The purpose of this work was to investigate the association between coffee drinking and diabetes development and potential mediation by oxidative stress and inflammatory biomarkers.In 2001-2002, a random sample of 1514 men (18-87 years old) and 1528 women (18-89 years old) were selected to participate in the ATTICA study (Athens metropolitan area, Greece). A validated food-frequency questionnaire was used to assess coffee drinking (abstention, casual, habitual) and other lifestyle and dietary factors. Evaluation of oxidative stress and inflammatory markers was also performed. During 2011-2012, the 10-year follow-up of the ATTICA study was carried out. The outcome of interest in this work was incidence of type 2 diabetes, defined according to American Diabetes Association criteria.During follow-up, 191 incident cases of diabetes were documented (incidence 13.4% in men and 12.4% in women). After various adjustments, individuals who consumed ⩾250 ml of coffee (≈1.5cup) had 54% lower odds of developing diabetes (95% confidence interval: 0.24, 0.90), as compared with abstainers. A dose-response linear trend between coffee drinking and diabetes incidence was also observed (P for trend=0.017). When controlling for several oxidative stress and inflammatory biomarkers, the inverse association between habitual coffee drinking and diabetes was found to be mediated by serum amyloid-A levels.This work highlights the significance of long-term habitual coffee drinking against diabetes onset. The anti-inflammatory effect of several coffee components may be responsible for this protection.