Empowering US healthcare delivery organizations: Cultivating a community of practice to harness AI and advance health equity.

Healthcare delivery organizations (HDOs) in the US must contend with the potential for AI to worsen health inequities. But there is no standard set of procedures for HDOs to adopt to navigate these challenges. There is an urgent need for HDOs to present a unified approach to proactively address the potential for AI to worsen health inequities. Amidst this background, Health AI Partnership (HAIP) launched a community of practice to convene stakeholders from across HDOs to tackle challenges related to the use of AI. On February 15, 2023, HAIP hosted an inaugural workshop focused on the question, "Our health care delivery setting is considering adopting a new solution that uses AI. How do we assess the potential future impact on health inequities?" This topic emerged as a common challenge faced by all HDOs participating in HAIP. The workshop had 2 main goals. First, we wanted to ensure participants could talk openly without reservations about challenging topics such as health equity. The second goal was to develop an actionable, generalizable framework that could be immediately put into practice. The workshop engaged 77 participants with 100% representation from all 10 HDOs and invited ecosystem partners. In an accompanying Research Article, we share the Health Equity Across the AI Lifecycle (HEAAL) framework. We invite and encourage HDOs to test the HEAAL framework internally and share feedback so that we can continue to refine and maintain the set of procedures. The HEAAL framework reveals the challenges associated with rigorously assessing the potential for AI to worsen health inequities. Significant investment in personnel, capabilities, and data infrastructure is required, and the level of investment needed could be beyond reach for most HDOs. We look forward to expanding our community of practice to assist HDOs around the world.

Development and preliminary testing of Health Equity Across the AI Lifecycle (HEAAL): A framework for healthcare delivery organizations to mitigate the risk of AI solutions worsening health inequities.

The use of data-driven technologies such as Artificial Intelligence (AI) and Machine Learning (ML) is growing in healthcare. However, the proliferation of healthcare AI tools has outpaced regulatory frameworks, accountability measures, and governance standards to ensure safe, effective, and equitable use. To address these gaps and tackle a common challenge faced by healthcare delivery organizations, a case-based workshop was organized, and a framework was developed to evaluate the potential impact of implementing an AI solution on health equity. The Health Equity Across the AI Lifecycle (HEAAL) is co-designed with extensive engagement of clinical, operational, technical, and regulatory leaders across healthcare delivery organizations and ecosystem partners in the US. It assesses 5 equity assessment domains-accountability, fairness, fitness for purpose, reliability and validity, and transparency-across the span of eight key decision points in the AI adoption lifecycle. It is a process-oriented framework containing 37 step-by-step procedures for evaluating an existing AI solution and 34 procedures for evaluating a new AI solution in total. Within each procedure, it identifies relevant key stakeholders and data sources used to conduct the procedure. HEAAL guides how healthcare delivery organizations may mitigate the potential risk of AI solutions worsening health inequities. It also informs how much resources and support are required to assess the potential impact of AI solutions on health inequities.

LXR signaling pathways link cholesterol metabolism with risk for prediabetes and diabetes.

BACKGROUNDPreclinical studies suggest that cholesterol accumulation leads to insulin resistance. We previously reported that alterations in a monocyte cholesterol metabolism transcriptional network (CMTN) - suggestive of cellular cholesterol accumulation - were cross-sectionally associated with obesity and type 2 diabetes (T2D). Here, we sought to determine whether the CMTN alterations independently predict incident prediabetes/T2D risk, and correlate with cellular cholesterol accumulation.METHODSMonocyte mRNA expression of 11 CMTN genes was quantified among 934 Multi-Ethnic Study of Atherosclerosis (MESA) participants free of prediabetes/T2D; cellular cholesterol was measured in a subset of 24 monocyte samples.RESULTSDuring a median 6-year follow-up, lower expression of 3 highly correlated LXR target genes - ABCG1 and ABCA1 (cholesterol efflux) and MYLIP (cholesterol uptake suppression) - and not other CMTN genes, was significantly associated with higher risk of incident prediabetes/T2D. Lower expression of the LXR target genes correlated with higher cellular cholesterol levels (e.g., 47% of variance in cellular total cholesterol explained by ABCG1 expression). Further, adding the LXR target genes to overweight/obesity and other known predictors significantly improved prediction of incident prediabetes/T2D.CONCLUSIONThese data suggest that the aberrant LXR/ABCG1-ABCA1-MYLIP pathway (LAAMP) is a major T2D risk factor and support a potential role for aberrant LAAMP and cellular cholesterol accumulation in diabetogenesis.FUNDINGThe MESA Epigenomics and Transcriptomics Studies were funded by NIH grants 1R01HL101250, 1RF1AG054474, R01HL126477, R01DK101921, and R01HL135009. This work was supported by funding from NIDDK R01DK103531 and NHLBI R01HL119962.

Frameworks for procurement, integration, monitoring, and evaluation of artificial intelligence tools in clinical settings: A systematic review.

Research on the applications of artificial intelligence (AI) tools in medicine has increased exponentially over the last few years but its implementation in clinical practice has not seen a commensurate increase with a lack of consensus on implementing and maintaining such tools. This systematic review aims to summarize frameworks focusing on procuring, implementing, monitoring, and evaluating AI tools in clinical practice. A comprehensive literature search, following PRSIMA guidelines was performed on MEDLINE, Wiley Cochrane, Scopus, and EBSCO databases, to identify and include articles recommending practices, frameworks or guidelines for AI procurement, integration, monitoring, and evaluation. From the included articles, data regarding study aim, use of a framework, rationale of the framework, details regarding AI implementation involving procurement, integration, monitoring, and evaluation were extracted. The extracted details were then mapped on to the Donabedian Plan, Do, Study, Act cycle domains. The search yielded 17,537 unique articles, out of which 47 were evaluated for inclusion based on their full texts and 25 articles were included in the review. Common themes extracted included transparency, feasibility of operation within existing workflows, integrating into existing workflows, validation of the tool using predefined performance indicators and improving the algorithm and/or adjusting the tool to improve performance. Among the four domains (Plan, Do, Study, Act) the most common domain was Plan (84%, n = 21), followed by Study (60%, n = 15), Do (52%, n = 13), & Act (24%, n = 6). Among 172 authors, only 1 (0.6%) was from a low-income country (LIC) and 2 (1.2%) were from lower-middle-income countries (LMICs). Healthcare professionals cite the implementation of AI tools within clinical settings as challenging owing to low levels of evidence focusing on integration in the Do and Act domains. The current healthcare AI landscape calls for increased data sharing and knowledge translation to facilitate common goals and reap maximum clinical benefit.

Assessment of Days Alive Out of Hospital as a Possible End Point in Trials of Stroke Prevention for Atrial Fibrillation: A ROCKET AF Analysis.

BACKGROUND: Days alive out of hospital (DAOH) is an objective and patient-centered net benefit end point. There are no assessments of DAOH in clinical trials of interventions for atrial fibrillation (AF), and it is not known whether this end point is of clinical utility in these populations. METHODS AND RESULTS: ROCKET AF (Rivaroxaban Once Daily Oral Direct Factor Xa Inhibition Compared With Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation) was an international double-blind, double-dummy randomized clinical trial that compared rivaroxaban with warfarin in patients with atrial fibrillation at increased risk for stroke. We assessed DAOH using investigator-reported event data for up to 12 months after randomization in ROCKET AF. We assessed DAOH overall, by treatment group, and by subgroup, including age, sex, and comorbidities, using Poisson regression. The mean±SD number of days dead was 7.3±41.2, days hospitalized was 1.2±7.2, and mean DAOH was 350.7±56.2, with notable left skew. Patients with comorbidities had fewer DAOH overall. There were no differences in DAOH by treatment arm, with mean DAOH of 350.6±56.5 for those randomized to rivaroxaban and 350.7±55.8 for those randomized to warfarin (P=0.86). A sensitivity analysis found no difference in DAOH not disabled with rivaroxaban versus warfarin (DAOH not disabled, 349.2±59.5 days and 349.1 days±59.3 days, respectively, P=0.88). CONCLUSIONS: DAOH did not identify a treatment difference between patients randomized to rivaroxaban versus warfarin. This may be driven in part by the low overall event rates in atrial fibrillation anticoagulation trials, which leads to substantial left skew in measures of DAOH.

Framework of the Strengths and Challenges of Clinically Integrated Trials: An Expert Panel Report.

The limitations of the explanatory clinical trial framework include the high expense of implementing explanatory trials, restrictive entry criteria for participants, and redundant logistical processes. These limitations can result in slow evidence generation that is not responsive to population health needs, yielding evidence that is not generalizable. Clinically integrated trials, which integrate clinical research into routine care, represent a potential solution to this challenge and an opportunity to support learning health systems. The operational and design features of clinically integrated trials include a focused scope, simplicity in design and requirements, the leveraging of existing data structures, and patient participation in the entire trial process. These features are designed to minimize barriers to participation and trial execution and reduce additional research burdens for participants and clinicians alike. Broad adoption and scalability of clinically integrated trials are dependent, in part, on continuing regulatory, healthcare system, and payer support. This analysis presents a framework of the strengths and challenges of clinically integrated trials and is based on a multidisciplinary expert "Think Tank" panel discussion that included representatives from patient populations, academia, non-profit funding agencies, the U.S. Food and Drug Administration, and industry.

The algorithm journey map: a tangible approach to implementing AI solutions in healthcare.

When integrating AI tools in healthcare settings, complex interactions between technologies and primary users are not always fully understood or visible. This deficient and ambiguous understanding hampers attempts by healthcare organizations to adopt AI/ML, and it also creates new challenges for researchers to identify opportunities for simplifying adoption and developing best practices for the use of AI-based solutions. Our study fills this gap by documenting the process of designing, building, and maintaining an AI solution called SepsisWatch at Duke University Health System. We conducted 20 interviews with the team of engineers and scientists that led the multi-year effort to build the tool, integrate it into practice, and maintain the solution. This "Algorithm Journey Map" enumerates all social and technical activities throughout the AI solution's procurement, development, integration, and full lifecycle management. In addition to mapping the "who?" and "what?" of the adoption of the AI tool, we also show several 'lessons learned' throughout the algorithm journey maps including modeling assumptions, stakeholder inclusion, and organizational structure. In doing so, we identify generalizable insights about how to recognize and navigate barriers to AI/ML adoption in healthcare settings. We expect that this effort will further the development of best practices for operationalizing and sustaining ethical principles-in algorithmic systems.

Interaction of AI-Enabled Quantitative Coronary Plaque Volumes on Coronary CT Angiography, FFRCT, and Clinical Outcomes: A Retrospective Analysis of the ADVANCE Registry.

BACKGROUND: Luminal stenosis, computed tomography-derived fractional-flow reserve (FFRCT), and high-risk plaque features on coronary computed tomography angiography are all known to be associated with adverse clinical outcomes. The interactions between these variables, patient outcomes, and quantitative plaque volumes have not been previously described. METHODS: Patients with coronary computed tomography angiography (n=4430) and one-year outcome data from the international ADVANCE (Assessing Diagnostic Value of Noninvasive FFRCT in Coronary Care) registry underwent artificial intelligence-enabled quantitative coronary plaque analysis. Optimal cutoffs for coronary total plaque volume and each plaque subtype were derived using receiver-operator characteristic curve analysis. The resulting plaque volumes were adjusted for age, sex, hypertension, smoking status, type 2 diabetes, hyperlipidemia, luminal stenosis, distal FFRCT, and translesional delta-FFRCT. Median plaque volumes and optimal cutoffs for these adjusted variables were compared with major adverse cardiac events, late revascularization, a composite of the two, and cardiovascular death and myocardial infarction. RESULTS: At one year, 55 patients (1.2%) had experienced major adverse cardiac events, and 123 (2.8%) had undergone late revascularization (>90 days). Following adjustment for age, sex, risk factors, stenosis, and FFRCT, total plaque volume above the receiver-operator characteristic curve-derived optimal cutoff (total plaque volume >564 mm3) was associated with the major adverse cardiac event/late revascularization composite (adjusted hazard ratio, 1.515 [95% CI, 1.093-2.099]; P=0.0126), and both components. Total percent atheroma volume greater than the optimal cutoff was associated with both major adverse cardiac event/late revascularization (total percent atheroma volume >24.4%; hazard ratio, 2.046 [95% CI, 1.474-2.839]; P<0.0001) and cardiovascular death/myocardial infarction (total percent atheroma volume >37.17%, hazard ratio, 4.53 [95% CI, 1.943-10.576]; P=0.0005). Calcified, noncalcified, and low-attenuation percentage atheroma volumes above the optimal cutoff were associated with all adverse outcomes, although this relationship was not maintained for cardiovascular death/myocardial infarction in analyses stratified by median plaque volumes. CONCLUSIONS: Analysis of the ADVANCE registry using artificial intelligence-enabled quantitative plaque analysis shows that total plaque volume is associated with one-year adverse clinical events, with incremental predictive value over luminal stenosis or abnormal physiology by FFRCT. REGISTRATION: URL: https://www.clinicaltrials.gov; Unique identifier: NCT02499679.

Methods, design, and initial results of an angiographic core lab from VOYAGER-PAD.

Background: Anatomy is critical in risk stratification and therapeutic decision making in coronary disease. The relationship between anatomy and outcomes is not well described in PAD. We sought to develop an angiographic core lab within the VOYAGER-PAD trial. The current report describes the methods of creating this core lab, its study population, and baseline anatomic variables. Methods: Patients undergoing lower-extremity revascularization for symptomatic PAD were randomized in VOYAGER-PAD. The median follow up was 2.25 years. Events were adjudicated by a blinded Clinical Endpoint Committee. Angiograms were collected from study participants; those with available angiograms formed this core lab cohort. Angiograms were scored for anatomic and flow characteristics by trained reviewers blinded to treatment. Ten percent of angiograms were evaluated independently by two reviewers; inter-rater agreement was assessed. Clinical characteristics and the treatment effect of rivaroxaban were compared between the core lab cohort and noncore lab participants. Anatomic data by segment were analyzed. Results: Of 6564 participants randomized in VOYAGER-PAD, catheter-based angiograms from 1666 patients were obtained for this core lab. Anatomic and flow characteristics were collected across 16 anatomic segments by 15 reviewers. Concordance between reviewers for anatomic and flow variables across segments was 90.5% (24,417/26,968). Clinical characteristics were similar between patients in the core lab and those not included. The effect of rivaroxaban on the primary efficacy and safety outcomes was also similar. Conclusions: The VOYAGER-PAD angiographic core lab provides an opportunity to correlate PAD anatomy with independently adjudicated outcomes and provide insights into therapy for PAD. (ClinicalTrials.gov Identifier: NCT02504216).

Hypertension, microvascular obstruction and infarct size in patients with STEMI undergoing PCI: Pooled analysis from 7 cardiac magnetic resonance imaging studies.

BACKGROUND: Mortality after ST-segment elevation myocardial infarction (STEMI) is increased in patients with hypertension. The mechanisms underlying this association are uncertain. We sought to investigate whether patients with STEMI and prior hypertension have greater microvascular obstruction (MVO) and infarct size (IS) compared with those without hypertension. METHODS: We pooled individual patient data from 7 randomized trials of patients with STEMI undergoing primary percutaneous coronary intervention (PCI) in whom cardiac magnetic resonance imaging was performed within 1 month after reperfusion. The associations between hypertension and MVO, IS, and mortality were assessed in multivariable adjusted models. RESULTS: Among 2174 patients (61.3 ± 12.6 years, 76% male), 1196 (55.0%) had hypertension. Patients with hypertension were older, more frequently diabetic and had more extensive coronary artery disease than those without hypertension. MVO and IS measured as percent LV mass were not significantly different in patients with and without hypertension (adjusted differences 0.1, 95% CI -0.3 to 0.6, P = .61 and -0.2, 95% CI -1.5 to 1.2, P = .80, respectively). Hypertension was associated with a higher unadjusted risk of 1-year death (hazard ratio [HR] 2.28, 95% CI 1.44-3.60, P < .001), but was not independently associated with higher mortality after multivariable adjustment (adjusted HR 1.04, 95% CI 0.60-1.79, P = .90). CONCLUSION: In this large-scale individual patient data pooled analysis, hypertension was not associated with larger IS or MVO after primary PCI for STEMI.

Impact of Smoking on Coronary Volume-to-Myocardial Mass Ratio: An ADVANCE Registry Substudy.

Purpose To examine the relationship between smoking status and coronary volume-to-myocardial mass ratio (V/M) among individuals with coronary artery disease (CAD) undergoing CT fractional flow reserve (CT-FFR) analysis. Materials and Methods In this secondary analysis, participants from the ADVANCE registry evaluated for suspected CAD from July 15, 2015, to October 20, 2017, who were found to have coronary stenosis of 30% or greater at coronary CT angiography (CCTA) were included if they had known smoking status and underwent CT-FFR and V/M analysis. CCTA images were segmented to calculate coronary volume and myocardial mass. V/M was compared between smoking groups, and predictors of low V/M were determined. Results The sample for analysis included 503 current smokers, 1060 former smokers, and 1311 never-smokers (2874 participants; 1906 male participants). After adjustment for demographic and clinical factors, former smokers had greater coronary volume than never-smokers (former smokers, 3021.7 mm3 ± 934.0 [SD]; never-smokers, 2967.6 mm3 ± 978.0; P = .002), while current smokers had increased myocardial mass compared with never-smokers (current smokers, 127.8 g ± 32.9; never-smokers, 118.0 g ± 32.5; P = .02). However, both current and former smokers had lower V/M than never-smokers (current smokers, 24.1 mm3/g ± 7.9; former smokers, 24.9 mm3/g ± 7.1; never-smokers, 25.8 mm3/g ± 7.4; P < .001 [unadjusted] and P = .002 [unadjusted], respectively). Current smoking status (odds ratio [OR], 0.74 [95% CI: 0.59, 0.93]; P = .009), former smoking status (OR, 0.81 [95% CI: 0.68, 0.97]; P = .02), stenosis of 50% or greater (OR, 0.62 [95% CI: 0.52, 0.74]; P < .001), and diabetes (OR, 0.67 [95% CI: 0.56, 0.82]; P < .001) were independent predictors of low V/M. Conclusion Both current and former smoking status were independently associated with low V/M. Keywords: CT Angiography, Cardiac, Heart, Ischemia/Infarction Clinical trial registration no. NCT02499679 Supplemental material is available for this article. © RSNA, 2024.

Bleeding Outcomes in Patients Treated With Asundexian in Phase II Trials.

BACKGROUND: Phase II trials of asundexian were underpowered to detect important differences in bleeding. OBJECTIVES: The goal of this study was to obtain best estimates of effects of asundexian vs active control/placebo on major and clinically relevant nonmajor (CRNM) and all bleeding, describe most common sites of bleeding, and explore association between asundexian exposure and bleeding. METHODS: We performed a pooled analysis of 3 phase II trials of asundexian in patients with atrial fibrillation (AF), recent acute myocardial infarction (AMI), or stroke. Bleeding was defined according to the International Society on Thrombosis and Hemostasis (ISTH) criteria. RESULTS: In patients with AF (n = 755), both asundexian 20 mg and 50 mg once daily vs apixaban had fewer major/CRNM events (3 of 249; incidence rate [IR] per 100 patient-years 5.47 vs 1 of 254 [IR: not calculable] vs 6 of 250 [IR: 11.10]) and all bleeding (12 of 249 [IR: 22.26] vs 10 of 254 [IR: 18.21] vs 26 of 250 [IR: 50.56]). In patients with recent AMI or stroke (n = 3,409), asundexian 10 mg, 20 mg, and 50 mg once daily compared with placebo had similar rates of major/CRNM events (44 of 840 [IR: 7.55] vs 42 of 843 [IR: 7.04] vs 56 of 845 [IR: 9.63] vs 41 of 851 [IR: 6.99]) and all bleeding (107 of 840 [IR: 19.57] vs 123 of 843 [IR: 22.45] vs 130 of 845 [IR: 24.19] vs 129 of 851 [IR: 23.84]). Most common sites of major/CRNM bleeding with asundexian were gastrointestinal, respiratory, urogenital, and skin. There was no significant association between asundexian exposure and major/CRNM bleeding. CONCLUSIONS: Analyses of phase II trials involving >500 bleeds highlight the potential for improved safety of asundexian compared with apixaban and similar safety compared with placebo. Further evidence on the efficacy of asundexian awaits the results of ongoing phase III trials.

Optimizing warfarin dosing for patients with atrial fibrillation using machine learning.

While novel oral anticoagulants are increasingly used to reduce risk of stroke in patients with atrial fibrillation, vitamin K antagonists such as warfarin continue to be used extensively for stroke prevention across the world. While effective in reducing the risk of strokes, the complex pharmacodynamics of warfarin make it difficult to use clinically, with many patients experiencing under- and/or over- anticoagulation. In this study we employed a novel implementation of deep reinforcement learning to provide clinical decision support to optimize time in therapeutic International Normalized Ratio (INR) range. We used a novel semi-Markov decision process formulation of the Batch-Constrained deep Q-learning algorithm to develop a reinforcement learning model to dynamically recommend optimal warfarin dosing to achieve INR of 2.0-3.0 for patients with atrial fibrillation. The model was developed using data from 22,502 patients in the warfarin treated groups of the pivotal randomized clinical trials of edoxaban (ENGAGE AF-TIMI 48), apixaban (ARISTOTLE) and rivaroxaban (ROCKET AF). The model was externally validated on data from 5730 warfarin-treated patients in a fourth trial of dabigatran (RE-LY) using multilevel regression models to estimate the relationship between center-level algorithm consistent dosing, time in therapeutic INR range (TTR), and a composite clinical outcome of stroke, systemic embolism or major hemorrhage. External validation showed a positive association between center-level algorithm-consistent dosing and TTR (R2 = 0.56). Each 10% increase in algorithm-consistent dosing at the center level independently predicted a 6.78% improvement in TTR (95% CI 6.29, 7.28; p < 0.001) and a 11% decrease in the composite clinical outcome (HR 0.89; 95% CI 0.81, 1.00; p = 0.015). These results were comparable to those of a rules-based clinical algorithm used for benchmarking, for which each 10% increase in algorithm-consistent dosing independently predicted a 6.10% increase in TTR (95% CI 5.67, 6.54, p < 0.001) and a 10% decrease in the composite outcome (HR 0.90; 95% CI 0.83, 0.98, p = 0.018). Our findings suggest that a deep reinforcement learning algorithm can optimize time in therapeutic range for patients taking warfarin. A digital clinical decision support system to promote algorithm-consistent warfarin dosing could optimize time in therapeutic range and improve clinical outcomes in atrial fibrillation globally.

Efficacy and Safety of Non-Vitamin-K Antagonist Oral Anticoagulants Versus Warfarin Across the Spectrum of Body Mass Index and Body Weight: An Individual Patient Data Meta-Analysis of 4 Randomized Clinical Trials of Patients With Atrial Fibrillation.

BACKGROUND: The efficacy and safety of non-vitamin-K antagonist oral anticoagulants (NOACs) across the spectrum of body mass index (BMI) and body weight (BW) remain uncertain. METHODS: We analyzed data from COMBINE AF (A Collaboration Between Multiple Institutions to Better Investigate Non-Vitamin K Antagonist Oral Anticoagulant Use in Atrial Fibrillation), which pooled patient-level data from the 4 pivotal randomized trials of NOAC versus warfarin in patients with atrial fibrillation. The primary efficacy and safety outcomes were stroke or systemic embolic events (stroke/SEE) and major bleeding, respectively; secondary outcomes were ischemic stroke/SEE, intracranial hemorrhage, death, and the net clinical outcome (stroke/SEE, major bleeding, or death). Each outcome was examined across BMI and BW. Because few patients had a BMI <18.5 kg/m2 (n=598), the primary analyses were restricted to those with a BMI ≥18.5 kg/m2. RESULTS: Among 58 464 patients, the median BMI was 28.3 (interquartile range, 25.2-32.2) kg/m2, and the median BW was 81.0 (interquartile range, 70.0-94.3) kg. The event probability of stroke/SEE was lower at a higher BMI irrespective of treatment, whereas the probability of major bleeding was lower at a higher BMI with warfarin but relatively unchanged across BMI with NOACs. NOACs reduced stroke/SEE overall (adjusted hazard ratio [HRadj], 0.80 [95% CI, 0.73-0.88]; P<0.001), with a generally consistent effect across BMI (Ptrend across HRs, 0.48). NOACs also reduced major bleeding overall (HRadj, 0.88 [95% CI, 0.82-0.94]; P<0.001), but with attenuation of the benefit at a higher BMI (trend test across BMI [Ptrend], 0.003). The overall treatment effects of NOACs versus warfarin for secondary outcomes were consistent across BMI, with the exception of the net clinical outcome and death. While these outcomes were overall reduced with NOACs (net clinical outcome, HRadj, 0.91 [95% CI, 0.87-0.95]; P<0.001; death, HRadj, 0.91 [95% CI, 0.86-0.97]; P=0.003), these benefits were attenuated at higher BMI (Ptrend, 0.001 and 0.08, respectively). All findings were qualitatively similar when analyzed across BW. CONCLUSIONS: The treatment effect of NOACs versus warfarin in atrial fibrillation is generally consistent for stroke/SEE across the spectrum of BMI and BW, whereas the reduction in major bleeding is attenuated in those with higher BMI or BW. Death and the net clinical outcome are overall reduced with NOACs over warfarin, although there remain uncertainties for these outcomes at a very high BMI and BW.