Effects of physiotherapy treatment on knee osteoarthritis gait data using principal component analysis Nathaly Gaudreault a,b, ⁎, Neila Mezghani a,c , Katia Turcot a , Nicola Hagemeister a,c , Karine Boivin a,d , Jacques A. de Guise a,c a Laboratoire de recherche en imagerie et orthopédie (LIO), Centre de recherche du Centre hospitalier de l'Université de Montréal, Quebec City, Canada b École de réadaptation, Faculté de médecine et des sciences de la santé de l'Université de Sherbrooke, Sherbrooke, Quebec, Canada c Département de génie de la production automatisée, École de technologie supérieure, Montreal, Québec, Canada d Département des sciences de l'activité physique de l'Université du Québec à Trois-Rivières, Canada abstract article info Article history: Received 5 May 2010 Accepted 20 October 2010 Keywords: Knee osteoarthritis Physiotherapy Gait Principal component analysis Background: Interpreting gait data is challenging due to intersubject variability observed in the gait pattern of both normal and pathological populations. The objective of this study was to investigate the impact of using principal component analysis for grouping knee osteoarthritis (OA) patients' gait data in more homogeneous groups when studying the effect of a physiotherapy treatment. Methods: Three-dimensional (3D) knee kinematic and kinetic data were recorded during the gait of 29 participants diagnosed with knee OA before and after they received 12 weeks of physiotherapy treatment. Principal component analysis was applied to extract groups of knee flexion/extension, adduction/abduction and internal/external rotation angle and moment data. The treatment's effect on parameters of interest was assessed using paired t-tests performed before and after grouping the knee kinematic data. Findings: Increased quadriceps and hamstring strength was observed following treatment (P b 0.05). Except for the knee flexion/extension angle, two different groups (G 1 and G 2 ) were extracted from the angle and moment data. When pre- and post-treatment analyses were performed considering the groups, participants exhibiting a G 2 knee moment pattern demonstrated a greater first peak flexion moment, lower adduction moment impulse and smaller rotation angle range post-treatment (P b 0.05). When pre- and post-treatment comparisons were performed without grouping, the data showed no treatment effect. Interpretation: The results of the present study suggest that the effect of physiotherapy on gait mechanics of knee osteoarthritis patients may be masked or underestimated if kinematic data are not separated into more homogeneous groups when performing pre- and post-treatment comparisons. © 2010 Elsevier Ltd. All rights reserved. 1. Introduction According to recent clinical guideline recommendations, regular physical activity and lower limb strengthening exercises are key components of knee osteoarthritis (OA) management (Zhang et al., 2007). Exercise has shown to have beneficial effects on decreasing symptoms of pain and improving physical function in knee OA patients. However, its effects on knee biomechanics are still unclear. Although changes in knee biomechanics during gait have been recently reported following a physiotherapy treatment (Turcot et al., 2009), other studies were not conclusive (Lim et al., 2008; Thor- stensson et al., 2007). Possible reasons for this discordance may be related to gait outcomes as well as to intersubject variability observed in both gait patterns and responses to the treatment of patients evaluated. In most gait studies, dynamic joint angles and moment data as a function of the gait cycle are presented in the form of curves (Astephen et al., 2008; Baliunas et al., 2002; Winter, 1988). Specific kinematic or kinetic gait parameters, such as the mean of peak values, are extracted at particular periods of the gait cycle and used for group comparison. However, limitations can be encountered using this technique. Although human gait is a cyclic and repeatable activity, every person has a fairly unique gait pattern, leading to intersubject variability in curve profiles (Winter, 1988). For example, temporal appearance of the parameter of interest can be different among persons being compared. Also, averaging can collapse information to the point of removing important intersubject variability within a given group, whether before or after physiotherapy. To recover the relevant intersubject variability in the gait data before and after physiotherapy Clinical Biomechanics 26 (2011) 284–291 ⁎ Corresponding author. École de réadaptation, Faculté de médecine et des sciences de la santé de l'Université de Sherbrooke, Sherbrooke, Quebec, Canada. E-mail addresses: nathaly.gaudreault@usherbrooke.ca (N. Gaudreault), Neila.Mezghani@etsmtl.ca (N. Mezghani), kturcot@gmail.com (K. Turcot), Nicola.Hagemeister@etsmtl.ca (N. Hagemeister), karine.boivin@polymtl.ca (K. Boivin), Jacques.deGuise@etsmtl.ca (J.A. de Guise). 0268-0033/$ – see front matter © 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.clinbiomech.2010.10.004 Contents lists available at ScienceDirect Clinical Biomechanics journal homepage: www.elsevier.com/locate/clinbiomech