ECSS Paris 2023: OP-AP12
INTRODUCTION: Estimation of physical activity intensity by translating accelerometry derived behaviors or counts to Metabolic Equivalence of Task (MET) values can be improved by accounting for the gradual changes, or slopes, in MET values that occur during transitions between different activities, rather than treating these transitions as abrupt, discrete shifts. This approach has the potential to better match the physiological response of different physical behaviors. This study aims to evaluate how adding exponential slopes to intensity classifications derived from thigh-worn accelerometer data influences the estimation of daily time spent in moderate and vigorous physical activities. The accuracy of these estimates will be validated using percentage heart rate reserve (HRR) as the criterion measure. METHODS: The analysis included data from 419 middle-aged adults as part of the Swedish CArdioPulmonary BioImage Study. Participants wore thigh-worn accelerometers and a portable electrocardiogram (ECG) for approximately 24h. Moderate-intensity (MPA) and vigorous-intensity physical activity (VPA) were estimated for each second by applying previously reported Metabolic Equivalents of Task (METs) values for various physical behaviors to thigh-worn accelerometry. Time spent in MPA and VPA were calculated using two acceleration processing methods: 1-second epochs and 90-second sloped METs. Time spent in MPA and VPA were also estimated using HRR and compared to each acceleration method. RESULTS: The HRR method estimated participants spent 27±36 minutes/day in MPA and 2.6±8.7 minutes/day in VPA. The 90-second sloped METs method overestimated participants spent 32±27 minutes/day in MPA, mean absolute error (MAE): 26±28 minutes/day; p<0.001) and 3.3±7.2 minutes/day in VPA (MAE: 4.3±9.8 minutes/day; p<0.001). The 1-second epoch METs method overestimated participants spent 45±25 minutes/day in MPA (MAE: 31±27 minutes/day; p<0.001) and 11±12 minutes/day in VPA (MAE: 10±12 minutes/day; p<0.001). Using Bland-Altman analyses, both acceleration processing methods demonstrated fixed and proportional biases compared to HRR (p<0.001). CONCLUSION: The findings highlight the impact of inclusion of exponential slopes into MET curves at behavior transitions on the accuracy of physical activity intensity estimates.
Read CV Johan RydgårdECSS Paris 2023: OP-AP12
INTRODUCTION: Detrended fluctuation analysis alpha-1 (DFAa1) has emerged as a promising non-invasive biomarker for exercise intensity assessment. However, the standard 2-min analysis window lacks temporal resolution necessary for real-time training applications. This study systematically investigated the validity of shortened DFAa1 windows (30s and 1min) versus the 2-min reference across different intensities. METHODS: Physically active males completed three continuous cycling protocols: low-intensity training at the first lactate threshold (LOW, n=19), moderate-intensity training at the second lactate threshold (MOD, n=19), and a 30-min self-paced time trial (TT30, n=18). DFAa1 was calculated using 30-s, 1-min, and 2-min moving windows, advancing in 1s increments. Validity was assessed using intraclass correlation coefficients (ICC), Bland-Altman analysis, and standard error of measurement (SEM). RESULTS: During LOW, both shortened windows showed poor agreement with the 2-min reference (30s: ICC=0.02, mean bias of –0.05; 1min: ICC=0.37, –0.02). During MOD, the 30-s window remained unreliable (ICC=0.32, –0.01), while the 1-min window achieved moderate reliability (ICC=0.63, 0.00). During TT30, both shortened windows substantially improved performance (30s: ICC=0.78, –0.02; 1min: ICC=0.95, –0.01), with the 1-min window achieving excellent reliability. CONCLUSION: DFAa1 analysis window validity is intensity-dependent, with shortened windows showing progressively improved agreement as exercise intensity and heart rate increases. While the 2-min window remains essential for low-intensity monitoring, 1-min or 30-s windows provide appropriate validity during high-intensity exercise, enabling more-responsive real-time feedback. These results support adaptive windowing strategies that dynamically adjust window length based on exercise intensity and the number of included data points, to optimize the analytical validity-temporal responsiveness trade-off.
Read CV Anton OlieslagersECSS Paris 2023: OP-AP12
INTRODUCTION: Heart rate variability (HRV) is a simple, non-invasive marker reflecting the influence of the autonomic nervous system on the cardiovascular system. Recent studies suggest its usefulness for monitoring training load in endurance athletes, where HRV trends may indicate physiological adaptation or maladaptation, such as overtraining. However, multiple factors beyond training volume, such as competitive season, stress and sleep, are integral part of a student-athlete’s life. The objective of this study was to investigate whether these factors modulate HRV changes measured by a Garmin wearable device (GWD) over a full year in varsity athletes. METHODS: This study included 18 varsity athletes (10 males, 8 females, age: 21.94 ± 2.3 years, weight: 70.09 ± 16.1 Kg, VO2max: 58.66 ± 9.7 mlO2.min-1.kg-1, body fat percentage: 15 ± 8%). Each athlete wore a GWD over a 12-month period to monitor nocturnal HRV. Subjective feeling was collected weekly with the Short Recovery and Stress Scale (SRSS) (1). ANOVA and linear mixed effect model were performed to measure interaction between HRV and GWD stress score, sleep score, sleep duration and competitive season. Spearman correlations were conducted to measure association between SRSS and GWD stress score. RESULTS: Individual HRV values were categorized into three groups relative to baseline: above (76.97 ± 28.97ms ), normal (73.63 ± 23.59ms) and below (62.68 ± 25.66ms). The effect of month on HRV was significant but small (η2 = 0.012, p < 0.05). November, December and March differed significantly from the preceding month (p < 0.05). Approximately 13% of the HRV variation was explained by the model (r2 marginal = 0.13). However, when the GWD stress score was excluded, the model explained only 6% of the variation (r2 conditional = 0, r2 marginal = 0.063). GWD stress score monitoring showed weak, non-significant, and scattered associations with self-reported recovery and stress (ρ range 0.00057-0.5; p range 0.009-0.99). CONCLUSION: These findings suggest that GWD HRV can capture context-dependent fluctuations across an academic year in varsity athletes. Although monthly HRV variations were statistically significant, their small magnitude implies that changes may relate to academic context, though multiple factors likely influence these results. In contrast, GWD stress score showed limited agreement with subjective stress and recovery, highlighting that they reflect distinct dimensions of athlete load. Collectively, theses findings suggest that GWD HRV monitoring alone has limited utility for interpreting physiological adaptation or stress in this population and should be used with caution, particularly without complementary contextual and subjective information. [1]Kellmann, M., & Kölling, S. (2019). Recovery and stress in sport: A manual for testing and assessment. Routledge, Taylor & Francis Group.
Read CV Philippe GendronECSS Paris 2023: OP-AP12