Inter-rater variability in 2D kinematic cycling analysis using Kinovea®: a cross-sectional study with 53 bike fitting professionals in Brazil

Abstract

Cycling kinematic analysis plays a central role in the bike fitting process, directly influencing decisions related to performance, comfort, and injury prevention. Consistency of measurements across evaluators is therefore essential for ensuring reliable outcomes in both clinical and performance contexts. This cross-sectional inter-observer agreement study evaluated inter-examiner variability in two-dimensional (2D) kinematic measurements obtained with Kinovea® software. A sample of 53 professional bike fitters from different regions of Brazil analyzed the same 40-second video of a cyclist pedaling on a stationary mountain bike. Each participant independently selected frames and measured seven predefined joint and positional angles. Statistical analyses included descriptive measures, Shapiro–Wilk normality testing, bootstrap confidence intervals, one-sided chi-square variance tests with Holm corrections, bias and empirical limits of agreement, Brown–Forsythe tests of dispersion, and Fleiss’ κ for categorical KOPS classification. The results showed notable inter-examiner variability, particularly for knee extension (CV = 6.2%), trunk flexion (CV = 5.8%), and plantar flexion (CV = 4.8%), which exceeded predefined tolerance thresholds of 2–4% of the mean. By contrast, hip flexion, knee flexion, and armpit angle showed greater consistency. Subgroup analyses revealed no significant effect of professional experience or software used on measurement variability. These findings highlight that, even under identical testing conditions, methodological differences among raters can substantially influence kinematic measurements. The study underscores the need for standardized protocols and structured training in 2D motion analysis to improve reliability in bike fitting practice and ensure safer, more effective adjustments for cyclists.