Multivariate Characterization of Gait Variability Under Simulated Ship Roll Motion.

Walking on a ship requires continuous adaptation to a support surface that changes orientation with vessel motion. However, conventional group-level and joint-by-joint analyses may not characterize the coordinated structure of this adaptation or the heterogeneity of individual responses. This secondary analysis examined multivariate lower-limb joint-angle variability during simulated ship roll in 30 healthy adults completing nine 2-min walking trials across roll amplitudes of 0-20° and roll periods of 6 and 12 s. OpenSim-derived joint angles were summarized as 28 trial-level variability features. Principal component analysis, k-means clustering, hierarchical density-based spatial clustering of applications with noise (HDBSCAN), and generalized estimating equations were applied. Five principal components explained 78.1% of the total variance, with PC1 alone explaining 53.4% and representing a coordinated increase in lower limb joint angle variability, particularly in the sagittal-plane. k-means derived operational partition classified trials into low- and high-variability groups. The odds of high-variability group membership increased 1.48-fold per 1° increase in roll amplitude (p < 0.001), whereas roll period was not significantly associated with group membership (p = 0.704). HDBSCAN showed weak density separation (DBCV = 0.061) and substantial participant-level heterogeneity. These findings highlight the potential value of individualized roll-sensitivity assessment and provide a biomechanical basis for future studies of shipboard gait safety and fall-risk assessment.