Reliability and validity of velocity measures and regression methods to predict maximal strength ability in the back-squat using a novel linear position transducer

Purpose: to examine the reliability of load-velocity profiles (LVPs) and validity of 1-repetition maximum (1-RM) prediction methods in the back-squat using the novel Vitruve linear position transducer (LPT). Methods: twenty-five men completed a back-squat 1-RM assessment followed by 2 LVP trials using 5 incremental loads (20%-40%-60%-80%-90% 1-RM). Mean propulsive velocity (MPV), mean velocity (MV), and peak velocity (PV) were measured via a (LPT). Linear and polynomial regression models were applied to the data. The reliability and validity criteria were defined a-priori as intraclass correlation coefficient (ICC) or Pearson correlation coefficient (r) > 0.70, coefficient of variation (CV) ≤ 10%, and effect size (ES) < 0.60. Bland-Altman analysis and heteroscedasticity of errors (r2) were also assessed. Results: the main findings indicated MPV, MV and PV were reliable across 20- 13 90% 1-RM (CV < 8.8%). The secondary findings inferred all prediction models had acceptable reliability (CV < 8.0%). While the MPV linear and MV linear models demonstrated the best estimation of 1-RM (CV < 5.9%), all prediction models displayed unacceptable validity and a tendency to overestimate or underestimate 1-RM. Mean systematic bias (-7.29 to 2.83 kg) was detected for all prediction models, along with little to no heteroscedasticity of errors for linear (r2 < 0.04) and polynomial models (r2 < 0.08). Furthermore, all 1-RM estimations were significantly different from each other (p < 0.03). Conclusions: MPV, MV, and PV can provide reliable LVPs and repeatable 1-RM predictions. However, prediction methods may not be sensitive enough to replace direct assessment of 1-RM. Polynomial regression is not suitable for 1-RM prediction.