Estimating the delay between onset and diagnosis of type 2 diabetes from the time course of retinopathy prevalence.

Abstract

By correlating known diabetes duration with the prevalence of retinopathy, more than 10 years have been estimated to lapse between the onset and diagnosis of type 2 diabetes. Such calculations, however, assumed a linear model, included stages of retinopathy not specific to diabetes, and allowed 5 years for retinopathy to occur after the onset of diabetes. We calculated the duration of undiagnosed type 2 diabetes in outpatients screened for retinopathy in a hospital-based diabetes clinic after correcting these assumptions. Diabetic patients (n = 12,074; 35,545 fundus examinations) were stratified into younger onset (YO; age at onset <30 years) or older onset (OO; age at onset ≥30 years), insulin treated (IT) or not IT (NIT), and with mild/more severe diabetic retinopathy (AnyDR) or moderate/more severe diabetic retinopathy (ModDR). The best-fitting equation correlating known duration among the OO-NIT group with the prevalence of ModDR was used to extrapolate time from appearance of retinopathy to diagnosis of type 2 diabetes. Time for retinopathy to develop after diabetes was calculated from the equation correlating the duration among the YO-IT group with appearance of ModDR. There were 1,719 patients in the OO-NIT group with AnyDR and 685 with ModDR and 756 in the YO-IT group with AnyDR and 385 with ModDR. A linear model showed ModDR appeared 2.66 years before diagnosis among those in the OO-NIT group. A quadratic model suggested that ModDR appeared 3.29 years after diagnosis among those in the YO-IT group. The resulting estimate was 6.05 years (2.66 + 3.29) between the onset and diagnosis of diabetes, compared with 13.36 years using standard criteria. Using best-fitting models and stratifying by glucose-lowering treatment and severity of retinopathy substantially lowers the estimated duration of undiagnosed type 2 diabetes.