Citations

ABSTRACT Intratumor heterogeneity arises from ongoing somatic evolution complicating cancer diagnosis, prognosis, and treatment. Here we present TEATIME (es t imating e volution a ry events t hrough s i ngle-ti m epoint s e quencing), a novel computational framework that models tumors as mixtures of two competing cell populations: an ancestral clone with baseline fitness and a derived subclone with elevated fitness. Using cross-sectional bulk sequencing data, TEATIME estimates mutation rates, timing of subclone emergence, relative fitness, and number of generations of growth. To quantify intratumor fitness asymmetries, we introduce a novel metric—fitness diversity—which captures the imbalance between competing cell populations and serves as a measure of functional intratumor heterogeneity. Applying TEATIME to 33 tumor types from The Cancer Genome Atlas, we revealed divergent as well as convergent evolutionary patterns. Notably, we found that immune-hot microenvironments constraint subclonal expansion and limit fitness diversity. Moreover, we detected temporal dependencies in mutation acquisition, where early driver mutations in ancestral clones epistatically shape the fitness landscape, predisposing specific subclones to selective advantages. These findings underscore the importance of intratumor competition and tumor-microenvironment interactions in shaping evolutionary trajectories, driving intratumor heterogeneity. Lastly, we demonstrate that TEATIME-derived evolutionary parameters and fitness diversity offer novel prognostic insights across multiple cancer types.

Abstract Distinguishing high-risk intraductal papillary mucinous neoplasms (IPMNs), pancreatic cysts requiring surgery, from low-risk lesions remains a clinical challenge, often resulting in unnecessary procedures due to limited specificity of current methods. While radiomics and deep learning (DL) have been explored for pancreatic cancer, cyst-level malignancy risk stratification of IPMNs remains untapped. We conducted a multi-institutional study (seven centers, 359 T2W MRI images) to assess the feasibility of AI for predicting IPMN dysplasia grade using cyst-level image features. We developed and compared 2D and 3D radiomics-only, deep learning (DL)-only, and radiomics-DL fusion models, using expert radiologist scoring as a baseline reference. Model performance was evaluated using held-out test data. The radiomics-DL fusion model showed the highest discriminatory ability on the test set (AUC 0.692), outperforming the radiomics-only model (AUC 0.665). Expert accuracy varied widely (37.4%-66.7%). The fusion model integrating deep learning and radiomics features from routine T2W MRI (AUC: 0.692) demonstrates potential for objective, cyst-level risk stratification of IPMNs in a multi-center cohort, outperforming both radiomics-only models and expert radiologists. While performance requires improvement for standalone clinical use, this approach offers a scalable, non-invasive method to potentially improve diagnostic accuracy and reduce unnecessary surgical interventions.

Background: Addressing social determinants of health in patient care helps hospitals better understand the non-medical factors influencing patients’ health outcomes. Objectives: The objective of this study was to evaluate the correlation between hospital characteristics, county determinants, and the systematic recording of health-related social needs among general and surgical acute care hospitals in the United States. It focused on the hospital’s routine collection of data on patients’ health-related social needs, such as transportation, housing, and food insecurity. Design: A cross-sectional retrospective study design was utilized. Methods: All hospitals that completed the American Hospital Association Annual survey (n = 2254) were included in the study. A series of multinomial logistic analyses were conducted. Results: The relative risk of hospitals routinely collecting health-related social needs data is 67% lower in for-profit hospitals and 90% higher in not-for-profit hospitals compared to government hospitals. Hospitals that are part of a system are 1.5 times more likely to routinely collect data on social needs. In addition, counties with higher household income have a statistically significant higher relative risk of hospitals collecting data on social needs, though the magnitude of the difference is small. The relative risk of hospitals collecting social needs data, but not routinely, is 2 times higher in teaching hospitals and 3 times higher among system hospitals. Conclusion: Our research strongly indicates that understanding and addressing these inherent hospital-related factors are essential for effectively integrating social determinants of health into routine healthcare data collection practices. Establishing more robust guidelines and standardization in these practices may enhance hospitals’ ability to document and utilize health-related social needs information, ultimately driving improved patient outcomes and supporting more equitable care.

e18064 Background: Cell-free circulating tumor DNA (ctDNA)-based test has emerged as a promising tool in informing treatment modification in solid tumors. The clinical applications of ctDNA in head and neck squamous cell carcinoma (HNSCC) has been explored and here we report the clinical utilities of ctDNA in disease monitoring in HNSCC. Methods: HNSCC patients treated at our institution with available ctDNA test (Signatera) results between 2022 and 2024 were identified. Clinical information including disease site, stage and imaging results were collected and analyzed. Results: A total of 59 patients were identified. 21 (36%) patients had P16 + oropharyngeal cancer, 16 (27%) had oral cavity cancer, 9 (15%) patients had laryngeal cancer, 7 (12%) had P16-oropharyngeal cancer, 6 (10%) had perinasal sinus, hypopharyngeal or nasopharyngeal cancers. At the time of ctDNA test, 22 (36%) patients completed curative-intent treatment and had no radiographic evidence of disease (NED), 16 (26%) patients were treatment-naïve and had local/regional (LR) disease only, and 22 (36%) patients completed definitive treatment and developed radiographic evidence of LR recurrence or distant (D) metastatic disease. ctDNA was detectable in less than half (44%) of the treatment-naïve patients and the detection rate was even lower (14%) in the subgroup of p16+ oropharyngeal cancer. However, among patients with LR recurrent and distant metastatic disease, majority (77%) had positive ctDNA. ctDNA was undetectable in most of the patients (80%) who completed definitive treatment and were NED radiographically. Among the 3 patients who had detectable ctDNA, 2 were subsequently found to have recurrent disease on images and 1 is still on surveillance (table). Conclusions: Positive ctDNA was detected in most of the HNSCC with LR recurrent or distant metastatic disease, supporting its potential role in treatment response monitoring in this group of patients. Detectable ctDNA preceded radiographic evidence of recurrent disease in a small number of patients, suggesting the possibility of minimal residual disease detection using ctDNA in a subgroup of patients. Treatment history S/P definitive treatment Treatment naive S/P definitive treatment Radiographic evidence of disease (patient number) No (21) Yes (16; LR only) Yes (21; LR only: 10, distant: 11) ctDNA negative 18 9 4 ctDNA positive 3 (recurrence confirmed on subsequent imaging in 2 patients) 7 17

e18031 Background: HPV+ OPSCC has better outcomes than HPV negative OPSCC but more distant metastatic potential. ICM in HPV+ OPSCC are rare. Data regarding risk factors, clinical presentation and outcomes after ICM diagnosis is limited. This study describes clinical characteristics and outcomes of patients (pts) with HPV+ OPSCC who develop ICM. Methods: Retrospective analysis of pts with HPV+ OPSCC and ICM between 2010 - 2024. Demographic, clinicopathologic, and survival data was collected. Skull base recurrences were excluded. Univariate and multivariate analyses used Cox proportional hazard models; continuous variables binarized by receiver operator curves. Survival assessed via Kaplan-Meier method; reported as median (95% CI). Calculations use BlueSky (v10.3.1). Results: Among 1,999 pts with HPV+ OPSCC, 16 (0.8%) had ICM. Pt characteristics and treatment are summarized in Table 1. Primary tumor sites were base of tongue (56%), tonsil (38%), and unknown (13%). Most pts (n=8, 50%) presented with stage IV disease (6 extracranial metastasis (ECM), 1 ICM only, 1 both). Frontal and parietal lobes were the most frequent ICM sites. Neurological symptoms were present in 10 (62%) pts; ICM was incidentally found in 6 (38%) pts. Median overall survival (mOS) from initial diagnosis was 34 months (21 – not reached) and median time from diagnosis to ICM was 24 months (range 1-112). mOS following ICM detection was 10.9 months (4 -18). On multivariate analysis, only surgical resection (hazard ratio [HR] 0.03, p<0.01) and radiotherapy (HR 0.01, p<0.02) were independently associated with post-ICM survival. With median follow up of 100 months, 14 pts (88%) had died; 2 (12%) were alive. Conclusions: ICM from HPV+ OPSCC are rare but devastating with a mOS <1 year from detection. Most pts were symptomatic, had de novo metastatic disease, and had other sites of metastases. Surgical resection and radiation were associated with improved OS. Multicenter efforts to investigate predictors, multiomic sequencing and identification of pt subsets who may benefit from ICM screening are warranted. Baseline characteristics of Pts with ICM in HPV+ OPSCC. Variable Summary statistics (n=16) Median age (range) 58 (34-71) Male 14 (88%) Smoking history 10 (60%) Clinical stage at diagnosis Stage I: 4 (25%) Stage II: 2 (12.5%) Stage III: 2 (12.5%) Stage IV: 8 (50%) Symptoms at ICM diagnosis Incidental: 6 (38%)Neurological: 10 (62%) Seizure (n=3) Motor/Sensory symptoms (n=3) Altered mental status (n=2) Cognitive symptoms (n=1) Dizziness (n=1) ICM site Frontal lobe: 7 (44%) Parietal lobe: 7 (44%) Occipital lobe: 4 (25%) Temporal lobe: 2 (13%) Cerebellum: 2 (13%) Concurrent ECM Bone: 11 (69%) Lung: 9 (56%) Lymph node: 6 (38%) Liver: 5 (31%) Kidney: 1 (6%) Cranial nerve: 1 (6%) Treatment after ICM Radiotherapy: 15 (94%) Chemotherapy: 10 (63%) Immunotherapy: 9 (56%) Surgery: 8 (50%)

e19554 Background: Adding daratumumab to standard multiple myeloma (MM) regimens improves response rates and clinical outcomes. Systematic reviews (meta-analyses) allow further evaluation of its effectiveness in MM patients with high-risk cytogenetic factors but are limited by model assumptions. his study addresses these limitations using a novel synthetic individual patient data (SynIPD) generation algorithm to enable a more robust analysis. Methods: Advances in AI have enabled the creation of synthetic data that mimics the statistical properties of real patient data, empowering aggregated analyses of clinical trials. This study focuses on newly diagnosed MM (NDMM) and relapsed/refractory MM (RRMM) patients with high-risk cytogenetic profiles. Six studies were included: ALCYONE, CASSIOPEIA, and MAIA for NDMM, and CANDOR, CASTOR, and POLLUX for RRMM. Large language models extracted clinical evidence, and SynIPD was generated, incorporating cytogenetic profiles. Random-effects meta-analyses were conducted, and results were compared with pooled analyses using synthetic data. Results: Synthetic data were generated for three studies for NDMM (ALCYONE, CASSIOPEIA, MAIA) and three studies for RRMM (CANDOR, CASTOR, POLLUX). The generated hazard ratios (HRs) for each study closely aligned with reported HRs for high-risk patients, with a maximum relative error of less than 2%. By pooling these datasets, we obtained synthetic datasets for NDMM and RRMM patients. Stratified Cox regression analyses yielded HRs (95% confidence interval) of 0.764 (0.529-1.104) for NDMM and 0.431 (0.294-0.613) for RRMM. Random-effects meta-analyses showed HRs (95% confidence interval) of 0.764 (0.529-1.104) for NDMM and 0.425 (0.287-0.621) for RRMM. Compared with traditional meta-analysis techniques, the synthetic data approach provides additional clinical insights that were not available in the original publications. For instance, the clinical results for the three NDMM studies are summarized as KM curves in Figure 1, which were not reported in the original studies. Conclusions: Our results suggest that the addition of daratumumab to standard regimens improves PFS for RRMM patients with high-risk cytogenetic factors, although its benefit for NDMM patients is not statistically significant. Pooling synthetic data enhances the robustness of conventional meta-analysis outcomes. The identification of HRs and corresponding confidence intervals through SynIPD improves the credibility of our proposed method, offering a reliable approach to evaluating the efficacy of daratumumab in high-risk patients. Since SynIPD includes individual covariate information, it provides deep clinical insights beyond merely enhancing meta-analysis. In other words, any statistical results derived from the original IPD can be effectively approximated using SynIPD.

8545 Background: Brain metastases (BM) correlate with poor prognosis, occurring in 10% of non-small cell lung cancer (NSCLC) patients (pts) at diagnosis and up to 40% over the disease course. Immunotherapy (IO) with or without chemotherapy has become the new standard of care for stage IV NSCLC. However, the survival benefit of IO in pts with BM remains unclear as most studies are limited to small sample sizes or highly selected pts with asymptomatic or treated BMs. This study represents the largest real-world data analysis evaluating the survival benefits of IO in stage IV NSCLC pts with BM. Methods: Demographics, clinical features, and survival were analyzed in stage IV NSCLC pts with BM from the National Cancer Database (NCDB) from 2014-2020. A multivariate Cox proportional hazards modeling assessed factors impacting mortality. Results: Of 204,249 pts with stage IV NSCLC, 30% had BM. The mean age was 68 years, with 54% male. Most pts were White (82%), followed by Black (12%) and Asian (3.1%). Government insurance covered 70% of pts, and 26% had private insurance. Adenocarcinoma was the predominant histology (62%), followed by squamous cell carcinoma (SCC) (20%). Liver and bone metastases were observed in 19% and 42% of pts, respectively. Among pts with BM, 18% received immunotherapy, 53% received chemotherapy, 66% received whole-brain radiation, and 18% received limited-brain radiation. Multivariate Cox analysis showed that pts receiving IO had a 46% lower mortality risk compared to not receiving IO (HR: 0.54, 95% CI: 0.51–0.56, p < 0.001), demonstrating the independent benefit of IO in pts with BM, regardless of brain radiation or chemotherapy. Females had a lower mortality risk than males (HR: 0.88, 95% CI: 0.85–0.91, p < 0.001). Asian (HR: 0.71, 95% CI: 0.64–0.79), Hispanic (HR: 0.76, 95% CI: 0.65–0.89), and Black pts (HR: 0.88, 95% CI: 0.84–0.93) had improved survival as compared to White. Pts with private insurance have lower mortality risk (HR: 0.94, 95% CI: 0.90–0.98), compared to lack of insurance (HR: 1.19, 95% CI: 1.08–1.31). SCC was linked to worse survival (HR: 1.28, 95% CI: 1.22–1.35). Conclusions: IO significantly improves survival in pts with NSCLC with BM, regardless of brain radiation therapy or chemotherapy. However, survival disparities based on histology, insurance status, and demographic factors persist, highlighting the need for more equitable treatment strategies.

6106 Background: Treatment options for pts with R/M SGC are limited. Responses to chemotherapy (CT) are low with high toxicity and responses to immune checkpoint inhibition (ICI) are <10%. Pemetrexed (PTX) is safe and tolerable with responses reported in pts with R/M SGC. 1 Given enhanced responses with PTX and pembrolizumab (PMB) for lung cancer, we hypothesized that PTX and PMB will have activity for SGC. Herein we present the efficacy results in pts with non-adenoid cystic carcinoma (ACC). Methods: MC200708 is a single arm phase II study of PTX and PMB in pts with R/M SGC (NCT04895735). Pts were treated in 2 cohorts: ACC (cohort A) and non-ACC (cohort B). Key eligibility criteria: ≥18 years, ECOG 0-1, measurable disease. Prior ICI and/or PTX was allowed. Key exclusion criteria: serious comorbidities, autoimmune disease, and brain metastases. Simon’s 2-stage design was used for each cohort. Primary endpoint was overall response rate (ORR). Secondary endpoints: progression free survival (PFS), overall survival (OS), and toxicity. All pts received PTX 500 mg/m2 IV + PMB 200 mg IV q3 weeks until progression or treatment intolerance. Imaging was q3 cycles. Results: 25 pts were enrolled Aug 2021-Feb 2024; 1 cancelled prior to treatment. Of 24 eligible and treated pts, median age was 59.5 years (46-77), 66.7% male, performance status 0 (62.5%) or 1 (37.5%). Histologies were salivary duct carcinoma (SDC, 11), acinic cell carcinoma (8), mucoepidermoid carcinoma (MEC, 3), myoepithelial carcinoma (1), and carcinoma NOS (1). 7 pts had no prior therapies (29.2%). The remaining pts had 1 (33.3%), 2 (29.2%), or ≥ 3 lines of therapy (8.3%). 83.3% of pts had no prior ICI. Median cycles of treatment was 8 (2-34), and duration of response was 12.5 months (4.1-20.6). 9 of 24 pts had a confirmed partial response (PR) for ORR of 37.5% (CI: 18.8-59.4). PRs were seen in 7 pts with SDC (7 of 11, 63.6% PR), 1 with MEC, and 1 with acinic cell. Stable disease (SD) was seen in 5 (20.8%), progressive disease in 8 (33.3%), and 2 didn’t have post-baseline imaging, but were considered non-responders per protocol. Clinical benefit rate was 58.3% (CI: 36.6-77.9). With median follow-up of 12.0 months (2-34.7), median OS is 20.7 months (CI: 17.3-not reached) and median PFS is 6.2 months (CI: 2.1-17.3). 1-year OS and PFS rate is 76.3% (95% CI: 59.9-97.2) and 36.1% (CI: 21.01-62.2), respectively. Common toxicities were grade 1 fatigue and nausea. 6 pts (25.0%) had ≥1 grade 3-4 toxicity possibly related to treatment, mostly hematologic. The 4 pts with grade 3-4 non-heme toxicity had grade 3 fatigue, rash, and heart failure, and 1 grade 4 hypokalemia. Correlative studies investigating biomarkers of response are underway. Conclusions: PTX and PMB has activity in pts with R/M SGC, with promising responses in pts with SDC and median response duration of 1 year. 1 Viscuse et al. Head Neck 2019;41(6):E99-103. Clinical trial information: NCT04895735 .

TPS5118 Background: The median overall survival of patients with mCRPC remains under 2 years even with newer therapies. Xaluritamig, an XmAb 2+1 T-cell engager that targets the six-transmembrane epithelial antigen of prostate 1 (STEAP1), facilitates lysis of STEAP1-expressing cancer cells, such as those in advanced prostate cancer. In a first-in-human study, xaluritamig demonstrated encouraging efficacy and a manageable safety profile for patients with mCRPC refractory to standard of care therapies (Kelly WK, et al. Cancer Discov. 2024;14(1):76-89). Methods: XALute is a randomized, multicenter, open-label, phase 3 study to evaluate the efficacy and safety of xaluritamig vs cabazitaxel or second ARDT in men with mCRPC previously treated with taxane chemotherapy. Enrollment in the control arm treatments will be split evenly between cabazitaxel and second ARDT. Stratification factors include LDH ≤ or > 260 IU/L, liver metastases (Y/N), prior prostate-specific membrane antigen radioligand therapy (PSMA-RLT) (Y/N) and the intention to treat with cabazitaxel or ARDT switch. Approximately 675 patients will be enrolled. Participants will be randomly assigned in a 2:1 ratio to xaluritamig monotherapy or standard care. Participants will receive treatment until radiographic disease progression per Prostate Cancer Clinical Trials Working Group 3 (PCWG3), unacceptable toxicity, initiation of other anticancer therapy, withdrawal of consent, death, or end of study as determined by the sponsor. The primary efficacy endpoint is overall survival. The key secondary efficacy endpoint is radiographic progression-free survival per PCWG3 by blinded independent central review. Key inclusion criteria are pathological/cytological confirmation of prostate adenocarcinoma; mCRPC with at least one metastatic lesion; evidence of progressive disease; prior treatment with at least one ARDT; one taxane therapy in the mCRPC setting, and ongoing androgen deprivation with serum testosterone levels (<50 ng/dL or <1.7 nmol/L). Prior treatment with PSMA-RLT, poly ADP-ribosylation inhibitors, and immune checkpoint inhibitors are permitted. Exclusion criteria include prior STEAP1-targeted therapy, any anticancer therapy within 4 weeks prior to first dose of study treatment (not including androgen deprivation therapy), prior PSMA-RLT within 2 months of first dose of study treatment unless less than 2 cycles received, and prior radionuclide therapy (radium-223) within 2 months of first dose of study treatment To mitigate risk of cytokine release syndrome, xaluritamig will be administered with step dosing. Cabazitaxel or second ARDT will be administered according to regional prescribing information. Funded by Amgen Inc. Clinical trial information: NCT06691984 .

7547 Background: Minimal residual disease (MRD) has been recently accepted by the Food and Drug Administration (FDA) as an endpoint for accelerated approval in Multiple Myeloma. However, emerging data from recent trials were not included in previous analyses. While literature-based meta-analyses on the correlation between MRD and approved clinical outcomes typically require extensive manual review, leveraging AI with expert-in-the-loop validation can efficiently generate reliable evidence from comprehensive clinical studies with up-to-date outcomes. Methods: An AI-assisted framework was developed to identify relevant studies and extract critical information via two independent objectives. The first objective examined trial-level associations, modeling treatment effects on MRD and clinical endpoints across patient populations using weighted least squares, with association strength measured by coefficients of determination (R²) and 95% confidence intervals (CIs). The second objective analyzed individual-level associations using synthetic individual patient data (SynthIPD) generated from the published Kaplan-Meier plots and summary statistics of patient subgroups. Results: AI-assisted screening identified eligible studies (>50 patients per treatment arm) reporting progression-free survival (PFS), overall survival (OS), and MRD-negative complete response rates (MRD-CR rate) using multi-parameter next-generation flow cytometry or sequencing methods (sensitivity threshold ≥10⁻⁵), expanding previous analyses from 15 to 20 two-arm studies. Trial-level analysis demonstrated an R² of 0.69 (95% CI 0.50–0.89) for PFS log hazard ratio versus MRD-CR rate log odds ratio. Analysis of synthetic individual data from Kaplan-Meier curves using a novel digitization method yielded a global odds ratio of 7.28 (95% CI 5.60–8.95) for individual-level correlation between MRD-CR rates and PFS outcomes. Conclusions: This study validates MRD-CR rate as an endpoint for accelerated approval in MM through rapid AI-assisted literature review and synthetic individual patient data. The findings demonstrate moderate correlation between MRD-CR rate and median PFS at both trial and individual levels, consistent with previous literature but incorporating additional eligible studies. The novel SynthIPD approach presents an efficient alternative to traditional data-sharing methods while maintaining analytical robustness. These results align with current Oncologic Drugs Advisory Committee (ODAC) surrogacy analysis methods and support the utility of MRD assessment in MM clinical trials.