Stage-to-Stage Changes in Heart Rate–Power Output Coupling During Cycling Stage Races: Associations with Race-Derived Performance and Stage Context

This study examines whether model-derived heart rate–power output (HR–PO) parameters changed across consecutive stages of elite under-23 stage races and whether their longitudinal trajectories were associated with race-derived performance characteristics and stage context. Race files from the 2025 Tour de l'Avenir and Giro Next Gen were analysed, comprising 127 valid stage observations from 23 cyclist–race cases and 19 unique cyclists. For each stage, τdel, τrec, and τmax were estimated from an HR–PO model, and longitudinal changes were analysed using linear mixed-effects models. The estimated decrease in τdel across race progression was not statistically significant after FDR correction across the three parameters (−1.33 s, 95% CI −2.45 to −0.20, unadjusted p = 0.021; q = 0.063), whereas τrec (0.68 s, 95% CI −6.58 to 7.93, p = 0.855) and τmax (−0.96 s, 95% CI −2.10 to 0.19, p = 0.101) showed no systematic change. The τdel trajectory differed between Avenir and Giro (race-by-progression interaction, p = 0.027) and was attenuated after exclusion of the most mountainous stages (−0.85 s, 95% CI −2.56 to 0.85, p = 0.327). In exploratory secondary analyses, higher race-derived PO at 5 s and 5–30 min (p ≤ 0.006, q ≤ 0.043) and better 5 min durability (p < 0.001, q = 0.007) were associated with a more negative τrec trajectory; however, cyclist-cluster bootstrap confidence intervals excluded zero only for the 5 s and 5 min PO interactions, and no association was retained for τdel or τmax. In stage-context analyses, higher normalized power relative to critical power was associated with higher τrec (8.09 s per 10 percentage points; p = 0.001, q = 0.044), whereas greater low-intensity exposure was associated with lower τmax (−0.82 s per 10 percentage points relative to moderate intensity; p = 0.003, q = 0.045); no stage-context association was retained for τdel. Overall, the estimated decrease in τdel did not reach statistical significance after FDR correction and was sensitive to race context and the exclusion of the most mountainous stages, whereas τrec showed no systematic longitudinal change but varied in association with race-derived PO, durability, and relative stage intensity. All secondary associations are hypothesis-generating and require independent validation.