Two hundred and fifty seven male Caucasian ultra-endurance athletes were recruited, pre-race, before different swimming, cycling, running and triathlon races. Fat mass and skeletal muscle mass were estimated using bioelectrical impedance analysis (BIA) and anthropometric methods in order to investigate whether the use of BIA or anthropometry would be useful under field conditions. Total body fat estimated using BIA was significantly high (P < 0.001) compared with anthropometry. When the results between BIA and anthropometry were compared, moderate to low levels of agreement were found. These results were in accordance with the differences found in the Bland-Altman analysis, indicating that the anthropometric equation of Ball et al. had the highest level of agreement (Bias = -3.0 ± 5.8 kg) with BIA, using Stewart et al. (Bias = -6.4 ± 6.3 kg), Faulkner (Bias = -4.7 ± 5.8 kg) and Wilmore-Siri (Bias = -4.8 ± 6.2 kg). The estimation of skeletal muscle mass using BIA was significantly (P < 0.001) above compared with anthropometry. The results of the ICC and Bland-Altman method showed that the anthropometric equation from Lee et al. (Bias = -5.4 ± 5.3 kg) produced the highest level of agreement. The combined method of Janssen et al. between anthropometry and BIA showed a lower level of agreement (Bias = -12.5 ± 5.7 kg). There was a statistically significant difference between the results derived from the equation of Lee et al. and Janssen et al. (P < 0.001). To summarise, the determination of body composition in ultra-endurance athletes using BIA reported significantly high values of fat and skeletal muscle mass when compared with anthropometric equations.