Alkhateeb, Hassan B.

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.

Abstract Recent updates to monocyte count thresholds recognize oligomonocytic chronic myelomonocytic leukemia (OM-CMML) as an early form of CMML. However, the clinical validity of these changes remains uncertain without incorporating biological and genomic factors. In this study, we analyzed a cohort of 911 patients (249 with OM-CMML, 359 with overt CMML, and 303 with myelodysplastic syndromes) using unsupervised clustering to evaluate the role of genomic determinants in refining CMML diagnosis. Our findings show that CMML molecular signatures (biallelic TET2 mutations or SRSF2-TET2 comutations) are linked to a distinct transcriptome, monocytic bias, classical monocytosis, and a higher risk of progression to overt CMML in OM-CMML cases. We developed a weighted genomic model and diagnostic workflow showing that combining genomic signatures with bone marrow monocyte frequencies in OM-CMML more accurately predicts progression to overt CMML. These findings support integrating genomic determinants and our clinic-ready diagnostic workflow into the CMML diagnostic framework to improve accuracy. Significance: Through comprehensive clinical and genomic profiling of a large patient cohort, alongside immunophenotypic and transcriptional cellular analyses, this study provides evidence that incorporating genomic determinants into the diagnostic criteria for OM-CMML improves diagnostic accuracy and refines the identification of early-stage CMML, thereby preventing misclassification.

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 Background IDH mutations are found in 15%–20% of acute myeloid leukemia (AML). Recent evidence suggests that IDH mutation subtypes may respond differently to intensive chemotherapy. We evaluated blast clearance patterns, remission outcomes, and overall survival (OS) in AML patients with IDH2 R140, IDH2 R172, and IDH1 R132 mutations to assess differences in treatment response. Methods We retrospectively reviewed AML patients diagnosed at Mayo Clinic, Rochester (2016–2023) with NGS data and treated with intensive chemotherapy (7+3 or CPX‐351). Bone marrow blasts were assessed at diagnosis, mid‐induction, and end‐of‐cycle; blast clearance was defined as < 5%. Composite remission (CRc) included CR/CRi per ELN 2022. IDH2 responders were classified as early (after Cycle 1) or late (≥ 2 cycles). Blast reduction was calculated as log 10 (diagnosis blasts/end‐induction blasts) and normalized per day. OS was measured from diagnosis. Results Among 381 patients, 31 had IDH2 mutations (23 R140, 8 R172) and eight had IDH1 R132. Baseline blasts were similar. Mid‐cycle blasts were lower in R140 versus R172 (4% vs. 10%), with higher clearance (68% vs. 38%). End‐induction blasts were lower in R140 versus R172 (2% vs. 4%) with higher clearance (100% vs. 75%). IDH1 and IDH2 showed similar clearance patterns. Log 10 analyses showed deeper and faster blast reduction in R140 versus R172, with comparable kinetics between IDH1 and IDH2 . OS did not differ across subtypes. Early CRc in IDH2 ‐mutated patients was associated with numerically longer OS. Conclusion IDH2 R140 demonstrated faster blast clearance than R172, while remission and survival were similar across IDH subtypes. Trial Registration : The authors have confirmed clinical trial registration is not needed for this submission.

Abstract Introduction Management of acute leukemia (AL) during pregnancy remains challenging. In the first trimester, chemotherapy is toxic to the fetus, while delaying treatment may endanger maternal health, often prompting recommendations for pregnancy termination. In later trimesters, although the risk of congenital malformations decreases, complications such as preterm labor remain prevalent. Delivery prior to chemotherapy may be considered in advanced gestation. This study aimed to describe maternal and fetal outcomes of patients diagnosed with AL during pregnancy. Methods Following institutional review board approval, we retrospectively identified pregnant patients diagnosed with AL between 1995 and 2025 across Mayo Clinic sites (Minnesota, Arizona, Florida). We recorded leukemia subtype, treatment course, maternal and fetal outcomes. Results We identified 18 patients diagnosed with AL during pregnancy, including 11 with acute myeloid leukemia (AML) and 7 with acute lymphoblastic leukemia (ALL). Ten were diagnosed in the first trimester, 3 in the second, and 5 in the third. In the AML group (n=11; median age 30 years, range 18-42), one patient had acute promyelocytic leukemia (APL). ELN 2022 risk stratification was favorable in 5 (45%), intermediate in 4 (36%), and adverse in 2 (18%). In the ALL group (n=7; median age 23 years, range 19-34), 5 had B-ALL (one Philadelphia chromosome-positive) and 2 had T-ALL. All ALL cases were diagnosed in the first trimester. Among first-trimester AMLs, 2 of 3 had core-binding factor (CBF) AML. All eight patients diagnosed in the second and third trimesters had AML (1 APL, 2 CBF AML, 1 KMT2A-rearranged, 1 with DEK::NUP214 fusion, 1 complex karyotype including t(8;21), 2 with normal karyotype). Overall, 15 (83%) patients achieved complete remission (CR) or CR with incomplete count recovery (CRi) after initial induction. One patient had refractory disease and achieved partial remission after salvage, another had refractory disease and achieved CR after multiple lines of therapy, and a third achieved CR after FLAG-idarubicin and revumenib. Five patients received chemotherapy during pregnancy (3 in the first trimester, 1 in the second, 1 in the third); two of these resulted in live births. Among the 10 patients diagnosed in the first trimester, all underwent therapeutic or spontaneous abortion. Of the 3 second-trimester diagnoses, 2 underwent spontaneous or therapeutic abortion at 15 weeks. The third received 7+3 induction at 24 weeks for KMT2A-rearranged AML, followed by preterm delivery at 26 weeks and re-induction with FLAG-idarubicin plus revumenib, achieving CR. Among the 5 third trimester diagnoses, 4 delivered preterm infants (median gestational age 33 weeks) prior to chemotherapy. One patient received 7+3 at 29 weeks and delivered a viable preterm infant at 36 weeks. At a median follow-up of 6.5 years, 5 of 11 patients with AML (45%) and 3 of 7 patients with ALL (43%) died, mostly due to disease progression. Median overall survival was not reached in either group. Causes of death were as follows: 1 of AML, 1 of ALL, 3 of acute stem cell transplant (SCT) complications, 1 of chronic SCT complications, 1 of Hodgkin lymphoma, and 1 of pneumonia. Among patients diagnosed in the second and third trimesters (n=8, all with AML), 6 (75%) resulted in live births: 5 of 5 (100%) third-trimester diagnoses and 1 of 3 (33%) second-trimester diagnoses. All received intensive induction: 7+3 (n=5), 7+3 + ATRA (n=1), 7+3 + gilteritinib (n=1), and 7+3 + nilotinib (n=1). Of the second-trimester group, only one received induction prior to delivery; she later achieved CR with FLAG-ida and revumenib. Among the third-trimester group, one patient received chemotherapy prior to delivery and achieved CR followed by SCT. All but one of the 8 AML patients achieved CR after initial therapy. Conclusions Successful pregnancy outcomes were achieved in 75% of patients diagnosed with AML during the late second or third trimester. Notably, 5 (28%) of patients received chemotherapy during pregnancy. Larger, multi-institutional cohorts are warranted to further characterize the incidence and outcomes of AL in pregnancy.

Abstract Introduction: CAR-T therapy has transformed cancer therapy. with its first Food and Drug Administration (FDA) approved product in August 2017. Given the expanding indications and growing experience, particularly with management of immediate toxicities requiring specialized expertise such as Cytokine Release Syndrome (CRS) and Immune Effector Cell-Associated Neurotoxicity syndrome (ICANS), FDA recently removed the Risk Evaluation and Mitigation Strategies (REMS) for CD19 and BCMA targeting CAR-T on June 26th2025. In particular, given that CRS and ICANS typically occur in the first 2 weeks post CAR-T infusion, the minimum required stay near CAR-T treatment center is reduced from one month to 2 weeks. While this could reduce treatment burden for patients (pts) who need to travel away from home to receive CAR-T, there is a need to update the transition of care from CAR-T center to pts' local oncology practice. Pts carry in their rate of recovery and have varying supportive care needs in the second half of the month post CAR-T infusion. To assist with care transition and planning, we examine the use of AI multimodal large language model (LLM) to predict patients' supportive care needs 2 weeks after CAR-T infusion based on the clinical data at the time of CAR-T evaluation, prior to leukapheresis. This timepoint was selected to also maximize travel, lodging and caregiver planning for pts and family as well. Methodology: Pts who were treated with FDA-approved CAR-T at Mayo Clinic Rochester between December 2022 and June 2025 were examined. Based on experiences with referral practices in the Midwest region, we defined patients with high monitoring needs as those who between day 14-30 post CAR-T had either hospital admission due to recurrent CRS or ICANS or required transfusion of more than 1 blood product per week (red blood cell or platelets). Others were classified as low monitoring needs. Clinical notes, labs, vital signs and electrocardiogram at the time of CAR-T evaluation were abstracted and summarized by Gemini LLM for association with patients with high versus low monitoring needs post CAR-T. Model was developed and tested for predicted needs and concordance with actual outcome of these pts. Results: Among the 362 pts who received FDA approved CAR-T in the timeframe, 84 pts had high monitoring needs. An additional 84 pts with low monitoring needs were randomly selected from 278 pts. These were divided into a training cohort and test cohort of 84 pts each, with 42 pts with high and 42 pts with low monitoring needs. Demographics of the two cohorts displayed no statistical difference. LLM analysis of the clinical data at the time of CAR-T eval identified 5 categories associated with risk for high monitoring needs: 1) Disease status (clinical notes about bulky, progressive disease or CNS involvement); 2) inflammatory and tumor burden markers (CRP, LDH); 3) hematologic status (platelet, plt; neutrophil count ANC); 4) renal function; 5) performance status. Using common clinical lab parameter convention (platelets &lt;100, ANC&lt; 1, CrCl&lt; 30ml/min), and prioritizing for not missing pts with high monitoring needs, LLM model was defined to predict patients for high monitoring needs if high risk feature was identified in category 1 or any category combinations; whereas patients with no high risk feature or only 1 category that is not disease status was predicted as low monitoring needs. In this model, the sensitivity was 85.7%, specificity 23.8%, F1 score 65.5%. Applying this model to test cohort 2, the sensitivity was 83.3%, specificity 28.6%, and F1 score 65.4%. Discussion: We demonstrate proof-of-concept that LLM can be used to analyze clinical data at the time of evaluation for CAR-T eligibility to predict for supportive care needs post treatment. Not surprisingly, features identified by LLM to be associated with high monitoring needs are those previously reported by us and others at the time of CAR-T evaluation and or pre-lymphodepletion to be associated with risk for cytopenia, infection, hospitalization and non-relapse mortality. Based on the summary output generated by LLM, additional opportunities for classification methods were identified to improve specificity and balanced accuracy of the LLM model. This is a promising tool for further development to assist clinicians and patients to plan their treatment journey.