Characterizing patient compliance over six months in remote digital trials of Parkinson’s and Huntington disease

The widespread global use of smartphones and connected sensor devices, in parallel with their continuously increasing capabilities, has begun to impact the clinical trial ecosystem. Digital wearable and non-wearable devices containing electronic sensors such as accelerometers, gyroscopes, and photosensors can now track diverse biomarkers including heart rate, blood pressure, heart rate variability, lung function, and gait to various degrees of accuracy [1]. Complementarily, smartphone applications enable frequent and facilitated interaction with patients in the form of reminders, self-reporting of drug intake, or acquisition of electronic patient-reported outcomes [2].

Remote data collection via connected devices may provide added value to clinical trials [3]. Statistically, remote technologies increase data collection frequency, consequently providing insight on variability, requiring less extrapolation, and potentially increasing study power. These insights typically rely on signal processing and machine learning methods developed using sensor collected (big) data. Additionally, sensors, while not exclusively used in remote settings, are inherently objective compared to clinician scoring of disease status [4]. Home-based monitoring may also be more objective from the patient perspective owing to white cloak phenomena, muddying response in the clinic-setting [5]. From the trial management perspective, real-time monitoring may enable early identification of safety, operational, and compliance issues. Not surprisingly, applications for wearable technologies have already been demonstrated in a wide spectrum of disorders, including cardiovascular, respiratory, metabolic, psychiatric, and neurological disease [6]. Eventually, these innovations have the potential to evolve into regulatory-approved, clinical trial endpoints [7, 8].

Parkinson’s disease (PD) and Huntington disease (HD) are both chronic, neurologically-based movement disorders. Motorically, PD is characterized by a diverse set of symptoms that present at successive stages of the disease, including slowing of gait, shuffling feet, reduced arm swing, freeze of gait, asymmetry, tremor, bradykinesia, and dyskinesia. In HD, the most notable movement impairment is chorea, which is often similar to PD dyskinesia but is generally stable in contrast to PD symptoms that may fluctuate throughout the day. Given their chronic nature, lack of adequate therapies, and unique motor aspects, PD and to a lesser extent HD have been the focus of many wearable studies in disease [7, 9].

Remote trials with smartphones and wearable devices have been conducted in movement disorders such as Parkinson’s disease (PD) and Huntington disease (HD). Beyond a large collection of mostly short-duration studies aimed at quantifying symptoms [10], several efforts have examined longer-term, home-based monitoring in PD [11‐16]. Four notable examples of large, multi-month, PD efforts that assessed aspects of remote compliance are the Parkinson@Home study which quantified the daily duration in which a patient-worn smartwatch was streaming data over a period of up to 13 weeks [13, 17], the mPower and HopkinsPD studies which published results on smartphone app usage and performance of home-based tasks over six months [14, 15], and the SMART-PD study which evaluated the impact of a smartphone app on medication adherence over four months [16]. In contrast, in HD existing remote studies have primarily focused on feasibility or symptom quantification over several days [9, 18, 19], as HD is an orphan disease which is naturally more challenging to recruit.

Despite its inherent advantages, remote patient-based data collection is prone to unique adoption and compliance issues [6, 20]. Technical factors include the need for frequent device charging, ease of device operation, and the simplicity of the user interface. Burden factors unique to remote study protocols include, for example, the imposition of recurrent performance of home-based tasks or requirement for consistent reporting on symptoms and drug intake times. These issues are notable, as often a high number of patients dropout of studies prior to completion, even in ‘traditional’ trials [3]. Moreover, the effect of multi-month duration is important as patient compliance in clinical trials of chronic conditions is lower than in acute conditions [21]. Collectively, this highlights the importance of understanding compliance dynamics and patient preferences with respect to the requirements of remote digital trials. Motivated by above, we analyzed two six-month studies of neurological movement disorders to better understand patient compliance patterns in remote settings for four digital study protocol metrics.

Two independent clinical trials were analyzed: the Clinician Input Study in PD patients (CIS-PD) and the observational digital health sub-study within Study TV7820-CNS-20016 in HD patients (Open PRIDE-HD) (Table 1 and Table 2, respectively). The CIS-PD data used in the preparation of this article were obtained from the Michael J. Fox Foundation database. The total duration of each study was six months, and both studies used the Intel® Pharma Analytics Platform for data collection, monitoring, and analysis [17, 22]. The mobile application interface was designed using User Interface/User Experience (UI/UX) expert input to emphasize ease of use (Fig. 1).

Remote monitoring is expected to gradually transform drug development procedures, with assessment of patient compliance using digital monitoring being increasingly useful for smart trial design. The current report includes a relatively long-duration follow-up of movement disorder patients from two unrelated trials with respect to four compliance measures that are specific to remote trials.

Several of our findings may assist in the design of similar clinical trials in the future. First, we observed that the variability of the compliance rates at the beginning of the trial are non-indicative of compliance patterns for the vast-majority of the remainder of the trial duration. This is not surprising given the need to adjust to new technology and may need to be considered during trial planning. Second, when given the option, study participants preferred to report their symptoms during the morning hours. Third, defining appropriate eligibility criteria and screening methods appears to be especially important in remote, technology-based trials given that compliant patients tend to exhibit high compliance across the different aspects of the remote trial protocols, rather than to only to a single protocol metric. This is likely influenced by factors such as age, technology savviness, and various additional factors which require further research.

As expected, there were differences in compliance rates over time between the two studies. While smartphone app-based interactions have been shown to increase compliance [16], the observation of gradually decreasing engagement in the PD study was not surprising as decreased compliance over time has been shown in numerous trials, for instance in a trial in which 50% of patients stopped taking anti-hypertensive drug within one year [24]. One potential reason for the observed difference between the trials may be that the HD trial was interventional, perhaps increasing the motivation of patients seeking experimental drug benefits. In addition, the HD trial implemented a higher level of interventional support calls which may have maintained higher compliance levels.

In this work we attempted to understand the adherence of HD and PD patients to remote clinical trial protocols over a multi-month duration. It is important to note that we defined compliance as patient participation levels rather than compliance as strictly defined by the study protocols. While both measures are tightly linked, participation levels are continuous and are thus better suited for understanding patient behavior. In contrast, protocol-defined compliance is often binary due to the definition of thresholds. For example, the PD study protocol defined smartwatch data streaming compliance as streaming data for at least 10 h per day for 25 days per month.

While HD is an orphan disease with little or perhaps no precedent of studies examining remote compliance, our work is not the first to investigate remote compliance in PD. In the large observational Parkinson@Home study, mean daily smartwatch data streaming durations of 14.8 h – 16.3 h were observed for two large cohort studies of 6 weeks (cohort 1, n = 304) and 13 weeks (cohort 2, n = 649) [13]. As in our PD trial, the researchers observed only a mild reduction in smartwatch data streaming of roughly 25% from start to finish, which they proposed may be attributed to the passiveness of, or the lack of interaction needed for, data collection. Another study, titled mPower, focused on data collection from thousands of users using home-assessments that leverage smartphone sensors and a corresponding mobile app [14]. In a paper describing the first six months of data collection, the authors observed a rapid, exponential-like drop in average data contribution per patient, which could be due to many factors including the remote recruitment, no interventional treatment, and no support or trial management intervention. Another effort, HopkinsPD collected passive data from the smartphones of PD patients and control subjects and asked for performance of two structured motor assessments at home during the morning hours [15]. While structured assessment compliance relative to protocol was not reported, the authors showed that the abundance of structured assessments at home as well as passive usage of the smartphone remained uniform across all days of the week, similar to our findings.

Patient compliance is a broad topic with many aspects and research challenges. One major challenge is obtaining large patient samples sizes given costs and recruitment challenges, especially in rare diseases. Our HD study had 17 patients, whereas a larger number may potentially impact conclusions. Despite the small sample size, this study provides an example for monitoring patient compliance in long-duration remote studies and can be used in the design, planning and execution of future studies in this patient population. Additionally, factors not addressed in this work such as mobile application user experience and interface design may influence compliance rates. Finally, a deeper analysis of digital trial economics, including device cost, data management and clinic visit savings, can help illustrate the financial considerations in such trials.

Understanding patterns of patient compliance in remote, technology-based clinical trials requires analysis of data collected using digital technologies. The insights from our work suggest that such remote trials are feasible, even when comprising multiple protocol requirements. Beyond our observations, examination of additional factors that impact compliance such as the impact of support, intervention, alternative application design interfaces, and additional covariates can further influence study design.

In the broader perspective, this report supports the growing trend of using mobile applications and wearable technologies to monitor, prompt, and encourage patient compliance with medication intake and performance of clinical assessments. This effort strengthens the notion that these data have the potential to provide further insight regarding patients’ daily lives outside the clinic and potentially evolve into novel endpoints for regulatory purposes.