Abstract: Background　Upper gastrointestinal bleeding（UGIB）is one of the common and serious clinical diseases，with threatening to life safety and relating to high medical burden；the influence of diabetes on the risk of UGIB has been attracting attention increasingly in recent years，but there is no research about the influence of fasting blood glucose（FBG）on the risk of new-onset UGIB.This study was conducted based on the data of Kailuan Research，which is with relatively fixed queue，large sample size，good follow-up and high reliability.Methods　A total of 100 003 in-service and retired employees of Kailuan Group who received 2006 physical examination were selected based on Kailuan study cohort and divided into normal group（FBG<6.1 mmol/L，n=79 966），pre-diabetes group（6.1 mmol/L≤FBG<7.0 mmol/L，n=7 645）and diabetes group（FBG≥7.0 mmol/L，n=12 392）according to the FBG. Incidence of new-onset UGIB in the three groups was recorded based on annual follow-up with a deadline to December 31，2018；Kaplan-Meier method was used to calculate the cumulative incidence of new-onset UGIB in the three groups and Log-rank test was performed；annual human incidence（incidence density）in the three groups was calculated，and COX proportional hazard regression model analysis and sensitivity analysis were carried out to analyze the influence of FBG on the risk of new-onset UGIB.Results　100 003 cases of physical examination were followed up from 2 days to 15 years，at average of（10.6±1.7）years with a rate of lost to follow-up of 9.96%（9 960/100 003），a total of 942 cases occurred new-onset UGIB with an incidence density of 8.91 per 10 000 person-years，thereinto the number of new-onset UGIB cases in the normal group，pre-diabetic group，and diabetes group were 711，84，and 147，and the incidence density was 8.35，10.45，and 11.67 per 10 000 person-years，respectively. Kaplan-Meier curve showed that cumulative incidence of new-onset UGIB increased with the FBG elevating，and Log-rank test results showed that there is significant difference of cumulative incidence of new-onset UGIB among the three groups（χ2=14.84，P<0.01）. COX proportional hazard regression model analysis results showed that，after adjusting for confounding factors，risk of new-onset UGIB in pre-diabetes group and diabetes group was 1.30 times〔HR=1.30，95%CI（1.02，1.66），P=0.03〕and 1.35 times〔HR=1.35，95%CI（1.09，1.67），P<0.01〕compared with normal group，moreover the results of sensitivity analysis were basically consistent.Conclusion　Cumulative incidence of new-onset UGIB increases with increase of FBG，and FBG≥6.1 mmol/L can increase the risk of new-onset UGIB，moreover pre-diabetes and diabetes are independent risk factors of new-onset UGIB.

Key words: Blood glucose, Hemorrhage, Gastrointestinal tract, Fasting blood glucose, Upper gastrointestinal bleeding, Cohort study

摘要： 背景　上消化道出血（UGIB）是一种临床常见重大疾病，会对患者生命安全构成威胁，且诊疗负担较重；近年来糖尿病对UGIB发病风险的影响备受关注，但目前尚未见关于空腹血糖（FBG）对新发UGIB发病风险影响的研究报道。开滦研究队列相对固定、样本量大、随访性好、可信度高，本研究通过利用开滦研究资料开展。目的　采用前瞻性队列研究探讨FBG对新发UGIB发病风险的影响。方法　基于开滦研究队列，共纳入参加2006年度第一次健康体检的开滦集团在职及离退休职工100 003例，根据FBG分为正常血糖组79 966例（FBG<6.1 mmol/L）、糖尿病前期组7 645例（6.1 mmol/L≤FBG<7.0 mmol/L）、糖尿病组12 392例（FBG≥7.0 mmol/L）。逐年随访并记录三组体检者新发UGIB发病情况，随访截止日期为2018-12-31；采用Kaplan-Meier法计算三组体检者新发UGIB累积发病率，并进行Log-rank检验；计算三组体检者UGIB人年发病率（发病密度）；采用COX比例风险回归模型分析及敏感性分析探究FBG对新发UGIB的影响。结果　100 003例体检者随访2 d~15年，平均随访时间为（10.6±1.7）年，失访率为9.96%（9 960/100 003），共新发UGIB 942例，发病密度为8.91/万人年，其中正常血糖组、糖尿病前期组、糖尿病组新发UGIB例数分别为711、84、147例，发病密度分别为8.35/万人年、10.45/万人年、11.67/万人年。Kaplan-Meier曲线显示新发UGIB累积发病率随FBG升高而升高，Log-rank检验结果显示三组体检者UGIB累积发病率比较，差异有统计学意义（χ2=14.84，P<0.01）。COX比例风险回归模型分析结果显示，校正混杂因素后糖尿病前期组、糖尿病组体检者新发UGIB发病风险分别是正常血糖组的1.30倍〔HR=1.30，95%CI（1.02，1.66），P=0.03〕、1.35倍〔HR=1.35，95%CI（1.09，1.67），P<0.01〕，且敏感性分析结果基本保持一致。结论　新发UGIB累积发病率随FBG升高而升高，FBG≥6.1 mmol/L即可增加新发UGIB发病风险，糖尿病前期和糖尿病均是新发UGIB发病的独立危险因素。

关键词: 血糖, 出血, 胃肠道, 空腹血糖, 上消化道出血, 队列研究