The Prognostic Value of Glycated Hemoglobin
The Prognostic Value of Glycated Hemoglobin
Selvin E, Steffes MW, Zhu H, et al.
N Engl J Med. 2010;362:800-811
The Atherosclerosis Risk in Communities (ARIC) study is a community-based assessment of 15,792 middle-aged adults in 4 US communities. After excluding individuals with established diabetes or cardiovascular disease (CVD), a sample size of 11,092 was followed for up to 15 years (4 visits at about 3-year intervals) for onset of new diabetes, new CVD, stroke, and all-cause mortality. Using data from visit 2 (1990-1992), the only visit for which stored whole-blood samples were available for measurement of glycated hemoglobin A1c (HbA1c), the investigators compared the prognostic value of HbA1c and fasting plasma glucose (FPG) for identifying adults at risk for diabetes or CVD. Statistical models examined categories of HbA1c as predictors of each of those outcomes, first adjusted for age, sex and race, then for the addition of lipid levels, body mass index, waist-to-hip ratio, hypertension, family history of diabetes, education, alcohol use, physical activity, smoking, and finally, FPG. This process was repeated for the examination of categories of FPG as predictors of the outcomes, with the final models adjusted for the addition of HbA1c. (Table)
Table. HbA1c Levels and Corresponding Multivariate Hazard Ratios
HbA1c = hemoglobin A1c
The table shows multivariable-adjusted hazard ratios (95% confidence intervals) for diagnosed diabetes for different HbA1c levels. The hazard ratios for stroke were similar, but for all-cause mortality, HbA1c displayed a J-shaped association curve. All associations remained significant after adjustment for the baseline FPG. In contrast, the associations between the FPG and the risk for CVD or death were not significant in models with adjustment for all covariates as well as for HbA1c. For coronary heart disease, measures of risk discrimination showed significant improvement when HbA1c was added to models including FPG.
Change rarely comes easily, and the American Diabetes Association's decision to recommend HbA1c as a diagnostic tool for diabetes will undoubtedly spawn a number of studies that alternatively confirm or deny the wisdom of the decision. The current study is a well-done analysis of well-known data (ARIC) that support the use of HbA1c, because the HbA1c and FPG values performed similarly for a diabetes diagnosis. It would have been nice to see the HbA1c and FPG values categorized in more directly comparable ways, with the same number of categories and each with an approximately similar level of dysglycemia.
Perhaps more important than the diabetes findings was that HbA1c values significantly predicted CVD or death, whereas FPG levels were not significant after adjustment for other risk factors. The additional usefulness of HbA1c may therefore justify the marginal cost of the test relative to the inexpensive FPG.
Furthermore, the recent ADVANCE [Action in Diabetes and Vascular Disease: Preterax and Diamicron MR Controlled Evaluation], ACCORD [Action to Control Cardiovascular Risk in Diabetes], and VADT [Veterans Affairs Diabetes Trial] trials left us wondering about the value of tight glycemic control in reducing CVD risk. One shortcoming of each of these trials was that most participants had had diabetes for many years, and the designs could not account for the long-term accumulation of glycemic burden. In other words, the damage from high HbA1c may have already been done, so that tight control during the trials had relatively little effect.
The current ARIC analysis demonstrates that higher HbA1c levels, even in the normal range, increase CVD risk. Although far from conclusive, these observational results suggest that keeping HbA1c as near normal as possible even before diabetes onset may help prevent CVD.
Abstract
Glycated Hemoglobin, Diabetes, and Cardiovascular Risk in Nondiabetic Adults
Selvin E, Steffes MW, Zhu H, et al.
N Engl J Med. 2010;362:800-811
Study Summary
The Atherosclerosis Risk in Communities (ARIC) study is a community-based assessment of 15,792 middle-aged adults in 4 US communities. After excluding individuals with established diabetes or cardiovascular disease (CVD), a sample size of 11,092 was followed for up to 15 years (4 visits at about 3-year intervals) for onset of new diabetes, new CVD, stroke, and all-cause mortality. Using data from visit 2 (1990-1992), the only visit for which stored whole-blood samples were available for measurement of glycated hemoglobin A1c (HbA1c), the investigators compared the prognostic value of HbA1c and fasting plasma glucose (FPG) for identifying adults at risk for diabetes or CVD. Statistical models examined categories of HbA1c as predictors of each of those outcomes, first adjusted for age, sex and race, then for the addition of lipid levels, body mass index, waist-to-hip ratio, hypertension, family history of diabetes, education, alcohol use, physical activity, smoking, and finally, FPG. This process was repeated for the examination of categories of FPG as predictors of the outcomes, with the final models adjusted for the addition of HbA1c. (Table)
Table. HbA1c Levels and Corresponding Multivariate Hazard Ratios
| HbA1c Level | Multivariate-Adjusted Hazard Ratio |
|---|---|
| < 5% | 0.52 (0.40-0.69) |
| 5% to < 5.5% | 1.00 (reference) |
| 5.5% to < 6% | 1.86 (1.67-2.08) |
| 6% to < 6.5% | 4.48 (3.92-5.13) |
| ≥ 6.5% | 16.47 (14.22-19.08) |
The table shows multivariable-adjusted hazard ratios (95% confidence intervals) for diagnosed diabetes for different HbA1c levels. The hazard ratios for stroke were similar, but for all-cause mortality, HbA1c displayed a J-shaped association curve. All associations remained significant after adjustment for the baseline FPG. In contrast, the associations between the FPG and the risk for CVD or death were not significant in models with adjustment for all covariates as well as for HbA1c. For coronary heart disease, measures of risk discrimination showed significant improvement when HbA1c was added to models including FPG.
Viewpoint
Change rarely comes easily, and the American Diabetes Association's decision to recommend HbA1c as a diagnostic tool for diabetes will undoubtedly spawn a number of studies that alternatively confirm or deny the wisdom of the decision. The current study is a well-done analysis of well-known data (ARIC) that support the use of HbA1c, because the HbA1c and FPG values performed similarly for a diabetes diagnosis. It would have been nice to see the HbA1c and FPG values categorized in more directly comparable ways, with the same number of categories and each with an approximately similar level of dysglycemia.
Perhaps more important than the diabetes findings was that HbA1c values significantly predicted CVD or death, whereas FPG levels were not significant after adjustment for other risk factors. The additional usefulness of HbA1c may therefore justify the marginal cost of the test relative to the inexpensive FPG.
Furthermore, the recent ADVANCE [Action in Diabetes and Vascular Disease: Preterax and Diamicron MR Controlled Evaluation], ACCORD [Action to Control Cardiovascular Risk in Diabetes], and VADT [Veterans Affairs Diabetes Trial] trials left us wondering about the value of tight glycemic control in reducing CVD risk. One shortcoming of each of these trials was that most participants had had diabetes for many years, and the designs could not account for the long-term accumulation of glycemic burden. In other words, the damage from high HbA1c may have already been done, so that tight control during the trials had relatively little effect.
The current ARIC analysis demonstrates that higher HbA1c levels, even in the normal range, increase CVD risk. Although far from conclusive, these observational results suggest that keeping HbA1c as near normal as possible even before diabetes onset may help prevent CVD.
Abstract