Browse Articles
Discover research articles across all indexed journals
Association of healthcare delivery model with stage at diagnosis of pancreatic cancer.
e23126 Background: Pancreatic cancer is a leading cause of cancer mortality in the U.S., with a 5-year relative survival rate of 13.3%. Prior studies suggest improved cancer outcomes within integrated healthcare systems compared with nonintegrated systems; however, the association between healthcare delivery model and stage at diagnosis for pancreatic cancer has not been assessed. We examined whether diagnosis within a vertically integrated healthcare system is associated with earlier stage at diagnosis in pancreatic cancer among a diverse patient population. Methods: We conducted a retrospective cohort study using the California Cancer Registry, identifying 12,187 insured patients diagnosed with pancreatic cancer in Southern California from 2015–2021. Patients were diagnosed in Kaiser Foundation (KF) Hospitals, part of a vertically integrated healthcare system, or in non-KF hospitals. Data included SEER summary stage at diagnosis and demographic characteristics (age, race and ethnicity, sex, and geocoded socioeconomic status). Multivariable logistic regression adjusted for these covariates assessed the association between hospital of diagnosis (KF vs non-KF) and stage at diagnosis (localized vs advanced). A subgroup analysis among KF members examined demographics associated with stage at diagnosis. Results: Among 12,187 patients, 3,194 (26.2%) were diagnosed within KF. The cohort was racially diverse: 3,080 (25.3%) Hispanic, 965 (7.9%) Black/African American, and 1,543 (12.7%) Asian/Pacific Islander patients. Overall, 2,128 (17.5%) patients presented with localized disease and 10,059 (82.5%) with advanced stage. Compared with diagnosis at non-KF hospitals, those diagnosed within KF had higher odds of localized stage at diagnosis (OR 1.15, 95% CI 1.03-1.28). In subgroup analyses by race and ethnicity, higher odds of presenting with localized disease were seen across patient groups, including Asian/Pacific Islander (OR 1.18; 95% CI 1.13 - 1.24), Black/African American (OR 1.36; 95% CI 1.29-1.43), Hispanic (OR 1.14; 95%CI 1.10 - 1.17), and non-Hispanic White (OR 1.13; 95% CI 1.11-1.16) when compared to their counterparts diagnosed at non-KF hospitals. In a subgroup analysis of patients within KF, Hispanic (OR 1.22; 95% CI 0.97-1.54) and Asian/Pacific Islander (OR 1.37; 95% CI 1.03-1.82) patients were at increased odds of being diagnosed with local disease compared to Non-Hispanic White individuals. Conclusions: Odds of localized stage at diagnosis in pancreatic cancer was greater in KF compared with diagnosis in non-integrated healthcare settings across a racially diverse patient population. Within KF, Asian/Pacific Islander patients had a statistically significant increased odds of being diagnosed with local disease. Further studies are warranted to elucidate mechanisms of how vertically integrated healthcare delivery models may facilitate earlier diagnosis of pancreatic cancer.
A phase 1/2 first-in-human study of TH9619 in patients with advanced refractory solid tumors (ODIN).
TPS3165 Background: TH9619 is a first-in-class, potent, small-molecule, and dual inhibitor of methylene-tetrahydrofolate dehydrogenase (MTHFD)1 and MTHFD2, both highly overexpressed in solid tumors and cancer-specific key enzymes within the one-carbon metabolic pathway. TH9619 kills cancer cells via a dual mechanism of action (1) inhibition of MTHFD1 traps folate leading to thymidine depletion, and (2) inhibition of nuclear MTHFD2 disrupts DNA damage response and repair pathways. With its unique characteristics, in preclinical models, TH9619 kills tumour cells, while sparing healthy tissues. The results from pre-clinical models, the novel mechanism of action, and the high unmet medical need in advanced refractory solid tumours support this investigation. Methods: ODIN (NCT07151040; EudraCT No. 2024-519639-40-00) is a first-in-human, multicentre, open-label, Phase 1/2 study. Eligible patients include adults with histologically confirmed advanced colorectal cancer, non–small cell lung cancer, head and neck squamous cell carcinoma, gastric cancer, or gastroesophageal junction cancer who have exhausted the institutional standard of care and have progressive and measurable disease per RECIST 1.1. Patients will receive TH9619 monotherapy via an intravenous infusion, weekly for 3 weeks, followed by one week without infusion, of a 28-day cycle, for up to 2 years. The Phase 1a dose-escalation, planning to recruit up to 80 patients, will assess the overall safety and tolerability profile and determine the maximum tolerated dose (MTD) and recommended Phase 2 dose(s) (RP2D(s)). Phase 1b is cohort expansion and plans to recruit up to 60 patients. Assessments include adverse events, pharmacokinetic (PK) and pharmacodynamic (PD) parameters, and preliminary anti-tumor activity per RECIST v1.1. Additionally, predictive biomarkers and metabolites related to the one-carbon metabolism and potential correlation with efficacy of TH9619 will be explored. The Phase 2 will further characterize safety, PK, PD, and efficacy at the RP2D(s). The ODIN phase 1/2 study is actively enrolling across leading academic and clinical research centres in the United Kingdom, France, and Spain, with expansion planned across additional European sites in the coming months. As of this submission, dose escalation is ongoing. Clinical trial information: NCT07151040 .
A deep-learning model for immunotherapy efficacy prediction directly from H&E slides in head-neck squamous cell carcinoma (HNSCC).
e18006 Background: Immunotherapy (IO) has emerged as the standard of care for recurrent and metastatic HNSCC. However, conventional PD-L1 expression levels offer limited utility for predicting therapeutic response. This study evaluates the application of a deep-learning framework, ENLIGHT-DP, to predict response to IO in HNSCC directly from common H&E slide scans. Methods: IO efficacy and pre-treatment tumor H&E slides were retrospectively collected from 113 HNSCC cases seen between 2020 and 2025 at the Taipei Veterans General Hospital (VGHTPE). On this data we trained, in cross-validation, an attention-based deep-learning model (ENLIGHT-DP) to predict a continuous response score to IO - the ENLIGHT Matching Score (EMS). For clinical practice, we also categorized all patients into three prediction classes: “EMS- H” (top 30% EMS), “EMS-I” (middle 40%) and “EMS-L” (bottom 30%). We defined "PDL1-I" as CPS 1-19, TC 1-9 if CPS is missing, and TPS 1-45 if both are missing. “PDL1-L” and “PDL1-H” were defined by values below and above these thresholds, respectively. Finally, we validated the model on two previously published cohorts collected at Hadassah Medical Center (HMC, n=25) and the BIO2 study from UHN, Toronto (BIO2, n=15). Results: The VGHTPE cohort had a median age of 57.8 years, with 90% males. Primary site distribution was 55% oral cavity, 17% oropharynx (5/19 cases HPV-associated), 15% hypopharynx, 8% larynx, 4% others. 15% of the cases were stage I/II, 9% stage III, 76% stage IV. 85% of cases had received prior radiation therapy, 48% had locoregional disease only, and 55% were treatment-naïve for the recurrent/metastatic disease. The cohort had an ORR of 34.5%, PFS of 5.6 months (3.3-7.9), and OS of 28.1 months (12.7-43.4). The EMS-H group had an ORR of 52.5%, 37% higher than the baseline. The EMS-I group had an ORR of 33.3%, and the EMS-L group 17.5%, 49.3% lower than baseline. The 35% ORR difference between EMS-H and EMS-L more than doubled the 16.3% ORR difference between PDL1-H and PDL1-L (41.3% and 25% respectively). ENLIGHT-DP achieved good stratification of the patient population with respect to PFS (HR: 0.447, p=0.005 for EMS-H vs. rest; HR: 0.343, p=0.0003 for EMS-H vs. EMS-L). This is superior to the PFS separation using PD-L1 (HR: 0.69, p=0.14 for PDL1-H vs. rest; HR: 0.523, p=0.0055 for PDL1-H vs. PDL1-L). ENLIGHT-DP was also borderline significant in stratifying with respect to OS (HR:0.55 , p=0.07 for EMS-H vs rest; HR: 0.577, p=0.16 for EMS-H vs. EMS-L), while PD-L1 was not predictive of OS (HR: 1.12, p=0.71 for PDL1-H vs. rest; HR: 1.1, p = 0.8 for PDL1-H vs PDL1-L). ENLIGHT-DP generalizes to the external cohorts with a ROC AUC of 0.70 in the HMC cohort and 0.67 in the BIO2 cohort, vs. 0.68 in the VGHTPE cohort. Conclusions: ENLIGHT-DP demonstrated a superior prediction performance to IO ORR, PFS, and OS in HNSCC, suggesting a potential clinical utility for treatment guidance.
Prevalence and risk of anemia and iron deficiency anemia in premenopausal women with breast cancer: A single-center retrospective study.
e24166 Background: Anemia is common in breast cancer patients at diagnosis and during treatment. It has been associated with reduced treatment response, decreased quality of life, and inferior survival outcomes. Iron deficiency anemia (IDA) is a particularly important contributor, especially in premenopausal women. Although studies from Malaysia and Saudi Arabia identified premenopausal status as a risk factor for IDA in breast cancer patients, data from developed countries, including the United States, remain limited. We conducted a single-center retrospective cohort study to measure the prevalence, incidence and risk factors for IDA in premenopausal women with breast cancer. Methods: We performed a retrospective chart review of premenopausal female breast cancer patients age ≤55 years who received care at Stony Brook University Hospital between 2005–2025. Demographic, clinical, and laboratory data were extracted. Baseline anemia prevalence, treatment-related anemia incidence, and iron deficiency anemia rates were determined. IDA was defined as hemoglobin < 12 g/dL with either ferritin < 30 ng/mL, or ferritin 30-100 ng/mL with transferrin saturation (Tsat) < 20%. Results: A total of 281 patients were included. Median age was 44 years (range, 23–55). Caucasian (74%) and African American (7.1%) were the most common races, and 51 patients (23%) were Hispanic. A total of 162 patients (57.7%) received chemotherapy, 269 (95.8%) received surgery, and 237 (84.3%) received endocrine therapy. At time of cancer diagnosis, 18.1% (51/281) of patients had anemia based on the above pre-defined parameters. During treatment, 59.6% (137/230) developed treatment-related anemia. Fifty-two patients had iron studies prior to cancer treatment and 152 had them during treatment. The prevalence of IDA was 15.4% (8/52) at baseline and higher at 23.0% (35/152) during treatment, of which 19 had ferritin < 30 ng/mL. Only 3 (1.1%) patients required treatment interruptions or suspension of systemic therapy due to anemia. There were 58 patients (21%) who received iron repletion and 31 patients (11%) who received packed red blood cell (pRBC) transfusion during cancer treatment. Among the 36 patients (62%) who had follow up iron workup after receiving iron repletion, 95% had ferritin increase, with an average increase of 495 ng/mL. Multivariable analysis showed that chemotherapy (p = 0.029), radiation therapy (p = 0.005) and inflammatory bowel disease (p = 0.023) were independently associated with presence of IDA. Conclusions: In this large cohort of premenopausal breast cancer patients in the United States, both anemia and IDA prevalence were high during therapy. Transfusion and iron repletion are frequently needed. Further investigation is needed to determine the impact of routine iron study and iron repletion on treatment adherence, patient quality of life, and overall clinical outcomes.
Trends and in-hospital outcomes of multiple myeloma in autoimmune rheumatic disease hospitalizations: A National Inpatient Sample analysis, 2016–2022.
e19577 Background: Multiple myeloma (MM) is associated with immune dysregulation and substantial inpatient morbidity. Autoimmune rheumatic diseases (ARDs), including systemic lupus erythematosus, rheumatoid arthritis, systemic sclerosis, Sjögren syndrome, and vasculitis, may further modify MM outcomes through chronic inflammation and immunosuppressive therapy. Contemporary national data describing temporal trends and inpatient outcomes of MM among ARD patients are limited. We evaluated national trends, inpatient outcomes, and resource utilization in MM hospitalizations with ARDs. Methods: We analyzed the National Inpatient Sample (2016–2022) using survey-weighted methods. Adult ARD hospitalizations were stratified by MM status. Temporal trends were assessed using year as a continuous variable. Outcomes included in-hospital mortality, sepsis, acute kidney injury (AKI), dialysis, disseminated intravascular coagulation (DIC), venous thromboembolism (DVT), anemia, thrombocytopenia, and tumor lysis syndrome (TLS). Multivariable survey logistic regression adjusted for demographics, comorbidities, and hospital factors. Results are reported as adjusted odds ratios (aOR) with 95% confidence intervals (CI). Results: Among an estimated 6.21 million ARD hospitalizations, 0.39% involved MM. MM prevalence increased from 0.31% in 2016 to 0.41% in 2022, with each year associated with a 3.9% higher odds of MM (aOR 1.04, 95% CI 1.02–1.06; p<0.001).MM patients were older (70.9 vs 65.4 years; p<0.001) and predominantly female (p<0.001). In-hospital mortality among MM hospitalizations rose from 2.8% to 3.6%, increasing 7.1% per year after adjustment (aOR 1.07, 95% CI 1.06–1.08; p<0.001). Compared with ARD hospitalizations without MM, MM was independently associated with higher odds of sepsis (11.9% vs 9.7%; aOR 1.19, 95% CI 1.08–1.30), AKI (34.2% vs 20.2%; aOR 1.79, 95% CI 1.67–1.91), DIC (0.53% vs 0.21%; aOR 2.36, 95% CI 1.59–3.50), DVT (4.9% vs 2.9%; aOR 1.59, 95% CI 1.39–1.83), anemia (62.5% vs 37.8%; aOR 2.51, 95% CI 2.36–2.67), thrombocytopenia (14.5% vs 6.3%; aOR 2.40, 95% CI 2.20–2.62), and TLS (0.51% vs 0.04%; aOR 11.89, 95% CI 7.95–17.80; all p<0.001). MM hospitalizations had longer length of stay (6.96 vs 5.45 days) and higher total charges ($84,594 vs $68,796). Conclusions: Hospitalizations involving MM among patients with ARDs are increasing and are associated with rising mortality, complication burden, and healthcare utilization. These findings highlight the need for early risk stratification and integrated oncologic-rheumatologic inpatient care.
From SMARCA4 deficiency to tumor-agnostic therapies: A single-center retrospective study for 66 patients with SMARCA4-deficient tumors.
e15206 Background: SMARCA4-deficient tumors represent a newly defined category of malignancies in recent years, characterized by high aggressiveness, rapid progression, and a lack of established treatment guidelines. Previous studies on SMARCA4-deficient thoracic tumors have suggested immune checkpoint inhibitor (ICI) therapy as a potential treatment strategy. However, the clinicopathological features and treatment approaches for SMARCA4-deficient tumors of other pathological types remain understudied. Tumor-agnostic therapy is a novel treatment approach that involves using the same molecularly targeted drug across different tumor types sharing a common molecular alteration. Whether this therapeutic strategy can be applied to SMARCA4-deficient tumors warrants investigation. Methods: We retrospectively analyzed the clinicopathological features, prognostic factors, treatment patterns, and clinical outcomes of 66 patients diagnosed with SMARCA4-deficient tumors at Xijing Hospital. Results: Tumors primarily occurred in the thoracic cavity, digestive system, nasal and paranasal sinuses, uterus, ovary, and other sites. Intermediate PD-L1 expression was observed in 42.9% (9/21) of patients. Over half of the tumors expressed Syn, CgA, CD56, and INSM1 (60%, 24/40). Genetic testing (n = 10) identified TP53 and KRAS as the predominant co-mutated genes, with other co-mutated genes including APC, CDKN2B, LRP1B, and PREX2. Thoracic tumors, two or more metastases, and stage IV disease were associated with poor prognosis, whereas surgical intervention was correlated with improved survival. First-line ICI-based combination therapy (n = 6) achieved an objective response rate (ORR) of 33.3%, a disease control rate (DCR) of 66.7%, and a median progression-free survival (mPFS) of 11.8 months. In contrast, later-line therapy (n = 8) resulted in an ORR of 0%, a DCR of 25.0%, and an mPFS of 2.4 months. First-line chemotherapy (n = 9) yielded an ORR of 22.2%, a DCR of 33.3%, and an mPFS of 1.8 months. First-line anti-angiogenic treatments (n = 5) showed an ORR of 0%, a DCR of 40%, and an mPFS of 3.2 months. Conclusions: SMARCA4-deficient tumors exhibit moderate PD-L1 expression and neuroendocrine differentiation features. Surgery can improve patient survival. For patients ineligible for surgery, first-line ICI-based combination therapy across different pathological types is associated with improved survival outcomes. This suggests that ICI-based combination therapy holds potential as a tumor-agnostic treatment strategy for SMARCA4-deficient tumors to overcome the histological and anatomical constraints.
Aspirin use and outcomes with immune checkpoint inhibitors in metastatic cancer.
e14584 Background: Aspirin (ASA) may modulate antitumor immunity. We evaluated the association between concomitant ASA use and outcomes in metastatic cancers treated with immune checkpoint inhibitors (ICI). Methods: We conducted a population-based cohort study of patients with metastatic cancers treated with ICIs from January 2015 to December 2024, stratified by concomitant ASA use (users vs non-users). ASA treatment was described as 100 mg aspirin continuously through therapy. In this series, the indication for ASA was: prophylaxis for coronary artery disease (CAD) or CAD-equivalent conditions and/or comorbidities. The primary endpoint was OS, and the secondary endpoints were PFS, ORR, and DCR. Results: Among 347 ICI-treated patients, ASA exposure status was available for 338 patients (82 users, 256 non-users); median age at ICI initiation was 63.6 years (IQR: 55.7–70.4), and 25.9% were female. The most recorded cancers were NSCLC (34.9%), RCC (25.1%), melanoma (14.4%), bladder cancer (9.2%), and SCLC (6.6%), with the remaining cancers accounting for 9.8%. ICI exposures were nivolumab (58.2%), pembrolizumab (21.3%), atezolizumab (7.8%), ipilimumab (6.6%), nivolumab–ipilimumab (2.6%), avelumab (2.6%), and durvalumab (0.9%). ASA users and non-users were generally well-balanced across malignancy and treatment distributions, baseline sites of metastasis, and ECOG PS; ASA users were older at the time of ICI commencement (median 66.6 vs 62.4 years; p = 0.001). Median PFS was 11.1 months (95% CI 8.2–14.1) and 3.9 months (95% CI 2.9–5.0) for ASA users vs non-users (p = 0.001). On multivariate analysis, ASA use (HR 0.565; 95% CI: 0.388–0.823; p = 0.003) and ECOG PS 0–1 (HR 0.330; 95% CI: 0.193–0.567; p < 0.001) were independent predictors of a favorable PFS outcome. The median follow-up for OS was 17.8 months (95% CI: 15.3-20.3). Median OS was 36.1 months (95% CI 11.9–60.2) in users as contrasted with 15.7 months (95% CI 11.1–20.3) in non-users (p = 0.047). ASA use (HR 0.597, 95% CI 0.379-0.939; p = 0.026) and ECOG PS 0-1 (HR 0.406, 95% CI 0.262–0.631; p < 0.001) were independently correlated with improved OS on a multivariable analysis whereas baseline lung metastases (HR 1.643, 95% CI 1.062–2.542; p = 0.025) and baseline liver metastases (HR 2.242, 95% CI 1.443–3.483; p < 0.001) were associated with worse OS. In the response-evaluable cohort (n = 240), ORR was 55.7% (34/61) in users vs 44.1% (79/179) in non-users (p = 0.117), and DCR was 83.6% (51/61) vs 62.6% (112/179), respectively (p = 0.002). Conclusions: Among patients with ICI-treated metastatic cancer, aspirin use was associated with longer PFS and OS, as well as higher DCR. These findings require prospective validation to ascertain causality, patient selection, and the safety–benefit profile of aspirin with ICIs.
Genomic clonality analysis of patients with metastatic lung-limited non–small cell lung carcinoma (NSCLC) who underwent double lung transplantation (DLT) across timepoints and tumor region.
8658 Background: Double lung transplantation (DLT) registry aimed for lung-limited malignancies (DREAM) study (NCT05671887) cohort A evaluates DLT as a therapeutic strategy for patients with advanced bilateral lung-limited NSCLC. This is the first translational report of the genomic clonality analysis of the DREAM study across timepoints and tumor regions. Methods: Patients with advanced bilateral lung-limited NSCLC who underwent DLT at Northwestern Memorial Hospital between September 2021 and December 2025 were included in the DREAM study cohort A. Patients with extrapulmonary disease were excluded. The primary indications for DLT were disease refractory to systemic therapies, with or without respiratory failure. Tissue next generation sequencing (tNGS) data was collected from diagnosis, bilateral explant and recurrence samples. Commercial tNGS panels utilized include Altera, Caris, Foundation Medicine, Tempus xT, and PGDx. Truncal genetic mutations were identified from genes that were common to all tNGS panels. Results: Among 18 patients enrolled in DREAM study cohort A, 11 patients with bilateral tissue NGS were analyzed. The median age was 55.5 years (range 37-74), 9 (81%) were female, and 5 (45%) had a history of smoking. 6 patients (54%) had invasive mucinous adenocarcinoma histology. All 11 patients harbored common gene mutations in bilateral tumors and had common oncogenic genomic clones indicating a truncal mutation. Details are outlined in Table 1. Conclusions: All patients harbored common oncogenic driver mutations across all available specimens, providing definitive genomic evidence of a monoclonal origin. These findings not only distinguish intrapulmonary metastasis from multiple primary tumors but also support a common ancestral tumor clone that has progressed over time and space. Further characterization of tumor evolution in advanced bilateral lung-limited NSCLC is warranted. Clinical trial information: NCT05671887 . Case Tissue NGS by timepoints and region (Dx/Explant/Recur) Shared variants 1 RML / Bilateral Exp / RUL *PIK3CA E110del, *SETD2 D2112fs, GRM3 R331C 2 LUL / Bilateral Exp / Pelvic bone *EGFR G719A 3 LLL / Bilateral Exp / T9 vertebra *KRAS G12D 4 Bilateral Exp / Pelvic bone *ETV6 splice, KMT2D L1443Q 5 Bilateral Exp *KRAS G12V, STK11 (*A200fs, *314fs) 6 LUL / Bilateral Exp *BAP1 Q595*, *GNAS R201C 7 Bilateral Exp *EGFR Ex19del 8 Bilateral Exp / LLL *KRAS G12V 9 Bilateral Exp EGFR (*Ex19del, C797S), LRP1B R531H, PTPRT I320N, SF3B1 A368V Additional Cases: Case 10 (Bilateral Exp): *KRAS G12C, *ARID1A Q507*, *ATM c.1A>T, *ATRX W1572*, *STK11 W332*, CHD2, MALT1, PHLPP1, PTPRT, RICTOR, SEMA3C, TAF1 (W457L, D169Y). Case 11 (RLL, Bilateral Exp): *MET Ex14 skip, *TP53 V143M, ERBB3, NF2. Asterisk (*) indicates oncogenic/likely oncogenic variants defined by OncoKB . Abbrev: Dx, diagnosis; Exp, explant; Recur, recurrence.
Publication pathways and impact metrics of African cancer research presented at major international conferences.
e23391 Background: Despite bearing a disproportionate share of the global cancer burden, African researchers remain under-represented in global oncology publications. We evaluated publication outcomes and scientific impact of cancer research abstracts presented by African researchers at major international oncology conferences. Methods: We conducted a retrospective analysis of abstracts presented by African researchers in the American Society of Clinical Oncology (ASCO), European Society for Medical Oncology (ESMO), and African Organization for Research and Training in Cancer (AORTIC) meetings over a seven-year period. Abstracts were tracked for subsequent peer-reviewed publication using PubMed and Google Scholar. Journal impact factors were obtained from Journal Citation Reports and categorized as high (≥10), medium (3–9.9), or low ( < 3). Citation counts were categorized as high ( > 30), medium (11–30), or low (≤10). Logistic regression was used to identify factors associated with publication. Descriptive statistics were used to summarize the data. Chi-square tests were performed to assess the association between last-author nationality and conference venue and publication status. Logistic regression was conducted to identify factors associated with publication. All analyses were performed using STATA v19, and a p -value < 0.05 was considered statistically significant. Results: A total 1,213 abstracts presented by African researchers affiliated with African institutions at major international oncology conferences were identified, the majority from AORTIC (91.1%), followed by ASCO (4.7%) and ESMO (4.2%), of which 333 (27.5%) were subsequently published. Abstracts with non-African last authors had significantly higher publication rates compared with those with African last authors (33.4% vs 16.5%, p = 0.001). In multivariable analysis, non-African last authorship was independently associated with higher odds of publication (adjusted OR 2.7, 95% CI 2.0–3.6; p = 0.001). Conference venue was not significantly associated with publication outcomes. Journal impact factor did not differ significantly by conference or last-author nationality; however, abstracts with non-African last authors were more likely to achieve higher citation counts (p = 0.014). Conclusions: Fewer than one-third of African cancer research abstracts progress to full publication, with authorship position strongly influencing dissemination and impact. Addressing structural inequities in authorship, mentorship, and publication support is essential to advance health equity in global oncology.
Intra-arterial chemotherapy for retinoblastoma: Treatment outcomes and globe preservation—A tertiary cancer center experience in India.
e12622 Background: Intra-arterial chemotherapy (IAC) represents an important globe-preserving strategy for advanced retinoblastoma. However, comprehensive real-world data from Indian centers examining IAC across varied clinical scenarios remain scarce. We present treatment outcomes, recurrence rates, globe preservation success, and safety profiles from our institutional experience. Methods: We conducted a retrospective review of pediatric retinoblastoma patients who underwent IAC at our tertiary cancer center in Hyderabad, India, during 2020-2024. Primary end point is Treatment response. Treatment response was classified as complete response (CR), near-complete response (NCR), partial response (PR), poor response, or no response. Secondary endpoints included response rates, disease recurrence, enucleation frequency, globe preservation, and adverse events. Results: The cohort comprised approximately 160 patients with over 200 treated eyes. Median age at IAC initiation ranged from 20-70 months, with bilateral involvement in 60%. IAC indications distributed as follows: primary treatment (30%), bridge therapy (25%), post-VEC salvage (35%), and post-IAC salvage (10%). The majority presented with advanced disease (IIRC Groups D/E). CR/NCR was attained in approximately 65% of eyes, PR in 25%, and poor/no response in 10%. Response correlated with disease stage: Group B achieved 90% CR/NCR, Group C 70%, Group D 55%, and Group E 30%. Disease recurrence developed in 35-40% of treated eyes, predominantly among Group E and salvage cases. Globe preservation rates by group were: B (100%), C (85%), D (70%), and E (45-50%), yielding an overall preservation rate of 65-70%. Enucleation became necessary in 20-25% of Group D eyes and 40-50% of Group E eyes. Treatment related toxicity remained minimal, with isolated events. No treatment-related deaths occurred. Conclusions: This substantial Indian real-world cohort demonstrates that IAC offers high efficacy with favorable safety across multiple clinical indications in retinoblastoma management. Globe preservation outcomes were excellent for Groups B-D and clinically significant even for Group E disease. While recurrence poses challenges in advanced and multiply-treated eyes, repeat IAC proved viable and effective in appropriately selected patients. Our findings validate IAC as a cornerstone therapy for retinoblastoma care in diverse resource settings. Keywords: Retinoblastoma, intra-arterial chemotherapy, globe preservation, disease recurrence, pediatric oncology, India.
Phase II study of liposomal irinotecan in combination with cisplatin and bevacizumab for high-grade recurrent glioma.
e14059 Background: For patients with high-grade glioma, the most common and aggressive primary brain tumor in adults, disease progression following first-line therapy—maximal safe resection, radiotherapy, and temozolomide—is nearly universal. Treatment options at recurrence remain limited and heterogeneous. This Phase II study evaluated the clinical efficacy of liposomal irinotecan combined with cisplatin and bevacizumab in patients with recurrent high-grade glioma. Methods: Eligible patients were aged 18–75 years with confirmed recurrent high-grade glioma and an ECOG performance status of 0–2. Patients received a regimen of liposomal irinotecan plus bevacizumab every two weeks, combined with cisplatin every four weeks, until disease progression or unacceptable toxicity. The primary endpoint was progression-free survival (PFS). Secondary endpoints included objective response rate (ORR), disease control rate (DCR), overall survival (OS). Results: From 2024 to 2025, 15 patients were enrolled (13 male, 2 female), with a median age of 54 years (range 32–63). The primary endpoint of median PFS has not yet been reached. Among 8 evaluable patients, the ORR was 75.0% (6/8). Grade ≥3 adverse events occurred in 40% (6/15) of patients and included leukopenia (13.3%), thrombocytopenia (6.7%), neutropenia (13.3%), and diarrhea (26.7%). Conclusions: The combination of liposomal irinotecan, cisplatin, and bevacizumab shows promising activity and manageable toxicity in recurrent high-grade glioma, offering new hope for this difficult-to-treat population. This clinical trial is ongoing and more data will be disclosed in the future. Clinical trial information: ChiCTR2500099331.
Different roles of SALL4 and OCT3/4 in the development and evolving processes of benign and malignant germ cells.
e17006 Background: In gonadogenesis, OCT3/4 and SALL4 are expressed in early germ cells and are critical for maintaining their pluripotency by suppressing differentiation. It is important to understand the variable expression of these two biomarkers throughout different stages of embryonic development as well as in tumor development. The primary aim of this study was to analyze SALL4 and OCT3/4 nuclear expression in various stages of fetal testes and examine their potential link to the risk of cryptorchidism in relation to later testicular tumor development. The second aim was to investigate the different expression patterns of SALL4 and OCT3/4 in metastatic germ cell tumors (GCT). Methods: Surgical and autopsy fetal testes ranging from 10 weeks of gestation to 11 months (n=10) were immunochemically stained for SALL4 and OCT3/4. OCT3/4 and SALL4 immunostains were further tested in 12 specimens from 11 patients with metastatic GCT. Results: SALL4 expression was maintained from early gonadogenesis within germ cells and persists regardless of age or malignant transformation. Whereas OCT3/4 expression started to diminish midway through the 3 rd trimester. Expression of OCT3/4 was retained in a case of a cryptorchid testicle, suggesting that continued expression of OCT3/4 may be linked to an increased risk of GCT in individuals with cryptorchid testicles. OCT3/4 expression returned in germ cell neoplasia in situ and undifferentiated GCTs such as seminoma and embryonal carcinoma. Only 2/12 (17%) metastatic GCT were positive for both SALL4 and OCT3/4, but the majority of metastatic GCT (83% 10/12 cases) continued to evolve by losing expressions of OCT3/4 and maintaining expression of SALL4 (Table 1). Conclusions: Our data demonstrates bridging associations from OCT3/4 and SALL4 expressions in embryonal testis development to their presence in metastatic GCT, indicating a continuous evolving process from benign germ cells to malignant GCT. SALL4 is an important marker for identifying GCT in metastatic sites, whereas metastatic GCTs often lose OCT3/4 expression during progression. SALL4 and OCT3/4 expressions in metastatic germ cell tumor (GCT). Age (years) Original tumor Diagnosis Metastatic location SALL4 OCT3/4 Follow-up period 1 50 Unknown Stage III GCT Peritoneum 3+ 0 NA 2 56 Seminoma Stage III GCT Cervical lymph node 3+ 0 4 years 3 61 Mixed GCT Stage III MGCT Cervical lymph node 3+ 3+ 4 years 4 26 Seminoma Stage III GCT Peritoneum 3+ 0 3 years 5 56 Mixed GCT with dominant seminoma Stage III GCT Gallbladder 3+ 0 NA 6 42 Unknown Stage III seminoma Peritoneum 3+ 3+ 3 years 7 49 Testicular GCT NOS Stage III yolk sac tumor Inguinal lymph node 3+ 0 2 years 8 63 Unknown Stage III GCT Adrenal 3+ 0 9 months 9 71 Unknown Stage III GCT Liver 3+ 0 8 months 10a 50 Unknown Stage III GCT duodenalampulla 3+ 0 8 months 10b 50 Unknown Stage III GCT Liver 3+ 0 8 months 11 68 Unknown Stage III GCT Kidney (12p+ by FISH) 3+ 0 4 months
Genotype-directed targeted therapy combined with HAIC and tislelizumab for microsatellite-stable colorectal cancer liver metastasis refractory to multiple-line systemic therapy (proof-of-concept SALVLIVE trial): An interim analysis.
e15545 Background: The prognosis of microsatellite-stable (MSS) colorectal cancer liver metastasis (CRCLM) refractory to systemic therapy is dismal. Previous studies had showed the efficacy of hepatic arterial infusion chemotherapy (HAIC) for CRCLM. Thus, this trial assesses the proof-of-concept for combining genotype-directed targeted therapy, HAIC, and tislelizumab in heavily pretreated MSS-CRCLM. Methods: This prospective, open-label, single-center trial screened participants with MSS-CRCLM refractory to multiple-line systemic therapy. Participants received the combination therapy of HAIC, tislelizumab (200 mg intravenously before HAIC on day 1), and targeted therapy every 4 weeks. The targeted therapy was determined based on genotype: fruquintinib (3 mg/day on day 3 - 21) was used for KRAS/NRAS/BRAF/EGFR mutation-type or failure of cetuximab in past 3 months (Arm A), while cetuximab (500 mg/m 2 ) was used for KRAS/NRAS/BRAF/EGFR wile-type without the usage of cetuximab in past 3 months (Arm B). The primary endpoint was 6-month progression-free survival (PFS) rate. PFS, overall survival (OS), hepatic PFS (HPFS), objective response rate (ORR), disease control rate (DCR), and safety were also investigated. Results: Between February 2024 and October 2025, 59 participants (57.4 ± 9.2 years old, 39 male) were enrolled (37 in Arm A and 22 in Arm B). The 6-months PFS rate was 51.7% (Arm A: 39.7%, Arm B: 67.5%), with the median PFS of 6.1 months (95% confidence interval [CI]: 5.524 – 6.676). The median OS and HPFS were 17 (95% CI: 14.969 – 19.031) and 8.1 months (95% CI: 7.402 – 8.798), respectively. The ORR was 64.4% (mRECIST)/47.5% (RECIST1.1), with a DCR of 96.6%. The incidence of ≥ grade 3 adverse events (AEs) was 25.4%. The most common ≥ grade 3 AEs in Arm A was elevated alanine transaminase (3/37, 8.1%), hyperbilirubinemia (3/37, 8.1%), and abdominal pain (3/37, 8.1%); and the most common ≥ grade 3 AEs in Arm B was thrombocytopenia (2/22, 9.1%). Conclusions: These interim results provided clinical proof-of-concept for the genotype-directed combination of targeted therapy, HAIC, and tislelizumab as a viable salvage strategy for heavily pretreated MSS-CRCLM. The trial is still ongoing, and the results will be updated in the future. Clinical trial information: NCT06199232 . Endpoints in Arm A and Arm B. Total Arm AFruq+HAIC+Tisl Arm BCet+HAIC+Tisl 6-month PFS rate 51.7% 39.7% 67.5% PFS 6.1 months(95% CI: 5.524 – 6.676) 5.7 months (95% CI: 5.037 – 6.363) 7.4 months (95% CI: 4.799 – 10.001) OS 17 months(95% CI: 14.969 – 19.031) 15.7 months (95% CI: 7.983 – 23.417) Not reach HPFS 8.1 months(95% CI: 7.402 – 8.798) 7.8 months (95% CI: 5.566 – 10.034) 8.2 months(95% CI: 7.094 – 9.306) ORR (mRECIST/RECIST1.1) 64.4% /47.5% 62.2% /37.8% 68.2% /63.6% DCR 96.6% 97.3% 95.5% Incidence of ≥ grade 3 AEs 25.4% 29.7% 18.2%
ADELA: A double-blind, placebo-controlled, randomized phase 3 trial of elacestrant (Ela) + everolimus (EVE) versus elacestrant + placebo in ER+/HER2− advanced breast cancer (aBC) patients with <i>ESR1</i> -mutated tumors progressing on endocrine therapy (ET) + CDK4/6i.
TPS1154 Background: ET+CDK4/6i is the standard-of-care (SOC) in 1L ER+/HER2- aBC; however, tumors eventually develop resistance. Constitutive activation of the PI3K/AKT/mTOR pathway can contribute to endocrine resistance in breast cancer. ESR1 mutations ( ESR1m ) are a common type of acquired resistance that emerges in 40-50% of patients (pts) in the metastatic setting after prolonged aromatase inhibitor exposure. There is an unmet need for novel therapeutic approaches to overcome resistance mechanisms and improve outcomes in pts with ER+/HER2- aBC with ESR1m tumors progressing after ET+CDK4/6i. Ela is a next-generation oral SERD that binds and degrades ERα. In the Ph3 EMERALD trial, single-agent Ela improved mPFS vs SOC ET in pts with ESR1m tumors (HR 0.55; 95% CI 0.39-0.77; P=0.0005) [Bidard 2022]. Among pts who received prior ET+CDK4/6i ≥12 months, mPFS with Ela was 8.6 vs 1.9 months with SOC ET (HR 0.41; 95% CI 0.26-0.63) [Bardia 2024]. The crosstalk between the ER and PI3K/AKT/mTOR pathways provides a rationale for evaluating Ela+EVE (a mTORC1 inhibitor). In the ELEVATE Ph2 trial (NCT05563220), the combination of Ela 345 mg + EVE 7.5 mg showed a clinically meaningful mPFS of 8.3 months (95% CI, 4.0-10.2) in all pts (N=50) with ER+/HER2- aBC who progressed after ET+CDK4/6i, regardless of ESR1 m status (Rugo, SABCS 2025). Safety was consistent with the known profile of EVE+SOC ET. Methods: ADELA (NCT06382948) is an international, multicenter, double-blind, placebo-controlled, randomized Ph3 trial that compares Ela+EVE vs Ela+placebo in pts who have ER+/HER2- aBC with ESR1m tumors progressing on ET+CDK4/6i. Eligible pts are adults (≥18 years) with ER+/HER2- aBC and centrally confirmed ESR1m who received 1-2 prior lines of ET for aBC and progressed on ET+CDK4/6i for aBC after ≥6 months. Pts receiving CDK4/6i-based adjuvant therapy are eligible if progression occurred after ≥12 months of treatment but <12 months following CDK4/6i completion. Exclusion criteria include prior chemotherapy for aBC and active uncontrolled/symptomatic brain metastases and/or leptomeningeal disease. Pts will be randomized 1:1 to 28-day cycles of Ela 345 mg + EVE 7.5 mg QD or Ela 345 mg + placebo QD until disease progression or unacceptable toxicity. Pts will receive dexamethasone mouthwash during the first 8 weeks. Stratification factors are visceral metastases (yes vs no) and duration of prior CDK4/6i therapy (≥12 vs <12 months). Primary objective is PFS assessed by BICR. Secondary objectives are investigator-assessed OS, PFS, ORR, CBR, DoR, TTR, best percentage change in tumor burden, safety, HRQoL. Planned enrollment is 240 pts. Recruitment is ongoing across Spain, France, Greece, Italy, Germany, Austria, Czech Republic, United Kingdom, and Brazil. Clinical trial information: NCT06382948 .
Impact of volunteer-integrated palliative home visits on psychological well-being among metastatic breast cancer patients in northern Malaysia.
12090 Background: Metastatic breast cancer patients face overlooked psychosocial challenges, mitigated through home visits. Universiti Sains Malaysia's volunteer-integrated palliative home visits bridge access gaps, support mental health (SDG 3), and foster partnerships (SDG 17). As Malaysia's most common female cancer (2017–2021: 16,430 cases, 22.2% stage IV), breast cancer highlights early palliative integration, yet local home-based outcomes remain unstudied. Methods: This prospective cross-sectional study (October 2024 to January 2026) in Northern Malaysia (Penang, Kedah) evaluated the impact of volunteer-integrated palliative home visits on Psychological Well-Being (Ryff 42-item scale, baseline and after 12 weeks). Healthcare professionals provided symptom and psychosocial care, while trained volunteers focused on emotional and practical support. Mixed teams (professionals and volunteers) conducted visits 1 and 6, with flexible staffing for intervening visits. Of 94 recruited patients, 81 completed the study and domain changes analyzed via descriptives, paired t-tests, and Cohen's d. Results: Participants were predominantly female (99%), aged 40 to 49 years (34.6%), low-income B40 group (64%) and multi-racial (78% Malay, 15% Chinese, 7% Indian). All had ECOG 0-3 at baseline, but 38% in ECOG 4 post-intervention. High-score groups (>30 on Ryff's subscale sums) increased for positive relations (70% to 95%), environmental mastery (47% to 69%), self-acceptance (45.5% to 68%), and personal growth (70% to 85%), but declined for autonomy (55.6% to 54.3%) and purpose in life (74% to 35%). Paired t-tests (Table 1) revealed significant small-to-moderate improvements in environmental mastery (p<0.001, d=0.46), positive relations (p<0.001, d=0.41), self-acceptance (p=0.042, d=0.23), and purpose in life (p=0.016, d=0.27). Changes in autonomy and personal growth were not statistically significant. Purpose-in-life declines reflect disease progression. Conclusions: The volunteer-integrated palliative care significantly improved environmental mastery, positive relations, self-acceptance, and purpose in life among breast cancer patients, despite ECOG deterioration. Findings support that the home visit enhanced psychological well-being amid disease progression, though refinement is needed for domains with non-significant changes. Paired t-test results for Ryff Psychological Well-Being (PWB) domains (pre- vs. post-intervention). PWB domains Mean Difference (SD) t(80) p value Cohen’s d Effect Size Autonomy 1.10 (6.51) 1.52 0.133 0.17 Small Self-Acceptance 1.99 (8.66) 2.07 *0.042 0.23 Small Positive Relations with Others 3.00 (7.31) 3.7 *0.000 0.41 Small to Medium Personal Growth 1.07 (8.75) 1.11 0.272 0.12 Small Environmental Mastery 3.27 (7.10) 4.15 *0.000 0.46 Medium Purpose in Life 1.59 (5.84) 2.46 *0.016 0.27 Small
Insights through the renal lens: Clinical outcomes of newly diagnosed multiple myeloma patients with and without renal impairment in a resource-limited setting.
e19531 Background: Renal impairment (RI) is a frequent and clinically significant complication of newly diagnosed multiple myeloma (NDMM), associated with increased tumor burden, treatment limitations, and inferior survival. However, real-world data on its prognostic impact and dynamic renal recovery from resource-limited settings remain scarce. Methods: We conducted a retrospective cohort study of NDMM patients diagnosed between December 2010 and March 2023 at Aga Khan University, Pakistan. RI was defined as estimated glomerular filtration rate (eGFR) <60 mL/min/1.73 m² and/or serum creatinine >2 mg/dL at diagnosis. Patients were stratified by RI status. Clinical, laboratory, treatment, and survival data were extracted. Renal recovery was defined as ≥25% improvement in eGFR at 6 months. Survival analyses were performed using Kaplan–Meier methods, and logistic regression identified predictors of disease progression. Results: Among 218 NDMM patients (median age 60 years; 66.5% male), 101 (46.3%) had RI at diagnosis. RI patients had significantly higher creatinine, β2-microglobulin, calcium, and free light chain levels, and were more frequently R-ISS stage III (p<0.001). Induction regimens were comparable between groups, with CYBORD and THAL-DEX most commonly used. Median baseline eGFR was markedly lower in RI patients (23 vs. 78 mL/min, p<0.001) and remained significantly reduced at 6 months. More than half of RI patients (53.5%) achieved ≥25% improvement in eGFR. Mean overall survival (OS) was 92 months in RI versus 111 months in non-RI patients, and mean progression-free survival (PFS) was 59 versus 104 months, respectively. Persistent eGFR <60 mL/min at 6 months independently predicted disease progression (OR 2.63, p<0.001). RI patients with ≥25% eGFR improvement had OS comparable to non-RI patients. Conclusions: Renal impairment at diagnosis is associated with higher disease burden and inferior survival in NDMM. Early renal recovery, reflected by ≥25% improvement in eGFR, mitigates this risk and is associated with outcomes comparable to patients without RI. Dynamic monitoring of eGFR provides a powerful, low-cost prognostic tool and should be integrated into routine management, particularly in resource-constrained settings.
RAINFOL-04 (ENGOT-OV96/GOG-3134): A phase 3, open-label, randomized study of rinatabart sesutecan (Rina-S) plus standard of care (SOC) vs SOC as maintenance treatment after second-line platinum-based chemotherapy in patients with recurrent platinum-sensitive ovarian cancer.
TPS5641 Background: Ovarian cancer (OC) is the eighth leading cause of cancer-related death in women globally. In patients with advanced OC, 70%-80% develop recurrent platinum-sensitive OC (PSOC) within 3 years after initial therapy. Standard treatment for recurrent PSOC includes platinum-based chemotherapy ± bevacizumab (BEV), but in most patients, disease progression occurs within a year. Rina-S is an antibody-drug conjugate targeting folate receptor alpha (FRα) with a novel hydrophilic protease-cleavable linker and a topoisomerase I inhibitor, exatecan, payload. In cohort B1 of the phase 1/2 RAINFOL-01 study (NCT05579366), Rina-S showed encouraging antitumor activity and was tolerated in patients with heavily pretreated advanced OC. We report the design of ENGOT-OV96/GOG-3134 (RAINFOL-04), a phase 3, randomized study of Rina-S + standard of care (SOC) vs SOC as maintenance treatment for patients with recurrent PSOC (NCT07225270). Methods: This phase 3 open-label study will enroll ~528 patients with platinum-sensitive high-grade serous or endometrioid epithelial OC, primary peritoneal cancer, or fallopian tube cancer regardless of their FRα expression. Patients must have disease progression >6 months after the last dose of first-line platinum chemotherapy and have achieved at least stable disease following completion of second-line platinum-based chemotherapy ± BEV. Patients will be randomized 1:1 to receive Rina-S 120 mg/m 2 intravenously every 3 weeks plus SOC or SOC (BEV or observation) until disease progression, unacceptable toxicity, or discontinuation. The primary endpoint is progression-free survival (PFS) per RECIST v1.1 by investigator assessment. The key secondary endpoint is overall survival. Additional secondary endpoints include PFS per RECIST v1.1 by blinded independent central review, PFS2 (time from randomization to the second progression or death), safety, and patient-reported outcomes. The trial is currently recruiting. Clinical trial information: NCT07225270 .
Immunonutrition intervention in driver gene–negative non–small cell lung cancer patients with sarcopenia: A randomized controlled trial.
8587 Background: Cancer-related sarcopenia impairs treatment tolerance and survival in advanced NSCLC. This Phase II trial evaluated IMPACT (arginine and omega-3 enriched formula) in driver gene-negative advanced NSCLC patients receiving chemoimmunotherapy. Methods: Single-center, open-label, randomized Phase II study. Eligible patients: driver gene-negative advanced NSCLC with sarcopenia (AWGS criteria). Randomization 1:1 to IMPACT plus chemoimmunotherapy (Arm A, n=59) versus chemoimmunotherapy alone (Arm B, n=49). IMPACT administered continuously during treatment. Primary endpoint: PFS. Secondary endpoints: body composition (BIA, CT-L3), inflammatory markers (NLR), nutritional status (PG-SGA), safety. Statistical design: 80% power, HR=0.60, alpha=0.05. Dropouts: 7 (11.9%) in Arm A, 5 (10.2%) in Arm B. Modified ITT: 96 patients (52 vs 44). Results: Baseline characteristics balanced. Median PFS: not reached (Arm A) versus 7.0 months (95% CI: 5.2-8.8, Arm B); HR=0.45 (95% CI: 0.23-0.88), P=0.018. Six-month PFS: 82.7% versus 54.5%; 12-month PFS: 59.6% versus 31.8%. NLR change: -1.28 versus +0.16 (P=0.0045). Lean mass change: +0.59 kg versus -1.00 kg (P<0.05). L3 skeletal muscle density: +1.93 HU versus +0.37 HU (P<0.05). PG-SGA improved in Arm A (9.1±5.6 to 6.4±5.1, P<0.001) but not Arm B (8.9±5.8 to 8.2±5.5, P=0.156). Elderly subgroup (≥65y) showed enhanced benefit: HR=0.23 (95% CI: 0.06-0.95), P=0.043. Grade 3-4 AEs: 38.5% versus 52.3% (P=0.156). No IMPACT-related SAEs. Conclusions: IMPACT immunonutrition with chemoimmunotherapy significantly improved PFS, preserved lean mass, reduced inflammation, and enhanced nutritional status in driver gene-negative advanced NSCLC with sarcopenia. Well-tolerated with favorable safety. Phase III evaluation warranted. Clinical trial information: ChiCTR2300078741. Key efficacy outcomes. Endpoint Arm A (Sustagen, n=52) Arm B (Control, n=44) P value edian PFS, months NR 7.0 (5.2-8.8) 0.018 6-month PFS, % 82.7 54.5 - 12-month PFS, % 59.6 31.8 - NLR change -1.28 +0.16 0.0045 Lean mass change, kg +0.59 -1.00 <0.05 L3 SMD change, HU +1.93 +0.37 <0.05
Foundation models to bridge the data scarcity and explainability gap in pancreatic cancer diagnosis.
e16003 Background: Deep learning (DL) for medical imaging has traditionally emphasized standard classification metrics, such as accuracy, sensitivity, and specificity. However, performance gains alone are not sufficient for high variability scenarios such as pancreatic cancer diagnosis using endoscopy ultrasound (EUS). In these settings, clinical confidence requires decisions grounded in anatomically and clinically meaningful image regions. Furthermore, AI-ready EUS data is limited. To mitigate this, we compared conventional Convolutional Neural Networks (CNNs) against a vision transformer-based Ultrasound Foundation Model (USFM) with minimal fine-tuning on a small EUS dataset. Methods: We used an open EUS dataset of 50 subjects (18 pancreatic ductal adenocarcinoma cases and 32 healthy controls). USFM along with two CNN models, ResNet-50 (RN50) and EfficientNet-V2 (EN2), were fine-tuned for classification of cancer vs normal and tested with subject-wise 5-fold cross-validation (CV). Specifically, we fine-tuned the USFM by appending a two-layer classifier decoder to the publicly available encoder and utilized the last attention layer for generating attention maps. We also implemented a quantitative explainability framework based on the spatial overlap between the ground-truth lesion and AI model saliency maps (i.e., Attention Maps for USFM and Gradient-weighted Class Activation Maps for RN50 and EN2) using the Saliency-to-Segmentation Area under the Curve (AUC) metric. Results: USFM had the best performance and stability. It achieved the highest Accuracy (92%) and PPV (0.90), surpassing the best supervised baseline EN2 (Accuracy: 90%; PPV: 0.87). Furthermore, the alignment between model saliency and ground truth improved substantially. USFM's attention maps achieved a mean Saliency-to-Segmentation AUC of 0.85, outperforming the GradCAM results of EN2 (0.70) and RN50 (0.61). Qualitative and quantitative analyses showed that, unlike conventional DL models, transformer-based USFM can consistently concentrate on lesion regions. Conclusions: Results show that USFM, once finetuned with a small amount of downstream data, can improve predictive performance and interpretability. Validated through a 5-fold CV, the USFM not only achieved superior classification metrics with significantly fewer training epochs but also demonstrated substantially better localization of pathological regions in saliency maps. These findings suggest that FMs offer a more explainable method, effectively mitigating constraint of small data sizes while increasing reliability by ensuring decisions are grounded in anatomically and clinically relevant features. Performance comparison of EfficientNetV2, ResNet50, and USFM. Model Acc PPV NPV Sensitivity F1 Score Saliency to Segmentation AUC Score EfficientNetV2 90% 0.87 0.94 0.86 0.85 0.70 ResNet50 87% 0.86 0.89 0.77 0.80 0.61 USFM 92% 0.9 0.93 0.86 0.87 0.85
A large-scale, multi-target deep learning model for virtual genomic and molecular risk profiling in colorectal cancer.
3526 Background: Molecular profiling of tumor biopsies is central to precision oncology, informing treatment selection, prognostication, and disease monitoring. Computational analysis of routine H&E slides offers a complementary, scalable approach that can accelerate biological insights and optimize downstream molecular testing. Recent advances in computational pathology have demonstrated that deep learning models can infer molecular features directly from digital H&E images, referred to as virtual genomics. However, most studies in colorectal cancer (CRC) have been constrained by limited cohort sizes, single-target predictions, and a lack of integration with longitudinal clinical outcomes. Here, we present a transformer-based deep learning framework designed for large-scale virtual genomic and molecular recurrence risk profiling in CRC. Methods: The model was trained on 45,155 patients with matched digital H&E images, whole-exome sequencing (WES), and longitudinal circulating tumor DNA (ctDNA) for molecular recurrence monitoring. We embedded all images using H-optimus-0 and trained a transformer-based multiple instance learning aggregation head for downstream virtual genomic predictions. We trained individual models to predict genes found in the MSK-IMPACT505 cancer gene panel and a single model that predicts all genes simultaneously. Lastly, we trained an independent model to predict the risk of molecular recurrence. All results were validated on an external cohort from The Cancer Genome Atlas (TCGA; n = 422 CRC patients). Results: Individual models predicted 379 mutated genes with an internal area under the receiver operating curve (AUROC) > 0.7. We validated 254 of these genes within TCGA with an AUROC > 0.7. Training a multi-task learning model to predict all genes simultaneously improved the AUROC for 85% of the genes. Based on NCCN guidelines in CRC, we further trained individual models to predict MSI status, BRAF V600E, KRAS G12D/V/G13D, and POLE/POLD1 exonuclease mutations with AUROCs of 0.96, 0.93, 0.84, and 0.86, respectively. Lastly, we developed an image-based molecular recurrence risk model trained directly on longitudinal ctDNA outcomes. The model stratified patients into low, medium, and high risk groups with a concordance index of 0.67. Compared to low-risk patients, the high-risk group exhibited a hazard ratio (HR) of 4.67, while the medium-risk group showed an HR of 1.98 (p-values < < 1E-5). Conclusions: This unified framework demonstrates that virtual genomics from routine H&E histopathology can enable scalable, cost-efficient inference of hundreds of clinically relevant genomic alterations and molecular recurrence risk in CRC. By leveraging universally available diagnostic slides, virtual genomics has the potential to expand access to precision oncology, optimize molecular testing strategies, and support earlier risk stratification.