BACKGROUND In karate, high percentages of fat mass (%FM) are often associated with poor sport performance. Dual-energy X-ray absorptiometry (DXA) can accurately measure the %FM (%FM-DXA), but it may not be practical in some sport settings, where anthropometry has long been used as an alternative. This study aimed to explore the ability of sixteen available anthropometric equations to estimate the %FM (%FM-AE) in karate athletes using the %FM-DXA as the criterion. Furthermore, two population-specific predictive equations to estimate the %FM-DXA were developed from anthropometric measurements. METHODS Forty-six male athletes aged 21.7±3.8 years underwent anthropometry and a whole-body DXA scan. RESULTS The results showed that in male karate athletes all the considered anthropometric equations are mostly inaccurate in estimating the %FM-DXA within a range of the limits of agreement in the Bland-Altman analysis, which ranged from 6.43% to 13.37%. Regression analysis yielded two statistically significant models (P<0.001 for both) to predict the %FM-DXA. In the first model (cross-validated estimation of R2=0.85), the predictors were the abdominal, triceps, calf, and biceps skinfolds and in the second model (cross-validated estimation of R2=0.77), the predictor was the sum of nine skinfolds (i.e., biceps, triceps, subscapular, chest, axilla, suprailiac, abdominal, anterior thigh and calf skinfolds). CONCLUSIONS These results underlined the need for sport-specific anthropometric equations to accurately estimate the %FM-DXA in male karate athletes. The two predictive anthropometric equations presented in this study provided a promising tool for professionals dealing with body composition in this athletic population to accurately estimate the %FM-DXA by means of anthropometry.