Ассоциация между нейрофильной цепью освещения и реальной функцией ампуляции в прогрессивных MS.

Introduction

Progressive multiple sclerosis (MS) is characterized by insidious neurological worsening driven by neuroaxonal damage, for which reliable markers of change remain limited [ 1 , 2 ]. Accessible biomarkers that capture neurodegeneration and predict clinical worsening are needed to accelerate therapeutic development and guide treatment decisions.

Cerebrospinal fluid (CSF) neurofilament light chain (NfL) levels reflect active neuroaxonal injury and serve as a quantifiable marker of neurodegeneration in MS [ 3 ]. However, obtaining CSF via lumbar puncture is invasive, costly, and impractical for serial monitoring in routine clinical practice. By contrast, serum NfL (sNfL) offers a minimally invasive, cost‐effective alternative. Obtained through routine blood sampling, sNfL also reflects neuroaxonal injury and correlates with MS disease activity [ 4 ], with elevated levels indicating ongoing neuroaxonal pathology [ 5 , 6 , 7 ]. However, the use of sNfL has important considerations for clinical implementation and research, including large differences between utilized assays and the potential for confounding by age, renal function, and other systemic factors [ 4 , 8 , 9 ]. Due to these considerations, the use of age‐ and BMI‐adjusted sNfL z ‐scores has become increasingly common. These considerations also highlight the importance of evaluating sNfL in relation to real‐world functional outcomes.

Average daily step count (STEPS), continuously captured using consumer‐grade wearable accelerometry, has emerged as an ecologically valid measure of ambulatory behavior in people with MS that complements episodic clinic‐based assessments such as the Expanded Disability Status Scale (EDSS) and standard timed walking tests [ 10 , 11 , 12 ]. STEPS is associated not only with disability level and common MS symptoms (including fatigue, mood, bladder dysfunction, and fall risk) but also with spinal cord gray matter atrophy, highlighting relationships between ambulatory function with both MS symptoms and structural pathology [ 10 , 13 ].

While there is preliminary evidence that self‐reported physical activity is associated with sNfL levels in frontotemporal lobar degeneration [ 14 ], the longitudinal relationship between sNfL and objective measures of real‐world ambulatory activity remains poorly characterized in MS. It is unclear how biological evidence of neuroaxonal injury (sNfL) and real‐world mobility (STEPS) evolve over time, and whether changes in sNfL precede subsequent changes in STEPS. Clarifying these temporal relationships may improve our understanding of disease progression and determine whether physical activity levels are temporally linked to biological markers of neurodegeneration, advancing implications for potential therapeutic intervention [ 9 , 15 , 16 , 17 ].

To address this question, data was leveraged from SPI2, a Phase 3 randomized controlled trial in progressive MS [ 18 ]. Hence, our primary objective was to evaluate the longitudinal relationship between sNfL and continuously measured STEPS to determine whether biological evidence of neuroaxonal injury is associated with real‐world ambulatory activity in people with progressive MS.

Methods

Participants

A post hoc analysis of the SPI2 trial (high‐dose biotin/MD1003 vs. placebo) was conducted [ 18 ]. Parexel International, an independent clinical research organization, conducted the trial. The Institutional review boards or ethics committees approved the protocol and amendments at all study centers. Adults (18–65 years) with primary or secondary progressive MS [ 19 ] were eligible if there was evidence of disability progression, a pyramidal functional score ≥ 2, EDSS 3.5–6.5, and a timed 25‐ft walk (T25FW) < 40 s, with no relapses in the preceding two years. Concomitant disease‐modifying therapies (DMTs) were permitted. The two treatment arms were pooled and analyzed together because a therapeutic benefit of high‐dose biotin was not found in the primary or secondary outcomes [ 18 ]. Written informed consent was obtained from all participants.

Biological Markers

sNfL z ‐scores were obtained from longitudinal blood samples at baseline and 6, 12, 15, and 27 months (when available). The measurement closest to each scheduled visit was used for analysis. sNfL concentrations were measured by NF‐light Advantage kit (Quanterix) and age‐ and body mass index (BMI)‐adjusted z scores were generated. sNfL z ‐scores are a measure of deviation from values observed in healthy controls: for example, an sNfL z ‐score of 1 means that the sNfL concentration deviates by 1 SD from values in the reference database (84.1st percentile) adjusted for relevant physiological factors [ 20 ].

A Fitbit Flex (or Flex 2) wearable accelerometer was provided. Anonymous accounts were created, and step counts were continuously recorded. Participants were instructed to wear the Fitbit continuously on the non‐dominant wrist every day and to synchronize the device regularly throughout the study. Uploaded Fitbit data were monitored by the STEPSCore at the primary study site to identify technical issues and adherence. Consistent with our previously published methodology [ 11 ], days with fewer than 128 recorded steps were classified as non‐wear days and excluded from the analyses. Average daily step count for each visit was calculated as the mean of daily steps within ±15 days of the scheduled visit date (STEPS) [ 21 ] at baseline and at 6, 12, 15, and 27 months; observations at 27 months were included when available.

Statistical Analysis

Statistical analysis was performed using STATA version 19 software (StataCorp LP, College Station, TX, USA) with significance set to p < 0.05. Descriptive statistics included means and SD for continuous variables and frequencies and percentages for categorical variables. Potential outliers were identified using an interquartile range (IQR)‐based approach and visual inspection of histograms. Extreme values corresponding to clear breaks in the distributions were excluded from the primary analyses (one observation for the sNfL z ‐score using an IQR ±2.5 criterion and nine observations for STEPS using an IQR > 2.5 criterion). Sensitivity analyses, including all observations, yielded similar results.

Contemporaneous associations between sNfL and STEPS were examined in separate models with each measure alternately specified as the outcome. The lagged analysis evaluated whether earlier changes in sNfL predicted subsequent changes in STEPS; the reverse temporal association was not evaluated.

sNfL Outcome

Overall Longitudinal Associations

Associations between STEPS and sNfL z ‐score were assessed using linear mixed‐effects models with random intercepts for participants and adjustment for age and sex. STEPS was log‐transformed prior to analysis due to a right‐skewed distribution; linearity with sNfL z ‐scores was assessed and supported. Model uncertainty was quantified using cluster bootstrap resampling at the participant level (with 1000 replications). Regression coefficients were interpreted as the expected change in sNfL z ‐score per 10% increase in daily steps. Sensitivity analyses tested interactions between STEPS and (1) sex, (2) treatment group, (3) baseline EDSS category (≤ 5.5 vs. > 5.5), and (4) use of DMT at screening, respectively, to assess potential effect modification. Associations between sNfL z ‐score and clinical disability measures (EDSS, T25FW) were evaluated using linear mixed‐effects models, adjusted for age and sex. EDSS was modeled both categorically and continuously to assess non‐linear associations and linear trends; T25FW was modeled as a continuous predictor. Effect modification was examined using interaction terms with sex, treatment group, baseline EDSS category, and baseline DMT use.

Group‐Based Change Analysis

Complementary group‐based analyses categorized participants by slope‐based change in STEPS over follow‐up (±15%) [ 22 ]. For each participant (≥ 2 observations required), a person‐specific linear regression of STEPS on study month was used to estimate the percent change between the earliest and latest timepoints. Participants were classified as having a ≥ 15% increase or ≤ −15% decrease (smallest clinically important change in people with MS = 15%) [ 22 ]; those not meeting either threshold were excluded. Associations between step‐change group and sNfL z ‐score were evaluated using linear mixed‐effects models adjusted for age and sex, with a month‐by‐group interaction. Uncertainty was quantified using participant‐level cluster bootstrap resampling (1000 replications). Predicted sNfL z ‐score trajectories were estimated over 0–27 months.

STEPS Outcome

Overall Longitudinal Associations

Associations between sNfL z ‐score and STEPS were evaluated using mixed‐effects negative binomial regression models [ 23 ] to account for the right‐skewed distribution of STEPS and repeated measures. Models were adjusted for age and sex; sNfL was modeled linearly after excluding non‐linearity. Uncertainty was quantified using cluster bootstrap resampling (1000 replications), with results reported as incidence rate ratios (IRRs). Sensitivity analyses evaluated effect modification by sex, baseline EDSS category (> 5.5 vs. ≤ 5.5), treatment group, and baseline DMT use through inclusion of interaction terms. Clinical disability measures were evaluated in separate models, with EDSS modeled categorically and continuously, and T25FW modeled continuously; interaction terms with sex, treatment group, and use of DMTs were included to assess effect modification.

Change‐Based Models

Lagged‐based approaches were repeated with STEPS as the outcome using linear regression models, evaluating whether 0 to 6‐month changes in sNfL predicted 6 to 12‐month changes in STEPS. Models were adjusted for age and sex, with bootstrapped inference (1000 replications), and assessed for non‐linearity.

Slope‐Based Analysis

To explore whether within‐person trajectories of sNfL and STEPS were related over time, separate mixed‐effects models with random intercepts and random slopes for time were fit for sNfL z ‐score and STEPS, adjusted for age and sex. Patient‐specific slope estimates were extracted from each model and merged by participant. The association between patient‐specific sNfL and STEPS slopes was then evaluated using Pearson correlation and linear regression.

Because of the post hoc exploratory nature of this study, all reported p ‐values are considered nominal rather than declarative.

Results

Participant Characteristics

Of 642 participants, 506 were included in the longitudinal analyses, defined as those with at least one paired non‐missing STEPS and sNfL measurement during follow‐up. Measurements were paired by visit, although some visit‐level observations lacked paired data because STEPS were unavailable within the predefined ±15‐day visit window. Among 1985 visit‐level sNfL observations eligible for pairing, 502 (25.3%) lacked a usable paired STEPS measurement within the predefined ±15‐day window, including 493 with missing STEPS data and nine excluded as outliers. This left 1483 (74.7%) paired observations available for analysis. Of the 630 participants with sNfL data, 506 (80.3%) had at least one STEPS observation and 124 (19.7%) had no STEPS data. Baseline EDSS was similar among participants with and without STEPS data (median [IQR], 6.0 [4.5–6.0] vs. 6.0 [5.0–6.5], respectively; p = 0.063).

The mean age was 52.7 years (SD 7.7), 53.8% ( N = 272) were female, 64.4% had secondary progressive MS ( N = 326), and the median EDSS score for the cohort was 6.0 [IQR 4.5–6.0, (range 2.5–6.5)]. Of those included in this analysis, 50.0% received MD1003 treatment ( N = 253) and 50.0% received placebo ( N = 253). At screening, 61.9% ( N = 313) were not currently treated with a DMT and 12.5% ( N = 63) were actively treated with a B cell depleting monoclonal antibody (further details available in Table S1 ). The mean baseline sNfL z ‐score was 1.00 (SD 0.95), and the median STEPS was 3029 (IQR: 1708–5210, Table 1 ).

sNfL Outcome

Overall Longitudinal Associations

Higher STEPS were associated with lower sNfL z ‐scores (Figures 1 and S1 shows the same model with raw data points overlaid). In adjusted mixed‐effects models, a 10% increase in STEPS was associated with a 0.015 lower sNfL z ‐score (95% CI: −0.022 to −0.008; p < 0.001). In sensitivity analyses, interaction testing showed that the association between STEPS and sNfL z ‐score did not differ by sex ( p = 0.14), treatment group ( p = 0.37), baseline EDSS category ( p = 0.82), or DMT at screening ( p = 0.26).

EDSS group (categorical) was associated with sNfL z ‐score (global test, p = 0.001), Figure 2 . Higher EDSS score (continuous) was associated with higher sNfL z ‐score ( β = 0.093 per 1‐point increase in EDSS, 95% CI 0.04 to 0.15; p = 0.001). There was no evidence of effect modification by sex ( p = 0.53), by treatment group ( p = 0.39), or by DMT use ( p = 0.10).

Slower walking speeds (during T25FW) were associated with higher sNfL z ‐scores ( β = 0.055 per 10‐s increase, 95% CI 0.016 to 0.094; p = 0.005). Sensitivity analyses showed no effect modification by sex ( p = 0.40), baseline EDSS category ( p = 0.53), treatment group ( p = 0.59), or DMT ( p = 0.77), Figure 2 .

Group‐Based Change Analysis

Using a slope‐based classification (±15% threshold), 112 participants were classified as having an increase, and 186 as having a decrease, in STEPS over 27 months. sNfL z ‐score increased over time in the group with decreasing STEPS ( β = 0.009 z ‐score units/month, 95% CI 0.002 to 0.015; p = 0.007), while the group with increasing STEPS showed a flatter trajectory ( β = −0.002 z ‐score units/month, 95% CI −0.012 to 0.008; p = 0.68). Although the difference in slopes between groups did not reach statistical significance (interaction p = 0.07), the estimated difference corresponded to ~0.160 lower sNfL z ‐scores over 15 months and ~0.290 lower sNfL z ‐scores over 27 months in the increasing‐STEP group relative to the decreasing‐STEP group, Figure 3 .

STEPS Outcome

Overall Longitudinal Associations

Higher sNfL z ‐scores were associated with lower STEPS (Figures 4 and S2 show the same model with raw data points overlaid). In adjusted models, each 1‐unit increase in sNfL z ‐score was associated with an approximately 3.4% decrease in STEPS (IRR = 0.97, 95% CI 0.94 to 0.99; p = 0.014). Based on marginal model predictions, participants at the 75th percentile of sNfL z ‐score had an estimated 176 fewer STEPS than participants at the 25th percentile (difference = −175.90 steps, 95% CI −315.72 to −36.09; p = 0.01). No evidence of effect modification was found (interaction p ‐values: EDSS category = 0.99, treatment group = 0.97, sex = 0.16, or DMT = 0.84). EDSS category was associated with STEPS overall (global test, p < 0.001), Figure 5 . Consistent with these findings, higher continuous EDSS was associated with significantly lower STEPS. Each 1‐point increase in EDSS was associated with a ~20% reduction in STEPS (IRR = 0.80; 95% CI 0.76–0.84; p < 0.001). There was no evidence of effect modification by sex, treatment group, or DMT use at baseline ( p > 0.05).

Slower walking speeds (during T25FW) were associated with lower STEPS (Figure 5 ). Each 1‐s increase in T25FW was associated with a ~1.3% lower STEPS (IRR = 0.987; 95% CI 0.983 to 0.991; p < 0.001). No evidence of effect modification by sex, treatment group, or DMT use was found, indicating that the association between T25FW and STEPS was consistent across subgroups.

Change‐Based Models

In lagged change analyses, there was a trend suggesting that greater increases in sNfL from 0 to 6 months were associated with lower subsequent changes in STEPS from 6 to 12 months ( β = −141; 95% CI −302 to 21; p = 0.088), adjusted for age and sex (Figure 6 ). Specifically, a 1 SD greater increase in sNfL over 0–6 months was associated with 141 fewer steps over the subsequent 6–12‐month interval (Figures 6 and S3 show the same model with raw data points overlaid).

Slope‐Based Analysis

In an analysis of patient‐specific trajectories, higher sNfL slopes were associated with lower STEPS slopes over time. Patient‐specific sNfL and STEPS slopes were modestly inversely correlated ( r = −0.18, p < 0.001). In linear regression, each 1‐unit greater increase in the patient‐specific sNfL slope was associated with a 618.90‐unit lower patient‐specific STEPS slope ( β = −618.90; 95% CI −910.42 to −327.38; p < 0.001; Figure S4 ). Conceptually, this analysis asks whether people whose NfL increases faster also tend to experience larger declines in STEPS over time. The modest negative correlation between the slopes suggests that participants with faster increases in NfL tend to experience greater declines in STEPS.

Discussion

In this post hoc, longitudinal analysis, sNfL was associated with daily ambulatory activity in people with progressive MS. Higher sNfL z ‐scores, adjusted for physiological confounders such as age and BMI to improve comparability across individual participants, were associated with lower real‐world mobility (STEPS) and vice versa in this progressive MS cohort. Notably, the association between higher STEPS and lower sNfL was similar across disability levels, suggesting that ambulatory function may reflect aspects of cumulative CNS injury not fully captured by traditional clinical metrics. Furthermore, group‐based change and exploratory participant‐specific slope analyses showed that decreases in STEPS over time were associated with increases in sNfL. As expected, and consistent with prior work, these relationships aligned with standard outcome measures: higher sNfL levels and lower STEPS associated with higher disability levels by EDSS and T25FW [ 10 , 24 , 25 , 26 ]. The apparent inflection of predicted sNfL around EDSS 4.5–5.5 may reflect either a true acceleration in tissue injury at higher disability, known measurement variability, or a low number of participants at this EDSS range. Similarly, the increase in predicted STEPS at EDSS 7.5 is likely due to low participant numbers, with only 16 of the 1483 total observations (1.1%) at EDSS 7.0–7.5.

Consistent with prior work across populations [ 4 , 14 , 17 ], sNfL and activity (STEPS) appear to reflect underlying neurodegeneration while capturing distinct and potentially complementary dimensions of disease progression. While sNfL reflects neuroaxonal injury as a unidirectional biomarker of CNS damage, STEPS captures both functional consequences of neurodegeneration and behavior. Declining activity could, in theory, contribute to worsening neurodegeneration through deconditioning or comorbidity burden. sNfL elevations may precede clinically detectable worsening by ~1 year [ 8 , 27 ], whereas the indirect effects of reduced mobility on neuroaxonal injury would be expected over timeframes beyond the duration of SPI2.

Although this study was not designed to assess causality, lagged analyses provide initial directional evidence: while not statistically significant, there was a trend towards greater increases in sNfL being followed by a subsequent decline in STEPS, consistent with neuroaxonal injury contributing to reduced physical activity. These findings support exploring sNfL as a potential biomarker in rehabilitation studies in MS. While exercise and physical activity improve MS symptoms, including fatigue and balance, their effects on disease progression remain limited [ 28 , 29 ]. Incorporating blood‐based biomarkers alongside traditional measures may enhance sensitivity in future trials [ 30 , 31 , 32 ].

While sNfL is associated with imaging markers, NfL from CSF, and clinical outcomes [ 31 , 33 , 34 ], frequent blood‐based monitoring may be logistically challenging in some clinical and research settings, particularly in resource‐limited or remote areas. Patient‐generated data (STEPS) from wearable sensors offers a scalable alternative for continuous monitoring of disease activity. In this context, real‐world mobility metrics may serve as accessible functional correlates of neuroaxonal injury where repeated biomarker sampling is not feasible.

This study has several limitations. The cohort was mainly recruited from Europe, North America, and Australia, limiting generalizability. SPI2 provided a unique opportunity to evaluate ongoing neurodegeneration in a progressive MS cohort who experienced at least moderate disability, low relapse activity, and minimal enhancing MRI activity [ 18 ], allowing evaluation of neurodegeneration with reduced confounding from inflammatory activity. However, this limits generalizability beyond patients with non‐active, progressive MS. Additional studies are needed to evaluate the associations in patients with lower disability or other MS subtypes. Although standardized STEPS collection procedures were implemented to maximize adherence and identify non‐wear, including regular synchronization of devices, monitoring of uploaded data, and exclusion of days with fewer than 128 recorded steps, occasional non‐wear or device misuse cannot be completely excluded. Any resulting measurement error would be expected to be non‐differential and therefore more likely to attenuate than exaggerate the observed associations. Further, the post hoc nature of this analysis is by its nature exploratory: all reported associations require validation in cohorts designed to test the hypotheses generated by this study.

Conclusion

These results suggest that, in this non‐active progressive MS cohort, reduced daily physical activity is associated with greater ongoing tissue injury measured by sNfL, and vice versa. These findings justify further exploration of the use of both biological and functional measures to detect MS progression in clinical and rehabilitation trials.

Author Contributions

G.B.J. performed the statistical analyses for this post hoc study and contributed to manuscript drafting, data interpretation, and critical revision of the manuscript. M.M. contributed to manuscript drafting and interpretation of the findings. R.B. contributed to the interpretation of the findings and critical revision of the manuscript. A.A. contributed to the conception of the present analysis, interpretation of the findings, and critical revision of the manuscript. B.A.C.C. served as the global principal investigator of the original SPI2 study and contributed to the conception and design of both the parent study and the present post hoc analysis, as well as to data interpretation and critical revision of the manuscript. V.J.B. contributed to data acquisition for the original SPI2 study and to the conception and design of the present analysis, development of the statistical analysis plan, manuscript drafting, and interpretation of the findings. All authors reviewed and approved the final manuscript.

Funding

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: DMT at baseline visit.

Figure S1: Higher daily step count was associated with lower serum NfL z ‐scores (sNfL). The figure displays predicted values from a mixed‐effects model, adjusted for age and sex, with raw data points overlaid. Error bars represent 95% confidence intervals. In adjusted models, a 10% increase in daily steps was associated with a 0.015 lower sNfL z ‐score (95% CI −0.022 to −0.008; p < 0.001).

Figure S2: Higher serum NfL z ‐score (sNfL) was associated with lower STEPS (IRR = 0.97; 95% CI 0.94 to 0.99; p = 0.01), corresponding to an approximately 3.4% decrease in STEPS per 1‐unit increase in sNfL z ‐score. The figure displays predicted values from a mixed‐effects binomial model, adjusted for age and sex, with raw data points overlaid. Error bars represent 95% confidence intervals.

Figure S3: Lagged association between 0 to 6‐month change in serum NfL z ‐score (sNfL) and subsequent 6 to 12‐month change in STEPS. Greater increases in sNfL were associated with subsequent declines in STEPS ( β = −141; 95% CI −302 to 21; p = 0.088), adjusted for age and sex. Points represent individuals; lines show model‐predicted values with 95% CIs.

Figure S4: Analysis of patient‐specific slopes for sNfL and STEPS. Separate age‐ and sex‐adjusted mixed‐effects models with random slopes for time were fit for sNfL z ‐score and STEPS, and patient‐specific slope estimates were extracted. Patient‐specific sNfL and STEPS slopes were modestly inversely correlated ( r = −0.18, p < 0.001). In linear regression, higher patient‐specific sNfL slopes were associated with lower patient‐specific STEPS slopes ( β = −618.90; 95% CI −910.42 to −327.38; p < 0.001), suggesting that individuals with greater increases in sNfL over time tended to show larger declines in STEPS.

Acknowledgements

The SPI2 study was sponsored and supported by MedDay Pharmaceuticals, although they were not involved in this post hoc analysis. The National Institutes of Health‐funded Eureka Research Platform (U2CEB021881) enabled step count collection.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

References

1 M. Ward and M. D. Goldman , “ Epidemiology and Pathophysiology of Multiple Sclerosis ,” Continuum (Minneapolis, MN) 28 ( 2022 ): 988 – 1005 . 10.1212/CON.0000000000001136 35938654

2 M. E. Baeva , I. Tottenham , M. Koch , and C. Camara‐Lemarroy , “ Biomarkers of Disability Worsening in Inactive Primary Progressive Multiple Sclerosis ,” Journal of Neuroimmunology 387 ( 2024 ): 578268 . 38157653 10.1016/j.jneuroim.2023.578268

3 M. S. Freedman , S. Gnanapavan , R. A. Booth , et al., “ Guidance for Use of Neurofilament Light Chain as a Cerebrospinal Fluid and Blood Biomarker in Multiple Sclerosis Management ,” eBioMedicine 101 ( 2024 ): 104970 . 38354532 10.1016/j.ebiom.2024.104970 PMC10875256

4 C. Barro , B. C. Healy , Y. Liu , et al., “ Serum GFAP and NfL Levels Differentiate Subsequent Progression and Disease Activity in Patients With Progressive Multiple Sclerosis ,” Neurology(R) Neuroimmunology & Neuroinflammation 10 ( 2023 ): e200052 . 36376097 10.1212/NXI.0000000000200052 PMC9749933

5 M. Khalil , C. E. Teunissen , S. Lehmann , et al., “ Neurofilaments as Biomarkers in Neurological Disorders ‐ Towards Clinical Application ,” Nature Reviews Neurology 20 ( 2024 ): 269 – 287 . 38609644 10.1038/s41582-024-00955-x

6 A. Abdelhak , C. Cordano , G. J. Duncan , et al., “ Markers of Axonal Injury in Blood and Tissue Triggered by Acute and Chronic Demyelination ,” Brain: A Journal of Neurology 148 ( 2025 ): 3011 – 3020 . 40278829 10.1093/brain/awaf144 PMC12316004

7 R. Aburashed , A. Eghzawi , D. Long , R. Pace , A. Madha , and J. Cote , “ Neurofilament Light Chain and Multiple Sclerosis: Building a Neurofoundational Model of Biomarkers and Diagnosis ,” Neurology International 17 ( 2025 ): 56 . 40278427 10.3390/neurolint17040056 PMC12029522

8 A. Abdelhak , P. Benkert , S. Schaedelin , et al., “ Neurofilament Light Chain Elevation and Disability Progression in Multiple Sclerosis ,” JAMA Neurology 80 ( 2023 ): 1317 – 1325 . 37930670 10.1001/jamaneurol.2023.3997 PMC10628837

9 E. Cantó , C. Barro , C. Zhao , et al., “ Association Between Serum Neurofilament Light Chain Levels and Long‐Term Disease Course Among Patients With Multiple Sclerosis Followed Up for 12 Years ,” JAMA Neurology 76 ( 2019 ): 1359 – 1366 . 31403661 10.1001/jamaneurol.2019.2137 PMC6692664

10 V. J. Block , A. Lizée , E. Crabtree‐Hartman , et al., “ Continuous Daily Assessment of Multiple Sclerosis Disability Using Remote Step Count Monitoring ,” Journal of Neurology 264 ( 2016 ): 1 – 11 . 27896433 10.1007/s00415-016-8334-6 PMC5292081

11 V. J. Block , R. Bove , C. Zhao , et al., “ Association of Continuous Assessment of Step Count by Remote Monitoring With Disability Progression Among Adults With Multiple Sclerosis ,” JAMA Network Open 2 ( 2019 ): e190570 . 30874777 10.1001/jamanetworkopen.2019.0570 PMC6484622

12 V. J. Block , M. Waliman , Z. Xie , et al., “ Making Every Step Count: Minute‐By‐Minute Characterization of Step Counts Augments Remote Activity Monitoring in People With Multiple Sclerosis ,” Frontiers in Neurology 13 ( 2022 ): 860008 . 35677343 10.3389/fneur.2022.860008 PMC9167929

13 V. J. Block , S. Cheng , J. Juwono , et al., “ Association of Daily Physical Activity With Brain Volumes and Cervical Spinal Cord Areas in Multiple Sclerosis ,” Multiple Sclerosis (Houndmills, Basingstoke, England) 29 ( 2022 ): 363 – 373 . 36573559 10.1177/13524585221143726 PMC9972237

14 K. B. Casaletto , J. Kornack , E. W. Paolillo , et al., “ Association of Physical Activity With Neurofilament Light Chain Trajectories in Autosomal Dominant Frontotemporal Lobar Degeneration Variant Carriers ,” JAMA Neurology 80 ( 2023 ): 82 – 90 . 36374516 10.1001/jamaneurol.2022.4178 PMC9664369

15 A. S. Gravesteijn , H. Beckerman , E. A. Willemse , et al., “ Brain‐Derived Neurotrophic Factor, Neurofilament Light and Glial Fibrillary Acidic Protein Do Not Change in Response to Aerobic Training in People With MS‐Related Fatigue ‐ a Secondary Analysis of a Randomized Controlled Trial ,” Multiple Sclerosis and Related Disorders 70 ( 2022 ): 104489 . 36621163 10.1016/j.msard.2022.104489

16 T. Berger , “ Biomarkers in the Evolution of Multiple Sclerosis ,” Neurodegenerative Disease Management 7 ( 2017 ): 3 – 6 . 29143589 10.2217/nmt-2017-0033

17 A. N. Nadkarni , K. J. Mukamal , X. Zhu , et al., “ Associations of Neurological Biomarkers in Serum With Gait Measures: The Cardiovascular Health Study ,” Journals of Gerontology, Series A, Biological Sciences and Medical Sciences 79 ( 2024 ): glae043 . 38334311 10.1093/gerona/glae043 PMC11005783

18 B. A. C. Cree , G. Cutter , J. S. Wolinsky , et al., “ Safety and Efficacy of MD1003 (High‐Dose Biotin) in Patients With Progressive Multiple Sclerosis (SPI2): A Randomised, Double‐Blind, Placebo‐Controlled, Phase 3 Trial ,” Lancet Neurology 19 ( 2020 ): 988 – 997 . 33222767 10.1016/S1474-4422(20)30347-1

19 A. J. Thompson , B. L. Banwell , F. Barkhof , et al., “ Diagnosis of Multiple Sclerosis: 2017 Revisions of the McDonald Criteria ,” Lancet Neurology 17 ( 2018 ): 162 – 173 . 29275977 10.1016/S1474-4422(17)30470-2

20 P. Benkert , S. Meier , S. Schaedelin , et al., “ Serum Neurofilament Light Chain for Individual Prognostication of Disease Activity in People With Multiple Sclerosis: A Retrospective Modelling and Validation Study ,” Lancet Neurology 21 ( 2022 ): 246 – 257 . 35182510 10.1016/S1474-4422(22)00009-6

21 V. J. Block , A. Lizée , E. Crabtree‐Hartman , et al., “ Continuous Daily Assessment of Multiple Sclerosis Disability Using Remote Step Count Monitoring ,” Journal of Neurology 264 ( 2017 ): 316 – 326 . 27896433 10.1007/s00415-016-8334-6 PMC5292081

22 R. W. Motl , L. A. Pilutti , Y. C. Learmonth , M. D. Goldman , and T. Brown , “ Clinical Importance of Steps Taken Per Day Among Persons With Multiple Sclerosis ,” PLoS One 8 ( 2013 ): e73247 . 24023843 10.1371/journal.pone.0073247 PMC3762863

23 Q. Ma , G. F. Dunton , and D. Hedeker , “ Negative Binomial Mixed Effects Location‐Scale Models for Intensive Longitudinal Count‐Type Physical Activity Data Provided by Wearable Devices ,” Biometrics 81 ( 2025 ): ujaf099 . 40742447 10.1093/biomtc/ujaf099 PMC12312402

24 S. Sahoo , C. Harper , A. K. Y. Tsui , et al., “ Serum Neurofilament Light Chain as a Biomarker in Multiple Sclerosis: A Cross‐Sectional Observation in Real‐World Clinical Practice ,” Multiple Sclerosis and Related Disorders 103 ( 2025 ): 106637 . 40840335 10.1016/j.msard.2025.106637

25 C. Ammitzbøll , T. B. Dyrby , L. Börnsen , et al., “ NfL and GFAP in Serum Are Associated With Microstructural Brain Damage in Progressive Multiple Sclerosis ,” Multiple Sclerosis and Related Disorders 77 ( 2023 ): 104854 . 37418931 10.1016/j.msard.2023.104854

26 M. D. Goldman , S. Chen , B. Kunisetty , J. M. Gelfand , B. A. C. Cree , and V. J. Block , “ Validating Wearable Step Counts in Multiple Sclerosis Research: A Replication Study ,” Frontiers in Neurology 16 ( 2025 ): 1709389 . 41245861 10.3389/fneur.2025.1709389 PMC12611647

27 A. Abdelhak , G. Cerono , F. Sheikhzadeh , et al., “ Myelin Injury Precedes Axonal Injury and Symptomatic Onset in Multiple Sclerosis ,” Nature Medicine 32 ( 2026 ): 362 – 368 . 10.1038/s41591-025-04014-w PMC13052337 41115958

28 U. Dalgas , M. Langeskov‐Christensen , E. Stenager , M. Riemenschneider , and L. G. Hvid , “ Exercise as Medicine in Multiple Sclerosis‐Time for a Paradigm Shift: Preventive, Symptomatic, and Disease‐Modifying Aspects and Perspectives ,” Current Neurology and Neuroscience Reports 19 ( 2019 ): 88 . 31720862 10.1007/s11910-019-1002-3

29 R. W. Motl and L. A. Pilutti , “ The Benefits of Exercise Training in Multiple Sclerosis ,” Nature Reviews Neurology 8 ( 2012 ): 487 – 497 . 22825702 10.1038/nrneurol.2012.136

30 R. W. Motl , B. Fernhall , K. K. McCully , et al., “ Lessons Learned From Clinical Trials of Exercise and Physical Activity in People With MS—Guidance for Improving the Quality of Future Research ,” Multiple Sclerosis and Related Disorders 68 ( 2022 ): 104088 . 35987111 10.1016/j.msard.2022.104088

31 R. Kapoor , K. E. Smith , M. Allegretta , et al., “ Serum Neurofilament Light as a Biomarker in Progressive Multiple Sclerosis ,” Neurology 95 ( 2020 ): 436 – 444 . 32675076 10.1212/WNL.0000000000010346 PMC7538221

32 M. Langeskov‐Christensen , L. Grøndahl Hvid , M. K. E. Nygaard , et al., “ Efficacy of High‐Intensity Aerobic Exercise on Brain MRI Measures in Multiple Sclerosis ,” Neurology 96 ( 2021 ): e203 – e213 . 33262230 10.1212/WNL.0000000000011241

33 G. Disanto , C. Barro , P. Benkert , et al., “ Serum Neurofilament Light: A Biomarker of Neuronal Damage in Multiple Sclerosis ,” Annals of Neurology 81 ( 2017 ): 857 – 870 . 28512753 10.1002/ana.24954 PMC5519945

34 C. Barro , P. Benkert , G. Disanto , et al., “ Serum Neurofilament as a Predictor of Disease Worsening and Brain and Spinal Cord Atrophy in Multiple Sclerosis ,” Brain: A Journal of Neurology 141 ( 2018 ): 2382 – 2391 . 29860296 10.1093/brain/awy154