Citations

Abstract Background Large language models (LLMs) offer promise for systematic review data extraction, but performance in complex multidisciplinary domains and utility for clinical statement generation remain insufficiently described. Objectives To evaluate Google NotebookLM for AI-assisted data extraction and RAND/UCLA consensus statement generation in a systematic review of IBD, obesity, and cardiometabolic comorbidities. Methods Studies were organized into domain-specific notebooks; structured prompts generated standardized evidence tables. Two independent reviewers validated outputs against full-text articles using a four-category error classification. Cell-level accuracy and critical accuracy (cells free of major factual errors) were the primary metrics; workflow time was compared against a published conventional extraction benchmark. Concordance between AI-generated and expert-finalized statements was assessed. Results Across 57 articles, 1,710 data cells were extracted; 151 (8.83%) were flagged, yielding 91.17% cell-level accuracy. Major factual errors occurred in only 4 cells (0.23%), for a critical accuracy of 99.77%. Most errors were minor omissions (59.6%) or incomplete extractions (30.5%); domain error rates ranged from 7.08% to 11.33%. The pipeline required 17.7 versus a projected 165.1 person-hours (89.3% reduction). PICO-structured prompting generated 70 candidate statements; 58 of 112 finalized panel statements (51.8%) were AI-derived, and 85.7% were retained in the finalized set. Conclusion Google NotebookLM demonstrates feasibility as a primary extraction and synthesis tool in a multidisciplinary systematic review, with extractive incompleteness as the principal limitation and substantial time savings over conventional approaches. Its novel application to RAND/UCLA consensus statement generation extends AI-assisted evidence synthesis to clinical consensus generation workflow.

Introduction Obesity and related cardiometabolic comorbidities, including hypertension, dyslipidaemia, diabetes, metabolic dysfunction-associated steatotic liver disease and atherosclerotic cardiovascular disease, are increasingly prevalent among individuals with inflammatory bowel disease (IBD). These conditions influence disease activity, therapeutic response, surgical outcomes and overall quality of life, yet evidence remains fragmented. The M odulate O besity and relate D metabolic compl I cations F or Y ielding improvements in IBD outcomes (MODIFY-IBD) initiative aims to synthesise evidence and generate consensus recommendations to guide practice and future research in this area. This study describes a protocol for a structured evidence synthesis and Research ANd Development/University of California, Los Angeles (RAND/UCLA) Appropriateness Method (RUAM) consensus process. Methods and analysis We will conduct three systematic reviews and a structured evidence synthesis organised into three domains: (1) the impact of obesity on IBD outcomes, (2) the burden of cardiometabolic complications in IBD and (3) the management of overweight, obesity and cardiometabolic comorbidities in IBD. A multidisciplinary international panel of gastroenterologists, surgeons, endocrinologists, hepatologists, cardiologists and dietitians will assess each statement using the RAND/UCLA appropriateness method. Panellists will rate the appropriateness of each statement (only those that fall within their area of expertise) on a 1–9 scale (1–3=inappropriate, 4–6=uncertain and 7–9=appropriate), with medians rounded up (eg, 6.5=appropriate). Agreement will be assessed using the RAND Disagreement Index (DI<1.0=agreement). Ethics and dissemination This study will not involve direct patient participation, as it is based on evidence synthesis and expert consensus; therefore, formal research ethics committee approval will not be required. Patient representatives will contribute to the consensus process to provide contextual perspectives but no identifiable data will be collected. Findings will be disseminated through publication in peer-reviewed journals, presentation at major gastroenterology and IBD conferences and communication with professional societies. A lay summary and patient-friendly infographic will also be developed to facilitate translation of recommendations into clinical practice. PROSPERO registration numbers CRD420251178843: a systematic review of the impact of obesity on inflammatory bowel disease outcomes. CRD420251178799: a systematic review of cardiometabolic complications in inflammatory bowel disease. CRD420251174653: management of overweight, obesity and cardiometabolic comorbidities in inflammatory bowel disease: a systematic review.

ABSTRACT Background Retrospective evaluations of multidisciplinary team (MDT) review of pancreatic cystic lesions (PCL) demonstrate improved PCL management. Endoscopic ultrasound (EUS) evaluates PCL for worrisome features; however, its influence on MDT is unclear and may be attenuated by anchoring/confirmation bias. We aimed to evaluate the impact of EUS and the overall performance of MDT evaluation of PCL. Methods Four mock MDT sessions at two institutions were presented imaging from 40 consecutive PCL cases (20 “high‐risk” with advanced neoplasia, 20 “low‐risk” with pathologic low‐grade dysplasia or ≥ 3 years of non‐progression on surveillance), with and without upfront EUS/FNA (cytology, CEA, Amylase), in a randomized, crossover design. Group 1 adjudicated without EUS, Group 2 adjudicated with EUS, and Group 3 adjudicated with EUS after initially adjudicating without it. Results The cohort included branch‐duct ( n = 20), mixed type ( n = 6), and main duct ( n = 4) IPMN, MCN ( n = 5), and non‐mucinous PCL ( n = 5). Group 2 (EUS) correctly identified mucinous PCL with 97% sensitivity and 70% specificity, significantly higher than Group 1 (No EUS) (Sensitivity 91%, Specificity 20%; p = 0.04). Group 3 (Unblinded to EUS) performed significantly worse in mucinous PCL identification compared with Group 2, indicating cognitive bias (Sensitivity 97%, Specificity 10% ( p = 0.0003)). Group 2 correctly identified high‐risk PCL that required surgical intervention with 85% sensitivity and 85% specificity, significantly higher than Group 1 (Sensitivity 60%, Specificity 75%; p = 0.0016) and Group 3 (Sensitivity 73%, Specificity 80%, p = 0.0485). Conclusion While susceptible to cognitive bias, EUS significantly improves MDT assessment of PCL. MDT establishes a high bar for iterative additions of biomarkers to PCL analysis.

Septal Myectomy (SM) and Alcohol Septal Ablation (ASA) improve symptoms in patients with Hypertrophic Cardiomyopathy with outflow tract obstruction (oHCM). However, outcomes data in this population is predominantly from specialized centers. The National Inpatient Database was queried from 2011 to 2019 for relevant international classification of diseases (ICD)-9 and -10 diagnostic and procedural codes. We compared baseline characteristics and in-hospital outcomes of patients with oHCM who underwent SM vs ASA. A p-value < 0.001 was considered statistically significant. We identified 15,119 patients with oHCM who underwent septal reduction therapies, of whom 57.4% underwent SM, and 42.6% underwent ASA. Patients who underwent SM had higher all-cause mortality (OR: 1.8 (1.3–2.5)), post-procedure ischemic stroke (OR: 2.3 (1.7–3.2)), acute kidney injury (OR: 1.4 (1.2–1.7)), vascular complications (OR: 3.6 (2.3–5.3)), ventricular septal defect (OR: 4.4 (3.2–6.1)), cardiogenic shock (OR: 1.7 (1.3–2.3)), sepsis (OR: 3.2 (1.9–5.4)), and left bundle branch block (OR: 3.5 (3–4)), compared to ASA. Patients who underwent ASA had higher post-procedure complete heart block (OR: 1.3 (1.1–1.4)), right bundle branch block (OR: 6.3 (5–7.7)), ventricular tachycardia (OR: 2.2 (1.9–2.6)), supraventricular tachycardia (OR: 1.6 (1.4–2)), and more commonly required pacemaker insertion (OR: 1.4 (1.3–1.7)) (p < 0.001 for all) compared to SM. This nationwide analysis evidenced that patients undergoing SM had higher in-hospital mortality and periprocedural complications than ASA; however, those undergoing ASA had more post-procedure conduction abnormalities and pacemaker implantation. The implications of these findings warrant further investigation regarding patient selection strategies for these therapies.

Getting around the blood–brain barrier The meninges comprise three membranes that surround and protect the central nervous system (CNS). Recent studies have noted the existence of myeloid cells resident there, but little is known about their ontogeny and function, and whether other meningeal immune cell populations have important roles remains unclear (see the Perspective by Nguyen and Kubes). Cugurra et al. found in mice that a large proportion of continuously replenished myeloid cells in the dura mater are not blood derived, but rather transit from cranial bone marrow through specialized channels. In models of CNS injury and neuroinflammation, the authors demonstrated that these myeloid cells have an immunoregulatory phenotype compared with their more inflammatory blood-derived counterparts. Similarly, Brioschi et al. show that the meninges host B cells that are also derived from skull bone marrow, mature locally, and likely acquire a tolerogenic phenotype. They further found that the brains of aging mice are infiltrated by a second population of age-associated B cells, which come from the periphery and may differentiate into autoantibody-secreting plasma cells after encountering CNS antigens. Together, these two studies may inform future treatment of neurological diseases. Science , abf7844, abf9277, this issue p. eabf7844 , p. eabf9277 ; see also abj8183, p. 396