Faculty Bibliography

PURPOSE OF THE REPORT/OBJECTIVE:To determine the prevalence, etiology, and clinical significance of incidental focal small bowel FDG uptake in patients undergoing PET/CT for staging of non-small bowel cancers. MATERIAL AND METHODS/METHODS:Retrospective review of consecutive FDG PET/CT examinations obtained for cancer staging with incidental focal small bowel radiotracer uptake was performed. Exclusion criteria included known small bowel pathology or insufficient reference standard. Imaging findings assessed included lesion location, number, CT correlate, SUVmax, and presence of metastases outside the bowel. Clinical data included age, sex, cancer clinical setting, origin, and stage. Focal small bowel FDG uptake etiology (benign vs. metastatic) was determined by composite reference standard (histopathology, clinical, and imaging follow-up). Statistical analyses included Wilcoxon rank-sum test, Pearson's χ2 test, Fisher exact test, and ROC curve analyses. RESULTS:In a review of 147,516 PET/CT examinations, incidental focal small bowel FDG uptake was rare, with a prevalence of 0.1% (88/147,516). Most cases were metastatic, 60.2% (53/88), most commonly spread from lymphoma [32.1% (17/53)] and melanoma [30.2% (16/53)]. Metastatic lesions were evenly distributed throughout the ileum [47.2% (25/53)] and jejunum [39.6% (21/53)]. Metastatic focal small bowel FDG uptake was associated with presence of other sites of distant metastases, higher SUVmax, and presence of a CT correlate (P <0.01). CONCLUSIONS:Incidental focal small bowel FDG uptake is rare. Most small bowel hypermetabolic foci are metastatic and are predominantly encountered with melanoma and lymphoma. Multiple imaging and clinical factors helped differentiate between benign and metastatic focal small bowel FDG uptake.

Radiography is an excellent first-line imaging technique for a wide range of conditions. However, there is insufficient evidence-based clinical guidance for ordering radiographic examinations in certain low-yield anatomic sites. To address this issue, a three-round modified Delphi consensus study was performed with a multidisciplinary, multi-institutional expert panel to determine the appropriate utilization of radiography at 12 often low-yield anatomic sites. The expert panel comprised 34 subspecialty faculty members (16 radiologists, nine emergency medicine physicians, four orthopedic surgeons, and five otolaryngology surgeons) from 21 academic centers. They evaluated 12 anatomic sites: rib, scapula, sacrum, coccyx, sternum, sternoclavicular joint, nasal bone, mandible, facial bones, sinuses, neck soft tissue, and skull. A structured literature review across three databases was conducted to assess the utilization and diagnostic test characteristics of radiography for each anatomic site. Teams drafted consensus statements and supporting text for their respective sites and submitted these statements to the panel for iterative rounds of anonymized voting. All 58 submitted statements achieved consensus by the end of round 3 and were endorsed by six major American medical societies. This study established expert consensus recommendations for the appropriate utilization of radiography at 12 anatomic sites through a multidisciplinary and multi-institutional deliberative process.

IMPORTANCE/UNASSIGNED:Screening by low-dose computed tomography can reduce lung cancer mortality among high-risk individuals, but many lung cancers occur among individuals with a smoking history who are not eligible for screening. OBJECTIVE/UNASSIGNED:To develop and validate the protein-based Integrative Analysis of Lung Cancer Risk and Etiology (INTEGRAL)-Risk model in individuals with a smoking history from the general population. DESIGN, SETTING, AND PARTICIPANTS/UNASSIGNED:Cohorts in the Lung Cancer Cohort Consortium recruited research participants in the US, Europe, Asia, and Australia between 1985 and 2009, who were followed up for lung cancer and other health outcomes until 2021. Fourteen case cohorts of 3695 participants with a smoking history within the Lung Cancer Cohort Consortium, including 2305 randomly sampled participants and 1390 patients diagnosed with lung cancer within 3 years after blood sample collection, were designed. Plasma or serum samples from each participant were assayed using the INTEGRAL protein panel in 2022. The INTEGRAL-Risk model was trained using 7 predefined case cohorts (training set; n = 1951) to estimate absolute risk of being diagnosed with lung cancer based on age, smoking history, and 13 proteins. The validity of the INTEGRAL-Risk model was assessed in 7 independent case cohorts (testing set; n = 1744) at 1, 2, and 3 years after blood collection. EXPOSURE/UNASSIGNED:Absolute risk estimates from the protein-based INTEGRAL-Risk model. MAIN OUTCOMES AND MEASURES/UNASSIGNED:The primary outcome was the validity of the INTEGRAL-Risk model in the testing set with respect to discrimination (area under the curve [AUC]) and calibration (ratio of expected-to-observed cases [E/O]). RESULTS/UNASSIGNED:A total of 3695 participants were included, with 1951 participants (including 807 with lung cancer) in the training set and 1744 participants (including 583 with lung cancer) in the testing set. In the combined 14 training and testing sets, after application of statistical weights, 323 570 participants were represented (185 016 [57%] female; median [IQR] age, 60 [51-67] years). In the independent testing set, discrimination of the INTEGRAL-Risk model was highest at 1 year of follow-up and exceeded that of the questionnaire-based PLCOm2012 (Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial) model (INTEGRAL-Risk AUC of 0.88 [95% CI, 0.85-0.91] vs PLCOm2012 AUC of 0.79 [95% CI, 0.75-0.83]; P value for difference <.001). Using a risk threshold to achieve the same specificity as US Preventive Services Task Force (USPSTF) 2021 criteria, the INTEGRAL-Risk model captured 85% of lung cancer cases compared with 63% by USPSTF 2021 and 70% by PLCOm2012. Discrimination of the INTEGRAL-Risk model decreased with longer prediction horizons, with a 2-year AUC of 0.84 (95% CI, 0.81-0.86) and 3-year AUC of 0.81 (95% CI, 0.79-0.83). The model was well calibrated (E/O over 3 years, 0.87 [95% CI, 0.69-1.14]). CONCLUSIONS AND RELEVANCE/UNASSIGNED:Compared with questionnaire-based approaches, the protein-based INTEGRAL-Risk model improved short-term prediction of lung cancer in people with a smoking history. This model has potential to improve selection of high-risk individuals who are most likely to benefit from lung cancer screening.

BACKGROUND AND OBJECTIVES/OBJECTIVE:General purpose vision-language models (VLMs) demonstrate impressive capabilities, but their opaque training on uncurated internet data poses critical limitations for high-stakes decision making, such as in neurosurgery. We present CNS-Obsidian, a neurosurgical VLM trained on peer-reviewed neurosurgical literature, and demonstrate its clinical utility compared with GPT-4o in a real-world setting. METHODS:We compiled 23 984 articles from Neurosurgery Publications journals, yielding 78 853 figures and captions. Using GPT-4o and Claude Sonnet-3.5, we converted these image-text pairs into 263 064 training samples across 3 formats: instruction fine-tuning, multiple-choice questions, and differential diagnosis. We trained CNS-Obsidian, a fine-tune of the 34-billion parameter Large Language and Visual Assistant-Next model. In a blinded, randomized deployment trial at NYU Langone Health (August 30-November 30, 2024), neurosurgeons were assigned to use either CNS-Obsidian or a Health Insurance Portability and Accountability Act-compliant GPT-4o end point as a diagnostic copilot after patient consultations. Primary outcomes were diagnostic helpfulness and accuracy, assessed through user ratings and presence of the correct diagnosis within the VLM-provided differential, respectively. RESULTS:CNS-Obsidian matched GPT-4o on synthetic questions (76.13% vs 77.54%, P = .235), but only achieved 46.81% accuracy on human-generated questions vs GPT-4o's 65.70% (P < 10-15). In the randomized trial, 70 consultations were evaluated (32 CNS-Obsidian, 38 GPT-4o) from 959 total consults (7.3% utilization). CNS-Obsidian received positive ratings in 40.62% of cases vs 57.89% for GPT-4o (P = .230). Both models included correct diagnosis in approximately 60% of cases (59.38% vs 65.79%, P = .626). CONCLUSION/CONCLUSIONS:Domain-specific VLMs trained on curated scientific literature can approach frontier model performance in specialized medical domains despite being orders of magnitude smaller and less expensive to train. This establishes a transparent framework for scientific communities to build specialized artificial intelligence models. However, low clinical utilization suggests chatbot interfaces may not align with specialist workflows, indicating need for alternative artificial intelligence integration strategies.

BACKGROUND:The frequency of residual angina and its impact on health status and death following anatomic complete revascularization in symptomatic patients with chronic coronary disease are unknown. METHODS:Data were analyzed from ISCHEMIA (International Study of Comparative Health Effectiveness With Medical and Invasive Approaches) trial participants randomized to invasive management with baseline angina (Seattle Angina Questionnaire Angina Frequency score <100), no prior coronary artery bypass graft surgery, and anatomic complete revascularization within 90 days of randomization. The primary outcome was frequency of residual angina after revascularization, defined as a Seattle Angina Questionnaire Angina Frequency score <100 within 6 months of randomization. Secondary outcomes included 6-month health status and medication use and 5-year all-cause and cardiovascular death. RESULTS:=0.006). Five-year all-cause and cardiovascular death did not differ significantly between groups. CONCLUSIONS:Residual angina is common (>40%) following anatomic complete revascularization for chronic coronary disease and is associated with reduced quality of life and greater antianginal medication use but no increase in death. REGISTRATION/BACKGROUND:Unique Identifier: NCT01471522.

The bone marrow niche (BMN) plays a central role in regulating hematopoietic stem-cell (HSC) maintenance, lineage commitment, and immune homeostasis, while also supporting osteogenesis and maintaining skeletal integrity. Once considered static, the BMN is now recognized as a dynamic and responsive microenvironment that integrates local signals and systemic cues to meet physiological demands and respond to stress. Aging causes profound and progressive changes to this niche, leading to functional decline across both hematopoietic and stromal compartments. Recent advances in high-resolution imaging, single-cell and spatial transcriptomics, and in vivo lineage tracing have revealed remarkable heterogeneity and plasticity within the vascular and mesenchymal elements of this niche. Yet, key questions remain unresolved, including the identity and hierarchy of mesenchymal and osteolineage cells, the specialization of subsets of endothelial cells, the integration of systemic regulation, and whether the aging bone marrow acts as a driver or a passenger in malignancy and chronic inflammation. This review revisits current models of the BMN, with a focus on the reciprocal interactions between osteogenic cells and specialized vasculature, and how their disruption during aging impairs hematopoietic output and skeletal remodeling. We also examine how systemic factors such as neural input, metabolic status, and inflammatory signaling influence the aging of the BMN. Finally, we highlight emerging translational platforms, including iPSC-derived bone marrow organoids, engineered niches/hydrogels, and vascularized organ-on-chip systems, that enable mechanistic testing of rejuvenation strategies. Together, these insights have the potential to pave the way toward targeted interventions that restore the function of the BMN and promote healthy aging of the bone and blood systems.

BACKGROUND:Persons with HIV have an increased risk of cardiovascular and metabolic diseases compared to HIV serostatus-negative individuals. OBJECTIVE:We conducted a pilot wait-list control randomized trial to assess the feasibility and acceptability of a virtual environment for CVD and metabolic disease prevention education in men living with HIV. METHODS:In phase one, we conducted interviews and virtual environment beta testing with (n=25) individuals. In phase two, reported here, we conducted a wait-list control trial with permuted block randomization. Participants were allocated to the virtual environment intervention (n = 40) or a wait-list control arm (n = 38). The primary outcomes were feasibility (defined by recruitment and retention metrics) and acceptability (defined by levels of engagement) with virtual environment. We also examined preliminary effect sizes across several cardiovascular health-related indicators. RESULTS:Participants had a mean age of 41.8 years (SD=10.4) and had been living with HIV for an average of 15.9 years (SD=9.7). Participants were 62.5% Black and 26.7% Hispanic. A total of 45.0% of the intervention arm engaged with the virtual environment, with an average session duration of 110 minutes. The most engaged content modules were nutrition (146 engagements), oral health (138 engagements), fitness tips and videos (82 engagements), and relaxation techniques (75 engagements). Retention was 61.5% at final assessment, which to our knowledge, is notably higher than the typical retention rates seen in remote digital health research and all behavioral intervention research. Preliminary efficacy analyses across 3- and 6-month changes revealed medium effect sizes favoring the intervention for key nutritional outcomes (vegetable consumption: d=0.59-0.66; whole grain intake: d=0.41-0.46); and vigorous physical activity (d=0.38-0.57). Small-to-medium effect sizes were observed for fast food reduction (d=0.37 at 6-months), total physical activity (d=0.37 at 6-months), and depressive symptoms (d=0.22 at 3-months), as well as HIV illness perceptions, including timeline acute/chronic (d=0.44 at 3-months), illness coherence (d=0.33 at 3-months), and emotional representations (d=0.23-0.33). CONCLUSIONS:The LEARN Study demonstrated the feasibility and acceptability of a virtual environment designed for cardiovascular and metabolic disease prevention education in persons with HIV. The high engagement rates with specific educational content and strong retention highlight the promise of this innovative approach. Preliminary effect sizes suggested positive trends in cardiovascular health indicators and mental well-being, indicating further potential benefits of the intervention. These findings support progression to a fully powered randomized controlled trial. Subsequent funding has been awarded, based on this work, to expand the research in LEARN2. We will integrate lessons learned and focus on HIV- related conditions with shared risk factors to enhance the study's impact. By addressing HIV conditions with interrelated risk factors, we aim to provide a more comprehensive intervention to improve the overall health and well-being of individuals living with HIV. CLINICALTRIAL/BACKGROUND:ClinicalTrials.gov NCT05242952. INTERNATIONAL REGISTERED REPORT/UNASSIGNED:RR2-10.2196/38348.

IMPORTANCE/UNASSIGNED:Bronchoscopic biopsy is conventionally performed with forceps, which can result in small specimen sizes and poor specimen quality due to crush artifact. Cryoprobe use localizes freezing at the probe tip, enabling retrieval of larger, more intact biopsy specimens. OBJECTIVE/UNASSIGNED:To evaluate the diagnostic yield of a 1.1-mm cryoprobe for transbronchial biopsy. DESIGN, SETTING, AND PARTICIPANTS/UNASSIGNED:This open-label, outcome assessor-masked, multicenter randomized clinical trial included 500 patients aged 18 years or older scheduled to undergo transbronchial biopsy for lung nodules or masses, lung transplant, or diffuse parenchymal lung disease. The trial was conducted in 9 US medical centers and enrolled patients between February 27, 2023, and September 11, 2024. The date of last follow-up was October 12, 2024. INTERVENTION/UNASSIGNED:Patients were randomized 1:1 to transbronchial biopsy using a 1.1-mm cryoprobe (n = 250) or 2.0-mm forceps (n = 250). MAIN OUTCOMES AND MEASURES/UNASSIGNED:The primary outcome was diagnostic yield, defined as the percentage of patients for whom the transbronchial biopsy sample led to a specific diagnosis based on histologic examination. Of the 8 prespecified secondary analyses, key secondary analyses were the diagnostic yield for each of the 3 conditions (lung nodules or masses, lung transplant, and diffuse parenchymal lung disease) and complication rates. RESULTS/UNASSIGNED:Of 774 patients assessed for eligibility, 609 provided consent, 500 were randomized, and 490 were included in the primary analysis; the mean age was 62.6 years (SD, 12.7 years) and 252 of 500 (50.4%) were male. The primary outcome of diagnostic yield was significantly higher in patients randomized to transbronchial biopsy with cryoprobes vs forceps (217 of 245 [88.6%] vs 193 of 245 [78.8%]; absolute difference, 9.8%; 95% CI, 3.3%-16.3%; P = .003). For the key secondary analyses, compared with that of forceps, the diagnostic yield of cryoprobes was significantly higher among patients with pulmonary nodules or masses (79 of 95 [83.2%] vs 68 of 97 [70.1%]; absolute difference, 13.1%; 95% CI, 1.0%-24.6%; P = .04) and lung transplant (120 of 125 [96.0%] vs 110 of 124 [88.7%]; absolute difference, 7.3%; 95% CI, 0.6%-14.4%; P = .03) but did not differ significantly in diffuse parenchymal lung disease (18 of 25 [72.0%] vs 15 of 24 [62.5%]; absolute difference, 9.5%; 95% CI, -16.0% to 33.6%; P = .55). For the secondary safety analysis, there were 4 pneumothoraces requiring chest tube placement in the forceps group (1.6%) vs none in the cryoprobe group; no patients experienced significant bleeding or respiratory failure events. CONCLUSIONS AND RELEVANCE/UNASSIGNED:Transbronchial lung biopsy performed with a 1.1-mm cryoprobe had a significantly higher diagnostic yield compared with 2.0-mm forceps in a group of patients with lung nodules or masses, lung transplant, and diffuse parenchymal lung disease. TRIAL REGISTRATION/UNASSIGNED:ClinicalTrials.gov Identifier: NCT05751278.