Citations

Background: Bipolar disorder (BD) is associated with a higher prevalence of chronic kidney disease (CKD) and increased mortality, yet individuals with both BD and CKD are less likely to receive kidney replacement therapy (KRT). Sodium-glucose cotransporter-2 inhibitors (SGLT2is) provide renal protection in diabetes, but their long-term effects in this high-risk population remain unknown. Objectives: To evaluate whether SGLT2i use reduces the risk of KRT and all-cause mortality in adults with BD and mild-to-moderate CKD (stage ⩽ 3). Design: Observational cohort study with propensity score matching (PSM). Method: In this study using TriNetX (2009–2024) real-world database, 89,369 adults with BD and mild-to-moderate CKD were classified as SGLT2i users ( n = 12,736) or nonusers ( n = 76,633). Primary outcomes were progression to KRT and all-cause mortality over 5 years. Adjusted hazard ratios (aHRs) were estimated using Cox proportional hazards models, and 1:1 PSM reduced confounding. Subgroup analyses examined sex, race/ethnicity, diabetes status, CKD stage, and mood stabilizer use. Results: SGLT2i use was associated with lower risk of KRT (aHR 0.47 (95% CI, 0.42–0.53)) and all-cause mortality (aHR 0.69 (95% CI, 0.65–0.73); both p < 0.001). Protective effects were consistent across subgroups, including individuals receiving lithium, lamotrigine, valproate, or antipsychotics. After PSM (10,967 matched pairs), 5-year KRT-free survival was 94.61% versus 90.99%, and overall survival was 78.67% versus 67.53% (both p < 0.001). Given the observational design, the possibility of residual or unmeasured confounding canot be excluded. Conclusion: SGLT2i therapy in individuals with BD and mild-to-moderate CKD is associated with substantially lower risks of KRT and all-cause mortality. Prospective trials are needed to confirm these findings.

Abstract Introduction Cancer clinical trials have historically faced challenges in achieving fully representative trial population, particularly in terms of including patients from diverse ethnic and racial groups, those with low socioeconomic status, and those living in rural regions. This has contributed to inequities in access to novel therapeutic options. Background The I-SPY adaptive platform trials have prioritized diversity since inception, consistently enrolling in &gt;10% of trial patients being Black and/or Hispanic individuals. The majority of participating institutions in the I-SPY 2.2 study are academic or large well-resourced health systems. In 2023, the I-SPY ACCESS (Achieving Cancer Clinical Trial Equity in Socioeconomically Diverse Sites) Initiative was launched to enhance representation by engaging and involving nontraditional trial sites- such as safety-net hospitals, community cancer centers, and rural cancer centers- that serve historically excluded patient populations. Methods The I-SPY ACCESS Initiative offers customized and personalized tailored support to under-resourced sites. This includes comprehensive feasibility and readiness assessments, onboarding guidance, enhanced site-specific training, centralized institutional review board (IRB) navigation, site peer mentorship, and educational materials for both patient and provider. Additionally, patients gain access to I-SPY advocate peer trial navigators. Sites are provided culturally responsive patient education tools- including multilingual brochures, a dedicated patient website, and content co-developed with community partners. A Clinical Trial Navigator supports both patients and sites, helping to locate logistical, social, and trial specific resources.Tools such as virtual Clinical Research Coordinators (vCRC's) and OneSource technology are deployed strategically based on assessed site need and resource availability. OneSource is an innovative platform that enables direct extraction of clinical data from electronic health records into trial databases, reducing site workload and improving data quality. Funding models are currently being explored to extend OneSource and vCRC services across more ACCESS sites. In the future, there are plans to pilot financial navigation services and integrative care resources for all patients enrolled in I-SPY with particular focus on those enrolled at ACCESS sites. Results and Progress As of June 2025, seven sites have been assessed as ACCESS sites for I-SPY 2.2, with two sites having successfully completed activation and currently enrolling participants. Beginning in mid-2024, one DCIS RECAST ACCESS site has been activated and 11 additional sites are in the onboarding process. The modest pace of I-SPY 2.2 activation reflects the complexity and operational demands associated with resource-intensive platform trial. In contrast, the DCIS RECAST protocol was designed with a lower burden, facilitating easier implementation by under-resourced sites, reflected by increased interest in the study among ACCESS sites. The ACCESS framework is being explored for potential future application in the Pre I-SPY Phase 1 platform trial, particularly considering the historical underrepresentation of such sites in the early-phase clinical research. Conclusion and Future Goals The I-SPY ACCESS initiative demonstrates that investing in thoughtful, innovative, and effective means for trial delivery, as well as incorporating community-based support, can strengthen representation in complex, adaptive breast cancer trials. The initiative offers a flexible, equity-driven framework that is impactful now and scalable for future innovation in oncology research. Citation Format: S. L. Horton, D. A. Idossa, K. C. Hewitt, J. Perlmutter, L. Van't Veer, N. Hylton, A. Borowsky, F. Symmans, D. Yee, C. Yau, P. Pohlmann, J. Boughey, L. Esserman. Expanding Diversity through Innovation: Development of the I-SPY ACCESS Initiative for the I-SPY2.2 and DCIS RECAST Clinical Trials [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS4-09-08.

Abstract Background: The Breast Cancer Index (BCI) is clinically validated to stratify patients with hormone receptor-positive (HR+), early-stage BC based on the risk of overall (0-10 years [y]) and late (5-10 y) distant recurrence (DR) with N0/N1 lymph node involvement. Validation of BCI in patients with HR+ HER2+ BC has been limited, especially in patients treated with adjuvant trastuzumab-based chemotherapy. The HOXB13 and IL17BR genes constitute the predictive component of BCI and are known to exhibit an ER-dependent correlation pattern with HER2. We evaluated the prognostic performance of BCI in patients with HR+ HER2+ disease enrolled in the NCCTG N9831 (Alliance) trial treated with the combination of chemotherapy and trastuzumab. Methods: All available tumor specimens from patients with HR+ HER2+ N0/N1 BC in arms B (sequential trastuzumab) and C (concurrent trastuzumab) of the trial were included. BCI testing was performed blinded to clinical outcomes. The primary endpoint was distant recurrence-free interval (DRFI). BCI (N0) and BCIN+ (N1) risk groups were calculated based on prespecified cut points. Kaplan-Meier analysis and log-rank testing were used to assess the prognostic significance of BCI/BCIN+ risk groups based on DR. Cox proportional hazard models with and without clinical covariates were used to estimate hazard ratios (HRs), and the associated 95% confidence intervals (CIs). Results: A total of 454 HR+ patients (234 arm B, 220 arm C), including 69 N0 and 385 N1 patients, were analyzed (median age was 47 y, 85% N1, 55% T2, 63% G3). BCIN+ scores were available for 378 of the 385 N1 patients due to missing tumor grade for 7 patients. The median follow-up was 12.7 years. In the overall cohort, BCI/BCIN+ stratified 51 (11%), 24 (5%) and 372 (83%) patients as low-, intermediate- and high-risk with 10-y DRFI rates of 96% (95% CI: 84-99%), 100% (100-100%) and 89% (86-92%), respectively. Since the low and intermediate risk groups had few patients and a similar risk of 10-y DRFI, they are combined into a single low-risk group hereafter. Thus, for overall 10-y DR, 75 (17%) and 372 (83%) of patients were classified by BCI/BCIN+ as low and high risk with 10-y DRFI rates of 97% (95% CI: 89-99%) and 89% (95% CI: 86-92%), respectively (HR=4.50, 95% CI: 1.09-18.61; p = 0.038). BCI/BCIN+ remained statistically significant in the multivariable analysis, adjusting for age, treatment, and nodal status (HR=7.54, 95% CI: 1.34-42.64; p = 0.022). For 69 N0 patients, BCI stratified 54 (78%) and 15 (22%) patients as low and high risk with 10-y DRFI rates of 100% (95% CI, 100-100%) and 77% (95% CI, 44-92%), respectively (p = 0.012). For 378 N1 patients, BCIN+ stratified 21 (6%) and 357 (94%) of patients as low and high risk with 10-y DRFI of 90% (95% CI, 66-97%) and 90% (95% CI, 86-93%), respectively (p = 0.07). Of the 21 patients classified as BCIN+ low-risk, two had BCIN+ scores of 6.75 and 6.73 in close proximity to the cut point (6.93) and experienced DR after 5 years, resulting in the same 10-y DRFI for both BCIN+ low- and high-risk patients. Conclusions: BCI is significantly prognostic for the risk of overall DR (0-10 y) in patients with HR+ HER2+ disease treated with adjuvant trastuzumab-based chemotherapy. Optimization of BCI cut points may further improve prognostic performance in HR+ HER2+. Additional validation in other HR+ HER2+ cohorts treated in the modern setting with anti-HER2-based therapy is warranted, given the small sample size of N0 patients and relatively low DR rate in the late recurrence cohort. In summary, these results provide further evidence of BCI’s utility to support clinical decision-making for adjuvant endocrine therapies for HR+ HER2+ BC. Support: U10CA180821, U10CA180882, U24CA196171, Genentech; Clinicaltrials.gov Id: NCT00005970; https://acknowledgments.alliancefound.org. Citation Format: S. Chumsri, E. Liu, K. Edwards, Y. Wen, Y. Zhang, A. K. Anderson, A. H. Partridge, L. Carey, K. Ballman, E. Perez, K. Treuner, M. Goetz, E. Thompson. Prognostic performance of Breast Cancer Index in patients with early-stage HR+ HER2+ breast cancer (BC) treated with adjuvant trastuzumab: NCCTG N9831 (Alliance) [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS3-08-08.

Abstract Background: Survivor plots permit adjustment by baseline demographics and clinical characteristics; log-normal survival analysis may be more robust than a Cox model under departures from model assumptions. We depicted distant disease-free survival (DDFS) for statistically standardized quantitated ER, PgR, and HER2. Methods: We utilized CCTG MA.27 (NCT00066573), an adjuvant phase III trial of exemestane versus anastrozole in postmenopausal women with ER+ and/or PgR+ tumours. IHC ER, PgR, and HER2 HSCORE were centrally assessed by machine image quantitation, and statistically standardized to mean of 0, standard deviation (SD) of 1 following Box-Cox variance stabilization transformations of square for ER, natural logarithm (ln) and square root for PgR, and ln for HER2. For the ln transformation, 0.1 was added to 0 HSCOREs for a stable transformation. The primary endpoint was DDFS at the longest trial follow-up of 4.1 years. For log-normal survival analysis, ln(time) is a linear function of covariates and had a normal distribution; the Cox models were log-linear for covariates. Survival was depicted with log-normal and Cox survivor plots adjusted at the means of age, T and N stage, grade, lymphovascular invasion, treatment, and baseline patient demographics. We examined ER and PgR ASCO/CAP cut-points at (0; &gt;0), HER2 cut-points at (IHC 0, 0 stain; IHC 0, (0,10%] stain; IHC 1+; IHC 2+; IHC 3+), and ER/PgR/HER2 standardized cut-points at standard deviations about mean of 0 (&lt;-1; (-1,0]; (0,1]; &gt;1). P was based on the likelihood ratio criterion (∼Χ2 (1)) for the addition of ER/PgR/HER2 to the model of baseline covariates, and had nominal significance if P&lt;0.05. Log-normal and Cox multivariable regressions with unstandardized and standardized categorical, and continuous, ER, PgR, and HER2 were adjusted by baseline covariates; 2-sided Wald tests had nominal significance if P&lt;0.05. Results: Of 7576 MA.27 patients, 2325 patients had quantitated image analysis assessments for ER, PgR, and HER2; 113 of 2325 (4.9%) patients experienced a DDFS event during follow-up. Multivariable log-normal survival analysis indicated DDFS was significantly longer for patients with T1 (p=0.01), N0 (p=.002), and grade 1/2 tumors (p=0.01); similar significance levels were found with a Cox model. Log-normal survivor plots were smooth, while Cox survivor plots were step-wise discontinuous at events. Adjusted log-normal survivor plots indicated similar 5-year DDFS for patients by standardized ER categories (p=0.43): 95.4% for those with ER &lt;-1.0 (n=401); 95.8%, for (-1.0,0.0] (n=737); 96.1%, for (0.0,1.0] (n=737); and 96.4%, for &gt;1.0 (n=450). The log-normal standardized PgR survivor plots indicated patients had significantly different 5-year DDFS (p&lt;.001) by categories, with both ln and square root transformations of HSCORE: with a ln transformation, 93.7% for &lt;-1.0 (n=607); 95.4%, for (-1.0,0.0] (n=390); 96.8%, for (0.0,1.0] (n=826); and 97.8%, for &gt;1.0 (n=502). Patients had similar 5-year DDFS by standardized HER2 categories (p=0.71): 96.2% for &lt;-1.0 (n=83); 96.0%, for (-1.0,0.0] (n=1201); 95.9%, for (0.0,1.0] (n=527); and 95.7%, for &gt;1.0 (n=514). In full fit multivariable log-normal survival analyses, DDFS was not significantly different for ER (p=0.51-0.89) and HER2 (p=0.12-0.98); higher PgR was associated with significantly better DDFS (p=0.01 to &lt;.001). Adjusted Cox survival analyses had similar results. Conclusions: Survivor plots adjusted for baseline demographic and clinical characteristics provided similar indications of significance for standardized ER, PgR, and HER2 as those seen in multivariable models with both log-normal and Cox survival analyses. Higher PgR was associated with significantly better DDFS. Citation Format: J. W. Chapman, J. Bayani, S. SenGupta, J. M. S. Bartlett, T. Piper, M. A. Quintayo, S. Virk, P. E. Goss, J. N. Ingle, M. J. Ellis, G. W. Sledge, G. T. Budd, M. Rabaglio, R. H. Ansari, R. Tozer, D. P. D'Souza, H. Chalchal, S. Spadafora, V. Stearns, E. A. Perez, K. A. Gelmon, T. J. Whelan, C. Elliott, L. E. Shepherd, B. E. Chen, K. J. Taylor. Survivor plots for quantitated adjunctive statistically standardized ER, PgR, and HER2 in adjuvant postmenopausal breast cancer: Canadian Cancer Trials Group MA.27 trial of exemestane versus anastrozole [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS3-07-23.

Abstract Purpose Background parenchymal enhancement (BPE), the contrast enhancement of normal fibroglandular tissue on dynamic contrast-enhanced breast MRI, is hormonally dependent and fluctuates with the menstrual cycle, lactation, and pregnancy. In patients with breast cancer, both endocrine therapy and chemotherapy can decrease BPE levels during treatment. Notably, a lack of BPE suppression was associated with inferior response after neoadjuvant chemotherapy in hormone receptor-positive (HR+) cancer. In HR+ breast cancer, BPE may be particularly relevant as estrogen fuels the tumor’s growth. However, there is limited clinical research directly comparing BPE and serum estradiol levels. This study aimed to investigate 1) the correlation between BPE and serum estradiol levels in treatment-naïve premenopausal patients with HR+ breast cancer and 2) the association between BPE suppression and serum estradiol suppression during neoadjuvant endocrine therapy (NET). Methods We retrospectively reviewed data from premenopausal patients enrolled in the Endocrine Optimization Pilot (EOP), a sub-study of the I-SPY 2 TRIAL, which tests new endocrine strategies in patients with stage 2/3, MammaPrint (MP) low-risk (index 0 to 1), HR+/HER2-negative (HER2-) breast cancer. Patients with MP High1 (index -0.57 to 0) tumors were eligible if clinically node negative. This study included 58 EOP patients who had both serial serum estradiol levels (drawn monthly) and BPE data available. These patients started ovarian function suppression (OFS) at various times based on their assigned treatment regimen: before NET start, n=28; at cycle 1, n=16; at cycle 2, n=14. The average estradiol levels over six months on NET were categorized into two groups: suppressed to the postmenopausal range (&lt;10 pg/mL) or non-suppressed (≥10 pg/mL). MRI scans at baseline (before NET) and pre-surgery (after six months of NET) were analyzed. BPE was calculated on the contralateral breast as the mean early (∼150s post-contrast injection) percent enhancement of the central 50% of the axial slices. The Spearman’s correlation coefficient and the Wilcoxon rank-sum test were used to examine the relationship between BPE and estradiol. Results Treatment-naïve estradiol levels were available for 25 of the 58 patients, with the median [1st, 3rd quartile] of 82.7 [32.4, 250] pg/mL. A positive correlation was observed between treatment-naïve estradiol and baseline BPE, with a correlation coefficient of ρ = 0.74 (p &lt; 10-5). There was no significant difference in the treatment-naïve estradiol levels between patients who later showed estradiol suppression to the menopausal range during NET (n=14 [56 %]) and those who did not (n=11 [44 %]) (81.3 pg/mL versus 86.5 pg/mL, respectively (p=0.39)). During NET, estradiol levels were suppressed to the postmenopausal range in 26 of the 58 patients (45%). Estradiol suppression was significantly associated with greater reduction in post-treatment BPE: -42.7% [-54.3, -21.8] BPE reduction for the suppressed estradiol group versus -15.0% [-40.6, 6.85] for the non-suppressed estradiol group (p=0.002). Conclusion This study demonstrates the estrogen dependency of BPE in premenopausal patients with HR+/HER2- breast cancer. Baseline estradiol levels positively correlated with baseline BPE, and the suppression of on-treatment estradiol was associated with greater BPE suppression following NET. Our findings further support exploration of the use of BPE as a reflection of estradiol level under NET and OFS in addition to other MRI biomarkers such as functional tumor volume. These complementary imaging biomarkers could independently reflect both non-malignant and tumor-specific biological processes to improve predicted performance. Citation Format: S. Choi, N. Onishi, P. Metanat, R. Mukhtar, K. Giridhar, M. P. Goetz, L. Huppert, A. Elias, O. Olopade, I-SPY2 Imaging Working Group, I-SPY2 Investigator Network, L. Esserman, J. Chien, N. Hylton. Correlation between MRI background parenchymal enhancement and serum estradiol in premenopausal patients on neoadjuvant endocrine therapy for breast cancer [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS3-07-22.

Abstract Background: The PALLAS trial (ABCSG-42, AFT-05, BIG-14-03) investigates adjuvant palbociclib plus endocrine therapy (ET) versus ET alone in patients (pts) with HR+/HER2− early breast cancer. While palbociclib improves outcomes in metastatic settings, final analysis from PALLAS did not demonstrate benefit in survival outcomes from the addition of palbociclib to ET. A number of pts discontinued palbociclib before the planned 2 year duration though degree of palbociclib exposure was not associated with clinical outcomes. This analysis explores factors predictive of palbociclib early discontinuation and dose reductions, including baseline characteristics, toxicities, and patient-reported outcomes (PROs). Methods: Using the safety population of 5736 pts (2840 receiving palbociclib + ET), we assessed early palbociclib discontinuation within a competing-risks framework (competing events: progression, death). Cumulative incidence curves (Aalen-Johansen estimates) and cause-specific hazard models were used to explore associations with clinical and demographic factors (age, menopausal status, stage, race, endocrine therapy type), key toxicities (neutropenia, anemia, nausea), and baseline PROs (patient assessed quality of life [QoL], anxiety, fatigue, pain, frailty/ dependency, depression, and low income. financial difficulties). An automated model selection based on the Akaike information criterion (AIC) starting from the cause-specific hazard model incorporating all considered predictors including PROs. Results: Early discontinuation of palbociclib not due to progression or death occurred in 1171 pts (41.3%); an additional 343 pts (12.1%) stopped due to results of a futility analysis. In 108 (3.8%) and 2 (0.1%) pts treatment was discontinued due to progression or death. Median time to early discontinuation was 7.7 months in pts that discontinued treatment (median interquartile range 3.45-13.60). Very early discontinuation (≤3 months) occurred in 261 pts (9.2%). Dose reductions to 100mg and 75mg occurred in 1566 pts (55.2%) and 951 pts (33.6%) of 2840 pts, respectively. In univariable analyses, higher age, lower N-stage, and specific baseline PROs (baseline QoL, fatigue, anxiety, depression, frailty/ dependency and financial difficulties) were linked with increased risk of discontinuation. The best-fitting multivariable model identified age, nodal stage (N-stage), low quality of life (QoL), and low income as key predictors of early treatment discontinuation. Conclusions: In the PALLAS trial, our findings highlight the importance of integrating PROs into risk prediction of early treatment discontinuation. Additionally, when starting pts on oral targeted therapies in the adjuvant setting, symptom control, social work and financial counseling resources may support pts at higher risk of non-adherence. This is in line with findings from the ESMO Breast QoL analysis (Bjelic-Radisic et al.), which emphasized the added value of subjective patient-reported outcomes alongside objective clinical parameters in understanding treatment burden and adherence. Support: AFT, ABCSG, Pfizer, ECOG-ACRIN, NSABP Foundation, GBG, BIG; https://acknowledgments.alliancefound.org; ClinicalTrials.gov Identifier: NCT02513394. Citation Format: M. Vetter, A. Wiencierz, M. Gnant, D. Hlauschek, C. Kurzeder, C. Grasic Kuhar, N. Zdenkowski, E. Shinn, J. Suga, D. Egle, J. Meisel, S. Antolin-Novoa, E. Munzone, T. Haddad, S. Loibl, A. Holynskyj, C. Fesl, A. Dueck, A. DeMichele, E. Mayer, on behalf of the PALLAS groups and investigators (ABCSG, AFT, BIG,PReCOG, GBG, NSABP). Predictors of early discontinuation of adjuvant palbociclib in early HR+/ HER2- breast cancer: final analysis of the PALLAS trial integrating patient-reported outcomes [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS3-07-14.

Abstract Background: (Z)-endoxifen (ENDX) is an oral SERM that dually targets ERα (&gt; 5 ng/ml) and PKCβ1 (&gt;500 ng/ml). It has been tested in the endocrine resistant setting at doses of 20-360 mg/day; however, the optimal dose in an endocrine naïve population is not defined. We explored 10 mg/day dose of neoadjuvant ENDX in the Endocrine Optimization Pilot (EOP) of I-SPY2. Methods: EOP eligibility included pts with Stage II/III HR+ HER2- breast cancer, MammaPrint (MP) low. Pts with MP High1 scores were eligible if clinically node-negative. Pts received ENDX 10 mg orally once daily for 6 cycles (each cycle = 28 days). Premenopausal pts started ovarian function suppression on cycle 2, day 1. ENDX was continued until the day prior to surgery. The primary study endpoint was feasibility, defined as &gt;75% of pts completing &gt;75% of ENDX. Baseline (T0), 3-wk (T1) biopsies, and the surgical specimen (T3) were assessed centrally for Ki-67. Rates of modified preoperative endocrine prognostic index (mPEPI) 0 were reported, defined as pT1-2 pN0, Ki67&lt; 2.7%, or pathologic complete response. A genomic index for sensitivity to endocrine therapy (SET) index was done at T0. Breast MRI measurement of functional tumor volume (FTV) was performed at T0, T1, 12 weeks (T2), and pre-operatively (T3). Blood was collected for tumor informed ctDNA at T0, T1, T2, T3 and for pharmacokinetic (PK) steady state concentrations (Css) at T1. Safety data using CTCAE V5 is reported. Results: Between March 2023 and May 2024, 20 pts were enrolled in this arm. The median age was 52.5 years (range 32 -73), 13 (65%) were premenopausal, and 55% had clinically positive lymph nodes. Tumors were MPHigh1 in 3 pts (15%) and SET2,3 index high in 16 pts (80%). The primary endpoint was met with 95% of study participants completing ≥75% of therapy. The T0 median Ki-67 was 10.5% (range 2-60%). At T1 (n=19) and T3 (n=18), median Ki-67 was 5% (range 0-35%) and 5% (range 0-45%), respectively. The 3-week rate of Ki-67 ≤10% at T1 was 61.5% in premenopausal and 71.4% in postmenopausal women. The rate of Ki-67 ≤10% at surgery was 81.8% in premenopausal and 57.1% in postmenopausal women. An mPEPI score of 0 was observed in 1 (5%) of participants. The median T0 FTV was 8.9 ccs (range 0.9 to 251.1). The median T3 FTV was 1.3 ccs (range 0 to 11.2) and the median change in FTV from T0 to T3 was -72.0% (range -88.5% to -61.6%). ctDNA analysis was performed in 17 pts longitudinally across 64 blood samples.10 pts had positive ctDNA at T0, with 7 converting to ctDNA undetectable during treatment with ENDX. 1/7 pts who was ctDNA negative at T0 became positive at T1 and then reverted to undetectable ctDNA. The other 6 pts remained ctDNA negative at all timepoints. The median ENDX T1 Css was 79.3 ng/mL (range 73.6 - 906.4 ng /mL). The most common side effects reported were Grade 1 hot flush (70%), nausea (40%), fatigue (40%), and insomnia (35%). Grade 2 side effects included fatigue (15%) and night sweats (5%). No dose reductions or discontinuations due to toxicity occurred. Conclusions: ENDX at 10 mg was well tolerated and demonstrated reductions in Ki-67, MRI FTV, and ctDNA. The ISPY2 EOP is currently enrolling patients to ENDX 40 mg/day (intended to dually target ERα and PKCβ1) in combination with abemaciclib (with and without OFS). Citation Format: K. V. Giridhar, M. P. Goetz, C. Yau, A. M. Wallace, C. Falkson, J. Boughey, M. Arora, E. Stringer-Reasor, C. Nangia, C. Gallagher, A. Elias, A. Zimmer, M. S. Trivedi, C. Omene, L. van 't Veer, L. Brown Swigart, G. L. Hirst, W. Symmans, A. D. Borowsky, N. Onishi, N. Hylton, C. Vakalvas, M. Wei, J. M. Reid, S. Alhafaji, N. Chen, R. M. Mukhtar, L. Huppert, O. Opopade, L. J. Esserman, A. Chein. Low dose (Z)-endoxifen in the I-SPY2 Endocrine Optimization Pilot [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS3-07-10.

Abstract Guideline-concordant breast cancer treatment selection is increasingly complex due to rapid evolution in biomarker-driven therapies. Large language models (LLMs), such as GPT-4, offer a potential tool to support clinical decision-making. However, hallucinations and deviation from evidence-based guidelines remain significant concerns. This study evaluated whether retrieval-augmented generation (RAG) using National Comprehensive Cancer Network (NCCN) guidelines improves GPT-4 performance in treatment recommendation tasks. We developed a retrieval-augmented GPT-4 agent (RAG-GPT) by indexing 2025 NCCN Breast Cancer Guidelines into a vector database using sentence embeddings. Forty clinical vignettes were created from published case reports to reflect diverse scenarios, including HER2-low, triple-negative, and BRCA-mutated disease across all stages. Two models—a baseline GPT-4 (without augmentation) and the RAG-GPT—were prompted with structured templates requesting treatment plans, rationale, clinical trial options, and citations.Outputs were independently scored by two blinded oncology-trained reviewers using the modified Generative AI Performance Score (mG-PS). The mG-PS scored outputs for 2 categories, guideline concordance (Gold standard = 1.0, Acceptable = 0.5, Non-concordant = 0.0) and hallucination errors (Severe = -1.0, Moderate = -0.5, Mild = -0.25, None = 0.0). Guideline Concordance was deemed based on NCCN guidelines and what an oncologist would due based on using the guidelines for each case. Hallucinations are a term used for falsehoods in Artificial Intelligence, like made-up facts or studies, incorrect information, improper doses, or inappropriate or wrong treatments. To ensure consistency and minimize inter-rater variability, reviewers were provided with comprehensive scoring guidelines and detailed instructions for each evaluation metric. Scores were normalized between -1.0 and 1.0 for improved evaluation of tools. Readability and rationale clarity were rated on a 5-point Likert scale. Inter-rater reliability was assessed using Cohen’s kappa. The RAG-GPT demonstrated a significantly higher mean mG-PS score (0.599 ± 0.102, 95% CI: 0.498-0.702) compared to the baseline GPT-4, which not only scored lower but had a negative mean (-0.21 ± 0.143, 95% CI: -0.353 to -0.067). For the mG-PS Cohen’s Kappa test resulted in a score of .69 with substantial aggreament. An independent sample t-test showed a difference (t(78) = 29.1, p &lt; 0.0001). Severe hallucinations were also absent in RAG-GPT outputs (0/40) but occurred in 8/40 baseline GPT-4 outputs. The evaluation of readability and rationality also showed notable differences, with the augmented LLM scoring significantly higher (M = 4.35, SD = 0.66) compared to the naïve LLM (M = 3.05, SD = 1.06). An independent samples t-test confirmed this difference t(65.3) = 6.58, p &lt; 0.0001, further supporting the benefit of augmented Large Language models. Cohen’s Kappa test resulted in a score of .76 for a substantial agreement for readability and rational clarity scale. Limitations include small sample size; further fine-tuning and larger real-world validation studies remain necessary. Linking GPT-4 to NCCN guidelines via retrieval-augmented generation significantly improves treatment accuracy and reduces hallucination severity in breast cancer care scenarios. While further refinement is needed, the reduced incidence of severe hallucinations, increase in guideline-concordant recommendations, and enhanced readability findings support further development and prospective validation of guideline-anchored LLMs for clinical use. Citation Format: C. Yost, A. Aseem, B. Callas, S. Monick, L. Bonilla, R. Nguyen, R. Galamaga, S. Osborn, Y. Kumar, N. Ertz-Archambault. Retrieval-augmented GPT-4 improves NCCN-concordant breast cancer treatment recommendations [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS3-06-29.

Abstract Background: While synoptic reports have standardized breast cancer data, a wealth of critical information remains locked in unstructured free-text, including nuanced biomarker details and procedural notes. This data is not easily computable, creating a significant barrier to understanding comprehensive population-level statistics and subtle clinical trends. Traditional data extraction methods are often insufficient to capture this complexity. We developed and validated a novel, flexible AI-powered system using an agent-based architecture to extract, normalize, and discretize this valuable unstructured data, enabling previously infeasible population-scale analytics. Methods: Our system employs a modular, agent-based framework where each "agent" is a specialized large language model (LLM) tasked with a specific function in a sequential workflow. The process begins with an initiator agent that parses the synoptic report. Subsequent agents then perform targeted tasks, such as identifying and normalizing tumor characteristics, procedural details, and biomarker information from free-text fields. A final agent aggregates this information into a structured, computable format. To rigorously test our approach, we first developed a synthetic data generator to create realistic, anonymized report snippets with known ground-truth values. We then evaluated the performance of several leading LLMs of various sizes (e.g., gemini, llama3.3, phi4, deepseek-r1, gemma3) on this synthetic dataset. Results: Our early analysis on simulated synoptic reports demonstrated high accuracy in data extraction and normalization. The gemini-1.5-pro-002 model achieved an overall accuracy of 96.9%, with a sensitivity of 99.3%, specificity of 94.3%, and an F1-score of 97.0%. The llama3.3:70b model also showed strong performance with a 98.7% accuracy, (F1: 98.8%, sensitivity: 99.8%, specifically: 97.6%). Gemma3:27b was the least performant model tested (F1: 66.2%), substantially worse than much smaller models (deepseek-r1:1.5b and phi4:14b). A detailed error analysis of false positives and negatives has allowed us to iteratively refine the agent logic. By the time of the symposium, we will have analyzed a large cohort of real, de-identified synoptic reports and will present comprehensive performance metrics and initial findings on population-level trends. Conclusion: This AI-powered, agent-based system offers a powerful and scalable solution to overcome the limitations of structured data in synoptic reports. By accurately extracting and normalizing previously inaccessible free-text information, our approach has the potential to provide clinicians and researchers with near real-time insights into population-level statistics and trends in breast cancer. This will facilitate a more profound understanding of disease patterns and ultimately support evidence-based advancements in patient care and clinical research. Citation Format: S. N. Hart. A Novel Agent-Based AI System to Unlock Unstructured Data in Synoptic Reports for Advanced Population Health Analysis in Breast Cancer [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS3-06-11.

Abstract Introduction: Accurate detection and prioritization of tumor-specific somatic variants is foundational to the development of personalized immunotherapies. Traditional pipelines for neoepitope identification rely heavily on short-read next-generation sequencing (NGS) and DNA-based variant calling, followed by RNA expression filtering and HLA-binding prediction. However, this approach is limited by time-intensive workflows, potential false positives from DNA-only calls, and poor capture of complex isoforms or RNA-level editing. Moreover, the accuracy of neoepitope prediction is constrained by conventional sequence-based HLA binding algorithms that do not account for structural and biophysical features of peptide-MHC interactions, particularly across underrepresented HLA alleles. Methods: To overcome these limitations, we developed a streamlined, sequencing-to-vaccine framework that leverages long-read RNA sequencing for direct detection of expressed somatic variants, including fusions, indels, and single nucleotide variants. Using Oxford Nanopore Technologies’ cDNA-PCR approach, we enable comprehensive, full-length transcript characterization from both fresh and frozen breast tumor samples. Our bioinformatic pipeline, NeoTarget, integrates alignment, variant calling, and neoepitope inference from tumor RNA in a single, modular workflow. A core innovation of this pipeline is HLA-Inception, a deep convolutional neural network trained to predict peptide-MHC class I binding using features derived from molecular dynamics-based modeling of electrostatic potential at the peptide-HLA interface, enhancing generalizability across diverse HLA subtypes. Results: Expanding on previous results, the current NeoTarget pipeline identifies, evaluates, and prioritizes single nucleotide variants, indels, and fusions. To validate this system, we used a set of 9 human breast cancer tumors benchmarked with short read sequencing. Identified neoantigens included most variants shown by clinical short read sequencing and also novel structural variants (i.e., fusions) only revealed with long read sequencing with high confidence. Conclusion: Here, we describe the design and performance of the NeoTarget pipeline, and show proof-of-concept feasibility of applying this fast and cost-effective workflow for HLA-restricted epitope discovery in breast tumors using long-read RNA sequencing. Overall, these methods can be applied to rapidly identify and prioritize neoantigens for personalized neoantigen-based immune therapies. Citation Format: F. Batalini, S. Henry, Y. Zhang, R. Foster, J. M. Gayton, A. Hostal, E. A. Wilson, A. Singharoy, M. Gustafson, B. Pockaj, J. Park, K. S. Anderson. Somatic variant detection using long read RNA sequencing reveals HLA-restricted targets for immunotherapy [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS3-06-08.