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.

Background: Hepatorenal syndrome–acute kidney injury (HRS-AKI) is associated with high mortality in cirrhosis. Recent FDA approval of terlipressin and updated clinical guidance have expanded management options, but real-world practice patterns remain unknown. Methods: We conducted a nationwide survey of U.S. hepatology providers regarding HRS-AKI management practices. The 33-question survey assessed provider demographics, vasoconstrictor selection, treatment timing, and adherence to guideline recommendations. Results: Among 162 respondents, most were hepatologists (94%) at academic centers (83%). Only 11% obtained a nephrology consultation at AKI detection, with 44% waiting until worsening renal function. Terlipressin was available at 80% of institutions but more commonly at academic centers (84% vs. 53%, p =0.003). Consistent with guideline recommendations, 77% of providers initiate vasoconstrictors only after completing a trial of volume expansion (if no renal improvement), particularly in academic centers (80% vs. 60% non-academic, p =0.03). Terlipressin (49%) and midodrine/octreotide (44%) were preferred first-line treatments, with providers at academic centers more likely to use midodrine/octreotide (46% vs. 30%, p =0.008). Mean arterial pressure was used by 62% of providers to guide vasoconstrictor dosing. Most providers (73%) discontinued vasoconstrictor treatment after 4 days if no improvement. Conclusions: Significant variations exist between guideline recommendations and real-world HRS-AKI management, especially first-line treatment choice, treatment monitoring, and nephrology consultation. These findings highlight opportunities to improve guideline implementation and identify areas where practice patterns might inform provider education and future guidance updates.

Abstract Introduction The effectiveness of temporary uterine artery ligation (TUAL) in reducing blood loss during laparoscopic myomectomy (LM) is uncertain, despite increased use as an adjunct to improve surgical safety. This study aims to evaluate the safety and efficacy of TUAL in LM for the management of uterine myomas. Material and Methods We conducted a systematic review and meta‐analysis of comparative studies evaluating LM with and without TUAL. Electronic databases were searched up to October 7, 2024. Eligible studies were randomized controlled trials or observational comparative studies published in English that reported at least one primary outcome. The primary outcomes were intraoperative estimated blood loss and perioperative change in hemoglobin. Secondary outcomes included operative time, number and size of myomas removed, length of hospital stay, recurrence, and complication rates. Data extraction and analysis were performed using a random‐effects model. This review was registered prospectively with PROSPERO (CRD42024595684). Results Seven studies met the inclusion criteria, comprising five randomized controlled trials and two retrospective case–control studies, with a total of 857 women (352 undergoing LM with TUAL and 505 undergoing LM alone). TUAL was associated with a reduction in intraoperative estimated blood loss (mean difference [MD]: −84.7 mL; 95% Cl: −89.0; −80.0) and postoperative reduction in hemoglobin level (MD: −0.4 g/dL; 95% Cl: −0.7; −0.2) compared with LM without TUAL. However, TUAL was associated with increased operative time (MD: 16.5 min; 95% Cl: 3.4; 29.6). Conclusions TUAL appears to be a possible adjunct to LM for selected patients in the management of uterine myomas. Our study demonstrated that TUAL with LM could be associated with a reduction in estimated blood loss, despite an increase in operative time. Nonetheless, further studies with a larger sample size and long‐term follow‐up, as well as stratification of outcomes based on ligation technique, are warranted to elucidate the benefits of TUAL.

Abstract Background and Aims Stroke and systemic embolism are recognized but poorly characterized complications of atrial fibrillation (AF) and left atrial flutter (AFL) catheter ablation. Their incidence, outcomes, and procedural associations remain insufficiently defined. Methods Retrospective global registry was initiated by the European Heart Rhythm Association Scientific Committee. A secure database captured data from ablation centres worldwide on general AF and left AFL ablation activity, as well as detailed information on patients with symptomatic embolic events between 2017 and 2024. Results Overall, 204 centres reported 335 743 ablation procedures and 550 embolic events (incidence 0.16%). Centre-level rate ranged from 0% to 3.1%. Most events were cerebral (94%), followed by peripheral (5%), and combined (1%). Coronary air embolism accounted for 62% of all peripheral events. Intraprocedural events occurred in 93 patients (17%), within 72 h in 357 (78%) and beyond 72 h in 100 (22%); 128 cases (23%) were diagnosed after hospital discharge. Nonparoxysmal AF (0.23% vs 0.15%; P < .001), centre experience, and ablation energy source (radiofrequency, 0.16%; cryoballoon, 0.16%; pulsed field ablation, 0.25%; laserballoon, 0.88%; P < .001) were associated with embolic risk. Median NIHSS score after stroke was 4 (interquartile range, 2–9). Three-month follow-up was available for 518 patients with embolism (95%): 181 (35%) had sequelae and 16 (3%) died. Conclusions Symptomatic embolic events are infrequent but associated with significant morbidity and mortality. Most events occur early, though a substantial proportion arise later. Higher event rates were observed with laser and possibly with pulsed field ablation. A greater centre experience was associated with a lower embolism incidence.

Abstract INTRODUCTION Visual perceptual difficulties occur in dementia with Lewy bodies (DLB). Normative data and cut scores for DLB and mild cognitive impairment with Lewy bodies (MCI‐LB) are needed for the 20‐item Noise Pareidolia Task. METHODS Participants included 344 cognitively unimpaired (CU), 112 DLB, and 105 MCI‐LB. Cumulative percentile ranking was used for normative classification. Receiver operating characteristic analyses, convergent validity, and associations with visual hallucinations (VH) were assessed. RESULTS In CU participants, a perfect score was average, one error was low average, and two or more errors fell below average. A cut score of one or more errors optimally distinguished CU participants from DLB (Total score area under the curve [AUC]: 0.82, Noise score AUC: 0.77) and MCI‐LB (Total score AUC: 0.75, Noise score AUC: 0.74). A cut score of two or more errors lowered sensitivity but increased specificity to 94%. Performance was associated with VH in DLB, but not MCI‐LB. DISCUSSION Results establish normative data for the 20‐item Noise Pareidolia Task and support its use in DLB and MCI‐LB.

Abstract Background High-sensitivity cardiac troponin (hs-cTn) assays are recommended for the diagnosis of acute myocardial infarction. Here, we characterize the analytical performance of a next-generation hs-cTn assay, Elecsys® Troponin T hs Gen 6 (Roche Diagnostics International). Methods Surplus lithium-heparin plasma or serum samples from patients or healthy volunteers were run on Cobas® e 801, e 402, and Pro analyzers. Limits of blank (LoB), limits of detection (LoD), and limits of quantitation (LoQ) were determined according to CLSI EP17-A2, with target values of 1.0 and 1.5 ng/L for LoB/LoD and 3.0 ng/L (10% CV) and 1.5 ng/L (20% CV) for LoQ, respectively. Precision was measured, per CLSI EP17-A2, using 3 QC samples (approximately 4, 30, and 220 ng/L), 12 native samples, and 3 reagent lots. Linearity, per CLSI EP06-Ed2, was determined by diluting samples with cardiac troponin T (cTnT) concentration above the measuring range with a low/blank sample. Interference (per Glick) with endogenous and assay components at 5 cTnT concentrations was assessed. Results Measured values for LoB, LoD, and LoQ at 10% and 20% CV were 0.1 to 0.7 ng/L, 0.3 to 1.4 ng/L, 1.0 to 2.9 ng/L, and 0.4 to 1.2 ng/L, respectively. Repeatability CVs were 1.0 to 5.8% for mean cTnT concentrations of 2.6 to 9230 ng/L in lithium-heparin plasma. High precision was shown across lots, and linearity was observed across the measuring range (1.5 to 9500 ng/L, all Pearson's r = 1.00). No interferences were observed, specified up to ≤1000 mg/dL hemoglobin, ≤50 mg/dL [≤855 µmol/L] icterus/bilirubin, and ≤1200 ng/mL biotin. Conclusions The analytical performance characterization of the assay demonstrated high sensitivity, high precision at the low end and across the measuring range, and resistance to interference.

Abstract Background IDH-mutant astrocytomas are classified as WHO grade 4 in the presence of conventional high-grade histologic features and/or homozygous CDKN2A/B deletion in the 5th edition of the WHO Classification of Central Nervous System Tumour guidelines. However, work over the past decade has indicated a number of other molecular alterations that warrant consideration as potential prognostic markers. Methods We used univariate Kaplan-Meier and multivariate Cox proportional hazards regression analysis to evaluate the prognostic effects of homozygous CDKN2A/B deletion, CDK4 amplification, CCND2 amplification, PDGFRA amplification/mutation, PIK3R1 mutation, PIK3CA mutation, MYCN amplification, EGFR amplification/mutation, TERT promoter mutation, and grade 4 histologic features in two independent cohorts of WHO grade 2-4 IDH-mutant astrocytoma (n = 840 and n = 367). Results The presence of CDK4 amplification, CCND2 amplification, PDGFRA alteration, PIK3R1 mutation, MYCN amplification, and EGFR alteration were each associated with reduced overall survival compared to WHO grade 2/3 astrocytomas without these molecular features. 17.7% (148/837) of otherwise grade 2/3 astrocytomas had one or more of these molecular criteria, with resulting intermediate clinical outcome in terms of overall survival (median survival of 67.3-82.0 months) compared to grade 2/3 astrocytomas without these molecular features (median survival of 135.0-140.7 months) and grade 4 astrocytomas (median survival of 35.3-45.0 months). Conclusions The presence of CDK4, CCND2, PDGFRA, PIK3R1, MYCN, and EGFR alterations result in an intermediate patient survival in IDH-mutant astrocytoma. Adding these molecular alterations should be considered in future diagnostic classification systems to improve stratification of high-risk patients.

Abstract Purpose Although the Woven EndoBridge (WEB) device is increasingly used for the treatment of wide-neck intracranial aneurysms, including in the acute rupture setting, comparative evidence assessing the impact of rupture status remains limited. This study compared angiographic, safety, and clinical outcomes between ruptured and unruptured intracranial aneurysms treated with WEB. Methods We conducted a retrospective analysis of prospectively collected data from the multicenter cohort registry WorldWideWEB, including consecutive adult patients with intracranial aneurysms treated with the WEB. Patients were stratified into groups of ruptured and unruptured aneurysms. Propensity score matching was used to balance baseline characteristics between both groups. Retreatment rate was the primary outcome. Secondary outcomes included mRS, safety events (thromboembolic complications) and angiographic outcomes (periprocedurally and last follow-up). Results Among 1,220 patients, 342 (28.0%) presented with ruptured aneurysms. There was no significant difference in the rate of thromboembolic complications between groups (11.8% vs 5.9%, p = 0.056). At follow-up, adequate occlusion was lower (82.2% vs 93.3%, p < 0.002) and functional outcomes were worse (mRS ≥2 in 34.1% vs 21.9%, p = 0.012) among patients with ruptured aneurysms. Periprocedural adequate occlusion (53.3% vs 53.8%, p>0.9) and retreatment rates (11.8% vs 7.1%, p = 0.14) did not differ between groups after matching. Conclusion Ruptured aneurysms demonstrated expected inferior follow-up functional and angiographic outcomes when compared with unruptured aneurysms, but no difference in retreatment rate and procedural safety. These findings support WEB as a safe and effective treatment option for appropriately selected ruptured intracranial aneurysms in routine clinical practice.