搜索结果：Yonsei medical journal

Primary aldosteronism (PA) is a common and treatable cause of hypertension, yet data on the durability of medical treatment response beyond the first year are limited. Using the PA medical treatment outcomes framework, we evaluated long-term biochemical and clinical outcomes and factors associated with treatment response. We conducted an international, multicenter, observational cohort study across 27 centers on 4 continents. Adults with PA treated with mineralocorticoid receptor antagonists and epithelial sodium channel blockers between 2016 and 2021 were included if follow-up data were available at ≥12 months. Biochemical and clinical responses were classified as complete, partial, or absent using PA medical treatment outcomes criteria. Ordinal logistic regression identified determinants of a favorable treatment response. The cohort included 1292 patients (47.5% women, mean age 52 years) with a median follow-up of 34 months (interquartile range, 21-51). At ≥12 months, 58.3% achieved a complete biochemical response, and 20.8% achieved a complete clinical response, with only 5% showing no response. Determinants of a favorable biochemical response included lower aldosterone and lateralization index, and higher renin and potassium levels at baseline. In those taking spironolactone, a higher daily dose was also associated with a favorable biochemical response. Determinants of a favorable clinical response included female sex, lower body mass index, shorter duration of hypertension, and absence of hypertension-mediated organ damage. Targeted medical therapy for PA can deliver sustained biochemical and clinical benefits. Early disease detection and adequate dose titration are highly actionable determinants of long-term treatment success.

Cricothyroidotomy is a critical procedure for managing difficult airways, yet it is infrequently performed, leading to limited familiarity among medical professionals. This study explores the effectiveness of augmented reality (AR)-based training using HoloLens compared to traditional non-head mounted device (HMD) training in teaching cricothyroidotomy. This study aims to confirm the effect of technical practice using AR on participants' performance ability, performance confidence, learning self-efficacy, and practical satisfaction, and to present basic data for using AR in medical education. All participants received initial guidance via PowerPoint (PPT) on laptops. The AR group performed the procedure with three-dimensional models via HoloLens, while the non-HMD group used the initial PPT guidance. Effectiveness was assessed through a pre- and post-test design, including cricothyroidotomy estimated time, skill confidence, NASA-Task Load Index, System Usability Scale, and participant satisfaction. The average pre-test procedure time for the AR group was 188 seconds, and 150 seconds for the non-HMD group. In the post-test, the AR group recorded 146 seconds and the non-HMD group 136 seconds, showing no statistically significant difference in procedure time (p>0.05). The HMD group showed a significant improvement in confidence levels (median increase from 1.0 to 5.0), and both groups demonstrated similar proficiency in cricothyroidotomy performance over time. Satisfaction rates were similar, but the AR group reported lower mental demand. Despite initial technical challenges, AR showed potential for enhancing learning and confidence without compromising performance, indicating the need for further research to optimize AR in medical training.

Test-time scaling has emerged as a promising method to enhance the reasoning capabilities of large language models (LLMs) and vision-language models (VLMs) during inference without additional training. While foundational studies established scaling paradigms in general domains, their applicability to the unique complexities of medical AI remains underexplored. This study aims to conduct a comprehensive investigation of test-time scaling in the medical domain. We evaluate the impact of scaling across different model sizes and task complexities. Furthermore, we seek to identify domain-specific bottlenecks and assess model robustness against user-driven perturbations, such as misleading clinical authority. This study evaluated a diverse set of general and medical-specific LLMs and VLMs. Experiments used five textual medical benchmarks comprising over 5500 questions and two multimodal benchmarks comprising 7000 samples. Performance was measured under three scaling conditions: increasing token budgets, iterative sequential scaling, and parallel scaling. Robustness was tested by embedding misleading hints with varying tones and levels of simulated clinical expertise into prompts. For nonreasoning LLMs, accuracy saturated quickly, with token usage often remaining under 500 tokens regardless of budget increases. Reasoning models demonstrated significant performance gains on complex tasks as token budgets increased. Notably, we identified distinct domain-specific behaviors. First, current VLMs showed a structural bottleneck in integrating visual clues and experienced limited benefit from token expansion. Second, medically fine-tuned LLMs excelled in clinical question answering but exhibited degraded scaling efficiency on calculation tasks compared to general-domain models. This reflects a disparity between qualitative clinical alignment and procedural logic. Third, while optimal scaling improved robustness, models exhibited a cognitive vulnerability by readily abandoning correct reasoning when confronted with misleading expert physician hints. Regarding scaling strategies, parallel scaling outperformed sequential scaling on easier tasks. Conversely, extended sequential scaling or increased budgets proved essential for complex problem-solving. Test-time scaling rules from general domains do not perfectly translate to medical AI. Longer reasoning is not universally beneficial. Concise reasoning with parallel scaling is optimal for simpler tasks. An extended chain of thought via sequential scaling or increased budgets is required for complex problems. Furthermore, safe clinical deployment requires addressing fundamental vision-language alignment, balancing clinical and procedural reasoning, and mitigating vulnerabilities to perceived clinical authority.

This corrects the article on p. 741 in vol. 61, PMID: 32882758.

This corrects the article on p. 408 in vol. 67, PMID: 42044982.

The single-port robot-assisted partial nephrectomy (SP-RAPN) procedure has emerged as a minimally invasive approach for the surgical treatment of localized renal tumors. Although numerous publications have been reported, the scientific structure of this research field has not been systematically evaluated. This bibliometric study aimed to identify publication characteristics, key contributors, collaborations, and research themes related to SP-RAPN. Records were retrieved from the Web of Science Core Collection database. After excluding non-English studies, publications outside the Science Citation Index Expanded, non-original article types, and studies unrelated to the topic, 62 original articles published between 2009 and 2025 were included. Bibliometric mapping and visualization analyses were performed using Biblioshiny and VOSviewer. Publication output, citation trends, journal productivity, institutional contributions, country productivity, co-authorship networks, and keyword patterns were evaluated. A marked increase in publications was observed after 2021, with the highest annual output recorded in 2025 (n = 15). The Journal of Endourology and Journal of Robotic Surgery were the most productive journals, each publishing 12 articles. Yonsei University was the leading institution with 29 publications. The United States contributed the highest number of publications, citations, and authors. These findings demonstrate growing scientific interest in SP-RAPN. Although research activity remains concentrated in a limited number of countries and institutions, collaborative interactions appear to be increasing. Further multicenter and prospective studies are needed to better define the role of SP-RAPN in minimally invasive renal surgery.

Nail unit melanoma (NUM) is a rare but potentially fatal malignancy often misdiagnosed as melanonychia. Because biopsy may cause permanent nail dystrophy, accurate noninvasive diagnosis is essential. This study aimed to develop and validate an artificial intelligence model to distinguish NUM from benign melanonychia using clinical images and to assess its diagnostic utility through human comparison and external validation. Clinical images from 172 patients with melanonychia and 122 patients with NUM were retrospectively collected. Pediatric melanonychia was excluded to minimize diagnostic ambiguity. Three convolutional neural networks (CNNs) were trained and validated with patient-wise splits. Human validation involved 40 representative images assessed by dermatologists, residents, and non-medical participants, with and without CNN assistance. ResNet-50 achieved the highest sensitivity (87.2 %). DenseNet-121 achieved the highest AUROC (area under the receiver operating characteristic curve) (0.954). CNN assistance improved the diagnostic accuracy of all human raters (accuracy increased from 70.0 % to 80.8 %), with dermatology residents showing the largest gain. The proposed CNN-based models demonstrated robust performance in differentiating NUM from melanonychia and improved diagnostic accuracy and agreement among human evaluators. While not intended to replace clinical judgment, this approach shows promise as a supportive screening tool in clinical settings, warranting further validation in larger, multi-institutional, and multi-ethnic cohorts.

Since the 2020 Korean guidelines for Helicobacter pylori treatment, clarithromycin resistance rates have risen from 17.8% to 33.3%, dual-priming oligonucleotide-based polymerase chain reaction-guided tailored therapy has been adopted, and potassium-competitive acid blockers (P-CABs) have become available. This fourth revision addressed these changes. Nine key questions were addressed through systematic review and meta-analysis. Thirteen recommendations were evaluated using a modified Delphi process involving 64 experts. Twelve recommendations achieved a first-round consensus; one required revision and achieved 73.9% agreement. Key changes included: 1) a dual-pillar strategy of tailored therapy and empirical quadruple therapy; 2) restricted use of empirical clarithromycin- based triple therapy under specific conditions; 3) removal of sequential therapy; 4) use of P-CABs as alternatives to proton pump inhibitors; 5) expansion of eradication indications to include gastric cancer prevention in H. pylori gastritis and regression of hyperplastic polyps ≤10 mm; and 6) positioning of bismuth quadruple therapy as a conditionally recommended first-line empirical option, with preference for reservation as salvage therapy, and introduction of modified bismuth quadruple therapy (addition of bismuth to conventional regimens) as an additional first-line empirical option. The revised guidelines provide updated evidence-based recommendations for the diagnosis and treatment of H. pylori infection, reflecting the rapidly changing antibiotic resistance landscape and the introduction of new diagnostic and therapeutic tools in Korea. These guidelines aim to assist clinicians, patients, policymakers, and medical educators in optimizing H. pylori management. They may differ from the current medical insurance standards and will be further revised based on emerging evidence.

Cancer registries remain the gold standard for global cancer monitoring, yet complementing them with electronic health records and claims can significantly enhance the understanding of the cancer burden by providing a more complete picture of the patient journey. The main aim of this project is to serve as a proof of concept for using real-world data mapped to the Observational Medical Outcomes Partnership (OMOP) common data model (CDM) to monitor cancer epidemiology over time and characterise patients' clinical history and outcomes. This study will be conducted as an observational cohort study using a multinational network of large real-world data sources mapped to the OMOP CDM. Electronic health records (EHR) from primary and secondary care, health insurance claims and cancer registry data will be included. To date, 20 databases from 16 countries, mainly from Europe but also North America and Asia, have committed to participate in the project.We will investigate the temporal trends in incidence, prevalence and survival of 36 cancers across haematopoietic and solid tumours from 2000 (or the start of accurate data if later) to the last year with complete data. Data from all individuals registered in each of the participating data sources will be eligible for inclusion in the study. For primary care EHR and claims, individuals will be required to have at least 1 year of prior observation to ensure the identification of incident cases and adequate capture of patient characteristics. We will estimate crude and age-standardised incidence and 5-year partial prevalence. Additionally, we will estimate crude and age-standardised overall survival at 1, 5 and 10 years for the total study period and by diagnosis year groups defined according to data availability. All study objectives will be investigated at the database level, with results stratified by age and sex. For incidence and survival analyses, additional stratifications will be performed by clinical conditions and smoking status (where available). We will use the National Cancer Institute (NCI) Joinpoint Regression Programme to model overall trends in cancer incidence and the NCI JPSurv software to estimate trends in survival. Finally, we will characterise individuals diagnosed with an incident cancer based on demographics, clinical conditions and medication use at different time windows.Findings will be presented separately for each database and further summarised through descriptive aggregation by country and data source type. Each data partner will obtain study approval from their local institutional review boards prior to study execution. Distributed queries will be employed, whereby standardised analytical code is shared and run at each site locally. Deidentified, aggregated results will be returned from all participating sites. A minimum cell count of five will be used when reporting results, depending on each collaborator's data governance requirements.All study code will be publicly available, and findings will be submitted to open science journals to promote transparency and reproducibility.

Intrapancreatic fat deposition (IPFD) has garnered appreciable attention across diverse medical disciplines amid the rising global burden of pancreatitis, pancreatic cancer and type 2 diabetes mellitus. Despite growing interest in the field, there remains no conceptual framework for it. To address this gap, 25 experts from 6 continents were brought together to develop guidance. Drawing on multiple peer-reviewed systematic reviews and applying an iterative, anonymous Delphi process, a consensus document was developed and ratified in Melbourne, Australia. The final document contains 58 recommendations - 30 reached unanimous agreement and 28 attained 90-99% agreement. The consensus formally defines fatty pancreas disorder (FPD) as a distinct pathological state, characterized by excessive IPFD that poses a risk to health. Assessment of IPFD is best performed with MRI. Expert-established diagnostic criteria enable identification and classification of FPD. Recognition of type 1 FPD (in individuals without excess body fat mass) and type 2 FPD (in individuals with excess body fat mass) is instrumental in gaining novel insights and tailoring preventive strategies to the individual. The Melbourne consensus is foundational in operationalizing the dissemination of new knowledge in the field, fostering intercontinental collaborations and, ultimately, addressing the global burden of diseases of the pancreas.

Traumatic cervical spinal cord injury (TCSCI) frequently necessitates mechanical ventilation, and early tracheostomy has been shown to reduce complications in patients requiring prolonged ventilation. This study aimed to develop a machine learning model to predict the need for tracheostomy in TCSCI patients, utilizing early clinical data to improve patient outcomes. This study was conducted using data from 2017 to 2024, obtained from the single institution database. A total of 267 TCSCI patients were included, of whom 49 underwent tracheostomy. Variables selected for the model included demographics, comorbidities, injury level, medical research council motor grading, Glasgow Coma Scale (GCS), and treatment details. This study also implemented SHapley Additive exPlanations analysis to interpret the predictive model and identify significant risk predictors contributing to the outcomes. The CatBoost model outperformed other models, achieving the highest performance metrics. The model that incorporated GCS and the American Spinal Injury Association (ASIA) impairment scale (AIS) yielded the highest area under the curve (AUC) score. Following variable selection, the CatBoost model, utilizing age, GCS, AIS score, surgery, and injury levels (C3/4, C4/5, and C2/3), achieved an AUC score of 0.8166 and an accuracy of 0.8652. Our machine learning model effectively predicted the need for tracheostomy in TCSCI patients. Important predictors were age, GCS, AIS score, cervical surgery, and injury level. This model may improve outcomes for patients requiring tracheostomy.

Evaluating clinical reasoning in large language models (LLMs) poses two open challenges: reference-oriented semantic metrics do not directly assess whether a model's stated diagnosis is supported by the evidence in its own justification, and the increasingly popular LLM-as-judge approach rests on a largely untested assumption-that independent verifier LLMs agree with one another. We assess three generator LLMs (HuatuoGPT-o1-8B, Meta-Llama-3.1-8B-Instruct, Meta-Llama-3.3-70B-Instruct) on 1,000 MIMIC-IV hospital-stay cases along four complementary axes (medical concept grounding, semantic similarity, semantic uncertainty, and evidence-conclusion coherence), with coherence judged independently by three frontier verifiers (Claude Sonnet 4.6, Gemini 2.5 Pro, GPT-5.4 mini). Two findings emerge. First, coherence reveals a dissociation that reference-oriented metrics do not capture: a model can score well on those axes yet still produce rationales that do not support its own conclusions. Second, inter-verifier agreement on coherence is consistently low (Fleiss' κ 0.087-0.223; disagreement 62.2%-74.3%), so the same rationale can be judged supported or unsupported depending on the verifier. A preliminary validation in which a physician adjudicated 50 cases echoed this: agreement with the physician varied across verifiers, underscoring that no single LLM reliably stands in for clinical assessment. Together, these results suggest a single LLM verifier lacks sufficient reliability to serve as a stand-alone judge of clinical reasoning at scale, and that structured human oversight remains essential. The unanimous-agreement tier offers a candidate for selective automation, but its clinical reliability remains to be confirmed in larger, multi-clinician adjudication studies.

The gut virome is emerging as a key contributor to host physiology, yet it remains underexplored relative to the gut bacteriome. Despite comprehensive viral catalogs for the human gut, comparable resources for mice, the principal model organism in biomedical research, remain lacking, limiting our understanding of virome-host interactions and their implications for host health and disease. Here, we present Mouse Reference Gut Virome (MRGV), a comprehensive catalog comprising 109,778 viral genomes with ≥50% completeness and representing 28,824 species, expanding known mouse gut viral diversity by ~67.6%. Viral genome binning accounted for ~36% of cataloged genomes and improved average completeness by ~60% relative to contig-level assemblies. Structure-based annotation assigned functions to nearly half of 8.2 million viral proteins. Comparison with human gut virome revealed pronounced taxonomic and functional divergence. MRGV enabled host linkage for 88% of viral genomes and validated diverse mouse-associated crAss-like lineages with lineage-specific host-range and lifestyle strategies. Finally, mouse gut virome features harbored host ageing-associated signals distinct from those captured by bacterial features, with most ageing markers derived from binned genomes, highlighting the importance of viral binning for elucidating virome-host interactions.

Numerous studies have shown that the atherogenic index of plasma (AIP) is a strong predictor of the risk of cardiovascular diseases (CVDs) and an independent predictor of cardiovascular events (CEs) and related mortality. The predictive value of AIP for in-hospital mortality (IHM) in individuals with sepsis remains uncertain. Therefore, this study aimed to investigate the association between AIP and IHM among patients with sepsis using a large sample from the Medical Information Mart for Intensive Care (MIMIC) database. Individuals with sepsis were identified from the MIMIC-IV database and categorized into four groups according to AIP quartiles. The IHM was the primary outcome. The association between AIP and IHM was evaluated using logistic regression and restricted cubic spline (RCS) models. A total of 2243 patients (59% male) were included this study. In univariable logistic regression, higher AIP was significantly associated with increased IHM (odds ratio: 1.18 [95% confidence interval: 1.05-1.33]; p of Wald test=0.005). The RCS model showed no evidence of a nonlinear association between AIP and IHM (p for nonlinearity>0.05). Sensitivity analysis demonstrated consistent findings regarding the magnitude and direction of the effects across different subgroups, indicating stable results. Elevated AIP was associated with higher IHM in individuals with sepsis.

The effect of postthrombectomy blood pressure (BP) management on the development of acute kidney injury (AKI) in patients with acute ischemic stroke remains largely unexplored. This secondary analysis of the OPTIMAL-BP trial (Outcome in Patients Treated With Intra-Arterial Thrombectomy-Optimal Blood Pressure Control) included patients with acute ischemic stroke due to large-vessel occlusion who achieved successful endovascular thrombectomy and had a systolic BP &#x2265;140 mm&#x2009;Hg. Patients were randomized to intensive (target systolic BP <140 mm&#x2009;Hg) or conventional (target systolic BP 140-180 mm&#x2009;Hg) BP management for 24 hours. The outcomes were AKI within 7 days and within 2 days, defined according to the Kidney Disease: Improving Global Outcomes criteria. In addition, we examined the associations between AKI and functional independence at 3 months, defined as a modified Rankin Scale score of 0 to 2. Multivariable logistic regression analyses were performed with adjustment for age, sex, time from stroke onset to enrollment, baseline National Institutes of Health Stroke Scale score, and baseline estimated glomerular filtration rate. Of 306 patients, 19 were excluded, and 287 patients were included in this analysis (mean age, 73.2 years; 117 [40.8%] women). AKI within 7 days occurred more frequently in the intensive management group than in the conventional group (20/147 [13.6%] versus 9/140 [6.4%]; adjusted odds ratio, 2.54 [95% CI, 1.10-6.35]). Most AKI events were stage 1 (20/29 [69.0%]). Early AKI within 2 days was also more common with intensive BP management. Patients with AKI had significantly lower rates of functional independence (4/29 [13.8%] versus 126/257 [49.0%]; adjusted odds ratio, 0.19 [95% CI, 0.05-0.55]) and higher stroke-related mortality at 3 months (11/29 [37.9%] versus 8/257 [3.1%]; adjusted odds ratio, 13.8 [95% CI, 4.14-49.64]). In a sensitivity analysis with equal creatinine ascertainment, the association with 48-hour AKI did not reach statistical significance (7/76 [9.2%] versus 2/66 [3.0%]; adjusted odds ratio, 4.20 [95% CI, 0.83-32.6]), although the absolute risk difference remained directionally consistent. Intensive BP lowering after successful endovascular thrombectomy was associated with a higher risk of AKI, even when kidney injury was predominantly mild. In addition, AKI was associated with worse neurological outcomes. These findings suggest that AKI is an important marker of systemic hemodynamic vulnerability after aggressive postendovascular thrombectomy BP lowering. URL: https://www.clinicaltrials.gov; Unique identifier: NCT04205305.

Understanding how photosynthetic microorganisms exchange electrons with electrodes is central to advancing biophotoelectrochemical systems. Here, we investigate bias-dependent photocurrent generation in biohybrid electrodes containing Limnospira indica immobilized in a PEDOT:PSS composite on fluorine-doped tin oxide (FTO). By systematically varying irradiance, applied potential, oxygen availability, and mediator composition, we show that photocurrent arises from multiple electron-transfer pathways whose engagement depends sensitively on electrical polarization. Under anodic polarization (+0.5 to +0.9&#xa0;V), photocurrent shows a trend of increase with light intensity while photon-to-current efficiency declines systematically, consistent with regulatory constraints within the photosynthetic chain. In contrast, cathodic polarization (-0.5 to -0.9&#xa0;V) induces distinct multiphasic transient behavior: at -0.5&#xa0;V, reproducible spike-peak-trough sequences and delayed post-illumination currents reflect competition between photosynthetic electron-transfer pathways and oxygen-dependent metabolic sinks, and oxygen removal simplifies but reduces the cathodic response. At -0.9&#xa0;V, cathodic currents become largely oxygen-independent yet remain strictly PSII-dependent, as DCMU abolishes both photocurrent development and dark recovery. These results demonstrate that L. indica engages multiple potential-selective electron-transfer regimes and that transient photocurrent features carry mechanistic information inaccessible from steady-state measurements, informing the rational design of improved biohybrid photoelectrodes.

The End-Stage Renal Disease Adherence Questionnaire (ESRD-AQ) evaluates adherence behaviors of patients on maintenance hemodialysis (e.g., hemodialysis attendance, medication use, fluid restrictions, and diet restrictions). The ESRD-AQ, with 46 items, is considered lengthy by some clinicians with time constraints. Therefore, this study was conducted to develop and psychometrically evaluate a 9-item short form of the ESRD-AQ (ESRD-AQ-SF) for faster and more efficient administration. Content validity, construct validity, and test-retest reliability were assessed. Results demonstrated that the ESRD-AQ-SF is a valid and reliable tool for evaluating treatment adherence in patients undergoing maintenance hemodialysis.

Mycobacterium avium subsp. paratuberculosis (MAP) is the causative agent of Johne's disease, a chronic enteritis in ruminants, and is capable of persisting within macrophages despite the activation of host immune defenses. Although this intracellular persistence is a key determinant of MAP pathogenicity, the bacterial factors and host responses that regulate this process remain poorly understood. In this study, we established the first CRISPR interference (CRISPRi) platform applied to bovine monocyte-derived macrophages (MDM) to evaluate the functions of MAP genes involved in intracellular survival and to perform an integrative analysis of host transcriptomic responses. MAP mutants were targeted to two genes (mdh and MAP1981c). The optimal concentration of anhydrotetracycline (ATc) was determined to be 2 &#x3bc;g/ml by measuring the survival of the cells and the downregulation of gene expression levels in the cells up to 72 h. The gene expression profiles and intracellular MAP levels were investigated using RNA-seq and colony-forming units, respectively. The survival rates of the MAP mutants significantly decreased with the time course of infection in MAP-mdhKD and MAP1981cKD (KD, knockdown). RNA-seq-based gene expression profiling suggested that target gene silencing in MAP mutants led to altered expression of host genes involved in lipid metabolism, T-cell activation reduction, and antimicrobial response in bovine MDM, contributing to reduced intracellular survival of MAP. Our study demonstrates that the downregulation of mdh and MAP1981c in MAP significantly alters the host transcriptomic landscape in bovine MDM, revealing their critical roles in subverting host immune defenses for intracellular persistence.

In patients with atrial fibrillation and chronic coronary syndrome beyond 12 months after percutaneous coronary intervention (PCI), oral anticoagulant monotherapy is recommended by the guideline; however, its efficacy and safety in patients with complex PCI remain uncertain. We conducted a post hoc analysis of the randomized ADAPT AF-DES trial (Appropriate Duration of Antiplatelet and Thrombotic Strategy After 12 Months in Patients With Atrial Fibrillation Treated With Drug-Eluting Stents), which compared non-vitamin K antagonist oral anticoagulant (NOAC) monotherapy with NOAC plus clopidogrel in patients with atrial fibrillation &#x2265;12 months after second- or third-generation drug-eluting stent implantation. Complex PCI was defined by one of the following characteristics: &#x2265;3 stents, &#x2265;3 lesions, bifurcation with 2 stents, total stent length &#x2265;60 mm, left main PCI, or chronic total occlusion PCI. Net adverse clinical events, ischemic composite outcomes, and bleeding composite outcomes were evaluated according to PCI complexity. Among 960 patients, 247 (25.7%) underwent complex PCI and 713 (74.3%) underwent noncomplex PCI. NOAC monotherapy was associated with a lower risk of net adverse clinical events compared with combination therapy in both the complex PCI group (9.5% versus 21.5%; hazard ratio, 0.42 [95% CI, 0.21-0.83]; P=0.01) and the noncomplex PCI group (9.6% versus 15.7%; hazard ratio, 0.59 [95% CI, 0.39-0.90]; P=0.02), with no significant interaction. Ischemic outcomes were infrequent and did not differ significantly between treatment strategies regardless of PCI complexity, whereas bleeding outcomes were consistently lower with NOAC monotherapy in both complex and noncomplex PCI groups. In this post hoc analysis of the randomized ADAPT AF-DES trial, NOAC monotherapy beyond 12 months after PCI was associated with a substantial reduction in bleeding regardless of PCI complexity. Although limited by low absolute event rates, ischemic outcomes did not differ significantly between treatment strategies. These findings suggest that a bleeding-focused long-term antithrombotic strategy may be reasonable in patients with atrial fibrillation and chronic coronary syndrome, including those with prior complex PCI, although prospective validation remains warranted. URL: https://www.clinicaltrials.gov; Unique identifier: NCT04250116.