Poster Spotlight 6: Tools for Predicting Response and Outcomes with Neoadjuvant Therapy

S. Bae¹, J. Kim², Y. Cha³, Y. Kook⁴, A. Kim⁴, J. Jeong⁴, S. Ahn⁴; ¹Surgery, Gangnam Severance Hospital, Yonsei University College of Medicine, Seoul, KOREA, REPUBLIC OF, ²Medical Oncology, Gangnam Severacne Hospital, Yonsei University College of Medicine, Seoul, KOREA, REPUBLIC OF, ³Pathology, Gangnam Severacne Hospital, Yonsei University College of Medicine, Seoul, KOREA, REPUBLIC OF, ⁴Surgery, Gangnam Severacne Hospital, Yonsei University College of Medicine, Seoul, KOREA, REPUBLIC OF.

Background: The incorporation of pembrolizumab into neoadjuvant chemotherapy (NAC) has significantly improved treatment response and prognosis in patients with triple-negative breast cancer (TNBC). Given the aggressive nature of the KEYNOTE-522 regimen, this study aimed to explore the potential for treatment de-escalation based on radiologic response during NAC. Methods: This retrospective study included 517 patients with TNBC who underwent serial breast MRI before, during, and after NAC. Radiologic complete response (rCR) was defined as the absence of any residual mass or enhancement on breast MRI. The association between MRI-based rCR and breast pathologic complete response (pCR, defined as ypT0) was analyzed according to imaging timepoints (mid-treatment vs. post-NAC) and chemotherapy regimen: anthracycline plus cyclophosphamide followed by weekly paclitaxel (AC-T), weekly paclitaxel plus carboplatin followed by anthracycline plus cyclophosphamide (TC-AC), or TC-AC plus pembrolizumab (KN-522 regimen). Results: At the mid-treatment timepoint, the overall rCR rate was 15.5%. Notably, the KN-522 regimen demonstrated an improved mid-treatment rCR rate compared to conventional regimens: 11.6% with AC-T, 12.9% with TC-AC, and 23.6% with the KN-522 regimen. In addition, the majority of patients with rCR at the mid-treatment subsequently achieved a breast pCR: 96.2% with AC-T, 94.7% with TC-AC, and 97.1% with the KN-522 regimen. At post-NAC, the overall rCR rate increased across all treatment groups (28.0% in AC-T, 34.5% in TC-AC, and 42.4% in the KN-522 regimen). However, the corresponding breast pCR rates showed a slight decline (88.9% in AC-T, 86.3% in TC-AC, and 83.6% in KN-522). In patients with high tumor-infiltrating lymphocytes (TILs, ≥ 30%) who received KN-522 regimen and subsequently achieved rCR, the breast pCR rate increased to 94.7%, a trend not observed in the other treatment groups. At the mid-treatment timepoint, the overall rCR rate was 15.5%. The KN-522 regimen resulted in a higher mid-treatment rCR rate (23.6%) compared to AC-T (11.6%) and TC-AC (12.9%). Among patients who achieved mid-treatment rCR, subsequent breast pCR rates were high across all regimens: 96.2% (AC-T), 94.7% (TC-AC), and 97.1% (KN-522). At the post-NAC timepoint, rCR rates increased across all groups (28.0% in AC-T, 34.5% in TC-AC, and 42.4% in KN-522), although corresponding breast pCR rates slightly declined (88.9%, 86.3%, and 83.6%, respectively). Notably, among patients with high tumor-infiltrating lymphocytes (TILs ≥30%) who received the KN-522 regimen and achieved rCR, the breast pCR rate rose to 94.7%, a pattern not observed in other treatment arms. Conclusion: The rCR on mid-treatment MRI was a strong predictor of breat pCR in TNBC. The addition of pembrolizumab to NAC significantly improved the rate of excellent radiologic response at the mid-treatment timepoint. Among patients receiving the KN-522 regimen, those with high TILs revealed improved diagnostic accuracy of post-NAC MRI in predicting breast pCR. These findings suggest the potential for KN-522 regimen de-escalation based on early radiologic response, particularly in patients with high TILs.

A. Carrasco Navas¹, O. Cano i Cano¹, M. Seguí Palmer¹, E. Dalmau Portulas¹, P. Andreu Cobo¹, M. Costillas Frías², M. Medarde Ferrer³, A. López Mestres⁴, N. F. Papaleo⁵, M. Busquets Godall¹, M. Fragío Gil¹, M. Sierra Boada¹, A. Ribas Bravo¹, O. Civit Figueras¹, C. Victoria Marin¹, J. Sandoval Alves¹, M. Monroe Rivera¹, I. Gallego Bernardo¹, S. Soriano Sánchez¹, J. Giner Joaquin¹, P. Ribera Fernández¹, L. Vila Martínez¹, R. Querol Niñerola¹, T. Bonfill Abella¹, E. Gallardo Díaz¹, L. Medina Ortega¹, B. Calvo Martínez¹, M. Rosich Peris¹, L. Fernández Morales¹; ¹Medical Oncology, Hospital Universitario Parc Taulí, Sabadell, SPAIN, ²Breast Imaging Department, Hospital Universitario Parc Taulí, Sabadell, SPAIN, ³General Surgery Department, Hospital Universitario Parc Taulí, Sabadell, SPAIN, ⁴Gynecology Department, Hospital Universitario Parc Taulí, Sabadell, SPAIN, ⁵Anatomic Pathology Department, Hospital Universitario Parc Taulí, Sabadell, SPAIN.

BACKGROUNDPrevious studies suggest that radiological complete response (rCR) to neoadjuvant chemotherapy (NAC), assessed by magnetic resonance imaging (MRI), may predict pathologic complete response (pCR) in patients with operable triple-negative breast cancer (TNBC) treated with anthracycline (AC) and taxane (TX)-based regimens. At our institution, serial MRI assessments are performed before NAC, mid-treatment, and prior to surgery.This study aimed to assess the predictive value of MRI response in patients receiving a novel NAC regimen with pembrolizumab (pembro) an its correlation with pCR. If a strong correlation is confirmed, this could support future de-escalation strategies to reduce toxicity without compromising efficacy.We also compared this group to a historical cohort treated with standard NAC, to evaluate whether adding pembro improves pCR rates and long-term outcomes such as overall survival (OS) and event-free survival (EFS). METHODSWe retrospectively analyzed 162 TNBC patients treated with NAC.Patients were divided into 2 cohorts:-Cohort 1 (n=140): historical cohort with patients treated from February 2000-March 2024 with sequential regimens: 116 received 4 cycles of AC every 3 weeks (wk) followed by TX weekly for 12 cycles, and 24 received the same adding carboplatin (in 21 patients with non-rCR, 2 with rCR and 1 without MRI after AC). Patients underwent MRI to assess tumour response before NAC, after 4 cycles of AC and after TX.-Cohort 2 (n=22): patients treated with new NAC based on KEYNOTE-522, from March 2023-March 2025. MRIs were performed before NAC, after 4 cycles of carboplatin + paclitaxel days 1,8,15 plus pembro day 1 every 3 wk, and after 4 cycles of AC + pembro every 3 wk.Treatment response was determined by histopathological examination. pCR was defined as absence of invasive tumor and nodal involvement. rCR was defined as no/minimal enhancement at the lesion site, comparable to normal tissue. Continuous variables were expressed as mean and standard deviation; categorical variables as absolute values and percentages. Chi-square or Fisher’s exact test was used to assess significance.Survival analysis was conducted using Kaplan-Meier curves and log-rank tests. A p-value <0.05 was considered statistically significant. RESULTSMedian age at diagnosis was 52 years (range 28–77). Most patients had stage III disease (97.5%, AJCC 8th edition). The overall pCR rate (breast and axilla) was 44.4%, significantly higher in cohort 2 vs. cohort 1 (68.2% vs. 40.7%, p=0.016).Early rCR on MRI was achieved in 28.6% of patients overall, significantly more frequently in cohort 2 (66.6% vs. 22.7%, p<0.001).Final rCR on MRI was achieved in 43.8% of patients overall (61.9% in cohort 2 vs. 40.8% in cohort 1, p=0.121).Among patients with early rCR (n=44), 84.1% achieved pCR (83.3% in cohort 1 vs. 85.7% in cohort 2). For those with rCR prior to surgery (n=64), the pCR rate was 82.8% (78.4% in cohort 1 vs. 100% in cohort 2).At 5 years, OS was 79% and EFS 81% for the overall cohort. Patients with pCR had significantly improved outcomes:-OS: 96.7% vs. 66.5% in non-pCR (p<0.001)-EFS: 98.5% vs. 64.4% in non-pCR (p<0.001) Patients with early rCR also had superior 5-year outcomes:-OS: 100% vs. 71.6% in non-responders (p<0.001)-EFS: 100% vs. 70.5% in non-responders (p<0.001)CONCLUSIONIn TNBC, adding pembro to NAC significantly increases early radiological and pathological complete response rates. Serial MRI during NAC are good predictors of pCR, correlating with improved long-term outcomes. The significant association between early rCR and pCR may serve as a non-invasive biomarker to guide therapeutic decisions and could be studied as a tool to de-escalate NAC especially in patients with comorbidities, older age or at high toxicity risk.

W. Li¹, N. Onishi¹, J. Kornak², C. Yau³, D. M. Wolf⁴, J. E. Gibbs¹, T. J. Bareng¹, N. N. Le¹, L. J. Wilmes¹, P. Metanat¹, M. Gibbons¹, E. Price¹, B. N. Joe¹, S. Venters⁴, K. Santos-Parker³, B. LeStage⁵, M. J. Magbanua⁴, L. J. van ‘t Veer⁴, I-SPY2 Imaging Working Group, I-SPY2 Network, A. M. DeMichele⁶, L. J. Esserman³, N. Hylton¹; ¹Radiology and Biomedical Imaging, University of California, San Francisco, San Francisco, CA, ²Epidemiology and Biostatistics, University of California, San Francisco, San Francisco, CA, ³Surgery, University of California, San Francisco, San Francisco, CA, ⁴Laboratory Medicine, University of California, San Francisco, San Francisco, CA, ⁵I-SPY 2 Advocacy Group, I-SPY 2 Advocacy Group, San Francisco, CA, ⁶Medicine, University of Pennsylvania, Philadelphia, PA.

Background: Functional tumor volume (FTV) measured from MRI serves as an imaging biomarker for redirecting treatment in the I-SPY 2 trial. This study aims to validate FTV-based MRI models used to predict favorable response early in treatment.Methods: The study cohort included patients enrolled and randomized to treatment in I-SPY 2 between May 2010 and June 2022. Treatment outcome was assessed by Residual Cancer Burden (RCB) at surgery. Receptor subtype-specific FTV-based models were developed at T2: 12 weeks) to predict the binary outcome RCB-0/I vs RCB-II/III, with RCB-0/I being the positive outcome. FTV at T0: baseline, as well as changes from baseline at T1: 3 weeks (dFTV0_1) and T2 (dFTV0_2), were included as variables in the model. The study cohort was divided into training and test sets balanced for key variables. All combinations of FTV variables were tested in the training set to select the highest performing model within each HR/HER2 subtype for predicting RCB 0/I based on maximizing area under the ROC curve (AUC). Results: The study cohort consisted of 2117 patients in total, divided into training (n=1474) and test sets (n=643). Table 1 shows sample sizes and estimated AUCs in the training and test sets for individual subtypes. AUCs estimated in the test set were lower than AUCs estimated in the training set for all subtypes except HR-/HER2- (Triple Negative). However, these differences did not reach statistical significance. Table 2 shows FTV reduction thresholds generated by maximizing the Youden Index of the ROC curve, specificity (sp), sensitivity (se), PPV, and NPV for predicting RCB 0/I estimated in the training set and in the test set. Conclusion: Our results show comparable performance in the training and test sets, with the highest AUC in the TNBC subtype and the highest PPV in the HR-/HER2+ subtype. The FTV-based models are currently being used in combination with core biopsy as part of the pre-RCB clinical algorithm for shortening therapy in I-SPY 2.

S. Ghose¹, S. Cho², C. Davis³, S. Gandhi⁴, L. Lan⁵, A. Mansuri⁶, H. Trivedi⁷, Y. Polar⁸, F. Ginty³, S. Badve⁹; ¹AI, GE HealthCare, Schenectady, NY, ²AI, GE HealthCare, Niskayuna, NY, ³HTIC, GE HealthCare, Niskayuna, NY, ⁴Emory Winship Cancer Institute, Emory University School of Medicine, Atlanta, GA, ⁵Department of Hematology and Medical Oncology, Emory School of Medicine, Atlanta, GA, ⁶Department of Radiology and Imaging Sciences, Emory School of Medicine, Atlanta, GA, ⁷Department of Radiology and Imaging Sciences, Emory University School of Medicine, Atlanta, GA, ⁸Department of Pathology, Pathology Cancer Program, Atlanta, GA, ⁹Department of Pathology and Laboratory Medicine, Emory University Hospital, Atlanta,, GA.

Recurrence Prediction in Triple Negative Breast Cancer (TNBC) Patients Using Tumor Boundary and Shape Features and Natural Language Based Explanation using Large Language Model (LLM/GPT) Background: TNBC patients exhibit an aggressive disease course with at least 25% developing recurrence or metastasis within 3-5 years, despite standard-of-care therapies. Clinicopathologic factors (CP; age, growth pattern, tumor size, margin status and grade) have limited value in identifying high risk TNBC patients. Early and accurate prediction of recurrence in TNBC patients from clinical mammograms would facilitate therapy optimization. However, new strategies to identify high-risk patients are needed. In this study, we used radiomics features at the tumor boundary to predict recurrence in TNBC patients and LLM/GPT to explain the prediction in natural language. Materials and methods: Mammograms from node negative TNBC patients (n=77; aged 30-90 years, lesion size of 2-45 mm, grades 2 and 3) who underwent adjuvant chemotherapy and had 5-yr follow-up were analyzed. Lesions were manually segmented and tumor boundary (1 mm containing both intra-tumoral and peritumoral regions) was automatically determined. Over 2,000 radiomics features from the tumor boundary and the central regions of the tumor were extracted to quantify heterogeneity, texture, shape, and size of the tumor. Recursive feature elimination using non-linear random forest classifier was used to reduce feature dimensionality and prevent over-fitting. A random forest classifier was employed to build an AI/machine learning model for risk of recurrence using the reduced feature dimension. Top ranking features along with clinicopathological variables were further used in a Bayesian network (BN) to create an interpretable machine learning model. To generate a patient-specific report, a sentence transformer model was further used for BN graph embedding, which aided in querying the graph in natural language inheriting the node/feature specific dependencies of the BN in LLM/GPT (BN LLM). Results: Recurrence prediction was significantly improved by analysis of radiomics features at the tumor boundary (using continuous gradient magnitude and texture features) and shape features, as compared to the same features extracted from the entire tumor. In a 3-fold validation framework, gradient and texture features at tumor boundary, along with tumor shape, better predicted recurrence (mean AUC of 0.78 (std 0.15)) as compared to the same features from the entire tumor (mean AUC of 0.35 (std 0.11)). The addition of clinical variables did not improve the AUC. Using clinical variables alone, AUC of 0.6 (std 0.09) was obtained. Using BN LLM an example patient specific report in natural language is summarized below: Patient A clinical profile: Tumor size – large; age – younger (<50 y); KI67 – high; lymph node invasion – yes; peritumoral probability – high; grade – high.Probability of recurrence based on tumor boundary radiomics: HighTop ranking clinical parameter: Younger age (<50 y) has high probability of recurrence.Prediction: Patient A is at high risk for recurrence. Conclusions: Heterogeneity features at the tumor boundary quantified by continuous gradient magnitude and texture features in routine clinical mammograms could predict TNBC recurrence. To the best of our knowledge, this is the first time that BN LLM has been used to generate patient specific clinically explainable report of risk for TNBC patients. This model will be validated in a larger cohort with a holdout validation set in future studies.

P. Sharma¹, R. Yoder², J. M. Staley², R. Salgado³, R. Madan⁴, H. Pathak⁴, A. K. Godwin², D. Koestler⁵, Q. J. Khan¹, J. O’Shaughnessy⁶, S. R. Stecklein⁷; ¹Internal Medicine, University of Kansas Medical Center, Kansas City, KS, ²The University of Kansas Cancer Center, University of Kansas Medical Center, Kansas City, KS, ³Pathology, ZAS Hospitals, Antwerp, BELGIUM, ⁴Pathology and Laboratory Medicine, University of Kansas Medical Center, Kansas City, KS, ⁵Biostatistics & Data Science, University of Kansas Medical Center, Kansas City, KS, ⁶Medical Oncology, Texas Oncology, Dallas, TX, ⁷Radiation Oncology, University of Kansas Medical Center, Kansas City, KS.

Background: Compared to immune-high TNBC, immune-low TNBC has lower pathologic complete response (pCR) rates and higher recurrence rates after neoadjuvant chemoimmunotherapy. Development of more effective therapies for immune-low early- stage TNBC is an area of unmet clinical need. Bispecific antibodies targeting PD-1/PD-L1 plus vascular endothelial growth factor (VEGF) are being evaluated in advanced TNBC. In this study we examined the prognostic role of an angiogenesis pathway expression signature in immune-high and immune-low early-stage TNBC treated with chemoimmunotherapy. Methods: 107 patients with TNBC treated with neoadjuvant chemoimmunotherapy (Carboplatin + Docetaxel + Pembrolizumab X 6 cycles) on the phase II NeoPACT trial (NCT03639948) with available pre-treatment +/- residual tumor whole exome RNA sequencing were included. Stromal tumor-infiltrating lymphocytes (sTILs) were scored by standard criteria, and tumors were defined as immune-high (Im-High; sTILs ≥30%) or immune-low (Im-Low; sTILs <30%). A previously published 10-gene angiogenesis signature score was computed from RNA sequencing data, and samples were classified as angiogenesis-high (Ang-High) or angiogenesis-low (Ang-Low) based on receiver operator characteristic analysis. Angiogenesis signature was tested for prediction of pCR in Im-High and Im-Low groups. Logistic regression was used to examine the independent prognostic utility of sTILs and angiogenesis signature for pCR. Impact of angiogenesis category on event-free survival (EFS) was assessed by Kaplan-Meier method, with groups compared by log-rank test. Results: 50/107 tumors (47%) were classified as Im-High, and 65/107 (61%) classified as Ang-High. Angiogenesis and sTILs were inversely correlated (P<0.001); compared to Im-High tumors, a higher proportion of Im-Low tumors were Ang-High (43% vs 74%, P=0.002). Angiogenesis signature was inversely associated with pCR both as a continuous variable (P=0.023) and when assessed as high/low categories (pCR 46% vs 75% in Ang-High vs Ang-Low; odds ratio [OR] 0.29, P=0.003). When included in a joint model with sTILs, angiogenesis was independently predictive of pCR (OR=2.48, P=0.046), and the two markers combined predicted pCR more accurately compared to each individually (AUC 0.676 for sTILs, AUC 0.642 for angiogenesis, and AUC 0.720 for joint model). Numerically lower EFS was noted in patients with Ang-High compared with Ang-Low tumors (3-year EFS 81% vs 94%, P=0.087). Dual classification based on both sTILs and angiogenesis was highly prognostic, with the highest pCR (83%) in the Im-High/Ang-Low group, lowest pCR (36%) in Im-Low/Ang-High group, and intermediate pCR in Im-High/Ang-High (68%) and Im-Low/Ang-Low (60%) groups, P<0.001. Paired analysis of pre-treatment and residual tumor showed increased angiogenesis expression in residual disease compared to pre-treatment samples in both Im-High (P=0.005) and Im-Low tumors (P<0.001). Conclusions: Correlative analysis from the NeoPACT trial shows that angiogenesis plays an important role in primary resistance to chemoimmunotherapy, and that this impact is independent of immune upregulation. Patients with Im-High/Ang-Low tumors experienced a pCR rate exceeding 80%, and those with Im-Low/Ang-High tumors had pCR rates below 40%. Furthermore, paired pre-treatment/residual tumor analysis shows that induction of angiogenesis appears to be an escape mechanism contributing to acquired chemoimmunotherapy resistance, irrespective of immune status. These findings lay a strong biological rationale for targeting VEGF in addition to PD-1/PD-L1 in combination with chemotherapy in early-stage TNBC.

M. Magbanua¹, D. Wolf¹, N. Manon¹, R. Sayaman¹, L. Brown Swigart¹, G. Hirst¹, C. Yau², W. Li³, C. Isaacs⁴, R. Shatsky⁵, A. Clark⁶, A. Zimmer⁷, A. Delson⁸, J. Perlmutter⁹, M. Liu¹⁰, P. Pohlmann¹¹, N. Hylton³, R. Nanda¹², D. Yee¹³, W. Symmans¹⁴, L. Esserman², H. Rugo¹⁵, A. DeMichele⁶, L. van ‘t Veer¹; ¹Laboratory Medicine, University of California San Francisco, San Francisco, CA, ²Department of Surgery, University of California San Francisco, San Francisco, CA, ³Department of Radiology, University of California San Francisco, San Francisco, CA, ⁴Lombardi Comprehensive Cancer Center, Georgetown University Medical Center, San Francisco, DC, ⁵5.Department of Medicine, University of California San Diego, San Diego, CA, ⁶Division of Hematology/Oncology, University of Pennsylvania, Philadelphia, PA, ⁷Division of Hematology/Oncology, Oregon Health and Science University, Portland, OR, ⁸Breast Science Advocacy Core, University of California San Francisco, San Francisco, CA, ⁹ISPY2 Advocacy Group, University of California San Francisco, San Francisco, CA, ¹⁰NA, Natera Inc, San Carlos, CA, ¹¹Department of Breast Medical Oncology, University of Texas MD Anderson Cancer Center, Houston, TX, ¹²Department of Hematology and Oncology, University of Chicago, Chicago, IL, ¹³Division of Hematology, Oncology, and Transplantation, University of Minnesota, Minneapolis, MN, ¹⁴Department of Pathology, University of Texas MD Anderson Cancer Center, Houston, TX, ¹⁵Division of Hematology/Oncology, University of California San Francisco, San Francisco, CA.

Background: ctDNA dynamics are emerging as a promising biomarker for predicting treatment response. In the I-SPY2 trial, early clearance of ctDNA during neoadjuvant therapy (NAT) has been shown to predict treatment response across receptor subtypes (Magbanua et al., Cancer Cell, 2023). Separately, transcriptional profiling of pretreatment tumors has enabled the classification of breast cancers into Response Predictive Subtypes (RPS), incorporating immune, DNA repair deficiency (DRD), and Luminal expression biomarkers in addition to HR and HER2, which, if used to allocate treatment decisions, are predicted to improve response and outcome (Wolf et al., Cancer Cell, 2022). However, ctDNA dynamics in RPS and its association with treatment response have yet to be explored. Methods: The study involved 719 patients (pts) with high-risk early-stage breast cancer who received NAT in I-SPY2. ctDNA levels were measured in blood collected at baseline and on-treatment at weeks (wk) 3 and 12 (post-paclitaxel before anthracycline) using a personalized tumor-informed 16-plex PCR-NGS assay (Signatera, Natera, Inc.). RPS were grouped into treatment-sensitive, i.e., those with high pathological complete response (pCR) rates [HER2-Immune+ (n=207, 29%), HER2-Immune-DRD+ (n=80, 11%), and HER2+Non-luminal (n=131, 18.%)] when treated with the best matched therapeutic option (e.g., immunotherapy for HER2-Immune+) vs. treatment-resistant, i.e., those that have shown low pCR rates to nearly every class of agent/combination tested in I-SPY2 to date [HR+HER2-Immune-DRD- (n=211, 29%), HR-HER2-Immune-DRD- (n=35, 5%), and HER2+Luminal (n=55, 8%)]. We evaluated the association between ctDNA dynamics and treatment response as measured by residual cancer burden (RCB) after NAT. ctDNA dynamics were defined as persistently ctDNA-negative, clearance at wk 3 (early clearance) or wk 12 (clearance), or no clearance at wk 12. Results: Baseline ctDNA positivity rates varied significantly across RPS [Fisher’s exact (FE) p<0.001], ranging from 56% (HER2+Luminal) to 95% (HER2-Immune-DRD+). There were significant differences in the median ctDNA concentration across RPS [Kruskal-Wallis (KW) p<0.001], with the lowest and highest median values observed in the HER2+Luminal and HER2-Immune-DRD+ groups, respectively. The proportion of pts who cleared ctDNA early varied significantly across subtypes (FE p<0.001), ranging from 21% (HR-HER2-Immune-DRD-) to 67% (HER2+Non-luminal). In the treatment-sensitive RPS, we observed a strong association between early ctDNA clearance and treatment response. In the HER2-Immune+ subtype, early ctDNA clearance was associated with significantly higher proportions of pts who eventually had no (RCB-0/pCR) or limited (RCB-I) residual cancer (86% vs. late clearance=57%, no clearance=32%, FE p<0.001). Early ctDNA clearance significantly improved the odds of achieving an RCB-0/RCB-I (OR=12.6, 95% CI 5.2-33.4). In addition, early clearance also markedly shifted the distributions toward lower RCB scores, with pts who cleared ctDNA early showing significantly lower median RCB scores compared to those who remained ctDNA-positive (KW p<0.001). Similar results were observed in other treatment-sensitive RPS (HER2-Immune-DRD+ and HER2+NonLuminal). These significant associations between ctDNA dynamics and treatment response were not observed in the treatment-resistant subtypes. Conclusions: Early ctDNA clearance is a strong predictor of favorable response in treatment-sensitive RPS. ctDNA monitoring and classifying tumors by RPS are being tested as part of an algorithm for treatment decisions in the neoadjuvant setting. These results offer insights into the potential contributions of RPS and ctDNA to imaging in optimizing therapy in the I-SPY2 trial.

J. Koh¹, K. Lee², A. Min³, H. Ryu¹, D. Lee², J. Kim², Y. Shin², Y. Kim³, C. Park², D. Ho⁴, H. Nguyen⁴, W. Lee⁵, H. Moon⁶, W. Han⁶, H. Lee⁶, T. Kim², N. Cho⁷, G. Cheon⁸, S. Im²; ¹Department of Pathology, Seoul National University Hospital, Seoul, KOREA, REPUBLIC OF, ²Department of Internal Medicine, Seoul National University Hospital, Seoul, KOREA, REPUBLIC OF, ³Cancer Research Institute, Seoul National University, Seoul, KOREA, REPUBLIC OF, ⁴Department of Cancer AI & Digital Health, National Cancer Center, Goyang-si, Gyeonggi-do, KOREA, REPUBLIC OF, ⁵Department of Microbiology and Immunology, Seoul National University College of Medicine, Seoul, KOREA, REPUBLIC OF, ⁶Department of Surgery, Seoul National University Hospital, Seoul, KOREA, REPUBLIC OF, ⁷Department of Radiology, Seoul National University Hospital, Seoul, KOREA, REPUBLIC OF, ⁸Department of Nuclear Medicine, Seoul National University Hospital, Seoul, KOREA, REPUBLIC OF.

Background We conducted MEDIOLA WOO trial with olaparib and durvalumab before standard neoadjuvant chemotherapy (NAC) in 54 stage II/III triple negative or ER-low breast cancer (NCT03594396) (Im SA et al., SABCS 2021, PD15-08; Koh J et al., SABCS 2024, PS12-01). Here, we present updated analyses. Methods Olaparib 300mg bid was given for 4 weeks with 1500 mg durvalumab (D15), followed by NAC. FDG PET and serial biopsies were taken at baseline (BL), 2 weeks (OT1), and 4 weeks (OT2). We conducted RAD51 foci assays on BL to assess functional homologous recombination deficiency (fHRD), and CT at OT2 for RECIST evaluation. At all timepoints, we performed PD-L1 (22C3) immunohistochemistry (IHC), RNAseq, multiplex IHC (mIHC), and artificial intelligence-guided segmentation (AI-S) of H&E images into tumor cells (TC) and inflammatory cells (IC). Metabolic (mRes) and pathologic (pRes) responses were defined as ≥ 30% reduction of SUVmax in PET and ≥50% reduction in TC. RNAseq analyses included gene set enrichment analysis (GSEA), and molecular subtyping by PAM50 and TNBC types into basal-like immune-activated (BLIA), basal-like immunosuppressed (BLIS), luminal androgen receptor (LAR), and mesenchymal (MES). Results Pathologic complete response (pCR) was achieved in 70.4%. At OT1, early mRes and pRes were observed in 44.4% and 63.0%, which were associated with fHRD status at BL (p < 0.001). pCR was associated with mRes, pRes, and RECIST responses at OT2 (p < 0.001, 0.043, and 0.027), while no such associations were observed with PD-L1 IHC or HRD status at BL. Longitudinal analyses on AI-S results showed gradual decreases in TC from BL to OT1 and OT2 (p < 0.01), while an increase in IC was noted at OT1 compared to BL (p < 0.001), which persisted into OT2. PAM50 subtypes at BL were predominantly Basal (64.8%), followed by LumB (13.0%), Normal-like (11.1%), HER2-enriched (9.3%), and LumA (1.9%). Over time, the proportions of Basal tumors shrank (40.7% at OT1; 14.8% at OT2), while those of LumA expanded (27.8% at OT1; 53.7% at OT2). Most tumors were BLIS (85.2%) at BL, with only 5.6% of BLIA; however, BLIS decreased to 68.5% at OT1, while BLIA increased to 25.9%. mIHC also confirmed the increased amounts of CD3⁺, CD4⁺, and CD8⁺ T-cells at OT1/OT2 compared to BL (p < 0.05), most pronouncedly on CD8⁺ subset (p < 0.0001). OT1 showed downregulations of MYC target, G2M checkpoint, and E2F target gene sets compared to BL (adj p < 0.001) by GSEA, possibly reflecting the effect of PARP inhibition. OT2 tumors showed significant enrichment of interferon-γ/α, IL2-STAT5, IL6-JAK-STAT3, and TNFα signaling pathways compared to BL (adj p < 0.001). When stratified by clinico-pathologic responses, the most striking differences were observed for pCR status, especially at OT1. Large transitions on PAM50 and TNBC subtypes at OT1 occurred nearly exclusively within those who achieved pCR. By mIHC, the boosting of CD8⁺ T-cells with immune checkpoint expression (PD1, LAG3, or TIM3) at OT1 was noted in those with pCR (p < 0.01), while negligible in non-pCR patients. Similar patterns were noted in other T-cell subsets, including CD4⁺ memory T-cells, resting/activated regulatory T-cells, activated CD8⁺ T-cells, CD8⁺ memory T-cells, CD4⁺CXCL13⁺, and CD8⁺CXCL13⁺ cells (p < 0.05). After a median follow-up duration of 57.5 (8.0-78.1) months, 5 patients relapsed, whose tumors showed upregulation of epithelial-mesenchymal transition signature at OT1 (adj p < 0.001), and G2M checkpoint, E2F, and MYC target gene sets at OT2 (adj p < 0.01). Conclusions Temporal tumor microenvironmental changes occurred as early as 2 weeks after olaparib, which was the most reflective of pCR. Our findings highlight the immune modulatory effects of PARP inhibition and the value of longitudinal biopsies for response prediction and prognostication.

D. Yee¹, C. Yau², P. Norwood³, D. M. Wolf⁴, P. R. Pohlmann⁵, N. Onishi⁶, L. van ‘t Veer⁷, A. Borowsky⁸, W. F. Symmans⁹, J. Perlmutter¹⁰, J. C. Boughey¹¹, E. Price⁶, W. Li⁶, I-SPY Investigators, N. Hylton⁶, L. J. Esserman²; ¹Masonic Cancer Center, University of Minnesota, Minneapolis, MN, ²Surgery, University of California San Francisco, San Francisco, CA, ³Biostatistics, Quantum Leap Health Care Collaborative, San Francisco, CA, ⁴Lab Medicine and Pathology, University of California San Francisco, San Francisco, CA, ⁵Breast Oncology, MD Anderson Cancer Center, Houston, TX, ⁶Radiology, University of California San Francisco, San Francisco, CA, ⁷Laboratory Medicine, University of California San Francisco, San Francisco, CA, ⁸Pathology and Laboratory Medicine, University of California Davis, Sacremento, CA, ⁹Anatomic Pathology, MD Anderson Cancer Center, Houston, TX, ¹⁰Gemini Group, Gemini Group, Ann Arbor, MI, ¹¹Surgery, Mayo Clinic, Rochester, MN.

Introduction: AC plus a taxane is standard neoadjuvant therapy for high-risk early breast cancer, but AC may be unnecessary for pathologic complete response (pCR) in many patients. I-SPY2 is a phase 2 trial evaluating the use of 12 weeks of paclitaxel +/- an investigational drug followed by AC. Serial measurements of Functional Tumor Volume (FTV) by MRI, inter-regimen biopsies and pathology from surgery were used to determine the contribution of AC to reduction in FTV or pCR. Methods: 1187 patients enrolled between 2010-2022 had serial measurements of FTV. 150 patients received an inter-regimen biopsy. The primary endpoint of I-SPY2 is pCR. To determine how much AC contributed to the reduction in FTV, we calculated the total FTV decrease between pre-treatment and pre-surgery, and then computed the percentage (%) of the total FTV decrease seen after 12 weeks of paclitaxel regimen. We attributed pCRs observed among patients for whom taxane contributed <90% FTV reduction to AC and estimated the % patients achieving pCR from AC. For the subset with inter-regimen biopsy, we estimated the percentage achieving pCR from those patients with invasive cancer present at inter-regimen biopsy. Results: We found that by week 12 (end of taxane), 64% (763/1187) of patients had achieved ≥90% of their total FTV reduction. In 424/1187 (36%) patients had with taxane <90% response, and subsequent AC contributed to more than 10% of the total decrease in FTV. 90 pCRs (21%) were observed among these 424 patients or, overall, approximately 8% of patients achieved pCR from AC after taxane. Of note, 54/1187 (4.5%) patients had no response to a taxane regimen and had decreased in FTV due to AC alone. Of the 54/1187 responding to AC only, there were 11 pCRs. These results varied by response predictive subtypes as shown. When the subpopulation with post-paclitaxel biopsy was examined, 63/150 (42%) patients had invasive cancer at this timepoint. 47 patients received AC and 4/47 (8.5%) achieved a pCR after AC. 13/47 (28%) achieved RCB 0/1. Conclusions: Serial measurements of FTV identifies patients who had a favorable response to a taxane regimen. For those with a suboptimal response, AC provided some contribution to further decrease in tumor volume. However, only a small fraction achieved a complete response. Immune positive and HER2+ (non-luminal) had the biggest benefit, with only 3-6% of other subtypes achieved pCR. Thus, AC can be reserved for those patients with suboptimal response to the initial taxane therapy. I-SPY2.2 utilizes these data to avoid AC by allowing patients to proceed to surgery after serial MRI and biopsy predict a pCR at an early timepoint. While AC remains the standard of care in many neoadjuvant regimens, the low response rates seen in I-SPY2 suggests that more active regimens need to be developed for patients with disease resistant to a taxane based therapy.

D. Gentile¹, W. Gatzemeier², S. Di Maria Grimaldi², A. Sagona², E. Barbieri², S. D’Amico², A. Bottini², V. Errico², A. Testori², G. Canavese², C. Tinterri²; ¹Breast Unit, IRCCS Humanitas Research Hospital, Milan, ITALY, ²Breast Unit, IRCCS Humanitas Research Hospital, Rozzano, Milan, ITALY.

Introduction: The advent of sentinel lymph node biopsy (SLNB) and neoadjuvant chemotherapy (NAC) has progressively reduced the extent of axillary surgery in breast cancer (BC). In early-stage BC, axillary lymph node dissection (ALND) is no longer recommended for patients with micrometastasis (ypN1mi) due to equivalent outcomes and reduced morbidity. However, the role of ALND remains unclear in patients initially cN+ who convert to cN0 after NAC but present residual axillary micrometastasis. Objectives: NEONOD2 is a prospective, multicenter, non-inferiority trial designed to assess whether omission of ALND in cN+ patients downstaged to cN0 post-NAC and found to have sentinel lymph node (SLN) micrometastasis (ypN1mi) leads to recurrence or survival outcomes comparable to those with negative SLN (ypN0), for whom ALND omission is current standard. Methods: Eligible patients were cT1-3/cN+ BC candidates for NAC. After completion of NAC, only patients with clinical conversion to cN0 underwent SLNB. Those with micrometastatic SLN (ypN1mi) were enrolled in the experimental group; those with ypN0 were assigned to the standard group. Primary endpoint was recurrence-free survival (RFS); secondary endpoints included overall survival (OS), loco-regional and distant RFS. Results: As of the current interim analysis, 449 patients have been enrolled across Italian Breast Units. The study population included 70 patients (15.6%) with SLN micrometastases (experimental group), 285 (63.5%) with negative SLN post-NAC (standard group), and 94 (20.9%) with macrometastatic SLN involvement. The median age was 50 years (range 26-75). Overall, 194 BC patients (43.2%) were post-menopausal. Magnetic resonance imaging was performed in 69.6% of patients. Most tumors were high-grade (G3: 52.7%) and of ductal histology (86.7%), with HER2-positive and triple-negative subtypes comprising 42.8% and 19.8% of the cohort, respectively. Following NAC, all patients achieved ycN0 status and underwent SLNB. A pathological complete response (pCR) was recorded in 35.1% of patients. The majority of patients (54.5%) were treated with breast-conserving surgery. Adjuvant radiotherapy, was administered in 46.8% of cases, with 18.9% receiving regional nodal irradiation. At a mean follow-up of 21.8 months (SD 13.0), nine recurrences (2.0%) were documented across the cohort. Three events (4.3%) occurred in the ypN1mi group, comprising one ipsilateral axillary-supraclavicular relapse, one ipsilateral breast recurrence following breast-conserving surgery, and one ipsilateral skin recurrence post-mastectomy. In the ypN0 group, four recurrences (1.4%) were observed, including two axillary (one supraclavicular), one ipsilateral breast, and one cutaneous recurrence. Two patients (2.1%) in the macrometastatic group experienced recurrence, including one breast and one distant brain metastasis. Notably, no isolated axillary relapse occurred in the micrometastatic group. Five deaths (1.1%) were recorded overall: three in the ypN0 group (1.1%) and two in the macrometastatic group (2.1%). No deaths occurred among patients in the micrometastatic group (0%). Conclusions: Preliminary findings from the NEONOD2 trial suggest that omission of ALND in patients with SLN micrometastases following NAC does not result in a higher incidence of regional recurrence or mortality at early follow-up. The absence of isolated axillary failure and the lack of deaths in the experimental group are reassuring signals supporting the safety of de-escalated axillary management in this subset. Continued accrual and extended follow-up are ongoing to confirm the long-term oncologic safety of this approach.