We compared the performance of hemoglobin A1c (HbA1c) versus the fasting plasma glucose (FPG) in diagnosing the metabolic syndrome and assessed the diagnostic accuracy of the metabolic syndrome definition using HbA1c in identifying insulin-resistant subjects. The cardiometabolic risk factors, HbA1c, and glucose tolerance were analyzed in 774 nondiabetic white subjects. Insulin sensitivity was estimated with an oral glucose tolerance test-derived insulin sensitivity index. Insulin resistance was defined as the lower quartile of insulin sensitivity index. A 90.9% agreement existed between the use of HbA1c and the FPG for diagnosis of the metabolic syndrome (κ coefficient = 0.813); however, the proportion of subjects who met the metabolic syndrome criteria using the HbA1c was greater (42.1% vs 39.7%). Compared to the subjects who met the metabolic syndrome criteria using the FPG alone, those with the metabolic syndrome using the HbA1c-alone criterion were younger, had greater visceral adiposity, greater levels of inflammatory markers and liver enzymes, and lower blood pressure. In a logistic regression analysis with adjustment for age and gender, the subjects with the metabolic syndrome using the HbA1c criterion only had a 3.6-fold increase risk of having insulin resistance, defined as the lowest quartile of the insulin sensitivity index. A similar risk (3.8-fold) was observed in those who met the metabolic syndrome criteria using FPG alone. Insulin-resistant subjects who did not meet the criteria for the metabolic syndrome using the HbA1c had an unfavorable cardiovascular disease risk profile. In conclusion, although a good agreement existed between the HbA1c and FPG criteria for the diagnosis of the metabolic syndrome, appreciably different groups of subjects were classified using each method.
In the present study, we evaluated whether the hemoglobin A1c (HbA1c) values can be used instead of the fasting plasma glucose (FPG) to identify subjects with the metabolic syndrome in a well-characterized sample of nondiabetic white subjects. Additionally, we determined the diagnostic accuracy of the HbA1c-based definition of the metabolic syndrome in identifying subjects with insulin resistance.
Methods
A total of 774 whites subjects recruited from the University of Rome and the University of Catanzaro areas participated in this cross-sectional study for an assessment of cardiometabolic risk factors. The recruitment mechanisms included word-of-mouth, fliers, and newspaper advertisements. The inclusion criteria were age ≥20 years, the absence of diabetes mellitus, defined as HbA1c ≥6.5%, FPG ≥126 mg/dl, or a 2-hour postload plasma glucose ≥200 mg/dl, and the presence of one or more cardiometabolic risk factors, including elevated blood pressure, dyslipidemia, overweight/obesity, and a family history of diabetes. On the first day, after a 12-hour fast, the subjects underwent an anthropometric evaluation, including body mass index and waist circumference, and a venous blood sample was drawn for laboratory determinations. Readings of the clinic blood pressure were obtained from the left arm of the supine patients, after 5 minutes of quiet rest, using a mercury sphygmomanometer. The pulse pressure was calculated as the difference between the systolic and diastolic blood pressure. On the second day, after a 12-hour fast, a 75-g oral glucose tolerance test was performed, with sampling for plasma glucose and insulin. The institutional ethics committees approved the protocol, and the participants provided written informed consent.
HbA1c was measured using high-performance liquid chromatography with a National Glycohemoglobin Standardization Program certified automated analyzer (Adams HA-8160 HbA1c analyzer, Menarini, Italy; normal range 4.3% to 5.9%). The glucose, triglyceride, total and high-density lipoprotein (HDL) cholesterol, alanine aminotransferase, and aspartate aminotransferase concentrations were measured using enzymatic methods (Roche, Basel, Switzerland). The plasma insulin concentration was determined by a chemiluminescence-based assay (Immulite, Siemens, Italy). Insulin sensitivity was estimated using the previously validated oral glucose tolerance test-derived insulin sensitivity index (ISI). The subjects were defined as having insulin resistance if their ISI value was in the lowest quartile; this definition was chosen from evidence from prospective studies suggesting a statistically significant increase in clinical outcomes in those with the lowest quartile of insulin sensitivity. The metabolic syndrome was defined according to the criteria of the consensus statement released in 2009. The use of HbA1c at a range of ≥5.7% to 6.4% or FPG ≥100 mg/dl in the definition of the glycemic component of the metabolic syndrome were compared. The variables with a skewed distribution, including the fasting and 2-hour postload insulin, triglyceride, alanine aminotransferase, and aspartate aminotransferase concentrations, were log transformed for analyses. Continuous data are expressed as the mean ± SD. Categorical variables were compared using the chi-square test. A general linear model with adjustment for age and gender was used to compare the phenotypic differences between the groups. The κ statistic was calculated as a measure of agreement between the 2 definitions of the metabolic syndrome using the HbA1c and FPG diagnoses, respectively. The diagnostic property of HbA1c and FPG was evaluated by calculating the receiver operating characteristic (ROC) curve with the 95% confidence intervals. To determine whether the areas under the ROC curve obtained using different variables were significantly different, we used the method of DeLong et al. Partial correlation coefficients adjusted for age and gender were computed between variables. A logistic regression analysis with adjustment for age and gender was used to test for an association between groups and insulin resistance, the metabolic syndrome, or its individual components. All analyses were performed using Statistical Package for Social Sciences, version 16.0, for Windows (SPSS, Chicago, Illinois).
Results
For the whole study group, the proportion of subjects who met the metabolic syndrome criteria using the FPG was 39.7% (n = 308), and it was significantly greater in men (45.6%) than in women (36.7%; chi-square test, p = 0.016). The proportion of subjects who met the metabolic syndrome criteria using the HbA1c was greater (42.1%, n = 326) and did not differ significantly between genders (44.9% in men and 40.6% in women; chi-square, p = 0.25). A 90.9% agreement was found between the HbA1c and FPG criteria for the diagnosis of the metabolic syndrome (κ coefficient = 0.813). No significant differences were observed between the 2 groups. The area under the ROC curve of HbA1c for identifying subjects with the metabolic syndrome according to the FPG criterion was 0.678 (95% CI 0.644 to 0.711; Figure 1 ). The clinical characteristics of the participants stratified according to the metabolic syndrome are listed in Table 1 . The data showed that significant differences were observed when the 44 subjects identified by HbA1c alone were compared to the 308 subjects identified using the FPG criterion. Furthermore, compared to the subjects who met the metabolic syndrome criteria using the FPG alone, those with the metabolic syndrome according to both FPG and HbA1c criteria had significantly greater visceral adiposity, greater levels of 2-hour insulin, triglycerides, and white blood cell count, and lower levels of FPG and HDL ( Table 2 ). Compared to the subjects who met the metabolic syndrome criteria using the FPG alone, those with the metabolic syndrome using the HbA1c criterion alone were significantly younger, had greater visceral adiposity, greater white blood cell counts, and lower blood pressure ( Table 2 ).
| Variable | Metabolic Syndrome Using Fasting Plasma Glucose Criterion | Metabolic Syndrome Using Hemoglobin HbA1c Alone | p Value vs 308 Subjects With Metabolic Syndrome | ||
|---|---|---|---|---|---|
| No Metabolic Syndrome | Metabolic Syndrome | p Value | |||
| Subjects | 466 | 308 | 44 | ||
| Gender | 0.016 | 0.067 | |||
| Male | 144 | 121 | 11 | ||
| Female | 322 | 187 | 33 | ||
| Age (years) | 40 ± 12 | 45 ± 11 | <0.0001 ⁎ | 41 ± 10 | 0.03 |
| Body mass index (kg/m 2 ) | 29.7 ± 7.6 | 36.8 ± 8.3 | <0.0001 | 38.8 ± 8.5 | 0.59 |
| Waist circumference (cm) | 94.2 ± 15.8 | 111.2 ± 15.1 | <0.0001 | 112.1 ± 15.1 | 0.96 |
| Systolic blood pressure (mm Hg) | 120 ± 14 | 135 ± 14 | <0.0001 | 128 ± 12 | 0.02 |
| Diastolic blood pressure (mm Hg) | 77 ± 10 | 84 ± 9 | <0.0001 | 82 ± 7 | 0.17 |
| Pulse pressure (mm Hg) | 43 ± 10 | 50 ± 11 | <0.0001 | 47 ± 10 | 0.12 |
| Fasting glucose (mg/dl) | 89 ± 10 | 100 ± 12 | <0.0001 | 89 ± 7 | <0.0001 |
| 2-Hour glucose (mg/dl) | 113 ± 33 | 137 ± 43 | <0.0001 | 127 ± 30 | 0.44 |
| Hemoglobin A1c (%) | 5.31 ± 0.46 | 5.60 ± 0.53 | <0.0001 | 5.95 ± 0.24 | <0.0001 |
| Fasting insulin (μU/ml) | 11 ± 7 | 17 ± 10 | <0.0001 | 17 ± 12 | 0.43 |
| 2-Hour insulin (μU/ml) | 62 ± 63 | 97 ± 95 | <0.0001 | 96 ± 95 | 0.84 |
| Total cholesterol (mg/dl) | 199 ± 38 | 208 ± 42 | 0.16 | 201 ± 41 | 0.54 |
| High-density lipoprotein (mg/dl) | 57 ± 14 | 44 ± 12 | <0.0001 | 51 ± 12 | <0.0001 |
| Triglycerides (mg/dl) | 97 ± 41 | 175 ± 106 | <0.0001 | 111 ± 30 | <0.0001 |
| Fibrinogen (mg/dl) | 311 ± 79 | 333 ± 85 | <0.0001 | 352 ± 85 | 0.49 |
| White blood cell count (×10 9 /ml) | 6,676 ± 1,825 | 7,266 ± 2,029 | <0.0001 | 7,176 ± 1,283 | 0.61 |
| Hemoglobin (g/dl) | 14.0 ± 1.4 | 14.2 ± 1.4 | 0.51 | 14.0 ± 1.5 | 0.57 |
| Alanine aminotransferase (UI/L) | 23.6 ± 14.8 | 33.4 ± 19.9 | <0.0001 | 25.5 ± 15.6 | 0.49 |
| Aspartate aminotransferase (UI/L) | 20.6 ± 7.2 | 25.0 ± 10.9 | <0.0001 | 32.2 ± 21.3 | 0.80 |
| Insulin sensitivity index (mg × L 2 × mmol −1 × mU −1 × min −1 ) | 95.5 ± 53.6 | 54.7 ± 35.4 | <0.0001 | 62.0 ± 34 | 0.05 |
| Odds ratio for lower insulin sensitivity (95% CI) | 1 (reference category) | 4.44 (3.07–6.37) | <0.0001 | 3.25 (1.61–6.55) | 0.001 |
⁎ p Values refer to results after analyses with adjustment for gender.
| Variable | No Metabolic Syndrome Using Either FPG or HbA1c | Metabolic Syndrome Using FPG Only | Metabolic Syndrome Using HbA1c Only | Metabolic Syndrome Using Both FPG and HbA1c | p Value |
|---|---|---|---|---|---|
| Subjects | 422 | 26 | 44 | 282 | |
| Gender | 0.046 | ||||
| Male | 133 | 13 | 11 | 108 | |
| Female | 289 | 13 | 33 | 174 | |
| Age (years) | 40 ± 12 | 48 ± 11 ⁎ | 41 ± 10 † | 44 ± 12 ⁎ | <0.0001 ‡ |
| Body mass index (kg/m 2 ) | 28.7 ± 6.8 | 34.3 ± 7.6 § | 38.8 ± 8.5 § | 37.0 ± 8.3 § | <0.0001 |
| Waist circumference (cm) | 92.3 ± 14.6 | 106.5 ± 14.8 § | 112.1 ± 15.1 † § | 111.6 ± 15.1 † § | <0.0001 |
| Systolic blood pressure (mm Hg) | 119 ± 14 | 140 ± 13 § | 128 ± 12 † § | 134 ± 14 § | <0.0001 |
| Diastolic blood pressure (mm Hg) | 76 ± 10 | 85 ± 10 § | 82 ± 7 ⁎ | 84 ± 12 § | <0.0001 |
| Pulse pressure (mm Hg) | 43 ± 10 | 54 ± 11 § | 47 ± 10 † | 50 ± 11 § | <0.0001 |
| Fasting glucose (mg/dl) | 89 ± 11 | 106 ± 7 § | 89 ± 7 ¶ | 98 ± 12 § ∥ # | <0.0001 |
| 2-Hour glucose (mg/dl) | 112 ± 33 | 134 ± 42 ⁎⁎ | 127 ± 30 ⁎⁎ | 137 ± 43 § | <0.0001 |
| Hemoglobin A1c (%) | 5.24 ± 0.43 | 5.32 ± 0.26 | 5.95 ± 0.24 § ¶ | 5.70 ± 0.52 § ∥ # | <0.0001 |
| Fasting insulin (μU/ml) | 10 ± 6 | 14 ± 7 ⁎ | 17 ± 12 § | 18 ± 11 § | <0.0001 |
| 2-Hour insulin (μU/ml) | 58 ± 57 | 74 ± 66 | 96 ± 95 § | 100 ± 98 † § | <0.0001 |
| Total cholesterol (mg/dl) | 199 ± 38 | 210 ± 40 | 201 ± 41 | 208 ± 43 | 0.80 |
| High-density lipoprotein (mg/dl) | 58 ± 14 | 52 ± 10 ⁎⁎ | 51 ± 12 § | 43 ± 12 § ¶ †† | <0.0001 |
| Triglycerides (mg/dl) | 94 ± 38 | 122 ± 71 | 111 ± 30 ⁎ | 180 ± 108 § ¶ # | <0.0001 |
| Fibrinogen (mg/dl) | 306 ± 77 | 312 ± 68 | 352 ± 85 ⁎ | 334 ± 87 ⁎⁎ | 0.25 |
| White blood cell count (×10 9 /ml) | 6,647 ± 1,849 | 5,839 ± 1,233 | 7,176 ± 1,283 § | 7,368 ± 2,039 † § | <0.0001 |
| Hemoglobin (g/dl) | 14.0 ± 1.3 | 14.6 ± 1.4 | 14.0 ± 1.5 | 14.1 ± 1.4 | 0.27 |
| Alanine aminotransferase (UI/L) | 22.1 ± 6.4 | 22.2 ± 10.6 | 25.5 ± 15.6 § | 28.3 ± 21.4 § | <0.0001 |
| Aspartate aminotransferase (UI/L) | 20.2 ± 11.7 | 27.4 ± 10.2 ⁎⁎ | 32.2 ± 21.3 § | 34.1 ± 20.5 § | <0.0001 |
| Insulin sensitivity index (mg × L 2 × mmol −1 × mU −1 × min −1 ) | 98 ± 52 | 55 ± 26 ⁎ | 62 ± 34 ⁎ | 54 ± 36 § | <0.0001 |
| Odds ratio for lower insulin sensitivity (95% CI) | 1 (reference category) | 3.90 (1.55–9.85) | 2.13 (1.01–4.46) | 3.67 (2.45–5.52) | <0.0001 |
| Odds ratio for high waist circumference ‡‡ (95% CI) | 1 (reference category) | 2.59 (1.09–6.13) | 11.8 (4.10–34.02) | 13.5 (8.50–21.45) | <0.0001 |
| Odds ratio for high triglycerides §§ (95% CI) | 1 (reference category) | 3.55 (1.27–9.92) | 1.60 (0.52–4.89) | 26.33 (16.25–42.64) | <0.0001 |
| Odds ratio for low HDL level ¶¶ (95% CI) | 1 (reference category) | 2.74 (1.02–7.36) | 3.83 (1.85–7.83) | 24.93 (15.97–38.89) | <0.0001 |
| Odds ratio for high blood pressure ∥∥ (95% CI) | 1 (reference category) | 10.34 (2.86–37.28) | 4.36 (2.10–9.05) | 13.02 (8.26–20.52) | <0.0001 |
‡ Refers to results after analyses with adjustment for gender.
§ p <0.0001 compared to subjects without metabolic syndrome after adjustment for age and gender;
¶ p <0.0001 compared to subjects with metabolic syndrome using FPG after adjustment for age and gender;
# p <0.0001 compared to subjects with metabolic syndrome using HbA1c alone after adjustment for age and gender.
‡‡ Waist circumference >102 cm for men and >88 cm for women.
¶¶ HDL <40 mg/dl in men and <50 mg/dl in women.
∥∥ Systolic blood pressure ≥130 mm Hg or diastolic blood pressure ≥85 mm Hg.
We also compared the correlation of the glycemic measures with the cardiometabolic variables and observed that HbA1c correlated more strongly with the body mass index (r = 0.31, p <0.001), waist circumference (r = 0.32, p <0.0001), HDL (r = −0.13, p <0.0001), triglycerides (r = 0.20, p <0.0001), fibrinogen (r = 0.22, p <0.0001), and white blood cell count (r = 0.21, p <0.0001) than did the FPG (with corresponding values of r = 0.24, p <0.0001; r = 0.23, p <0.0001; r = −0.08, p = 0.03; r = 0.14, p <0.0001; and r = 0.04, p = 0.86). The correlations of the systolic blood pressure (r = 0.16, p <0.0001), pulse pressure (r = 0.11, p <0.0001), 2-hour postchallenge plasma glucose (r = 0.32, p <0.0001), fasting insulin (r = 0.32, p <0.0001), and insulin sensitivity (r = −040, p <0.0001) with FPG were stronger than the corresponding correlations with HbA1c (r = 0.12, p <0.0001; r = 0.08, p = 0.04; r = 0.24, p <0.0001; r = 0.33, p <0.0001; and r = −0.22, p <0.0001).
Next, we evaluated the diagnostic property of HbA1c and FPG to identify insulin-resistant subjects (lowest ISI quartile) by calculating a ROC curve. The area under the ROC curve of FPG for identifying insulin-resistant subjects was 0.771 (95% CI 0.739 to 0.800; Figure 2 ). The area under the ROC curve of HbA1c for identifying insulin-resistant subjects was 0.648 (95% CI 0.613 to 0.681). The difference between the 2 areas under the ROC curve (0.123, 95% CI 0.0700 to 0.176) was statistically significant (p <0.0001; Figure 2 ). The relatively low sensitivity of the HbA1c criterion for identifying subjects with insulin resistance might indicate that a significant number of subjects who were insulin resistant but were not considered to have the metabolic syndrome. To further explore this possibility, we assessed the cardiometabolic risk factors in 3 groups of subjects: those who met the criteria for the metabolic syndrome using HbA1c alone, insulin-resistant subjects (lowest ISI quartile) who did not meet the criteria for the metabolic syndrome, and those who did not meet the criteria for the metabolic syndrome and were insulin sensitive (ISI quartile 2 to 4). The insulin-resistant subjects were older and had a worse cardiometabolic risk profile than did the insulin-sensitive group ( Table 3 ).
