Browse Articles

A randomized, double-blind, placebo-controlled pilot study of add-on beta-sitosterol supplementation on treatment response in metastatic prostate cancer patients.

e17073 Background: Metastatic prostate cancer remains difficult to manage due to treatment resistance and cumulative toxicity of systemic therapies. Beta-sitosterol, a dietary phytosterol, has demonstrated anticancer activity in preclinical prostate cancer models and symptomatic benefit in benign prostatic hyperplasia, but clinical evidence in prostate cancer is lacking. Additionally, circulating ceramides are emerging as potential biomarkers of tumor burden and disease progression. This pilot trial was designed to evaluate the feasibility, safety, and preliminary clinical effects of beta-sitosterol in metastatic prostate cancer. The aim was to assess the effect of add-on beta-sitosterol supplementation on prostate-specific antigen (PSA) levels, urinary symptoms, safety, and tolerability in patients with metastatic prostate cancer, and to explore the relationship between baseline plasma ceramide levels and PSA. Methods: This was a randomized, double-blind, placebo-controlled, parallel-group pilot trial conducted. Forty-one men with biopsy-proven metastatic prostate cancer were randomized to receive either beta-sitosterol 320 mg once daily (n≈20) or placebo (n≈21) for three months, in addition to standard therapy. The primary outcome was the least-squares mean difference in log₁₀ PSA at three months. Secondary outcomes included changes in International Prostate Symptom Score (IPSS) and safety. Plasma ceramide levels were measured at baseline using ELISA. Results: At three months, there was no statistically significant difference in log₁₀ PSA levels between the beta-sitosterol and placebo groups (p > 0.05). However, patients receiving beta-sitosterol showed a greater reduction in IPSS scores over follow-up, with a higher proportion shifting from moderate-to-severe symptoms to milder categories compared to placebo (p < 0.05). Beta-sitosterol was well tolerated, with no serious adverse events and only mild gastrointestinal complaints reported. Baseline plasma ceramide levels showed a significant positive correlation with PSA (Pearson correlation, p < 0.05), suggesting an association with tumor burden. Conclusions: Add-on beta-sitosterol was safe and improved urinary symptoms but did not significantly reduce PSA levels. Larger trials are required to confirm efficacy and validate ceramide as a prognostic biomarker. Clinical trial information: CTRI/2024/02/063329.

Expression of Concern: Mapping Quantitative Trait Loci of Resistance to Tomato Spotted Wilt Virus and Leaf Spots in a Recombinant Inbred Line Population of Peanut (Arachis hypogaea L.) from SunOleic 97R and NC94022

Exploring mango (Mangifera indica) fruit peel extract mediated bio-preparation of CuO nanoparticles for biological, dye degradation and sensor applications

Effect of Atomic Layer Deposition of Ultra‐Thin Oxide on Reactivity and Durability of Perovskite Oxygen Electrodes (Adv. Mater. 32/2026)

Crystallinity‐Engineered Three‐Dimensional Graphitic Carbon Tube Grids as Load‐Tolerant Electrodes for AC Line‐Filtering Capacitors

ABSTRACT Three‐dimensionally (3D) architected carbons with oriented nanopores provide a promising platform for AC line‐filtering electric double‐layer capacitors (EDLCs). However, their performance, particularly at high electrode loading, is fundamentally constrained by insufficient electronic conduction—an intrinsic but largely overlooked limitation. Here, we develop a crystallinity‐engineered, highly conductive 3D graphitic carbon tube grid (3D‐GCTG) using a 3D nickel nanorod grid (3D‐NiNRG) as both structural template and catalytic framework. A central advance lies in elucidating and resolving structural collapse and granulation in 3D‐NiNRG during catalytic graphitization, enabling a fully interconnected, well‐graphitized carbon network with a predefined 3D microstructure. A direct comparison between two carbon tube grids with identical structures and thicknesses but different crystallinities unambiguously reveals the crystallinity‐enabled enhancement in frequency response. This synergistic ion‐electron transport allows the 3D‐GCTG to function as a load‐tolerant electrode, effectively decoupling areal capacitance from phase angle. In a two‐electrode configuration, the 3D‐GCTG maintains a phase angle below −80° at 120 Hz even at 40 µm, delivering a high areal capacitance of 3.77 mF cm −2 , a 3.6‐fold improvement over the previously reported non‐graphitized counterpart. This work establishes graphitization‐enabled transport engineering as a general strategy for overcoming the long‐standing capacitance‐response trade‐off, offering a versatile platform for high‐performance AC‐filtering EDLCs.

Galvanic Replacement Synthesis Enabled by Gallium‐Based Liquid Metal: A Powerful Route for Material Design and Versatile Applications

ABSTRACT The precise engineering of core–shell or hollow metallic nanoarchitectures has become a central pursuit in fundamental research and advanced technologies. Among the available synthetic methods, galvanic replacement reaction (GRR) holds immense potential for engineering the composition/structure tunable nanomaterials, yet conventional solid templates impose fixed geometry, and limited reactivity, restricting both versatility and functional tunability. The emergence of liquid metal (LM) with highly reactive, fluidic, dynamic, and reconfigurable attributes offers a transformative alternative to conventional GRR. By establishing a liquid‐liquid interface, LM enables the deposition of metals previously inaccessible with solid templates, while simultaneously creating a significantly milder synthetic paradigm. This review is dedicated to extracting the unique features of LM‐enabled GRR (LM‐GRR) to highlight a powerful and facile materials‐synthesis route for diverse application scenarios. The fundamentals and mechanisms of the LM‐GRR will be first introduced, followed by a detailed explanation of the uniqueness and overall programmability in terms of composition, structure, processable template, and interfacial characteristics. Representative reaction systems are systematically surveyed to illustrate their specificity and chemical versatility. Finally, we discuss functional metal materials obtained via LM‐GRR for use in high‐performance catalysis, flexible electronics, biomedicine, biosensing, and electromagnetic shielding areas, and conclude with perspectives on future directions.

Improving seismic fault detection through fault-balanced patch extraction and deep learning networks (UNet, Efficient-UNet, and VGG19-UNet)

Structural basis for inhibition of the voltage-gated sodium channel NaV1.7 by the tarantula toxin HWTX-I

Addressing health-related social needs for cancer patients: An implementation assessment.

1575 Background: Health-related social needs (HRSNs) are associated with delayed diagnosis and treatment, worsened quality of life, and higher recurrence and mortality rates for cancer patients. Health systems are beginning to recognize the tremendous impact of HRSNs by recommending standardized HRSNs screening and referral in cancer care. Further research is needed to effectively design and implement HRSNs interventions for cancer patients. We evaluated the implementation of a novel community health worker (CHW) referral and navigation intervention to address unmet HRSNs in a National Cancer Institute (NCI)-designated comprehensive cancer center serving a highly disadvantaged community of patients in Bronx County, NY. Methods: We conducted a retrospective study of cancer patients utilizing data from HRSNs screenings, clinician CHW referral orders, and CHW navigation services between October 2023 and March 2025. Patients were identified as eligible for CHW assistance if they were screened for and self-reported at least one unmet HRSN. We organized process and outcome measures using the RE-AIM (Reach, Effectiveness, Adoption, Implementation, and Maintenance) implementation framework and evaluated measures through descriptive statistics. Results: There were 30 cancer center clinical practices who adopted the HRSNs screening tool. A total of 8,591 patients were screened for unmet HRSNs, with 20.5% self-reporting at least one unmet HRSN. Of those with at least one unmet HRSN (n=1,758), 43.4% were referred to CHWs by clinicians. Referral rates to CHWs ranged from 26-55% by clinical practice. Of patients referred to CHWs, 794 (64.5%) were assisted with HRSNs navigation by CHWs, with the most common HRSNs identified as financial (50.3%), housing (48.7%), and food insecurities (45.5%). CHWs were successful in outreaching referred patients (95.5%) in a timely manner (Median: 3 days from referral order, IQR: 1-6 days), connecting patients to at least one social service (98.9%), and resolving or improving at least one social need (97.6%). The median annual cost to implement and maintain the CHW referral and navigation intervention to address unmet HRSNs was $345.58 per patient. Conclusions: Implementation of CHW referral and navigation services within an NCI-designated comprehensive cancer center was feasible and effective in addressing unmet HRSNs within a safety-net cancer population. CHWs were successful in reaching patients and connecting them with essential social services that helped to resolve their HRSNs. Further research is needed to improve clinician screening and referral and understand the impact of CHW navigation on oncologic care delivery and clinical outcomes.

Palliative care utilization in metastatic lung cancer at an urban safety-net hospital.

e20651 Background: Early integration of palliative care has been associated with improved survival, quality of life, and reduced healthcare costs among patients with metastatic lung cancer. ASCO and NCCN guidelines recommend early referral to interdisciplinary palliative care for patients with advanced lung cancer. Despite these recommendations, rates of early palliative care referrals remain low. Methods: We conducted a retrospective cohort analysis of patients diagnosed with metastatic lung cancer (SCLC and NSCLC) between January 1, 2021 and September 1, 2025 at an urban safety-net oncology clinic. Patients with non-small cell neuroendocrine tumors and those evaluated for a single clinic visit were excluded. Clinical and referral data, including diagnosis date, referral dates, outpatient encounters, and inpatient palliative care consultations, were obtained from the electronic medical record. Results: A total of 162 patients met inclusion criteria. Fewer than half of patients with advanced lung cancer (n=67, 41.3%) were referred to outpatient palliative care at any point during their disease course. Among those referred, 36 (53.7%) completed at least one outpatient palliative care visit. The mean time from stage IV diagnosis to outpatient palliative care referral was 136 days, with an additional mean interval of 59 days between referral order placement and first outpatient palliative care visit. Patients treated with oral targeted therapies (n=10) demonstrated a significantly longer time to palliative care referral compared with those receiving other treatment modalities (n=57) (mean 351 vs 99.2 days, mean difference 251.8, 95% CI 120.2 to 382.8; p = 0.0003). Inpatient palliative care utilization was common, with 91 of 162 patients (55%) receiving inpatient consultation; however, continuity after discharge was limited, with only 23% of these patients subsequently seen in outpatient palliative care clinic. Conclusions: Despite demonstrated survival benefits and guideline recommendations supporting early palliative care integration for patients with metastatic lung cancer, outpatient palliative care utilization remains low in this urban safety-net setting. Significant delays were observed throughout the referral pathway, including prolonged intervals from diagnosis to referral, delays between referral placement and initial outpatient visits, low rates of completed referrals, and missed opportunities to transition patients from inpatient to outpatient palliative care. Future implementation efforts should prioritize systematic, electronic medical record-based interventions such as automated referral triggers, standardized order sets, and care transition workflows to facilitate timely and equitable integration of outpatient palliative care to improve patient-centered outcomes. The impact of palliative care in patients receiving oral targeted therapy also requires further study.

Real-world clinicopathologic characteristics and outcomes of MSI-high colon cancer in community oncology practice from western India.

e15585 Background: Microsatellite instability–high (MSI-H) colon cancer represents a distinct molecular subset with prognostic and therapeutic implications particularly in the era of immunotherapy. However, real-world data from community oncology settings in India remains limited. Methods: This retrospective study evaluated MSI-H colon cancer patients from March 2018 to May 2025 at Western India community oncology centers. PFS and OS were analyzed using Kaplan-Meier. Results: A total of 1100 colon cancer patients were evaluated out of which 358 underwent MSI testing & 66/358 (18.43%) were MSI-H. The median age of the MSI-H cohort was 66 years (IQR, 43.5–72), with M:F ratio of 1.7:1; most patients had an ECOG PS of 0–1 (87.8%)& stage distribution was: Stage I -3.0%, Stage II- 40.9%, Stage III -48.5%,Stage IV- 7.6%. MSI testing was performed at baseline in 80.3% & at progression in 19.7%, predominantly by IHC (86.3%), followed by NGS (10.6%) and PCR (3%). Loss of expression was most frequently observed for hMLH-1 (47%) followed by hPMS-2 (45.5%), hMSH-2 (18.2%) and hMSH-6 (16.7%). In 31.8% (n = 21), expression status was unknown, with variable instability detected across microsatellite markers NR-21 (n = 2), NR-24 (n = 2), BAT-25 (n = 2), BAT-26 (n = 2), and MONO-27 (n = 1). MSI-H incidence was most common in adenocarcinoma (95%) most commonly involved the ascending (n = 34) and descending colon (n = 11) with a predominance of right-sided over left-sided disease (54.5% vs 42.4%).PD-L1 status was available in 7.6% (n = 5), with positivity in 4/5 and TMB-high ( > 10muts) in n = 1. Common co-mutation included KRAS (n = 2). First line therapy was mainly surgery (n = 61), chemotherapy (n = 35) and targeted therapy (n = 4). Immunotherapy was administered in combination or as single agent to 9.3% (n = 6) with distribution: Pembrolizumab, nivolumab and dostarlimab in n = 2 each with 60% ORR. Among first line recipients, best responses included complete response in 39.1%, partial response in 14.1%, and stable disease in 10.9%. For MSI tested patients, the median follow-up was 19.9 months (95% CI 17.4-22.4) and Median OS (mOS) was not reached in either the MSI-stable or MSI-H groups with 23.9%(n = 70) and 6.1% (n = 4)events reported respectively (NA; 95% CI, NA–NA) (p = 0.001). Median PFS was 25.9 months (95% CI, 15.96–35.8) in MSI-stable cohort vs 71.9 months (95% CI, 0.0–168.02) in the MSI-H cohort (p = 0.006). Conclusions: This real-world analysis from Western India identifies MSI-H colon cancer as a biologically distinct entity with right-sided predominance and superior survival outcomes; however, underutilization of MSI testing and immunotherapy underscores the need for systematic MSI assessment to facilitate timely, biomarker-driven treatment in community oncology settings.

Physician preferences for metastatic hormone-sensitive prostate cancer treatment: A discrete choice experiment in China.

e17086 Background: Androgen receptor pathway inhibitors (ARPIs) improve disease control and overall survival (OS) in metastatic hormone-sensitive prostate cancer (mHSPC). Unclear benefit-risk trade-offs may yield heterogeneous treatment selection and uneven clinical outcomes. This study explored Chinese physicians’ preferences and benefit-risk trade-offs for mHSPC therapies. Methods: Two-phase Discrete Choice Experiment (DCE) was conducted. Phase 1 included a systematic literature review (SLR) and qualitative interviews with 20 urologists to identify key DCE attributes and levels. Phase 2, online survey of 100 urologists in tertiary hospitals, each completed 36 choice sets comparing two hypothetical mHSPC treatment profiles defined by six attributes (2-3 levels). Random parameters logit model estimated preference weights, relative attributes importance (RAI) assessed attribute influence, marginal rates of substitution quantified efficacy-risk trade-offs. Results: A total of 100 urologists completed the survey, median years in practice were 15. Consistent with prior interviews and the systematic review, efficacy (PSA-related) and fall risk were key decision drivers. Physicians favored higher efficacy and lower risks. Greater OS gains (median: 5 years + 12 months vs. 5 years + 6 months; estimate 0.61, 95% CI: 0.44-0.77) and higher PSA response rates (50% vs. 30% ≤0.2 ng/mL at 3 months; estimate 0.56, 95% CI: 0.31-0.81) were preferred, while serious falls were least favored (≥5% vs none; estimate -0.83, 95% CI: -1.06 to -0.59). All attribute levels except rash differed significantly from baseline ( p < 0.05). Attributes ranking: efficacy attributes accounted for over 50% of total influence, with depth of PSA decline (PSA 0.02, PSA 0.2, PSA 90, RAI 22%), OS prolongation (RAI 20%), and PSA decline rate (RAI 18%). Among risks, serious-fall likelihood was the most influential (RAI 27%), followed by fatigue/asthenia (RAI 7%) and rash (RAI 6%). Trade-offs: to lower serious-fall risk, physicians would forgo 8.2 months of OS (95% CI: 5.2-11.1), 2.2 months for fatigue/asthenia (95% CI: 0.6-3.9) and 1.8 months for rash (95% CI: -0.2-3.9). Conclusions: Efficacy - particularly PSA-related indicators and OS- dominates mHSPC treatment choice. Serious-fall risk emerged as the most critical safety concern, physicians were willing to forgo meaningful OS to mitigate it. These findings highlight the importance of balancing efficacy and quality-of-life when selecting ARPI therapies, expected to inform clinical treatment selection in China.

A randomized phase III trial of dexamethasone mouthwash to prevent chemotherapy-induced oral mucositis in patients with breast cancer.

12017 Background: Oral mucositis is a common adverse event in patients undergoing chemotherapy for breast cancer. A previous randomized phase II trial showed that a dexamethasone-based mouthwash significantly reduced the incidence and severity of chemotherapy-induced oral mucositis. The SMASH-BC trial was designed to confirm the efficacy and safety of dexamethasone-based mouthwash for preventing chemotherapy-induced oral mucositis in patients with breast cancer. Methods: This multicenter, open-label, randomized phase III trial was conducted in Japan. Patients with early breast cancer scheduled to receive standard adjuvant or neoadjuvant chemotherapy (AC, EC, dose-dense AC/EC, TC, or pembrolizumab-containing regimens) were randomized 1:1 to receive either a dexamethasone-based mouthwash (10 mL of dexamethasone elixir 0.01%) or tap-water mouthwash as a control. The mouthwash was swished for 2 min and spit, four times daily for 8 weeks, starting on day 1 of chemotherapy. Randomization was stratified by chemotherapy regimen, smoking status, and institution. The primary endpoint was the incidence of all-grade oral mucositis at 8 weeks, assessed using electronic patient-reported outcomes (ePRO). Assuming an all-grade oral mucositis incidence of 55% in the control group and 35% in the intervention group, 212 patients were required to provide 80% power with a two-sided α of 0.05. The planned sample size was 230, allowing for 10% attrition. Results: Overall, 230 patients were randomized (intervention, n = 117; control, n = 113), and 220 patients (intervention, n = 114; control, n = 106) were included in the analysis. The incidence of ePRO-assessed all-grade oral mucositis at 8 weeks was 58.8% and 63.2% in the intervention and control groups, respectively (risk ratio [RR], 0.93; 95% confidence interval [CI], 0.75–1.15; P = 0.59). The incidence of ePRO-assessed moderate-to-very severe oral mucositis was 13.2% and 18.9% in the intervention and control groups, respectively (RR, 0.70; 95% CI, 0.38–1.29). During the 56-day study period, mucositis reported by ePRO on a median of 2 and 6 days in the intervention and control groups, respectively (median difference −4 days; 95% CI, −5.3 to 8.3 days). The incidence of physician-assessed oral mucositis was 31.9% vs. 41.5% in the intervention and control groups, respectively (RR, 0.77; 95% CI, 0.54–1.09). The incidence of grade ≥ 3 adverse events was 6.2% and 6.6% in the intervention and control groups, respectively (RR, 0.94; 95% CI, 0.34–2.58). However, the incidence of oral candidiasis was higher in the intervention group (5.4% vs 0%). Conclusions: Prophylactic dexamethasone mouthwash did not significantly reduce all-grade oral mucositis at 8 weeks. Overall safety was comparable between groups; however, the incidence of oral candidiasis was higher in the intervention group. Clinical trial information: jRCTs071240036.

Knowledge characterization of newly diagnosed breast cancer patients.

e12738 Background: Patient decision making is understudied and seldom focused on those diagnosed with cancer. This NCI-supported knowledge acquisition study sought to characterize patient knowledge of those newly diagnosed with breast cancer and examine how it informs decision making. Methods: We recruited participants in the US, > 18 years, within three months of a breast cancer diagnosis to participate in 1-hour semi-structured interviews on Zoom with compensation. An interview guide was developed to elicit patient knowledge state, as well as acquisition and sources regarding diagnosis and next steps. Interviews were audio-recorded, transcribed, and analyzed using Braun and Clarke’s (2006) thematic analysis of qualitative data. Results: We interviewed 26 female participants with a mean age of 51. Approximately 54% (n = 14), 31% (n = 8), and 4% (n = 1) were White, Black, and Asian, respectively and 4% (n = 1) were Hispanic. Below, we highlight a selection of themes. Diagnosis and next steps: Patients reported inconsistent experiences of receiving their diagnosis. Eighteen patients received a report via MyChart with follow-up communication from a provider (n = 5) on a timeline of hours to days, or not at all. Only some patients reported receipt of an explanation from their provider. Patients often sought clarification through external resources (e.g. Google, Reddit, TikTok, Instagram, Facebook patient communities, survivors in their networks). Four participants used AI to help understand their pathology reports, medical terminology, treatment options, and next steps. Roles of clinical care team: Some patients demonstrated incomplete understanding of the structure of their care team. Half (n = 14) used the term “doctor” to refer to either the primary care provider or oncologist (without clarity to what role). Only a few patients mentioned having a patient navigator and, when prompted, some patients reported that they either did not have one or that they did not know. Second opinion: Some patients did not talk about seeking a second opinion (suggesting they did not), while others reported feeling empowered to do so (n = 9), influenced by their social network and embodied knowledge, yet another reported discomfort associated with doing so. Conclusions: During the diagnostic phase, patients report lack of support in acquiring credible foundational knowledge. Findings suggest that patient-driven knowledge acquisition is an active and iterative process that begins at diagnosis, not at treatment selection. Specifically, patients continuously assess what they know, do not know, and how to fill the gaps. Highlighted by the variability in reported patient experiences in both receiving and seeking information, many report inconsistent support. Implications include that without effective information transmission, patients may not have decision making agency.

Concurrent RNA- and DNA-NGS testing for the detection of clinically-relevant fusions in pediatric solid-tumor patients.

10035 Background: Identifying structural rearrangements and gene fusions is critical to providing high quality, precision-driven clinical care for many types of pediatric cancers. RNA next-generation sequencing (NGS) provides a functional readout of the genome that enables superior detection of chimeric transcripts and novel driver fusions, particularly when genomic breakpoints reside in complex intronic regions. This study demonstrates the benefit of concurrent DNA- and RNA-NGS for fusion detection in a real world pediatric cohort of 1,050 patients, one of the largest such studies to date. Methods: We used the Tempus de-identified multimodal database to select a cohort of pediatric solid tumor cancer patients (all stages) who received successful DNA (Tempus xT, 648 gene panel with enhanced detection of structural variants [SVs] in 22 genes) and RNA (Tempus xR, whole-transcriptome) NGS sequencing. All patients had a minimum tumor purity of 20% and were aged 0-21 at the time of sample collection (n=1,050). All assessed fusions appeared on clinical reports. Results: Overall, we detected a fusion in 35.1% of patients (n=369). The top 3 cancer types in our cohort were soft tissue sarcoma (n=321), brain/CNS cancer (n=280), and bone cancer (n=120), and fusion prevalence in these types was 53% (n=170), 30.4% (n=82) and 30% (n=36), respectively. The fusion types with the highest overall prevalence were: EWSR1 (9.0%, n=95), BRAF (6.1%, n=64), PAX3-FOXO1 (2.7%, n=28), ALK (2.2%, n=23), and RET (2.0%, n=21) fusions. In assessing the subset of genes that appear on both the DNA-NGS and RNA-NGS panel, 303 patients harbored one of these fusions and 38.6% (117/303) of those fusions were detected only via RNA-NGS. Among fusions with highest prevalence, the percentage detected only via RNA-NGS ranged from 3.1% (EWSR1 fusions) to 100% (PAX3-FOXO1); BRAF (53/64, 82.8%) and ALK (4/23, 17.4%) fusions both had comparatively high proportions detected only via RNA-NGS. RNA-NGS alone detected gene fusions in 65 additional patients where neither partner appears on the DNA-NGS panel. Considering only fusions associated with an indication-matched FDA-approved targeted therapy, we observed a prevalence of 9.0% (94/1050), and of these, 46.8% (44/94) were detected only by RNA-NGS. Overall, 49.3% (182/369) of fusion-positive patients would have been missed if RNA-NGS were not performed, representing 17.3% of the total cohort. Conclusions: Pediatric solid tumors are frequently driven by structural rearrangements and gene fusions that are difficult to characterize using DNA sequencing alone. This study demonstrates that performing combined DNA-NGS and RNA-NGS substantially improves the identification of patients with a clinically relevant fusion in a large real-world data set.

Hair relaxer use and breast cancer risk by tumor molecular subtypes.

10596 Background: Women of West African ancestry, including US populations, are disproportionately diagnosed with aggressive breast cancers including triple negative tumors. Previously, hair relaxer use has been associated with increased breast cancer risk in the Ghana Breast Health study (GBHS), with former users at an over 2-fold increased risk compared to never users. We aimed to determine if relaxer use risk associations differed by tumor molecular subtypes suggesting possible biological mechanisms of carcinogenesis. Methods: The GBHS is a population-based case-control study conducted 2013–2015 in Accra and Kumasi, with 1,071 pathologically confirmed invasive breast cancer cases (87% had available estrogen receptor (ER) status) and 2,106 matched controls, aged 18-74 years. Molecular subtypes were defined using St Gallen’s criteria using molecular marker data for ER, progesterone receptor, human epidermal growth factor receptor-2 (HER2), and grade (G). Molecular tumor subtypes defined as: Lum A (ER/PR+, HER2- and G1/G2, N = 260), Lum B- (ER/PR+, HER2- and G3, N = 468), Lum B+ (ER/PR+, HER2+ and G1/G2, N = 195), HER2-Enriched (ER-, PR- and HER2+, N = 59), Triple-Negative (ER, PR, and HER2-, N = 129). We imputed missing data using chained equations. Polytomous logistic regression models were used to estimate odds ratios (ORs) with 95% confidence intervals (CIs) for associations of relaxer use with breast cancer subtypes. Models were adjusted for matching factors (age, hospital site), and potential confounders: education, parity, age at first birth, and median breastfeeding months per child. Results were combined across imputed datasets. Heterogeneity (p-het) in breast cancer risk by tumor molecular subtypes using polytomous logistic regression case-only models. Results: Relaxer use was common (ever-use of relaxers 96% of cases and 94% of controls). Compared to never users, increased risk for former users compared to never users was observed across subtypes. While magnitudes of risk differed (OR LumA = 2.64, 95% CI [0.91—6.60], OR LumB+ = 2.69, 95% CI [0.47—15.47], OR LumB- = 1.74, 95% CI [0.44—6.91], OR HER2 = 3.66, 95% CI [0.75—17.98], OR Triple-Negative = 1.21, 95% CI [0.37—4.00]), there was no statistical evidence of etiologic heterogeneity by subtype (p-het = 0.59). Similarly, other measures of relaxer use (duration, age at first use, lye or non-lye) did not show any evidence of heterogeneity. Conclusions: In this population with a high proportion of aggressive tumor subtypes, breast cancer risk associated with relaxer use did not support etiologic heterogeneity by molecular subtype, however, power was limited. These findings suggest that relaxer exposure is unlikely to explain subtype-specific breast cancer risk patterns. Future investigations should pool studies for improved power, evaluate other molecular tumor characteristics and relaxer products’ formulations to clarify carcinogenic mechanisms to reduce risk.

p53 R248Q upregulation of VTCN1 via dual mechanisms as a driver of immune evasion and adaptive resistance to CDK4/6 inhibitors: Rationale for combination immunotherapy in endometrial cancer.

5618 Background: TP53-mutated endometrial cancer (EC) is a heterogeneous disease with variable responses to immunotherapy. While generic p53 mutation status is a standard biomarker, the distinct clinical impact of specific hotspot mutants remains underappreciated. Our bioinformatics analysis of TCGA data suggested that the R248Q hotspot—but not others—is specifically linked to an "immune-desert" phenotype. We aimed to mechanistically define this subtype and validate its diagnostic value as a predictor for immune exclusion and CDK4/6 inhibitor sensitivity. Methods: We performed a comprehensive bioinformatic analysis of the TCGA-UCEC cohort to correlate specific p53 hotspot mutations with immune landscapes. To elucidate the underlying mechanism, we engineered a panel of isogenic EC cell lines (p53-null, R248Q, R175H, R248W, R273H/C) and assessed VTCN1 (B7-H4) regulation via biochemical assays. To validate these findings clinically, we established an independent cohort of 254 EC patients with tissue microarrays (TMA); whole-exome sequencing (WES) is currently underway to pinpoint hotspot mutations and correlate them with VTCN1 expression and immune infiltration. Results: Discovery of Diagnostic Target: In the TCGA cohort, stratifying p53 mutants revealed that R248Q carriers exhibited the most profound "cold" tumor microenvironment compared to other hotspots, suggesting it as a potential negative predictive biomarker. Mechanistic Validation: Using our isogenic models, we observed that R248Q specifically drives high VTCN1 expression. Mechanistically, R248Q appears to promote VTCN1 via a dual pathway: transcriptional upregulation and, uniquely, protein stabilization by inhibiting proteasomal degradation. This indicates VTCN1 as a potential surrogate marker for the R248Q subtype. Therapeutic Implications: In preclinical models, abemaciclib treatment paradoxically upregulated VTCN1 in R248Q cells, suggesting adaptive resistance. However, this induced vulnerability was associated with enhanced sensitivity to combined CDK4/6 inhibition and immunotherapy (anti-VTCN1), which elicited superior tumor regression compared to monotherapy. Conclusions: Our study suggests the potential value of refining p53 diagnostics from binary (mutant/WT) to hotspot-specific resolution. Current data indicate that p53 R248Q may function as a distinct driver that contributes to an immunosuppressive "cold" TME in EC, likely through synergistic transcriptional and post-translational upregulation of VTCN1. Furthermore, abemaciclib-induced VTCN1 elevation points to a possible mechanism of monotherapy resistance. These findings provide a strong preclinical rationale to explore combining CDK4/6 inhibitors with immunotherapy, specifically for patients with the high-risk p53 R248Q subtype.

Comparative mortality following intracranial hemorrhage in patients with brain metastases from melanoma vs renal cell carcinoma: A propensity score–matched analysis.

e14027 Background: Melanoma and renal cell carcinoma (RCC) frequently metastasize to the brain and carry high hemorrhagic risk, yet comparative outcomes following intracranial hemorrhage (ICH) remain poorly characterized. Methods: Using TriNetX (111 healthcare organizations), we identified adults with nontraumatic ICH (2016-2025) and brain metastases within one year prior. Patients were stratified by primary malignancy (melanoma vs RCC) after excluding concurrent diagnoses, vascular malformations, and coagulopathies. Propensity score matching (1:1) balanced cohorts on demographics, comorbidities, and laboratory values. Primary analysis examined 1-day to 1-year outcomes; landmark analyses at 90 and 14 days assessed survivors. Cox and logistic regression evaluated outcomes. Results: After matching, 561 patients per cohort were analyzed. Melanoma patients had significantly higher one-year mortality (52.9% vs 37.9%; HR 1.69, 95% CI 1.42-2.02, p<0.001; number needed to harm 7, 95% CI 5-11). This difference persisted among 90-day survivors (HR 1.60, 95% CI 1.17-2.20, p=0.003) and in the 90-day subgroup (HR 1.75, 95% CI 1.41-2.16, p<0.001), with consistent findings at the 14-day landmark. No significant differences were observed in thromboembolic events, neurosurgical interventions, or other secondary outcomes; however, statistical power was limited for these endpoints. Conclusions: Patients with ICH and brain metastases from melanoma experience significantly higher mortality compared to RCC, persisting beyond the acute period, suggesting tumor biology rather than ICH severity drives this disparity. Cox proportional hazards regression analysis (melanoma vs RCC). Outcome Primary Index–1yr Landmark 90d–1yr Subgroup Index–90d Landmark 14d–90d Mortality 1.69 (1.42–2.02)* 1.60 (1.17–2.20)* 1.75 (1.41–2.16)* 1.67 (1.30–2.15)* ED Visits 0.96 (0.60–1.53) 1.00 (0.47–2.12) 0.94 (0.52–1.71) 0.63 (0.31–1.30) Inpatient Visits 1.25 (0.63–2.48) NR NR NR Hospice Enrollment 1.10 (0.56–2.16) NR 1.34 (0.63–2.85) NR Ischemic Stroke 1.26 (0.69–2.29) NR 1.05 (0.54–2.06) 1.19 (0.51–2.80) Acute VTE 1.14 (0.71–1.83) NR 1.45 (0.82–2.56) 1.61 (0.83–3.13) STEMI/NSTEMI 1.01 (0.51–2.03) NR 1.20 (0.53–2.71) NR Blood Transfusions 0.83 (0.47–1.46) NR 0.94 (0.47–1.86) NR Craniotomy/EVD/Shunt 1.09 (0.57–2.08) NR 1.15 (0.60–2.21) NR Mechanical Ventilation 1.29 (0.77–2.15) NR 1.18 (0.67–2.09) 1.08 (0.45–2.59) Values represent Hazard Ratio (95% CI). Reference group: RCC. *p<0.05. NR=not reported due to insufficient events (<10 in one group). Proportional hazards assumption satisfied for all mortality analyses (Schoenfeld test p>0.35). Secondary outcomes were underpowered; null findings should be interpreted cautiously.

Engineering and evaluating intelligent information retrieval systems for hepatocellular carcinoma clinical question answering.

e16009 Background: Large language models (LLMs) help navigate complex clinical guidelines but often suffer from hallucinations. Retrieval-augmented generation (RAG) systems aim to mitigate this, yet systematic evaluations in medical contexts are rare. This study benchmarks four AI architectures against the AASLD hepatocellular carcinoma guidelines to address this gap. Methods: Four AI architectures were evaluated: a baseline LLM; a custom RAG system; a custom multimodal RAG system integrating figures; and a custom agentic multimodal RAG system with autonomous multi-step retrieval. All systems used GPT-5.1. Two test sets were used: 39 guideline-derived question–answer pairs spanning epidemiology, diagnostics, and therapeutics, and 10 complex clinical vignettes requiring integration of multiple guideline components. Each query was run five times and scored using an LLM-as-a-judge three-point scale: 0.0 (incorrect), 0.5 (partially correct), and 1.0 (correct). Mann–Whitney U tests compared systems with the baseline LLM. Results: For guideline questions, performance increased with architectural complexity. Agentic multimodal RAG achieved the highest mean score (0.81±0.31, SD), with 28/39 questions (71.8%) scoring 1.0 and 3/39 (7.7%) scoring 0.0. Multimodal RAG scored 0.74±0.36, traditional RAG 0.62±0.41, and the base LLM 0.51±0.40. All RAG systems significantly outperformed the base LLM (p < 0.001 for multimodal and agentic; p = 0.004 for traditional). Agentic multimodal RAG showed the lowest variability across guideline questions, indicating more consistent performance than other architectures. In contrast, for clinical vignettes (n = 10), the base LLM performed best (0.90±0.24), with 8/10 questions scoring 1.0. All RAG systems significantly underperformed compared to base LLM (p≤0.005): traditional RAG scored 0.63±0.46 (6/10 perfect), agentic multimodal RAG scored 0.73±0.35 (5/10 perfect), and multimodal RAG scored only 0.42±0.44, with 5/10 questions (50%) scoring 0.0. These findings suggest complex clinical reasoning relies more on parametric knowledge and general reasoning rather than document retrieval. Conclusions: These findings reveal a critical dichotomy in medical AI performance. While agentic and multimodal RAG architectures excel at factual extraction and visual data interpretation for specific guideline queries, they falter in complex clinical vignettes, where the base model’s holistic reasoning proves superior. This suggests that retrieval mechanisms can inadvertently fragment context or introduce noise during synthesis tasks. Comparative performance metrics across four AI system architectures. System Guideline Questions (n=39) Mean (SD) Vignettes (n=10) Mean (SD) Base LLM 0.51 (0.40) 0.90 (0.24) RAG Pipeline 0.62 (0.41) 0.63 (0.46) Multimodal RAG 0.74 (0.36) 0.42 (0.44) Agentic Multimodal RAG 0.81 (0.31) 0.73 (0.35)

Prevalence of high Claudin 18 expression in gallbladder cancer and association with survival: Implications for CLDN18.2-targeted therapy.

e16299 Prevalence of high Claudin 18 expression in gallbladder cancer and association with survival: implications for CLDN18.2-targeted therapy Background: Claudin 18.2, an isoform of Claudin 18 (CLDN18) is a therapeutically actionable tight-junction target in upper gastrointestinal malignancies, with biomarker thresholds driving patient selection. Data in gallbladder cancer (GBC) are scarce. We evaluated the prevalence of high CLDN18 expression assessed by immunohistochemistry (IHC) on tissue microarrays (TMA) and explored its association with survival outcomes in GBC. Methods: We conducted a retrospective cohort analysis of patients with GBC with available clinicopathologic and survival data. CLDN18 expression was assessed by IHC on TMA from GBC specimens. High CLDN18 expression (CLDN18-high) was defined as membranous staining ≥2+ in ≥75% of tumor cells. Overall survival (OS) was estimated using Kaplan–Meier methods and compared by log-rank testing for (i) all-cause mortality and (ii) GBC-specific cause of death. Cox proportional hazards models evaluated the independent association of stage group and CLDN18-high status with OS. Results: Among patients with available CLDN18 assessment (N = 122), CLDN18-high was observed in 9.0% (11/122). Baseline clinicopathologic characteristics were broadly similar between CLDN18-high and non-high tumors (all p > 0.05). In the survival cohort (N = 159), median OS was 11.0 months for both all-cause and GBC-specific mortality analyses. In CLDN18-evaluable cases, CLDN18-high tumors showed numerically longer median OS compared with non-high, without statistical significance: all-cause OS 16.0 vs 9.0 months (p = 0.24) and GBC-specific OS 21.0 vs 10.0 months (p = 0.21). In multivariable analysis, stage III/IV was independently associated with worse OS (HR 3.06, 95% CI 1.41–6.65; p = 0.005), whereas CLDN18-high was not (HR 0.48, 95% CI 0.07–3.56; p = 0.476). Conclusions: Using IHC on TMA, high CLDN18 expression was detected in 9% of GBC tumors, identifying a biomarker-defined subgroup that could be considered for CLDN18-targeted approaches. CLDN18-high status was not independently prognostic after adjustment for stage, but the prevalence of high expression supports provides a rationale for biomarker-selected clinical trials of CLDN18.2-directed therapies in GBC.