White Paper Clinical Trials & Research September 15, 2023 Choosing a commercially available wearable device for your research study With so many options available, including Fitbit, Apple Watch, Garmin, Oura, and more, how do you decide which devices—and their corresponding digital biomarkers—are right for your study? Abstract Wearable devices have gained significant popularity in clinical research due to their high scientific value and low participant burden. With a variety of commercially available wearable devices on the market, however, selecting a device to distribute for your research study can be challenging and overwhelming. These guiding principles for making such a selection cover three major categories of considerations: scientific priorities, the participant experience, and operational and protocol needs. We also discuss some specifics for a number of popular devices to help readers put these principles into practice. The promise of wearable devices in clinical research: longitudinal, real-time, and participant-friendly data collection for more complete data Wearable devices have garnered significant popularity among consumers over the past decade. Between 20-30% of adult Americans own a wearable, with a large percentage of those individuals using such devices on a daily basis. [1-3] This prevalence of devices and the benefits consumers have found from them have contributed to their growing popularity in clinical research, as well. Traditional clinical research protocols utilize in-clinic, episodic data collection, which limits the ability to collect longitudinal and contextual data. In recent years—and further spurred on by the pandemic—clinical research has begun to adopt virtual approaches to study protocols. These virtual or hybrid protocols incorporate novel, participant-friendly data collection methods—including wearable technology—to collect continuous, real-time data (Figure 1). This real-time data is of high scientific value for many research applications. When combined with their existing popularity with consumers, this form of high scientific value and low participant burden data collection has resulted in a significant increase in popularity of wearables among researchers. Figure 1: Wearable devices facilitate continuous or real-time data collection in virtual or hybrid clinical research models. Lower participant burden A critical aspect of the popularity of wearable devices in clinical research is the low burden on participants. While the setup of devices may require some digital literacy, the use of such devices following setup involves little effort from participants. Most commercially available wearables passively collect data throughout the individual’s daily life and only require attention when the battery needs to be charged. Thus, the overall participant burden when incorporating wearables in clinical research is lower than the majority of other protocol elements. Enable longitudinal, multi-parameter monitoring Commercially available devices are designed for long-term use—longer than six months—and often collect measurements in multiple data domains simultaneously. The most common data domains include sleep, physical activity, and heart health. This wealth of detailed information has implications across a wide variety of health domains (Figure 2), including cardiology, mental health and wellness, epidemiology, and more. [4-12] Figure 2: Wearable devices collect health information with utility across a wide variety of scientific domains. Image adapted from McCarthy, et al. [13] Deliver real-time interventions While multi-parameter monitoring is the most common use of wearable devices in clinical research, many devices are also being explored for delivering interventions. For example, patients with chronic fatigue syndrome with post-exertional malaise may use a wearable device to help with pacing—they can receive an alert when their heart rate rises to a level that indicates they have reached their anaerobic threshold. [14] Other wearable-based interventions include optimizing physical activity based on heart rate zones during exercise [15] and identifying the need for medical assistance due to sleep apnea [16] or irregular heart rhythms [17]. Device categories In general, wearable devices fit into two major categories: prescription or research-grade sensors and general wellness use devices. These categories differ greatly in how you might acquire the devices, the data you receive, and how the participant uses the device. Prescription sensors and research-grade devices Prescription sensors include wearables that are only available to a patient with a prescription; these devices will also carry clearance from the Food and Drug Administration (FDA), indicating their intended application and use. These sensors are often designed to collect data within a single scientific domain and include patches, wearables, and other devices like pulse oximeters, blood pressure cuffs, and continuous glucose monitors. Research-grade devices (e.g., Actigraph) are sensors designed and validated for research purposes. These devices are typically meant to be worn for short periods of time rather than used for longitudinal data collection and offer no return of results to the wearer. General wellness use devices General wellness use devices include commercially available, over-the-counter (OTC) wearables that do not require a “physician order” or a clinical indication for use. OTC devices are designed for long-term wear and simultaneously collect data in multiple scientific domains. These multi-parameter wearables include smartwatches, fitness trackers, rings, and other such devices. A number of studies have evaluated the accuracy of commercially available devices compared to research-grade devices. [18, 19] While research-grade devices may be more accurate in specific populations or use cases (e.g., older, sedentary adults), OTC wearables offer valid, longitudinal data collection and return of value to the consumer, which makes them popular choices for research studies. The guiding principles outlined below are intended for use when considering commercially available devices, not research-grade devices or prescription sensors. With a focus on general wellness use devices, it is important to acknowledge the existing popularity of these wearables. Up to 30% of the United States population already owns a wearable device. Bring-your-own-device (BYOD) model With a focus on general wellness use devices, it is important to acknowledge the existing popularity of these wearables. Up to 30% of the United States population already owns a wearable device. [1] Therefore, there are likely participants in your study that may not wish to take your selected device if they have their own. Incorporating a bring-your-own-device (BYOD) model would help accommodate that population. BYOD is an established model, with a number of studies demonstrating the ability to recruit BYOD participants at a large scale. [20, 21] While the upcoming guiding principles are framed around selecting a wearable device, the information provided may also help with understanding the differences in data collected from these devices if your research study includes a BYOD model. Guiding principles for selecting a general wellness use device When selecting a commercially available wearable device there are three major areas of consideration: your scientific priorities, your participants’ experience, and operational implications. Guiding questions in each of these areas are provided in Figure 3, including a scale by which to evaluate each device. On this scale, 0 indicates that the device does not satisfy that criteria for your study, and 4 indicates that the wearable would be a good selection based on that guiding principle. Guiding principle Evaluation scale 0 1 2 3 4 Incorporating scientific priorities Data domains: Does the device collect data within domains that are scientifically prioritized for your study? Granular/Raw data access: Does the device provide researchers access to granular/raw data? Considering the participant experience Consumer acceptance: Is the device largely accepted by consumers? Multi-platform: What platforms (e.g., iOS, Android) support the device? Battery life: How long does the device’s battery life typically last? Operational and protocol components Researcher familiarity: How familiar is your study team with the device and associated data? How frequently has the device been used in other studies? Cohorts supported: Does your study include pediatric participants? Are all participants approved to use the device? Cost: How much does each device cost? Is there an additional “premium” monthly cost? Figure 3: Guiding principles for selecting a commercially available wearable device for your research study. Commercially available wearable device market leaders produce similar base data When considering scientific priorities, there are several data domains for which the three wearable device market leaders—Fitbit, Apple Watch, and Garmin—provide similar base data. [22-24] These base data are listed in table 1. Table 1: Similar base data for market leaders Fitbit, Apple Watch, Garmin Heart Activity Sleep Other Heart rate Resting heart rate HRV (type varies significantly) Steps Distance Flights of stairs Calories Workouts / logged activities Sleep duration Light sleep duration Deep sleep duration REM sleep duration “Awake” or “In Bed” duration SpO2 Respiratory rate VO2Max For these base data, the three market-leading devices all produce some level of “intraday” data—not just daily processed metrics. In addition, each device produces its own set of unique parameters, some of which are outlined below. Fitbit: a popular selection for sleep studies Fitbit devices are, by far, the most widely used or distributed wearables in clinical research. The ease of procurement and shipping of devices to participants may be a contributing factor to this popularity for researchers. With the Fitbit Dropship API, Fitbit can receive device ordering data from a study app, like MyDataHelps™, and ship the device directly to study participants while providing the ordering details to the study team. [25] This does not require study teams to hold any inventory of devices or manually distribute them to participants, which eliminates burden on the team and prioritizes participant convenience. In addition to operational positives for using Fitbit devices, these wearables are typically well regarded for sleep tracking. [26] It is important to note, though, that HRV is only measured while sleeping. So while HRV is included in the base data seen from all three market leaders, its context is limited to that measured during sleep when using a Fitbit. Cost $150 (Charge 5) $300 (Sense 2) Platforms iOS/Android Studies 800+ from clinicaltrials.gov Battery life ~7 days Raw accel/gyro data? No Additional parameters Heart rate zone minutes Light/moderate/strenuous activity time Sleeping skin temperature Apple Watch: a popular selection for granular data One major Apple Watch feature that appeals to researchers is the availability of continuous, 24/7 raw accelerometer and gyroscope data. These data do require approval to collect via SensorKit, but they can be incredibly valuable for relevant research studies. [27] Outside of SensorKit, base data from Apple Watch, like HRV, is contextually different than other market leaders. In fact, Apple Watch HRV is infrequently recorded outside of the participant doing a “mindfulness session”. There are a few common concerns with using Apple Watch for research studies, including that they are only available to iPhone users and tend to have a shorter battery life than other wearables devices. The platform compatibility issue often raises health equity concerns, which can limit the generalizability of study results. In addition, a short battery life has resulted in many consumers charging their devices over night, which limits sleep tracking capabilities. Cost $249 (SE) $400 (Series 8) Platforms iOS Studies 100+ from clinicaltrials.gov Battery life ~18 hours Raw accel/gyro data? Yes Additional parameters Heart rate recovery Stand time Wrist temperature Low / High / Irregular heart rate events Garmin: a popular selection for continuous daytime HRV Garmin devices are highly regarded for their “body battery” feature. This feature incorporates continuous daytime HRV—available as an indication of “stress”, measured as the inverse of HRV—along with other activity data to provide an estimate of how much energy a person has throughout the day. Garmin devices are not well known for quality sleep tracking, however, and it is unclear if this has been resolved with newer models. Cost $150 (Vivosmart 5) $200 (Venu Sq) Platforms iOS/Android Studies 130+ from clinicaltrials.gov Battery life ~7 days Raw accel/gyro data? Limited Additional parameters Stress periods / duration Body battery Moderate / Vigorous activity time Other popular devices While Fitbit, Apple Watch, and Garmin are current leaders in the market for commercially available wearable devices for clinical research, there are other popular devices that may better suit your study or your participant population. Table 2 provides a few high-level notes on a handful of other devices. Table 2: Notable considerations for additional wearable devices Wearable Device Notes Oura Ring [28] Well regarded for sleep accuracy Must be sized to the individual Mostly exposes processed, daily metrics WHOOP [29] Well regarded sleep and “readiness score” tracking Very targeted toward exercise / fitness use case Requires a subscription per device Missing some basic parameters like steps; mostly exposes highly process parameters like “readiness scores” Polar HR [30] No on-device screen; designed for exercise use case Supports an SDK for accessing more “raw” data and HRV Primarily focused on heart monitoring, but does produce some sleep / activity data Withings [31] Supports “Advanced Research API” for getting raw accel/PPG data Used in 68 studies on clinicaltrials.gov Long 25+ day battery life Samsung Watch [32] Android only Limited prior use in studies Data availability may be improved with Google Health Connect (currently in beta) Discussion Looking collectively at the data, participant experience, and operational implications for a handful of popular wearable devices, it’s clear that some of these selection criteria are not trivial. For example, if your study intends to use HRV as a primary outcome measure, the context for this measure is important. You may choose to distribute Fitbit devices because HRV measured during sleep is considered more reliable. You may instead choose to provide participants with a Garmin device if you aim to associate changes in HRV throughout the day with corresponding events. In addition to selecting a wearable, it is important to consider what’s next. Using digital research technology, there are three methods that can have a large impact on keeping your participants engaged with the device after distribution (Figure 4): Sending automated notifications to participants to remind them to wear their device and sync their data is a low lift for study teams that can help increase adherence. Returning results or value to participants not only empowers patients with their own health data, but also actively engages them with your study over time. In this vein of providing value, distributing wearables to participants can also serve as an organic incentive. Allowing participants to keep their device after completing your study can ultimately increase adherence; studies have seen up to 98% compliance with this strategy. [33] Allowing participants to select a device model (e.g., smartwatch vs. fitness tracker) gives them a stronger sense of ownership and results in longer use of the device. Related to participant selection and ownership, incorporating BYOD models can also increase engagement because participants are already familiar with their devices and committed to using them. Figure 4: Three methods to increase participant engagement with study-provided wearable devices. While BYOD models can be considered more participant-inclusive, the analysis of data coming from various devices and models is complicated. [34] Due to the differences in data received from different wearables, researchers may not be able to compare point-in-time values across participants. Rather, a more scientifically rigorous approach might include processing individual trends that could then be compared across devices and between participant cohorts. Despite the popularity of wearables in clinical research and the benefit of receiving longitudinal, real-time data, some researchers still hesitate to incorporate these devices into their studies with questions around accuracy of the data. Data received from wearable devices do have significant scientific value, but when compared to research-grade devices specifically designed for a single measurement, there may be discrepancies. Some studies may therefore wish to use commercially available devices to capture broad data for long study durations, but use these data to trigger in-depth, shorter term sub-studies using more specialized devices. Conclusion Wearable devices are a source of great promise for enhancing clinical research, but not all wearables are created equally. When selecting a device to distribute to participants, it’s important to consider your scientific priorities (data domains, raw data access), the participant experience (consumer acceptance, platform compatibility, battery life), and your operational needs (researcher familiarity, cohorts supported, cost). These guiding principles can be used as a starting point for understanding the nuances of each wearable device, both when selecting a primary device to distribute for your research study and when incorporating a BYOD model. MyDataHelps Designer lets you build robust digital clinical trials with real-world data Create a free account to start designing your study today! Get Started – Create a Free Account You may occasionally receive account-related emails. We promise not to share your email with anyone. By registering, you agree to the MyDataHelps Designer Privacy and Use Policy. References Chandrasekaran, R., Katthula, V., & Moustakas, E. (2020). Patterns of Use and Key Predictors for the Use of Wearable Health Care Devices by US Adults: Insights from a National Survey. Journal of Medical Internet Research, 22(10): e22443. https://doi.org/10.2196/22443 Holko, M., et al. (2022). Wearable fitness tracker use in federally qualified health center patients: strategies to improve the health of all of us digital health devices. NPJ Digital Medicine, 5:53. https://doi.org/10.1038/s41746-022-00593-x Vogels, E. A. (2020) About one-in-five Americans use a smart watch or fitness tracker. Pew Research Center. https://www.pewresearch.org/short-reads/2020/01/09/about-one-in-five-americans-use-a-smart-watch-or-fitness-tracker/ de Zambotti, M., et al. (2019). Wearable Sleep Technology in Clinical and Research Settings. Medicine & Science in Sports & Exercise, 51(7), 1538–1557. https://doi.org/10.1249/mss.0000000000001947 Grandner, M. A. (2017). Sleep, Health, and Society. Sleep Medicine Clinics, 12(1), 1–22. https://doi.org/10.1016/j.jsmc.2016.10.012 Grandner, M. A. (2017). Sleep, Health, and Society. Sleep Medicine Clinics, 12(1), 1–22. https://doi.org/10.1016/j.jsmc.2016.10.012 Grandner, M. A. (2017). Sleep, Health, and Society. Sleep Medicine Clinics, 12(1), 1–22. https://doi.org/10.1016/j.jsmc.2016.10.012 Pevnick, J. M., et al. (2018). Wearable technology for cardiology: An update and framework for the future. Trends in Cardiovascular Medicine, 28(2), 144–150. https://doi.org/10.1016/j.tcm.2017.08.003 Reimers, A., Knapp, G., & Reimers, C. D. (2018). Effects of Exercise on the Resting Heart Rate: A Systematic Review and Meta-Analysis of Interventional Studies. Journal of Clinical Medicine, 7(12), 503. https://doi.org/10.3390/jcm7120503 Sett, N., et al. (2019). Are You in Pain? Predicting Pain and Stiffness from Wearable Sensor Activity Data. In M. Bramer & M. Petridis (Eds.), Artificial Intelligence XXXVI (SGAI 2019. Lecture Notes in Computer Science ed., Vol. 11927, pp. 183–197). Springer. https://doi.org/10.1007/978-3-030-34885-4_15 Shaffer, F., & Ginsberg, J. P. (2017). An Overview of Heart Rate Variability Metrics and Norms. Frontiers in Public Health, 5(258). https://doi.org/10.3389/fpubh.2017.00258 Warburton, D. E. R., Nicol, C. W., & Bredin, S. S. D. (2006). Health benefits of physical activity: the evidence. Canadian Medical Association Journal, 174(6), 801–809. https://doi.org/10.1503/cmaj.051351 McCarthy, M., Ballinger, R., & Lewis, H. (2020) Advancing digital endpoints: An end-to-end approach to managing wearable devices through clinical development. ICON Plc. Retrieved February 14, 2022 from https://www.iconplc.com/insights/digital-disruption/mHealth-wearables/digital-endpoints/ Jackson, S. Using a Heart Rate Monitor to Prevent Post-Exertional Malaise in ME/CFS | Solve M.E. Published August 15, 2019. Retrieved July 19, 2023 from https://solvecfs.org/using-a-heart-rate-monitor-to-prevent-post-exertional-malaise-in-me-cfs/ Duking, P., et al. (2016). Comparison of Non-Invasive Individual Monitoring of the Training and Health of Athletes with Commercially Available Wearable Technologies. Frontiers in Physiology, 7: 71. https://doi.org/10.3389/fphys.2016.00071 Ganglberger, W., et al. (2022). Sleep apnea and respiratory anomaly detection from a wearable band and oxygen saturation. Sleep and Breathing, 26(3): 1033-1044. https://doi.org/10.1007/s11325-021-02465-2 New Fitbit feature makes AFib detection more accessible | Google Products. Published April 11, 2022. Retrieved July 19, 2023 from https://blog.google/products/fitbit/irregular-heart-rhythm-notifications/ O’Driscoll, R., et al. (2020). The validity of two widely used commercial and research-grade activity monitors, during resting, household and activity behaviours. Health and Technology, 10: 637-648. https://doi.org/10.1007/s12553-019-00392-7 Tedesco, S., et al. (2019). Accuracy of consumer-level and research-grade activity trackers in ambulatory settings in older adults. PLoS ONE, 14(5): e0216891. https://doi.org/10.1371/journal.pone.0216891 Radin, J. M., et al. (2022). Sensor-based surveillance for digitising real-time COVID-19 tracking in the USA (DETECT): a multivariable, population-based, modelling study. The Lancet Digital Health, 4(11): e777-e786. https://doi.org/10.1016/S2589-7500(22)00156-X All of Us Research Program | National Institutes of Health (NIH). (n.d.). National Institutes of Health (NIH) — All of Us. Retrieved July 19, 2023, from https://allofus.nih.gov/ Web API Reference | Fitbit Developer. Retrieved July 19, 2023, from https://dev.fitbit.com/build/reference/web-api/ WatchKit | Apple Developer. Retrieved July 19, 2023, from https://developer.apple.com/documentation/watchkit/ Garmin Connect Developer Program | Garmin Developer. Retrieved July 19, 2023, from https://developer.garmin.com/gc-developer-program/overview/ Fitbit: Device Distribution Options | MyDataHelps Designer Help Center. Retrieved July 19, 2023, from https://support.mydatahelps.org/hc/en-us/articles/8617448886163-Fitbit-Device-Distribution-Options Haghayegh, S., et al. (2019). Accuracy of Wristband Fitbit Models in Assessing Sleep: Systematic Review and Meta-Analysis. Journal of Medical Internet Research, 21(11): e16273. https://doi.org/10.2196/16273 SensorKit | Apple Developer Documentation. Retrieved July 19, 2023, from https://developer.apple.com/documentation/sensorkit API Documentation | Oura Ring Developer. Retrieved July 19, 2023, from https://cloud.ouraring.com/docs/ WHOOP API Docs | WHOOP for Developers. Retrieved July 19, 2023, from https://developer.whoop.com/api/ Polar Accesslink API v3 | Polar. Retrieved July 19, 2023, from https://www.polar.com/accesslink-api/#polar-accesslink-api Withings developer documentation (2.0) | Withings. Retrieved July 19, 2023, from https://developer.withings.com/api-reference/ Galaxy Watch | Samsung Developers. Retrieved July 19, 2023, from https://developer.samsung.com/galaxy-watch Michigan Predictive Activity and Clinical Trajectories (MIPACT): 98.3 percent adherence in a 6,735 participant, daily, at-home blood pressure monitoring study | CareEvolution Case Studies. Retrieved August 15, 2023, from https://careevolution.com/case-studies/michigan-predictive-activity-and-clinical-trajectories-mipact/ Demanuele, C., et al. (2022). Considerations for Conducting Bring Your Own “Device” (BYOD) Clinical Studies. Digital Biomarkers, 6:47-60. https://doi.org/10.1159/000525080
