Assessing True Variability And Mean Changes To Two Distinct Resistance Training Protocols

Abstract

Millions of dollars are spent analyzing inter-individual differences in response to resistance exercise, but the lack of a non-exercise control group makes it possible that these studies may simply be examining random error. Furthermore, it has been hypothesized that the magnitude of variability may differ depending upon the exercise protocol employed, but this yet to be appropriately tested. PURPOSE: To determine differences in two distinct resistance training protocols and whether true variability could be detected after accounting for random error. METHODS: Individuals (n=151) were randomly assigned to one of three groups: (1) a traditional exercise group performing four sets of elbow flexion exercise to failure; (2) a one-repetition maximum (1RM) performing a 1RM elbow flexion test; and (3) a time-matched non-exercise control group. Both exercise groups performed 18 sessions over six weeks. A Bayesian ANCOVA was used to test for mean changes across groups while adjusting for pre-values. To assess whether the variability in response to each exercise intervention differed from that of the control group, Bayesian Levene’s tests were computed. Bayes Factors (BF10) were used to quantify evidence for or against the null hypothesis. RESULTS: Both 1RM (2.3kg; BF10 = 4.791e+6) and traditional training groups (2.4kg; BF10 = 11,915) increased 1RM strength similarly (BF10 = 0.21), but only the 1RM group increased untrained arm 1RM strength (1.5kg; BF10 = 271). Only the traditional exercise group increased ultrasound measured muscle thickness (~0.23 cm across all sites; all BF10 ≥ 224). Across both training groups, the only differential responses were found in the change in 1RM strength of the trained arm in the traditional training group (BF10 = 5.381). This resulted in a true variability of 1.8 kg after the removal of random error. CONCLUSION: These findings demonstrate the importance of taking into consideration the magnitude of random error when determining response heterogeneity, as many studies may be classifying individuals based on random error. Additionally, our mean results demonstrate that strength is largely driven by task specificity, and the cross-over effect of strength may be load dependent.