The development of new outcome measures with high sensitivity in capturing time-related changes may improve the design of clinical trials allowing the reduction of treatment time and population size. In the ataxia field, the use of wearable accelerometers has been employed to capture specific gait and posture characteristics that could serve as quantitative ataxia biomarkers (Supplementary Table S1). Wearable sensors have been primarily tested in cross-sectional studies conducted in laboratory or in-clinic setting to assess gait and balance characteristics in a controlled environment (Schmitz-Hübsch et al., 2016; Terayama et al., 2018; Caliandro et al., 2019; Ilg et al., 2020; Shah et al., 2021; Velázquez-Pérez et al., 2021; Kadirvelu et al., 2023). Sensors were used to analyse several features from body movements, like gait spatial-temporal parameters, body sway, velocity and angles of body movements. These studies showed that wearable sensor technologies may enable to capture specific features differentiating patients from healthy controls and measures were correlated with clinical scales (Ilg et al., 2020; Shah et al., 2021; Kadirvelu et al., 2023).

Another approach is to use wearable sensors to assess physical activity performed by the subjects during free living conditions. This type of remote monitoring allowed the collection of large sets of data with continuous activity recording for several days (Subramony et al., 2012; Milne et al., 2021; Mueller et al., 2021; Khan et al., 2022; Eklund et al., 2023; Gupta et al., 2023). Several cross-sectional studies, applying different techonolgies and protocols, collected information regarding patient’s ability to perform daily physical activity, energy expenditure, and numbers of steps. Two studies tested wearables both in-clinic and in a free-living environment (Ilg et al., 2020; Thierfelder et al., 2022). Patients were required to perform a defined in-clinic protocol and then to wear the devices in the subsequent few hours in a free-living context. The results of these studies showed a high correlation between laboratory-based and free-walking conditions, with the highest effect size observed in real-life walking (Ilg et al., 2020; Thierfelder et al., 2022).

Despite the large number of cross-sectional studies, only few longitudinal studies have been performed in patients with ataxia (Milne et al., 2021; Gupta et al., 2023). Longitudinal data are essential to support the use of wearable technologies in therapeutical clinical trials to complement traditional outcome measures in quantifying treatment effect on patient daily life.

Here we present a 12-month longitudinal study monitoring real-life physical activity in patients with Friedreich ataxia (FRDA). FRDA is an autosomal recessive multi-systemic neurodegenerative disorder that represents the most frequent hereditary ataxia in the Caucasian population. The disease is caused in >95% of cases by a repeated GAA expansion in intron 1 of the frataxin (FXN) gene that encodes for the frataxin protein. The patients usually manifest gait instability, limb incoordination and dysmetria, before the age of 25 years (Reetz et al., 2015). The first treatment for patients with FRDA, Omaveloxolone, was approved by FDA in 2023 (Subramony and Lynch, 2023), and additional other compounds and different therapeutical strategies are currently being tested. The response to treatments in clinical trials is currently measured using neurological scales, such as the Scale for the Assessment and Rating of Ataxia (SARA; Schmitz-Hübsch et al., 2006) and the Friedreich’s Ataxia Rating Scale (FARS; Rummey et al., 2019). At present, only one previous longitudinal study with wearable sensors has been performed in FRDA (Milne et al., 2021).

In our prospective study, we evaluated a population of patients with FRDA with the following aims: (i) assess feasibility of 1-week remote free-living monitoring, (ii) detect effective changes in accelerometer-based measures at 1-year interval, (iii) identify useful measures of disease progression, and (iv) evaluate possible correlations between accelerometer-derived measures and clinical scores.

Disorders of the cerebellum cause several neurological motor signs such as disequilibrium, limb incoordination and speech difficulties. The severity of motor impairment is currently evaluated and graded by the scores obtained using specific clinical scales or functional tests measuring walking speed or manual abilities. Recently several devices and new tools have been developed to obtain quantitative measures that could improve the accuracy of the scales and tests performed in a clinical setting during the neurological examination, and allow the monitor of a larger time frame of the patient real life (Minen and Stieglitz, 2021). Some of these assessments may be performed remotely and digital devices may record the patient activities in a home environment. These new tools are of great interest for establishing the impact of potential therapeutic intervention of the real-life activity of patients with movement disorders (Lipsmeier et al., 2022). To assess the use of accelerometer devices as effective outcome measures in interventional trials it is important to evaluate feasibility and reliability of activity measurements in tracing longitudinal changes during disease progression.

Remote real-life physical monitoring has been previously assessed in ataxic patients both in FRDA (Mueller et al., 2021) and in other types of ataxias (Khan et al., 2022; Eklund et al., 2023; Gupta et al., 2023) (Supplementary Table S1). Mueller et al. tested different home-based digital endpoints evaluating speech, hand function and gait (Mueller et al., 2021) in a cross-sectional study versus controls. Several parameters discriminated FRDA patients from controls, including activity monitoring parameters recorded from feet and wrist sensors, but not from trunk (Mueller et al., 2021). In a second study, Milne et al. performed a 1-year longitudinal study including measures assessing gait, balance and remote activity monitoring. A significant change in activity duration, daily step count and distance walked compared to baseline was observed after 6 months, while significant decline was present only in daily step count and distance walked after 12 months (Milne et al., 2021).

We decided to analyse very simple accelerometer-derived measures: the overall level of physical activity, from sedentary to vigorous activity, the VM3 indicating the total amount of movement in the three-dimensional space (3 axes), and the step counts per minute. Each study participant worn two Actigraph sensors, one at waist and one at wrist of the non-dominant hand. As expected, overall activity monitoring showed less and lighter movement activities in subjects with FRDA than in healthy controls. Patients were significantly more sedentary than controls, and were also less likely to perform sustained physical activities (i.e., activity bouts) compared to controls. Only light activity calculated from wrist sensors was similar in the two groups of participants. While differences in activity measures between patients and controls were expected, our major interest was to observe longitudinal data in the patient group, and test the hypothesis that real-life activity could reliably capture changes related to disease progression in 1-year interval.

We observed greater levels of activity at wrist-recorded sensors in comparison with waist sensors. In a previous study, the same observation in ataxic patients has been interpreted as an overestimation of activity related to patient’s upper limb dysmetria and tremor (Mueller et al., 2021). In the present study, the increased movement activity at upper-limb was not disease-specific but was accounted in both patient and control participants, suggesting a physiological occurrence of more activity at upper limbs than at the trunk level.

Activity monitoring in ataxic patients has been tested with different type of sensors that have been have been placed at dominant wrist (Khan et al., 2022; Gupta et al., 2023), ankle (Subramony et al., 2012), dominant wrist and ankle (Eklund et al., 2023), triceps (Milne et al., 2021), feet, trunk and wrists (Mueller et al., 2021) (Supplementary Table S4). Some authors have compared the performances of sensors according to placement, with different results. Mueller and co-authors (Mueller et al., 2021) found that trunk sensor did not discriminate between FRDA patients and controls, while differences between disease and control groups were better captured using feet and wrist-worn sensors.

Digital measures were significantly related with ataxia rating scales (Subramony et al., 2012; Mueller et al., 2021; Gupta et al., 2023) and with patient perception of disability measured via a patient reported questionnaire (Eklund et al., 2023). We confirm that activity measures in FRDA are highly correlated with disease duration, clinical scales, and functional tests, while are not influenced by demographic variables, such as age, sex and BMI.

At 1-year interval, all sensor-derived measures (both from waist- and wrist-worn sensors) showed a trend toward reduced activity in patients, and remained unchanged in controls. After correction for multiple comparisons, we demonstrated a significant effect of time in VM3, sedentary and light activity time (waist) and MVPA (wrist). VM3 measures had lower variability compared to the other activity measures, and were associated with the highest effect size. Milne and colleagues described a reduced activity in terms of physical activity duration, step count and distance walked in FRDA patients after a 6-month interval. However, the same authors reported that at 12 months-follow-up only step count continued to show a progressive decline, suggesting that clinical assessment with FARS outclassed digital measures in the sensitivity to longitudinal changes (Milne et al., 2021). Gupta and colleagues (Gupta et al., 2023), conducted a longitudinal study in patients with Ataxia-telangiectasia (A-T) showing reduced activity indexes after a 1-year interval. These authors provided a novel approach in the analyses of the data allowing the extraction of characteristic movements from wrist sensor and identifying several sub-movement features that distinguished patients from controls. Although the recorded movements were highly disease specific, their measures showed no substantial changes at 1-year interval.

In our study we assessed physical activity during daily life, not limited to a specific task or setting. Our results suggest that the reduction in physical activity observed at follow-up in FRDA patients is mostly related to a worsening in gait performances, as showed by the presence of significant correlation of the change in SARA gait sub-score with VM3 and sedentary time. A longitudinal study in amyotrophic lateral sclerosis, that adopted the same monitoring setting of our study, showed that VM3 was the most reliable digital indicator of functional progression (van Eijk et al., 2019). In our study, we confirmed that VM3 measure represent the most valuable digital measure for future clinical trials, as showed by the highest effect size (Table 2).

As new treatments are becoming available for FRDA (Subramony and Lynch, 2023), it is crucial to link the improvement observed during clinical assessments with the improvements in patients’ ability to perform daily activities. The activity measures presented here, such as the time spent in different levels of physical activities and step count, could represent sensitive measures linked to day-to-day functioning of patients. Importantly, we used waist- and wrist-worn sensors to collect data from ambulatory patients, however the same protocol could also apply to patients with reduced or absent ambulatory capacity.

Using wearable devices during real-life activities comes with some limitations. Due to the free-living context where these data were collected, the exact nature of the activities recorded is unknown. Moreover, the analysis of activity was based on proprietary software (Actilife) that lacks an algorithm specific for the ataxic population, over- or under-estimating performances and introducing other possible bias. Due to the limited number of subjects, the present work has to be considered as a pilot exploratory study. Possibly a larger number population size would be needed to prove that the observed longitudinal changes are above measurement noise (Supplementary Table S1). Our data, however, suggest that VM3 could represent a viable solution to assess total level of physical activity, as it represents the raw value of the vector recorded by the devices, and does not introduce bias related to post-processing.

In conclusion, our results indicate that activity monitoring using Actigraph devices represent a well-tolerated and sensitive tool for individuals with FRDA. Based on the longitudinal data and the preliminary results on the test-retest-reliability, the waist sensors seem to be more reliable than wrist sensors. In addition waist sensor measures showed good correlations with. clinical scale and daily life activity rating scores, supporting the usefulness of waist-based accelerometer to measure activity in future clinical trials in FRDA.