Citations

Objective Closed-source large language models (LLMs) like generative pre-trained transformer 4o (GPT-4o) have shown promise for clinical information extraction but are potentially limited by cost, data security concerns and inflexibility. Open-source models are an attractive alternative with various adaptation strategies with no consensus on best practices. This study aims to rigorously identify optimal adaptation strategies for open-source models and evaluate their performance relative to closed-source alternatives. Methods and Analysis We studied three LLM adaptation strategies: chain-of-thought prompting, few-shot prompting and fine-tuning. Our target for information extraction was the Mayo Endoscopic Subscore (MES). We applied those strategies in all combinations to six open-source models (8–70 billion parameters) using an annotated set of colonoscopy procedure reports from the University of California, San Francisco (N=608) and San Francisco General Hospital (N=217). We analysed the relationship of these strategies to several performance metrics with a mixed-effects model, accounting for the variability between centres and LLMs. GPT-4o served as a closed-source oracle and provided in-depth commentary on the cost-effectiveness of these options. Results Quantised low-rank adaptation (QLoRA) statistically improves ( p §amp;lt; 0.001 ) ) the performance of open-source LLMs by 9.1–15.7 percentage points across accuracy, precision recall and annotation eligibility accuracy. However, GPT-4o with prompt engineering outperforms the best open-source model by 4.9%–11.2%. A simple cost-effectiveness analysis suggests that GPT-4o is more affordable compared with open-source alternatives. Conclusion GPT-4o is currently the most efficient LLM for MES extraction. If unavailable, QLoRA-optimised open-source models are a competitive alternative. However, results also suggest that current instruction-following LLMs including GPT-4o do not fully follow user-provided instructions, leaving room for improvement. More work is needed to achieve consistent, near-perfect performance in clinical information extraction by LLMs.

ABSTRACTWith the global rollout of COVID‐19 vaccines, vaccine safety remains a priority. Emerging concerns have raised the potential risk of a long COVID‐like syndrome following vaccination, informally called long Vax and provisionally termed post‐COVID‐19 vaccination syndrome (PCVS). Our narrative review describes the putative manifestation, pathophysiology, and therapeutic approaches of PCVS based on the available evidence, mostly from case reports/series and observational studies. Our review noted that PCVS typically manifests within days to weeks post‐vaccination, with symptoms lasting months to years. PCVS may present as recognized diagnoses such as postural orthostatic tachycardia syndrome (POTS), small‐fibre neuropathy (SFN), myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS), or as long‐term sequelae of myocarditis, vaccine‐induced thrombotic thrombocytopaenia (VITT), or immune thrombocytopaenia purpura (ITP). Symptomatically, PCVS overlaps with long COVID, such as fatigue and brain fog, but PCVS may involve more frequent paraesthesia and less dyspnoea. We also review pathophysiological hypotheses of PCVS, focussing on the vaccine‐derived spike protein and related immune responses. Finally, we discuss potential therapies used to treat patients with PCVS or related conditions, primarily documented in case reports/series, which could guide future clinical research. Overall, PCVS remains a poorly understood condition that requires more research to elucidate its prevalence, prognosis, risk factors, and treatments.

Abstract Objective This study aims to better characterize the long‐term neurological quality of life (QOL) outcomes (using the Neuro‐QOL scale) in survivors of new onset refractory status epilepticus (NORSE), including its subtype febrile infection‐related epilepsy syndrome (FIRES), and provide guidance for psychological and social support strategies. Methods Utilizing data from a multicenter prospective study of NORSE/FIRES led by Yale University, we enrolled patients who completed the validated, patient‐reported Neuro‐QOL scale at least once at 3–6 months ( n = 37), 12 months ( n = 29), 24 months ( n = 23), or ≥36 months ( n = 9) following discharge. The Neuro‐QOL scale assesses physical, mental, and social health in patients with neurological disorders. QOL impairment (QOL‐I) scores were calculated, with higher scores indicating greater impairment. T ‐scores enabled comparisons with reference populations. Results In adults, median QOL‐I improved from 44.1% at 3–6 months to 37.6% at 36+ months. Paired analysis showed significant improvement in QOL‐I between 3–6 and 24 months ( p = .016), with specific improvements in communication, satisfaction with social roles, fatigue, and mobility. Greater improvement was also observed for participation in social roles (5.5‐point T ‐score gain) compared to the reference population, suggesting meaningful change. A gradual improvement in overall QOL‐I scores was also observed in pediatric participants, despite a modest sample size ( n = 5 with data at 3–6 and 12 months). Measures of fatigue and anxiety persisted in adults, and cognitive difficulties persisted in both adults and children. In adults, longer status epilepticus duration and intensive care unit stay were associated with poorer QOL. Additionally, a higher number of antiseizure medications was associated with more depression, cognitive impairments, and perceived stigma. Significance These findings highlight the potential for recovery following an acute episode of NORSE, although many patients continue to face challenges requiring ongoing support, and the clinical meaning of the reported QOL improvement remains unclear. Furthermore, the findings underscore the importance of strategic multidisciplinary support systems in the years following discharge.

Abstract Extracellular matrix (ECM) stiffening correlates with tumor invasion in various cancer types, including glioblastoma (GBM). Increased matrix stiffness promotes a migratory phenotype through dysregulation of cell-ECM bidirectional communication. Exposure to stiffer environments is sensed by cells, which then adapt in ways that promote invasive behavior. These adaptive changes are imprinted onto the cells and persist even after they are placed in new, softer microenvironments via a process known as “mechanical memory.” Mechanical memory is believed to be driven by mechanosensitive transcription factor activity and epigenetic remodeling. Glioblastoma recurrence is linked to the ability of cells to disperse and infiltrate the surrounding healthy tissue. Extracellular matrix stiffness in GBM is heterogeneous; it starts with a softer tumor core and becomes progressively stiffer toward the tumor’s edges, potentially promoting sustained tumor invasion through mechanical memory. This review discusses the role of ECM stiffness in cancer cell behavior and the implications of ECM stiffening in GBM. We then describe the findings associated with mechanical memory and relay underlying mechanisms currently understood to drive the preservation of mechanically primed phenotypes. Finally, we discuss how matrix stiffness can drive migratory phenotypes in GBM cells and the potential role that progressive ECM dysregulation at the tumor periphery can link the formation of invasive tumor niches to the aggressive, resistant, and mesenchymal-like phenotypes present in GBM recurrent tumors.

Abstract Early diagnosis and treatment are essential for improving outcomes of coccidioidal meningitis. While the detection of IgG by serologic testing of the cerebrospinal fluid (CSF) has been a mainstay of diagnosis for years, the diagnosis of coccidioidal meningitis in clinical practice can be very challenging due to suboptimal sensitivity of laboratory tests. We reviewed the results of the CSF diagnostic test results from the initial lumbar puncture in 110 patients with proven, probable, and likely coccidioidal meningitis from 1989 to 2024. One hundred four patients were diagnosed with coccidioidal meningitis on the initial CSF examination. The positivity rate of the first CSF testing was 89.1% (n = 64) for IgG by enzyme immunoassay, 62.2% (n = 98) for IgG by immunodiffusion, and 70.2% (n = 104) for IgG by complement fixation, 4.4% (n = 90) for fungal culture, 18.8% (n = 69) for polymerase chain reaction, and 33% (n = 21) for Coccidioides antigen. The non-specific fungal marker 1,3-β-d-glucan was positive in the CSF in 70.4% (n = 27) of samples. Contrasted brain magnetic resonance imaging identified leptomeningeal enhancement in 53.1%. Optimal detection of coccidioidal meningitis requires a combination of diagnostic modalities.

ABSTRACT Background & Aims Steatotic liver disease (SLD) is characterised by liver fat accumulation exceeding 5%. Lean body weight (BMI ≤ 25 kg/m 2 ) with non‐alcoholic SLD is a rare phenotype, and the balance of risks related to metabolic conditions or genetic predisposition, or the risk of progressive liver disease, is not known. This study evaluates clinical and genetic predictors of non‐alcoholic SLD and advanced liver disease in lean individuals. Methods We used International Classification of Disease codes, radiology and pathology review to identify 177 lean non‐alcoholic SLD cases (47 cryptogenic, 130 MASLD), 677 matched lean controls, and 3090 overweight/obese SLD cases in the Mayo Clinic Biobank and Tapestry databases. We performed case–control and cross‐sectional comparisons of clinical and genetic factors between these groups, using univariable and multivariable analyses. Results Lean individuals with non‐alcoholic SLD exhibited metabolic and genetic profiles that were intermediate between those of lean controls without SLD and overweight/obese SLD individuals, including intermediate rates of diabetes, hypertension, hyperlipidaemia and of homozygosity for the risk allele of PNPLA3 . Multivariable analysis indicated that diabetes was an independent predictor of SLD among lean individuals. Among lean individuals with non‐alcoholic SLD, those with metabolic‐associated SLD (MASLD), as compared to those with cryptogenic SLD (without metabolic risk factor), were more likely to be homozygous for risk alleles in GCKR . The Fib‐4 score, a tool for screening advanced liver disease in SLD, accurately predicted advanced fibrosis among lean individuals with non‐alcoholic SLD. Conclusion Diabetes serves as a primary predictor of non‐alcoholic SLD in lean individuals. These results support the recommendation to screen for SLD in patients with diabetes, regardless of BMI. The GCKR risk allele is associated with MASLD in lean individuals, and the risk factors for cryptogenic SLD remain unclear.

Background: Ischemic stroke remains a leading cause of morbidity and mortality worldwide. Cardioembolic sources account for a significant proportion of these events, necessitating accurate identification of underlying structural heart abnormalities. Traditionally, transthoracic echocardiography (TTE) and transesophageal echocardiography (TEE) have been utilized for this purpose, although both have limitations. Cardiac computed tomography (CCTA) has emerged as a potential alternative, offering non-invasive imaging with high sensitivity and specificity for detecting intracardiac thrombi. Objective: This study aimed to evaluate the diagnostic yield of CCTA compared to TTE and TEE in detecting high-risk cardioaortic sources of embolism, including left ventricular (LV) thrombus, left atrial (LA) thrombus, valvular thrombus, and ulcerated aortic plaques, in patients with ischemic stroke. Methods: We conducted a retrospective, single-center cohort study at Mayo Clinic, Rochester, MN, including 426 ischemic stroke patients who underwent CCTA within 30 days of stroke onset between March 2019 and June 2023. We compared the diagnostic yields of CCTA, TTE, and TEE in detecting structural cardiac sources of embolism. Results: The overall diagnostic yield of CCTA for detecting high-risk cardioaortic sources was 15.5%. Specifically, CCTA identified LV thrombus in 2.3% of patients, LA thrombus in 4.0%, valvular thrombus in 2.8%, and ulcerated aortic plaques in 7.0%. In comparative analysis, CCTA demonstrated a higher yield than TTE (16.0% vs 3.5%, p < 0.001) and a comparable yield to TEE (25.3% vs 20.7%, p = 0.47) in detecting these sources. Conclusions: CCTA proved to be a viable option for identifying high-risk structural sources of cardioembolism in ischemic stroke patients. Its non-invasive nature, coupled with the ability to perform imaging soon after stroke onset, underscores its clinical utility. However, further prospective and blinded studies are required to validate these findings and fully establish the efficacy of CCTA in this context.

ABSTRACTThis report details a severe complication of phenol neurolysis in a diabetic patient, where phenol induced tissue necrosis can lead to severe cellulitis with ulceration that may prove difficult to treat. It emphasizes the importance of considering comorbidities when selecting phenol neurolysis for treatment of chronic pain.