Enhancing the ability of gait analyses to differentiate between groups: scaling gait data to body size.

One goal of gait analysis is to distinguish clearly between a set of abnormal gait values measured from a patient referenced to a comparable population. However, the comparable population is often composed of individuals of various heights and weights, which increases inter-subject gait value variation which reduces the ability of a statistical test to identify a set of gait data outcomes with evaluative properties. Therefore, scaling gait data, based on subject leg length and mass, is commonly used to decrease the inter-subject variation but the efficacy of these methods is unknown. In this paper each of eight scaling strategies (none, ad hoc, dimensionless numbers, and five connected strategies based on similarity, dimensional analysis and muscle properties) were used to modify a set of gait data outcomes acquired from 10 individuals spanning a wide range of height (1.33-1.96 m) and mass (42.3-148.8 kg). These data were then examined to select that strategy and those scaling factors which maximally reduced inter-subject variation. The ad hoc, dimensionless numbers, and dynamic/mechanical/elastic (diameter of a limb (D) proportional to it's length (L) to the 1.5 power; time proportional to L(2) D(-1)) scaling strategies reduced the global inter-subject gait data outcome variation to 44% of its un-scaled value. Considering ten commonly reported gait data outcomes (temporal and spatial (stride time, stride length, progression velocity), kinematic (angles in the sagittal and frontal planes, angles in the transverse plane), external kinetic (ground reaction force and moment), and internal kinetic (joint force, moment, and power)) these three scaling strategies provided the largest number of minimum inter-subject variations (10, 10, and 9, respectively). Reduced inter-subject variation in gait data outcomes increases the ability of a statistical tool to detect a difference between a patient and a comparable group. With a statistically significant difference a clinician can then decide if this patient's gait pattern clinically deviates from that of the comparable group and an appropriate intervention warranted. The ad hoc, dimensionless numbers, and the dynamic/mechanical/elastic scaling strategies all reduce maximally the inter-subject variation in gait data outcomes.