Multiple sclerosis (MS) is a chronic inflammatory disease of the central nervous system that affects more than two million people worldwide [1
]. MS can cause various neurological and functional disorders, such as alterations in sensation, vision, cognition, balance, and walking capacities [2
]. Among the functional limitations encountered by people with MS, those concerning walking are of particular importance due to their negative impact on physical activity and quality of life [3
Assessing walking capacities, both with walking distance metrics (e.g., total distance during a six-minute walking test) and walking speed metrics (e.g., usual walking pace), has clinical interests in people with MS, such as understanding to what extent mobility may be altered during daily life, and quantifying functional benefits from therapies that would aim at improving mobility [4
]. Although self-perception of walking capacities in people with MS is clinically relevant information, objective measurements are needed to have a more accurate picture of the capacities of the patient, especially since subjective measurements using scales may not reflect the results obtained from objective measurements [6
Several laboratory-based walking tests have been used to objectively measure walking capacities of people with MS [6
]. However, these tests have some drawbacks, including the need of an appropriate space to conduct the test, and the fact that capacities in real-life contexts may not be well reflected by laboratory-based measurements [6
]. In contrast, outdoor evaluations, that can be implemented in large areas and for a relatively long duration, may allow us to better reproduce and more directly measure the daily life outdoor walking capacities of patients in terms of walking speed, endurance, and pattern (e.g., inter-walking bout variability in walking speed, variability in recovery duration, etc. [7
]), with the possibility to perform the evaluations close to home [8
] and thus with better accessibility for the patients with sufficient functional capabilities. Thus, some studies [9
] have proposed an evaluation of walking capacities of people with MS in natural (i.e., outdoor) contexts where walking capacities were characterized using the greatest distance, measured by Global Positioning System (GPS), that a patient could walk before stopping due to symptoms (e.g., fatigue), also called “maximal walking distance”. Unfortunately, some concerns could be raised from these studies regarding the validity of the GPS-measured maximal walking distances in view of the implemented methodologies. Indeed, in the study by Créange et al. [9
], the patients were asked to perform a single maximal walking bout (i.e., the greatest distance as possible) at usual pace around a hospital. However, it is unknown as to whether allowing several walking bouts to be performed, rather than a single one as in the study by Créange et al. [9
], would allow the patients to reach a higher maximal walking distance, as it has been observed in other diseases inducing walking limitations [11
]. In the study by Dalla-Costa et al. [10
], the result used to characterize patients’ walking capacity was the mean of the daily greatest walking distances measured during several days in real-life contexts. However, in this last study, there was no information about how GPS data were processed to detect walking bouts and then calculate walking distance during daily life. Moreover, some studies conducted in other populations with walking limitations [7
] suggest that, in addition to maximal walking distance, other outcomes could be of interest when measuring outdoor walking capacities in people with MS, such as outdoor usual walking speed, which is a parameter that may be needed to reveal the improvement of the functional status following a treatment procedure such as surgery [12
While GPS allows easy outdoor measurements of speed and distance, one of the methodological challenges raised when analyzing data from an evaluation that includes multiple walking bouts is to correctly discriminate the actual walking bouts from the actual stopping bouts for then calculating walking distance. Previous works [13
] have shown that a semi-automatic speed processing methodology can correctly detect walking bouts and accurately estimate walking speed and distance. While this procedure has been successfully implemented in patients with intermittent claudication [7
], it is unknown as to whether it is usable in MS patients with various functional profiles, as reflected for example by various usual walking speeds or capacities of maintaining balance and pace on nonlinear paths. Thus, the objective of the present study was three-fold:
illustrate, using fully open materials, how GPS data, in particular speed, obtained during an outdoor evaluation allowing multiple walking bouts, could be used along with a previously validated speed processing methodology [13
] to characterize walking capacities in people with MS;
highlight methodological issues that may occur when implementing an outdoor walking evaluation with GPS measurements in people with MS; and
explore the construct validity of outdoor maximal walking distance and outdoor usual walking speed as functional outcomes in people with MS.
The present study aimed at exploring the GPS-measured outdoor walking capacities of people with MS assessed during a session that had to be performed at usual pace, with free recovery durations, and allowing multiple walking bouts. While this kind of evaluation has been implemented several times amongst people with intermittent claudication [7
], this is, to our knowledge, the first results obtained in people with MS using such a procedure.
Unfortunately, it is difficult to compare our GPS-measured outdoor maximal walking distances with the two previous GPS studies conducted in people with MS [9
]. Indeed, Dalla-Costa et al. [10
] did not provide summary statistics for the recorded outdoor maximal walking distance. Moreover, while Créange et al. [9
] obtained results from participants who had to walk until being forced to stop due to fatigue whatever the time needed to do this, we proposed in the present study an evaluation with an a priori fixed upper time limit. Thus, in our study, the least disabled participants did not reach a maximal walking distance during the session. This could explain why Créange et al. [9
] obtained much higher values of outdoor maximal walking distance (median [IQR]: 1540 m [50–4550]) than in the present study.
Regarding outdoor usual walking speed, our participants could be considered as “slow walkers” in comparison with healthy people. Indeed, when considering all the walking time of the session, our participants walked at a median speed of 2.52 km/h (0.70 m/s), while a recent meta-analysis [23
] estimated outdoor usual walking speed in healthy adults at a mean (95% CI) of 1.31 m/s (1.27; 1.35). The outdoor usual walking speed of our MS participants was also lower than that observed in other chronic disease populations, such as patients with intermittent claudication where a median [IQR] outdoor walking speed of 3.6 km/h [3.1–3.9 km/h] has been observed in a sample of 203 patients [8
]. Our results, at the group level, are also lower than a recent estimate of the comfortable walking speed in people with MS [4
], where authors found a mean speed (95% CI) of 1.12 m/s (1.05; 1.18). Thus, our sample of MS participants could be viewed as particularly disabled if we considered usual walking speed as a functional outcome [4
]. Of note, because our methodology implemented a minimum bout duration of 15 s, the estimate of session mean walking speed could be underestimated because very short walking bouts that initially remained following speed data processing were related to low walking speeds. Including these bouts in ≥15-s bouts could have led to more weight being given to a walking bout with a lower mean speed in the final calculation of the session mean speed. To assess the impact of our methodology on the estimate of session mean speed, we conducted a sensitive analysis using no minimum bout duration to detect a given bout. This resulted in a median [IQR] (90% CI) walking speed of 2.64 km/h [2.22–3.05] (2.19; 2.93), that is, 0.12 km/h higher than following our initial analysis. Thus, the effect of the minimum bout duration of 15 s on session mean speed could be considered as not substantial because such variation corresponded to the typical error of estimate when speed is measured with the DG100 GPS during walking [14
Interestingly, outdoor maximal walking distance was not always recorded during the first walking bout of the session. This was true for 40% of the participants among those who had to stop walking due to symptoms in the present study. While the reasons for which the walking distance performed between two stops could greatly vary from a waking bout to another are unclear, this result suggests that implementing an evaluation session allowing multiple walking bouts may be of interest to have a more reliable result of the outdoor maximal walking distance in people with MS. However, such a procedure could be worthwhile only in patients who present sufficient disability to have the possibility to perform several walking bouts during a session of a reasonable duration. Unfortunately, due to a small sample size, it was not possible here to reliably determine which disability status would be associated with the need to perform several walking bouts during a session, nor the session duration that would be needed to achieve the “true” maximal walking distance of the session. This last information could be worthwhile to avoid the implementation of an unnecessarily time and effort-consuming walking session.
Regarding the methodological issues we encountered, the absence of a resting phase in the GPS data at the beginning of the session for two participants was of importance since it could threaten the validity of the maximal walking distance recorded for those participants. This problem highlighted the need of graphical analysis when dealing with GPS data recorded in the present context, even if the R code we provided could work and provided results despite this loss of data. It could be good practice to determine a precise starting point of the walking session that could be recognized on a map, and to compare it, when necessary, with the recorded starting point to have confidence in the true beginning of the session observed in the GPS data. In the present study, we did not have such a precise starting point but we were able to situate the actual starting point with an approximated error of 20–30 m and ensure that the loss of data for the two concerned participants was not an issue in confirming their outdoor maximal walking distance.
The second methodological issue we met was that the walking courses we proposed could influence the walking speed in some participants. This fact could be observed in some of the figures created to show all the GPS data related to a given participant (e.g., Figure 5
; see also all figures included in the OSF repository: https://osf.io/tp37b
, lastly updated on 17 April 2021), with a walking speed that sometimes fell to nearly 0 km/h where turns were steep. The fall in walking speed in such cases could reveal a strategy of the concerned participants to maintain balance in these situations, as it has been observed and explained during 6-min walking tests with courses including 180° turns [24
]. This issue may have influenced the walking distances performed over the walking bouts and necessarily led to a decrease in mean walking speed over the concerned walking bouts. Such falls in walking speed could have resulted in an increase in the number of detected bouts, but the minimal durations of 15 s we used to validate a bout seem to have prevented this. Thus, good practice for future implementations of GPS-based measurements of outdoor walking capacities in people with MS would be to design walking courses without steep turns to allow all participants to keep a relatively constant walking speed. This would allow easier analysis and easier comparisons of outdoor maximal walking distances and mean walking speeds between the participants in future studies.
Our results suggest that outdoor maximal walking distance might not well reflect walking endurance capacities as assessed using a 6-min walking test, with r and rho coefficients < 0.50, while mean outdoor walking speed could more likely discriminate against people with MS as it can be done based on the 6-min total walking distance (r and rho > 0.65). While our small sample of people with MS does not allow us to make accurate and definitive conclusions about the correlations between outdoor walking capacities and those assessed using the 6-min walking test, both outdoor maximal walking distance and walking speed may be interesting clinical information regarding how a patient is able to walk in natural contexts. Of note, a previous study [25
] has shown that comfortable walking speed chosen during a maximal walking bout performed inside a clinic is a more stable parameter over different days in people with MS than maximal walking distance, meaning that walking speed would be more likely to allow the detection of change in disability status. This could also be the case for outdoor walking speed in comparison with outdoor maximal walking distance, but this should be investigated in reliability studies.
The present work presents several strengths. First, our results were obtained under normal working conditions related to three different healthcare or sports structures and highlight the possibility, and also the challenges, of conducting GPS-measurements in real-world contexts. Second, all of our GPS results are provided in a context that is transparently presented for all the participants, with the possibility to make judgements about the characteristics of the walking courses (i.e., based on corrected altitude data, longitude/latitude data, and the corresponding map) and about the quality of the GPS signals. Third, the present paper proposes fully open materials with de-identified information, allowing the reproduction of GPS data analysis and statistical analysis. Of note, if used for new investigations, the R code could need some adaptations depending on the structure of the gpx. files, and on the use or not of corrected altitude data.
Several limits have also to be acknowledged. First, during the outdoor walking session, some participants choose to use a walking aid although they were able to walk without it. Moreover, some participants could perform their walk relatively close to other participants. We do not know to what extent our GPS results may have been influenced by these elements, but this should be taken into account for potential comparisons with future studies. Second, our outdoor walking sessions had an upper time limit, forcing the participants to stop walking even if they could have walked a longer distance. Thus, our ability to detect and quantify outdoor maximal waking distance was restricted to the most disabled participants. Third, due to the fact that we used a convenience sample of small size and an exploratory approach, any estimate we provided should be confirmed and more robustly determined with appropriate confirmatory studies. Fourth, despite the fact that we sought to conduct the outdoor walking sessions in places where the walking courses were as flat as possible, we could observe some variations in the altitude profile of the walking courses depending on where the measurements were performed (see SDC 1). This might have differently influenced the walking capacities of the participants (i.e., outdoor maximal walking distance and mean walking speed). Future studies should prefer standardized environments (e.g., an athletic track) to assess outdoor walking capacities in people with MS, at least if the aim is to establish reference values and some relationships between outdoor walking capacities and other variables. However, clinicians could be interested in how patients adapt to variations in the physical environment (e.g., various grades, various terrain surfaces). In this case, evaluations could be performed throughout the day with no constraints about the features of the terrain to observe then how maximal walking distance or walking speed would vary depending on the difficulty of the terrain. Such evaluation would require further software development to put in relation distance, speed, and topography information, for example. Fifth, as we have used a recording rate (1 Hz) that was different from that used in previous validation works (0.5 Hz), it is difficult to precisely predict the accuracy with which we have detected the walking and stopping bouts when using the speed processing methodology.
To allow a more user-friendly experience than using R software when analyzing GPS walking data, we developed a shiny app available on the Web. Of note, a web platform just recently released (https://mapam.ens-rennes.fr
, accessed on 2 April 2021) along with a published paper [26
] now allows a full automatisation of speed data analysis, rather than manually building speed data filters that would be the best suited for the detection of walking and stopping bouts performed during the session. Unfortunately, the web platform does not yet provide contextual information (map, coordinates) nor a full summary of the walking session results as proposed when using the shiny app. Moreover, the overall workflow we used to get the final results may still be difficult to implement during clinical routines. This stresses the need of engineering work to develop software solutions, that would ideally be device-agnostic, personal computer-based, standalone and open source, to visualize GPS data and related context and to easily get the results of interest. Such software solutions could be combined with devices and applications allowing for patients’ tele-monitoring using GPS data to get the results in real-time. Some projects are already ongoing and conducted by other research teams to take up these kinds of methodological and technical challenges (https://project.inria.fr/sherpam
, accessed on 2 April 2021).
From a clinical point of view, future studies could more precisely determine the proportion of people with MS that really need several walking bouts to reach their maximal walking distance, as well as investigate relationships between the well-established EDSS score and outdoor walking capacities. Moreover, it could be interesting to determine to what extent the ability to keep a constant pace over challenging outdoor walking courses (e.g., with 180° turns) could reflect the disability status of the patients; this could be valuable information when dealing with outdoor walking capacities in people with MS. Finally, future works could consider the assessment of gait quality during outdoor walking (e.g., using gait asymmetry or gait variability metrics) to better understand its potential effect on the quantity of walking that people with MS can perform (e.g., assessed using maximal walking distance as dependent variable).