Agreement between measurements of stance width using motion capture and center of pressure in individuals with and without Parkinson’s disease

Background Many individuals with Parkinson’s disease exhibit narrow stance width during balance and gait. Because of this, stance width is an important biomechanical variable in many studies. Measuring stance width accurately using kinematic markers in parkinsonian patients can be problematic due to occlusions by research staff who must closely guard patients to prevent falls. Methods We investigated whether a measure of stance width based on the mediolateral distance between the center of pressure under each foot could approximate stance width measured with kinematic data. We assessed the agreement between estimates of stance width obtained from simultaneous kinematic and center of pressure measures during quiet standing in 15 individuals (n=9 parkinsonian, n=6 age-similar neurotypical). The source data (1363 unique trials) contained observations of stance width varying between 75–384 mm (≈25-150% of hip width). Findings Stance width estimates using the two measures were strongly correlated (r = 0.98). Center of pressure estimates of stance width were 48 mm wider on average than kinematic measures, and did not vary across study groups (F2,12=1.81, P<0.21). The expected range of differences between the center of pressure and kinematic methods was 14–83 mm. Agreement increased as stance width increased (P<0.02). Interpretation It is appropriate to define stance width based on center of pressure when it is convenient to do so in studies of individuals with and without Parkinson’s disease. When comparing results across studies with the two methodologies, it is reasonable to assume a bias of 48 mm.


Abstract 31
Background 32 Many individuals with Parkinson's disease exhibit narrow stance width during balance and gait. Because 33 of this, stance width is an important biomechanical variable in many studies. Measuring stance width 34 accurately using kinematic markers in parkinsonian patients can be problematic due to occlusions by 35 research staff who must closely guard patients to prevent falls. 36

Methods 37
We investigated whether a measure of stance width based on the mediolateral distance between the center 38 of pressure under each foot could approximate stance width measured with kinematic data. We assessed 39 the agreement between estimates of stance width obtained from simultaneous kinematic and center of 40 pressure measures during quiet standing in 15 individuals (n=9 parkinsonian, n=6 age-similar 41 neurotypical). The source data (1363 unique trials) contained observations of stance width varying 42 between 75-384 mm (≈ 25-150% of hip width). 43

Findings 44
Stance width estimates using the two measures were strongly correlated (r = 0.98). Center of pressure 45 estimates of stance width were 48 mm wider on average than kinematic measures, and did not vary across 46 study groups (F 2,12 =1.81, P<0.21). The expected range of differences between the center of pressure and 47 kinematic methods was 14-83 mm. Agreement increased as stance width increased (P<0.02). 48

Interpretation 49
It is appropriate to define stance width based on center of pressure when it is convenient to do so in 50 studies of individuals with and without Parkinson's disease. When comparing results across studies with 51 the two methodologies, it is reasonable to assume a bias of 48 mm. 52

Introduction 56
Many individuals with Parkinson's disease (PD) exhibit narrow stance width during balance and gait 57 (1). Clinically, "narrow stance" is a postural abnormality in which the feet are placed substantially medial 58 to the anterior superior iliac spines (ASIS) (2). Stance width is therefore an important variable in many 59 studies of parkinsonian posture and balance (e.g., (3-5)). It is typically treated as a nominal single value or 60 as a range of values described by the mediolateral distance between kinematic markers placed on the 61 heels, or between the medial malleoli (3-5). 62 Due to repeated protective steps, dyskinesias, and other practical concerns when studying 63 parkinsonian balance, it is difficult to control stance width precisely during experiments -and so ideally, 64 stance width should be measured as a continuous covariate throughout an experiment. However, doing so 65 with kinematic markers can be problematic due to occlusions by research staff who must carefully guard 66 patients to prevent falls. 67 Here, we investigated whether a proxy measure of stance width based on the mediolateral distance 68 between the centers of pressure (CoP) beneath each foot could approximate stance width measured 69 kinematically. As typically defined (6), the CoP is the point location of the vertical ground reaction force 70 vector beneath the entire body, and represents a weighted average of all the pressures over the surface 71 area in contact with the ground (6). Whole-body CoP location is often calculated as an important outcome 72 variable in clinical balance studies (5, 7, 8). If bilateral force plates are used, CoP can be calculated 73 separately for each foot (e.g., as it is in instrumented treadmill studies (9)). Since the CoP of each foot 74 must be located within its boundaries, the mediolateral distance between them must be considerably 75 associated with the stance width between the heels during bipedal standing. 76 We used the approach suggested by Bland and Altman (10) to assess agreement between stance width 77 estimated from foot CoP and measured kinematically in neurotypical individuals (NT) and in 78 parkinsonian individuals in the ON (PD-ON) (8) and OFF (PD-OFF) (11) medication states. We 79 quantified the bias and expected range of differences associated with using stance width estimates from 80 foot CoP rather than kinematic measures. Then, we tested whether differences between methods were associated with group membership (NT vs. PD-ON vs. PD-OFF), and whether differences varied with 82 stance width (12) We used baseline measurements from a convenience sample of participants in previous (3) and 87 ongoing cohort studies investigating the effects of rehabilitation on balance responses (Table 1). PD 88 participants were mild-moderate with bilateral symptoms (Hoehn and Yahr stage 2-3 (13)). All 89 participants provided written informed consent and all study procedures were approved by Institutional 90 Review Boards at the Georgia Institute of Technology and Emory University. 91

EXPERIMENT 92
As in previous studies (3,14), participants stood barefoot on two laboratory-grade force plates 93 (AMTI-OR6-6-1000, AMTI, Watertown, MA, USA). The force plates were mounted onto a custom 94 translation platform; however, analyses here considered only periods during which the platform was 95 stationary. Force and moment data were sampled at 1080 Hz and used to calculate the locations of the 96 center of pressure beneath each foot using calibration values supplied with the plates (15-17). Kinematic 97 data were collected at 120 Hz using a Vicon motion capture system (Centennial, CO, USA) and a 25-98 marker set including reflective markers placed on the left and right heels. Average foot CoP locations and 99 heel marker positions were calculated over the first 250 ms of each trial. 100 Stance width was controlled by requesting participants press an object (typically a book) between the 101 medial surfaces of their feet, which was subsequently removed before data collection (≈87% of trials), or 102 by manipulating participant's feet so that kinematic markers on the heels were aligned in the mediolateral 103 direction with tape marks on the floor (≈13%).

DATA ANALYSIS 105
Stance width measurements derived from CoP and kinematic data were plotted against each other and 106 examined visually. After visual assessment of outliers, trials were excluded due to: 1) tension in a ceiling-107 mounted fall arrest tether interfering with CoP calculation (17 trials in one participant), and 2) absent 108 video records preventing trial review (2 trials in one participant). After applying exclusions, 1363 trials 109 The SAS Institute, Cary, NC, USA) and considered significant at P = 0.05. 121

Results 122
Stance widths measured from kinematic data varied between 75 -348 mm, corresponding to 24.9 -123 154.1% of inter-ASIS distance. CoP and kinematic stance width measurements are presented in Figure  124 1A. The two measures were strongly correlated (r = 0.98). The mean difference d between methods was 125 48 mm, and the standard deviation of the differences (s) was 17 mm. Differences d i did not vary across 126 groups (F 2,12 =1.81, P<0.21). The limits of agreement, defined as the range d-2s to d+2s (10), was 14-83 127 mm. A "Bland-Altman plot" of the differences between the two methods d i against their means m i is

Discussion 130
Stance width is an important variable in many studies of parkinsonian (4, 5) and neurotypical (18,19) 131 posture and balance. We found that stance width estimates from foot CoP and kinematic markers were 132 strongly linearly correlated, and that on average, measures of stance width derived from CoP were 48 mm 133 wider than those derived from kinematic markers. This bias that can be explained by the externally-134 rotated "toe out" posture used by most participants, in which a substantial portion of the foot plantar 135 surface lies lateral to the posterior face of the heel. Overall, these results suggest that foot CoP location, a 136 commonly calculated variable in clinical biomechanics studies (5,7,8) can be used to approximate stance 137 width in healthy aging and in individuals with PD in the ON and OFF medication states. 138 We noted that differences between methods were non-negligible -ranging from 14 to 83 mm. 139 However, this precision is adequate to discriminate between nominal stance widths used in the literature, 140 which are typically separated by 100 mm or more (4, 18). Due to the high precision of CoP calculation 141 with laboratory force plates (2-5 mm (17)), the primary source of variability in differences is probably 142 trial-to-trial variability in weight distribution, rather than instrumentation error. 143 There are two notable limitations to this approach. First, differences between methods were highest at 144 the narrow stance widths preferred by PD subjects, a fact that should be considered carefully during study 145 design. Second, because these participants were allowed to adopt a comfortable "toe out" orientation 146 during testing, the agreement between the methods in experimental paradigms in which foot orientation is 147 enforced (e.g., parallel (4); 20° (18)) remains to be established. 148

Conclusion 149
In summary, these results suggest that: 1) it is appropriate in studies of individuals with and without 150 PD to define stance width based on CoP, and 2) when comparing results across studies with the two 151 methods, it is reasonable to assume a bias of 48 mm.

Competing Interests 164
The author has declared that no competing interests exist.  Table 1

1A. Comparison of stance width (mm) measured from heel markers and estimated from CoP
Difference between two methods (CoP-markers)