An Analysis of Variability in “CatWalk” Locomotor Measurements to Aid Experimental Design and Interpretation

Abstract

Preclinical studies in models of neurological injury and disease rely upon behavioral outcomes to measure intervention efficacy. For spinal cord injury, the CatWalk system provides unbiased quantitative assessment of subtle aspects of locomotor function in rodents and so can powerfully detect significant differences between experimental and control groups. Although clearly of key importance, summary group-level data can obscure the variability within and between individual subjects and therefore make it difficult to understand the magnitude of effect in individual animals and the proportion of a group that may show benefit. Here we calculate 'reference change intervals' that define boundaries of normal variability for measures of rat locomotion on the CatWalk. Our results indicate that many commonly-used outcome measures are highly variable, such that differences of up to 70% from baseline value must be considered normal variation. Many CatWalk outcome variables are also highly correlated and dependent upon run speed. Application of calculated reference change intervals to open access data (odc-sci.org) on hindlimb stride length in spinal cord-injured rats illustrates the complementarity between group-level (16mm change; P=0.0009) and individual-level (5/32 animals show change outside reference change interval boundaries) analysis between week 3 and week 6 after injury. We also conclude that interdependence amongst CatWalk variables implies that test 'batteries' require careful composition to ensure that different aspects of defective gait are analyzed. Calculation of reference change intervals aids in experimental design by quantifying variability and enriches overall data analysis by providing details of change at an individual level that complement group-level analysis. Significance statement Selection of robust candidate interventions for translation from experimental animals into the neurology clinic requires meticulous examination of behavioral effects observed in the laboratory. Although analysis of group-level data, the current mainstay, is critically 4 important, analysis of individual-level data provides a complementary viewpoint that, bearing in mind the immense variability in neurological deficits in people with spinal cord injury, has high relevance to the interpretation of studies on putative therapies. Here we describe the derivation of specific 'reference change intervals' and, using example data, show how these augment interpretation of overall effect and can aid in effective experimental design. The combination of group-level and individual-level analysis will provide more stringent analysis of intervention effects in neurological injury and disease research.

Gait Analysis Using Animal Models of Peripheral Nerve and Spinal Cord Injuries

“…[ [13][14][15][16] Stand duration Stand time of the paw; stand time/stance phase (sec). [17,18] Swing duration Duration of the step cycle phase; swing time/swing phase.…”

“…[17,18] Swing duration Duration of the step cycle phase; swing time/swing phase. [13][14][15][16][17][18] Step cycle duration Duration of the step cycle (stance and swing duration) (sec). [14] Duty cycle (%) Stand time divided by stand time + swing time.…”

“…The distance between the two forelimbs or between the two hind limbs. [13,14,[16][17][18] Stride length Length (mm) of a stride; distance between two consecutive placements of a paw.…”

“…The total floor area contacted by the paw during the stance phase; expressed in square pixel. [13,14,16,17] Minimum paw print intensity Minimum intensity of the paw. [17,19] Mean paw print intensity Mean intensity of the paw.…”

Combination of Defined CatWalk Gait Parameters for Predictive Locomotion Recovery in Experimental Spinal Cord Injury Rat Models

“…Since SCI affects several motor and sensory pathways, multiple gait parameters need to be analyzed simultaneously (Lakes and Allen, 2016). However, in the majority of published work, CatWalk parameters are evaluated individually (Aceves et al, 2020;Bhimani et al, 2017;Bieler et al, 2018;Fagoe et al, 2016;Garcia-Ovejero et al, 2014;Heinzel et al, 2020;Kappos et al, 2017;Slusarczyk et al, 2017). The large number of gait parameters as well as the variated gait effect of an injury often bring complications in the analysis of gait.…”