Study design, Participants, and Setting
For this prospective observational, non-interventional trial, 25 participants were recruited between March 11 and of May 11, 2022, from the neuromuscular outpatient clinic at the Department of Neurology, RWTH Aachen University Hospital, a designated center for chronic immune-mediated neuropathies. Data collection was completed on April 12, 2023. Prior to inclusion, patients gave written informed consent to participate in the study. The study was approved by the Ethics Committee of the University Hospital Aachen and was conducted in accordance with the Declaration of Helsinki.
Patients were eligible, if they had been diagnosed with CIDP or MMN according to the latest guidelines of the European Federation of Neurological Societies/Peripheral Nerve Society (EFNS/PNS) [4, 5]. They needed to be at least 18 years old and demonstrate clinical stability before enrollment, as determined by an experienced board-certified neurologist. Exclusion criteria included any confounding conditions that affect motor function of the hand (e.g., rheumatoid arthritis), other serious illnesses that would prevent participation in the study (e.g., dementia), and pregnancy.
Some patients with chronic immune-mediated neuropathies show motor impairment of the distal upper extremities, while others do not [2, 5]. At the initial timepoint, an experienced neurologist identified individuals with clinically relevant hand motor impairment to differentiate subgroups for subsequent analyses. This allocation was determined through a detailed neurological examination, with particular emphasis on motor functionality, and aligned with clinical outcome parameters. The assessment of motor impairment applied to the dominant hand, or in cases of focal CIDP to the only affected hand, highlighting their importance for everyday life.
In this proof-of-concept study, we collected data during ongoing standard-of-care IVIg maintenance therapy. Patients were followed-up at five consecutive time points (T0 – T4). The interval between the different time points was determined by the individual treatment regimens.
To introduce the technique and to avoid confounding factors such as learning, the time point T0 was declared as a learning time point to familiarize with the assessments and especially with the glove. Consequently, we only included the data from T1-T4 in all subsequent analyses.
Glove design and sensors
Patients underwent repeated assessments with the data glove, manufactured by Cynteract® GmbH, Aachen, Germany at all consecutive time points. In combination with its corresponding software (Figure (Fig.) 1), the glove is certified as a Class I Conformité Européenne (CE) medical device and measures the range of motion (ROM) of the hand.
The data glove consists of spandex material and weighs 60 grams. It provides a USB cable connection to connect the glove to a laptop. For this study, there was one pair of unisize gloves of for the left and the right hand.
To measure the angles of the finger joints, the data glove contains a total of eleven Bosch BNO055 nine-axis sensors that are sewn into the glove layers according to the hand anatomy (Fig. 1) and the sensors are connected by flexible silicone cables. A built-in accelerometer measures acceleration and a built-in gyroscope measures angular velocity. The sensor also embeds a magnetometer to measure the orientation of the sensor relative to the Earth´s magnetic field. It was switched off for this study due to interference from the clinical environment. Furthermore, each sensor contains an integrated processor that automatically fuses the data of the sensor components.
In the first step, the global rotation of each sensor is given as three-dimensional quaternions and is recorded approximately 25 times per second. From the quaternion information of two adjacent sensors, the angle of the finger joint situated in between is calculated in degrees (°). For this purpose, the rotation of the more distal sensor is relatively calculated to the rotation of the more proximal sensor.
Afterwards, the three-dimensional relative rotation is reduced to a specific rotation axis to determine a two-dimensional angle (in °). This axis varies depending on the particular type of angle (e.g., bending-/spread angle, or angle during opposition movement) of interest.
Movement patterns of the data glove
We evaluated the ROM of the hand using the data glove in three different glove movement patterns, referred to as finger spread, thumb opposition and fist opening (Fig. 1), out of which finger spread is mainly mediated by the ulnar nerve, thumb opposition by the median nerve, and the fist opening by the radial nerve.
Figure 1: Overview on the three data glove movement patterns. The sequence of the three different movement patterns: finger spread (A), thumb opposition (B), and fist opening (C) is depicted. As merely one movement direction is mediated by one of the three hand nerves, only this specific direction was relevant for the data analyses of the range of motion (ROM). The arrow between the right and left images indicates the movement direction, which is essentially mediated by one of the three hand nerves.
The specific movements (Fig. 1), performed while wearing the data glove, control an interactive computer game. The computer game includes a standardized course of 50 alternating obstacles, which must be traversed by passing above or below the obstacles
(Fig. 2). For this, the user had to adjust the size of a sphere by performing specifically defined hand movements in an alternating sequence.
Before the computer game starts, there is a Reset to determine the zero-degree position, that served as a reference for the subsequent angle determination (Additional file 1). To adapt the interactive game control to the individual ROM, the patients also need to calibrate the movement patterns before starting the course. Depending on the extent of hand motor impairment, a whole procedure takes approximately eight to 15 minutes per hand.
Figure 2: Overview on the technical and functional background of the glove. The data glove was connected to a computer/laptop running the corresponding software (A). Patients had to complete a course and its alternating obstacles by adjusting the size of a sphere, that was controlled by finger movements. To the left of the screen, a green bar indicated the current motion amplitude based on calibration. At the top, a progress bar specified the position within the course. The button in the upper right corner allowed for cancellation of the game and for returning to the main menu (B). Sensors for data acquisition were located over the dorsal sides of the proximal, resp., the middle phalanges of the long fingers. For the thumb, they were fixed at the first metacarpal bone and the proximal phalanx. The larger main board sensor was placed on the dorsum of the hand in the region of the metacarpus and the carpus and served as the central reference point for the proximal phalanges of the long fingers, as well as for the first metacarpal bone. Corresponding to the anatomical bone structures, on the right, the glove is shown as being worn (C).
Moreover, we included a classical computer game called the rocket game, that was based on the movement pattern fist opening (Additional file 2) at the end of each study time point.
Angle measurement and ROM analysis
The specific angles, measured by the data glove during the three distinct movement patterns about 25 times per second, and their corresponding axes of rotation are depicted in Fig. 3A-C. To determine the maximum ROM of the hand, we calculated the difference between the angle in the maximum and the minimum position (in °) for each of the alternating movements during glove assessment. In the following, this difference is referred to as the Δ - angle (Fig. 3D). As for each run through the course 50 obstacles had to be passed, half of them required a movement, which is essentially mediated by one of the three main hand nerves. In theory, this resulted in 25 Δ - angles for each glove assessment. The resulting Δ - angles were revised and their mean was the parameter of interest for all analyses concerning the glove (Additional file 1).
Figure 3: Angle measurement and analysis of ROM by the data glove. The current spread angle between the little and the index finger, measured during the finger spread movement pattern (light blue area), results from the current alignment of the index and the little finger (blue arrows) (A). The current bending angle of the metacarpophalangeal (MCP) joint and the proximal interphalangeal (PIP) joint of the long fingers (light blue areas), results from the alignment of the different sensors during the fist opening movement pattern (B). During the opposition movement of the thumb, the angle to of the carpometacarpal (CMC) joint was determined. The orange arrow signifies the longitudinal alignment of the sensor at the first metacarpal bone in the so-called zero-degree position). The blue arrow indicates the longitudinal alignment of the first metacarpal bone in the current opposition position. In between, the current opposition angle of the CMC joint is illustrated (light blue area). (C) The maximum range of motion (ROM) is calculated based on the time-angle signal, exemplarily shown for a single movement during the fist opening movement pattern. The maximum angle was measured when forming a fully clenched fist (a). The minimum angle was measured when the long fingers were fully extended (b). The difference between these two points (a-b) is referred to as the delta (Δ) - angle (in °) and represents the maximum ROM (D). The rotation axis is always shown in red.
The corresponding data revision and processing software is described in Additional file 1.
Patients’ satisfaction survey
One part of the feasibility assessment was a standardized questionnaire that evaluated whether and why the patients would (not) recommend the examinations with the data glove to others based on their individual experience in this study. Furthermore, we asked the patients whether they would prefer the data glove assessment to any of the routine outcome measures (Vigorimeter, NCS, HRUS, everyday-life questionnaires), and if so, why.
Clinical outcome parameters and PROMs
We performed clinical examination at each time point and followed the current recommendations by using multidimensional array of established outcome measures [1, 5, 13]. To address the domain of impairment, the examination included the assessment of grip strength by the Martin Vigorimeter [16, 40]. The mean of three consecutive measurements was calculated [16]. Furthermore, we applied MRC Sum Score to examine muscle strength [18]. The disability domain was represented by both the INCAT disability score [17, 41] and the disease-specific R-ODS, quantified in log-odds (logits) [20, 25]. As the focus of this study was on the hand motor function of the patients, the respective arm sub-score was extracted from the MRC Sum Score and the INCAT disability score and used for the analyses. The symptoms domain was referred to by assessing the Beck Depression Inventory (BDI) [21] and the Fatigue Severity Scale (FSS) [22], which were not mainly focused on in this study.
Nerve conduction studies
We performed NCS at T2 and T4 to assess nerve conduction velocity (NCV) of the three hand nerves (median – motor NCV, ulnar – motor NCV, and (superficial) radial nerve – sensory NCV) at the forearm. We therefore used standard neurophysiology devices (Natus Neurology, Nicolet EDX), and performed all studies with a surface stimulator and surface recording electrodes.
High-resolution ultrasound
To assess the morphology of the peripheral nerves, we measured the cross-sectional area (CSA) of specific nerve sections at T2 and T4. We used a Mindray TE7 ultrasound scanner with a 3–13 MHz linear transducer. Maximum CSA of the median and ulnar nerve were measured at the upper arm and for the radial nerve at the radial sulcus.
An overview of the assessments during the study course can be found in Table 1.
Table 1
Assessments during the study course
time points | measurements |
T0 / T1 / T3 | MRC (arm sub-score), INCAT disability (arm sub-score), Vigorimeter, R-ODS (logits), BDI, FSS |
T2 / T4 | MRC (arm sub-score), INCAT disability (arm sub-score), Vigorimeter, R-ODS (logits), BDI, FSS, NCS, HRUS |
(MRC = Medical Research Council, INCAT = Inflammatory Neuropathy Cause and Treatment, R-ODS = Rasch-built Overall Disability Scale, BDI = Beck Depression Inventory, FSS = Fatigue Severity Scale, NCS = nerve conduction studies, HRUS = high-resolution ultrasound) |
Statistical analyses
We conducted statistical analyses using RStudio (Version 4.2.2) and Graph Pad Prism (Version 9.5.1) on MacOS. The Shapiro-Wilk test, supported by Q-Q plots, assessed data normality. Data were reported as mean (SD) for normally distributed metrics and median (IQR) for ordinal or non-normal metrics. Data analysis included information up to the point of participant withdrawal for those lost to follow-up. The level of statistical significance was defined as α = 0.05.
Analyses focused solely on data from the dominant, affected side. To assess test-retest reliability, we calculated Intraclass correlation coefficients (ICCs) for metrically scaled parameters, using a two-way mixed-effects model for absolute agreement (guidelines by Koo and Li [42]). For ordinal data, we applied the linearly weighted Cohen’s Kappa, following Landis and Koch [43]. To compare the ROM of patients with and without relevant hand motor impairment, we applied unpaired, two-sided Student´s t-tests including Welch correction at T2. We performed receiver operating characteristics (ROC) analyses for the three glove movement patterns and ascertained the area under the curve (AUC), as well as a specific cut off value and the corresponding sensitivity and specificity. AUC was then rated according to the literature [44]. Additionally, we performed a composite ROC analysis comparing the performance of the glove movement patterns with that of NCV and CSA. We therefore calculated z-scores and applied the mean of each parameter.
For analyses targeting the data glove's ability to assess hand motor function, we only included patients with clinically relevant hand motor impairment. To validate the data glove, correlations between glove data and clinical parameters at T2 were measured using Pearson’s correlation for normally distributed data and Kendall-Tau for ordinal or non-normal data, without Bonferroni correction for multiple testing. Correlation effect sizes were interpreted following Cohen [45]. Additional correlation analyses between glove movement patterns and NCV/CSA of hand nerves were detailed in Additional file 5. A Linear Mixed Model (LMM) with time as a fixed effect and person as a random effect handled longitudinal data, with more details in Additional file 3.