[Metformin use and risk of ischemic stroke in patients with type 2 diabetes: A cohort study].

OBJECTIVE: To explore the association between the use of metformin and the risk of ischemic stroke in patients with type 2 diabetes. METHODS: A prospective cohort study was designed from the Fangshan family cohort in Beijing. According to metformin use at baseline, 2 625 patients with type 2 diabetes in Fangshan, Beijing were divided into metformin group or non-metformin group and the incidence of ischemic stroke between the different groups during follow-up was estimated and compared by Cox proportional hazard regression model. The participants with metformin were first compared with all the parti-cipants who did not use metformin, and then were further compared with those who did not use hypoglycemic agents and those who used other hypoglycemic agents. RESULTS: The patients with type 2 diabetes were with an average age of (59.5±8.7) years, and 41.9% of them were male. The median follow-up time was 4.5 years. A total of 84 patients developed ischemic stroke during follow-up, with a crude incidence of 6.4 (95%CI: 5.0-7.7) per 1 000 person-years. Among all the participants, 1 149 (43.8%) took metformin, 1 476 (56.2%) were metformin non-users, including 593 (22.6%) used other hypoglycemic agents, and 883 (33.6%) did not use any hypoglycemic agents. Compared with metformin non-users, the Hazard ratio (HR) for ischemic stroke in metformin users was 0.58 (95%CI: 0.36-0.93; P = 0.024). Compared with other hypoglycemic agents, HR was 0.48 (95%CI: 0.28-0.84; P < 0.01); Compared with the group without hypoglycemic agents, HR was 0.65 (95%CI: 0.37-1.13; P=0.13). The association between metformin and ischemic stroke was statistically significant in the patients ≥ 60 years old compared with all the metformin non-users and those who used other hypoglycemic agents (HR: 0.48, 95%CI: 0.25-0.92; P < 0.05). Metformin use was associated with a lower incidence of ischemic stroke in the patients with good glycemic control (0.32, 95%CI: 0.13-0.77; P < 0.05). In the patients with poor glycemic control, and the association was not statistically significant (HR: 0.97, 95%CI: 0.53-1.79; P>0.05). There was an interaction between glycemic control and metformin use on incidence of ischemic stroke (Pinteraction < 0.05). The results of the sensitivity analysis were consistent with the results in the main analysis. CONCLUSION: Among patients with type 2 diabetic in rural areas of northern China, metformin use was associated with lower incidence of ischemic stroke, especially in patients older than 60 years. There was an interaction between glycemic control and metformin use in the incidence of ischemic stroke.

[Genotype-environment interaction on arterial stiffness: A pedigree-based study].

OBJECTIVE: To utilized the baseline data of the Beijing Fangshan Family Cohort Study, and to estimate whether the association between a healthy lifestyle and arterial stiffness might be modified by genetic effects. METHODS: Probands and their relatives from 9 rural areas in Fangshan district, Beijing were included in this study. We developed a healthy lifestyle score based on five lifestyle behaviors: smoking, alcohol consumption, body mass index (BMI), dietary pattern, and physical activity. The measurements of arterial stiffness were brachial-ankle pulse wave velocity (baPWV) and ankle-brachial index (ABI). A variance component model was used to determine the heritability of arterial stiffness. Genotype-environment interaction effects were performed by the maximum likelihood methods. Subsequently, 45 candidate single nucleotide polymorphisms (SNPs) located in the glycolipid metabolism pathway were selected, and generalized estimated equations were used to assess the gene-environment interaction effects between particular genetic loci and healthy lifestyles. RESULTS: A total of 6 302 study subjects across 3 225 pedigrees were enrolled in this study, with a mean age of 56.9 years and 45.1% male. Heritability of baPWV and ABI was 0.360 (95%CI: 0.302-0.418) and 0.243 (95%CI: 0.175-0.311), respectively. Significant genotype-healthy diet interaction on baPWV and genotype-BMI interaction on ABI were observed. Following the findings of genotype-environment interaction analysis, we further identified two SNPs located in ADAMTS9-AS2 and CDH13 might modify the association between healthy dietary pattern and arterial stiffness, indicating that adherence to a healthy dietary pattern might attenuate the genetic risk on arterial stiffness. Three SNPs in CDKAL1, ATP8B2 and SLC30A8 were shown to interact with BMI, implying that maintaining BMI within a healthy range might decrease the genetic risk of arterial stiffness. CONCLUSION: The current study discovered that genotype-healthy dietary pattern and genotype-BMI interactions might affect the risk of arterial stiffness. Furthermore, we identified five genetic loci that might modify the relationship between healthy dietary pattern and BMI with arterial stiffness. Our findings suggested that a healthy lifestyle may reduce the genetic risk of arterial stiffness. This study has laid the groundwork for future research exploring mechanisms of arterial stiffness.

[Revision rate of periprosthetic joint infection post total hip or knee arthroplasty of 34 hospitals in China between 2015 and 2017: a multi-center survey].

Objective: To investigate the rate of periprosthetic joint infection (PJI) revision surgeries and clinical information of hip-/knee- PJI cases nationwide from 2015 to 2017 in China. Methods: An epidemiological investigation. A self-designed questionnaire and convenience sampling were used to survey 41 regional joint replacement centers nationwide from November 2018 to December 2019 in China. The PJI was diagnosed according to the Musculoskeletal Infection Association criteria. Data of PJI patients were obtained by searching the inpatient database of each hospital. Questionnaire entries were extracted from the clinical records by specialist. Then the differences in rate of PJI revision surgery between hip- and knee- PJI revision cases were calculated and compared. Results: Total of 36 hospitals (87.8%) nationwide reported data on 99 791 hip and knee arthroplasties performed from 2015 to 2017, with 946 revisions due to PJI (0.96%). The overall hip-PJI revision rate was 0.99% (481/48 574), and it was 0.97% (135/13 963), 0.97% (153/15 730) and 1.07% (193/17 881) in of 2015, 2016, 2017, respectively. The overall knee-PJI revision rate was 0.91% (465/51 271), and it was 0.90% (131/14 650), 0.88% (155/17 693) and 0.94% (179/18 982) in 2015, 2016, 2017, respectively. Heilongjiang (2.2%, 40/1 805), Fujian (2.2%, 45/2 017), Jiangsu (2.1%, 85/3 899), Gansu (2.1%, 29/1 377), Chongqing (1.8%, 64/3 523) reported relatively high revision rates. Conclusions: The overall PJI revision rate in 34 hospitals nationwide from 2015 to 2017 is 0.96%. The hip-PJI revision rate is slightly higher than that in the knee-PJI. There are differences in revision rates among hospitals in different regions.

[Effect of laparoscopic ovarian cystectomy on ovarian reserve and the outcomes of in vitro fertilization patients with benign ovarian cysts].

Objective: To investigate the effect of laparoscopic ovarian cystectomy on anti-Mullerian hormone (AMH) level, ovarian response to gonadotropin stimulation and pregnancy rate for in vitro fertilization (IVF) patients with benign ovarian cysts. Methods: Patients with benign ovarian cysts who were admitted for cystectomy and had undergone IVF treatment were enrolled in the study. There were 373 participants with ovarian cysts underwent laparoscopic ovarian cystectomy in the experimental group. According to duration of post-surgery, there were four sub-groups: 1 year post-surgery (1Y POST), 2 years post-surgery (2Y POST), 4 years post-surgery (4Y POST) and ≥5 years post-surgery (≥5Y POST) in the experimental group. According to histopathologic types of ovarian cysts, there were two sub-groups: ovarian endometriotic cysts and ovarian non-endometriotic cysts. Two hundreds and three patients with no history of ovarian cysts and ovarian surgery were in the control group. The level of AMH and basic concentrations of follicle stimulating hormone (FSH), luteinizing hormone (LH), progesterone (P), estradiol (E2) were measured. Antral follicle counts (AFC) were calculated. There were other study variables: total dose of gonadotropins, duration of ovarian stimulation, the number of oocyte retrieved, the number of embryo obtained, blastocyst transfer rate and pregnancy rate. Results: The control group was matched as closely as possible to the experimental group, including age, body mass index and menstrual cycle (all P>0.05). Compared to the women in control group, the women in ovarian endometriotic cystectomy sub-group had significantly higher levels of basal FSH and basal P, lower level of AMH (all P<0.05); the women in ovarian endometriotic cysts sub-group had significantly higher dose of gonadotropins (all P<0.05); the women in ovarian endometriotic cysts ≥5Y POST sub-group had significantly lower number of oocyte retrieved, lower number of embryo obtained, lower blastocyst transfer rate, and lower pregnancy rate (all P<0.05). Compared to the women in control group, the women in ovarian non-endometriotic cysts sub-group had a significantly higher level of basal FSH and basal P (all P<0.05). The women in ovarian non-endometriotic cysts sub-group had lower level of AMH, higher dose of gonadotropins, lower number of oocyte retrieved, lower number of embryo obtained, lower rate of blastocyst transfer and lower rate of pregnancy than the control group but there were no statistically significant differences among them (all P>0.05). The women with unilateral ovarian endometriotic cysts had significantly lower number of oocyte retrieved on the side of surgery than another side (P<0.05). Conclusions: In short term laparoscopic ovarian cystectomy has no significant effect on ovarian reserve. But with long-term follow-up ovarian reserve, ovarian response to gonadotropin stimulation and pregnancy rate are decreased. The effect of laparoscopic ovarian cystectomy in benign cysts on ovarian is associated with whether or not it is the surgical side.

[Effect of the varicella vaccination on the clinical characteristics of herpes zoster cases aged 20 years and under].

To discuss the effect of varicella vaccination on the clinical characteristics of herpes zoster (shingles) cases aged 20 years and under, and analyze its clinical features. Based on the Yichang Health Big Data Platform, a descriptive study was conducted to collect the information of cases aged 20 years and under in three medical institutions of Yichang Central People's Hospital, Yichang First People's Hospital and Yichang Second People's Hospital from March 2019 to September 2020. According to the history of varicella vaccine, cases were divided into vaccination group and non-vaccination group, and their clinical features and outcomes were compared. The results showed that 46 shingles cases, aged from 7 to 20 years old, were included in this study. 26 males (56.5%), 20 females (43.5%), 15 cases in vaccination group (32.6%) and 31 cases in non-vaccination group (67.4%). 28 cases had thoracic involvement, followed by lumbar (n=8), cranial (n=7) involvements and extremities (n=7). The spread of herpes skin area: 2 cases involved too large area, 21 cases of 10 cm×10 cm, 14 cases of 5 cm×5 cm, 9 cases of 1 cm×1 cm. Herpes number: 26 cases had 10-49 herpes, followed by <10 herpes (n=9), uncountable herpes (n=7) and 50-99 herpes (n=4). The clinical course[M(Q1,Q3)] lasted 20.5 (13.5,24.8) d averagely, 5 cases had postherpetic neuralgia (PHN) and 1 case had respiratory complications. Shingles decrustation time was significantly shorter in vaccination group (Z=-2.01, P<0.05), and there was no significant difference in other characteristics by vaccination. In conclusion, the number and spread of shingles in most children and adolescents are less, and the complications such as PHN are less. Varicella vaccination can reduce the decrustation time and relieve shingles cases with some clinical symptoms.

[Analysis of big data characteristics of allergic rhinitis patients in Beijing City from 2016 to 2021].

To explore the characteristics of big data of patients with allergic rhinitis, including the time, population and spatial distribution of allergic rhinitis in Beijing from 2016 to 2021, so as to provide reference for the prevention and treatment of this disease. Descriptive epidemiological methods were used to analyze the distribution (including gender, age and location)and trend of allergic rhinitis patients in 30 pilot hospitals from January 2016 to December 2021, T test and Kruskal-Wallis rank sum test were used to test the statistical differences. The results showed that the number of patients with allergic rhinitis in 30 hospitals increased year by year from 2016 to 2019, with an increase of 97.9%. In 2020, the number of patients decreased. In 2021, the number of visits returned to the pre-epidemic level (461 332); The number of patients with allergic rhinitis was the highest in September, with a seasonal index of 177.6%, while the lowest number was in February, accounting for only 47.2%; a significant difference was observed in the number of patients in different age groups(H=45 319.48, P<0.05), and patients under 15 years old accounted for the highest proportion(819 284 visits); There were significant differences between patients of different genders in the 45-59 year old group (t=-4.26, P<0.05).There were relatively more patients with allergic rhinitis in Dongcheng District(31.1%) than in Huairou District and Miyun District (0.4%). In conclusion, since 2016, the number of patients increased significantly, with a varied trend in different seasons. Most patients were children. There were more patients in the central urban area than in the outer suburbs.

[Epidemiological survey of hepatitis B and analysis of hepatitis B vaccine coverage rate among children aged 1-14 years in Lhasa in 2006, 2014 and 2020].

In 2006, 2014 and 2020, the positive rates of HBsAg in 560, 384 and 402 children aged 1 to 14 years were 4.5%, 2.6% and 2.5%, respectively, with no statistically significant differences (P>0.05). The positive rate of anti-HBs was highest in 2014 (57.8%) and lowest in 2006 (34.1%) (P<0.05). The positive rate of anti-HBc was highest in 2006 (15.7%), and decreased in 2014 (7.8%) and 2020 (5.7%) (P<0.001). The timely rate of the first dose of hepatitis B vaccine for children in Lhasa in 2006, 2014 and 2020 was 7.7% (43/560), 50.3% (193/384) and 94.8% (381/402), respectively. The overall vaccination rates were 15.4% (86/560), 35.2% (135/384) and 96.0% (386/402), respectively, showing a trend of gradual increases (χtrend values were 718.63 and 589.59, both P values<0.001).

[Incidence and risk factors of ischemic stroke in patients with type 2 diabetes among urban workers in Beijing, China].

OBJECTIVE: To explore the incidence of ischemic stroke after the onset of type 2 diabetes, and further analyze the risk factors, so as to provide a basis for further research. METHODS: The data were obtained from the database of the Beijing Urban Employee Basic Medical Insurance Database. The study used a prospective design to describe the incidence of ischemic stroke in patients with type 2 diabetes. In our study, these patients were followed up for seven years. Multivariate Logistic regression models were used to analyze the risk factors of ischemic stroke in patients with type 2 diabetes. RESULTS: A total of 185 813 newly diagnosed type 2 diabetes patients were enrolled, with an average age of (58.5±13.2) years, and 49.0% of them were males. A total of 10 393 patients with newly diagnosed ischemic stroke occurred in 7 years, with a cumulative incidence of 5.6% and an incidence density of 8.1/1 000 person-years. Ischemic stroke occurred in all age groups in patients with type 2 diabetes. The cumulative incidence was 1.5% (95%CI: 1.3%-1.6%) in group ≤44 years old, 3.6% (95%CI: 3.4%-3.7%) in group 45-54 years old, 5.4% (95%CI: 5.2%-5.5%) in group 55-64 years old, and 9.2% (95%CI: 9.0%-9.4%) in group ≥65 years old, and the cumulative incidence increased with age (P < 0.05). Cumulative incidence rate of the males (6.8%, 95%CI: 6.7%-7.0%) was higher than the females (4.4%, 95%CI: 4.3%-4.6%). Among the patients < 80 years old, the cumulative incidence rate of the males was higher than that of the females in all the age groups. In the patients ≥80 years of age, the cumulative incidence was higher in the females (9.2%) than in the males (7.9%). Further analysis revealed that complications, such as coronary heart disease (OR=3.18, 95%CI: 2.72-3.72), heart failure (OR=1.53, 95%CI: 1.32-1.79) and kidney failure (OR=1.45, 95%CI: 1.20-1.75) were associated with ischemic stroke in the patients with type 2 diabetes. CONCLUSION: The incidence level of ischemic stroke in patients with type 2 diabetes is high. It is necessary to strengthen the management of risk factors in elderly patients, screen the complications of type 2 diabetes as early as possible, and take active preventive and control measures.

[Association study between haplotypes of WNT signaling pathway genes and nonsyndromic oral clefts among Chinese Han populations].

OBJECTIVE: To explore whether WNT signaling pathway genes were associated with non-syndromic oral clefts (NSOC) based on haplotypes analyses among 1 008 Chinese NSOC case-parent trios. METHODS: The genome-wide association study (GWAS) data of 806 Chinese non-syndromic cleft lip with or without cleft palate (NSCL/P) trios and 202 Chinese non-syndromic cleft palate (NSCP) case-parent trios were drawn from the International Consortium to Identify Genes and Interactions Controlling Oral Clefts (ICOCs) study GWAS data set, whose Chinese study population were recruited from four provinces in China, namely Taiwan, Shandong, Hubei, and Sichuan provinces. The process of DNA genotyping was conducted by the Center for Inherited Disease Research in the Johns Hopkins University, using Illumina Human610-Quad v.1_B Bead Chip. The method of sliding windows was used to determine the haplotypes for analyses, including 2 SNPs haplotypes and 3 SNPs haplotypes. Haplotypes with a frequency lower than 1% were excluded for further analyses. To further assess the association between haplotypes and NSOC risks, and the transmission disequilibrium test (TDT) was performed. The Bonferroni method was adopted to correct multiple tests in the study, with which the threshold of statistical significance level was set as P < 0.05 divided by the number of tests, e.g P < 3.47×10-4 in the current stu-dy. All the statistical analyses were performed by using plink (v1.07). RESULTS: After quality control, a total of 144 single nucleotide polymorphisms (SNPs) mapped in seven genes in WNT signaling pathway were included for the analyses among the 806 Chinese NSCL/P trios and 202 Chinese NSCP trios. A total of 1 042 haplotypes with frequency higher than 1% were included for NSCL/P analyses and another 1 057 haplotypes with frequency higher than 1% were included for NSCP analyses. Results from the TDT analyses showed that a total of 69 haplotypes were nominally associated with the NSCL/P risk among Chinese (P < 0.05). Another 34 haplotypes showed nominal significant association with the NSCP risk among Chinese (P < 0.05). However, none of these haplotypes reached pre-defined statistical significance level after Bonferroni correction (P>3.47×10-4). CONCLUSION: This study failed to observe any statistically significant associations between haplotypes of seven WNT signaling pathway genes and the risk of NSOC among Chinese. Further studies are warranted to replicate the findings here.

[Interaction between ischemic stroke risk loci identified by genome-wide association studies and sleep habits].

OBJECTIVE: To explore the relationship between sleep habits (sleep duration, sleep efficiency, sleep onset timing) and ischemic stroke, and whether there is an interaction between sleep habits and ischemic stroke susceptibility gene loci. METHODS: A questionnaire survey, physical examination, blood biochemical testing and genotyping were conducted among rural residents in Beijing, and the gene loci of ischemic stroke suggested by previous genome-wide association studies (GWAS) were screened. Multivariable generalized linear model was used to analyze the correlation between sleep habits, sleep-gene interaction and ischemic stroke. RESULTS: A total of 4 648 subjects with an average age of (58.5±8.7) years were enrolled, including 1 316 patients with ischemic stroke. Compared with non-stroke patients, stroke patients with sleep duration ≥9 hours, sleep efficiency < 80%, and sleep onset timing earlier than 22:00 accounted for a higher proportion (P < 0.05). There was no significant association between sleep duration and risk of ischemic stroke (OR=1.04, 95%CI: 0.99-1.10, P=0.085). Sleep efficiency was inversely associated with the risk of ischemic stroke (OR=0.18, 95%CI: 0.06-0.53, P=0.002). The risk of ischemic stroke in the subjects with sleep efficiency < 80% was 1.47-fold (95%CI: 1.03-2.10, P=0.033) of that in the subjects with sleep efficiency ≥80%. Falling asleep earlier than 22:00 was associated with 1.26 times greater risk of stroke than falling asleep between 22:00 and 22:59 (95%CI: 1.04-1.52, P=0.017). Multifactorial adjustment model showed that rs579459 on ABO gene had an interaction with sleep time (P for interaction =0.040). When there were two T alleles for rs579459 on the ABO gene, those who fell asleep before 22:00 had 1.56 times (95%CI: 1.20-2.04, P=0.001) the risk of stroke compared with those who fell asleep between 22:00 and 22:59, and there was no significant difference when the number of pathogenic alleles was 0 or 1. In the model adjusted only for gender, age and family structure, sleep duration and the number of T allele rs2634074 on PITX2 gene had an interaction with ischemic stroke (P for interaction=0.033). CONCLUSION: Decreased sleep efficiency is associated with increased risk of ischemic stroke, and falling asleep earlier than 22:00 is associated with higher risk of ischemic stroke. Sleep onset timing interacted with rs579459 in ABO gene and the risk of ischemic stroke. Sleep duration and PITX2 rs2634074 may have a potential interaction with ischemic stroke risk.

[Exploring the association between de novo mutations and non-syndromic cleft lip with or without palate based on whole exome sequencing of case-parent trios].

OBJECTIVE: To explore the association between de novo mutations (DNM) and non-syndromic cleft lip with or without palate (NSCL/P) using case-parent trio design. METHODS: Whole-exome sequencing was conducted for twenty-two NSCL/P trios and Genome Analysis ToolKit (GATK) was used to identify DNM by comparing the alleles of the cases and their parents. Information of predictable functions was annotated to the locus with SnpEff. Enrichment analysis for DNM was conducted to test the difference between the actual number and the expected number of DNM, and to explore whether there were genes with more DNM than expected. NSCL/P-related genes indicated by previous studies with solid evidence were selected by literature reviewing. Protein-protein interactions analysis was conducted among the genes with protein-altering DNM and NSCL/P-related genes. R package "denovolyzeR" was used for the enrichment analysis (Bonferroni correction: P=0.05/n, n is the number of genes in the whole genome range). Protein-protein interactions among genes with DNM and genes with solid evidence on the risk factors of NSCL/P were predicted depending on the information provided by STRING database. RESULTS: A total of 339 908 SNPs were qualified for the subsequent analysis after quality control. The number of high confident DNM identified by GATK was 345. Among those DNM, forty-four DNM were missense mutations, one DNM was nonsense mutation, two DNM were splicing site mutations, twenty DNM were synonymous mutations and others were located in intron or intergenic regions. The results of enrichment analysis showed that the number of protein-altering DNM on the exome regions was larger than expected (P < 0.05), and five genes (KRTCAP2, HMCN2, ANKRD36C, ADGRL2 and DIPK2A) had more DNM than expected (P < 0.05/(2×19 618)). Protein-protein interaction analysis was conducted among forty-six genes with protein-altering DNM and thirteen genes associated with NSCL/P selected by literature reviewing. Six pairs of interactions occurred between the genes with DNM and known NSCL/P-related genes. The score measuring the confidence level of the predicted interaction between RGPD4 and SUMO1 was 0.868, which was higher than the scores for other pairs of genes. CONCLUSION: Our study provided novel insights into the development of NSCL/P and demonstrated that functional analyses of genes carrying DNM were warranted to understand the genetic architecture of complex diseases.

[The relationship between triglyceride glucose index and risk of cardiovascular disease among Kazakh and Uygur population in Xinjiang: a retrospective cohort analysis].

Objective: To analyze whether triglyceride-glucose (TyG) index is associated with increased risk of cardiovascular diseases (CVD) and the value of TyG index in predicting CVD risk among Kazakh and Uighur population of Xinjiang. Methods: In this study, 5 375 Kazakh and Uygur people of Xinyuan county and Jiashi county were selected as the research objects. Subjects were divided into four groups based on the quartile of the TyG index level. Cox regression model was used to analyze the association between TyG index with the risk of CVD. The dose-response relationship between TyG index and CVD risk was described by restricted cubic splines. The area under the receiver operating characteristic curve, net reclassification improvement (NRI), and integrated discrimination improvement (IDI) were used to estimate the value of TyG index for predicting CVD. Mediating effect analysis was conducted to analyze the mediating effect of TyG index in the association between body mass index and CVD. Results: The age of subjects was 41.06（30.11，53.00）years old, with 46.30%（2 489/5 375）was male. After multivariate adjustment, there was an increasing trend between the risk of CVD and the higher TyG index Ptrend<0.001, compared with subjects of TyG index in Q1, the HR (95%CI) of Q2, Q3, and Q4 groups was 1.53, 1.23 and 1.73, respectively. Restricted cubic splines showed that TyG index was the linearly associated with the risk of CVD. TyG index could improve the prediction ability of Framingham model for the risk of CVD (NRI=0.106,P=0.010; IDI=0.003,P=0.030). The mediating effect analysis showed that in the relationship between body mass index and CVD, the TyG index had a mediating effect (P<0.001), and the ratio of mediating effect was 12.69%. Conclusion: TyG index is an independent predictor of CVD risk among kazakh and Uygur population in Xinjiang and has a good predictive value for the risk of CVD.

[Prediction of intensive care unit readmission for critically ill patients based on ensemble learning].

OBJECTIVE: To develop machine learning models for predicting intensive care unit (ICU) readmission using ensemble learning algorithms. METHODS: A publicly accessible American ICU database, medical information mart for intensive care (MIMIC)-Ⅲ as the data source was used, and the patients were selected by the inclusion and exclusion criteria. A set of variables that had the predictive ability of outcome including demographics, vital signs, laboratory tests, and comorbidities of patients were extracted from the dataset. We built the ICU readmission prediction models based on ensemble learning methods including random forest, adaptive boosting (AdaBoost), and gradient boosting decision tree (GBDT), and compared the prediction performance of the machine learning models with a conventional Logistic regression model. Five-fold cross validation was used to train and validate the prediction models. Average sensitivity, positive prediction value, negative prediction value, false positive rate, false negative rate, area under the receiver operating characteristic curve (AUROC) and Brier score were used as performance measures. After constructing the prediction models, top 10 predictive variables based on importance ranking were identified by the model with the best discrimination. RESULTS: Among these ICU readmission prediction models, GBDT (AUROC=0.858) had better performance than random forest (AUROC=0.827), and was slightly superior to AdaBoost (AUROC=0.851) in terms of AUROC. Compared with Logistic regression (AUROC=0.810), the discrimination of the three ensemble learning models was much better. The feature importance provided by GBDT showed that the top ranking variables included vital signs and laboratory tests. The patients with ICU readmission had higher mean arterial pressure, systolic blood pressure, diastolic blood pressure, and heart rate than the patients without ICU readmission. Meanwhile, the patients readmitted to ICU experienced lower urine output and higher serum creatinine. Overall, the patients having repeated admissions during their hospitalization showed worse heart function and renal function compared with the patients without ICU readmission. CONCLUSION: The ensemble learning based ICU readmission prediction models had better performance than Logistic regression model. Such ensemble learning models have the potential to aid ICU physicians in identifying those patients with high risk of ICU readmission and thus help improve overall clinical outcomes.

[Predicting prolonged length of intensive care unit stay via machine learning].

OBJECTIVE: To construct length of intensive care unit (ICU) stay (LOS-ICU) prediction models for ICU patients, based on three machine learning models support vector machine (SVM), classification and regression tree (CART), and random forest (RF), and to compare the prediction perfor-mance of the three machine learning models with the customized simplified acute physiology score Ⅱ(SAPS-Ⅱ) model. METHODS: We used medical information mart for intensive care (MIMIC)-Ⅲ database for model development and validation. The primary outcome was prolonged LOS-ICU(pLOS-ICU), defined as longer than the third quartile of patients' LOS-ICU in the studied dataset. The recursive feature elimination method was used to do feature selection for three machine learning models. We utilized 5-fold cross validation to evaluate model prediction performance. The Brier value, area under the receiver operation characteristic curve (AUROC), and estimated calibration index (ECI) were used as perfor-mance measures. Performances of the four models were compared, and performance differences between the models were assessed using two-sided t test. The model with the best prediction performance was employed to generate variable importance ranking, and the identified top five important predictors were pre-sented. RESULTS: The final cohort in our study consisted of 40 200 eligible ICU patients, of whom 23.7% were with pLOS-ICU. The proportion of the male patients was 57.6%, and the age of all the ICU patients was (61.9±16.5) years.Results showed that the three machine learning models outperformed the customized SAPS-Ⅱ model in terms of all the performance measures with statistical significance (P < 0.01). Among the three machine learning models, the RF model achieved the best overall performance (Brier value, 0.145), discrimination (AUROC, 0.770) and calibration (ECI, 7.259). The calibration curve showed that the RF model slightly overestimated the risk of pLOS-ICU in high-risk ICU patients, but underestimated the risk of pLOS-ICU in low-risk ICU patients. Top five important predictors for pLOS-ICU identified by the RF model included age, heart rate, systolic blood pressure, body tempe-rature, and ratio of arterial oxygen tension to the fraction of inspired oxygen(PaO2/FiO2). CONCLUSION: The RF algorithm-based pLOS-ICU prediction model had a best prediction performance in this study. It lays a foundation for future application of the RF-based pLOS-ICU prediction model in ICU clinical practice.

[Prevalence and risk factors of osteoarthritis in patients with type 2 diabetes in Beijing, China from 2015 to 2017].

OBJECTIVE: To explore the prevalence and related factors of osteoarthritis in patients with type 2 diabetes mellitus, and provided a scientific basis for the prevention of the comorbidity. METHODS: The data were obtained from the database of all designated medical institutions in Beijing from 2015 to 2017. Data of the adult patients with type 2 diabetes mellitus were collected for descriptive analysis, and a Logistic regression model was used to explore the related factors of osteoarthritis in the patients with type 2 diabetes mellitus. RESULTS: A total of 1 046 264 diagnosed type 2 diabetes mellitus adult patients were included in our study, with an average age of 63.07 years, and 50.78% were males. Among the patients with type 2 diabetes mellitus, there were 341 561 cases with osteoarthritis, and the prevalence of osteoarthritis was 32.65%. The prevalence of females (38.05%) was higher than that of males (27.41%), and the difference was statistically significant (P < 0.05). Osteoarthritis occurred in all age groups among the patients with type 2 diabetes mellitus, with the highest prevalence of osteoarthritis in the age group of 65-69 years (36.76%), and the lowest prevalence in the age group ≤44 years (14.3%). Before the age of 70, the prevalence increased with age. Further analysis of related factors for osteoarthritis in the patients with type 2 diabetes mellitus showed that female (OR=1.62, 95%CI: 1.61-1.63), age (OR=1.01, 95%CI: 1.01-1.01), had other comorbidities (OR=1.19, 95%CI: 1.18-1.21), used hypoglycemic drugs (OR=0.79, 95%CI: 0.78-0.80), having the cardiovascular disease (OR=1.13, 95%CI: 1.11-1.15), having cerebrovascular disease (OR=1.25, 95%CI: 1.23-1.28), and having nephropathy (OR=1.61, 95%CI: 1.51-1.71) were associated with the osteoarthritis in the type 2 diabetic mellitus patients. CONCLUSION: Our study revealed that the prevalence of osteoarthritis in patients with type 2 diabetes mellitus is high in Beijing area. Health education and disease monitoring should be strengthened in middle-aged and elderly patients. Screening for comorbidities should be carried out as soon as possible, with the focus on menopausal women.

[Association analysis of famine exposure during early life and risk of hypertension in adulthood].

Objective: To explore the relationship between exposure to famine in early life and the risk of hypertension in adulthood. Methods: The medical data of Yichang Health Management Big Data Center from 2018 to 2019 were analyzed. A retrospective cohort study design was adopted, with hypertension as the study outcome, and different life periods exposed to the Great Famine in China were divided into groups. Multivariate logistic regression model was used to analyze the relationship between famine exposure in early life and hypertension in adulthood. At the same time, the interaction between gender and famine exposure was analyzed. Results: The age of 142 016 subjects was (60. 56±4.43). Among them, men accounted for 46.36% (65 845/142 016) and women accounted for 53.64% (76 171/142 016). There are 42 575(29.98%), 19 644(13.83%), 28 405(20.00%), 28 305(19.93%), 23 087 (19.93%) in non-famine exposure group, fetal famine exposure group, early childhood famine exposure group and late childhood famine exposure group, respectively. The prevalence of hypertension was 17.57% (24 947 cases). Multivariate logistic regression model analysis showed that after adjusting for related confounding factors, compared with non-famine exposure group, the risk of hypertension in fetal, early childhood, middle childhood and late childhood famine exposure group was higher and the OR (95%CI) values were 1.16 (1.11-1.22), 1.27 (1.21-1.33), 1.54 (1.47-1.60) and 1.84 (1.76-1.92), respectively. There was an interaction between sex and famine exposure group (P<0.001). The above association is stronger among women than among men. Conclusion: Famine exposure in early life may increase the risk of hypertension in adulthood, and the risk of women is greater than that of men.

[Study on the incidence of adult herpes zoster in Yichang city and its association with early-life famine exposure].

Based on Yichang health big data platform, 850 608 patients from September 2018 to September 2019 were included in this study. According to the date of birth, the participants were divided into early childhood famine exposure group, fetal famine exposure group and non-famine exposure group. The incidence of adult herpes zoster (HZ) in Yichang city was analyzed, and the correlation between early life famine exposure and adult HZ was analyzed. In 2019, the crude incidence rate of adult HZ in Yichang was 6.83‰. The crude incidence rate of adult HZ in females (7.26‰) was higher than that in males (6.40‰). Compared with the non-famine exposure group, fetal famine exposure was associated with the incidence of adult HZ (OR=1.21; 95%CI: 1.01-1.45, P=0.041). After stratification by sex, fetal famine exposure was only found to be associated with the onset of adult HZ in females (OR=1.28, 95%CI:1.02-1.61, P=0.034).

[Programmed death ligand 1 expression and CD8+T lymphocyte infiltration in salivary gland lymphoepithelial carcinoma].

Objective: To study the expression of programmed death ligand-1 (PD-L1) in tumor cells and CD8+T lymphocytes in tumor infiltrating lymphocytes, and to analyze the correlation of PD-L1 expression with infiltration of CD8+T lymphocytes and clinicopathologic features in salivary gland lymphoepithelial carcinoma (LEC). Methods: Forty-two cases of primary salivary LECs and 21 cases of secondary salivary LECs were enrolled at the Department of Oral Pathology, Shanghai Ninth People's Hospital, Shanghai Jiaotong University between 2015 and 2017. The expression of Epstein-Barr (EB) virus, PD-L1 and CD8 was examined using chromogenic in situ hybridization (CISH) and immunohistochemistry (IHC) staining. The data were analyzed using SPSS 23.0 software package. Results: EB virus was detected in 61 cases (61/63, 96.8%), including 42 (42/42, 100%) primary LECs and 19 (19/21, 90.5%) secondary LECs. The PD-L1 positive rate (score ≥1) was 97.6% (41/42), and its high-expression rate (score ≥20) was 78.6% (33/42) in primary LECs. The PD-L1 positive rate (score ≥1) was 71.4% (15/21), and its high-expression rate (≥20) was 38.1% (8/21) in secondary LECs. However, the PD-L1 positive rate (score ≥1, P=0.004) and high-expression rate (score ≥20, P=0.001) in primary LECs were higher than those in secondary LECs. There was no difference in the infiltration degree of CD8+T lymphocytes between primary and secondary LECs. There was a significant correlation between the expression of PD-L1 and CD8 in primary LECs (P=0.001) and in secondary LECs (P=0.048), respectively. Conclusions: There is PD-L1 expression in primary and secondary salivary LECs, while the expression rate is higher in primary LECs than secondary LECs. The combination of PD-L1 expression and CD8+T lymphocytes' presence suggest that most LEC patients might be responsive to immunotherapy, and primary LECs might be more significantly responsive than secondary LECs.

[Application of the China-PAR stroke risk equations in a rural northern Chinese population].

OBJECTIVE: To validate five-year risk prediction models for stroke in a contemporary rural Northern Chinese population. METHODS: Totally 6 483 rural adults aged 40 to 79 years without cardiovascular diseases were enrolled at baseline between June and August 2010, and followed up through January 2017. Expected prediction risk using the China-PAR (prediction for atherosclerotic cardiovascular disease risk in China) stroke risk equations were compared with the new Framingham stroke risk profile (FSRP). The recalibrated models were applied by adjusting the five-year baseline survival rate and the mean score to our rural northern Chinese population, while keeping other coefficient parameters the same as the original models. Kaplan-Meier analysis was used to obtain the observed event (nonfatal or fatal stroke) rate for the five years, and the expected-observed ratios were calculated to evaluate overestimation or underestimation in the cohort. The models were assessed by discrimination C statistic, calibration χ2, and calibration charts and plots for illustration as well. RESULTS: Over an average of (5.83 ± 1.14) years of the follow-up in this validation cohort with 6 483 rural Chinese participants, 438 subjects deve-loped a first stroke event. Recalibrated China-PAR stroke risk equations and FSRP well-performed for predicting five-year stroke risk in men, and had C statistics of 0.709 (95%CI, 0.675 - 0.743) and 0.721 (95%CI, 0.688 - 0.754), with calibration χ2 values being 5.7 (P = 0.770) and 13.6 (P = 0.137), respectively. However, both China-PAR and FSRP overestimated stroke events by 11.6% and 30.0% in women, and had C statistics of 0.713 (95%CI, 0.684-0.743) and 0.710 (95%CI, 0.679-0.740), respectively. Calibration χ2 values in women were 12.5 (P = 0.188) for China-PAR and 24.0 (P = 0.004) for FSRP. In addition, the calibration charts and plots illustrated good agreement between the observations and the predictions only in the China-PAR stroke risk equations, especially for men. CONCLUSION: In this validation cohort of rural northern Chinese adults, the China-PAR models had better performance of five-year stroke risk prediction than the FSRP, indicating that recalibrated China-PAR stroke risk equations might be appropriate tools for risk assessment and primary prevention of stroke in China.

[Evaluating the effect of WNT pathway genes considering interactions on the risk of non-syndromic oral clefts among Chinese populations].

OBJECTIVE: In this study, we used genome-wide association study (GWAS) data to explore whether WNT pathway genes were associated with non-syndromic oral clefts (NSOC) considering gene-gene interaction and gene-environment interaction. METHODS: We conducted the analysis using 806 non-syndromic cleft lip with or without cleft palate (NSCL/P) case-parent trios and 202 non-syndromic cleft palate (NSCP) case-parent trios among Chinese populations selected from an international consortium established for a GWAS of non-syndromic oral clefts. Genotype data and maternal environmental exposures were collected through DNA samples and questionnaires. Conditional Logistic regression models were adopted to explore gene-gene interaction and gene-environment in teraction using trio package in R software. The threshold of significance level was set as 3.47×10-4 using Bonferroni correction. RESULTS: A total of 144 single nucleotide polymorphisms (SNPs) in seven genes passed the quality control process in NSCL/P trios and NSCP trios, respectively. Totally six pairs of SNPs interactions showed statistically significant SNP-SNP interaction (P < 3.47×10-4) after Bonferroni correction, which were rs7618735 (WNT5A) and rs10848543 (WNT5B), rs631948 (WNT11) and rs556874 (WNT5A), and rs631948 (WNT11) and rs472631 (WNT5A) among NSCL/P trios; rs589149 (WNT11) and rs4765834 (WNT5B), rs1402704 (WNT11) and rs358792 (WNT5A), and rs1402704 (WNT11) and rs358793 (WNT5A) among NSCP trios, respectively. In addition, no significant result was found for gene-environment interaction analysis in both of the NSCL/P trios and NSCP trios. CONCLUSION: Though this study failed to detect significant association based on gene-environment interactions of seven WNT pathway genes and the risk of NSOC, WNT pathway genes may influence the risk of NSOC through potential gene-gene interaction.