摘要目的：研究糖化血红蛋白在不同年龄、不同性别的糖尿病及糖尿病前期患者中的风险值。方法回顾性研究。收集浙江省台州医院2012年检测HbA1c的健康体检者10840例，使用决策树CHAID法进行初步分层，根据初步分层结果将年龄进一步划分为5层，最终分20～30岁、31～40岁、41～50岁、51～60岁及＞60岁五层（F＝434.342，P＝0.000）。再采用单因素方差分析和线性回归分别确定各年龄组间差异性及HbA1c值的95％预测区间（ PI）；以HbA1c值的95％PI上限值为糖尿病风险值（DRVs），分别用HbA1c的参考区间上限值及DRVs从1889例HbA1c值在6.0％～6.5％的人群中筛查DM患者，对比两者的筛查灵敏度来验证DRVs对于DM患者的筛查价值；再用HbA1c的DRVs对5年前（2008年） HbA1c值在6.0％～6.5％的人群进行糖尿病筛查并跟踪其复查记录来验证其对于PreD患者的筛查价值。结果（1）决策树根节点按照年龄初步分为9层（F＝231.462，P＝0.000），计算各层95％置信区间后对上限值相同的进行合理合并。（2）采用线性回归得到的年龄5层式HbA1c的95％PI如下：20～30岁：男性4.6％～5.9％，女性4.6％～5.9％；31～40岁：男性4.8％～6.0％，女性4.7％～6.0％；41～50岁：男性4.9％～6.2％，女性4.8％～6.1％；51～60岁：男性5.0％～6.3％，女性4.9％～6.2％；＞60岁：男性5.1％～6.4％，女性5.0％～6.3％；传统的HbA1c参考区间为：4.8％～6.3％。（3）得到的DRVs与传统HbA1c参考区间上限值在HbA1c值6.0％～6.5％人群筛查出糖尿病患者的灵敏度分别为：48.1％，34.5％。（4）回顾2008年HbA1c值在6.0％～6.5％而未被诊断为DM的人群采用传统HbA1c参考区间上限值进行糖尿病筛查为阴性结果而采用本文得出的DRVs筛查为阳性结果的有49人，至今真正发展为DM的有13人，占26.5％。结论分层设定不同年龄和不同性别人群之间的HbA1c的DRVs，作为传统参考区间的补充，对于DM患者的诊治及PreD患者的筛查和干预具有重要意义。（中华检验医学杂志，2014，37：701-704）

abstractsObjective To study the risk values of glycosylated hemoglobin ( HbA1c) screening patients with diabetes mellitus(DM) and pre-diabetes(PreD) in different age and sex.Methods HbA1c results from10 840 in Taizhou Hospital in 2012 were retrospectively collected , and CHAID analysis of decision tree was used in preliminary hierarchy .Age was further divided into five groups according to the preliminary results , then one-way ANOVA and linear regression were utilized to determine the differences between age groups and the prediction intervals (PI) of HbA1c,respectively.The upper limit value of 95%PI of HbA1c was set as diabetes risk values ( DRVs).By comparison of the sensitivity of DM patients screening by DRVs in 1 889 cases whose HbA1c were between 6.0%to 6.5%and the upper limit value of traditional HbA1c reference intervals to verify the screening value ofDRVs .Then the DRVs of HbA1c were utilized to screen DM in the population whose HbA 1c is were 6.0%-6.5% in 2008,and followed their re-examination records toverify the screening value of the DRVs toin PreD patients. Results (1) On the basis of the initially root of decisiontree by age , the populationwas divided into 9 groups ( P =0.000, F =231.462).By calculatingthe 95% confidence interval(CI) of each group and merging reasonably,it was finally divided into 5 groups:20-30years,31-40years,41-50years,51-60years and >60years (F=434.342, P=0.000).(2)Using the method of linear regression , the 95% PI of HbA1c of the 5 groups showed as following:20-30 years, males 4.6%-5.9%, females 4.6%-5.9%.31-40 years, males 4.8%-6.0%, females 4.7%-6.0%.41-50 years, males 4.9%-6.2%, females 4.8%-6.1%.51-60 years, males 5.0%-6.3%, females 4.9%-6.2%.>60 years, males 5.1%-6.4%, females 5.0%-6.3%.The traditional HbA1c reference interval was4.8%-6.3%.(3) The screening sensitivity of DM by the upper limit value of traditional HbA1c reference interval and DRVs in group which HbA1c was between 6.0%-6.5% were 34.5%and 48.1%,respectively.(4) Reviewing of the group with HbA1c between 6.0%-6.5% and not diagnosed as DM in 2008, 49 hadnegative DM screening results by using the upper limit value of traditional HbA1c reference interval but were positive in our setting DRVs , and 13 have developed to DM now , which accounted for 26.5%.Conclusion HbA1c DRVs need to be set hierarchically between different age and sex groups as a supplement to the traditional reference interval , which has a great significance for the diagnosis and treatment of DM and also forthe screening and intervention of PreD patients.