Data Management And Sharing System of China National Clinical Research Center for Neurological Diseases

Journal：CNS Neuroscience & Therapeutics

Publication type: Original Contributions

IF(IF): 7.035

PMID: 36152306 Link To Pubmed Full text Download

1

Statistical analysis Continuous variables will be expressed as mean± SD for normally distributed data and median [interquartile range] for non-normally distributed data. Categorical variables will be expressed as number (percentage). Table 1 We determined the normal FBG level 3.9-5.6 mmol/L (70-100mg/dl) as the reference group. In univariate analyses, normally distributed data will be compared using the ANOVA analysis and non-normally distributed data will be compared using Kruskal–Wallis test. Categorical variables will be compared using χ2 test, as appropriate. A multiple logistic regression analysis will be used to test the correlation between FBG and in-hospital mortality and conducted model 1. Continuous variables will be transformed into categorical variables (FBG: ＜3.9 mmol/L (70mg/dl); 3.9-5.6 mmol/L (70-100mg/dl) (REF) ; 5.7-6.9 mmol/L (100-125mg/dl)≥7 mmol/L (125mg/dl). HbA1c: ≥ 7.0% (REF.); ＜7.0%) before entering into the binary logistic regression analyses. Considering the multicollinearity, if the linear correlation coefficient≥0.7, Model 2 will be conducted to test the correlation between FBG and in-hospital mortality adjusted for the covariates with ASD ≥10 in the comparison of baseline characteristics apart from HbA1c, and Model 3 will be conducted to test the correlation between HbA1c and in-hospital mortality adjusted for the covariates with ASD ≥10 in the comparison of baseline characteristics apart from FBG. If the linear correlation coefficient＜0.7, only model 2 will be conducted to test the correlation between FBG and in-hospital mortality adjusted for the covariates with ASD ≥10 in the comparison of baseline characteristics. ORs and 95% CIs will be recommended to be presented. Next, Model 4 will be used to test the correlation between FBG and percentage of the evacuation of intracranial hematoma. Considering the multicollinearity, if the linear correlation coefficient≥0.7, Model 5 will be conducted to test the correlation between FBG and percentage of the evacuation of intracranial hematoma adjusted for the covariates with ASD ≥10 in the comparison of baseline characteristics apart from HbA1c, and Model 6 will be conducted to test the correlation between HbA1c and percentage of the evacuation of intracranial hematoma adjusted for the covariates with ASD ≥10 in the comparison of baseline characteristics apart from FBG. If the linear correlation coefficient＜0.7, only model 4 will be conducted to test the correlation between FBG and in-hospital mortality adjusted for the covariates with ASD ≥10 in the comparison of baseline characteristics. The first subgroup analysis will be conducted in patients with a history of diabetes mellitus or not Figure 3; the second subgroup analysis will be conducted in patients with a medication history of antidiabetic agents or not Figure 4; the third subgroup analysis will be conducted in patients with a HbA1c ≥ 7.0% or ＜7.0% Figure 5, to determine the impact of different fasting blood glucose levels on in-hospital mortality and percentage of the evacuation of intracranial hematoma respectively, presented with ORs and 95% CI. The cut-off value for FBG level to predict in-hospital mortality will be determined through ROC curve analysis.