Objective Monitoring of Motor Symptom Severity and their Progression in Parkinson’s Disease Using a Digital Gait Device

Objective Monitoring of Motor Symptom Severity and their Progression in Parkinson’s Disease Using a Digital Gait Device | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Objective Monitoring of Motor Symptom Severity and their Progression in Parkinson’s Disease Using a Digital Gait Device Tamara Raschka, Jackrite To, Tom Hähnel, Stefano Sapienza, Alzhraa Ibrahim, and 7 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4521747/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 15 Jul, 2025 Read the published version in Scientific Reports → Version 1 posted 14 You are reading this latest preprint version Abstract Digital technologies for monitoring motor symptoms of Parkinson’s Disease (PD) have underwent a strong evolution during the past years. Although it has been shown for several devices that derived digital gait features can reliably discriminate between healthy controls and people with PD, the specific gait tasks best suited for monitoring motor symptoms and especially their progression, remain unclear. Furthermore, the potential benefit as endpoint in a clinical trial context has not been investigated so far. In this study we employed a digital gait device manufactured by Portabiles HCT, which has been used by 339 patients within the LuxPark cohort (n = 161, Luxembourg) as well as within routine clinical care visits at the University Medical Center Erlangen (n = 178, Erlangen, Germany). Linear (mixed) models were used to assess the association of task-specific digital gait features with disease progression and motor symptom severity measured by several clinical scores. Furthermore, we employed machine learning to evaluate whether digital gait assessments were prognostic for patient-level motor symptom progression. Overall, digital gait features derived from Portabiles digital gait device were found to effectively monitor motor symptoms and their longitudinal progression. At the same time the prognostic performance of digital gait features was limited. However, we could show a strong reduction in required sample size, if digital gait features were employed as surrogates for traditional endpoints in a clinical trial context. Thus, Portabiles digital gait device provides an effective way to objectively monitor motor symptoms and their progression in PD. Furthermore, the digital gait device bears strong potential as an alternative and easily assessable endpoint predictor in a clinical trial context. Biological sciences/Neuroscience/Diseases of the nervous system/Parkinsons disease Health sciences/Biomarkers Health sciences/Diseases/Neurological disorders/Neurodegenerative diseases/Parkinsons disease Health sciences/Medical research/Biomarkers Figures Figure 1 Figure 2 Figure 3 Introduction Gait abnormalities in people with Parkinson’s Disease (PD) are a major motor symptom, along with tremor, bradykinesia, and rigidity, and have a major impact on patients’ mobility and quality of life [ 1 – 3 ]. Signs of impaired gait patterns are reduced gait speed and step length, increased axial rigidity, impaired rhythmicity, and freezing of gait [ 2 , 4 ]. In the early stages of the disease, slower gait speed, shortened step length, and reduced arm swing during walking are the most common signs of gait abnormalities, while in the later stages, movements become even more slow, gait shuffling and freezing of gait appears and patients are thus prone to falls [ 4 ]. With progression into an advanced stage, gait abnormalities, including balance and postural control, become worse [ 4 ]. Thus, investigating patient’s gait impairment is crucial for evaluating the stage of disease progression. Furthermore, objective monitoring of gait could help defining an optimal schedule of when to treat a patient, whether to adjust the dose or whether to switch from one treatment to another. In the past, digital technologies for monitoring motor symptoms of PD have evolved from non-portable in-lab devices to wearable sensors and at-home phone applications [ 5 ]. While each technology has its advantages, wearable sensors are able to measure gait parameters like stride length, gait speed, or the lateral excursion of each patient quantitatively and potentially also in an outpatient setting, while being portable, light weight, less expensive, and easy to use, for patients, as well as, for clinicians [ 5 ]. However, before any application in clinical care, such technologies must be carefully assessed regarding technical robustness as well as medical diagnostic validity [ 6 ]. Importantly, a digital gait sensor can only reliably be used in routine medical care after regulatory approval, requiring proof of its safety. Besides that, trustable digital gait features require clear evidence that these sensor measurements correlate significantly with established clinical outcomes such as MDS-Unified Parkinson’s Disease Rating Scale (MDS-UPDRS) [ 7 ]. Although digital gait features have been explored in several studies in PD research, most studies solely focus on distinguishing PD patients from healthy controls in cross-sectional manner [ 5 , 8 – 13 ]. Some studies also classify patients into multiple disease stages [ 14 – 17 ], but studies rarely describe the associations between PD progression and gait directly [ 4 – 6 ]. Furthermore, there is a lack of studies investigating the reproducibility of findings from the same gait device across several PD cohorts. Taken together, it is still unclear to date whether digital gait sensors can reproducibly and objectively monitor motor symptoms and their progression over time and whether in that regard there is a demonstrable benefit to classical clinical scales. The ERA PerMed funded EU-wide project DIGIPD (Validating DIGI tal biomarkers for better personalized treatment of P arkinson's D isease— https://www.digipd.eu ) tries to fill a gap by evaluating the potential benefit of digitally assessed PD symptoms compared to established clinical outcomes across several cohorts in Europe. In this study, we focused on a sensor-based digital gait device clipped to the shoe of a patient while walking ( https://www.portabiles-hct.de/en/product/ ). The device is currently registered for regulatory approval at the US Food and Drug Administration (FDA) and a certified class I medical device (MDR) with a CE mark in Europe. We investigated data recorded by the device in two independent longitudinal cohorts (one from Germany, one from Luxembourg) and while performing different gait tasks in the clinic. More specifically, we initially modelled PD progression based on several established outcome measures derived from the MDS-UPDRS by fitting a latent time joint mixed-effect model (LTJMM) [ 18 ] to longitudinal data from both studies. This allowed us to estimate each patients’ true rate of progression and to align disease trajectories on a common disease timescale, taking into account the fact that patients may have progressed differently at baseline. Following that, linear (mixed) models were used to evaluate the association of digital gait features with estimated progression rates as well as UPDRS derived outcomes. In that regard we also explored, in how far a machine learning model could forecast an individual’s disease progression rate solely based on digital gait features. Finally, we investigated on the potential benefit of using a digital gait predicted UPDRS III compared to the original UPDRS III as an endpoint in a clinical trial. Material and Methods Cohorts We analyzed data from two independent PD cohort studies: (1) the Luxembourg Parkinson’s Study (LuxPARK, NCT05266872) [ 19 ] and (2) from the University Medical Center Erlangen, Germany (Erlangen). LuxPARK is an ongoing observational cohort of patients at all disease stages. Data was collected periodically up to four years of follow-up between 2015 and 2022. Data from Erlangen contained information from patients at all disease stages collected during multiple irregular clinical routine visits between 2011 and 2022. For both studies, we used multiple clinical scores as outcomes in our analysis. In LuxPARK these scores were MDS-Unified Parkinson's Disease Rating Scale (MDS-UPDRS) I, MDS-UPDRS II, MDS-UPDRS III [ 7 ], Tremor dominance (TD) score [ 20 ], Postural Instability and Gait Dysfunction (PIGD) score [ 20 ], and an axial score. The axial score was calculated by summing MDS-UPDRS III sub-items 1 (speech), 2 (facial expression), 9 (arising from chair), 10 (gait), 11 (freezing of gait), 12 (postural instability), and 13 (posture), which is a customized version of the axial score described in [ 21 ] to integrate axial changes in the face and freezing of gait more comprehensively. While LuxPARK employed the MDS-UPDRS, the Erlangen data assessed PD symptoms of patients with the original UPDRS [ 22 ]. Differences are described in the MDS-UPDRS publication [ 7 ]. For the sake of simplicity, we refer to both, the MDS-UPDRS, as well as the original UPDRS, as UPDRS. The Erlangen data evaluated less clinical scores and thus, only UPDRS III and the axial score (without freezing of gait, as this was not part of the UPDRS, but MDS-UPDRS only) were used as outcomes. For both datasets, patients were selected, such that each patient had data from at least two visits for each outcome. This resulted in 612 PD patients from the LuxPARK cohort and 264 PD patients from the Erlangen cohort. Characteristics of these patients can be found in Supplementary Table S1 in detail. Gait Measurements Additional to the traditional clinical outcome scores described above, gait measurements using Portabiles digital gait device were available in both, LuxPARK and Erlangen. While in LuxPARK only one digital gait assessment, at a single visit, was performed, multiple digital gait assessments over time were available for the Erlangen patients (2.84 ± 2.37 visits per patient on average, total: 506 gait visits). Digital gait assessments in Erlangen were collected irregularly, similar to the clinical data (difference between first and second digital gait assessment: 503.11 ± 488.98 days, median = 364 days). Digital gait data was available for 343 (LuxPARK) and 802 (Erlangen) patients for multiple exercises such as Time-Up-and-Go (TUG), the Turn, manual TUG (Tray), and cognitive TUG (Count), as well as the 2x10m, and 4x10m walking tasks. A detailed description of the different tasks, including differences between Erlangen and LuxPARK, can be found in the Supplementary Material . All analysis regarding digital gait were conducted only on the subset of patients for which both longitudinal clinical features (as described above), as well as, digital gait features were available. In total these were 161 patients from LuxPARK and 178 patients from Erlangen. Characteristics of these patients can be found in Table 1 . A detailed description of the used device, the raw signal processing, and the derived features for further analysis along with a visualization of their distribution can be found in the Supplementary Material ( Supplementary Table S2 , Supplementary Figure S1 , Supplementary Figure S2 ). A data catalogue visualizing all clinical, as well as, digital gait features of both cohorts can be found here: https://p-data.molekulare-neurologie.uk-erlangen.de/ . Table 1 Demographic and clinical characteristics of patients Patients characteristics for LuxPARK and Erlangen datasets. Only the subset used for the gait analysis is described here. These are all patients having longitudinal clinical features plus digital gait features. Mean and standard deviations are shown for all characteristics but sex and Hoehn & Yahr stages, where absolute (and relative) values are shown. Values for disease duration, age, and clinical scores are reported at baseline. PIGD, TD, UPDRS I, and UPDRS II are not available in the Erlangen cohort. LuxPARK Erlangen Number of patients 161 178 Number of visits 5.32 ± 1.53 12.08 ± 9.12 Disease duration, years 4.92 ± 5.29 5.31 ± 4.57 Age, years 64.71 ± 10.08 62.39 ± 10.37 Sex Male 117 (73%) 110 (62%) Female 44 (27%) 68 (38%) Axial score 5.71 ± 3.79 5.35 ± 3.93 PIGD 0.56 ± 0.58 N/A TD 0.5 ± 0.41 N/A UPDRS I 9.28 ± 5.68 N/A UPDRS II 10.32 ± 6.8 N/A UPDRS III 29.89 ± 14.2 20.2 ± 11.89 H&Y 0 1 1 1 19 25 1.5 11 14 2 93 35 2.5 20 18 3 12 35 4 4 11 5 1 2 N/A - 37 Modeling approaches Latent time joint mixed-effect model The latent time joint mixed-effect model (LTJMM) was first published by Li et al. [ 18 ]. Its rational is to align trajectories of patients on a common latent (i.e. unobserved) disease timescale using a multivariate linear mixed effects model. This model is defined by fixed and random effects describing the deviation of each patient from a “mean” reference trajectory, both, on the time axis, as well as with respect to the actual outcome. The model can thus describe piecewise linear progression of multiple clinical outcomes over time by estimating how far the timescale of each patient is shifted away from the timescale of the reference. In our case the timescale was the reported time since PD diagnosis. The inferred patient-specific time shift can be seen as an estimate how far a particular patient is advanced in the severity of symptoms compared to the other patients in the same cohort. Additionally, the model generates an estimate of the patient-specific slope of disease symptoms, meaning that a large “random” slope indicates a higher speed of symptom progression compared to all other patients in the same cohort. Used outcomes were in our case the UPDRS I-III, TD, PIGD, and axial score, depending on the availability in the two different datasets. Additionally, we adjusted for age at diagnosis, sex and medication status (ON/OFF) as fixed effects. All longitudinal data, irrespective of gait data availability, was used. Mathematical details of the model are described in the Supplementary Material . Validation of LTJMM model Distributions of Hoehn and Yahr (H&Y) stages were examined to validate the alignment of disease trajectories on a common disease timescale. This was based on the assumption that the shifts of each individuals’ trajectory should follow an order, which roughly agrees with the H&Y stages. Patients with lower H&Y stages, who are at the beginning of the development of the disease, should be shifted further to the left on the common disease time scale than those with a higher HY stage who are more advanced in their disease. Additionally, the accuracy of the LTJMM approach was checked by calculating the correlation between real and predicted outcomes at the last observed visit, similar to [ 23 ]. For that, a LTJMM was fitted on all outcomes, but the last measurements of each patient were excluded and predicted back by the LTJMM. Linear mixed models to evaluate gait parameters Linear mixed models (LMMs) were used for associating digital gait features with different outcomes at the same visit. Therefore, clinical data at the same day of gait data assessment was matched with gait data. Only data from patients having both longitudinal clinical features, as well as, digital gait features available were used in this analysis. Characteristics of this patient subset is shown in Table 1 . Modeled outcomes are patient-specific a) latent disease time, b) clinical outcome scores, and c) progression rate. Both latent disease time and progression rate are a direct outcome of the LTJMM model described above. When modeling patient-specific latent disease time (interpreted as estimated time since disease onset), age at diagnosis and sex were used as adjustment covariates. In case of modeling clinical outcome scores or the patient-specific progression rate, we employed the latent disease time elapsed at the time of the clinical assessment as adjustment covariate to adjust for temporal differences between observed patient trajectories. Furthermore, we added a patient-specific random intercept in case of the Erlangen data, as digital gait features were available at several time points. Different sets of digital gait features were integrated as features in the model. First, all digital gait features were used jointly ( allGait ), then features were selected according to the individual gait tasks, and lastly, we fitted LMMs for each digital gait feature separately ( singleGait ). All these models were tested against a null model containing only adjustment covariates via a likelihood ratio test. Results were adjusted for multiple testing with the Benjamini & Yekutieli method [ 24 ]. Simulation of a randomized controlled study A randomized controlled trial was simulated over a one-year observation period with either original or gait predicted UPDRS III as study endpoints. The gait predicted UPDRS III was estimated from 4x10m tests’ digital gait features using a linear mixed effect model. We assumed visits every 60 (bi-monthly), 30 (monthly) or 7 (weekly) days to assess the sample sizes needed for evaluating the efficacy of a potentially disease-modifying drug. We employed the Erlangen data for this task, as longitudinal digital gait features were required. Simulations were inspired by an ongoing trial [ 25 ] and assumed a treatment effect on disease progression of 30%. Control and treatment groups were equally sized without different treatment dosage arms. A power of 80%, and a significance level of 0.01 was set as parameters for the outcome simulation of the disease progression difference between treatment and control group. Power and sample size was calculated using linear mixed-effects model, based on the method from Edland [ 26 ], implemented in the R package longpower [ 27 ]. Predictive ML models for disease progression Random Forest [ 28 ], XGBoost [ 29 ] and Lasso [ 30 ] were used to predict the LTJMM random slope from digital gait features. In Erlangen cohort, where multiple gait visits were conducted, measurements from first and first plus second gait visit were used for prediction. Multiple feature sets were tested: only sex and age, all digital gait features ( allGait ), and task-specific digital gait feature sets. Implementations of algorithms of scikit-learn [ 31 ] and xgboost [ 29 ] packages were used. Hyperparameter optimization was performed using a randomized search within an inner 5-fold cross validation procedure. Hyperparameter spaces for the different algorithms can be found in the Supplementary Table S3 . Performance metrics were evaluated in an outer 5-fold repeated cross validation procedure. Performance of the models including gait features was additionally tested with a Wilcoxon signed-rank sum test comparing its \({R}^{2}\) values to those from a model integrating only sex and age as predictors. The best performing model was selected based on these statistical test results. Results Overview Disease trajectories of both studies, LuxPARK and Erlangen, were first aligned on a common disease timescale via LTJMM (Fig. 1 A ) with the aim to adjust trajectories of each patient according to their disease status and to remove bias introduced by including patients from different stages of Parkinson’s disease. After aligning the disease trajectories on that common disease timescale, digital gait features were used as predictors in multiple settings, either for monitoring the disease severity or the disease progression (Fig. 1 B ) and additionally for predicting disease severity and progression (Fig. 1 C ) . Latent time joint mixed-effect models for alignment of disease trajectories LTJMM models were fitted to Erlangen and LuxPARK data. Both the original trajectories, as well as, the trajectories on the common disease timescale for each study and outcome are shown in the Supplementary Figure S3 and Supplementary Figure S4 . Supplementary Figure S5 and Supplementary Table S4 confirms that the model orders patient trajectories correctly according to the H&Y stages. Additionally, on average Pearson correlations of real and predicted clinical outcomes were ρ = 0.73 for Erlangen and ρ = 0.42 for LuxPARK, indicating a sufficient fit of the LTJMM models. Monitoring disease stage, motor symptom severity and motor symptom progression through digital gait Statistical associations of clinical outcomes with features derived from the Portabiles digital gait device were analyzed with linear (mixed) models after aligning observed clinical outcome trajectories on a common disease timescale. We found a significant association of digital gait features with traditional clinical outcome measures in both datasets (LuxPARK: axial score, PIGD, TD and UPDRS III; Erlangen: axial score, UPDRS III) (Table 2 and Table 3 ). Clear task-specificity can be observed in LuxPark cohort, with the TUG being the most informative task. Here, the axial score was found to be highest correlated with digital gait features. When modeling the axial score with digital gait features, stance time (Count: p = 0.04), swing time (Tray: p = 0.009), gait speed (Tray: p = 0.029), stride length (Tray: p = 0.015) and the landing impact (TUG: p = 0.047) were found to have significant effect sizes in the models. TUG’s landing impact shows also significant effect size when modelling the PIGD score (p = 0.048). For the PIGD score as model outcome, we have also found the gait speed (p = 0.014) and stride length (p = 0.010) significant for the Count task. Similar as in Tray task in LuxPark, we could find the swing time in Erlangen’s 4x10m task significant (p = 0.045) when modelling the axial score as outcome. Table 2 Associations of digital gait features with dedicated clinical scores, their slope and latent time in LuxPARK cohort. Adjusted p-values of likelihood ratio testing the association of multiple digital gait feature sets, either all measured digital gait features or task-specific (Count, Tray, TUG, Turn), with clinical outcomes, adjusted for age at diagnosis and sex. Significant (adjusted p < 0.05) tests are marked in bold and weakly significant (adjusted p < 0.1) tests in italic. Measured outcomes are the clinical scores (top), their slope (i.e. progression, middle) and latent time (i.e. disease stage, bottom). Outcome All Gait Count Tray TUG Turn Clinical score Axial score 0.0055 0.0006 0.0006 0.0001 0.0175 PIGD 0.2152 0.0068 0.0068 < 0.0001 0.0629 TD 0.0178 0.4882 0.4882 1 1 UPDRS I 0.2717 0.0783 0.2384 0.2504 0.0783 UPDRS II 0.4154 0.1427 0.1427 0.1427 0.3601 UPDRS III 0.0337 0.2149 0.0538 0.0476 0.1058 Slope (progression) Axial score 0.2765 0.5264 0.0534 0.0990 0.4970 PIGD 0.6788 1 1 1 1 TD 0.4895 1 0.1709 0.1709 0.6128 UPDRS I 0.3035 0.8514 0.0621 0.0621 0.5288 UPDRS II 0.3864 0.9143 0.1260 0.0686 0.4988 UPDRS III 0.6779 1 1 1 1 Latent time (disease stage) Latent time 0.001 0.0288 0.0726 0.0005 0.0023 Table 3 Associations of digital gait features with dedicated clinical scores, their slope and latent time in Erlangen data. Adjusted p-values of likelihood ratio testing the association of multiple digital gait feature sets, either all measured digital gait features or task-specific ( TUG, 2x10m With Stop, 4x10m Without Stop) with clinical outcomes, adjusted for age at diagnosis and sex. Significant (adjusted p < 0.05) tests are marked in bold and weakly significant (adjusted p < 0.1) tests in italic. Measured outcomes are the clinical scores (top), their slope (i.e. progression, middle) and latent time (i.e. disease stage, bottom). Outcome All Gait TUG 2x10m With Stop 4x10m Without Stop Clinical score UPDRS III 0.1574 0.0361 0.0203 0.0110 Axial score 0.0010 0.0068 0.0068 0.0068 Slope (progression) UPDRS III 1 1 1 1 Axial score 1 1 1 < 0.0001 Latent time (disease stage) Latent time 0.0047 < 0.0001 0.0001 < 0.0001 Furthermore, a significant association with the disease stage, represented via the latent time, could be observed in LuxPARK, as well as in Erlangen (Table 2 and Table 3 ). While we could find the maximal sensor lift significant in LuxPARK for both the TUG (p = 0.031) and Turn (p = 0.048) task, the stride time in 2x10m (p = 0.001) and TUG (p = 0.045) task was the most informative digital gait feature in Erlangen. Estimates for each of the models can be found in Supplementary Table S6 . A visualization of those, also showing the significance, can be found in the Supplementary Information . The association with the slope of the axial score (i.e. progression) was clearly significant in the Erlangen data for the 4x10m test and demonstrated a weak significance in LuxPARK for the Tray and TUG test. Further weakly significant associations with digital gait included the slope of the UPDRS I (Tray, TUG tests) and the slope of the UPDRS II (TUG test). Altogether, significant associations of digital gait features were thus identifiable with traditional clinical motor scores, their slopes as well as disease stage. Results of single digital gait features tests can be found in Supplementary Table S5 , generally supporting the findings reported in this section. Machine learning based prediction of motor symptom progression with digital gait We trained different machine learning algorithms to predict the patient-specific slope (i.e. progression) of motor symptom related clinical outcomes. The strongest results could be observed via the Random Forest algorithm in a 10-fold cross-validation setup (see Supplementary Figure S7 ) in the Erlangen data. When using only the first gait assessment in the Erlangen data for the prediction of the axial scores’ slope, a significant improvement compared to only using age and sex as predictors could be observed, specifically when using data from the 4x10m task (Fig. 2 ). Integrating also a second assessment of gait improved the prediction even more and specifically when using data from all gait tasks. Additionally, the UPDRS III slope could now be predicted significantly better compared to only using age and sex as covariates. It should be noted that similar findings could not be made in LuxPARK, because digital gait features were only available at a single visit (see Supplementary Figure S6 ). Altogether, our findings reported in this section support those made in the previous section, namely that digital gait is associated with traditional motor symptom scores from the UPDRS and is sensitive to changes in those scores. Randomized controlled trial simulation shows benefit of digital gait features as endpoint To further evaluate the potential benefit of digital gait features compared to traditional clinical outcome scores, we simulated a longitudinal clinical trial with different visit frequencies. Notably, digital gait features can be collected more easily and cheaper at higher frequency compared to traditional clinical outcome scores, and data collection could potentially even be performed within a patient’s home environment. Estimates of the UPDRS III from digital gait features correlated well with the original data (ρ = 0.77, see Supplementary Figure S8 ). According to our statistical sample size calculation, which assumes a treatment effect size of 30%, bi-monthly assessments of the UPDRS III require 690 patients in total to reach 80% statistical power at a significance level of 1%. At this visit frequency the same sample size of 690 patients would be needed for a digital endpoint estimating the UPDRS III from digital gait features via the 4x10m task (Fig. 3 ). However, already a monthly assessment of this gait predicted clinical endpoint can reduce the number of patients to 550, and a weekly assessment even to 380 patients, showing a potential reduction of almost 44%. We would like to highlight once more that assessments of the traditional UPDRS III at a similar frequency in clinic would be highly challenging. Discussion Digital Gait Assessments for Symptom Monitoring and Prognosis Features derived from the Portabiles digital gait device were significantly associated with disease stage (represented by latent time), disease severity assessed via the Unified Parkinson's Disease Rating Scale (UPDRS), specifically the axial score, and the slope of the axial score. Although, the association between axial abnormalities and digital gait with more severe motor symptoms, as well as with lower step lengths and gait speed, has been described by several authors before [ 32 , 33 ], correlation analysis with digital gait features was still lacking. Additionally, most studies analyzing digitally assessed gait features, provide only micro analyses [ 8 , 13 , 16 , 34 , 35 ], looking at single digital gait features, but no overall analysis of multiple digital gait features. Here, a comprehensive analysis of single, but also multiple combined features – on several levels – is provided. In comparison to other studies, which used only one or even no standardized tasks, here, multiple tasks had to be performed by patients and task specificity was examined, providing a systematic and unique overview about most useful tasks for digital gait analysis. Although, combining all digital gait features turned out to be useful, there were notable task-specific differences between gait tasks. Interestingly, we found the traditional ‘Time Up And Go’ (TUG) task consistently effective for monitoring the motor symptom severity in both studies. Additionally, it was the only task that could monitor the UPDRS III in the LuxPARK cohort. Digital gait features derived from Tray and Count tasks could also be associated with PIGD score, indicating that these dual tasks, combining motor and cognitive exercises, are efficient in monitoring disease severity in terms of postural instability. In the past, dual tasks also have been seen to trigger Freezing of Gait [ 36 – 38 ]. Thus, investigating the ability to monitor cognitive impairments should be subject for further research. As expected, a clearly significant association of digital gait features with UPDRS derived motor symptom progression could only be established in the Erlangen dataset, where gait measurements have been taken over a longer period of time. However, this association could only be established for the 4x10m task. This highlights the importance of longer distance walks to identify changes of motor symptoms within an individual patient over several visits. Despite this, prediction of the UPDRS III from digital gait data only is a feasible task, as such estimates correlate well with the original data (ρ = 0.88), although not integrating any information about symptoms other than gait (e.g. hypomimia, dysarthria, or tremor). Potentially, the severity of these other symptoms is correlated to digital gait, or they are relatively stable in terms of disease progression such that their contribution to the UPDRS III progression is low. Right now, the reasons are unclear and further research might help to investigate on this relationship between digital gait and other PD symptoms. Although other studies have shown relationships between digital gait features and multiple disease stages, none of them quantify disease progression, but only classifying patients into disease stage groups [ 16 , 34 , 35 , 39 ]. Often the analysis focuses only on control vs. PD classification [ 5 , 8 , 13 , 40 , 41 ]. In contrast, our analysis focused on the association of digitally assessed gait features with motor symptoms and their progression. However, machine learning based prediction of the patient-level slope of UPDRS derived motor symptoms, demonstrated only a limited accuracy. When combining data from two subsequent clinical visits the coefficient of determination ( \({R}^{2}\) ) for predicting the slope of the axial score reached 0.45 and for the UPDRS III approximately 0.35. Hence, digital gait features alone have only limited prognostic value despite their ability to objectively monitor motor symptoms and their progression. Overall, the approach presented here demonstrates the feasibility of monitoring motor symptoms through digital gait assessment and its task specificity for individual symptoms across multiple independent studies. Potential Value for Clinical Trials Significant association between the patient-specific slope of the axial score and digital gait features, was found. We thus performed a randomized control trial simulation, with a predicted UPDRS III outcome using the Portabiles digital gait device derived features from the 4x10m test. Interestingly, a digital gait predicted outcome (i.e. a digitized version of the UPDRS III), although not integrating any information about other symptoms than gait, as mentioned above, requires the same number of patients to show the efficacy of a treatment as the standard UPDRS III, if gait features are collected at the same frequency. However, a large reduction of the number of needed patients could be shown when assessing the “digital” UPDRS III via the Portabiles digital gait device in a weekly manner. This shows the potential for using gait sensors and associated data in clinical studies, where measurements with a digital device could be taken remotely at a much higher frequency than in a traditional in-hospital setting. This may decrease the burden for patients’ quality of life and increase their willingness to contribute to a clinical study as they are not required to visit the clinic for each assessment. Additionally, data in general can be recorded via the 4x10m task much easier and faster than performing a complete UPDRS exam. Only limited studies have already investigated the benefit of using digital gait device data for clinical trials. While sensor-based digital gait analysis has shown its potential to overcome limitations of UPDRS based gait assessment in a phase II trial recently [ 42 ], most studies only refer to a subjective potential endpoint in studies, but do not test the validity of this hypothesis. Here, for the first time, an objective comparison of both outcomes, original and gait predicted UPDRS III, quantifies the potential of using sensor-based digital gait data as secondary endpoint in trials. Although the potential of the sensor-based digital gait data could clearly be seen, it does not capture quality of life aspects during a trial, which is a limitation for phase III studies and still needs to be addressed by other instruments. Conclusion Overall, we have demonstrated the ability to objectively monitor the severity and progression of motor symptoms using the Portabiles HCT digital gait device. Additionally, we investigated the potential benefit of using digital gait as an outcome measure in a clinical trial context. Our work thus adds to the increasing body of literature demonstrating the benefit of digital solutions in the PD field. A distinction point is our focused digital gait device, the analysis of symptom progression and the exploration of the benefit for clinical studies. We would like to emphasize that the use of any medical device, including the tested Portabiles digital gait device, outside pure research settings in routine medical care requires approval by a regulatory agency, and in that regard Portabiles has recently launched a request to the FDA in the USA. Besides that, it is already certified as a class I medical device (MDR) with a CE mark in Europe. Declarations Competing Interests Ralph Steidl is the CEO of Portabiles HCT. Jean-Christophe Corvol has served in advisory boards for Alzprotect, Bayer, Biogen, Denali, Ferrer, Idorsia, iRegene, Prevail Therapeutic, Servier, Theranexus, UCB; and received grants from Sanofi and the Michael J Fox Foundation outside of this work. Holger Fröhlich has received grants from UCB and AbbVie outside this work.Tamara Raschka, Jackrite To, Tom Hähnel, Stefano Sapienza, Alzhraa Ibrahim, Enrico Glaab, Heiko Gaßner, Jürgen Winkler, and Jochen Klucken declare no competing interests. Funding This study was funded by the ERA PerMed EU-wide project DIGIPD (01KU2110). This project is additionally supported by the Luxembourg National Research Fund (FNR) (INTER/ERAPerMed 20/14599012). LuxPARK is part of the National Centre of Excellence in Research on Parkinson's Disease (NCER-PD), which is funded by the Luxembourg National Research Fund (FNR/NCER13/BM/11264123). The funders played no role in study design, data collection, analysis and interpretation of data, or the writing of this manuscript. Data from Erlangen is approved under (IRB-approval-Re. -No. 4208, 21.04.2010, IRB, Medical Faculty, Friedrich-Alexander University Erlangen-Nürnberg, Germany) and by local ethics committee (reference number: 166_18 B, Medical Faculty, FAU Erlangen-Nürnberg, Germany). Author Contribution R.S., J.W., J.C., J.K., and H.F. designed the study. S.S., A.I., E.G., R.S., J.W., E.G., H.G., J.C., J.K., and H.F. contributed to the acquisition of data. T.R., J.T., and T.H. analyzed and interpreted the data. T.R. wrote the initial draft of the paper. H.F. substantively revised it. All authors discussed the results and contributed to revising the paper. All authors approved the final version of the manuscript. Data Availability As this study is a retrospective analysis, availability of the clinical data depends on the individual study groups (LuxPARK: [email protected] ; Erlangen: [email protected] ). References Postuma RB, Berg D, Stern M, Poewe W, Olanow CW, Oertel W, et al. MDS clinical diagnostic criteria for Parkinson’s disease: MDS-PD Clinical Diagnostic Criteria. Mov Disord. 2015;30: 1591–1601. doi: 10.1002/mds.26424 Moustafa AA, Chakravarthy S, Phillips JR, Gupta A, Keri S, Polner B, et al. Motor symptoms in Parkinson’s disease: A unified framework. Neurosci Biobehav Rev. 2016;68: 727–740. doi: 10.1016/j.neubiorev.2016.07.010 Xia R, Mao Z-H. Progression of motor symptoms in Parkinson’s disease. Neurosci Bull. 2012;28: 39–48. doi: 10.1007/s12264-012-1050-z Mirelman A, Bonato P, Camicioli R, Ellis TD, Giladi N, Hamilton JL, et al. Gait impairments in Parkinson’s disease. Lancet Neurol. 2019;18: 697–708. doi: 10.1016/S1474-4422(19)30044-4 Chandrabhatla AS, Pomeraniec IJ, Ksendzovsky A. Co-evolution of machine learning and digital technologies to improve monitoring of Parkinson’s disease motor symptoms. Npj Digit Med. 2022;5: 32. doi: 10.1038/s41746-022-00568-y Fröhlich H, Bontridder N, Petrovska-Delacréta D, Glaab E, Kluge F, Yacoubi ME, et al. Leveraging the Potential of Digital Technology for Better Individualized Treatment of Parkinson’s Disease. Front Neurol. 2022;13: 788427. doi: 10.3389/fneur.2022.788427 Goetz CG, Tilley BC, Shaftman SR, Stebbins GT, Fahn S, Martinez-Martin P, et al. Movement Disorder Society‐sponsored revision of the Unified Parkinson’s Disease Rating Scale (MDS‐UPDRS): Scale presentation and clinimetric testing results. Mov Disord. 2008;23: 2129–2170. doi: 10.1002/mds.22340 Kirk C, Zia Ur Rehman R, Galna B, Alcock L, Ranciati S, Palmerini L, et al. Can Digital Mobility Assessment Enhance the Clinical Assessment of Disease Severity in Parkinson’s Disease? J Park Dis. 2023;13: 999–1009. doi: 10.3233/JPD-230044 Nair P, Shojaei Baghini M, Pendharkar G, Chung H. Detecting early-stage Parkinson’s disease from gait data. Proc Inst Mech Eng [H]. 2023; 09544119231197090. doi: 10.1177/09544119231197090 Deng K, Li Y, Zhang H, Wang J, Albin RL, Guan Y. Heterogeneous digital biomarker integration out-performs patient self-reports in predicting Parkinson’s disease. Commun Biol. 2022;5: 58. doi: 10.1038/s42003-022-03002-x Djurić-Jovičić M, Belić M, Stanković I, Radovanović S, Kostić VS. Selection of gait parameters for differential diagnostics of patients with de novo Parkinson’s disease. Neurol Res. 2017;39: 853–861. doi: 10.1080/01616412.2017.1348690 Adams JL, Kangarloo T, Tracey B, O’Donnell P, Volfson D, Latzman RD, et al. Using a smartwatch and smartphone to assess early Parkinson’s disease in the WATCH-PD study. Npj Park Dis. 2023;9: 64. doi: 10.1038/s41531-023-00497-x Pistacchi M. Gait analysis and clinical correlations in early Parkinson’s disease. Funct Neurol. 2017;32: 28. doi: 10.11138/FNeur/2017.32.1.028 Cai G, Shi W, Wang Y, Weng H, Chen L, Yu J, et al. Specific Distribution of Digital Gait Biomarkers in Parkinson’s Disease Using Body-Worn Sensors and Machine Learning. Duque G, editor. J Gerontol Ser A. 2023;78: 1348–1354. doi: 10.1093/gerona/glad101 Beswick E, Fawcett T, Hassan Z, Forbes D, Dakin R, Newton J, et al. A systematic review of digital technology to evaluate motor function and disease progression in motor neuron disease. J Neurol. 2022;269: 6254–6268. doi: 10.1007/s00415-022-11312-7 Zhu S, Wu Z, Wang Y, Jiang Y, Gu R, Zhong M, et al. Gait Analysis with Wearables Is a Potential Progression Marker in Parkinson’s Disease. Brain Sci. 2022;12: 1213. doi: 10.3390/brainsci12091213 Diao JA, Raza MM, Venkatesh KP, Kvedar JC. Watching Parkinson’s disease with wrist-based sensors. Npj Digit Med. 2022;5: 73, s41746-022-00619–4. doi: 10.1038/s41746-022-00619-4 Li D, Iddi S, Thompson WK, Donohue MC, Alzheimer’s Disease Neuroimaging Initiative. Bayesian latent time joint mixed effect models for multicohort longitudinal data. Stat Methods Med Res. 2019;28: 835–845. doi: 10.1177/0962280217737566 Hipp G, Vaillant M, Diederich NJ, Roomp K, Satagopam VP, Banda P, et al. The Luxembourg Parkinson’s Study: A Comprehensive Approach for Stratification and Early Diagnosis. Front Aging Neurosci. 2018;10: 326. doi: 10.3389/fnagi.2018.00326 Aleksovski D, Miljkovic D, Bravi D, Antonini A. Disease progression in Parkinson subtypes: the PPMI dataset. Neurol Sci. 2018;39: 1971–1976. doi: 10.1007/s10072-018-3522-z Hayashi Y, Mishima T, Fujioka S, Morishita T, Inoue T, Nagamachi S, et al. Unilateral GPi-DBS Improves Ipsilateral and Axial Motor Symptoms in Parkinson’s Disease as Evidenced by a Brain Perfusion Single Photon Emission Computed Tomography Study. Front Hum Neurosci. 2022;16: 888701. doi: 10.3389/fnhum.2022.888701 Fahn S. Unified Parkinson’s disease rating scale. Recent Dev Park Dis. 1987; 153–163. Hähnel T, Raschka T, Sapienza S, Klucken J, Glaab E, Corvol J-C, et al. Progression subtypes in Parkinson’s disease identified by a data-driven multi cohort analysis. Npj Park Dis. 2024;10: 95. doi: 10.1038/s41531-024-00712-3 Benjamini Y, Yekutieli D. The control of the false discovery rate in multiple testing under dependency. Ann Stat. 2001;29. doi: 10.1214/aos/1013699998 UCB Biopharma SRL. A Double-Blind, Placebo-Controlled, Randomized, 18-Month Phase 2a Study to Evaluate the Efficacy, Safety, Tolerability, and Pharmacokinetics of Oral UCB0599 in Study Participants With Early Parkinson’s Disease. clinicaltrials.gov. 2023. Available: https://clinicaltrials.gov/study/NCT04658186 (accessed July 16, 2023) Ard MC, Edland SD. Power Calculations for Clinical Trials in Alzheimer’s Disease. Ashford JW, Rosen A, Adamson M, Bayley P, Sabri O, Furst A, et al., editors. J Alzheimers Dis. 2011;26: 369–377. doi: 10.3233/JAD-2011-0062 Iddi S, C Donohue M. Power and Sample Size for Longitudinal Models in R -- The longpower Package and Shiny App. R J. 2022;14: 264–282. doi: 10.32614/RJ-2022-022 Breiman L. Random Forests. Mach Learn. 2001;45: 5–32. doi: 10.1023/A:1010933404324 Chen T, Guestrin C. XGBoost: A Scalable Tree Boosting System. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. San Francisco California USA: ACM; 2016. pp. 785–794. doi: 10.1145/2939672.2939785 Tibshirani R. Regression Shrinkage and Selection Via the Lasso. J R Stat Soc Ser B Methodol. 1996;58: 267–288. doi: 10.1111/j.2517-6161.1996.tb02080.x Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, et al. Scikit-learn: Machine learning in Python. J Mach Learn Res. 2011;12: 2825–2830. Cao S, Cui Y, Jin J, Li F, Liu X, Feng T. Prevalence of axial postural abnormalities and their subtypes in Parkinson’s disease: a systematic review and meta-analysis. J Neurol. 2023;270: 139–151. doi: 10.1007/s00415-022-11354-x Pongmala C, Fabbri M, Zibetti M, Pitakpatapee Y, Wangthumrong T, Sangpeamsook T, et al. Gait and axial postural abnormalities correlations in Parkinson’s disease: A multicenter quantitative study. Parkinsonism Relat Disord. 2022;105: 19–23. doi: 10.1016/j.parkreldis.2022.10.026 Schlachetzki JCM, Barth J, Marxreiter F, Gossler J, Kohl Z, Reinfelder S, et al. Wearable sensors objectively measure gait parameters in Parkinson’s disease. Toft M, editor. PLOS ONE. 2017;12: e0183989. doi: 10.1371/journal.pone.0183989 Welzel J, Wendtland D, Warmerdam E, Romijnders R, Elshehabi M, Geritz J, et al. Step Length Is a Promising Progression Marker in Parkinson’s Disease. Sensors. 2021;21: 2292. doi: 10.3390/s21072292 Snijders AH, Haaxma CA, Hagen YJ, Munneke M, Bloem BR. Freezer or non-freezer: Clinical assessment of freezing of gait. Parkinsonism Relat Disord. 2012;18: 149–154. doi: 10.1016/j.parkreldis.2011.09.006 Spildooren J, Vercruysse S, Desloovere K, Vandenberghe W, Kerckhofs E, Nieuwboer A. Freezing of gait in Parkinson’s disease: The impact of dual-tasking and turning. Mov Disord. 2010;25: 2563–2570. doi: 10.1002/mds.23327 Raccagni C, Nonnekes J, Bloem BR, Peball M, Boehme C, Seppi K, et al. Gait and postural disorders in parkinsonism: a clinical approach. J Neurol. 2020;267: 3169–3176. doi: 10.1007/s00415-019-09382-1 Micó-Amigo ME, Kingma I, Heinzel S, Rispens SM, Heger T, Nussbaum S, et al. Potential Markers of Progression in Idiopathic Parkinson’s Disease Derived From Assessment of Circular Gait With a Single Body-Fixed-Sensor: A 5 Year Longitudinal Study. Front Hum Neurosci. 2019;13: 59. doi: 10.3389/fnhum.2019.00059 Di Biase L, Di Santo A, Caminiti ML, De Liso A, Shah SA, Ricci L, et al. Gait Analysis in Parkinson’s Disease: An Overview of the Most Accurate Markers for Diagnosis and Symptoms Monitoring. Sensors. 2020;20: 3529. doi: 10.3390/s20123529 Coates L, Shi J, Rochester L, Del Din S, Pantall A. Entropy of Real-World Gait in Parkinson’s Disease Determined from Wearable Sensors as a Digital Marker of Altered Ambulatory Behavior. Sensors. 2020;20: 2631. doi: 10.3390/s20092631 Payne T, Appleby M, Buckley E, Van Gelder LMA, Mullish BH, Sassani M, et al. A Double-Blind, Randomized, Placebo‐Controlled Trial of Ursodeoxycholic Acid (UDCA) in Parkinson’s Disease. Mov Disord. 2023;38: 1493–1502. doi: 10.1002/mds.29450 Additional Declarations Competing interest reported. Ralph Steidl is the CEO of Portabiles HCT. Jean-Christophe Corvol has served in advisory boards for Alzprotect, Bayer, Biogen, Denali, Ferrer, Idorsia, iRegene, Prevail Therapeutic, Servier, Theranexus, UCB; and received grants from Sanofi and the Michael J Fox Foundation outside of this work. Holger Fröhlich has received grants from UCB and AbbVie outside this work. Tamara Raschka, Jackrite To, Tom Hähnel, Stefano Sapienza, Alzhraa Ibrahim, Enrico Glaab, Heiko Gaßner, Jürgen Winkler, and Jochen Klucken declare no competing interests. Supplementary Files GaitManuscriptSupplement.pdf SupplementaryTableS5singleFeatureResults.xlsx SupplementaryTableS6modelCoefficients.xlsx Cite Share Download PDF Status: Published Journal Publication published 15 Jul, 2025 Read the published version in Scientific Reports → Version 1 posted Editorial decision: Revision requested 12 Feb, 2025 Reviews received at journal 26 Dec, 2024 Reviewers agreed at journal 15 Dec, 2024 Reviews received at journal 30 Nov, 2024 Reviewers agreed at journal 18 Nov, 2024 Reviewers agreed at journal 06 Sep, 2024 Reviewers agreed at journal 06 Sep, 2024 Reviewers agreed at journal 21 Jul, 2024 Reviewers agreed at journal 15 Jul, 2024 Reviewers invited by journal 15 Jul, 2024 Editor assigned by journal 10 Jul, 2024 Editor invited by journal 10 Jun, 2024 Submission checks completed at journal 06 Jun, 2024 First submitted to journal 03 Jun, 2024 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4521747","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":316056887,"identity":"8cfac23c-2a90-4223-9056-8a874b42adbb","order_by":0,"name":"Tamara Raschka","email":"","orcid":"","institution":"Fraunhofer Institute for Algorithms and Scientific Computing (SCAI)","correspondingAuthor":false,"prefix":"","firstName":"Tamara","middleName":"","lastName":"Raschka","suffix":""},{"id":316056888,"identity":"a24202c8-27f4-4b4b-8578-47a42ca1ce03","order_by":1,"name":"Jackrite To","email":"","orcid":"","institution":"Fraunhofer Institute for Algorithms and Scientific Computing (SCAI)","correspondingAuthor":false,"prefix":"","firstName":"Jackrite","middleName":"","lastName":"To","suffix":""},{"id":316056889,"identity":"bf85940f-039f-4703-b90e-d0a6653c1551","order_by":2,"name":"Tom Hähnel","email":"","orcid":"","institution":"Department of Neurology, University Hospital and Faculty of Medicine Carl Gustav Carus, TUD Dresden University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Tom","middleName":"","lastName":"Hähnel","suffix":""},{"id":316056890,"identity":"5de72926-2e82-4538-a043-ef6136990b48","order_by":3,"name":"Stefano Sapienza","email":"","orcid":"","institution":"Luxembourg Centre for Systems Biomedicine, University of Luxembourg","correspondingAuthor":false,"prefix":"","firstName":"Stefano","middleName":"","lastName":"Sapienza","suffix":""},{"id":316056891,"identity":"dc837a72-c39f-4983-a906-31b51040fb92","order_by":4,"name":"Alzhraa Ibrahim","email":"","orcid":"","institution":"Department of Molecular Neurology, University Hospital Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg","correspondingAuthor":false,"prefix":"","firstName":"Alzhraa","middleName":"","lastName":"Ibrahim","suffix":""},{"id":316056892,"identity":"b8796498-b385-4ad2-84c1-2a46c490aa3d","order_by":5,"name":"Enrico Glaab","email":"","orcid":"","institution":"Luxembourg Centre for Systems Biomedicine, University of Luxembourg","correspondingAuthor":false,"prefix":"","firstName":"Enrico","middleName":"","lastName":"Glaab","suffix":""},{"id":316056893,"identity":"37c2f172-c3fc-4751-9637-99989e9b3757","order_by":6,"name":"Heiko Gaßner","email":"","orcid":"","institution":"Department of Molecular Neurology, University Hospital Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg","correspondingAuthor":false,"prefix":"","firstName":"Heiko","middleName":"","lastName":"Gaßner","suffix":""},{"id":316056894,"identity":"5ba8b625-2d9e-480d-814b-6683f0df5b13","order_by":7,"name":"Ralph Steidl","email":"","orcid":"","institution":"Portabiles HealthCare Technologies","correspondingAuthor":false,"prefix":"","firstName":"Ralph","middleName":"","lastName":"Steidl","suffix":""},{"id":316056895,"identity":"684d1cf9-a1cc-43d9-92d8-74392bfc6b44","order_by":8,"name":"Jürgen Winkler","email":"","orcid":"","institution":"Department of Molecular Neurology, University Hospital Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg","correspondingAuthor":false,"prefix":"","firstName":"Jürgen","middleName":"","lastName":"Winkler","suffix":""},{"id":316056896,"identity":"a1bbe175-a606-443c-a7c6-9323533be869","order_by":9,"name":"Jean-Christophe Corvol","email":"","orcid":"","institution":"Sorbonne Université, Paris Brain Institute – ICM, Inserm, CNRS, Assistance Publique Hôpitaux de Paris","correspondingAuthor":false,"prefix":"","firstName":"Jean-Christophe","middleName":"","lastName":"Corvol","suffix":""},{"id":316056897,"identity":"bab921cb-7b8b-4429-abd0-b7eda9e34654","order_by":10,"name":"Jochen Klucken","email":"","orcid":"","institution":"Luxembourg Centre for Systems Biomedicine, University of Luxembourg","correspondingAuthor":false,"prefix":"","firstName":"Jochen","middleName":"","lastName":"Klucken","suffix":""},{"id":316056898,"identity":"6a97a0a5-6eb2-4cf3-9384-73534dbd796d","order_by":11,"name":"Holger Fröhlich","email":"data:image/png;base64,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","orcid":"","institution":"Fraunhofer Institute for Algorithms and Scientific Computing (SCAI)","correspondingAuthor":true,"prefix":"","firstName":"Holger","middleName":"","lastName":"Fröhlich","suffix":""}],"badges":[],"createdAt":"2024-06-03 12:24:47","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4521747/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4521747/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41598-025-09088-7","type":"published","date":"2025-07-15T16:05:36+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":59054811,"identity":"01b7c905-f3d4-48a8-85bd-019361bfc7b5","added_by":"auto","created_at":"2024-06-25 20:41:15","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":134439,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eWorkflow overview of digital gait analysis.\u003c/strong\u003e\u003cbr\u003e\n Depicted is the whole workflow done in this study. A – Random Slope and Time Shift Estimation: First, disease trajectories are shifted on a common disease timescale via a latent time joint mixed-effect model (LTJMM). Here, the time shift, showing the divergence from a ‘mean’ PD patient, and random slopes, a random effect measuring the progression of each patient compared to the mean patient, are calculated. B – Disease Monitoring: In a next step, the variance of the random slopes is modeled with a linear (mixed) model using either only age and sex (purple) as adjustment covariates or age, sex, and digital gait features (blue). Here, multiple different sets of digital gait parameters are tested: all, task-specific, and single. Models are compared via likelihood ratio tests assessing the statistical significance of digital gait features. C – Disease Prediction: Lastly, scores are predicted via a machine learning algorithm (Random Forest) with multiple digital gait feature sets (blue: all digital gait features plus age and sex, orange: TUG task plus age and sex, green: 2x10m task plus age and sex, red: 4x10m task plus age and sex, and purple: only Age and Sex).\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-4521747/v1/490c980990815afeb16b2c8c.png"},{"id":59054810,"identity":"d85fe939-7e93-4f7d-abc7-9ae266d07b05","added_by":"auto","created_at":"2024-06-25 20:41:15","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":83923,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePrediction of patient-specific slopes of traditional clinical outcomes from digital gait features\u003cbr\u003e\n \u003c/strong\u003eBoxplots show the squared coefficient of determination (R²) of multiple repeats of cross-validating a Random Forest machine learning model. The model predicts the (random) slope of the axial and UPDRS III score of each patient trained on data from first gait assessment only (left) or from first plus second gait assessment (left). Different colours indicate the used feature set: all digital gait features plus age and sex (blue), TUG task plus age and sex (orange), 2x10m task plus age and sex (green), 4x10m task plus age and sex (red), and only Age and Sex (purple).\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-4521747/v1/bda2ccc15558fbe77d89ea99.png"},{"id":59054813,"identity":"e86e0c88-6af4-4897-99ae-80a626b8748b","added_by":"auto","created_at":"2024-06-25 20:41:15","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":248249,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSimulation of randomized clinical trial with multiple visit frequencies\u003c/strong\u003e\u003cbr\u003e\nCurves describe the power a clinical study can reach with a specific sample size while expecting a treatment effect of 30% during a one year observation period and a significance level of α=0.01. Dashed line indicates 80% power. Different randomized clinical studies with either original or gait predicted UPDRS III as outcome were simulated with multiple visit frequencies. For gait predicted UPDRS III weekly (blue), monthly (orange), and bi-monthly (green) visits were simulated, while only bi-monthly visits were simulated for original UPDRS III as outcome (red).\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-4521747/v1/5e0c9cab7c589a94eeb75292.png"},{"id":87221026,"identity":"9fb0d2a5-9419-414f-8fe4-fe075378d2ee","added_by":"auto","created_at":"2025-07-21 16:13:22","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1767878,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4521747/v1/a64c5a92-0fd1-4422-9539-30b86241d1f2.pdf"},{"id":59054815,"identity":"9385ba0f-b5e2-44e9-bf2e-04f7ad3fe706","added_by":"auto","created_at":"2024-06-25 20:41:16","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":5573784,"visible":true,"origin":"","legend":"","description":"","filename":"GaitManuscriptSupplement.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4521747/v1/b7df2e73e7a1792161ebae87.pdf"},{"id":59055135,"identity":"36e48d8f-dddf-48c6-8eb9-2acf88227bb0","added_by":"auto","created_at":"2024-06-25 20:49:15","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":27976,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryTableS5singleFeatureResults.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-4521747/v1/564f6aa583032fa7afe5b13d.xlsx"},{"id":59054814,"identity":"93360b72-84a6-429e-a6a6-ab273a5e5d63","added_by":"auto","created_at":"2024-06-25 20:41:15","extension":"xlsx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":269301,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryTableS6modelCoefficients.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-4521747/v1/480db00c74d3e796f9ba4636.xlsx"}],"financialInterests":"Competing interest reported. Ralph Steidl is the CEO of Portabiles HCT. Jean-Christophe Corvol has served in advisory boards for Alzprotect, Bayer, Biogen, Denali, Ferrer, Idorsia, iRegene, Prevail Therapeutic, Servier, Theranexus, UCB; and received grants from Sanofi and the Michael J Fox Foundation outside of this work. Holger Fröhlich has received grants from UCB and AbbVie outside this work.\nTamara Raschka, Jackrite To, Tom Hähnel, Stefano Sapienza, Alzhraa Ibrahim, Enrico Glaab, Heiko Gaßner, Jürgen Winkler, and Jochen Klucken declare no competing interests.","formattedTitle":"Objective Monitoring of Motor Symptom Severity and their Progression in Parkinson’s Disease Using a Digital Gait Device","fulltext":[{"header":"Introduction","content":"\u003cp\u003eGait abnormalities in people with Parkinson\u0026rsquo;s Disease (PD) are a major motor symptom, along with tremor, bradykinesia, and rigidity, and have a major impact on patients\u0026rsquo; mobility and quality of life [\u003cspan additionalcitationids=\"CR2\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Signs of impaired gait patterns are reduced gait speed and step length, increased axial rigidity, impaired rhythmicity, and freezing of gait [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. In the early stages of the disease, slower gait speed, shortened step length, and reduced arm swing during walking are the most common signs of gait abnormalities, while in the later stages, movements become even more slow, gait shuffling and freezing of gait appears and patients are thus prone to falls [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. With progression into an advanced stage, gait abnormalities, including balance and postural control, become worse [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. Thus, investigating patient\u0026rsquo;s gait impairment is crucial for evaluating the stage of disease progression. Furthermore, objective monitoring of gait could help defining an optimal schedule of when to treat a patient, whether to adjust the dose or whether to switch from one treatment to another.\u003c/p\u003e \u003cp\u003eIn the past, digital technologies for monitoring motor symptoms of PD have evolved from non-portable in-lab devices to wearable sensors and at-home phone applications [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. While each technology has its advantages, wearable sensors are able to measure gait parameters like stride length, gait speed, or the lateral excursion of each patient quantitatively and potentially also in an outpatient setting, while being portable, light weight, less expensive, and easy to use, for patients, as well as, for clinicians [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. However, before any application in clinical care, such technologies must be carefully assessed regarding technical robustness as well as medical diagnostic validity [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. Importantly, a digital gait sensor can only reliably be used in routine medical care after regulatory approval, requiring proof of its safety. Besides that, trustable digital gait features require clear evidence that these sensor measurements correlate significantly with established clinical outcomes such as MDS-Unified Parkinson\u0026rsquo;s Disease Rating Scale (MDS-UPDRS) [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Although digital gait features have been explored in several studies in PD research, most studies solely focus on distinguishing PD patients from healthy controls in cross-sectional manner [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan additionalcitationids=\"CR9 CR10 CR11 CR12\" citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Some studies also classify patients into multiple disease stages [\u003cspan additionalcitationids=\"CR15 CR16\" citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e], but studies rarely describe the associations between PD progression and gait directly [\u003cspan additionalcitationids=\"CR5\" citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. Furthermore, there is a lack of studies investigating the reproducibility of findings from the same gait device across several PD cohorts. Taken together, it is still unclear to date whether digital gait sensors can reproducibly and objectively monitor motor symptoms and their progression over time and whether in that regard there is a demonstrable benefit to classical clinical scales.\u003c/p\u003e \u003cp\u003eThe ERA PerMed funded EU-wide project DIGIPD (Validating \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eDIGI\u003c/span\u003etal biomarkers for better personalized treatment of \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eP\u003c/span\u003earkinson's \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eD\u003c/span\u003eisease\u0026mdash;\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.digipd.eu\u003c/span\u003e\u003cspan address=\"https://www.digipd.eu\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) tries to fill a gap by evaluating the potential benefit of digitally assessed PD symptoms compared to established clinical outcomes across several cohorts in Europe. In this study, we focused on a sensor-based digital gait device clipped to the shoe of a patient while walking (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.portabiles-hct.de/en/product/\u003c/span\u003e\u003cspan address=\"https://www.portabiles-hct.de/en/product/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). The device is currently registered for regulatory approval at the US Food and Drug Administration (FDA) and a certified class I medical device (MDR) with a CE mark in Europe. We investigated data recorded by the device in two independent longitudinal cohorts (one from Germany, one from Luxembourg) and while performing different gait tasks in the clinic. More specifically, we initially modelled PD progression based on several established outcome measures derived from the MDS-UPDRS by fitting a latent time joint mixed-effect model (LTJMM) [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e] to longitudinal data from both studies. This allowed us to estimate each patients\u0026rsquo; true rate of progression and to align disease trajectories on a common disease timescale, taking into account the fact that patients may have progressed differently at baseline. Following that, linear (mixed) models were used to evaluate the association of digital gait features with estimated progression rates as well as UPDRS derived outcomes. In that regard we also explored, in how far a machine learning model could forecast an individual\u0026rsquo;s disease progression rate solely based on digital gait features. Finally, we investigated on the potential benefit of using a digital gait predicted UPDRS III compared to the original UPDRS III as an endpoint in a clinical trial.\u003c/p\u003e"},{"header":"Material and Methods","content":"\u003cp\u003eCohorts\u003c/p\u003e \u003cp\u003eWe analyzed data from two independent PD cohort studies: (1) the Luxembourg Parkinson\u0026rsquo;s Study (LuxPARK, NCT05266872) [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e] and (2) from the University Medical Center Erlangen, Germany (Erlangen). LuxPARK is an ongoing observational cohort of patients at all disease stages. Data was collected periodically up to four years of follow-up between 2015 and 2022. Data from Erlangen contained information from patients at all disease stages collected during multiple irregular clinical routine visits between 2011 and 2022. For both studies, we used multiple clinical scores as outcomes in our analysis. In LuxPARK these scores were MDS-Unified Parkinson's Disease Rating Scale (MDS-UPDRS) I, MDS-UPDRS II, MDS-UPDRS III [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e], Tremor dominance (TD) score [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e], Postural Instability and Gait Dysfunction (PIGD) score [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e], and an axial score. The axial score was calculated by summing MDS-UPDRS III sub-items 1 (speech), 2 (facial expression), 9 (arising from chair), 10 (gait), 11 (freezing of gait), 12 (postural instability), and 13 (posture), which is a customized version of the axial score described in [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e] to integrate axial changes in the face and freezing of gait more comprehensively. While LuxPARK employed the MDS-UPDRS, the Erlangen data assessed PD symptoms of patients with the original UPDRS [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. Differences are described in the MDS-UPDRS publication [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. For the sake of simplicity, we refer to both, the MDS-UPDRS, as well as the original UPDRS, as UPDRS. The Erlangen data evaluated less clinical scores and thus, only UPDRS III and the axial score (without freezing of gait, as this was not part of the UPDRS, but MDS-UPDRS only) were used as outcomes.\u003c/p\u003e \u003cp\u003eFor both datasets, patients were selected, such that each patient had data from at least two visits for each outcome. This resulted in 612 PD patients from the LuxPARK cohort and 264 PD patients from the Erlangen cohort. Characteristics of these patients can be found in \u003cb\u003eSupplementary Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e\u003c/b\u003e in detail.\u003c/p\u003e \u003cp\u003eGait Measurements\u003c/p\u003e \u003cp\u003eAdditional to the traditional clinical outcome scores described above, gait measurements using Portabiles digital gait device were available in both, LuxPARK and Erlangen. While in LuxPARK only one digital gait assessment, at a single visit, was performed, multiple digital gait assessments over time were available for the Erlangen patients (2.84\u0026thinsp;\u0026plusmn;\u0026thinsp;2.37 visits per patient on average, total: 506 gait visits). Digital gait assessments in Erlangen were collected irregularly, similar to the clinical data (difference between first and second digital gait assessment: 503.11\u0026thinsp;\u0026plusmn;\u0026thinsp;488.98 days, median\u0026thinsp;=\u0026thinsp;364 days). Digital gait data was available for 343 (LuxPARK) and 802 (Erlangen) patients for multiple exercises such as Time-Up-and-Go (TUG), the Turn, manual TUG (Tray), and cognitive TUG (Count), as well as the 2x10m, and 4x10m walking tasks. A detailed description of the different tasks, including differences between Erlangen and LuxPARK, can be found in the \u003cb\u003eSupplementary Material\u003c/b\u003e. All analysis regarding digital gait were conducted only on the subset of patients for which both longitudinal clinical features (as described above), as well as, digital gait features were available. In total these were 161 patients from LuxPARK and 178 patients from Erlangen. Characteristics of these patients can be found in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. A detailed description of the used device, the raw signal processing, and the derived features for further analysis along with a visualization of their distribution can be found in the \u003cb\u003eSupplementary Material\u003c/b\u003e (\u003cb\u003eSupplementary Table \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e, Supplementary Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e, Supplementary Figure \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e\u003c/b\u003e). A data catalogue visualizing all clinical, as well as, digital gait features of both cohorts can be found here: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://p-data.molekulare-neurologie.uk-erlangen.de/\u003c/span\u003e\u003cspan address=\"https://p-data.molekulare-neurologie.uk-erlangen.de/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003e\u003cb\u003eDemographic and clinical characteristics of patients\u003c/b\u003e\u003c/p\u003e \u003cp\u003ePatients characteristics for LuxPARK and Erlangen datasets. Only the subset used for the gait analysis is described here. These are all patients having longitudinal clinical features plus digital gait features. Mean and standard deviations are shown for all characteristics but sex and Hoehn \u0026amp; Yahr stages, where absolute (and relative) values are shown. Values for disease duration, age, and clinical scores are reported at baseline. PIGD, TD, UPDRS I, and UPDRS II are not available in the Erlangen cohort.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLuxPARK\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eErlangen\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eNumber of patients\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e161\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e178\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eNumber of visits\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e5.32\u0026thinsp;\u0026plusmn;\u0026thinsp;1.53\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e12.08\u0026thinsp;\u0026plusmn;\u0026thinsp;9.12\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDisease duration, years\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4.92\u0026thinsp;\u0026plusmn;\u0026thinsp;5.29\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e5.31\u0026thinsp;\u0026plusmn;\u0026thinsp;4.57\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAge, years\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e64.71\u0026thinsp;\u0026plusmn;\u0026thinsp;10.08\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e62.39\u0026thinsp;\u0026plusmn;\u0026thinsp;10.37\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eSex\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eMale\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e117 (73%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e110 (62%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eFemale\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e44 (27%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e68 (38%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAxial score\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e5.71\u0026thinsp;\u0026plusmn;\u0026thinsp;3.79\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e5.35\u0026thinsp;\u0026plusmn;\u0026thinsp;3.93\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003ePIGD\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.56\u0026thinsp;\u0026plusmn;\u0026thinsp;0.58\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eTD\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.5\u0026thinsp;\u0026plusmn;\u0026thinsp;0.41\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eUPDRS I\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e9.28\u0026thinsp;\u0026plusmn;\u0026thinsp;5.68\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eUPDRS II\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e10.32\u0026thinsp;\u0026plusmn;\u0026thinsp;6.8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eUPDRS III\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e29.89\u0026thinsp;\u0026plusmn;\u0026thinsp;14.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e20.2\u0026thinsp;\u0026plusmn;\u0026thinsp;11.89\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eH\u0026amp;Y\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e0\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e1\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e19\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e25\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e1.5\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e14\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e2\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e93\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e35\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e2.5\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e3\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e35\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e4\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e11\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e5\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eN/A\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e37\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eModeling approaches\u003c/p\u003e \u003cp\u003eLatent time joint mixed-effect model\u003c/p\u003e \u003cp\u003eThe latent time joint mixed-effect model (LTJMM) was first published by Li et al. [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. Its rational is to align trajectories of patients on a common latent (i.e. unobserved) disease timescale using a multivariate linear mixed effects model. This model is defined by fixed and random effects describing the deviation of each patient from a \u0026ldquo;mean\u0026rdquo; reference trajectory, both, on the time axis, as well as with respect to the actual outcome. The model can thus describe piecewise linear progression of multiple clinical outcomes over time by estimating how far the timescale of each patient is shifted away from the timescale of the reference. In our case the timescale was the reported time since PD diagnosis. The inferred patient-specific time shift can be seen as an estimate how far a particular patient is advanced in the severity of symptoms compared to the other patients in the same cohort. Additionally, the model generates an estimate of the patient-specific slope of disease symptoms, meaning that a large \u0026ldquo;random\u0026rdquo; slope indicates a higher speed of symptom progression compared to all other patients in the same cohort. Used outcomes were in our case the UPDRS I-III, TD, PIGD, and axial score, depending on the availability in the two different datasets. Additionally, we adjusted for age at diagnosis, sex and medication status (ON/OFF) as fixed effects. All longitudinal data, irrespective of gait data availability, was used. Mathematical details of the model are described in the \u003cb\u003eSupplementary Material\u003c/b\u003e.\u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eValidation of LTJMM model\u003c/h2\u003e \u003cp\u003eDistributions of Hoehn and Yahr (H\u0026amp;Y) stages were examined to validate the alignment of disease trajectories on a common disease timescale. This was based on the assumption that the shifts of each individuals\u0026rsquo; trajectory should follow an order, which roughly agrees with the H\u0026amp;Y stages. Patients with lower H\u0026amp;Y stages, who are at the beginning of the development of the disease, should be shifted further to the left on the common disease time scale than those with a higher HY stage who are more advanced in their disease. Additionally, the accuracy of the LTJMM approach was checked by calculating the correlation between real and predicted outcomes at the last observed visit, similar to [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. For that, a LTJMM was fitted on all outcomes, but the last measurements of each patient were excluded and predicted back by the LTJMM.\u003c/p\u003e \u003cp\u003eLinear mixed models to evaluate gait parameters\u003c/p\u003e \u003cp\u003eLinear mixed models (LMMs) were used for associating digital gait features with different outcomes at the same visit. Therefore, clinical data at the same day of gait data assessment was matched with gait data. Only data from patients having both longitudinal clinical features, as well as, digital gait features available were used in this analysis. Characteristics of this patient subset is shown in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. Modeled outcomes are patient-specific a) latent disease time, b) clinical outcome scores, and c) progression rate. Both latent disease time and progression rate are a direct outcome of the LTJMM model described above. When modeling patient-specific latent disease time (interpreted as estimated time since disease onset), age at diagnosis and sex were used as adjustment covariates. In case of modeling clinical outcome scores or the patient-specific progression rate, we employed the latent disease time elapsed at the time of the clinical assessment as adjustment covariate to adjust for temporal differences between observed patient trajectories. Furthermore, we added a patient-specific random intercept in case of the Erlangen data, as digital gait features were available at several time points. Different sets of digital gait features were integrated as features in the model. First, all digital gait features were used jointly (\u003cb\u003eallGait\u003c/b\u003e), then features were selected according to the individual gait tasks, and lastly, we fitted LMMs for each digital gait feature separately (\u003cb\u003esingleGait\u003c/b\u003e). All these models were tested against a null model containing only adjustment covariates via a likelihood ratio test. Results were adjusted for multiple testing with the Benjamini \u0026amp; Yekutieli method [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eSimulation of a randomized controlled study\u003c/p\u003e \u003cp\u003eA randomized controlled trial was simulated over a one-year observation period with either original or gait predicted UPDRS III as study endpoints. The gait predicted UPDRS III was estimated from 4x10m tests\u0026rsquo; digital gait features using a linear mixed effect model. We assumed visits every 60 (bi-monthly), 30 (monthly) or 7 (weekly) days to assess the sample sizes needed for evaluating the efficacy of a potentially disease-modifying drug. We employed the Erlangen data for this task, as longitudinal digital gait features were required. Simulations were inspired by an ongoing trial [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e] and assumed a treatment effect on disease progression of 30%. Control and treatment groups were equally sized without different treatment dosage arms. A power of 80%, and a significance level of 0.01 was set as parameters for the outcome simulation of the disease progression difference between treatment and control group. Power and sample size was calculated using linear mixed-effects model, based on the method from Edland [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e], implemented in the R package \u003cem\u003elongpower\u003c/em\u003e [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e].\u003c/p\u003e \u003cp\u003ePredictive ML models for disease progression\u003c/p\u003e \u003cp\u003eRandom Forest [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e], XGBoost [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e] and Lasso [\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e] were used to predict the LTJMM random slope from digital gait features. In Erlangen cohort, where multiple gait visits were conducted, measurements from first and first plus second gait visit were used for prediction. Multiple feature sets were tested: only sex and age, all digital gait features (\u003cb\u003eallGait\u003c/b\u003e), and task-specific digital gait feature sets. Implementations of algorithms of scikit-learn [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e] and xgboost [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e] packages were used. Hyperparameter optimization was performed using a randomized search within an inner 5-fold cross validation procedure. Hyperparameter spaces for the different algorithms can be found in the \u003cb\u003eSupplementary Table \u003cspan refid=\"MOESM3\" class=\"InternalRef\"\u003eS3\u003c/span\u003e\u003c/b\u003e. Performance metrics were evaluated in an outer 5-fold repeated cross validation procedure. Performance of the models including gait features was additionally tested with a Wilcoxon signed-rank sum test comparing its \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({R}^{2}\\)\u003c/span\u003e\u003c/span\u003e values to those from a model integrating only sex and age as predictors. The best performing model was selected based on these statistical test results.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003eOverview\u003c/p\u003e \u003cp\u003eDisease trajectories of both studies, LuxPARK and Erlangen, were first aligned on a common disease timescale via LTJMM (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA\u003cb\u003e)\u003c/b\u003e with the aim to adjust trajectories of each patient according to their disease status and to remove bias introduced by including patients from different stages of Parkinson\u0026rsquo;s disease. After aligning the disease trajectories on that common disease timescale, digital gait features were used as predictors in multiple settings, either for monitoring the disease severity or the disease progression (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB\u003cb\u003e)\u003c/b\u003e and additionally for predicting disease severity and progression (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eC\u003cb\u003e)\u003c/b\u003e.\u003c/p\u003e \u003cp\u003eLatent time joint mixed-effect models for alignment of disease trajectories\u003c/p\u003e \u003cp\u003eLTJMM models were fitted to Erlangen and LuxPARK data. Both the original trajectories, as well as, the trajectories on the common disease timescale for each study and outcome are shown in the \u003cb\u003eSupplementary Figure \u003cspan refid=\"MOESM3\" class=\"InternalRef\"\u003eS3\u003c/span\u003e\u003c/b\u003e and \u003cb\u003eSupplementary Figure S4\u003c/b\u003e. \u003cb\u003eSupplementary Figure S5\u003c/b\u003e and \u003cb\u003eSupplementary Table S4\u003c/b\u003e confirms that the model orders patient trajectories correctly according to the H\u0026amp;Y stages. Additionally, on average Pearson correlations of real and predicted clinical outcomes were ρ\u0026thinsp;=\u0026thinsp;0.73 for Erlangen and ρ\u0026thinsp;=\u0026thinsp;0.42 for LuxPARK, indicating a sufficient fit of the LTJMM models.\u003c/p\u003e \u003cp\u003eMonitoring disease stage, motor symptom severity and motor symptom progression through digital gait\u003c/p\u003e \u003cp\u003eStatistical associations of clinical outcomes with features derived from the Portabiles digital gait device were analyzed with linear (mixed) models after aligning observed clinical outcome trajectories on a common disease timescale. We found a significant association of digital gait features with traditional clinical outcome measures in both datasets (LuxPARK: axial score, PIGD, TD and UPDRS III; Erlangen: axial score, UPDRS III) (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e and Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Clear task-specificity can be observed in LuxPark cohort, with the TUG being the most informative task. Here, the axial score was found to be highest correlated with digital gait features. When modeling the axial score with digital gait features, stance time (Count: p\u0026thinsp;=\u0026thinsp;0.04), swing time (Tray: p\u0026thinsp;=\u0026thinsp;0.009), gait speed (Tray: p\u0026thinsp;=\u0026thinsp;0.029), stride length (Tray: p\u0026thinsp;=\u0026thinsp;0.015) and the landing impact (TUG: p\u0026thinsp;=\u0026thinsp;0.047) were found to have significant effect sizes in the models. TUG\u0026rsquo;s landing impact shows also significant effect size when modelling the PIGD score (p\u0026thinsp;=\u0026thinsp;0.048). For the PIGD score as model outcome, we have also found the gait speed (p\u0026thinsp;=\u0026thinsp;0.014) and stride length (p\u0026thinsp;=\u0026thinsp;0.010) significant for the Count task. Similar as in Tray task in LuxPark, we could find the swing time in Erlangen\u0026rsquo;s 4x10m task significant (p\u0026thinsp;=\u0026thinsp;0.045) when modelling the axial score as outcome.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003e\u003cb\u003eAssociations of digital gait features with dedicated clinical scores, their slope and latent time in LuxPARK cohort.\u003c/b\u003e\u003c/p\u003e \u003cp\u003eAdjusted p-values of likelihood ratio testing the association of multiple digital gait feature sets, either all measured digital gait features or task-specific (Count, Tray, TUG, Turn), with clinical outcomes, adjusted for age at diagnosis and sex. Significant (adjusted p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) tests are marked in bold and weakly significant (adjusted p\u0026thinsp;\u0026lt;\u0026thinsp;0.1) tests in italic. Measured outcomes are the clinical scores (top), their slope (i.e. progression, middle) and latent time (i.e. disease stage, bottom).\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOutcome\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAll Gait\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCount\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTray\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eTUG\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eTurn\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colspan=\"6\" nameend=\"c6\" namest=\"c1\"\u003e \u003cp\u003eClinical score\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAxial score\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.0055\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0006\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.0006\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.0001\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.0175\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003ePIGD\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.2152\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e0.0068\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e0.0068\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.0001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cem\u003e0.0629\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eTD\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.0178\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.4882\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.4882\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eUPDRS I\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.2717\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cem\u003e0.0783\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.2384\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.2504\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cem\u003e0.0783\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eUPDRS II\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.4154\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.1427\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.1427\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.1427\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.3601\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eUPDRS III\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.0337\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.2149\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003e0.0538\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e0.0476\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.1058\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"6\" nameend=\"c6\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eSlope (progression)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAxial score\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.2765\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.5264\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003e0.0534\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cem\u003e0.0990\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.4970\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003ePIGD\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.6788\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eTD\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.4895\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.1709\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.1709\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.6128\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eUPDRS I\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.3035\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.8514\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003e0.0621\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cem\u003e0.0621\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.5288\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eUPDRS II\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.3864\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.9143\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.1260\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cem\u003e0.0686\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.4988\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eUPDRS III\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.6779\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"6\" nameend=\"c6\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eLatent time (disease stage)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eLatent time\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e0.0288\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003e0.0726\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e0.0005\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e0.0023\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003e\u003cb\u003eAssociations of digital gait features with dedicated clinical scores, their slope and latent time in Erlangen data.\u003c/b\u003e\u003c/p\u003e \u003cp\u003e\u003cem\u003eAdjusted p-values of likelihood ratio testing the association of multiple digital gait feature sets, either all measured digital gait features or task-specific (\u003c/em\u003eTUG, 2x10m With Stop, 4x10m Without Stop) \u003cem\u003ewith clinical outcomes, adjusted for age at diagnosis and sex. Significant (adjusted p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) tests are marked in bold and weakly significant (adjusted p\u0026thinsp;\u0026lt;\u0026thinsp;0.1) tests in italic. Measured outcomes are the clinical scores (top), their slope (i.e. progression, middle) and latent time (i.e. disease stage, bottom).\u003c/em\u003e\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOutcome\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAll Gait\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTUG\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2x10m With Stop\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e4x10m Without Stop\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003eClinical score\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eUPDRS III\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.1574\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e0.0361\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e0.0203\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e0.0110\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAxial score\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.0010\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e0.0068\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e0.0068\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e0.0068\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eSlope (progression)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eUPDRS III\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAxial score\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.0001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eLatent time (disease stage)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eLatent time\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.0047\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.0001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e0.0001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.0001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eFurthermore, a significant association with the disease stage, represented via the latent time, could be observed in LuxPARK, as well as in Erlangen (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e and Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). While we could find the maximal sensor lift significant in LuxPARK for both the TUG (p\u0026thinsp;=\u0026thinsp;0.031) and Turn (p\u0026thinsp;=\u0026thinsp;0.048) task, the stride time in 2x10m (p\u0026thinsp;=\u0026thinsp;0.001) and TUG (p\u0026thinsp;=\u0026thinsp;0.045) task was the most informative digital gait feature in Erlangen. Estimates for each of the models can be found in \u003cb\u003eSupplementary Table S6\u003c/b\u003e. A visualization of those, also showing the significance, can be found in the \u003cb\u003eSupplementary Information\u003c/b\u003e.\u003c/p\u003e \u003cp\u003eThe association with the slope of the axial score (i.e. progression) was clearly significant in the Erlangen data for the 4x10m test and demonstrated a weak significance in LuxPARK for the Tray and TUG test. Further weakly significant associations with digital gait included the slope of the UPDRS I (Tray, TUG tests) and the slope of the UPDRS II (TUG test).\u003c/p\u003e \u003cp\u003eAltogether, significant associations of digital gait features were thus identifiable with traditional clinical motor scores, their slopes as well as disease stage. Results of single digital gait features tests can be found in \u003cb\u003eSupplementary Table S5\u003c/b\u003e, generally supporting the findings reported in this section.\u003c/p\u003e \u003cp\u003eMachine learning based prediction of motor symptom progression with digital gait\u003c/p\u003e \u003cp\u003eWe trained different machine learning algorithms to predict the patient-specific slope (i.e. progression) of motor symptom related clinical outcomes. The strongest results could be observed via the Random Forest algorithm in a 10-fold cross-validation setup (see \u003cb\u003eSupplementary Figure S7\u003c/b\u003e) in the Erlangen data. When using only the first gait assessment in the Erlangen data for the prediction of the axial scores\u0026rsquo; slope, a significant improvement compared to only using age and sex as predictors could be observed, specifically when using data from the 4x10m task (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Integrating also a second assessment of gait improved the prediction even more and specifically when using data from all gait tasks. Additionally, the UPDRS III slope could now be predicted significantly better compared to only using age and sex as covariates. It should be noted that similar findings could not be made in LuxPARK, because digital gait features were only available at a single visit (see \u003cb\u003eSupplementary Figure S6\u003c/b\u003e). Altogether, our findings reported in this section support those made in the previous section, namely that digital gait is associated with traditional motor symptom scores from the UPDRS and is sensitive to changes in those scores.\u003c/p\u003e \u003cp\u003eRandomized controlled trial simulation shows benefit of digital gait features as endpoint\u003c/p\u003e \u003cp\u003eTo further evaluate the potential benefit of digital gait features compared to traditional clinical outcome scores, we simulated a longitudinal clinical trial with different visit frequencies. Notably, digital gait features can be collected more easily and cheaper at higher frequency compared to traditional clinical outcome scores, and data collection could potentially even be performed within a patient\u0026rsquo;s home environment. Estimates of the UPDRS III from digital gait features correlated well with the original data (ρ\u0026thinsp;=\u0026thinsp;0.77, see \u003cb\u003eSupplementary Figure S8\u003c/b\u003e). According to our statistical sample size calculation, which assumes a treatment effect size of 30%, bi-monthly assessments of the UPDRS III require 690 patients in total to reach 80% statistical power at a significance level of 1%. At this visit frequency the same sample size of 690 patients would be needed for a digital endpoint estimating the UPDRS III from digital gait features via the 4x10m task (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). However, already a monthly assessment of this gait predicted clinical endpoint can reduce the number of patients to 550, and a weekly assessment even to 380 patients, showing a potential reduction of almost 44%. We would like to highlight once more that assessments of the traditional UPDRS III at a similar frequency in clinic would be highly challenging.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eDigital Gait Assessments for Symptom Monitoring and Prognosis\u003c/p\u003e \u003cp\u003eFeatures derived from the Portabiles digital gait device were significantly associated with disease stage (represented by latent time), disease severity assessed via the Unified Parkinson's Disease Rating Scale (UPDRS), specifically the axial score, and the slope of the axial score. Although, the association between axial abnormalities and digital gait with more severe motor symptoms, as well as with lower step lengths and gait speed, has been described by several authors before [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e], correlation analysis with digital gait features was still lacking.\u003c/p\u003e \u003cp\u003eAdditionally, most studies analyzing digitally assessed gait features, provide only micro analyses [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e, \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e], looking at single digital gait features, but no overall analysis of multiple digital gait features. Here, a comprehensive analysis of single, but also multiple combined features \u0026ndash; on several levels \u0026ndash; is provided. In comparison to other studies, which used only one or even no standardized tasks, here, multiple tasks had to be performed by patients and task specificity was examined, providing a systematic and unique overview about most useful tasks for digital gait analysis. Although, combining all digital gait features turned out to be useful, there were notable task-specific differences between gait tasks. Interestingly, we found the traditional \u0026lsquo;Time Up And Go\u0026rsquo; (TUG) task consistently effective for monitoring the motor symptom severity in both studies. Additionally, it was the only task that could monitor the UPDRS III in the LuxPARK cohort. Digital gait features derived from Tray and Count tasks could also be associated with PIGD score, indicating that these dual tasks, combining motor and cognitive exercises, are efficient in monitoring disease severity in terms of postural instability. In the past, dual tasks also have been seen to trigger Freezing of Gait [\u003cspan additionalcitationids=\"CR37\" citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e]. Thus, investigating the ability to monitor cognitive impairments should be subject for further research.\u003c/p\u003e \u003cp\u003eAs expected, a clearly significant association of digital gait features with UPDRS derived motor symptom progression could only be established in the Erlangen dataset, where gait measurements have been taken over a longer period of time. However, this association could only be established for the 4x10m task. This highlights the importance of longer distance walks to identify changes of motor symptoms within an individual patient over several visits. Despite this, prediction of the UPDRS III from digital gait data only is a feasible task, as such estimates correlate well with the original data (ρ\u0026thinsp;=\u0026thinsp;0.88), although not integrating any information about symptoms other than gait (e.g. hypomimia, dysarthria, or tremor). Potentially, the severity of these other symptoms is correlated to digital gait, or they are relatively stable in terms of disease progression such that their contribution to the UPDRS III progression is low. Right now, the reasons are unclear and further research might help to investigate on this relationship between digital gait and other PD symptoms.\u003c/p\u003e \u003cp\u003eAlthough other studies have shown relationships between digital gait features and multiple disease stages, none of them quantify disease progression, but only classifying patients into disease stage groups [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e, \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]. Often the analysis focuses only on control vs. PD classification [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e, \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e]. In contrast, our analysis focused on the association of digitally assessed gait features with motor symptoms and their progression. However, machine learning based prediction of the patient-level slope of UPDRS derived motor symptoms, demonstrated only a limited accuracy. When combining data from two subsequent clinical visits the coefficient of determination (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({R}^{2}\\)\u003c/span\u003e\u003c/span\u003e) for predicting the slope of the axial score reached 0.45 and for the UPDRS III approximately 0.35. Hence, digital gait features alone have only limited prognostic value despite their ability to objectively monitor motor symptoms and their progression.\u003c/p\u003e \u003cp\u003eOverall, the approach presented here demonstrates the feasibility of monitoring motor symptoms through digital gait assessment and its task specificity for individual symptoms across multiple independent studies.\u003c/p\u003e \u003cp\u003ePotential Value for Clinical Trials\u003c/p\u003e \u003cp\u003eSignificant association between the patient-specific slope of the axial score and digital gait features, was found. We thus performed a randomized control trial simulation, with a predicted UPDRS III outcome using the Portabiles digital gait device derived features from the 4x10m test. Interestingly, a digital gait predicted outcome (i.e. a digitized version of the UPDRS III), although not integrating any information about other symptoms than gait, as mentioned above, requires the same number of patients to show the efficacy of a treatment as the standard UPDRS III, if gait features are collected at the same frequency. However, a large reduction of the number of needed patients could be shown when assessing the \u0026ldquo;digital\u0026rdquo; UPDRS III via the Portabiles digital gait device in a weekly manner.\u003c/p\u003e \u003cp\u003eThis shows the potential for using gait sensors and associated data in clinical studies, where measurements with a digital device could be taken remotely at a much higher frequency than in a traditional in-hospital setting. This may decrease the burden for patients\u0026rsquo; quality of life and increase their willingness to contribute to a clinical study as they are not required to visit the clinic for each assessment. Additionally, data in general can be recorded via the 4x10m task much easier and faster than performing a complete UPDRS exam.\u003c/p\u003e \u003cp\u003eOnly limited studies have already investigated the benefit of using digital gait device data for clinical trials. While sensor-based digital gait analysis has shown its potential to overcome limitations of UPDRS based gait assessment in a phase II trial recently [\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e], most studies only refer to a subjective potential endpoint in studies, but do not test the validity of this hypothesis. Here, for the first time, an objective comparison of both outcomes, original and gait predicted UPDRS III, quantifies the potential of using sensor-based digital gait data as secondary endpoint in trials. Although the potential of the sensor-based digital gait data could clearly be seen, it does not capture quality of life aspects during a trial, which is a limitation for phase III studies and still needs to be addressed by other instruments.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eOverall, we have demonstrated the ability to objectively monitor the severity and progression of motor symptoms using the Portabiles HCT digital gait device. Additionally, we investigated the potential benefit of using digital gait as an outcome measure in a clinical trial context. Our work thus adds to the increasing body of literature demonstrating the benefit of digital solutions in the PD field. A distinction point is our focused digital gait device, the analysis of symptom progression and the exploration of the benefit for clinical studies.\u003c/p\u003e \u003cp\u003eWe would like to emphasize that the use of any medical device, including the tested Portabiles digital gait device, outside pure research settings in routine medical care requires approval by a regulatory agency, and in that regard Portabiles has recently launched a request to the FDA in the USA. Besides that, it is already certified as a class I medical device (MDR) with a CE mark in Europe.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eCompeting Interests\u003c/h2\u003e\u003cp\u003eRalph Steidl is the CEO of Portabiles HCT. Jean-Christophe Corvol has served in advisory boards for Alzprotect, Bayer, Biogen, Denali, Ferrer, Idorsia, iRegene, Prevail Therapeutic, Servier, Theranexus, UCB; and received grants from Sanofi and the Michael J Fox Foundation outside of this work. Holger Fr\u0026ouml;hlich has received grants from UCB and AbbVie outside this work.Tamara Raschka, Jackrite To, Tom H\u0026auml;hnel, Stefano Sapienza, Alzhraa Ibrahim, Enrico Glaab, Heiko Ga\u0026szlig;ner, J\u0026uuml;rgen Winkler, and Jochen Klucken declare no competing interests.\u003c/p\u003e\u003ch2\u003eFunding\u003c/h2\u003e \u003cp\u003eThis study was funded by the ERA PerMed EU-wide project DIGIPD (01KU2110). This project is additionally supported by the Luxembourg National Research Fund (FNR) (INTER/ERAPerMed 20/14599012). LuxPARK is part of the National Centre of Excellence in Research on Parkinson's Disease (NCER-PD), which is funded by the Luxembourg National Research Fund (FNR/NCER13/BM/11264123). The funders played no role in study design, data collection, analysis and interpretation of data, or the writing of this manuscript. Data from Erlangen is approved under (IRB-approval-Re. -No. 4208, 21.04.2010, IRB, Medical Faculty, Friedrich-Alexander University Erlangen-N\u0026uuml;rnberg, Germany) and by local ethics committee (reference number: 166_18 B, Medical Faculty, FAU Erlangen-N\u0026uuml;rnberg, Germany).\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eR.S., J.W., J.C., J.K., and H.F. designed the study. S.S., A.I., E.G., R.S., J.W., E.G., H.G., J.C., J.K., and H.F. contributed to the acquisition of data. T.R., J.T., and T.H. analyzed and interpreted the data. T.R. wrote the initial draft of the paper. H.F. substantively revised it. All authors discussed the results and contributed to revising the paper. All authors approved the final version of the manuscript.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eAs this study is a retrospective analysis, availability of the clinical data depends on the individual study groups (LuxPARK: [email protected]; Erlangen: [email protected]).\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003ePostuma RB, Berg D, Stern M, Poewe W, Olanow CW, Oertel W, et al. MDS clinical diagnostic criteria for Parkinson\u0026rsquo;s disease: MDS-PD Clinical Diagnostic Criteria. Mov Disord. 2015;30: 1591\u0026ndash;1601. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/mds.26424\u003c/span\u003e\u003cspan address=\"10.1002/mds.26424\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMoustafa AA, Chakravarthy S, Phillips JR, Gupta A, Keri S, Polner B, et al. Motor symptoms in Parkinson\u0026rsquo;s disease: A unified framework. Neurosci Biobehav Rev. 2016;68: 727\u0026ndash;740. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.neubiorev.2016.07.010\u003c/span\u003e\u003cspan address=\"10.1016/j.neubiorev.2016.07.010\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXia R, Mao Z-H. Progression of motor symptoms in Parkinson\u0026rsquo;s disease. Neurosci Bull. 2012;28: 39\u0026ndash;48. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s12264-012-1050-z\u003c/span\u003e\u003cspan address=\"10.1007/s12264-012-1050-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMirelman A, Bonato P, Camicioli R, Ellis TD, Giladi N, Hamilton JL, et al. Gait impairments in Parkinson\u0026rsquo;s disease. Lancet Neurol. 2019;18: 697\u0026ndash;708. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/S1474-4422(19)30044-4\u003c/span\u003e\u003cspan address=\"10.1016/S1474-4422(19)30044-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChandrabhatla AS, Pomeraniec IJ, Ksendzovsky A. Co-evolution of machine learning and digital technologies to improve monitoring of Parkinson\u0026rsquo;s disease motor symptoms. Npj Digit Med. 2022;5: 32. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41746-022-00568-y\u003c/span\u003e\u003cspan address=\"10.1038/s41746-022-00568-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFr\u0026ouml;hlich H, Bontridder N, Petrovska-Delacr\u0026eacute;ta D, Glaab E, Kluge F, Yacoubi ME, et al. Leveraging the Potential of Digital Technology for Better Individualized Treatment of Parkinson\u0026rsquo;s Disease. Front Neurol. 2022;13: 788427. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fneur.2022.788427\u003c/span\u003e\u003cspan address=\"10.3389/fneur.2022.788427\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGoetz CG, Tilley BC, Shaftman SR, Stebbins GT, Fahn S, Martinez-Martin P, et al. Movement Disorder Society‐sponsored revision of the Unified Parkinson\u0026rsquo;s Disease Rating Scale (MDS‐UPDRS): Scale presentation and clinimetric testing results. Mov Disord. 2008;23: 2129\u0026ndash;2170. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/mds.22340\u003c/span\u003e\u003cspan address=\"10.1002/mds.22340\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKirk C, Zia Ur Rehman R, Galna B, Alcock L, Ranciati S, Palmerini L, et al. Can Digital Mobility Assessment Enhance the Clinical Assessment of Disease Severity in Parkinson\u0026rsquo;s Disease? J Park Dis. 2023;13: 999\u0026ndash;1009. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3233/JPD-230044\u003c/span\u003e\u003cspan address=\"10.3233/JPD-230044\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNair P, Shojaei Baghini M, Pendharkar G, Chung H. Detecting early-stage Parkinson\u0026rsquo;s disease from gait data. Proc Inst Mech Eng [H]. 2023; 09544119231197090. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1177/09544119231197090\u003c/span\u003e\u003cspan address=\"10.1177/09544119231197090\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDeng K, Li Y, Zhang H, Wang J, Albin RL, Guan Y. Heterogeneous digital biomarker integration out-performs patient self-reports in predicting Parkinson\u0026rsquo;s disease. Commun Biol. 2022;5: 58. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s42003-022-03002-x\u003c/span\u003e\u003cspan address=\"10.1038/s42003-022-03002-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDjurić-Jovičić M, Belić M, Stanković I, Radovanović S, Kostić VS. Selection of gait parameters for differential diagnostics of patients with \u003cem\u003ede novo\u003c/em\u003e Parkinson\u0026rsquo;s disease. Neurol Res. 2017;39: 853\u0026ndash;861. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1080/01616412.2017.1348690\u003c/span\u003e\u003cspan address=\"10.1080/01616412.2017.1348690\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAdams JL, Kangarloo T, Tracey B, O\u0026rsquo;Donnell P, Volfson D, Latzman RD, et al. Using a smartwatch and smartphone to assess early Parkinson\u0026rsquo;s disease in the WATCH-PD study. Npj Park Dis. 2023;9: 64. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41531-023-00497-x\u003c/span\u003e\u003cspan address=\"10.1038/s41531-023-00497-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePistacchi M. Gait analysis and clinical correlations in early Parkinson\u0026rsquo;s disease. Funct Neurol. 2017;32: 28. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.11138/FNeur/2017.32.1.028\u003c/span\u003e\u003cspan address=\"10.11138/FNeur/2017.32.1.028\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCai G, Shi W, Wang Y, Weng H, Chen L, Yu J, et al. Specific Distribution of Digital Gait Biomarkers in Parkinson\u0026rsquo;s Disease Using Body-Worn Sensors and Machine Learning. Duque G, editor. J Gerontol Ser A. 2023;78: 1348\u0026ndash;1354. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/gerona/glad101\u003c/span\u003e\u003cspan address=\"10.1093/gerona/glad101\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBeswick E, Fawcett T, Hassan Z, Forbes D, Dakin R, Newton J, et al. A systematic review of digital technology to evaluate motor function and disease progression in motor neuron disease. J Neurol. 2022;269: 6254\u0026ndash;6268. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s00415-022-11312-7\u003c/span\u003e\u003cspan address=\"10.1007/s00415-022-11312-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhu S, Wu Z, Wang Y, Jiang Y, Gu R, Zhong M, et al. Gait Analysis with Wearables Is a Potential Progression Marker in Parkinson\u0026rsquo;s Disease. Brain Sci. 2022;12: 1213. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3390/brainsci12091213\u003c/span\u003e\u003cspan address=\"10.3390/brainsci12091213\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDiao JA, Raza MM, Venkatesh KP, Kvedar JC. Watching Parkinson\u0026rsquo;s disease with wrist-based sensors. Npj Digit Med. 2022;5: 73, s41746-022-00619\u0026ndash;4. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41746-022-00619-4\u003c/span\u003e\u003cspan address=\"10.1038/s41746-022-00619-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi D, Iddi S, Thompson WK, Donohue MC, Alzheimer\u0026rsquo;s Disease Neuroimaging Initiative. Bayesian latent time joint mixed effect models for multicohort longitudinal data. Stat Methods Med Res. 2019;28: 835\u0026ndash;845. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1177/0962280217737566\u003c/span\u003e\u003cspan address=\"10.1177/0962280217737566\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHipp G, Vaillant M, Diederich NJ, Roomp K, Satagopam VP, Banda P, et al. The Luxembourg Parkinson\u0026rsquo;s Study: A Comprehensive Approach for Stratification and Early Diagnosis. Front Aging Neurosci. 2018;10: 326. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fnagi.2018.00326\u003c/span\u003e\u003cspan address=\"10.3389/fnagi.2018.00326\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAleksovski D, Miljkovic D, Bravi D, Antonini A. Disease progression in Parkinson subtypes: the PPMI dataset. Neurol Sci. 2018;39: 1971\u0026ndash;1976. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s10072-018-3522-z\u003c/span\u003e\u003cspan address=\"10.1007/s10072-018-3522-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHayashi Y, Mishima T, Fujioka S, Morishita T, Inoue T, Nagamachi S, et al. Unilateral GPi-DBS Improves Ipsilateral and Axial Motor Symptoms in Parkinson\u0026rsquo;s Disease as Evidenced by a Brain Perfusion Single Photon Emission Computed Tomography Study. Front Hum Neurosci. 2022;16: 888701. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fnhum.2022.888701\u003c/span\u003e\u003cspan address=\"10.3389/fnhum.2022.888701\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFahn S. Unified Parkinson\u0026rsquo;s disease rating scale. Recent Dev Park Dis. 1987; 153\u0026ndash;163.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eH\u0026auml;hnel T, Raschka T, Sapienza S, Klucken J, Glaab E, Corvol J-C, et al. Progression subtypes in Parkinson\u0026rsquo;s disease identified by a data-driven multi cohort analysis. Npj Park Dis. 2024;10: 95. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41531-024-00712-3\u003c/span\u003e\u003cspan address=\"10.1038/s41531-024-00712-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBenjamini Y, Yekutieli D. The control of the false discovery rate in multiple testing under dependency. Ann Stat. 2001;29. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1214/aos/1013699998\u003c/span\u003e\u003cspan address=\"10.1214/aos/1013699998\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUCB Biopharma SRL. A Double-Blind, Placebo-Controlled, Randomized, 18-Month Phase 2a Study to Evaluate the Efficacy, Safety, Tolerability, and Pharmacokinetics of Oral UCB0599 in Study Participants With Early Parkinson\u0026rsquo;s Disease. clinicaltrials.gov. 2023. Available: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://clinicaltrials.gov/study/NCT04658186\u003c/span\u003e\u003cspan address=\"https://clinicaltrials.gov/study/NCT04658186\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (accessed July 16, 2023)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eArd MC, Edland SD. Power Calculations for Clinical Trials in Alzheimer\u0026rsquo;s Disease. Ashford JW, Rosen A, Adamson M, Bayley P, Sabri O, Furst A, et al., editors. J Alzheimers Dis. 2011;26: 369\u0026ndash;377. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3233/JAD-2011-0062\u003c/span\u003e\u003cspan address=\"10.3233/JAD-2011-0062\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIddi S, C Donohue M. Power and Sample Size for Longitudinal Models in R -- The longpower Package and Shiny App. R J. 2022;14: 264\u0026ndash;282. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.32614/RJ-2022-022\u003c/span\u003e\u003cspan address=\"10.32614/RJ-2022-022\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBreiman L. Random Forests. Mach Learn. 2001;45: 5\u0026ndash;32. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1023/A:1010933404324\u003c/span\u003e\u003cspan address=\"10.1023/A:1010933404324\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen T, Guestrin C. XGBoost: A Scalable Tree Boosting System. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. San Francisco California USA: ACM; 2016. pp. 785\u0026ndash;794. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1145/2939672.2939785\u003c/span\u003e\u003cspan address=\"10.1145/2939672.2939785\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTibshirani R. Regression Shrinkage and Selection Via the Lasso. J R Stat Soc Ser B Methodol. 1996;58: 267\u0026ndash;288. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1111/j.2517-6161.1996.tb02080.x\u003c/span\u003e\u003cspan address=\"10.1111/j.2517-6161.1996.tb02080.x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, et al. Scikit-learn: Machine learning in Python. J Mach Learn Res. 2011;12: 2825\u0026ndash;2830.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCao S, Cui Y, Jin J, Li F, Liu X, Feng T. Prevalence of axial postural abnormalities and their subtypes in Parkinson\u0026rsquo;s disease: a systematic review and meta-analysis. J Neurol. 2023;270: 139\u0026ndash;151. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s00415-022-11354-x\u003c/span\u003e\u003cspan address=\"10.1007/s00415-022-11354-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePongmala C, Fabbri M, Zibetti M, Pitakpatapee Y, Wangthumrong T, Sangpeamsook T, et al. Gait and axial postural abnormalities correlations in Parkinson\u0026rsquo;s disease: A multicenter quantitative study. Parkinsonism Relat Disord. 2022;105: 19\u0026ndash;23. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.parkreldis.2022.10.026\u003c/span\u003e\u003cspan address=\"10.1016/j.parkreldis.2022.10.026\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchlachetzki JCM, Barth J, Marxreiter F, Gossler J, Kohl Z, Reinfelder S, et al. Wearable sensors objectively measure gait parameters in Parkinson\u0026rsquo;s disease. Toft M, editor. PLOS ONE. 2017;12: e0183989. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1371/journal.pone.0183989\u003c/span\u003e\u003cspan address=\"10.1371/journal.pone.0183989\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWelzel J, Wendtland D, Warmerdam E, Romijnders R, Elshehabi M, Geritz J, et al. Step Length Is a Promising Progression Marker in Parkinson\u0026rsquo;s Disease. Sensors. 2021;21: 2292. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3390/s21072292\u003c/span\u003e\u003cspan address=\"10.3390/s21072292\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSnijders AH, Haaxma CA, Hagen YJ, Munneke M, Bloem BR. Freezer or non-freezer: Clinical assessment of freezing of gait. Parkinsonism Relat Disord. 2012;18: 149\u0026ndash;154. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.parkreldis.2011.09.006\u003c/span\u003e\u003cspan address=\"10.1016/j.parkreldis.2011.09.006\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSpildooren J, Vercruysse S, Desloovere K, Vandenberghe W, Kerckhofs E, Nieuwboer A. Freezing of gait in Parkinson\u0026rsquo;s disease: The impact of dual-tasking and turning. Mov Disord. 2010;25: 2563\u0026ndash;2570. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/mds.23327\u003c/span\u003e\u003cspan address=\"10.1002/mds.23327\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRaccagni C, Nonnekes J, Bloem BR, Peball M, Boehme C, Seppi K, et al. Gait and postural disorders in parkinsonism: a clinical approach. J Neurol. 2020;267: 3169\u0026ndash;3176. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s00415-019-09382-1\u003c/span\u003e\u003cspan address=\"10.1007/s00415-019-09382-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMic\u0026oacute;-Amigo ME, Kingma I, Heinzel S, Rispens SM, Heger T, Nussbaum S, et al. Potential Markers of Progression in Idiopathic Parkinson\u0026rsquo;s Disease Derived From Assessment of Circular Gait With a Single Body-Fixed-Sensor: A 5 Year Longitudinal Study. Front Hum Neurosci. 2019;13: 59. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fnhum.2019.00059\u003c/span\u003e\u003cspan address=\"10.3389/fnhum.2019.00059\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDi Biase L, Di Santo A, Caminiti ML, De Liso A, Shah SA, Ricci L, et al. Gait Analysis in Parkinson\u0026rsquo;s Disease: An Overview of the Most Accurate Markers for Diagnosis and Symptoms Monitoring. Sensors. 2020;20: 3529. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3390/s20123529\u003c/span\u003e\u003cspan address=\"10.3390/s20123529\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCoates L, Shi J, Rochester L, Del Din S, Pantall A. Entropy of Real-World Gait in Parkinson\u0026rsquo;s Disease Determined from Wearable Sensors as a Digital Marker of Altered Ambulatory Behavior. Sensors. 2020;20: 2631. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3390/s20092631\u003c/span\u003e\u003cspan address=\"10.3390/s20092631\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePayne T, Appleby M, Buckley E, Van Gelder LMA, Mullish BH, Sassani M, et al. A Double-Blind, Randomized, Placebo‐Controlled Trial of Ursodeoxycholic Acid (UDCA) in Parkinson\u0026rsquo;s Disease. Mov Disord. 2023;38: 1493\u0026ndash;1502. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/mds.29450\u003c/span\u003e\u003cspan address=\"10.1002/mds.29450\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-4521747/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4521747/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eDigital technologies for monitoring motor symptoms of Parkinson\u0026rsquo;s Disease (PD) have underwent a strong evolution during the past years. Although it has been shown for several devices that derived digital gait features can reliably discriminate between healthy controls and people with PD, the specific gait tasks best suited for monitoring motor symptoms and especially their progression, remain unclear. Furthermore, the potential benefit as endpoint in a clinical trial context has not been investigated so far.\u003c/p\u003e \u003cp\u003eIn this study we employed a digital gait device manufactured by Portabiles HCT, which has been used by 339 patients within the LuxPark cohort (n\u0026thinsp;=\u0026thinsp;161, Luxembourg) as well as within routine clinical care visits at the University Medical Center Erlangen (n\u0026thinsp;=\u0026thinsp;178, Erlangen, Germany). Linear (mixed) models were used to assess the association of task-specific digital gait features with disease progression and motor symptom severity measured by several clinical scores. Furthermore, we employed machine learning to evaluate whether digital gait assessments were prognostic for patient-level motor symptom progression.\u003c/p\u003e \u003cp\u003eOverall, digital gait features derived from Portabiles digital gait device were found to effectively monitor motor symptoms and their longitudinal progression. At the same time the prognostic performance of digital gait features was limited. However, we could show a strong reduction in required sample size, if digital gait features were employed as surrogates for traditional endpoints in a clinical trial context. Thus, Portabiles digital gait device provides an effective way to objectively monitor motor symptoms and their progression in PD. Furthermore, the digital gait device bears strong potential as an alternative and easily assessable endpoint predictor in a clinical trial context.\u003c/p\u003e","manuscriptTitle":"Objective Monitoring of Motor Symptom Severity and their Progression in Parkinson’s Disease Using a Digital Gait Device","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-06-25 20:41:10","doi":"10.21203/rs.3.rs-4521747/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-02-12T06:12:10+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-12-26T15:13:37+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"194422509395302552764421161023811967168","date":"2024-12-15T15:11:57+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-11-30T06:12:46+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"19588483627493729151349556191387250659","date":"2024-11-18T21:37:53+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"281372932263439452771890557515410991541","date":"2024-09-06T15:02:07+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"136907032855068691878054858683667618112","date":"2024-09-06T12:52:57+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"224441573471707552201790665531668097103","date":"2024-07-21T15:40:40+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"278050973226117294155909095357113123863","date":"2024-07-15T23:52:20+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2024-07-15T20:40:05+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-07-10T20:13:35+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2024-06-10T08:30:04+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-06-06T04:47:52+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2024-06-03T12:23:32+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"92000c4f-f177-4011-8938-0c0961d6ed9e","owner":[],"postedDate":"June 25th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":33423445,"name":"Biological sciences/Neuroscience/Diseases of the nervous system/Parkinsons disease"},{"id":33423446,"name":"Health sciences/Biomarkers"},{"id":33423447,"name":"Health sciences/Diseases/Neurological disorders/Neurodegenerative diseases/Parkinsons disease"},{"id":33423448,"name":"Health sciences/Medical research/Biomarkers"}],"tags":[],"updatedAt":"2025-07-21T16:13:17+00:00","versionOfRecord":{"articleIdentity":"rs-4521747","link":"https://doi.org/10.1038/s41598-025-09088-7","journal":{"identity":"scientific-reports","isVorOnly":false,"title":"Scientific Reports"},"publishedOn":"2025-07-15 16:05:36","publishedOnDateReadable":"July 15th, 2025"},"versionCreatedAt":"2024-06-25 20:41:10","video":"","vorDoi":"10.1038/s41598-025-09088-7","vorDoiUrl":"https://doi.org/10.1038/s41598-025-09088-7","workflowStages":[]},"version":"v1","identity":"rs-4521747","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4521747","identity":"rs-4521747","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}