Abstract: Background　Diabetes is a disease seriously threatening the health of middle-aged and elderly people in the community.Although many studies have explored the predictive effect of obesity indicators on diabetes，the results are still controversial due to different participant populations and obesity indicators involved.Objective　To explore the values of six obesity indicators in diabetes risk prediction among middle-aged and elderly population in the community，so as to provide a reference for early detection of high-risk groups of diabetes.Methods　We recruited 2 070 middle-aged and elderly individuals，1 000 men and 1 070 women，from the entire population of two communities in Pearl River Delta attending a medical examination.We used univariate and banal logistic regression to perform a sex-based analysis of the relationships of diabetes with six obesity indicators〔BMI，waist circumstance（WC），waist-to-hip ratio（WHR），waist-to-stature ratio（WSR），ponderal index（PI） and body adiposity index（BAI）〕，and conducted an ROC curve analysis and area under curve（AUC） to examine the value of the six indicators in diabetes risk prediction.Results　Of the six obesity indicators，WC was most closely associated with diabetes，either in men or in women.Moreover，the risk of diabetes increased with the growth of WC in both men〔OR（95%CI）=1.601（1.322，1.938）〕and women〔OR（95%CI）=1.496（1.242，1.802）〕.ROC curve showed that the AUC of WC and WSR was the largest（0.645）in male middle-aged and elderly population，and PI was the largest（0.654） in female middle-aged and elderly population.WC was 87.95 cm，WSR was 0.54，and PI was 14.81 kg/m3.Conclusion　The risk of diabetes as well as diabetes prediction indicators differ significantly in male and female middle-aged and elderly individuals in the community.WC and WSR are more suitable to be used as obesity index for predicting diabetes in male population and PI is more suitable to be used as obesity index for predicting diabetes in female population..

Key words: Community health services；Diabetes, type 2；Sex factors；Obesity index；Middle-aged and elderly people；Sensitivity；Specificity；Predictive value

摘要： 背景　糖尿病是严重威胁中国社区中老年人群健康的疾病之一。目前虽有许多研究探索了肥胖指标对糖尿病的预测效果，但由于研究对象的不同和肥胖指标选择的差别，使得研究结果仍具有一定的争议性。目的　探讨不同肥胖指标对社区中老年人群患糖尿病风险的预测效果，为尽早发现糖尿病高危人群提供参考。方法　以珠江三角洲某两个社区进行全人群体检的2 070例中老年人为研究对象，其中男1 000例，女1 070例。按照性别分类，采用二分类Logistic回归分析不同肥胖指标〔体质指数（BMI）、腰围（WC）、腰臀比（WHR）、腰围身高比（WSR）、重量指数（PI）、身体肥胖指数（BAI）〕对糖尿病患病的影响，并用受试者工作特征曲线（ROC曲线）及曲线下面积（AUC）分析各个肥胖指标预测糖尿病患病风险的效果。结果　不论是在男性还是女性社区中老年人群中，与其他肥胖指标相比，WC是糖尿病最大的危险因素，糖尿病患病风险随着WC的增加而增加，在男性中OR（95%CI）=1.601（1.322，1.938），在女性中OR（95%CI）=1.496（1.242，1.802）。ROC曲线结果显示，男性中老年人群中，WC和WSR的AUC最大（0.645），WC最佳切点为87.95 cm，WSR最佳切点为0.54；女性中老年人群中，PI的AUC最大（0.654），PI的最佳切点为14.81 kg/m3。结论　在社区中老年人群中，预测糖尿病患病情况的肥胖指标在不同性别是有差异的。在男性人群中，WC和WSR更适合作为预测糖尿病的肥胖指标；在女性人群中，则PI更适合作为预测指标。

关键词: 社区卫生服务；糖尿病, 2型；性别因素；肥胖指标；中老年人；灵敏度；特异度；预测价值