Paludo, Jonas

Abstract Background: De novo secondary central nervous system (CNS) lymphoma (SCNSL), characterized by concurrent systemic and CNS aggressive B-cell lymphoma at diagnosis, is a rare entity with a dismal prognosis. The prognostic significance of the initial anatomical compartment of CNS involvement (parenchymal vs. leptomeningeal) is not well-defined. We sought to analyze outcomes based on CNS localization in a single-institution cohort. Methods We retrospectively reviewed 34 patients with de novo SCNSL (including diffuse large B-cell lymphoma [DLBCL], high-grade B-cell lymphoma, not otherwise specified (HGBCL NOS), and double-/triple-hit lymphoma) treated at the Mayo Clinic between 2002 and 2025. Clinicopathological features, radiologic findings, treatment details, and survival outcomes were abstracted from the medical records. Patients were categorized by the primary site of CNS involvement. Overall survival (OS) and progression-free survival (PFS) were estimated using the Kaplan-Meier method, and groups were compared with the log-rank test. Results The study cohort had a median age of 61.5 years (range 19-79) and a median International Prognostic Index score of 3 (range 2-4). Pathological subtypes included double-/triple-hit lymphoma (n=14, 41.2%), non-germinal center B-cell like (non-GCB) DLBCL (n=11, 32.4%), germinal center B-cell like (GCB) DLBCL (n=8, 23.5%), and HGBCL NOS (n=1, 2.9%). Among patients with available data, 14 (43.8%) were positive for MYC rearrangement, 13 (54.2%) for BCL2 rearrangement, and 10 (45.5%) for BCL6 rearrangement. First-line regimens received by patients included MR-CHOP (n=18, 52.9%), DA-R-EPOCH (n=5, 14.7%), combined CODOX-M/IVAC (n=4, 11.7%), R-CHOP with intrathecal methotrexate (n=3, 8.8%), miscellaneous therapies (n=3, 8.8%), and MRT (n=1, 2.9%). The complete response rate was 58.8% (n=20). Intrathecal chemotherapy was administered to 61.8% of patients, and consolidation with autologous stem cell transplant was performed in 35.3%. Based on the CNS compartment involved, de novo SCNSL (DN-SCNSL) were classified into two subtypes: the leptomeningeal subtype, defined as leptomeningeal involvement with or without parenchymal CNS disease (LM-DN-SCNSL; n=22, 64.7%), and the parenchymal subtype, defined as parenchymal-only CNS involvement (P-DN-SCNSL; n=12, 35.3%). Parenchymal disease was multifocal in 75% of cases, supratentorial in 58.3%, and associated with a midline shift in 30%. Although no tested factors (including pathological subtype, age, sex, and bone marrow involvement) were significantly associated with the anatomical compartment, a non-significant trend was noted for BCL6 rearrangement to be more common in the leptomeningeal group (8/13, 61.5%) than the parenchymal group (2/8, 25.0%) (P = 0.18). Survival outcomes were analyzed for both anatomical and pathological classifications. For the parenchymal versus leptomeningeal cohorts, 2- and 5-year Overall Survival (OS) rates were 75.0% vs. 45.1% and 41.0% vs. 38.7%, respectively (P = 0.34). For Progression-Free Survival (PFS), the 2- and 5-year rates were 48.6% vs. 32.5% and 29.2% vs. 26.0%, respectively (P = 0.53). The 2-year OS rates by pathological subtype were 81.8% for non-GCB DLBCL, 45.8% for DHL/THL, and 37.5% for GCB DLBCL (P = 0.27). Survival outcomes (OS/PFS) were not significantly associated with patient age, sex, pathological subtype, or anatomical compartment. Conclusion DN-SCNSL can be anatomically classified into two subtypes: LM-DN-SCNSL and P-DN-SCNSL. While a prognostic trend favoring the parenchymal subtype was observed, the difference did not achieve statistical significance, a finding likely attributable to the small cohort size. This classification may hold clinical utility for guiding risk stratification, and larger studies are needed to validate these findings and identify the most optimal treatments for each subtype of DN-SCNSL.

Abstract Description Cellular immune responses are critical for protection from environmental and genetic insults, including pathogens and cancer. However, immunity can also drive pathogenic responses that damage organs and drive pathogenesis. Immune response quality is influenced by several factors including genetics, age, sex and the environment. Environmental influences including everyday activities and lifestyle have been shown to modulate the immune system but are often difficult to monitor. We have developed a system of serial monitoring of peripheral blood mononuclear cells that coincided with wearable device activity monitoring to further understand how biometric data correlates with the composition of blood derived immune cells. Using a 31-marker CyTOF phenotyping panel we monitored blood derived immune cell composition over a 14-week period in an individual volunteer while biometric data was collected via wearable technology. To determine whether interventions can be introduced that change peripheral blood immune cell phenotypes our volunteer introduced lifestyle changes at specific intervals in this period of monitoring. These included intermittent fasting and the introduction of regular exercise. In addition, the individual was immunized against COVID-19 during this time to provide an immune challenge that would perturb immune homeostasis. In this study, we find that lifestyle changes can promote alterations in immune cell phenotypes identified using high parameter mass cytometry. Topic Categories Translational and Interventional Immunology (TI)

592 Background: While not all fevers occurring during hematopoietic stem cell transplantation (HCT) are dangerous, episodes of febrile neutropenia tend to be managed aggressively given the significant mortality and morbidity associated with infectious complications of transplantation. This can potentially lead to unnecessary diagnostic testing, antibiotic treatment, and prolonged hospital stays for non-infectious fever etiologies such as engraftment. Our group has previously demonstrated that high-frequency body temperature monitoring (HFTM) using a wearable device detects fevers earlier in transplant recipients (Flora et al, Cancer Cell, 2021). We hypothesized that applying supervised machine learning to HFTM data collected from inpatient allogeneic HCT patients would allow for accurately distinguishing between non-infectious and infectious etiologies of fever and could potentially guide safe de-escalation of antibiotics. Methods: We analyzed 74 independent fever events detected in HCT patients (n=42) outfitted with an FDA-cleared, wireless sensor measuring axillary temperatures every 2 minutes for the duration of their respective hospitalizations. After data filtering, independent fever events were then retrospectively grouped by fever etiology using physician-determined annotation as previously described (Flora et al., 2021). A split ratio of 80:20 was used to create training and test sets from the parent data set, consisting of 59 and 15 fever events for each split, respectively. Training sets were used to create a classification algorithm identifying bacterial infections requiring antibiotics using the machine learning model XGBoost. The performance of this model was then evaluated using test sets of the remaining fevers (n=15). Results: Averaging over 1000 test sets of 15 independent fevers comprising 5 bacterial infection-related fevers and 10 non-infectious fevers, our model correctly identified 82% (mean recall) of cases requiring antibiotics with a mean precision of 0.41. The model also correctly identified 42% of non-infectious fevers (mean true negative rate, 0.42) that would otherwise have been treated with unnecessary antibiotics. The mean AUCPR was 0.58, with the best performing individual model achieving an AUCPR of 0.97. Conclusions: The preliminary results of this study demonstrate that our classification model shows promise for differentiating fevers requiring antibiotics due to bacterial infections from other non-infectious fever etiologies. Prospective studies further developing supervised machine learning models for subcategorization of febrile neutropenia could pave the way for safely de-escalating antibiotics in HCT and other immunocompromised patients undergoing cancer treatment monitored using HFTM.

207 Background: CAR-T & bispecific antibodies (BsAb) for large B-cell lymphoma offer deep responses, but have unique toxicities requiring close monitoring. Different practice settings have varied strategies for managing these toxicities, leading to increased costs, healthcare resource use, & reduced quality of life. With no direct comparison, rationale for therapy selection may vary. This study explores provider choices, sequencing strategies & access barriers. Methods: A survey of US-based Hematologists was conducted on practice demographics, therapy availability, preferences, & barriers in LBCL. Responses were analyzed across clinical scenarios. Results: Responses were received from 45 physicians. Availability of CAR-T & BsAb varied with 75% prescribing both, & only BsAb being available at 9% of providers’ centers. There was a strong preference for CAR-T over BsAb—higher among academic vs. nonacademic providers (96% vs. 78%; p = 0.03) & at CAR-T vs. non-CAR-T centers (100% vs. 72%; p = 0.001). Even at BsAb-only centers, preference leaned toward CAR-T, though notably lower compared to CAR-T centers. Among centers without either, 86% preferred CAR-T. Practice volume had no impact. Reasons for favoring CAR-T included perceived better outcomes (55%), shorter treatment (26%), center choice (19%), more experience (48%), & curative potential (21%). BsAb selection was driven by CAR-T unavailability & logistical barriers (distance to specialized center, insurance delays, referral complexity, manufacturing time, & hospitalization needs). Axi-cel was preferred by 69% ( p = 0.0001) vs. other CAR-T for third-line or later due to experience with use, response rates, center choice, & logistical ease. Liso-cel (26%) & Tisa-cel (2%) were chosen for perceived favorable toxicity over Axi-cel. Axi-cel was also preferred second-line vs. Liso-cel (73% vs. 24%; p < 0.0001) for similar reasons. Among BsAb, Epcoritamab was favored by 56% for administration ease (subcutaneous), experience, & center preference; Glofitamab by 42% for shorter duration, experience, & center preference. There was no difference by academic vs nonacademic practice & volume for both CAR-T & BsAb. In patients with prior non-CAR-T anti-CD19 therapy, providers chose CD19 CAR-T over CD20 BsAb (80% vs. 18%; p < 0.0001). For relapse after second-line CAR-T, 78% providers preferred BsAb as the next line over other options ( p < 0.0001) due to more favorable response rates, toxicity profile, & experience with use. Providers prioritized response rates, survival, & toxicity profile in their treatment decisions. Logistics, duration, frequency, & cost were less important. There were no differences based on practice setting & CAR-T experience. Conclusions: Despite higher preference for CAR-T due to perceived superior outcomes, access barriers exist. Efforts to streamline referral pathways, logistics, & expand availability are crucial to improve outcomes.

1652 Background: Tarlatamab is a bispecific T-cell engager targeting DLL3 that has shown promise as a therapy in small cell lung cancer. Administering tarlatamab requires careful consideration of immune-related adverse events (AE), including cytokine release syndrome (CRS) and immune effector cell-associated neurotoxicity syndrome (ICANS). Prescribing information recommends patients to be monitored for 22-24 hours in an appropriate healthcare setting for cycle 1 (C1), days 1 and 8. At Mayo Clinic in Rochester MN (MCR), tarlatamab is initially administered in a hospital-based outpatient (HBO) setting utilizing an innovative remote patient monitoring system (RPM). In this study, we describe our RPM process and patient outcomes thus far. Methods: We retrospectively reviewed records of patients treated with tarlatamab from August through December 2024. All patients received tarlatamab in the HBO setting and were enrolled into RPM for C1, days 1–10. Patients attended daily HBO visits on days 1–3 and 8–10 of C1 and were required to stay within 30 minutes of the hospital with a 24-hour caregiver for 48 hours following infusion days in C1-2. Data on patient characteristics, frequency of RPM alerts, escalations, rate of CRS & ICANS, and need for hospitalizations were reported using descriptive metrics. Results: As part of RPM, patients are provided with a kit containing preconnected Bluetooth-enabled devices to measure vital signs (VS), including blood pressure (BP), heart rate (HR), temperature (T), and pulse oximetry (SpO2). The kit also includes a cellular-enabled tablet for electronic ICANS questionnaires, which upload directly into our electronic medical record system. Patients are required to log VS four times daily, monitored in real-time by a centralized virtual RPM nursing (RN) team. Embedded decision trees trigger alerts, prompting the RN team to contact patients and follow care pathways for escalations as needed. Among the 16 patients treated, the median age was 68 years, and the median ECOG PS was 1. During the 10-day RPM period, a total of 2,233 VS entries were recorded. Of these, 88% (14/16) of patients had alerts triggered, with a median of 2 alerts per patient. Alerts were mostly for out-of-range systolic BP (56%), HR (56%), and T (25%), with none for SpO2. The RN team escalated alerts in 50% (8/16) of the patients. Overall, 38% (6/16) of patients were managed entirely in the HBO setting. Among the 63% hospitalized, 56% had CRS limited to grade (G) 2, and 31% had ICANS limited to G3. Two patients were hospitalized for other indications. The mean length of stay was 2.6 days, with no ICU admissions. Conclusions: Our experience shows that frequent monitoring required with tarlatamab can be safely and effectively executed in an outpatient setting with the utilization of an RPM system. It can potentially minimize the burden of hospitalizations for patients and the healthcare system.

e13622 Background: Chimeric antigen receptor T-cell (CAR-T) therapy has revolutionized treatment of hematologic malignancies; however, severe toxicities like cytokine release syndrome (CRS) remain a limitation to broader utilization. Using a large dataset of patients treated with CAR-T, we developed a machine learning (ML) model leveraging multi-model patient characteristics to predict the risk of CRS onset with clinically meaningful lead times. Methods: A prospectively maintained dataset of CAR-T patients treated at Mayo Clinic was used. Baseline disease/patients characteristics, labs, ECG, echocardiogram, and vital signs (including data from remote patient monitoring program) were collected. A 24-hour rolling window was used to generate daily averages, maximum, minimum, and deltas for each parameter. ECG and echocardiogram data were forward-imputed; vital signs and labs were not imputed. Each day-post-infusion served as a potential event (CRS onset vs. no CRS). A gradient boosting classifier was trained and tested on a subset of patients (n = 351) using 1,000 bootstrap iterations, evaluating 24-, 48-, and 72-hour lead times. Shapley-based feature importance guided simplification to an 11-feature model, externally validated on an independent cohort (n = 51). Models’ performance assessed using the area under the receiver operating characteristic curve (AUC). Results: A total of 402 patients treated with CAR-T were examined. Median age was 64 years and 36% were female. Primary diagnosis was B-cell lymphoma (n = 240) or multiple myeloma (n = 162). A total of 226,332 vital sign measurements and 372,871 lab values, as well as ECG and echocardiogram data, were collected for 30 days post-infusion. Overall, 79.8% of patients experienced CRS within 30 days. The full ML model demonstrated AUCs of 0.90 (95% CI 0.86–0.93), 0.83 (95% CI 0.77–0.88), and 0.82 (95% CI 0.77–0.87) for 24-, 48-, and 72-hour predictions of CRS, respectively. Including “day of CRS onset post CAR-T” as a feature improved the 72-hour AUC to 0.87 (95% CI 0.83–0.90). A model restricted to vital signs alone showed AUCs of 0.88, 0.75, and 0.71 for 24-, 48-, and 72-hour horizons. An 11-feature model—focusing on temperature, blood pressure, fibrinogen, CRP, serum sodium level, QRS interval, and lymphocyte counts—maintained high accuracy at 24 hours (AUC 0.90, 95% CI 0.86–0.93) and was validated externally using an independent dataset with an AUC of 0.92. Conclusions: Our ML model provides a reliable approach for predicting if/when CRS will occur, with up to three days of lead time before clinical manifestation, offering an opportunity for proactive management. The robust performance of a parsimonious 11-feature model supports practical adoption across clinical settings and real-time risk stratification to enhance patient safety. Prospective validation and integration into clinical workflows will be critical to translating these findings into practice.