Abstract
Sport coaches increasingly rely on external load metrics for designing effective training programs. However, their accuracy in estimating internal load is inconsistent, and their ability to predict autonomic nervous system (ANS) deterioration is unknown. This study aimed to evaluate the relationships between internal and external training load metrics and ANS recovery and function in college football players. Football athletes were recruited from a D1 college in the southeastern US and prospectively followed for 27 weeks. Internal load was estimated via exercise cardiac load (ECL; average training heartrate (HR) × session duration) and measured with an armband monitor equipped with electrocardiographic capabilities (Warfighter MonitorTM (WFM), Tiger Tech Solutions, Miami, FL, USA). External load was estimated via the summation and rate of acceleration and decelerations as measured by a triaxial accelerometer using the WFM and an accelerometer-based (ACCEL) device (Catapult Player Load, Catapult Sports, Melbourne, Australia) worn on the mid-upper back. Baseline HR, HR variability (HRV) and HR recovery served as the indicators for ANS recovery and function, respectively. For HRV, two, time-domain metrics were measured: the standard deviation of the NN interval (SDNN) and root mean square of the standard deviation of the NN interval (rMSSD). Linear regression models evaluated the associations between ECL, ACCEL, and the indicators of ANS recovery and function acutely (24 h) and cumulatively (one- and two-week). Athletes (n = 71) were male and, on average, 21.3 ± 1.4 years of age. Acute ECL elicited stronger associations for 24 h baseline HR (R2 0.19 vs. 0.03), HR recovery (R2 0.38 vs. 0.07), SDNN (R2 0.19 vs. 0.02) and rMSSD (R2 0.19 vs. 0.02) compared to ACCEL. Similar results were found for one-week: 24 h baseline HR (R2 0.48 vs. 0.05), HR recovery (R2 0.55 vs. 0.05), SDNN (R2 0.47 vs. 0.05) and rMSSD (R2 0.47 vs. 0.05) and two-week cumulative exposures: 24 h baseline HR (R2 0.52 vs. 0.003), HR recovery (R2 0.57 vs. 0.05), SDNN (R2 0.52 vs. 0.003) and rMSSD (R2 0.52 vs. 0.002). Lastly, the ACCEL devices weakly correlated with ECL (rho = 0.47 and 0.43, p < 0.005). Our findings demonstrate that ACCEL poorly predicted ANS deterioration and underestimated internal training load. ACCEL devices may “miss” the finite window for preventing ANS deterioration by potentially misestimating training loads acutely and cumulatively.
