Validação e sensibilidade de ferramentas integradas em plataforma web para monitoração do estado fisiológico de atletas

Abstract

Introduction: Heart rate monitors connected to smartphone apps have emerged as an alternative to electrocardiograms (ECG) for recording heart rate variability (HRV). However, these devices face limitations, such as app dependency and the need for coaching staff to access athletes' individual devices. As an alternative, our research group is developing a web-based platform, including a module for recording HRV using a heart rate monitor, which requires validation of its accuracy. Tools with high administrative practicality are preferred for daily monitoring of training load responses. While it is widely accepted that monitoring should not rely on a single tool, it remains necessary to investigate which variables best reflect athletes' psychophysiological states. Objective: To verify the accuracy of HRV indices recorded by the web platform and heart rate monitor compared to ECG and test the sensitivity of sRPE, PRS, well-being, HRV, and CMJ variables over a mesocycle with intentionally variable training loads in track and field athletes. Methods: For the validation phase, 22 track and field athletes (11 men, mean age 17.1 ± 6.58 years) participated. HRV was recorded for five minutes using ECG and the web platform connected to a heart rate monitor simultaneously in the supine position. Time-domain (HR, MeanRR, SDNN, pNN50, rMSSD), frequency-domain (LF, HF), and nonlinear (SD1, SD2) indices were compared using t-tests, Pearson correlation, intraclass correlation coefficient (ICC), simple linear regression, and Bland-Altman analyses. To assess variable responses to training load fluctuations, 26 athletes (national and continental levels, mean age 18.6 ± 7.98 years; middle-distance runners [n=18; 10 men] and sprinters [n=8; 5 men]) completed four microcycles (six training sessions each): one ordinary, two overload (30% volume and 20% intensity increases), and one regenerative (50% load reduction). Measurements were taken daily before (PRS, well-being, HRV, and CMJ) and after (sRPE) training sessions, or twice a week (creatine kinase [CK]). Results: In the validation phase, time-domain p-values ranged from 0.94 to 0.99, with ICC and Pearson correlations above 0.99, EPE < 1.82%, and EC between 0.00% and 0.69%. In the frequency domain, p-values ranged from 0.89 to 0.96, ICC from 0.93 to 0.98, Pearson correlation from 0.89 to 0.96, and EPE was 26% (LF) and 17% (HF), with EC of -1.1% (LF) and 3.4% (HF). For Poincaré indices, SD1 and SD2 showed p-values between 0.94 and 0.96, ICC and Pearson correlations of 0.99, EPE < 1.80%, and EC < 0.67%. In the sensitivity phase, weekly mean analysis revealed that during the overload microcycle, sRPE and fatigue scale scores increased, while PRS significantly decreased. In the regenerative microcycle, sRPE and fatigue decreased, while PRS, rMSSD, and CMJ significantly increased, with CK remaining unchanged. Daily data analysis showed that HRV indices (rMSSD, pNN50, and SD1) and CK were sensitive to training load variations. Conclusion: The web platform and heart rate monitor provided valid HRV data in time, frequency, and nonlinear domains compared to ECG. sRPE, PRS, and the fatigue scale from the well-being questionnaire were sensitive to accumulated fatigue and recovery, while CMJ and rMSSD were sensitive to recovery alone.