San Diego, CA—Screening for diabetes and prediabetes in primary care using hemoglobin (Hb)A1c misclassified a substantial percentage of patients, whereas oral glucose tolerance testing (OGTT) was more accurate, according to a large study of veterans at risk for diabetes. HbA1c screeningfailed to diagnose 10% of patients with diabetes and 42% of patients with prediabetes who were correctly identified by OGTT.
“There is a clear need for screening of high-risk individuals. In this study, HbA1c misclassified a larger percentage of patients, missing the disease when present and classifying patients as normal when they had diabetes,” stated Sandra L. Jackson, MD, Research Fellow at Emory University, Atlanta, GA, at the 2011 Scientific Sessions of the American Diabetes Association (ADA).
The study was based on a sample of 789 veterans who were part of a screening study with a planned enrollment of 1800 veterans. “Veterans are at risk for diabetes, because there are more minorities and older people in this group,” Dr Jackson pointed out. All enrollees were considered at high risk for diabetes, because of age, weight, or other risk factors. Of them, 95% were men, 73.9% were black, 51% were obese, 40% were overweight, and 9% were normal weight. The mean age was 57.5 years.
OGTT has some drawbacks, including the requirement for fasting, morning testing, and imposing a burden on the healthcare system, because it takes 2 hours to administer. Furthermore, reproducibility is poor. HbA1c screening is more convenient and has less day-to-day variation, but it can be unreliable, and there is lack of agreement on cutoffs. Also, black people and older people may have higher HbA1c at baseline.
All participants were screened using OGTT and HbA1c. In accordance with criteria from the ADA, Veterans Affairs (VA), and the International Expert Committee (IEC), 3 different cutoffs for HbA1c were used. The cutoffs were:
- ADA: 5.7% to 6.4% for prediabetes, ≥6.5% for diabetes
- VA: 5.7% to 6.9% for prediabetes, ≥7.0% for diabetes
- IEC: 6.0% to 6.4% for prediabetes, ≥6.5% for diabetes.
Less than 50% of patients showed normal results with OGTT; 10% had diabetes, and 41.7% had prediabetes. Subgroup analysis showed more prediabetes in men and in non-Hispanic whites, and a progressive increase was seen across body mass index (BMI), going from 1.5% in those with BMI of 18.5 to 24.9 kg/m2 to 15% in those with the largest BMI (>40 kg/m2).
Using OGTT, 9.9% had diabetes versus 6.7% using the cutoffs from the IEC and ADA, and 1.5% using the VA cutoff. “HbA1c had specificity of 96% to 99%, but sensitivity of 12% to 32%,” Dr Jackson said.
Of patients correctly diagnosed with diabetes with OGTT, only 32% were identified by the IEC and ADA cutoffs and only 12% by the VA cutoff. In this group, false-negatives were 38%, 50%, and 17%, respectively.
Looking at patients diagnosed as prediabetic using OGTT, only 36% were correctly diagnosed by the IEC cutoff, 61% by the ADA cutoff, and 66% by the VA cutoff. HbA1c screening incorrectly classified prediabetic patients as having normal glucose levels in 59%, 33%, and 33% of patients, respectively.
Dr Jackson pointed out that using all 3 HbA1c cutoffs resulted in 68% (ADA and IEC) and 88% false-negative rates (VA).