Determinants of gait dystonia severity in cerebral palsy

Abstract

Aim: To determine the movement features governing expert assessment of gait dystonia severity in individuals with cerebral palsy (CP).

Method: In this prospective cohort study, three movement disorder neurologists graded lower extremity dystonia severity in gait videos of individuals with CP using a 10-point Likert-like scale. Using conventional content analysis, we determined the features experts cited when grading dystonia severity. Then, using open-source pose estimation techniques, we determined gait variable analogs of these expert-cited features correlating with their assessments of dystonia severity.

Results: Experts assessed videos from 116 participants (46 with dystonia aged 15 years [SD 3] and 70 without dystonia aged 15 years [SD 2], both groups ranging 10-20 years old and 50% male). Variable limb adduction was most commonly cited by experts when identifying dystonia, comprising 60% of expert statements. Effect on gait (regularity, stability, trajectory, speed) and dystonia amplitude were common features experts used to determine dystonia severity, comprising 19% and 13% of statements respectively. Gait variables assessing adduction variability and amplitude (inter-ankle distance variance and foot adduction amplitude) were significantly correlated with expert assessment of dystonia severity (multiple linear regression, p < 0.001).

Interpretation: Adduction variability and amplitude are quantifiable gait features that correlate with expert-determined gait dystonia severity in individuals with CP. Consideration of these features could help optimize and standardize the clinical assessment of gait dystonia severity in individuals with CP.

FIGURE 1

Frequencies of expert-cited features justifying the presence (a), absence (b), or severity grading (c) of lower extremity dystonia in gait videos of individuals with cerebral palsy (CP), broken down by Global Dystonia Severity Rating Scale (GDS) score (d).

FIGURE 2

Quantitative analogs of expert-cited features of dystonia severity. Example video labels are shown under the leftmost column at the midpoint of the patella (squares), midpoint between the medial and lateral malleoli (triangles), and tip of the third toe (circles). The quantitative analogs assessed were the minimum values and variances across all frames in each video of: (first row) the aggregate knee-ankle-toe angles for the right and left foot; (second row) inter-knee distance in the horizontal plane/maximum knee-ankle distance in the same frame; (third row) inter-ankle distance in the horizontal plane/maximum knee-ankle distance in the same frame; and (fourth row) inter-toe distance in the horizontal plane/maximum knee-ankle distance in the same frame. Comparisons were made between videos of individuals with (average Global Dystonia Severity Rating Scale [GDS] = 1+) and without dystonia (average GDS = 0 to <1). For data with unequal SDs between groups, a t-test with Welch’s correction was used. For data that were not normally distributed, a Mann–Whitney U test was used (*these comparisons are significant at p < 0.05).

FIGURE 3

Correlation between Global Dystonia Severity Rating Scale leg subscores and quantitative analogs of expert-cited features of dystonia severity. Only quantitative analogs that were significantly different between videos of individuals with and without dystonia with regard to amplitude of adduction (a) or variability of adduction (b) (Figure 2) were analysed using stepwise multiple linear regression. Pearson correlation R² values and p-values are shown (*correlations are significant when p < 0.05). Lines are the mean with standard (bold black) or with standard error (dotted black).