This study aimed to assess the combination of video-based kinematic variables adjusted by intrinsic covariates to predict the relative eccentric force (RelF) during the Nordic curl. The participants (n = 21) performed Nordic curls (3 trials; 3-min rest) on a device measuring the eccentric force. The peaks were normalized by body weight. Kinovea software was used to track angular and linear velocity and acceleration from recorded videos. Two prediction models with multiple linear regression equations associated kinematic, anthropometric, and age variables to adjust the actual RelF. The equations obtained the predicted RelF. The actual RelF was inversely correlated with height (r = -.52), tangential (r = -.50) and centripetal accelerations (r = -.715), and angular velocity (r = -.70). The best prediction models combined angular velocity with age (F2,18 = 15.1, P = .001, r = .792, r2 = .627) and with height (F2,18 = 14.5, P = .001, r = .785, r2 = .616). No differences were observed between actual and predicted values (P = .993-.994), with good levels of agreement and consistency (intraclass correlation coefficient = .77-.78; Cronbach α = .86-.87). Bland-Altman results showed high levels of agreement and low biases. The standard error of measurement and minimal detectable change ranges were 0.46 to 0.49 N/kg and 1.28 to 1.36 N/kg, respectively. Also, the percentage of standard error of measurement was below 10% (7.92%-8.35%). The coefficient of variation analysis returned a 14.54% and 15.13% for each model, respectively. Kinematic analysis offers portability and low cost to current expensive or technical impaired dynamometry-based techniques to assess the RelF.
Keywords: dynamometry; exercise; strain injury; technology assessment.