Prediction of body fat in male athletes from ultrasound and anthropometric measurements versus DXA. Abstract Background: To compare the accuracy of body fat percentage (BF%) measured by an ultrasound portable device and anthropometric measurements with a Dual Energy X-ray Absorptiometry (DXA) as the reference technique in male athletes. Methods: A total of 100 athletes: 16 boxers, 4 rowers, 5 gymnasts, 6 base ball players, 19 judo players, 10 taekwondo players, 7 basket-ball players, 21 wrestlers, 6 cyclists on track and 6 karate expert aged from 18 to 30 years participated. All athletes were selected from the French National Institute of Sports and Physical Education . Ultrasound measurements were made with a sonographic US BOX at the mid-thigh level. We developed a multi -linear model of body fat estimation from ultrasound and anthropometric dimensions (height, weight, waist circumference) using the DXA reference method. A cross-validation study was then performed with this linear regression on 62 males athletes proportionally stratified across the sports. Results: The best accuracy of BF was obtained using a multi-linear model from ultrasonic and anthropometric measurements with a concordance correlation ρ c = 0.941. This model was then used to estimate BF on the 62 males athletes. The concordance correlation ρ c = 0.931 and SEE =1.60. The 95% limits of agreement for individual BF% were [-4.1;3.6%] with symmetrically distributed deviations.Conclusions: Comparing to DXA, ultrasonic and anthropometric measurements are both accurate techniques to estimates BF%. Our results suggest that this regression model is practical to apply to different sports.