ECSS Paris 2023: OP-AP42
INTRODUCTION: Load–velocity (L–V) profiling is widely used to estimate the one-repetition maximum (1RM) on a session-by-session basis. However, it remains unclear whether a baseline profile can accurately predict strength changes following a resistance training intervention without recalibration. This study examined the longitudinal stability of L–V profiles in the free-weight back squat and whether training-induced changes in profile characteristics impair 1RM prediction. METHODS: Thirty-six strength-trained participants (15 females, 21 males) completed free-weight back squat 1RM testing before and after a six-week resistance training intervention. Barbell velocity was recorded using a linear position transducer. Individual L–V profiles were modelled using both multiple-point (all loads) and two-point models (30% and 70% 1RM) regression models. Terminal velocity was determined using individual and general values obtained during the 1RM and a repetition-to-failure test. To assess longitudinal validity, two prediction models were constructed: (A) pre-intervention slope and intercept combined with post-intervention terminal velocity, and (B) post-intervention slope and intercept combined with pre-intervention terminal velocity, both predicting post-training 1RM. Agreement was evaluated using Bland–Altman analyses, and mixed-effects models were used to examine changes in slope, intercept, and terminal velocity. RESULTS: When baseline L–V profiles were applied to post-intervention terminal velocities (Model A), 1RM was systematically underestimated (–3.4 to –7.9 kg), with wide limits of agreement (LoA span: 32.9–56.2 kg). This underestimation reflects a pivoting of the L–V profile following the intervention: slopes flattened in two-point models (p = 0.009), whereas intercepts increased in multiple-point models (p = 0.011). Conversely, combining post-intervention profiles with baseline terminal velocities (Model B) reduced prediction bias (–2.7 to +1.7 kg), but LoA remained larger than ±5 kg for all models (span: 19.3–42.4 kg). Two-point models consistently showed larger random error and lower concordance than multiple-point models. Mixed-effects models revealed a significant reduction in terminal velocity over time (p = 0.041), indicating a lower failure threshold post-intervention. These results indicate that changes in terminal velocity contributed to the prediction error, but to a lesser extent than longitudinal changes in slope and intercept. CONCLUSION: L–V parameters are not stable enough to predict 1RM changes over a training block. Resistance training notably alters the L-V profile configuration, leading to unacceptable prediction accuracy. Consequently, L–V profiling should not replace direct testing for tracking progress; regular recalibration is essential to maintain accuracy.
Read CV Erik HobeinECSS Paris 2023: OP-AP42