STOCKHOLM – Use of a "rule-in" hemoglobin A1c cut point of 6.8% and a "rule-out" value of 5.8%, with glucose testing for individuals who fall in the middle of the diagnostic cutoff, was more accurate in diagnosing type 2 diabetes than was a single cutoff value of 6.5%.
The finding from a multiethnic cohort study of 8,696 previously undiagnosed primary care patients addresses some of the concerns about false-positive and false-negative diagnoses associated with using a single measure of hemoglobin A1c. Multiple studies have shown that the 6.5% cutoff may be discordant with the results of an oral glucose tolerance test (OGTT), which is considered to be the standard diagnostic test for type 2 diabetes, said Dr. Samiul A. Mostafa, a clinical research fellow in the diabetes research unit of the University of Leicester (England).
In July 2009, an international expert committee recommended the use of hemoglobin A1c for diagnosing diabetes, with a diagnostic cutoff of 6.5% or above following a repeat confirmatory A1c test (Diabetes Care 2009;32:1327-34). In January 2010, the American Diabetes Association endorsed that recommendation (Diabetes Care 2010;33[suppl. 1]:S62-9). The European Association for the Study of Diabetes and the World Health Organization are expected to issue similar statements soon.
The study participants were identified from two systematic screening programs during 2002-2008. Three-quarters (75%) were white Europeans and 23% were South Asians from Pakistan, Bangladesh, and India. The mean A1c for the entire cohort was 5.7%. All underwent an OGTT and also had their HbA1c levels measured. Using the WHO criteria (a 2-hour plasma glucose level of 200 mg/dL or above, following a 75-g glucose load), the OGTT detected 291 individuals (3.3% of 8,696 study participants) with type 2 diabetes.
Among the white Europeans, use of the 6.5% A1c cutoff had a sensitivity of 62% and a positive predictive value of 45%. Based on an Australian study published earlier this year, the investigators chose to compare those values with a rule-out A1c cutoff of 5.5% and a rule-in cutoff of 7.0%, with a confirmatory OGTT used for those falling in between (Diabetes Care 2010;33:817-9).
That method gave an improved sensitivity of 98% and positive predictive value of 76% in the white European group. With either method, specificity and negative predictive values were close to 100%. For the South Asians, the 6.5% cutoff gave a sensitivity of 79% and positive predictive value of 36%, both of which improved to 99% and 68%, respectively, with the two–cut-point criteria. Again, specificity and negative predictive values were strong with either method, Dr. Mostafa reported.
"Impaired HbA1c," the term used for the values between the two cutoffs (5.6%-6.9%), was found in 59% of the total cohort, who thus required confirmatory tests. Noting that those in the impaired HbA1c group (55% of the total cohort) had A1c values between 5.6% and 6.4% (that is, lower than 6.5%), they tried various cut points and arrived at a rule-out value of 5.8% or below and a rule-in value of 6.8% or above. That left 28% of the total cohort in the "impaired HbA1c" category when defined as an A1c of 5.9%-6.7%.
"We believe [a rule-out value of 5.8% and a rule-in value of 6.8%] would be a more feasible strategy to implement in clinical practice," Dr. Mostafa said.
These cutoffs gave sensitivities of 92% for white Europeans and 98% for South Asians, and positive predictive values of 70% and 54%, respectively, while maintaining the nearly 100% specificity and negative predictive values for both ethnic groups. Despite the slight reductions in positive predictive values, "overall, we feel using the cut points of 5.8% and 6.8% is still diagnostically accurate, with the major advantage that only a quarter of the population would have to return for a subsequent test," he said.
In a final analysis, the investigators looked at mean HbA1c values in various undiagnosed populations. Compared with the U.K. cohort’s mean of 5.7%, the Australian cohort had a mean of 5.1%, which resulted in 24% falling into their 5.6%-6.9% "impaired HbA1c" category. That led to the hypothesis that broader cut points are acceptable when mean A1c is relatively low, but a tighter range is required when mean A1c is higher.
Studies conducted in Denmark and in India showed relatively high mean A1c values of 5.8% and 5.6%, respectively, similar to the U.K. group. Those populations would probably need to use narrower cut points to reduce the number of people who would require subsequent testing. In contrast, like the Australians, data from the U.S. National Health and Nutrition Examination Survey showed a 5.2% mean A1c, meaning that a broader cut point range might be feasible, Dr. Mostafa said.