Abstract
For decades, indirect submaximal tests using heart rate (HR) to estimate maximal oxygen consumption (VO2max) have been used for assessing cardiorespiratory fitness without pushing individuals to their limits. However, the optimal combination of submaximal workloads to use for maximizing estimation performance remains unclear. The study reprocessed data from 18 adolescent athletes undergoing a cycle ergometer incremental test with step-wise increments of 15 Watt/min until volitional exhaustion, sourced from a publicly available dataset. Multiple HR-derived metrics were computed over six distinct combinations of increasing workloads (50, 65, 80, 95, 110, 125 Watt). Principal component analysis was employed for dimensionality reduction. The top-performing regression model was chosen after training and validating various regression models, including machine learning-based ones. The HR data recorded at a single workload of 50 Watt was already adequate for estimating group VO2max, exhibiting similar scores (p = 0.80) to actual group values. Utilizing three consecutive workloads (50, 65, and 80 Watt) provided the most accurate individual VO2max prediction, revealing the highest correlation coefficient (0.71) along with the smallest bias (0.019 L/O2) and standard deviation (0.39 L/O2) across all six combinations. The project identifies optimal workloads for constructing new submaximal VO2max estimation tests. Additionally, it introduces new models for estimating VO2max for adolescents, each with varying performance based on the number of workloads utilized.