BackgroundPrognostic information is essential for decision-making in breast cancer management. In recent years, trials and clinical practice have emphasized genomic prognostication tools, despite clinicopathological methods being more affordable and accessible. PREDICT v3 is one such tool with promising results across cohorts. Advances in machine learning (ML), transfer learning, and ensemble methods provide opportunities to enhance these approaches, especially where missing data and model assumptions differ across diverse populations. ObjectiveThis study evaluates the potential to improve survival prognostication in breast cancer. More precisely, we compare de novo ML, transfer learning from the pretrained prognostication model PREDICT v3, and a stacked ensemble approach. MethodsData from the MA.27 trial (NCT00066573) were used for model training, with external validation on data from the Tamoxifen Exemestane Adjuvant Multinational trial (NCT00279448 and NCT00032136) and a US Surveillance, Epidemiology, and End Results cohort. Transfer learning was applied by re-estimating the parameters of the pretrained prognostic tool PREDICT v3. De novo ML included random survival forests and extreme gradient boosting, and the ensemble was implemented using weighted linear stacking of model predictions. Internal and external validation was assessed in terms of the integrated calibration index and discrimination. Shapley Additive Explanations values were used to explain model predictions and decision-curve analysis to facilitate the interpretation of performance differences. ResultsTransfer learning, de novo random survival forest, and the stacked ensemble improved calibration in MA.27 over the pretrained model (integrated calibration index reduced from 0.042 in PREDICT v3 to ≤0.007) while discrimination remained comparable (AUROC increased from 0.738 in PREDICT v3 to 0.744-0.799). In decision-curve analysis, these approaches demonstrated consistently positive net benefit across clinically relevant thresholds, while PREDICT v3 lost net benefit beyond 7.5% predicted risk. Invalid PREDICT v3 predictions were observed in 23.8% to 25.8% of MA.27 individuals due to missing information. In contrast, ML models and the stacked ensemble predicted survival despite missing data. Across all models, patient age, nodal status, pathological grading, and tumor size had the highest Shapley Additive Explanations values, indicating their importance for survival prognostication. External validation in the US Surveillance, Epidemiology, and End Results cohort confirmed the benefits of transfer learning, RSF, and ensemble in terms of calibration while maintaining discrimination at comparable levels. In contrast, generalizability was limited in the Tamoxifen Exemestane Adjuvant Multinational trial, a cohort with a substantially different distribution of clinicopathological characteristics. ConclusionsThis study demonstrates that transfer learning, de novo RSF, and a stacked ensemble can improve prognostication compared with the pretrained PREDICT v3, particularly in the presence of missing or uncertain inputs. Transportability may be limited in cohorts with different clinicopathological profiles, requiring local validation before clinical deployment. Ultimately, better survival estimation can provide more meaningful guidance in breast cancer care. Trial RegistrationClinicalTrials.gov NCT00066573; https://clinicaltrials.gov/study/NCT00066573, NCT00279448; https://clinicaltrials.gov/study/NCT00279448, NCT00032136; https://clinicaltrials.gov/study/NCT00032136

e12633 Background: The Keynote-522 trial established neoadjuvant chemo-immunotherapy as a new standard of care for patients with early-stage triple negative breast cancer (TNBC). However, questions remain regarding the optimal chemotherapy backbone, duration of immunotherapy and the potential for chronic toxicities in this curative population. This study aimed to identify Canadian practices for the management of early-stage TNBC, areas of clinical uncertainty and interest in future clinical trials. Methods: A voluntary electronic survey was distributed to Canadian medical oncologists via an email database compiled by the Ottawa Hospital Research Institute, REthinking Clinical Trials (REaCT) program. The survey focused on prescribing practices prior to and following the Keynote-522 trial and the incorporation of immunotherapy into the neoadjuvant treatment of early-stage TNBC. The analysis is reported descriptively. Results: Out of the 50 oncologists from across Canada who responded to the survey, 36 were eligible to complete the survey in its entirety. Among respondents, 83.3% adopted the Keynote-522 chemotherapy backbone as their first choice for early-stage TNBC. However, only 47% reported following the trial protocol exactly. Common modifications included utilizing dose-dense (every 2-week) administration of anthracycline and cyclophosphamide (AC) (50%) with 6-week pembrolizumab dosing intervals (50%). Concerns regarding toxicity (39%) and schedule impracticality (25%) were primary drivers for protocol divergence. While 86% of physicians prescribe immunotherapy for eligible stage II/III TNBC, 50% would also offer it for Stage I (T1N0) disease. Areas of clinical uncertainty were highlighted as the management of stage I disease and the tolerability of the Keynote-522 protocol in the elderly and comorbid patients. Optimal adjuvant treatment practices also remain uncertain, particularly in the setting of patients with pathological complete response (pCR), with 39% of respondents indicating they felt there was currently insufficient data to support the continuation of pembrolizumab in the adjuvant setting. Finally, 100% of respondents indicated a need for more response-adapted regimens to mitigate treatment-related toxicities. Conclusions: Canadian oncologists have rapidly adopted neoadjuvant immunotherapy for early-stage TNBC as standard of practice but frequently modify the chemotherapy backbone to improve tolerability and clinical efficiency. Significant uncertainty persists regarding the treatment of stage I disease and elderly or comorbid patients, in addition to optimal adjuvant treatment practices. While research is ongoing to address these issues, more is needed, particularly efforts to support response-adapted treatment approaches.

Despite the paucity of high-quality data supporting its benefits, routinely scheduled, in-person post-treatment surveillance of early breast cancer (EBC) patients remains common. Evaluation of different follow-up strategies is required. We present the feasibility phase of an ongoing randomized controlled trial (RCT) comparing two different follow-up strategies. Patients with EBC who completed the acute phase of their treatment were randomized to receive either personalized or ASCO guideline-based follow-up care alone. Feasibility endpoints, including rate of accrual, physician participation, patient acceptance of randomization arm, and patient retention 1 year after randomization, are presented. Of 279 patients approached, 261 (93.5%) were eligible and provided consent. Median rate of accrual was 34.5 patients per month, and all healthcare providers who agreed to study participation (n = 11) approached patients. Patients were randomized to receive personalized (n = 131) or guideline-based (n = 130) follow-up. No patients declined their randomization arm. For all 261 randomized patients, the 1-year participant retention rate was 92.0% (240/261). This RCT confirms both patient and healthcare-provider enthusiasm for studies comparing different strategies for post-treatment surveillance. No patients withdrew consent post-randomization due to a preference for one study arm over the other. While clinical effectiveness and patient satisfaction remain to be analyzed, our reported rates of attrition can be used by others when designing similar studies.

BACKGROUND:Three-year data from a randomized trial comparing a single zoledronate infusion with infusions every 6 months were previously reported showing that a single infusion was associated with increased patient convenience, less toxicity, and lower rates of treatment discontinuation at 3 years. We now report the trial's final, prespecified 5-year follow-up data. METHODS:Patients who were postmenopausal (either naturally or treatment induced) with early breast cancer were randomly assigned to receive a single infusion of zoledronate (4 mg intravenously) or an infusion every 6 months for 3 years. Final 5-year preplanned secondary outcomes included recurrence-free survival, bone metastasis-free survival, overall survival, confirmed osteonecrosis of the jaw, and fragility fracture rates. RESULTS:A total of 211 patients were randomly assigned to receive either a single infusion (n=107) or an infusion every 6 months (n=104). After 5 years of follow-up, there were no clearly apparent differences between the groups for recurrence-free survival (88.7% vs. 84.1%; hazard ratio, 0.71; 95% confidence interval [CI], 0.34-1.51), bone metastasis-free survival (90.6% vs. 86.0%; hazard ratio, 0.68; 95% CI, 0.30-1.52), or overall survival (91.6% vs. 88.0%; hazard ratio, 0.79; 95% CI, 0.34-1.82) in the single-infusion group versus the group receiving infusions every 6 months, respectively. After 5 years of follow-up, there were no significant differences in the incidence of osteonecrosis of the jaw (0% vs. 0%) or fragility fractures (5% vs. 2%) in the single-infusion group versus the 6-monthly infusion group, respectively. CONCLUSIONS:At 3 years of follow-up, a single infusion of zoledronate had previously been shown to be associated with increased patient convenience, less toxicity, and lower rates of treatment discontinuation compared with treatment every 6 months. In the current analysis, extended to 5 years of follow-up, there was no clear evidence of a difference in survival outcomes between the treatment approaches. (Funded by the CURE Foundation, REthinking Clinical Trials Program at the Ottawa Hospital Research Institute, and the Ottawa Hospital Foundation and its generous donors. Participating sites also received accrual support from the Canadian Cancer Clinical Trials Network [3CTN].).

Abstract Drug-related UVA-induced photoreactions have been reported for several therapeutic compounds, including fluoroquinolones. CX5461 is a clinically relevant quinolone-derived anti-cancer small molecule with documented UVA-sensitising activity. Here we compared, by bulk and clonal whole genome sequencing (WGS) under light-protected conditions, the mutational signatures in human retinal pigment epithelial cells (RPE1) exposed to UVA, CX5461, or co-exposed to UVA and CX5461. Treatment with CX5461 or UVA alone resulted in a low SNV burden and background-like mutational profiles. In contrast, bulk sequencing of human cells co-exposed to UVA and CX5461 had a markedly higher SNV burden characterized by T>A and T>C substitutions. Furthermore, single-cell clonal expansion and sequencing of CX5461 alone, UVA alone or CX5461+UVA treatments confirmed that the pattern was only observed when cells were exposed to both UVA and CX5461. The CX5461+UVA-associated SNV signature we report arises only when CX5461-treated cells are exposed to UVA, and is not observed when CX5461-treated cells are shielded from light. We do not observe strong single base mutagenic activity of CX5461 alone, under light protected conditions. Our data emphasise the need for appropriate controls and light-exposure precautions when studying base mutagenesis activity of known photosensitiser molecules.

6017 Background: Primary immunotherapy (IO) with radiotherapy remains an investigational approach in HPV+ LA-OPSCC. The CCTG-led international phase II, randomized, non-comparative, HN.9 (NCT03410615) trial evaluated the efficacy of the of durva+RT as a chemo-sparing approach in patients (pts) with intermediate-risk, HPV+, LA-OPSCC. Methods: Pts with newly diagnosed, PDL1-unselected, pathologically proven, treatment-naïve HPV+ LA-OPSCC (UICC/AJCC 8th Edition T1-2N1 or T3N0-1 smokers [≥10 pack years] or T1-3 N2 with any smoking history) eligible for definitive CRT were randomized (1:2) to Arm A (CRT: 70Gy/35F + cisplatin 100 mg/m2 Q3W on days 1,22,43 of RT) vs Arm B (durva IV 1500 mg, days -7, 22 of RT, followed by adjuvant durva for 6 doses). Primary objective was 3-year EFS in Arm B (efficacy if one-sided 90% CI lower bound [LB] >83%). Secondary objectives: QoL (MDADI, FACT-HN at baseline, end of RT, 3,6,12,24 and 36 mo), OS, safety, distant metastasis-free survival (DMFS) and locoregional control (LRC). Safety was assessed per CTCAE v5. The trial closed early based on emerging external efficacy data of IO in HN LA setting, enrolling 129 pts overall (80 of the planned 120 in Arm B) with approximately 80% power retained for the primary analysis. Results: 129 pts were randomized across 21 Canadian and European sites. Baseline characteristics were well balanced between arms. At the data cutoff (Sep 19, 2025), median follow-up time was 55.7 mo. In both arms, all pts received the planned RT schedule. In Arm A, the median number of cisplatin cycles was 2 (21% received <200m/m 2 and 79%≥ 200 mg/m 2 ). In arm B, all pts (100%) received durva concurrent to RT. 126 pts were eligible for efficacy analysis. The estimated 3-year EFS in Arm B was 80% (one-sided 90% CI: LB 73%); the prespecified efficacy criterion (LB >83%) was not met. The 3-year EFS in Arm A was 89%. The estimated 3-year OS rates were 92% in both arms. At 3 years, LRC were 97 and 91% in Arms A and B respectively. DMFS was 89% in Arm A and 86% in Arm B. Grade ≥ 3 adverse events any time during trial were similar between Arm B (69%) and Arm A (63%). QoL completion was 96% at baseline and 80%+ throughout; acute worsening during RT followed by gradual recovery (slower in arm B during adjuvant durva), with longer-term persistence of isolated issues (hearing loss, dry mouth) was similar to published experience for drug+RT regimens. Conclusions: Durva+RT did not improve 3-year EFS in the overall patient population of intermediate-risk HPV+ OPSCC. No new safety signal was observed. Specific toxicities varied by arm, but overall QoL experience, including swallowing, was similar between arms and with prior trials. Subgroup analysis, blood, tissue and microbiome-based correlates are ongoing. Clinical trial information: NCT03410615 .

Drug-related UVA-induced photoreactions have been reported for several therapeutic compounds, including fluoroquinolones. CX-5461 is a clinical stage quinolone-derived anti-cancer small molecule with documented UVA-sensitizing activity. Here, we compared, by bulk and clonal whole-genome sequencing under extraneous light-protected conditions, the mutational signatures in human retinal pigment epithelial cells (RPE1) exposed to UVA, CX-5461, or co-exposed to UVA and CX-5461. Treatment with CX-5461 or UVA alone resulted in a low single-base mutation burden and background-like mutational profiles. In contrast, bulk sequencing of human cells co-exposed to UVA and CX-5461 had a markedly higher mutation burden characterized by T > A and T > C substitutions. Furthermore, single-cell clonal expansion and sequencing of CX-5461 alone, UVA alone, or CX-5461+UVA treatments confirmed that the pattern was only observed when cells were exposed to both UVA and CX-5461. The CX-5461+UVA-associated SBS signature we report arises only when CX-5461-treated cells are exposed to UVA, and is not observed when CX-5461-treated cells are shielded from light. We do not observe strong single-base mutagenic activity of CX-5461 alone, under light-protected conditions. Our data define a human SBS mutagenesis signature for CX-5461 potentiated UVA mutations and emphasize the need for appropriate controls and light-exposure precautions when studying SBS activity of known photosensitizer molecules.

Despite widespread use of clinico-pathologic and genomic risk scores in early breast cancer (EBC), questions remain as to whether they are predictive, prognostic, or both. We evaluated risk score performance with actual patient outcomes in a trial dataset. Discrimination and calibration of PREDICT 2.1 and PREDICT v3 for overall survival (OS) and RSClin for distant metastasis-free survival (DMFS) were compared with actual patient outcomes in 645 postmenopausal, node-negative, hormone-positive patients with EBC from the TEAM pathology substudy. Estimated chemotherapy benefit (low < 3 ClinicalTrials.gov , NCT00279448, NCT00032136, and NCT00036270; NTR 267; Ethics Commission Trial 27/2001; and UMIN, C000000057

Abstract Background: Over two dozen histological types of salivary gland cancers are recognized comprising 3-5% of all head and neck cancers. Response to chemotherapy and actionable target expression in patients with metastatic disease vary by histological type e.g. HER2 and/or androgen receptor are often overexpressed by salivary duct carcinomas and NTRK fusions often present in secretory carcinomas. Overall results of this phase II basket trial have been previously reported (Gupta 2025). Herein we report additional data from the metastatic salivary gland cohort by histological subtype. Methods: IND.228 was a prospective multi-center, non-blinded, open-label phase II basket trial that enrolled 8 cohorts of patients with rare cancers including a salivary gland cohort. Eligible patients had incurable disease with no known life-prolonging treatment options or contraindications to immunotherapy and ECOG performance status 0 or 1. Tumor PD-L1 expression was not required and patients with adenoid cystic carcinoma were excluded. Patients received durvalumab 1500 mg IV plus tremelimumab 75 mg IV q4 weeks for 4 cycles followed by durvalumab q4 weeks until disease progression. The primary outcome was objective response by RECIST 1.1 criteria. Results: 21 patients with incurable salivary gland cancer were enrolled. Of 20 patients evaluable for response there were 6 females and 14 males, median age 61 years (range, 40-72) with a median of 3 metastatic sites (range, 1-5). Histological types were salivary duct carcinoma (8), acinic cell carcinoma (5), adenocarcinoma (3), clear cell carcinoma (2), and others (2). All had diagnosis confirmed by central pathology review. 190 cycles of treatment were given with a median of 7 cycles per patient (range, 1-40). Four patients had partial responses, 8 had stable disease, and 8 had progressive disease as best response. The proportions of response and response plus stable disease by histological type, respectively, were salivary duct carcinoma (2/8, 4/8), acinic cell carcinoma (1/5, 4/5), and adenocarcinoma (1/3, 3/3). 10/13 patients with avaialable tumor grade had high-grade cancers. Progression-free survival ranged from 7.3-59.3+ months in responding patients, and from 3.8-11.4 months in stable disease patients. One patient with salivary duct carcinoma received 40 cycles of treatment and had a response duration of over 5 years. Conclusions: Patients with salivary duct carcinoma, acinic cell carcinoma, and adenocarcinoma appeared to benefit from combined CTLA-4 and PD-L1 immune checkpoint blockade. Immune checkpoint blockade deserves further study and consideration as a treatment option in these histological types of incurable salivary gland cancer. Citation Format: Eric Winquist, John Hilton, Christian Kollmannsberger, Danielle Charpentier, Dorie-Anna Dueck, Hal Hirte, Sebastien Hotte, Rahima Jamal, Raymond Jang, Andrew Maksymiuk, Randeep Sangha, Stephanie Snow, Osama Souied, Jennifer Spratlin, Ralph Wong, Abha Gupta, Thierry Alcindor, Quincy Chu, Derek Jonker, Torsten Nielsen, Ming Tsao, Tricia Cottrell, Joana Sederias, Siwei Zhang, Wei Tu, Janet Dancey. Durvalumab and tremelimumab in patients with metastatic non-adenoid cystic salivary gland cancer treated in the CCTG IND.228 phase II basket trial [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Breaking Barriers in the Fight against Rare Cancers; 2026 Jul 18-20; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2026;86(14_Suppl):Abstract nr A029.

Understanding real-world treatment patterns and their effectiveness in HR + HER2- advanced breast cancer (aBC) in Canadian patients. This was a multi-center, observational, prospective cohort study including men and pre-/peri-/postmenopausal women with HR + HER2- aBC receiving endocrine therapy (ET) or ET + targeted therapy (ET + TT). The primary objective was duration of treatment (DOT) with ET and ET + TT. Sequence of therapies, treatment patterns, and Overall Survival (OS) were also evaluated. DOT was prolonged in patients receiving ET + TT compared to ET (median DOT: ET + TT 397 days vs ET 192 days; Log-Rank test p value < .0001; HR = 0.66; 95

Clinico-pathologic (including nomograms) and genomic tools are widely used to determine prognosis and predict benefit from treatment in early breast cancer (EBC). However, little is reported on patient perceptions of these tools and whether they enhance understanding of their individual risk of recurrence or benefits from therapy. Patients with EBC were surveyed to evaluate the use of prognostic/predictive tools in clinical practice and their self-reported recurrence risk. Their actual risk of local, contralateral and distant recurrence was estimated using the INFLUENCE 2.0 tool. Information was also collected on key aspects that patients wanted these tools to address, as well as the anticipated benefits that would make receiving chemotherapy worthwhile from their perspective. Completed surveys were received from 210 patients. Despite the use of NHS PREDICT 2.1 in 50

BACKGROUND:Granulocyte colony-stimulating factors (G-CSFs), including filgrastim and pegfilgrastim, are associated with bone pain, potentially impacting treatment adherence. This study hypothesized that a 5-day regimen of filgrastim would result in less bone pain than single-dose pegfilgrastim in patients receiving chemotherapy for early breast cancer. METHODS:In this multicenter, open-label, randomized controlled trial, patients requiring prophylactic G-CSF during chemotherapy were randomly assigned 1:1 to receive either 5-day filgrastim or pegfilgrastim. The primary outcome was patient-reported bone pain, assessed as area under the curve of daily pain scores (0 = no pain to 10 = worst pain) over the first 5 days following G-CSF in cycle 1. Secondary outcomes included bone pain in cycles 2-4, febrile neutropenia, hospitalizations, chemotherapy delays, dose reductions, early discontinuations, chemotherapy-related deaths, health-related quality of life, and health-care resource utilization. RESULTS:From June 2021 to March 2023, a total of 233 patients were randomly assigned, with 219 analyzed (110 filgrastim and 109 pegfilgrastim) after excluding those who withdrew before receiving treatment. Adjusting for stratification factors and prespecified baseline covariates using repeated measures linear regression, the mean area under the curve (0-40) for cycle 1 bone pain was 10.2 (11.2) for 5-day filgrastim and 10.2 (9.81) for pegfilgrastim, with an adjusted mean difference of 0.70 (95% confidence interval = 1.62 to 3.02; P = .556). Although no clinically significant differences were observed in most secondary outcomes, the 5-day filgrastim group exhibited a numerically higher incidence of febrile neutropenia (6.4% vs 0.9%, P = .065) and hospitalization (10.0% vs 3.7%, P = .106). CONCLUSION:There was no significant difference in bone pain between 5-day filgrastim and pegfilgrastim.

The combination of CDK4/6 inhibitor+endocrine therapy (CDK4/6i+ET) is standard 1st-line (1L) systemic treatment for patients with ER+/HER2-negative metastatic breast cancer (MBC). Numerous agents are available / in development for subsequent lines of treatment. Circulating tumor DNA (ctDNA) via liquid biopsy presents a promising, non-invasive approach for blood-based tumor genotyping, patient stratification and response assessment with the potential to enhance biomarker-driven strategies and aid in development of new therapeutics. IND.241 is a master protocol platform design consisting of independent substudies monitoring patients with ER+/HER2- MBC prior to progression (PD) on CDK4/6i+ET and investigating novel agents or drug combinations in 2nd/3rd lines (2L, 3L) after progression on CDK4/6i+ET. The primary objective of the novel drug/combination substudies is to centrally interrogate ctDNA (Tempus xF+, a 523-gene liquid biopsy panel) and evaluate whether biomarker selection improves ORR or CBR (RECIST 1.1). Secondary objectives include safety and toxicity profile, PFS, and OS. The monitoring substudy (Substudy A) enrolls patients currently on CDK4/6i+ET treatment and aims to characterize molecular profile, clinical features, and ctDNA dynamics of acquired resistance. This platform trial enables creation and maintenance of a tissue and data bank including clinical data, genomics, and radiomics to evaluate surrogates of treatment outcomes and potential biomarkers of response, resistance, and disease progression. Patients with specific biomarkers detected in ctDNA will be enrolled to corresponding biomarker positive cohorts of substudies. Patients with no substudy-specific biomarkers are randomized to biomarker negative cohorts of available substudies. Treatment substudies follow a 2-stage design. Currently, the monitoring substudy A is actively accruing. Substudy B is evaluating lunresertib (PKMYT1 inhibitor) + gemcitabine in patients +/-CCNE1 overexpression / amplification. Substudy C is evaluating niraparib (PARP inhibitor) + fulvestrant (ET) in patients +/- alterations in BRCA1/2 (germline/somatic) or PALB2 (germline). These latter two substudies have now closed to accrual, with efficacy and safety evaluation ongoing. Substudy D, recently added, is evaluating lunresertib + camonsertib (ATR inhibitor) in patients +/- CCNE1 overexpression/amplification, FBXW7 or PPP2R1A alterations. Additional substudies are in development for this platform trial. IND.241 employs a biomarker-driven platform trial designed to evaluate CDK4/6i resistant ER+/HER2- MBC, using non-invasive ctDNA tumor genotyping to select patients for biomarker-driven substudies with the goal of defining the ctDNA landscape across 1L to 3L treatments. Moira Rushton, Dave W. Cescon, Nathalie Levasseur, Michelle B. Nadler, Daniel Rayson, Brooke E. Wilson, Sara K. Taylor, Rossanna Pezo, Arif Awan, Zahi Mitri, Vikaash Kumar, Jennifer Melvin, John Hilton, Stephen Chia, Philippe L. Bedard, Spencer Finn, Wei Tu, Courtney Coschi, Philippe Jamme, Elia Aguado-Fraile, Lesley Katie Seymour, Laura Pearce. IND.241: A Canadian cancer trials group liquid-biopsy informed platform trial to evaluate treatment in CDK4/6-inhibitor resistant ER+/HER2- metastatic breast cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_2):Abstract nr CT239.

Prognostic information is essential for decision-making in breast cancer management. Recently trials have predominantly focused on genomic prognostication tools, even though clinicopathological prognostication is less costly and more widely accessible. Machine learning (ML), transfer learning and ensemble integration offer opportunities to build robust prognostication frameworks. We evaluate this potential to improve survival prognostication in breast cancer by comparing de-novo ML, transfer learning from a pre-trained prognostic tool and ensemble integration. Data from the MA.27 trial was used for model training, with external validation on the TEAM trial and a SEER cohort. Transfer learning was applied by fine-tuning the pre-trained prognostic tool PREDICT v3, de-novo ML included Random Survival Forests and Extreme Gradient Boosting, and ensemble integration was realized through a weighted sum of model predictions. Transfer learning, de-novo RSF, and ensemble integration improved calibration in MA.27 over the pre-trained model (ICI reduced from 0.042 in PREDICT v3 to <=0.007) while discrimination remained comparable (AUC increased from 0.738 in PREDICT v3 to 0.744-0.799). Invalid PREDICT v3 predictions were observed in 23.8-25.8

Background: Vepdegestrant (ARV-471), an oral PROTAC ER degrader, was well tolerated and showed evidence of clinical activity in the first-in-human phase 1/2 study in heavily pretreated patients (pts) with ER+/HER2- advanced breast cancer (NCT04072952). In xenograft models, vepdegestrant plus the cyclin-dependent kinase (CDK)4/6 inhibitor abemaciclib showed greater tumor growth inhibition vs fulvestrant plus abemaciclib. The open-label, phase 1b/2 TACTIVE-U umbrella study is evaluating the safety, efficacy, and pharmacokinetics (PK) of vepdegestrant in combination with other anticancer treatments in pts with ER+ advanced or metastatic breast cancer. Here, we report preliminary phase 1b results from the ongoing TACTIVE-U sub-study A (NCT05548127) investigating vepdegestrant plus abemaciclib. Methods: Eligible pts were aged ≥18 years with confirmed ER+/HER2- advanced or metastatic breast cancer and had received up to 2 lines of prior therapy, including a CDK4/6 inhibitor-based regimen in any setting (pts with prior CDK4/6 inhibitor dose reductions due to adverse events [AEs] were excluded). Following a 7-day PK lead-in with abemaciclib administered alone, pts received vepdegestrant 200 mg orally once daily and abemaciclib 150 mg orally twice daily (both continuously). The primary endpoint of the phase 1b portion of the study was dose-limiting toxicities (DLTs) during the first cycle. Secondary endpoints of phase 1b are assessments of safety (type, frequency, and severity of AEs), PK (plasma concentrations of study drugs), and antitumor activity (objective response, clinical benefit rate). Results: As of April 4, 2024, 16 female pts (median age: 56 years [range: 38–79]) received vepdegestrant plus abemaciclib, with 9 pts treated for ≥5 cycles; 12 remain on treatment. All pts (100%) received prior CDK4/6 inhibitors (ribociclib [n=8], palbociclib [n=7], and abemaciclib [n=1]), 14 (88%) prior aromatase inhibitors, 11 (69%) prior chemotherapy, and 5 (31%) prior fulvestrant. There were no DLTs or treatment discontinuations due to treatment-emergent AEs (TEAEs). TEAEs led to dose reduction of vepdegestrant and abemaciclib in 1 pt (neutropenia) and dose reductions of abemaciclib in 6 pts (fatigue [n=2] and fatigue/muscular weakness, neutropenia, diarrhea, and anemia [n=1 each]). Treatment-related AEs (TRAEs) to either vepdegestrant or abemaciclib that occurred in ≥20% of pts were diarrhea (69%), nausea (44%), fatigue (44%), and decreased appetite (25%). The majority of TRAEs were grade 1/2; grade 3 TRAEs were neutropenia or neutrophil count decreased in 5 pts and fatigue in 2 pts. There were no grade ≥4 TRAEs. Observed PK of vepdegestrant and abemaciclib was consistent with historical data reported in single-agent clinical studies. The exposure of abemaciclib (Cmax and AUCtau) increased slightly (15% and 13%, respectively) when dosed with vepdegestrant compared with abemaciclib alone. Conclusions: The safety profile of vepdegestrant plus abemaciclib in pts with ER+/HER2- advanced or metastatic breast cancer was generally consistent with the known profiles of each agent, and no DLTs were observed. Neutropenia was manageable with dose modifications. The impact of vepdegestrant on abemaciclib exposure was minor and indicated no significant drug interaction. These data support further evaluation of this combination at the full standard doses of vepdegestrant 200 mg once daily and abemaciclib 150 mg twice daily in the phase 2 portion of this sub-study. Additional ongoing sub-studies for TACTIVE-U are evaluating vepdegestrant plus ribociclib (sub-study B; NCT05573555) and samuraciclib (sub-study C; NCT06125522). Citation Format: John Hilton, Katarzyna J. Jerzak, A. Jo Chien, Jose L.M. Guimaraes, Colombe Chappey, Hechuan Wang, Stefanie K. Drescher, Elisa Dall’O’, Julia Perkins Smith, Swapnil Parmar, Rachel M. Layman, Cynthia Ma. Vepdegestrant, a PROteolysis TArgeting Chimera (PROTAC) Estrogen Receptor (ER) Degrader, Plus Abemaciclib in ER-Pos/Human Epidermal Growth Factor Receptor 2 (HER2)-Negative Advanced or Metastatic Breast Cancer: TACTIVE-U Prelim Phase 1b Results [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2024; 2024 Dec 10-13; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(12 Suppl):Abstract nr P4-12-03.

PURPOSE:An important goal of routinely scheduled physical examination in the post-treatment surveillance of patients with early breast cancer (EBC) is to allow for earlier detection of potentially curable recurrences. We present an updated analysis of recurrence detection, and its potential for curative management, at a single-centre survivorship program that follows ASCO recommendations. METHODS:Patients with EBC referred to the Wellness Beyond Cancer Program who had breast cancer recurrence between February 2013 and June 2023 were reviewed. Descriptive analyses were used to present patient and disease characteristics stratified by type of recurrence and mode of cancer detection. RESULTS:Of 389 first recurrences, 250 were distant (64.3%), 75 local (19.3%), and 64 (16.5%) contralateral new breast primaries. Distant recurrences were primarily detected via patient-reported symptoms (220/250, 88.0%). 42.7% (32/75) of ipsilateral breast recurrences were detected by patients and 53.3% (40/75) by routine imaging. Contralateral breast primaries were primarily detected by routine imaging 71.9% (46/64) and patient-reported symptoms 25.0% (16/64). 2.1% (8/389) recurrences were detected by healthcare providers, 3 were coincidental intraoperative findings and 1.3% (5/389) were detected by healthcare providers at routinely scheduled follow-up visits. Of the 5 recurrences detected at routinely scheduled follow-up visits, only 0.3% (1/389) was potentially curable. CONCLUSION:Despite following ASCO guidelines, potentially curable recurrences are rarely detected by healthcare providers at routinely scheduled follow-up appointments. Our data suggests that approximately 87,670 follow-up visits were required for healthcare providers to detect one curable recurrence. Updated evidence-based guidelines are required to optimise the post-treatment surveillance procedures.