Motion Analysis Using the Gaitsmarttmsystem in the Imi-Approach Cohort

Purpose: Disease modifying therapy of osteoarthritis (OA) represents an unmet clinical need, and appropriate outcome measures are required to accurately identify patients with different OA phenotypes who may benefit from specific therapies. APPROACH is an exploratory, European, 5-centre, 2-year prospective follow-up cohort study to identify such phenotypes. One of the possible new OA markers used in APPROACH is motion analysis using the GaitSmartTM system. The objective of this study is to evaluate the additional value of GaitSmartTM as outcome measure in knee OA. Methods: 297 knee OA patients were included in the APPROACH cohort study (age; 66.5±7.1, female; 230 (77%), BMI; 28.1±5.3). For this specific study, baseline and six months follow-up (M6) data were used. The GaitSmartTM system uses six inertial measurement units, comprising 3 tri-axial accelerometers and three tri-axial gyroscopes. For these analyses 15 GaitSmartTM measurements were selected based on previous research and clinical expertise (Table 1). Firstly, using solely baseline data, principal component analysis (PCA) was performed to explore structure in relationships between individual GaitSmartTM parameters and reduce the total set of parameters to a limited set of underlying domains, which were used in further analyses. Next, logistic regression was used to evaluate the relationship of identified GaitSmartTM domains with the presence of radiographic knee OA (ROA), defined as Kellgren & Lawrence grade ≥2 in at least one knee, in addition to demographics; age, BMI, and gender, and commonly used patient reported outcome measures (PROMs); pain and activities of daily living (ADL) subscales of the Knee injury and Osteoarthritis Outcome Score (KOOS). Independent variables were entered stepwise starting with demographics, then KOOS subscales, and finally GaitSmartTM domains. It was also evaluated whether the association between GaitSmartTM domains with ROA depended on pain severity, by testing interaction terms. Statistically significant interactions were added as a fourth step. The area under the receiver operating characteristics curves (AUC-ROC) was calculated for all models as a measure of (increase in) model fit. Subsequently, linear regression was used to explore whether GaitSmartTM domains were associated with commonly used function measures for knee OA. As there is no ‘gold standard’ for function, we evaluated whether six commonly used outcome measures (KOOS ADL, KOOS sports and recreational function (sport/recr), SF-36 physical functioning (SF-36 physical function), SF-36 role limitations due to physical health (SF-36 role physical), 30s chair-stand test, and 40m self-paced walk test) could represent one or multiple function domains, using PCA. Resulting domain(s) were used as outcome variables for linear regression. Linear regression modelling started with a full model including all possible predictors and interaction terms. Then all predictors with a p-value >0.2 were removed stepwise. If interactions were relevant predictors, the individual parameters of these interactions were retained as well. Resulting regression formula(s) were used to construct GaitSmartTM based function scores (GS total function, GS performance-based function and GS self-reported function). Lastly, using M6 data in addition to baseline data, patients were divided based on an in- or decrease of at least the minimal detectable change in 30s chair-stand test or 40m self-paced walk test. Standardized response mean within the subgroup was calculated for the other function outcome measures. The differences in changes after six months between subgroups were compared using t-tests and Hedges’ g effect sizes, to evaluate responsiveness of GaitSmartTM to clinical change. Results: PCA of baseline GaitSmartTM parameters identified five underlying domains: one mainly related to ROM in hips (GS Hip), one mainly related to ROM of knee and calves (GS Knee), one mainly related to differences either ROM of knees and calves in swing phase (GS Difference knee), ROM in hips (GS Difference hip), and ROM of knees during stance phase (GS Difference stance) (Table 1). 159 patients (56%) had ROA in at least one knee. Logistic regression showed that addition of GaitSmartTM domains to the model improved the association with the presence of ROA. The final model, including statistically significant interaction terms between pain and GaitSmartTM domains, resulted in a Nagelkerke R2 of 0.212, AUC-ROC of 0.724 (95%CI 0.665, 0.783), sensitivity of 74.2%, and specificity of 51.2% (table 2, figure 1). Using the default setting of an eigenvalue>1 in the PCA, all six function outcome measures loaded on one domain: total function. When extracting two components, a division into performance-based function and self-reported function was found (Table 3.) Using these three function domains as outcome variable in linear regression modelling resulted in final models with adjusted R2 of 0.121, 0.227, and 0.316 for self-reported, performance-based and total function, resp. (Table 4) After 6 months, when there is a decrease in sit-to-stand activity (table 5a), it is most prominently detected by the GS self-reported and total function score. When there is an improvement in sit-to-stand activity, it is most prominently found in the GS performance-based function score. A decrease in walking activity (Table 5b) is most prominently detected by the GS total function score, whereas an improvement in walking activity is most prominently found in the GS performance-based function score. Effect sizes for worsening compared to improving in 30s chair-stand test, as well as 40m self-paced walk test, are highest for all three GaitSmartTM based function scores, meaning these are more responsive to detect an actual change in sit-to-stand or walk activity, as compared to commonly used function measures (Table 5a And B). Conclusions: GaitSmartTM outcome is associated with the presence of radiographic knee OA. The increase in adjusted R2 compared to models using only demographics and PROMs is considerably. Besides, GaitSmartTM is related to commonly used function measures in knee OA, combining evaluation of self-reported and performance-based function. Motion analysis using the GaitSmartTM system might serve as additional non-invasive and easily applicable measure to assess OA, responsive to short-term changes in functional activity. Acknowledgements: This work has received support from the EU/EFPIA Innovative Medicines Initiative Joint Undertaking (APPROACH grant n° 115770 ). This communication reflects the views of the authors and neither IMI nor the European Union and EFPIA are liable for any use that may be made of the information contained herein. See www.imi.europe.eu and www.approachproject.euView Large Image Figure ViewerDownload Hi-res image Download (PPT)View Large Image Figure ViewerDownload Hi-res image Download (PPT)View Large Image Figure ViewerDownload Hi-res image Download (PPT)View Large Image Figure ViewerDownload Hi-res image Download (PPT)