Citations

815 Background: EV, alone or combined with pembrolizumab (EVP), is a standard of care for patients (pts) with aUC. Markers predictive of early or durable treatment response in aUC remain limited. Here, we investigate serial ctDNA monitoring and TP53 VAF as complementary predictors of early EV/EVP response. Methods: This single-institution IRB-approved study included pts with aUC treated with EV/EVP between 1/2021 and 9/2025. Pt demographics, clinicopathologic features, serial Signatera ctDNA results, and somatic tissue next-generation sequencing (NGS) data were abstracted from medical records. Primary refractory disease (PRD) was defined as progressive disease on first restaging scan or earlier clinical progression resulting in death from disease. PRD and progression-free survival (PFS) were assessed using Fisher’s exact and log-rank tests. Results: Out of 158 pts, 67 pts were included in study cohort. Median follow-up was 13.2 months, with 52% of pts alive at time of data cutoff. Clinical characteristics at EV start include median age 69 years (range 41-99), 78% male, 91% white, 15% variant-predominant histologic subtype, 15% upper tract disease, 43% with visceral metastasis (24 lung, 7 liver), 66% frontline treatment (42 EVP, 2 EV), and 34% refractory setting (6 EVP, 17 EV). At first restaging scan, observed response rate was 50%; 23% had PRD. TP53 alterations on NGS were identified in 39 pts (58%); TP53 VAF was available for 36 pts (92%), with median VAF 50.7% (range 8-82%). While TP53 mutation presence did not significantly associate with PFS (hazard ratio [HR] 1.28, 95% CI 0.68-2.42; log-rank p = 0.45), pts with TP53 VAF ≥ 50% had significantly longer PFS compared with TP53 VAF < 50% (HR 0.44, 95% CI 0.18-1.04; log-rank p = 0.03). PRD was more commonly observed with TP53 VAF < 50% than with TP53 VAF ≥ 50% (odds ratio [OR] 0.06, 95% CI 0.01-0.48, p = 0.01). 23 pts (21 EVP, 2 EV) had ≥ 2 serial ctDNA draws ≤ 120 days apart, including baseline draw within 2 months of EV start. Median time between 1 st and 2 nd ctDNA draws was 43 days. 70% (16 pts) had ≥ 90% ctDNA decline and 17% (4 pts) had ctDNA rise. The 2 pts with PRD exhibited ctDNA rise, whereas no pts with ctDNA decline had PRD (p = 0.04). 10 of these pts (43%) had TP53 mutations with available VAF. Of those 5 pts with TP53 VAF ≥ 50%, 3 had ctDNA clearance over 90%, 1 had 0-90% ctDNA clearance, and 1 had ctDNA rise. Of those 5 pts with TP53 VAF < 50%, 2 had ctDNA clearance over 90%, 1 had 0-90% ctDNA clearance, and 2 had ctDNA rise. Conclusions: Integrated serial ctDNA monitoring and tumor TP53 VAF may capture early treatment sensitivity to EV/EVP. Limitations include retrospective analysis, small sample size/power, combined EV/EVP cohort, and lack of centralized imaging review. Further validation in large, prospective cohorts is warranted.

In this 5-year follow-up from the CheckMate 743 study in patients with unresectable pleural mesothelioma (PM), we evaluated updated efficacy and safety outcomes and biomarkers and performed treatment-switching analyses with first-line nivolumab plus ipilimumab versus chemotherapy. With a median follow-up of 66.8 months, nivolumab plus ipilimumab demonstrated continued overall survival (OS) benefit versus chemotherapy in all randomly assigned patients (5-year OS rates, 14% v 6%; hazard ratio [HR], 0.74 [95% CI, 0.62 to 0.88]); similar benefit was observed regardless of tumor histology. Of biomarker-evaluable patients treated with nivolumab plus ipilimumab (n = 242), high baseline monocytic myeloid-derived suppressor cell (M-MDSC) levels correlated with worse OS versus low M-MDSC levels (HR, 1.25 [95% CI, 1.09 to 1.43]). After adjusting for 24% of patients in the chemotherapy arm who received subsequent immunotherapy, nivolumab plus ipilimumab demonstrated continued OS benefit versus chemotherapy (HR, 0.64 [95% CI, 0.53 to 0.78]). No new safety signals were observed. These results demonstrate long-term, durable clinical benefit with nivolumab plus ipilimumab versus chemotherapy, which continued to be preserved even after treatment-switching adjustment in the chemotherapy arm. Exploratory analyses suggested greater benefit with nivolumab plus ipilimumab in the low M-MDSC subgroup. These results further support first-line nivolumab plus ipilimumab as standard of care for unresectable PM.

Abstract BACKGROUND: EOP is an I-SPY2 sub-study designed to test novel endocrine-based strategies in patients (pts) predicted to have modest benefit from chemotherapy. Vepdegestrant is a selective oral PROTAC estrogen receptor (ER) degrader that has demonstrated efficacy in mutant ER+ HER2-negative MBC. METHODS: EOP eligibility included pts with Stage 2/3 HR+ HER2-negative, MammaPrint (MP) Low Risk BC. Patients with MP High1 signatures were eligible if clinically node-negative or SET 2,3 High. Pts were randomized to one of 3 oral treatment (tx) arms: vepdegestrant 200 mg/d (V), vepdegestrant 200 mg/d +letrozole (VL), vepdegestrant 200 mg/d + abemaciclib 150 mg bid (VA). Pts received tx for six 28-day cycles followed by surgery. Premenopausal (PreMN) pts received ovarian function suppression (OFS) starting C1D1 in the VL arm, and C2D1 in the V and VA arms. The primary endpoint was feasibility, defined as &gt;75% of pts completing &gt;75% of study therapy. Breast MRI functional tumor volume (FTV) and ctDNA (tumor-informed assay) were assessed at baseline (T0), 3 weeks (T1), 12 weeks (T2), and 6 months (T3). Ki67 and ER/PR by IHC were centrally assessed from tumor biopsies at T0, T1, and the surgical specimen. Results: Between 4/2023 and 1/2025, 121 pts were enrolled: 40 in V, 41 in VL, and 40 in VA. Median age was 54 years, 50% preMN. 60% of pts had ductal, 35% lobular, 5% mixed histology. 55% were cN+, and 59% were not breast conservation surgery (BCS) candidates at baseline. Mean baseline Ki67 was 12.5%, 85.1% had MP Low Risk signature, 79.3% were SET 2,3 High. All 3 tx arms met the primary endpoint with 88%, 88%, and 83% completing &gt; 75% study therapy in V, VL, and VA (thus far), respectively. Most common all grade (AG) TRAEs in V and VL arms include fatigue (53% V, 71% VL), hot flashes (48% V, 71% VL), arthralgias (33% V, 49% VL), all G1 or G2 except for 1 G3 fatigue in VL arm. G2 and G3 neutropenia was seen in 1 pt each in V and VL arms. The most common AG TRAEs in VA include diarrhea (80%), fatigue (65%), nausea (58%), hot flashes (50%), neutropenia (43%). High grade TRAEs in VA include neutropenia (23% G3, 5% G4), ovarian cyst requiring surgery (5% G3), diarrhea (15% G3), nausea (3% G3), transaminitis (3% G3), thromboembolic event (8% G3), acute kidney injury (3% G3). Table 1 summarizes key response endpoints by tx arm. CONCLUSION: Six months of neoadjuvant vepdegestrant, alone or in combination with letrozole or abemaciclib, is feasible and demonstrates robust ER/PR degradation. Encouraging anti-tumor activity through reduction of Ki67 and MRI FTV, ctDNA clearance, and achievement of BCS was seen in pre- and postMN women. The neoadjuvant setting provides a rich platform to identify the dynamic range of response and non-response using a multi-modal approach and can be leveraged to inform future trial design and adjuvant therapy. Citation Format: J. Chien, R. M. Mukhtar, C. Yau, A. D. Elias, A. M. Wallace, N. Chan, C. Omene, N. Chen, J. Tseng, M. S. Trivedi, E. Stringer-Reasor, D. Yee, A. S. Clark, A. Thomas, H. Han, M. Arora, C. Nangia, K. Albain, C. Falkson, C. Isaacs, A. Zimmer, M. Rozenblit, L. van't Veer, L. Brown Swigart, G. Hirst, W. Symmans, A. D. Borowsky, N. Onishi, N. Hylton, S. Alkhafaji, K. V. Giridhar, C. Vaklavas, M. Wei, M. P. Goetz, O. Olopade, L. Huppert, C. Wu, L. J. Esserman. I- SPY2 Endocrine Optimization Pilot (EOP): Neoadjuvant vepdegestrant monotherapy or in combination with letrozole or abemaciclib in molecularly selected patients with stage 2/3 HR+ HER2-negative breast cancer (BC) [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 PD10-02.

Abstract Background Immune checkpoint inhibitors can result in serious, long-lasting immune-related adverse events (irAEs). Early identification of symptoms predictive of irAEs could enhance monitoring and timely intervention. This study assessed whether symptoms within the first 8 weeks of treatment could predict subsequent development of immune-related adrenal insufficiency(AI) or hypothyroidism. Methods This retrospective cohort study analyzed prospectively collected data from the I-SPY2 trial, a phase 2 platform trial for high-risk stage II/III breast cancer across 30 U.S. sites. The cohort included 482 women treated with experimental immunotherapy agents concurrent with weekly paclitaxel neoadjuvant chemotherapy. The primary outcomes were grade ≥ 1 hypothyroidism or AI, adjudicated by an independent safety group, up to 1-year post-treatment. Symptoms and irAEs were assessed using the Common Terminology Criteria for Adverse Events. Symptom burden was quantified as area under the curve (AUC) based on symptom grade and duration. Predictive modeling was performed using logistic regression and ROC analysis; symptom enrichment between cases and controls was evaluated using Fisher’s exact tests. Results Among 482 participants, 107 (22.2%) developed irAEs, with hypothyroidism (n = 61, 12.7%) occurring more frequently than AI (n = 38, 7.9%) at medians of 99 and 105 days from treatment initiation, respectively. Symptom enrichment analysis identified early predictive symptoms. Fatigue (17.2% vs 6.8%, p = 0.011) and rash (20.7% vs 7.8%, p = 0.0037) were predictive of hypothyroidism, while diarrhea (45.9% vs 31%, p = 0.048), constipation (5.4% vs 0.2%, p = 0.018), and taste changes (5.4% vs 0.5%, p = 0.034) were associated with AI. A predictive model demonstrated moderate performance (AUC 0.65 for AI, p < 0.0001; AUC 0.61 for hypothyroidism, p = 0.012). Model accuracy in an external validation cohort was 72.8% for AI and 74.7% for hypothyroidism. Conclusions This study presents a predictive framework to identify patients at risk for adrenal insufficiency and hypothyroidism as irAEs, enabling personalized care and proactive intervention to improve treatment outcomes and safety.

Convergent framework for future POC cancer diagnostics aligning clinical needs (why), enabling technologies (how), and translational bridges. It integrates lab-on-chip biosensing and analytics for decentralized, early, longitudinal care.

Background and Aims: Cholangiocarcinoma (CCA) is a rare but aggressive malignancy with poorly understood genetic susceptibility. To date, genome-wide association studies (GWAS) investigating germline variants associated with CCA risk remain limited. We aimed to identify genetic risk loci for CCA and its clinical subtypes through comprehensive GWAS and post-GWAS analyses. Approach and Results: We conducted a GWAS of 2366 CCA cases and 11,750 controls of European ancestry. Genome-wide significant loci ( p <5×10 −8 ) were identified and further examined through fine-mapping, functional annotation, and HLA imputation. Subgroup analyses were conducted by CCA subtypes and primary sclerosing cholangitis (PSC) status. Cross-trait linkage disequilibrium score regression and Mendelian randomization were employed to investigate the shared genetic architecture and potential causal relationships with a diverse range of traits. We identified 1 new genome-wide significant variant, rs535777 (OR=1.44), near HLA-DRB1/DQA1 associated with CCA, and 2 variants associated with extrahepatic CCA: rs116224263 (OR=0.17) in LINC02506 at 4p15.1 and rs6914950 (OR=1.63) near HLA-DRB1/DQB1 . Stratified analyses revealed rs2395184 (OR=3.51) near HLA-DRA/DRB5 associated with PSC-related CCA, and rs142674434 (OR=2.98) in THSD7A at 7p21.3 associated with non-PSC-related CCA. HLA imputation uncovered new amino acid residues associated with disease risk. Cross-trait analyses identified shared genetic signals between CCA and anthropometric, lipidemic, lifestyle, and medical traits. Mendelian randomization supported putative causal associations for 12 traits with CCA or its subtypes. Conclusions: Our large-scale GWAS highlights new genetic variants and HLA-linked mechanisms underlying CCA susceptibility. Integrating multi-step post-GWAS approaches enhances understanding of CCA pathogenesis and may facilitate the development of risk biomarkers for early detection and precision prevention strategies.

Abstract Deep learning models enable the prediction of clinical endpoints from whole-slide images (WSIs), but many such models function as “black boxes”, lacking transparency about whether and which histomorphological patterns drive their predictions, hindering interpretability and clinical adoption. Here we propose a human-in-the-loop explanation framework, MorphoXAI, which provides both local and global interpretability for deep learning models by incorporating human-expert interpretations. At the global level, it reveals the histomorphological patterns on which the model consistently relies to distinguish between classes of WSIs, as well as the patterns associated with confusion between classes. At the local level, it indicates which of these patterns are used in the prediction of an individual WSI and which regions within the slide correspond to such patterns. We validated our method on a deep learning model trained for ovarian tumor histologic subtype prediction. The results show that our framework generates explanations that accurately reflect the histomorphology underlying the model’s predictions at both global and local levels. For interpretability and clinical utility in diagnostic contexts, human evaluation results showed that our explanations were easy to interpret, rich in diagnostic features, and directly helpful for diagnostic decision-making, thereby enhancing pathologist-AI collaboration. Our work highlights that unifying global and local explanations and grounding them in expert-interpreted morphology enhances the interpretability and verifiability of deep learning models, thereby facilitating the transparent deployment of such models in clinical practice.

Background Patient complaints provide a complementary lens on surgical safety, yet prior analyses have been small and specialty specific. We aimed to characterise themes, perioperative processes, system factors and outcomes in surgical complaints using a scalable, large language model (LLM)-assisted approach. Methods We performed a national, retrospective, cross-sectional, sequential methods study of publicly available investigation reports from the Aotearoa New Zealand Health and Disability Commissioner. All reports published since 1998 were retrieved and screened for relevance across all surgical specialties. A fixed-parameter LLM workflow extracted demographics, clinical context, complications, breached patient rights, system factors and outcomes. An inductive LLM-supported thematic synthesis generated complaint themes, followed by rescoring of theme relevance across reports. Human validation supported the LLM-assisted process (Cohen’s κ >0.85). Results Of 1827 reports screened, 650 involved surgical care. Postoperative complications were frequent (84.2%), commonly accompanied by permanent disability (33.2%) or death (24.2%). Delays in recognition of deterioration (76.1%), escalation (58.3%) and definitive management (49.5%) of complications were common and associated with higher mortality. Prominent themes concerned postoperative clinical management/monitoring, communication/informed consent and professional conduct/competence. Technical/procedural errors and medication errors were comparatively less common. Breaches are most often related to reasonable care and skill, and to informed consent, with marked variation across specialties. Conclusions Complaints about surgical care predominantly reflect postoperative monitoring, escalation and communication rather than intraoperative technical errors. System priorities should include robust deterioration recognition and response, reliable handover and escalation pathways, and strengthened consent and communication processes. An LLM-assisted pipeline can scale complaint analysis to support organisational learning and quality improvement.

Key Points Inclisiran for LDL cholesterol reduction was analyzed post hoc in patients with CKD. Mean percentage LDL cholesterol reduction from baseline was around 50% regardless of eGFR in phase 3 trials. Inclisiran showed sustained and effective LDL cholesterol–lowering in patients across various levels of eGFR values, without new safety findings. Background Lowering LDL cholesterol reduces the risk of atherosclerotic cardiovascular disease in patients with CKD. The efficacy and safety of inclisiran versus placebo in patients without and with CKD were investigated in a post hoc pooled analysis of three phase 3 trials (ORION-9, ORION-10, and ORION-11). Methods Patients with heterozygous familial hypercholesterolemia, atherosclerotic cardiovascular disease or its risk equivalent, and elevated LDL cholesterol were randomized 1:1 to subcutaneous inclisiran or placebo on days 1 and 90 and every 6 months thereafter for 540 days. Patients were stratified based on baseline eGFR (by CKD Epidemiology Collaboration equation): ≥90, 60 to <90, 45 to <60, and 15 to <45 ml/min per 1.73 m 2 . Coprimary end points were percentage change in LDL cholesterol at day 510 and time-adjusted percentage change after day 90 and through day 540. Safety was also evaluated. Results Of 3660 patients, 1610 (44%) had eGFR ≥90, 1608 (44%) 60 to <90, 300 (8%) 45 to <60, and 142 (4%) 15 to <45 ml/min per 1.73 m 2 . The mean (95% confidence interval) placebo-corrected percentage changes in LDL cholesterol from baseline at day 510 in patients with eGFR ≥90, 60 to <90, 45 to <60, and 15 to <45 ml/min per 1.73 m 2 were −49.9% (−53.2 to −46.6), −51.2% (−54.4 to −48.0), −54.7% (−62.5 to −47.0), and −44.7% (−57.6 to −31.8), respectively ( P < 0.001); the corresponding mean (95% confidence interval) time-adjusted placebo-corrected percentage changes in LDL cholesterol from baseline after day 90 through day 540 were −48.4% (−50.8 to −46.1), −51.8% (−54.2 to −49.4), −55.6% (−61.0 to −50.2), and −50.4% (−59.3 to −41.5; each P < 0.001). Significant decreases in total cholesterol, apolipoprotein B, non-HDL cholesterol, and lipoprotein(a) occurred in all eGFR groups. Inclisiran was well tolerated without new safety findings. Conclusions Inclisiran demonstrated sustained and effective LDL cholesterol reduction in patients with or at risk of atherosclerotic cardiovascular disease, regardless of baseline eGFR as low as 15 ml/min per 1.73 m 2 , without new safety findings. Clinical Trial registry name and registration number: ClinicalTrials.gov, ORION-9 (NCT03397121), ORION-10 (NCT03399370), and ORION-11 (NCT03400800).

OBJECTIVE The aim of this study was to determine predictors of the minimum clinically important difference (MCID) in the Neck Disability Index (NDI) following cervical spinal deformity surgery. METHODS A retrospective review was performed of a prospective, multicenter adult cervical spinal deformity database. All patients had baseline and 1-year NDI scores. Patients met MCID with an improvement of NDI by 7 points between baseline and 1 year, as previously established. Baseline demographics, comorbidities, and both baseline and 1-year spinopelvic parameters were evaluated for statistical significance in a univariate logistic regression analysis. Significant variables, in addition to baseline NDI, were analyzed in a multivariable logistic regression model by backward selection with Akaike information criterion minimization. RESULTS A total of 122 patients were included with a median age of 62 (IQR 56, 69) years; 62% of patients were female. Of the 122 patients, 72 (59%) achieved NDI MCID at 1 year. Predictors of achieving MCID on univariate analysis included a lower Charlson Comorbidity Index (CCI) total score (OR 0.70, p = 0.03), depression as a comorbidity (OR 2.9, p = 0.02), lower C2 tilt at the 1-year follow-up (OR 0.92, p = 0.02), and a greater difference between 1-year postoperative C2–7 sagittal vertical axis (SVA) and preoperative C2–7 SVA (OR 0.98, p = 0.0495). On multivariable logistic regression analysis, predictors of achieving MCID included a lower CCI (OR 0.62, p = 0.03), depression as a comorbidity (OR 3.1, p = 0.059), a greater change in C2–7 SVA at the 1-year follow-up compared with baseline (OR 0.97, p = 0.055), and baseline NDI (OR 1.02, p = 0.24) with an area under the curve of 0.74. CONCLUSIONS The best-fit multivariable model included higher baseline NDI, a greater change in C2–7 SVA, patient-reported baseline depression, and lower CCI as important factors in predicting NDI MCID.