Ted Olsen

07/18/2018

Beyond fitness tracking: The use of consumer-grade wearable data from normal volunteers in

cardiovascular and lipidomics research.

The trend of wearable fitness trackers ie the Smart watch, Fitbit and others have grown

significantly in recent years with the accessibility and expanded features of devices. Recent

market analysis expects to see the global market to exceed $34 billion US dollars by the year

2020. From the most basic of trackers that act as simple step counters to watches that are capable

of even monitoring heart rate (HR). These two basic functions have the potential to act as good

indicators to the wearers health. For instance, resting heart rate (RHR) is significant to a person’s

cardiovascular health. Step counts can be used to determine a person’s physical activity level.

Both RHR and step count can show prevention and prediction of risk of cardiovascular and

metabolic disorders (CVMDs).

With the importance of physical activity related to health outcomes, there has been an interest in

the use of fitness trackers in healthcare. Most research on fitness trackers have been focused on

increasing physical activity in healthy and diseased populations. Most studies have shown an

increase in physical activity after people start wearing fitness trackers, but no research has been

conducted to see if this increase results in clinically significant health outcomes. Researchers

conducted a year-long study on corporate employees showed that although fitness tracker

introduction increased physical activity, it did not find an improved health. It also found that

around 10% of participants were still using the fitness trackers after the year mark. More recent

studies have also started exploring how wearable data correlate with clinical and biological

markers. One study monitored fitness tracker data from 43 individuals and found that disease

states and physiological differences between individuals could be discerned from the data.

Another study conducted on personal data collected from 108 individuals determined physiology
Ted Olsen
07/18/2018

and disease did not identify any significant correlations with fitness tracker data. There are also

studies being conducted on time series HR data from trackers in the detection of conditions

associated with cardiovascular disease such as atrial fibrillation (AF), sleep apnea (SA), and

hypertension. They found that, deep neural networks (DNNs) trained on HR and step count data

obtained from the Apple Watch (Apple, www.apple.com) were able to detect AF, SA, and

hypertension at accuracies of 97%, 90%, and 82%, respectively.

The lack of comprehensive datasets that compare fitness tracker data with other data types means

that the utility of fitness trackers and clinical research, as well as personalized health, remains

unknown. In this study, the goal was to investigate how well fitness trackers could be used in

cardiovascular and lipidomics research. The researchers generated multidimensional data from

233 people recruited for a longitudinal study. Subjects generated data using a consumer-grade

fitness and HR tracker (Fitbit Charge HR; Fitbit, www.fitbit.com), in addition to basic health

knowledge through lifestyle questionnaires, clinical measurements (e.g., weight, height, waist

circumference [WC], blood pressure, etc.), lipid panel values, blood glucose test, cardiac

magnetic resonance imaging (CMR), and lipidomic profiling. The researchers then performed

analysis of the data in order to answer three specific questions. The first, can tracker data

obtained from the subjects provide information into their behavioral and demographic

characteristics. The second, do tracker metrics (both step- and HR-based) relate with CVMD risk

that they are useful in the areas of clinical research and personalized health monitoring. The last

question was can tracker metrics be used to support basic research in the analysis of cardiac

imaging and lipidomic profiling data.

One limitation of this study was the short duration of the tracking periods, thus compromising

power to detect associations between activity and CVMD markers. Their cardiac imaging and
Ted Olsen
07/18/2018

lipidomics analyses suggest that longer tracking periods especially if volunteers will share data

from their personal devices, would prove to be even more useful. Additionally, volunteers

recruited into this cohort may gain a higher level of regard for their health and well-being. They

did have a limited examination of time series wearable data. They determined that further studies

on the utility of features derived from the time series data (e.g., HR variability, HR recovery,

activity intensity) are needed.

In conclusion, the researchers found a sizeable relationship between fitness tracker data and a

wide range of volunteer phenotypes including lifestyle patterns, demographics, CVMD clinical

markers, cardiac imaging, and serum sphingolipid profiles. Their findings showed that apart

from fitness tracking, consumer-grade fitness trackers can play a role in both basic and clinical

research. The trackers could also provide a low-cost means for early detection of changes in an

individual’s personal CVMD risk profile, resulting in more timely detection and intervention of

CVMDs.