The Insulin Resistance Atherosclerosis Study (IRAS) is an epidemiologic study of the relationships among insulin resistance, cardiovascular disease, and its known risk factors (sociodemographic variables, family history, blood pressure, dyslipidemia, obesity, and chronic inflammation) in three ethnic groups (non-Hispanic whites, African Americans, and Hispanics) and different states of glucose tolerance (normal glucose tolerance, IGT, and type 2 diabetes). The design and methods of this study have been described previously (¹⁴). In brief, the study was conducted at four clinical centers: Oakland and Los Angeles, California, San Antonio, Texas, and San Luis Valley, Colorado. A total of 1,625 individuals were enrolled between October 1992 and April 1994 (mean age 54.6 years, range 40–69 years; 56% women). After an average of 5.2 years (range 4.5–6.6 years), follow-up examinations were conducted using the same baseline protocol. The response rate was 81%, and those who attended the follow-up examination were similar to those who did not in terms of ethnicity, sex, baseline glucose tolerance status, and BMI (all comparisons, P > 0.32). The IRAS protocol was approved by local institutional review committees, and all participants provided written informed consent.

Subjects described in this study have been part of many previous reports. Among 1,065 participants who were free of diabetes (2003 ADA criteria), 22 died during the follow-up period and 890 (85.3%) returned to the follow-up examination. We excluded 13 participants because of treatment with glucose-lowering agents and 22 others because of missing information. This report is limited to 855 of 1,030 eligible participants (83.0%) who attended the follow-up examination, because A1C was only measured at this time point (mean age 59.8 years; range 44–75 years).

Characteristics of participants by A1C categories are presented in Table 1. Most metabolic variables including S_i and AIR worsened with increasing A1C.

A1C 5.7– 6.4% identifies a much smaller proportion of individuals at increased risk of diabetes than do IFG and IGT, particularly among non-Hispanic whites and lean individuals. A1C is a less precise correlate of insulin resistance and insulin secretion in studies of metabolism than single determinations of FPG and 2-h plasma glucose and, in general, is worse for other metabolic disorders.

In participants in the 2003–2006 National Health and Nutrition Examination Survey (NHANES), A1C between 6.0 and 6.4% missed 90% of the individuals at increased risk of diabetes (¹¹). Our results suggest that the newly proposed category, A1C 5.7–6.4%, still misses three-quarters of them. Similarly, in a recent study from Qingdao, China, A1C was unable to distinguish individuals with IFG and/or IGT from those with normal glucose tolerance (¹⁷). Conversely, in a study from Chennai, India, A1C ≥5.7% detected two-thirds of individuals with IFG and/or IGT (¹⁸). The performance of A1C 5.7–6.4% may differ among studies owing to differences in the target population. Ginde et al. (¹⁹) have reported that risk stratification improves the ability of A1C to screen for undiagnosed diabetes. In our study, A1C 5.7–6.4% performs poorly among non-Hispanic whites and lean individuals. The 2010 ADA report indicated that the risk of diabetes associated with A1C 5.7–6.4% is comparable to that in participants in the Diabetes Prevention Program (DPP) (¹). Because DPP participants in the control arm had an 11% per year incidence of diabetes (²⁰), it would be beneficial to detect individuals at an earlier stage of the disease process so that they may benefit from lifestyle changes. In clinical and epidemiological settings, a significant proportion of individuals even with A1C <5.5% have either IFG or IGT (²¹). Thus, A1C 5.7–6.4% may be inadequate as the only criterion for detecting individuals at increased risk of the disease.

Inoue et al. (⁴) have reported than both IFG and A1C between 5.5 and 6.4% are independent predictors of conversion to diabetes (⁴). In nondiabetic individuals, little of the S_i (and AIR) variance is explained by A1C; in this regard, FPG and 2-h plasma glucose seem to be somewhat better. FPG correlates better with fasting insulin and first-phase insulin secretion compared with A1C. Similarly, 2-h plasma glucose correlates better with directly measured insulin resistance than does A1C. Nevertheless, each of the three glycemic measures is related to both S_i and AIR, independently of the effect of the other two measures. This is probably so because each glycemic measure reflects unique domains of glucose metabolism. Furthermore, FPG and 2-h plasma glucose may be superior to A1C in their relationship with metabolic variables other than insulin resistance and insulin secretion. The preeminence of A1C as a better indicator of future complications, at least relative to fasting glucose (¹,⁵), is in agreement with the concept of “metabolic memory,” which postulates that hyperglycemia leaves a very early imprint on the progression to complications (²²). However, A1C is inferior for metabolic abnormalities (insulin resistance and secretion, obesity, and triglycerides), which may be more determined by acute levels of glycemia.

A1C ≥6.5% identifies one-third of patients with diabetes. A larger percentage is detected by FPG ≥126 mg/dl and 2-h plasma glucose ≥200 mg/dl (45 and 87%, respectively). These results are almost identical to those derived from 2003–2006 NHANES data (30% of diabetic individuals detected by A1C ≥6.5%, 46% by FPG ≥126 mg/dl, and 90% by 2-h plasma glucose ≥200 mg/dl) (¹¹). In Qingdao, China, the 6.5% A1C cut point also detects 30% of individuals with diabetes (2003 ADA criteria) (¹⁷). In Chennai, India, however, A1C ≥6.5% detects 78% of individuals with newly diagnosed diabetes (¹⁸). Furthermore, the 6.5% A1C threshold seems to be specific for detecting undiagnosed diabetes defined by a single measure of FPG (¹²,²³) and identifies retinopathy better than do FPG ≥126 mg/dl (¹) and 2-h plasma glucose ≥200 mg/dl (¹³). It seems logical to use A1C ≥6.5% as a criterion for diagnosing diabetes provided that the plasma glucose criteria stay in use. The 6.5% A1C threshold misses a large percentage of undiagnosed diabetes, and the clinical consequences of A1C screening remain unknown. Less sensitivity could lead to undertreatment for cardiovascular disease if future recommendations integrate the risk of diabetes as part of the algorithm.

There are practical considerations as well. Although A1C is convenient, given that it does not require the fasting state, it is substantially more expensive to analyze than plasma glucose tests. Patients are commonly asked to come in fasting for other types of tests (e.g., LDL cholesterol and triglycerides), and there is little added effort required if the alternative FPG is paired with other fasting tests. There is also the concern of potential racial differences in the interpretability of A1C. The DPP and others have reported ethnic differences in the way A1C correlates to glucose levels among individuals with IGT, suggesting the potential for further detection disparities (²⁴). In this regard, our results suggest that A1C does not reflect the same domain of glucose metabolism across ethnic groups. Furthermore, our study indicates that A1C 5.7–6.4% detects at-risk individuals better as BMI worsens. Others have reported that A1C in diabetic patients with chronic kidney disease (stages 3–4) has been shown to underestimate actual glucose values (²⁵), nor is A1C accurate in individuals with anemia and hemoglobinopathies (¹). Whether this is a concern in detection of at-risk individuals is unclear.

A significant limitation of our study is the use of single determinations of plasma glucose values. Nevertheless, single determinations of FPG and 2-h plasma glucose seem to be more precise correlates of insulin resistance and secretion than A1C and, in general, better for other metabolic disorders.

In summary, the 6.5 and 5.7% A1C thresholds have a low sensitivity for detecting individuals with diabetes and at increased risk for the disease, respectively. A1C 5.7–6.4% performs poorly for identifying at-risk individuals among non-Hispanic whites but is more effective as BMI increases. A1C is a less precise correlate of insulin resistance and insulin secretion in studies of metabolism than FPG and 2-h plasma glucose and, in general, is worse for other metabolic disorders. Further studies are needed for assessing the relation of glycemic measures to metabolic abnormalities.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

This study was supported by the National Heart, Lung, and Blood Institute (grants HL-47887, HL-47889, HL-47890, HL-47892, and HL-47902) and the General Clinical Research Centers Program (National Center for Research Resources M01-RR431 and M01-RR-01346).

No potential conflicts of interest relevant to this article were reported.

C.L. contributed to discussion, wrote the manuscript, and reviewed/edited the manuscript. L.E.W. and S.M.H. researched data, contributed to discussion, and reviewed/edited the manuscript. A.J.G.H. and M.J.R. contributed to discussion and reviewed/edited manuscript. A.J.K. researched data and reviewed/edited the manuscript.

Parts of this study were presented in poster form at the 70th annual meeting of the American Diabetes Association, Orlando, Florida, 25–29 June 2010.