Citations

Background: Accurate characterization of tumor stiffness and adhesion is crucial when surgically resecting meningiomas. Currently, such determinations are difficult to ascertain through standard preoperative imaging diagnostics and clinical experience. Herein, we explored magnetic resonance elastography (MRE) as a means of assessing these parameters and thereby improving neurosurgical resections. Methods: This prospective study of 193 patients with meningiomas involved three distinct patient subsets (cohort 1, 88 [2017–2021]; cohort 2, 55 [2022–2023]; cohort 3, 50 [2023–2024]. We analyzed preoperatively obtained MRE data on meningioma stiffness (cohorts 1 and 2) and adhesion (cohort 2) for accuracy, while matching subjects with or without MRE studies (cohort 3, 1:1 ratio) to define the role of preoperative MRE in improving operative outcomes. Results: Areas under the curve (AUCs) for MRE in gauging meningioma stiffness and adhesion were 0.87 and 0.89, respectively. Both parameters significantly surpassed empiric predictions of stiffness and adhesion rendered by surgeons and radiologists, based on routine magnetic resonance imaging (MRI). MRE-determined stiffness measures independently correlated with prolonged resection duration, whereas MRE-assessed adhesion showed independent associations with increased blood loss, longer duration of tumor resection, and higher risk of pial injury. Ultimately, results based on cohort 3 demonstrated that MRE has great potential in this setting to improve neurosurgical procedures by significantly reducing intraoperative bleeding, resection duration, and risks of pial injury. Conclusion: MRE is a promising, noninvasive tool for accurately assessing meningioma stiffness and adhesion, standing to improve the precision and safety of needed surgical intervention.

Abstract Background: For large language models (LLMs) to reach their potential as information technology tools that make medication use safer, clinically relevant benchmarks capable of automated grading and designed specifically to measure the performance of LLMs for medication tasks are required. The purpose of this study was to design a suite of benchmarking tests reflective of Comprehensive Medication Management (CMM; the standard of care for medication optimization) and quantify the baseline performance of the latest LLMs. Methods: We established six benchmarks representing critical stages of the CMM process: drug formulation matching, drug order (sig) generation, drug route matching, drug-drug interaction identification, renal dose identification, and drug-indication matching. For each benchmark, we curated a clinician-annotated dataset comprising 250 standardized input-output pairs including both inpatient and outpatient medications. We evaluated the clinical knowledge retrieval capabilities of three LLMs: GPT-4o-mini, MedGemma-27B, and LLaMA3-70B. We employed a zero-shot prompting strategy, excluding in-context examples, to assess the models’ internal clinical knowledge rather than their few-shot learning potential. To check reliability, each model was run three times using a temperature of 0.7 (a mid-range value of an LLM setting controlling text generation randomness). Performance was assessed using task-specific evaluation metrics including precision (positive predictive value), recall (sensitivity), F1-score, accuracy, and correctness consistency across trials. Results: Across six benchmarks, LLaMA3-70B demonstrated the highest performance in four tasks: drug-formulation matching (F1, 54.0% [95 CI: 50.1-58]), drug-order generation (accuracy, 88.0%), drug-route identification (F1, 74.3% [95 CI: 71-78]), and drug-indication identification (accuracy, 97.6% [95 CI: 95.6-99.2]). In the drug–drug interaction task, GPT-4o-mini achieved the highest overall accuracy (70.4% [95 CI: 64.8-75.7]). For renal dose–adjustment identification, GPT-4o-mini demonstrated the highest F1 score (83.3% [95 CI: 77.6-88]). Correctness-consistency scores ranged from 8.0% to 97.6% across benchmarks, with no model exhibiting uniformly superior consistency. Conclusions: Model performance varied substantially across medication-related tasks. LLaMA3-70B demonstrated promising baseline performance in tasks involving formulation, ordering, route, and indication. GPT-4o-mini showed potential advantages in drug–drug interaction detection and renal dose adjustment. These findings underscore the need for task-specific evaluation when deploying models for medication-focused clinical decision support.

Abstract Immune checkpoint inhibitor–associated immune thrombocytopenia (ICI-ITP) has been described in case reports and small case series, but comprehensive data on its incidence, risk factors, clinical features, treatment, and outcomes are lacking. We reviewed medical records of all adults initiating ICI therapy between 2016 and 2023 at 29 US hospitals across 7 major cancer centers to identify patients with ICI-ITP. Multivariable logistic regression was used to identify risk factors, and Cox modeling was performed to assess the association between ICI-ITP, its severity, and mortality. Among 86 467 patients, ICI-ITP occurred in 214 (0.25%). Independent risk factors included lower baseline platelet count, combination ICI therapy, stage IV cancer, and additional immune-related adverse events. ICI-ITP occurred at a median of 8 weeks (interquartile range [IQR], 4-18) after ICI initiation, with a median nadir platelet count of 41 × 109/L (IQR, 17 × 109/L to 64 × 109/L). Patients were treated with glucocorticoids (n = 106 [49.5%]), immune globulin (n = 39 [18.2%]), and thrombopoietin receptor agonists (n = 29 [13.6%]). Recovery occurred in 161 patients (75.2%) at a median of 2.3 weeks (IQR, 1.0-5.3). Of 76 patients rechallenged with ICIs, 23 (30.3%) developed recurrent ICI-ITP. ICI-ITP and its severity were associated with higher all-cause mortality, with a nearly threefold increase in risk among patients with severe ICI-ITP than those without ICI-ITP (adjusted hazard ratio, 2.96 [95% confidence interval, 2.14-4.08]). These findings establish ICI-ITP as a rare but clinically significant complication of ICI therapy, provide, to our knowledge, the first large-scale description of its risk factors and clinical course, and underscore the importance of timely recognition and management.

Abstract Background The purpose of this study was to develop a series of benchmarks specifically for large language model (LLM) performance in the identification of drug-drug interactions (DDIs). Methods We evaluated six LLMs using a clinician-designed dataset of 750 DDI scenarios. Tasks included: (1) a pointwise two-drug classification task (2) a pairwise three-drug discrimination task and (3) a listwise 4–6 drug selection task. Zero-shot prompting was applied and performance was assessed using precision, recall, F1 score, accuracy, and self-consistency. Results Model performance varied by task structure and interaction severity. GPT-5-Chat demonstrated the highest recall and F1 score in the pointwise task (80.7% and 80.2%), whereas DrugGPT achieved superior accuracy in the pairwise (93.7%) and listwise task (92.6%). All models had improved performance with more severe compared to less severe DDIs (accuracy 72.0-99.2% vs. 58.3-86.7%). All models tended to have higher self-consistency in the pairwise and listwise experiments compared to the pointwise experiment (68.8-93.6% vs. 31.0-74.2%). No model exhibited uniformly high reliability across all tasks. Conclusions Current LLMs show inconsistent capabilities in identifying DDIs. Performance degrades as reasoning space expands with limited stability. These findings emphasize the need for multi-format evaluation frameworks and reliability-aware assessment when considering LLMs for medication-safety applications.

ABSTRACT Background Older adults are at high risk for severe injuries and death from falls. Physical therapist (PT) and occupational therapist (OT) evaluations have been introduced into Emergency Department (ED) care to assist with fall risk evaluation and fall prevention care. An evaluation of current evidence was undertaken to inform the Geriatric ED Guidelines 2.0. Methods Systematic review of physical and/or occupational therapy evaluation for fall prevention in the ED. The intervention was evaluation by a rehabilitation therapist (PT, OT, or both) for fall prevention, fall assessment, or mitigation of fall risk factors. Studies that did not perform the evaluation during the ED or ED Observation Unit visit were excluded, such as referrals for home consultation. The published literature was searched using strategies created by a medical librarian and was implemented in Embase and PubMed in March 2025. Covidence was used for article collation and review. Risk of bias was assessed using the Cochrane Risk of Bias v2 and the Ottawa–Newcastle scale. Results The search resulted in 387 articles with 52 duplicates, for 335 unique citations. Ten articles on 6 datasets were included (4 abstracts and 6 published manuscripts). One article was at high risk of bias as very few intervention patients received therapist evaluation. The remaining 5 datasets had varied co‐interventions in addition to PT and/or OT evaluations. PT or OT evaluation in the ED was associated with reduced ED revisits at 1, 2, and 6 months. Evaluations in the ED were not associated with increased hospital admission rates or ED length of stay. A meta‐analysis was infeasible due to varied outcome timeframes. Conclusions PT and/or OT evaluations for fall prevention were associated with reduced subsequent ED revisits for older adults. However, the impact of co‐interventions and heterogeneity limits strong conclusions.

Background. As cancer therapeutics improve, more liver transplant (LT) candidates will have a history of prior to transplant malignancy (PTM), an established risk for cancer posttransplant. Methods. This study provides granular analysis (with focus on nonliver PTM) of posttransplant cancer risk relating to PTM in 1105 consecutive adult LT recipients (2009–2019, N = 542 with PTM). Results. PTM prevalence increased from 37.8% to 63.2% during this timeframe. Post-LT survival was lower (10 y: 61.2% versus 75.1%; P < 0.001) with higher cancer-related mortality (17.6% versus 2.2%; P < 0.001) in patients with any PTM. The higher cumulative incidence of cancer post-LT associated with PTM (59.7% versus 37.6%; P < 0.001) was not attributable to de novo cancer (38.5% versus 37.7%; P = 0.21) but to recurrence. Cancer recurrence rate (10 y) was lowest in nonliver/nonskin (“other”) PTM (10.6%) with mortality rates from recurrence of only 4% in this cohort. De novo cancers may present at more advanced stage in patients with PTM and lung cancer the most common nonskin de novo malignancy (PTM 21%, no PTM 7%). Family history of cancer was an independent predictor for de novo cancer. Conclusions. PTM associated post-LT cancer outcomes are dominated by skin and liver cancer recurrence and not recurrence of “other” cancers or de novo cancer. This data supports the guidance proposals for cautious liberalization of “other” cancer remission requirements before transplantation. As post-LT cancer is highly prevalent regardless of PTM status, adherence to posttransplant cancer screening and surveillance is critical to identify early cancer after transplant.

Abstract Background: Routine laboratory testing prior to chemotherapy administration is a cornerstone of hematology and oncology practice, aimed at ensuring the safety and efficacy of treatment. However, over time, standardized protocols may accumulate excessive or low-yield laboratory tests that offer limited value to clinical decision-making. In 2018, Mayo Clinic implemented standardized treatment protocols using Epic Beacon, an electronic health record module that integrates chemotherapy regimens with associated laboratory testing and clinical workflows. While well-intentioned, these protocols may include redundant tests that delay treatment initiation, increase patient discomfort and anxiety, and burden providers with excessive in-basket communications. We hypothesize that many laboratory tests currently embedded within prechemotherapy workflows are unnecessary and can be safely omitted without compromising patient care. Methods: We conducted a retrospective cohort study evaluating the pre-treatment laboratory testing associated with the R-CHOP (rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone) chemotherapy protocol within the Mayo Clinic Rochester Division of Hematology. The study included 100 patients diagnosed with diffuse large B-cell lymphoma who received treatment between January 1 and December 31, 2024. Patients were excluded if they had incomplete medical records or received treatment outside the Mayo Clinic system. Data was abstracted via chart review using the Mayo Data Explorer. We analyzed complete blood counts (CBC), which include 12 individual tests, and comprehensive metabolic panels (CMP), which include 16 individual tests, obtained prior to each treatment cycle. For each cycle, we assessed the proportion of abnormal laboratory results and the proportion of abnormal results deemed clinically relevant which were defined as those that prompted a treatment delay, dose reduction or early termination of a cycle. Results: The cohort comprised 100 patients with a mean age of 68.2 years (SD 13.4) and a median age of 71 years (IQR 58.75–77); 57 were male and 43 were female. Based on available documentation, disease stage was classified as limited (stage I/II; n=41) or extensive (stage III/IV; n=59). A total of 519 R-CHOP cycles were administered, including 419 subsequent cycles (cycles 2–6). Fifteen patients discontinued treatment early, although none of the discontinuations were attributable to abnormalities in pre-infusion laboratory testing. Among the 419 subsequent cycles, 34 (8.1%) experienced treatment delays, with only one delay resulting from a clinically significant laboratory abnormality (hyperglycemia). Additionally, 23 patients underwent regimen modification (dose reduction or drug holding), but none of these were due to pre-infusion lab abnormalities.Of the 411 comprehensive metabolic panels (CMPs) obtained prior to these cycles, 315 (76.6%) contained at least one abnormal value, yielding a total of 786 abnormal results. Similarly, 405 of 415 (97.6%) complete blood counts (CBCs) demonstrated at least one abnormality, generating 2,696 abnormal values in total. Despite the high frequency of abnormal findings, the overall clinical yield was minimal, with only 1 of 826 pre-infusion laboratory panels (0.12%) leading to a modification in clinical management. Conclusion: Routine pre-infusion laboratory testing prior to R-CHOP chemotherapy frequently identifies abnormal results. However, these abnormalities are rarely clinically significant and do not affect treatment decisions. These findings support the reconsideration of current testing protocols to reduce unnecessary testing, enhance workflow efficiency, and minimize patient and provider burden without compromising safety.

Abstract Introduction: Patients with triple-class exposed (TCE) relapsed/refractory multiple myeloma (RRMM) have historically experienced poor clinical outcomes, prompting the development of novel therapeutic options. Idecabtagene vicleucel (ide-cel) was the first chimeric antigen receptor (CAR) T cell therapy approved for this population, while teclistamab became the first approved bispecific antibody. Both target B-cell maturation antigen (BCMA) and are increasingly used in real-world (RW) settings. Most comparative evidence to date is derived from single-arm clinical trials with cross-trial adjustments. RW studies offer an opportunity to directly compare these therapies and assess their relative effectiveness in routine clinical practice. Methods: We conducted a comparative effectiveness analysis of ide-cel and teclistamab in patients with TCE RRMM using RW data from the United States Flatiron Health electronic health record-derived, deidentified database (January 2011 to April 2025 data cut). Adults who received either index therapy at any line of treatment after becoming TCE (ie, previously exposed to an immunomodulatory agent, a proteasome inhibitor, and an anti-CD38 monoclonal antibody) were included. Exclusion criteria were prior exposure to CAR T cell therapy, bispecific antibodies, or belantamab mafodotin; absence of index treatment; or lack of post-index follow-up. Index dates (T0) represent the date of infusion for ide-cel and the start of treatment for teclistamab. Seventeen prognostic covariates were selected a priori based on published rankings and used to balance treatment groups via inverse probability of treatment weighting (IPTW) with average treatment effect weights. Cox proportional hazards models were used to estimate unadjusted and IPTW-adjusted HRs with 95% CIs for progression-free survival (PFS) and overall survival (OS). As the ide-cel cohort only included infused patients, additional analyses were performed to emulate a real-world intention-to-treat population given the potential immortal time bias (patients in the ide-cel cohort were required to survive between apheresis and infusion). Specifically, patients in the teclistamab cohort who died within the first 24 days (median turnaround time for ide-cel manufacturing in United States in 2024) were excluded and 2 scenarios were performed to re-align T0 based on this 24-day period (either shifting teclistamab T0 forward or ide-cel T0 backward). Lastly, sensitivity analyses assessed differences in restricted mean survival time (dRMST) at 12 and 24 months. Results: A total of 281 patients met the inclusion criteria (ide-cel, N=106; teclistamab, N=175), with median follow-up of 22.4 and 13.7 months, respectively. IPTW adjustment balanced all covariates between groups (effective sample size: ide-cel, n=68; teclistamab, n=135). Adjusted analyses showed improved PFS with ide-cel versus teclistamab (HR, 0.67; 95% CI, 0.47–0.97), with median PFS of 11.3 versus 6.5 months. OS was also improved with ide-cel (HR, 0.59; 95% CI, 0.35–0.98); median OS was not reached in either group due to insufficient follow-up. There was a consistent trend of HRs favoring ide-cel versus teclistamab for both PFS and OS in both scenario analyses addressing the potential immortal time bias. Further, sensitivity analyses using dRMST confirmed statistically significant improvements in both PFS and OS with ide-cel across all scenarios at all time points evaluated. Conclusions: In this RW analysis, ide-cel was associated with superior PFS and OS compared with teclistamab in patients with TCE RRMM. These findings support emerging RW evidence favoring BCMA-directed CAR T cell therapies over bispecific antibodies and may inform treatment sequencing decisions in this high-risk population.

Abstract BACKGROUND Children exposed to poverty during frontline therapy for B-lymphoblastic leukemia (B-ALL) experience inferior overall survival (OS), higher relapse rates, and earlier relapse than their peers. Real-world data describing treatment patterns and survival after relapse of B-ALL are lacking and, as a result, the impact of poverty following relapse remains poorly understood. This study evaluated the association of poverty exposure, proxied by insurance status, with post-relapse survival and explored potential explanatory mechanisms, including disease biology and therapy utilization. METHODS Data from Retrospective Study of Contemporary Approaches to First Relapse in B-ALL (ReCALL-1), a multicenter retrospective cohort study of patients &lt; 30 years old with first relapse of B-ALL from 2018-2022, were used. The primary exposure was poverty, proxied by insurance status at relapse, categorized as any private insurance (poverty unexposed), public-only insurance (poverty exposed), no insurance, or unknown; the latter two were not included in comparative analyses due to low numbers. The primary outcome was OS. Secondary outcomes included measurable residual disease (MRD)-negative second complete remission (CR2), relapse after CR2, non-relapse mortality (NRM), with relapse treated as a competing risk, and receipt of cellular and immunotherapies. RESULTS Among 462 patients from 31 U.S. institutions, 41% had private insurance, 47% public-only insurance, 2% no insurance, and 10% unknown. Compared to privately insured patients, those with public insurance were more likely to be Hispanic (56% vs 19%, p&lt;0.001) and have non-English preferred language (30% vs 6%, p&lt;0.001). Non-Hispanic Black patients comprised 9% of those with public insurance and 5% of those with private insurance. Disease characteristics that differed by insurance status at relapse included: Ph-like cytogenetics, more common in publicly insured (23% vs 11%, p=0.002), and infant ALL, more common in privately insured patients (10% vs 5%, p=0.06). Publicly insured patients were more likely to have experienced major organ toxicity during frontline therapy (32% vs 18%, p=0.002) and very early relapse (&lt;18m from diagnosis: 43% vs 32%, p=0.07). Other demographic and disease characteristics, including relapse site and cytomolecular features, were similar by insurance status. With a median follow-up of 51 months from relapse, 4-year OS differed by insurance (log-rank p=0.02): 72% for private vs 56% for public insurance. After adjusting for race/ethnicity and language, public insurance increased the hazard of death by 65% (aHR 1.65 95% CI 1.13-2.42, p=0.004). The association remained, but was attenuated, when unbalanced disease characteristics (time to relapse, infant ALL, Ph-like cytogenetics, prior organ dysfunction) were included (aHR 1.47, 95% CI 1.00-2.16, p=0.05). Rates of MRD-negative CR2 were similar (85% public vs 87% private). However, point estimates suggested publicly insured patients had more relapse after CR2 (44% vs 37%, p=0.42) and higher 4-year NRM (16% vs 10%, Gray's test p=0.12), though neither were significant. Treatment choices for first relapse did not differ by insurance status. Utilization of intensive reinduction (91% vs 95% for public vs private, p=0.18) and clinical trial enrollment (22% for both) were similar. Receipt of cellular therapies did not significantly differ by public vs private insurance: Chimeric Antigen Receptor-T cells (CAR-T: 38% vs 34%, p=0.41), hematopoietic cell transplant (HCT: 42% vs 47%, p=0.32). Use of blinatumomab trended higher in those with private insurance (36% vs 45%, p=0.07), and inotuzumab was similar (11% for both, p=0.87). Median time from relapse to CAR-T or HCT was comparable (to CAR-T 102 vs 100 days, to HCT 117 vs 114 days). CONCLUSION This is the first study to demonstrate differential survival by insurance type, a proxy for poverty exposure, after first relapse of B-ALL. In this large multicenter, real-world cohort, patients with public insurance had worse OS after relapse than those with private insurance. The survival disparity existed independent of other SDOH factors. Adjustment for known high-risk disease features only partially accounted for the observed differences in OS. In contrast, rates of MRD-negative CR2 and receipt of CAR-T and HCT were comparable across insurance types. These findings highlight the urgent need to identify effective interventions that will ensure equitable outcomes after relapse.