The new diagnostic threshold of hemoglobin A(1c) was made based on evidence from cross-sectional studies, and no longitudinal study supports its validity. To examine whether hemoglobin A(1c) of 6.5% or higher defines a threshold for elevated risk of incident retinopathy, we analyzed longitudinal data of 19,897 Japanese adults who underwent a health checkup in 2006 and were followed up 3 years later. We used logistic regression models and restricted cubic spline models to examine the relationship between baseline hemoglobin A(1c) levels and the prevalence and the 3-year incidence of retinopathy. The restricted cubic spline model indicated a possible threshold for the risk of incident retinopathy at hemoglobin A(1c) levels of 6.0-7.0%. Logistic regression analysis found that individuals with hemoglobin A(1c) levels of 6.5-6.9% were at significantly higher risk of developing retinopathy at 3 years compared with those with hemoglobin A(1c) levels of 5.0-5.4% (adjusted odds ratio, 2.35 [95% CI 1.08-5.11]). Those with hemoglobin A(1c) levels between 5.5 and 6.4% exhibited no evidence of elevated risks. We did not observe a threshold in the analysis of prevalent retinopathy. Our longitudinal results support the validity of the new hemoglobin A(1c) threshold of 6.5% or higher for diagnosing diabetes.