Back grey_arrow_rt.gif
 
 
Identifying Adults at High Risk for Diabetes and Cardiovascular Disease Using Hemoglobin A1c: (pdf attached)
 
 
  Download the PDF here
 
National Health and Nutrition Examination Survey 2005-2006
 
Am J Prev Med 2011;40(1)11-17
 
"In summary, A1c-based testing appears to have value for identifying adults who are at high predicted risk for the future development of type 2 diabetes and CVD. Because of its practical nature and wide availability, A1c-based testing should be considered as a means to identify greater numbers of high-risk adults in clinical settings. If one important goal is to approximate the composite risk for diabetes and CVD among people who are already included in the 2003 ADA definition for prediabetes, then A1c ≥5.5% is a suitable test threshold for defining high risk. Alternatively, if the goal is to approximate composite risk levels among people enrolled in previous prevention trials such as the DPP, then a test threshold of ≥5.7% may be more appropriate. Recommendations to use A1c in the identification of adults at high risk for type 2 diabetes will alter the current prevalence estimates for high risk in the population. Though important, this concern should remain secondary to the importance of adopting the A1c test as a practical strategy for identifying the large population of high-risk individuals who remain unidentified today and for whom evidence-based preventive interventions are likely to be cost effective."
 
Ronald T. Ackermann, MD, MPH, Yiling J. Cheng, MD, PhD, David F. Williamson, PhD, Edward W. Gregg, PhD From the Department of Medicine, Indiana University School of Medicine (Ackermann), Indianapolis, Indiana; Division of Diabetes Translation (Cheng, Gregg), CDC; Hubert Department of Global Health, Rollins School of Public Health, Emory University (Williamson), Atlanta, Georgia
 
Background
 
The American Diabetes Association (ADA) recently proposed the use of hemoglobin A1c as a practical and valid strategy to identify high-risk people for whom delivery of an intensive lifestyle intervention to prevent type 2 diabetes is likely to be cost effective.
 
Purpose
 
To estimate composite risks of developing diabetes and cardiovascular disease (CVD) for adults with different hemoglobin A1c test results and to compare those risks with those of adults who met the 2003 ADA definition for prediabetes.
 
Methods
 
Cross-sectional data from the 2005-2006 National Health and Nutrition Examination Survey were analyzed in 2009. The method of Stern and colleagues was used to estimate the 7.5-year probability of type 2 diabetes, and the Framingham General CVD Risk Engine was used to estimate the 10-year probability of CVD for adults with different A1c results. Sample weights were used to account for sampling probability and to adjust for noncoverage and nonresponse.
 
Results
 
Among adults meeting the 2003 ADA definition for prediabetes, the probabilities for incident type 2 diabetes (over 7.5 years) and CVD (over 10 years) were 33.5% and 10.7%, respectively. Use of A1c alone, in the range of 5.5% to <6.5%, would identify a population with comparable risks for diabetes (32.4% [SE=1.2%]) and CVD (11.4% [SE=0.6%]). A slightly higher cutoff (≥5.7%) would identify adults with risks of 41.3% (SE=1.5%) for diabetes and 13.3% (SE=0.8%) for CVD-risks that are comparable to people enrolled in the Diabetes Prevention Program.
 
Conclusions
 
A1c-based testing in clinical settings should be considered as a means to identify greater numbers of adults at high risk of developing type 2 diabetes and CVD.
 
Intensive lifestyle interventions and select medications are cost effective for delaying or preventing the development of type 2 diabetes among high-risk adults.1 Lifestyle interventions offer benefit by not only preventing type 2 diabetes but also reducing the burden of cardiovascular risk factors.2, 3 Identifying people who are at high risk for both type 2 diabetes and cardiovascular disease (CVD) is a pivotal step toward maximizing the cost effectiveness of services to prevent diabetes in both clinical practice and the community. However, the selection of a particular test to define "high risk" can have tremendous implications for prevalence estimates, health policies, costs of care, and for individualized decisions about preventive therapy.
 
Recent meta-analyses show that adults with fasting plasma glucose (FPG) concentrations in the range of 110-125 mg/dL or 2-hour plasma glucose (2hPG) concentrations in the range of 140-199 mg/dL are at increased risk for the development of both type 2 diabetes4 and CVD.5 Unfortunately, several practical considerations limit the routine use of FPG and 2hPG tests in primary care settings. Both tests require a patient to return on a separate day after an overnight fast, which is a barrier to test completion. In the U.S. today, the FPG test remains underutilized and the 2hPG test is rarely performed.6 As a result, only ∼7% of American adults with prediabetes are actually aware of their status.7
 
Given these challenges, the hemoglobin A1c (A1c) test is an attractive alternative for identifying adults at high risk for type 2 diabetes. The A1c does not require an overnight fast and is already used routinely in primary care settings to guide therapeutic decisions for patients with diagnosed diabetes. Recent studies also have strengthened the case for expanding the use of A1c to identify high-risk people by demonstrating that higher levels of A1c across the nondiabetic range (i.e., 5.0% to 6.5%) are associated with both a higher prevalence of cardiovascular risk factors8, 9 and greater numbers of incident cardiovascular events.10, 11, 12
 
In June 2009, the International Expert Committee (IEC), which represents the American Diabetes Association (ADA), the European Association for the Study of Diabetes, and the International Diabetes Federation, recommended adoption of the A1c test for the diagnosis of diabetes.13 The IEC also stated that "individuals whose A1c values are close to the 6.5% A1c threshold of diabetes (i.e., 6.0%) should receive demonstrably effective [prevention] interventions." Unfortunately, this recommendation does not appear to have considered carefully whether a cutoff of 6.0% would exclude a large proportion of truly high-risk adults. With concern for this possibility, the ADA recently recommended use of a lower A1c test threshold of ≥5.7% to identify adults who are at particularly high risk for developing diabetes.14
 
Although a handful of prospective studies have evaluated differences in the future development of diabetes associated with different A1c test results, these studies were largely not population-based, used varying methods to define diabetes onset, and did not explore differences in risk associated with small (e.g., 0.1%) increments in A1c across a wide range of subdiabetic levels (5.0%-6.5%).12, 15, 16, 17, 18 Many past studies also focused largely on whether A1c is an independent predictor of incident diabetes or cardiovascular events and not whether it has utility as a single test for identifying people in whom composite risk is high because of multiple co-occurring risk factors. Because a composite risk score can be challenging to calculate from multiple clinical variables at the point of care, the validation of a single and readily available test to guide decisions about whether to offer more intensive resources for lifestyle change could advance diabetes prevention efforts.
 
To address this issue, the present study used cross-sectional data from the 2005-2006 U.S. National Health and Nutrition Examination Survey (NHANES) to explore the associations of A1c with existing models for estimating the composite risks for future CVD and type 2 diabetes from multiple traditional risk factors in U.S. adults.19, 20 This analysis provides a comparison between the use of A1c and FPG testing for identifying adults with high levels of composite risk, as well as the impact on U.S. prevalence estimates for prediabetes with different possible A1c test thresholds to define high risk.
 
Results
 
Nonselective A1c Testing to Identify Adults at High Risk for Diabetes or Cardiovascular Disease

 
Among all U.S. adults aged ≥18 years without self-reported diabetes or prior CVD, 30.0% (SE=1.6%) met the 2003 ADA criteria for prediabetes. Within this large group, the mean predicted probabilities for incident diabetes (over 7.5 years) and CVD (over 10 years) were 33.5% (SE=1.0%) and 10.7% (SE=0.8%), respectively. Associations of different A1c test thresholds with predicted risk for diabetes and CVD in adults without self-reported diabetes or prior CVD are shown in Table 1These data show a nonlinear, graded relationship of A1c with risk for both incident diabetes over 7.5 years and incident CVD over 10 years. Use of A1c alone, in the range of 5.5% to <6.5%, would result in identification of a subpopulation with levels of predicted risk for incident diabetes (32.4% [SE=1.2%]) and CVD (11.4% [SE=0.6%]) that are comparable to those of adults meeting the 2003 ADA criteria.
 
After excluding people with self-reported diabetes or prior CVD, ∼10.2% (SE=1.1%) of U.S. adults met the more-focused high-risk glycemic criteria required for enrollment in the DPP. Within this group, the mean risks for incident diabetes (over 7.5 years) and CVD (over 10 years) were 42.2% (SE=2.8%) and 13.6% (SE=1.4%), respectively. Use of A1c in the range of 5.7% to <6.5% Table 1 would result in identification of a population with a similar predicted risk for incident diabetes (41.3% [SE=1.5%]) and CVD (13.3% [SE=0.8%]).
 
Targeted A1c Testing to Identify Adults at High Risk for Diabetes or Cardiovascular Disease
 
Predicted risk levels for incident diabetes and CVD were higher when testing was applied only to adults with a BMI ≥25 kg/m2 who also had evidence of either high blood pressure or high cholesterol. After excluding people with self-reported diabetes or prior CVD, an estimated 42.4% (SE=2.5%) of adults in this targeted subgroup met the 2003 ADA criteria for prediabetes. Among this high-risk group, the mean predicted risks for incident diabetes (over 7.5 years) and CVD (over 10 years) were 43.2% (SE=1.1%) and 13.4% (SE=1.3%), respectively. Targeted testing with the use of A1c alone, in the range of 5.5% to <6.5%, would identify people with comparable levels of predicted risk for incident diabetes (41.3% [SE=1.7%]) and CVD (13.9% [SE=0.9%]; Table 2.
 
Impact of A1c Use on Estimates of High-Risk Prevalence Among U.S. Adults Without excluding people with self-report diabetes or prior CVD, 33.9% (SE=1.7%) of the U.S. adult population met the 2003 ADA criteria for prediabetes (Table 3). If people who have an A1c of 5.5% to <6.5% are also considered high-risk, the total prevalence would increase to 47.2% (SE=1.7%). This larger group would be made up of 13.3% of adults meeting only the A1c criteria, 16.9% meeting only the 2003 ADA criteria, and 17.0% meeting both. Conversely, if people who have an A1c of 5.7% to <6.5% are considered high-risk, the total prevalence would increase to 39.1% (SE=1.7%). This larger group would be made up of 5.2% of adults meeting only the A1c criteria, 23.9% meeting only the 2003 ADA criteria, and 10.0% meeting both.
 
Conclusion
 
A large majority of people who would meet the 2003 ADA classification for prediabetes are unaware of their risk because of challenges to performing FPG and 2hPG tests routinely.6, 7 A simpler form of high-risk testing could improve diabetes prevention efforts by substantially increasing the numbers of individuals who complete testing. The current study indicates that the A1c test may provide a badly needed, clinically practical indicator of the composite risk for incident diabetes and CVD. Within a test range of 5.5% to <6.5%, the A1c test identifies a population with levels of predicted risk for future diabetes and CVD that are similar to people identified by the 2003 ADA classification criteria for prediabetes. This concordance of risk persists regardless of whether case-finding efforts target a general population or focus on high-risk subgroups that are similar to those recommended by the USPSTF.25
 
An additional important finding of the present study is that A1c testing identifies some high-risk people who would not have been identified by currently recommended strategies that are based on FPG and 2hPG results alone. The ADA recently recommended a new definition for high-risk individuals that includes everyone meeting the 2003 ADA definition for prediabetes or those with A1c of 5.7% to <6.5%. This definition increases the total prevalence of high-risk Americans from 33.9% to an estimated 39.1% of the U.S. adult population. When evaluating strategies to classify people at high risk, it is important to consider whether those strategies identify people for whom preventive interventions are likely to be most cost effective. Although the ADA defined prediabetes in 2003 to include people with either a 2hPG between 140 and 199 mg/dL or an FPG between 100 and 125 mg/dL, people with elevated FPG alone (i.e., 2hPG <140 mg/dL) have not been included in most randomized prevention trials to date26 and are at lower risk for future diabetes and CVD than participants in those studies. In this context, a decision regarding the most appropriate A1c range for identifying high-risk people should probably not be based on estimates of risk for people with only modestly elevated FPG 100-109 mg/dL and no elevation of 2hPG. This analysis found that an A1c test threshold of ≥5.7% identifies people for whom the predicted risk for diabetes and CVD are comparable to those who meet the DPP enrollment criteria of both elevated fasting glucose and IGT.
 
This research has some limitations. First, cross-sectional data were used to predict the probability of future diabetes and CVD. These predictions should be interpreted as indicators of the aggregate risk imposed by multiple co-occurring risk factors among people with different levels of A1c and not as exact estimates of future event prevalence. Actual event prevalence may be higher or lower than predicted. Second, the Stern method was derived using data from participants in the San Antonio Heart Study,20 which enrolled a proportionately large number of Mexican Americans and excluded adults aged <25 years and >64 years. The Framingham Study enrolled a predominantly non-Hispanic white population and excluded adults aged <30 years and >74 years.19 Both methods may have lower predictive validity in other population subgroups. Unfortunately, no alternative published prediction models were identified for either diabetes or CVD that do not incorporate the A1c result in the estimation and have been validated across a broader array of population subgroups.
 
Third, the present study did not evaluate (because of limited sample sizes) whether the predictive value of A1c test cutoffs varies across different population subgroups. One recent study27 found that older adults are more likely to meet the 2hPG criterion for high risk compared to younger adults. This suggests that the identification of high-risk people by only one form of testing can vary by age and that it would be prudent to encourage FPG or 2hPG testing in adults with multiple risk factors who are found to have a subthreshold A1c result. In another recent publication,28 average A1c levels at the time of screening for the DPP were 0.4% higher among African-American adults with IGT than in their non-Hispanic white counterparts, even after adjusting for differences in other characteristics.
 
This finding implies that use of a single A1c cutoff would identify greater numbers of African Americans than using only FPG or 2hPG testing. It has been hypothesized that A1c levels in some racial-minority groups might be elevated, in part, because of undiagnosed disorders of hemoglobin glycation or red cell survival. However, it is still unclear whether adults of minority race/ethnicity are at higher or lower overall risk for diabetes and CVD events when compared to non-Hispanic whites with the same level of A1c. Last, decisions to expand efforts that identify people at high risk for developing type 2 diabetes presume that those individuals will benefit from knowing their risk. However, these efforts may be premature unless parallel efforts are undertaken to expand the availability of prevention programs that will support high-risk people in efforts to reduce their risk.
 
In summary, A1c-based testing appears to have value for identifying adults who are at high predicted risk for the future development of type 2 diabetes and CVD. Because of its practical nature and wide availability, A1c-based testing should be considered as a means to identify greater numbers of high-risk adults in clinical settings. If one important goal is to approximate the composite risk for diabetes and CVD among people who are already included in the 2003 ADA definition for prediabetes, then A1c ≥5.5% is a suitable test threshold for defining high risk. Alternatively, if the goal is to approximate composite risk levels among people enrolled in previous prevention trials such as the DPP, then a test threshold of ≥5.7% may be more appropriate. Recommendations to use A1c in the identification of adults at high risk for type 2 diabetes will alter the current prevalence estimates for high risk in the population. Though important, this concern should remain secondary to the importance of adopting the A1c test as a practical strategy for identifying the large population of high-risk individuals who remain unidentified today and for whom evidence-based preventive interventions are likely to be cost effective.
 
Support for this research was provided by the CDC and the Robert Wood Johnson Foundation Physician Faculty Scholars Program. The findings and conclusions in this report are those of the authors and do not necessarily represent the official positions of the CDC.
 
No financial disclosures were reported by the authors of this paper.
 
 
 
 
  icon paper stack View Older Articles   Back to Top   www.natap.org