Recruitment for Remote Decentralized Studies in Parkinson’s Disease

INTRODUCTION

Traditional, in-person clinical research studies are plagued by slow and often unsuccessful recruitment [1]. The location of research sites, travel burden, and reliance on busy clinicians to identify potential participants [2] are all factors that contribute to recruitment difficulties. These obstacles are exacerbated in Parkinson’s disease (PD) research, where physical disability, cognitive impairment, and decreased driving ability interfere with research participation [3, 4].

One potential solution is the decentralized research study, in which study visits occur outside of the typical clinical trial setting and data is collected directly from participants [5]. With this model, assessments may be conducted remotely and digital measures can complement traditional measures. Interest in decentralized research models has been increasing, driven not only by interest in reducing participant burden and enhancing recruitment, but also by the ongoing COVID-19 pandemic, which has delayed and interrupted clinical research [6]. One COVID-19 treatment trial has already been conducted using a decentralized framework [7]. Prior remote PD research studies have demonstrated high satisfaction with and interest in this model [8]. However, remote decentralized research studies cannot rely on the same recruitment tactics as in-person studies, raising the question of how best to identify and enroll participants.

Here, we examine recruitment processes and outcomes for three remote decentralized PD research studies with the aim of jump-starting the development of best practices for future remote decentralized clinical trials.

METHODS

We examined recruitment for three independent remote decentralized, observational PD clinical research studies: Fox Insight Validation Effort (FIVE), Assessing Tele-Health Outcomes in Multiyear Extensions of Parkinson’s Disease trials (AT-HOME PD), and Virtual Assessment of LRRK2 carriers to Optimize Research in Parkinson’s disease (VALOR-PD) (Table 1). All studies were approved by the University of Rochester Institutional Review Board. All participants provided informed consent. Detailed study methods have been previously reported for each study [9–11].

RESULTS

Reasons for non-participation

Reasons for non-participation, which were not systematically assessed, varied among the studies (Fig. 1). Across the three studies, 38 (range 6–23) were deemed ineligible. Among the eligible participants for whom reasons for non-participation were known, the most common reasons were being too busy (6/12, 50.0%) in FIVE, being too busy (8/33, 24.2%) in VALOR-PD, and lacking interest in continued research participation (57/114, 50.0%) in AT-HOME PD.

Fig. 1

Reasons for non-participation, by study.

Speed of recruitment and impact of adaptations

Study enrollment windows were a mean (range) 51 (33–61) weeks and the mean (range) number enrolled per week was 4.9 (3.7–6.2) (Table 1, Fig. 2). In FIVE, after changing the invitation email language and adding a reminder email, the weekly average consent to contact rate increased from 8%over the first six weeks to 19%over the subsequent 25 weeks. In AT-HOME PD, after two months of phone recruitment, only 29.9%of potential participants had been pre-screened, prompting the addition of emails and mailers. After adding emails and mailers, the mean number of pre-screens per week increased from 4.7 in the first 12 weeks to 14 in the subsequent three weeks. In VALOR-PD, the average email click rate of opened emails was 24%over the first five months and after adding language targeting individuals without PD, it did not improve.

Fig. 2

Enrollment timeline, by study.

Geographic distribution of participants

The geographic distribution of participants for each study was widespread with 45 states (range 34–42 states) and one Canadian province represented across the studies (Fig. 3). No participants were recruited from five states—Alaska, Idaho, North Dakota, Kansas, and Delaware—in any of the three studies. The number of participants living in Health Professional Shortage Areas, which in this case refers to U.S. zip codes with shortages of primary care providers, ranged from 30.3–42.9%(Table 1).

Fig. 3

Geographic distribution of participants.

Satisfaction and future research participation

Participants in all three studies were highly satisfied with video visits (> 97%overall satisfaction in all studies) (Fig. 4). Across the studies, the most frequent concerns about video visits related to confidentiality, which occurred in a minority of participants —FIVE (6.6%), VALOR-PD (3.5%), and AT-HOME PD (3.0%). Questions asked regarding interest in future research opportunities were similar but not the same across the studies. Most participants in FIVE (98.5%) and VALOR-PD (99.6%) reported being willing to participate in future observational research studies that include video visits. Most participants in FIVE (76.1%) and VALOR-PD (87.6%) reported they would be willing to participate in future clinical trials that include video visits. Compared with studies with all in-person visits, the majority of those willing to participate in future research were more interested in participating in research studies with video visits, both observational (FIVE (62%) and VALOR-PD (65%)) and interventional (FIVE (52%) and VALOR-PD (60%)). These questions were not asked of AT-HOME PD participants. Among AT-HOME PD participants, compared to a trial without any remote assessments, 65%were more willing to participate in a future clinical trial with some remote assessments and 40%with all remote assessments.

Fig. 4

Percentage of participants “satisfied” or “highly satisfied” with various aspects of virtual visits.

DISCUSSION

These three studies—FIVE, VALOR-PD, and AT-HOME PD—demonstrate that nationwide recruitment of participants with PD, without PD, and with a genetic risk factor for PD into remote decentralized research studies is feasible and well-liked by participants. These results inform and support the future of remote decentralized research studies in PD and likely beyond, however, more work is needed to identify best practices for recruitment.

Remote studies enable research to be conducted using a single IRB and from a single location, which in turn reduces costs, complexity, and enrollment time, while also affording greater flexibility. In the three studies presented, a mean (range) of 4.9 (3.7–6.2) participants were enrolled per week. In-person PD research studies may exceed these enrollment rates, however, they require a large number of sites and substantial resources. For example, OBSERVE-PD (an in-person, cross-sectional study of individuals with PD) recruited an average of 0.4 participants per week at each of its 128 sites during 48.5 weeks of recruitment [15, 16] and PPMI (an in-person, longitudinal, observational study of individuals with and without PD including blood, CSF and tissue sampling as well as neuroimaging) enrolled an average of 1 PD participant and 0.5 non-PD participants per month at each of its 24 sites [17]. While it is not realistic to compare recruitment between observational and interventional studies, as a reference, STEADY-PD III, with its comprehensive recruitment “tool kit” for sites, averaged 7.0 participants enrolled per week for twelve months across its 57 sites combined [18]. We also recognize that there are trade-offs with remote video visits that should be considered; some testing cannot be performed, and biospecimen collection is limited to what participants can collect themselves.

All three studies had expansive geographical spread despite being conducted from Rochester, NY. The VALOR-PD cohort represented the fewest states (34) and had the lowest percentage of participants residing in primary care Health Professional Shortage Areas (HPSA) (30%). This is likely because only individuals interested in genetic testing and, with few exceptions, possessing disposable/higher income entered into 23andMe’s pool of potential participants. In contrast, FIVE had participants from the most states (40) and the highest percentage of participants living in primary care HPSAs (42.9%). The potential geographic spread for AT-HOME PD participants was dictated by the distribution of parent study sites (which were located in 36 states, Puerto Rico, and one Canadian province). In order to achieve comparable geographic spread, in-person studies require multiple geographically dispersed sites. For example, PPMI has 34 sites (mean 20.1 participants per site) [17], STEADY-PDIII 56 sites (mean 6.0 participants per site) [13], SURE-PD3 54 sites (mean 5.5 participants per site) [14], and the LRRK2 Cohort Consortium 20 sites (mean 75.3 participants per site) [19]. Furthermore, the nature of in-person studies restricts participation to individuals living in proximity to a research center. While anecdotal, we note that we conducted visits with participants who were located in offices, parks, parking lots, and even on boats. By shifting research from clinics to participant-selected locations, decentralized studies allow for enrollment of individuals without prior access to research centers.

Across the three studies, five states go unrepresented: Alaska, Idaho, North Dakota, Kansas, and Delaware. All of these states but Delaware are among the ten least populated states in the U.S. Difficulties recruiting from these states may reflect the persistence of a geographically-defined digital divide. The digital divide reflects differential access to technologies based on socio-economic, geographic, or other factors [20]. Indeed, only 63%of rural households and 56%of individuals making $30,000 or less per year have access to broadband [21]. Across the three studies only 4.7%–14.3%of participants were located in rural zip codes (defined as having > 50%of its population residing in either a Non-Metro County and/or a rural Census Tract [22]) and 30.3–42.9%of participants were located in primary care HPSA. We used the latter as a proxy to identify underserved areas but recognize that a broad range of social determinants of health influence access to healthcare beyond the availability of providers [23]. The COVID-19 pandemic has forced the transition of many aspects of daily life, education [24], clinical care [25], and research online, highlighting the critical need for expansion of broadband internet access to enable equal opportunity for participation in future studies.

While remote studies stand to reduce geographic barriers to research participation, the three studies examined here failed to recruit a racially, ethnically, or educationally diverse sample of participants. Across the three studies, less than 5%were non-white, less than 7.5%were Hispanic/Latino, and 90.7–98.5%had more than a high school education. In part, this reflects lack of diversity of the population pools from which each study recruited. The challenge of recruiting diverse participants is not unique to remote studies [26]. Between 1985 and 2007, only 17%of PD trials reported the race and ethnicity of participants [27]. Of those that did report this information, only 0.9%of participants were Black, 0.7%Asian, and 1.1%Hispanic [27]. In addition to the expansion of broadband internet access, possible solutions include collaboration with community groups and trusted organizations and geo-targeted Facebook and Google recruitment efforts. The latter was used by Fox Insight to boost recruitment of non-white and Hispanic individuals [28]. Once individuals are identified, the free provision of devices and data plans may enable participation of those otherwise unable to participate. More research is needed in this area in particular to develop best practices.

While each study recruited a large number of participants, two of the three studies failed to meet pre-specified recruitment targets. FIVE was the only study to meet (and exceed) its target enrollment and featured the fastest recruitment. We posit that this was because individuals were already familiar with online PD research, initial contact came from a familiar source (MJFF), the duration of commitment was short, and there was an easily accessible website with study information. Recruitment does not appear to have been hindered by the lack of compensation. Of those who consented to study team contact, VALOR-PD was the most successful in reaching participants for a pre-screen—a critical step in the recruitment process—and converting consents to contact into enrollment. This may reflect enthusiasm for PD prevention research studies in a population with a genetic risk factor for PD. In contrast, AT-HOME PD, may have been hindered by a lack of email and mailing addresses or updated phone numbers, individuals’ unfamiliarity with the study team, disappointment with the negative results of the parent clinical trials, and/or lack of enthusiasm for the study’s aims. We saw an increase in response rate after collaborating with local parent study sites, suggesting that better cooperation from the outset may have enhanced enrollment. In addition, while FIVE and VALOR-PD drew from pools of individuals who were already engaged in online research, AT-HOME PD drew from a pool of individuals engaged in a traditional in-person research study.

A comparison of the three studies presented here offers preliminary suggestions for how to successfully recruit individuals into remote PD studies. One, initial contact about the study should come from a familiar source where possible. Large-scale outreach, such as podcasts and webinars publicized by a well-known organization in the PD community (e.g. MJFF), may be effective, as evidenced by the efficient recruitment in STEADY-PDIII [18]. Second, recruitment materials should be multimodal. Recommended modalities to include are online (website), phone, email, mail, and webinar/podcast to reach the broadest possible audience. Third, as with traditional studies, recruitment success should be tracked and continuously evaluated with strategies adapted as necessary. For example, in AT-HOME PD, we identified a low rate of successful telephone contacts and successfully adapted our approach to include email and mail-based recruitment strategies.

Our study is not without limitations. First, all three study designs featured a pre-specified pool of potential participants who were already engaged in some form of research. As such, our results may not be generalizable to the identification of potential participants from the general population. Second, none of the studies systematically tracked reasons for non-participation and analyses on the impact of changes in recruitment strategies were not pre-planned. Third, our analyses are based on only three studies. Given the large number of variables that may impact recruitment, our outcomes may reflect other aspects of study design unrelated to the remote decentralized design. Lastly, we do not have recruitment outcome data to enable direct comparisons with traditional, in-person research studies. Nonetheless, our experiences may inform recruitment strategies for future PD clinical trials.

Recruitment of geographically widespread individuals with and without PD into remote decentralized observational PD studies is feasible. Notably, initial satisfaction with the examined studies was uniformly high and participants expressed interest in participating in future remote PD studies supporting future expansion of such models. However, two of the studies failed to meet recruitment targets and none of the studies were racially, ethnically, or educationally diverse. More work is needed to develop best recruitment strategies, particularly in the recruitment of Black and Hispanic/Latino participants.