBC. A prospective ctDNA sub-study collected serial blood during the first two years after randomization. In the recent 2025 San Antonio Breast Cancer Symposium, we presented the primary endpoint demonstrating that post-surgery ctDNA positivity was associated with an ∼30-fold higher risk of distant recurrence. Here, post- NAT correlation with pCR and distant recurrence is presented. METHODS: Blood was collected at baseline, at completion of NAT prior to surgery, 3-6 weeks post-surgery, and 12- and 24-months after randomization. Primary tumors were analyzed via whole-exome sequencing for variant discovery. For ctDNA detection, libraries containing 15-60 ng of cell-free DNA from plasma samples were hybridized to patient-specific probes to enrich variant-containing regions, then sequenced. The association of post-NAT ctDNA status with pathological complete response, pCR (ypT0/is, ypN0) status and distant recurrence-free interval (dRFI) was determined. Additionally, baseline plasma ctDNA concentration was analyzed by T stage and nodal status RESULTS: At baseline, ctDNA was detected in 153 of 160 patients (96%). ctDNA was detected in 14 of 155 (9%) patients after completion of NAT. Positive ctDNA status after NAT was associated with residual invasive disease at surgery, and positive predictive value for non-pCR was 85.7%. Among the 97 patients with pCR after NAT, 95 (97.9%) were ctDNA-negative prior to surgery. Post-NAT pre-surgery ctDNA status was strongly associated with dRFI, with an HR 10.2 (3.8- 27.3; p<0.0001). Baseline plasma ctDNA concentration significantly increased by T stage (p<0.0001) and nodal status (p=0.0146). CONCLUSIONS: Building on our prior findings, post-NAT, pre-surgery ctDNA-positivity was strongly associated with inferior dRFI, mirroring the negative prognostic impact previously observed for post-surgery ctDNA. Moreover, post-NAT ctDNA positivity showed a high positive predictive value for non-pCR, highlighting its potential utility as an early indicator of treatment resistance and risk stratification. Citation Format: Marija Balic, Gong Tang, Priya Rastogi, Gregory Young, Matthew Wallace, Joshua Acosta, Andreas Schneeweiss, Christie J. Hilton, Tanner J. Freeman, Jiahe Li, Carston Denkert, Mattea Reinisch, Melanie R. Palomares, Bradley A. Arrick, Matthew Petitt, Sujatha Murali, Jean-François Boileau, Dominique Boudreau, Peter J. Polewski, Saima Hassan, Jorge Garces, Gina Costa, Janine LoBello, João Mouta, Walter C. Darbonne, Frederick L. Baehner, Eleftherios P. Mamounas, Norman Wolmark, Sibylle Loible, Charles E. Geyer. Whole-exome sequencing tumor-informed circulating tumor DNA detection after completion of neoadjuvant treatment predicts non-pCR and distant recurrence in patients with early triple-negative breast cancer (TNBC) — Results from a sub-study of the NSABP B-59/GBG-96-GeparDouze Trial [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(8_Suppl):Abstract nr CT013.

TPS615 Background: The TAILORx and RxPONDER trials demonstrated that RS identifies many postmenopausal pts with node-neg and node-pos BC and RS ≤25, who do not benefit from addition of ACT to endocrine therapy (ET). Both trials also showed that certain subsets of premenopausal pts (node-neg/high clinical risk/RS 16-20, node-neg/RS 21-25, and node-pos/RS ≤25) benefited from adding ACT to ET. Most premenopausal pts in these trials did not receive ovarian function suppression (OFS) as part of their ET regimen. Given the observed benefit from OFS in high-risk premenopausal pts with HR+/HER2- BC in the SOFT/TEXT trials, many questioned whether all or part of the observed ACT benefit in the TAILORx/RxPONDER trials may have been the result of chemotherapy-induced OFS. To address this question, we developed OFSET, a phase III, multicenter clinical trial comparing OFS+ET v ACT+OFS+ET. Methods: We hypothesize that addition of ACT to OFS+ET is superior to OFS+ET in improving invasive breast cancer-free survival (IBCFS) among premenopausal, early-stage BC pts with HR+/HER2- tumors, and a 21-gene RS between 16-25 (for pN0 pts) and 0-25 (for pN1 pts). Secondary objectives include invasive disease-free survival, overall survival, distant recurrence-free interval, breast cancer-free interval, and health-related quality of life (HRQOL). Pts must be node-neg with RS 16-20 (plus high clinical risk), or RS 21-25, or have 1-3 positive nodes with RS ≤25. Stratification is by nodal status/RS status (pN0 RS 16-25 v pN1 RS 0-15 and pN1 RS 16-25), intent to receive CDK4/6 inhibitor (yes; no), and age (18-39 v ≥40). Pts are randomized after surgery to either OFS+ET or ACT+OFS+ET v ET is an aromatase inhibitor (AI). Choice is per investigator discretion; tamoxifen is allowed if AI is not tolerated or if OFS is incomplete. Radiotherapy will be administered per investigator discretion per protocol guidelines. The HRQOL sub-study will assess differences in severe menopausal symptoms, measured by the FACT ESS-19 score between arms, as well as increased pain severity (PROMIS). Blood and tumor specimens will be collected for future research. Accrual of 3,960 pts is anticipated to be completed in 7 yrs, 7 mos. Per NSABP B-28 and RxPONDER data, 5yr IBCFS of pN1 pts on the ACT+OFS+ET arm is estimated at 92.3%. Based on TAILORx data, 5yr IBCFS of pN0 pts on the ACT arm is ~95%. Assuming 56% of pts to be pN0 and 44% pN1, and a 0.5% annual loss-to-follow-up rate, the definitive analyses to detect a hazard ratio: 0.75 with ACT+OFS+ET v OFS+ET, with one-sided α of 0.025 and 80% power, will require 380 IBCFS events, expected to occur~11 yrs after study initiation. OFSETwas activated Aug 2023. As of 1-6-25, accrual is: 188/3,960. NCT #: NCT05879926. Support: U10CA180868, -80822, UG1CA189867, U24CA196067. Clinical trial information: NCT05879926 .

Prognostic performance of MI Clarity M3T model for breast cancer-free interval (BCFI), disease-free survival (DFS) and recurrence-free interval (RFI).

Comparison of HER2 expression intensity in paired pre-NAT (i.e., before surgery) and post-NAT surgical samples. HER2, human epidermal growth factor receptor 2; IHC, immunohistochemistry; NAT, neoadjuvant therapy.

Comparison of IDFS outcomes in the ITT population compared to the (A) RNA-evaluable population (REP) (including pre- and post-NAT surgical samples) and in patients with (B) pre-NAT REP samples (left) or post-NAT surgical REP samples (right). HER2, human epidermal growth factor receptor 2; IDFS, invasive disease-free survival; IHC, immunohistochemistry; ITT, intent-to-treat; NAT, neoadjuvant therapy; T-DM1, trastuzumab emtansine.