Faculty Bibliography

Acute myeloid leukemia (AML) is an aggressive myeloid malignancy with a poor prognosis. Venetoclax (Ven), a BCL2 inhibitor, has shown promising results but often leads to relapse due to mitochondrial dysregulation, particularly due to upregulation of the anti-apoptotic protein MCL1. Overexpression of the transcription factor STAT3 has been linked to poor survival in AML patients. Overexpression of STAT3 in a transgenic murine model induces a myeloid malignancy with a short latency period and inflammatory upregulation. The current study identifies STAT3 upregulation as a key mechanism of Ven resistance. A clinically relevant STAT3 degrader effectively reduces both total and phosphorylated STAT3, corrects mitochondrial structural and functional dysregulation, and induces apoptosis in Ven-resistant AML cell lines. KT-333 significantly decreases STAT3 and MCL1 protein levels and improves survival in Ven-resistant (Ven-Res) AML murine models. In summary, STAT3 hyperactivation is leukemogenic, is further potentiated in Ven-resistance and can be clinically targeted with a novel and specific STAT3 degrader. Pictorial representation depicting upregulation of STAT3 and MCL1 in venetoclax resistant myeloid malignancies such as MDS and AML causing mitochondrial structural abnormalities and dysfunction. By using specific STAT3 degrader, STAT3 inhibition, and thereby indirect downregulation of MCL1 can be a promising therapeutic intervention to target drug resistant clones in MDS and AML.

PROBLEM/OBJECTIVE:Undergraduate medical education (UME) often lacks detailed data on student learning in the clinical learning environment, instead relying on self-reported and observational assessments of student involvement in patient care. This reliance on subjective data can lead to inconsistencies and gaps in understanding student experiences during clinical encounters. The electronic health record (EHR) contains a wealth of data that could address these limitations but is underused in UME, limiting objective analysis of student encounters and hindering the ability to monitor and ensure consistent experiences across different clinical sites. APPROACH/METHODS:In 2020, a multidisciplinary team at the Kaiser Permanente Bernard J. Tyson School of Medicine used business intelligence software to develop dashboards that enhance analysis of student experiences in the clinical learning environment. Student encounters were identified using a unique EHR profile that enabled the capture of encounter-level data, which were then exported to a centralized dataset, facilitating creation of dashboards for comprehensive visualization and analysis of student experiences. OUTCOMES/RESULTS:By 2024, 17 dashboards were created that included visit- and patient-specific data. The EHR-linked dashboards featured encounter-specific details (specialty, preceptor, visit type and specialty, chief concern, diagnoses) and patient-specific details (age, race, sex, language, interpreter use). This allowed the capture of student experiences and facilitated analysis of student quality and patient-reported experience metrics. The dashboards also served as feedback tools to ensure comparability between students and cohorts across clinical sites. NEXT STEPS/CONCLUSIONS:The dissemination of individualized student dashboards enables insights into clinical experiences and identifies student contributions to patient care. By sharing rich data, students can pinpoint learning opportunities and faculty can better support curricular goals, advancing precision medical education strategies. This approach can serve as a model for empirical studies on how clinical learning environments shape student development and marks a necessary step toward personalized learning systems in UME.

Despite advances in endoscopic techniques, many colorectal surgeries in the United States are still performed for non-malignant colorectal polyps (NMCRPs). This study evaluated trends, demographic variations, and outcomes of surgeries for NMCRPs among all colorectal surgeries over the past decade. Using the TriNetX nationwide database, we identified adults (≥ 18 years of age) who underwent colectomy or proctectomy for NMCRPs or colorectal cancer between 2013 and 2023. We evaluated the proportion of surgeries performed for NMCRPs, stratified by demographic factors, and compared postoperative adverse events (AEs) between NMCRP and colorectal cancer surgeries. Among 136,721 surgeries, 52,480 (38.4%) were for NMCRPs. The proportion of NMCRP surgeries decreased from 59% in 2013 to 33% in 2023, with the most significant decline between 2013 and 2016. Black individuals showed the highest decrease. Compared with colorectal cancer surgeries, NMCRP surgeries were associated with significantly lower risks of wound, infectious, urinary, pulmonary, gastrointestinal, and cardiac AEs. Although the proportion of NMCRP surgeries has declined, ongoing efforts in education and training are needed to further reduce unnecessary surgeries and improve patient outcomes.

Lungs recovered from donation after circulatory death (DCD) are markedly underutilized for transplantation in the U.S. Evidence demonstrating excellent outcomes after DCD lung transplantation underpins the need for strategies to mitigate barriers to DCD lung utilization, including data driven revisions to allocation policy, revising program-specific quality metrics, removing financial barriers, and minimizing logistical disincentives. Expanded DCD transplantation is vital to reduce waitlist mortality and increase lung transplant rate.

BACKGROUND/UNASSIGNED:Academic dermatologists manage a broad array of responsibilities, often extending beyond clinical duties to include education, research, mentorship, and administrative work. OBJECTIVE/UNASSIGNED:This study aimed to quantify the scope of paid and unpaid responsibilities among academic dermatologists in the United States and identify disparities based on gender, career stage, and geography. METHODS/UNASSIGNED:A cross-sectional survey of 63 board-certified academic dermatologists from 12 institutions collected data on demographics, workweek allocation, after-hours tasks, compensation, burnout, and resilience from January to May 2025. RESULTS/UNASSIGNED:Respondents reported spending 58.3% of their time on clinical care, 15.9% on administrative duties, 13.8% on education, and 11.9% on research. Nearly half received 11 to 50 daily patient messages via the clinical inbox, and over 80% took hospital call duties, with most receiving no extra compensation. Burnout affected 63.5% of respondents and was significantly more prevalent among women, early-career physicians, and those in the Mid-Atlantic and Northeast regions. Resilience scores were lower among women. Consulting opportunities, often associated with supplemental income and career advancement, were more common among male and senior physicians. Parental leave policies were inconsistently applied and poorly understood; nearly 20% of respondents had 8 or fewer weeks of paid parental leave. LIMITATIONS/UNASSIGNED:Study limitations include the relatively small sample size and overrepresentation of the Northeast region, potentially hindering generalizability. Additionally, no significant race-based differences were observed, which may be due to insufficient sample sizes within comparator groups. CONCLUSION/UNASSIGNED:These findings highlight the cumulative burden of undercompensated labor, inadequate institutional support, and gender disparities in workload and opportunity. Structural changes in compensation transparency, workload distribution, recognition of academic labor, and support for early-career and female physicians are essential to fostering sustainability and equity within academic dermatology.

OBJECTIVE:To evaluate whether the risk of long COVID among individuals infected with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) during pregnancy differs from that of individuals who were not pregnant at time of virus acquisition. METHODS:We conducted a multicenter observational cohort study at 79 NIH RECOVER (Researching COVID to Enhance Recovery) sites. Individuals assigned female at birth aged 18-45 years with an index (first) SARS-CoV-2 infection on or after December 1, 2021, were included. The exposure was pregnancy (any gestational age) at the time of index SARS-CoV-2 infection. The primary outcome was long COVID 6 months after index infection , defined as RECOVER-Adult Long COVID Research Index score 11 or higher based on a detailed symptom survey. To account for confounding and differential selection between participants who were pregnant and not pregnant at infection, propensity score-matching methods were used to balance the groups on variables potentially associated with both pregnancy status and long COVID. RESULTS:Overall 2,423 participants were included; 580 (23.9%) were pregnant at index SARS-CoV-2 infection. The median age at infection was 33 years (interquartile range 28-38 years), and 2,131 of participants (90.0%) with known vaccination status were vaccinated. After propensity score matching, the adjusted long COVID prevalence estimates 6 months after index infection were 10.2% (95% CI, 6.2-14.3%) among those pregnant at infection and 10.6% (95% CI, 8.8-12.4%) among those not pregnant at infection. Pregnancy was not associated with a difference in adjusted risk of long COVID (adjusted risk ratio 0.96, 95% CI, 0.63-1.48). CONCLUSION/CONCLUSIONS:Acquisition of SARS-CoV-2 during pregnancy was not associated with a differential risk of long COVID at 6 months compared with similar-aged individuals who acquired SARS-CoV-2 outside of pregnancy.

BACKGROUND/UNASSIGNED:It remains unknown how often, and to which extent, patients with diabetes and non-ST-segment-elevation myocardial infarction are revascularized. This study aimed to identify practice patterns and clinical outcomes in patients with diabetes and non-ST-segment-elevation myocardial infarction according to completeness of revascularization. METHODS/UNASSIGNED:All patients with diabetes and multivessel disease hospitalized for non-ST-segment-elevation myocardial infarction in Ontario, Canada, between April 2009 and March 2020 and undergoing coronary angiography were included. Patients with previous coronary artery bypass grafting (CABG) at any time, percutaneous coronary intervention (PCI) in the previous 90 days, or an ST-segment-elevation myocardial infarction in the previous 90 days were excluded, as were patients with hemodynamic instability at hospital admission. Patients were classified into 4 groups, from the most to the least complete revascularization strategy: CABG, complete revascularization with PCI, incomplete PCI, or no revascularization. The primary outcome was all-cause death at 5 years. Cox regression adjusted for multiple baseline characteristics was used to compare outcomes between groups. RESULTS/UNASSIGNED:We included 14 511 patients (mean age: 68.7±11.5 years; 69.6% male patients); 4525 (31.2%) were treated with CABG, 3008 (20.7%) with complete PCI, 3624 (25.0%) with incomplete PCI, and 3354 (23.1%) did not receive any revascularization procedure. Adjusted 5-year risks of all-cause death after CABG, complete PCI, incomplete PCI, and no revascularization were progressively increased: 25.9%, 29.8%, 32.2%, and 39.4% respectively. Compared with no revascularization, a 46% reduction in all-cause death was observed after CABG (hazard ratio, 0.54 [95% CI, 0.50-0.58]), 35% after complete PCI (hazard ratio, 0.65 [95% CI, 0.60-0.71]), and 27% after incomplete PCI (hazard ratio, 0.73 [95% CI, 0.68-0.79]), over a median follow-up of 5.8 years (interquartile range, 3.3-8.6). CONCLUSIONS/UNASSIGNED:Almost a quarter of the patients with diabetes and non-ST-segment-elevation myocardial infarction did not receive any revascularization procedure. There was an incremental mortality reduction associated with CABG or complete revascularization with PCI, compared with incomplete revascularization with PCI or no revascularization.

BACKGROUND:Foundation models have shown remarkable potential in medical imaging by leveraging extensive pretraining on general datasets to enable fine-tuning for specific tasks. This is thought to be particularly beneficial for tasks where annotated data is scarce. A key underlying assumption, however, is that these models can learn from small amounts of training data more efficiently than existing state-of-the-art models. PURPOSE/OBJECTIVE:This study aims to characterize the performance of two major foundation segmentation models (SAM and MedSAM) when fine-tuned to segment neuroanatomic structures across a spectrum of dataset sizes, compared to a standard fully-supervised UNet model. METHODS:This study used 1,113 T1-weighted 3D MRIs from the Human Connectome Project's Young Adult cohort with corresponding Freesurfer-generated, manually-refined segmentations of 93 gray and white matter regions. The dataset was divided into 891 (80%) training MRIs, 111 (10%) validation MRIs, and 111 (10%) testing MRIs. SAM and MedSAM models were first fine-tuned and compared against a standard UNet model using Dice score to establish the baseline performance using all training 3D volumes. Subsequently, MedSAM and UNet models were fine-tuned across a varying number of training volumes to assess performance with diminishing dataset size, down to a single MRI, as well as no MRIs (zero-shot) for the MedSAM and SAM models. RESULTS:Using the entire training set, UNet outperformed MedSAM and SAM across most regions, with median Dice scores of 0.88 versus 0.82 and 0.84, respectively (p < 0.001). With diminishing dataset size, UNet continued to perform as well as or better than MedSAM in the three studied regions, down to even a single 3D volume. In the zero-shot setting, SAM and MedSAM showed some ability to segment with overall median Dice scores of 0.66 and 0.59, respectively. CONCLUSIONS:SAM and MedSAM did not outperform a standard UNet model in segmentation tasks, even in extremely limited training data settings, contrary to the foundation model hypothesis, suggesting that foundation models do not necessarily yield superior fine-tuned performance compared to standard segmentation models in the low data setting. Instead, the potential benefit of foundation models will depend on the characteristics of the task at hand and the behavior and capacity of the specific foundation model in question. Thus, it will be essential to benchmark against standard supervised deep learning methods for each distinct application to demonstrate the added value of using a foundation model.