| ObjectivesTo evaluate the performance of HbA1c in diagnosing diabetes and pre-diabetes among a general population in Chinese large cities, and to identify the optimal cut-off for diabetes diagnosis. To evaluate the performance of different ways of combination of HbA1c and FPG in detecting diabetes, and to find out an optimal strategy of combination of HbAlc and FPG for detecting diabetes and pre-diabetes, so to provide a scientific basis for choosing feasible strategy to detect diabetes in the population.Subjects and MethodsThe 2010 National Nutrition and Health Surveillance were conducted in 34 large-size cities in China in 2010. After the pregnant and juveniles under 18 excluded, 26,548 subjects participated in this survey. All participants were interviewed to get demographic and health information, and were asked to take part in the physical examination and the oral glucose tolerance test. Besides, we tested the HbA1c for those who have blood drawn. Then, test measurements with poor laboratory quality control of glucose, with HbAlc, FPG or 2hPG missing and those with diagnosed diabetes were excluded. Finally, 12604 subjects were included in this analysis.The oral glucose tolerance test was taken as the gold standard, and the accuracy of HbA1c and combination of HbA1c and FPG in diagnosing diabetes and impaired glucose regulation was evaluated using receiver-operating characteristic curve.Sensitivity, specificity and other indicators of accuracy was calculated. The optimal cut-off was chosen when related to the maximal Youden's Index. And the threshold of fasting plasma glucose to exclude diabetes by 2hPG was chosen based on the accuracy indicators and the ROC curve. Logistic model was used to calculated the odds ratio of different HbAlc level to predict diabetes and impaired glucose regulation.ResultsPart I1. The average HbA1c level was 5.7%. At the age of 55 and lower, the HbA1c level increased with age, and after 55-year-old, HbAlc did not increased significantly.However, the level of HbAlc increased with age in each age group in the female. The average HbAlc levels of male and female subjects in the north district were 5.67%and 5.65%, and those of southern male and female were 5.78% and 5.72%. The HbAlc level of sujects in northern district was higher than that of subjects in the southern district (P<0.0001).2. The area under the ROC curve for HbA1c diagnosing diabetes by oral glucose tolerance test was 0.836 (95% CI 0.819-0.852). The optimal cut-off with the maximal Youden's Index was at HbA1c level of 6.0%, related with sensitivity and specificity of 74.0% and 79.5%. The HbAlc level of 6.5% was associated with sensitivity of 48.0%and specificity of 95.0%?Compared with using HbA1c of 6.5% and greater,54.2%more diabetes would be detected using HbA1c?6.0%. But 17.1% of the normal glucose tolerance and 39.2% of the impaired glucose regulation by oral glucose tolerance test would be misdiagnosed as diabetes using HbAlc?6.0%. The kappa coefficient between the criteria of OGTT and HbAlc?6.0% in men and women was 0.2359 and 0.2134(P<0.0001),respectively. Using HbAlc?6.5%, 52.0% diabetes by OGTT would be missed diagnosis, but the proportions of which NGT and IGR by OGTT would be misdiagnosed as diabetes were 4.1% and 10.1%. The kappa coefficient between the criteria of OGTT and HbAlc?6.5% was 0.3756 and 0.3951(P<0.0001).3. When male subjects were divided into groups according to gender, age, district,BMI, waist and lipid respectively, the areas under the ROC curve were not significantly different between groups (P>0.05). While in the female subjects, the area under ROC curves for subjects from the north was higher than that for those from the south, and that for the mid-aged was higher than that for the elderly, and that for abnormal lipid was higher than that for those without abnormal lipid (P<0.05).4. The area under the ROC curve for HbAlc detecting diabetes by OGTT for the female subjects from the north was 0.894(95% CI 0.856-0.932), which was higher than that for the male subjects from the north (P<0.05), and was also higher than that for the female subjects from the south (P<0.05). The maximal Youden's Index was associated with the optimal HbA1c cut-off of 6.1%. The coefficients of chance-corrected proportion agreement (kappa) for classification of diabetes between the criteria of the OGTT and any other HbA1c cut-offs (6.0, 6.3, 6.5) for the northern female subjects were all higher than the corresponding values for the southern female subjects (P<0.05). Besides, the AUC for the northern female was highest among the high-risk subjects (P<0.05).5. The AUC for HbA1c detecting impaired glucose regulation was 0.678(95%CI 0.665-0.691), and the optimal cut-off was 5.7%, with sensitivity and specificity of 67.5% and 59.7%.6. HbA1c was positively correlated with fasting plasma glucose and 2-hour post-load plasma glucose. The Spearmen coefficients between HbAlc and FPG, 2hPG was 0.308, 0.270 (P<0.0001). The coefficients between HbA1c and FPG,2hPG among the subjects with diabetes by OGTT was 0.460, 0.586 (P<0.0001).7. The subjects diagnosed by HbA1c?6.0% had lower BMI,waist,FPG,2hPG,HbA1c and HDL_C than those diagnosed by HbA1 ?6.5% or by OGTT (P<0.0001).Compared with diabetes diagnosed by OGTT, those with HbAlc?6.5% had lower BMI, FPG, 2hPG and triglyceride, but higher HbA 1 c (P<0.0001).Part ?1. Among the newly diagnosed diabetes, subjects with only fasting plasma glucose?7.00mmol/L accounted for 37.2%, and with only 2-hour plasma glucose11.1mmol/L accounted for 13.6%. And 43.7% of the impaired glucose regulation was impaired fasting glucose alone, and 39.4% was impaired glucose tolerance.2. The AUC for FPG detecting 2hPG-DM was 0.895(95% CI 0.877-0.914), and the optimal cut-off was 6.2mmol/l. The sensitivity of the FPG level of 5.4mmol/ and greater in detecting 2hPG-DM was 90.5% and when FPG value was lower 5.4 mmol/I,the likelihood to predict non-diabetes was 99.4%. The proportion of 2hPG-DM was 0.63% among the subjects with fasting plasma glucose lower than 5.4mmol/l. The AUC for HbAlc predicting 2hPG-DM was 0.882(95%CI 0.864-0.900), the optimal cut-off was 6.2%.3. The AUC for FPG predicting impaired glucose tolerance was 0.713(95% CI 0.696-0.731), the optimal cut-off was 5.5mmol/I. The AUC for HbA1c predicting IGR was 0.673 (95%CI 0.655-0.690).4. The strategy of combination of HbA1c and FPG in detecting diabetes was that:firstly, FPG was tested for all participants and those with FPG?7.0mmol/I were defined as diabetes; then, HbAlc was tested among those with 5.4mmol/l?FPG<7.0mmol/l and those with HbAlc?6.5% were diagnosed with diabetes. Using this method to diagnose diabetes defined by OGTT, the sensitivity was 82.9% and the specificity was 97.4%,with the Youden's Index of 80.4%. The proportion of diabetes detected by this strategy was 7.4%, and also 4910 subjects would have their HbAlc tested.5. The strategy of combination of HbA1c and FPG in detecting IGR was that: firstly,FPG was tested for all participants and those with 6.1mmol/l?FPG<7.0mmol/l were defined as IGR; then, HbA1c was tested among those with 5.4mmol/l?FPG<6.1mmol/l and those with 6.0%?HbA1c<6.5% were diagnosed with IGR.Using this method to diagnose IGR defined by OGTT, the sensitivity was 68.9% and the specificity was 93.4%, with the Youden's Index of 62.3%. The proportion of diabetes detected by this strategy was 15.3%.ConclusionSex, age and geographical affected the HbA1c level. There was a positive relationship between HbA1c and FPG, HbA1c and 2hPG. As the level of HbA1c increased, the risk of diabetes, pre-diabetes and abnormal lipid profiles increased. HbAlc was useful to diagnose diabetes, but it was not consistent with oral glucose tolerance test. The optimal HbA1c cut-off of HbA1c for detecting diabetes which correspoongded to the maximal Youden's Index, had low specificity and moderate sensitivity,and it was not ideal to diagnose diabetes in the current population. However, HbAlc of 6.5% and greater had high specificity though low sensitivity, and subjects with HbA1c of this level had higher risk of diabetes and poorer lipid profiles. Therefore, when FPG or 2hPG was not available, HbAlc?6.5% recommended by ADA might be suitable as a diabetes diagnostic criterion. Neverthless, the performance of HbA1c detecting pre-diabetes was poor.The combination of FPG and HbA1c improved the efficiency of detecting OGTT-DM and IGR, and also the sensitivity and specificity increased. The better strategy of the combination of these two indicators was a two-step method, that is,firstly, FPG was tested among all subjects and those with FPG?7.0mmol/l were diagnosed as diabetes and those with 6.1mmol/l?FPG<7.0mmol/l were pre-diabetes; secondly, HbAlc would be tested among those with 7.0mmol/l>FPG?5.4mmol/l and those who had HbA1c ?6.5% were diabetes and those who had 6.5%>HbA1c ?6.0% were pre-diabetes. |
PDF Full Text Request