Author(s): MAHY, T., BENOIT, N., WARNIER, G., PRÉMONT, C., FRANCAUX, M., Institution: UCLOUVAIN, Country: BELGIUM, Abstract-ID: 1722

INTRODUCTION:
Laboratory-based cycling race simulations are commonly used to study fatigue, recovery, and durability. However, those protocols often fail to reflect the real demands of road racing due to inadequate intensity anchoring and insufficient representation of power variability and intensity domains (Peeters, 2025). This study aimed to develop an ecologically valid fatigue protocol derived directly from power data acquired during national-level races. The goal was to reproduce the distribution of intensities, workload, and variability of real competition in highly trained cyclists.
METHODS:
Power data were collected from five male cyclists (21.0 ± 1.7 yrs; 67.2 ± 4.0 kg; VO₂max: 77.5 ± 9.5 ml·kg⁻¹·min⁻¹; LT1: 275 ± 21 w; LT2 (Lactate Minimum Test): 328 ± 6 w), classified as Tier 3 (McKay, 2022), during four stages of a national Belgian road race (2.12 IC1) (20 files; 19 retained). All riders completed an incremental test to determine VO₂max, LT1, and LT2. 30-s moving average power data (1 Hz) acquired during the races were segmented into five intensity zones based on lactate thresholds following current intensity prescription consensus (Bishop, 2024). High-intensity surges were identified from raw power and integrated into the protocol by substituting equivalent work within corresponding zones. An R script automated segmentation and protocol generation. Intensity was expressed relative to zone anchors (Z1: %LT1; Z2: LT1 + % Δ LT1–LT2; Z3: %LT2; surges: w·kg⁻¹), and volume as work per body mass (J·kg⁻¹). Protocol outputs were compared with race data using percentage differences and bilateral t-tests.
RESULTS:
The 200-segment exercise protocol created matched the total relative work observed during the races (42.5 ± 4.1 vs 42.9 ± 3.5 kJ·kg⁻¹, p = 0.51) but elicited lower normalized power (276 ± 12 vs 289 ± 15 w, p < 0.001), intensity factor (0.84 ± 0.04 vs 0.88 ± 0.06, p < 0.001) and training stress score (251 ± 33 vs 270 ± 41, p < 0.001). This was associated with fewer but longer surges (11 ± 5 vs 14 ± 7, p < 0.001; mean duration: 11 ± 2 vs 8 ± 1, p < 0.001), and a redistribution of work toward low intensity (Z1: 21.2 ± 3.2 vs 13.1 ± 2.4 kJ·kg⁻¹, p < 0.001), with reduced work above LT2 (Z3: 13.8 ± 4.3 vs 22.8 ± 4.5 kJ·kg⁻¹, p < 0.001), while work in Z2 remained similar (7.4 ± 5.9 vs 7.0 ± 3.1 kJ·kg⁻¹, p = 0.37). Despite these differences, power output variability was unchanged (variability index: 1.24 ± 0.05 vs 1.25 ± 0.05, p = 0.23).
CONCLUSION:
We present a race-derived, threshold-anchored fatigue protocol integrating intensity domains and power variability. Although temporal simplification and segment aggregation alter distributional features, especially work in the severe domain, the protocol preserves overall workload and variability of road races. Balancing ecological validity and feasibility, this approach is suited to investigate cyclist fatigue, recovery, and durability under laboratory conditions that better reflect real-world demands than model published previously.