Browse Articles
Discover research articles across all indexed journals
Impact of pleural effusion on mortality and hospitalization in stage IV <i>EGFR</i> -mutated NSCLC treated with osimertinib: A real-world analysis.
e20646 Background: Pleural effusion (PE) is common in advanced non–small cell lung cancer (NSCLC) and may reflect aggressive disease biology. However, its prognostic significance among patients with stage IV EGFR-mutated NSCLC treated with osimertinib in real-world settings remains incompletely defined. We evaluated long-term mortality and hospitalization outcomes in propensity-matched cohorts with and without PE. Methods: Using the TriNetX US Collaborative Network, adults with stage IV EGFR-mutated NSCLC treated with osimertinib were identified. Stage IV disease was defined by the presence of distant metastatic ICD-10 codes. Inclusion criteria required confirmed EGFR mutation and osimertinib exposure. Patients with non-malignant causes of pleural effusion, end-stage renal disease, heart failure, or prior exposure to other EGFR tyrosine kinase inhibitors were excluded. Patients were stratified by presence versus absence of PE using diagnostic and procedural codes. Cohorts were matched 1:1 using propensity scores based on age at index, sex, race, and ethnicity. Outcomes included all-cause mortality and hospitalization, assessed at 1-, 3-, and 5-year time points. Risk ratios (RR), odds ratios (OR), and Kaplan–Meier survival analyses were performed. Results: A total of 1,061 patients were included in each cohort after application of inclusion and exclusion criteria and 1:1 propensity score matching. Baseline characteristics were well balanced (mean age ~67 years; ~68% female). Mortality was consistently higher in patients with PE compared with those without PE at 1 year (23.9% vs 13.2%; RR 1.81; OR 2.07; HR 2.01), 3 years (36.6% vs 23.8%; RR 1.54; OR 1.85; HR 1.88), and 5 years (38.8% vs 26.4%; RR 1.47; OR 1.77; HR 1.84), all p <0.001. Hospitalization rates were also significantly higher in the PE cohort at 1 year (31.0% vs 18.9%; RR 1.64; OR 1.92), 3 years (37.3% vs 26.0%; RR 1.44; OR 1.69), and 5 years (38.5% vs 27.7%; RR 1.39; OR 1.63), all p <0.001. Survival curves diverged early and remained separated across all time horizons. Conclusions: In this large real-world analysis of stage IV EGFR-mutated NSCLC treated with osimertinib, the presence of pleural effusion was associated with persistently higher mortality and hospitalization rates at 1, 3, and 5 years. Pleural effusion represents a high-risk clinical phenotype despite targeted therapy and may warrant closer monitoring and intensified supportive and oncologic strategies. Propensity-matched demographics. Characteristic Pleural Effusion (n = 1,061) No Pleural Effusion (n = 1,061) Age at index, mean ± SD (years) 67.2 ± 12.0 67.7 ± 11.2 Female sex, n (%) 722 (68.1) 731 (68.9) Male sex, n (%) 339 (31.9) 330 (31.1) Race, n (%) White 617 (58.1) 622 (58.6) Asian 263 (24.8) 262 (24.7) Black or African American 89 (8.4) 92 (8.7) Ethnicity, n (%) Not Hispanic or Latino 838 (79.0) 839 (79.1) Hispanic or Latino 62 (5.8) 56 (5.3)
Genetic associations with chemotherapy-induced nausea and vomiting: A genome-wide association study.
12050 Background: Nausea and vomiting affect up to 80% of patients treated with chemotherapy, and the costs of healthcare utilization due to chemotherapy-induced nausea are estimated to cost over $6 billion each year in the United States. Existing antiemetic regimens targeting 5-HT3 and NK1 antagonists are suboptimal, as up to 40% of continue to report nausea despite antiemetic treatment. Further, existing risk prediction algorithms have poor discrimination. Methods: In this original research study, we conducted a genome-wide association study of genetic factors associated with patient- and clinician-reported nausea and vomiting at 24 hours after infusion (acute) and 5 days after infusion (delayed) in patients treated with moderately- or highly-emetogenic chemotherapy. Participants were recruited from five ORIEN centers: City of Hope, Moffitt Cancer Center, Ohio State University, Rutgers University, and University of Colorado-Denver. Patient-reported CINV was assessed using the MASCC Antiemesis Tool and clinician-reported CINV was assessed using Common Terminology Criteria for Adverse Events (CTCAE) ratings. The study cohort was divided into a training set (70% of the sample) and a test set (30% of the sample). Machine learning was used to evaluate the predictive performance of a polygenic score along with clinical variables, over and above models with clinical variables only. Areas under the curve (AUC) were used to compare the accuracy of each predictive model. Results: Participants (n=1,271) were 58 years of age on average (SD=13), and most were female (74%), White (97%), with breast cancer (41%). Half of participants (50%) had early stage cancer and most were treated with a guideline-concordant prophylactic regimen (70%). Any acute and delayed self-reported nausea was reported by 42% and 70% of participants, respectively. For acute nausea, one gene, C12orf50 , met the threshold for suggestive significance (P = 4.58 x 10 -6 ). For vomiting, the presence of the G allele at rs9712639 (nearest gene: OR7E23P ) was linked to an increased likelihood of clinician-reported acute vomiting (P = 1.83 x 10 -8 ). A further 11 SNPS reached the threshold for suggestive significance and were associated with both acute and delayed vomiting reported by clinicians; these were mapped to the FTLP18 gene. Polygenic scores did not improve predictive risk models based on clinical factors only. Conclusions: Results of our GWAS indicate that there are different genetic associations for acute and delayed nausea and vomiting, suggesting potentially distinct genetic factors underlying each outcome. However, regardless of outcome, genetic information did not significantly improve models predicting the development and intensity of CINV.
The phase I therapeutic landscape in pancreatic ductal adenocarcinoma: Molecular profiling and emerging strategies.
e15114 Background: Pancreatic ductal adenocarcinoma (PDAC) is a lethal malignancy with limited treatment options. Phase I trials explore novel therapies, yet the investigational landscape and genomic characteristics of enrolled patients (pts) remain poorly defined. We analyzed therapeutic strategies and genomic profiles of pts in early-phase trials to better inform future precision oncology approaches. Methods: This retrospective study included pts with advanced PDAC at a single center. Demographic, clinical, molecular, and outcome data were extracted from electronic health records. Tumor genomic profiling was performed using next-generation sequencing. Investigational agents were classified by mechanism of action. Response was assessed per RECIST v1.1. Overall survival (OS) was calculated from trial consent to death or last follow-up. Responses were compared using chi-square tests, and associations of OS with mechanisms and responses were evaluated using Kaplan-Meier and log-rank analyses. Results: Between Jan 2015 and Apr 2025, 617 pts with advanced PDAC enrolled in phase 1 trials. Median age was 61 years (range 22–87), 44% were female. Median prior lines of therapy were 2 (range 1–6). Genomic profiling revealed alterations of KRAS in 42% (G12D 19%, G12V 11%, G12R 8%), TP53 in 37%, and CDKN2A/B in 19%. Other actionable alterations included BRCA 1/2 mutations (10%), MTAP deletions (3%), ERBB2 amplifications (2%), and BRAF V600E mutations (0.6%). Across 236 unique agents tested, targeted therapies (TT) were 61% and immunotherapies (IO) were 33%. Monotherapy and combinations were 64% and 36%, respectively. Leading mechanisms included RAS pathway inhibition (18%), immune checkpoint modulation (16%), tumor microenvironment targeting (14%), and DNA damage response inhibition (12%). Best responses included complete response (CR, 0.2%), partial response (PR, 7.6%), stable disease (SD, 31.8%), and progressive disease (PD, 60.4%). The objective response rates (CR + PR) were highest for TT (13.9%) followed by combinations (4%), and IO (1.3%) (p < 0.001). Median OS (mOS) for the entire cohort was 5.0 mo (95% CI, 4.5–5.4). Pts receiving TT had highest mOS (5.5 mo) vs. combinations (5.1 mo), and IO (2.8 mo) (p = 0.002). Response correlated significantly with OS (CR: 49.1 mo, PR: 18.0 mo, SD: 7.6 mo, PD: 3.7 mo) (p < 0.001). KRAS alteration status was not significantly associated with OS. Conclusions: This analysis highlights a shift in PDAC phase I trials from cytotoxic therapy toward precision oncology, with targeted therapies demonstrating superior efficacy. However, the low mOS underscores refractory nature of PDAC. Our findings emphasize that comprehensive genomic profiling is essential for precision therapeutic matching and biomarker-driven trial enrollment in PDAC.
Clinical validation of circulating tumor cell synaptophysin and androgen receptor expression in metastatic prostate cancer phenotypes.
5041 Background: Chronic exposure to androgen receptor (AR) targeted therapies can drive lineage plasticity, promoting the emergence of neuroendocrine (NE) prostate cancer (NEPC) and treatment resistance. NEPC is characterized by a phenotypic shift, with loss of AR expression and emergence of markers (e.g., synaptophysin, SYP). To definitively diagnose NEPC, a metastatic biopsy is typically required; however, there are numerous barriers to obtaining a timely tissue diagnosis. As such, there is an unmet need for blood-based assays that can accurately identify treatment-emergent NEPC. In this study, we evaluate concordance between circulating tumor cell (CTC) SYP and AR expression patterns in patients with and without evidence of NE differentiation on tissue histology. Methods: Blood for CTC analysis was prospectively collected from patients with advanced prostate cancer. Patients were retrospectively assigned to either an adenocarcinoma (AR-positive) or NE differentiated (SYP-positive) group based on clinical and patho-genomic factors. Amphicrine PC was included in NE differentiated cohort. The RareCyte system was used to generate slides for CTC-based immunofluorescent (IF) analysis for SYP and AR. Associations between CTC SYP and AR expression, PSA at time of CTC collection, and overall survival (OS) were assessed for patients with ≥ 5 CTCs using Wilcoxon rank-sum test, Fisher’s exact test, and Kaplan-Meier analyses. Results: We enrolled 52 patients with metastatic castration-resistant prostate cancer (mCRPC); 26 were classified as NE differentiated and 26 as AR-positive adenocarcinoma. Median time between biopsy and CTC collection was 10.6 months (IQR 4.1–34.3) and 19.4 months (IQR 3.5–66.4) for the NE differentiated and AR-positive adenocarcinoma cohorts respectively. NE differentiated patients had a significantly higher fraction of SYP-positive CTCs (median 28.6 % vs 0%, p = 0.0084), whereas patients in the adenocarcinoma cohort had a higher fraction of AR-positive CTCs (median 96.7% vs 0.0%, p = 0.0004). Phenotype classification based on high AR or SYP expression ( > 80% of CTCs positive) significantly differentiated cohorts and was strongly associated with tissue histology (p = 0.0048). PSA level at the time of CTC collection was significantly lower in the NE differentiated cohort (median 3.8 vs 486.9 ng/mL, p = 0.0001). There was no difference in OS from time of diagnosis or time of metastasis between groups. Conclusions: SYP and AR CTC expression demonstrated strong concordance with tissue histology and established clinical features of NEPC and AR-positive adenocarcinoma. These early findings support the feasibility of CTC expression profiling to identify NE biomarker expression. However, results are only hypothesis-generating and prospective validation with matched tissue biopsy in a larger cohort with adequate follow up is warranted.
Leptomeningeal disease in ALK-positive NSCLC: Survival impact of third-generation ALK inhibitors.
8644 Background: As survival outcomes for patients with ALK-positive non-small cell lung cancer (NSCLC) continue to improve with successive generations of ALK tyrosine kinase inhibitors (TKIs), leptomeningeal disease (LMD) has emerged as a growing and critical area of unmet need. While next-generation ALK TKIs were developed to enhance central nervous system (CNS) penetration and improve control of parenchymal brain metastases, their role in the treatment of LMD remains less well defined. Methods: We performed a global multi-center retrospective analysis of 141 patients with a diagnosis of ALK-positive NSCLC and radiographically or cytologically confirmed LMD at 5 academic centers across the United States and China (MD Anderson Cancer Center, Zhejiang University, Guandong Lung Cancer Institute, National Cancer Centre Singapore, and Union Hospital at Tongji Medical College) between 2007-2024. Baseline clinical characteristics, treatment history, and outcomes were collected, and subgroup analyses were performed to identify clinical and treatment-related factors associated with leptomeningeal overall survival (LMOS). Results: Of the 141 patients with ALK+ NSCLC, most patients had radiographic LMD (79.4%), while 41.1% had positive CSF cytology and 59.6% were neurologically symptomatic. The median time from metastatic lung cancer to LMD diagnosis was 40.6 months among patients who were previously treated with a 3 rd generation (3G) ALK TKI (lorlatinib), compared to 23.7 months among those without prior lorlatinib treatment (p < 0.001). The median overall survival following LMD diagnosis (LMOS) was 22.5 months. Use of a 3G ALK TKI following LMD diagnosis was associated with improved LMOS in both treatment-naïve patients and those previously treated with earlier-generation TKIs. Among TKI-naïve patients, LMOS was 42.4 months with second-generation (2G) TKI alone, 49.6 months with escalation from 2G to 3G TKI, and not reached with 3G TKI alone (p < 0.001). Among patients previously treated with a 2G TKI, LMOS was significantly improved with 3G TKI (32.6 months) compared with an alternative 2G TKI (7.0 months) or no ALK TKI (6.8 months; p = 0.01). Whole-brain radiotherapy, intrathecal therapy, and VEGF inhibition were not associated with improved LMOS. On multivariable analysis, ECOG performance status < 2 and female sex were associated with prolonged LMOS. Conclusions: This study represents the largest multi-institutional analysis of LMD in patients with ALK+ NSCLC. Highly CNS-penetrant 3G ALK TKIs are associated with delayed LMD development and significantly improved survival following LMD diagnosis, whereas WBRT and intrathecal therapies confer limited benefit. These findings support the need for potent CNS-active ALK TKIs and prospective studies including patients with LMD in ALK-driven NSCLC.
Genome-wide association study of tyrosine kinase inhibitor–induced hepatotoxicity in All of Us.
e15130 Background: Tyrosine kinase inhibitors (TKIs) are widely used as standard therapies for a variety of malignancies; however, clinically significant hepatotoxicity has been frequently reported in patients receiving these agents. Such adverse effects may necessitate treatment interruption or dose modification, thereby imposing a substantial clinical burden. This study aimed to investigate genetic susceptibility to TKI-induced hepatotoxicity by analyzing genetic polymorphisms in patients treated with TKIs. Methods: In this study, a cohort of patients treated with tyrosine kinase inhibitors (TKIs) was constructed using data from the All of Us Research Program. Hepatotoxicity was defined as grade ≥2 liver function abnormalities according to the Common Terminology Criteria for Adverse Events (CTCAE). To identify genetic variants associated with TKI-induced hepatotoxicity, a genome-wide association study (GWAS) was performed. Given the limited sample size, a suggestive significance threshold of P < 1 × 10⁻⁵ was applied. Based on the variants identified in the GWAS, additional multivariable logistic regression analyses and polygenic risk score (PRS) analyses were conducted. Results: Among 811 participants exposed to TKIs, 98 developed grade ≥2 hepatotoxicity, while 713 served as controls. Genome-wide association analysis identified several loci exceeding the suggestive significance threshold (P < 1 × 10⁻⁵), with the strongest association observed at rs72734306 on chromosome 1. Additional suggestive associations were detected at loci near genes involved in immune regulation and endoplasmic reticulum stress–related pathways. In multivariable logistic regression analyses adjusting for age, sex, and concomitant chemotherapy use, several GWAS-identified variants remained significantly associated with TKI-induced hepatotoxicity, whereas clinical covariates were not independently associated with the outcome. Polygenic risk score (PRS) analysis showed a trend toward an increased risk of hepatotoxicity with higher PRS values; however, this association did not reach conventional statistical significance (P = 0.051). Conclusions: These research findings may help predict liver toxicity side effects in patients using TKIs and contribute to personalized treatment.
Genomic correlates of leptomeningeal disease after brain metastasis resection.
2032 Background: Leptomeningeal disease (LMD) remains a frequent and devastating pattern of failure following surgical resection of brain metastases (BM), occurring in up to one-third of patients despite optimal local therapy. Although mechanical disruption of BM tissues may increase the risk of subsequent LMD, the underlying biologic determinants of postoperative LMD remain undefined. This study aims to identify genomic alterations in resected BM and their association with LMD development using the largest craniotomy cohort to date with paired next-generation sequencing and longitudinal CNS outcomes data. Methods: We retrospectively identified patients who underwent surgical resection and MSK-IMPACT next-generation sequencing of one or more BM. LMD was diagnosed using brain and spine MRI and CSF cytology and classified as either classical LMD (cLMD; diffuse coating of meningeal spaces and/or positive CSF cytology), nodular LMD (nLMD; discrete nodular deposits with negative CSF cytology), or both. Fine–Gray subdistribution hazard (sHR) estimated the cumulative incidence of LMD, treating death as a competing risk. Results: Among 1,006 patients, NSCLC (360, 36%), breast (181, 18%), and melanoma (128, 13%) predominated; most patients had single (530, 53%), supratentorial (796, 79%), and large (>2cm, 886, 88%) BM, and underwent postoperative cavity radiotherapy (890, 88%). Median follow-up and OS were 24.8 and 21.5 months, respectively. The cumulative incidence of LMD was 31% (cLMD 18%, nLMD 19%) at 2 years, with the highest rates seen in breast cancer patients (45% LMD; cLMD 28%, nLMD 23%). Median time to cLMD and nLMD were 7.7 and 6.4 months, respectively. Median OS after cLMD and nLMD diagnosis were 5.7 and 12 months, respectively. The most frequently altered genes in resected BM were TP53 (62%), CDKN2A (25%), TERT (23%), KRAS (20%), ERBB2 (12%), PIK3CA (12%), PTEN (12%), and RB1 (11%). In a pan-cancer analysis, alterations in CDH1 (sHR 3.0, q=0.03), EGFR (HR 2.5, q<0.001), GATA3 (sHR 2.2, q=0.03), or RB1 (sHR 1.7; q=0.04) were associated with an increased risk of cLMD; alterations in PTEN were associated with an increased risk of nLMD (sHR 1.8, q=0.02). Within breast and lung cohorts, alterations in RB1 or EGFR, respectively, were associated with an increased risk of cLMD. BM from patients who developed LMD exhibited greater genomic instability, with a higher fraction of the genome altered compared to those who did not (0.47 vs 0.40, p<0.001). Conclusions: LMD is a common and deadly form of progression following BM resection. Distinct genomic features, present in the resected BM specimen, may identify patients at increased risk of LMD. These findings support the development of a clinicogenomic model to select BM patients for intensified surveillance and postoperative care.
Comparative yield of three models to evaluate germline genetic risk in patients (pts) with pancreatic cancer (PCA): A single-institution study.
e22647 Background: ~10% of PCA patients (pts) have a germline pathogenic variant (gPV) in a hereditary CA risk gene. After the POLO trial reported efficacy of PARP inhibitor olaparib in BRCA1/2+ metPAC at ASCO 2019, updated NCCN guidelines recommended universal genetic counseling/testing (GC/GT) in all PAC pts. Various care delivery models have examined approaches to improve rates of GC/GT in PAC. In the current study, we compared 3 models in PAC pts: 1) non-universal GC/GT provider-initiated referral (PR) to genetics; 2) universal mainstream genetic testing (MT) conducted by oncologists in clinic; 3) universal direct scheduling (DS) of all new PAC pts to genetics (no referral). We hypothesized that universal DS would improve uptake of GC/GT and improve yield of gPV identified. Methods: Demographic, GC/GT, and clinical follow-up data were retrospectively extracted from the EMR for 3 unique one-year time periods: 1) PR (11/1/2017-10/31/2018), 2) MT (11/1/2022-10/31/2023) and 3) DS (11/1/2024-10/31/2025). Universal GT for PAC began 11/2019. Summary descriptive statistics and chi-square tests used in data reporting. Results: Under PR model, 117 eligible new PAC pts were evaluated (mean age 69 yrs, 50% F, 19% non-White) and 19% (n = 21) had pretest GC followed by GT. Notably 9% pts were referred to genetics but declined. gPVs in ATM, BRCA1, CDKN2A, RAD51D, MUTYH and CFTRx2 were identified, and 9% had a gVUS. Overall, 84% with a gPV+/gVUS+ had posttest GC. Under MT model, 56% (57/102) PAC pts (mean age 67 yrs, 49% F, 32% non-White) had GT sent by their oncologist: 8% had a gPV (ATM, BRCA2, PALB2, CFTRx3, MUTYHx2, PALB2) and 12% had a gVUS. Overall, 30% pts were referred to genetics, but only 19% pts had GC, and only 50% of those with a gPV+/gVUS+. Nearly 1 in 3 untested pts lacked EMR documentation by oncology explaining lack of GT. Under DS model, 235 pts were EMR routed to genetics. Of these, 60% (n = 141) were appropriate for DS GT (mean age 64 yrs, 50% F, 32% non-White) and 50% (n = 71) were scheduled for GC/GT. Reasons for non-scheduling included: previous GT (23%), GC/GT active decline (33%) and passive decline/non-response (11%), pursuit of care elsewhere (10%) rapid decline/death (7%), or non-PAC diagnosis (17%). gPV were identified in 8%: APC, APC/FH, ATMx3, BRCA1, BRIP1, MSH2, MUTYHx2, PMS2 and 7% had gVUS. 95% pts with gPV+/gVUS+ had post-test GC. Conclusions: Universal GT + direct scheduling improves rates of GC/GT in PAC over the non-universal provider referral model and improves completion of post-test GC in PV+/VUS+ pts over universal mainstream GT; nonetheless, nearly half of eligible PAC pts fail to receive GC/GT with universal direct scheduling. Provider referral (PR) n=117 Universal mainstream (MT) n=102 Universal direct scheduling (DS) n=141 p Had GT 19%* 56% 50%* <0.001* Any GC 19% 19%* 50%* <0.001* Post-test GC for PV+VUS+ 84% 50%* 95%* <0.05* PV n(%) 7 (6%) 9 (8%) 12 (8%) NS VUS 9% 12% 7% NS
Online biographies of gynecologic oncologists: Analysis of linguistic differences based on gender, graduation year, and region.
e17535 Background: Selecting a physician is a universal act for patients. This process has been reshaped by the digital transformation of healthcare. The widespread availability of online information has empowered patients with unprecedented access to data about potential providers. In response, physicians increasingly rely on online professional biographies as a means of engaging prospective patients. These biographies humanize providers and foster relatability for patients navigating healthcare choices. Our objective is to examine linguistic differences in online biographies of gynecologic oncologists across gender, geographic region, and year of graduation. Methods: A list of Gynecologic Oncology fellowship graduates from 2014-2024 was collected from the programs’ websites or requested from fellowship administrators. Their online biographical text was recorded in a spreadsheet, along with their gender, geographic region, and medical school graduation year. The texts were uploaded to Linguistic Inquiry and Word Count software (LIWC-22), which is a validated software that parses text word-by-word to create “summary variables.” These variables describe broader psychologic dimensions of the text. The dependent linguistic variables measured included analytic language, emotional tone, clout, authenticity, and big words. RStudio was used to analyze the LIWC-22 output. A MANOVA, followed by Type I ANOVAs, were conducted to examine these linguistic differences across gender, graduation year, and region. Results: 124 bios were collected for analysis. Significant linguistic differences were found by geographic region, F(3, df ) = value , p < .05. Post hoc analyses revealed that oncologists in different regions significantly varied in their use of analytic language, emotional tone, and big words (all p < .05). The South region exhibited more analytic language than both the Northeast and Midwest, but less emotional language than all other regions. No significant differences were observed by gender or graduation year. Conclusions: Gynecologic oncologists of different genders or ages marketed their practice without significant linguistic variability. This finding deviates from multiple prior studies that noted significant linguistic divergence in other descriptors of male versus female healthcare providers (ie., LORs). Moreover, linguistic differences in biographies varied by region of the physician–particularly in tone, emotional language, and use of big words. With the knowledge that physicians tailor their online biographies to their prospective customers, we can conclude that there is geographic variability in how patients prefer their doctors discuss healthcare. As the digital presence of physicians grows, understanding how to optimize language in patient-facing materials will be increasingly important in patient-physician engagement.
Large language models as decision support in the absence of genomic profiling: A simulated prospective study on early-stage breast cancer.
e12553 Background: Genomic profiling guides adjuvant chemo in early-stage BC but remains inaccessible in resource-constrained settings. We evaluated whether LLMs could optimize oncologist decision-making in the absence of genomic data. Methods: In this simulated prospective, multi-reader study, two board-certified oncologists independently evaluated clinicopathological vignettes of 200 HR+/HER2- early-stage BC cases. Adjuvant treatment recommendations and confidence ratings were documented in two phases separated by a 4-week washout: an initial unaided assessment utilizing only clinicopathological variables, and a subsequent review assisted by the Gemini 3 Pro LLM. We evaluated changes in chemo recommendations, inter-rater agreement, decision confidence, and concordance with NCCN guideline-based recommendations derived from Oncotype DX recurrence scores. Results: LLM assistance increased inter-rater reliability from moderate (κ = 0.473) to substantial (κ = 0.565). This improvement was statistically significant in the node-positive subgroup (n = 56), where the kappa value increased from 0.257 to 0.546 (Δκ = 0.290; p = 0.031). The oncologist with a higher baseline recommendation rate significantly reduced chemo proposals from 62% to 56% (p = 0.045), indicating a treatment de-escalation effect. Physician confidence in decision-making also increased significantly for both oncologists (p < 0.01). However, despite improved consensus and confidence, concordance with the Oncotype DX-based reference standard did not statistically improve for either oncologist (Oncologist 1: Δκ = 0.066, p = 0.124; Oncologist 2: Δκ = −0.084, p = 0.114). Conclusions: In the absence of genomic profiling, LLMs effectively standardized clinical reasoning and mitigated inter-observer variability, serving as a pragmatic decision-support tool to reduce potential overtreatment. While AI enhances physician confidence and consensus, it cannot replicate the biological risk stratification provided by molecular assays. Impact of LLM assistance on key metrics (n=200). Key Metric Unaided Phase LLM-Aided Phase p- value Oncologist 1 chemo rec. (%) 62.0 56.0 0.045 Oncologist 2 chemo rec. (%) 37.0 35.5 0.678 Inter-rater agreement (κ) 0.473 0.565 0.131* High confidence - Onc 1 (%) 39.5 66.5 <0.001 High confidence - Onc 2 (%) 53.0 64.5 0.007 Concordance w/ ref. (κ) - Onc 1 0.253 0.319 0.124* Concordance w/ ref. (κ) - Onc 2 0.461 0.377 0.114* *Bootstrap analysis.
Serum albumin as an early predictor of survival in patients receiving antibody-drug conjugates.
e15041 Background: Decreased baseline and cycle 1 (C1) serum albumin (ALB) predicts shorter overall survival (OS) in pts with NSCLC treated with IgG based immune checkpoint inhibitors (ICIs) but not in pts on ICI + chemotherapy. Decreased ALB may serve as a marker of hyper catabolism associated with cancer cachexia and elevated clearance (CL) of IgG drugs. It is unknown if ALB associates with OS in pts treated with antibody drug conjugates (ADCs), which combine IgG and chemotherapy. In this single-center retrospective analysis, we aimed to assess ALB and OS association. Methods: 601 cancer pts who received ADC, with at least one pretreatment and one on-treatment (C1) ALB value were included. Normal ALB was ≥ 3.5 g/dL per institution cut-off. In a subset of pts with pretreatment ALB measured within 30 days prior to start of ADC (n = 503), ALB percent change was calculated relative to the C1 value. ALB change was categorized as Increase/No change/Decrease and ( > 0%, 0% to –5%, –5% to –10%, < –10%). OS was defined as time from first ADC dose to death/censoring at most recent contact. Cox proportional hazards regression assessed ALB percent change vs. OS, adjusted for ADC and cancer type. Effect modification by ADC and cancer was evaluated with interaction terms (e.g. ADC x ALB change). Kaplan-Meier plots were used to visualize differences in OS. Results: In the full dataset (n = 601), pretreatment and C1 ALB were strongly associated with OS. Pts with normal (n = 490) vs. low ( < 3.5 g/dL, n = 111) C1 ALB had median OS of 31 mo vs. 9.8 mo, respectively (p < 0.0001). Among pts with relevant pre-treatment ALB (n = 503), pts with ALB increase had median OS of 29.3 mo, while pts with 0 – 5%, 5-10% or ≥10% decrease had median OS of 23.4 mo, 17.7 mo, and 12.1 mo, respectively ( p < 0.001). In multivariable analysis, each 10% ALB increase associated with 22% reduction in hazard of death (HR = 0.78, 95% CI 0.67–0.92). Compared with pts experiencing ≥10% ALB decrease, those with smaller decrease or increase had significantly lower hazards (adjusted HRs 0.45–0.49, all p < 0.001); highlighting the strong association of large ALB drop on survival. Borderline and significant interactions were observed comparing ALB (% change) vs cancer type (p = 0.06) and ALB change (3 categories) vs ADC type (p = 0.046), respectively. Conclusions: Early ALB change was independently associated with OS in cancer pts receiving ADCs, suggesting it may serve as an early indicator of mechanisms underlying IgG-based therapy resistance. Given known link between ALB and CL of IgG-based drugs, further investigation into the CL patterns, toxicities and outcomes of pts treated with ADCs is warranted. Age; Male(n), Female (n) 58 (24–89); 91, 510 Breast cancer, n (%) 402 (67) Non-breast, n (%) 199 (33) Advanced stage (3 - 4), n (%) 231 (38) Stage ≤ 2, n (%) 284 (47) Stage unknown, n (%) 86 (14) Trastuzumab Emtansine, n (%) 210 (35) Trastuzumab Deruxtecan, n (%) 147 (25) Enfortumab Vedotin, n (%) 91 (15) Other ADC, n (%) 153 (26)
Tislelizumab (Tisle) combined with POFI (irinotecan, paclitaxel, oxaliplatin, and 5-FU/levoleucovorin) as first-line treatment for advanced gastric/gastroesophageal junction adenocarcinoma (AGC): Update from a single-arm, open-label phase I/II trial (SYLT-023).
4030 Background: Tisle is an anti-PD-1 antibody. The combination of Tisle + XELOX/FP is a first-line treatment of AGC in China. Both irinotecan and paclitaxel have also shown antitumor activity in AGC. This Phase I/II study evaluated the efficacy and safety of Tisle + POFI in this population. Methods: This single-center, single-arm phase I/II study enrolled treatment-naïve patients with pMMR/MSS AGC, aged 18-75, who have at least one measurable lesion according to RECIST 1.1 and an ECOG PS of 0-1. Eligible participants received POFI (irinotecan/paclitaxel (mg/m2): 135/45 (dl #1), 150/45 (dl #2), 135/67.5 (dl #3), and 135/90 (dl #4), oxaliplatin 85 mg/m 2 , levoleucovorin 200 mg/m 2 , and 5-FU 2400 mg/m 2 for 46 hours every 2 weeks) in combination with Tisle 200mg every 2 weeks. The primary endpoint was the objective response rate (ORR) per RECIST 1.1. Results: As of Dec. 23, 2025, 47 patients were enrolled (median age: 58 years, range: 31–72; 26%(12/47) ECOG PS 1; 74% poorly differentiated; 68%(32/47) with ≥2 metastatic sites; 34% with programmed death-ligand 1 combined positive score [PD-L1 CPS] ≥1). The confirmed ORR was 76.6% (36/47), and the disease control rate was 95.7%. Median progression-free survival (PFS) was 11.1 months (95% CI: 8.8, 13.4) (versus 6.9 months (5.7-7.2) for ITT population for RATIONALE 305, ESMO 2023). Median duration of response was 9.7 months (95% CI: 6.7, 12.2). Median overall survival was 16.0 months (95% CI: 12.4, 23.0). All patients experienced treatment-emergent adverse events. The most common grade 3/4 adverse events included neutropenia (53.2%), leukopenia (27.7%), and anemia (21.3%). No new safety signals were identified. Conclusions: Tisle + POFI showed promising efficacy and a manageable safety profile as first-line treatment for HER2-negative, pMMR/MSS AGC, warranting further investigation. Clinical trial information: NCT05319639 .
Role of iron supplementation in the efficacy of immune checkpoint inhibitors (ICIs): A real-world propensity-matched analysis.
11131 Background: Iron plays a crucial role in the induction of ferroptosis mediated by immune checkpoint inhibitors (ICIs) in cancer. However, it plays a complex role in T-cell and macrophage function in the tumor immune microenvironment, which could influence anti-cancer immunity. Currently, the role of iron supplementation in affecting the therapeutic outcomes in patients receiving ICIs is unknown. Methods: A retrospective analysis utilizing TriNetX, a global federated health research network database, of patients receiving ICIs from January 2006 to January 2026, was conducted. Patients were divided into 3 cohorts based on whether they received intravenous (IV) iron (cohort A), oral (PO) iron (cohort B), or no iron supplementation (cohort C), 1 month prior to or up to 3 months after receiving ICIs. Propensity score matching (PSM) was used to balance the cohorts based on age, race, ethnicity, hemoglobin (Hb), ferritin, CRP, albumin, metastatic disease, line of treatment and underlying comorbidities. Kaplan-Meier analyses were used to analyze overall survival (OS). Results: A total of 4464, 5299, and 201,014 patients were identified across cohorts A, B and C, respectively. The most common primary sites of malignancies were the lung and/or bronchus, the gastrointestinal tract (and associated glands), and the urinary tract. Although IV iron recipients had better survival outcomes than PO iron recipients, both groups had worse outcomes than patients receiving no supplementation (Table 1). Similar trends were seen in subgroups receiving anti-PD1 and anti-PDL1. In the subgroup of patients with likely true iron deficiency (Hb≤11g/dl and ferritin<100ng/ml) ( n =508 after PSM), IV iron recipients had better outcomes than PO iron recipients (2Yr OS HR (95% CI): 0.822 (0.691,0.978) p=0.027). In the subgroup of patients having Hb≤11g/dl and ferritin≥300ng/ml, likely anemia of chronic disease ( n =2483 after PSM), iron supplement recipients, whether IV or PO, had worse survival outcomes than patients receiving no supplementation (2Yr OS HR (95% CI): 1.22 (1.137,1.308) p<0.0001). Conclusions: Our study highlights the adverse impact of concurrent iron supplementation in patients receiving ICIs, especially in those without true iron deficiency. When clinically indicated, the IV route is preferred. Clinical studies are planned or ongoing to determine the role of iron supplementation in patients receiving ICIs, given its potential to induce ferroptosis. We suggest careful reconsideration or appropriate stratification of the cohort based on iron deficiency and/or type of iron supplementation in such studies. 2-year overall survival (OS). Comparison groups Number of patients after PSM 2-Year OSHR (95% CI) p-value (log-rank test) IV Iron v/s PO Iron 2762 0.898 (0.83,0.97) p=0.006 IV Iron v/s No Iron 3808 1.199 (1.121,1.283) p<0.0001 PO Iron v/s No Iron 5210 1.405 (1.327,1.488) p<0.0001
Long term outcomes of central neurocytoma in a multi-institutional cohort: The role of radiotherapy and Ki-67.
2089 Background: Central neurocytomas (CNs) are rare, WHO grade 2 brain tumors that can have relatively high recurrence rates following surgical resection. Owing to the rarity of CNs and limited robust data, the optimal use and timing of radiotherapy (RT) remain undefined, particularly in young adults with low mortality. Tumor proliferation index (Ki-67) may also predict CN recurrence rates, but there is no consensus on its utility in clinical practice. The present collaboration between Northwestern University and the Mayo Clinic describes long-term outcomes of CN as it relates to RT timing and Ki-67 level. Methods: Patients established between 1990-2025 with a tissue diagnosis of CN at Northwestern University and the Mayo Clinic were reviewed. Demographic, treatment, and tumor characteristics (histologic/molecular) were recorded. Recurrence free survival (RFS) and overall survival (OS) were calculated according to the Kaplan Meier-method. Results: Fifty-one patients with CN (51% male, 49% female)--with a median age of 26 years and 9.7 years median follow-up--were included. RFS at 2, 5, and 10 years for the cohort was 79%, 65%, and 59% respectively; OS at 10 years for the cohort was 94%, and no deaths were attributed to uncontrolled CN. Subtotal resection (STR) yielded 2- and 5-year RFS of 74% and 65%, while gross total resection (GTR) yielded 2- and 5-year RFS of 90% and 66%. Median Ki-67 of the cohort was 3%; RFS at 2 and 5 years for Ki-67 <3% was 82% and 76%, compared to 76% and 43% for Ki-67 ≥3% (p<0.02). In patients who underwent resection without adjuvant RT (n=35), those with STR had a 2- and 5-year RFS of 61% and 45%, and those with GTR had a 2- and 5- year RFS of 88% and 61%. Additionally, without adjuvant RT, RFS at 2 and 5 years for Ki-67 <3% was 73% and 64% versus 71% and 24% for Ki-67 ≥3%. Among patients without adjuvant RT and Ki-67 ≥3% (n=14), 2-year RFS was 60% after STR and 78% after GTR. Salvage RT after first recurrence (n=16) had a 5-year RFS of 81%, with a median follow-up of 4.1 years. Conclusions: We report one of the largest cohorts of CN with long-term follow-up. STR and/or Ki-67 ≥3% were associated with shorter RFS. Salvage RT at recurrence achieved durable long-term tumor control, and long-term (~10-year) mortality was low. In appropriately selected patients, observation after resection may be a reasonable initial strategy, with RT reserved for salvage. General cohort characteristics (n=51) with Ki-67 (n=39). Variable Number (%) Median Age (years) 26 (range: 5-57) Gender, Male:Female (M/F) 26:25 (51/49) Subtotal Resection 29 (57) Gross Total Resection 22 (43) Adjuvant Radiotherapy 16 (31) Salvage Radiotherapy 16 (31) No Radiotherapy 19 (38) Ki-67 ≥3 21 (56) Ki-67 <3 18 (44)
Efficacy and hematopoietic recovery of high-dose melphalan with stem cell rescue as bridging to CAR-T compared with non-intensive bridging in relapsed/refractory multiple myeloma.
7544 Background: Bridging therapy is required for disease control in relapsed/refractory multiple myeloma (RRMM) patients awaiting CAR-T. Recent data suggest intensive cytotoxic bridging impairs post-CAR-T hematopoietic recovery. We compared melphalan-based therapy with stem cell rescue (MBT) versus talquetamab (T), a GPRC5D-directed bispecific antibody. Methods: We performed a retrospective single-center study of RRMM patients bridged with MBT or T prior to CAR-T (June 2021–December 2024) post apheresis. Cytopenias were graded per CTCAE v5.0 during early (within 30 days of CAR-T) and late (beyond 30 days of CAR-T) periods. Kaplan Meier analysis was performed to evaluate survival outcomes. Results: Of 242 CAR-T recipients, 34 received bridging either with T (n = 21, 62%) or MBT (n = 13, 38%). MBT patients were younger (61.9 vs 67.2 years; P = 0.08); high-risk cytogenetics by FISH were similar (92% vs 85%; P = 0.58). Bridging response rates were comparable (92% vs 82%; P = 0.36).At 20 months, 64% MBT and 58.7% T patients remained progression-free. Early grade 3–4 neutropenia for MBT vs T: 92% vs 95% (P = 0.37); late: 46% vs 33% (P = 0.84). Early grade 3–4 thrombocytopenia MBT vs T: 31% vs 24% (P = 0.28); late: 15% vs 14% (P = 0.40). Early grade 3 anemia MBT vs T: 69% vs 38% (P = 0.18); late grade 3–4: 17% vs 10% (P = 0.35). Median recovery times (MBT vs T): ANC 10 vs 5 days early (P = 0.28), 19 vs 17 days late (P = 0.11); platelets 7 vs 7.5 days early (P = 0.30), 12 vs 49 days late (P = 0.20); hemoglobin 13.5 vs 9 days early (P = 0.49), 35 vs 93 days late (P = 0.65). CRS rates were comparable: 53.8% vs 63.6% (P = 0.57). ICANS occurred only in the T cohort (13.6%) (p = 0.08). Conclusions: MBT yielded hematopoietic recovery, toxicity, and outcomes comparable to bispecific antibody bridging. These findings contrast with prior studies showing impaired recovery after intensive cytotoxic bridging, likely because stem cell rescue improves hematopoietic reserve. Prospective studies are warranted.
Identifying barriers to prescribing PARP inhibitors, including olaparib, in advanced ovarian cancer.
e17560 Background: Despite improvements in survival with poly (ADP-ribose) polymerase inhibitors (PARPi) as first-line maintenance (1LM) therapy in advanced ovarian cancer (AOC), research has shown suboptimal utilization. This study assessed PARPi prescribing behavior and key barriers to prescribing PARPi, including olaparib, for approved indications in AOC. Methods: A cross-sectional online survey was completed in Fall 2025 by board-certified/-eligible medical, hematologic, and gynecologic oncologists (oncs) with experience treating AOC with PARPi in the US. The survey assessed PARPi prescribing practices, influences, and beliefs. Barriers to prescribing olaparib in AOC were assessed with a Best Worst scaling (BWS) exercise; oncs were shown 12 tasks with 4 items each and asked to select the largest and smallest challenge to prescribing olaparib. Hierarchical Bayesian modeling was used to estimate BWS scores (range 0-100, higher score = larger barrier relative to the others); means and standard errors were reported. All other data were reported descriptively. Results: Analysis included 192 oncs (38 gynecologic oncs); 44% were from academic settings. Most oncs agreed there is high scientific evidence demonstrating PARPi benefits in approved AOC settings (82%), PARPi benefits outweigh risks, motivating them to prescribe PARPi (81%), and PARPi benefits are superior to non-PARPi in BRCA-mutated (BRCAm) patients (pts) (81%). Median percentage of pts they reported treating with a PARPi in 1LM was 75% BRCAm and 70% BRCA wild type/ homologous recombination deficiency positive (HRD+). Among treatment decision makers (n = 182), reasons selected for not having prescribed a PARPi in the past 6 months included pt ineligibility (49% of oncs), pt concern about side effects (46%), results of genetic testing (34%), out-of-pocket costs (33%), pt unwillingness for other reasons (28%), and prior authorization denial (27%). BWS results showed that the largest challenges to prescribing olaparib to eligible pts with AOC were managing side effects that impact quality of life and severe side effects, followed by patient adherence and complex prior authorizations (Table). Conclusions: While most oncs believe in the benefits of PARPi for treatment of AOC, barriers to prescribing PARPi are multifactorial and for many, emphasize management of side effects. Strategies to mitigate barriers should feature multiple targets. Barriers to prescribing olaparib in AOC BWS score(higher = larger barrier) Managing side effects that impact quality of life 18.5 (0.6) Severe side effects 18.3 (0.7) Patient adherence 11.7 (0.5) Complex prior authorization process 11.3 (0.6) Potential detriment outweighs the benefit 10.6 (0.4) Other PARPi have a more favorable side effect profile 8.3 (0.4) Patient choice 8.0 (0.5) Getting the relevant genetic testing 7.6 (0.5) Other manufacturer provides better financial assistance 5.7 (0.3)
Gastric cancer outcomes in adolescents and young adults in Alberta, Canada.
e16032 Background: Gastric cancer (GC) poses a significant challenge in Canada, with survival rates at 29%. Adolescents and Young Adults (AYA, ages 18-49) account for about 10% of GC cases but face challenges due to limited treatment guidelines. The biological behaviour of GC differs in younger patients compared with older adults. Understanding these disparities is crucial for developing targeted interventions for this vulnerable population. This study aimed to characterize GC outcomes in AYA patients compared to non-AYA patients in Alberta, Canada. Methods: We used the Alberta Cancer Registry to identify patients diagnosed with primary gastric or gastro-esophageal junction (GEJ) cancer from 2008 to 2023, ensuring a minimum of two years of follow-up. Patients with multiple malignancies or missing staging information were excluded. Those under 50 were classified as AYA (n = 345), while those 50 and older were categorized as non-AYA (n = 2888). We further stratified patients into four groups: early-stage AYA (n = 125), early-stage non-AYA (n = 1200), advanced-stage AYA (n = 220), and advanced-stage non-AYA (n = 1688). Key demographics and clinical variables (sex, gastric region, histology, treatments offered, overall survival (OS)) were compared. Results: In the early-stage cohort, histologic differences emerged, with signet ring cell histology more prevalent in AYA patients (23.2% vs. 10.6%; p < 0.001). AYA patients were also more likely to receive curative-intent systemic therapy (68.0% vs. 55.0%; p = 0.005). In the advanced setting, signet ring cell histology was again more prevalent in AYA patients (27.7% vs. 11.8%; p < 0.001). They were more likely to receive palliative therapy (53.2% vs. 37.4%; p < 0.001) and immunotherapy/chemotherapy combinations (17.1% vs. 11.4%; p = 0.042). Among those starting systemic therapy, more AYA patients progressed to second-line treatment (21.8% vs. 15.2%; p = 0.011). Conclusions: AYA patients with gastric cancer in Alberta exhibit aggressive disease characteristics and receive extensive treatment in both early and advanced stages. However, while outcomes are better in the early stages, they deteriorate in advanced stages despite high rates of second-line treatment, resulting in poorer survival rates for those undergoing systemic treatment in the metastatic setting. This highlights the urgent need for age-specific diagnostic strategies along with comprehensive, age-appropriate supportive care tailored to their distinct clinical and psychosocial needs.
Determinants of recurrence after curative gastrectomy in young-onset gastric cancer: A Latin American cohort.
e16132 Background: Young-onset gastric cancer with curative surgery have limited data on prognostic factors. We evaluated clinicopathological, nutritional and inflammatory features associated with recurrence in young patients undergoing gastrectomy for gastric adenocarcinoma. Methods: This is a retrospective observational cohort study with patients ≤45 years and gastric adenocarcinoma who underwent curative-intent gastrectomy between 2009 - 2020 at a Latin American cancer center. Collected variables included: T and N stage, overall pathological stage, Lauren classification, presence of signet ring cell component, lymphovascular invasion (LVI), and perineural invasion. Pathological variables were dichotomized as follows: T1–2 vs T3–4, N0 vs N+, and stage I–II vs stage III. Preoperative inflammatory and nutritional biomarkers (albumin, neutrophil, lymphocyte, and platelet counts) were used to calculate the neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), systemic immune-inflammation index (SII), and prognostic nutritional index (PNI). Biomarkers were summarized as medians and dichotomized using cohort-specific medians for survival analyses. The primary outcome was recurrence-free survival (RFS), estimated using the Kaplan–Meier method and analyzed using Cox proportional hazards regression. Results: 87 patients were included (median age 40 years, IQR 37–44), and 52% were female. Tumor recurrence occurred in 36 patients (41%). Patients with recurrence had a significantly higher prevalence of advanced pathological disease, including T3–4 tumors (100% vs 84%, p = 0.019), nodal involvement (94% vs 73%, p = 0.009), and stage III disease (86% vs 62%, p = 0.014). Lymphovascular invasion was also more frequent among patients with recurrence. No significant differences were observed according to age, sex, Lauren classification, or presence of signet ring cell component. Median preoperative levels of albumin, NLR, PLR, SII, and PNI did not differ significantly. The median RFS was 4.9 years (95% CI 3.0–not reached). Estimated RFS rates at 1, 3, and 5 years were 89.8%, 60.9%, and 49.8%, respectively. Node-positive patients exhibited significantly worse RFS compared with node-negative patients (log-rank p = 0.001). In univariate Cox regression analysis, nodal involvement and lymphovascular invasion were significantly associated with increased risk of recurrence. However, in multivariate analysis, pathological nodal status remained the only independent predictor of recurrence. Conclusions: In young patients undergoing curative gastrectomy for gastric cancer, recurrence is primarily driven by pathological indicators of tumor dissemination, particularly nodal involvement. Preoperative inflammatory and nutritional biomarkers were not independently associated with recurrence in this cohort.
Leader-member exchange (LMX) and workforce outcomes in radiation oncology: A preliminary generational comparison.
9038 Background: Job satisfaction and turnover intention remain crucial outcomes for healthcare organizations incurring substantial costs associated with recruiting and training new employees. Continual workplace stressors continue to challenge retention efforts across clerical and clinical settings. Leader-member exchange (LMX) theory provides a relationship-based framework for understanding these outcomes, emphasizing the quality of the manager-employee relationship. As healthcare workforce increasingly spans generational cohorts, differences in career expectations and workplace values may influence how leadership relationships are perceived. This study examines the association between LMX quality, job satisfaction, and turnover intention, with specific attention to generational differences between older (Baby Boomers and Generation X) and younger (Millennials and Generation Z) employees within a Radiation Oncology department. Methods: This quantitative study surveyed healthcare professionals (> 18 years) in a Radiation Oncology department at an academic medical center using purposive sampling. An anonymous, web-based survey was distributed via email and included standardized measures of leader-member exchange (LMX-7), job satisfaction, and turnover intention (TIS-6), along with demographic items. After electronic informed consent was obtained, the average for survey completion was 7.5 minutes. Data was collected over three months, de-identified, and analyzed using RStudio. Results: The survey achieved a 74% response rate, indicating broad representation of the department. Mean LMX scores were comparable between older (M = 4.04, SD = 1.06) and younger (M = 3.91, SD = 1.13) employees. Job satisfaction was modestly higher among younger respondents (M = 4.18, SD = 0.73) compared with older respondents (M = 4.00, SD = 0.82), whereas turnover intention was slightly higher in the older group (M = 1.56, SD = 0.87 versus M = 1.45, SD = 0.83). LMX was positively correlated with job satisfaction in both older (r = .58, p = .002) and younger (r = .59, p < .001) professionals. LMX demonstrated an inverse association with turnover intention in both groups, reaching statistical significance among older employees (r = -.57, p = .003) but not among younger employees (r = -.32, p = .069). Conclusions: Higher-quality LMX was associated with greater job satisfaction and lower turnover intention among both older and younger healthcare professionals. This emphasizes the importance of leadership relationships in workforce development and stability. These findings support the value of authentic leadership practices grounded in mutual respect and trust. Additionally, these findings highlight the need for future analyses to examine individual generational groups (e.g., Generation X versus Millennials) to better understand variations.
GRACE: A conversational AI platform for geriatric oncology risk and capability evaluation.
1658 Background: Comprehensive geriatric assessment (CGA) is a multidimensional evaluation of health in older adults and endorsed by the American Society of Clinical Oncology (ASCO) and the National Comprehensive Cancer Network (NCCN) for improved outcomes in geriatric cancer patients. However, CGA is underutilized due to time constraints and staffing shortages. Artificial intelligence (AI) can provide autonomous geriatric assessments (GA) but require validation. Methods: GRACE, an AI platform for GA via natural language interaction, was tested against manual GA. 70 adults aged ≥65 were enrolled from UCLA using the portal MyChart and at affiliated active senior living centers. 15 clinicians contributed qualitative perspectives. Concordance was evaluated using sensitivity, specificity, predictive values, and agreement (Cohen’s κ, Gwet’s AC1). Usability was measured with the System Usability Scale (SUS), and patient and provider interviews were thematically analyzed. Results: GRACE showed high diagnostic performance (sensitivity 0.73, specificity 0.94, κ=0.63, AC1=0.82) with highest agreement in polypharmacy (κ≈0.97; AC1≈0.98). Usability was high (SUS >80th percentile) and 84% of participants reported confidence using GRACE. Clinicians endorsed GRACE as a practical pre-visit intake tool that integrates with electronic health records and facilitates timely referrals. Conclusions: GRACE demonstrates that conversational AI can replicate key elements of clinician-administered GA while reducing burden on oncology teams. Larger multi-site studies are warranted to evaluate clinical impact and integration into routine oncology practice. Performance of GRACE versus manual geriatric assessment. Domain Sensitivity (CI) Specificity (CI) PPV (CI) NPV (CI) Cohen's Kappa (CI) Gwet's AC1 (CI) General Health 0.872 (0.748, 0.940) 0.931 (0.891, 0.957) 0.719 (0.592, 0.819) 0.973 (0.943, 0.988) 0.741 (0.637, 0.845) 0.887 (0.839, 0.935) Physical 0.859 (0.760, 0.922) 0.971 (0.956, 0.981) 0.753 (0.649, 0.834) 0.985 (0.973, 0.992) 0.781 (0.704, 0.858) 0.953 (0.935, 0.970) Functional 0.563 (0.332, 0.769) 0.985 (0.971, 0.993) 0.529 (0.310, 0.738) 0.987 (0.974, 0.994) 0.532 (0.298, 0.765) 0.972 (0.957, 0.986) Social Support 0.629 (0.530, 0.718) 0.873 (0.839, 0.900) 0.508 (0.420, 0.596) 0.918 (0.889, 0.940) 0.458 (0.359, 0.558) 0.753 (0.702, 0.805) Psychological 0.570 (0.460, 0.673) 0.920 (0.894, 0.940) 0.506 (0.404, 0.607) 0.937 (0.913, 0.955) 0.465 (0.353, 0.576) 0.839 (0.802, 0.876) Comorbidity 0.641 (0.484, 0.773) 0.996 (0.989, 0.998) 0.862 (0.694, 0.945) 0.985 (0.975, 0.991) 0.726 (0.601, 0.851) 0.980 (0.971, 0.989) Polypharmacy 0.974 (0.865, 0.995) 1.000 (0.893, 1.000) 1.000 (0.906, 1.000) 0.970 (0.847, 0.995) 0.971 (0.915, 1.027) 0.972 (0.916, 1.027) Overall 0.721 (0.674, 0.763) 0.956 (0.949, 0.963) 0.649 (0.603, 0.692) 0.968 (0.962, 0.974) 0.645 (0.604, 0.687) 0.917 (0.907, 0.927)