This study aims to develop cancer-associated venous thromboembolism (CA-VTE) risk prediction models using survival machine learning (ML) algorithms.

This study employed a double-cohort study design (retrospective and prospective). The retrospective cohort (n ​= ​1036) was used as training set (70.0%, n ​= ​725) and internal validation set (30.0%, n ​= ​311); while the prospective cohort (n ​= ​321) was used as external validation set. Seven survival ML algorithms, including COX regression, classification, regression and survival tree, random survival forest, gradient boosting survival machine tree, extreme gradient boosting survival tree, survival support vector analysis, and survival artificial neural network, were applied to train CA-VTE models.

Univariate analysis and LASSO-COX regression both selected five predictors: age, previous VTE history, ICU/CCU, CCI, and D-dimer. The seven survival ML models (C-index: 0.709-0.760; Brier Score: 0.212-0.243) all outperformed Khorana Score (C-index: 0.632; Brier Score: 0.260) in external validation set. Among all models, the COX_DD model (COX regression ​+ ​D-dimer) performed best. However, ML models and Khorana Score predicted CA-VTE risk on ≥ 7 days of hospitalization with an increase in Brier Score ≥ 0.25, showing poor calibration.

In this study, the CA-VTE risk prediction models developed in seven survival ML algorithms outperformed Khorana Score. Combining with D-dimer can improve model performance. Applying the nomogram based on the optimal COX_DD model allows oncology nurse to reassess CA-VTE risk once a week. The prediction models developed using survival ML algorithms in this study may contribute to the dynamic and accurate risk assessment of CA-VTE for cancer survivors.

Previous studies have reported that renal function is associated brain structure and cognitive dysfunction. However, the association between renal function and memory-related disease was not well characterized.

Altogether, 5,282 individuals were included in this study based on China Longitudinal Study of Health and Retirement. Four estimated glomerular filtration rate indicators (eGFR), including CG, CKD-EPIscr, CKD-EPIscr-cys, and CKD-EPIcys were used to evaluate the association between renal function and memory-related disease.

The multivariable-adjusted HRs (95% CIs) of the memory-related disease in the low eGFR group (eGFR < 90 mL/min/1.73m2) were 1.56 (1.13-2.16) for CG, 1.56 (1.19-2.06) for CKD-EPIscr, 1.45 (1.06-1.99) for CKD-EPIscr-cys and 1.27 (0.91-1.77) for CKD-EPIcys, respectively. Similarly, each SD increase of eGFR was associated with reduced risk of memory-related disease on continuous analyses. Subgroup analyses further confirmed these associations. Moreover, the addition of eGFR to conventional risk factors improved the predictive power for memory-related disease (net reclassification improvement: 13.90% for CG, 19.83% for CKD-EPIscr and 30.65% for CKD-EPIscr-cys).

In conclusion, impaired renal function was associated with the increasing risk of memory-related disease, indicating that renal function may be a potential indicator for memory-related disease. Further studies from other races and populations are needed to replicate our findings and to clarify the potential mechanisms.

Predicting medium-term survival after admission is necessary for identifying end-of-life patients who may benefit from goals of care (GOC) discussions. Considering that several patients have multiple hospital admissions, this study leverages patients' longitudinal data and information collected routinely at admission to predict the Hospital One-year Mortality Risk.

We propose the Ensemble Longitudinal Network (ELN) to predict one-year mortality using patients' longitudinal records. The model was evaluated: (i) with only predictors reported upon admission (AdmDemo); and (ii) also with diagnoses available later during patients' stay (AdmDemoDx). Using records of 123,646 patients with 250,812 hospitalizations from 2011 to 2021, our dataset was split into a learning set (2011-2017) to compare models with and without longitudinal information using nested cross-validation, and a holdout set (2017-2021) to assess clinical utility towards GOC discussions.

The ELN achieved a significant increase in predictive performance using longitudinal information (p-value < 0.05) for both the AdmDemo and AdmDemoDx predictors. For randomly selected hospitalizations in the holdout set, the ELN showed: (i) AUROCs of 0.83 (AdmDemo) and 0.87 (AdmDemoDx); and (ii) superior decision-making properties, notably with an increase in precision from 0.25 for the standard process to 0.28 (AdmDemo) and 0.36 (AdmDemoDx). Feature importance analysis confirmed that the utility of the longitudinal information increases with the number of patient hospitalizations.

Integrating patients' longitudinal data provides better insights into the severity of illness and the overall patient condition, in particular when limited information is available during their stay.

Hemogram inflammatory markers, including neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), red-cell distribution width (RDW), and mean platelet volume (MPV) have been associated with type 2 diabetes mellitus (T2DM) and its complications, namely diabetic kidney diseases (DKD). We aimed to develop and validate logistic regression (LR) and CatBoost diagnostic models and study the role of adding these markers to the models.

All individuals who were managed in our secondary care center from March 2020 to December 2023 were identified. After excluding the ineligible patients, train-test splitting, and data preprocessing, two baseline LR and CatBoost-based models were developed using demographic, clinical, and laboratory features. The AUC-ROC of the models with biomarkers (NLR, PLR, RDW, and MPV) was compared to the baseline models. We calculated net reclassification improvement (NRI) and integrated discrimination index (IDI).

One thousand and eleven T2DM patients were eligible. The AUC-ROC of both LR (0.738) and CatBoost (0.715) models was comparable. Adding target inflammatory markers did not significantly change the AUC-ROC in both LR and CatBoost models. Adding RDW to the baseline LR model reclassified 41.7% of patients without DKD, in the cost of misclassification of 38.4% of DKD cases. This change was absent in CatBoost models, and other markers did not achieve improved NRI or IDI.

The basic models with demographical and clinical features had acceptable performance. Adding RDW to the basic LR model improved the reclassification of the non-DKD participants. However, adding other hematological indices did not significantly improve the LR and CatBoost models' performance.

The online version contains supplementary material available at 10.1007/s40200-024-01523-2.

A novel marker left atrioventricular coupling index (LACI) has been proved to be associated with cardiovascular events in patients without history of cardiovascular disease. However, the studies on cardiac magnetic resonance-derived LACI in hypertrophic cardiomyopathy (HCM) patients are limited, and the prognostic value of LACI has still not been studied thoroughly, so we aimed to explore the association between LACI and adverse clinical outcomes in HCM patients.

A total of 206 HCM patients underwent cardiac magnetic resonance examination were retrospectively enrolled. LACI is defined by the ratio between the left atrial (LA) volume and the left ventricular (LV) volume in LV end-diastolic phase. The composite endpoint was categorized into death-related, heart failure-related, and arrhythmia-related events, reflecting mortality risk, heart failure progression, and arrhythmia burden, respectively. Receiver operating characteristics curve analysis was used to determine the optimal cut-off value for LACI to distinguish HCM patients at high risk of adverse clinical outcome. Multivariable Cox regression models were built including significant clinical variables, LA ejection fraction (LAEF), LA volume index (LAVI), late gadolinium enhancement (LGE) extent and LACI. The improvement of discrimination by adding LACI to a clinical model was assessed using C-statistic, net reclassification improvement (NRI) and integrated discrimination improvement (IDI).

Thirty-four HCM patients reached the endpoint during a median follow-up time of 60 [interquartile range (50-68)] months. In the multivariate Cox regression analysis, LACI [hazard ratio 1.054, 95% confidence interval (CI): 1.037, 1.071; P < 0.001] was an independent predictor of the composite events after adjustment for age and atrial fibrillation. Then 40.09% was identified as an optimal cut-off for LACI in the risk stratification. Integrating LACI to the clinical model yielded higher C-statistic 0.892 with 95% CI (0.861, 0.922) compared with LA diameter, LAEF, LAVI and LGE extent, providing an improvement in prediction of high-risk patients (NRI = 0.627, 95% CI: 0.112-0.934; IDI = 0.295, 95% CI: 0.016-0.709).

LACI is an independent risk factor for clinical adverse outcome and is superior to conventional LA parameters and LGE extent for the identification of high-risk HCM patients.

We investigated the prognostic value of cardiac myosin-binding protein C (cMyC), a novel cardiospecific marker, both independently and in combination with N-terminal pro-B-type natriuretic peptide (NT-proBNP), for predicting 6-month all-cause mortality in patients without acute coronary syndrome (ACS) treated at medical (nonsurgical) cardiac intensive care units (CICUs). Admission levels of cMyC, high-sensitivity cardiac troponin T (hs-cTnT), and NT-proBNP were measured in 1032 consecutive patients (mean age; 70 years) without ACS hospitalized acutely in medical CICUs for the treatment of cardiovascular disease. Serum cMyC was closely correlated with hs-cTnT and moderately with NT-proBNP (r = 0.92 and r = 0.49, respectively, p < 0.0001). During the 6-month follow-up period after admission, there were 109 (10.6%) all-cause deaths, including 72 cardiovascular deaths. Both cMyC and NT-proBNP were independent predictors of 6-month all-cause mortality (all p < 0.05). Combining cMyC and NT-proBNP with a baseline model of established risk factors improved patient classification and discrimination beyond any single biomarker (all p < 0.05) or the baseline model alone (both p < 0.0001). Moreover, patients were divided into nine groups using cMyC and NT-proBNP tertiles, and the adjusted hazard ratio (95% confidence interval) for 6-month all-cause mortality in patients with both biomarkers in the highest vs. lowest tertile was 9.67 (2.65-35.2). When cMyC was replaced with hs-cTnT, similar results were observed for hs-cTnT. In addition, the C-indices for addition of cMyC or hs-cTnT to the baseline model were similar (0.798 vs. 0.800, p = 0.94). In conclusion, similar to hs-cTnT, cMyC at admission may be a potent, independent predictor of 6-month all-cause mortality in patients without ACS treated at medical CICUs, and their prognostic abilities may be comparable. Combining cMyC or hs-cTnT with NT-proBNP may substantially improve early risk stratification of this population.

The level of evidence (LOE) grading system for European Society of Cardiology (ESC) Clinical Practice Guidelines (CPG) classifies the quality of the evidence supporting a recommendation. However, the current taxonomy does not fully consider the optimal study design necessary to establish evidence for different types of recommendations in ESC guidelines. Therefore, two separate task forces of clinical and methodological experts were appointed by the CPG Committee, with the first tasked with updating the LOE grading system for therapy and prevention and the second responsible for developing a LOE grading system for diagnosis and prediction. This report from the second of these Task Forces develops a new system for diagnostic tests and prediction models which maintains the three-level grading structure to classify the quality of the evidence but introduces new definitions specific for diagnosis and prediction. For diagnostic tests, LOE A represents conclusive evidence of adequate diagnostic ability from at least two high-quality studies. Level of evidence B represents suggestive evidence from one high-quality or at least two moderate-quality studies. Level of evidence C represents preliminary evidence not classified as A or B, including evidence from less than two moderate-quality studies, or from expert consensus. For prediction models, LOE A represents conclusive evidence of adequate predictive ability from at least one high-quality derivation and two or more external validation studies of at least moderate quality. Level of evidence B represents suggestive evidence in one or more derivation studies and one or more external validation studies of at least moderate quality. Level of evidence C represents preliminary evidence not classified as A or B, including evidence from a derivation study of at least moderate quality, but with low quality or no external validation, or a derivation study of low quality.

Atherogenicity indices have emerged as promising markers for cardiometabolic disorders, yet their relationship with prediabetes risk remains unclear. This study aimed to comprehensively evaluate the associations between six atherogenicity indices and prediabetes risk in a Chinese population, and explore the predictive value of these atherosclerotic parameters for prediabetes.

This retrospective cohort study included 97,151 participants from 32 healthcare centers across China, with a median follow-up of 2.99 (2.13, 3.95) years. Six atherogenicity indices were calculated: Castelli's Risk Index-I (CRI-I), Castelli's Risk Index-II (CRI-II), Atherogenic Index of Plasma (AIP), Atherogenic Index (AI), Lipoprotein Combine Index (LCI), and Cholesterol Index (CHOLINDEX). To address the natural relationships between the atherogenicity indices and risk of prediabetes, we applied Cox proportional hazards regression with cubic spline functions and smooth curve fitting, using a recursive algorithm to calculate inflection points. Machine learning approach (XGBoost and Boruta methods) to address the high collinearity among indices and assess their relative importance, combined with time-dependent ROC analysis to evaluate the predictive performance at 3-, 4-, and 5-year follow-up.

During follow-up, 11,199 participants developed prediabetes (incidence rate: 3.71 per 100 person-years). Significant nonlinear associations were observed between all atherogenicity indices and prediabetes risk. Through Z-score standardization of atherogenicity indices and comprehensive Cox proportional hazards regression and advanced machine learning techniques, we identified AIP as the most significant predictor of prediabetes [HR = 1.057 (95% CI 1.035-1.080, P < 0.0001)], with LCI emerging as a secondary important marker [HR = 1.020 (95% CI 1.002-1.038, P = 0.0267)]. Our innovative XGBoost and Boruta analysis uniquely validated these findings, providing robust evidence of AIP and LCI's critical role in prediabetes risk assessment. Time-dependent ROC analysis further validated these findings, with LCI and AIP demonstrating comparable discrimination, with overlapping AUC ranges of 0.5952-0.6082. Notably, the combined indices model achieved enhanced predictive performance (AUC: 0.6753) compared to individual indices, suggesting the potential benefit of using multiple atherogenicity indices for prediabetes risk prediction.

This study identifies statistically significant associations between atherogenicity indices and prediabetes risk, highlighting their nonlinear relationships and combined effects. While the predictive performance of these indices is modest (AUC 0.55-0.68), these findings may contribute to improved risk stratification when incorporated into comprehensive assessment strategies.

BackgroundOdor identification deficits predict Alzheimer's disease (AD) in epidemiological studies.ObjectiveTo compare the accuracy of a short odor identification test with a short cognitive screening test for the detection of dementia and cognitive impairment in elderly persons with cognitive concerns.MethodsThis was a cross-sectional study of 600 participants 65 years and older, without known mild cognitive impairment (MCI) or dementia, with cognitive concerns, attending primary care practices in New York City. The odor identification test was the Brief Smell Identification Test (BSIT). The comparator test was the Mini Mental Status Exam II (MMSE). Cognitive diagnoses were made using the National Alzheimer's Coordinating Center Uniform Data set (NACC-UDS) version 3 forms with slight modifications. Test performance was compared using Receiver Operating Characteristic analyses.ResultsThe mean age was 72.65 ± 6.31 years, 73.3% were female, 63.3% were Hispanic, 13.5% non-Hispanic Black, and 20.8% non-Hispanic White; 23.5% were classified as normal cognition, 27.7% as cognitive impairment-not mild cognitive impairment (MCI), 31.2% as amnestic MCI, 5.7% as non-amnestic MCI, and 12% as dementia. The MMSE was superior to the BSIT in detecting dementia and any cognitive impairment. Combining abnormal scores in the BSIT (≤8) to MMSE (≤24) improved the MMSE's specificity and positive predictive value (PPV) in detecting cognitive impairment.ConclusionsThe MMSE was superior to the BSIT in detecting dementia and cognitive impairment in primary care but using both tests improved specificity and PPV for identifying persons with subjective complaints needing further cognitive and biomarker evaluation.

Strategies to estimate risk of cancer therapy-related cardiac dysfunction (CTRCD) before initiating cardiotoxic cancer treatment are needed. We hypothesized that baseline ECG markers could identify patients at risk for CTRCD.

In this retrospective cohort study, 1278 female patients with stage I-III HER2 (human epidermal growth factor receptor 2)-positive breast cancer meeting the following inclusion criteria were included: baseline ECG with QRS <120 milliseconds, baseline echocardiogram, and ≥1 follow-up echocardiogram. Quantitative measurements of ECG waveform parameters were performed using MUSE (GE Healthcare). The primary outcome of interest was CTRCD at 1 year, defined by left ventricular ejection fraction decline (≥10% to <53% or ≥16% from baseline), or clinical heart failure (New York Heart Association class III/IV). Mean age was 51.7±11.1 years, 990 (77%) received anthracyclines, and all received HER2-targeted therapy. CTRCD occurred in 160 (13%) patients. In a multivariable Cox proportional hazards model adjusting for our previously published CTRCD risk score (composed of patient and treatment-specific factors), 4 ECG markers remained independently associated with CTRCD risk: QRS axis, R-wave duration (lead II), ST segment deviation (lead II), and Sokolow-Lyon voltage (all P<0.05). Compared with a model using only clinical CTRCD risk variables, addition of ECG parameters provided incremental value for predicting CTRCD risk (P<0.001, likelihood ratio test) with continuous net reclassification improvement of 34.9% and integrated discrimination improvement of 3.4%.

Baseline ECG variables are predictive of subsequent CTRCD and provide incremental value to established clinical risk factors for CTRCD risk classification.

With different infarct morphological characteristics included, the relationship between single subcortical infarction type and neurological deterioration (ND) remains unclear. Similarly, the diagnostic value of the known risk factors is also uncertain.

We conducted a prospective observational study at a tertiary teaching hospital affiliated with Fudan University, enrolling patients with anterior circulation single subcortical infarction within 24 hours of symptom onset from 2017 to 2018. Clinical data, magnetic resonance imaging infarct characteristics, and echocardiographic indices were analyzed using a multivariable logistic regression model to identify independent ND predictors. The receiver operating characteristic curve with multiple testing corrections was performed to assess the discriminatory abilities of different models and the calibration curve for the accuracy of the optimal model.

The study included 298 patients, with 80 (26.85%) experiencing ND. Multivariate analysis identified admission National Institutes of Health Stroke Scale score (odds ratio [OR], 1.197 [95% CI, 1.067-1.343], P=0.002), proximal single subcortical infarction (OR, 3.311 [95% CI, 1.608-6.817], P=0.001), maximal diameter on axial DWI (OR, 1.651 [95% CI, 1.042-2.617], P=0.033), and left ventricular fractional shortening (OR, 0.001 [95% CI, 0.000-0.282], P=0.021) as independent predictors of ND. The optimal model, including the independent predictors and parent artery disease, demonstrated improved discrimination (area under the curve=0.762) and good calibration (Hosmer-Lemeshow P=0.51). left ventricular fractional shortening contributed positively to this model's performance (net reclassification improvement: 24.8%, P=0.051; integrated discrimination index: 2.1%, P=0.018).

When considering different infarct morphological characteristics simultaneously, SSI type remains an independent predictor of ND in patients with anterior circulation SSI. Furthermore, our research indicated left ventricular fractional shortening as a novel predictor, which can improve the discriminative ability of the prediction model.

Cirrhotic cardiomyopathy (CCM) is an underrecognized risk factor for cardiac events in patients undergoing liver transplantation (LT). Blunted cardiac reserve (BCR) is an emerging indicator of CCM, although it has not been integrated into diagnostic guidelines. This study assesses post-transplant cardiac outcomes and mortality in patients with BCR compared to current CCM diagnostic guidelines, focusing on diastolic indices. Consecutive patients undergoing liver transplant assessment were included. Of 978 patients screened with dobutamine stress echocardiography between 2010-2023, 481 (58.0%) progressed to LT, with 183 (38.0%) meeting BCR criteria and 117 (24.3%) meeting existing CCM diagnostic criteria. Thirty (6.2%) patients suffered a 30-day major adverse cardiovascular event (MACE), and 92 patients (19.1%) died on long-term follow-up. Following multivariate regression analysis, BCR was the strongest independent risk factor for post-operative MACE (HR 2.57 (1.13-5.85), p=0.024), heart failure exacerbations (HR 6.93 (1.46-33.01), p=0.015), and 30-day mortality (HR 9.69 (1.04-92.33), p=0.049). Addition of BCR to the existing guidelines improved MACE prediction (HR 5.81 (1.71-19.76) vs 2.59 (1.15-5.87), p=0.006), with a net reclassification improvement index of 41.9% (p=0.004) compared to existing guidelines alone. These results support the integration of a cardiac reserve assessment into CCM diagnostic criteria, and use in risk stratification of patients undergoing LT.

To develop an explainable fusion model that combines clinical, radiomic, and habitat features to predict postoperative early recurrence in hepatocellular carcinoma (HCC).

The bicentric retrospective study included 370 patients with surgically confirmed early-stage HCC who underwent gadoxetic acid-enhanced MRI. The patients were stratified into a training cohort (n=296) and an external validation cohort (n=74). From the hepatobiliary phase images, habitat and radiomics features were extracted across the entire tumor and used to construct radiomics and habitat models. Additionally, a clinical model was established utilizing relevant clinical features. Subsequently, all previously mentioned features were merged to construct the fusion model (HabRad_FB). Diagnostic performance of these models was assessed and compared using the area under the receiver operating characteristic curve (AUC), net reclassification index (NRI), and integrated discrimination improvement (IDI). The fusion model was then interpreted using SHapley Additive exPlanations (SHAP) analysis.

Tumor recurrence was observed in 73 out of 370 patients (19.7%; 55.2±11.3 years; male=333). Among all study cohorts, the HabRad_FB model showed the highest AUC (0.820-0.959), outperforming the clinical (0.517-0.729), radiomics (0.707-0.815), and habitat (0.729-0.861) models. The HabRad_FB model also demonstrated significant improvement in IDI in the training cohort and NRI in the validation cohort. SHAP force plots provided valuable insights into the interpretation of HabRad_FB model's predictions for early recurrence.

The HabRad_FB, an explainable fusion model, aids clinicians in accurately and non-invasively predicting the early recurrence of HCC preoperatively. This model might provide great potential in prognostic prediction and clinical management.

This study aimed to investigate the association of platelet-to-high-density lipoprotein cholesterol ratio (PHR) and its cumulative exposure with cardiovascular disease (CVD) risk.

The investigation utilized data from the China Health and Retirement Longitudinal Study (CHARLS). Platelet-to-high-density lipoprotein cholesterol ratio was calculated as platelet count (×10⁹/L)/high-density lipoprotein cholesterol (mmol/L), and a cumulative platelet-to-high-density lipoprotein cholesterol ratio (Cumulative PHR) was derived for longitudinal assessment. Multivariable logistic regression models were used to evaluate the association between PHR, cumulative PHR, and CVD risk across three models with increasing adjustments for confounders. Restricted cubic splines (RCS) regressions were utilized to examine if there were non-linear relationships. Subgroup analyses were conducted to enhance the reliability of the study findings. Furthermore, predictive performance was assessed using concordance index (C-index), net reclassification improvement (NRI), and integrated discrimination improvement (IDI).

A total of 7,063 participants aged 45 and older were included, of whom 1,433 (20.29%) experienced a cardiovascular disease. Participants with CVD had higher PHR (167.93 vs. 156.84, P < 0.001) and Log PHR (5.12 vs. 5.06, P < 0.001) values compared to non-CVD participants. Multivariable logistic regression revealed that Log PHR was independently associated with CVD risk [Odds ratio (OR) per-unit: 1.30, 95% confidence interval (CI): 1.13-1.49, P < 0.001; OR per- standard deviation (SD): 1.13, 95% CI: 1.06-1.21, P < 0.001]. Log cumulative PHR showed similar associations (OR per-unit: 1.34, 95% CI: 1.05-1.71, P = 0.02). Participants in the highest quartile of Log PHR had a nearly 1.32-fold higher risk of CVD compared to the lowest quartile (OR: 1.32, 95% CI: 1.10-1.57, P = 0.002). Addition of Log PHR and Log cumulative PHR slightly improved predictive performance metrics of baseline model.

Both Log PHR and Log cumulative PHR are independently associated with increased CVD risk and slightly improved the predictive performance of the baseline risk model. Future research should focus on its clinical implementation and integration into existing risk assessment frameworks.

The association of disturbance in gut microbiota with hypertension (HTN) defined on three separate occasions among children and adolescents remains unclear. In this study, we aimed to compare the differences in gut microbiota composition and diversity between children with HTN and those with normal blood pressure (BP).

Data and stool samples were collected from the second follow-up of a childhood cardiovascular health cohort study in 2021. 16 S ribosomal RNA gene sequencing was conducted to determine the relative abundance of microbial taxa in 51 children aged 10-14 years with HTN and 51 children with normal BP.

Compared with children with normal BP, those with HTN had decreased gut microbiome diversity. At the genus level, after adjusting for the false discovery rate (FDR), the proportions of several gut microbiota such as Blautia (PFDR=0.042), Coprococcus (PFDR=0.042), Eubacterium_ventriosum_group (PFDR=0.027), Christensenellaceae_R-7_group (PFDR=0.027), and norank_f__Lachnospiraceae (PFDR=0.015) significantly decreased in children with HTN compared to those with normal BP. Receiver operating characteristic analysis, net reclassification improvement (NRI) and integrated discrimination improvement (IDI) were performed and showed that the genera norank_f__Lachnospiraceae and Dorea significantly enhanced the ability of body mass index to differentiate between children with HTN and those with normal BP (area under the receiver operating characteristic curve: 0.95, 95% confidence interval 0.91-0.99; NRI > 0; IDI = 0.12, P < 0.05). Phylogenetic Investigation of Communities by Reconstruction of Unobserved States showed that the mean proportions of cofactors and vitamins metabolism pathway and the glycan anabolism pathway were higher in children with HTN.

Disturbances in the abundance and diversity of gut microbiota may contribute to the development of HTN in children. Gut microbiota biomarkers may be of significant importance in the early identification and diagnosis of childhood HTN.

Not applicable.

Peripheral artery disease (PAD) and dyslipidemia are both independent predictors of cardiovascular disease, but the association between individual lipid species and subclinical PAD, assessed by ankle-brachial index (ABI), is lacking in large-scale longitudinal studies.

We used liquid chromatography-mass spectrometry to repeatedly measure 1,542 lipid species from 1,886 American Indian adults attending 2 clinical examinations (mean ~5 years apart) in the Strong Heart Family Study. We used generalized estimating equation models to identify baseline lipid species associated with change in ABI and the Cox frailty regression to examine whether lipids associated with change in ABI were also associated with incident coronary heart disease (CHD). We also examined the longitudinal association between change in lipid species and change in ABI and the cross-sectional association of individual lipids with ABI. All models were adjusted for age, sex, body mass index, smoking, alcohol use, hypertension, estimated glomerular filtration rate, diabetes, and lipid-lowering medication.

Baseline levels of 120 lipid species, including glycerophospholipids, glycerolipids, fatty acids, and sphingomyelins, were associated with change in ABI. Among these, higher baseline levels of 3 known lipids (phosphatidylinositol[16:0/20:4], triacylglycerol[48:2], triacylglycerol[55:1]) were associated with a lower risk of CHD (hazard ratios [95% CIs] ranged from 0.67 [0.46-0.99] to 0.76 [0.58-0.99]), while cholesterol was associated with a higher risk of CHD (hazard ratio [95% CI] = 1.37 [1.00-1.87]). Longitudinal changes in 32 lipids were significantly associated with change in ABI during 5-year follow-up. Plasma levels of glycerophospholipids, triacylglycerols, and glycosylceramides were significantly associated with ABI in the cross-sectional analysis.

Altered plasma lipidome is significantly associated with subclinical PAD in American Indians beyond traditional risk factors. If validated, the identified lipid species may serve as novel biomarkers for PAD in this high-risk but understudied population.

Quick Sequential Organ Failure Assessment (qSOFA) is a simple and easy tool for identifying patients with suspected infection, who are at a high risk of poor outcome. However, its predictive performance is still insufficient. The Eastern Cooperative Oncology Group Performance Status (ECOG-PS) score, a tool to evaluate physical function, has been recently reported to be useful in predicting the prognosis of patients with pneumonia. We aimed to evaluate the added value of ECOG-PS to qSOFA in predicting 30-day mortality in older patients admitted with suspected infections.

Between 2018 and 2019, we prospectively collected data from adults aged 65 years or older, admitted with suspected infection at two acute care hospitals. Predictive performance was compared between two logistic regression models: one using qSOFA score alone (qSOFA model) and the other in which ECOG-PS was added to qSOFA (extended model).

Of the 1536 enrolled patients, 135 (8.8%) died within 30 days. The area under the curve of the extended model was significantly higher than that of the qSOFA model (0.67 vs. 0.64, p = 0.008). When the risk groups were categorized as follows: low (<5%), intermediate (5%-10%), and high (≥10%), 5.0% of those who died and 2.1% of those who survived were correctly reclassified by the extended model with an overall categorized net reclassification improvement of 0.03 (95% confidence interval: -0.06 to 0.30).

Adding the ECOG-PS score could improve the performance of qSOFA in predicting mortality in older patients admitted with suspected infection.

Our study aimed at evaluating the association between plasma human cartilage glycoprotein-39 (YKL-40) and depressive symptoms at 3 months among acute ischemic stroke patients.

Plasma YKL-40 levels were measured in 619 patients with ischemic stroke who participated in the China Antihypertensive Trial in Acute Ischemic Stroke (CATIS). The patients' depressive symptoms at 3 months after stroke were assessed using the Hamilton Rating Scale for Depression (HRSD-24).

During the 3-month follow-up period, 242 (39.1 %) participants were classified as experiencing depressive symptoms. Patients in the highest quartile of YKL-40 had a 1.98-fold (95 %CI: 1.19-3.30, P for trend = 0.02) risk of depressive symptoms compared with those in the lowest quartile. Per 1-SD increase of logarithm-transformed YKL-40 was associated with a 32 % (95 % CI: 10 %-58 %) increased risk for the depressive symptoms. The multiple-adjusted spline regression model confirmed dose-response relationships between YKL-40 levels and depressive symptoms (P for linearity = 0.02). Adding YKL-40 to a model containing conventional risk factors significantly improved the discriminatory power (area under the receiver operating characteristic curve improved by 0.02, P = 0.04) and reclassification power for depressive symptoms (net reclassification improvement = 18.77 %, P = 0.02; integrated discrimination improvement = 1.30 %, P = 0.005).

Elevated YKL-40 levels might be a potential risk marker of depressive symptoms at 3 months among acute ischemic stroke patients.

URL: https://www.

gov; Unique identifier: NCT01840072.

To determine local OCT structural correlates of deep visual sensitivity defects (threshold of ≤10 decibels on microperimetry) in early atrophic age-related macular degeneration (AMD).

Prospective observational study.

Forty eyes from 40 participants, with at least incomplete retinal pigment epithelium (RPE) and outer retinal atrophy, or more advanced atrophic lesion(s).

Participants underwent ≥2 targeted, high-density microperimetry tests of atrophic lesions of interest in 1 eye, and high-density 3×3-mm volume scans of that region on a swept-source OCT angiography device, all at a single visit. Seven OCT-defined features of atrophy were manually annotated: hypertransmission, RPE attenuation/disruption, complete RPE loss, ellipsoid zone disruption, external limiting membrane (ELM) disruption, subsidence of the outer plexiform layer and inner nuclear layer, and/or hyporeflective wedge-shaped band, and outer nuclear layer (ONL) thickness.

Association between OCT-defined features of atrophy and presence of a deep visual sensitivity defect at a local, pointwise level.

All OCT-defined features of atrophy were individually associated with the presence of a deep visual sensitivity defect at a pointwise level in univariable mixed-effects logistic regression analyses (P < 0.001 for all). However, only hypertransmission, complete RPE loss, ELM disruption, and ONL thickness remained significantly and independently associated with deep visual sensitivity defects in a multivariable analysis (P ≤ 0.011). A prediction model incorporating these 4 OCT features (partial area under the curve [pAUC] at ≥90% specificity = 0.80) outperformed models using any single feature alone in predicting the presence of deep visual sensitivity defects (pAUC = 0.65 to 0.78, respectively; P ≥ 0.040).

The study identified hypertransmission, complete RPE loss, ELM disruption, and ONL thickness as key OCT-defined features of atrophy independently associated with deep visual sensitivity defects. These findings are important when considering anatomical outcome measures for evaluating interventions for early atrophic AMD that are most likely to capture beneficial treatment effects that will be accompanied by evidence of functional preservation if measured directly.

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Current treatment paradigms assume aortic regurgitation (AR) patients to be a homogenous population, but varied courses of disease progression and outcomes are observed clinically.

The aim of this study was to first use unsupervised machine learning to identify unique patient phenoclusters in AR, and subsequently evaluate their prognostic relevance.

Clinical and cardiac magnetic resonance (CMR) characterization of moderate or severe AR patients was performed across 4 U.S.

Data from 2 centers were used for derivation of phenoclusters and validation was performed in the other 2. The outcome was all-cause death. An unsupervised clustering pipeline, Partition Around Medoids, used 23 clinical and CMR variables to derive patient clusters independent of outcomes.

Included were 972 patients with mean age 62 ± 23.2 years, 754 (78%) male, 680 (70%) trileaflet valve, and 330 (34%) underwent valve surgery. Over a median follow-up of 2.58 years (Q1-Q3: 1.03-5.50 years), the overall mortality rate was 12%. Four clusters were derived: 1) a younger predominantly male phenotype with majority of bicuspid aortic valve and high extent of left ventricular (LV) remodeling (1% mortality); 2) older male patients with predominantly tricuspid valves and intermediate outcomes (10% mortality); 3) older predominantly male patients with the highest burden of comorbidities, LV scarring, and dysfunction (22% mortality); and 4) a phenotype of predominantly female patients with high mortality and relatively higher symptoms burden, relatively lower extent of LV remodeling, and rate of aortic valve replacement (20% mortality). The clustering algorithm was independently associated with survival after adjustment for time-dependent aortic valve replacement and traditional risk markers of prognosis in patients with AR (C statistic 0.77 vs 0.75; P = 0.009 in the validation cohort).

Unique patient phenoclusters of AR are described using a machine learning approach leveraging comprehensive CMR and clinical characterization. This approach may be an opportunity for a precision medicine approach to enhance risk stratification of patients with AR. Female patients with AR pose a unique phenotype with high mortality, which deserves greater attention.

Recent changes in the regulatory assessment of in vitro medical tests reflect a growing recognition of the need for more stringent clinical evidence requirements to protect patient safety and health. Under current regulations in the United States and Europe, when needed for regulatory approval, clinical performance reports must provide clinical evidence tailored to the intended purpose of the test and allow assessment of whether the test will achieve the intended clinical benefit. The quality of evidence must be proportionate to the risk for the patient and/or public health. These requirements now cover both commercial and laboratory developed tests (LDT) and demand a sound understanding of the fundamentals of clinical performance measures and study design to develop and appraise the study plan and interpret the study results. However, there is a lack of harmonized guidance for the laboratory profession, industry, regulatory agencies and notified bodies on how the clinical performance of tests should be measured. The Working Group on Test Evaluation (WG-TE) of the European Federation of Clinical Chemistry and Laboratory Medicine (EFLM) is a multidisciplinary group of laboratory professionals, clinical epidemiologists, health technology assessment experts, and representatives of the in vitro diagnostic (IVD) industry. This guidance paper aims to promote a shared understanding of the principles of clinical performance measures and study design. Measures of classification performance, also referred to as discrimination, such as sensitivity and specificity are firmly established as the primary measures for evaluating the clinical performance for screening and diagnostic tests. We explain these measures are just as relevant for other purposes of testing. We outline the importance of defining the most clinically meaningful classification of disease so the clinical benefits of testing can be explicitly inferred for those correctly classified, and harm for those incorrectly classified. We introduce the key principles and a checklist for formulating the research objective and study design to estimate clinical performance: (1) the purpose of a test e.g. diagnosis, screening, risk stratification, prognosis, prediction of treatment benefit, and corresponding research objective for assessing clinical performance; (2) the target condition for clinically meaningful classification; (3) clinical performance measures to assess whether the test is fit-for-purpose; and (4) study design types. Laboratory professionals, industry, and researchers can use this checklist to help identify relevant published studies and primary datasets, and to liaise with clinicians and methodologists when developing a study plan for evaluating clinical performance, where needed, to apply for regulatory approval.

Little is known about the role of electrocardiography (ECG) in the community population independent of physical and laboratory examinations. Thus, this study developed and validated several ECG-based models for cardiovascular disease (CVD) risk assessment, with or without simple questionnaire-based variables.

Using a derivation cohort of 3734 Chinese participants aged ≥40 years, we developed the ECG-based models to predict the risk of developing CVD (comprising fatal and non-fatal coronary heart disease, unstable angina, stroke, and heart failure). Candidate predictors associated with CVD were screened from hundreds of ECG characteristics using a hybrid algorithm. By incorporating the questionnaire-based predictors, we constructed the ECG-questionnaire model. All models were tested in an external validation cohort (n = 1224) to determine their discrimination and calibration. Over a maximum follow-up of 7 years, 433 CVD events occurred in the derivation cohort. The ECG model with 37 selected features achieved comparable performance concerning the clinical model using traditional cardiovascular risk factors (C-statistic: 0.690, 95% confidence interval [CI]: 0.638-0.743) in the external validation cohort. Such performance significantly improved when the questionnaire-based predictors were added (C-statistic: 0.734, 95% CI: 0.685-0.784; calibration χ2: 3.334, P = 0.950). Compared with the clinical model, 17.4% of the participants were correctly assigned to the corresponding risk groups, with an absolute integrated discrimination index of 0.048 (95% CI: 0.016-0.080).

The ECG model with/without questionnaire-based variables can accurately predict future CVD risk independent of physical and laboratory examinations, suggesting its great potential in routine clinical practice.

To assess whether muscle power (force times velocity) outperforms strength as a risk indicator and predictor of mortality.

Anthropometric, clinical and vital status, muscle power, and strength data were assessed in 3889 individuals aged 46 to 75 years (2636 [67.8%] men) who were participants in the CLINIMEX Exercise prospective cohort between February 13, 2001, and October 31, 2022. Study participants were stratified by sex and categorized into 4 groups according to the distribution of the results of relative muscle power and strength (adjusted for body weight) measured, respectively, by handgrip and upper row movement tests.

Death rates were 14.2% (373 of 2636) and 8.9% (111 of 1253) for men and women, respectively, during a median (IQR) follow-up of 10.8 years (6.7 to 15.5 years). In multivariable Cox proportional hazards regression analyses, the hazard ratios (95% CIs) for mortality comparing the lowest vs highest categories of relative muscle power were 5.88 (2.28 to 15.17; P<.001) and 6.90 (1.61 to 29.58; P=.009) for men and women, respectively. The corresponding hazard ratios (95% CIs) for relative strength were 1.62 (0.89 to 2.96; P=.11) and 1.71 (0.61 to 4.80; P=.31), respectively. Sex-specific results of risk prediction analyses revealed that improvements in C index provided by relative power over relative strength were 0.0110 (95% CI, 0.0039 to 0.0182) in men and 0.0112 (95% CI, -0.0040 to 0.0265) in women.

In this large prospective study, relative muscle power was a stronger predictor of mortality than relative strength in middle-aged and older men and women. Evaluating and training muscle power could be of clinical and practical relevance.

The role of inflammation in shaping death risk in diabetes is still unclear.

To study whether inflammation is associated with and helps predict mortality risk in patients with type 2 diabetes. To explore the intertwined link between inflammation and tryptophan metabolism on death risk.

There were 2 prospective cohorts: the aggregate Gargano Mortality Study (1731 individuals; 872 all-cause deaths) as the discovery sample, and the Foggia Mortality Study (490 individuals; 256 deaths) as validation sample. Twenty-seven inflammatory markers were measured. Causal mediation analysis and in vitro studies were carried out to explore the link between inflammatory markers and the kynurenine to tryptophan ratio (KTR) in shaping mortality risk.

Using multivariable stepwise Cox regression analysis, interleukin (IL)-4, IL-6, IL-8, IL-13, RANTES, and interferon gamma-induced protein-10 (IP-10) were independently associated with death. An inflammation score (I score) comprising these 6 molecules is strongly associated with death in both the discovery and the validation cohorts HR (95% CI) 2.13 (1.91-2.37) and 2.20 (1.79-2.72), respectively. The I score improved discrimination and reclassification measures (all P < .01) of 2 mortality prediction models based on clinical variables. The causal mediation analysis showed that 28% of the KTR effect on mortality was mediated by IP-10. Studies in cultured endothelial cells showed that 5-methoxy-tryptophan, an anti-inflammatory metabolite derived from tryptophan, reduces the expression of IP-10, thus providing a functional basis for the observed causal mediation.

Adding the I score to clinical prediction models may help identify individuals who are at greater risk of death. Deeply addressing the intertwined relationship between low-grade inflammation and imbalanced tryptophan metabolism in shaping mortality risk may help discover new therapies targeting patients characterized by these abnormalities.

Plasma volume status (PVS) is recognized as a marker of systemic congestion, but its clinical utility in patients with mitral regurgitation (MR) undergoing transcatheter edge-to-edge mitral valve repair (M-TEER) has not been well established. This study aimed to evaluate the prognostic significance of PVS in these patients.

Data from 3763 patients who underwent M-TEER were analysed from a Japanese multicentre registry. Patients were classified into functional MR (FMR) and degenerative MR (DMR) according to MR aetiology, and the median PVS values for each were calculated (FMR 12.7, DMR 14.4). The median value was used as the cut-off, stratifying the cohort into a high PVS group (n = 1882) and a low PVS group (n = 1881). All-cause mortality, cardiovascular death, and heart failure (HF) hospitalization between these two groups were compared up to 3 years in the overall, FMR, and DMR populations. The cumulative incidence rates of all-cause mortality, cardiovascular death, and HF hospitalization were higher in the high PVS group than in the low PVS group (47.0% vs. 22.2%, P < 0.001, 31.6% vs. 13.6%, P < 0.001, and 35.9% vs. 24.7%, P < 0.001, respectively). Similar trends in terms of all-cause mortality, cardiovascular death, and HF hospitalization were observed in the FMR and DMR cohorts (all P < 0.05). In the multivariate Cox regression analysis, the high PVS compared with the low PVS group was independently associated with the increased risk of all-cause death (hazard ratio [HR], 1.02; 95% confidence interval [CI], 1.01-1.03; P < 0.001), cardiovascular death (HR, 1.02; 95% CI, 1.01-1.03, P < 0.001) and HF hospitalization (HR, 1.02; 95% CI, 1.01-1.02, P < 0.001). An independent association between a high PVS and all-cause death, cardiovascular death, and HF hospitalization was also found in FMR and DMR sub-groups (all P < 0.05) while reducing MR severity to moderate or less after M-TEER was associated with improved outcomes in both the high and low PVS groups.

Preoperative PVS is a strong independent prognostic marker in patients undergoing M-TEER, correlating with increased risk of mortality and HF hospitalization. PVS may provide valuable clinical insights for patient stratification and management strategies in M-TEER patients.

The application of artificial intelligence (AI) in laboratory medicine will revolutionize predictive modeling using clinical laboratory information. Machine learning (ML), a sub-discipline of AI, involves fitting algorithms to datasets and is broadly used for data-driven predictive modeling in various disciplines. The majority of ML studies reported in systematic reviews lack key aspects of quality assurance. In clinical laboratory medicine, it is important to consider how differences in analytical methodologies, assay calibration, harmonization, pre-analytical errors, interferences, and physiological factors affecting measured analyte concentrations may also affect the downstream robustness and reliability of ML models. In this article, we address the need for quality improvement and proper validation of ML classification models, with the goal of bringing attention to key concepts pertinent to researchers, manuscript reviewers, and journal editors within the field of pathology and laboratory medicine. Several existing predictive modeling guidelines and recommendations can be readily adapted to the development of ML models in laboratory medicine. We summarize a basic overview of ML and key points from current guidelines including advantages and pitfalls of applied ML. In addition, we draw a parallel between validation of clinical assays and ML models in the context of current regulatory frameworks. The importance of classification performance metrics, model explainability, and data quality along with recommendations for strengthening journal submission requirements are also discussed. Although the focus of this article is on the application of ML in laboratory medicine, many of these concepts extend into other areas of medicine and biomedical science as well.

Over the past decade, coronary computed tomographic angiography (CCTA) has been the most robust non-invasive method for evaluating significant coronary stenosis. Thanks to new technologies, it is now possible to determine the fractional flow reserve (FFR) non-invasively using computed tomographic (CT) images. The aim of this work was to evaluate the incremental diagnostic value of CT-derived FFR for ischemia detection. In this retrospective monocentric study, we investigated 421 patients who underwent CCTA and subsequent ischemia testing between 04/2009 and 06/2020. Endpoint was ischemia on a coronary vessel level assessed by CMR (n = 20), SPECT (n = 225), invasive angiography (stenosis ≥ 90%; n = 80) or invasive FFR (positive if ≤ 0.8; n = 96). CT-FFR was derived from CCTA images by a machine learning (ML) based software prototype. Patients averaged 66.5 [58.2-73.6] years of age and 72.7% (n = 306) were male. Overall, 52.5% (n = 221) had hypertension and 67.9% (n = 286) had hypercholesteremia. Logistic regression analysis on a per vessel base showed that the diagnostic model with CT-FFR plus CCTA had significantly better-fit criteria than the diagnostic model with CCTA alone (log-likelihood χ2 230.21 vs. 192.17; p for difference < 0.001). In particular, the area under curve (AUC) by receiver operating characteristics curve (ROC) analysis for CT-FFR plus CCTA (0.87) demonstrated greater discrimination of hemodynamic ischemia compared to CCTA alone (0.83; p for difference < 0.0001). Combined CCTA and CT-FFR have improved diagnostic accuracy compared to CCTA alone in detecting ischemia on the coronary vessel level and thus could reduce the use of invasive coronary angiography in the future.

Kidney stone disease is a multifactorial disease with increasing incidence worldwide. Trans-ancestry GWAS has become a popular strategy to dissect genetic structure of complex traits. Here, we conduct a large trans-ancestry GWAS meta-analysis on kidney stone disease with 31,715 cases and 943,655 controls in European and East Asian populations. We identify 59 kidney stone disease susceptibility loci, including 13 novel loci and show similar effects across populations. Using fine-mapping, we detect 1612 variants at these loci, and pinpoint 25 causal signals with a posterior inclusion probability >0.5 among them. At a novel locus, we pinpoint TRIOBP gene and discuss its potential link to kidney stone disease. We show that a cross-population polygenic risk score, PRS-CSxEAS&EUR, exhibits superior predictive performance for kidney stone disease than other polygenic risk scores constructed in our study. Relative to individuals in the third quintile of PRS-CSxEAS&EUR, those in the lowest and highest quintiles exhibit distinct kidney stone disease risks with odds ratios of 0.57 (0.51-0.63) and 1.83 (1.68-1.98), respectively. Our results suggest that kidney stone disease patients with higher polygenic risk scores are younger at onset. In summary, our study advances the understanding of kidney stone disease genetic architecture and improves its genetic predictability.

Coronary radial wall strain (RWS) represents a novel approach enabling discrimination of vulnerable plaques with prognostic significance.

This study sought to evaluate the prognostic impact of RWS in diabetic patients with non-flow-limiting coronary stenosis when compared with optical coherence tomography-detected vulnerability features (OCT-VFs).

This was a post hoc analysis of the COMBINE OCT-FFR dataset. The primary endpoint was lesion-oriented composite endpoint (LOCE), a composite of cardiac death, target vessel-related myocardial infarction, and clinically driven target lesion revascularization.

RWS was assessed in 435 eligible non-flow-limiting lesions from 366 patients. Lesion-level maximal RWS was predictive of lesions with any OCT-VFs (area under the curve: 0.630 [95% CI: 0.571-0.688]; P < 0.001). The median follow-up was 3.2 years (Q1-Q3: 2.2-4.1 years). With a prespecified cutoff of ≥13.0%, the incidence of LOCE was 17.0% (15/88; 95% CI: 9.0%-25.1%) in RWS-positive vs 6.8% (19/278; 95% CI: 3.8%-9.8%) in RWS-negative patients (HR: 2.70; 95% CI: 1.37-5.32; P = 0.004). Positive RWS predicted LOCE independently from any OCT-VFs (direct effect β = 0.099 [95% CI: 0.029-0.168]; P = 0.006; indirect effect β = 0.004 [95% CI: -0.008 to 0.015]; P = 0.555; mediation proportion 3.9% [95% CI: -5.0% to 20.3%]). Adding RWS to any OCT-VFs mainly improved the reclassification for LOCE in the lower-risk strata (positive continuous net reclassification improvement [cNRI] -0.060 [95% CI: -0.420 to 0.318]; P = 0.749; negative cNRI 0.583 [95% CI: 0.474-0.681]; P < 0.001; integrated discrimination improvement 0.066 [95% CI: 0.013-0.182]; P = 0.010).

In diabetic patients with non-flow-limiting stenosis, RWS can help to localize stenoses with OCT-VFs. RWS predicts increased risk for LOCE, both independently from-and incrementally beyond-OCT-VFs. (Combined Optical Coherence Tomography Morphologic and Fractional Flow Reserve Hemodynamic Assessment of Non-Culprit Lesions to Better Predict Adverse Event Outcomes in Diabetes Mellitus Patients [COMBINE OCT-FFR]; NCT02989740).

The value of carotid ultrasound in real-world practice remains controversial. We investigated the outcomes of people with vascular risk factors according to an easy carotid-plaque burden scale (CPB-scale). Predictive yield of the addition CPB-scale to ESC-SCORE2 (CPB-SCORE2 table) was assessed.

A cohort of participants without preexisting cardiovascular disease (CVD) was evaluated for clinical outcomes according to the number of plaques by segment. The usefulness of the CPB-SCORE2 table was investigated.

A total of 1004 patients were followed for a mean of 12.5 years for major adverse cardiovascular events (MACEs) and death. The CPB-scale was independently associated with MACEs; compared to those in the low-risk group, the corresponding adjusted hazard ratios (95% confidence intervals) for MACEs among the intermediate and high-risk groups were 13.1 (4.87-35.5) and 19.4 (7.27-51.9), respectively. Similarly, the risk of death was greater for participants stratified as high-risk than for those in the low-risk group (adjusted HR 3.36 [1.58-7.15]). According to our CPB-SCORE2 table, 149 of 178 (84%) CV events were detected in the high-risk group and exhibited greater sensitivity than did the SCORE2 Table, 84%; vs. 62%; but slightly less specificity, 62%; vs. 68%. Our table shows the improved performance of SCORE2; c-statistics: 0.74 vs. 0.68; p<0.001 for net reclassification index and integrated discrimination index.

A simple prognostic CPB-scale was strongly associated with the long-term risk of developing a first MACE and all-cause death. Adding the CPB-scale to the SCORE2 may improve risk prediction with easy applicability in clinical practice.

Real-life binary classification problems often involve imbalanced datasets, where the majority class outnumbers the minority class. We previously developed the U-smile method, which comprises the U-smile plot and the BA, RB and I coefficients, to assess the usefulness of a new variable added to a reference prediction model and validated it under class balance. In this study, we evaluated the U-smile method under class imbalance, proposed a three-level approach of the U-smile method, and used the I coefficients as a weighting factor for point size in the U-smile plots of the BA and RB coefficients. Using real data from the Heart Disease dataset and generated random variables, we built logistic regression models to assess four new variables added to the reference model (nested setting). These models were evaluated at seven pre-defined imbalance levels of 1%, 10%, 30%, 50%, 70%, 90% and 99% of the event class. The results of the U-smile method were compared to those of certain traditional measures: Brier skill score, net reclassification index, difference in F1-score, difference in Matthews correlation coefficient, difference in the area under the receiver operating characteristic curve of the new and reference models, and the likelihood-ratio test. The reference model overfitted to the majority class at higher imbalance levels. The BA-RB-I coefficients of the U-smile method identified informative variables across the entire imbalance range. At higher imbalance levels, the U-smile method indicated both prediction improvement in the minority class (positive BA and I coefficients) and reduction in overfitting to the majority class (negative RB coefficients). The U-smile method outperformed traditional evaluation measures across most of the imbalance range. It proved highly effective in variable selection for imbalanced binary classification, making it a useful tool for real-life problems, where imbalanced datasets are prevalent.