Abstract
Background:
Besides cognitive and psychiatric abnormalities, motor symptoms are the most prominent in Huntington’s disease. The manifest disease is preceded by a prodromal phase with subtle changes such as fine motor disturbances or concentration problems.
Objective:
Movement disorders show a high variation in their clinical manifestation depending on condition and external influences. Therefore, devices for continuous measurements, which patients use in their daily life and which can monitor motor abnormalities, in addition to the medical examination, might be useful. The aim of current scientific efforts is to find markers that reflect the prodromal phase in gene carriers. This is important for future interventional studies, as future therapies should be applied at the stage of neuronal dysfunction, i.e., before the clinical manifestation.
Methods:
We performed a software-supported, continuous monitoring of keyboard typing on the participants’ own computer to evaluate this method as a tool to assess the motor phenotype in HD. We included 40 participants and obtained sufficient data from 25 participants, 12 of whom were manifest HD patients, 7 HD gene expansion carriers (HDGEC) and 6 healthy controls.
Results:
In a cross-sectional analysis we found statistically significant higher typing inconsistency in HD patients compared to controls. Typing inconsistency compared between HDGEC and healthy controls showed a trend to higher inconsistency levels in HDGEC. We found correlations between typing cadence and clinical scores: the UHDRS finger tapping item, the composite UHDRS and the CAP score.
Conclusion:
The typing cadence inconsistency is an appropriate marker to evaluate fine motor skills of HD patients and HDGEC and is correlated to established clinical measurements.
INTRODUCTION
Huntington’s disease (HD) is an inherited neurodegenerative disease leading to a clinical triad of movement disorder, cognitive decline and psychiatric symptoms. The disease is inherited autosomal dominant and is caused by an increased number of repeats of the base triplet sequence cytosine-adenine-guanine (CAG) in the Huntingtin (HTT) gene on the short arm of chromosome 4 [1, 2]. A high number of CAG triplet repeats correlates with an early disease manifestation [3]. The precise point of disease diagnosis is poorly characterized but is mainly based on the occurrence of the first certain motor abnormalities. However, subtle motor abnormalities emerge gradually over many years during a “pre-manifest” or prodromal phase [4, 5].
As of now, HD can only be treated symptomatically. Thus, medication is given to improve the individual symptoms of the disease. Disease modifying therapies are underway but are not yet available for all patients. First strategies to “silence” the gene in question are currently tested in clinical studies [6, 7]. First therapies using this mode of action will likely be available soon. Such therapies make it much more important to divide pre-manifest gene carriers (HD gene expansion carriers, HDGEC) into persons in the prodromal stage, i.e., close to the onset of the disease, and persons far away from the onset. Making this distinction is particularly important when deciding whether a HDGEC should receive therapy. Thus, it is important to find measurements which can reliably make this differentiation.
The severity of motor symptoms of HD vary in the short term depending on various external and internal factors, like emotional status, stress, time of day and mood. To date, the severity and course of clinical symptoms, are rated by a physician during a visit in the medical center or study center. However, such ratings can be unreliable, as they represent only a snapshot in time and the patient’s symptoms may be more or less pronounced given the unusual situation of a consultation. For many years, efforts have been made to complement or replace these scores, which depend on the subjective impression of the rater and are therefore susceptible to variation, with more objective and quantitative measurements [8].
A functionally relevant motor sign besides chorea in HD is a progressive impairment of fine motor skills, for example in the finger tapping task of the Unified Huntington’s Disease Rating Scale (UHDRS), in which the index finger and thumb of the respective hand should be brought together as quickly as possible [9]. A number of studies have already addressed the question of subtle motor changes in HD patients and in gene carriers in a prodromal stage. Van Vugt and colleagues measured reaction time and movement time to a button 6.5 cm away in HD patients compared to healthy controls. They found a slower movement initiation and execution in patients and could confirm a correlation with clinical rating scales [10]. Saft and colleagues had measured tapping on a tapping board and compared the tapping speed of different groups: tapping speed was impaired in HD patients and tapping outcomes were related to caudate atrophy [11]. The same group was able to find evidence that a modified tapping task - speeded tapping and metronome tapping - was also impaired in premanifest HD gene carriers. They could also show a correlation of tapping variability with disease burden as well as with the clinical motor phenotype and even with grey and white matter atrophy [12]. Michell and colleagues investigated HD patients and controls with a specially developed tapping device (number of taps in 30 seconds) and could provide evidence that HD patients had a lower number of total taps and a higher variability in between repeated testing intervals [13]. An investigation of rhythmic movements and reaction time in 30 patients compared to 24 controls of Martinez Pueyo et al. showed a significant difference of reaction time in self-paced timing precision [14].
Bilney and colleagues as well as Delval and colleagues could find similar changes when examining gait in HD patients: An increased variability in footstep cadence, due to disturbances in footstep timing, has been shown in HD patients [15, 16].
Writing on a computer keyboard is, compared to a tapping task, a more complex fine motor action. Finger movements when writing on a keyboard of a computer (cadence) represent a highly complex process, mechanically comparable to the two-handed playing on a piano. The finger movements in computer writing require: 1) the almost automated use of the 5 fingers on both hands at the same time; 2) depending on which word is written, the writing operation is performed by the use of the fingers of both hands in a defined order predetermined by the distribution of the letters on the keyboard. This means that in rapid succession, the motor cortex is activated for the movement control of one or more fingers of the right hand and for the use of one or more fingers of the left hand.
To our knowledge, no attempts have ever been made to investigate whether keyboard typing in patients with HD differs from healthy individuals. Additionally, it has not been investigated whether keyboard typing could be altered even before other manifest motor symptoms emerge in HDGEC. The examination of typing cadence (TC) is an attempt to find such a measurement instrument and to investigate its suitability as a digital biomarker.
NeuraMetrix TC measurement
NeuraMetrix developed the software NeuraMetrix TC to analyze and evaluate individual typing behavior; under the assumption that disease-specific factors individually influence typing on a keyboard.
We assumed that motor impairment in HD leads to changes or a higher variability in keyboard typing. In addition, we hypothesised that abnormalities in the typing cadence are present before other motor symptoms occur. Given that the effort associated with a regular examination in an HD center is enormous for many patients, this study is intended to evaluate this software for its suitability as a new tool for measuring and investigating motor function. After installation of the software, the participants experience no significant additional burden since the measurement process takes place in the background, while the patients carry out their regular, everyday writing at home or at work on their own PC. Here, we provide the results in cross-section of this new approach.
MATERIALS AND METHODS
Enroll HD
Enroll-HD is a global clinical research platform designed to facilitate clinical research in HD. Core datasets are collected annually from all research participants as part of this multi-center longitudinal observational study. Data are monitored for quality and accuracy using a risk-based monitoring approach. All sites are required to obtain and maintain local ethical approval.
All participants received a full clinical examination at the beginning of participation. The clinical examination was done in connection with the annual study visit of the Enroll-HD study. The group of participants in Enroll-HD consist of HD patients at stage I and II of the disease, as well as HDGEC and healthy controls. The controls are usually recruited from the family environment of affected persons.
Enroll-HD assessments
The Enroll-HD assessment battery principally consists of the UHDRS, which is assessed annually with each participating person, combined with some additional questionnaires. The UHDRS is a clinical tool which has been developed by the Huntington Study Group [17] to provide a uniform assessment of the clinical features and course of HD. The UHDRS has undergone extensive reliability and validity testing and has been used as an outcome measure in numerous clinical trials. The components include the Total Motor Score (TMS), the Total Functional Capacity Score (TFC), Cognitive Assessments (verbal fluency, symbol digit modalities test, Stroop word, color, and interference tests), the Functional Assessment Score and the Independence Score. The original UHDRS Behavior Score has been replaced by a short Version of the Problems Behavior Assessments (PBA-s) in the Enroll-HD study.
Demographic data, medical history data, pharmacotherapy, and nutritional supplementation are also collected. Based on the collected clinical assessment data, we calculated two scores that were developed in recently published works: the CAP score (CAP score, cytosine-adenine-guanine age product score) and the composite UHDRS [18, 19].
CAP score
The CAP score has been developed as a marker which generally indicates disease burden. The score was created from a large cohort of presymptomatic gene carriers called TRACK-HD considering magnetic resonance imaging-based structural markers. The score is calculated from age and number of CAG repeats using the below formula [18]:
CAP (30) is the version of the CAP score with L = 30 and K = 6.27, then the CAP score will be equal to 100 at the expected age of onset of motor symptoms.
Composite UHDRS
The composite UHDRS (cUHDRS) is built on the basis of the original scale and retains the most reliable tests. It has been developed by using data of 1,668 individuals with early-stage manifest HD by combining data out of 4 studies (Co-Enzyme Q10 and Remacemide: Evaluation in HD [CARE-HD] [20], TRACK-HD [21], Cooperative Huntington’s Disease Observational Research Trial [COHORT] [22], and Coenzyme Q10 in Huntington Disease [2CARE] [19, 23].
It includes the TFC, the TMS, the Symbol Digit Modality Test (SDMT), and the Stroop Word Reading Test (SWR) and shows enhanced sensitivity to clinical change in early symptomatic HD compared to the original score.
NeuraMetrix TC
In order to analyze the typing cadence of participants, a memory stick containing the software and installation instructions was given to each participant after informed consent.
This programme automatically separates the content of the written text from the information about 1) consistency, 2) the speed of movement of individual fingers, 3) the dwell time of each finger on certain keys, 4) the speed with which fingers of the right hand are moved in relation to fingers of the left hand - for right- or left-handed people. It is important to underline that the speed or accuracy with which a participant types was not measured, rather the consistency of the keystrokes.
The NeuraMetrix TC software was installed by the participants at home on their PC. After installation, the program began collecting data, whenever participants typed on their PC. This measurement ran in the background and was not noticed by the patients, i.e., the patients are not restricted in their daily activities.
The collected data was copied by the software before further processing by the operating system. The data was bundled on the PC and transmitted to the NeuraMetrix servers every time 1000 keystrokes had been made. This data transmission contained no sequence of what was typed so that reconstruction of the original written text was not possible. For evaluation purposes, the raw data was converted into a so-called Sincor. The Sincor value is a proprietary measure of the typing inconsistency.
Mechanical aspects, which are determined by the type of keyboard and the effects of wear and tear, can influence the typing cadence. Thus, the NeuraMetrix TC system sets up a profile for each keyboard the person uses. By calculating the Sincor from the consistency of the dwell times, the typing data can be made interindividually comparable.
The graphical representation of the data was shown in a user-specific portal of the company NeuraMetrix. Here, the assignment was anonymized by the identification number. The current status of the data collection per participant with the number of collected data points and the Sincor values could be viewed by the investigator.
Participants
Participants were recruited from the study center in the University Hospital of Ulm. The classification into the three groups HD patients, HDGEC and healthy controls was adopted from the classification of the Enroll-HD study. Recruitment was done during the annual visits for the Enroll-HD study.
The following persons could not be included in the study: a) Participants that take part in any interventional clinical study; b) persons who share a PC with other members of the household without their own user account; c) or persons who use a computer from the company “Apple”.
The simultaneous participation in non-interventional studies was allowed. Patients should continue to take their usual medication unchanged during the observation period. There were no forbidden medications.
This study was conducted according to the Declaration of Helsinki. Likewise, the applicable legal requirements were followed. Both the Enroll-HD study and the keyboard typing were conducted in accordance with good clinical practice and US FDA guidelines. The Enroll-HD study is registered at clinicaltrials.gov as NCT01574053. All protocols had been approved by the local ethics committee.
The detailed inclusion and exclusion criteria are provided in the Supplementary Material.
Statistical analysis
We received the raw data including the Sincor values displaying the typing inconsistency from NeuraMetrix for the statistical calculations. The calculations were performed with the SPSS Statistics 25 computer program from IBM. We used a mean value from the inconsistency values of the first 3 months after installation of the program for all participants in order to weight each participant equally, since the participants naturally generated different amounts of data.
Nonparametric tests were performed for the statistical analysis. We tested the three groups (HD patients, HDGEC, and healthy controls) by comparing them in pairs with a non-parametric test using the Mann-Whitney test, including the Wilcoxon rank sum test, so that three tests were calculated on the basis of the mean values. Additionally, the effect sizes were calculated [24]. In order to investigate whether there is a correlation between Sincor values and the results of the clinical evaluation, we performed correlation calculations. We compared the Sincor values with four different clinical values or characteristics; we calculated a correlation with the cUHDRS and the CAP score, additionally with a single item of the UHDRS, the Finger Tapping item, where we used the sum of the item for the right and left hand. As a further correlation parameter, we used age to investigate a possible correlation with typing cadence. The Spearman coefficient was used for all correlation calculations.
We could only calculate the CAP score for HDGEC and HD patients, as we had no information about the CAG repeat length in our healthy controls. For this reason, a smaller group size of n = 19 is given in this correlation calculation.
RESULTS
Descriptive statistics
We included 40 participants, of whom 20 were patients with manifest HD symptoms in stage I or II, 8 were HDGEC and 12 were healthy controls. In summary, 15 participants dropped out as we had collected few or no data from them. From the remaining 25 participants, we evaluated data of 12 HD patients, 7 HDGEC and 6 healthy controls. The group of participants who did not provide typing data, named as “drop-out” group, did not substantially differ in TFC Score or number of CAG repeats. However, in the HD patient group and the control group were on average older than the comparable group whose data were used for the analysis. Descriptive statistics are shown in Table 1.
Description of study population
Description of study population
Descriptive statistics of the study population, the group of participants that were analysed is shown seperately from the group of included participants from whom we have not received data about their keyboard writing. The two groups are divided into three subgroups. Data is always shown as mean with standard deviation in brackets. HD, Huntington’s disease, HDGEC, Huntington’s disease gene expansion carriers; TFC, total functional capacity; CAG, cytosin-adenin-guanin; n.a., not applicable.
All HDGEC had a positive genetic report with an expansion of the CAG repeat sequence to 40 or more (Minimum: 40, Maximum: 48). The total motor score for this group was 0 for five participants, 1 for one participant and 7 for one participant, which indicates that no or only very mild, unspecific motor abnormalities were found in any of the participants in this group in the clinical examination.
The group of HD patients was assessed to have had symptoms of the disease for an average of 7.33 years (minimum 1 year, maximum 18 years) in the regular clinical examination; assessed on the basis of the item: “rater’s estimate of symptom onset”. All participants in this group were in stage I or II, which means the TFC score was in between 7 and 13. The number of CAG repeats on the extended allele in the patient group ranged from 41 to 56 with a mean of 46.67 repeats.
Since the activity of how much the participants wrote on their computer was very different, we calculated an average value of the typing inconsistency for each participant. Recruitment started in February 2018. For the analyses we used the collected data from a data cut dated March 2019. We used an averaged value per participant. A graphical example for the distribution of the average Sincor for one participant is shown in Fig. 1. Table 2 shows the mean and median Sincor values divided in groups.
Example of data collection for one individual using the NeuraMetrix TC software.
Typing inconsistency (Sincor) mean and median values divided in groups
Mean and median of the typing inconsistency value (Sincor) recorded by the NeuraMetrix TC software divided in three subgroups.
The subgroups were compared in pairs, the results are presented in Table 3. The analysis showed a significant result when comparing the HD patient group with the control group at a significance level of α= 5%with a large effect size. In the comparison of the inconsistency between patients and HDGEC a p-value of 0.063 was calculated, which is just below the significance level. The effect size was medium with an r of 0.43. Between the typing inconsistency of healthy controls and HDGEC, no significant differences were found. The effect size was low with r = 0.12. The single data points for each individual divided in groups are shown in Fig. 2.
Comparison of typing inconsistency values (Sincor) between groups
Comparison of typing inconsistency values (Sincor) between groups
Results of the nonparametric statistical test (Mann-Whithney-U and Wilcoxon test), in each column two of the three groups are compared in terms of typing inconsistency (Sincor). Classification and interpretation of effect size measure r according to Cohen. R = 0.1 small effect, r = 0.3 medium effect, r = 0.5 large effect [24]. Significant results are shown in bold type, the significance level was determined at 5%. HD, Huntington’s disease; HDGEC, Huntington’s disease gene expansion carrier.
Mean inconsistency of the three subgroups. HDGEC, Huntington’s disease gene expansion carrier; HD, Huntington’s disease.
We compared the Sincor values with the Finger Tapping Task after calculating the sum of the values for the right and left side. The Spearman correlation analysis showed a significant result with a coefficient of 0.450. Figure 3 provides an overview of the distribution of the values for each individual.
The correlation calculation between the CAP score and Sincor values resulted in a statistically significant correlation with a Spearman coefficient of 0.505.
Distribution of the values of the Finger Tapping Item from the Unified Huntington’s Disease Rating Scale (sum of items for left and right side) with associated inconsistency value per individual and linear trend line.
A similar, also significant result was found in the calculation of the correlation between Sincor and cUHDRS, with a Spearman coefficient of 0.432. An overview of the distribution of the values is shown in Fig. 4 for the CAP score and in Fig. 5 for the cUHDRS.
Distribution of the values of the CAP score with associated inconsistency value per individual and linear trend line. CAP score, cytosine-adenine-guanine age product score.
Distribution of the values of the cUHDRS with associated inconsistency value per individual and linear trend line. cUHDRS, composite Unified Huntington’s Disease Rating Scale.
To verify whether the typing inconsistency is also related to age, we also calculated the correlation between Sincor and age. No correlation between the parameters could be detected here. All correlation calculations are shown in Table 4.
Correlations of typing inconsistency (Sincor) with clinical scores
Correlation calculations with typing inconsistency (Sincor) and clinical scores as well as typing inconsistency and age. Significant results are shown in bold type, the significance level was set at 5%. CAP score, cytosine-adenine-guanine age product score; cUHDRS, composite Unified Huntington’s Disease Rating Scale. The Finger Tapping Item for the right and left side from the Unified Huntington’s Disease Rating Scale were added to one sumscore before the correlation was calculated.
DISCUSSION
The objective of this study was to investigate a new method for detecting the smallest possible early changes in motor function in Huntington’s patients and HD gene carriers at the prodromal stage. The key objective was to determine the typing cadence of each participant individually and to investigate whether early motor changes can be detected and recorded with this method.
This study - albeit with a small number of cases - achieved statistical significance regarding the ability to distinguish between healthy controls and HD patients by measuring the keyboard typing inconsistency. The comparison of the HDGEC group with healthy controls did not show statistical significance with a low effect size, suggesting that a higher number of subjects would be needed to find a difference. Due to the small number of cases, the group was not further subdivided, so it was not defined more precisely whether the participant in the HDGEC were likely to be close to the beginning of the disease or not. This fact lets interpret that a difference could be found either in a larger group of gene carriers or in a group of gene carriers that are all close to the onset. However, the typing inconsistency correlated with the CAP score, which is a measure for the burden of disease. We interpret this as a sign that the measurement of keyboard typing is a marker that can show disease severity. This leads to the assumption, that changes in typing cadence are even present in prodromal gene carriers, which could already be shown for finger tapping [12].
As mentioned above, several studies have already addressed finger tapping as a potential marker showing changes like impairment of tapping speed and tapping consistency in HD [10–14]. These studies have provided evidence that tapping speed decreases and variability increases.
With the similar aim as in our study, to find a marker that reveals alterations already in the prodromal stages of HD, Blekher et. al. investigated eye movements in terms of various occulomotor tests. They found, although with a considerably larger number of participants (n = 215), a high sensitivity of saccade measures in discriminating prediagnostic and early HD gene carriers from nongene carriers. It can therefore be assumed that the occulomotor system, similar to the measurement of fine motor skills, shows early changes in the course of HD [25].
There have been attempts to supplement the physician’s examinations with digital measurements in everyday life. In the PRIDE-HD study, a subgroup of the participants (n = 10) were monitored at home using wearable sensor devices such as mobile phones and smartwatches, thereby measuring the chorea objectively [26]. An advantage of both wearables and the monitoring of keyboard typing in comparison to others is that consistency of typing can be monitored over a long time period and deliver information continuously. This is particularly important since the motor phenotype of HD is subject to fluctuations depending on many factors such as stress.
To our knowledge, keyboard typing as a diagnostic marker in HD has not yet been investigated. Some studies on various diseases and their impact on typing cadence measured by NeuraMetrix TC are underway. This is the first study on Huntington’s disease using this software.
Several circumstances kept persons from being able to participate in the measurement. We could not include potential participants that share their computer with other family members and do not have an own account, because the software measures every typing activity. Yet, the possibility that another person also typed, could not be excluded with absolute certainty. However, the typing inconsistency is intra-individually different, so that a change of the typing persons would be conspicuous by a high variance of the inconsistency values. We have not investigated if this technique is suitable for patients in later stages of the disease, when motor impairment is severe. We assume that, with the increasing functional and cognitive decline in the course of HD, patients in later stages rarely use a computer and therefore may not be reliably measured with this method. Yet our focus was on the prodromal phase and the early stages of the disease.
We conclude that keyboard writing is a complex motor skill that deteriorates early in the course of HD and which can be measured with the method tested here. Monitoring this motor skill over a longer period of time from a distance is easily possible. There are indications that there is also a change in the prodromal stage of the disease, which we could not find evidence for, most probable due to the small number of subjects. This aspect as well as the rates of change over time must be verified by larger and longitudinal studies.
Footnotes
ACKNOWLEDGMENTS
Enroll-HD is a clinical research platform and longitudinal observational study for Huntington’s disease families intended to accelerate progress towards therapeutics; it is sponsored by CHDI Foundation, a nonprofit biomedical research organization exclusively dedicated to collaboratively developing therapeutics for HD. Enroll-HD would not be possible without the vital contribution of the research participants and their families. We would especially like to thank all participants.
CONFLICT OF INTEREST
B. G. Landwehrmeyer owns stocks of the company NeuraMetrix and works as consultant for NeuraMetrix. All other authors have no conflicts of interest to report.
