Citations

Abstract Large Language Models (LLM) are being widely adopted into the medical field for their impressive ability to analyze and summarize large amounts of text data. These models enable clinicians and researchers to extract meaningful insights from complex datasets and may assist with decision making. Here we present our workflows and application of LLM for the interpretation and summarization of clinical data related to CAR-T cell therapy. Using an LLM (Gemini 2.5 pro) in the Google Cloud Computing (GCP) environment, two applications were developed to analyze CAR-T cell therapy clinical data: 1) extracting and summarizing CRS and ICANS event-related data to streamline the compliance team workflow, and 2) identifying features available at time of CAR-T infusion able to classify patients into high- or low-monitoring needs 14 days post CAR-T. Patient data (vitals, labs, hematology notes, and EKGs) were extracted from the electronic medical record (EMR) using Google BigQuery into SQL tables in GCP. For each application, relevant data fields were retrieved, formatted into JSON objects, and embedded in the LLM prompt for context-aware processing. Both applications have undergone iterative prompt engineering after analyzing the LLM output against ground truth data in the EMR. For the application extracting and summarizing CRS and ICANS events, the LLM was optimized on Mayo Clinic Rochester data. When compared to the IEC compliance database, the LLM achieved 100% accuracy and F1 score for CRS events, and 96% accuracy and 82% F1 score for ICANS events. The match rate for CRS and ICANS grades were 83% and 89% respectively. We applied the same LLM to Mayo Clinic Arizona (MCA) and Mayo Clinic Florida (MCF) cases, who document clinical notes and toxicity flowsheet differently, and achieved an accuracy (MCA: 93%, MCF:93%) and F1 score (MCA: 96%, MCF: 96%) for CRS and accuracy (MCA: 81%, MCF:82%) and F1 score (MCA: 76%, MCF: 75%) for ICANS. LLM was able to capture events missed by manual reviews. For most of the discrepancies where compliance team final adjudication is needed, LLM will be updated to flag discrepancies for review by the compliance team. For the application related to the monitoring needs 14 days post CAR-T, the LLM identified 5 categories predictive of high or low monitoring needs post CAR-T infusion (disease status, inflammatory and tumor burden markers, hematologic status, renal function and performance status). Our model achieved a sensitivity of 85.7%, specificity of 23.8%, and F1 score of 65.5% for our first cohort, compared with data extracted by the LLM from the EMR. For our second cohort, with demographics statistically similar to cohort 1 and using the same 5 categories, the LLM achieved a sensitivity of 83.3%, a specificity of 28.6%, and F1 score of 65.4%. Our workflow and applications of LLM’s provide examples and guidance to others interested in applying LLM’s for clinical and research applications. Citation Format: Emmanuel Contreras Guzman, Matthew Jankowski, Andre De Menezes Silva Corraes, Malvika Gupta, Monica L. Shaw, Madiha Iqbal, Talal Hilal, Saurabh Chhabra, Ricardo Daniel Parrondo, Jody K. Mclean, Kim R. Riester, Kayla Joseph, Melinda Tan, Holly Ross, Cleyonia Barnett, Sylvia Carter, Semy Girmay, Rachel Wolan, Milana Ramsey, Christian Downhour, Kristy Morgan, Shae Sibley, Erica Rushing, Lucy Holmes, Allison Burgstahler, Stephen M. Ansell, Hassan Alkhateeb, Matthew Hathcock, Ramona Bruno, Allison C. Rosenthal, Hemant Murthy, Patrick B. Johnston, Jonas Paludo, Yi Lin. Applications of large language models to CAR-T cell therapy clinical data using Google cloud computing [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 2740.

Measurable residual disease (MRD) has established prognostic significance in patients with myeloid disorders; however, there is limited data regarding its prognostic and predictive value and the optimal technique for measurement in the setting of post allogeneic hematopoietic cell transplantation (allo-HCT). A multi-gene next generation sequencing (NGS) panel can overcome the limitations of conventional techniques for measurement of MRD and can shed light onto the longitudinal evolution and genomic complexity of patients undergoing allo-HCT for acute myeloid leukemia (AML) and myelodysplastic syndrome (MDS). In this study, patients with AML or MDS who underwent their first allo-HCT and completed NGS testing at day +100 and were in a state of morphological remission were identified from the transplant registry of Mayo Clinic in Florida. The primary aim was to evaluate the prognostic impact of somatic variants detected via NGS at day +100 post allo-HCT on long-term outcomes. In total, 105 patients who underwent allo-HCT for AML and MDS and completed NGS testing on day +100 were included in the study population. Seventeen patients were found to have detectable variants, and 88 patients were without a detectable variant. At a median follow up of 1.4 years (0.1–5.5), presence of any variant at day +100 was associated with inferior 4-year overall survival (OS) (32% vs. 69%, P < 0.001), decreased progression free survival (PFS) (34% vs. 61%, P < 0.001), and a higher incidence of relapse (64% vs. 23%, P < 0.001). In univariate models, a lower Karnofsky performance score and presence of variants post allo-HCT remained significantly associated with inferior OS, higher relapse incidence, and a decreased PFS. The presence of variants at day +100 post allo-HCT remained significant in multivariate models for relapse ( P = 0.002) but not for OS ( P = 0.06) and PFS ( P = 0.08). These results indicate that the presence of detectable somatic variants at day +100 is associated with poor long-term outcomes and are predictive of a higher incidence of relapse.

Abstract Allogeneic hematopoietic cell transplantation (allo-HCT) is a potentially curative treatment for many hematological malignancies, but graft-versus-host disease (GVHD) is a common complication. Low gut microbiome diversity is associated with higher GVHD risk and shorter survival in multiple studies. Recently, the BMT CTN 1703 clinical trial demonstrated superiority of a GVHD-prophylaxis regimen including post-transplant cyclophosphamide (PTCy) compared to the standard prophylaxis (tacrolimus and methotrexate, Tac/MTX) in terms of GVHD-free, relapse-free survival at one year among reduced intensity conditioning allo-HCT recipients. However, the effect of PTCy on the gut microbiome and its association with clinical outcome have not been described. Here, we report on a companion randomized clinical controlled trial (BMT CTN 1801), which collected 2575 longitudinal stool samples from 304 study participants. Samples were obtained up to weekly up to day 84 post allo-HCT and at less frequent intervals thereafter, up to 2 years. Microbiome diversity and absolute microbial load were lower in the PTCy group compared to the Tac/MTX group on days 14–28 post-HCT. However, diversity at the timepoint closest to neutrophil engraftment was not significantly associated with non-relapse mortality after one year or other clinical outcomes, contrary to expectations from previous studies. Microbial domination events, when a single species exceeds 30% relative abundance, were comparable across treatment arms and reflected both pathogen blooms as well as less severe disruptions of the microbial community. Clostridium scindens and secondary bile acid metabolism pathways were less prevalent in the PTCy arm than in the Tac/MTX arm post-HCT, yet presence of secondary bile acid metabolism pathways was associated with a lower risk of chronic GVHD. Given that PTCy was associated with a greater disruption of the microbiome as measured by diversity, absolute microbial abundance, and bile acid metabolism capability, but better clinical outcomes overall, these data suggest that the importance of the microbiome in modulating the host immune systems after allo-HCT is specific to different types of GVHD prophylaxis.

This study uncovers a previously unrecognized role for NAV3 in TGF-β1-driven cardiac fibroblast activation. We show that NAV3 facilitates profibrotic remodeling through noncanonical signaling and cell cycle arrest, independently of SMAD2/3. These findings position NAV3 as a novel regulator of fibroblast phenotype and a potential modulator of cardiac fibrosis.

Introduction: Gastroesophageal reflux disease (GERD) plays an important role in lung transplant (LT) outcomes. This study aims to evaluate the impact of a novel sleep positioning device (SPD) on lung function in LT recipients with GERD. Methods: In this single-center cohort study, LT recipients (2014 to 2019) with GERD who were prescribed an SPD within 2 years of LT were included. A historical control group of LT recipients with GERD on PPI only (2011 to 2013) was selected to compare forced expiratory volume in 1 second (FEV1) as a marker of lung function. Results: Twenty LT recipients using SPD (cases) and 54 patients in the historical control group with proven GERD were included. Mean (SD) interval between LT and SPD prescription was 8.6 (5.7) months. Mean % predicted FEV1 was higher in SPD users than controls post-LT at 6 months (80% vs. 69%) and 1 year (80% vs. 71%), respectively. Among SPD users, post-SPD FEV1 was higher than pre-SPD FEV1 (82% vs. 70%). Conclusion: Use of a novel SPD can stabilize or improve pulmonary function in patients with GERD after LT. SPD may be a safe and cost-effective adjunctive therapy for the management of GERD after LT.

Abstract Background: As with most Sub-Saharan African countries, Nigeria has a rising incidence of cancer, with disproportionate mortality rates. The financial burden of cancer care often results in catastrophic healthcare spending, leading to treatment refusal, disruption, and discontinuation. This is particularly significant in Nigeria, where nearly all patients are uninsured, and out-of-pocket costs often exceed households’ ability to pay. Financial Navigation Programs (FNPs) have been shown to mitigate treatment-related financial toxicity in cancer care and reduce treatment abandonment, but there is a paucity of high-quality data on this intervention in resource-constrained settings. Here, we present a randomized controlled trial to evaluate the impact of a novel FNP in Nigeria. Methods: We designed the COST-FIN trial, a multi-site pragmatic single-blinded randomized controlled trial of newly diagnosed (<6 weeks from diagnosis) adults (≥18 years) with breast, colorectal, or prostate cancer at two tertiary cancer centers in Nigeria. Participants (n=200) will be randomized (1:1) to either the intervention (FNP) or the control arm and followed for 12 months. Data on key individual, treatment, and financial parameters will be collected via structured interviews and chart abstraction at baseline, 3, 6, and 12-month follow-up. In addition, participants randomized to the FNP will receive a tailored financial literacy assessment, financial planning support, and enhanced access to resources by trained financial navigators. Primary and secondary outcomes are Financial Catastrophe (FC) and Financial Distress (FD), respectively. Exploratory outcomes will include cost-related non-adherence and cost-effectiveness of the program. An interim analysis will be conducted when 50% of the estimated accruals reach 6 months of follow-up, with crossover if compelling evidence of benefit is demonstrated at that time point. All participants will be followed for 12 months from recruitment. Discussion: This first-of-its-kind study will provide evidence on the role of FNP in potentially eliminating financial barriers to cancer care in Nigeria. Given the country’s renewed interest in cancer control through the passage of the National Cancer Control Plan, findings from this study have the potential to influence policy reform and set the stage for further studies to evaluate the scalability and implementation of similar interventions in resource-limited settings. Trial registration : ClinicalTrials.gov NCT06630962. Oct 8, 2024 (https://clinicaltrials.gov/study/NCT06630962)

Abstract The comparative value of liposomal cytarabine/daunorubicin (CPX-351) versus venetoclax plus a hypomethylating agent (Ven-HMA) in the frontline treatment of older adults with primary ( de novo ) or secondary acute myeloid leukemia (AML) remains uncertain. In the current study, we retrospectively examined outcomes of 600 patients with newly diagnosed AML treated with CPX-351 (N = 112) or Ven-HMA (N = 488). AML subtypes included de novo (46%), post-myelodysplastic syndrome (MDS) (19%), post-myeloproliferative neoplasm (MPN) (12%), post-MDS/MPN (6%), therapy-related (17%) and AML with myelodysplasia-related genetic abnormalities (AML-MR; 46%). Patients receiving CPX-351 were younger (median 65 vs. 73 years; p < 0.01), predominantly female (50% vs. 38%; p = 0.02), more likely to have secondary AML (68% vs. 51%; p < 0.01), and less likely to harbor NPM1 MUT (5% vs. 12%; p = 0.02). Rates of complete response with or without count recovery (CR/CRi) were comparable between CPX-351 and Ven-HMA (55% vs. 60%; p = 0.30), including in AML-MR (60% vs. 63%; p = 0.70). Ven-HMA use was associated with fewer infectious complications (62% vs. 83%; p < 0.01) and yielded higher CR/CRi rates in males (60% vs. 45%; p = 0.04), de novo AML (68% vs. 50%; p = 0.03), and in the presence of STAG2 MUT (86% vs. 44%; p = 0.02), or CEBPA MUT (88% vs. 50%; p = 0.03). Overall survival censored for transplant, was similar (median 10 vs. 13 months; p = 0.90), with Ven-HMA being superior in post-MDS AML (median 12 vs. 7 months; p = 0.02) and CPX-351 in the presence of SF3B1 MUT (not reached vs. 14 months; p < 0.01). Our findings suggest that Ven-HMA is as effective and less toxic than CPX-351 in newly diagnosed AML, including AML-MR, despite selection of younger, fitter patients for CPX-351.

Abstract Venetoclax (Ven) in combination with hypomethylating agents (HMA) is widely used as frontline treatment in older or unfit patients with acute myeloid leukemia (AML). Liposomal cytarabine and daunorubicin (CPX-351) is FDA-approved for the treatment of secondary AML, including therapy-related (t-AML) and AML with myelodysplasia-related changes (AML-MR). The current study aimed to provide a comprehensive comparison between the two treatment strategies to better inform clinical decisions. Methods The current retrospective analysis was performed under an institutional review board approved minimal risk protocol at Mayo Clinic (MN, AZ, FL). AML-MR gene mutations or cytogenetic abnormalities were categorized according to the International Consensus Classification (Arber et al. Blood 2022;140:1200). Treatment response was assessed according to the 2022 European LeukemiaNet (ELN) criteria (Dohner et al. Blood 2022;140:1345). Overall survival (OS) analysis was censored for allogeneic stem cell transplant (ASCT). Statistical analyses were performed using JMP Pro (Version 18.0.0). Results A total of 600 patients with newly diagnosed AML (ND-AML) treated with CPX-351 (N-112, 19%), or Ven-HMA (N=488, 81%) were included. AML subtypes included de novo (N=277, 46%), post-myelodysplastic syndrome (post-MDS, N=114,19%), post-myeloproliferative neoplasm (post-MPN, N=70, 12%), post-MDS/MPN (N=36, 6%), t-AML (N=103, 17%) and AML-MR (N=274, 46%). Compared with those treated with Ven-HMA, patients receiving CPX-351 were younger (median age 65 vs 73 years; p&lt;0.01), more likely to be female (50% vs 38%; p=0.02), more frequently diagnosed withsecondaryAML (68% vs 51%; p&lt;0.01), and less likely to have NPM1MUT (5% vs 12%; p=0.02) whereas ELN 2022 cytogenetic risk distribution was similar between the two (adverse risk; 43% vs 40%; p=0.5). Comparison of treatment response and survival Complete response rates with (CR) or without (CRi) count recovery were similar between CPX-351 and Ven-HMA overall (55% vs 60%) as well as specifically in patients with AML-MR (60% vs 63%), adverse karyotype (44% vs 44%), NPM1MUT (100% vs 89%), TP53MUT(32% vs 45%), IDH1MUT (60% vs 76%), IDH2MUT (63% vs 77%), SF3B1MUT(71% vs 45%), and CBLMUT (33% vs 69%) [p&gt;0.05 in all instances]. Compared to those with CPX-351, CR/CRi rates with Ven-HMA were superior in males (60% vs. 45%; p=0.04) and in patients with de novo AML (68% vs. 50%; p=0.03), STAG2MUT (86% vs. 44%; p=0.02), or CEBPAMUT (88% vs. 50%; p=0.03); infectious complications were significantly more frequent with CPX-351 (83% vs 62%, p&lt;0.01). At a median follow-up of 12 months (range; 1-88), 416 (69%) deaths, 179 relapses (44% in CPX-351 and 52% in Ven-HMA responders) and 149 ASCTs (53% in CPX-351 and 18% in Ven-HMA) were documented. OS rates were similar between CPX-351 and Ven-HMA, in general (median 10 vs 13 months), as well as in patients with adverse karyotype, TP53MUT, IDH1MUT, and IDH2MUT (p&gt;0.05 in all instances). By contrast, OS was superior in Ven-HMA-treated patients with post-MDS AML (median 12 vs. 7 months; age-adjusted p=0.02) and in CPX-351-treated patients harboring SF3B1MUT(median not reached vs 14 months; age-adjusted p&lt;0.01). In addition, a borderline-significant OS advantage for CPX-351 was apparent in patients with PTPN11MUT (13.5 vs 8 months; p=0.08). Comparison in ICC defined AML-myelodysplasia-related CR/CRi and OS rates were comparable between patients treated with CPX-351 (N=57) or Ven-HMA (N=217), with response rates of 60% vs 63% (p=0.68) and median OS of 10 vs 14 months (p=0.61), respectively. However, OS was superior with Ven-HMA in patients with post-MDS AML (15 vs 8 months; age-adjusted p=0.02) and with CPX-351 in those with SF3B1MUT (NR vs 14 months; age-adjusted p=0.02). Conclusions: The current study suggests that Ven-HMA was as effective and possibly less toxic than CPX-351 for the treatment of ND-AML, including AML-MR, despite the bias for younger and fitter patients treated with the latter regimen. Observed differences involving patients with SF3B1MUT or post-MDS AML require validation in prospective studies and provide parameters for developing treatment-specific predictive models that will be presented at the meeting.

Abstract Introduction: Isocitrate dehydrogenase (IDH) mutations are observed in ~20% of patients with acute myeloid leukemia (AML). In the U.S., ivosidenib (ivo) is approved with or without azacitidine (aza) in the frontline (FL) and as monotherapy in relapsed/refractory (R/R) IDH1-mutated (mut) AML. Olutasidenib (oluta) is approved as monotherapy for R/R IDH1-mut AML. Enasidenib (ena) is approved as monotherapy for R/R IDH2-mut AML. This study describes real-world IDH inhibitor (IDHi) administration patterns and evaluates their outcomes in on-label settings. Methods: A retrospective review of patients with IDH1/2-mut AML who received ivosidenib, olutasidenib, or enasidenib was conducted at 12 academic institutions in the U.S. Best response at any time during treatment, including complete response (CR) and complete response with partial hematologic recovery (CRh), were graded using European Leukemia Net (ELN) 2022 criteria. Results: A total of 387 patients with IDH1- and IDH2-mut AML were treated with an IDHi in the FL or R/R setting. Median follow-up was 9.6 months (range, 0.03-90.7). Demographics include 79% White, 14% Black, 3% Asian, and 4% other/unknown race; 5% were Hispanic ethnicity; 46% were female; the median age was 70 years (range, 28-95). 70% had an Eastern Cooperative Oncology Group (ECOG) performance status of 0-1, 26% had secondary AML, and 10% had therapy-related AML. Per the ELN 2024 less-Intensive risk classification, 58% were favorable risk, 29% intermediate, 7% adverse, and 6% unknown. Overall, 209 of 387 patients (54%) received IDHi regimens on-label. In FL IDH1-mut AML (n=45), 11 received ivo monotherapy, 25 aza/ivo, and 9 used off-label combinations: 3 hypomethylating agent (HMA) + venetoclax (ven) + ivo, 5 decitabine/ivo, 1 intensive chemotherapy (IC)/ivo. Among R/R IDH1-mut AML (n=109), 47 received ivo monotherapy, 7 oluta monotherapy, and 55 off-label (31 HMA/ivo, 16 HMA/ven/ivo, 3 IC/ivo, and 5 other). 44 received off-label ena for FL treatment of IDH2-mut AML (25 ena monotherapy, 9 HMA/ena, 8 HMA/ven/ena, and 2 IC/ena). In R/R IDH2-mut AML (n=189), 119 patients received ena monotherapy and 70 received off-label combinations (25 HMA/ena, 30 HMA/ven/ena, 7 IC/ena, 6 ena/gilteritinib, and 2 other). In on-label treatment of FL IDH1-mut AML, one of four patients with response assessment of FL ivo monotherapy had a CR/CRh. Of 19 patients with response assessment on FL aza/ivo, 13 patients (68.4%) had a CR as best response (95% CI, 46.0-84.6%). Median OS was 14.9 mo (95% CI, 4.17-NR) for ivo monotherapy (n=11) and 31.9 mo (95% CI, 9.1-NR) for aza/ivo (n=25), with 0 and 2 patients receiving allogeneic stem cell transplant (allo-SCT), respectively. In R/R IDH1-mut AML, the rate of CR/CRh was 35.5% (95% CI, 21.1-53.1) for on-label ivo monotherapy. All 4 patients with response assessment for on-label oluta monotherapy had a CR. Median OS was 14.0 mo (95% CI, 10.1-22.4) for ivo monotherapy and NR for oluta monotherapy (95% CI, 10.1-NR), with 10 and 0 receiving allo-SCT, respectively. The CR/CRh rate for ivo was 38.1% (8 of 21 assessed; 95% CI, 20.8-59.1) in age &lt; 70, and 30.0% (3 of 10 assessed; 95% CI, 10.8-60.3) in age ≥ 70. Median OS was 15.2 mo (95% CI 9.3-NR) in age &lt; 70 (n=27) and 10.4 mo (95% CI 8.0-18) in age ≥ 70 (n=20). In R/R IDH2-mut AML, the rate of CR/CRh for on-label ena monotherapy was 38.2% (95% CI, 27.6-50.1). Median OS was 12.1 mo (95% CI, 8.2-16.6), with 5 having received allo-SCT. In demographic subgroups for ena monotherapy, 52 non-Hispanic White (NHW) patients with response assessment had a CR/CRh rate of 42.3% (95% CI, 0.30-0.56). Only 1 of 13 non-Hispanic Black (NHB) patients with response assessment had a CRh (7.7%; 95% CI, 1.4-33.0) as best response, with none attaining CR. Median OS was 13.7 mo (95% CI 10.3-26.3) in NHW (n=82), 11.2 mo (95% CI, 6.1-NR) in NHB (n=18), and 69.1 mo (95% CI, 52.6-NR) in Hispanic patients (n=8). The CR/CRh rate was 48.6% (17 of 35 assessed; 95% CI, 33.0-64.4) for age &lt; 70, and 27.3% (9 of 33 assessed; 95% CI, 15.1-44.2) for age ≥ 70. Median OS was 12.0 mo (95% CI, 9.9-22.1) in age &lt; 70 (n=48) and 5.8 mo (95% CI, 4.1-8.7) in age ≥ 70 (n=69). Conclusion: Practice patterns for IDH inhibitors in AML vary considerably, with high amounts of off-label use. Response rates for ivosidenib and enasidenib in the real-world setting are comparable to those of clinical trials. Descriptive analyses of demographic subgroups highlight areas in need of further research.