Prediction of Exercise Tolerance in the Severe and Extreme Intensity Domains by a Critical Power Model

This study aimed to assess the predictive capability of different critical power (CP) models on cycling exercise tolerance in the severe- and extreme-intensity domains. Nineteen cyclists (age: 23.0 ± 2.7 y) performed several time-to-exhaustion tests (Tlim) to determine CP, finite work above CP (W'), and the highest constant work rate at which maximal oxygen consumption was attained (I(HIGH)). Hyperbolic power-time, linear power-inverse of time, and work-time models with three predictive trials were used to determine CP and W'. Modeling with two predictive trials of the CP work-time model was also used to determine CP and W'. Actual exercise tolerance of I(HIGH) and intensity 5% above I(HIGH) (I(HIGH+5%)) were compared to those predicted by all CP models. Actual I(HIGH) (155 ± 30 s) and I(HIGH+5%) (120 ± 26 s) performances were not different from those predicted by all models with three predictive trials. Modeling with two predictive trials overestimated Tlim at I(HIGH+5%) (129 ± 33 s; p = 0.04). Bland-Altman plots of I(HIGH+5%) presented significant heteroscedasticity by all CP predictions, but not for I(HIGH). Exercise tolerance in the severe and extreme domains can be predicted by CP derived from three predictive trials. However, this ability is impaired within the extreme domain.