INTRODUCTION: Subjects with type 2 diabetes have an excess risk of cancer. The potential role of advanced glycation end products (AGEs) accumulated during long-term hyperglycemia in cancer development has been suggested by biological studies but clinical data are missing. AGEs can be estimated by measuring the skin autofluorescence. We searched whether the skin autofluorescence could predict new cancers in persons with type 2 diabetes. RESEARCH DESIGN AND METHODS: From 2009 to 2015, we measured the skin autofluorescence of 413 subjects hospitalized for uncontrolled or complicated type 2 diabetes, without any history of cancer. The participants were followed for at least 1 year and the occurrences of new cancers were compared according to their initial skin autofluorescences. RESULTS: The participants were mainly men (57.9%), with poorly controlled (HbA1c 72±14 mmol/mol or 8.7%±1.8%) and/or complicated type 2 diabetes. Their median skin autofluorescence was 2.6 (2.2-3.0) arbitrary units. Forty-five new cancer cases (10.9%) were registered during 4.8±2.3 years of follow-up: 75.6% of these subjects had skin autofluorescence higher than the median (χ(2): p=0.001). By Cox regression analysis adjusted for age, gender, body mass index, history of smoking and renal parameters, skin autofluorescence >2.6 predicted a 2.57-fold higher risk of cancer (95% CI 1.28 to 5.19, p=0.008). This association remained significant after excluding the eight cancers that occurred in the 4 years after inclusion (OR 2.95, 95% CI 1.36 to 6.38, p=0.006). As a continuous variable, skin autofluorescence was also related to new cancers (OR 1.05, 95% CI 1.01 to 1.10, p=0.045). CONCLUSIONS: Skin autofluorescence, a potential marker of glycemic memory, predicts the occurrence of cancer in subjects with type 2 diabetes. This relation provides a new clinical argument for the role of AGEs in cancer. Their estimation by measuring the skin autofluorescence may help select subjects with diabetes in cancer screening programs.