Cluster analysis to classify gait alterations in rheumatoid arthritis using peak pressure curves.

OBJECTIVE: To detect gait alterations in rheumatoid arthritis (RA) patients using peak pressure curves (PPC) and normalized force curves (NFC) instead of clinical classification based on the health assessment questionnaire (HAQ).
METHODS: Three RA groups--30 patients each--were classified according to their HAQ score. Cluster analysis based on a k-means algorithm was applied to PPCs and NFCs in order to classify RA patients with respect to amplitude and shapes of such gait parameters.
RESULTS: The best gait pattern identification was obtained by clustering PPCs into three clusters. Peak pressures of the three identified patterns were 1169.5+/-99.4 kPa for cluster 1, 885.8+/-165.2 kPa for cluster 2 and 402.0+/-128.5 kPa for cluster 3 (statistically different, Student's t-test, p<0.001). 41 patients were included in cluster 3, 31 in cluster 2 and only 18 patients in cluster 1. Most RA3 patients--17 out of 30--showed low peak pressures and almost normal PPCs (cluster 3). Cluster 2, which incorporated altered PPCs, was mainly formed by RA1 and RA2 patients.
CONCLUSIONS: PPC appears as a reliable gait parameter for a shape-based clustering algorithm. The proposed cluster analysis was proved to be reliable and the delivered classifications stable. The distribution of RA patients among the three identified PPC clusters showed only a partial agreement between clinical and functional classification, thus revealing the development of specific gait strategies related to the pathology more than to its clinical level of severity. This finding may be clinically relevant and support effective treatment of RA gait related pathologies.