Association of A1C and fasting plasma glucose levels with diabetic retinopathy prevalence in the US population implications for diabetes diagnostic thresholds

Abstract OBJECTIVE To examine the association of A1C levels and fasting plasma glucose (FPG) with diabetic retinopathy in the U.S. population and to compare the ability of the two glycemic measures to discriminate between people with and without retinopathy. RESEARCH DESIGN AND METHODS This study included 1,066 individuals aged ≥40 years from the 2005–2006 National Health and Nutrition Examination Survey. A1C, FPG, and 45° color digital retinal images were assessed. Retinopathy was defined as a level ≥14 on the Early Treatment Diabetic Retinopathy Study severity scale. We used joinpoint regression to identify linear inflections of prevalence of retinopathy in the association between A1C and FPG. RESULTS The overall prevalence of retinopathy was 11%, which is appreciably lower than the prevalence in people with diagnosed diabetes (36%). There was a sharp increase in retinopathy prevalence in those with A1C ≥5.5% or FPG ≥5.8 mmol/l. After excluding 144 people using hypoglycemic medication, the change points for the greatest increase in retinopathy prevalence were A1C 5.5% and FPG 7.0 mmol/l. The coefficients of variation were 15.6 for A1C and 28.8 for FPG. Based on the areas under the receiver operating characteristic curves, A1C was a stronger discriminator of retinopathy (0.71 [95% CI 0.66–0.76]) than FPG (0.65 [0.60 – 0.70], P for difference = 0.009). CONCLUSIONS The steepest increase in retinopathy prevalence occurs among individuals with A1C ≥5.5% and FPG ≥5.8 mmol/l. A1C discriminates prevalence of retinopathy better than FPG. Tests of glycemia and their thresholds for diabetes diagnosis is an area of long-standing debate. The presence of diabetic retinopathy is arguably the best criterion from which to compare glycemic measures because it is a specific and early clinical complication usually related to diabetes, and it represents a specific and relevant clinical end point for judging an alternative test (1). For these reasons, diabetic retinopathy has served as the basis for diagnostic criteria of type 2 diabetes (2–4) and provides the rationale for the American Diabetes Association's recommendation of a threshold of a fasting plasma glucose (FPG) of 7.0 mmol/l to define the presence of diabetes (4,5). However, an analysis of three recent population-based cross-sectional studies suggested that there may be considerable variation across populations and that the association of FPG with retinopathy prevalence may be more of a continuous relationship than previously thought (5). A1C levels are being considered as an alternative diagnostic tool for diabetes diagnosis (6). Unlike FPG, A1C does not require an overnight fast, is not affected by short-term lifestyle changes, and has less variability within individuals than FPG (7–9). Nevertheless, few studies have examined the prevalence of retinopathy across the spectrum of A1C levels, which could assist in the designation of ideal A1C diagnostic cut points (2,3). The newly released National Health and Nutrition Examination Survey (NHANES) 2005–2006 incorporated a multiple-field retinal photograph examination, presenting an opportunity to reassess the selection of glucose and A1C cut points for diabetes diagnosis. Our objectives were to examine the relation between levels of A1C and FPG and prevalence of retinopathy in the U.S. population and to compare the ability of both measures to differentiate people with and without retinopathy