To investigate the protective effects of mitochondrial preconditioning in pericytes (PC) against pulmonary vascular leakage in septic rats.

① 128 specific-pathogen-free SD rats (equal gender, 200±20 g) were randomized into: Sham group (postoperative tail vein injection of 0.5 mL saline), Sepsis (Sep) group (CLP+saline), PC group (CLP+untreated PC: 10⁶ cells/rat), and Mito-PC group (CLP+PC preconditioned with 2 μg mitochondria/10⁴ cells for 12 h). Assessments included: PC lung colonization (flow cytometry), pulmonary barrier function (Evans blue assay), lung histopathology (HE staining), serum organ injury markers [cTnT (cardiac), urea/creatinine (renal)], and inflammatory cytokines (TNF-α, IL-6). ② MSC-derived mitochondria were validated by electron microscopy/flow cytometry; primary retinal PCs from weaned SD rats were purity-verified (confocal microscopy). In vitro groups: Control (PC), Mito-PC, PC+H₂O₂ (0.5 μmol/L, 4 h), and Mito-PC+H₂O₂. Antioxidant capacity was assessed via pentose phosphate pathway (PPP) activity, NADPH levels, G6PD activity, and NADP⁺/NADPH ratio.

① Compared with Sham, Sep group showed significant increase in pulmonary leakage (Evans blue P<0.05), severe lung injury, elevated serum markers (TNF-α, IL-6, cTnT, urea, creatinine all P<0.05), 0% 72 h survival. PC group showed partial improvement (P<0.05). Mito-PC group demonstrated significant reduction in vascular leakage (P<0.05 vs PC group), improved histopathology and organ function (P<0.05), attenuated inflammation (P<0.05), higher 72 h survival rate (P<0.05).② Mitochondrial preconditioning significantly enhanced PPP activation and NADPH-mediated antioxidative capacity, Mito-PC+H₂O₂ vs PC+H₂O₂ showed improved cell viability (P<0.05), Mito-PC vs PC showed increased G6PD activity (P<0.05), decreased NADP⁺/NADPH ratio (P<0.05).

Mitochondrial preconditioning potentiates pericyte-mediated protection against sepsis-induced pulmonary vascular leakage through enhanced pentose phosphate pathway activity. Mitochondrial preconditioning potentiates pericyte-mediated protection against sepsis-induced pulmonary vascular leakage through enhanced pentose phosphate pathway activity.

To investigate the critical role of hypoxia in impairing spermatogenesis via M1 macrophage polarization and subsequent induction of lysosome-dependent cell death (LDCD) in Leydig cells.

Ten eight-week-old male SD rats (weight 220~250 g) were randomly divided into normoxia (n=5) and hypoxia groups (n=5). Hypoxic models were established by housing rats in hypoxic chambers simulating 5 000 m altitude for 4 weeks. TM3 mouse Leydig cells underwent 1% O₂ exposure for 2 d. A transwell co-culture system integrated RAW 264.7 macrophages and TM3 cells. Testicular histopathology was evaluated via HE staining (spermatogenic cells/tubular area). Testosterone levels were quantified by ELISA (BCA-quantified). Immunofluorescence assessed Leydig cells (CYP11A1⁺), M1 macrophages (CD68⁺/CD86⁺ in tissue; iNOS⁺ in cells), and lysosomal integrity (LysoTracker/CTSB staining). Single-cell RNA-seq data (PRJNA926018) with GSEA identified programmed cell death pathways.

Hypoxia significantly reduced spermatogenic cell density (P<0.01), testosterone levels (P<0.01), and Leydig cell counts (P<0.01). scRNA-seq revealed decreased proportions of elongated spermatids, Leydig cells, and sertoli cells (P<0.01), but increased spermatocytes/spermatogonia (P<0.01), with significant LDCD enrichment in Leydig cells (P=0.002 7). Hypoxia induced testicular M1 macrophage infiltration (CD68⁺/CD86⁺) and elevated iNOS⁺ in RAW 264.7 cells (P<0.01). Transwell co-culture showed hypoxia-treated macrophages,but not hypoxia alone,disrupted lysosomal membrane stability in TM3 cells.

Hypoxia impairs spermatogenesis by driving M1 macrophage polarization, triggering LDCD in Leydig cells, and reducing testosterone synthesis.

To construct CD19-targeted chimeric antigen receptor natural killer (CAR-NK) cells armored with interleukin15 (IL15) or IL15 receptor-linker (RLI) and preliminarily validate their proliferative capacity and anti-tumor activity in vitro.

Natural killer cells (NK92 and NK92MI) from patients with human malignant non-Hodgkin lymphoma were cultured, and CD19-targeted CAR-NK cells armored with IL15 or RLI were prepared using retroviral vector particles. IL15 secretion was measured by ELISA, and proliferative capacity was assessed via CFSE dilution assays. Human B-lymphocytic leukemia cells (Nalm6-GFP-Luc) and human colon cancer cells overexpressing CD19 (hCD19-SW620-GFP-Luc) were cultured, with surface CD19 expression confirmed (>99% positivity for both). Anti-tumor activity was evaluated by measuring cytotoxicity at effector-to-target (E:T) ratios using luciferase-based assays (4/12 h), detecting surface CD107a expression via flow cytometry, and quantifying cytokine release using CBA assays.

NK92/NK92MI-CD19 CAR cells armored with IL15 or RLI were successfully generated. IL15 secretion was significantly higher in armored groups versus non-armored controls (P<0.01). Without IL-2 stimulation, IL15/RLI enhanced CAR-NK proliferation (P<0.05). Both armored designs significantly increased tumor-killing efficiency (P<0.05) and CD107a degranulation. IL15/RLI-armored NK92-CD19 CAR cells exhibited elevated release of IL2, IL10, IL6, TNF-α, sFas, IFN-γ, and Granulysin (P<0.05), while armored NK92MI-CD19 CAR cells showed additional increases in Granzyme A, Granzyme B, and Perforin (P<0.05).

IL15/RLI-armored NK92/NK92MI-CD19 CAR cells demonstrate potent anti-tumor activity, supporting their combined clinical therapeutic potential for tumors.

To investigate the anti-inflammatory effects of Ophiopogon japonicus polysaccharides (OJP) on submandibular gland inflammation in a mouse model of Sjögren’s syndrome (SS) and its impact on the tumor necrosis factor-α (TNF-α)/nuclear factor kappa-B (NF-κB)/NOD-like receptor family pyrin domain containing 3 (NLRP3) signaling pathway.

① Eight C57BL/6J mice served as the control group (Control group), while 32 C57BL/6J mice were subjected to immune induction to establish the SS model and then randomly divided into 4 groups (n=8 each): model group (SS group), low-dose OJP group [OJP (10 mg/kg) group], high-dose OJP group [OJP (20 mg/kg) group], and hydroxychloroquine group (Hyd group). The Control and SS groups received intragastric administration of saline, the OJP (10 mg/kg) and OJP (20 mg/kg) groups received 0.2 mL of OJP solution at 10 mg/kg or 20 mg/kg, respectively, and the Hyd group received 0.2 mL of hydroxychloroquine solution at 100 mg/kg, once daily for 4 weeks. Water consumption and salivary flow rate were measured. Serum levels of inflammatory cytokines interleukin-6 (IL-6), interleukin-1β (IL-1β), and TNF-α were detected by ELISA. Pathological changes in submandibular gland tissue were observed by HE staining. Protein expression related to the TNF-α/NF-κB/NLRP3 pathway was assessed by Western blot, and TNF-α and NLRP3 protein expression was detected by immunofluorescence. ② RAW264.7 macrophages were divided into blank control group (Control group), model group (SS group), low-dose OJP group (OJP1 group), medium-dose OJP group (OJP2 group), and high-dose OJP group (OJP4 group). After 24 hours of culture, the medium was removed; the Control group received 2 mL of complete DMEM medium, the SS group received 2 mL of complete DMEM medium containing 16 μg LPS, and the different OJP dose groups received 2 mL of complete DMEM medium containing 16 μg LPS plus OJP at concentrations of 1, 2, or 4 mg/mL, respectively. Inflammatory cytokine levels were measured by ELISA, TNF-α/NF-κB/NLRP3 pathway-related protein expression was detected by Western blot, and TNF-α protein expression was detected by immunofluorescence.

① Compared with the Control group, the SS group showed increased water consumption and decreased salivary flow rate (P<0.05), elevated serum levels of IL-1β, IL-6, and TNF-α (P<0.01), aggravated pathological damage observed by HE staining in submandibular glands, and increased expression of NLRP3, apoptosis-associated speck-like protein containing a CARD (ASC), cysteinyl aspartate specific proteinase (Caspase-1), IL-1β, TNF-α, TNF receptor type 1 (TNFR1)/glyceraldehyde-3-phosphate dehydrogenase (GAPDH), and phosphorylated NF-κB p65 (p-NF-κB p65)/NF-κB p65 (P<0.01), with immunofluorescence also showing elevated TNF-α and NLRP3 protein expression. Compared with the SS group, the Hyd group and both OJP dose groups exhibited decreased water consumption, increased salivary flow rate (P<0.01), reduced serum levels of IL-1β, IL-6, and TNF-α, alleviated submandibular gland pathological damage, decreased expression of NLRP3, ASC, Caspase-1, IL-1β, TNF-α, TNFR1/GAPDH, and p-NF-κB p65/NF-κB p65 proteins by Western blotting (P<0.01), and reduced TNF-α and NLRP3 protein expression by immunofluorescence. ② Compared with macrophages treated solely with LPS, those treated with OJP intervention showed reduced levels of inflammatory cytokines and decreased expression of NLRP3, ASC, Caspase-1, IL-1β, TNF-α, TNFR1/GAPDH, and p-NF-κB p65/NF-κB p65 (P<0.01).

OJP may exert anti-inflammatory effects by inhibiting the activation of the TNF-α/NF-κB/NLRP3 signaling pathway, thereby alleviating the inflammatory response in SS mice.

To investigate the mechanism by which SLC38A1 promotes hepatocellular carcinoma (HCC) progression through glutamine transport-mediated activation of the PI3K/AKT signaling pathway.

Bioinformatics analysis was employed to assess the correlation between SLC38A1 expression and clinicopathological features/prognosis in HCC patients, with functional enrichment analysis of SLC38A1 conducted. SLC38A1 was silenced or over expressed via transfection in Huh-7 cells, with glutamine inhibited using DRP-104 and the PI3K/AKT pathway suppressed using LY294002. Cell viability, proliferation, migration, invasion, and glutamine concentration were evaluated using CCK-8, EdU staining, scratch wound healing assay, Transwell chamber assay, and glutamine assay kits, respectively. PI3K/AKT pathway activity was assessed by Western blotting.

SLC38A1 mRNA and protein levels were significantly higher in HCC tumor tissues than in adjacent normal tissues (P<0.05). HCC patients with high SLC38A1 expression exhibited significantly lower overall survival than those with low expression (P<0.05). SLC38A1 expression correlated significantly with pathologic T-stage (pT), N-stage (pN), M-stage (pM), survival status, and immune infiltration in HCC patients (P<0.05). SLC38A1 silencing markedly reduced glutamine uptake in HCC cells (P<0.001), suppressing cell viability, proliferation, migration, invasion, and PI3K/AKT pathway activity (P<0.001). Conversely, SLC38A1 overexpression promoted proliferation, migration, invasion, and PI3K/AKT activation (P<0.001). DRP-104-mediated glutamine inhibition suppressed HCC cell malignancy and PI3K/AKT signaling while abolishing the oncogenic effects of SLC38A1 overexpression (P<0.001). PI3K/AKT pathway inhibition blocked the pro-tumorigenic effects of SLC38A1 overexpression (P<0.001).

SLC38A1 promotes proliferation, migration, and invasion in HCC by activating the PI3K/AKT pathway through enhanced glutamine transport.

To demonstrate that neferine (Nef) alleviates Parkinson’s disease (PD) by inhibiting microglial pyroptosis mediated through the reactive oxygen species (ROS)/NOD-like receptor protein 3 (NLRP3)/Caspase-1 pathway.

BV2 microglial cells were divided into: control group, lipopolysaccharides (LPS)-adenosine triphosphate (ATP) group, and LPS-ATP+Nef group. Pyroptosis was induced by 1 μg/mL LPS + 5 mmol/L ATP, with 2 mmol/L Nef pretreatment. Eighteen 10-12-week-old male C57BL/6 mice (22~25 g) were randomly assigned to: control (n=6), 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) (n=6), and MPTP+Nef (n=6) groups. Detection methods included: flow cytometry for pyroptosis, Cell Counting Kit-8 (CCK-8) for viability, 2',7'-dichlorodihydrofluorescein diacetate (DCFH-DA) for ROS, commercial kits for malondialdehyde (MDA), superoxide dismutase (SOD), glutathione (GSH), ELISA/Western blot for interleukin-1β (IL-1β)/IL-18, immunofluorescence/immunohistochemistry for NLRP3/Caspase-1, tyrosine hydroxylase (TH) immunohistochemistry, hematoxylin-eosin staining for neuropathology, and modified neurological severity score (mNSS).

Versus control, LPS-ATP group showed decreased viability (P=0.002), increased pyroptosis (P<0.001), elevated ROS (P<0.001)/MDA (P<0.001)/IL-1β (P<0.001)/IL-18 (P<0.001), upregulated NLRP3 (P<0.001)/Caspase-1 (P<0.001), and reduced GSH (P<0.001)/SOD (P<0.001). Nef treatment reversed these effects (all P<0.05). According to the results of murine studies, compared with the control group, the MPTP group had increased mNSS (P<0.001)/tissue ROS (P<0.001), downregulated TH (P<0.001), upregulated NLRP3 (P<0.001)/Caspase-1 (P<0.001). Nef treatment significantly attenuated the MPTP-induced deleterious effects (P<0.05). Histopathological analysis revealed that control group exhibited uniformly distributed hippocampal neurons with distinct nuclear morphology; MPTP group showed neuronal swelling, interstitial edema, and nuclear atrophy; MPTP+Nef group demonstrated ameliorated neuronal damage.

Nef inhibits microglial pyroptosis via ROS/NLRP3/Caspase-1 axis, ameliorating PD neuroinflammation and pathology.

To investigate polyp miss rate and influencing factors during colonoscopy using the opportunity of short-interval repeat examinations in day surgery settings.

This single-center retrospective cohort study consecutively enrolled 1 005 patients undergoing polypectomy at the Endoscopy Center of Army Medical University First Affiliated Hospital(March-June 2023). All patients received initial diagnostic colonoscopy followed by therapeutic colonoscopy(repeat examination). Using repeat examination findings as reference, we calculated patient miss rate, polyp miss rate, and adenoma patient miss rate. Multivariable logistic regression analyzed operator- and patient-related factors influencing patient miss rates.

Patient miss rate, polyp miss rate, and adenoma miss rate were 26.67%(95%CI: 23.95~29.52), 16.45%(95%CI:15.26~17.70), and 15.52%(95%CI:13.34~17.91), respectively. Multivariable analysis identified independent risk factors— operator-related: initial procedure by junior endoscopists(OR=2.07, 95%CI: 1.42~3.03, P<0.001) and afternoon procedures(13:30-16:00; OR=1.68, 95%CI: 1.01~2.81, P=0.047); Patient-related: inadequate bowel preparation(OR=4.16, 95%CI: 2.28~7.61, P<0.001) and flat-type lesions(OR=4.33, 95%CI: 2.49~7.52, P<0.001). Reduced miss rates occurred with polyps 0.3~0.5 cm(OR=0.30, 95% CI: 0.21~0.41, P<0.001) and 0.6~0.9 cm(OR=0.10, 95%CI: 0.04~0.22, P<0.001). Inadequate preparation during therapeutic examination decreased polyp detection(OR=0.23, 95%CI: 0.10~0.53, P<0.001).

Colonoscopy exhibits significant polyp miss rates influenced by operator and patient factors. Standardized repeat examinations in day surgery settings effectively detect missed lesions, supporting long-term quality control.

To investigate the role of mitochondrial antiviral signaling protein (MAVS) in viral infection-triggered generalized pustular psoriasis (GPP) in children.

This retrospective case-control study enrolled 80 GPP patients aged 0~18 years from Children’s Hospital of Chongqing Medical University (from October 2013 to April 2019). Whole-exome sequencing identified rare MAVS variants associated with GPP. Pathogenicity of variants was predicted using Mutation Taster, Disease Association, SIFT, and CADD bioinformatics tools. Sanger sequencing validated variants, followed by construction of wild-type (WT) and mutant MAVS expression plasmids transfected into HEK 293 cells. Protein expression was assessed by Western blot. Dual-luciferase reporter gene assays measured IFNB1 and NF-κB transcriptional activity. Genotype distribution of the MAVS c.171dupT/p.H57fs variant was analyzed using Fisher’s exact test.

This study enrolled 80 pediatric GPP patients (aged 0~18 years). Whole-exome sequencing identified five rare MAVS variants, with bioinformatics analyses predicting deleterious effects on protein stability and function. Western blot demonstrated that the c.171dupT mutation in GPP patients significantly reduced full-length MAVS expression (P<0.001); dual-luciferase assays further revealed this variant impaired MAVS-mediated IFNB1 transcriptional activation by 85% (P<0.001), abrogated NF-κB signaling pathway activation (P<0.001), but exhibited no dominant-negative effect on wild-type MAVS function（P>0.05）.

The MAVS c.171dupT frameshift variant may contribute to infection-triggered GPP in children, suggesting its potential as a genetic biomarker for GPP susceptibility.

To develop and validate a nomogram incorporating postoperative blood pressure and other multifactorial predictors for the risk of delayed cerebral ischemia(DCI) following surgical clipping in patients with aneurysmal subarachnoid hemorrhage(aSAH).

This retrospective cohort study consecutively enrolled 272 aSAH patients who underwent aneurysm clipping at the Department of Neurosurgery, Second Affiliated Hospital of Chongqing Medical University from January 2018 to January 2022. Demographic, clinical, and imaging data were collected. Predictors were screened using LASSO regression, and a multifactorial logistic regression-based nomogram was constructed. The dataset was randomly divided into training(n=190) and validation(n=82) sets at a 7∶3 ratio. Model discrimination, calibration, and clinical utility were evaluated using the area under the receiver operating characteristic curve(AUC), calibration curves, and decision curve analysis(DCA), respectively.

Among 272 enrolled patients, 98(36%) developed DCI. Multivariate analysis identified age(OR=0.97; 95%CI: 0.94~0.99; P=0.036), World Federation of Neurological Surgeons(WFNS) grade(OR=1.36; 95%CI: 1.02~1.82; P=0.038), Fisher grade(OR=1.44; 95%CI: 1.01~2.05; P=0.048), and mean arterial pressure on postoperative day 2(OR=1.06; 95%CI: 1.03~1.09; P<0.001) as independent risk factors for DCI. The model achieved C-indices of 0.739 (95%CI : 0.668~0.810) in the training set and 0.760 (95%CI: 0.647~0.872) in the validation set. Calibration curves demonstrated high agreement between predicted and observed probabilities(mean absolute error=0.027), while DCA confirmed significant net clinical benefit at threshold probabilities of 15%~65%.

We successfully developed a predictive model incorporating age, WFNS grade, Fisher grade, and postoperative day 2 mean arterial pressure, with elevated postoperative blood pressure identified as an early critical indicator for DCI development.

To investigate the efficacy of low-frequency repetitive transcranial magnetic stimulation (rTMS) combined with vestibular rehabilitation training (VRT) for persistent postural-perceptual dizziness (PPPD).

In this retrospective case-control study, 48 PPPD patients admitted to our rehabilitation center (September 2023 to September 2024) were divided into control group (n=24, VRT only) and study group (n=24, VRT + rTMS). Assessments at baseline and post-treatment included Dizziness Handicap Inventory (DHI), Vestibular Symptom Index (VSI), Hamilton Depression Rating Scale (HAMD), Pittsburgh Sleep Quality Index (PSQI), Numeric PPPD Questionnaire (NPQ), and overall response rate.

After 6 weeks, the study group showed significantly higher overall response rate compared to control group (91.67% vs 66.67%, P=0.033). The study group demonstrated significantly lower scores than control group: DHI (37.85±6.46 vs 48.53±7.54, P=0.007), VSI (3.85±0.46 vs 5.53±1.54, P=0.002), HAMD (10.20±6.30 vs 15.74±7.05, P=0.006), PSQI (6.24±1.88 vs 8.89±2.79, P=0.003), and NPQ (15.44±3.26 vs 20.81±9.14, P=0.009).

‍Combined low-frequency rTMS and VRT effectively alleviates dizziness symptoms and improves treatment efficacy in PPPD patients.

To analyze clinical characteristics affecting survival outcomes in gastric adenocarcinoma (GAC) patients with second primary malignancies (SPM) and construct a predictive model with a web-based calculator.

Patients diagnosed with GAC between January 2010 and December 2017 in the SEER database (n=24 085) were analyzed, comparing non-SPM (n=22 963) and SPM cohorts (n=1 122). SPM patients were randomized (3:1) into training (n=842) and internal validation cohorts (n=280). Univariate/multivariate Cox regression identified prognostic factors for model construction. Model performance was evaluated via ROC curves, calibration plots, and decision curve analysis (DCA). A web-based calculator was deployed using DynNom (https://kunma697.shinyapps.io/dynnomapp-1/). External validation used 192 SPM patients diagnosed at Yantai Yuhuangding Hospital (2010-2017).

χ² tests revealed SPM patients had higher age (56.3%), earlier T-stage (T1:29.2%; T2:10.5%), predominant gastric cardia involvement (43.7%), fewer distant metastases (12.3%), and higher rates of radiotherapy (32.5%) and surgery (77.2%) vs. non-SPM (P<0.05). Cox analyses identified GAC primary site, T-stage, SEER stage, radiotherapy/surgery history, plus SPM grade/stage/treatment history as significant predictors (P<0.05). AUCs in the training cohort were 0.771 (95%CI:0.722~0.820), 0.839 (95%CI:0.796~0.882), and 0.836 (95%CI: 0.792~0.879) for 1-/3-/5-year survival; internal validation showed 0.751 (95%CI: 0.700~0.801), 0.746 (95%CI: 0.695~0.797), and 0.772 (95%CI: 0.723~0.821); external validation yielded 0.713 (95%CI: 0.648~0.778), 0.805 (95%CI: 0.749~0.861), and 0.851 (95%CI: 0.801~0.901). Calibration indicated high prediction-actuality concordance; DCA confirmed clinical utility.

The model and web calculator incorporating GAC/SPM characteristics effectively predict SPM patient prognosis.

To evaluate the predictive value of deep learning features from chest CT images combined with clinical and radiomics features for epidermal growth factor receptor (EGFR) mutations in non-small cell lung cancer (NSCLC).

This case-control study retrospectively analyzed clinical and imaging data of 1 070 NSCLC patients from radiology departments at three hospitals (January 2013 to October 2023). Patients were divided into: a training set (n=502) and internal validation set (n=217) via 7∶3 randomization of 719 cases from the First Affiliated Hospital of Army Medical University; external validation set 1 (n=173) from General Hospital of Eastern Theater Command; external validation set 2 (n=178) from Daping Hospital of Army Medical University. Deep learning features were extracted using a 2.5D convolutional neural network (CNN) with ResNet101 backbone, radiomics features were derived from CT images, and clinical risk factors were identified to construct models. An integrated model combined deep learning, clinical, and radiomics features. All four models were developed using random forest (RF) classifiers. Calibration curves assessed goodness-of-fit, and decision curve analysis (DCA) evaluated clinical utility.

The deep learning model achieved AUCs of 0.833 7 (95%CI: 0.770 6~0.884 7), 0.815 1 (0.741 6~0.882 8), and 0.810 1 (0.745 2~0.873 6) in the internal and two external validation sets, respectively. Clinical models yielded AUCs of 0.731 0 (0.660 2~0.802 1), 0.746 0 (0.666 4~0.824 9), and 0.813 4 (0.743 1~0.883 6); radiomics models showed AUCs of 0.762 4 (0.692 4~0.825 1), 0.745 4 (0.671 1~0.814 3), and 0.724 7 (0.639 7~0.801 6). The integrated model demonstrated optimal performance with AUCs of 0.905 5 (0.857 0~0.945 4), 0.832 7 (0.763 3~0.896 4), and 0.889 0 (0.834 4~0.934 3). DCA indicated significant net benefit for EGFR prediction at threshold probabilities of 0.15~0.85 using the integrated model.

Deep learning features from CT images effectively predict EGFR mutation status in NSCLC. The integrated model combining deep learning, clinical, and radiomics features further enhances predictive performance.

To develop and validate a risk prediction model for infiltration/extravasation in peripheral intravenous catheter therapy.

This retrospective study analyzed 942 patients who completed the Infiltration/Extravasation Risk Assessment Form between January and June 2023 at the First Affiliated Hospital of Army Medical University (including Departments of Neurology, Endocrinology, Gastroenterology and Hepatobiliary Surgery). Patients were allocated to a derivation cohort (n=628) and validation cohort (n=314) in a 2∶1 ratio based on catheterization chronology. The derivation cohort served for model development and internal validation, while the validation cohort underwent external validation. Logistic scoring method constructed the risk model, with Hosmer-Lemeshow (HL) test assessing goodness-of-fit and ROC curve evaluating predictive performance. Twenty-one potential risk factors were assessed, including age, gender, chronic diseases, clinician experience, treatment compliance, and total infusion volume.

Infiltration/extravasation occurred in 48 cases (5.10% incidence: 31 derivation/17 validation). Among 21 factors, 15 showed significant association (P<0.05), with 6 independent predictors: junior high school education or below (OR=5.2), chronic disease history (OR=3.1), poor compliance (OR=2.8), lower extremity venipuncture (OR=4.1), total infusion ≥1 000 mL (OR=3.5), and hyperosmotic/corrosive medications (OR=6.7). The final prediction model was: Y=2×(low education)+1×(chronic disease)+1×(poor compliance)+1×(lower extremity puncture)+1×(volume ≥1 000 mL)+2×(corrosive agents). For the derivation cohort, AUC was 0.967 (95%CI: 0.936~0.998), specificity 0.911, sensitivity 0.935, with good calibration (χ²=4.135, P=0.845). Validation cohort showed AUC 0.939 (0.853~1.000), specificity 0.919, sensitivity 0.941, and acceptable calibration (χ²=8.998, P=0.085).

This model demonstrates excellent discriminative ability and calibration, providing an effective tool for identifying high-risk patients and guiding targeted preventive strategies.