To evaluate the effect of physical activity levels on 10-year diabetes incidence and investigate the potential mechanism.In 2001-2002, a random sample of apparently healthy 3042 men and women (18-89 years) was selected to participate in the ATTICA study. Several socio-demographic, clinical and lifestyle characteristics were recorded. Physical activity level was recorded through a translated, validated, version of International Physical Activity Questionnaire (IPAQ); MET min/week was calculated and quartiles constructed. Diabetes diagnosis was defined according to the ADA criteria. During 2011-2012, a 10-year follow-up was performed.n = 191 cases were recorded, yielding an incidence of 12.9%. In multivariable analysis, moderate physical activity level (331-1484 MET min/week) was found to decrease 10-year diabetes incidence by 53% compared to very low physical activity (< 150 MET min/week) (OR = 0.47; 95% CI 0.24, 0.93). For high physical activity level (> 1484 MET min/week), the results were not significant. The antidiabetic effect was found to be mediated by oxidized LDL and total antioxidant capacity.The current work revealed the significant beneficial role of moderate physical activity against diabetes development, potentially through attenuating oxidative stress.