Gait & Posture 84 (2021) 120–126 Contents lists available at ScienceDirect Gait & Posture journal homepage: www.elsevier.com/locate/gaitpost Full length article U-turn speed is a valid and reliable smartphone-based measure of multiple sclerosis-related gait and balance impairment Wei-Yi Cheng a, Alan K. Bourke a, Florian Lipsmeier a, Corrado Bernasconi a, Shibeshih Belachew a, 1, Christian Gossens a, *, Jennifer S. Graves b, c, Xavier Montalban d, e, Michael Lindemann a, f a F. Hoffmann-La Roche Ltd, Basel, 4070, Switzerland b Department of Neurosciences, University of California, San Diego, San Diego, CA, 92093, USA c University of California, San Francisco, San Francisco, CA, 94143, USA d Division of Neurology, University of Toronto, Toronto, ON, M5S 1A1, Canada e Department of Neurology-Neuroimmunology, Multiple Sclerosis Centre of Catalonia (CEMCAT), Vall d’Hebron University Hospital, Barcelona, 08035, Spain f Department of Economics, Baden-Wuerttemberg Cooperative State University, Loerrach, 79539, Germany A R T I C L E I N F O A B S T R A C T Keywords: Background: People living with multiple sclerosis (MS) experience impairments in gait and mobility, that are not Consumer wearable technology fully captured with manually timed walking tests or rating scales administered during periodic clinical visits. We Digital health have developed a smartphone-based assessment of ambulation performance, the 5 U-Turn Test (5UTT), a Gait and balance impairment quantitative self-administered test of U-turn ability while walking, for people with MS (PwMS). Multiple sclerosis Turning while walking Research question: What is the test-retest reliability and concurrent validity of U-turn speed, an unsupervised self- Turn speed assessment of gait and balance impairment, measured using a body-worn smartphone during the 5UTT? Smartphone Methods: 76 PwMS and 25 healthy controls (HCs) participated in a cross-sectional non-randomised interventional Accelerometer feasibility study. The 5UTT was self-administered daily and the median U-turn speed, measured during a 14-day Gyroscope session, was compared against existing validated in-clinic measures of MS-related disability. Results: U-turn speed, measured during a 14-day session from the 5UTT, demonstrated good-to-excellent test- retest reliability in PwMS alone and combined with HCs (intraclass correlation coefficient [ICC] = 0.87 [95 % CI: 0.80–0.92]) and moderate-to-excellent reliability in HCs alone (ICC = 0.88 [95 % CI: 0.69–0.96]). U-turn speed was significantly correlated with in-clinic measures of walking speed, physical fatigue, ambulation impairment, overall MS-related disability and patients’ self-perception of quality of life, at baseline, Week 12 and Week 24. The minimal detectable change of the U-turn speed from the 5UTT was low (19.42 %) in PwMS and indicates a good precision of this measurement tool when compared with conventional in-clinic measures of walking performance. Significance: The frequent self-assessment of turn speed, as an outcome measure from a smartphone-based U-turn test, may represent an ecologically valid digital solution to remotely and reliably monitor gait and balance impairment in a home environment during MS clinical trials and practice. 1. Introduction cognitive disturbances [1], may contribute to gait and balance impair­ ment. People with MS (PwMS) experience a higher incidence of falls [2], Symptoms of multiple sclerosis (MS), including impaired lower- with over 50 % reporting a fall in a 3–6-month period [1]. Most falls take extremity muscle strength, sensory, cerebellar, vestibular and place at home while standing, walking or turning [2]. To turn while * Correspondence author at: F. Hoffmann-La Roche Ltd, 124 Grenzacherstrasse, Basel, CH 4070, Switzerland. E-mail addresses: wei-yi.cheng@roche.com (W.-Y. Cheng), alan.bourke@roche.com (A.K. Bourke), florian.lipsmeier@roche.com (F. Lipsmeier), corrado. bernasconi@roche.com (C. Bernasconi), belachew@me.com (S. Belachew), christian.gossens@roche.com (C. Gossens), jgraves@ucsd.edu (J.S. Graves), xavier. montalban@cem-cat.org (X. Montalban), michael.lindemann@roche.com (M. Lindemann). 1 During completion of the work related to this manuscript, S. Belachew was an employee of F. Hoffmann-La Roche Ltd; his current affiliation is Biogen, Cambridge, MA, USA. https://doi.org/10.1016/j.gaitpost.2020.11.025 Received 29 October 2019; Received in revised form 13 November 2020; Accepted 22 November 2020 Available online 25 November 2020 0966-6362/© 2020 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). W.-Y. Cheng et al. Gait & Posture 84 (2021) 120–126 walking, the body must reduce its forward momentum, rotate and then participating in the study protocol. Further details regarding the inclu­ accelerate in the new direction [3]. The ability to adapt gait and balance, sion/exclusion criteria were previously published [14]. to perform a turn or change in direction safely, is thus an essential part of The trial protocol was approved by the institutional review board/ independent living [4]. Studies suggest that the neural systems involved ethics committee at each participating site. These include the Multiple in turning may be more vulnerable to impairments than during forward Sclerosis Centre of Catalonia (CEMCAT), Vall d’Hebron University walking [5]. Characteristics of turning while walking have thus been Hospital, Barcelona, Spain; and the University of California, San Fran­ shown to be important predictors of balance confidence, gait impair­ cisco (UCSF), San Francisco, California, USA. The study protocol con­ ment and falls in PwMS [6]. formed to Good Clinical Practice guidelines and the principles outlined Existing best practice methods of assessing ambulation in PwMS in the Declaration of Helsinki. The study is registered at ClinicalTrials. primarily involve clinical observations using trained clinical staff, with gov (NCT02952911) [14]. Written informed consent was obtained from manual stopwatch-timed walking-based assessments (e.g., Timed Up all participants or participants’ legally authorised representatives prior and Go [TUG], Timed 25-Foot Walk [T25FW], Two-Minute Walk Test to assessment. [2MWT] and Six-Minute Walk Test [6MWT]) [7]. However, such in-clinic assessments are administered infrequently and can be 2.2. Protocol time-consuming, thus limiting their utility in some clinical practice settings and are prone to rater-dependent error. Furthermore, the Eligible PwMS and HCs were assessed clinically at three time points biomechanical aspects of participants’ mobility function remain over the 24-week study duration. They were provided with a smart­ uncaptured. phone at enrolment and asked to perform the 5UTT once daily (Fig. 1). Recently, body-worn inertial sensor-based systems have been used to assess physical function in PwMS in laboratory and clinical settings, 2.3. The 5 U-Turn Test showing differences in mobility between PwMS and healthy controls (HCs) [6,8–11]. While performing a modified TUG test, PwMS had a The 5UTT is a self-administered test, where participants walk at a longer U-turn duration [8]. A follow-up longitudinal study, conducted self-selected comfortable pace and perform five consecutive 180◦ turns over 18 months, demonstrated that turning duration was important but (U-turns) at least 4 m apart within 60 s (Fig. 2a-c). not significant in separating HCs versus patients with mild and moderate MS disability [11]. The magnitude of turns measured during the TUG 2.4. The in-clinic assessments test was also a relevant feature in a regression model used to distinguish PwMS from HCs [9]. The inclusion of turn parameters collected during At the three in-clinic visits, all participants performed both the the TUG and 6MWT improved modelling of patient-reported balance T25FW and the 5UTT, and were assessed using the Fatigue Scale for confidence and walking limitations in PwMS [6]. A longitudinal study Motor and Cognitive Functions (FSMC) [16]. In addition, PwMS were conducted by Chitnis et al. [10] reported significant moderate-to-good assessed at each scheduled visit with the following disability rating correlations between turn speed during the TUG and 2MWT, and con­ scales and patient-reported outcomes: EDSS [17], Multiple Sclerosis ventional in-clinic MS disability metrics, including the Expanded Impact Scale-29 (MSIS-29; version 2) [18] and Patient Determined Disability Status Scale (EDSS) and T25FW. Disease Steps (PDDS) (Fig. 1) [19]. Advances in inertial-sensor technology, combined with the wide­ The T25FW test score was calculated as the average time of two spread adoption of smartphones, have made the ubiquitous monitoring successive timed walking bouts over 25 feet (7.62 m) and has demon­ of human body movement, using body-worn inertial sensors, in a free- strated reliability, validity and clinical meaningfulness in MS [20]. living situation more feasible. Inertial sensor-based smartphones and FSMC is a validated 20-item scale developed for the assessment of body-worn sensors have facilitated the instrumentation of existing MS-related motor and cognitive fatigue [16]. Here we examined the physical functional tests [12] and the harvesting of rare real-world 10-item subscale related to motor fatigue, specifically items 2–3, 5–6, events (e.g., falls) [13]. In the study “Monitoring of Multiple Sclerosis 9–10, 12, 14, 16 and 19 (FSMC motor subscale, range 10 [no fatigue]–50 (MS) Participants With the Use of Digital Technology (Smartphones and [severe fatigue]). Smartwatches) - A Feasibility Study” (NCT02952911) [14], we introduced MSIS-29 is a 29-item MS-specific questionnaire to assess patient- a daily self-administered U-turn test that can be performed indepen­ reported quality of life [18]. Scores were summed to form a physical dently at home using a smartphone. The 5 U-Turn Test (5UTT) aims to subscale (20 items, range 20–80) and psychological impact subscale (9 examine a person’s ability to perform five 180◦ turns (U-turns) while items, range 9–36), with higher values indicating lower perceived walking at a comfortable pace. quality of life. The physical MSIS-29 subscale was used (items 1–20), This study assessed the test-retest reliability and concurrent validity along with the ambulation summed items, which include items 4 and 5. of the 5UTT to assess gait and balance impairment, a surrogate for dis­ All scales and summed items were transformed to a 0–100 scale. ease state, in PwMS in both supervised and unsupervised settings, using PDDS is a validated ordinal scale that ranks MS severity from inertial-sensor data from a preconfigured smartphone. 0 (normal) to 8 (bedridden) [21] according to patients’ perceptions, primarily based on physical motor function and gait impairment. 2. Materials and methods 2.5. Equipment This cross-sectional analysis examined the longitudinal intra-session test-retest reliability and concurrent validity of the 5UTT, a self- Participants were provided with a Samsung Galaxy S7 smartphone. administered functional assessment of gait and balance impairment in The inertial-sensor embedded in the smartphone contained tri-axial PwMS. accelerometer and gyroscope sensors and sampled using a variable sampling rate of approximately 50 Hz. The sensor data were encrypted 2.1. Participants before wireless transmission to a secure server for subsequent data analysis [22]. In total, 76 PwMS and 25 HCs were enrolled in this study. PwMS were included if they had a diagnosis of MS (2010 revised McDonald 2.6. Sensor data analysis criteria [15]; treated or untreated) and an EDSS score of 0–5.5 (inclu­ sive). A maximum EDSS score of 5.5 was used to ensure any patient with Data analysis was performed in Python 2.7 and 3.6. The inertial- relapsing or progressive MS would not have any significant difficulty in sensor data were first resampled, using linear interpolation, to a fixed 121 W.-Y. Cheng et al. Gait & Posture 84 (2021) 120–126 Fig. 1. Study protocol. PwMS and HCs were assessed clinically at baseline (enrolment), Week 12 and study termination/early discontinuation (Week 24). At the enrolment visit, PwMS and HCs were provided with a preconfigured smartphone. Participants were requested to attach the smartphone and perform the 5UTT once daily, at approximately the same time, during the entire 24-week study duration [15]. 5UTT = 5 U-Turn Test; HC = healthy control; PRO = patient-reported outcome; PwMS, people with multiple sclerosis. Fig. 2. The 5UTT is a self-administered test, which can be activated by the participant through the touchscreen interface. The test is recorded using the same smartphone attached to the body. After answering safety questions, reading the instructions (a) and attaching the smartphone, participants were requested to walk at a self-selected comfortable pace and perform five consecutive 180◦ turns (U-turns) at least 4 m apart within 60 s (a-b). Audio and vibration cues indicate the start and the end of the test. Feedback on how participants perform the test is not provided. A 4m distance was chosen to allow a clear separation between U-turns while also reducing the risk of dizziness and fatigue. The test was performed on a flat level surface, either indoors or outdoors. Participants were allowed to wear regular footwear and an assistive and/or orthotic device as needed. The smartphone was carried at the front of the waist in a belt bag or alternatively in a trouser pocket (c). (a) Instruction screens presented to the participant prior to completing the 5UTT. (b) Participants are instructed to walk safely and perform five successive U-turns going back and forth between two points at least 4 m apart, safely and at a normal comfortable walking speed within 60 s. (c) The attachment methods. The smartphone is either worn at the front of the waist, in a waist-worn belt bag (provided to the participant) or in a trouser pocket. 5UTT = 5 U-Turn Test. sampling rate of 50 Hz. The vertical axis of the sensor coordinate frame, was assessed using the second (baseline session excluded to allow attached to the body, was aligned with gravity via matrix rotation [23], familiarisation) and subsequent complete 14-day session per partici­ and the yaw gyroscope signal was integrated to calculate the turn angle. pant. A two-way analysis of variance using a mixed-effects model with Individual turns were included if they exceeded 90◦ . The turn speed was absolute agreement (intraclass correlation coefficient [ICC]; ICC [2,1]) calculated as the turn angle divided by the duration of the turn event was chosen [24]. [23]. Using the algorithm by El-Gohary et al. [5], the beginning and end The standard error of measurement (SEM) provides an absolute of each turn was identified using thresholds of ±5◦ /s on the angular index of reliability and is used to quantify the precision of the mea­ velocity profile. If turns were less than 0.05 s apart and in the same di­ surement instruments [25]. The SEM was calculated, where SD is the rection, they were considered part of the same turn. The median turn standard deviation of all turn speed values as: speed of the recorded turns was taken for each test, with a minimum of √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ one turn required per test. SEM = SD. 1 − ICC The minimal detectable change, with a 95 % confidence interval (CI) 2.7. Statistical analysis (MDC95), was calculated to determine how much measured change is likely to reflect true change [25]. The MDC95 [4] was calculated as: A session consists of a 14-day non-overlapping period, chosen to √̅̅̅ reduce intra-day, day-to-day/weekend-weekday fluctuations and vari­ MDC95 = z − score (95% CI)SEM. 2 ations in the daily test schedule/location, which may be considered The SEM and MDC95 provide a longitudinal assessment of the irrelevant to the chronic temporal pattern of MS disability progression. sensitivity to change in signal versus noise from repeated tests. A minimum of three tests performed during this 14-day period was The SEM and MDC95 can be expressed as percentages that are in­ required to register a valid session. The median turn speed during each dependent of the units of measurement, calculated as: valid session was compiled for each participant. Test-retest reliability 122 W.-Y. Cheng et al. Gait & Posture 84 (2021) 120–126 SEM Table 2 SEM% = 100. x Test-retest reliability for the turn speed (◦ /s) of the 5UTT for PwMS and HCs. Category ICC (95 % CI) SEM SEM MDC95 MDC95% MDC95 % MDC95 % = 100. x 14-day session This allows comparison of the amount of random error between median measurements using the above two equations, where x is the mean for all PwMS & HC 0.87 5.24 6.60 14.51 18.29 observations [26]. (0.81–0.91) PwMS 0.87 5.54 7.00 15.36 19.42 To examine concurrent validity, the median turn speed during 14- (0.80–0.92) day sessions from 7 days before to 7 days after each of the three clin­ HC 0.88 3.75 4.68 10.40 12.97 ical visits (baseline, Week 12 and Week 24), and during all three visits (0.69–0.96) (‘All visits’), was correlated against the in-clinic disability outcome ICC(2,1) calculated for the 14-day session median. Reliability is indicated using measures and patient-reported quality-of-life and fatigue scales (EDSS, the ICC values classified as poor (ICC < 0.5), moderate (ICC = 0.5–0.75), good T25FW, PDDS, MSIS-29 ambulation summed items, MSIS-29 physical (ICC = 0.75–0.9) and excellent (ICC > 0.9) [24]. SEM and MDC95 can be subscale and FSMC motor subscale) using non-parametric Spearman’s expressed as percentages that are independent of the units of measurement. correlation. The two-sided p-value, for significant correlation, was 5UTT = 5 U-Turn Test. CI = confidence interval. HC = healthy control. adjusted using Holm–Bonferroni correction applied across all primary ICC = intraclass correlation coefficient. MDC95 = minimal detectable change outcome measures, with statistical significance set at p < 0.05. with a 95 % CI. PwMS = people with multiple sclerosis. SEM = standard error of measurement. 3. Results 3.2. Concurrent validity A total of 9628 tests were recorded, with an average of 90.46 (SD: 56.01) tests recorded per participant. Average 5UTT turn speed was Median turn speed was correlated against in-clinic disability 77.91◦ /s ± 3.27◦ /s (mean ± 95 % CI, n = 76) for PwMS and 80.63◦ /s ± outcome measures and patient-reported quality-of-life and fatigue scales 3.72◦ /s (mean ± 95 % CI, n = 25) for HCs. Participants’ baseline de­ using Spearman’s rho correlation. All correlations were statistically mographics are described in Table 1. significant at all study time points (baseline, Week 12 and Week 24) individually and collectively, and varied between fair and moderate-to- 3.1. Test-retest reliability good (r=–0.331 to –0.603, all p-values from p < 0.001 to p < 0.025). Examining across all three time points collectively (‘All visits’), median Test-retest reliability data were available for 85 participants (67 turn speed produced significant moderate-to-good negative PwMS, 18 HCs). ICC values demonstrated good-to-excellent reliability correlations with the T25FW and MSIS-29 ambulation summed items for the PwMS alone (ICC [2,1] [95 % CI]: 0.87 [0.80–0.92]) and com­ (r=–0.506 to –0.516, p < 0.001). The strength of correlations with the bined with HCs (ICC [2,1] [95 % CI]: 0.87 [0.81–0.91]). Moderate-to- EDSS, PDDS, MSIS-29 physical summed items and the FSMC motor excellent reliability was demonstrated for the HCs (ICC [2,1] [95 % subscale was fair, approaching moderate-to-good (r=–0.452 to –0.498, CI]: 0.88 [0.69–0.96]) (Table 2). Low MDC95 (10.40◦ /s–15.36◦ /s) and p < 0.001). The details of the observed correlations are summarised in MDC95% (12.97 %–19.42 %) values were found along with low SEM Table 3 and Fig. 3. (3.75◦ /s–5.54◦ /s) and SEM% (4.68 %–7.00 %) values. 4. Discussion This study provides evidence of the reliability and concurrent val­ Table 1 idity for PwMS to self-assess gait and balance impairment through a Cohort demographics and characteristics for PwMS and HCs at baseline. novel smartphone sensor-based outcome measure of turn speed, the Parameter PwMS HCs p-value 5UTT. (N = 76) (N = 25) A recent study demonstrated that turning is an important marker of Age, mean ± SD, years 39.5 ± 7.9 34.9 ± 9.3 0.043 balance confidence and walking limitation in PwMS. They promote the Female, n (%) 53 (69.7) 7 (28.0) NA inclusion of turning in tests of longer durations with multiple turns, as MS diagnosis (PPMS, SPMS, RRMS), % 3.9, 5.3, 90.8 NA NA performed in the 5UTT, which may be used to model longitudinal Time since MS symptom onset, 11.3 ± 7.0 NA NA change more accurately [6]. mean ± SD, years EDSS, mean ± SD 2.4 ± 1.4 NA NA To the best of our knowledge, this is the first study to demonstrate 9HPT, mean ± SD, seconds excellent test-retest reliability for turn-speed, derived from a Dominant hand 22.1 ± 4.6 18.9 ± 2.1 <0.001 smartphone-based self-assessment of gait and balance impairment, in an Non-dominant hand 22.8 ± 4.9 19.5 ± 2.0 <0.001 MS population using the embedded inertial sensors. Attaching the T25FW, mean ± SD, seconds 6.0 ± 2.1 5.0 ± 1.0 0.19 BBS, mean ± SD 52.5 ± 5.7 56.0 ± 0 <0.001 smartphone using a belt bag at the waist, or in a trouser pocket, allows PDDS, mean ± SD 1.5 ± 1.6 NA NA for the approximation of the body’s centre-of-mass movement and FSMC total score, mean ± SD 59.1 ± 22.7 25.5 ± 6.0 <0.001 therefore a better indication of general body movement. It is anticipated PHQ-9, mean ± SD 8.3 ± 6.1 2.4 ± 2.9 <0.001 that these convenient sensor attachment options will ensure adherence MSIS-29 (version 2), mean ± SD for future 5UTT implementations. Physical subscale 26.2 ± 24.4 NA NA Psychological subscale 36.4 ± 26.5 NA NA An analysis of the ICC results demonstrated excellent intra-session test-retest reliability in both PwMS and HCs. In addition to demon­ 9HPT = 9-Hole Peg Test. BBS = Berg Balance Scale. EDSS = Expanded Disability strating acceptable ICC values, a measurement tool should exhibit small Status Scale. FSMC = Fatigue Scale for Motor and Cognitive Functions. measurement error and be capable of identifying real changes within a HC = healthy control. MS = multiple sclerosis. MSIS-29 = Multiple Sclerosis Impact Scale-29. NA = not applicable. PDDS = Patient Determined Disease group and in individuals [27]. The absolute reliability analysis per­ Steps. PHQ-9 = Patient Health Questionnaire-9. PPMS = primary progressive formed in this study corroborates the good relative reliability of the multiple sclerosis. PwMS = people with multiple sclerosis. RRMS = relapsing- 5UTT from the 14-day session median. Specifically, the low SEM and remitting multiple sclerosis. SD = standard deviation. SPMS = secondary pro­ SEM% values together with low MDC95 and MDC95% values indicate gressive multiple sclerosis. T25FW = Timed 25-Foot Walk. good precision of the measurement tool. In particular, the MDC95% 123 W.-Y. Cheng et al. Gait & Posture 84 (2021) 120–126 Table 3 value was 19.42 % for PwMS, which compares favourably with values of Spearman’s correlations between the turn speed (◦ /s) of the 5UTT and in-clinic 36 % for the in-clinic T25FW [28]. More precisely, for 95 % of PwMS, disability and patient-reported outcome measures in PwMS at baseline, Week differences in repeated session measurements of ≥15.36◦ /s would 12, Week 24 and for all three visits (‘All visits’). reflect a real difference, reflective of a change in disease status. In-clinic outcome measures r p-value* n The concurrent validity [27] of turn speed measured during the T25FW self-administered 5UTT was demonstrated against conventional in-clinic All visits –0.506 <0.001 70 MS disability outcome measures (EDSS and T25FW), patients’ percep­ Baseline –0.525 <0.001 53 tion of their disability (PDDS), physical impact of MS (physical subscale Week 12 –0.501 <0.001 55 and ambulation summed items of the MSIS-29) and physical fatigue Week 24/early discontinuation 0.008 50 (motor subscale of the FSMC). –0.416 EDSS All visits –0.452 <0.001 70 Our results compare favourably with a recent longitudinal study Baseline –0.529 <0.001 54 conducted by Chitnis et al. [10]. They reported fair and Week 12 –0.546 <0.001 57 moderate-to-good significant correlations for turn speed measured at the Week 24/early discontinuation 0.025 51 –0.347 chest, with the EDSS (r=–0.563 to –0.588, p < 0.05, for maximum, mean PDDS All visits –0.498 <0.001 70 and standard deviation) and with the T25FW (r = 0.476, p < 0.05, for Baseline –0.582 <0.001 54 standard deviation and r = 0.552 for maximum, p < 0.05). Similarly, in Week 12 –0.591 <0.001 57 this study, we observed fair and moderate-to-good significant correla­ Week 24/early discontinuation –0.463 0.003 51 tions with the EDSS (r=–0.347 to –0.546, p < 0.05) and with the T25FW MSIS-29 ambulation summed items (r=–0.416 to –0.525, p < 0.05). However, Chitnis et al. [10] analysed All visits –0.516 <0.001 70 Baseline –0.576 <0.001 54 turns recorded during clinically based assessments in a smaller cohort of Week 12 –0.529 <0.001 57 PwMS (n = 23). In our study, by comparison, we analysed turns recorded Week 24/early discontinuation –0.490 0.002 51 during a novel self-administered U-turn test, the 5UTT, performed MSIS-29 physical subscale independently at home using a smartphone, in a larger cohort of PwMS All visits –0.470 <0.001 70 Baseline –0.603 <0.001 54 (n = 76). Week 12 –0.556 <0.001 57 The 5UTT is a quick and convenient smartphone-based test that can Week 24/early discontinuation –0.331 0.018 51 be incorporated into the daily routine of PwMS, as it does not require a FSMC motor subscale laboratory setting, external timing/camera-based measurement equip­ All visits 70 ment or trained clinical/technical staff. Instead, it can be performed –0.496 <0.001 Baseline –0.564 <0.001 53 Week 12 –0.537 <0.001 56 remotely on an even walking surface, either indoors at home or out­ Week 24/early discontinuation –0.416 0.010 51 doors. In addition, the test enables patients to self-assess gait and bal­ ance impairment at a comfortable pace and therefore does not increase 5UTT = 5 U-Turn Test. EDSS = Expanded Disability Status Scale. FSMC = Fa­ tigue Scale for Motor and Cognitive Functions. MSIS-29 = Multiple Sclerosis the risk of falling. Consequently, this self-administered test allows cli­ Impact Scale-29. n = number of participants. PDDS = Patient Determined Dis­ nicians to perform continuous longitudinal monitoring of MS disease ease Steps. PwMS = people with multiple sclerosis. r = Spearman’s correlation. state. T25FW = Timed 25-Foot Walk. * Significant correlation after Holm–Bonferroni adjusted p-values with an 4.1. Limitations and future work alpha value of 0.05 for 24 statistical significance tests. The strengths of the correlations were classified as good-to-excellent (r>0.75), moderate-to-good (r = 0.5–0.75), fair (r = 0.25–0.49) or no correlation (r<0.25) [25]. PwMS from this study encompassed mostly patients with relapsing forms of MS, with mild levels of overall disability and minimal Fig. 3. A graphical representation of the Spearman’s correlations between the turn speed and in-clinic disability outcome measures in PwMS. Spearman’s corre­ lations between the turn speed and EDSS and T25FW at baseline, Week 12 and Week 24 (or early discontinuation). EDSS = Expanded Disability Status Scale. n = number of participants. PwMS = people with multiple sclerosis. r = Spearman’s correlation. s = seconds. T25FW = Timed 25-Foot Walk. 124 W.-Y. Cheng et al. Gait & Posture 84 (2021) 120–126 ambulatory impairment. Correlations observed in this preliminary editing. Shibeshih Belachew: Conceptualization, Methodology, Inves­ work, between 5UTT turn speed and all in-clinic disability and patient- tigation, Writing - review & editing, Supervision. Christian Gossens: reported outcome measures, need to be further substantiated across the Conceptualization, Resources, Writing - review & editing, Supervision, complete spectrum of MS clinical phenotypes. The longitudinal sensi­ Funding acquisition. Jennifer S. Graves: Conceptualization, Method­ tivity to change of turn speed and its responsiveness to treatment ology, Investigation, Resources, Writing - review & editing. Xavier intervention, will be investigated in the ongoing Phase III CONSO­ Montalban: Conceptualization, Validation, Investigation, Supervision, NANCE trial (NCT03523858). Finally, by construction, the 5UTT turn Writing - review & editing. Michael Lindemann: Conceptualization, speed assesses only the speed of turning while walking at a self-selected Methodology, Writing - review & editing, Supervision. comfortable pace (due to safety concerns) and may not fully capture aspects of gait ataxia that could be better reflected by other combina­ tions of 5UTT sensor-based features. Declaration of Competing Interest In addition to motor skills, participant-specific confounding factors, such as motivation, may impact test performance [29], which may The authors declared the following potential conflicts of interest with explain the decrease in correlations observed at Week 24. respect to the research, authorship and/or publication of this article: W. The work presented here formed part of a pilot study, which aimed to Y.C. is an employee of F. Hoffmann-La Roche Ltd. A.K.B. is an employee evaluate the feasibility of conducting remote patient monitoring, using of F. Hoffmann-La Roche Ltd. F.L. is an employee of F. Hoffmann-La digital technology in PwMS [14]. As the primary aims of this study were Roche Ltd. C.B. is a contractor for F. Hoffmann-La Roche Ltd. S.B. was considered exploratory in nature, a duration of 24 weeks was deemed an employee of F. Hoffmann-La Roche Ltd during the completion of the adequate to assess the feasibility of performing smartphone- and work related to this manuscript. S.B. is now an employee of Biogen smartwatch-based assessments. It is also sufficient to assess the (Cambridge, MA, USA), which was not in any way associated with this test-retest reliability of the 5UTT and its correlation with in-clinic study. C.G. is an employee and shareholder of F. Hoffmann-La Roche disability and patient-reported outcome measures at three clinical Ltd. J.S.G. has received grants or research support from Biogen, Gen­ time points. entech, Inc. and S3 Group, and has received compensation for a non­ We noted significant inter-patient differences in turn speed in PwMS, branded resident and fellow education seminar supported by Biogen. X. who otherwise had similar performance-based disability scores, such as M. has received speaker honoraria and travel expense reimbursement the ambulation speed captured by the T25FW (Fig. 3). This suggests that for participation in scientific meetings, and has been a steering com­ sensor-grade measures of turn speed, recorded in a patient’s own envi­ mittee member of clinical trials or served on advisory boards of clinical ronment, may capture valuable incremental dimensions in character­ trials for Actelion, Biogen, Celgene, Merck, Novartis, Oryzon, Roche, ising MS-related disability. A similar phenomenon, of higher separation Sanofi Genzyme and Teva Pharmaceutical. M.L. is a consultant for F. of performance distribution in inertial sensor-based measures of phys­ Hoffmann-La Roche Ltd via Inovigate. ical function, within EDSS subgroups, was also observed by Block et al. [30] when measuring daily step count in PwMS. Turn speed thus war­ Acknowledgements rants further evaluation as an outcome measure in longitudinal MS clinical trials and could provide an assessment of gait and balance We would like to thank all patients, their families and the in­ impairment at higher temporal resolution, compared to the relatively vestigators who participated in this trial. We would also like to thank the infrequent periodic snapshots of function, currently captured by following employees from F. Hoffmann-La Roche Ltd who supported and in-clinic assessments. contributed to the study: Atieh Bamdadian, Alessandro Barbato, Jan Future work will further establish the qualitative and quantitative Beckmann, Sandro Fritz, Nicholas Pierce Heinemeier, Sven Holm, validity of the 5UTT for assessing gait and balance impairment using Timothy Kilchenmann, Lito Kriara, Bernd Laub, Grégoire Pointeau, data from CONSONANCE and explore the application of this algorithm Caroline Polakowska, Marcin Puhacz, Cedric Simillion, Jens Schjodt- to turns performed during free-living activity. The longitudinal analysis Eriksen, Alf Scotland, Jöerg Sprengel, Ralf Stubner and Krzysztof Try­ of turn speed recorded during walking in a free-living situation, may bus. The authors disclosed receipt of the following financial support for expose potential subtle insidious changes in MS state that may occur the research, authorship and/or publication of this article: F. Hoffmann- early in disease onset. Two types of turning strategy while walking, La Roche Ltd, Basel, Switzerland provided financial support for the study namely the spin turn and the step turn [3], are typically described. and publication of this manuscript. Writing and editorial assistance for Further biomechanical investigation is required to assess which turn this manuscript was provided by Heather Latimer, PhD of Articulate strategy is adopted by PwMS, and what is the relationship to speed and Science, UK, and funded by F. Hoffmann-La Roche Ltd. smoothness of turns in the 5UTT and during free-living gait. The algo­ Qualified researchers may request access to individual patient level rithm for turn speed measurement, implemented here, was validated in data through the clinical study data request platform (https://vivli.org/). a Parkinson’s disease cohort by El-Gohary et al. [5] using a motion Further details on Roche’s criteria for eligible studies are available here capture system. Future work will determine the accuracy of this algo­ (https://vivli.org/members/ourmembers/). For further details on Roche’s rithm, using ground truth data, harvested from an MS cohort. Global Policy on the Sharing of Clinical Information and how to request In conclusion, turn speed collected from a smartphone-based self- access to related clinical study documents, see here (https://www.roche. assessment in PwMS, demonstrated excellent intra-session test-retest com/research_and_development/who_we_are_how_we_work/clinical_tr reliability and correlates significantly with all in-clinic MS disability and ials/our_commitment_to_data_sharing.htm). patient-reported outcome measures. The 5UTT is reliable and showed concurrent validity to assess gait and balance impairment in PwMS. References CRediT authorship contribution statement [1] J.J. Sosnoff, R.W. Motl, S. Morrison, Multiple sclerosis and falls—an evolving tale, Eur. Neurol. Rev. 9 (2014) 4. [2] R. Mazumder, C. Murchison, D. Bourdette, M. Cameron, Falls in people with Wei-Yi Cheng: Methodology, Software, Validation. Alan K. Bourke: multiple sclerosis compared with falls in healthy controls, PLoS One 9 (2014) Conceptualization, Methodology, Software, Validation, Formal analysis, e107620. https://www.ncbi.nlm.nih.gov/pubmed/25254633. Resources, Data curation, Writing - original draft, Writing - review & [3] K. Hase, R.B. Stein, Turning strategies during human walking, J. Neurophysiol. 81 editing, Visualization, Project administration. Florian Lipsmeier: (1999) 2914–2922. https://www.ncbi.nlm.nih.gov/pubmed/10368408. [4] R.L. Robinson, S.S.M. Ng, The timed 180 degrees turn test for assessing people with Conceptualization, Methodology, Writing - review & editing. Corrado hemiplegia from chronic stroke, Biomed Res. Int. 2018 (2018) 9629230. https Bernasconi: Conceptualization, Formal analysis, Writing - review & ://www.ncbi.nlm.nih.gov/pubmed/29568774. 125 W.-Y. Cheng et al. Gait & Posture 84 (2021) 120–126 [5] M. El-Gohary, S. Pearson, J. McNames, M. Mancini, F. Horak, S. Mellone, et al., [18] J. Hobart, S. Cano, Improving the evaluation of therapeutic interventions in Continuous monitoring of turning in patients with movement disability, Sensors multiple sclerosis: the role of new psychometric methods, Health Technol. Assess (Basel) 14 (2013) 356–369. https://www.ncbi.nlm.nih.gov/pubmed/24379043. 13 (iii) (2009) 1–177, ix-x, https://www.ncbi.nlm.nih.gov/pubmed/19216837. [6] G. Adusumilli, S. Lancia, V.A. Levasseur, V. Amblee, M. Orchard, J.M. Wagner, et [19] R.A. Marrie, M. Goldman, Validity of performance scales for disability assessment al., Turning is an important marker of balance confidence and walking limitation in in multiple sclerosis, Mult. Scler. 13 (2007) 1176–1182. https://www.ncbi.nlm. persons with multiple sclerosis, PLoS One 13 (2018) e0198178. https://www.ncbi. nih.gov/pubmed/17623733. nlm.nih.gov/pubmed/29879144. [20] R.W. Motl, J.A. Cohen, R. Benedict, G. Phillips, N. LaRocca, L.D. Hudson, et al., [7] S.E. Bennett, L.E. Bromley, N.M. Fisher, M.R. Tomita, P. Niewczyk, Validity and Validity of the timed 25-foot walk as an ambulatory performance outcome measure reliability of four clinical gait measures in patients with multiple sclerosis, Int. J. for multiple sclerosis, Mult. Scler. 23 (2017) 704–710. https://www.ncbi.nlm.nih. MS Care 19 (2017) 247–252. https://www.ncbi.nlm.nih.gov/pubmed/29070965. gov/pubmed/28206828. [8] R.I. Spain, R.J. St George, A. Salarian, M. Mancini, J.M. Wagner, F.B. Horak, et al., [21] Y.C. Learmonth, R.W. Motl, B.M. Sandroff, J.H. Pula, D. Cadavid, Validation of Body-worn motion sensors detect balance and gait deficits in people with multiple patient determined disease steps (PDDS) scale scores in persons with multiple sclerosis who have normal walking speed, Gait Posture 35 (2012) 573–578. https sclerosis, BMC Neurol. 13 (2013) 37. https://www.ncbi.nlm.nih.gov/pubmed ://www.ncbi.nlm.nih.gov/pubmed/22277368. /23617555. [9] B.R. Greene, S. Rutledge, I. McGurgan, C. McGuigan, K. O’Connell, B. Caulfield, et [22] A.K. Bourke, A. Scotland, F. Lipsmeier, C. Gossens, M. Lindemann, Gait al., Assessment and classification of early-stage multiple sclerosis with inertial Characteristics Harvested During a Smartphone-Based Self-Administered 2-Minute sensors: comparison against clinical measures of disease state, IEEE J. Biomed. Walk Test in People with Multiple Sclerosis: Test-Retest Reliability and Minimum Health Inform. 19 (2015) 1356–1361. https://www.ncbi.nlm.nih.gov/pubmed Detectable Change, Sensors 20 (2020) 5906. https://www.mdpi.com/1424-822 /26087505. 0/20/20/5906. [10] T. Chitnis, B.I. Glanz, C. Gonzalez, B.C. Healy, T.J. Saraceno, N. Sattarnezhad, et [23] W.-Y. Cheng, F. Lipsmeier, A. Creigh, A. Scotland, T. Kilchenmann, L. Jin, et al., al., Quantifying neurologic disease using biosensor measurements in-clinic and in Large-scale continuous mobility monitoring of Parkinson’s disease patients using free-living settings in multiple sclerosis, NPJ Digit. Med 2 (2019) 123. https smartphones, in: P. Perego, A. Rahmani, N. TaheriNejad (Eds.), Wireless Mobile ://www.ncbi.nlm.nih.gov/pubmed/31840094. Communication and Healthcare, MobiHealth 2017, Lecture Notes of the Institute [11] R.I. Spain, M. Mancini, F.B. Horak, D. Bourdette, Body-worn sensors capture for Computer Sciences, Social Informatics and Telecommunications Engineering, variability, but not decline, of gait and balance measures in multiple sclerosis over vol 247, Springer, Cham, 2018, pp. 12–19. 18 months, Gait Posture 39 (2014) 958–964. https://www.ncbi.nlm.nih.gov/pub [24] T.K. Koo, M.Y. Li, A guideline of selecting and reporting intraclass correlation med/24405749. coefficients for reliability research, J. Chiropr. Med. 15 (2016) 155–163. https [12] S. Mellone, C. Tacconi, L. Chiari, Validity of a smartphone-based instrumented ://www.ncbi.nlm.nih.gov/pubmed/27330520. timed up and go, Gait Posture 36 (2012) 163–165. https://www.ncbi.nlm.nih. [25] L.G. Portney, M.P. Watkins, Foundations of Clinical Research: Applications to gov/pubmed/22421189. Practice, 3rd edition, Pearson/Prentice Hall, Upper Saddle River, New Jersey, NJ, [13] J. Klenk, L. Schwickert, L. Palmerini, S. Mellone, A. Bourke, E.A.F. Ihlen, et al., The USA, 2009. FARSEEING real-world fall repository: a large-scale collaborative database to [26] P.M. Nair, T.G. Hornby, A.L. Behrman, Minimal detectable change for spatial and collect and share sensor signals from real-world falls, Eur. Rev. Aging Phys. Act. 13 temporal measurements of gait after incomplete spinal cord injury, Top. Spinal (2016) 8. https://www.ncbi.nlm.nih.gov/pubmed/27807468. Cord Inj. Rehabil. 18 (2012) 273–281. https://www.ncbi.nlm.nih.gov/pubmed [14] L. Midaglia, P. Mulero, X. Montalban, J. Graves, S.L. Hauser, L. Julian, et al., /23459394. Adherence and satisfaction of smartphone- and smartwatch-based remote active [27] V. Hadzic, E. Ursej, M. Kalc, E. Dervisevic, Reproducibility of shoulder short range testing and passive monitoring in people with multiple sclerosis: nonrandomized of motion in isokinetic and isometric strength testing, JESF 10 (2012) 83–89. interventional feasibility study, J. Med. Internet Res. 21 (2019) e14863. https http://www.sciencedirect.com/science/article/pii/S1728869X12000329. ://www.ncbi.nlm.nih.gov/pubmed/31471961. [28] Y.C. Learmonth, D.D. Dlugonski, L.A. Pilutti, B.M. Sandroff, R.W. Motl, The [15] C.H. Polman, S.C. Reingold, B. Banwell, M. Clanet, J.A. Cohen, M. Filippi, et al., reliability, precision and clinically meaningful change of walking assessments in Diagnostic criteria for multiple sclerosis: 2010 revisions to the McDonald criteria, multiple sclerosis, Mult. Scler. 19 (2013) 1784–1791. https://www.ncbi.nlm.nih. Ann. Neurol. 69 (2011) 292–302. https://www.ncbi.nlm.nih.gov/pubmed gov/pubmed/23587605. /21387374. [29] P. Rabbitt, P. Diggle, F. Holland, L. McInnes, Practice and drop-out effects during a [16] I.K. Penner, C. Raselli, M. Stöcklin, K. Opwis, L. Kappos, P. Calabrese, The Fatigue 17-year longitudinal study of cognitive aging, J. Gerontol. B Psychol. Sci. Soc. Sci. Scale for Motor and Cognitive Functions (FSMC): validation of a new instrument to 59 (2004) 84–97. https://www.ncbi.nlm.nih.gov/pubmed/15014091. assess multiple sclerosis-related fatigue, Mult. Scler. 15 (2009) 1509–1517. https [30] V.J. Block, A. Lizée, E. Crabtree-Hartman, C.J. Bevan, J.S. Graves, R. Bove, et al., ://www.ncbi.nlm.nih.gov/pubmed/19995840. Continuous daily assessment of multiple sclerosis disability using remote step [17] J.F. Kurtzke, Rating neurologic impairment in multiple sclerosis: an expanded count monitoring, J. Neurol. 264 (2017) 316–326. https://www.ncbi.nlm.nih. disability status scale (EDSS), Neurology 33 (1983) 1444–1452. https://www.ncbi. gov/pubmed/27896433. nlm.nih.gov/pubmed/6685237. 126