ObjectiveTo investigate the value of predictive models established based on machine learning methods in predicting the risk of metabolic associated fatty liver disease （MAFLD）， and to analyze its key risk factors. MethodsA retrospective analysis was performed for the 50 variables of 2 168 healthy individuals who underwent physical examination in Department of Health Assessment， Xiyuan Hospital， China Academy of Chinese Medical Sciences， from January 2021 to December 2024， including body composition， past history， and laboratory tests， and according to whether they were diagnosed with MAFLD or not， they were divided into MAFLD group with 265 individuals and non-MAFLD group with 1 903 individuals. The Mann-Whitney U test was used for comparison of continuous data between two groups， and the chi-square test was used for comparison of categorical data between two groups. Randomly split the research data into a training set and a validation set in a 70% to 30% ratio. Predictive factors were screened from the training set data using univariate analysis， LASSO regression， and multivariate Logistic regression analysis. Predictive models were then constructed using seven machine learning methods： Logistic regression， decision tree， random forest （RF）， eXtreme gradient boosting， light gradient boosting machine， support vector machine， and artificial neural network. Model performance was evaluated by plotting receiver operating characteristic curve for the validation set and calculating the area under the curve （AUC）， sensitivity， specificity， and Youden index for each model. Furthermore， the SHapley Additive exPlanation （SHAP） method was used to analyze the contribution of variables in the optimal model. ResultsThe prevalence rate of MAFLD among the 2 168 subjects was 12.22% （265/2 168）. Smoking， diastolic blood pressure， phase angle， visceral fat area， muscle fat ratio， waist-to-hip ratio， aspartate aminotransferase， non-HDL-C/HDL-C ratio， triglyceride-glucose index， and gallstones were independent risk factors for MAFLD （all P<0.05）. The seven predictive models of support vector machine， eXtreme gradient boosting， decision tree， light gradient boosting machine， artificial neural network， RF， and Logistic regression had an AUC of 0.738， 0.754， 0.757， 0.786， 0.795， 0.796， and 0.815， respectively， in the validation set， among which the RF model had the best discriminatory ability （AUC=0.796， 95% confidence interval： 0.754 — 0.839）， with a sensitivity of 81.01%， a specificity of 63.16%， and a Youden index of 44.17%. The SHAP analysis showed that visceral fat area， waist-to-hip ratio， and diastolic blood pressure were the top three predictive factors in terms of importance. ConclusionThe RF model， constructed based on body composition and clinical indicators， has a good performance in predicting the risk of MAFLD， and its interpretability can help to identify high-risk individuals in the early stage in clinical practice.

ObjectiveTo investigate the patterns of chronic disease comorbidity among permanent elderly residents in Xuhui District of Shanghai, and to analyze the impact of different comorbidity patterns on the elderly’s ability to perform activities of daily living. MethodsBased on data from the 2015 and 2021 Surveys on Health Status and Health Service Utilization among Permanent Residents in Xuhui District of Shanghai, the study included permanent residents who fully participated in both surveys, were aged 60 years or older at the 2015 survey, self-reported having a chronic disease, and reported no significant changes in their chronic disease status between the two surveys. A prospective cohort study design was adopted, using the 2015 data as the baseline and the 2021 survey data as follow-up data. Physical examinations and questionnaire surveys were conducted by the Shanghai Xuhui District Center for Disease Control and Prevention in the subdistricts and towns within the jurisdiction from April to July in both 2015 and 2021. Latent class analysis (LCA) was used to classify comorbidity patterns across 11 types of chronic diseases. The smaller the values of the Akaike information criterion (AIC), the Bayesian information criterion (BIC), and the sample-corrected Bayesian information criterion (aBIC), the better the model fits. The Barthel index (BI) and the Lawton and Brody Instrumental Activities of Daily Living Scale (LB-IADL) were used to assess the participants’ activities of daily living (ADL) and instrumental activities of daily living (IADL). Multiple linear regression and multivariate logistic regression models were used to analyze the association between different comorbidity patterns and ADL and IADL. ResultsA total of 1 608 study participants were enrolled. Among them, 671 (41.73%) self-reported having no chronic diseases. The prevalence of 11 types of chronic diseases was as follows: hypertension in 743 participants (46.21%), heart disease in 148 participants (9.20%), cerebrovascular disease in 37 participants (2.30%), diabetes or abnormal blood glucose in 310 participants (19.28%), other endocrine system diseases (non-diabetic) in 39 participants (2.43%), malignant tumors in 37 participants (2.30%), chronic lung disease in 16 participants (1.00%), gout in 8 participants (0.50%), musculoskeletal disorders in 34 participants (2.11%), gastric or digestive system diseases in 27 participants (1.68%), and dyslipidemia in 15 participants (0.93%). LCA was used to classify comorbidity patterns of chronic diseases among older adults into 1 to 9 latent category combinations. Model 5 (i.e., 5 latent categories) had the lowest BIC (7 112.69) and aBIC (6 909.38), with a relatively low AIC (6 768.20), indicating the best model fit; thus, five latent categories represented the optimal combination in this study. The five latent categories (and their respective proportions of study participants) were the healthy control group (41.73%), the hypertension group (31.28%), the diabetes-hypertension group (19.28%), the severe illness group (5.29%), and the other metabolic diseases group (2.43%). According to multiple linear regression analyses, compared with the healthy control group, the diabetes-hypertension group had significantly lower BI-20 scores (b=-0.952, 95%CI: -1.317‒ -0.587, P<0.001) and lower IADL scores (b=-0.744, 95%CI: -0.999‒ -0.489, P<0.001); the severe illness group (b=-0.644, 95%CI: -1.072‒ -0.217, P=0.003) and the hypertension group (b=-0.344, 95%CI: -0.562‒ -0.125, P=0.002) had lower IADL scores. According to multivariate logistic regression analyses, compared with the healthy control group, older adults in the diabetes-hypertension group had a higher risk of disability (OR=2.835, 95%CI: 1.392‒5.774), impaired ADL function (OR=2.470, 95%CI: 1.637‒3.726), and impaired IADL function (OR=1.739, 95%CI: 1.264‒2.392); older adults in the severe illness group had a higher risk of impaired ADL function (OR=2.206, 95%CI: 1.171‒4.155). ConclusionAmong the elderly participating in the Survey on Health Status and Health Service Utilization of Permanent Residents in Xuhui District of Shanghai, diabetes-hypertension is one of the key chronic disease comorbidity combinations and has adverse effects on both ADL and IADL. Community health management should prioritize the elderly with diabetes-hypertension comorbidity, and strengthen early screening and comprehensive interventions to slow the decline in ADL and IADL.

ObjectiveTo investigate the application of central auditory testing combined with electroencephalographic features in the early screening of mild cognitive impairment due to Alzheimer's disease. MethodsA convenience sampling method was used to conduct a field survey in the Qixia District of Nanjing City, Jiangsu Province, from July to September, 2025. A total of 40 older adults diagnosed as mild cognitive impairment due to Alzheimer's disease and 23 cognitively normal older adults were recruited. Auditory N-back and dichotic listening paradigms were employed for assessment, while the scores of Montreal Cognitive Assessment Beijing Version (MoCA-BJ) and 32-channel electroencephalogram (EEG) data were simultaneously collected. Binary logistic regression analysis was applied to identify predictive indicators, receiver operating characteristic (ROC) curves were plotted, and the area under the curve (AUC) was calculated. ResultsThe total score of dichotic listening was positively correlated with overall cognitive function (r = 0.475, P < 0.01), while right-ear advantage showed negative correlations with overall cognition, visuospatial ability, abstract reasoning, and delayed recall (r < -0.276, P < 0.05). Task performance across all conditions of the N-back task was positively correlated with overall cognitive function (r > 0.249, P < 0.05). Regarding EEG features, δ and θ band power and power peaks were positively correlated with delayed recall (r > 0.258, P < 0.05), whereas (α+θ)/(α+β) and θ/(α+β) were positively correlated with naming ability and visuospatial executive function, respectively (r > 0.261, P < 0.05). The results of the binary logistic regression analysis showed that the total score of binaural dichotic listening, the accuracy of 1-back task, the average power of δ waves, and the ratio of θ/(α+β) were independent influencing factors for mild cognitive impairment due to Alzheimer's disease (P < 0.05). The ROC analysis indicated that the AUC for auditory task performance and EEG features were 0.845 and 0.735 respectively, and when combined, the AUC increased to 0.862. ConclusionThe screening model based on the integration of auditory task performance and EEG features can be used in the early identification of mild cognitive impairment due to Alzheimer's disease.

It is believed that there is a pathogenesis of yang deficiency in polycystic ovary syndrome （PCOS） infertility， and it is concluded that warm-yang method has a better effect in improving endometrial abnormality， enhancing the quality of follicles， correcting endocrine disorders， and resolving or alleviating clinical symptoms in PCOS infertility. Based on Yanghe Decoction （阳和汤）， a representative traditional Chinese medicine decoction for warming yang， Yanghe Xiaonang Decoction （阳和消囊汤） was formulated， combining with warm medicinals according to symptoms， and aerobic exercise was also advocated to help generate and develop yang qi， in order to provide ideas for clinical treatments.

Objective To analyze the virus subtypes, molecular evolutional and molecular transmission network features of the first confirmed mpox case in Huai’an City, Jiangsu Province, so as to provide insights into understanding of the transmission and evolution dynamics of mpox virus and formulation of the mpox control strategy in the city. Methods Genomic DNA was extracted from swabs of the first confirmed mpox case’s skin lesions in Huai’an City, and the amplicon sequencing library was constructed using the hypersensitive mpox virus whole-genome capture kit. High-throughput sequencing was performed using the GridION X5 nanopore sequencer on the Nanopore sequencing platform, and single nucleotide polymorphism (SNP) analysis of mpox virus genome sequences was performed following sequence assembly. In addition, phylogenetic analysis, genetic genealogy and molecular traceability analysis were performed. Results The virus whole genome sequence of the first confirmed mpox case was successfully obtained by high-throughput sequencing, with a full length of 197 182 bp, and was named hMpxV/China/JS-HA01/2023, which belonged to the clade IIb (West African clade) lineage B.1.3. Compared with the mpox virus reference sequence MPXV-M5312_HM12_Rivers-001 (GenBank accession number: NC_063383), the genome sequence of the Huai’an virus isolate carried 86 SNPs, including 40 SNPs in the coding region as non-synonymous mutations and 73 SNPs as nucleotide mutations caused by APOBEC3 (APOBEC3). Of the 97 mpox virus gene sequences, 79 sequences were included in the molecular network (81.44%), and the threshold of the genetic distance accessed to the network was 0.35/10⁵. There were two large molecular transmission clusters and one scattered cluster in the molecular transmission network of the mpox virus, andthehMpxV/China/JS-HA01/2023 sequence was located in the large cluster. The 97 gene sequences formed 92 haplotypes, including three shared haplotypes Hap_4, Hap_6 and Hap_38, and an exclusive haplotype Hap_1 of hMpxV/China/JS-HA01/2023 generated from mutation of the exclusive haplotype Hap_43, while the exclusive haplotype Hap_43 was generated from mutation of the shared haplotype Hap_38. Conclusions The whole genome sequence of the mpox virus isolated from the first confirmed mpox case in Huai’an City has been successfully obtained, and the molecular evolutionary and molecular transmission network characteristics of the virus have been preliminarily understood.

Objective:To investigate the role of serotonergic (5-HT) neurons activation in the dorsal raphe nucleus (DRN) in heat-induced seizures and its influence on breathing patterns in Scn1a + /- mice. Methods:24-day-old male Scn1a + /- mice were used for experiments, and sudden unexpected death in epilepsy(SUDEP) model was established through heat induction for 5 consecutive days. (1) Fluoxetine intervention experiment: 20 male mice were randomly divided into model group( n=10) and fluoxetine group ( n=10) according the weight-matched method. The fluoxetine group received intraperitoneal injection of fluoxetine (10 mg/kg) 45 min before heat induction each day for 5 consecutive days, while the model group received equal volume of 0.9% NaCl solution. (2) Chemogenetic activation experiment: 20 male mice were randomly divided into vector control group( n=10) and chemogenetic activation group ( n=10) according the weight-matched method. Empty vector or rAAV-TPH2-hM3d(Gq)-EGFP-WPREs was stereotaxically injected into DRN 14 d prior to seizure induction, and deschloroclozapine (5 mg/kg) was intraperitoneally injected 30 min before heat induction. Seizure characteristics and survival were assessed through video monitoring, and respiratory parameters were monitored. Immunofluorescence was used to detect colocalization of tryptophan hydroxylase 2(TPH2) and c-Fos in DRN. Statistical analysis was performed using SPSS 24.0 and GraphPad Prism 9.0. The independent sample t-test or Mann-Whitney test was used for inter-group comparison. Results:(1) In the fluoxetine intervention experiment: the survival rates between the model group and fluoxetine group showed no statistically significant difference ( χ2=2.23, P>0.05). As for the frequency of grade Ⅳ seizures, the model group (1.50(1.25, 2.35)min) demonstrated higher frequency than the fluoxetine group (0.43(0.20, 0.67)min) ( t=-3.40, P<0.05).With respect to respiratory parameters, the model group demonstrated shorter expiratory time ((0.10±0.02) s) and inspiratory time ((0.15±0.02) s) compared to the fluoxetine group ((0.16±0.05) s, (0.19±0.04) s) ( t=-3.47, -3.73, both P<0.01). The respiratory rate in the model group ((269.96±44.84) times/min) was significantly higher compared to the fluoxetine group ((195.04±52.37) times/min) ( t=3.44, P<0.01). The tidal volume in the model group ((0.10±0.02) mL) was significantly lower than the fluoxetine group ((0.13±0.04) mL) ( t=-2.19, P<0.05).The number of TPH2+ /c-Fos+ co-expressing cells in the model group (11.00±4.00) was lower than that in the fluoxetine group (33.00±8.39)( t=-4.16, P<0.05). (2) In the chemogenetic activation experiment: compared to the vehicle group, the chemogenetic activation group demonstrated significantly enhanced survival rates ( χ2=5.83, P<0.05). As for the frequency of grade Ⅴ seizures, the vehicle group (2.11(1.62, 3.44) times/min) showed higher frequency compared to the chemogenetic activation group (0.81(0.00, 1.62) times/min) ( t=18.00, P<0.05). In terms of respiratory parameters, the vehicle group showed shorter expiratory time ((0.10±0.01) s) and inspiratory time ((0.14±0.01) s) compared to the chemogenetic activation group ((0.12±0.01) s, (0.15±0.01) s) ( t=-2.78, -2.50, both P<0.05). The respiratory rate in the vehicle group ((208.37±9.73) times/min) was significantly higher than the chemogenetic activation group ((191.85±8.83) times/min) ( t=3.98, P<0.01). The tidal volume in the vehicle group ((0.09±0.01) mL) was significantly lower than the chemogenetic activation group ((0.12±0.02) mL) ( t=-4.77, P<0.001).The number of TPH2+ /c-Fos+ co-expressing cells in the vehicle group (9.00±3.46) was lower than that in the chemogenetic activation group (43.00±11.02)( t=-5.20, P<0.01). Conclusion:Specific activation of serotonergic neurons in the DRN can ameliorate heat-induced epileptic symptoms, improve respiratory function, and prolong survival time in Scn1a + /- mice.

Objective To investigate whether Piceatannol(PIC)improves diabetic retinopathy(DR)mediated by microglial polarization and to elucidate the underlying molecular mechanisms.Methods Network pharmacology and bioinformatics were used to analyze the common targets of DR,PIC,and microglia.Human retinal vascular endothelial cells(HRVECs)were cultured in vitro and randomly divided into the NG-HRVECs group,HG-HRVECs group,and HG+PIC-HRVECs group.Cell viability was assessed by the CCK-8 assay,apoptosis was detected by the TUNEL assay,and the concentrations of tumor necrosis factor-α(TNF-α)and interleukin-6(IL-6)were measured using ELISA kits.BV-2 cells were cultured in vitro and randomly divided into the NG-BV-2 group,HG-BV-2 group,HG+PIC-BV-2 group,HG+Si-NC-BV-2 group,HG+Si-CXCR4-BV-2 group,and HG+Ibrutinib-BV-2 group.The levels of arginase-1(Arg-1)and inducible ni-tric oxide synthase(iNOS)were measured using ELISA kits.Furthermore,conditioned medium(CM)from BV-2 cells of each group was collected to treat HRVECs,after which the viability,apoptosis rate,and TNF-α and IL-6 concentrations of the HRVECs were measured.A DR rat model was established and intervened with PIC to investigate the ameliorative effects of PIC on retinal pathology.Results Bioinformatics analysis identified CXCR4 as the key target of this study.Compared with the NG-HRVECs group,the apoptosis rate and the concentrations of TNF-α and IL-6 were increased in the HG-HRVECs group.Compared with the HG-HRVECs group,the HG+PIC-HRVECs group showed a decreased apoptosis rate and reduced concentrations of TNF-α and IL-6(all P＜0.05).Compared with the NG-BV-2 group,the HG-BV-2 group ex-hibited decreased Arg-1 levels and increased iNOS levels.Compared with the HG-BV-2 group,the HG+PIC-BV-2,HG+Si-CXCR4-BV-2,and HG+Ibrutinib-BV-2 groups all showed increased Arg-1 levels and decreased iNOS levels(all P＜0.05).Compared with the NG-BV-2-CM group,the HG-BV-2-CM group led to decreased viability,increased apoptosis rate,and in-creased concentrations of TNF-α and IL-6 in HRVECs.In contrast,the HG+PIC-BV-2-CM,HG+Si-CXCR4-BV-2-CM,and HG+Ibrutinib-BV-2-CM groups reversed the effects induced by HG-BV-2-CM on HRVECs(all P＜0.05).Animal experiment results showed that compared with DR model rats,rats treated with different doses of PIC exhibited significantly ameliora-ted retinal histopathological damage,and the protein expressions of Arg-1,iNOS,CXCR4,and p-BTK were reversed.Con-clusion PIC ameliorates DR progression mediated by microglial polarization by inhibiting the CXCR4/BTK pathway.

Objective:To evaluate the effect of laser-assisted laser in situ keratomileusis (LASIK) surgery for optimized monovision correction in the treatment of age-related accommodative deficiency with myopia on regulatory function.Methods:A cohort study was conducted.A total of 74 consecutive patients (74 eyes) with age-related accommodative insufficiency and myopia who underwent optimized monovision LASIK at Lanzhou Bright Eye Hospital from October 2019 to October 2021 were enrolled.Only non-dominant eye data were analyzed.Patients were randomized into a control group (37 cases, 37 eyes) treated with FS-LASIK and an optimization adjustment group (37 cases, 37 eyes) treated with Q-value adjusted monovision FS-LASIK.Preoperative and postoperative (within 1 year) data were collected for both groups, including distance visual acuity, near visual acuity, central 3 mm corneal equivalent spherical refraction (SE), anterior corneal Q-value, amplitude of accommodation (AMP), monocular positive relative accommodation (PRA), and lens thickness (LT).The differences of different indicators (AMP, PRA, LT) related to accommodation power at different time points were compared between the two groups.The correlation between near visual acuity, PRA, LT and AMP in the optimized adjustment group was analyzed.This study adhered to the Declaration of Helsinki.The study protocol was approved by the Ethics Committee of LanZhou Bright Eye Hospital (No.LP-202302).Written informed consent was obtained from each subject.Results:There were significant overall differences in distance visual acuity, near visual acuity, Q value, and SE in non-dominant eyes at different time points in the two groups ( Ftime=244.98, 69, 843.65, 454.80; all P<0.05).There was a significant overall difference in Q value between the two groups ( Fgroup=2.49, P<0.05).Compared with preoperative, the postoperative distance visual acuity and near visual acuity of patients in both groups were improved, with an increase in Q value and a decrease in SE, with statistically significant differences (all P<0.05).One year after surgery, the near visual acuity and Q value of the optimized adjustment group were better than those of the control group, with statistically significant differences (both P<0.05).There were significant differences in AMP and PRA in non-dominant eyes at different time points between the two groups (AMP: Fgroup=19.03, P<0.05; Ftime=4.99, P<0.05.PRA: Fgroup=42.67, P<0.05; Ftime=28.64, P<0.05).At 1 month, 6 months and 1 year after surgery, AMP and PRA increased in the optimization adjustment group compared with the preoperative and control group, and the differences were statistically significant (all P<0.05).There were significant differences in LT before and after surgery in non-dominant eye between the two groups ( Ftime=4.94, P<0.05).Among them, the LT in the optimization adjustment group became thinner at different time points after surgery than before surgery, and the differences were statistically significant (all P<0.05).In the optimization adjustment group, near visual acuity at 1 year after surgery in non-dominant eye was positively correlated with AMP ( r=0.269, P<0.05), and PRA was negatively correlated with AMP ( r=-0.601, P<0.05), and preoperative LT was positively correlated with preoperative AMP ( r=0.276, P<0.05). Conclusions:Optimized monovision LASIK for the treatment of age-related accommodative insufficiency with myopia significantly improves the myopia-related accommodation parameters AMP, PRA, and LT.

Objective To investigate the effect and mechanism of Pollen Typhae extract on renal tubular epithelial cell(HK-2)injury induced by high glucose through microRNA-148b-3p(miR-148b-3p)/nuclear factor-κB(NF-κB)pathway.Methods HK-2 cells were divided into normal control(NC)group,hypertonic control(MG)group,high glucose(HG)group,low dose Pollen Typhae extract(HG+L)group,medium dose Pollen Typhae extract(HG+M)group,high dose Pollen Typhae extract(HG+H)group,HG+inhibitor negative control(anti-miR-NC)group,HG+miR-148b-3p inhibitor(HG+anti-miR-148b-3p)group,HG+mimic negative control+high-dose Pollen Typhae extract(HG+miR-NC+H)group,HG+miR-148b-3p mimic+high-dose Pollen Typhae extract(HG+miR-148b-3p+H)group.Apoptosis rate was detected by flow cytometry.The levels of tumor necrosis factor α(TNF-α)and interleukin 6(IL-6)were detected by ELISA.Malondialdehyde(MDA)level and superoxide dismutase(SOD)activity were detected by spectrophotometry.Reactive oxygen species(ROS)level was detected by DCFH-DA method.qRT-PCR was used to detect the expression of miR-148b-3p.The expression of phosphorylated nuclear factor κB P65(p-p65)and phosphorylated NF-κB inhibitor(p-IκBα)proteins were detected by western blot.Results Compared with NC group,the apoptosis rate,TNF-α and IL-6 secretion levels,MDA level,ROS level,miR-148b-3p expression,Cleaved caspase-3,p-p65,and p-IκBα protein expression were increased(P＜0.05),while cell survival rate and SOD level were decreased in HG group(P＜0.05).Compared with the HG group,HG+L,HG+M,HG+H group increased cell survival rate,SOD(P＜0.05),apoptosis rate,TNF-α,IL-6,MDA,ROS,miR-148b-3p expression,Cleaved-caspase-3,p-p65,p-IκBα protein expression decreased(P＜0.05).Compared with HG+anti-miR-NC group,the cell survival rate and SOD were increased(P＜0.05),while apoptosis rate,expression of TNF-α,IL-6,MDA,ROS,miR-148b-3p,Cleaved caspase-3,p-p65,and p-IκBα protein expression were decreased in HG+anti-miR-148b-3p group(P＜0.05).Compared with HG+miR-NC+H group,apoptosis rate,TNF-α,IL-6,MDA,ROS,miR-148b-3p expression,Cleaved caspase-3,p-p65,and p-IκBα protein expression were increased(P＜0.05),while cell survival rate and SOD were decreased in HG+miR-148b-3p+H group(P＜0.05).Conclusions Pollen Typhae extract can reduce renal tubular epithelial cell injury induced by high glucose through down-regulating the expression of miR-148b-3p and inhibiting NF-κB pathway.

Objective:To investigate the association between urinary arsenic, selenium, and chromium levels and the risk and predisposing factors of diabetes mellitus in people with previous endemic arsenic exposure.Methods:From September to December 2024, 240 residents in the drinking-water-borne endemic arsenic disease area in Hohhot were taken as the study subjects. They were divided into an exposed group ( n = 91) and a non-exposed group ( n = 149) based on whether they had suffered from arsenism in the past. The exposed group was further divided into diabetes and non diabetes subgroups ( n = 54, 37) based on the prevalence of diabetes, and the diabetes subgroup was further divided into type 1 and type 2 diabetes subgroups ( n = 23, 31) based on the type of diabetes. Questionnaire survey was conducted to investigate the basic situation, measure fasting blood glucose, and determine the levels of arsenic, selenium, and chromium in urine by inductively coupled plasma mass spectrometry. Multivariate logistic regression analysis was used to assess the risk factors of diabetes mellitus. Results:The difference of prevalence of diabetes mellitus was statistically significant between the exposed group and non-exposed group [59.3% (54/91) vs. 41.6% (62/149), χ2 = 7.11, P = 0.008]. The levels of urinary arsenic, selenium, and chromium in the exposed group were higher than those in the non-exposed group ( t = - 2.00, - 2.14, - 2.18, P < 0.05). There were statistically significant differences in urinary arsenic level, body mass index (BMI), the distribution of age, smoking status, and gender between the diabetes patients and non-diabetes patients in the exposed group ( t = 2.20, 3.57, χ2 = 10.76, 5.23, 4.01, P < 0.05). The difference of urinary arsenic levels were statistically significant between patients with type 1 diabetes and type 2 diabetes in the exposed group ( t = - 2.06, P = 0.048). Multivariate logistic regression analysis showed that BMI, age, sex, smoking, and urinary arsenic levels were risk factors for diabetes ( P < 0.05). For every 1-unit increase in urinary arsenic, fasting blood glucose levels increased by 0.057 times (95% CI: 0.018 - 0.103, P = 0.029). Conclusions:There is a significant correlation between the urine arsenic level of people with previous endemic arsenic exposure and diabetes, especially type 2 diabetes. Men, smoking, overweight, age ≥65 years, and high urinary arsenic level are risk factors for diabetes.