Feeding intolerance (FI) is a common complication in premature infants, often resulting in delayed achievement of full enteral feeding and prolonged hospitalization. This study aimed to construct an interpretable risk-prediction model for FI using machine-learning algorithms based on routine electronic medical-record data. We prospectively collected clinical data for 420 preterm infants admitted to a tertiary hospital between June 2023 and August 2024. Time zero was defined as birth, the intended prediction time point was admission to the Neonatal Intensive Care Unit, and the prediction horizon was the first week of life. To avoid data leakage, the cohort was split 70:30 (stratified by outcome) before any further processing; missForest-style imputation, LASSO feature selection (10-fold CV, λ_min rule), and hyper-parameter tuning were performed within the training set only (9 predictors retained). Nine machine-learning algorithms were compared: support vector machine (SVM), random forest (RF), artificial neural network (ANN), logistic regression (LR), extreme gradient boosting (XGBoost), elastic net (EN), light gradient boosting machine (LightGBM), k-nearest neighbor (KNN), and decision tree (DT). Model performance was assessed by area under the receiver-operating characteristic curve (AUC), accuracy, recall, precision, F1 score (all with 95% bootstrap CIs from 2000 resamples), calibration intercept and slope, and Brier score. SHapley Additive exPlanations (SHAP; KernelExplainer with k-means background of 100 training samples, nsamples = 2000) were used for model interpretation. The incidence of FI was 43.6% (183/420). After LASSO selection, 9 predictors remained. On the held-out test set (n = 126, 55 events), the KNN model achieved the most balanced performance: AUC = 0.862 (95% CI 0.795–0.922), F1 = 0.733 (95% CI 0.621–0.820), Brier = 0.146 and calibration slope = 1.062 (95% CI 0.711–1.643). Internal validation by 100 bootstrap optimism correction showed an optimism-corrected AUC of 0.865 for KNN, indicating only modest overfitting. SHAP analysis identified gastric-content withdrawal, caffeine therapy, tube feeding, PICC insertion, antibiotic use, defecation interval, gastric injury, birth weight, and sex as the most influential contributors to the model-predicted risk. A leakage-free KNN-based FI risk-prediction model constructed from routine clinical data demonstrated good discrimination and acceptable calibration in internal validation. Given the single-center design, the limited sample size, and the uncertain temporal precedence of some predictors, the model is exploratory and is not ready for clinical implementation; prospective studies with a predefined prediction time point, standardized predictor collection, and external validation are required.
While the frequency of constipation increases sharply with age, the genetic links between constipation and aging-related phenotypes remain largely unexplored. So far, research has typically focused only on individual characteristics while neglecting broader genetic connections. In this study, we evaluated the shared genetic architecture of constipation and aging-related phenotypes in order to better understand the causes of aging-related intestinal dysfunction. Using publicly available genome-wide association study summary statistics, we developed a multi-stage analysis pipeline to identify common signals at the variation and genetic levels. We studied 4 aging-related phenotypes: frailty index (FI), telomere length, facial aging and GrimAge acceleration. Our approach combines linkage disequilibrium score regression and high-definition likelihood with Pleiotropy Analysis under Composite Null Hypothesis, functional mapping and annotation and Bayesian colocalization to identify pleiotropic loci, followed by tissue-specific mapping via Multi-marker Analysis of GenoMic Annotation and stratified linkage disequilibrium score regression. Constipation showed a moderate positive genetic correlation with FI in both linkage disequilibrium score regression and high-definition likelihood. The inverse correlation with TL was weak and method-dependent. No significant genome-wide genetic correlation was detected with facial aging or GrimAge acceleration. Multi-layer locus mapping and gene-based analyses prioritized several candidate genes, including MSL2 , STAG1 , and SRSF9, through different combinations of association-based evidence. However, no gene showed multiple-testing-corrected summary-data-based Mendelian randomization evidence shared between constipation and any of the 4 aging-related phenotypes. Multi-marker Analysis of GenoMic Annotation suggested cerebellar enrichment for genes associated with the constipation–FI pair, whereas stratified linkage disequilibrium score regression identified broader central nervous system enrichment for FI but no significant cell-type enrichment for constipation. These findings support a shared genetic component between constipation and selected aging-related phenotypes and provide candidate loci, genes and tissues for further functional investigation.
Constipation is a common gastrointestinal disorder that can substantially impair quality of life. Environmental exposure to heavy metals has been implicated in gastrointestinal dysfunction; however, population-based evidence regarding their associations with constipation remains limited. This study aimed to evaluate the associations of blood lead, mercury, and cadmium levels with constipation among US adults. This cross-sectional study used data from the National Health and Nutrition Examination Survey (NHANES) 2005–2010 and included 11,014 adults aged ≥20 years. Survey-weighted multivariable logistic regression models were constructed, with metal levels analyzed as log-transformed continuous variables and quartile categories based on the original concentrations. Restricted cubic spline (RCS) analyses were performed to assess dose–response relationships. Subgroup analyses, interaction tests, and sensitivity analyses were also conducted. In the fully adjusted model, blood lead and mercury levels were each significantly inversely associated with constipation. Each 1-unit increase in log-transformed blood lead was associated with 50.4% lower odds of constipation (OR = 0.496, 95% confidence interval [CI]: 0.331–0.745), whereas each 1-unit increase in log-transformed blood mercury was associated with 29.9% lower odds (OR = 0.701, 95% CI: 0.542–0.905). Blood cadmium showed no significant association. RCS analyses further confirmed linear inverse associations of both lead and mercury with constipation, whereas no significant association was identified for cadmium. A significant interaction was identified between educational level and the association between blood lead and constipation ( P for interaction = .04). The sensitivity analyses yielded generally consistent results. Blood lead and mercury levels were significantly and linearly inversely associated with constipation among US adults, whereas blood cadmium showed no significant association. Educational attainment may modify the association between blood lead exposure and constipation. Further prospective and mechanistic studies are needed to clarify the nature and underlying mechanisms of these associations.
Background: Immune-related pancreatitis is a rare but clinically significant adverse event associated with immune checkpoint inhibitor therapy. Early recognition and appropriate management are essential to prevent potentially serious outcomes. Herein, we report a case of suspected immune-related pancreatitis during nivolumab-based immunotherapy for metastatic gastric cancer. Case presentation: An 82-year-old female patient, post-gastrectomy for gastric cancer with subsequent cerebral metastases, multiple leptomeningeal metastases, subcutaneous metastases, and retroperitoneal lymph node metastases, received “nivolumab + apatinib” therapy for 4 cycles. On the second day of the fourth cycle of treatment, the patient developed mild epigastric pain, acid regurgitation, and elevated pancreatic enzyme levels, suggestive of immune-related pancreatitis. Alternative etiologies, including biliary disease, alcohol use, and hypertriglyceridemia, were excluded. The patient was classified as Grade 1 according to the Common Terminology Criteria for Adverse Events v5.0 and was managed with fasting, parenteral nutrition, octreotide, enzyme inhibition, and electrolyte balance adjustment. Pancreatic enzyme levels normalized, and symptoms resolved. Nivolumab and apatinib were subsequently withheld. Conclusion: This case highlights that immune-related pancreatitis may present with only mild abdominal symptoms despite marked pancreatic enzyme elevation during immune checkpoint inhibitor therapy. Clinicians should maintain a high index of suspicion for pancreatic immune-related adverse events, even in the absence of typical imaging findings, and systematically exclude alternative causes to ensure timely and appropriate management.
Mediastinal nodal classification in non–small cell lung cancer (NSCLC) at the N0/N1–N2/N3 treatment-defining boundary guides therapeutic strategy and is commonly finalized in multidisciplinary tumor board (MDT) meetings. We compared positron emission tomography/computed tomography (PET/CT) and endobronchial ultrasound–guided transbronchial needle aspiration (EBUS-TBNA) against MDT assignment and quantified discordance at this threshold. We retrospectively analyzed 114 consecutive NSCLC patients who underwent both PET/CT and EBUS-TBNA before MDT review. Nodal stage was dichotomized as N0/N1 vs N2/N3. Because the MDT decision incorporates both modalities, agreement with MDT is reported as observed agreement and Cohen κ together with the direction of discordance, rather than as sensitivity and specificity. In the subgroup with surgical mediastinal staging, both modalities and the MDT category were compared with histopathology. MDT assigned N2/N3 in 84 patients (73.7%). Relative to the MDT category, PET/CT over-classified 16 patients (14.0%) and under-classified 9 (7.9%). EBUS-TBNA under-classified 15 of the 43 patients it classified N0/N1 (34.9%), and under-classification persisted among the 25 patients in whom every sampled station was diagnostic (8/25, 32.0%); 12 of these 15 assignments rested on clinical–radiological course rather than histopathology. Agreement with the MDT category was 85.1% (κ = 0.662) for EBUS-TBNA and 78.1% (κ = 0.389) for PET/CT; because malignant cytology was definitional for the MDT category, 67 of 97 concordant EBUS-TBNA classifications (69.1%) were determined by that rule and do not measure EBUS-TBNA performance. Against surgical histopathology in 18 patients, PET/CT classified 11 of 13 pathologically N0 patients as N2/N3 (specificity 15.4%; 95% CI, 4.3–42.2) and EBUS-TBNA identified 2 of 5 with pathological N2 disease (sensitivity 40.0%; 95% CI, 11.8–76.9). At the treatment-defining nodal boundary, the two modalities diverged from the multidisciplinary decision in opposite directions: PET/CT over-classified nodal status, whereas EBUS-TBNA under-classified in approximately 1 in 3 patients it had assigned N0/N1, including those in whom sampling was complete, although most of these assignments rested on clinical course rather than tissue. These findings are consistent with guideline recommendations that PET-positive mediastinal nodes require tissue confirmation and that a negative endosonographic result be interpreted alongside pretest probability and station coverage.
