Browse Articles

Comparison of seprafilm and pitavastatin treatments in an experimental model of peritoneal adhesion

The HypA and HypB metallochaperones from Methanococcus maripaludis have unique metal-binding properties and a distinct nickel transfer mechanism

Prognostic value of perioperative ctDNA monitoring in patients with esophageal squamous cell carcinoma (ESCC) undergoing neoadjuvant chemotherapy (NAC) and surgery.

4094 Background: Recent results from a multicenter phase II clinical trial (jRCTs031200094) demonstrated that neoadjuvant FLOT (fluorouracil, leucovorin, oxaliplatin, and docetaxel) improves survival outcomes of patients with resectable ESCC. In this analysis, we evaluated the value of circulating tumor DNA (ctDNA) dynamics after FLOT and curative-intent surgery to predict treatment response and long term outcomes. Methods: The jRCTs031200094 clinical study enrolled 37 patients with ESCC to receive neoadjuvant FLOT followed by curative-intent surgery. Plasma samples were collected NAC, post-NAC (prior to surgery), and within 2–12 weeks post surgery [molecular residual disease (MRD) window]. A clinically validated, personalized, tumor-informed mPCR-NGS assay (Signatera, Natera, Inc.) was used on the banked specimens for ctDNA detection. Survival analyses were performed to calculate progression-free survival (PFS) stratified by ctDNA status at the post-NAC and post surgery timepoints. Results: This retrospective analysis included the 34 patients with ESCC and available ctDNA results. Patients had a median age of 66 years (range: 44–80), and the majority were male (79% [27/34]). Tumor locations included the middle (53%, 18/34), lower (41%, 14/34), and upper esophagus (5.9%, 2/34). The clinical stage distribution was 8.8% (3/34) stage I, 21% (7/34) stage II, 56% (19/34) stage III, and 15% (5/34) stage IV. R0 resection was achieved in 94% (29/31) of resected patients; 3 patients did not undergo esophagectomy due to disease progression or patient preference. Baseline ctDNA positivity was observed in 97.1% (33/34) of patients. Among 33 patients with ctDNA data available at the post-NAC timepoint, ctDNA-positivity was associated with significantly inferior PFS (HR: 3.17, 95% CI: 1.22-8.21; P=0.018), higher rate of lymph node positivity and higher tumor regression grade (TRG). Of the 29 patients with ctDNA data in the MRD window 31.3% (9/29) were ctDNA-positive and ctDNA-positivity in the MRD window was associated with significantly inferior PFS (HR: 5.89; P=0.001) and overall survival (OS; HR: 7.23; P=0.019). In addition, ctDNA dynamics in response to NAC and surgery were highly predictive of long term outcomes. Sustained clearance after NAC correlated with 80% 24m PFS. Inferior survival was observed in patients who remained ctDNA positive after NAC but cleared after surgery (HR 3.17) and in patients without clearance after surgery (HR 7.63). Conclusions: This study demonstrates prognostic and predictive value of single timepoint and longitudinal ctDNA status in the neoadjuvant and postsurgical management of patients with ESCC. Persistent ctDNA positivity following NAC and in the MRD window was strongly associated with inferior PFS, and early ctDNA clearance conferred the most favorable outcomes. Clinical trial information: jRCTs031200094.

Serum albumin and toxicity burden in patients with solid tumors receiving tyrosine kinase inhibitors: A real-world cohort study from Saudi Arabia.

e23359 Background: Tyrosine kinase inhibitors (TKIs) are widely used across multiple malignancies; however, treatment-related toxicities remain a major cause of hospitalization, treatment modification, and reduced quality of life. Serum albumin is a routinely measured biomarker reflecting nutritional status, systemic inflammation, and disease burden, and may influence drug exposure for highly protein-bound TKIs. Real-world data examining the relationship between baseline albumin levels and clinically meaningful toxicity outcomes with TKIs remains limited. Methods: We conducted a retrospective cohort study of adult cancer patients treated with TKIs at King Fahad Medical City, a tertiary care hospital in Saudi Arabia. Patients were stratified by baseline serum albumin (< 3.5 vs. ≥3.5 g/dL). Treatment-related toxicities were identified through detailed medical record review and categorized chronologically as first, second, and third documented adverse events. Outcomes included toxicity incidence and severity, toxicity-related hospitalization or emergency department (ED) visits, treatment discontinuation, and time to first documented toxicity. Time-to-event analyses were performed using Kaplan–Meier methods with log-rank testing, and Cox proportional hazards regression was used for multivariable analyses. Results: A total of 369 patients were included; 121 (31.7%) had baseline hypoalbuminemia. Patients with hypoalbuminemia had more advanced disease, poorer performance status, and greater metastatic burden. The overall incidence of first documented toxicity was similar between groups (61.2% vs. 61.7%; p = 0.92). However, toxicity-related hospitalization or ED visits following the first toxicity were more frequent among patients with hypoalbuminemia (29.0% vs 17.7%; p = 0.014). Median time to first toxicity was shorter in the hypoalbuminemia group (29.4 months) compared with patients with normal albumin (not reached; log-rank p = 0.015). In univariate analysis, normal albumin was associated with a lower risk of toxicity (HR 0.67, 95% CI 0.49–0.93), though this association was attenuated after multivariable adjustment. No significant differences in overall survival were observed between groups. Conclusions: Baseline hypoalbuminemia was associated with earlier onset of TKI-related toxicity and increased toxicity-related hospitalization or ED visits, despite similar overall toxicity rates. Serum albumin may serve as a pragmatic clinical marker to identify patients at higher risk of early, clinically consequential toxicity in real-world TKI practice, supporting closer monitoring and proactive supportive care strategies at treatment initiation.

Clinical fidelity of large language models in chronic myeloid leukemia: A multimodel comparative study.

6585 Background: Large language models (LLMs) are emerging as promising tools for clinical decision-making. These models are increasingly investigated for their potential in diagnosis, prognosis, and personalized treatment of chronic myeloid leukemia (CML). This study evaluates the efficacy of five leading LLMs and examines the impact of incorporating up-to-date clinical guidelines into LLM prompts to enhance clinical decision-making. Methods: We developed fifty standardized clinical vignettes on CML, staging, tyrosine kinase inhibitor (TKI) selection, monitoring, adverse events, and treatment-free remission. Five LLMs (DeepSeek-R1, Gemini-2.5-Pro, Claude 4.0, GPT-OSS-120B, and LLAMA 4) generated responses under two distinct prompt conditions. In the baseline prompt, the clinical vignette was entered, and a response was obtained. In the guideline-augmented prompt, models were instructed to extract information from the NCCN v1.2026 and ESMO 2017 guidelines. A blinded expert panel evaluated the responses using a peer-reviewed answer key, assigning scores of 0 for incorrect, 1 for partial, and 2 for correct answers. The maximum possible score was 100. Results: The use of guideline-augmented prompting increased accuracy across all models, raising performance from 79.8% to 97.6% (mean difference: 17.8 percentage points, 95% CI: 3.9-31.7, p=0.024). Improvements for individual models ranged from 5 to 33 percentage points. Notably, only Gemini-2.5-Pro achieved perfect accuracy (100%). Knowledge drift was observed in four out of five models (80%), primarily due to reliance on outdated WHO classification criteria instead of current NCCN standards. This discrepancy resulted in staging errors in 32% of baseline cases. The diagnostic workup category exhibited the highest baseline accuracy (94%), whereas the treatment-free remission criteria showed the most substantial improvement following guideline integration, increasing from 72% to 96%. Conclusions: LLMs possess substantial medical knowledge that can facilitate clinical decision-making. However, these AI systems remain vulnerable to errors such as knowledge drift and hallucinations. This scalable framework demonstrates that mandatory compliance with guidelines, supported by structured prompting, is essential for safe LLM integration into oncology workflows. Institutions implementing LLM-based clinical decision support should require guideline integration rather than relying solely on pre-trained knowledge. Future studies should validate these findings in prospective clinical settings and evaluate mechanisms for automated guideline updating. Comparative accuracy of LLM responses under baseline vs. guideline-augmented prompts. LLM Baseline response accuracy Guideline augmented response accuracy Delta Claude 4.0 82 97 +15 DeepSeek-R1 72 97 +25 Gemini-2.5 Pro 89 100 +11 GPT-OSS-120B 90 95 +5 LLAMA-4 66 99 +33

Burnout among early-career oncologists in Latin America: A regional survey.

9013 Background: Burnout in healthcare professionals negatively impacts patient care, job satisfaction, and personal well-being. Early-career oncologists are especially vulnerable due to emotional strain and heavy clinical and administrative demands. These challenges are driven by structural constraints. This study aimed to determine the frequency of burnout among early-career oncologists in LATAM and to describe workload conditions and perceived institutional resources. Methods: A multicenter, observational, cross-sectional, voluntary, web-based survey was conducted between Nov/2025 and Jan/2026 among young oncologists in LATAM. Participants were physicians aged <40 years or <10 years of clinical practice. Participants were invited electronically via email and social media. The survey included demographic, professional, and workplace-related variables. An adapted version of the Maslach Burnout Inventory (MBI) was performed. Results: A total of 105 surveys were included from oncologists across 15 Latin American countries, with the highest representation from Colombia (37.1%), Mexico (23.8%), and Argentina (8.6%). Fifty-one percent were aged ≥35 years, were female (58.1%), and worked in non-academic settings, including public (36.2%), private (32.4%), or mixed public–private practice (29.5%). 19.1% were trainees. Regarding burnout, 55.2% reported feeling emotionally exhausted due to work at least once per week (median 4; IQR 3–5), and 64.8% reported end-of-day exhaustion at least weekly (median 4; IQR 3–5). Depersonalization symptoms were infrequent (median 1; IQR 0–3), and participants reported generally high levels of personal accomplishment (median 5; IQR 4–6). Overall, 46% of participants reported adequate technological, pharmacological, and institutional support resources. Only 21% indicated that their clinical and administrative workload was manageable; 40% considered it unmanageable. Most respondents reported misalignment between salary and workload, with 61% expressing disagreement. Conclusions: This survey provides one of the first regional overviews of working conditions among early-career oncologists in LATAM. A proportion of participants reported limited institutional resources and unmanageable clinical and administrative workloads, highlighting structural challenges within the region’s oncology workforce. Burnout survey among EOCs in LATAM. Questions Median, (IQR) I feel exhausted at the end of the workday. 4, (3-5) I feel emotionally exhausted by my work. 4, (3-5) I feel I work too hard at my job. 3, (2-5) I feel frustrated by my work. 3, (1-4) I have become more insensitive to people since I’ve been working. 1, (0-3) I’m afraid that this job is making me uncaring. 1, (0-3) I really don’t care about what happens to some of my patients. 0, (0,1) Through my work, I feel that I have a positive influence on people. 5, (4-6) I am easily able to create a relaxed atmosphere with my patients. 5, (4-5)

Distinct prognostic roles of <i>IDH</i> mutation and high-risk genomic alterations in patients with primary chondrosarcoma.

11550 Background: Prior studies report mixed associations between IDH mutation and survival in chondrosarcoma. We evaluated the prognostic impact of IDH mutation and high-risk genomic alterations in one of the largest multi-institutional cohorts while accounting for established prognostic factors. Methods: We retrospectively analyzed patients with histologically confirmed chondrosarcoma treated at the University of Miami and University of Florida. Patients with available molecular data were included. High-risk genomic alterations were defined as TP53, TERT, or cell-cycle pathway alterations. Endpoints included median overall survival (mOS), progression-free survival (mPFS) after treatment and distant metastasis–free survival (mMFS). Patients with metastatic disease at diagnosis were excluded from mMFS analyses. Multivariable Cox regression models were adjusted for age, sex, tumor grade and tumor type. Results: A total of 144 patients were included; 85 (59%) had NGS data. Median age was 60 years, and 46% had grade 3 disease. The majority of patients had conventional chondrosarcoma; 52 patients (36.1%) had dedifferentiated pathology. Among 64 IDH-mutant and 62 IDH-wildtype tumors, high-risk genomic alterations were identified in 48 patients, with comparable positive margin rates between IDH-mutant (14.1%) and IDH-wildtype (14.5%) cases. After multivariable adjustment for age, sex, tumor grade, and tumor type, IDH mutation was not independently associated with mOS (HR 1.26; p=0.48). However, IDH mutation demonstrated a trend toward inferior mPFS that did not reach statistical significance (HR 1.52; p=0.07) and was independently associated with inferior mMFS (HR 2.60; p=0.004). In contrast, high-risk genomic alterations were independently associated with inferior mOS (HR 2.45; p=0.025), mPFS (HR 2.94; p=0.003) and mMFS (HR 2.71; p=0.009) as compared to those patients without high-risk genomic alterations. Conclusions: After multivariable adjustment, IDH mutation conferred increased metastatic risk without an overall survival disadvantage, whereas high-risk genomic alterations consistently predicted poor outcomes across all survival endpoints. Multivariable Cox regression analyses by survival endpoint in chondrosarcoma. Endpoint Comparison Adjusted HR (95% CI) p-value Overall Survival IDH mutant vs IDH wildtype 1.26 (0.66–2.40) 0.48 High-risk NGS vs Low-risk NGS 2.45 (1.12–5.38) 0.025 Progression-Free Survival IDH mutant vs IDH wildtype 1.52 (0.95–2.43) 0.077 High-risk NGS vs Low-risk NGS 2.94 (1.45–5.96) 0.003 Metastasis-Free Survival IDH mutant vs IDH wildtype 2.60 (1.35–5.01) 0.0043 High-risk NGS vs Low-risk NGS 2.71 (1.28–5.74) 0.009

Impact of baseline performance status on survival and toxicity with ipilimumab plus nivolumab: A 10-year institutional experience across solid tumors.

e23340 Background: Dual immune checkpoint blockade with ipilimumab plus nivolumab (Ipi+Nivo) provides durable benefit across multiple solid tumors and is increasingly used in routine practice. However, patients with poor performance status (PS) are largely excluded from immunotherapy trials, limiting the generalizability of these results to this population. In a prior small institutional analysis, we observed significantly worse overall survival in patients with poor PS compared with good PS. Building on this work, we analyzed a decade-long institutional cohort to evaluate survival, response, toxicity, and hospitalization outcomes with Ipi+Nivo across advanced solid tumors, stratified by baseline PS. Methods: We conducted a retrospective cohort study of adults with advanced/metastatic solid tumors treated with Ipi+Nivo at an NCI-designated cancer center between January 2014 and January 2025. Patients were stratified by baseline ECOG PS into good (0–1) and poor (≥2) groups. The primary endpoint was overall survival (OS). Secondary endpoints included objective response rate (ORR), grade 3-4 immune-related adverse events (irAEs), and hospitalization. Survival was estimated using Kaplan-Meier methods and multivariable time-dependent Cox regression. Multivariable logistic regression was used for secondary outcomes. Results: Among 270 patients, 219 (81%) had ECOG PS 0–1 and 51 (19%) had ECOG PS ≥2. Median OS was significantly shorter in patients with poor PS compared with good PS (4.6 vs 21.2 months; p&lt;0.001). In multivariable time-dependent analyses, poor PS was associated with a markedly increased mortality risk at treatment initiation (Hazard Ratio=4.36), with attenuation of this effect over time (p=0.047). ORR was lower in the poor PS group (18% vs 36%; p=0.013). Patients with poor PS had higher baseline neutrophil-to-lymphocyte ratios (NLR; median 6.6 vs 4.3; p=0.001). Grade 3-4 irAEs were less frequent in poor PS patients (29% vs 51%; p=0.005), yet hospitalization occurred more often (78% vs 58%; p=0.007), with higher admission frequency. Conclusions: In a large cohort, patients with advanced malignancies and poor PS treated with Ipi+Nivo had markedly worse survival and response rates and substantially higher hospitalization despite fewer severe irAEs. These results highlight a disconnect between toxicity and clinical benefit, identify poor performance status as a marker of limited clinical benefit from dual immune checkpoint blockade, and underscore the need for improved patient selection and alternative therapeutic strategies in this high-risk population. Key clinical outcomes. Outcome ECOG 0–1 (n = 219) ECOG ≥2 (n = 51) p-value Median OS, months 21.2 4.6 &lt;0.001 ORR (CR + PR), % 36% 18% 0.013 Grade 3–4 IrAEs, % 51% 29% 0.005 Any Hospitalization, % 58% 78% 0.007 Baseline NLR (IQR) 4.3 (2.4–6.7) 6.6 (3.5–10.6) 0.001

Impact of concomitant GLP-1 receptor agonist use on overall survival and chemotherapy-related adverse events in colorectal cancer: A real-world pharmacoepidemiologic cohort study.

3661 Background: Glucagon-like peptide-1 receptor agonists (GLP-1 RAs) are increasingly prescribed for diabetes and obesity and exert pleiotropic metabolic and anti-inflammatory effects. While prior studies have focused on their role in cancer prevention, limited data exist regarding their impact on cancer treatment outcomes. We evaluated the association between concomitant GLP-1 RA use and overall survival (OS) in patients with colorectal cancer (CRC) receiving standard first-line chemotherapy. Methods: Using the Blue Diamond Global Collaborative Network, adult patients with CRC initiating first-line FOLFOX or FOLFIRI between 2015 and 2024 were identified. Patients were classified as GLP-1 RA users if they received a GLP-1 RA (e.g., semaglutide, liraglutide, dulaglutide) within 90 days before or after chemotherapy initiation, and non-users otherwise. Cohorts were 1:1 propensity score matched for age, sex, BMI, diabetes status, line of therapy, comorbidities, and baseline corticosteroid use. The primary endpoint was OS. Kaplan–Meier methods and Cox proportional hazards models were used for analysis. Results: After matching, 4,824 patients were included (2,412 per cohort). Baseline characteristics were well balanced (mean age 61.8 years; 46% female; mean BMI 31.2 kg/m²; 58% with diabetes). At a median follow-up of 24.6 months, patients receiving concomitant GLP-1 RAs demonstrated significantly improved overall survival compared with non-users. Median OS was 28.4 months among GLP-1 RA users versus 23.6 months among non-users (log-rank p &lt;0.001). GLP-1 RA use was independently associated with improved OS (HR 0.82, 95% CI 0.76–0.89). Regarding safety outcomes, GLP-1 RA users experienced a lower incidence of neutropenia (18.9% vs 23.7%; RR 0.80, p =0.002) and reduced chemotherapy-induced neuropathy (14.2% vs 18.6%; RR 0.76, p =0.004). Subgroup analyses demonstrated consistent OS benefit across chemotherapy backbone: FOLFOX (HR 0.81) and FOLFIRI (HR 0.84), and the most pronounced effect in patients with BMI ≥30 kg/m². Sensitivity analyses restricting sustained GLP-1 RA exposure (&gt;6 months) yielded similar results. Conclusions: In this large real-world cohort of patients with colorectal cancer, concomitant GLP-1 receptor agonist use was associated with improved overall survival and reduced chemotherapy-related toxicity. These findings suggest a potential adjunctive therapeutic role for GLP-1 RAs during active cancer treatment, potentially mediated through metabolic stabilization, reduced systemic inflammation, and improved treatment tolerance. Prospective studies are warranted to confirm these observations.

Palliative care for advanced cancer: Medicare Advantage and provider networks.

11050 Background: Palliative care is a specialized medical approach to improve symptoms and quality of life for patients with serious illnesses, including cancer, yet its utilization remains low. While enrollment in Medicare Advantage (MA) has grown and surpassed Traditional Medicare (TM), less is known about differences in palliative care between MA and TM and the extent to which MA plan design, particularly provider networks, contributes to these differences. Methods: Using the 2025 SEER–Medicare linkage, we identified Medicare beneficiaries aged ≥66 years diagnosed with distant-stage breast, colorectal, lung, pancreatic, or prostate cancer between 2016 and 2021. We included patients who survived at least 2 months after diagnosis and had continuous enrollment in TM or MA from 1 year before diagnosis to death/end of follow-up. The primary outcome was the cumulative incidence of palliative care billing within 6 months of diagnosis. To assess the role of provider networks, we first assigned each patient a treating oncologist based on the plurality of visits with cancer diagnosis codes; we then conducted 1:1 matching of patients enrolled in a given MA plan to patients in TM treated by the same oncologists within the same county. Multivariable Cox proportional hazards models estimated differences in palliative care billing between TM and MA overall and by MA plan type, before and after matching, adjusting for sociodemographic and clinical characteristics. Results: Among 135,402 beneficiaries with advanced cancer, 67.7% were enrolled in TM and 32.3% in MA; over half were female, 7.3% Hispanic, and 9.5% non-Hispanic Black. The 6-month cumulative incidence of palliative care billing was 13.3% among MA beneficiaries and 9.3% among TM beneficiaries. In adjusted analyses, MA enrollment was associated with higher palliative care billing (hazard ratio [HR], 1.39; 95% CI, 1.34–1.44). When stratified by plan types, palliative care was particularly higher among patients in health maintenance organization (HMO) plans (HR=1.66, 95%CI=1.60-1.72) but not other plan types. After matching on treating oncologists, 23,033 TM and 23,033 MA patients were included. Matched MA beneficaries were less likely to have HMO plans than those not matched (49.1% vs. 76.7%). Differences in palliative care were attenuated and statistically non-significant, with cumulative incidence of 11.0% and 10.4% among MA and TM, and an adjusted hazard ratio of 1.05 (95%CI=0.99, 1.11). Similar attenuation was observed in HMO plans (1.16, 95%CI=1.08, 1.24). Conclusions: Palliative care billing was significantly higher among MA than TM beneficiaries, but these differences were substantially reduced after accounting for provider networks. These findings suggest that between-provider variation accounts for the majority of the difference in palliative care receipt between TM and MA beneficiaries.

Heterogeneity of cancer prevalence in young adults in the post–COVID-19 era: A multicenter real-world analysis.

e22597 Background: COVID-19 exposure has been linked to persistent immune dysregulation, inflammatory signaling, and epigenetic changes whose role in cancer biology remains unclear. This study evaluated temporal differences in early-onset cancer (EOc) prevalence across major cancer groups in the pre- and post-pandemic eras. Methods: De-identified patient data from the TriNetX database; a federated electronic medical record was used to collect the number of patients diagnosed with various cancers. Data was divided into pre-COVID (2014-2019) and post-COVID (2020-2025) groups. The period between 7/1/2019-6/30/20 was excluded given ambiguity of COVID-19 diagnoses in the US during this time. Patients ≥18 years with malignant neoplasms of digestive organs (C15-26), bronchus and lung (C34), soft tissue sarcomas (C49), breast (C50), female genital organs (C51-58), male genital organs (C60-63), urinary tract (C64-68), reticuloendothelial groups (C81-96) were selected in each period. Groups were further stratified by age ≤50 years (EOc) and &gt;50 years. Chi-square and rate ratios were used for statistical analysis. Results: A total of 2,106,086 patients were included in the 2014–19 group and 2,904,405 patients in the 2020-25 group. Patients with &gt;1 malignancy of interest were excluded in individual group analyses. EOc accounted for 6.686% of 2014-19 group and 9.181% of 2020-25 group. Due to large sample sizes, p &lt; 0.001 was deemed significant. The proportion of EOc was significantly increased in the 2020-25 across all groups (p &lt; 0.001) except in the soft tissue sarcoma group ( p = 0.0247). Upon stratification, breast cancer had the highest EOc burden in post-COVID period, followed by gastrointestinal and lung cancers. Urinary tract and female genital groups showed a moderate rise in EOc, while male genital and reticuloendothelial groups had mild increases (p &lt; 0.001, rate ratio ~1). To account for earlier breast cancer diagnoses compared to other groups, the age cut-off for EOc was lowered to ≤40 years. The EOc rate ratio for breast cancer remained high at 2.824 [CI: 2.727–2.925] in 2020-25 with this lowered cut-off. Conclusions: The aftermath of the COVID-19 pandemic has seen increasing cases of younger adults with primary solid malignancies. Heterogeneity among cancer sub-groups, especially a two-fold rise in proportion of early-onset breast cancer, is suggestive of a molecular mechanism rather than system-level differences in health care access. Further studies are needed to clarify mechanisms to improve current screening guidelines. Cancer type EOc rate ratio – post vs pre-COVID [95% CI] Breast 2.0908 [2.0619 - 2.1200] Gastrointestinal 1.8952 [1.8598 - 1.9312] Lung 1.7486 [1.6690 - 1.8320] Urinary tract 1.4781 [1.4376 - 1.5197] Female Genital organs 1.4031 [1.3793 - 1.4273] Male Genital organs 1.0822 [1.0564 - 1.1086] Soft tissue 1.0293 [1.0037 - 1.0557] Reticuloendothelial 1.0275 [1.0167 - 1.0384]

Care utilization and inpatient mortality among women younger than 50 years hospitalized with breast cancer: Racial and socioeconomic disparities.

e13757 Background: Racial and socioeconomic disparities in breast cancer outcomes are well recognized, but it remains unclear whether these inequities are evident at the time of hospital presentation. Women aged less than 50 years represent a distinct subgroup as they are presumed healthier with a lower co-morbidity burden. We conducted a retrospective cross-sectional analysis to evaluate whether racial and socio-economic status are associated with illness severity and need for ICU-level care among hospitalized women with breast cancer younger than 50 years. Methods: A retrospective cross-sectional analysis of the 2018–2022 National Inpatient Sample was conducted. Adult women younger than 50 years hospitalized with a principal diagnosis of breast cancer were identified. The primary outcome was ICU-level care, defined as a composite of mechanical ventilation, shock, or extreme APR-DRG severity. Secondary outcomes included in-hospital mortality, length of stay (LOS) and hospitalization cost. National estimates accounted for NIS stratification, clustering, and discharge-level weighting to generate nationally representative estimates. Multivariable survey-weighted logistic regression assessed associations of race, insurance status, and ZIP-code income quartile with ICU utilization and mortality, adjusting for age, year, elective admission and hospital characteristics. Race-by-insurance interactions were explored using collapsed payer categories to address sparse cells. Results: The cohort included 8,129 unweighted hospitalizations, representing 40,645 weighted admissions nationally. Overall ICU composite utilization was 2.6% (95% CI 2.3–3.0). ICU use varied by race (White 1.4%, Black 3.4%, Hispanic 2.2%, Asian/Pacific Islander 2.1%), insurance (private 2.1%, Medicare 5.0%, Medicaid 3.3%) and neighborhood income (lowest vs highest quartile: 4.0% vs 1.0%). In adjusted analyses, Asian/Pacific Islander women had higher odds of ICU-level care compared to White women (OR 2.18, 95% CI 1.05–4.54), while residence in the highest income quartile was protective (OR 0.45, 95% CI 0.24–0.82). Overall in-hospital mortality was 2.2% but was substantially higher among hospitalizations meeting ICU composite criteria (28.3% vs 1.5%). After adjustment, ICU-level care was the strongest predictor of mortality (OR 8.69, 95% CI 4.99–15.13). Mortality also demonstrated racial and socioeconomic gradients. Conclusions: Among women younger than 50 years hospitalized with breast cancer, racial and socioeconomic disparities are evident at the point of hospital presentation, reflected by differential need for ICU-level care. ICU utilization identifies a small but extremely high-risk subgroup with markedly elevated in-hospital mortality, highlighting severity at presentation as a critical and underrecognized dimension of cancer health equity.

A feasibility and usability study of a virtual exercise platform in patients with non–small cell lung cancer in a large academic center.

e13794 Background: Exercise during cancer treatment improves cardiorespiratory fitness, reduces fatigue, and enhances quality of life, with post-diagnosis physical activity associated with 24-31% reduction in cancer-specific mortality for lung, prostate and colorectal cancer (Ligibel et al. JCO, 2022). However, sustained adherence remains challenging for patients with lung cancer. We conducted a feasibility study evaluating the integration of a physician-prescribed virtual exercise platform into the care of patients with NSCLC and its impact on health outcomes. Methods: This single-center, prospective cohort study enrolled patients with stage I–IV NSCLC. Eligible patients had an ECOG score 0–2 and were deemed appropriate for exercise participation by PM&amp;R. Individualized exercise programs were prescribed by PM&amp;R physicians and delivered through a virtual platform supported by Salaso. The primary outcome was feasibility, defined by enrollment and continued participation over 12 months. Secondary outcomes included patient and provider satisfaction assessed by the System Usability Scale (SUS), Technology Acceptance Model (TAM) for perceived usefulness, and Net Promoter Score (NPS). Exploratory outcome included pulmonary function test (PFT), which were compared at baseline and 12 months using the paired t-test . Results: Among the 41 patients screened, 24 (59%) remained active at follow-up, 7 (17%) failed initial screening, and 10 (24%) withdrew. Median NPS was 33.3, indicating favorable patient satisfaction. Median TAM for perceived usefulness score was 4.65, reflecting patients’ agreement regarding the value of the platform. Median SUS score was 55, suggesting suboptimal usability of the platform with room for improvement. Provider acceptance was high, with 90% reporting they were very or extremely likely to use the platform. Only 5% of patients evaluated by PM&amp;R were deemed inappropriate for participation due to safety concerns, suggesting the potential for streamlining the onboarding process by selective PM&amp;R screening. Fourteen active patients (58.3%) completed baseline PFTs, and five completed repeat testing at 12 months. Mean predicted FVC, predicated FEV1, FEV1/FVC, and DLCO were 89.2%, 85.8%, 74%, and 15.2 at enrollment and 86.2%, 84.2%, 74%, and 11.7 at 12 months, respectively. There is no significant difference between pre and post exercise PFTs (p &gt; 0.05 for all four metrics), suggesting a potential role of exercise in maintaining pulmonary function in this patient group. Conclusions: A physician-prescribed virtual exercise platform is feasible for patients with NSCLC across disease stages. Improvements in platform usability and streamlining of the onboarding process may enhance platform scalability. Virtual exercise delivery represents a promising strategy to support survivorship and functional status in patients with NSCLC. Clinical trial information: NCT06540495 .

Risk of immune-related adverse events from checkpoint inhibitors in cancer patients with autoimmune disease: An emulated target trial.

11120 Background: Immune checkpoint inhibitors (ICIs) have revolutionized cancer treatment by activating the immune system to target and eliminate cancer cells, leading to durable responses and improved survival. However, patients with preexisting autoimmune diseases are excluded from ICI clinical trials due to concerns about triggering immune-related adverse events (irAEs). Consequently, there is limited real-world data comparing the risk of irAEs between ICIs and other systemic cancer therapies in this high-risk population. Methods: We conducted a target trial emulation using a new-user, active-comparator design to compare the risk of developing de novo irAEs following ICI initiation compared to chemotherapy or targeted therapy. Using Merative MarketScan data (2014–2023), we identified adults with cancer and a preexisting autoimmune disease who initiated treatment with ICIs, chemotherapy, or targeted therapy. Propensity score matching (1:1) was performed for two comparisons: ICI versus chemotherapy and ICI versus targeted therapy. The primary outcome was de novo irAEs affecting organ systems distinct from the patient’s baseline autoimmune disease. Cox proportional hazards models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs), with subgroup analyses by age, sex, cancer type, preexisting autoimmune disease category, and ICI class. Results: A total of 3,601 matched patient pairs were included in the ICI–chemotherapy comparison and 1,479 matched pairs in the ICI–targeted therapy comparison. Patients receiving ICI had a significantly higher risk of developing de novo irAEs compared to those receiving chemotherapy (HR=1.71; 95% CI, 1.49–1.97) or targeted therapy (HR=1.43; 95% CI, 1.20–1.70). This increased risk was consistent across age and sex. By ICI class, CTLA-4 inhibitors were associated with the highest risk (HR=2.64; 95% CI, 1.97–3.53 vs chemotherapy; HR=1.72; 95% CI, 1.18–2.50 vs targeted therapy), followed by ICI combination therapy (HR=2.46; 95% CI, 1.80–3.37 vs chemotherapy). By cancer type, patients with liver cancer had the greatest risk (HR=2.75; 95% CI, 1.96–3.85 vs chemotherapy; HR=2.67; 95% CI, 1.68–4.24 vs targeted therapy), followed by breast cancer (HR=2.11; 95% CI, 1.31–3.42 vs chemotherapy; HR=1.97; 95% CI, 1.06–3.68 vs targeted therapy). By baseline autoimmune disease category, patients with endocrine autoimmune diseases consistently showed elevated risk across both comparisons (HR=1.74; 95% CI, 1.42–2.14 vs chemotherapy; HR=1.92; 95% CI, 1.43–2.58 vs targeted therapy). Conclusions: ICI use in cancer patients with preexisting autoimmune diseases is associated with a higher risk of de novo irAEs than chemotherapy or targeted therapy, with risk varying by ICI type, cancer type, and underlying autoimmune condition, highlighting the need for individualized risk assessment and close monitoring.

Development and external validation of a recursive attention-based multimodal deep learning model (RAMs-NPC) for prognostic stratification in nasopharyngeal carcinoma: A retrospective study.

e18005 Background: Accurate prognostication is essential for personalized therapy in nasopharyngeal carcinoma (NPC), yet conventional risk stratification based on TNM staging and EBV DNA fails to capture tumor heterogeneity. We developed and validated a recursive attention mechanism (RAMs)-enhanced multimodal deep learning model (RAMs-NPC) that integrates MRI, digital pathology and clinical data to improve survival prediction for NPC. Methods: This single-center retrospective study enrolled 618 patients with non-metastatic NPC treated from 2017 to 2020, all with pretreatment MRI, H&amp;E slides and ≥5 years of follow-up. Patients were randomly divided into training (n=433) and validation (n=185) cohorts at a 7:3 ratio. The RAMs-NPC model adopted U-Net for MRI tumor segmentation, ResNet for pathological feature extraction and MLP for clinical variable processing. RAMs dynamically focused on prognostically salient regions and weighted multimodal feature contributions adaptively; fused features via an attention gate were input into a Cox proportional hazards model for survival prediction. The primary endpoint was the concordance index (C-index) for overall survival (OS). Results: In the validation cohort, the model achieved an OS C-index of 0.84 (95% CI: 0.80-0.88), a progression-free survival (PFS) C-index of 0.81, a 3-year OS time-dependent AUC of 0.86 and a Dice similarity coefficient of 0.82 for MRI tumor segmentation. Ablation experiments demonstrated that RAMs significantly improved the model performance, with a 6.3% increase in OS C-index and a 6.6% increase in PFS C-index (all p&lt;0.01). Kaplan-Meier analysis revealed significant separation of survival curves between high- and low-risk groups (log-rank p&lt;0.001). Decision Curve Analysis (DCA) confirmed the model’s clinical utility, and the model-derived risk score was an independent prognostic factor for OS (HR=2.95, 95% CI: 1.88-4.63, p&lt;0.001). Conclusions: The RAMs-NPC model synthesizes multimodal data via a recursive attention mechanism, providing a highly accurate and interpretable tool for NPC prognostic stratification. It outperforms conventional staging systems, shows promising potential to guide personalized treatment decisions, and warrants further external validation.

Cervical cancer in Armenia: Current situation and challenges.

e17529 Background: In Armenia, cervical cancer (CC) is the fifth most common cancer among women. Among 15 to 44-year-old women, CC is the second most common cancer. Every year, approximately 250 new cases are diagnosed; of all cases, 43% are in stage III, and 19% are in stage IV. Methods: We collected statistics from Armenian oncology registers during the 2016 to 2024 period, GLOVOCAN, and WHO. Also Information about CC statistics was collected from published articles, the HPV international centers article. Treatment availability was assessed against international guidelines. Information about CC screening statistics is very low, and we collect information from a few articles. Results: Despite high prevention and early diagnosis opportunities, CC in Armenia was diagnosed in early stages in 27,5% of cases in the last 8 years. More cases were detected in the locally advanced and unresectable stage. Armenia started doing the PAP screening test in Jan 2015, targeting women aged 30 to 60 years. From 2007 to 2022 period the proportion of women undergoing PAP testing increased from 5.6 % to 31%. From 2015 to 2021, approximately 329.000 women were screened with coverage of the national program. From a cell-based survey examination among women aged 21 to 39, just 18% had any type of HPV, 11% had high-risk HPV, and 3.3% had HPV 16 and 18 types. In Armenia, because of cultural features, the median age of first sexual intercourse is 21 years among women in Armenia, which is late compared to many countries. The HPV vaccination coverage program in Armenia started in Dec 2017. In 2024, HPV vaccination coverage among, last dose, 15-year old girls was only 41%, and in 2020, just 2% was vaccinated. We compared three countries: Australia, Bulgaria, and Armenia, where HPV vaccination program coverage, last dose, among females in 2022 was 57%, 3%, and 14%, respectively. In 2022, in these three countries, age-standardized CC mortality rates per 100.000 women were 1.4, 6.2, and 4.0, respectively, which illustrated the impact of an effective vaccination program on CC. We don't have national guidelines in Armenia, and so we use international ones, which are not adapted for our country. Some medications, for example, Pembrolizumab is not registered in Armenia, and there is no governmental reimbursement. As a result, many patients refuse to receive immunotherapy because of the high cost and inaccessibility. Conclusions: In Armenia, CC is mostly diagnosed in the locally advanced stage, despite early, highly effective screening and prevention programs. We cannot collect how many cases were diagnosed in the precancer stage, such as CIN1, CIN2, and CIN3, because of the absence of national data. We can say that despite a higher median age of first sexual intercourse, the circulation of HPV remains sufficient to sustain CC. In Armenia, anti-vaccination campaigns are very active, but the number of vaccinated people in society is increasing year by year and not decreasing, striving to reach the WHO target.

Pembrolizumab vs nivolumab plus chemotherapy in metastatic gastric cancer: TriNetX propensity score–matched analysis.

e16066 Background: Pembrolizumab and nivolumab are PD-1 inhibitors used with chemotherapy for first-line HER2-negative advanced gastric/gastroesophageal junction (G/GEJ) adenocarcinoma; comparative real-world outcomes are limited. We performed a propensity score-matched (PSM) analysis comparing OS and safety in the federated TriNetX Research Network. Methods: Retrospective cohort. Adults ( &gt; = 18) with advanced/metastatic gastric cancer (ICD-10 C16 with C77-C79) treated with pembrolizumab (C1) or nivolumab (C2) plus platinum-fluoropyrimidine; index was first PD-1 dose. Cohorts were 1:1 PSM on demographics, comorbidities (hypertension, diabetes, CKD, cardiovascular disease), baseline hemoglobin, and TNM stage. Primary: all-cause mortality. Secondary: noninfective colitis, AKI, sepsis, pneumonia, acute respiratory failure, neutropenia, anemia, GI hemorrhage, VTE. Timepoints: 3m, 6m, 205d, 1y, 2y, 3y, and 5y. Statistical analyses: Kaplan–Meier (log-rank), Cox regression, and t-tests (event instances). Results: After PSM, n = 153/cohort; baseline was balanced (all SMD &lt; 0.1): age 57.3±13.6 vs 57.5±13.0 years; White 66.0% vs 64.1%; secondary lymph node involvement 50.3% vs 45.8%; hypertension 48.4% vs 47.7%; anemia 47.1% vs 45.8%; diabetes 20.9% vs 20.3%. Median follow-up was 226d (IQR 485) for C1 and 354d (IQR 560) for C2. Mortality was higher with C1 at 205d (43.4% vs 34.4%; HR 1.448, 95% CI 1.007–2.083; log-rank P = 0.0447), with nonsignificant elevations at 3m (22.3% vs 15.0%; HR 1.599; P = 0.085), 6m (37.5% vs 31.1%; HR 1.368; P = 0.110), and 1y (53.3% vs 46.4%; HR 1.368; P = 0.054); no differences at 2y (64.5% vs 64.2%; HR 1.229; P = 0.150), 3y (69.1% vs 71.5%; HR 1.164; P = 0.268), or 5y (71.7% vs 72.9%; HR 1.161; P = 0.270). Noninfective colitis was lower in C1, significant at 3y (16.3% vs 25.5%; risk difference P = 0.049) and 5y (16.3% vs 26.1%; P = 0.036), while HR remained non-significant (5y HR 0.714, 95% CI 0.433–1.177; P = 0.184). Sepsis incidence was similar, but C1 had fewer mean sepsis episodes at 6m (1.2 vs 2.3; P = 0.008), 205d (1.3 vs 2.2; P = 0.010), 2y (1.6 vs 3.5; P = 0.022), 3y (1.6 vs 3.5; P = 0.020), and 5y (1.6 vs 3.5; P = 0.020). No differences were observed for AKI (5y 32.7% vs 37.9%; P = 0.339), pneumonia (5y 26.1% vs 27.5%; P = 0.796), neutropenia (5y 22.2% vs 29.4%; P = 0.151), anemia (5y 48.2% vs 48.2%; P = 1.000), GI hemorrhage (5y 19.0% vs 19.0%; P = 1.000), or VTE (5y 17.2% vs 24.0%; P = 0.163). Conclusions: In this real-world PSM analysis, pembrolizumab plus chemotherapy showed higher 205-day mortality vs nivolumab plus chemotherapy, while 2–5-year survival was similar. Pembrolizumab showed lower noninfective colitis rates and fewer recurrent sepsis episodes. The early survival difference warrants evaluation for confounding (PD-L1/HER2 status) and early toxicity management; prospective head-to-head trials are needed.

Prognostic factors associated with overall survival in patients with Merkel cell carcinoma: A retrospective analysis in a Mexican cohort at a tertiary referral center in Latin America.

e21571 Background: Merkel cell carcinoma is a rare but highly aggressive neuroendocrine neoplasm of the skin. It predominantly affects older people and immunocompromised individuals. Furthermore, it is known for its high risk of recurrence and metastatic potential. This retrospective cohort aims to identify clinical, pathologic, and therapeutic variables as prognostic factors associated with overall survival. Methods: Demographics, clinical and pathological features, treatment modality, progression, and functional status were recorded. Associations with mortality were assessed using contingency analyses (χ2/Fisher). Survival analysis was estimated using Kaplan–Meier with log-rank testing for PFS and OS. Cox proportional hazards models produced univariate and multivariable hazard ratios. Results: In this retrospective cohort of 63 patients, the median age was 70 years (IQR 62–80). Stages I–IV comprised 15.9%, 20.6%, 30.2%, and 33.3%, respectively. Forty-five patients (71.4%) were alive, and 18 (28.6%) were deceased; progression occurred in 22 (34.9%). Contingency analyses linked mortality to metastasis (p &lt; 0.001), ECOG (p = 0.001), clinical stage (p &lt; 0.001), tumor border status (p &lt; 0.001), symptoms (p = 0.037), superinfection (p = 0.005), and receipt of institutional systemic therapy (p = 0.021). Univariate Cox findings included progression HR 7.00 (95% CI 2.09–23.38; p = 0.002), metastatic disease HR 32.5 (95% CI 7.19–146.89; p &lt; 0.001), ECOG ≥2 HR 8.20 (95% CI 2.05–32.76; p = 0.003), and stage IV HR 22.5 (95% CI 2.32–218.35; p = 0.007). Defined surgical margins were protective (HR 0.098; 95% CI 0.027–0.352; p &lt; 0.001). Receipt of extra-institutional treatment correlated with a higher unadjusted hazard (HR 3.9; 95% CI 1.18–12.83; p = 0.025). Multivariable Cox regression retained progression as an independent predictor of reduced OS (adjusted HR 4.76; 95% CI 1.16–19.58; p = 0.030); metastasis showed a strong trend after adjustment (adjusted HR 9.41; 95% CI 0.94–94.10; p = 0.056). Median PFS was 17.0 months (95% CI 0.0–47.9), and median OS was 39.0 months (95% CI 16.7–61.3); metastatic patients had markedly inferior OS (median 7.0 months vs NR for M0; p &lt; 0.001), and age &gt; 55 years was associated with shorter OS (median 39.0 vs 113.0 months; p = 0.033). Conclusions: Progression was the principal independent determinant of survival. Metastatic disease, poor performance status, and advanced stage strongly predicted mortality in univariate analyses; the limited sample size likely constrained detection of additional independent effects. Clear surgical margins were associated with improved unadjusted survival, underscoring the importance of early control and accurate staging.

The impact of COVID-19 mRNA vaccination on bispecific antibody outcomes in hematologic malignancies: A retrospective multicenter cohort study.

7040 Background: Bispecific antibody (BsAb) therapies are universally available, off-the-shelf therapies and are increasingly being used in hematologic malignancies. However, long term efficacy, durability and survival data are lacking. Furthermore, widespread expansion is limited by immune-mediated toxicities, including cytokine release syndrome (CRS), immune effector cell-associated neurotoxicity syndrome (ICANS), and infections. The real-world impact of mRNA SARS-CoV-2 vaccines on outcomes in patients receiving BsAb therapies remain unknown. Methods: We conducted a retrospective, propensity score-matched cohort study using the TriNetX federated electronic health record network. Adult patients with hematologic malignancies who received BsAb therapy were included and stratified by receipt of mRNA SARS-CoV-2 vaccines. Patients with documented COVID-19 infection or post-COVID-19 conditions were excluded. Cohorts were matched 1:1 on baseline demographic and clinical characteristics, yielding 238 patients per group, and followed longitudinally. Outcomes included all-cause mortality, any-grade CRS grade, and ICANS, graded per the ASTCT criteria. Associations were assessed using risk estimates and time-to-event analyses with Kaplan-Meier methods and Cox proportional hazards models. Event burden was evaluated by the number of events per patient among those experiencing CRS or ICANS. Results: All-cause mortality was lower in the vaccinated cohort compared with the unvaccinated cohort (21.4% vs 29.0%; p=0.057), with significantly improved survival on time-to-event analysis (hazard ratio [HR] 0.55; 95% CI, 0.38-0.79; log-rank p=0.001). The incidence of CRS grade 1/2 was similar between cohorts (25.2% vs 24.0%), with no difference in time to first CRS event (HR 0.98; 95% CI, 0.68-1.41). ICANS incidence was also comparable (10.9% vs 11.3%), with no significant difference in time-to-event (HR 0.87; 95% CI, 0.51-1.50). Among patients who developed ICANS, the vaccinated cohort experienced a higher mean number of ICANS episodes (2.58 vs 1.67; p=0.036) see (table1). Conclusions: In this propensity score-matched real-world analysis, mRNA SARS-CoV-2 vaccination was associated with significantly reduced all-cause mortality (HR, 0.55) among patients receiving BsAb therapy, without an increased risk of CRS grade 1/2 or ICANS. These findings support the safety of mRNA vaccination in this high-risk population and suggest a potential survival benefit necessitating prospective validation. Outcome Vaccinated (n=238) Unvaccinated (n=238) Key Effect All-cause mortality 21.4% 29.0% HR 0.55 (95% CI 0.38-0.79) CRS grade 1/2 25.2% 24.0% HR 0.98 (95% CI 0.68-1.41) CRS grade 1/2 episodes (mean) 2.98 2.35 P=0.16 ICANS 10.9% 11.3% HR 0.87 (95% CI 0.51-1.50) ICANS episodes (mean) 2.58 1.67 p=0.036

Bispecific antibodies in breast cancer: A systematic landscape analysis of clinical trials.

e14503 Background: Bispecific antibodies (BsAbs) engage two targets to enhance antitumor activity and potentially overcome resistance in breast cancer, but their mechanistic diversity has not been systematically mapped. Methods: ClinicalTrials.gov was systematically searched for therapeutic trials evaluating BsAbs with breast cancer as an eligible cohort. Studies were excluded if breast cancer was ineligible, the agent was not bispecific, or the study was non-therapeutic. Trials were categorized by target pair, class, and phase. Descriptive analyses were performed. Results: We identified 34 clinical trials evaluating 21 BsAbs that included patients with breast cancer. HER2 was the most frequently targeted pathway (n = 18), with HER2×HER2 constructs most common (n = 8; 3 BsAbs). Immune checkpoint BsAbs were prominent, particularly PD-1×CTLA-4 (n = 8; 4 BsAbs). CD3×HER2 T-cell–engaging BsAbs were also frequently studied (n = 7; 3 BsAbs). Clinical efficacy/safety outcomes were reported in a limited subset. Across three phase II HER2-directed BsAb studies, objective response rates ranged 27–76%, with complete responses observed in at least five patients (zenocutuzumab n = 2, KN026 n = 3). Partial responses comprised most responses, with at least 47 patients achieving partial response. Median progression-free survival ranged 5.5–27.7 months. Grade ≥3 treatment-related adverse events occurred in approximately 51–67%, most commonly neutropenia and diarrhea. Additionally, a phase II immune checkpoint BsAb targeting PD-L1×CTLA-4 demonstrated ORR 44% and mPFS 7.3 months in metastatic triple-negative breast cancer. Conclusions: The BsAb clinical landscape in breast cancer is rapidly expanding with marked mechanistic heterogeneity and enrichment of HER2-directed strategies. Early outcomes from select agents show meaningful antitumor activity, supporting continued development of oncogene-targeted and immune-modulating BsAbs. Distribution of bispecific antibody clinical trials in breast cancer by target combination and phase. Target N Phase Example agents HER2×HER2 8 I–III Zanidatamab; KN026;TQB2930 CD3×HER2 7 I–II Ertumaxomab; ISB1302; HER2Bi HER2×HER3 1 II Zenocutuzumab 4-1BB×HER2 1 I YH32367 PD-1×CTLA-4 8 I–II QL1706; SI-B003; KN046; Vudalimab CTLA-4×LAG-3 1 I Pavunalimab CD3×MUC1 1 II Anti-CD3×MUC1 Other PD-1-based 5 I–III See footnote† CD47-based 2 I See footnote* †PD-1–based checkpoint bispecific antibodies include: PD-1×TIGIT: Rilvegostomig; PD-1×VEGF: Ivonescimab; PD-L1×VEGF-A: Pumitamig; PD-1×CD73: AK131; ICOS×PD-1: XmAb23104. *CD47-based include: CD47×HER2: IMM2902; CD47×MSLN: NI1801.