Optimal cut point of glycosylated serum protein and glycated albumin in diagnosis of diabetes mellitus in high risk populations

Objective To study the optimal cut point of receiver operating characteristic（ROC） curves for glycosylated serum protein（GSP）,glycated albumin（GA-L） and fasting plasma glucose（FPG） when glycated hemoglobin（HbA1c） was used in diagnosis of diabetes mellitus（DM）.Methods ROC curves were plotted for 88 patients（54 males and 34 females） with suspicious DM diagnosed following the 2010 American Diabetes Association（ADA） diagnostic criteria.The optimal cut point,sensitivity and specificity of GSP,GA-L and FPG in diagnosis of DM were comparatively analyzed.Results The optimal cut point of ROC curves for GSP in diagnosis of DM was 186.1μmol/L,with a sensitivity of 55%,a specificity of 98%,area of 0.804 under the curve,a positive likelihood ratio of 27.63,a negative likelihood ratio of 0.46,a positive predictive value of 95% and a negative predicative value of 74%.The optimal cut point of ROC curves for GA-L in diagnosis of DM was 16.86%,with a sensitivity of 61%,a specificity of 98%,area of 0.849 under the curve,a positive likelihood ratio of 30.26,a negative likelihood ratio of 0.40,a positive predictive value of 96% and a negative predicative value of 77%.The optimal cut point of ROC curves for FPG in diagnosis of DM was 5.67mmol/L,with a sensitivity of 58%,a specificity of 78% and area of 0.736 under the curve.Conclusion Among the GA-L,GSP and FPG,the specificity and sensitivity of GA-L are the highest in diagnosis of DM.The accuracy of GA-L in diagnosis of DM is higher than that of GSP and FPG since the area under the curve is GA-L＞GSP ＞FPG.