LAUSR

The accuracy of type 2 diabetes diagnosis is important for optimal clinical care. Expert bodies, including the American Diabetes Association and the World Health Organization, recommend basing the diagnosis on 1 of 3 measures of glycemia: hemoglobin A1c (HbA1c) level, fasting glucose level, or 2-hour plasma glucose level after an oral glucose tolerance test (13). Recommendations advocate repeated testing within a short period of time to confirm the diagnosis (13), with the goal of reducing the probability of a false-positive diagnosis. The American Diabetes Association has not expressed a preference for the choice of confirmatory test; however, it recommends that if results of 2 different tests of glycemia are discordant for the diagnosis, the test that yielded the positive result should be repeated soon after on a subsequent visit (4). Using a repeated test on a new sample from a subsequent visit to confirm diabetes can be logistically cumbersome, inconvenient, and expensive and can delay patient care. In clinical settings, it is common for 2 different tests (for example, fasting glucose and HbA1c measurement) to be performed simultaneously on the same blood sample. Thus, whether 2 tests from the same blood sample can be used for both screening and confirmation is an intriguing question with practical relevance. In this issue, Selvin and colleagues report the prognostic capability of single-sample definitions of confirmed undiagnosed diabetes (elevated levels of both fasting glucose and HbA1c) and unconfirmed undiagnosed diabetes (elevated fasting glucose level or elevated HbA1c level) to identify future risk for diagnosed diabetes and its complications (5). They used data from the ARIC (Atherosclerosis Risk in Communities) study, a prospective epidemiologic study of 15792 white and black Americans aged 45 to 64 years who had a baseline examination in 1987 to 1989, 3 subsequent follow-up visits every 3 years, and a fifth visit in 2011 to 2013. This analysis used visit 2 (1990 to 1992) as baseline, when levels of fasting glucose and HbA1c were measured concurrently. Of the 13346 participants included, 978 had at least 1 elevated level (HbA1c level 6.5% or fasting glucose level 7.0 mmol/L [126 mg/dL]) at baseline; among them, 383 (39.2%) had both and 595 (60.8%) had 1 elevated level. Selvin and colleagues found that elevated levels of fasting glucose and HbA1c from a single baseline blood sample had moderate sensitivity (54.9%) but high specificity (98.1%) to identify patients who had diabetes diagnosed during 5 years of follow-up (5), with specificity increasing to 99.6% at 15 years. The 15-year positive predictive value for the confirmed diabetes definition was 88.7% compared with only 71.7% for the unconfirmed definition. Of note, the 5- and 15-year adjusted cumulative incidences were significantly higher among persons with elevated levels of both fasting glucose and HbA1c compared with those with only 1 elevated level. Further, elevated levels of both fasting glucose and HbA1c in a single sample were more strongly associated with risk for cardiovascular disease and peripheral artery disease (but not kidney disease or all-cause mortality) than only 1 elevated level. Although the authors' use of data from the large, prospective ARIC study to attempt to answer an important practical question was creative, a few limitations warrant mention. The baseline data in their analysis were obtained after the cohort had been followed for 3 to 5 years. Glucose measurements were done at 5-year intervals in the ARIC study. Ideally, an effort to determine whether the diagnostic accuracy of a single-sample 2-test approach approximates that of a repeated test at a subsequent visit would use a confirmatory test within a much shorter period to replicate clinical practice. The study cohort included black and white Americans aged 45 to 64 years, and the generalizability to other groups with different prevalence of diagnosed and undiagnosed diabetes is not clear. Most important, the concordance between elevated HbA1c and fasting glucose levels from the single-sample determination was approximately 40% in this study but approximately 70% for repeated measurements of fasting glucose based on conventional criteria (6). Therefore, the criteria proposed by Selvin and colleagues may miss a large number of persons with diabetes. This tradeoff between sensitivity and specificity is important because the evidence now favors early intervention for persons with undiagnosed diabetes and even prediabetes (7, 8). Diabetes is a leading contributor to increasing health care costs in the United States. It cost the country $327 billion in 2017, a 26% increase from 2012 (9). Further, diabetes is the leading contributor to increasing health care costs among Medicare recipients (10). In 2014, the U.S. Preventive Services Task Force recommended screening for impaired fasting glucose, impaired glucose tolerance, and undiagnosed diabetes so that patients with elevated blood glucose levels could be referred for management starting with intensive behavioral counseling to promote a healthful diet and physical activity (7). As data from Selvin and colleagues' study underscore, hyperglycemia exists on a continuum, and even a single elevation in HbA1c or fasting glucose level is associated with increased risk for progression to overt diabetes and its complications. Early detection and diagnosis of diabetes are critical for appropriate initiation of interventions to prevent or delay complications. Simplifying methods to identify and confirm diabetes will facilitate this important clinical and public health priority. Selvin and colleagues offer an innovative and practical use of a single blood sample at 1 visit, rather than 2 samples obtained at 2 visits, to confirm the diagnosis of diabetes. This approach has appeal, especially for resource-challenged settings, but it needs replication in other populations before becoming accepted clinical practice.