Combining glycosylated hemoglobin A_1c and fasting plasma glucose for diagnosis of type 2 diabetes in Chinese adults

Type 2 diabetes mellitus (T2DM) is a common disease reflecting metabolic disorders characterized with hyperglycemia, which may lead to specific long-term complications affecting heart, brain, eyes, kidneys and nervous system [1]. Currently, diagnosis of T2DM in Chinese adults is principally according to the 1999 World Health Organization (WHO) diagnostic criteria [2], namely, a person can be diagnosed as T2DM when he or she has 1) a random plasma glucose ≥ 11.1 mmol/l accompanied by typical symptoms of diabetes such as thirst and polyuria; or 2) a fasting plasma glucose (FPG) ≥ 7.0 mmol/l; or 3) a 2-hour plasma glucose (2hPG) after an oral glucose tolerance test (OGTT) ≥ 11.1 mmol/l [3]. These diagnostic criteria are determined based on the relationship between glucose levels and the risk of subsequent long-term complications.

The accuracy of glucose measurement, however, is more or less affected by the pre-analytic instability and biologic variability of glucose concentrations within and between days. Moreover, OGTT requires a second blood sample and is therefore more costly and time-consuming, bringing inconvenient in logistics and causing uncomfortable feeling in individuals. Compared with 2hPG in OGTT, FPG measurement is more reproducible and simpler to perform in clinics and communities [4]. Therefore FPG measurement is usually used alone to screen T2DM in China [5]. Due to that a certain proportion of T2DM patients have a normal FPG but an elevated 2hPG (≥ 11.1 mmol/l) [6], almost half diabetes patients remain undiagnosed in China [7]. A more efficient and practical method is needed to identify the individuals with diabetes in general population.

Hemoglobin A_1c (HbA_1c), a glycosylated form of hemoglobin, reflects one’s plasma glucose levels over past 2–3 months and is less influenced by recent diet and emotional stress [8]. HbA_1c level is a convenient and reliable measurement of chronic glycemia [9], and can be tested concurrently with FPG at fasting. HbA_1c is associated with the risk of long-term diabetic complications and has been used to monitor glycemic control status in prevalent patients with diabetes for about 30 years. Substantial studies [10‐12], including one conducted in Chinese population [12], have proved that HbA_1c may be a helpful tool for diagnosing T2DM. The American Diabetes Association (ADA) has used an HbA_1c cut-point of 48 mmol/mol (6.5%) as one of diagnostic criteria for T2DM in its 2010 Clinical Practice Recommendations [1].

Although HbA_1c has been applied to identify T2DM in the U.S. and European countries, it has not been used in China, mainly due to lack of a standardized approach to measure HbA_1c, short of knowledge about racial-specific standard and deficiency of an optimal cut-point for detecting T2DM in the population. HbA_1c values between 37 mmol/mol (5.5%) and 48 mmol/mol (6.5%) have been associated with a substantially increased risk for developing T2DM [13]. Due to racial disparity in HbA_1c level [14], however, the optimal threshold varies across populations. In Chinese adults at high risk of glucose intolerance, combining use of an FPG ≥ 6.1 mmol/l and an HbA_1c ≥ 43 mmol/mol (6.1%) has been shown to predict subsequent T2DM accurately [15]. Bao et al.[12] reported that an HbA_1c ≥ 45 mmol/mol (6.3%) can be used as a diagnostic criterion for diabetes in Chinese adults when FPG and OGTT are not available. Given that the HbA_1c value reflects average level of glycemia over the preceding 2–3 months and the FPG represents the current glucose concentration, combining use of the two indicators may improve the accuracy of diagnosis.

In this study, we included 6,661 Chinese adults with no prior history of diabetes to evaluate whether the combination of FPG and HbA_1c can be used as a tool for screening and diagnosis of T2DM in Chinese community settings.

Table 1 shows the demographic and clinical characteristics of participants of the survey. Compared with the men, the women had a lower average level of BMI, WC, SBP, DBP or TG but a higher concentration of TC, HDL-C, LDL-C or 2hPG (all P values < 0.05). No significant difference was observed for age and average levels of FPG and HbA_1c between men and women.

Presented in Additional file 1: Table S1 are the average levels of FPG, 2hPG and HbA_1c by demographic and clinical characteristics of the subjects. In both sexes, the levels of FPG, 2hPG and HbA_1c increased with age and increasing BMI, with Spearman correlation coefficients varying from 0.10 to 0.25 (all P < 0.05). The participants with elevated WC, hypertension, elevated TG, reduced HDL-C or HW phenotype had a higher concentration of FPG, 2hPG or HbA_1c than those without (P < 0.05), but with the difference in the FPG level between subjects with and without reduced HDL-C reaching significant only among women.

In this population, HbA_1c was closely correlated with FPG (r = 0.613, P < 0.0001) and 2hPG (r = 0.299, P < 0.0001) after adjusting for age and sex. As shown in Figure 2, however, nonlinear does-response relationship was observed for HbA_1c level with the levels of FPG and 2hPG after accounting for age and sex (P values for non-linear association < 0.0001).

Using the 1999 WHO diagnostic criteria as the gold standard, the sensitivity of FPG ≥ 7.0 mmol/l in diagnosing T2DM was 52.6%, while the sensitivity and specificity of HbA_1c of 43 mmol/mol (6.1%), the optimal cut-point of the index in this population, reached 75.6% and 89.3%, respectively (Table 2). The optimal threshold of HbA_1c for detecting T2DM ranged from 40 to 45 mmol/mol (5.8-6.3%) in the subgroups stratified by sex, age, BMI, WC, TG, HDL-C or presence of hypertension or HW phenotype. These cut-points increased with age and appeared higher in subjects with an elevated TG level or HW phenotype. All the optimal cut-points performed better in sensitivity than did the FPG ≥ 7.0 mmol/l. It is of note that the PPV of the optimal threshold increased with age and increasing BMI, and was higher in the subjects with elevated WC, hypertension, elevated TG, reduced HDL-C or HW phenotype (Table 2).

As shown in Additional file 2: Table S2, 239 T2DM were identified by elevated FPG (≥ 7.0 mmol/l) and 215 by increased 2hPG (≥ 11.0 mmol/l). Of the two groups of diabetes patients, 74.1% (177 of 239) and 45.6% (98 of 215) were identified as patients using an HbA_1c cut-point of 48 mmol/mol (6.5%) recommended by ADA, producing a sensitivity of 60.6% and a specificity of 96.2%. By using an HbA_1c of 45 mmol/mol (6.3%) recommended by Bao et al. for Chinese adults [12], the sensitivity increased to 67.2% and the specificity decreased to 94.0%. HbA_1c of 43 mmol/mol (6.1%) achieved a sensitivity of 75.6%, a specificity of 89.3%, a PPV of 34.1% and an NPV of 98.0% in diagnosis of T2DM.

As presented in Table 3, combining use of an FPG ≥ 7.0 mmol/l and an HbA_1c ≥ 43 mmol/mol (6.1%) had a higher sensitivity (83.7%) and PPV (36.4%) than using either one of the indicators alone in identify T2DM, together with a specificity of 89.3% and an NPV of 98.0%. Combining use of FPG and HbA_1c performed much better among the subjects with an elevated TG level or HW phenotype in the sensitivity (87.4% and 86.6% versus 83.7%) and PPV (42.7% and 44.9% versus 36.4%) than in all subjects, but with a certain decrease in the specificity (85.1% and 82.9% versus 89.3%) and NPV (98.2% and 97.4% versus 98.7%). Combining use of the two indicators was observed to classify 26.9% (272/1011) of pre-diabetes defined by 1999 WHO criteria as diabetes (Additional file 3: Table S3).

In this study including 6,661 Chinese adults randomly selected from a community setting, we find that an HbA_1c threshold of 43 mmol/mol (6.1%) performs best in diagnosing T2DM when combining with an FPG ≥ 7.0 mmol/l, particularly among those at a high risk of T2DM. Our results provide additional evidence on the value of HbA_1c measurement in identifying diabetes patients in Chinese population.

For diabetes patients, it is diabetic complications occurred subsequently that damage their health and quality of life. The value of OGTT in diagnosing diabetes lies in the close relationship of 2hPG with the risk of diabetic complications. Elevated HbA_1c level has also been linked to a higher risk of diabetic retinopathy [19], nephropathy [20], cardiovascular diseases [21, 22] and premature death [23, 24]. Moreover, HbA_1c possesses much lower intra-individual CV (3.6%) than 2hPG (16.7%) and FPG (5.7%), and has been used as the most reproducible and repeatable measurement of glycemic status [25]. Therefore, HbA_1c has great potential as a screening or diagnosing tool for T2DM.

Several cross-sectional [11, 12] and prospective [10, 26] studies have provided evidence on the superior performance of HbA_1c to FPG in screening or diagnosing T2DM. Consistent with these reports, we find that the optimal HbA_1c threshold of 43 mmol/mol (6.1%) had a higher sensitivity (75.6%) than did the FPG ≥7.0 mmol/l (52.6%) in our population, and produced a specificity of 89.3%, a PPV of 34.1% and an NPV of 98.0%. It is of note that, in this population, the HbA_1c cut-point of 45 mmol/mol (6.3%) achieved a slightly higher sensitivity (67.2% vs. 62.8%) and a slightly lower specificity (94.0% vs. 96.1%) than in Bao et al’s report [12]. The older average age of our population may partly explain the difference.

Interestingly, when combining the cut-point of 43 mmol/mol (6.1%) for HbA_1c and 7.0 mmol/l for FPG, the sensitivity and the PPV increased to 83.7% and 36.4%, respectively, with the specificity and the NPV as high as by using HbA_1c alone. Based on these results and the fact that HbA_1c is more stable and convenient than OGTT in practice [12], we recommend that the combination of an HbA_1c ≥ 43 mmol/mol (6.1%) and an FPG ≥ 7.0 mmol/l can be used as screening and diagnostic criteria for T2DM in Chinese population. We also find that, among the subjects with an elevated TG level or with HW phenotype, the combination of an HbA_1c ≥ 43 mmol/mol (6.1%) and an FPG ≥ 7.0 mmol/l achieved a satisfactory sensitivity and a greatly improved PPV, implicating that the criteria are more efficient in detecting T2DM among subgroups at high risk of T2DM.

Although it has been suggested that HbA_1c alone has potential to be used as a diagnostic criterion for T2DM in Chinese adults [12], combining use of the measurement with FPG has two additional advantages. Firstly, genetic variants of hemoglobin (e.g. sickle cell trait, HbS trait, HbC trait, etc.) and some medical conditions that could shorten or conversely prolong the survival of erythrocytes may affect the accuracy of HbA_1c measurements [27]. Combining use of HbA_1c and FPG can reduce the misdiagnosis of T2DM caused by using HbA_1c alone. Secondly, as HbA_1c value reflects average level of blood glucose over the preceding 2–3 months, subjects having diabetes in 3 months cannot be detected by testing HbA_1c alone. Combining use of HbA_1c with FPG can compensate for this weak point.

The main strength of the study is its large sample size. However, the response rate of 67.2% arouses a concern on the representation of the sample, and the lack of characteristics information on the non-participants limits our ability to evaluate the potential selection bias. In this population, age appeared to be the most important characteristic factor linked to different optimal HbA_1c cut-off points (Table 2). The older average age of our sample population due to non-response of the subjects who refused to participate or were absent during the enrolment period may lead to a higher cut-off point of HbA_1c for diagnosis. However, the non-inclusion of these individuals is unlikely matter in the potential application of the public health tool. Additionally, the biochemical assays were conducted in several Community Healthcare Centers, which may have led to inter-lab bias in measurements of HbA_1c. The unified protocol and strict quality control process may help to release our concerns. Moreover, although the sensitivity and specificity of combining use of HbA_1c and FPG were as high as 83.7% and 89.3%, respectively, false positive and negative cannot be neglected. Finally, the high cost of HbA_1c test may limit its widespread use in China. Compared with FPG assay, HbA_1c test is much more expensive. Considering that the cost for HbA_1c assay almost equals to that for the OGTT [12, 28], however, combining use of HbA_1c and FPG will not lead to an extra economic burden comparing with the current application of 1999 WHO diagnostic criteria.

In conclusion, a threshold of HbA_1c ≥ 43 mmol/mol (6.1%) in combination with an FPG ≥ 7.0 mmol/l produce a satisfactory sensitivity and specificity for diagnosis of T2DM in our population. Our findings support the combining use of FPG and HbA_1c as a screening and diagnosis tool of T2DM in Chinese adults, especially among those at a high risk of the disease. Further epidemiological and clinical studies are warranted to validate our results.