Simulations found within-subject measurement variation in glycaemic measures may cause overdiagnosis of prediabetes and diabetes

Abstract

ObjectiveTo determine the impact of test measurement variation on misclassification of prediabetes and diabetes in the US adult population. Study Design and SettingData from adults with no prior diagnosis of diabetes in the 2015 to 2016 National Health and Nutrition Examination Survey (NHANES) were used to simulate populations of US adults eligible for screening. Estimates of measurement variation were applied to each simulated individual's true values to generate observed values for up to five repeated screens. Results: UnderdiagnosisFor 100,000 people assessed according to ADA or USPSTF guidelines, no people with true values in the diabetes range would be underdiagnosed as normal after one screen, and between 64 and 138 people would be misclassified with prediabetes after five re-screens (depending on guideline and test used). OverdiagnosisFor 100,000 people assessed according to the guidelines, between 1,602 and 2,233 people with true values in the normal range would be over diagnosed with prediabetes after 3 re-screens. A further 627 to 1,672 people with true values in the prediabetes range would be over diagnosed with diabetes after 5 re-screens. ConclusionMeasurement variation may cause overdiagnosis of prediabetes and diabetes, as well as apparent “progression” or “regression” of either condition.