Browse Articles

CNN-based classifier for automated identification of magnetic states in spin dynamics simulations

Abstract The identification and classification of different magnetic states are essential for understanding the complex behavior of magnetic systems. Traditional approaches that rely on handcrafted features or manual inspection often fall short, particularly when dealing with subtle or topologically complex spin textures. In this study, we present an automated deep learning model that employs an EfficientNetV1B0 Convolutional Neural Network to classify nine distinct magnetic states, including both ferromagnetic (FM) and antiferromagnetic (AFM) spin textures such as AFM skyrmions and AFM stripe domains. The spin configurations are generated through atomistic spin dynamics simulations using the Spirit code, then visualized with VFRendering to produce RGB images, which serve as inputs to the classification model. To train and evaluate the model, we created a new dataset of manually labeled RGB images. Evaluation results show that the proposed model achieves an accuracy and F1-score of 99%, outperforming the other deep learning baselines evaluated in this study.

HIRA-SETDB1-H3K9me3 axis regulates chromatin architecture in leukemia cells

USP43 downregulation as driver of acquired resistance to dabrafenib and trametinib in melanoma via the HIF1A-PDK1 axis-mediated oxidative phosphorylation.

e21517 Background: Acquired resistance to combined BRAF and MEK inhibition (Dabrafenib+Trametinib) limits long-term survival in BRAF-mutant melanoma. Metabolic reprogramming toward oxidative phosphorylation (OXPHOS) is a key driver of this resistance, yet the post-translational mechanisms triggering this switch remain elusive. We investigated the role of the deubiquitinase USP43 in regulating metabolic plasticity and drug sensitivity. Methods: Dabrafenib+Trametinib-resistant melanoma cell lines were established. Transcriptomic profiling and paired clinical sample analysis were performed to identify key regulators. Metabolic phenotypes were assessed using Seahorse extracellular flux analysis. Mechanistic studies included Co-IP, K48-linkage specific ubiquitination assays, cycloheximide (CHX) chase assays, and nucleocytoplasmic fractionation. Therapeutic potential was validated in xenograft models with rescue arms using HIF1A inhibitors. Results: 1) We successfully generated resistant melanoma cell lines with significantly elevated IC50 values. RNA-sequencing and immunohistochemistry of paired clinical samples revealed a specific downregulation of USP43 in resistant tumors compared to pre-treatment baselines. 2) Metabolic profiling and Seahorse assays demonstrated that resistant cells underwent a metabolic switch characterized by elevated oxygen consumption rates (OCR) and mitochondrial ATP production, distinct from the glycolytic profile of sensitive cells. 3) Mechanistically, USP43 was identified to directly interact with HIF1A. USP43 specifically cleaved K48-linked poly-ubiquitin chains from HIF1A, thereby extending its protein half-life and preventing proteasomal degradation. 4) In resistant cells, USP43 deficiency led to rapid HIF1A degradation, resulting in impaired nuclear translocation and reduced transcriptional activation of PDK1. This downregulation of PDK1 facilitated pyruvate entry into mitochondria, fueling hyper-active oxidative phosphorylation. 5) Overexpression of USP43 in resistant cells restored the HIF1A-PDK1 axis, suppressed mitochondrial respiration, and re-sensitized cells to treatment both in vitro and in xenograft models. Crucially, this re-sensitization was abrogated by pharmacological inhibition of HIF1A or PDK1, confirming that USP43 functions strictly through this metabolic axis. Conclusions: Our study elucidates a novel USP43-HIF1A-PDK1 signaling axis that safeguards against metabolic drug resistance. The loss of USP43 unleashes oxidative phosphorylation by destabilizing HIF1A, allowing melanoma cells to survive targeted therapy. These findings highlight USP43 as a potential predictive biomarker and therapeutic target for overcoming resistance in BRAF-mutant melanoma.

Comparative outcomes of red blood cell transfusion versus erythropoiesis-stimulating agents in lung cancer patients with chemotherapy-induced anemia: A real-world analysis.

e20781 Background: Chemotherapy-induced anemia affects up to 80% of lung cancer patients and is associated with significant morbidity, yet real-world comparative outcomes of red blood cell transfusion versus erythropoiesis-stimulating agents (ESAs) as mutually exclusive options remain poorly characterized. The study objective was to explore primary end point of all-cause mortality as well as secondary end point of cardiopulmonary and neurological outcomes. Methods: In this multi-centre retrospective study using TriNetX, adults (≥18 years) with lung cancer and chemotherapy-induced anemia were identified via ICD codes. Participants were stratified by receipt of RBC transfusion or ESA therapy (epoetin alfa, darbepoetin alfa, or methoxy polyethylene glycol-epoetin beta). Participants receiving both RBC transfusion and ESA therapy were excluded to ensure mutually exclusive exposure groups. Propensity score matching adjusted for demographics, comorbidities, cancer stage, chemotherapy regimens, and hemoglobin level. Generalized linear models estimated odds ratios (ORs) with 95% confidence intervals (CIs). Results: Before matching, 4,171 patients were identified. After propensity score matching, 1,972 participants (986 per cohort) had similar baseline characteristics (mean age of 73, 52% female; 73% White, 12% Black, 5% Asian). No significant difference in all-cause mortality was observed between cohorts [0.993 (0.827-1.191)]. ESA therapy was associated with a lower risk of sepsis [OR: 0.760 (0.579–0.996)] and myocardial infarction [OR: 0.648 (0.423–0.991)], but higher incidence of brain metastases [OR: 1.408 (1.054-1.881)] compared with RBC transfusion. No significant difference was observed for deep vein thrombosis or pulmonary embolism. Conclusions: In this real-world analysis, ESA therapy was associated with lower risks of sepsis and myocardial infarction, but a higher incidence of brain metastases, potentially due to erythropoietin receptor-mediated tumor progression. These findings may inform individualized risk-benefit discussions when selecting anemia management strategies in lung cancer. Prospective studies are needed to validate these associations.

Exposure–response (E–R) analyses of efficacy and safety with raludotatug deruxtecan (R-DXd), a CDH6-directed antibod-drug conjugate (ADC), to inform dose selection for phase (Ph) 3 development in platinum-resistant ovarian cancer (PROC).

5570 Background: R-DXd, an ADC, comprises a humanized cadherin 6 (CDH6) IgG1 monoclonal antibody attached to a topoisomerase I inhibitor payload (DXd) via a tumor-selective cleavable linker. R-DXd monotherapy demonstrated promising antitumor activity and a manageable safety profile at doses up to 6.4 mg/kg IV Q3W in patients (pts) with heavily pretreated ovarian cancer (OC) in a Ph 1 study and in pts with PROC in an ongoing Ph 2/3 study. We report findings from E–R analyses of key efficacy and safety endpoints that supported R-DXd dose selection for Ph 3 clinical investigation in patients with PROC. Methods: Data were included from (i) the first-in-human Ph 1 study in pts with advanced OC (N=156) or renal cell carcinoma (RCC; N=23) evaluating R-DXd 1.6–9.6 mg/kg Q3W (NCT04707248), and (ii) the dose-optimization part of the Ph 2/3 REJOICE-Ovarian01 study in pts with PROC (N=107) (NCT06161025), in which pts were randomized 1:1:1 to receive R-DXd 4.8, 5.6, or 6.4 mg/kg Q3W. E–R analysis for efficacy was conducted based on pooled data from 241 pts with PROC using BICR-assessed endpoints: ORR (primary endpoint in the Ph 2 part of REJOICE-Ovarian01), best change in target lesion size, DOR, and PFS. Further, E–R analysis for safety was performed for 11 AEs of clinical interest using data from 286 pts treated with R-DXd across tumor types (OC, RCC). ORR and all safety endpoints, except interstitial lung disease (ILD), were analyzed by logistic regression; DOR, PFS, and ILD were analyzed using the Kaplan–Meier method. Numerous patient-specific covariates were evaluated. Results: E–R analyses demonstrated positive dose–response relationships for both efficacy and safety. Higher R-DXd exposure was associated with increased probability of achieving objective response, greater reduction in target lesion size from baseline, longer PFS, and longer DOR. Covariate analysis identified that baseline CDH6 expression was positively associated with ORR and tumor response. Higher R-DXd or DXd payload exposure was associated with an increase in drug-related TEAEs, including Grade ≥3 (G3+) AEs, any-grade ILD, G2+ gastrointestinal AEs (nausea/vomiting), G3+ cytopenias (anemia, neutropenia, thrombocytopenia), and AEs leading to dose reduction. While higher doses of R-DXd are predicted to maximize efficacy based on the E–R analysis, the 5.6 mg/kg dose provides an overall optimal balance of safety, tolerability, and efficacy compared to 6.4 mg/kg and 4.8 mg/kg. Conclusions: Integrated E–R analyses, combined with the totality of data from 286 pts, support selection of 5.6 mg/kg Q3W as the optimal R-DXd monotherapy dose for the Ph 2 extension and Ph 3 parts of REJOICE-Ovarian01, in accordance with the principle of benefit/risk for oncology dose optimization and guidelines from the FDA’s Project Optimus.

Machine learning risk stratification in a US-based database to identify subgroups of patients with PD-L1-high NSCLC who benefit from adding chemotherapy to pembrolizumab.

e20523 Background: First-line treatment options for advanced non-small cell lung cancer (aNSCLC) with PD-L1 TPS ≥50% include pembrolizumab or pembrolizumab plus platinum-doublet chemotherapy. Without head-to-head randomized data, optimal patient selection remains unclear. We hypothesized that machine learning–predicted baseline prognosis modifies chemotherapy benefit, with higher-risk patients more likely to benefit from chemotherapy’s rapid clinical effects. Methods: Using the Flatiron Health Research Database, we identified patients with aNSCLC and PD-L1 CPS TPS ≥50% treated with first-line pembrolizumab or pembrolizumab plus chemotherapy. A gradient-boosted survival model predicted 6-month survival from baseline clinical variables (demographics, cancer features, ECOG, laboratory values, and comorbidities) using cross-validation, then calibrated via isotonic regression. Heterogeneity of absolute treatment benefit was evaluated using overlap-weighted regression with 2-year restricted mean survival time (RMST) pseudo-observations. We summarized the continuous treatment-effect function using a crossover point, defined as the baseline 6-month survival probability at which the estimated RMST benefit of adding chemotherapy reached a clinically meaningful magnitude (≥30 days). Patients were stratified by this survival probability, and survival was compared between treatments using inverse probability of treatment weighting (IPTW). Results: Among 1,434 eligible patients, 930 received pembrolizumab and 504 received pembrolizumab plus chemotherapy. Median age was 71 years, 53% were male, and median follow-up was 38 months. The model achieved 6-month AUC 0.76 with good calibration (Brier score 0.17). Chemotherapy benefit increased as baseline predicted survival worsened: for every 10 percentage-point decrease in predicted 6-month survival, patients gained 17 days in 2-year RMST with combination therapy (p = 0.051). The crossover point corresponded to a baseline 6-month survival probability of 70%. Patients below the crossover survival probability (30.8%)—characterized by worse ECOG, weight loss, bone and liver metastases, anemia and hypoalbuminemia—derived significant benefit from adding chemotherapy in the IPTW-adjusted survival analysis, while those above (69.2%) showed no benefit (Table 1). Conclusions: A machine learning model trained on nationally-representative data identified subgroups of patients with PD-L1-high aNSCLC who benefit from adding chemotherapy to pembrolizumab. RMST differences stratified by crossover survival probability. 6-month Survival <70% 6-month Survival ≥70% 1-year RMST Δ 57.9 (22.7-88.7) 2.0 (-15.7-20.4) 2-year RMST Δ 93.7 (22.9-155.1) 3.6 (-36.2-44.3) RMST differences (Pembro+Chemo - Pembro) in days; 95% CI in parentheses.

Differential clinical and economic signatures of CAR T-cell therapy in DLBCL across demographic and payer strata: A national analysis.

7064 Background: Chimeric antigen receptor (CAR) T cell therapy has revolutionized care for relapsed/refractory diffuse large B cell lymphoma (DLBCL), yet granular delineation of real-world immune effector toxicities, organ-specific sequelae and economic burden across demographic and socioeconomic strata requires further exploration. Methods: Using the National Inpatient Sample (2017–2022), we identified adult DLBCL hospitalizations receiving CAR T through ICD-10-PCS codes. Cytokine release syndrome (CRS) [all grades - 2021–2022], immune effector cell–associated neurotoxicity syndrome (ICANS) [all grades – 2022], infections, cardiac events and acute kidney injury (AKI) were identified through ICD-10 codes. Age was stratified into 21-40, 41-60, 61-80 and >81 years. Payer groups were Medicare, Medicaid, Private and Uninsured. Races were White, African American, Hispanic and others [Pacific Islanders, Native Indians, Asians]. Chi-square analysis was performed, and multivariate regression adjusted for demographics, comorbidities, and hospital characteristics. Results: Across >11,000 weighted hospitalizations, inpatient mortality, all-grade CRS and ICANS were comparable across age, race, and payer groups (p>0.05). Inpatient mortality was numerically higher in other minor races (6.6%) vs White (3.3%), African American (3.1%), and Hispanic (0.9%; p=0.230). Any grade CRS was numerically higher in other minor races (72.5%) than in White (62.3%) and African American (60.6%; p=0.614). Any grade ICANS dispersion included 29.5% (White), 39.1% (African American), 22.5% (Hispanic) and 11.1% (Other; p=0.204). A race-specific signal for ICANS grade 5 (p=0.030) was seen, but numbers were small. Among cardiac events, arrhythmias rose from 7.6% (age 21–40 years) to 57.1% ( age ≥81 years) (p<0.001). By payer, arrhythmias were highest in Medicare (31.2%; p<0.001). Other cardiac events including myocardial infarction, stroke and cardiogenic shock were similar across groups. Pneumonia rates varied by payer, peaking in the Uninsured (23.8%; p=0.010). Sepsis and septic shock rates showed no differences. AKI (p=0.007) rates showed differences but attenuated after adjustment. Hispanic and other minor races had lower adjusted costs than White (Odds ratio [OR] 0.9, p=0.046; OR 0.8, p=0.006), while ages 61–80 had higher adjusted costs vs 21–40 (OR 1.2, p=0.023). Conclusions: In this national cohort, CRS, ICANS and mortality were similar across demographic and payer strata, indicating consistent immune effector safety. Cardiac arrhythmia rates were higher among the elderly and Medicare beneficiaries. There were lower expenditures among minor race groups and higher costs in elderly. These findings highlight discrete domains of variability and suggest intensified cardiac surveillance in elderly and equitable resource allocation to optimize outcomes with CAR-T.

Mortality trends for co-occurring colon carcinoma and metabolic syndrome in the United States: A CDC WONDER analysis, 1999-2020.

e15614 Background: Colon carcinoma remains a leading cause of cancer morbidity and mortality in the United States, with persistent disparities despite advances in screening and treatment. Metabolic syndrome, affecting nearly one-third of U.S. adults, is increasingly linked to colorectal cancer risk and poorer outcomes. However, population-level data on their combined burden remain limited. Methods: Mortality data from the CDC WONDER database were analyzed for adults aged ≥45 years from 1999-2020. Deaths listing malignant neoplasms of the colon, rectosigmoid junction, and rectum (C18.0-C18.9, C19, and C20), type 2 diabetes mellitus (E11.0–E11.9), obesity (E66.0–E66.9), lipoprotein metabolism disorders and other lipidemias (E78.0–E78.9), and hypertensive diseases (ICD-10: I10–I15) as underlying or contributing causes were included. Crude and age-adjusted mortality rates (AAMRs) per 100,000 persons were standardized to the 2000 U.S. population, and temporal trends were assessed using Joinpoint regression to estimate annual (APC) and average annual percent changes (AAPC), stratified by age, race/ethnicity, state, and urban-rural status. Results: From 1999-2020, 138,431 deaths occurred among adults with colon cancer and metabolic syndrome, with overall AAMRs rising from 3.7 to 6.1 per 100,000. Age-stratified analysis showed the highest mortality in adults ≥85 years (APC 21.53, p = 0.155169) and the lowest in 45-54 years (from 0.2 to 0.8). Sex-wise, males had higher AAMRs than females (from 4.2 to 7.7 vs. 3.3 to 4.8); females exhibited sharp increases 1999-2001 (APC 16.82%, p = < 0.000), while males rose 1999-2005 (APC 5.24%, p = 0.008) and 2018-2020 (APC 12.82%, p = 0.003). Race-wise, Non-Hispanic (NH) Black or African Americans had the highest AAMRs (from 6.7-10.8), followed by NH Whites (from 3.4 to 5.9) and Hispanics (rapid increase, APC 20.48%, p = 0.045); NH American Indians showed no significant trend. Non-Hispanics exceeded Hispanics (from 3.7 to 6.1 vs. 2.1 to 5.6). Rural noncore areas had the highest mortality (from 4.0 to 8.4), and large fringe metros the lowest (from 3.3 to 5.0), with varying trends across micropolitan and other metropolitan areas. Overall mortality rose rapidly, with the steepest increase in 1999-2001 (APC 16.24%, p < 0.001) and another steady increase 2016-2020 (APC 5.07%, p < 0.001) and stable rates during 2001-2016. States with age-adjusted mortality rates above the 90th percentile included Mississippi (9.3), Nebraska (9.3), Ohio (8.4), Oklahoma (8.3), and West Virginia (8.1). Conclusions: Colorectal cancer mortality among U.S. adults with cardiometabolic comorbidities rose sharply from 1999–2020, with higher risk in males, older adults, NH Black individuals, and rural populations. Findings highlight a growing synergy between metabolic dysfunction and cancer, underscoring the need for integrated prevention and care strategies.

Differential clinical and economic signatures of CAR T-cell therapy in multiple myeloma across demographic and payer strata: A national analysis.

e19537 Background: Chimeric antigen receptor (CAR) T cell therapy is a paradigm shifting modality for relapsed/refractory multiple myeloma (MM). Yet national scale delineation of immune effector toxicities, cardiopulmonary and renal sequelae, and economic burden across demographic and payer strata has not been adequately characterized. Methods: The National Inpatient Sample (2018–2022) was queried to identify adult MM hospitalizations receiving CAR T through ICD-10-PCS codes. Cytokine release syndrome (CRS) [all grades - 2021–2022], immune effector cell–associated neurotoxicity syndrome (ICANS) [all grades – 2022], cardiac events and infections were identified through ICD-10 codes. Age was stratified into 21-40, 41-60, 61-80, >81 years. Payer groups were Medicare, Medicaid, Private and Uninsured. Races were White, African American, Hispanic and others [Pacific Islanders, Native Indians, Asians]. Chi-square analysis was performed and multivariate regression adjusted for demographics, comorbidities and hospital characteristics. Results: Across >7,000 weighted hospitalizations, mortality, any grade CRS and ICANS were statistically similar across age, race, and payer (p>0.05). However, CRS was numerically higher in Hispanic patients (78.6%) than in White (63.2%), African American (68.1%), and other minor races (60.0%) (p=0.339). ICANS grade 4 was numerically higher in Hispanic (5.3%) versus White (0.8%) (p=0.278). Mortality was numerically highest in other minor races (9.5%) versus White (3.6%), Hispanic (2.8%), and African American (0.0%) (p=0.179). Age-stratified analyses showed significant heterogeneity in arrhythmias (22.2% [21–40 years] to 28.6% [≥81 years]; p=0.021), attenuating after adjustment. By race, arrhythmias ranged from 8.3% (Hispanic) to 25.0% (White) (p=0.113). Other cardiac events including myocardial infarction, stroke and cardiogenic shock were similar across groups. Payer stratified pneumonia varied (p=0.010) with uninsured 23.8% versus Private 5.1%, Medicare 7.3%, and Medicaid 6.4%, without persistence in adjusted models. Sepsis and septic shock rates showed no differences. Hispanic patients had higher adjusted costs versus White (Odds ratio [OR] 1.3, 1.04–1.7; p=0.023). Age-stratified costs differed (p=0.032), with elevated cost in ≥81 years (USD 394,617) versus 21–40 (USD 229,862), 41–60 (USD 214,974), and 61–80 (USD 269,844) but it was not significant in adjusted analysis. Conclusions: In this national cohort, CRS, ICANS and mortality were broadly stable across strata, while cardiac events displayed select divergences with elevated rates of arrhythmias among the elderly population suggesting robust cardiac surveillance in these patients. Cost differentials were evident with higher costs among Hispanic and elderly patients warranting validation in larger, prospectively harmonized datasets.

Dysphagia-associated inpatient risk phenotypes and mortality in hospitalized head and neck cancer: A national analysis.

11085 Background: Dysphagia is common in patients with head and neck cancer (HNC) but is often viewed as a quality-of-life issue rather than a driver of acute inpatient outcomes. Its contribution to airway failure, mortality, and resource utilization is poorly defined, particularly when dysphagia is uncoded. We evaluated whether dysphagia-associated clinical signals define inpatient risk phenotypes, including a high-risk subgroup with silent dysphagia. Methods: We conducted a survey-weighted analysis of the National Inpatient Sample (NIS), 2016 to 2023, including adults hospitalized with HNC identified using ICD-10-CM codes C00 to C14 and C30 to C32. Tumors were categorized by first-hit anatomic site, and palliative admissions (Z51.5) were excluded. Dysphagia-associated signals included dysphagia (R13*), aspiration pneumonia (J69*), and feeding tube placement (ICD-10-PCS 0DH6*), which were combined into mutually exclusive dysphagia phenotypes. Silent dysphagia was defined as aspiration pneumonia or feeding tube placement without dysphagia coding. The primary outcome was in-hospital mortality. Secondary outcomes included mechanical ventilation, tracheostomy, shock, length of stay (LOS), and hospitalization cost. Survey-weighted regression models adjusted for demographics, payer, socioeconomic status, admission characteristics, hospital factors, tumor site, comorbidity burden, and year. Results: Among 112,495 unweighted HNC hospitalizations (weighted national estimate approximately 562,000), 26.6% had coded dysphagia, 8.7% aspiration pneumonia, and 16.3% underwent feeding tube placement; 11.4% met criteria for silent dysphagia. Mortality increased stepwise across phenotypes, from 1.82% (95% CI 1.72 to 1.93) in patients without dysphagia-associated signals to 8.59% (95% CI 7.80 to 9.47) with aspiration alone. Aspiration plus feeding tube placement was associated with the longest LOS (20.2 days, 95% CI 18.9 to 21.6) and highest mean cost ($77,615, 95% CI $70,531 to $84,700). In adjusted analyses, aspiration alone was associated with higher mortality (OR 3.77, 95% CI 3.26 to 4.36), as was aspiration with feeding tube placement (OR 2.40, 95% CI 1.70 to 3.40). Dysphagia coding alone was not associated with increased mortality (OR 0.71, 95% CI 0.60 to 0.83). Silent dysphagia was independently associated with mortality (OR 2.31, 95% CI 2.04 to 2.61), mechanical ventilation (OR 3.77, 95% CI 3.51 to 4.03), and tracheostomy (OR 4.69, 95% CI 4.40 to 4.99). Conclusions: Dysphagia-associated clinical signals define inpatient risk phenotypes in hospitalized patients with HNC. While dysphagia coding alone does not confer excess mortality risk, aspiration pneumonia and feeding tube placement when dysphagia is uncoded identify a high-risk population with increased mortality, airway failure, LOS, and cost, supporting earlier inpatient recognition and intervention.

Germline pathogenic variants in early-onset colorectal cancer and limitations of tumor-based screening.

e15719 Background: In the United States, the incidence of early-onset colorectal cancer (EOCRC) is increasing, yet real-world utilization of genetic counseling and germline testing remains variable. Historically, tumor mismatch repair (MMR) and microsatellite instability (MSI) testing have guided evaluation for Lynch syndrome; however, this approach may fail to identify the broader spectrum of hereditary cancer syndromes now recognized in EOCRC. Methods: We performed a retrospective analysis of patients diagnosed with colorectal cancer before age 50 who presented to Moffitt Cancer Center between 2016-2025 (n = 140). Rates of genetic counseling referral, tumor MSI/MMR testing, and germline testing results were abstracted from the electronic medical record. Pathogenic or likely pathogenic germline variants and variants of uncertain significance (VUS) were categorized by gene and associated hereditary cancer syndrome. Descriptive statistics were used to assess mutation prevalence and identify gaps in genetic counseling referral and completion of germline testing. Results: Of the 140 patients with EOCRC, 77 (55%) were referred for genetic counseling, of whom 66 (85.7%) completed germline genetic testing. Referral for genetic counseling in this cohort rose significantly over time, from 56.4% in 2019-2022 to 80.6% in 2023-2025 (p = 0.013). Pathogenic germline variants were identified in 12 of 66 tested patients (18%), all of whom had MSI-stable/pMMR tumors. Detected variants included APC (6.1%), MUTYH (6.1%), BRCA2 (1.5%), MSH2 (1.5%), MLH1 (1.5%), and NF1 (1.5%). Alterations in APC or MUTYH accounted for 66.7% of the detected pathogenic variants, compared to 16.7% involving mismatch repair genes, representing a 4:1 predominance of non–Lynch-associated hereditary colorectal cancer syndromes. Tumor MSI/MMR status was available for all tested patients; only 2 of 66 (3%) had MSI-high/dMMR tumors, neither of which harbored a pathogenic germline variant. An additional 10 patients harbored at least one VUS. Extrapolating from the observed mutation prevalence suggests that EOCRC patients who did not undergo genetic counseling or testing likely harbor undetected pathogenic germline variants. Conclusions: In this EOCRC cohort treated at a large tertiary referral center, nearly one in five tested patients carried a pathogenic germline variant, most involving non–Lynch-associated hereditary colorectal cancer syndromes. These findings support universal germline genetic testing for EOCRC regardless of tumor MSI/MMR status to reduce missed hereditary cancer diagnoses and improve cascade testing for at-risk relatives.

Knowledge and practice toward hereditary lung cancer in Brazil: A GBOT (Grupo Brasileiro de Oncologia Torácica) national survey.

e20780 Background: Non–small cell lung cancer (NSCLC) in never smokers represents 10–25% of cases and shows distinct epidemiologic and molecular features. Although hereditary predisposition to lung cancer has been increasingly recognized, there is still no consensus regarding which patients may benefit from germline genetic testing. This study assessed the knowledge, perceptions, and clinical practices of Brazilian healthcare professionals dedicated to lung cancer regarding inherited lung cancer. Methods: We conducted a national, anonymous, electronic survey with 20 questions distributed to professionals involved in lung cancer care in Brazil. The questionnaire assessed demographics, clinical practice, access to genetic testing and counseling, referral criteria, and knowledge related to hereditary lung cancer. Results: A total of 86 participants completed the survey. Most respondents were aged 30–50 years (76.7%), and 42.9% were from Southeastern Brazil. Medical oncologists comprised 54.6% of respondents, followed by thoracic surgeons (20.8%). Nearly 90% worked in the private or mixed settings, and 94.7% currently treated lung cancer patients.Most respondents (96%) believed lung cancer may have a hereditary component, although estimates of its frequency varied widely. Family history was routinely collected by 97%, and 85.7% had previously treated patients with a family history of lung cancer. Approximately 60% reported access to germline genetic testing, 50% had managed patients with known pathogenic germline variants, and 70% had access to genetic counseling, with 65.8% having referred at least one lung cancer patient for evaluation. Common referral criteria included diagnosis before age 50 (67%), never-smoking status (62.6%), Ashkenazi ancestry (67%), family history of lung cancer (68%), and multiple primary tumors (64.8%). The most cited genes were TP53 (62.6%), EGFR (53.8%), and BRCA1/2 (44%). Li–Fraumeni syndrome was the most frequently recognized hereditary syndrome associated with lung cancer (62.6%). Almost all participants (98.7%) agreed that further studies are needed to better define the role of inherited predisposition in lung cancer. Conclusions: Brazilian healthcare professionals widely recognize the possibility of hereditary predisposition to lung cancer; however, knowledge regarding specific genes, syndromes, and referral criteria remains heterogeneous. These results highlight essential gaps in education and reinforce the need for broader awareness initiatives and prospective studies to better define indications for germline testing in lung cancer, particularly among never smokers.

Trends in cancer drug revenue retained by providers under buy-and-bill reimbursement, 2010-2024.

11032 Background: Cancer care providers, including physician offices and hospitals, retain “markup” fees on clinician-administered drugs (such as chemotherapies and immunotherapies) which are proportional to drug price. Under this system, commonly known as “buy and bill,” multiple external factors may impact provider revenue: increasing cancer drug prices, increasing care delivery in hospitals (where markups are higher) rather than offices, the 340B Drug Pricing Program, and payer “bagging” strategies to avoid paying markups to providers. Study goals were to estimate changes in provider revenue retained on clinician-administered cancer drugs from 2010-2024 and to quantify the contribution of four factors: 1) cancer drug sales, 2) site of care (office vs. hospital), 3) 340B participation, and 4) payer “bagging” strategies. Methods: We included all clinician-administered cancer drugs with available US sales data, including biosimilars and cancer-specific supportive care drugs (e.g., growth factors). We obtained cancer drug sales from manufacturer SEC filings (aggregated in SSR Health data), inflated to 2024 USD. We estimated the proportion of cancer drug sales administered within each of nine payer-care settings (three major payers: Medicare, Medicaid, commercial, and three major sites of care: physician office, 340B-participating hospital outpatient, and non-340B hospital outpatient), using Surveillance, Epidemiology, and End Results (SEER), the National Health Interview Survey (NHIS), and CMS and commercial claims. To estimate markup revenue, we applied setting-specific markup rates (from statute for Medicare/Medicaid, from peer-reviewed literature for commercial) to the drug sales in each setting. Plausible estimates of “bagging” prevalence were obtained from the literature. Results: Cancer drug sales increased from $27.3 billion (B) in 2010 to $64.9B in 2024, a 138% increase. Provider markup revenue on cancer drugs increased from $8.7B in 2010 to $35.7B in 2024, a 309% increase. Of the 2010-2024 increase, increasing drug sales accounted for 73.3%, shifting site-of-care to hospitals for 24.2%, and 340B participation growth for 2.5%. During this period, drug revenue retained by physician offices increased by 30%, non-340B hospitals by 49%, and 340B-participating hospitals by 739%. Accounting for bagging reduced 2024 markup revenue from $35.7B to $28.4B. Conclusions: Provider revenue from cancer drug markups grew substantially during 2010-2024. The largest single factor was growing cancer drug sales, although shifts in site of care also contributed. Revenue retained by 340B-participating hospitals increased much more than offices or other hospitals. Growth in provider markups was reduced modestly by growth in payer bagging strategies. Policymakers should consider whether growing markups on clinician-administered drugs reflects an efficient use of healthcare spending.

Body mass index (BMI) and overall survival (OS) among Hispanic patients treated with immune checkpoint inhibitors (ICIs).

11171 Background: Prior studies have suggested improved OS among cancer patients receiving ICIs with higher BMI, despite obesity being a known cancer risk factor; however, these findings are largely derived from predominantly non-Hispanic cohorts. In Austin, Texas, Hispanic individuals comprise approximately 30% of the population, limiting the external validity of prior studies. We evaluated whether the previously described association between BMI and OS extends to Hispanic patients treated with ICIs. Methods: We retrospectively identified Hispanic patients with solid tumors treated with ICIs at Ascension/UT Health Austin between November 2013 and December 2025 using Research Electronic Data Capture (REDCap) data. BMI at ICI initiation was categorized as normal (18.5–24.9 kg/m²), overweight (25–29.9 kg/m²), or obese (≥30 kg/m²); underweight patients were excluded. Overall survival (OS) was defined as time from ICI initiation to death, with censoring at last follow-up or 5 years. Missing 5-year survival status was addressed using single stochastic regression imputation. Stabilized inverse probability of treatment weighting from a multinomial propensity score model was used to adjust for baseline differences, including age, sex, ECOG status, cancer type and stage, ICI agent, steroid exposure, and antibiotic exposure. Weighted Cox proportional hazards models estimated associations between BMI and OS, with prespecified subgroup analyses by sex and stage. Results: Among 177 Hispanic patients included, average age was 54 years and 53% were male; 32% had normal BMI, 31% were overweight, and 37% were obese. Median overall survival (OS) was 497 days. After inverse probability weighting, BMI was not associated with OS (p = 0.90). Compared with normal BMI, overweight patients had a hazard ratio (HR) of 0.94 (95% CI 0.49–1.80) and obese patients an HR of 1.09 (95% CI 0.60–2.00). No significant differences were observed between obese and overweight patients (HR 1.16; 95% CI 0.64–2.09). Subgroup analyses by sex and cancer stage showed no survival advantage associated with higher BMI. Among women, obese patients had an HR of 1.58 (95% CI 0.49–5.13), while among men the HR was 1.30 (95% CI 0.66–2.58). When stratified by stage, obese patients had an HR of 1.68 (95% CI 0.44–6.40) in stages I–III and 1.13 (95% CI 0.53–2.40) in stage IV. Conclusions: In this cohort of Hispanic patients with solid tumors treated with immune checkpoint inhibitors, we did not find a survival advantage for those with higher BMI. These findings raise the possibility that BMI associated survival advantages reported in predominantly non-Hispanic cohorts may not extend to the Hispanic population. Larger studies are needed to determine whether these findings represent true population-specific differences.

Patient-reported quality of life and breast specific symptoms among long term survivors of breast cancer in India: A cross-sectional analysis using EORTC QLQ C30 and QLQ BR42.

e12743 Background: Understanding how cancer therapies affect day to day living requires listening to patients, not just measuring tumour response. The EORTC QLQ-C30 provides a global view of quality of life and functioning, while the breast specific BR42 module captures body image, sexual well-being and treatment-related symptoms. There are few data from Indian patients receiving modern treatments. Methods: We analysed questionnaires from 110 long term breast cancer survivors treated at a tertiary centre in New Delhi,India. Each completed the QLQ-C30 and QLQ-BR42. Raw scale scores were calculated as the mean of constituent items, then linearly transformed to a 0–100 scale; functional scales were inverted so that higher scores denote better functioning, and symptom scales were not . Scores were only computed when at least half the items in a domain were answered. Results: Participants reported reasonably good overall quality of life: mean (±SD) global health score was 73.6 ± 25.8. Physical, role and emotional functioning were also relatively high (78.4 ± 20.4, 89.1 ± 17.4 and 75.2 ± 26.2, respectively), and body-image and sexual-functioning scores were well above two thirds of the maximum (85.1 ± 17.7 and 86.6 ± 18.0). Fatigue was the only core C30 symptom with a low mean (27.4 ± 22.8). In contrast, several BR42 symptom domains showed substantial burden: endocrine-therapy side effects averaged 77.6 ± 18.7, arm-related symptoms 75.9 ± 20.0 and endocrine-sexual side effects 90.3 ± 16.2 (scores > 66.7 indicate clinically significant problems ). Endocrine-sexual side effects showed a modest negative correlation with global health (r = –0.34), whereas body-image scores were positively associated with overall quality of life (r = 0.39). Associations between endocrine-therapy side effects and physical functioning were small (r = –0.18). Conclusions: In this cohort, global quality of life and functional domains were largely preserved despite cancer treatment, and women generally felt satisfied with their bodies and sexual relationships. However, endocrine therapy toxicity and arm symptoms were strikingly high and were linked to lower overall well being. Routine assessment of these patient reported outcomes can help clinicians anticipate problems and tailor supportive care. Further studies that include treatment details and larger samples are needed to confirm these findings and identify modifiable predictors.

Prognostic value of post-operative circulating tumor DNA (ctDNA) in real-world treatment and outcomes among patients (pts) with muscle invasive (MIBC) and locally advanced (LA) bladder cancer (BC).

4600 Background: BC has high recurrence risk despite curative intent surgery and perioperative therapy. As ctDNA guided strategies are investigated in trials, evidence from routine practice is needed to understand treatment decisions and how post-surgery ctDNA status relates to outcomes. This study evaluated real-world survival outcomes and treatment patterns by post-surgery ctDNA status among pts with MIBC/LA BC and assessed prognostic value beyond pathologic response (pCR). Methods: This retrospective study used the US-based, EHR-derived deidentified Flatiron Health Research Database (cutoff 08/31/25). Pts with MIBC or LA BC diagnosed on or after January 2015 who underwent surgery and ctDNA testing were included. Pt characteristics were summarized, including ctDNA testing patterns. Adjuvant therapy initiation was described by post-surgery ctDNA status. Real-world overall survival (rwOS) and disease-free survival (rwDFS) were estimated from the date of surgery using Kaplan–Meier methods, stratified by pCR status and post-surgery ctDNA status (results within 90 days post-surgery), and summarized descriptively with landmark survival probabilities. Results: Among 388 pts in the study population who underwent surgery and received ctDNA testing, median age at diagnosis was 70 years (IQR: 63.8-75.0); 63% received neoadjuvant therapy, 49% received adjuvant therapy, and 11% had residual disease post-surgery. Use of ctDNA testing increased substantially over time in the study cohort, from 2.6% (10/388) in 2021 to 57% (223/388) in 2025. Of ctDNA-tested pts, 97 pts had evaluable post-surgery ctDNA results (ctDNA+: 10.3% [n=40]; ctDNA–: 14.7% [n=57]). Of these, adjuvant therapy was given to 43% of post-surgery ctDNA+ pts and 35% of ctDNA− pts. The 12-month rwOS (95% CI) was 69.4% (53.8%-89.6%) for ctDNA+ vs 96.5% (91.8%-100.0%) for ctDNA–; 12-month rwDFS (95% CI) was 34.4% (20.8%-56.8%) vs 72.7% (61.2%-86.4%), respectively. Stratified by pCR, outcomes were favorable among pts achieving pCR. Although limited by small sample sizes, ctDNA status appeared to further differentiate prognosis within each stratum, particularly among pts without pCR, where ctDNA+ pts had worse rwOS (43.2% vs 100.0%, P < .01) and worse rwDFS (9.9% vs 64.2%, P = .014) over time compared with ctDNA– pts. Conclusions: Rapid growth in ctDNA testing highlights increasing clinical adoption. Post-surgery ctDNA positivity was linked to worse rwOS/rwDFS in pts with MIBC/LA BC and further stratified prognosis beyond pCR. Importantly, ctDNA may complement pathologic response in identifying pts at highest risk for recurrence and informing post-surgical management. Future work is needed to translate these findings into clinically actionable insights and treatment strategies.

Benefit of chemotherapy on survival in women aged ≥ 70 with HR-positive, HER2-negative breast cancer and N2-like nodal burden: A population-based analysis.

e12703 Background: High-risk localized hormone receptor (HR)-positive, HER2-negative breast cancer is often treated with chemotherapy regardless of age. We evaluated whether chemotherapy improves survival in women ≥70 with high nodal burden. Methods: This is a retrospective cohort study using SEER plus database (2000–2022) from 17 registries. Women diagnosed with non-metastatic HR–positive, HER2-negative breast cancer and 4–9 positive lymph nodes between 2010 and 2021 were included. Patients were stratified by age and categorized based on receipt of chemotherapy. The primary outcome was breast cancer–specific survival (BCSS). Covariates included age, race, tumor grade, marital status, tumor T stage, and radiation therapy. Cox models assessed the association between chemotherapy and BCSS, including age interactions. Adjusted HRs with 95% CIs are reported and p ≤ 0.05 was significant. Results: Out of 4208 female patients, 57.6% received chemotherapy. Median BCSS was longer with chemotherapy (54 vs. 45 months, HR 0.64, 95% CI 0.45–0.58, p<0.001), though the association was not significant on multivariate analysis (HR 0.90, 95% CI 0.74–1.09, p=0.27). There was significant effect modification by age (p for interaction <0.05). Among patients aged 70–74 years, chemotherapy was associated with improved survival (HR 0.65, 95% CI 0.48–0.88; p = 0.006). In contrast, no survival benefit was observed among patients aged 75–79 years (HR 1.02, 95% CI 0.77–1.35; p = 0.90) or ≥80 years (HR 1.12, 95% CI 0.82–1.52; p = 0.47). Relative to patients aged 70–74 years, the association between chemotherapy and survival was significantly attenuated in patients aged 75–79 years (interaction HR 1.56, p = 0.036) and ≥80 years (interaction HR 1.72, p = 0.018). In contrast, no evidence of effect modification was observed for hormone receptor subtype, tumor stage, receipt of radiation therapy, or histologic grade, indicating that the association between chemotherapy and survival did not differ across these tumor-related characteristics. Conclusions: In this analysis, chemotherapy improved BCSS only in women aged 70–74 with hormone receptor–positive, HER2-negative breast cancer with 4-9 lymph nodes involved. Lack of SEER data on endocrine therapy, frailty, comorbidities, treatment details, and toxicity, plus possible underreporting of chemotherapy, limits interpretation. These findings support age-informed treatment decisions and warrant validation in prospective and retrospective studies. Association between chemotherapy and breast cancer–specific survival by age group. Age group (years) Chemotherapy HR (95% CI) P value Interaction HR (95% CI) Interaction P value 70 - 74 (reference) 0.65 (0.48–0.88) 0.006 1.00 (reference) - 75 - 79 1.02 (0.77–1.35) 0.90 1.56 (1.03–2.37) 0.036 > 80 1.12 (0.82–1.52) 0.47 1.72 (1.10–2.71) 0.018

Neoadjuvant nivolumab plus ipilimumab and adjuvant nivolumab in patients with localized microsatellite instability-high (MSI)/mismatch repair deficient (dMMR) oeso-gastric adenocarcinoma: Long-term follow-up of the GERCOR NEONIPIGA phase II study.

4099 Background: Neoadjuvant nivolumab and ipilimumab in localized MSI/dMMR gastric cancer were evaluated in the NEONIPIGA study (NCT04006262). The primary endpoint was pathological complete response (pCR) and has been previously reported. Here, we report on the long-term follow-up analysis and the secondary objectives of event-free survival (EFS) and overall survival (OS). Methods: This phase II, single-arm study evaluated neoadjuvant nivolumab 240 mg q2w x 6 and ipilimumab 1 mg/kg q6w x 2, followed by surgery 5 weeks (±1 week) after the last injection of nivolumab and adjuvant nivolumab 480 mg q4w x 9 in patients with resectable MSI/dMMR, T2-T4 NxM0 oeso-gastric adenocarcinoma. EFS and OS were evaluated through Kaplan-Meier curves. Results: 32 patients were included: 16 (50%) with gastric location and 16 (50%) with gastro-esophageal junction; 28 (88%) were initially classified as usT3 and four (12%) as usT2; 23 (72%) were lymph node positive. Three patients did not undergo surgery (one metastatic progression, two refusals). 27 (84%) patients completed the planned 6 cycles of neoadjuvant therapy, and 14 (44%) received the full neoadjuvant and adjuvant treatment. Median follow-up was 48.3 months (44.8-52.0). In the ITT population, 4-year OS was 84.1% (65.8-93). In the population eligible for surgery, 4-year EFS was 83.5% (64.8-92.8). Only one patient (classified as ypT0N1 and TRG1b at surgery) relapsed with cerebral metastases 29 months after initiation of neoadjuvant treatment. Six deaths were reported, including two related to gastric cancer (one post-operative due to surgery complications, one from cerebral metastases). The four remaining deaths were unrelated to gastric cancer or treatment toxicity and were due to pulmonary infections (n = 2; 33 and 60 months after initiation of neoadjuvant treatment), perforation of a strangulated hernia (13 months), and metastatic tongue cancer (30 months). All three patients without surgery achieved a clinical complete response after immunotherapy (including one deceased patient reported above). Combined with the 17 (59%) patients in pCR previously reported, 20 of 32 (62.5%) patients achieved complete response. No new safety signals were reported during extended follow-up. Conclusions: Neoadjuvant nivolumab and ipilimumab before surgery and adjuvant nivolumab in localized MSI/dMMR oeso-gastric adenocarcinoma achieved a high curative rate. These findings strongly support further investigation of a watch-and-wait approach in case of clinical complete response after immune checkpoint inhibitors, which is currently being evaluated in the DEWI GERCOR phase II study (NCT06059495). Clinical trial information: NCT04006262 .

Time to treatment initiation for triple-negative breast cancer in Brazil: A call for public health action.

1544 Background: Timely initiation of treatment is a critical determinant of outcomes in breast cancer. In Brazil, national legislation mandates treatment beginning within 60 days of histopathological diagnosis. Adherence to this interval may be particularly relevant in triple-negative breast cancer (TNBC), an aggressive subtype associated with rapid disease progression and poor prognosis. Methods: Data were collected from patients (pts) diagnosed with primary TNBC, stages I-III, treated at two large public cancer centers in Brazil between 2010-2016 and in 2022 and one private oncology network in Brazil between 2020-2023. Treatment delay was defined as an interval ≥60 days. Data were compared between public and private settings. Logistic regression models were used to identify factors associated with treatment delay. Results: A total of 1,644 pts were included, 1,347 (82%) in the public and 297 (18%) in the private setting. Pts aged <50 years accounted for 40.3% of those in the public healthcare system and 67.7% of those in the private sector. Patients in the private setting presented with earlier-stage disease (49.5% diagnosed at stage I-IIA vs. 28.1% in the public setting), whereas a higher proportion of stage IIB-III disease was observed in the public setting (71.9% vs. 50.5%; P < 0.001). Treatment initiation beyond 60 days occurred in 75.4% of pts in the public setting, compared with 14.1% in the private setting (P < 0.001). In multivariable logistic regression analysis, factors independently associated with treatment delay included public healthcare service (OR 19.26, 95% CI 13.3–27.9; P < 0.001), older age compared to ≤30 ys, ranging from OR = 2.06 among patients aged 31–39 years (95% CI 1.02–4.16; p = 0.045) to OR = 5.48 among those aged ≥70 years (95% CI: 2.55–11.81; p < 0.001) and earlier stage (I–IIA vs. IIB-III OR = 1.45, 95% CI 1.10--1.91; p = 0.008). Conclusions: Rates of treatment initiation beyond 60 days after diagnosis are alarmingly high in the Brazilian public health system, indicating that the 60-day law implementation remains suboptimal. Lower odds of treatment delay among younger patients and those with more advanced disease likely reflect clinical prioritization of individuals with more aggressive or symptomatic presentations. These findings underscore the urgent need for health policy interventions aimed at reducing delays in cancer care, particularly for high-risk subtypes such as TNBC.

Multiomic spatial analysis of tumor microenvironment during neuroendocrine (NE) transformation in <i>EGFR</i> -mutant LUAD.

8085 Background: Histological transformation of EGFR mutant lung cancer adenocarcinoma (LUAD) to small-cell lung cancer (SCLC) is a well-known resistance mechanism and has been found in about 14% resistant cases. This phenomenon is associated with extremely poor prognosis and limited therapeutic options. This underscores the need to identify the biological mechanisms triggering lineage plasticity in EGFR+ LUADs. We and others have shown that several genomic and epigenetic alterations play a central role in the activation of small cell like neuroendocrine (NE) states. However, the impact of these changes on the immune microenvironment and vice versa is not understood, limiting the identification of novel immunotherapy approaches in this setting. Methods: We describe the evolution of the tumor microenvironment (TME) during SCLC transformation by comparing tissues from de novo LUAD ( n = 10), de novo SCLC ( n = 15), and transformed SCLC (n =12) including 11 cases with EGFR -mutant LUAD/SCLC mixed histology and 1 pre-transformed LUAD. Additionally, we used de novo lung squamous cancer (LUSC, n = 11), and transformed LUSC (n = 11) including 8 mixed LUSC/LUAD histology and 3 post-transformed samples to inform the LUSC transformation program as a comparator. All samples were analyzed by RNA sequencing, DNA methylation, and whole-exome sequencing. Spatial proteomics (ORION, n = 4) and transcriptomics (XENIUM, n = 6) were performed on pre- and post-transformed paired samples from EGFR+ patients. Results: Differential gene expression and pathway analysis highlight the upregulation of G2M, E2F, and MYC targets during both SCLC and LUSC transformation, occurring at an earlier stage for SCLC compared to LUSC. Conversely, a more pronounced and broader and earlier downregulation of immune pathways was detected during SCLC, compared to LUSC, transformation. Analysis of the epigenetic regulators potentially driving these signaling changes pointed to EZH2 as master regulators of SCLC transformation, while the LUSC program was mainly driven by the TP63 and FOXA1 transcription factors. Spatial proteomics and transcriptomics highlighted a decrease in immune-related programs with an increased proportion of immune suppressive subsets and NE features in SCLC-transformed samples compared to baseline specimens. Sub-clustering analysis showed a higher percentage of pro-tumoral macrophages and exhausted T-cells in transformed samples. Conclusions: Our results suggest that a broader transcriptional reprogramming is associated with the transition from LUAD to SCLC than from LUAD to LUSC. The TF profile associated with SCLC histological transformation negatively regulates immune pathways leading to profound remodeling of the immune compartment and a cold TME that supports the aggressive phenotype of transformed tumors.