Development Cohort Research Articles

A new preprocedural predictive risk model for post-endoscopic retrograde cholangiopancreatography pancreatitis: The SuPER model.

Post-endoscopic retrograde cholangiopancreatography (ERCP) pancreatitis (PEP) is a severe and deadly adverse event following ERCP. The ideal method for predicting PEP risk before ERCP has yet to be identified. We aimed to establish a simple PEP risk score model (SuPER model: Support for PEP Reduction) that can be applied before ERCP. This multicenter study enrolled 2074 patients who underwent ERCP. Among them, 1037 patients each were randomly assigned to the development and validation cohorts. In the development cohort, the risk score model for predicting PEP was established via logistic regression analysis. In the validation cohort, the performance of the model was assessed. In the development cohort, five PEP risk factors that could be identified before ERCP were extracted and assigned weights according to their respective regression coefficients: -2 points for pancreatic calcification, 1 point for female sex, and 2 points for intraductal papillary mucinous neoplasm, a native papilla of Vater, or the pancreatic duct procedures (treated as 'planned pancreatic duct procedures' for calculating the score before ERCP). The PEP occurrence rate was 0% among low-risk patients (≤0 points), 5.5% among moderate-risk patients (1-3 points), and 20.2% among high-risk patients (4-7 points). In the validation cohort, the C statistic of the risk score model was 0.71 (95% CI 0.64-0.78), which was considered acceptable. The PEP risk classification (low, moderate, and high) was a significant predictive factor for PEP that was independent of intraprocedural PEP risk factors (precut sphincterotomy and inadvertent pancreatic duct cannulation) (OR 4.2, 95% CI 2.8-6.3; p<0.01). The PEP risk score allows an estimation of the risk of PEP prior to ERCP, regardless of whether the patient has undergone pancreatic duct procedures. This simple risk model, consisting of only five items, may aid in predicting and explaining the risk of PEP before ERCP and in preventing PEP by allowing selection of the appropriate expert endoscopist and useful PEP prophylaxes. No external funding was received for this work.

Development and validation of a score for clinical deterioration in patients with cerebral venous thrombosis.

Patients with cerebral venous thrombosis (CVT) may experience poor response to anticoagulant therapy and delayed surgical treatment may lead to clinical deterioration. However, the factors contributing to clinical deterioration remain poorly understood. Patients with CVT from three centers between January 2017 and October 2023 were included and grouped as the development cohort and validation cohort. The danger triangle was defined as the posterior two-thirds of the superior sagittal sinus, confluence of sinuses, straight sinus, and deep venous system. The primary endpoint was clinical deterioration, characterized by new or progressive bleeding or infarctions or worsened neurological conditions post-admission. Using the results of multivariable logistic analysis, the Cerebral venOus thrombosis DEterioration (CODE) score was developed within the development cohort and validated within the validation cohort. The score' performance in predicting clinical deterioration was evaluated using the area under the receiver operating characteristic curve (AUC). The development cohort included 194 CVT patients (101 males, and the median age was 38 years). clinical deterioration occurred in 45 (23.2%) patients. Multivariate logistic analysis defined D-dimer > 1.5mg/L, Glasgow coma scale ≤ 8, intracerebral hemorrhage, danger triangle as risk factors of clinical deterioration. The CODE score integrating these factors performed well to predict patients suffering from clinical deterioration within the validation cohort (n = 79, 11 deteriorations) with an AUC of 0.83 (95%CI, 0.71-0.96) and an accuracy of 88.6% (95%CI, 81.4-95.8%). CODE score could evaluate the risk of clinical deterioration in CVT patients and may serve as a useful tool for decision-making for CVT. Clinical Deterioration in Cerebral Venous Thrombosis: A Predictive Study (CVT deterioration cohort), ClinicalTrials.gov number: NCT06266585, link: https://classic.clinicaltrials.gov/ct2/show/NCT06266585 ).

Development of a disease diagnostic model to predict the occurrence of central precocious puberty of female.

To develop a clinical model for predicting the occurrence of Central Precocious Puberty based on the breast development outcomes in chinese girls. This is a retrospective study, which included a total of 1,001 girls aged 6-9 years old who visited the outpatient clinic of Beijing Children's Hospital from January 2017 to October 2022 for "breast development". Participants were categorized into pubertal development (PD) cohort and simple premature breast development (PT) according to the criteria, and information was collected and tested for relevant indicators. After dealing with missing data, logistic regression, LASSO regression and random forest were used to screen the variables, and support vector machine models were built with SMOTE oversampling and ten-fold cross-validation to assess the effectiveness of the models in the training and validation sets. 1,001 girls were included in the analysis, of whom 369 (36.9 %) were diagnosed with PD and 632 (63.1 %) with PT. Body mass index (BMI), bone age (BA), luteinizing hormone (LH), follicle stimulating hormone (FSH), estradiol (E2), uterine diameter, and ovary volume were identified as the final predictor variables by three variable screening methods. The AUC of the constructed disease diagnostic model was 0.9457 in the developmental cohort and 0.8357 inthe external validation group, and sensitivity analyses revealed that the performance of the constructed models with different variable selection strategies was similar. A disease diagnostic model was developed that may help predict a girl's risk of diagnosing central precocious puberty.

Personalized Dynamic Prediction Model for Biopsy Timing in Patients With Prostate Cancer During Active Surveillance

Active surveillance (AS) for patients with prostate cancer (PC) often includes fixed repeat prostate biopsies that do not account for the varying risk of reclassification to significant disease. Given the invasive nature and potential complications of biopsies, a personalized approach is needed to balance the burden of biopsies with the risk of missing disease progression. To develop and externally validate a dynamic model that predicts an individual's risk of PC reclassification during AS. This prognostic study developed a dynamic prediction model using data from the Prostate Cancer Research International: Active Surveillance (PRIAS) study, which was initiated in 2006. Follow-up was truncated until April 2023. External validation was conducted using cohorts from the world's largest centralized AS database, the Global Action Plan Prostate Cancer Active Surveillance initiative database. The PRIAS study is a multicenter, prospective, web-based cohort study monitoring patients undergoing AS, involving more than 175 academic, nonacademic, and private centers across 23 countries worldwide. For the development and external validation of the model, all patients diagnosed with Grade Group 1 PC who underwent at least 1 baseline or follow-up magnetic resonance imaging (MRI) and 1 follow-up biopsy were included. Data were analyzed from September 2023 to January 2024. AS, including prostate-specific antigen (PSA) tests, MRI, and prostate biopsies according to a fixed follow-up schedule. A joint model for longitudinal and time-to-event data was used to predict reclassification to Grade Group 2 or greater on repeat biopsy using predefined baseline and repeated clinical characteristics. Performance was assessed using time-dependent area under the receiver operating characteristic curve and negative predictive value. The development cohort included 2512 patients (median [IQR] age, 65 [59-69] years). Characteristics significantly associated with a higher risk of reclassification were increased age, higher PSA and velocity, lower prostate volume, a suspicious lesion on MRI, and no previous negative biopsy findings. Depending on the threshold and time point used, the model demonstrated a negative predictive value of 86% to 97%. External validation included 3199 patients from 9 other cohorts. The time-dependent area under the curve ranged from 0.81 to 0.84 in the development cohort and 0.52 to 0.90 at external validation. In this prognostic study, the developed dynamic risk model effectively identified patients at low risk of PC reclassification during AS. After prospective validation, this model may support personalized, risk-based AS and reduce the burden of unnecessary biopsies.

A Supervised Explainable Machine Learning Model for Perioperative Neurocognitive Disorder in Liver-Transplantation Patients and External Validation on the Medical Information Mart for Intensive Care IV Database: Retrospective Study.

Patients undergoing liver transplantation (LT) are at risk of perioperative neurocognitive dysfunction (PND), which significantly affects the patients' prognosis. This study used machine learning (ML) algorithms with an aim to extract critical predictors and develop an ML model to predict PND among LT recipients. In this retrospective study, data from 958 patients who underwent LT between January 2015 and January 2020 were extracted from the Third Affiliated Hospital of Sun Yat-sen University. Six ML algorithms were used to predict post-LT PND, and model performance was evaluated using area under the receiver operating curve (AUC), accuracy, sensitivity, specificity, and F1-scores. The best-performing model was additionally validated using a temporal external dataset including 309 LT cases from February 2020 to August 2022, and an independent external dataset extracted from the Medical Information Mart for Intensive Care Ⅳ (MIMIC-Ⅳ) database including 325 patients. In the development cohort, 201 out of 751 (33.5%) patients were diagnosed with PND. The logistic regression model achieved the highest AUC (0.799) in the internal validation set, with comparable AUC in the temporal external (0.826) and MIMIC-Ⅳ validation sets (0.72). The top 3 features contributing to post-LT PND diagnosis were the preoperative overt hepatic encephalopathy, platelet level, and postoperative sequential organ failure assessment score, as revealed by the Shapley additive explanations method. A real-time logistic regression model-based online predictor of post-LT PND was developed, providing a highly interoperable tool for use across medical institutions to support early risk stratification and decision making for the LT recipients.

Predicting the risk of hematoma expansion in acute intracerebral hemorrhage: the GIVE score

BackgroundNumerous noncontrast computed tomography (NCCT) markers have been reported and validated as effective predictors of hematoma expansion (HE). Our objective was to develop and validate a score based on NCCT markers and clinical characteristics to predict risk of HE in acute intracerebral hemorrhage (ICH) patients.MethodsWe prospectively collected spontaneous ICH patients at the First Affiliated Hospital of Chongqing Medical University to form the development cohort (n = 395) and at the Third Affiliated Hospital of Chongqing Medical University to establish the validation cohort (n = 139). We adopted a revised HE definition, incorporating the standard definition of HE (> 6 mL or > 33%) and intraventricular hemorrhage (IVH) expansion (any new IVH or IVH expansion ≥ 1 ml). The predictive score was formulated based on the parameter estimates derived from the multivariable logistic regression analysis.ResultThe Glasgow Coma Scale, island sign, ventricular hemorrhage and time elapsed from onset to NCCT scan (GIVE) score was created as a total of individual points (0–6) based on Glasgow Coma Scale (2 points for ≤ 11), island sign (1 point for presence), ventricular hemorrhage (1 point for presence), and time elapsed from onset to NCCT scan (2 points for ≤ 2.5 h). The c statistic was 0.72(95% confidence interval [CI], 0.66–0.78) and 0.73(95% CI, 0.63–0.82) in the development and validation cohorts, respectively.ConclusionA six-point scoring algorithm has been developed and validated to assess the risk of HE in patients with ICH. This scoring system facilitates the rapid and accurate identification of patients at increased risk for HE.

Derivation and validation of a clinical predictive model for longer duration diarrhea among pediatric patients in Kenya using machine learning algorithms

BackgroundDespite the adverse health outcomes associated with longer duration diarrhea (LDD), there are currently no clinical decision tools for timely identification and better management of children with increased risk. This study utilizes machine learning (ML) to derive and validate a predictive model for LDD among children presenting with diarrhea to health facilities.MethodsLDD was defined as a diarrhea episode lasting ≥ 7 days. We used 7 ML algorithms to build prognostic models for the prediction of LDD among children < 5 years using de-identified data from Vaccine Impact on Diarrhea in Africa study (N = 1,482) in model development and data from Enterics for Global Health Shigella study (N = 682) in temporal validation of the champion model. Features included demographic, medical history and clinical examination data collected at enrolment in both studies. We conducted split-sampling and employed K-fold cross-validation with over-sampling technique in the model development. Moreover, critical predictors of LDD and their impact on prediction were obtained using an explainable model agnostic approach. The champion model was determined based on the area under the curve (AUC) metric. Model calibrations were assessed using Brier, Spiegelhalter’s z-test and its accompanying p-value.ResultsThere was a significant difference in prevalence of LDD between the development and temporal validation cohorts (478 [32.3%] vs 69 [10.1%]; p < 0.001). The following variables were associated with LDD in decreasing order: pre-enrolment diarrhea days (55.1%), modified Vesikari score(18.2%), age group (10.7%), vomit days (8.8%), respiratory rate (6.5%), vomiting (6.4%), vomit frequency (6.2%), rotavirus vaccination (6.1%), skin pinch (2.4%) and stool frequency (2.4%). While all models showed good prediction capability, the random forest model achieved the best performance (AUC [95% Confidence Interval]: 83.0 [78.6–87.5] and 71.0 [62.5–79.4]) on the development and temporal validation datasets, respectively. While the random forest model showed slight deviations from perfect calibration, these deviations were not statistically significant (Brier score = 0.17, Spiegelhalter p-value = 0.219).ConclusionsOur study suggests ML derived algorithms could be used to rapidly identify children at increased risk of LDD. Integrating ML derived models into clinical decision-making may allow clinicians to target these children with closer observation and enhanced management.

Development and validation of a prognostic nomogram forsevere postpartum hemorrhage after cesarean delivery: Atwo-center retrospective study.

To develop and validate a nomogram to predict severe postpartum hemorrhage following cesarean delivery. This is a two-center retrospective cohort study. Cesarean delivery patients from the First Affiliate Hospital of Jinan University were divided into a development cohort (n = 11 137) and an internal validation cohort (n = 4739). Cesarean delivery patients from the Dongguan Maternal and Child Health Care Hospital (n = 13 775) were enrolled in the external validation cohort. The nomogram was based on independent risk factors for severe postpartum hemorrhage obtained by multivariate logistic regression. We evaluated the discrimination and calibration of the nomogram in the development and validation cohorts. The nomogram used data including previous cesarean delivery, pre-pregnancy weight, preterm birth, placenta previa, placenta accreta spectrum disorders, placental abruption, and mode of anesthesia. The area under the curves of the nomogram in the internal and external validation cohorts were 0.922 (95% confidence interval [CI] 0.897-0.947) and 0.813 (95% CI 0.785-0.841), respectively. Consistency between the predicted and actual probabilities was observed in both validation cohorts. The nomogram displayed good calibration and discrimination and can be used for screening in clinical practice to enable clinicians to intervene appropriately.

Preliminary Development and Validation of Automated Nociception Recognition Using Computer Vision in Perioperative Patients.

Effective pain recognition and treatment in perioperative environments reduce length of stay and decrease risk of delirium and chronic pain. We sought to develop and validate preliminary computer vision-based approaches for nociception detection in hospitalized patients. Prospective observational cohort study using red-green-blue camera detection of perioperative patients. Adults (≥18 years) admitted for surgical procedures to the San Francisco Veterans Affairs Medical Center (SFVAMC) were included across 2 study phases: (1) algorithm development phase and (2) internal validation phase. Continuous recordings occurred perioperatively across any postoperative setting. We inputted facial images into convolutional neural networks using a pretrained backbone, to detect (1) critical care pain observation tool (CPOT) and (2) numerical rating scale (NRS). Outcomes were binary pain/no-pain. We performed external validation for CPOT and NRS classification on data from University of Northern British Columbia-McMaster University (UNBC) and Delaware Pain Database. Perturbation models were used for explainability. We included 130 patients for development, 77 patients for validation cohort and 25 patients from UNBC and 229 patients from Delaware datasets for external validation. Model area under the curve of the receiver operating characteristic for CPOT models were 0.71 (95% confidence interval [CI] 0.70, 0.74) on the development cohort, 0.91 (95% CI 0.90, 0.92) on the SFVAMC validation cohort, 0.91 (0.89, 0.93) on UNBC and 0.80 (95% CI 0.75, 0.85) on Delaware. NRS model had lower performance (AUC 0.58 [95% CI 0.55, 0.61]). Brier scores improved following calibration across multiple different techniques. Perturbation models for CPOT models revealed eyebrows, nose, lips, and foreheads were most important for model prediction. Automated nociception detection using computer vision alone is feasible but requires additional testing and validation given small datasets used. Future multicenter observational studies are required to better understand the potential for automated continuous assessments for nociception detection in hospitalized patients.

Prediction of mortality risk in patients with severe community-acquired pneumonia in the intensive care unit using machine learning

The aim of this study was to develop and validate a machine learning-based mortality risk prediction model for patients with severe community-acquired pneumonia (SCAP) in the intensive care unit (ICU). We collected data from two centers as the development and external validation cohorts. Variables were screened using the Recursive Feature Elimination method. Five machine learning algorithms were used to build predictive models. Models were evaluated through nested cross-validation to select the best one. The model was interpreted using Shapley Additive Explanations. We selected the optimal model to generate the web calculator. A total of 23 predictive features were selected. The Light Gradient Boosting Machine (LightGBM) model had an area under the receiver operating characteristic curve (AUC) of 0.842 (95% CI: 0.757–0.927), with an external 5-fold cross-validation average AUC of 0.842 ± 0.038, which was superior to the other models. External validation results also demonstrated good performance by the LightGBM model with an AUC of 0.856 (95% CI: 0.792–0.921). Based on this, we generated a web calculator by combining five high importance predictive factors. The LightGBM model was confirmed to be efficient and stable in predicting the mortality risk of patients with SCAP admitted to the ICU. The web calculator based on the LightGBM model can provide clinicians with a prognostic evaluation tool.

Development and external validation of a machine learning model for brain injury in pediatric patients on extracorporeal membrane oxygenation

BackgroundPatients supported by extracorporeal membrane oxygenation (ECMO) are at a high risk of brain injury, contributing to significant morbidity and mortality. This study aimed to employ machine learning (ML) techniques to predict brain injury in pediatric patients ECMO and identify key variables for future research.MethodsData from pediatric patients undergoing ECMO were collected from the Chinese Society of Extracorporeal Life Support (CSECLS) registry database and local hospitals. Ten ML methods, including random forest, support vector machine, decision tree classifier, gradient boosting machine, extreme gradient boosting, light gradient boosting machine, Naive Bayes, neural networks, a generalized linear model, and AdaBoost, were employed to develop and validate the optimal predictive model based on accuracy and area under the curve (AUC). Patients were divided into retrospective cohort for model development and internal validation, and one cohort for external validation.ResultsA total of 1,633 patients supported by ECMO were included in the model development, of whom 181 experienced brain injury. In the external validation cohort, 30 of the 154 patients experienced brain injury. Fifteen features were selected for the model construction. Among the ML models tested, the random forest model achieved the best performance, with an AUC of 0.912 for internal validation and 0.807 for external validation.ConclusionThe Random Forest model based on machine learning demonstrates high accuracy and robustness in predicting brain injury in pediatric patients supported by ECMO, with strong generalization capabilities and promising clinical applicability.

Macroelements Intake in Chinese Adults - 10 PLADs, China, 2022-2023.

The Chinese Nutrition and Health Surveillance in 2015-2017 reported that among adults aged 18-59 years, the average daily intake was 328.3 mg of calcium, 251.8 mg of magnesium, 5,681.4 mg of sodium, and 1,474.1 mg of potassium. For adults aged 60 years and above, the corresponding intakes were 333.2 mg, 242.5 mg, 5,412.1 mg, and 1,392.6 mg per day, respectively. Based on the China Development and Nutrition Health Impact Cohort Survey (2022-2023) across 10 provincial-level administrative divisions (PLADs), the median daily intakes among Chinese adults were calcium (288.3 mg), magnesium (228.9 mg), sodium (4,182.4 mg), and potassium (1,397.6 mg), with a median Na/K ratio of 3.0. The study revealed that 96.1% of adults had calcium intake below the estimated average requirement (EAR), 64.4% had magnesium intake below the EAR, 81.9% had potassium intake below the adequate intake (AI), and 89.4% had sodium intake above the PI-NCD. Urban residents demonstrated generally better macroelement intake patterns compared to rural residents. There is an urgent need for effective policies and interventions to enhance diet quality through increased consumption of dairy products, fruits, and vegetables, thereby improving calcium, magnesium, and potassium intake while reducing sodium intake and the sodium-to-potassium ratio among Chinese adults. Priority should be given to improving the nutritional status of rural residents and reducing urban-rural disparities.

Development, validation, and prognostic evaluation of LiverPRO for the prediction of significant liver fibrosis in primary care: a prospective cohort study.

Clinically significant liver fibrosis is associated with future adverse events in patients with steatotic liver disease. We designed a software tool for detection of clinically significant liver fibrosis in primary care. In this prospective cohort study, we developed and validated LiverPRO using six independent cohorts from Denmark, Germany, and England that included patients from primary and secondary care with steatotic liver disease related to alcohol or metabolic dysfunction. We used clinically significant fibrosis (histology stage ≥F2) and advanced fibrosis (≥F3) as outcomes for variable selection in the development cohort and built the model with fractional polynomial regression. In all cohorts, we independently validated the tool for prediction of elevated liver stiffness by transient elastography (≥8 kPa and ≥12 kPa) and for the 2-year and 5-year risk of liver-related events. Diagnostic performance was assessed using the area under the receiver operating curve (AUC), with clinical performance evaluated through sensitivity, specificity, and Harrell's C-statistic for prognostic purposes. In the development cohort (n=462), we derived 466 multivariable models consisting of age in combination with three to nine variables from a list of nine blood tests (aspartate aminotransferase, alkaline phosphatase, gamma-glutamyl transferase, international normalised ratio, albumin, sodium, bilirubin, platelet count, and cholesterol). In the development cohort, LiverPRO diagnosed clinically significant fibrosis with good accuracy (transient elastography ≥8 kPa area under the receiver operating characteristic curve [AUC] 0·86 [95% CI 0·83-0·90]). In the DECIDE validation cohort (n=6468), LiverPRO detected participants with a transient elastography of 8 kPa or higher with good accuracy (AUC 0·80 [95% CI 0·78-0·82]), comparable to enhanced liver fibrosis testing (0·78 [0·75-0·80]) and the LiverRisk score (0·81 [0·79-0·84]), but superior to the Fibrosis-4 index (0·69 [0·66-0·72]) and NAFLD Fibrosis Score (0·74 [0·72-0·77]). Findings were consistent in three other validation cohorts (n=2554), albeit accuracy was slightly lower. Using a rule-out cutoff of less than 25% (indicating no further examinations required), LiverPRO had a rule-out sensitivity of 80·6% (95% CI 76·4-84·3) and a rule-out negative predictive value of 98·0% (95% CI 97·5-98·4) in the DECIDE cohort. Similarly, with a rule-out cutoff of less than 1·3, FIB-4 had a rule-out sensitivity of 53·8% (48·5-58·9) and a rule-out negative predictive value of 95·8% (95·1-96·4). For rule-in thresholds, using a cutoff of more than 65% (indicating referral to a hepatologist required) LiverPRO had a rule-in specificity of 95·5% (95% CI 94·9-96·0) and a rule-in positive predictive value of 33·0% (95% CI 28·5-37·8) in the DECIDE cohort whereas FIB-4, with a rule-in threshold of 2·67, had a rule-in specificity of 98·7% (94·9-96·0) and a rule-in positive predictive value 35·6% (27·0-44·9). Using UK Biobank data, LiverPRO predicted liver-related events with a C-statistic of 0·80 (0·77-0·84) at 2 years. LiverPRO reliably identifies clinically significant liver fibrosis and elevated liver stiffness, predicts the risk of liver-related events in primary care, and is adaptable to the availability of different liver blood test analytes. On the basis of these results LiverPRO was certified according to IVDR class b, obtaining European CE approval in 2024. EU Horizon 2020 research and innovation programme and Novo Nordisk Foundation.

Development and external validation of the SMA2SH2ERS risk prediction model for aneurysmal subarachnoid haemorrhage in the general population: a population-based prospective cohort study

ObjectivesAneurysmal subarachnoid haemorrhage (ASAH) is a severe stroke type, preventable by screening for intracranial aneurysms followed by treatment in high-risk individuals. We aimed to develop and validate a risk prediction...

A Simple Nomogram for Predicting the Development of ARDS in Postoperative Patients with Gastrointestinal Perforation: A Single-Center Retrospective Study.

Acute respiratory distress syndrome (ARDS) is a severe form of organ dysfunction and a common postoperative complication. This study aims to develop a predictive model for ARDS in postoperative patients with gastrointestinal perforation to facilitate early detection and effective prevention. In this single-center retrospective study, clinical data were collected from postoperative patients with gastrointestinal perforation admitted to the ICU in Hebei Provincial People's Hospital from October 2017 to May 2024. Univariate analysis and multifactorial logistic regression analysis were used to determine the independent risk factors for developing ARDS. Nomograms were developed to show predictive models, and the discrimination, calibration, and clinical usefulness of the models were assessed using the C-index, calibration plots, and decision curve analysis (DCA). Two hundred patients were ultimately included for analysis. In the development cohort, 38 (27.1%) of 140 patients developed ARDS, and in the internal validation cohort, 13 (21.7%) of 60 patients developed ARDS. The multivariate logistic regression analysis revealed the site of perforation (OR = 0.164, P = 0.006), the duration of surgery (OR = 0.986, P = 0.008), BMI (OR = 1.197, P = 0.015), SOFA (OR = 1.443, P = 0.001), lactate (OR = 1.500, P = 0.017), and albumin (OR = 0.889, P = 0.007) as the independent risk factors for ARDS development. The area under the curve (AUC) was 0.921 (95% CI: 0.869, 0.973) for the development cohort and 0.894 (95% CI: 0.809, 0.978) for the validation cohort. The calibration curve and decision curve analysis (DCA) demonstrate that the nomogram possesses good predictive value and clinical practicability. Our research introduced a nomogram that integrates six independent risk factors, facilitating the precise prediction of ARDS risk in postoperative patients following gastrointestinal perforation.

Prenatal exposure to phthalates and phthalate replacements in relation to chorionic plate surface vasculature at delivery.

Pregnant people are ubiquitously exposed to endocrine-disrupting phthalates through consumer products and food. The placenta may be particularly vulnerable to the adverse effects of phthalates, with evidence from animal models suggesting impacts on placental development and vascularization. We translate this research to humans, examining gestational exposure to phthalates and phthalate replacements in relation to novel markers of chorionic plate surface vascularization. Phthalate and phthalate replacement metabolites were measured in first trimester urine from pregnant participants in the Understanding Pregnancy Signals and Infant Development (UPSIDE) cohort (n=154). At delivery, placentae underwent specialized 2D and 3D digital imaging to quantify chorionic plate surface vasculature. Using weighted quantile g-computation mixtures methods as well as multivariable linear regression models examining individual metabolites, we evaluated associations with overall chorionic plate surface area and five chorionic plate surface vascular measures, adjusting for covariates. We additionally examined interactions with placental sex. Exposure to a phthalate mixture was associated with longer total arterial arc length (β=9.64cm; 95%CI: 1.68, 17.59), shorter mean arterial arc length (β=-0.07cm; 95%CI: -0.14, -0.01), and more arterial branch points (β=5.77; 95%CI: 1.56, 9.98), but not chorionic plate surface area. In models considering individual metabolites and their molar sums, results were strongest for the metabolites of Di-isobutyl phthalate (DiBP), Di-isononyl phthalate (DiNP), and Di(2-ethylhexyl) phthalate (DEHP). Associations with metabolites of phthalate replacements tended to be in the same direction but weaker. Few sex differences were observed. Gestational phthalate exposure may be associated with alterations in placental chorionic plate surface vasculature characterized by more branching and shorter segments. These alterations may have implications for placental perfusion and suggest a placental mechanism by which phthalates may impact fetal development.

US-Guided Thermal Ablation for Secondary Hyperparathyroidism: A Prospective Multicenter Study.

Background Interest in microwave ablation (MWA) and radiofrequency ablation (RFA) use for treating secondary hyperparathyroidism (SHPT) is rising; however, ablation outcomes in patients with SHPT are not well characterized. Purpose To assess the response of parathyroid hormone (PTH), calcium, phosphorus, and alkaline phosphatase (ALP) levels to US-guided parathyroid MWA and RFA and the safety of these treatments in participants with SHPT. Materials and Methods This prospective multicenter cohort study, conducted from September 2017 to March 2022, included participants with SHPT. The primary end point was the proportion of participants achieving the target PTH level (≤585 pg/mL). The secondary end points included PTH, calcium, phosphorus, and ALP levels before ablation and time points for follow-up assessments after ablation (2 hours, 1 day, 1 month, 3 months, and 6 months, and then every 6 months) and complications and technical success rates. Mixed-effects logistic regression models were used to identify factors associated with treatment failure. Results A total of 215 participants (median age, 53 years [IQR, 43-60 years]; 109 [50.7%] male participants) were evaluated, and 183 (85.1%) achieved target PTH levels. Compared with baseline levels, there was an 85.9%, 6.3%, 15.3%, and 37.4% reduction in PTH, calcium, phosphorus, and ALP levels at 24 months after ablation, respectively. For major complications, one (0.5%) participant experienced persistent hoarseness, and severe hypocalcemia (<1.87 mmol/L) was present in 74 (34.4%) participants. After adjustments, predictors associated with treatment failure included the preablation PTH level (adjusted odds ratio [OR], 3.78; 95% CI: 1.19, 12.04; P = .03), maximum tumor volume (adjusted OR, 5.02; 95% CI: 1.74, 14.53; P = .003), and number of glands ablated (adjusted OR, 0.32; 95% CI: 0.11, 0.98; P = .046). The prediction model showed good discrimination ability in the development and validation cohorts (area under the receiver operating characteristic curve, 0.78 [95% CI: 0.66, 0.90] and 0.73 [95% CI: 0.55, 0.91], respectively). Conclusion US-guided thermal ablation techniques were effective and safe treatments in participants with SHPT because they effectively reduced PTH, calcium, phosphorus, and ALP levels. © RSNA, 2025 Supplemental material is available for this article. See also the editorial by Gemmete in this issue.

Development and Validation of a Comprehensive Predictive Model for Surgical Planning in Patients with Renal Cell Carcinoma and Inferior Vena Cava Tumor Thrombus

ObjectiveTo investigate preoperative predictors for selecting different surgical approaches in patients with renal cell carcinoma with inferior vena cava (IVC) tumor thrombus (RCC-IVCTT), and to establish and validate corresponding predictive models. MethodsClinical data of 583 RCC-IVCTT patients were retrospectively analyzed. Of these, 465 cases were used to construct predictive models, and 118 cases were used for validation. Univariate and multivariate analyses identified independent predictors for surgical strategies. Two nomogram prediction models were established based on relevant independent factors to predict surgical approach. ResultsIn the development cohort, 342 patients underwent IVC thrombectomy (IVCT), 91 underwent IVC cavectomy (IVCC), and 32 underwent IVC reconstruction (IVCR). Multivariate logistic regression analysis identified the following predictors for inability to perform IVCT: higher Mayo classification of tumor thrombus (TT), IVC wall invasion, presence of bland thrombus, pan-immune-inflammation value (PIV) > 358×109, and maximum anteroposterior (AP) diameter of IVC at renal vein ostium (RVo) > 24mm. Platelet > 170×109/L and inadequate collateral circulation were predictors for IVCR. The developed model predicted capacity of the nomogram was evaluated in terms of its calibration, discrimination, and clinical utility. The validation set confirmed these findings. ConclusionThe comprehensive preoperative predictive model for RCC-IVCTT patients aids in preoperative determination of the required surgical approach and necessity for IVC angiography, facilitating perioperative preparation and reducing unnecessary invasive examinations.

Deep Learning Model for Predicting Immunotherapy Response in Advanced Non−Small Cell Lung Cancer

Only a small fraction of patients with advanced non-small cell lung cancer (NSCLC) respond to immune checkpoint inhibitor (ICI) treatment. For optimal personalized NSCLC care, it is imperative to identify patients who are most likely to benefit from immunotherapy. To develop a supervised deep learning-based ICI response prediction method; evaluate its performance alongside other known predictive biomarkers; and assess its association with clinical outcomes in patients with advanced NSCLC. This multicenter cohort study developed and independently validated a deep learning-based response stratification model for predicting ICI treatment outcome in patients with advanced NSCLC from whole slide hematoxylin and eosin-stained images. Images for model development and validation were obtained from 1 participating center in the US and 3 in the European Union (EU) from August 2014 to December 2022. Data analyses were performed from September 2022 to May 2024. Monotherapy with ICIs. Model performance measured by clinical end points and objective response rate (ORR) differentiation power vs other predictive biomarkers, ie, programmed death-ligand 1 (PD-L1), tumor mutational burden (TMB), and tumor-infiltrating lymphocytes (TILs). A total of 295 581 image tiles from 958 patients (mean [SD] age, 66.0 [10.6] years; 456 [48%] females and 502 [52%] males) treated with ICI for NSCLC were included in the analysis. The US-based development cohort consisted of 614 patients with median (IQR) follow-up time of 54.5 (38.2-68.1) months, and the EU-based validation cohort, 344 patients with 43.3 (27.4-53.9) months of follow-up. The ORR to ICI was 26% in the developmental cohort and 28% in the validation cohort. The deep learning model's area under the receiver operating characteristic curve (AUC) for ORR was 0.75 (95% CI, 0.64-0.85) in the internal test set and 0.66 (95% CI, 0.60-0.72) in the validation cohort. In a multivariable analysis, the deep learning model's score was an independent predictor of ICI response in the validation cohort for both progression-free (hazard ratio, 0.56; 95% CI, 0.42-0.76; P < .001) and overall survival (hazard ratio, 0.53; 95% CI, 0.39-0.73; P < .001). The tuned deep learning model achieved a higher AUC than TMB, TILs, and PD-L1 in the internal set; in the validation cohort, it was superior to TILs and comparable with PD-L1 (AUC, 0.67; 95% CI, 0.60-0.74), with a 10-percentage point improvement in specificity. In the validation cohort, combining the deep learning model with PD-L1 scores achieved an AUC of 0.70 (95% CI, 0.63-0.76), outperforming either marker alone, with a response rate of 51% compared to 41% for PD-L1 (≥50%) alone. The findings of this cohort study demonstrate a strong and independent deep learning-based feature associated with ICI response in patients with NSCLC across various cohorts. Clinical use of this deep learning model could refine treatment precision and better identify patients who are likely to benefit from ICI for treatment of advanced NSCLC.

Identification of two distinct clusters in membranous lupus nephritis patients: recognition of a high-risk profile based on unsupervised analysis.

Membranous lupus nephritis (MLN) traditionally includes class V (alone) and may be associated with other classes (III or IV). The clinical, therapeutic and prognosis relevance of the classification remains controversial. A retrospective cohort of 412 MLN patients from the First Affiliated Hospital of Sun-Yat Sen University was followed for a median of 65.68 (IQR: 23.13-131.70) months. The primary outcomes were adverse renal events including all-cause death and end-stage renal diseases (ESRD). Phenotypes were identified and validated using unsupervised clustering analysis (K-means), Principal component analysis (PCA) and decision tree analysis. Distinct clinical and pathological differences were noted for the traditional Class IV+V, the traditional Class V+III and Class V exhibited high similarities in clinical features and prognosis (P=0.074). K-means clustering revealed high-risk (n=180) and low-risk groups (n=232), with significant differences in adverse renal outcomes (9.2%vs. 4.1%, P<0.001). To recognize the high-risk profile of MLN patients, a decision tree based on Systemic Lupus Erythematosus Disease Activity Index (SLEDAI) Score, hemoglobin, serum creatinine, traditional classification, and activity index of renal biopsy accurately clustered patients in the development (95.8% accuracy) and validation (87.1% accuracy) cohorts. Two novel phenotypic clusters, more predictive than traditional classifications, enhance high-risk profile identification and prognostic accuracy.