Are There Clinical Implications of Racial Differences in HbA_1c? A Difference, to Be a Difference, Must Make a Difference

In a major change to clinical practice guidelines, the International Expert Committee first recommended the use of hemoglobin A_1c (HbA_1c) for the diagnosis of diabetes in 2009 (1). This recommendation was codified in the American Diabetes Association’s Clinical Practice Recommendations in 2010 (2) and has been adopted by the World Health Organization and numerous other professional groups across the globe (3,4). Given the long-standing use of HbA_1c for diabetes control and its strong link to complications, the use of HbA_1c in diagnosis of diabetes seemed advisable and advantageous. Nevertheless, the 2009 recommendations for the use of HbA_1c as a diagnostic test for diabetes were met with considerable controversy (5). Central to this controversy has been the interpretation of racial differences in HbA_1c levels. The relevance of racial differences in HbA_1c for its use in screening, diagnosis, and management of diabetes is the focus of this commentary.

In the U.S., racial and ethnic minority groups are disproportionately burdened by adverse social and economic conditions that can profoundly influence disease risk. Society-level factors such as social position, residence, material conditions (including wealth), social connections, environment, and food and physical insecurity are particularly important factors influencing risk of obesity and diabetes (6,7). There are well-documented racial disparities in the risk of diabetes, with African Americans approximately twice as likely to develop diabetes as compared with their white counterparts (8). Racial and ethnic minority groups are also disproportionately burdened by the complications of diabetes including retinopathy (9,10), chronic kidney disease (11), and lower-extremity peripheral vascular disease (12,13), with an especially high risk of amputation (14). Indeed, racial differences in end-stage renal disease represent one of the most striking racial disparities in health in the U.S. (15,16).

In numerous cohorts and in national data, it has been shown that blacks have higher HbA_1c values than whites in both the presence and absence of diabetes (17–21). Mexican Americans have values of HbA_1c that are intermediate between blacks and whites (9,22). Comparisons of HbA_1c in other racial/ethnic groups and non-U.S. populations are scarce but suggest higher nondiabetic levels of HbA_1c in some groups, e.g., South Asians, black Brazilians, and Inuit populations, compared with whites or Caucasians (23–25). It is unclear what factors might be driving these differences.

Racial differences in HbA_1c have been widely cited as a potential shortcoming of HbA_1c testing for diagnosis of diabetes (26–29). On the basis of evidence for racial differences in HbA_1c, current clinical guidelines from the American Diabetes Association state: “It is important to take…race/ethnicity…into consideration when using the A1C to diagnose diabetes” (29). However, it is not clear from the guidelines how this recommendation might be actualized. Some argue that racial differences are nonglycemic in nature, i.e., a result of factors that influence HbA_1c via pathways independent of glucose or hyperglycemia, and have suggested that HbA_1c is “invalid” or “misleading” as a diagnostic test in African Americans (30,31). Clearly this claim is not trivial: HbA_1c is widely considered the gold standard measure of chronic hyperglycemia in diabetes care. Treatment and diagnostic decisions are routinely based on HbA_1c levels. If the higher HbA_1c in blacks compared with whites is primarily due to nonglycemic factors, then HbA_1c is falsely high in blacks. If this claim is substantiated, it suggests potential disparities in diabetes care may not be real, efforts to reduce hyperglycemia in blacks may be unwarranted and could cause harm, and that we might need race-specific diagnostic and treatment thresholds.

The debate is not whether racial differences in HbA_1c exist: they do. What is not clear is why levels of HbA_1c are somewhat higher in blacks compared with whites.

HbA_1c is an indirect measure of hyperglycemia (32–35), but it is well established that the primary determinant of HbA_1c is circulating glucose level (33). It has been postulated that racial differences in HbA_1c might be explained by differences in hemoglobin-related factors. Red cell turnover may be the most important unmeasured nonglycemic determinant of HbA_1c (36), but there is currently no direct evidence of racial differences in red cell turnover that might explain racial differences in HbA_1c. The impact of red cell turnover on HbA_1c in the general population is not well understood because of major difficulties in its measurement (37). Certain conditions such as glucose-6-phosphate dehydrogenase deficiency and specific hemoglobin variants (e.g., sickle cell) are more common in African Americans than in whites. Glucose-6-phosphate dehydrogenase deficiency causes hemolysis and can result in a lowering of HbA_1c, and sickle cell trait (and other hemoglobin variants) can falsely lower or raise HbA_1c or may have no effect depending on the method of HbA_1c measurement (38).

The racial differences we see in HbA_1c levels across populations may reflect real differences in circulating average (nonfasting) glucose that are reflected in HbA_1c but not (or not as much) in fasting glucose or 2-h glucose. Given the considerable black–white disparities in risk of diabetes and other major health conditions, perhaps it is not so surprising that there are racial differences in HbA_1c even after adjusting for fasting glucose. A single measurement of fasting glucose or 2-h glucose does not fully reflect average glycemia and would not account for possible differences in true circulating average glucose between races. Differences in body composition, physical activity, diet, lifestyle, stress, and/or environmental and neighborhood-level factors might affect circulating nonfasting glucose levels and contribute to the racial differences in HbA_1c. Such parameters may not be fully captured by standard assessments in large epidemiologic studies, leaving open the possibility that racial disparities in HbA_1c are not artifactual but reflect black–white differences in true circulating nonfasting glucose.

Importantly, the higher levels of HbA_1c are also seen for other biomarkers of chronic hyperglycemia, specifically fructosamine and glycated albumin (39–42). Because fructosamine and glycated albumin are unaffected by the hematologic factors that might affect HbA_1c, racial differences in erythrocyte turnover or hemoglobin glycation cannot explain racial differences in these hemoglobin-independent serum biomarkers of hyperglycemia. The racial differences in fructosamine and glycated albumin support a difference in glycemia itself.

Genetic differences undoubtedly contribute to both glycemic and nonglycemic variation in measures of hyperglycemia including HbA_1c. The clinical significance of a nonglycemic genetic contribution is uncertain, particularly in persons without genetic hemoglobin abnormalities. We have previously shown that genetic ancestry does not contribute substantially (<1%) to variability in HbA_1c among African Americans (43). Furthermore, no known genetic variants differ substantially enough between persons of African compared with Caucasian ancestry to explain racial differences in HbA_1c in the general population (44–46). Although the current evidence does not rule out the possibility of genetic nonglycemic determinants of HbA_1c, there is no clear evidence that genetic differences contribute substantially to racial differences in HbA_1c. Race is primarily a social construct (47,48), and the literature does suggest that we should not treat race like a biological factor that should be used to adjust HbA_1c values.

The cause or causes of racial disparities in HbA_1c are incompletely understood, and we cannot rule out a small but systematic nonglycemic difference. Research is needed to understand the full determinants of HbA_1c, particularly the impact of red cell turnover on differences across population subgroups. Nonetheless, we should recognize that, in the diabetic range, the primary determinant of HbA_1c is circulating ambient glucose; other factors are likely to have a relatively small influence.

Are there nonglycemic determinants of HbA_1c? Certainly. Do these nonglycemic determinants play a large role at diagnostic or higher (diabetic) levels of HbA_1c in most of the population? Probably not. Are there studies that provide direct evidence that nonglycemic factors explain racial differences in HbA_1c? No. The question is then not whether there are racial differences in HbA_1c as an accurate index of chronic hyperglycemia. The important question now is: Are the observed racial differences in HbA_1c level clinically meaningful?

A major justification for using HbA_1c as a diagnostic test for diabetes is the strong evidence linking it to future diabetes and major clinical complications in ethnically diverse populations (19,49–54). If the observed systematically higher HbA_1c levels in African Americans as compared with whites stem from racial differences not in glucose exposure but from nonglycemic factors, then HbA_1c should be a weaker predictor of diabetic complications in African Americans, especially compared with fasting glucose. The current diabetes diagnostic cut point of HbA_1c 6.5% is supported by epidemiologic evidence for a high prevalence of retinopathy beginning above this threshold (1,55,56), with key studies in multiethnic U.S. study populations (56,57), Malay adults in Singapore (58), and Australian (59), Pima Indian (60), Egyptian (55,61), Korean (62), Chinese (63), and Japanese (64,65) populations. In analyses of data from the National Health and Nutrition Examination Survey (NHANES), investigators have directly compared the prognostic value of clinical categories of HbA_1c in populations of Mexican Americans, African Americans, and whites. These analyses found no evidence for racial/ethnic differences in the relative association of HbA_1c with prevalent retinopathy, suggesting that current diabetes clinical cut points should be interpreted similarly in whites, African Americans, and Mexican Americans (9,10). Randomized clinical trials in persons with diabetes have further demonstrated that lowering HbA_1c reduces the risk of microvascular disease, regardless of race/ethnicity (66).

For this report we conducted analyses of two population-based studies, the community-based Atherosclerosis Risk in Communities (ARIC) Study and the nationally representative NHANES, to compare associations of diagnostic categories of HbA_1c and fasting glucose with major long-term diabetic complications in black, Mexican American, and white persons. We analyzed HbA_1c and fasting glucose data from 11,018 participants aged 48–68 years with no history of cardiovascular disease who attended the second examination of the ARIC Study from 1990 to 1992. During a median of approximately 20 years of follow-up, there were 279 peripheral vascular disease events, 1,550 cases of chronic kidney disease, 2,205 cardiovascular (coronary heart disease or stroke) events, and 2,999 deaths. Comparing the hazard ratios across clinical categories of HbA_1c and fasting glucose reveals that, in general, HbA_1c is more strongly associated with future clinical outcomes as compared with fasting glucose and the relative risk associations appear similar in blacks and whites (Table 1). NHANES III, which is linked to national mortality data (but not nonfatal outcomes), also included measurements of HbA_1c and fasting glucose in non-Hispanic black, non-Hispanic white, and Mexican American adults. Thus, similar analyses can be conducted in this nationally representative cohort. In an analysis of 12,722 NHANES III (1988–1994) participants aged 20 years or older with HbA_1c measurements (and 5,676 with fasting glucose), there were 804 total deaths of which 363 were from cardiovascular causes during a median of approximately 19 years of follow-up. In NHANES III, clinical categories of HbA_1c in non-Hispanic blacks were similarly or more strongly associated with cardiovascular and all-cause mortality as compared with non-Hispanic whites (Table 2).

These data from ARIC and NHANES demonstrate that HbA_1c ≥6.5% is a risk factor for future development of peripheral vascular disease, kidney disease, cardiovascular disease, and death across racial/ethnic groups. We see patterns of association of HbA_1c diagnostic categories that are generally similar or stronger than those for fasting glucose; our results do not support the contention that HbA_1c is a weaker predictor of outcomes compared with fasting glucose in African Americans compared with whites. These results extend and update prior publications (18,19,53,67–69), and, taken as a whole, the current literature demonstrates that race-specific HbA_1c cut points for diagnosis of diabetes would not be consistent with long-term risk associations.

In other studies in ARIC, we also observed that associations of nontraditional biomarkers of hyperglycemia (fructosamine and glycated albumin) with clinical outcomes were also similar in blacks and whites (39). Ultimately, the literature suggests that HbA_1c is a similarly valid diagnostic and prognostic tool in persons of different races/ethnicities and supports recommendations for using the same HbA_1c diagnostic cut points across racial/ethnic groups (10,70). To quote a saying commonly attributed to Gertrude Stein: “A difference, to be a difference, must make a difference.”

As with any clinical test, the strength and limitations of HbA_1c need to be understood and communicated. Each HbA_1c test result needs to be interpreted in the context of the individual patient. Although population-level evidence is critical to guide individual decision-making, diabetes clinical practice guidelines have increasingly recognized the need for individualization of diabetes treatment (71–73). To most effectively address the diabetes epidemic, we need to improve our approaches to preventing and treating diabetes and tailor these approaches to each individual.

The evidence from population-based studies suggests that HbA_1c is a useful and valid test of hyperglycemia across racial/ethnic groups. Indeed, studies using modern HbA_1c assays have now shown that HbA_1c is more strongly associated with outcomes as compared with fasting glucose or 2-h postprandial glucose (19,54). There is robust evidence that HbA_1c is associated with microvascular and macrovascular outcomes in diverse populations. There is no compelling evidence that the validity of HbA_1c as a measure of hyperglycemia and the prognostic value of clinical categories of HbA_1c differ substantially according to race.

Certainly more work needs to be done to understand the causes of racial differences in HbA_1c and the contribution of nonglycemic factors. But race is a poor surrogate for differences in underlying causes of disease risk, and suggestions for racially based medical decisions are disquieting. If anything, we need less emphasis on using race to define health and guide medical decision making. With respect to HbA_1c, we need to understand what might be causing disparities lest we inappropriately withhold a useful and prognostic test from a subgroup of the population known to be at high risk for diabetes and its complications. There is no evidence that HbA_1c testing will lead to “overdiagnosis” of diabetes in African Americans. There is, however, a real concern that recommendations to avoid or interpret HbA_1c results differently in racial/ethnic minority populations may actually increase health disparities.

Acknowledgments. E.S. is indebted to Drs. Larry Appel and Morgan Grams, Welch Center for Prevention, Epidemiology and Clinical Research, Johns Hopkins University, Baltimore, MD, for their valuable comments on an early draft of this article and to the staff and participants of the ARIC Study for their important contributions to this work. E.S. also thanks Yuan Chen, Welch Center for Prevention, Epidemiology and Clinical Research, Johns Hopkins University, Baltimore, MD, for assistance with statistical analyses.

Funding. This work is supported by National Institutes of Health National Institute of Diabetes and Digestive and Kidney Diseases grants 2R01-DK-089174 and K24-DK-106414 to E.S. The ARIC Study is carried out as a collaborative study supported by National Heart, Lung, and Blood Institute contracts (HHSN268201100005C, HHSN268201100006C, HHSN268201100007C, HHSN268201100008C, HHSN268201100009C, HHSN268201100010C, HHSN268201100011C, and HHSN268201100012C).

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