Associations and condition sensitivity of smartphone-and force-plate-derived sway metrics during VR-CTSIB

Abstract

Smartphone-based sensors may provide a complement to laboratory-based force-plate balance assessment; however, validation studies have not established whether smartphone–force-plate associations and agreement vary across combined manipulations of visual input and support-surface reliability, or which smartphone-derived metrics are most sensitive to sensory challenges. Therefore, this study compared smartphone-derived and force-plate-derived sway metrics during virtual reality-based clinical test of sensory interaction and balance (VR-CTSIB) conditions. This secondary analysis included 40 adults (20 with chronic ankle instability and 20 healthy controls). Participants completed 30 s double-leg stance trials under six VR-CTSIB conditions combining firm or foam surfaces with NormVision, DisplayOff, and SwayRef visual environments. Sway data were collected using a force plate and a smartphone-based head-mounted display. Root mean square (RMS) and mean absolute value (MAV) were calculated in the mediolateral (ML) and anterior–posterior (AP) directions. Pearson correlations with false discovery rate (FDR) correction were used to examine device associations, Bland–Altman analyses assessed agreement, and Cohen’s dz quantified condition sensitivity. In the primary analyses, ML smartphone metrics showed larger correlation coefficients with force-plate measures than AP metrics. The largest correlations were observed in the firm NormVision condition for ML RMS with force-plate ML velocity RMS (r = 0.705, FDR-adjusted p < 0.001) and ML MAV with force-plate ML velocity MAV (r = 0.679, FDR-adjusted p < 0.001). Force-plate outcomes demonstrated greater condition sensitivity than smartphone-derived outcomes. Among smartphone metrics, ML MAV and RMS in the SwayRef condition showed the largest effects (dz = 0.895 and 0.876, respectively). Bland–Altman analyses showed systematic bias, with agreement varying across metrics and conditions. Overall, smartphone-derived sway metrics may provide complementary information during VR-based balance assessment, particularly in the ML direction. However, associations, agreement, and condition sensitivity varied by metric and sensory condition, indicating that smartphone- and force-plate-derived measures should not be interpreted as interchangeable.