Faculty Bibliography

IMPORTANCE/UNASSIGNED:Large language models (LLMs) are crucial for medical tasks. Ensuring their reliability is vital to avoid false results. Our study assesses two state-of-the-art LLMs (ChatGPT and LlaMA-2) for extracting clinical information, focusing on cognitive tests like MMSE and CDR. OBJECTIVE/UNASSIGNED:Evaluate ChatGPT and LlaMA-2 performance in extracting MMSE and CDR scores, including their associated dates. METHODS/UNASSIGNED:Our data consisted of 135,307 clinical notes (Jan 12th, 2010 to May 24th, 2023) mentioning MMSE, CDR, or MoCA. After applying inclusion criteria 34,465 notes remained, of which 765 underwent ChatGPT (GPT-4) and LlaMA-2, and 22 experts reviewed the responses. ChatGPT successfully extracted MMSE and CDR instances with dates from 742 notes. We used 20 notes for fine-tuning and training the reviewers. The remaining 722 were assigned to reviewers, with 309 each assigned to two reviewers simultaneously. Inter-rater-agreement (Fleiss' Kappa), precision, recall, true/false negative rates, and accuracy were calculated. Our study follows TRIPOD reporting guidelines for model validation. RESULTS/UNASSIGNED:For MMSE information extraction, ChatGPT (vs. LlaMA-2) achieved accuracy of 83% (vs. 66.4%), sensitivity of 89.7% (vs. 69.9%), true-negative rates of 96% (vs 60.0%), and precision of 82.7% (vs 62.2%). For CDR the results were lower overall, with accuracy of 87.1% (vs. 74.5%), sensitivity of 84.3% (vs. 39.7%), true-negative rates of 99.8% (98.4%), and precision of 48.3% (vs. 16.1%). We qualitatively evaluated the MMSE errors of ChatGPT and LlaMA-2 on double-reviewed notes. LlaMA-2 errors included 27 cases of total hallucination, 19 cases of reporting other scores instead of MMSE, 25 missed scores, and 23 cases of reporting only the wrong date. In comparison, ChatGPT's errors included only 3 cases of total hallucination, 17 cases of wrong test reported instead of MMSE, and 19 cases of reporting a wrong date. CONCLUSIONS/UNASSIGNED:In this diagnostic/prognostic study of ChatGPT and LlaMA-2 for extracting cognitive exam dates and scores from clinical notes, ChatGPT exhibited high accuracy, with better performance compared to LlaMA-2. The use of LLMs could benefit dementia research and clinical care, by identifying eligible patients for treatments initialization or clinical trial enrollments. Rigorous evaluation of LLMs is crucial to understanding their capabilities and limitations.

Recent studies have shown that a pro-inflammatory diet and dysbiosis, especially a high level of trimethylamine-N-oxide (TMAO), are associated with various adverse health conditions. Cardiovascular diseases and pancreatic diseases are two major morbidities in the modern world. Through this narrative review, we aimed to summarize the association between a pro-inflammatory diet, gut microbiota, and cardiovascular and pancreatic diseases, along with their underlying mechanisms. Our review revealed that TMAO is associated with the development of cardiovascular diseases by promoting platelet aggregation, atherosclerotic plaque formation, and vascular inflammation. TMAO is also associated with the development of acute pancreatitis. The pro-inflammatory diet is associated with an increased risk of pancreatic cancer and cardiovascular diseases through mechanisms that include increasing TMAO levels, activating the lipopolysaccharides cascade, and the direct pro-inflammatory effect of certain nutrients. Meanwhile, an anti-inflammatory diet decreases the risk of cardiovascular diseases and pancreatic cancer.

IMPORTANCE/UNASSIGNED:Patients with large ischemic core stroke have poor clinical outcomes and are frequently not considered for interfacility transfer for endovascular thrombectomy (EVT). OBJECTIVE/UNASSIGNED:To assess EVT treatment effects in transferred vs directly presenting patients and to evaluate the association between transfer times and neuroimaging changes with EVT clinical outcomes. DESIGN, SETTING, AND PARTICIPANTS/UNASSIGNED:This prespecified secondary analysis of the SELECT2 trial, which evaluated EVT vs medical management (MM) in patients with large ischemic stroke, evaluated adults aged 18 to 85 years with acute ischemic stroke due to occlusion of the internal carotid or middle cerebral artery (M1 segment) as well as an Alberta Stroke Program Early CT Score (ASPECTS) of 3 to 5, core of 50 mL or greater on imaging, or both. Patients were enrolled between October 2019 and September 2022 from 31 EVT-capable centers in the US, Canada, Europe, Australia, and New Zealand. Data were analyzed from August 2023 to January 2024. INTERVENTIONS/UNASSIGNED:EVT vs MM. MAIN OUTCOMES AND MEASURES/UNASSIGNED:Functional outcome, defined as modified Rankin Scale (mRS) score at 90 days with blinded adjudication. RESULTS/UNASSIGNED:A total of 958 patients were screened and 606 patients were excluded. Of 352 enrolled patients, 145 (41.2%) were female, and the median (IQR) age was 66.5 (58-75) years. A total of 211 patients (59.9%) were transfers, while 141 (40.1%) presented directly. The median (IQR) transfer time was 178 (136-230) minutes. The median (IQR) ASPECTS decreased from the referring hospital (5 [4-7]) to an EVT-capable center (4 [3-5]). Thrombectomy treatment effect was observed in both directly presenting patients (adjusted generalized odds ratio [OR], 2.01; 95% CI, 1.42-2.86) and transferred patients (adjusted generalized OR, 1.50; 95% CI, 1.11-2.03) without heterogeneity (P for interaction = .14). Treatment effect point estimates favored EVT among 82 transferred patients with a referral hospital ASPECTS of 5 or less (44 received EVT; adjusted generalized OR, 1.52; 95% CI, 0.89-2.58). ASPECTS loss was associated with numerically worse EVT outcomes (adjusted generalized OR per 1-ASPECTS point loss, 0.89; 95% CI, 0.77-1.02). EVT treatment effect estimates were lower in patients with transfer times of 3 hours or more (adjusted generalized OR, 1.15; 95% CI, 0.73-1.80). CONCLUSIONS AND RELEVANCE/UNASSIGNED:Both directly presenting and transferred patients with large ischemic stroke in the SELECT2 trial benefited from EVT, including those with low ASPECTS at referring hospitals. However, the association of EVT with better functional outcomes was numerically better in patients presenting directly to EVT-capable centers. Prolonged transfer times and evolution of ischemic change were associated with worse EVT outcomes. These findings emphasize the need for rapid identification of patients suitable for transfer and expedited transport. TRIAL REGISTRATION/UNASSIGNED:ClinicalTrials.gov Identifier: NCT03876457.

B-cell-maturation-antigen (BCMA)-directed therapies are highly active for multiple myeloma, but infections are emerging as a major challenge. In this retrospective, single-center analysis we evaluated infectious complications after BCMA-targeted chimeric-antigen-receptor T-cell therapy (CAR-T), bispecific-antibodies (BsAb) and antibody-drug-conjugates (ADC). The primary endpoint was severe (grade ≥ 3) infection incidence. Amongst 256 patients, 92 received CAR-T, 55 BsAb and 109 ADC. The incidence of severe infections was higher with BsAb (40%) than CAR-T (26%) or ADC (8%), including grade 5 infections (7% vs 0% vs 0%, respectively). Comparing T-cell redirecting therapies, the incidence rate of severe infections was significantly lower with CAR-T compared to BsAb at 1-year (incidence-rate-ratio [IRR] = 0.43, 95%CI 0.25-0.76, P = 0.004). During periods of treatment-emergent hypogammaglobulinemia, BsAb recipients had higher infection rates (IRR:2.27, 1.31-3.98, P = 0.004) and time to severe infection (HR 2.04, 1.05-3.96, P = 0.036) than their CAR-T counterparts. During periods of non-neutropenia, CAR-T recipients had a lower risk (HR 0.44, 95%CI 0.21-0.93, P = 0.032) and incidence rate (IRR:0.32, 95% 0.17-0.59, P < 0.001) of severe infections than BsAb. In conclusion, we observed an overall higher and more persistent risk of severe infections with BsAb. Our results also suggest a higher infection risk during periods of hypogammaglobulinemia with BsAb, and with neutropenia in CAR-T recipients.

BACKGROUND:Xenotransplantation clinical trials may begin soon. A persistent risk of xenotransplantation, known for decades, is the possibility that a xenozoonotic infection could be transferred from a xenograft to its recipient and then to other human contacts. Because of this risk, guidelines and commentators have advocated for xenograft recipients to agree to either long-term or lifelong surveillance mechanisms. METHODS:For the past few decades, one solution that has been proposed to ensure that xenograft recipients will comply with surveillance protocols is the use of a heavily modified Ulysses contract, which we review. RESULTS:These contracts are most often used in psychiatry, and their application to xenotransplantation has been espoused several times with minimal criticism. CONCLUSIONS:In this article, we argue against the applicability of Ulysses contracts in xenotransplantation based upon (1) the telos of the advance directive that may not be applicable to this clinical context, (2) the suspect nature of enforcing Ulysses contracts in xenotransplantation, and (3) the ethical and regulatory hurdles that such enforcement would require. Although our focus is on the US regulatory landscape in preparation for clinical trials, there are applications globally.

INTRODUCTION/BACKGROUND:Data suggest atherosclerotic-related inflammation may play a role in the pathogenesis of inflammatory bowel disease (IBD), but large-scale studies are missing. METHODS:In this nationwide case-control study, we used the Swedish Patient Register and the Epidemiology Strengthened by histoPathology Reports in Sweden cohort to identify adult cases of incident IBD between 2002 and 2021, with each case matched to up to 10 general population controls. We used conditional logistic regression to calculate odds ratios (OR) for exposure to an atherosclerotic-related condition (myocardial infarction, thromboembolic stroke, or atherosclerosis itself) before being diagnosed with IBD. RESULTS:There were a total of 56,212 individuals with IBD and 531,014 controls. Of them, 2,334 (4.2%) cases and 18,222 (3.4%) controls had a prior diagnosis of an atherosclerotic-related condition, corresponding to an OR of 1.30 (95% confidence interval [CI] 1.24-1.37). Results were statistically significant for both Crohn's disease (OR 1.37, 95% CI 1.26-1.48) and ulcerative colitis (OR 1.27, 95% CI 1.20-1.35) and for individuals who developed IBD at 40-59 years of age and 60 years or older. In addition, associations persisted when adjusting for underlying comorbidities, including the presence of immune-mediated diseases and prior aspirin and/or statin use. The highest odds of an atherosclerotic-related condition were seen in the 6-12 months before IBD diagnosis, though odds were increased even ≥5 years before. A higher magnitude of odds was also observed when having 2 or more atherosclerotic-related conditions when compared with having only 1 condition. DISCUSSION/CONCLUSIONS:A history of an atherosclerotic-related condition is associated with increased odds of developing IBD, particularly among older adults. Future studies should investigate whether drugs targeting atherosclerotic-related inflammation may prevent IBD in higher-risk individuals.

IMPORTANCE/UNASSIGNED:Stent retriever-based thrombectomy is highly beneficial in large vessel occlusion (LVO) strokes. Many stent retriever designs are currently available, but comparison of these technologies in well-conducted studies is lacking. OBJECTIVE/UNASSIGNED:To determine whether thrombectomy for LVO stroke with the pRESET stent retriever is noninferior to treatment with the Solitaire stent retriever. DESIGN, SETTING, AND PARTICIPANTS/UNASSIGNED:This study was a multicenter, prospective, randomized, controlled, open-label, adaptive, noninferiority trial with blinded primary end point evaluation. Between October 2019 and February 2022, multicenter participation occurred across 19 research hospitals and/or universities in the US and 5 in Germany. Patients with LVO stroke were enrolled and included up to 8 hours after symptom onset. INTERVENTIONS/UNASSIGNED:Patients underwent 1:1 randomization to thrombectomy with the pRESET or Solitaire stent retriever. MAIN OUTCOMES AND MEASURES/UNASSIGNED:The primary outcome was the difference in the rate of 90-day functional independence across the 2 devices, using a -12.5% noninferiority margin for the lower bound of the 1-sided 95% CI of the difference between pRESET and Solitaire retrievers. RESULTS/UNASSIGNED:Of 340 randomized patients, 170 (50.0%) were female, and the median (IQR) age was 73.0 (64.0-82.0) years. The study procedure was completed in 322 of the 340 randomized patients. The primary end point of 90-day functional independence was achieved by 95 patients (54.9%; 95% CI, 48.7-61.1) in the pRESET group and in 96 (57.5%; 95% CI, 51.2-63.8) in the Solitaire group (absolute difference, -2.57%; 95% CI, -11.42 to 6.28). As the lower bound of the 95% CI was greater than -12.5%, the pRESET retriever was deemed noninferior to the Solitaire retriever. The noninferiority of pRESET over Solitaire was also observed in the secondary clinical end point (90-day shift in modified Rankin Scale score) and in both angiographic end points (Expanded Treatment in Cerebral Infarction [eTICI] score of 2b50 or greater within 3 passes: 146 of 173 [84.4%] vs 149 of 167 [89.2%]; absolute difference, -4.83%; 95% CI, -10.84 to 1.19; eTICI of 2c or greater following the first pass: 76 of 173 [43.7%] vs 74 of 167 [44.3%]; absolute difference, -0.63%; 95% CI, -9.48 to 8.21). Symptomatic intracranial hemorrhage occurred in 0 patients in the pRESET group and 2 (1.2%) in the Solitaire group. Mortality occurred in 25 (14.5%) in the pRESET group and in 24 (14.4%) in the Solitaire group at 90 days. Findings of the per-protocol and as-treated analyses were in concordance with findings of the intention-to-treat analysis. CONCLUSIONS AND RELEVANCE/UNASSIGNED:In this study, among patients with LVO stroke, thrombectomy with the pRESET stent retriever was noninferior to thrombectomy with the Solitaire stent retriever. Findings suggest that pRESET offers a safe and effective option for flow restoration and disability reduction in patients with LVO stroke.

BACKGROUND:The Financial Incentives for Weight Reduction (FIReWoRk) clinical trial showed that financial incentive weight-loss strategies designed using behavioral economics were more effective than provision of weight-management resources only. We now evaluate cost-effectiveness. METHODS:Cost-effectiveness analysis of a multisite randomized trial enrolling 668 participants with obesity living in low-income neighborhoods. Participants were randomized to (1) goal-directed incentives (targeting behavioral goals), (2) outcome-based incentives (targeting weight-loss), and (3) resources only, which were provided to all participants and included a 1-year commercial weight-loss program membership, wearable activity monitor, food journal, and digital scale. We assessed program costs, time costs, quality of life, weight, and incremental cost-effectiveness in dollars-per-kilogram lost. RESULTS:Mean program costs at 12 months, based on weight loss program attendance, physical activity participation, food diary use, self-monitoring of weight, and incentive payments was $1271 in the goal-directed group, $1194 in the outcome-based group, and $834 in the resources-only group (difference, $437 [95% CI, 398 to 462] and $360 [95% CI, 341-363] for goal-directed or outcome-based vs resources-only, respectively; difference, $77 [95% CI, 58-130] for goal-directed vs outcome-based group). Quality of life did not differ significantly between the groups, but weight loss was substantially greater in the incentive groups (difference, 2.34 kg [95% CI, 0.53-4.14] and 1.79 kg [95% CI, -0.14 to 3.72] for goal-directed or outcome-based vs resources only, respectively; difference, 0.54 kg [95% CI, -1.29 to 2.38] for goal-directed vs outcome-based). Cost-effectiveness of incentive strategies based on program costs was $189/kg lost in the goal-directed group (95% CI, $124/kg to $383/kg) and $186/kg lost in the outcome-based group (95% CI, $113/kg to $530/kg). CONCLUSIONS:Goal-directed and outcome-based financial incentives were cost-effective strategies for helping low-income individuals with obesity lose weight. Their incremental cost per kilogram lost were comparable to other weight loss interventions.