AIMS: As diabetic retinopathy (DR) can occur even in well-controlled patients with type 2 diabetes (T2D), our study sought to determine whether it might be related to 'glucose memory' by evaluating patients' HbA1c over previous years and their skin autofluorescence (SAF). METHODS: In 334 patients with T2D and HbA1c levels≤8%, their available values of HbA1c from previous years were collected, and their SAF measured by an advanced glycation end-product (AGE) reader. Binary logistic regression analysis was then used to correlate DR with previously recorded HbA1c levels and to SAF, with adjustment for DR risk factors [age, gender, BMI, duration of diabetes, arterial hypertension, diabetic kidney disease (DKD), blood lipid levels and statin treatment]. RESULTS: Our patients were mostly men (58.4%) aged 63±10years, with a duration of diabetes of 13±10years and HbA1c=7.1±0.7%. Of these patients, 84 (25.1%) had DR, which was associated with longer duration of diabetes and greater prevalence of DKD. A total of 605 HbA1c values from previous years were collected for time periods -4±3 months (n=255), -16±4months (n=152), -30±4months (n=93) and -62±26 months (n=105). After adjustment, the association between DR and having an HbA1c higher than the median was significant only for the oldest previous HbA1c values: OR=6.75, 95% CI: 1.90-23.90. Moreover, SAF values were higher in those with DR [2.95±0.67 arbitrary units (AU)] vs 2.65±0.65 AU with no DR (P<0.01) and were also associated with the oldest previous HbA1c values (P<0.01). CONCLUSION: Our study found that 25.1% of our well-controlled T2D patients had DR, which was related to both their HbA1c levels from 5years prior to study admission and their SAF values, a marker of glucose memory.