Browse Articles

Cultivating the microbiome to enhance cancer immunotherapy

Ultra‐High Dielectric Acceptor Enables 21% Efficiency and Thickness‐Insensitive Organic Solar Cells

ABSTRACT Dielectric constant (ε r ) of non‐fullerene acceptors (NFAs) is a crucial parameter in organic solar cells (OSCs), significantly influencing exciton dissociation efficiency and charge recombination dynamics. However, the state‐of‐the‐art NFAs typically exhibit relatively low dielectric constants (ε r ≈ 3–5), which inherently limit device performance by promoting substantial recombination losses. Herein, a new guest acceptor named L8‐BO‐FO is reported by replacing the branched alkyl chains on L8‐BO with dimethyl ether units, which presents a higher ε r value of 7.41 than that of L8‐BO (4.57). Subsequently, the high‐ε r L8‐BO‐FO was introduced into the PM6:L8‐BO system to improve the dielectric property of the active layer, which, encouragingly, increases the ε r value of the active layer, accelerates exciton dynamics, and suppresses the nonradiative charge recombination. As a result, the power conversion efficiency (PCE) of the ternary OSCs is boosted up to 21.0%. Surprisingly, the ternary OSCs maintained an outstanding PCE of 19.1% even when the active‐layer thickness increased to 300 nm. This work brings new insights into the design of high‐performance OSCs with excellent thickness tolerance by dielectric engineering, which is essential for high‐throughput and scale‐up fabrication of OSCs.

Holter-ECG findings after acute ischemic stroke and TIA: A systematic analysis of the MonDAFIS randomized trial

Abstract Holter-ECG monitoring is a critical component of post-stroke diagnostics, guiding cardiac work-up and secondary stroke prevention. Abnormal ECG findings beyond atrial fibrillation (AF) in stroke patients remain understudied. The prospective multicenter MonDAFIS trial randomized patients with acute ischemic stroke or transient ischemic attack (TIA) without known AF to Holter-ECG recording up to 7 days or usual care. Holter-ECG findings from the first 72 h of the intervention arm were analyzed to provide a reference guide in clinical practice. Furthermore, 24-hour and 72-hour Holter-ECG monitoring were compared to analyze the value of prolonged monitoring. 24-hour Holter-ECGs from 1,665 patients (median age 67; 40.4% women) identified supraventricular tachycardia (SVT) in 4.1% and newly-diagnosed AF in 2.2% of patients. Premature ventricular complexes were common (85.8%), ventricular couplets (28.0%) or bigeminy (14.0%) less common. Non-sustained ventricular tachycardia (nsVT) was detected in 1.7% of patients. Extended 72-hour-monitoring in 1,283 patients led to higher detection rates across all abnormalities, doubling nsVT (4.4%) and SVT (8.8%) detection rates. Generally, we observed higher detection rates with older age. Detection rates of supraventricular arrhythmias were higher in women, whereas men exhibited higher rates of ventricular abnormalities. Post-stroke ECG monitoring detects various arrhythmias beyond AF in a substantial proportion of individuals. Longer monitoring and older age are associated with increased detection rates, with notable sex-specific differences.

A cryptic-site ligand stabilizes a non-canonical interface and blocks membrane insertion of the chloride intracellular channel CLIC1

Belantamab mafodotin with daratumumab, lenalidomide, and dexamethasone in transplant-ineligible, newly diagnosed multiple myeloma patients: Phase 1/2 BelaDRd study.

7512 Background: The MAIA trial established daratumumab with lenalidomide, dexamethasone (DRd) as standard therapy for transplant-ineligible (TI) NDMM. Belantamab mafodotin (belamaf) showed high efficacy in relapsed/refractory disease. Quadruplet regimens improved clinical outcomes, providing a strong rationale to evaluate belamaf in combination with DRd in TI NDMM. Methods: BelaDRd is a phase 1/2, open-label study (EUCT-2024-515634-32) evaluated this combination in TI-NDMM. Part 1 (dose-finding phase) assessed the safety and tolerability of two belamaf doses (1.9/1.4 mg/kg Q8W) with DRd. This part established belamaf 1.9 mg/kg Q8W, extended to Q12W to account as the recommended phase 2 dose (RP2D). Part 2 (dose-expansion phase) evaluates the safety and efficacy of the RP2D in two cohorts Groups A and B, guided by ophthalmologist and hematologist (using Vision-Related Anamnestic, VRA tool), respectively. Results: In Part 1, 24 pts (median age: 73 years; male: 54.2%) entered the study;12.5% had high-risk cytogenetics (HRC). At a median follow-up of 34.3 months, 18 (75%) pts remained on treatment, while 6 (25%) discontinued; 3 (12.5%) death, 1 (4.2%) AE/SAE, 1 (4.2%) withdrew consent, and 1 (4.2%) due to PD. The median dose intensity (MDI) of belamaf was 0.5 mg/kg/Q4W. MRD negativity was achieved in 13 (81.3%) of evaluable patients (n=16). The ORR was 22 (91.7%). Sixteen patients (69.6%) achieved both ≥CR and ≥VGPR. In Part 2 (n=12; median age: 74 years; male: 58.3%),16.7% of pts had HRC. At a median follow-up of 17.9 months, 11 pts (91.7%) remained on treatment; 1 pt (8.3%) discontinued due to PD. Of 113 planned belamaf doses 13 doses were skipped due to OAEs. The MDI was 0.8 mg/kg/Q4W. MRD negativity was achieved in 5 (71.4%) of evaluable patients (n=7). ORR was 11 (91.7%) in evaluable pts. The median time to ≥PR was 1.1 months for both parts. At 18 months, PFS was 91.7% (95% CI 76.4-97.2) and 87.5% (95% CI 66.1-95.8) in the overall and RP2D populations, respectively; corresponding TTP rates were 97.1% (95% CI 80.9-99.6) and 95.5% (95% CI 71.9-99.4). Grade ≥3 BCVA decline was observed in 74(10.4%) of ocular assessments in part 1 and 4(3.9%) and 5(5.3%) in part 2 groups A and B, respectively. Grade ≥2/≥3 keratopathy occurred in 20.2%/0.6% of assessments in Part 1; in Part 2, rates were 6.7%/0% (Group A) and 10.6%/0% (Group B). Median ≥Grade 2 BCVA OAE/keratopathy resolution time was 1.1/1.0 months (Part 1) and 1.9/1.0 months (Part 2). Conclusions: BelaDRd demonstrated robust clinical activity, with rapid and deep responses. Low frequency of ≥Grade 3 OAEs was observed and were rapidly resolved. Moreover, a similar frequency of OAEs was noted when dosing was guided by haematologist vs ophthalmologist ocular assessment. The remarkable PFS supports further evaluation of this quadruplet in a phase 3 study versus other novel quadruplet combinations in NDMM. Clinical trial information: EUCT-2024-515634-32.

Validation of an integrated statistical framework for reconstructing individual-patient data from time-to-event curves.

e23432 Background: Published time-to-event results are most often presented as Kaplan-Meier (KM) plots, but the underlying individual participant data (IPD) are rarely accessible, limiting secondary analyses and evidence synthesis. Methods for reconstructing IPD from KM curves exist but remain technically demanding, non-reproducible, or slow to scale. We developed a provenance-preserving computational platform that enables accurate and reproducible IPD reconstruction directly from published survival figures. Methods: We developed and implemented an open-source, modular, semi-supervised architecture linking each study, figure, and curve to its calibration parameters and metadata. Study PDFs from published randomized control trials (RCTs) are uploaded and metadata including DOI and PMID are extracted automatically. Plots containing KM curves are identified and cropped by the user to establish boundaries based on Cartesian coordinates. KM curves are then masked to isolate each individual treatment curve, at which point the digitized KM plots are geometrically normalised, parsed for number-at-risk tables, and reconstructed using the validated Wei-Royston algorithm, which produces the patient-level data that would have been in the original arm in the RCT. This process was done systematically across an expert-curated selection of clinical trials relevant in urological oncology. Accuracy was assessed by comparing hazard ratios (HRs) and median survival values against those reported in 30 published clinical trials with 73 total curves spanning oncology and non-oncology indications. Comparative evaluation was performed against the fully automated KM-GPT pipeline published elsewhere for three additional studies. Results: Our approach reproduced published HRs with near-perfect concordance (slope = 1·0059, R² = 0·994). The mean difference between reconstructed and reported HRs was -0·0058 (SD 0·0176). Reconstructed survival curves visually overlapped their published counterparts, and absolute error distributions were normally distributed, indicating random rather than systematic variance. Compared with KM-GPT, our approach achieved equivalent or superior accuracy while maintaining full provenance control such that users can run simulations and counterfactual analyses to analyze different scenarios, compare arms from different trials, or conduct population-level analysis to measure benefit at the population level. Conclusions: Our approach demonstrates that reliable, low-latency reconstruction of time-to-event data can be achieved at scale with transparent, auditable provenance. This platform provides a foundation for clinician-guided, reproducible evidence synthesis from the published biomedical literature.

Multimodal risk phenotyping of early-onset colorectal cancer.

10561 Background: Early-onset colorectal cancer (EOCRC) is rising rapidly, yet its risk profile is incompletely defined. Prior studies have largely focused on isolated metabolic or lifestyle factors and have rarely integrated broader sociodemographic context or objective functional measures. We performed multimodal phenotyping to characterize integrated clinical, behavioral, cardiometabolic, and wearable-derived profiles among individuals with EOCRC compared with average-onset CRC (AOCRC). Methods: Adults with CRC in the NIH All of Us Controlled Tier Dataset v8 were identified (≥2 EHR diagnosis codes ≥3 months apart). The primary exposure was age at diagnosis, defined as EOCRC (<50 years) vs AOCRC (≥50 years). Covariates included sociodemographic factors (sex, race/ethnicity, education, employment, health insurance, household income, marital status, neighborhood deprivation). Outcomes included health behaviors (smoking, alcohol use), anthropometrics (BMI, waist circumference), cardiometabolic conditions (dyslipidemia, metabolic syndrome, diabetes, hypertension), laboratory measures (lipids), and wearable-derived (Fitbit) activity, sleep, and heart rate, anchored to the most recent pre-diagnosis timestamp. Results: Among 3,915 CRC participants, 710 (18.1%) had EOCRC. The cohort was predominantly male (54.5%), White (64.4%), unemployed (68.6%), with annual income <$50k (45.5%), and married (53.6%). Participants with EOCRC were less often female (38.7% vs 47.0%) and White (52.7% vs 67.0%) (both p <0.001). EOCRC was associated with higher obesity (42.0% vs 37.0%), higher LDL (101 vs 86 mg/dL), and lower HDL (51 vs 54 mg/dL) (all p <0.001). Alcohol use (60.6% vs 50.5%, p <0.001) and heavy drinking (22.0% vs 16.1%, p =0.003) were more common among participants with EOCRC. In the wearable sub-cohort, EOCRC participants had higher heart rate (76.8 vs 73.8 bpm , p =0.03) and lower sedentary time (773 vs 894 min/day , p =0.016) with no significant differences in sleep metrics or heart rate variability. Female sex was associated with lower odds of EOCRC (OR 0.45, 95% CI 0.25–0.83; p =0.01), while Hispanic ethnicity was associated with higher odds (OR 4.04, 95% CI 1.51–10.86; p =0.006). Conclusions: EOCRC is characterized by higher obesity, dyslipidemia, and alcohol use, as well as differences in wearable-derived physiologic and behavioral profiles. Integrating real-world clinical data with digital biomarkers offers a scalable framework for refined risk stratification and precision prevention in EOCRC. Multivariable logistic regression examining sociodemographic predictors for EOCRC. Variable OR (95% CI) P value Sex Male Ref Female 0.45 (0.25–0.83) 0.010 Race/ethnicity White Ref Black 2.43 (0.99–5.97) 0.052 Hispanic 4.04 (1.51–10.86) 0.006 Employment Unemployed Ref Employed 3.38 (1.87–6.10) <0.001

Does where you live matter?: State-level variation in liver cancer mortality.

e16344 Background: In the United States, liver cancer is among the most lethal cancers, highlighting the need for improved systems of care. Survival for patients with liver cancer may vary by state of residence, and factors related to this difference remain unexplored. The objective of this study was to assess differences in mortality for liver cancer by state and examine factors associated with any observed variation in mortality. Methods: Age-adjusted liver cancer incidence and mortality rates for 2022 were obtained for 50 U.S. states and the District of Columbia from the Centers for Disease Control and Prevention Wide-Ranging Online Data for Epidemiologic Research (CDC WONDER). Liver cancer was defined using ICD-10 codes and included hepatocellular carcinoma and intrahepatic cholangiocarcinoma. For each state, the mortality to incidence ratio (MIR) was calculated as age-adjusted mortality divided by incidence. The state with the lowest MIR was designated as the top-performer and used as a benchmark. Delta MIR (dMIR) was defined as the difference between a state’s MIR and the benchmark. Variation in cancer mortality relative to incidence was visually depicted using heatmaps. Factors associated with mortality variation were examined using Spearman correlation analyses of 12 publicly available, state-level economic and health system indicators, such as healthcare expenditure per capita, insurance coverage and life expectancy. Weighted linear regression analyses were subsequently performed using inverse-variance weights from dMIR standard errors. Results: The median dMIR was 0.142, with wide state-level variation ranging from 0.005 to 0.501, representing an approximately 96-fold difference between the lowest and highest dMIR. New York demonstrated the best performance with the lowest MIR and served as the benchmark. Maine, Utah and Minnesota ranked as the next top 3 performers. In contrast, the highest dMIR values, indicating the greatest excess mortality, were observed in North Dakota (0.400), Rhode Island (0.403), and South Dakota (0.501), ranking these as the worst performing states for liver cancers. Univariate analysis showed that higher dMIR (worse performance) was significantly associated with higher percentage of minority population (p = 0.009) and fewer cancer centers (p = 0.025). In multivariable analysis, higher dMIR was significantly associated with lower health expenditure per capita (p = 0.046), lower life expectancy (p = 0.006) and higher percentage of minority population (p = 0.046). Conclusions: Liver cancer mortality outcomes vary substantially across states, with some states outperforming others. dMIR can be used as a benchmarking tool to explain geographic disparity and guide targeted cancer control efforts and policy interventions to improve mortality outcomes. Our study suggests that lower investment in public health infrastructure is associated with higher relative mortality for liver cancer.

Common symptoms, quality of life, and lifestyle factors reported by cancer survivors attending the Sydney Cancer Survivorship Multidisciplinary Clinic.

12113 Background: Sydney Cancer Survivorship Clinic (SCSC) provides multidisciplinary care for cancer survivors. During an initial visit, survivors consult with a medical oncologist or hematologist, cancer nurse specialist, dietitian, clinical psychologist, and exercise physiologist. Over 12 years, we assessed the health status of cancer survivors after curative-intent treatment. Methods: Survivors completed patient-reported outcomes assessing symptoms (PT-DATA), quality of life (QOL) (FACT-G), distress (Distress thermometer), fear of cancer recurrence (FCR-7, and psychologist assessment), diet, and physical activity (LTEQ) before initial clinic visit. Clinical data were obtained from the medical record. Simple descriptive statistical methods were used. FACT-G scores were standardized and compared to Australian general population and population with cancer (T-score 50, SD 10). Results: Between September 2013 and December 2025, 1181 survivors attended initial SCSC; 1081 (95%) consented to the use of their data. Of these, 68% were female; median age 58 years (range 18-91). Median time from diagnosis 10.7 months (IQR 8.4-16.6). Tumor types: 38% breast, 33% colorectal, 16% hematological, 14% other. Primary treatment: surgery 84%, chemotherapy 90%, radiotherapy 40%. Overall, 47% had 5+ symptoms of at least moderate severity (4+/10). Most frequently reported moderate-severe symptoms were: fatigue (50%), insomnia (40%), anxiety (36%), pain (35%), numbness (33%), sore hands/feet (32%), trouble concentrating (32%). Moderate-severe distress was reported in 44%. On the FCR-7, 32% had moderate-severe FCR. The psychologist rated 46% as having moderate-severe FCR. Mean QOL (FACT-G): total 79.9: domains physical 22.1; social 21.2, emotional 18, functional 18.6. Comparison to Australian populations: Total T-score 46.0: physical and emotional domains both 0.8 SD below. 59% were overweight or obese (mean BMI 27.2kg/m 2 (range 16.2-59.1)). Only 7% met national exercise guidelines for both resistance and aerobic exercise (18% resistance exercise alone, 24% aerobic exercise alone). Conclusions: Symptom burden remains high almost a year after diagnosis, with 47% reporting at least five moderate-severe symptoms; most commonly distress, fatigue, insomnia, anxiety and FCR. Overweight/obesity (59%) and physical inactivity remain lifestyle risk factors.

Evaluating CT-derived body composition and colorectal cancer risk: A group-based trajectory modeling morphomics study.

e22530 Background: Obesity is an known risk factor for colorectal cancer (CRC), yet body mass index (BMI) incompletely captures metabolic risk, particularly in females and racial and ethnic minorities. Body composition is a more precise measure, as it describes visceral, subcutaneous, and intramuscular fat, and lean and skeletal muscle mass, which have different implications for CRC risk, but trajectories of body composition are under-investigated. CT-based analytic morphomics can quantify body composition and may enable “opportunistic” risk stratification using routine imaging. Methods: We conducted a retrospective cohort study within the Veterans Health Administration (VHA) using paired CT scans with morphomic measures at L3. We included CRC patients with age- and sex-matched controls with at least 2 CT scans > 1 year prior to CRC diagnosis or last known follow-up. Scan phases were selected using a standardized preference hierarchy: non-contrast, delayed, venous, arterial (all patients analyzed had matching phases). Two group-based trajectory modeling (GBTM) strategies were evaluated: age as the time scale (GBTM-AGE) and CT timepoint while adjusting for age. The primary outcome was time to CRC from the first CT. Associations between trajectory group membership and CRC were estimated using Cox proportional hazards models adjusted for age, sex, and BMI. Results: The final cohort included 39 patients (22 no CRC; 17 CRC), predominantly male and non-Hispanic White, with median age 60 years at first CT (Table 1). Among CRC cases, most cancers were colon primaries (64.7%), and over half presented with stage III/IV disease (64.7%). In GBTM-AGE models, membership in the second trajectory group for mean skeletal muscle attenuation, a marker of myosteatosis, was associated with significantly higher CRC hazard (HR 4.21, 95% CI 1.38–12.91; p = 0.012), independent of age, sex, and BMI. Other evaluated trajectory groupings were not significantly associated with CRC risk in adjusted models. Conclusions: In this VHA cohort with longitudinal CT morphomics, a trajectory class of skeletal muscle attenuation identified patients at elevated future CRC risk independently from age and BMI. These findings suggest that CT-based morphomics may be a feasible avenue for CRC risk stratification and emphasize the need for further collection and evaluation of morphomics data for opportunistic screening for CRC. Demographic characteristics. Variable No CRC (n = 22) CRC (n = 17) P-value Age at 1st CT Scan; Median [Q1,Q3] 60.0 [50.3, 66.5] 60.0 [52.0, 71.0] 0.403 Age at 2nd CT Scan; Median [Q1,Q3] 62.0 [52.8, 69.0] 63.0 [54.0, 72.0] 0.505 Age at CRC; Median [Q1,Q3] 69.0 [57.0, 79.0] Male; n (%) 18 (81.8%) 16 (94.1%) 0.363 Non-Hispanic White; n (%) 22 (100%) 17 (100%) 0.423 BMI at 1st CT Scan; Median [Q1,Q3] 29.4 [26.1, 34.4] 28.9 [25.7, 32.7] 0.865 BMI at 2nd CT Scan; Median [Q1,Q3] 29.1 [26.3, 34.0] 28.6 [24.4, 36.2] 0.821

Age-dependent pathogenic variant risk in female breast cancer.

1624 Background: Germline pathogenic and likely pathogenic variants (GPVs) in BRCA1 , BRCA2 , and PALB2 raise female breast cancer risk. Genetic counseling and screening guidelines rely on fixed age estimates. Using two large biobanks, All of Us Research Program (AoURP) and BioVU biobank, we aim to characterize age-dependent breast cancer risk trajectories and associations with family history (FH). Methods: In AoURP, we extracted data on female participants with genetic data to identify those with BRCA1 , BRCA2 , and PALB2 GPVs, FH surveys to identify those with first degree breast cancer FH (BCFH), and electronic health records (EHR) to determine presence of breast cancer diagnosis and diagnosis age. Age served as the time scale, with left truncation at first EHR entry and censoring at last EHR event. Gene-specific breast cancer risk was estimated using Cox models with age-dependent genetic effects, adjusted for genetic ancestry (first five principal components) and FH. Analyses in AoURP were stratified by self-reported race and BCFH. We validated results in BioVU using the same methodology excluding FH which was unavailable as discrete data. Results: In AoURP, 100,916 women with genetic, FH, and EHR data were available. The majority (67%) were White, based on self-reported race, and 5208 had breast cancer. The sample included 291 BRCA1 GPVs, 505 BRCA2 GPVs, and 139 PALB2 GPVs, with breast cancer diagnosed in 36%, 23%, and 19% respectively. BioVU included 127,147 women, of whom 85.5% were White and 5,664 had breast cancer. The sample had 262 BRCA1 GPVs, 515 BRCA2 GPVs, and 210 PALB2 GPVs, with breast cancer diagnosed in 29%, 22%, and 16% respectively. Across the genes, we observed a bimodal distinct age-dependent breast cancer risk pattern. Bimodal peaks for BRCA1 in AoURP were at 26 and 36 years, while in BioVU were at 35 and 62 years. The bimodal peaks for BRCA2 in AoUPR were at 32 and 51 years, while in BioVU were at 32 and 55 years. Finally, bimodal peaks for PALB2 in AoUPR were 35 and 51 years, while in BioVU were at 41 and 64 years. Age-specific risk trajectories also varied by FH where risk peaks for participants with BCFH were earlier by 2-7 years compared to those without BCFH. Conclusions: We presented population based study providing unbiased estimates of breast cancer risk associated with GPVs in BRCA1, BRCA2, and PALB2 in two enrollment-based cohorts. Leveraging these large datasets, we identified novel bimodal age-dependent patterns of breast cancer risk across genes, with consistent peak ages for BRCA2 and PALB2 and cohort specific differences for BRCA1, highlighting the importance of population context. Replication in BioVU confirmed similar bimodal trends with later peak ages, supporting our finding robustness. Our results show that breast cancer risk in GPVs genes is more dynamic than previously recognized. Integrating age and FH risk estimates into genetic counseling and screening may improve personalized risk assessment and clinical decision making.

County-level palliative care physician density and socioeconomic determinants of place of cancer death in the United States, 2021-2023.

12032 Background: Palliative care, a specialized approach focused on care for individuals with serious illness including cancer, has been shown to improve end-of-life outcomes, including higher rates of death in preferred settings. However, access to palliative care remains inequitable due to significant geographic disparities in workforce distribution. This study examines the association between county-level palliative care physician density and place of death among US decedents died from cancer. Methods: We conducted a retrospective cohort study using CDC WONDER multiple cause of death data (2021-2023) linked with palliative care physician supply data from the American Medical Association (AMA) Physician Masterfile. The primary exposure was county-level palliative care physician density, categorized as zero, below-median, or above-median (among non-zero counties). The primary outcome was the county-level percentage of cancer deaths (ICD: C00-C97) occurring in non-hospital settings (home, hospice, long-term care, or nursing home). Covariates include county sociodemographic characteristics (race/ethnicity, median income, region, metropolitan status) and hospice facility availability. Multivariable regression models estimated associations between physician density and place of death, adjusting for other covariates. All models incorporated county population weighting and state-level clustering. Results: Most U.S. counties (n = 2056, 66.27%) had no palliative care physicians in 2020. Counties with palliative care physicians showed higher median household incomes, more hospice facilities and metropolitan status, and were concentrated in the Northeast and West (all p < 0.001). The average percentage of non-hospital facility deaths across all counties was 56.6% (SD = 7.1) for all causes, 60.5% (SD = 8.1) for heart disease, and 70.7% (SD = 8.0) for malignant neoplasms. Multivariable regression models revealed a dose-response relationship: counties with non-zero below median and above-median palliative care physician density showed 0.9 percentage points (ppts, p < 0.05) and 1.1 ppts (p < 0.05) higher proportions of cancer death at non-hospital facility-based facilities than counties without palliative care physicians. In addition, Metropolitan and Micropolitan counties had 2.5 ppts (p < 0.001) and 2.1 ppts (p < 0.001) higher proportions of cancer death at non-hospital facility-based facilities than non-metropolitan counties respectively. Conclusions: Higher county-level palliative care physician density is associated with increased non-hospital facility-based deaths for cancer, demonstrating a clear dose-response relationship. These findings suggest that ensuring access to palliative care, through workforce expansion or care delivery model innovation may facilitate end-of-life care planning aligned with patient preferences.

Association of neoadjuvant therapy with perineural invasion and postoperative survival in locally advanced colon cancer.

e15634 Background: Perineural invasion (PNI) is a marker of aggressive tumor biology and adverse prognosis in colon cancer. The impact of neoadjuvant therapy on PNI and its relationship with postoperative survival remains poorly defined. Methods: We conducted a retrospective cohort study using the National Cancer Database including patients with clinically locally advanced, non-metastatic colon cancer (cT3-4, cN+, cM0) who underwent definitive resection between 2010 and 2022. Neoadjuvant therapy was defined as receipt of chemotherapy, immunotherapy, or radiation therapy prior to surgery. Multivariable logistic regression evaluated the association between neoadjuvant therapy and PNI. Overall survival (OS) was measured from surgery and analyzed using Cox proportional hazards models adjusted for baseline demographic, clinical, and tumor characteristics. Postoperative pathologic features were not included in primary survival models. Results: Among 15,471 patients, receipt of neoadjuvant therapy was independently associated with lower odds of PNI (aOR 0.56, p < 0.001). Neoadjuvant therapy was also associated with improved postoperative OS (HR 0.84, p = 0.001). In treatment-sequence sensitivity analyses, neoadjuvant therapy alone, adjuvant therapy alone, and combined neoadjuvant plus adjuvant therapy were associated with improved survival compared with surgery alone, with the greatest benefit observed among patients receiving combined therapy. In stratified analyses, neoadjuvant therapy was associated with improved survival among PNI-negative tumors, whereas in PNI-positive disease, survival benefit was primarily observed among patients receiving adjuvant or combined therapy. Conclusions: Neoadjuvant therapy is associated with reduced perineural invasion and improved postoperative survival in locally advanced colon cancer. In PNI-positive disease, survival benefit appears dependent on receipt of systemic therapy rather than neoadjuvant timing alone.

TTFields with maintenance temozolomide (TMZ) and pembrolizumab versus TTFields with maintenance TMZ and placebo for newly diagnosed glioblastoma: The phase 3 EF-41/KEYNOTE D58 trial.

TPS2096 Background: Glioblastoma remains a lethal disease with limited treatment options. With current standard-of-care, consisting of maximal safe resection followed by radiotherapy with concurrent TMZ, then maintenance TMZ plus TTFields, median overall survival (OS) is 20.9 months, as reported in the pivotal EF-14 trial completed more than a decade ago. While immune checkpoint inhibitors (ICIs) have improved outcomes in multiple malignancies, they have not demonstrated meaningful clinical benefit in glioblastoma, in part because the tumor microenvironment (TME) is typically profoundly immunosuppressive. Strategies that reprogram the TME toward immune activation may therefore enable ICIs to elicit more effective anti-tumor immunity. Beyond antimitotic effects, TTFields have shown preclinical evidence of inducing immunogenic cell death and activating type 1 interferon signaling via DNA sensor inflammasomes, with downstream increases in dendritic cell activation and cytotoxic T cell infiltration. In a single-arm phase 2 study (NCT03405792), TTFields plus TMZ and pembrolizumab was associated with improved progression-free survival (PFS) and OS compared to historical controls in newly diagnosed glioblastoma. Here, we describe the design of a randomized phase 3 trial evaluating this regimen. Methods: EF-41/KEYNOTE D58 is a randomized, double-blind, placebo-controlled, phase 3 study (NCT06556563). Eligible patients have newly diagnosed glioblastoma (WHO 2021 Classification), have completed concurrent chemoradiotherapy, and can initiate treatment 4–7 weeks thereafter. Additional criteria include ECOG performance status 0–1 and availability of tumor tissue for central MGMT methylation analysis. Key exclusion criteria include prior anti-PD(L)1 or anti-PDL2 therapy and ongoing dexamethasone >2 mg/day. Patients are randomized 2:1 to TTFields (200 kHz for ≥18 hours/day) plus maintenance TMZ (150–200 mg/m 2 /day PO, days 1-5 of each 28 day cycle for 6-12 cycles) with either pembrolizumab 200 mg IV Q3W (up to 35 cycles) or matching placebo. TTFields is continued until second progression. At first progression, TTFields therapy is maintained and patients may receive standard salvage therapy, including re-resection and/or radiotherapy as well as systemic therapy. The target enrollment is 741 patients, providing 85% power to detect an OS improvement at a two-sided alpha of 0.05 using a 2-sided log-rank test. The primary endpoint is OS. Secondary endpoints include PFS per RANO 2.0 and RANO, PFS6 and PFS12 rates (RANO 2.0), PFS2 (RANO 2.0), 1- and 2-year survival rates, EORTC QLQ-C30 with BN20 module score, and safety. MRI assessments are performed every 9 weeks and evaluated per RANO 2.0. Adverse events are monitored throughout the study. Enrollment is ongoing. Clinical trial information: NCT03405792 .

Associations between self-efficacy and symptom burden in young women with breast cancer initiating endocrine therapy (ET) on SWOG S2010.

521 Background: Self-efficacy (SE) is the belief in one's ability to succeed in specific situations or accomplish a task, influencing motivation, behavior, and resilience. Both symptom (sx) burden at the time of ET initiation and lower SE for sx management are associated with ET nonadherence. The SWOG S2010 clinical trial enrolled premenopausal women starting adjuvant ovarian function suppression (OFS) plus ET for breast cancer. We examined associations between sx burden and self-efficacy at the time of ET initiation. Methods: All participants completed the FACT GP5 question evaluating bother from side effects of treatment, individual symptom PROMIS measures, and the PROMIS SE for managing sx and for managing medication (meds)/treatment (tx). GP5 is measured on a 5 point scale ranging from “Not at all” to “Very much.” T scores were calculated for PROMIS measures (Table), with a mean score of 50 and higher scores reflecting more of the item being assessed. We evaluated associations between baseline sx and SE using logistic regression. Results: Of 528 participants with data, mean age was 44.1 years (SD 5.8), 23.5% were Hispanic, 5.5% Black, 6.4% Asian, and 77.7% White. The majority (69.9%) had received chemotherapy. On the FACT GP5 (n=503), only 31.6% reported no treatment side effect burden. Patients reporting a higher side effect burden from tx immediately prior to ET initiation had lower SE for sx (odds ratio (OR) 0.93 (95% CI 0.91-0.96, p<.0001)) and lower SE for meds & tx (OR 0.97 (95% CI 0.95-1.00, p=.04). Increased baseline sx and worse physical function were also statistically significantly associated with worse SE for managing both sx and meds & tx (Table). For example, patients with greater than average fatigue had lower SE for sx (OR 0.32 (95% CI 0.23-0.46, p<.001)). Conclusions: Symptom burden at the time of ET initiation is high for premenopausal women starting OFS plus ET, which may partially explain high rates of early ET discontinuation in this patient population. Lower SE was strongly associated with worse individual symptoms. Interventions to improve SE for sx and med management starting at or before ET initiation may improve ET adherence, which could thereby improve disease outcomes and quality of life for young women with breast cancer. Funding: NIH/NCI/NCORP grant UG1CA189974. Clinical trial information: NCT05568472 . Associations between self-efficacy and symptoms. PROMIS Self Efficacy Sx PROMIS Self Efficacy Meds & Tx PROMIS domain <50 ≥50 OR (CI) P value (Chi square) <50 ≥50 OR (CI) P value (Chi square) Fatigue <50 97 144 0.32 (0.23,0.46) <.001 82 158 0.52 (0.36,0.74) <.001 ≥50 194 93 143 143 Anxiety/Fear <50 56 105 0.30 (0.20,0.44) <.001 42 118 0.36 (0.24,0.54) <.001 ≥50 235 132 183 183 Sleep disturbance <50 85 120 0.40 (0.28,0.58) <.001 69 135 0.54 (0.38,0.78) .001 ≥50 206 117 154 166 Physical function <50 195 82 3.84 (2.67,5.51) <.001 146 131 2.40 (1.68,3.42) <.001 ≥50 96 155 79 170 CI: Confidence intervals; OR: Odds ratio.

When autoimmunity meets immunotherapy: Immune checkpoint inhibitors in cancer patients with rheumatoid arthritis.

11168 Background: Immune checkpoint inhibitors (ICIs) improve outcomes across multiple malignancies but are associated with immune-related adverse events (irAEs). Patients with rheumatoid arthritis (RA) have been largely excluded from clinical trials, and disease-specific real-world data remain limited. We evaluated irAEs and survival outcomes in cancer patients with and without preexisting RA treated with ICIs. Methods: We conducted a retrospective cohort study using the TriNetX Research Network. Patients with RA diagnosed prior to ICI initiation were compared with patients without RA. Propensity score matching (1:1) was performed on age, sex, race, ethnicity, cancer type, ICI type, comorbidities, metastatic disease, other autoimmune conditions, and preexisting irAE codes. Outcomes were measured beginning 1 day after ICI initiation. Overall survival (OS) and organ-specific irAEs were assessed. Sensitivity analyses were conducted excluding patients with baseline glucocorticoid or disease-modifying antirheumatic drug (DMARD) exposure to account for differences in preexisting autoimmune treatment. Results: Before matching, 3,172 patients with RA and 156,543 without RA were identified. After matching, 3,162 patients were included in each cohort with balanced baseline characteristics and laboratory values. OS was numerically lower in the RA cohort but not statistically significant (29% vs 30%; median 21.9 vs 23.7 months; p = 0.1861). Early survival differed within 6 months (76% vs 78%; p = 0.0415) but not at 1 year. RA patients experienced higher rates of gastrointestinal (8.4% vs 6.9%; p = 0.0234) and pulmonary irAEs (5.4% vs 4.0%; p = 0.0106), with no significant differences in hepatic, dermatologic, or endocrine irAEs. Glucocorticoid and DMARD use were more frequent in RA patients. In sensitivity analyses excluding prior exposure, new DMARD use remained significantly higher in RA patients (4.5% vs 1.0%; p < 0.0001), while other differences attenuated. Conclusions: Cancer patients with preexisting RA treated with ICIs experience higher rates of select irAEs without a significant long-term OS disadvantage. These findings suggest increased immune toxicity risk with preserved anticancer efficacy, supporting cautious but feasible ICI use in RA patients with appropriate monitoring.

Real-world experience of cancer patients with clonal hematopoiesis in an urban county hospital system.

e18630 Background: Clonal hematopoiesis (CH), including CH of indeterminate potential (CHIP) and clonal cytopenia of undetermined significance (CCUS), is associated with increased comorbidities and risk of myeloid transformation. Most CH studies focus on predominantly white, healthy populations. Outcomes in cancer patients (pts), particularly those treated in safety-net settings, remain poorly defined. We evaluated clinical characteristics of cancer pts with CH in an underserved county hospital. Methods: We retrospectively identified oncology pts with CH detected by next-generation sequencing (NGS) panel xF+ (Tempus AI, Chicago, IL) at a county hospital from November 2024 to October 2025. Pt characteristics at CH detection were assessed and compared with those of cancer pts with CH from a tertiary cancer hospital. Survival was updated in December 2025. Results: Out of 256 pts with xF+ data, 40 pts (16%) had CH (CHIP n=19 [48%], CCUS n=21 [52%]). Median age was 61 years (range: 24, 85); 23 were female (58%). Race was predominantly white (n=35, 88%) and ethnicity Hispanic (n=29, 73%). Median state Area Deprivation Index (ADI) was 6 (range: 1, 10), national ADI 65 (range: 2, 93), and Rural-Urban Commuting Area code 1 (range: 1, 9). The most common primary tumors were colorectal (n=8, 20%), lung (n=6, 15%), and cervical (n=4, 10%). Median Adult Comorbidity Evaluation (ACE)-27 score was 3 (range: 2, 3), with cardiovascular disease most prevalent (n=27, 68%). Median absolute neutrophil count (ANC) was 4.7 (range: 2.0, 18.8), hemoglobin 11.9 (range: 7.0, 14.8), platelet count 271 (range: 108, 696), and mean corpuscular volume (MCV) 90 (range: 67, 100). The most common mutations were TP53 (n=11, 27.5%), DNMT3A (n=11, 27.5%), and PPM1D (n=8, 20%); 3 (8%) pts had >1 CH mutation. By CH Risk Score (CHRS), 4 (10%), 17 (43%), and 19 (47%) pts were low, intermediate, and high risk, respectively. With median follow-up time 5.06 months (95% confidence interval [CI]: 4.37, 6.31), no pts developed myeloid malignancy, and median overall survival was not reached. Seven pts (18%) died, all from complications of their primary cancer. Compared with 59 tertiary-center cancer pts with CH, county hospital pts were younger (p=0.001), more often Hispanic (p<0.001), more socioeconomically disadvantaged (p=0.007), and from more metropolitan areas (p=0.048). They had higher ACE-27 scores (p<0.001), higher ANC (p=0.003) and platelet count (p<0.001), and lower MCV (p=0.005) and trended toward higher CHRS scores (12.0 vs 11.0, p=0.054). The cohorts underwent different NGS panels, and cancer center pts had a longer median follow-up time of 24.02 months (95% CI: 18.86, 34.04). Conclusions: Cancer pts with CH in an underserved county hospital had higher comorbidity burdens and CHRS scores than that of a traditional CH pt cohort. Longer follow-up time is needed to define transformation and survival outcomes in this pt population.

Physician preferences in large language models for breast cancer management.

e12555 Background: As large language models (LLMs) are increasingly used as decision-support tools in breast oncology, their comparative performance in management plan guidance remains unclear. We evaluated whether breast medical oncologists perceive measurable quality differences across LLMs and which models are preferred. Methods: We created 11 synthetic breast cancer vignettes representing common clinical scenarios with specified tumor characteristics, stage, prior therapy, performance status, and medical context. Six LLMs were queried August 10-15, 2025 for management plans: ChatGPT-5 Fast, ChatGPT-5 Thinking, Claude Sonnet 4, DeepSeek-V3, Grok 4, and OpenEvidence. In an electronic survey, breast medical oncologists at University of California, San Francisco and affiliated sites were shown two de-identified, randomized treatment plans per vignette and prompted to select the superior option. We estimated relative performance using Elo ratings (starting 1500; K=32), which summarize pairwise preferences where higher scores indicate greater preference probability. We calculated 95% confidence intervals using 1000 bootstrap resamples. Results: Five oncologists completed 49 head-to-head comparisons across 11 vignettes. ChatGPT-5 Thinking and Grok achieved the highest Elo ratings and were preferred in 68.8% (11/16) and 66.7% (12/18) of comparisons, respectively. Remaining models: DeepSeek 53.3% (8/15), ChatGPT-5 Fast 50.0% (6/12), OpenEvidence 38.1% (8/21), and Claude 25.0% (4/16). Bootstrap 95% CIs supported this ranking. The rank order was the same using unbootstrapped Elo ratings. Conclusions: Breast medical oncologists consistently preferred management plans from ChatGPT-5 Thinking and Grok, demonstrating measurable perceived quality differences across LLM-generated treatment recommendations. Large language model rankings by Elo rating for breast cancer management recommendations. Large Language Model(in order of highest to lowest Elo rating and win rate) Elo Rating (Bootstrap Mean, 95% CI) Win Rate ChatGPT-5 Thinking 1573.3 (1486.2-1658.7) 11/16 wins (68.8%) Grok 4 1571.9 (1488.8-1649.8) 12/18 wins (66.7%) DeepSeek-V3 1505.6 (1423.7-1587.6) 8/15 wins (53.3%) ChatGPT-5 Fast 1496.1 (1421.2-1568.6) 6/12 wins (50.0%) OpenEvidence 1448.9 (1362.9-1536.1) 8/21 wins (38.1%) Claude Sonnet 4 1404.2 (1332.4-1486.1) 4/16 wins (25.0%)

Phase III trial of brain MRI surveillance in stage IV breast cancer.

TPS1160 Background: The incidence and prevalence of breast cancer patients who develop brain metastasis is increasing. Intracranial surveillance of asymptomatic stage IV breast cancer patients with brain MRIs is not currently recommended by the National Comprehensive Cancer Network (NCCN) Guidelines. In a previously conducted phase II brain MRI surveillance trial, approximately 25% of patients with stage IV breast cancer, independent of tumor subtype, were found to have newly detected brain metastases on the 6-month follow-up MRI from study entry. Methods: The study is designed as a randomized phase III, multi-institutional prospective trial evaluating the role of surveillance brain MRIs in neurologically asymptomatic patients with metastatic breast cancer, based on subtypes (triple negative (TN), HER2+, and hormone receptor (HR)+ breast cancer). Following study enrollment, patients will be randomized in a 1:1 fashion to surveillance brain MRIs at baseline and q6 months for 24 months compared to a baseline MRI followed by standard-of-care brain MRI surveillance for symptomatic presentation. Newly diagnosed stage IV patients (within ≤ 60 days of starting systemic therapy for HER2+ or TN or within ≤ 60 days of initiating 1 st or 2 nd line therapy for HR+/HER2- disease) with an ECOG ≤ 2 and life expectancy ≥ 6 months are eligible. The primary objective is to evaluate the treatment characteristics of brain metastases diagnosed via brain MRI surveillance or standard of care imaging. Secondary objectives include the frequency of asymptomatic brain metastases and leptomeningeal disease, the number and size of brain metastases at diagnosis, quality of life, and overall and brain metastases specific survival following brain metastasis diagnosis in patients randomized to brain MRI surveillance compared to standard of care imaging. A total of 156 patients will be enrolled with an equal distribution of TN, HER2+, and HR+/HER2- subtypes, respectively. Funding: Florida Department of Health. Clinical trial information: NCT07357298 .

The California End of Life Option Act: A case series of 77 consecutive cancer patients.

e24041 Background: California is 1 of 10 US states with legalized medical assistance in dying for the terminally ill. There is little published on the actual experience of assisting cancer patients with this intervention. Methods: 77 consecutive cancer patients for whom “End of Life Option Act” (EOLOA) prescriptions were written by one MD (onc/palliative/hospice) from 2019-2026. Results: Of 77 prescriptions written, 43 took the drugs. There were 14 prostate, 12 lung, 6 breast, 5 colorectal, 4 oral/pharynx. “Second MDs” included med oncs, PCPs, rad oncs, hospitalists, nursing home MDs, surgeons. Most frequently prescribed was a combination of 1 gm morphine, 1 gm diazepam, 50 mg digoxin, and 3.75 gm amitriptyline, all powdered by one pharmacy. At the home an 8 oz slurry was made, with vanilla ice cream and chocolate syrup to mitigate bitterness. All patients took antiemetics 30 min prior. All were on hospice at time of ingestion, and none were motivated by uncontrolled pain. The MD was at the home for 58% of patients at time of ingestion, and 21% at time of death. Patients were asleep 20 min after ingestion (5-90) and died a mean of 4.2 hours (range 1-22) after ingestion. Several patients were unable to ingest the full slurry; all succumbed. There were 3 subgroups: those who took the meds immediately (days), those who intended to after addressing affairs (weeks to months), and those who never intended to but wanted the security of having them. There was gratitude at all stages; “relief” at having the meds was the word used most. No patient who took the medications had a “bad death.” At times there was unexpected value. Family were often brought together when it would not have been feasible otherwise. The dying process was “normalized.” Physician and/or hospice team presence and involvement were instrumental in providing bereavement counseling. Minutes after one woman died, her son in law, silent throughout the process, opened up for the first time regarding the recent suicide of his own son. One woman with metastatic lung cancer had decisively quit treatment and follow up, convinced it was ineffective; after obtaining the EOLOA medications, she gradually though reluctantly was convinced to reconsider re-evaluation, which ultimately led to a return to treatment (with a 4 year-long response to immunotherapy), still alive and NED. Conclusions: 1) The EOLOA intervention was without serious complication in all 77 cancer patients; 2) There were 3 subpopulations of patients with regard to intent and outcome; 3) Virtually all patients and families were grateful; the anxiolytic effect of having the meds was obvious; 4) There appeared to be much benefit in the MD assisting with prep of meds and facilitating on-site bereavement; 5) There were patients/families for whom the process was clearly therapeutic; 6) MDs and their teams are encouraged to consider playing a meaningful role in their patients’ end of life care for patients who wish to implement the EOLOA.