The use of wearable/portable digital sensors in Huntington’s disease: a systematic review

1 In chronic neurological conditions, wearable/portab le devices have potential as innovative 2 tools to detect subtle early disease manifestations and disease fluctuations for the purpose of 3 clinical diagnosis, care and therapeutic developmen t. Huntington’s disease (HD) has a unique 4 combination of motor and non-motor features which, combined with recent and anticipated 5 therapeutic progress, gives great potential for suc h devices to prove useful. The present work 6 aims to provide a comprehensive account of the use of wearable/portable devices in HD and 7 of what they have contributed so far. We conducted a systematic review searching 8 MEDLINE, Embase, and IEEE Xplore. Thirty references were identified. Our results 9 revealed large variability in the types of sensors used, study design, and the measured 10 outcomes. Digital technologies show considerable pr omise for therapeutic research and 11 clinical management of HD. However, more studies wi th standardized devices and 12 harmonized protocols are needed to optimize the pot ntial applicability of wearable/portable 13 devices in HD. 14


Introduction
Huntington's disease (HD) is an autosomal dominant neurodegenerative disorder caused by 2 an expanded trinucleotide CAG repeat in the HTT gene. [1] Clinically it is characterized by 3 motor, behavioural, and cognitive signs and symptoms. 4 The natural history of HTT expansion carriers is divided into premanifest and manifest 5 phases, with "clinical onset" diagnosed on the basis of "unequivocal" motor signs such as 6 chorea.[2, 3] However, a long prodromal phase, lasting a decade or more, frequently precedes 7 this point and brings subtle motor, cognitive and behavioural features that can nonetheless be 8 disabling.
[4] 9 Furthermore, signs and symptoms in HD can be extremely heterogeneous among patients and 10 can also vary over time in the same patient in a non-linear manner. For example, motor 11 impairment can range from the classical hyperkinetic involuntary movements to a more 12 subtle hypokinetic impairment of voluntary movements, as well as impairment of motor 13 coordination.
[5] Additionally, signs and symptoms can also display short-term fluctuations. 14 Phenotypic variability and the difficulty in consistently detecting subtle early clinical 15 manifestations pose challenges to therapeutic development as well as clinical management. 16 The Unified Huntington's Disease Rating Scale Total Motor Score (UHDRS TMS), has been 17 "recommended" by the International Parkinson and Movement Disorder Society (MDS) for 18 the assessment of motor signs in HD [6] and included by the National Institute for 19 Neurological Disorder and Stroke HD group in a list of recommended sensitive outcome 20 measures to be used as primary or secondary endpoints HD clinical trials.
[7] However, its use 21 in clinical trials has shown limited sensitivity, especially in the pre-manifest stage of HD. [3,22 8] It is also unreliable in capturing day-to-day or minute-to-minute variability of motor signs 23 which could easily dwarf any treatment effect. In addition to more reliable measures, there is 24 J o u r n a l P r e -p r o o f therefore face value in assessing manifestations of HD over a longer period, with high-1 frequency or continuous monitoring. 2 Quantitative measures of motor and cognitive alterations in HD can be an optimal tool to 3 detect and monitoring subtle modifications even in pre-manifest HD. [9,10] However, such 4 quantitative assessment is mainly based on expensive and cumbersome technology that can 5 only be used in-clinic settings for limited time periods. [11] 6 Recently, advances in wearable/portable sensors, information and communication 7 technologies, have enabled a continuous monitoring of chronic diseases. The use of 8 wearable/portable sensors allows the collection of high-dimensional data from multiple 9 domains and during everyday activities, in order to obtain a detailed, objective and precise 10 picture of disease manifestations. In addition, GPS data can provide evidence on real-world 11 mobility and be a surrogate of social activity, while smartphones and other devices can also 12 be used to implement questionnaires about symptoms or cognitive tasks. The high spatial and 13 temporal resolution of the registered data allows the monitoring of long-term trends and 14 short-term fluctuations of symptoms, as well as the detection of "soft" signs and symptoms of 15 disease onset/progression, or of therapeutic response that would otherwise go unnoticed. By 16 improving signal to noise ratios, this could be useful to increase the power of clinical trials 17 for new drug discovery. The term 'digital biomarkers' is sometimes used to denote the 18 meaningful outputs derived from electronic sensor data, whether or not the equipment used is 19 wearable/portable. 20 Such technologies are still in their infancy when it comes to implementation in such settings. in AD clinical trials in parallel with other already accepted and widely-used measures. [14] 2 We undertook a systematic review to provide a comprehensive overview of the use of such 3 devices in HD and provide an evidence basis to comment on possible future directions.

Review process
3 Two review authors independently screened for eligibility the titles and abstracts of all 4 identified references. The full-text of all potentially eligible reports were retrieved and 5 screened using the same procedure. Disagreements were resolved by discussion, or by 6 consulting a third author. 7 Validity analysis 8 We conducted a validity analysis of the included wearable/portable devices/tools. We 9 followed the strategy proposed by the Movement Disorder Society Committee on Rating 10 Scales Development to appraise clinical assessment tools in HD. [6,[20][21][22][23] We included 11 seven criteria with a Yes/No/Not Applicable response, namely: 1-used in HD, 2-used in HD 12 by more than one group, 3-test-retest reliability, 4-ability to discriminate cases from 13 controls, 5-ability to capture disease stage/severity, 6-ability to capture changes over time, 14 7-ability to detect therapeutic response. The answer "Not Applicable" referred to the fact that 15 that criterion has never been investigated for that specific device/tool in HD.

18
Search results 19 The electronic search returned 2489 records (MEDLINE 382; Embase 1711; IEEE Xplore to study outcomes (1 did not specify the inertial/wearable sensors used, and another presented 1 no results). Additional four original articles were included after a manual search across the 2 references of the assessed full-texts. At the end of the evaluation process, according to the 3 eligibility criteria, 30 references were included in the final review ( Figure 1). Two 4 references[24, 25] refer to the same study, but present different analyses and results, so we 5 did not consider them as duplicates. 6 General characteristics of the included studies 7 The main characteristics of the included studies are listed in Table 1. Twenty-one of them 8 were published in indexed journals, while 9 were presented at international conferences.[26-  11 The majority of the studies were focused on manifest HD, with only six including pre-  Types of sensors 1 Accelerometers were the type of sensors used most, initially uniaxial and later mainly tri-2 axial. Only Saadeh and colleagues [32] proposed the use of a Flexi-force sensing resistor 3 (FSR: https://ww.tekscan.com/products-solutions/force-sensors/a201. Figure 2a), a thin, 4 flexible piezoresistive force sensor. The sensor was placed unobtrusively into the shoe sole, 5 and was able to translate the force applied in a designed sensing area into gait data, 6 subsequently acquired and processed in a detection processor able to extract the 7 discriminating features to classify different neurodegenerative diseases. The acquired 8 information was then transferred to a mobile phone through a Bluetooth/Cloud network. [32] 9 The studies of Waddel and colleagues and Lauraitis and colleagues didn't use any kind of 10 motor sensor and they were based on an app for smartphone or tablet.[34, 47] 11 With advances in technology, the tested devices became lighter, smaller, and characterized by  questionnaires about mood, quality of life, and general wellbeing; cognitive tests, namely the 10 Symbol-digit Modalities Test and the Stroop Word Reading Test; motor tasks, such as the 11 Speed Tapping Test, the Draw a Shape Test, the Chorea Test, the Balance Test, and the U-12 Turn Test. Furthermore, the smartphone was equipped with a GPS, in order to register the 13 daily activities of the participants (Figure 3). Both the devices were designed for long-term showed that the majority of participants found the sensors "comfortable" and "easy to  [49] In the study of 1 Kegelmeyer and colleagues, wearable accelerometers were used for rehabilitation purposes in 2 order to adjust trunk movements and reflexes in HD patients. [50] Youdan and colleagues 3 showed dual-task impairment in HD, reporting an increased total sway area, decreased gait 4 speed and decreased correct response to cognitive tasks in HD participants who performed 5 motor and cognitive tasks at the same time.
[30] 6 Extracting meaningful and useful outcomes from high-dimension datasets is a major 7 challenge as digital biomarker technology becomes ever more complex. That was the reason 8 why some of the studies focused on advanced machine learning approaches and new 9 algorithms or analysis methods to extract parameters with the best discrimination ability and 10 increase the classification accuracy between HD and controls.[25, 32, 37, 44, 51, 52] 11 However, none of the proposed algorithms has been reproduced in a replication cohort. 13 The results are reported in Supplementary Table 1   as of other disease characteristics like trunk sway or sleep patterns/movements using 8 wearable/portable devices can be a reliable approach to identify patients in the manifest stage 9 of the disease and they are promising in the characterization of the pre-manifest and early 10 manifest phases as well. This is of a huge interest because advanced wearable technologies 11 represent a revolutionary approach in collecting data. They are able to measure objective 12 parameters in a tolerable way and to collect a large amount of data in "ecological" open-source, modular, scalable and secure platforms for data analysis, integration, and 16 visualization (What to display), 4-Establish a roadmap for regulatory approval and adoption 17 into health care systems (How to disseminate). Subsequently they proposed a roadmap to 18 satisfy those needs, but discussed that several challenges must be fought before the roadmap 19 could be transferred to the real world.

Validity analysis
[61] 20 Aspects of HD that are currently under-investigated, such as non-motor symptoms, have the 21 potential to be studied using wearable technologies as well, adopting a more comprehensive 22 and holistic approach with the aim to measure a broader spectrum of HD features. HD. 20 In summary, there is great promise that wearable and portable devices will contribute to a