Laboratory versus daily life gait characteristics in patients with multiple sclerosis, Parkinson’s disease, and matched controls

We recruited people with MS, age-matched MS-Ctl, people with PD, and age-matched PD-Ctl for this study which is the part of a larger, longer study to identify gait measures that predict prospective falls in MS and PD. Our cohort with MS was younger than the cohort with PD, so we recruited younger and older control groups for comparison. Inclusion and exclusion criteria were the same as described in Shah et al. [35]. Specifically, inclusion criteria for PD were a diagnosis of idiopathic Parkinson’s disease from movement disorders neurologist with the United Kingdom Parkinson’s disease Society Brain Bank criteria, Hoehn & Yahr scores of II-IV, and complaints about gait. Inclusion criteria for MS were a diagnosis of relapsing–remitting MS confirmed by a neurologist specialist, a mild-to-moderate MS-associated disability (patient-reported EDSS score ≤ 6.0), and complaints about gait. Exclusion criteria for all subjects included the inability to follow protocol instructions, other factors affecting gait such as musculoskeletal disorders, uncorrected vision or vestibular problems, or inability to stand or walk in the home without an assistive device. The experimental protocol was approved by the Institutional Review Board of the Oregon Health & Science University (eIRB #15578). The experimental protocol was carried out in accordance with the institution’s ethical committee, and all the participants provided informed written consent.

In the laboratory, subjects were asked to wear 3 inertial sensors (Opals by APDM Wearable Technologies, Portland, OR, USA; Fig. 1a); one sensor on top of each foot, and one over the lower lumbar on an elastic belt. Each Opal sensor includes a tri-axial accelerometer, gyroscope, and magnetometer with a sampling rate of 128 Hz. The Opal is lightweight (22 g), has a battery life of 16 h, and includes 8 GB of storage, that can record over 30 days of data. All the subjects performed the Instrumented Stand and Walk test (ISAW) [36]. The ISAW consists of standing quietly for 30 s, followed by a verbal instruction to initiate gait, walk 7 m, turn 180 degrees after crossing a line on the ground, and return to the initial starting position [36]. The ISAW test was designed to measure postural sway, step initiation, gait, and turning, all in one short test. Individuals with MS completed estimates of the severity of MS with Patient-Reported Expanded Disability Status Scale (PREDSS) [37, 38], walking ability with Multiple Sclerosis Walking Scale (MSWS-12) [39], and fatigue with Modified Fatigue Index Scale (MFIS) [40]). PD severity was assessed by a certified researcher using the Movement Disorders Society-Unified Parkinson’s Disease Rating Scale (MDS-UPDRS), Part III [41]). All subjects were tested in the laboratory 1 h after taking their regular medication intake (ON-medication state), because most of the time, subjects attempt to be in the ON-medication state in daily life.

Subjects were asked to wear instrumented socks (prototype developed by APDM Wearable Technologies, Portland, OR, USA; Fig. 1b) on each foot, and one Opal sensor over the lower lumbar area with an elastic belt. They wore the sensors for at least 8 h/day for a week of daily life. To reduce the burden on participants of trying to securely attach the Opal monitors to the outside of their shoes, APDM Wearable Technologies designed an instrumented sock that wraps around the participant’s foot and ankle with the inertial sensor inserted into a small, lightweight pocket in the foot area. The inertial sensor within the sock is located on the dorsum of the foot like the Opal sensors worn in the laboratory, The main unit containing the battery is located in a second pocket just above the lateral malleolus. To maximize fit, the socks come in different sizes, and the Velcro attachment around the foot and ankle is adjustable to ensure that a snug fit and that the sensor does not move on the foot while being worn. Thus the inertial system fits into the instrumented sock for ease of application and safe, unobtrusive use. The instrumented socks and Opals have the same inertial sensors with the same sensor specifications but different form factor. The instrumented socks are synchronized with the Opal worn on the lumbar area. This allowed subjects to comfortably wear the instrumented socks in their shoes or slippers, and without a visually-distracting external strap attachment. It also eliminated the task of securing the monitors on the shoes, especially important for people (such as people with PD), who have difficulty with fine motor movements, such as tying shoelaces. The subjects removed the socks and the belt at night to recharge the batteries. Data were stored in the 8 GB internal memory of the Opals and socks. Subjects returned instrumented socks either by mail using a pre-paid box after completion of a week of data collection or to a research assistant who met subjects at their homes. Data were uploaded to a secure cloud-based database upon return of the devices and downloaded to a local computer for further processing using the same gait algorithms for laboratory and daily life gait (after identification of appropriate length gait bouts).

The algorithms used for extracting spatial and temporal measures of gait and turning are identical for laboratory and daily life gait and have been detailed previously [35, 42]. In addition, the daily life algorithm first searches for possible bouts of walking using a time-domain approach to inertial sensor data from the feet and for turns based on yaw rotational orientation of the pelvis. Second, individual steps are combined into potential bouts of walking, as long as the duration from one step to the next step is no longer than 2.5 s. Finally, each possible bout that contains at least 3 steps and is at least 3 s in duration is processed with the commercial gait analysis algorithms included in Mobility Laboratory (APDM Wearable Technologies, Portland, Oregon) [36, 43, 44]. Our gait analysis algorithm uses the Unscented Kalman Filter to fuse information from the accelerometers, gyroscopes, and magnetometers to precisely estimate the orientation and position trajectory of each foot between quiet stance periods [45, 46]. This approach reduces the problem of tracking over a long period of time.

To compare between in-laboratory and daily life gait measures, we used only gait bouts that had a similar number of strides during the ISAW test in the laboratory and during free-living daily activities (4–15 strides in about). Specifically, we calculated number valid of strides for the ISAW test in the laboratory for all participants and found that it ranged between 4 to 15 strides, so for the bout analysis in daily life we used the same range (4–15 strides) corresponding to the stride range observed in the laboratory ISAW test. In total, we chose 13 gait measures (see Fig. 3 and Additional file 1: Table S1) that are commonly used to characterize gait in the laboratory and grouped them into three aspects of gait for simplicity: 3 upper-body, 5 spatial, and 5 temporal measures of gait [47]. We avoided comparing the coefficient of variation measures as it is affected by the environment and by bout length [48], and also the number of strides in a bout for our analysis was less than the minimum number of strides required to accurately calculate a coefficient of variation (20 strides) [49, 50].

The normality of data was determined with Shapiro–Wilk tests and parametric analysis was used, unless otherwise stated. Independent t-tests or Mann–Whitney U tests (if not normally distributed) were used to compare between-group differences in subject characteristics, adherence, and total weekly walking bouts. As not all gait quality measures in the laboratory and during daily life were normally distributed, to be consistent, non-parametric test (Mann–Whitney U test) was used to compare differences between groups for all gait measures (i.e., MS vs MS-Ctl and PD vs PD-Ctl). In addition, we computed Receiving Operating Characteristics (ROC) and calculated the Area Under Curve (AUC) [51] to discriminate gait measures in people with MS from MS-Ctl and people with PD from PD-Ctl, and ordered measures from the highest to lowest AUC value. Paired Wilcoxon tests were used to compare the laboratory and daily life gait measures for MS, MS-Ctl, PD, and PD-Ctl. All statistical analysis was performed using R Version 1.1.456 software. The statistical significance was set to p < 0.05, and we used threshold of p ≤ 0.004 based on the Bonferroni’s correction (due to 13 multiple comparisons) just to prioritize gait measures.

Fifteen people with MS, 16 age-match MS-Ctl, 16 people with PD (Hoehn and Yahr stage with II (n = 26), III (n = 2), and IV (n = 1)), and 15 age-matched PD-Ctl participated in the study. Table 1 shows the demographics and activity characteristics of subjects who participated in this study. Age, height, and weight were similar between the MS and MS-Ctl and between the PD and PD-Ctl groups. Adherence to the weekly recordings for each subject group was similar with 60.19 ± 11.02 (mean ± SD) hours in MS, 64.15 ± 9.59 h in MS-Ctl, 67.66 ± 12.53 h in PD, and 64.67 ± 10.13 h in PD-Ctl of daily life data. The histogram in Fig. 2 illustrates the number of strides in each bout during daily life, and it is evident that the stride range (4–15 strides) in a bout considered for the analysis in this study captures the major portion of participants’ daily activity.

Measures discriminating gait characteristics in people with MS from MS-Ctl in laboratory versus daily life were different (see Table 2 and Fig. 3a). Specifically, the toe-off angle was the most discriminative in the laboratory (p = 0.004; AUC [95% CI] = 0.796 [0.628–0.964]) whereas gait speed was in daily life (p = 0.001; AUC = 0.842 [0.686–0.998]). Stride length was the second best discriminative measure in the laboratory (p = 0.027; AUC = 0.735 [0.556–0.915]) whereas the duration of swing phase as percent of gait cycle was in daily life (p = 0.002; AUC = 0.812 [0.648–0.977). Furthermore, more gait measures discriminated MS from MS-Ctl in daily life (n = 3) compared to the laboratory (n = 1), after the Bonferroni’s correction (13 multiple tests).

Most of the measures discriminating gait characteristics in people with PD from PD-Ctl in laboratory versus daily life were different (see Table 3 and Fig. 3b). Specifically, the lumbar coronal range of motion was the most discriminative in the laboratory (p = 0.010; AUC = 0.775 [0.589–0.961]) whereas the foot-strike angle was in daily life (p = 0.001; AUC = 0.838 [0.697–0.978]). The foot-strike angle was the second best discriminative in the laboratory (p = 0.011; AUC = 0.767 [0.578–0.955]) whereas gait speed was in daily life (p = 0.003; AUC = 0.804 [0.632–0.976]). Furthermore, more daily life (n = 2) than laboratory (n = 0) measures discriminated gait in the PD group from the PD-Ctl group, after the Bonferroni’s correction.

Generally, the gait characteristics in daily life compared to the laboratory reflected a slowing down behavior in all groups, and subjects performed better in the laboratory compared to daily life (see Fig. 4). Specifically, looking at the gait measures that were significant after the Bonferroni's correction in PD and MS groups, gait speed (except MS-Ctl), swing duration (% of the gait cycle), and the toe-off angle were significantly lower in daily life compared to the laboratory in all groups. Further, double support time was significantly longer in daily life compared to the laboratory in all groups. In contrast, the foot strike angle was not statistically significant between laboratory and daily life in all groups except for MS-Ctl.

Long double-support time, slow gait speed, and short swing time (all affected by balance impairment and fatigue) [7, 9, 52] were significantly different daily life gait measures in MS from MS-Ctl. Indeed, gait speed double-support time and swing time as a percent of the gait cycle all discriminated gait in people with MS from gait in healthy control people over a week of daily life with a similar, excellent area under the curve [35]. In contrast, in the laboratory, the toe-off angle was the only laboratory gait measure that discriminated our mild-moderate MS from MS-Ctl group during comfortable-pace gait after Bonferroni’s correction for 13 gait characteristics. This result is consistent with our previous report of a small toe-off angle in a separate group of people with MS during a 2-min walk in the laboratory [53]. The toe-off angle is a surrogate for the push-off phase of gait produced by the power in the gastrocnemius-soleus complex, responsible for stride length and gait speed.

Slow gait speed (representing hypokinetic gait) and small foot strike angle (representing shuffling of gait) were significantly different daily life gait measures in the PD group compared to the PD-Ctl group. Previous studies of gait in daily life agree that foot strike angle [35], and gait speed [14] discriminated gait in PD from healthy control groups. Surprisingly, none of the laboratory gait measures discriminated gait characteristics in mild-moderate PD (ON state), from the PD-Ctl group, after Bonferroni’s correction, suggesting that monitoring gait during daily life is more sensitive to impairments from PD than gait test in the laboratory. The participants with PD showed much larger changes in their gait parameters between the laboratory and daily life than the controls or people with MS. This difference in performance in a laboratory test and daily life in people with PD may be due to their reliance on less automatic, more attention demanding gait mechanisms that would make gait in daily life more challenging [54]. The difference could also be due to people with PD being more prone to placebo effects and white coat effects than the other groups, so they perform better when their performance is observed. Alternatively, it might be that we picked up the ON and OFF fluctuations during daily life that influenced the averaged gait measures over a week. Nevertheless, assessing mobility during daily life resulted in more sensitive and specific differences in gait characteristics than laboratory gait between the PD and control groups.

Interestingly, the lumbar coronal range of motion was one of the top gait measures discriminating both the MS and PD groups from their age-matched controls in the laboratory, but not in daily life. The inability of lumbar motion to discriminate during daily life might be due to lumbar sensor measures being affected by the exact location of the sensor. In the laboratory, the researchers make sure the lumbar sensor location is consistent and stays securely attached throughout the testing, for all subjects, but it is hard to maintain a consistent sensor location placed by the subject in daily life conditions, and thus might not a reliable measure during daily life. Reduced lumbar range of motion while walking may reflect axial rigidity and loss of arm swing in the PD group [55], and may reflect the compensatory strategy to truncal ataxia in the MS group. In contrast to the reduced lumbar range of motion in the MS group here, our previous study found an excessive lumbar motion in people with very early MS who had normal gait speed [52].

Longer bout lengths, such as in 1-min laboratory tests, are known to result in faster gait speed and other accompanying measures [5, 14]. There are various ways to measure the bout length. Researchers have used bout duration [5, 14, 31], and the distance traveled during a particular walk test [32] as bout length measures. We chose to define bout length in terms of a number of strides in the bout because it helps to eliminate the effect of gait speed, per se, on the bout length. Most gait bouts during daily life have < 15 strides in all 4 groups so the 7 m × 2 in the ISAW test reflected the most common bout lengths people actually take during daily life.

Our results suggest that clinicians should consider quantitative daily life gait behavior as an integral part of a functional clinical assessment. Furthermore, this study provides encouraging results to support the use of instrumented socks for daily life gait evaluation in people with PD and MS, and also a potential to use in clinical trials, with a possibility that fewer subjects will be required for clinical trials using this quantitative measurement of mobility in daily life.

There are several limitations of the current study. First, we had a modest sample size of only 15–16 subjects in each group. This also resulted in a modest statistical power for detecting differences. If a larger number of subjects had been included, additional measures would have been able to discriminate between the neurological groups from their matched controls. Further analysis is needed with larger cohorts to test the generalizability of the findings. Second, we used a conservative correction for multiple hypothesis tests. Many of the tests we performed were on measures of gait that at correlated and not statistically independent. The Bonferroni correction assumes these tests are independent, so the correction may have reduced the power of the statistical tests so that additional measures are actually statistically significant. Thirdly, for daily life data, we assumed that subjects attempt to be in the ON-medication state most of the time, and hence, we compared with laboratory walking test only with subjects with their ON state. Further, future studies need to determine the test–retest reliability and sensitivity of the top mobility measures to a treatment and disease progression in daily life to be useful as digital biomarkers for clinical trials. Finally, with larger cohorts, we can investigate if the paired ROCs in a laboratory and daily life are statistically significant.