Background: Visceral adiposity index (VAI) is a novel marker of visceral adipose tissue accumulation and dysfunction. The study aim was to explore the association of VAI with the 10-year type 2 diabetes mellitus (T2DM) incidence in apparently healthy individuals and compare its T2DM predictive ability against common anthropometric indices. Methods: In 2001 to 2002, the ATTICA study (Greece) recruited a random sample of 1514 and 1528 CVD-free men (18-87 years old) and women (18-89 years old), respectively. Sociodemographic, lifestyle, clinical, and biochemical characteristics of participants were measured at baseline, and the 10-year follow-up was performed during 2011 to 2012. After excluding participants with diabetes at baseline and participants without complete follow-up information regarding diabetes status and/or baseline VAI values, the working sample consisted of 1049 participants. In this sample, the predictive value of baseline VAI value was studied in relation to 10-year diabetes incidence. Results: One hundred thirty-three incident cases of diabetes were documented (10-year incidence: 12.7%). In the fully adjusted model, VAI significantly increased diabetes risk by 22% (OR per 1-unit increase =1.22; 95%CI, 1.09-1.37). Markers of oxidative stress and inflammation were found to, at least partly, mediate this relationship. Also, a moderating effect of menstruation status was revealed among women. VAI showed the highest predictive ability and contributed the most, along with waist-to-height ratio, to the correct classification of participants who developed diabetes. Conclusions: The present findings suggest that VAI may be a useful index for predicting long-term diabetes development and may exhibit better predictive ability to commonly used anthropometric indices.