Wearable device and smartphone data can track ALS disease progression and may serve as novel clinical trial outcome measures

doi:10.21203/rs.3.rs-2248487/v1

Amyotrophic lateral sclerosis (ALS) therapeutic development has largely relied on staff-administered functional rating scales to determine treatment efficacy. We sought to determine if mobile applications (apps) and wearable devices can be used to quantify ALS disease progression through active (surveys) and passive (sensors) data collection. Forty ambulatory adults with ALS were followed for 6-months. The Beiwe app was used to administer self-entry ALS functional rating scale-revised (ALSFRS-RSE) and the Rasch Overall ALS Disability Scale (ROADS) surveys every 2-4 weeks. A wrist-worn activity monitor (ActiGraph's Insight Watch) or an ankle-worn activity monitor (Modus' StepWatch) was used continuously by each participant. Wearable device wear and app survey compliance were adequate. ALSFRS-R highly correlated with ALSFRS-RSE. Several wearable data daily physical activity measures demonstrated statistically significant change over time and associations with ALSFRS-RSE and ROADS. Active and passive digital data collection hold promise for novel ALS trial outcomes development.

Health sciences/Biomarkers

Health sciences/Neurology/Neurological disorders/Motor neuron disease/Amyotrophic lateral sclerosis

Health sciences/Diseases/Neurological disorders/Motor neuron disease/Amyotrophic lateral sclerosis

Health sciences/Medical research/Biomarkers

Amyotrophic lateral sclerosis (ALS) therapeutic development has largely relied on staff-administered functional rating scales to gauge interventional efficacy in large scale clinical trials, with limited success. Alternative methods of assessing progressive functional decline in ALS, such as using mobile applications (apps) to collect patient-reported outcome measures (PROMs) and wearable devices for measurement of physical activity (PA), may be valuable.

Several factors contribute to the difficulty in finding efficacious ALS treatments. Although ALS fits well into the motor neuron disease category, it remains a heterogeneous disorder with respect to onset location, pattern of progression, degree of upper motor neuron involvement, underlying genetic architecture, patient experience, and prognosis.^1,2 It is uncommon in the general population despite being the most common motor neuron disease,³ has a poorly understood pathobiology,⁴ and remains a clinically defined syndrome. Traditional ALS outcome measure limitations may also play a role. Factors outside of ALS itself and germane to all clinical trial research add to the difficulty of understanding whether an intervention works. The need for a substantial research infrastructure, experienced staff, and adequate coordinator support for recruitment and logistical operations are obstacles.⁵ Trial visit frequency is designed factoring in staff availability, participant willingness and ability to travel, and cost. While higher trial outcome sampling frequency is typically advantageous, it usually requires additional resources. All these aspects slow progress towards finding a cure.

The Amyotrophic Lateral Sclerosis Functional Rating Scale-Revised (ALSFRS-R) is a validated, traditionally staff-administered, 12-item, 0–4 points/item scale used to assess function and disease progression over time. It has served as the primary longitudinal ALS clinical trial outcome measure for years;^6–9 however, the scale has limitations. The ALSFRS-R does not meet criteria for unidimensionality,^8,10 it is ordinal meaning point-wise decline is nonuniform with each score decrease, and it may fail to detect decline in people with confirmed ALS (PALS) even over the course of 18-months based on Pooled Resource Open-Access ALS Clinical Trials Database (PRO-ACT) data. Over the course of 6-months (the duration of many clinical trials), 25% of participants on placebo or a non-efficacious intervention did not decline in the PRO-ACT dataset.⁹

Though extensive data have been collected to date using traditional functional rating scales, different assessment approaches may have advantages, especially PROMs. Self-entry functional rating scales can reduce the burden of frequent trial center visits for patients and allow home-monitoring of disease progression.¹¹ A self-entry version of the ALSFRS-R, the ALSFRS-RSE, has been developed and has repeatedly demonstrated excellent correlation with the ALSFRS-R (staff-administered).^11–16 Other scales are being developed using modern test theory techniques and specifically for self-entry. One such example is the Rasch-built Overall ALS Disability Scale (ROADS).^10,17 Prior studies have demonstrated the feasibility of both in-clinic and non-clinic ALS PROM collection.^15,18

Remote, passive data collection offers another avenue for outcome measure refinement without the need for participants’ concentration or effort. Modern digital wearable monitors allow for detailed, objective, and near-continuous motor and other biometric measurements.¹⁹ Furthermore, they permit assessment of novel and previously challenging to quantify human functional domains. Digital biomarkers have been used already in the neuromuscular and neurodegenerative disease sphere, with significant expansion expected.²⁰ Viability of periodic non-clinic device and sensor-based data collection has been demonstrated in measuring heart rate variability, accelerometry-based activity, speech, spirometry, hand-grip strength, and electrical impedance myography.^21–24 Few studies have examined continuous monitoring of PALS using wearable devices. Some have measured 3-day/month samples of PA and heart rate variability using a chest-worn and ECG sensor Faros device.^21,23 Kelly et al. reported four PA level endpoints showing moderate or strong between-patient correlations with ALSFRS-R total score and gross motor domain score.²³ Van Eijk et al.²² employed a hip-worn ActiGraph device for 7 days every 2–3 months for 12 months and reported four accelerometer-based endpoints of PA levels to be strongly associated with the ALSFRS-R.

Mobile devices may also represent a valuable mode of digital biomarker collection. Smartphones are becoming nearly ubiquitous, with 85% of the United States (US) population owning one and over 7 billion smartphones worldwide.^25,26 Smartphone sensors coupled with mobile apps allow for versatile data collection of both active (e.g., surveys, audio recordings) and passive (e.g., global positioning system (GPS), accelerometry) data.²⁷ They have also shown promise collecting these data longitudinally from PALS.²⁸

We evaluated whether an app and wearable devices might serve as important tools for monitoring ALS disease progression through active (app surveys) and passive (wearable device sensors) data collection.

We aimed to (1) compare (smartphone) ALSFRS-RSE with (staff-administered) ALSFRS-R at baseline and longitudinally, (2) quantify the baselines and change over time of wearable daily measures of PA, and (3) quantify the association between those measures and both the ALSFRS-RSE and ROADS total scores.

Participants, enrollment, and compliance

From January through December 2021, 46 participants with ALS were enrolled and followed for 6 months. Of these, 40 met analysis sample criteria. One participant was unable to download the Beiwe app. Participants chose their wearable device and did not favor one over the other (ActiGraph = 20, Modus = 20). Clinicodemographic characteristics were similar between groups (Table 1). There was a higher proportion of male (62.5%), white (87.5%), and non-Hispanic participants (87.5%), overall. Most participants smartphones ran iOS (82.5%). Median baseline ALSFRS-R was 33 (range 11 to 47).

Table 1

Baseline demographic and clinical characteristics.
	ActiGraph	Modus	Combined
	N = 20	N = 20	N = 40
Age
Mean (SD)	62.9 (13.4)	60.6 (10.7)	61.8 (12.0)
Median [min, max]	64 [34, 98]	60 [35, 81]	63 [34, 98]
Sex
Males (%)	12 (60.0%)	13 (65.0%)	25 (62.5%)
Females (%)	8 (40.0%)	7 (35.0%)	15 (37.5%)
Ethnicity
Not Hispanic or Latino (%)	18 (90.0%)	17 (85.0%)	35 (87.5%)
Hispanic or Latino (%)	2 (10.0%)	2 (10.0%)	4 (10.0%)
Unknown / Not Reported (%)	0 (0.0%)	1 (5.0%)	1 (2.5%)
Race
White (%)	17 (85.0%)	18 (90.0%)	35 (87.5%)
More Than One Race (%)	2 (10.0%)	1 (5.0%)	3 (7.5%)
Native Hawaiian or Other Pacific Islander (%)	1 (5.0%)	0 (0.0%)	1 (2.5%)
Other (%)	0 (0.0%)	1 (5.0%)	1 (2.5%)
Phone operating system
iOS (%)	15 (75.0%)	18 (90.0%)	33 (82.5%)
Android (%)	5 (25.0%)	2 (10.0%)	7 (17.5%)
ALSFRS-R
Mean (SD)	31.4 (8.6)	31.4 (7.9)	31.4 (8.1)
Median [min, max]	33 [12, 47]	32 [11, 45]	33 [11, 47]
ALSFRS-R -- baseline staff-administered Amyotrophic Lateral Sclerosis Functional Rating Scale-Revised; n -- number; min -- minimum; max – maximum; SD -- standard deviation

Compliance statistics are summarized in Table 2 and were robust for both groups.

Table 2

Phone survey and wearable device compliance. (Aggregated -- median [min, max])
	ActiGraph	Modus	Combined
ALSFRS-RSE submissions	5 [2, 10]	6 [2, 13]	5 [2, 13]
ROADS submissions	4 [2, 10]	5 [2, 13]	5 [2, 13]
Days in observation period	178 [23, 191]	180 [61, 189]	179 [23, 191]
Valid days in observation period*	158 [21, 191]	136 [16, 183]	146 [16, 191]
Average number of valid hours on a valid day*	21 [15, 24]	12 [10, 17]	16 [10, 24]
ALSFRS-RSE -- smartphone self-entry Amyotrophic Lateral Sclerosis Functional Rating Scale-Revised; ROADS – Rasch-built Overall ALS Disability Scale; valid hour/day -- hour/day passing the conditions outlined in Methods section; * -- different methods used to define valid hour/day across the ActiGraph and Modus groups (see Methods section).

Correlation between ALSFRS-R and ALSFRS-RSE

Smartphone self-administered ALSFRS-RSE was highly correlated with staff-administered ALSFRS-R at baseline, 3-, and 6-months (r ≥ 0.93; Fig. 1). The population mean for ALSFRS-RSE was higher than that of ALSFRS-R at each time point (3.3, 2.2, and 3.7 points, respectively).

Survey baselines and change over time

Mean baseline scores were 31.6 (ALSFRS-R), 34.3 (ALSFRS-RSE) (max 44, lower is worse), and 84.9 (ROADS) (max 146, lower is worse); and monthly rates of decline ([95% CI], p-value) were − 0.37 ([-0.62,-0.11], 0.007), -0.48 ([-0.63,-0.32], 0.001), and − 1.26 ([-1.71,-0.81], < 0.001), respectively (Fig. 2a,b,c; Supplementary Table 1, models 1–3). The ALSFRS-RSE baseline was 2.86 points higher than the ALSFRS-R baseline score (95% CI [2.26,3.47]; p < 0.001), but the monthly change in slope between the scales did not significantly differ (Fig. 2d; Supplementary Table 1, model 4).

Comparing ActiGraph and Modus groups, there was no significant difference in baseline or monthly change in either ALSFRS-RSE or ROADS.

Wearable daily Physical Activity (PA) measures: baseline values and a change over time

Table 3 summarizes the average baseline and monthly change of wearable daily PA measures. Estimates are provided for 13 ActiGraph vendor-provided daily (24-hour period) summary measures (VDMs), 7 ActiGraph investigator-derived daily summary measures (IDMs), and 12 Modus VDMs. The majority of measures demonstrated significant change over time. All measures reflected declining PA. Participants’ individual conditional means and the population mean for each VDM/IDM are depicted (Supplementary Figs. 2–12, first columns). Log total activity counts (LTAC), active-to-sedentary transition probability (ASTP) (ActiGraph), and mean steps/minute (Modus) had very high (≥ 0.85) conditional coefficient of determination (R2c) values.^29,30

Table 3

Average baseline and monthly change in wearable daily PA measures.
Mod.	Device	Data	Daily measure	Baseline est. [95% CI]	Monthly change	R2m	R2c
no.	group	set			est. [95% CI] (p-val.)
1	AG	VDMs	Light activity [minutes]	1283 [1217, 1349]	-24.75 [-48.29, -1.215] (0.040)	0.033	0.459
2	AG	VDMs	Moderate activity [minutes]	51.12 [35.00, 67.25]	-0.581 [-3.949, 2.788] (0.719)	0.000	0.585
3	AG	VDMs	Vigorous activity [minutes]	4.032 [-1.130, 9.194]	-0.069 [-0.363, 0.226] (0.643)	0.000	0.507
4	AG	VDMs	MVPA [minutes]	55.17 [36.28, 74.06]	-0.691 [-4.201, 2.818] (0.682)	0.000	0.647
5	AG	VDMs	Sedentary [minutes]	1058 [987.3, 1130]	-13.06 [-32.25, 6.122] (0.170)	0.009	0.560
6	AG	VDMs	Non-sedentary [minutes]	279.7 [204.5, 354.9]	-13.08 [-22.50, -3.662] (0.010)	0.022	0.819
7	AG	VDMs	Locomotion [minutes]	23.17 [10.23, 36.12]	-0.526 [-2.143, 1.091] (0.502)	0.001	0.772
8	AG	VDMs	Non-locomotion [minutes]	1315 [1250, 1380]	-24.80 [-48.39, -1.219] (0.040)	0.033	0.464
9	AG	VDMs	Steps	2661 [760.1, 4561]	-165.7 [-321.1, -10.21] (0.038)	0.005	0.791
10	AG	VDMs	Calories	2109 [1840, 2377]	-53.95 [-100.8, -7.073] (0.027)	0.023	0.772
11	AG	VDMs	METs	1525 [1354, 1697]	-38.42 [-71.32, -5.525] (0.025)	0.033	0.663
12	AG	VDMs	Total activity counts	1292056 [988272, 1595841]	-56625 [-100107, -13144] (0.014)	0.025	0.798
13	AG	VDMs	Sleep [minutes]	659.5 [538.4, 780.7]	-1.691 [-31.44, 28.06] (0.905)	0.000	0.558
14	AG	IDMs	Total activity counts	1362438 [1046388, 1678489]	-58631 [-102202, -15059] (0.012)	0.025	0.824
15	AG	IDMs	Total log(activity counts)	5456 [4808, 6104]	-127.7 [-236.9, -18.47] (0.026)	0.029	0.707
16	AG	IDMs	log(total activity counts)	13.90 [13.50, 14.31]	-0.046 [-0.078, -0.013] (0.011)	0.009	0.934
17	AG	IDMs	Active to sedentary TP	0.389 [0.299, 0.479]	0.009 [0.003, 0.014] (0.003)	0.005	0.890
18	AG	IDMs	Sedentary to active TP	0.074 [0.057, 0.091]	-0.003 [-0.006, -0.000] (0.028)	0.026	0.723
19	AG	IDMs	Non-sedentary [minutes]	271.4 [200.6, 342.2]	-12.88 [-22.23, -3.530] (0.010)	0.023	0.809
20	AG	IDMs	Sedentary [minutes]	1169 [1098, 1239]	12.88 [3.530, 22.23] (0.010)	0.023	0.809
21	M	VDMs	Steps recorded/minute - total	1871 [1137, 2605]	-57.79 [-89.26, -26.32] (0.001)	0.003	0.836
23	M	VDMs	Steps recorded/minute − 95th perc.	25.92 [21.19, 30.65]	-0.533 [-1.015, -0.052] (0.032)	0.005	0.734
24	M	VDMs	Steps recorded/minute - mean	9.170 [7.050, 11.29]	-0.149 [-0.265, -0.034] (0.015)	0.003	0.858
25	M	VDMs	Steps recorded/minute - median	6.793 [4.566, 9.020]	-0.081 [-0.183, 0.020] (0.110)	0.001	0.845
26	M	VDMs	Time % with 1–15 steps recorded/minute	9.395 [7.334, 11.46]	-0.187 [-0.315, -0.060] (0.006)	0.004	0.731
27	M	VDMs	Time % with 16–40 steps recorded/minute	2.743 [1.451, 4.036]	-0.120 [-0.182, -0.057] (0.001)	0.005	0.810
28	M	VDMs	Time % with 41 + steps recorded/minute	0.293 [0.026, 0.560]	-0.001 [-0.012, 0.010] (0.863)	0.000	0.754
29	M	VDMs	Max 60	8.776 [5.947, 11.61]	-0.256 [-0.502, -0.009] (0.043)	0.003	0.708
30	M	VDMs	Max 20	14.89 [10.60, 19.18]	-0.424 [-0.904, 0.057] (0.081)	0.003	0.669
31	M	VDMs	Max 5	24.55 [18.98, 30.11]	-0.736 [-1.276, -0.195] (0.011)	0.007	0.729
32	M	VDMs	Max 1	35.56 [29.71, 41.41]	-0.785 [-1.374, -0.196] (0.012)	0.007	0.793
33	M	VDMs	Peak performance index	23.21 [17.92, 28.49]	-0.593 [-1.017, -0.170] (0.009)	0.005	0.810
AG -- ActiGraph device; CI -- estimate's confidence interval obtained from LMM estimation; est. -- estimate obtained from LMM estimation; IDMs -- investigator-derived daily measures; M -- Modus device; Mod. no. -- model number: ordering index of a LMM fit, assigned to a particular daily measure; R2c -- LMM conditional coefficient of determination; R2m -- LMM marginal coefficient of determination; PA – physical activity; VDMs -- vendor-provided daily measures

Wearable daily Physical Activity (PA) measures: association with ALSFRS-RSE and ROADS

The mixed-effects models (LMM) associations between each daily measure (VDMs,IDMs) and each survey (ALSFRS-RSE, ROADS) are summarized (Supplementary Tables 1–3) and graphically depicted (Supplementary Figs. 2–12, 2nd and 3rd columns). The daily measures were the covariates in the LMM and were standardized to have means equal to 0 and standard deviations equal to 1. 13/32 daily measures were significantly associated with both ALSFRS-RSE and ROADS: total activity counts (TAC), moderate activity, moderate and vigorous physical activity (MVPA), non-sedentary minutes, locomotion minutes, steps, calories – VDMs, and TAC, LTAC, ASTP, sedentary-to-active transition probability (SATP), sedentary and non-sedentary minutes – IDMs. In the LMMs of wearable PA measures and surveys, higher ALSFRS-RSE and ROADS scores associated with measures reflecting higher PA volume and activity fragmentation (active/inactive transitions).

Sensitivity analysis of wearable device wear protocol

Using all weeks’ (6-months) data, the estimated baseline IDM-TAC were 1,362,438 and the monthly change was − 58,631. Across the 9 differing IDM-TAC data collection frequency scenarios examined using LMM (Supplementary Table 4), results did not vary substantially (≤ 1.5%). The monthly change did not vary substantially (< 4%) for scenarios with 2 weeks of data collection and up to 6 weeks break but did for the others compared to using all weeks’ data. Nonetheless, all scenarios detected significant associations between declining TAC over time.

Principal results

Forty ambulatory adults with ALS were followed for 6 months during which they continuously wore a wearable device (the wrist-worn ActiGraph Insight Watch or the ankle-worn Modus StepWatch) and regularly contributed ALSFRS-RSE and ROADS surveys via the Beiwe smartphone app. Daily measures of physical activity (PA) from the wearables allowed quantification of baseline values, monthly PA volume decline, and other characteristics.

Though ambulatory status was an inclusion criterion, PALS with a range of functional abilities met this criterion (baseline ALSFRS-R ranged from 11 to 47), increasing generalizability. Given that PALS frequently have limb weakness and fatigue affecting activity levels, 8 hours of activity logged per day was chosen as the daily device wear compliance threshold. The number of compliant days and active hours per day differing between the ActiGraph and Modus groups should not be interpreted as meaningfully different because the devices collected data differently, thus methodologies used for valid hour determination differed. Despite the observational nature of the study, older age group, and use of multiple technologies, device wear- and survey submission-compliance were robust in both device groups. Compliance is often quite variable in many digital health studies, both within and between studies: levels range from 25–80%.^31–34 The number of Beiwe app survey submissions was much higher than the number obtained in traditional in-person clinical trials (Table 2).²⁴

Excellent correlation between staff-administered ALSFRS-R and smartphone self-administered ALSFRS-R (≥ 0.93), along with higher (2.86 points) self-report baseline scores, are consistent with prior studies,^11,15,16 reinforcing that valuable functional outcome data can be reliably obtained remotely and that the ALSFRS-R, the mainstay of outcome assessment in ALS trials, can be reliably obtained in this manner.

The significant difference between ALSFRS-R and ALSFRS-RSE at baseline but not over time suggests that self-report and staff-administered scores should be compared separately to avoid introducing unwanted variability, something to be considered during trial design.

In general, all three scales (ALSFRS-R, ALSFRS-RSE, ROADS) declined as expected and performed well in detecting functional decline despite our participants having a milder rate of decline as a group compared to those in interventional trials.^35–38 Observational studies tend to enroll PALS with more slowly progressing disease.^22,24

Notably, neither ActiGraph nor Modus VDMs explicitly account for missing data due to non-wear, charging, etc. For example, ActiGraph's non-sedentary and sedentary minutes add up to 1440 minutes for IDM but do not for VDM. Consequently, PA volume and performance intensity measures may be biased downward. To explore what results might show if accounting for missing data, we created a set of IDMs that imputed missing minute-level data. As expected, the baseline TAC were lower for VDM compared to IDM (1,292,056 vs 1,362,438). The code to derive such IDMs is relatively straightforward and easily shared (our code linked in Methods), developments in this sphere can be easily accessible and quickly scaled with straightforward data postprocessing. The IDMs also had much stronger associations and were statistically significant. These findings also suggest there may be other, even better, measures for detecting functional decline. Similarly, in a Nuedexta study, the effects of the drug on speech features were not detected by speech-language pathologists but were detected using objective measures.³⁹

The model estimation process converged for all LMM discussed. The marginal coefficient of determination (R2m) represents the proportion of the variance explained by fixed effects and R2c represents the proportion of the variance explained by both fixed and random effects.

The majority of the wearable measures significantly changed over time. We calculated R2c, the proportion of the variance explained by both fixed and random effects, to identify measures with very high R2c (≥ 0.85). Measures of PA volume and fragmentation had very high R2c and were well characterized with the use of relatively simple LMMs: LTAC, ASTP, and mean steps/minute.

The 13 wearable PA measures with significant associations with both surveys (ALSFRS-RSE and ROADS) centered around higher PA volume. This suggests that the specific type of PA monitoring may not matter as much as the volume of activity being monitored. There appears to be a trend towards more intense activity. Further exploration of whether certain PA intensities may serve as useful measures for detecting functional decline is needed. Modus measures were not significantly associated with either survey. It should be noted that this study was not designed to assess whether one device performs better than the other and such conclusions cannot be drawn from the results.

Results obtained in the sensitivity experiment comparing different scenarios of data collection protocols indicated that for the IDM TAC, which had an R2c of 0.82, as long as 2 weeks of data were collected, there was no substantial change in the baseline value or monthly change value with a break in data collection up to 6 weeks. Change in slope varied substantially in the less frequent monitoring scenarios (up to 24.5%) but all data collection frequencies found significant functional decline associations, even 1 week of data with 8 week breaks in between (1wD + 8wB). This is a very important finding, as it (1) points to the ability to identify thresholds of data collection needed to accurately track functional decline and (2) shows that days-long continuous data collection may be an adequate substitute for months-long data collection in some cases.

These findings affirm the feasibility and utility of remote monitoring with the use of apps and wearable devices for the collection of PROMS and daily measures of physical activity in PALS.

Benefits of digital data collection

Remote/at-home digital biometric data collection has several benefits for the field, and which make it attractive for outcome measure development. First, it enables more frequent sampling, which means that trials need not enroll as many participants to know whether an intervention is efficacious. This accelerates trial results and is also important in relatively uncommon diseases where recruitment may be difficult. Second, remote digital biometric data collection decreases participant burden, especially important in a population that experiences mobility challenges. Third, it may also increase access and inclusion, allowing those who are geographically or socioeconomically isolated to participate. Fourth, there is also likely benefit to assessing characteristics of daily life activities, like walking, in one’s lived-in environment. Fifth, digital data collection can reduce artificial changes in performance that might occur during an in-person visit and eliminates interrater differences. Sixth, objectively measured physical function data could be valuable, especially in a disease that profoundly affects quality of life and for which we rely on patient responses to generate metrics. Seventh, digital technology and data can also be highly accessible to participants, staff, and other researchers. Depending on the technology and vendors, it can be low-cost and quite scalable.⁴⁰ Finally, remote outcome assessment can reduce participants’ exposure risk to pathogens (i.e. COVID-19).

Challenges in digital data collection

There are challenges with remote, digital data collection, some of which may be unexpected. For example, we found that one participant’s partner often carried her phone on his person. One participant downloaded the Beiwe app onto a tablet rather than smartphone, despite instruction. One participant using the Modus wearable routinely used stationary bicycle pedals (desk-cycle). Device feature design is of paramount importance, especially in a population that frequently loses hand strength and dexterity. Participants can change the side on which they wear the device. Compliance is frequently an issue, despite the fact that research participants are likely more adherent than non-research patients. Technological difficulties, such as how to use the app (log back in, resolve glitches, etc.) all affect data collection and can require study coordinator support. In addition, for some participants, experiencing a high battery drain was very bothersome. Usability factors can be anticipated, such as designing straps and devices so that PALS with hand weakness can utilize them without undue burden. The advent of data sharing and open-source code enables a collaborative, transparent research environment. Caution is also needed to ensure systems perform as expected and that participant privacy is protected.⁴¹

Limitations

This study did not have a healthy control arm to which to compare results. There was missing data, resulting from imperfect compliance, though not enough to substantively affect data integrity. It is a reminder, however, that digital interface design, beta testing, and robust support for technical issues are critically important for ensuring data quality in digital studies. The study enrolled a more slowly progressing population than interventional trials, thus testing digital biomarkers in trial populations should be performed. Our ethnoracial diversity was low, but more inclusive than most US ALS studies. This may suggest that, properly applied, digital technology can bridge the gap between underrepresented populations in ALS research.

Future directions

This work generates many avenues for future directions in trial application validation, novel outcome measure development, and existing PROM and digital outcome measure refinement. Understanding how data is affected by the use of various assistive devices is also important. Ultimately, incorporation of digital outcome measures into interventional clinical trials alongside standard functional assessments will help clarify their role in boosting trial efficiency and acting as key outcome measures.

Until there is a cure for ALS, we must work to better understand the disease and improve our efficiency at investigating potential therapeutics. Having the best diagnostic and monitoring biomarkers and outcome measures to quantify disease progression are key. Just as ALS diagnostic research criteria have iteratively evolved over time,^42–45 so too, should our outcome measures if there is room for improvement. The COVID-19 pandemic has shown us that there are opportunities for enhanced research and care through technology incorporation.^46,47

This study demonstrated in a non-rapidly progressing, functionally heterogeneous group of PALS that passively collected sensor data from wearable devices (either wrist-worn or ankle-worn) can characterize daily physical activity and its change over time. These measures were also found to be significantly associated with ALSFRS-RSE and ROADS scores. Evaluation of digital outcome measures in interventional clinical trials is needed to understand whether they may have a role as primary outcome measures and enhance clinical trial efficiency.

Ethics

Massachusetts General Brigham (MGB) Institutional Review Board (IRB) and Massachusetts General Hospital (MGH) Information Security Office approvals were obtained prior to study initiation. All data were collected and maintained in accordance with MGB, state, and national policies and regulations. Participants provided written informed consent prior to participation in study procedures.

Study design and population

This was a single-center, non-interventional, remote study enrolling ambulatory participants over the age of 18 years with a diagnosis of ALS by El-Escorial Criteria⁴⁴ who were able to provide informed consent, comply with study procedures, and operate their smartphone without assistance – as determined by the site investigator’s assessment. The study was advertised on ALS social media accounts, an institutional research recruitment website, and to patients attending the ALS multidisciplinary clinic. Study procedures were performed remotely, with rare exception (participant on site). Participants were observed for 6-months and received $50 at 3- and 6-months if still contributing data.

Data collection and devices

Data were collected in three ways: (1) staff-administered ALSFRS-R surveys by phone/televisit, (2) remote survey completion using the Beiwe app, (3) and wearable device passive data collection (Fig. 3).

Following consent, study staff screened participants, and, if eligible, performed baseline visit procedures: demographics, baseline ALSFRS-R, download of and instruction on the use of the Beiwe platform (Beiwe) app on his/her smartphone (iOS or Android), and wearable device selection. Participants were called after receipt of their chosen device to review proper use and maintenance.

Beiwe is an open-source digital phenotyping research platform designed specifically for the collection of research-grade, raw (unprocessed) data from smartphone sensors, logs and self-entry measures (surveys, voice recordings).⁴⁸ Data collected with the Beiwe app are encrypted and stored on the smartphone until Wi-Fi transfer to the HIPAA-compliant Amazon Web Services (AWS) cloud occurs. Data remains encrypted at each stage.

The app was configured to administer all surveys at baseline, "key events" surveys every 12 weeks, and the ALSFRS-RSE and ROADS surveys every 2–4 weeks. Contacting participants was permitted if it seemed they were having technical difficulties or asked for help, though these calls were not systematically logged. The app was removed following study completion.

Two wearable devices were used: a wrist-worn ActiGraph Insight Watch (ActiGraph) purchased from ActiGraph LLC and an ankle-worn StepWatch 4 (Modus) from Modus Health LLC. Participants were instructed to wear their device as much as possible (preferably 24h); Modus participants were told they could remove it during sleep. Participants were not instructed to wear their device on a specific side. Devices were returned using prepaid, pre-addressed packaging.

The ActiGraph device is an activity monitor using a micromachined microelectromechanical systems triaxial-accelerometer that collected raw accelerometry data at 32 Hz. The ActiGraph CentrePoint Data Hub securely transmits de-identified data linked by subject ID using cellular data transfer to the CentrePoint cloud (Fig. 3).

The Modus device uses a custom mechanical biaxial accelerometer measuring acceleration at 128 Hz. A proprietary algorithm determines if a step was taken and records the corresponding timestamp. It has been validated in various patient populations, including those with impaired gait.⁴⁹ This study used Modus’ Clinical Research Trials app software. The device securely transmits de-identified data linked by a subject ID via Bluetooth to the Modus app on participants’ smartphones, which then transmits to Modus’ server using cellular data or Wi-Fi.

Vendor-provided data for both devices were accessible through secure, permissions-based Web Portals. Raw data were provided upon request.

Wearable daily PA measures

ActiGraph provided the following processed data: (1) "raw" data -- subsecond-level acceleration measurements; (2) minute-level data: AC (also known as vector magnitude counts) -- derived from "raw" data with the ActiGraph's algorithm recently made open-source,⁵⁰ (x-, y-, and z-) axis counts, steps, calories, metabolic equivalents (METs), wear status, awake status; (3) vendor-provided daily summary measures (VDMs) of time awake and asleep; sedentary and non-sedentary; in locomotion and non-locomotion; in light (< 3 METs), moderate (3 to < 6 METs), vigorous (> 6 METs), and MVPA; and each of the minute-level data measures. VDMs filtered for awake, wear, and awake + wear status were provided. Wear-filtered VDMs (those calculated from periods when the device was worn) were used in our analyses. Published algorithms are used for calculation of calories and METs⁵¹ and most of the remaining VDMs.⁵² VDMs do not explicitly account for missing data from sensor non-wear or interruptions of raw accelerometry data during device charging and/or data upload. As a result, VDMs may underrepresent PA.⁵³

Using ActiGraph minute-level AC, we created a set of investigator-derived daily measures (IDMs). Missing AC data was imputed prior to IDM calculation using participants’ wear-days’ corresponding minutes’ mean AC.⁵³ IDMs included total activity counts (TAC; 24-hour AC sum), LTAC (log of TAC), total log activity counts (TLAC; daily sum of log(1 + AC)),⁵⁴ minutes spent active (minutes with AC > 1853)⁵⁵ and inactive, ASTP (fragmentation measure representing the conditional probability a given minute is sedentary given a previously active minute),⁵⁶ and SATP (conditional probability a given minute is active given a previously sedentary minute).⁵⁶

Modus provided the following processed data: (1) second-level step count data; (2) minute-level step sums; (3) daily-level (VDM) step counts; percent time in low (1–15 steps/minute), medium (16–40 steps/minute), and high (41 + steps/minute) activity; (c) mean, median, 95th percentile, peak performance index (PPI), and max consecutive (60, 20, 5, and 1 minute) cadences. Cadence is defined as steps/minute and does not signify that there was activity during the entire given minute. PPI is the mean cadence of the day’s most intensive, non-contiguous 30 minutes. Modus did not have a wear status VDM. Wear was indirectly ascertained through step timestamp examination.

Data analysis sample

The analysis sample consisted of participants with at least two fully completed ALSFRS-RSE and ROADS surveys, used normed ROADS scores,¹⁰ and used only “valid days” for wearable data (VDMs and IDMs). Valid days were defined as days with at least 8, not necessarily consecutive, “valid hours.” Due to device differences, valid hours were defined uniquely for each wearable. For Actigraph, a valid hour was defined as 60 consecutive minutes without missing data and vendor-provided wear status indicating device wear. For Modus, a valid hour was one with at least one step logged.

Statistical data analysis

The number of complete ALSFRS-RSE and ROADS submissions were computed for each participant and then aggregated (median and range) across participants by device type (ActiGraph, Modus) and both groups combined. To characterize device wear compliance, the number of days in the observation period, valid days, and average valid hours on a valid day were computed and aggregated.

To quantify ALSFRS-RSE, ALSFRS-R, and their differences at baseline and longitudinally, four LMMs were fitted. Each model assumed time as a fixed effect, participant-specific random intercept and random slope, and differed in outcome (surveys’ total scores): (1) ALSFRS-RSE, (2) ROADS, (3) ALSFRS-R, and (4) ALSFRS-RSE and ALSFRS-R values. Model 4 included an indicator term for the ALSFRS-RSE outcome and a term for the indicator’s interaction with time.

To investigate whether differences exist between the ActiGraph and Modus groups’ survey baselines and change over time, two additional LMMs were fitted using time as a fixed effect, participant-specific random intercept and random slope, an indicator for Modus users, and the interaction between the Modus indicator and time with outcomes as ALSFRS-RSE (5), and ROADS (6). In both, the "time" variable reflects participant-specific elapsed time (in months) from the beginning of the observation period (coinciding with the ALSFRS-R baseline date).

ALSFRS-R and ALSFRS-RSE correlation was calculated at each ALSFRS-R administration: baseline, 3-, and 6-months using participants’ closest matching ALSFRS-RSE within 28 days.

To estimate the average baseline values and the change over time of the wearable daily PA measures, LMMs were fitted separately for each VDM and IDM. Each of the 32 LMM had a daily measure as the outcome, time as a fixed effect, and participant-specific random intercept and random slope.

To quantify the association between wearable daily PA measures and ALSFRS-RSE and ROADS scores, LMMs were fitted for each VDM and IDM (32 unique measures) to both ALSFRS-RSE and ROADS separately. The covariates (daily measures) were set as fixed effects, participant-specific random intercept and random slopes were used, and the ALSFRS-RSE/ROADS scores were the outcomes. Covariates were constructed by taking the average of the daily measure’s values spanning the 7-days before and after the survey date for a given participant and survey. To facilitate model comparison, covariates were standardized to have zero means and unit standard deviations.

R2c and R2m for generalized LMMs are reported.^29,30

A sensitivity analysis was performed to understand how less frequent monitoring might change wearable results using the TAC IDM. A series of models with different monitoring durations were fitted using TAC as the outcome, time as a fixed effect, and participant-specific random intercepts and random slopes: all wD (weeks data), 2wD + 2wB (weeks break), 2wD + 4wB, 2wD + 6wB, 2wD + 8wB, 1wD + 2wB, 1wD + 4wB, 1wD + 6wB, and 1wD + 8wB.

All analyses were performed using R software (version 4.2.0; The R Project). R code for all data preprocessing and data analysis is publicly available on GitHub repository (https://github.com/onnela-lab/als-wearables).⁵⁷

Study Funding: Mitsubishi Tanabe Pharma Holdings America, Inc., ALS Association

Search Terms: Amyotrophic lateral sclerosis, wearables, smartphones, digital biomarkers, ActiGraph, Modus StepWatch, outcomes, ALSFRS-RSE, ALSFRS-R, ROADS

Contributions –

Stephen A. Johnson – Concept and design, data collection, data analysis, drafting of the manuscript, critical review and revision of manuscript

Marta Karas – Concept and design, led data and statistical analysis, implemented the final version of the data preprocessing and analyses, major edits to the manuscript draft.

Katherine M. Burke – Concept and design, critical review of the manuscript

Marcin Straczkiewicz – critical review of the manuscript

Zoe A. Scheier – participant enrollment, data collection, critical review of the manuscript

Alison P. Clark – participant enrollment, data collection, Figure 3 creation, critical review of the manuscript

Satoshi Iwasaki – concept and design, critical review of the manuscript

Amir Lahav – concept and design, critical review of the manuscript

Amrita S. Iyer – data collection, Figure 3 creation, critical review of the manuscript

Jukka-Pekka Onnela – Concept and design, data analysis, critical review of the manuscript

James D. Berry – Concept and design, data analysis, critical review of the manuscript, study supervision and guidance

Acknowledgements

We would like to first and foremost thank the participants and their caregivers who devoted their time and energy, the device vendors for the services provided, and both Mitsubishi Tanabe Pharma Holdings America, Inc. and the ALS Association for their support. We also want to thank Zach Clement (Research Assistant in the Onnela Lab) and Kenzie W. Carlson (Senior Manager in the Onnela Lab) for their support with operations related to data access and processing infrastructure.

Data and code availability

The de-identified data are available upon reasonable request. R code for all data preprocessing and data analysis is publicly available on GitHub repository (https://github.com/onnela-lab/als-wearables).⁵⁷

Disclosures

The authors declare the following competing interests:

Stephen A. Johnson – reports research support from the ALS Association

Marta Karas – reports no competing interests

Katherine M. Burke – reports no competing interests

Marcin Straczkiewicz – reports no competing interests

Zoe A. Scheier – reports no competing interests

Alison P. Clark – reports no competing interests

Satoshi Iwasaki – employee of Mitsubishi Tanabe Pharma Holdings America, Inc.

Amir Lahav – paid strategic advisor for digital health at Mitsubishi Tanabe Pharma Holdings America, Inc. He has been paid for consulting services for Redenlab, Curebase, Castor, Lapsi Health, Solo AI, Quantificare, and the Global Platform Alzheimer’s Foundation. He has an unpaid role on the advisory board for the non-profit Everything ALS.

Amrita S. Iyer – reports no competing interests

Jukka-Pekka Onnela – reports no competing interests

James D. Berry – reports research support from Biogen, MT Pharma Holdings of America, Transposon Therapeutics, Alexion, Rapa Therapeutics, ALS Association, Muscular Dystrophy Association, ALS One, Tambourine, ALS Finding a Cure. He has been a paid member of an advisory panel for Regeneron, Biogen, Clene Nanomedicine, Mitsubishi Tanabe Pharma Holdings America, Inc., Janssen, RRT. He received an honorarium for educational events for Projects in Knowledge and the Muscular Dystrophy Association. He has unpaid roles on the advisory boards for the non-profits Everything ALS and ALS One.

Mitsubishi Tanabe Pharma Holdings America, Inc. (MTPHA) was a sponsor of this study. Listed authors from MTPHA were involved in protocol development and manuscript review. MTPHA has not restricted in any way publication of the full data set nor the authors’ right to publish. The authors retained full editorial control throughout manuscript preparation.

Goutman, S. A. et al. Recent advances in the diagnosis and prognosis of amyotrophic lateral sclerosis. Lancet Neurol 21, 480–493 (2022). https://doi.org:10.1016/s1474-4422(21)00465-8
Shatunov, A. & Al-Chalabi, A. The genetic architecture of ALS. Neurobiol Dis 147, 105156 (2021). https://doi.org:10.1016/j.nbd.2020.105156
Mehta, P. et al. Prevalence of amyotrophic lateral sclerosis (ALS), United States, 2016. Amyotroph Lateral Scler Frontotemporal Degener 23, 220–225 (2022). https://doi.org:10.1080/21678421.2021.1949021
Mejzini, R. et al. ALS Genetics, Mechanisms, and Therapeutics: Where Are We Now? Frontiers in Neuroscience 13 (2019). https://doi.org:10.3389/fnins.2019.01310
Kiernan, M. C. et al. Improving clinical trial outcomes in amyotrophic lateral sclerosis. Nat Rev Neurol 17, 104–118 (2021). https://doi.org:10.1038/s41582-020-00434-z
Cedarbaum, J. M. et al. The ALSFRS-R: a revised ALS functional rating scale that incorporates assessments of respiratory function. BDNF ALS Study Group (Phase III). J Neurol Sci 169, 13–21 (1999). https://doi.org:10.1016/s0022-510x(99)00210-5
Kollewe, K. et al. ALSFRS-R score and its ratio: a useful predictor for ALS-progression. J Neurol Sci 275, 69–73 (2008). https://doi.org:10.1016/j.jns.2008.07.016
Franchignoni, F., Mora, G., Giordano, A., Volanti, P. & Chio, A. Evidence of multidimensionality in the ALSFRS-R Scale: a critical appraisal on its measurement properties using Rasch analysis. J Neurol Neurosurg Psychiatry 84, 1340–1345 (2013). https://doi.org:10.1136/jnnp-2012-304701
Bedlack, R. S. et al. How common are ALS plateaus and reversals? Neurology 86, 808–812 (2016). https://doi.org:10.1212/WNL.0000000000002251
Fournier, C. N. et al. Development and Validation of the Rasch-Built Overall Amyotrophic Lateral Sclerosis Disability Scale (ROADS). JAMA Neurol 77, 480–488 (2020). https://doi.org:10.1001/jamaneurol.2019.4490
Montes, J. et al. Development and evaluation of a self-administered version of the ALSFRS-R. Neurology 67, 1294–1296 (2006). https://doi.org:10.1212/01.wnl.0000238505.22066.fc
Maier, A. et al. Online assessment of ALS functional rating scale compares well to in-clinic evaluation: a prospective trial. Amyotroph Lateral Scler 13, 210–216 (2012). https://doi.org:10.3109/17482968.2011.633268
Bakker, L. A. et al. Development and assessment of the inter-rater and intra-rater reproducibility of a self-administration version of the ALSFRS-R. J Neurol Neurosurg Psychiatry 91, 75–81 (2020). https://doi.org:10.1136/jnnp-2019-321138
Manera, U. et al. Validation of the Italian version of self-administered ALSFRS-R scale. Amyotroph Lateral Scler Frontotemporal Degener 22, 151–153 (2021). https://doi.org:10.1080/21678421.2020.1813307
Berry, J. D. et al. Design and results of a smartphone-based digital phenotyping study to quantify ALS progression. Ann Clin Transl Neurol 6, 873–881 (2019). https://doi.org:10.1002/acn3.770
Chew, S. et al. Patient reported outcomes in ALS: characteristics of the self-entry ALS Functional Rating Scale-revised and the Activities-specific Balance Confidence Scale. Amyotroph Lateral Scler Frontotemporal Degener 22, 467–477 (2021). https://doi.org:10.1080/21678421.2021.1900259
Johnson, S. A. et al. Longitudinal comparison of the self-entry amyotrophic lateral sclerosis functional rating scale-revised (ALSFRS-RSE) and rasch-built overall amyotrophic lateral sclerosis disability scale (ROADS) as outcome measures in people with amyotrophic lateral sclerosis. Muscle Nerve (2022). https://doi.org:10.1002/mus.27691
De Marchi, F. et al. Patient reported outcome measures (PROMs) in amyotrophic lateral sclerosis. J Neurol 267, 1754–1759 (2020). https://doi.org:10.1007/s00415-020-09774-8
Karas, M. et al. Accelerometry Data in Health Research: Challenges and Opportunities. Statistics in Biosciences 11, 210–237 (2019). https://doi.org:10.1007/s12561-018-9227-2
Youn, B. Y. et al. Digital Biomarkers for Neuromuscular Disorders: A Systematic Scoping Review. Diagnostics (Basel) 11 (2021). https://doi.org:10.3390/diagnostics11071275
Garcia-Gancedo, L. et al. Objectively Monitoring Amyotrophic Lateral Sclerosis Patient Symptoms During Clinical Trials With Sensors: Observational Study. JMIR Mhealth Uhealth 7, e13433 (2019). https://doi.org:10.2196/13433
van Eijk, R. P. A. et al. Accelerometry for remote monitoring of physical activity in amyotrophic lateral sclerosis: a longitudinal cohort study. J Neurol 266, 2387–2395 (2019). https://doi.org:10.1007/s00415-019-09427-5
Kelly, M. et al. The use of biotelemetry to explore disease progression markers in amyotrophic lateral sclerosis. Amyotroph Lateral Scler Frontotemporal Degener 21, 563–573 (2020). https://doi.org:10.1080/21678421.2020.1773501
Rutkove, S. B. et al. Improved ALS clinical trials through frequent at-home self-assessment: a proof of concept study. Ann Clin Transl Neurol 7, 1148–1157 (2020). https://doi.org:10.1002/acn3.51096
The Mobile Economy 2022. (GSM Association, 2022).
Mobile Fact Sheet. (Pew Research Center, Online, 2021).
Onnela, J.-P. Opportunities and challenges in the collection and analysis of digital phenotyping data. Neuropsychopharmacology 46, 45–54 (2021). https://doi.org:10.1038/s41386-020-0771-3
Beukenhorst, A. L. et al. Using Smartphones to Reduce Research Burden in a Neurodegenerative Population and Assessing Participant Adherence: A Randomized Clinical Trial and Two Observational Studies. JMIR Mhealth Uhealth 10, e31877 (2022). https://doi.org:10.2196/31877
Johnson, P. C. D. Extension of Nakagawa & Schielzeth's R2GLMM to random slopes models. Methods in Ecology and Evolution 5, 944–946 (2014). https://doi.org:https://doi.org/10.1111/2041-210X.12225
Nakagawa, S. & Schielzeth, H. A general and simple method for obtaining R2 from generalized linear mixed-effects models. Methods in Ecology and Evolution 4, 133–142 (2013). https://doi.org:https://doi.org/10.1111/j.2041-210x.2012.00261.x
Brusniak, K. et al. Challenges in Acceptance and Compliance in Digital Health Assessments During Pregnancy: Prospective Cohort Study. JMIR Mhealth Uhealth 8, e17377 (2020). https://doi.org:10.2196/17377
Merilahti, J. et al. Compliance and technical feasibility of long-term health monitoring with wearable and ambient technologies. Journal of Telemedicine and Telecare 15, 302–309 (2009). https://doi.org:10.1258/jtt.2009.081106
Cohen, S. et al. Characterizing patient compliance over six months in remote digital trials of Parkinson’s and Huntington disease. BMC Medical Informatics and Decision Making 18, 138 (2018). https://doi.org:10.1186/s12911-018-0714-7
Martinez, G. J. et al. Predicting Participant Compliance With Fitness Tracker Wearing and Ecological Momentary Assessment Protocols in Information Workers: Observational Study. JMIR Mhealth Uhealth 9, e22218 (2021). https://doi.org:10.2196/22218
Cudkowicz, M. E. et al. Safety and efficacy of ceftriaxone for amyotrophic lateral sclerosis: a multi-stage, randomised, double-blind, placebo-controlled trial. Lancet Neurol 13, 1083–1091 (2014). https://doi.org:10.1016/s1474-4422(14)70222-4
Paganoni, S. et al. Trial of Sodium Phenylbutyrate–Taurursodiol for Amyotrophic Lateral Sclerosis. New England Journal of Medicine 383, 919–930 (2020). https://doi.org:10.1056/NEJMoa1916945
Writing, G. & Edaravone, A. L. S. S. G. Safety and efficacy of edaravone in well defined patients with amyotrophic lateral sclerosis: a randomised, double-blind, placebo-controlled trial. Lancet Neurol 16, 505–512 (2017). https://doi.org:10.1016/S1474-4422(17)30115-1
Atassi, N. et al. The PRO-ACT database: design, initial analyses, and predictive features. Neurology 83, 1719–1725 (2014). https://doi.org:10.1212/WNL.0000000000000951
Green, J. R. et al. Additional evidence for a therapeutic effect of dextromethorphan/quinidine on bulbar motor function in patients with amyotrophic lateral sclerosis: A quantitative speech analysis. Br J Clin Pharmacol 84, 2849–2856 (2018). https://doi.org:10.1111/bcp.13745
Stephenson, D., Badawy, R., Mathur, S., Tome, M. & Rochester, L. Digital Progression Biomarkers as Novel Endpoints in Clinical Trials: A Multistakeholder Perspective. J Parkinsons Dis 11, S103-s109 (2021). https://doi.org:10.3233/jpd-202428
Stegmann, G. M. et al. Repeatability of Commonly Used Speech and Language Features for Clinical Applications. Digit Biomark 4, 109–122 (2020). https://doi.org:10.1159/000511671
Shefner, J. M. et al. A proposal for new diagnostic criteria for ALS. Clin Neurophysiol 131, 1975–1978 (2020). https://doi.org:10.1016/j.clinph.2020.04.005
Brooks, B. R., Miller, R. G., Swash, M. & Munsat, T. L. El Escorial revisited: revised criteria for the diagnosis of amyotrophic lateral sclerosis. Amyotroph Lateral Scler Other Motor Neuron Disord 1, 293–299 (2000). https://doi.org:10.1080/146608200300079536
Brooks, B. R. El Escorial World Federation of Neurology criteria for the diagnosis of amyotrophic lateral sclerosis. Subcommittee on Motor Neuron Diseases/Amyotrophic Lateral Sclerosis of the World Federation of Neurology Research Group on Neuromuscular Diseases and the El Escorial "Clinical limits of amyotrophic lateral sclerosis" workshop contributors. J Neurol Sci 124 Suppl, 96–107 (1994). https://doi.org:10.1016/0022-510x(94)90191-0
de Carvalho, M. et al. Electrodiagnostic criteria for diagnosis of ALS. Clin Neurophysiol 119, 497–503 (2008). https://doi.org:10.1016/j.clinph.2007.09.143
Govindarajan, R., Berry, J. D., Paganoni, S., Pulley, M. T. & Simmons, Z. Optimizing telemedicine to facilitate amyotrophic lateral sclerosis clinical trials. Muscle Nerve 62, 321–326 (2020). https://doi.org:10.1002/mus.26921
Beukenhorst, A. L. et al. Smartphone data during the COVID-19 pandemic can quantify behavioral changes in people with ALS. Muscle Nerve 63, 258–262 (2021). https://doi.org:10.1002/mus.27110
Jukka-Pekka Onnela, C. D., Keary Griffin, Tucker Jaenicke, Leila Minowada, Sean Esterkin, Alvin Siu, Josh Zagorsky, and Eli & Jones. Beiwe: A data collection platform for high-throughput digital phenotyping.. Journal of Open Source Software 6, 1–6 (2021). https://doi.org:https://doi.org/10.21105/joss.03417
Treacy, D. et al. Validity of Different Activity Monitors to Count Steps in an Inpatient Rehabilitation Setting. Phys Ther 97, 581–588 (2017). https://doi.org:10.1093/ptj/pzx010
Neishabouri, A. et al. Quantification of acceleration as activity counts in ActiGraph wearable. Scientific Reports 12, 11958 (2022). https://doi.org:10.1038/s41598-022-16003-x
Hildebrand, M., VT, V. A. N. H., Hansen, B. H. & Ekelund, U. Age group comparability of raw accelerometer output from wrist- and hip-worn monitors. Med Sci Sports Exerc 46, 1816–1824 (2014). https://doi.org:10.1249/mss.0000000000000289
Staudenmayer, J., He, S., Hickey, A., Sasaki, J. & Freedson, P. Methods to estimate aspects of physical activity and sedentary behavior from high-frequency wrist accelerometer measurements. J Appl Physiol (1985) 119, 396–403 (2015). https://doi.org:10.1152/japplphysiol.00026.2015
Catellier, D. J. et al. Imputation of missing data when measuring physical activity by accelerometry. Med Sci Sports Exerc 37, S555-562 (2005). https://doi.org:10.1249/01.mss.0000185651.59486.4e
Varma, V. R. et al. Re-evaluating the effect of age on physical activity over the lifespan. Prev Med 101, 102–108 (2017). https://doi.org:10.1016/j.ypmed.2017.05.030
KOSTER, A. et al. Comparison of Sedentary Estimates between activPAL and Hip- and Wrist-Worn ActiGraph. Medicine & Science in Sports & Exercise 48, 1514–1522 (2016). https://doi.org:10.1249/mss.0000000000000924
Di, J. et al. Patterns of sedentary and active time accumulation are associated with mortality in US adults: The NHANES study. Preprint at https://www.biorxiv.org/content/10.1101/182337v1.full.pdf. bioRxiv, 182337 (2017). https://doi.org:10.1101/182337
Karas, M., Onnela, Jukka-Pekka. Wearable device and smartphone data can track ALS disease progression and may serve as novel clinical trial outcome measures’ R code for all data preprocessing and data analysis, <https://github.com/onnela-lab/als-wearables> (2022).

There is a conflict of interest as discussed at end of manuscript:

The authors declare the following competing interests: Stephen A. Johnson – reports research support from the ALS Association Marta Karas – reports no competing interests Katherine M. Burke – reports no competing interests Marcin Straczkiewicz – reports no competing interests Zoe A. Scheier – reports no competing interests Alison P. Clark – reports no competing interests Satoshi Iwasaki – employee of Mitsubishi Tanabe Pharma Holdings America, Inc. Amir Lahav – consultant for Mitsubishi Tanabe Pharma Holdings America, Inc. Amrita S. Iyer – reports no competing interests Jukka-Pekka Onnela – reports no competing interests James D. Berry – reports research support from Biogen, MT Pharma Holdings of America, Transposon Therapeutics, Alexion, Rapa Therapeutics, ALS Association, Muscular Dystrophy Association, ALS One, Tambourine, ALS Finding a Cure. He has been a paid member of an advisory panel for Regeneron, Biogen, Clene Nanomedicine, Mitsubishi Tanabe Pharma Holdings America, Inc., Janssen, RRT. He received an honorarium for educational events for Projects in Knowledge and the Muscular Dystrophy Association. He has unpaid roles on the advisory boards for the non-profits Everything ALS and ALS One.

Mitsubishi Tanabe Pharma Holdings America, Inc. (MTPHA) was a sponsor of this study. Listed authors from MTPHA were involved in protocol development and manuscript review. MTPHA has not restricted in any way publication of the full data set nor the authors’ right to publish. The authors retained full editorial control throughout manuscript preparation.

Supptablesfigures.docx

Wearable device and smartphone data can track ALS disease progression and may serve as novel clinical trial outcome measures

Status:

Journal Publication

Version 1

Abstract

Figures

Introduction

Results

Discussion

Conclusions

Methods

Declarations

References

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1