Self-reported training variables are poor predictors of laboratory measures in cyclists.

Purpose: Cycling is an activity that depends on a range of physiological attributes, as well as genetic, dietary, lifestyle and training factors. The aim of this study was to determine what self-reported training-related factors (e.g. intensity, frequency, supervision, etc) might predict laboratory-measured physiological and performance characteristics of a heterogeneous group of male and female self-classified cyclists. Methods: Forty-eight male and fourteen female cyclists completed all aspects of the study including a training questionnaire, incremental cycling test to determine maximal oxygen uptake (VO2max), 30-s Wingate test and a 4-km cycling time-trial. Principle component analysis and LASSO regression modelling were used to analyse laboratory-measures and training variables and the predictive capacity of the latter. Results: Total distance covered across all intensities was the only training variable included in most bootstrap models (63.8%), although the actual contribution was very low with a median f2 effect size equal to 0.01. Self-reported classification of cycling level was weakly correlated to guideline classification of relative VO2max in men (r=0.396, p=0.004), but not women (r=0.024, p=0.925). Conclusions: Self-reported training variables were poor predictors of laboratory-based physiological and performance variables in this heterogeneous group of cyclists. Total distance covered was the only training variable included in most regression models, but the predictive capability of outcomes was low. This suggests that most of these self-report variables are not useful pre-screening tools for categorising non-elite cyclists or raises the potential that non-elite cyclists cannot accurately quantify their own training intensities. Researchers and coaches should be wary that self-reported classification may not directly reflect the level of the cyclist.