Background:The prognostic value of preoperative skeletal muscle mass index (SMI) in lung cancer patients who undergo the surgery remains unclear. This meta-analysis aimed to clarify prognostic value of preoperative SMI with long-term survival among operated lung cancer patients. Methods:PubMed, Cochrane Library, and Web of Science databases were searched up to March 21, 2026. Endpoints for survival included the overall survival (OS), disease-free survival (DFS) and cancer-specific survival (CSS). Subgroup analyses stratified by the neoadjuvant therapy and tumor type for OS were further performed. Results:Fourteen retrospective studies with 8,008 participants were included. After combining available data, it was identified that a lower SMI was associated with poorer OS [hazard ratio (HR) =1.35, 95% confidence interval (CI): 1.30-1.52, P<0.001] and DFS (HR =2.10, 95% CI: 1.06-4.15, P=0.03) and potentially poorer CSS (HR =1.64, 95% CI: 0.91-2.95, P=0.10). Subgroup analyses for the OS based on the neoadjuvant therapy (no: HR =1.90, P=0.003; yes: HR =2.50, P=0.03; mixed: HR =1.27, P=0.004) and tumor type (non-small cell lung cancer: HR =1.28, P<0.001; lung cancer: HR =2.03, P<0.001) manifested consistent results. Conclusions:Lower preoperative SMI was associated with worse survival among surgical lung cancer patients and may have potential value in prognostic assessment.
Background:Preserved ratio impaired spirometry (PRISm) is spirometric pattern characterized by reduced forced expiratory volume in 1 second (FEV1) despite a preserved FEV1/forced vital capacity (FVC) ratio. PRISm has been associated with respiratory symptoms, impaired lung function, and increased mortality; however, data from Southeast Asian populations remain limited. This study aimed to determine the prevalence, spirometric trajectory, contributing factors, and mortality of PRISm in a Thai hospital-based cohort. Methods:We conducted a retrospective cohort study of adults who underwent spirometry at Siriraj Pulmonary Function Laboratory between April 2017 and February 2018, with follow-up through March 2024. PRISm was defined using either: (I) lower limit of normal (LLN) criteria: FEV1 < LLN, FVC ≥ LLN and FEV1/FVC ≥ LLN; or (II) fixed-ratio criteria: FEV1 <80% predicted, FVC ≥80% predicted and FEV1/FVC ≥75% for age ≤45 years or ≥70% for age >45 years. Demographic characteristics, co-morbidities, trajectories and mortality were analyzed. Logistic regression and survival analyses were performed. Results:Among 1,300 subjects, the overall prevalence of PRISm was 8.4% when either definition was applied. The prevalence was 3.5% using LLN criteria, and 6.0% using fixed-ratio criteria. Compared with normal spirometry, those with PRISm were older age and had a higher prevalence of congestive heart failure. PRISm was associated with increased all-cause mortality compared with normal [hazard ratio (HR) 2.37, 95% confidence interval (CI): 1.19-4.72, P=0.001] and remained independently associated after adjustment (aHR 2.08, 95% CI: 1.04-4.15, P=0.04). Among 52 subjects with follow-up spirometry, 30.8% transitioned to normal, 28.9% developed obstructive impairment, 19.2% developed restrictive impairment, and 21.1% remained PRISm. Conclusions:PRISm is a clinically relevant spirometric pattern associated with increased mortality in this Thai cohort. Its dynamic trajectory highlights the importance of longitudinal monitoring and further research to clarify its natural history.
Background:Narcolepsy is a neurological sleep disorder associated with immune response. However, the autoimmune basis for narcolepsy remains unclear. This study aimed to evaluate the causal relationship between immune cells and narcolepsy type 1 (NT1). Methods:We used a two-sample Mendelian randomization (MR) method to investigate the associations between 731 immune cell traits and NT1 based on a genome-wide association study (GWAS) database from the FinnGen consortium. The inverse-variance weighted (IVW) method was used as the primary method, followed by sensitivity analyses, including the MR-Egger intercept test, Cochran's Q test, and MR pleiotropy residual sum and outlier (MR-PRESSO). Additional mediation analysis was conducted to investigate the mediating effect of 91 cytokines on immune cells to facilitate the immune processes in NT1. Results:Immune cell traits showed significant causal associations with NT1. Risk-associated traits mainly involved human leukocyte antigen-DR (HLA-DR)-related monocyte phenotypes, T-cell-related traits, natural killer (NK) cell-related traits, and natural killer T (NKT) cell-related traits, whereas protective traits mainly involved CD4+ T-cell-related, plasmacytoid dendritic cell-related, monocyte-related, and B-cell-related phenotypes. Overall, ten immune cell traits were associated with an increased risk of narcolepsy, whereas five were associated with a reduced risk. Mediation analysis further indicated that interleukin-6 (IL-6) mediates the immune-inflammatory pathway linking monocyte-related traits to NT1. These findings remained consistent in all sensitivity analyses. Conclusions:Our study identified immunophenotypes that are related to the development of NT1, providing insight into the autoimmune pathogenesis of NT1 and subsequent immunotherapy.
Background:Despite curative resection for early-stage lung adenocarcinoma (LUAD), postoperative recurrence remains a significant concern, with substantial outcome heterogeneity even among patients with stage IA disease. An accurate tool to identify individuals at high risk of early postoperative recurrence after curative resection is urgently needed to guide intensified surveillance and adjuvant strategies. Therefore, the objective of this study was to develop and validate an interpretable machine learning model integrating clinical-radiological features, radiomics, and delta-radiomics to predict early postoperative recurrence in stage IA LUAD. Methods:This retrospective study included 463 patients with pathologically confirmed stage IA LUAD who underwent curative surgery. Preoperative serial computed tomography (CT) scans (baseline and follow-up) were used to extract radiomics and delta-radiomics features, the latter quantifying temporal changes in tumor phenotype. Clinical and conventional radiological variables were also collected. After random allocation to training and validation cohorts, the Synthetic Minority Oversampling Technique (SMOTE) was applied to address class imbalance. Feature selection was performed using univariate analysis, correlation analysis, and Random Forest with five-fold cross-validation. A fusion model integrating clinical-radiological features, radiomics score (Rad-score), and Delta Rad-score was developed using a Random Forest algorithm. Model performance was assessed by area under the curve (AUC), calibration, and decision curve analysis. Shapley additive explanations (SHAP) were used for model interpretability. Prognostic value for recurrence-free survival (RFS) and overall survival (OS) was evaluated using Kaplan-Meier analysis and compared with traditional tumor-node-metastasis (TNM) staging. Results:Among 463 patients (median age 57 years; 156 males), 23 (5.0%) experienced early recurrence. The fusion model demonstrated excellent discrimination in the validation cohort (AUC: 0.881), significantly outperforming the clinical model (AUC: 0.725), with comparable discrimination to the radiomics model (AUC: 0.845) and delta-radiomics model (AUC: 0.827). Calibration was good (Brier score: 0.021), and decision curve analysis confirmed the highest net clinical benefit for the fusion model. SHAP analysis identified Rad-score and Delta Rad-score as the most important predictors, with higher values associated with increased recurrence risk. The fusion model stratified patients into high- and low-risk groups with significant differences in both RFS and OS (log-rank P<0.001 for both), whereas TNM staging (T1c vs. T1a/T1b) was associated with RFS (P=0.02) but not OS (P=0.32). Conclusions:The interpretable machine learning model integrating clinical-radiological features, radiomics, and delta-radiomics demonstrated promising predictive performance in internal validation of early postoperative recurrence in stage IA LUAD compared with traditional models and TNM staging. Delta-radiomics captures dynamic tumor evolution and emerges as a key prognostic biomarker. This model offers a non-invasive tool for individualized risk stratification to guide postoperative management.
Background:Some studies have investigated the prognostic value of tumor markers in small-cell lung cancer (SCLC), but results have been inconsistent, and the confounding effects across these markers remain unknown. The prognostic value of tumor markers in the era of immune, anti-angiogenesis, and targeted therapies has been less studied. This study aimed to analyze the relationship between serum tumor markers and overall survival (OS) in patients with treatment-naive SCLC. Methods:We retrospectively collected clinical data from untreated SCLC patients and those with less missing information who visited Tianjin Medical University General Hospital between June 2019 and October 2022. Serum concentrations of four tumor markers, carcinoembryonic antigen (CEA), neuron-specific enolase (NSE), a fragment of cytokeratin 19 (CYFRA 21-1), and pro-gastrin-releasing peptide (proGRP), which were measured by electrochemiluminescence assay, were extracted from their medical records. Missing data were imputed using the MICE package. OS information was obtained via telephone follow-up. Survival curve, Cox proportional hazards model, restricted cubic spline (RCS), and time-dependent receiver operating characteristic (ROC) curve were used to analyze the relationship between tumor markers and patients' OS. Results:A total of 160 patients were included in the study. CEA was positively correlated with CYFRA21-1, and proGRP was positively correlated with NSE and CYFRA21-1. In the multivariate Cox proportional hazards model, only CYFRA21-1 was independently associated with OS after adjusting for stage. RCS analysis showed a linear relationship between tumor markers and OS, with a hazard ratio (HR) around 1.5 at high marker concentrations. Time-dependent ROC curves showed that the area under the curve (AUC) for tumor markers in predicting OS was around 0.6. Conclusions:CEA, NSE, proGRP, and CYFRA21-1 are linearly associated with OS in patients with SCLC. Among the four investigated tumor markers, CYFRA21-1 shows the strongest association.
Background:The recent expansion of anaplastic lymphoma kinase (ALK)-targeted therapies into curative-intent settings is a major advance in the management of ALK-rearranged lung adenocarcinomas. However, real-world data on the prevalence and postoperative outcome reflecting this shift are needed to establish clinical baselines. This study aimed to characterize the prevalence, recurrence patterns, and prognostic factors of ALK-positive lung cancer in a modern surgical cohort. Methods:We conducted a single-center, retrospective cohort study of 5,783 patients with pathological stage I-III non-small cell lung cancer (NSCLC) who underwent curative-intent resection. Among the 3,288 patients with resected lung adenocarcinoma who underwent ALK testing, ALK prevalence and postoperative outcomes, including disease-free survival (DFS) and recurrence patterns, were evaluated. A multivariate Cox regression analysis was used to identify factors associated with recurrence. Results:ALK rearrangements were identified in 150 of 3,288 patients (4.6%). The median age was 56 years; 60.7% were never-smokers and 56.0% were female. Most patients had stage I disease (58.0%) and 30.0% had stage III disease. At a median follow-up of 45.7 months, the median DFS was not reached, and the 1-, 3-, and 5-year DFS rates were 93.3%, 77.6%, and 65.7%, respectively. N2 nodal involvement was the strongest independent predictor of recurrence [hazard ratio (HR), 14.64; 95% confidence interval (CI): 4.59-46.76; P<0.001]. Distant recurrence was predominant (81.1%), with the brain being the most frequent site (29.7%). Overall survival (OS) data were immature, with one death (0.7%) observed during follow-up. Conclusions:ALK rearrangements represent a clinically meaningful subset of resectable lung adenocarcinomas, supporting routine molecular testing even in early-stage disease. Recurrence risk is primarily driven by nodal burden, particularly N2 disease, and is characterized by frequent distant and intracranial recurrences. These findings define the postoperative risk landscape of resected ALK-positive lung cancer and support CNS-active perioperative ALK-targeted strategies, particularly in patients with high-risk nodal disease.
Background:Pneumothorax is a common and clinically significant complication during computed tomography (CT)-guided preoperative lung nodule localization. This study aimed to evaluate the clinical outcomes of a novel strategy-combining extrapleural local fluid application with gravity-dependent needle insertion-for reducing the risk of pneumothorax in an observational setting. Methods:A total of 166 patients who underwent preoperative CT-guided percutaneous pulmonary nodule (PN) localization and completed thoracoscopic surgery at The First Affiliated Hospital of Kunming Medical University between June 1, 2024, and November 1, 2024, were included in this study. Patients were grouped based on whether they received the CT-guided injection of 2% lidocaine outside the parietal pleura and the area of the localization needle insertion point. Baseline information, nodule-related details, localization-related data, and complications were collected and compared between groups and subgroups. Results:The pneumothorax incidence was analyzed among the four groups: FD (fluid application + gravity-dependent area) group: 1/33 case (3%) < FnD (fluid application + non-dependent area) group: 4/44 cases (9%) < nFD (non-fluid application + gravity-dependent area) group: 6/36 cases (17%) < nFnD (non-fluid application + non-dependent area) group: 19/37 cases (51%), with statistically significant differences (P<0.001). Multivariate analysis indicated that lidocaine injection beneath the pleura [P<0.001, odds ratio (OR) =0.07, 95% confidence interval (CI): 0.02-0.24] and needle insertion in gravity-dependent areas (P<0.001, OR=0.16, 95% CI: 0.05-0.48) were independent risk factors for pneumothorax. In the F group (FD + FnD), 53% of patients reported no pain (score 0), and 47% experienced mild pain (score 1-3). In the nF group (nFD + nFnD), 91% of patients experienced mild pain, and 9% reported moderate pain (score 4-6), with P<0.001. Complication-related analysis showed a positive correlation between pneumothorax and pain, as well as between pain and pleural reactions, with statistically significant differences. Conclusions:Prior to percutaneous CT-guided lung nodule localization, the precise administration of 10 mL of lidocaine external to the parietal pleura, coupled with the selection of an appropriate needle insertion site within the gravity-dependent region, can reduce the incidence of pneumothorax during CT-guided localization of PNs. The use of lidocaine outside the parietal pleura can reduce the pain felt by patients during the localization process, increase comfort, and enhance the embodiment of humanistic care.
Background:Sodium-glucose cotransporter-2 inhibitors (SGLT2i) provide cardiovascular and renal benefits in non-transplant populations, but their safety in lung transplant recipients remains poorly defined, particularly during the peri-transplant period. We evaluated early postoperative complications, long-term allograft outcomes, and adverse events associated with SGLT2i exposure before transplantation or within the first 90 days after lung transplantation (LTx). Methods:We performed a single-center retrospective cohort study of 443 adult recipients who underwent primary isolated LTx at Northwestern University between January 2018 and May 2024. Patients were classified according to peri-transplant SGLT2i exposure, defined as either ongoing use at the time of transplantation or de novo initiation within 90 days after transplantation. Early outcomes included primary graft dysfunction (PGD) grade 3 at 72 hours and major postoperative complications. Long-term outcomes included chronic lung allograft dysfunction (CLAD) and overall survival. Time-to-event analyses for CLAD and survival used a postoperative day-90 landmark. Multivariable logistic and Cox regression analyses were performed. Results:Among 443 recipients, 11 (2.5%) were exposed to SGLT2i either before transplantation or within 90 days after transplantation; 7 were receiving SGLT2i at the time of transplant and 4 initiated therapy de novo within 90 days post-transplant. Pre-transplant SGLT2i use was not associated with increased early postoperative complications or PGD grade 3. Among the small number of exposed patients, no clear excess in early complications, CLAD, or death was observed. Among SGLT2i-exposed patients, 10 of 11 continued therapy without interruption. No cardiovascular deaths or genitourinary infections were observed during follow-up. One patient developed diabetic ketoacidosis (DKA) 27 months after transplantation, leading to permanent drug discontinuation. Conclusions:In this single-center cohort, peri-transplant SGLT2i exposure was uncommon but was not associated with worse early graft outcomes, CLAD, or survival among treated patients. These findings suggest a reassuring early safety and feasibility signal for carefully selected lung transplant recipients and support the need for larger multicenter studies.
Background:Drug-induced respiratory depression is a severe and potentially life-threatening adverse event (AE); however, its systematic risk profile based on real-world data remains incomplete. Using real-world data from the U.S. Food and Drug Administration (FDA) Adverse Event Reporting System (FAERS), this study systematically identified drugs associated with respiratory depression and generated reporting safety signals to provide exploratory evidence for optimizing clinical medication safety. Methods:The reporting odds ratio (ROR) was used to evaluate reports of drug-induced respiratory depression from the first quarter of 2015 to the third quarter of 2025. Single-factor, least absolute shrinkage and selection operator (LASSO), and multi-factor regression models were employed to analyze the association between various medications and respiratory depression events. Odds ratios (ORs) and 95% confidence intervals (CIs) were used to assess the associated risks. Results:A total of 869 drugs associated with respiratory depression were identified, with nervous system drugs being the predominant category. Multivariate analysis identified 16 of 17 LASSO-screened drugs and male patients under 41 years of age as significant safety signals for respiratory depression, exhibiting strong disproportionate reporting signals. The median duration of drug-related respiratory depression was 1 day [interquartile range (IQR), 23 days], with approximately 75% of AEs occurring within this 23-day period. Conclusions:These findings assist clinicians in the early identification of the risk of drug-related respiratory depression, underscoring the necessity for proactive assessment and management of this risk in both analgesia and disease treatment.
Background:Tuberculosis (TB) remains a major public health challenge, and nutritional deprivation continues to shape TB susceptibility, progression, and outcomes. However, the population-level TB burden associated with specific suboptimal dietary exposures has not been systematically quantified across countries, sociodemographic settings, and time. We estimated global, regional, and national TB mortality and disability-adjusted life years (DALYs) associated with six suboptimal dietary exposures from 1990 to 2021 using the Global Burden of Disease (GBD) Study 2021. Methods:We extracted GBD 2021 estimates of TB deaths, DALYs, age-standardized mortality rates (ASMRs), and age-standardized DALY rates (ASDRs) associated with diets high in processed meat, high in red meat, high in sugar-sweetened beverages, low in fruits, low in vegetables, and low in whole grains across 204 countries and territories. The GBD comparative risk assessment (CRA) framework integrates multisource population, dietary exposure, disease surveillance, vital registration, demographic, and published epidemiological data to estimate modelled population-attributable burden. Rates were calculated using population denominators and expressed per 100,000 persons. Temporal trends were assessed using estimated annual percentage changes (EAPCs). Additionally, we applied age-period-cohort (APC) models to isolate the independent contributions of age, calendar period, and birth cohort to both TB mortality and DALY rates associated with suboptimal dietary exposures. Results:In 2021, diet low in whole grains and diet low in fruits were the two largest contributors to diet-associated TB DALYs, accounting for 177,304 DALYs and 164,027 DALYs, respectively. Smaller absolute burdens were associated with low vegetable intake, high red meat intake, high processed meat intake, and high sugar-sweetened beverage intake. From 1990 to 2021, all six exposures showed declining ASDRs and ASMRs, with ASDR EAPCs ranging from -0.94% for diet high in sugar-sweetened beverages to -2.94% for diet low in vegetables and ASMR EAPCs ranging from -1.33% to -3.24%. Globally, the ASDR of diet high in processed meat decreased from 1.92 [95% uncertainty interval (UI): 0.45 to 3.38] to 0.92 (95% UI: 0.22 to 1.58) per 100,000 persons, and the ASMR decreased from 0.06 (95% UI: 0.01 to 0.11) to 0.03 (95% UI: 0.01 to 0.05) per 100,000 persons. Most exposures showed the highest ASDR and ASMR in low-sociodemographic index (SDI) regions and the lowest in high-SDI regions. Men and older adults had higher DALY and mortality rates across most exposures. APC analyses showed negative net drifts for both DALY and mortality rates across all exposures. Conclusions:The age-standardized TB burden associated with suboptimal dietary exposures declined globally from 1990 to 2021, but substantial absolute burden and socioeconomic disparities persisted. Population-level nutrition-sensitive strategies, including improved access to whole grains, fruits, and vegetables, nutrition support within TB care, and structural reforms addressing food insecurity and inequities in TB prevention and treatment, may complement biomedical TB control programs, particularly in low-SDI settings.
Background:Heart failure (HF) remains a leading cause of mortality and morbidity worldwide, posing substantial challenges for early and accurate diagnosis as well as personalized therapeutic management. With the rapid evolution of artificial intelligence (AI), substantial progress has been made in the clinical translation and interdisciplinary research of HF. AI offers unprecedented potential to improve the diagnostic accuracy and therapeutic efficacy of HF. However, there is a lack of systematic reviews and analyses in the current research landscape, hotspots, and development trends in this field. This study aimed to employ bibliometric methods to systematically clarify the research status, core research directions, and future prospects of AI applications in HF. Methods:Using the Web of Science Core Collection as the data source, we systematically retrieved literature on AI applications in HF published from 2005 to 2025 and conducted a comprehensive bibliometric analysis via VOSviewer, CiteSpace, and SCImago Graphica. Results:A total of 4,133 records were retrieved initially, among which 4,110 eligible publications were finally included after strict screening. The annual publication output in this field showed accelerated growth since 2019, reaching 1,067 articles in 2025. The United States (1,508 publications) and China (952 publications) ranked as the top two contributing countries. Frontiers in Cardiovascular Medicine was the most prolific journal, whereas Circulation recorded the highest citation frequency. High-frequency keywords included "heart failure, machine learning, AI, risk, and mortality". The mainstream research hotspots concentrated on machine learning algorithms, medical image analysis, and clinical feature extraction. Conclusions:Research on AI applications in HF has undergone rapid development and established a comprehensive research framework covering disease diagnosis, prognosis assessment, and mechanism investigations. Future research priorities should incorporate more multicenter data for external validation, as well as promoting the construction and sharing of multicenter, multimodal datasets. These steps will further enhance the clinical applicability and credibility of AI-driven models in HF management.
Acute respiratory distress syndrome (ARDS) remains one of the most severe forms of acute lung injury, characterized by diffuse alveolar damage and refractory hypoxemia secondary to non-cardiogenic pulmonary edema. After many years of research and incremental improvements in ventilatory and supportive strategies, mortality rates are still disappointingly high and no pharmacological agent has convincingly demonstrated a mortality benefit in large trials. Elucidating the rapid molecular processes driving the acute phase of lung injury, particularly post-translational modifications (PTMs), may be critical for devising targeted therapies and potentially lowering mortality. In this review, we provide a comprehensive synthesis of the multifaceted roles of PTMs in ARDS pathogenesis, bridging molecular mechanisms to clinical relevance. We begin by examining canonical modifications, including phosphorylation and ubiquitination, which serve as swift molecular switches coordinating the cytokine storm, endothelial barrier disruption, and defective alveolar fluid clearance. We then discuss emerging PTMs, including citrullination, lactylation, and succinylation, and highlight their contributions to neutrophil extracellular trap (NET) formation and inflammatory amplification. We examine how metabolic reprogramming of the ARDS lung directly governs PTM enzyme activity and substrate availability, thereby acting as an upstream regulatory layer linking cellular metabolism to PTM dynamics. We further discuss how PTMs do not operate independently but engage in obligatory sequential and competitive crosstalk. Finally, we explore the therapeutic promise of modulating specific PTM-regulating enzymes alongside precision medicine approaches, underscoring the value of multi-omics integration in surmounting current translational barriers against this deadly syndrome.
Background:Extracorporeal membrane oxygenation (ECMO) is a crucial intervention for patients with refractory cardiogenic shock, though its management is complex and carries substantial risks such as thromboembolism, hemorrhage, and infection. This study aims to provide a comprehensive overview of global research trends and to assess the clinical utility of established prognostic tools. Methods:A literature search was conducted on the Web of Science Core Collection from 1988 to 2024. The bibliometric tools VOSviewer 1.6.20, CiteSpace 6.3.R1, and the R 4.3.3 were employed to visualize collaboration networks, keyword co-occurrences, and emerging research trends. Additionally, a retrospective cohort of adult patients with refractory cardiogenic shock who received veno-arterial extracorporeal membrane oxygenation (VA-ECMO) was identified from the Medical Information Mart for Intensive Care IV (MIMIC-IV) (v2.2) database. The Survival After Veno-Arterial ECMO (SAVE) score's discrimination and calibration were externally validated in this cohort. Results:The bibliometric analysis showed that the USA led in publications [129], followed by France [96] and China [72], with Assistance Publique Hôpitaux Paris (APHP) and Institut National de la Santé et de la Recherche Médicale (INSERM) among key institutions. Major journals included ASAIO Journal, Perfusion-UK, and Artificial Organs, and leading authors were Alain Combes, Pascal Leprince, and Guillaume Lebreton. Since 2020, research hotspots have shifted toward "ventricular assist device", "guidelines", and "hospital cardiac arrest". In the MIMIC-IV validation cohort (n=101), the SAVE score demonstrated limited discrimination [area under the receiver operating characteristic curve (AUC-ROC) 0.578] and significant miscalibration for predicting in-hospital mortality among VA-ECMO patients. Decision curve analysis indicated minimal net clinical benefit, underscoring the importance of external validation and model adaptation for real-world practice. Conclusions:This study underscores the performance variations of established prognostic tools across different clinical settings, suggesting that tools like the SAVE score require cautious interpretation when applied to localized cohorts. Future ECMO research in cardiogenic shock must refine management strategies, standardize guidelines, and improve prediction models through rigorous multi-center external validation. Leveraging large clinical databases such as MIMIC-IV will be essential for advancing precision medicine and optimizing outcomes in this high-risk population.
Background:Acute kidney injury (AKI) is a common and serious complication after cardiac surgery, significantly impacting patient outcomes and healthcare systems. This study aimed to develop and validate a nomogram for predicting postoperative AKI in adults undergoing elective cardiac valve and coronary artery bypass graft surgery. Methods:A clinical prediction study was conducted. The primary outcome was the occurrence of postoperative AKI. Potential predictors analyzed included demographic characteristics, comorbidities, preoperative laboratory values, anticoagulant medication usage, and intraoperative factors (transfusion volume, transfusion incidence, cardiopulmonary bypass duration, surgery type, perioperative bleeding). Multivariable logistic regression was used to identify independent predictors in a training cohort, and a nomogram was constructed. Model performance was assessed using the area under the receiver operating characteristic (ROC) curve (AUC) and validated in an independent cohort. Results:Multivariable analysis identified older age [odds ratio (OR) =1.03; 95% confidence interval (CI): 1.02-1.04], preoperative hemoglobin (OR =0.98; 95% CI: 0.98-0.98), creatinine (OR =1.03; 95% CI: 1.02-1.03), higher intraoperative red blood cell transfusion volume (OR =1.07; 95% CI: 1.02-1.12), and increased perioperative bleeding (OR =1.00; 95% CI: 1.00-1.00) as independent predictors of AKI. This predictive nomogram demonstrated good discriminatory ability, with an AUC of 0.880 (95% CI: 0.867-0.894) in the training cohort and an AUC of 0.883 (95% CI: 0.863-0.904) in the internal validation cohort, and an AUC of 0.690 (95% CI: 0.671-0.709) in the external validation set. Conclusions:A nomogram incorporating five readily available clinical variables effectively predicts the risk of postoperative AKI in adults undergoing elective cardiac valve and bypass surgery. This tool may assist in preoperative risk stratification and guide perioperative management strategies.
Background:Non-small cell lung cancer (NSCLC) remains the leading cause of cancer-related mortality worldwide, and accurate survival prediction is essential for individualized treatment planning. Existing prognostic models often rely on single algorithms with limited capacity to capture the complexity inherent in clinical data. This study aimed to develop, validate, and interpret an attention-gated stacking ensemble model for predicting 1-, 3-, 5-year overall survival (OS) in NSCLC patients. Methods:A retrospective cohort of 67,400 NSCLC patients from the Surveillance, Epidemiology, and End Results (SEER) database (2004-2021) was analyzed. A dual-filtering strategy combining Pearson and Spearman correlation analysis identified 11 prognostic features. An attention-gated stacked ensemble model was constructed, consisting of eleven diverse base learners plus a feature attention-gated mechanism, which ultimately generated survival probability estimate of 1, 3, and 5 years. Model performance was assessed using area under the receiver operating characteristic curve (AUC), confusion matrices, calibration curves with Brier scores, and decision curve analysis (DCA). An independent external validation cohort of 623 patients from Xinqiao Hospital (2015-2023) was used to evaluate generalizability. SHapley Additive exPlanations (SHAP) analysis was employed for model explanation. A web-based tool was developed for clinical application. Results:On the internal test set, the attention-gated stacking ensemble model achieved AUC values of 0.799, 0.799, and 0.778 for 1-, 3-, and 5-year OS prediction, respectively, outperforming all eleven individual base learners and a conventional stacking model without the attention gate. On external validation, AUC values reached 0.817, 0.807, and 0.825, with Brier scores of 0.127, 0.107, and 0.095, demonstrating robust cross-population generalizability. SHAP analysis identified metastasis (M) stage, tumor size, and lymph node (N) stage as the three most influential predictors, and the learned attention gate weights exhibited concordant importance rankings. A publicly accessible web-based calculator was deployed to facilitate personalized survival estimation at the point of care. Conclusions:The proposed attention-gate stacking ensemble model provides accurate and explainable survival predictions for NSCLC patients, outperforming conventional single-algorithm and standard stacking approaches. The integration of a feature-wise attention gate mechanism enhances both predictive performance and clinical transparency, supporting its potential utility in personalized oncological decision-making. The deployed web-based clinical decision support tool further bridges the gap between algorithmic development and bedside application.
Background:Coronary artery disease is a leading global cause of death, and acute coronary syndrome (ACS) is its most critical acute form. Percutaneous coronary intervention (PCI) is the standard reperfusion treatment for this condition, but no‑reflow occurs in 30-40% of emergency PCI cases, and this phenomenon is associated with higher mortality and poor prognosis. Coagulation, cardiac, and thyroid function markers, reflecting thrombus burden, myocardial injury, and vascular dysfunction, are closely correlated with no‑reflow, but single biomarkers have limited predictive value. This study aimed to screen independent no-reflow risk factors, build and validate a multivariate model for clinical risk assessment and prevention in ACS patients. Methods:This single‑center retrospective study enrolled a total of 136 patients with ACS admitted to The Second Hospital of Tianjin Medical University in 2024 as the development cohort. Patients were divided into a normal blood flow group (110 cases) and a no-reflow group (26 cases). Preoperative levels of D-dimer, fibrin monomer (FM), von Willebrand factor (VWF), N-terminal pro-B-type natriuretic peptide (NT-proBNP), cardiac troponin I (cTnI) and thyroid function indices were measured. We analyzed intergroup differences and constructed a multivariate logistic regression model to screen independent risk factors for no-reflow. Receiver operating characteristic (ROC) curves were used to evaluate the predictive performance of both individual indicators and the established model, while bootstrap resampling combined with calibration analysis was adopted for internal validation. Results:The proportion of hypertension and the levels of D-dimer, FM, VWF, cTnI, and NT-proBNP were higher in the no-reflow group compared with the normal reflow group (P<0.05), while FT3 concentration was lower (P<0.05). Multivariate logistic regression showed that high levels of D-dimer, FM, VWF, cTnI, and NT-proBNP, along with a low level of FT3, were independent risk factors for no-reflow (P<0.05). ROC curve analysis indicated that the prediction model (incorporating all six independent risk factors) had a diagnostic value for no-reflow. Individual indicators also demonstrated good predictive performance. Conclusions:The preliminary prediction model based on preoperative D-dimer, FM, VWF, cTnI, NT-proBNP, and FT3 levels shows promising predictive performance for no‑reflow, but these findings are exploratory. It should not yet be considered a definitive clinical tool, but may inform future risk‑stratification research.
Background:While artificial intelligence (AI) offers unprecedented capabilities for predictive modeling and precision asthma management, there is an urgent clinical necessity to successfully translate these rapid algorithmic innovations into real-world respiratory care. The exponential growth of cross-disciplinary AI literature has paradoxically created information overload for clinicians, obscuring underlying translational friction and hindering evidence-based implementation. Consequently, bibliometric analysis serves as the optimal quantitative vehicle to decode this vast scientific architecture. This study aims to objectively map the evolutionary trajectory, global research landscape, and emerging hotspots of AI in asthma, providing actionable roadmaps to reconcile computational development with clinical practice. Methods:A comprehensive literature search was conducted utilizing the Web of Science Core Collection database for studies related to AI in asthma, with the retrieval timeframe updated from January 1, 2016, to June 4, 2026. Following strict inclusion and exclusion criteria (restricted to English-language original articles and peer-reviewed reviews), a finalized dataset of 1,967 publications was extracted. Raw metadata parameters, including citations and bibliographic information, were exported for network topology analysis. Data synthesis was executed using specific algorithmic parameters in CiteSpace (for structural centrality and citation burst detection), VOSviewer (for co-authorship and keyword clustering), and the Bibliometrix R-package (for thematic evolution mapping). Results:The field has experienced robust exponential growth (22.24% per annum), with a pivotal inflection point in 2019. Geographically, a dual-centric geopolitical landscape exists between the United States and China in publication volume; yet, the United Kingdom serves as the ultimate global hub with superior network centrality, alongside highly integrated networks from nations like Australia and France. Institutional analysis highlights a productive yet fragmented landscape driven by prolific powerhouses such as Harvard Medical School and Imperial College London, while high-centrality nodes like the University of Zurich and Johns Hopkins University cross-link clinical and algorithmic clusters. Thematically, the field has undergone a distinct three-epoch technological trajectory: from early statistical clustering [2016-2019] to machine learning-based electronic health record mining [2020-2023], and recently to advanced deep learning architectures and multi-modal integration [2024-2026]. Current research frontiers focus on multi-omics precision phenotyping, real-time exacerbation prediction via wearables, and causal inference for personalized therapy. Conclusions:While AI paradigms in asthma research have rapidly advanced, current literature is fundamentally constrained by profound translational friction stemming from an over-reliance on retrospective datasets, which introduces critical structural biases and limits clinical generalizability. To effectively translate algorithms into clinical utility, future research must urgently deploy federated learning frameworks to securely overcome global data silos, and prioritize prospective, multicenter pragmatic trials to validate AI-driven predictive interventions in real-world respiratory care.
Background:Tuberculosis (TB) is one of the leading causes of death worldwide. The United Nations' Sustainable Development Goals include a target to reduce TB burden inequalities among countries. However, the current extent of TB burden inequality among countries remains unclear. This study aims to analyze the global burden inequality of TB. Methods:This study utilized the Global Burden of Disease (GBD) 2021 database to extract TB-related data from 204 countries and territories, including age-standardized rates (ASRs) of disability-adjusted life years (DALYs) and population from 1990 to 2021 and annual DALY rates across different age groups. By conducting statistical analyses, we calculated the slope index of inequality (SII) and the Gini index for each country and territory to assess the inequalities in the burden of TB. Results:The global SII reached an estimated 3,107.34×10-5 in 2021, a decrease of 56.32% from 1990. In contrast, the Gini index exhibited an upward trend, rising from 0.57 in 1990 to 0.60 in 2021. At the regional level, the burden of TB showed significant disparities. The SII and Gini index generally showed a pattern of first declining and then increasing with age. Conclusions:While absolute inequalities in TB burden have declined, relative inequalities persist as a significant concern, especially in specific regions and age groups. Future research should investigate the underlying drivers of these inequalities to better guide TB control strategies.
