Faculty Bibliography

Generative AI, particularly large language models (LLMs), is reshaping clinical workflows in dermatology. However, cloud-based commercial models pose persistent challenges to Health Insurance Portability and Accountability Act (HIPAA) compliance, especially in dermatology, where protected health information (PHI) extends beyond text to clinical photographs, dermoscopic images, and total-body photography that may capture identifiable anatomical features and document conditions carrying social stigma. Locally hosted, open-weight LLMs that are run within the institution's own infrastructure offer dermatology practices a pathway to leverage AI capabilities while retaining full control of their data. This viewpoint synthesizes evidence on when locally hosted, open-weight LLMs should be preferred for dermatologic workflows, when cloud deployment may remain preferable, and how multimodal AI fits into a coherent local deployment strategy. We advance 4 arguments. First, model compression techniques (knowledge distillation, structured pruning, and low-bit quantization) together with mixture-of-experts architectures have lowered hardware thresholds enough that 7- to 33-billion-parameter models now run on consumer-grade workstations with modest neural processing units or graphics processing units. Second, dermatology is fundamentally a visual specialty, and a credible local deployment strategy must integrate LLMs with vision models, including convolutional neural networks, vision transformers, vision-language models, and dermatology-specific foundation models such as PanDerm and medical multimodal models such as MedGemma. Third, locally hosted, open-weight models confer specific advantages for dermatology, including complete institutional control of clinical images, freedom from vendor model deprecation that disrupts validated workflows, and the ability to audit and fine-tune models to address well-documented performance gaps in skin of color. Fourth, local deployment is not a panacea; cloud models remain preferable for some tasks, and local deployment introduces governance challenges (heterogeneity across practices, model drift, and quantization-induced accuracy loss) that require structured mitigation through validated reporting frameworks such as CONSORT-AI (Consolidated Standards of Reporting Trials), SPIRIT-AI (Standard Protocol Items: Recommendations for Interventional Trials), DECIDE-AI (Developmental and Exploratory Clinical Investigations of Decision support systems driven by AI), and TRIPOD+AI (Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis), as well as retrieval-augmented generation and federated learning approaches. We situate these arguments within the international regulatory landscape, including the European Union's General Data Protection Regulation, the European Union AI Act, and Germany's Digitale Gesundheitsanwendungen (DiGA) framework, in addition to HIPAA. We provide quantitative cost examples showing that current consumer hardware capable of running 14- to 33-billion-parameter models can be acquired for roughly the price of 1 to 2 years of enterprise cloud-AI subscriptions. We close by mapping a practical implementation pathway and identifying near-term research priorities. Locally hosted, open-weight LLMs that are deployed thoughtfully and within governance frameworks offer dermatology practices a credible route to harness generative AI while preserving regulatory compliance, equity across skin types, and the dermatologist-patient relationship.

BACKGROUND:Neural regulation contributes to pancreatic ductal adenocarcinoma (PDAC) development but effects of neural-targeting medications on presentation or outcomes remain unclear. METHODS:We conducted a retrospective study using the Veterans Affairs (VA) Corporate Data Warehouse (CDW) to identify patients with pancreatic cancer (2000-2020). Exposure to beta blocker, cholinergics or statins was defined by active prescriptions within 6 months before or 1 month after diagnosis. Outcomes included histologic subtype, stage, and overall survival (OS). Propensity score matching was performed for each medication class, with survival assessed using Kaplan-Meier and Cox regression models. RESULTS:Among 7,578 Veterans with pancreatic cancer, 76% had adenocarcinoma and 60% presented with stage IV disease. Beta blocker use was associated with lower odds of advanced stage (OR 0.55, 95% CI 0.5-0.63, p<0.0001) and improved OS (HR 0.89, 95%CI 0.84-0.95, p<0.0001). Cholinergic agonist use was associated with reduced likelihood of adenocarcinoma histology (OR 0.64, 95% CI 0.41-1.01, p=0.051) and advanced stage (OR 0.53, 95%CI 0.34-0.85, p=<0.0001), but not OS. Statin use was associated with adenocarcinoma histology (OR 1.24, 95%CI 1.09-1.43, p=0.002) and lower odds of advanced stage (OR 0.77, 95%CI 0.68-0.88, p=<0.0001). CONCLUSIONS:Beta blockers were associated with earlier stage and improved survival. Cholinergic agonists and statins were associated with earlier stage. Cholinergic agonists were linked to lower likelihood of adenocarcinoma histology, while statins to higher likelihood. These findings suggest neural and metabolic pathways may shape early PDAC biology. IMPACT/CONCLUSIONS:Autonomic and metabolic pathways may influence PDAC biology. Beta blockers merit mechanistic and clinical evaluation as adjunctive therapies.

OBJECTIVES/UNASSIGNED:To better understand the implications of tumor size on auditory brainstem implant (ABI) categories of auditory performance (CAP) score to facilitate more precise patient counselling. DESIGN/UNASSIGNED:Single-center retrospective chart review. SETTING/UNASSIGNED:Tertiary referral center. PARTICIPANTS/UNASSIGNED:Patients > 18 years old with neurofibromatosis type 2 who underwent ABI placement between the years 2009 and 2023 were included. Patients with prior surgical resection were excluded to reduce the potential of previous cochlear nucleus trauma confounding results. MAIN OUTCOME MEASURES/UNASSIGNED:Ipsilateral tumor volume was measured from the preoperative MRI, and the primary endpoint was the 1-year CAP score. RESULTS/UNASSIGNED: = 0.010). CONCLUSION/UNASSIGNED:Increased tumor size has a negative correlation with ultimate ABI performance. There were patients in this study with larger tumor sizes who still benefited from ABI placement. While larger tumor size should not be a contraindication for ABI placement, understanding the role tumor size can have on performance variation can serve to improve tumor management strategies and preoperative counselling.

PURPOSE OF REVIEW/OBJECTIVE:The coronavirus disease 2019 pandemic, shifting health policy and increasing vaccine hesitancy have all increased the likelihood healthcare practitioners will encounter vaccine-preventable diseases in children. Both front-line and consulting clinicians need to be able to identify these previously rare diseases and their neurologic complications, particularly in un- or under-vaccinated children. RECENT FINDINGS/RESULTS:As diseases previously considered rare or eliminated reemerge, more is known about direct and indirect consequences of peripheral and central nervous system infection in children-both in general and specific to individual vaccine-preventable diseases. Primary and secondary neurologic complications can be acute, subacute, chronic, or arise years later. Advances in imaging and molecular identification along with better understanding of underlying disease pathophysiology can all aid earlier identification, treatment, prognostication, and recovery. SUMMARY/CONCLUSIONS:While the benefits of childhood vaccination programs clearly outweigh the risks for the majority of children, in light of current realities the goal of this guide is to better prepare front-line and consulting clinicians to identify and manage vaccine-preventable diseases and their neurologic complications in un- or under-vaccinated children as they increasingly encounter them now and in the future.

The United States continues to grapple with the shifting substance use landscape among the emerging concurrent use of opioids and stimulants. Polysubstance use (PSU) of opioids and stimulants, whether intentionally or unintentionally, has increased substance use related morbidity and mortality. Changes at the policy level are needed to effectively curb overdose and other adverse outcomes. This article outlines 3 policy changes that are particularly relevant to opioid-stimulant PSU: (1) reforms in opioid agonist treatment policy, (2) reforms in payment for evidence-based treatment, and (3) reforms of drug paraphernalia laws. The current overdose crisis requires a multidimensional paradigm shift toward patient-centered and harm reduction focused policies that will expand access to comprehensive evidence-based treatments and services for individuals engaged in opioid and stimulant PSU.

Mesenchymal stromal cells (MSCs) support tissue homeostasis and regeneration, yet their molecular signals remain largely enigmatic. In skeletal muscle (SkM), MSCs, known as fibroadipogenic progenitors (FAPs), are essential for maintenance and repair, orchestrating these processes through intricate cellular communication networks. Given the critical role of SkM in lifelong health and longevity, FAP signaling has drawn significant interest as a potential therapeutic target and a model for MSC interactions. However, deciphering FAP-derived regulatory signals remains challenging due to their pleiotropic complexity. Here, we employ a systems-level approach to construct a comprehensive FAP interactome in both homeostatic and regenerating SkM. By integrating unique single-cell RNA sequencing atlases with advanced computational analyses, we identify putative FAP-mediated signaling pathways and validate their biological relevance through FAP depletion experiments, assessing disruptions in key pathways. This approach reveals novel signaling networks across diverse SkM cell populations, corroborates key FAP interactions from recent studies, and provides a valuable dataset for modeling MSC interactions and their roles in SkM homeostasis and regeneration.

BACKGROUND AND OBJECTIVES/OBJECTIVE:Alzheimer's disease affects mostly women over 65, who make up nearly two-thirds of cases. Subjective cognitive complaints (SCC) can precede Alzheimer's disease by years. While physical activity supports healthy aging, the long-term impact of midlife physical activity (PA) on late-life SCCs in women is limited. RESEARCH DESIGN AND METHODS/METHODS:Using data from 4,397 participants (mean age=46.2 years at baseline; 78.7 years at SCC assessment) in the New York University Women's Health Study (NYUWHS), a prospective cohort of 14,274 women enrolled in 1985-1991, we examined the association between midlife PA and SCC reported nearly 30 years later. PA was assessed at baseline using self-reported weekly hours of mild, moderate, and vigorous activity, converted to metabolic equivalent hours per week (MET-hours/week). Odds ratios and 95% CIs were estimated using unconditional logistic regression, adjusting for demographics and midlife and late-life health conditions. Multiple imputation and inverse probability weighting addressed missing data and potential selection bias. RESULTS:Women in the highest tertile of total midlife PA had 20% lower odds of reporting ≥2 SCCs compared with those in the lowest tertile (OR = 0.80; 95% CI = 0.68-0.95). The inverse association was stronger among never smokers (OR = 0.68; p for interaction=0.029). The inverse association was consistent for moderate and vigorous, but not mild PA. Findings were consistent across sensitivity analyses using alternative SCC definitions, exclusion criteria, and inverse probably weights. DISCUSSION AND IMPLICATIONS/CONCLUSIONS:Higher level of midlife PA was associated with fewer late-life SCCs, suggesting that midlife PA may be an important behavioral correlate of later-life cognitive health in women.