Describing The Percent Inactivation Versus Force Relationships And Predicting True Force With Polynomial Regression

Abstract

PURPOSE: The purpose of this study was to examine the percent voluntary inactivation versus submaximal force relationships with polynomial regression to predict each subjects' true maximal voluntary contraction (MVC) capabilities. METHODS: Nine participants (mean±SD age=21±3 yrs) completed nine randomly-ordered isometric plantar flexions from 10 to 90% of the MVC. Transcutaneous electrical stimuli were delivered to the tibial nerve using a high-voltage constant-current stimulator (Digitimer DS7AH, Herthfordshire, UK). Doublets were administered at a supramaximal stimulus intensity during the submaximal MVC plateau and then again 3-5 s after the submaximal MVC trial at rest. Percent inactivation was calculated for each submaximal MVC. The nine subjects selected for this study demonstrated < 5% voluntary inactivation at the 90% MVC, which suggested that these subjects were highly activated. Percent voluntary inactivation versus force relationships were analyzed on a subject-by-subject basis using polynomial regression (linear, quadratic, and cubic models). The predicted true MVC for each subject was extrapolated with their best fit polynomial model. A paired-samples t test was used to compare the actual MVC to the predicted true MVC. RESULTS: Polynomial regression indicated that 3 of 9 subjects (33%) were best fitted with a linear model, 2 of 9 subjects (22%) were best fitted with a quadratic model, and 4 of 9 subjects (44%) were best fitted with a cubic model. The t test indicated that the actual MVC (mean±SD 446±114 N) was greater than the predicted MVC (mean±SD 405±124 N) values (P = 0.003). Percent difference for each subject between actual MVC versus predicted MVC values ranged from 0.7% to 20.5%. CONCLUSIONS: Fitting the percent inactivation versus force relationships of highly activated subjects resulted in 33% linear and 66% nonlinear relationships. These results suggested that these patterns should be analyzed on a subject-by-subject basis. However, the predicted true MVC values from the best fit polynomial models were significantly less than the actual MVC values.