Background/Aims: Our aim was to evaluate whether visceral adiposity index (VAI) could predict the risk of type 2 diabetes (T2D) in different genders and to compare the predictive ability between VAI and other fatness indices. Methods: Four thousand seventy-eight participants including 1,817 men and 2,261 women, aged 18 and older and free of T2D at baseline were enrolled in 2010 and followed up for 4 years. New cases of T2D were identified via the annual medical examination. Cox regression analysis was used to assess the association between VAI and incidence of T2D. Receiver operating characteristic curve and area under the curves (AUC) were applied to compare the prediction ability of T2D between VAI and other fatness indices. Results: During the 4-year follow-up, 153 (8.42%) of 1,817 men and 88 (3.89%) of 2,261 women developed T2D. The multivariable-adjusted hazards ratios for developing T2D in the highest tertile of VAI scores were 2.854 (95% CI 1.815-4.487) in men and 3.551 (95% CI 1.586-7.955) in women. The AUC of VAI was not higher than that of other fatness indices. Conclusions: VAI could predict the risk of T2D among Chinese adults, especially in women. However, the prediction ability of T2D risk for VAI was not higher than that of the other fatness indices.