Using Continuous Glucose Monitoring to Motivate Physical Activity in Overweight and Obese Adults: A Pilot Study

Regular physical activity (PA) is associated with a reduced risk of many types of cancer, with the strongest evidence for endometrial cancer (>20% risk reduction) and breast and colorectal cancer (both >10% risk reduction; ref. 1). However, nearly half of American adults do not meet the national guidelines for aerobic PA, which call for 30 minutes of moderate-intensity PA 5 days a week (2). The physical inactivity rate is even higher among overweight and obese [body mass index (BMI) ≥25 kg/m²] American adults (3–5); in combination with the negative impact of excess weight, further increases the cancer risk in this population (6, 7). An estimated four in five American adults may be considered as overweight or obese in the next decade (8). Therefore, it is critical to promote regular PA in this population as a cancer prevention strategy.

Interventions that aim to increase PA levels often incorporate feedback as a behavior change strategy (9, 10). Most feedback provided in PA interventions is performance based, which gives data about recorded activity levels or evaluates PA performance in relation to a set goal (11). The effects of feedback-based interventions on behavior change are highly variable (12, 13). One reason that sedentary individuals are not physically active is that they are not sufficiently motivated to change their current behavior. According to the self-determination theory, individuals are more likely to engage in a behavior if they perceive it as personally important (14, 15). Other popular theoretical constructs, such as perceived benefits and outcome expectancy, also address the importance of individuals' beliefs about the positive outcomes of engaging in PA (16–18). However, one of the main challenges in motivating sedentary individuals to become more active is that PA is often characterized by immediate effort for possible future gain, with no obvious short-term benefit (19). Thus, there is a need to find innovative ways to help individuals recognize the more immediate health benefits of PA to enhance the feedback technique, optimize behavior change motivation, and subsequently increase their PA levels.

Providing feedback on individuals' biological indicators of health has been used to increase motivation and promote behavior change (20). Traditionally, such biological indices were obtained in a clinical setting (e.g., blood panel) at infrequent measurement points (e.g., annual physical examinations). However, recent advances in wearable sensor technology have made personal biological data available in real time (21); for example, continuous glucose monitors (CGM), which measure glucose concentrations in the interstitial fluid in real time through a tiny sensor inserted under the skin, have been used by patients with type 1 diabetes to promote glycemic control (22). Given that PA has an acute impact on glucose patterns in both diabetic and nondiabetic populations (23–27), data from CGMs could be used for providing biological feedback that demonstrates the immediate benefits of PA.

In patients with type 2 diabetes, CGM-based PA counseling in intervention settings has been shown to increase PA (28, 29). Incorporating CGMs into lifestyle interventions for nondiabetic individuals has also been explored (30, 31). The results suggest that leveraging data from biosensors is highly promising. However, no studies have investigated the acceptability of CGM-incorporated PA intervention in overweight and obese adults. The goals of this pilot study, My Moves, were to (i) determine the acceptability of a PA intervention that incorporated the use of CGMs in sedentary overweight and obese adults without diabetes and (ii) evaluate the changes in exercise motivation.

The My Moves intervention study included two in-person visits, with a 10-day self-monitoring period in between. During the first visit, participants completed the baseline assessment, received the study devices, and received a one-on-one PA education session. The 10-day self-monitoring period started the day after this first visit. At the second visit, participants returned the study devices, reviewed their data, and completed the postintervention assessment and an exit interview. More details about each study procedure are given below. All study participants provided written informed consent. The study was approved by the Institutional Review Board at the University of Texas MD Anderson Cancer Center.

Participants were recruited through public announcements (e.g., flyers, e-mail listserv) around the Texas Medical Center in Houston, Texas. They included individuals who worked or lived in communities near the medical center, as well as patients and visitors at MD Anderson who did not have cancer or diabetes. Interested individuals were asked to contact the study team to obtain a more detailed description of the study, which included a brief introduction to the CGM. Individuals who remained interested after reading the study description were sent a screening questionnaire to determine their initial eligibility.

Eligible individuals were men and women ages 18 to 65 years who had a BMI ≥25 kg/m²; engaged in less than 150 minutes of moderate-intensity PA per week in the previous month; could speak, read, and write English; and had a mobile phone with daily internet access. Individuals were excluded if they reported being diagnosed with diabetes; reported using any medication known to affect glucose levels (e.g., corticosteroids, antidepressants, metformin); reported health issues that might limit unsupervised PA; worked overnight shifts; were pregnant or lactating; were already using a wearable activity tracker or CGM; or were unable or unwilling to use a CGM.

Individuals who passed the initial eligibility screening were scheduled for an in-person visit with an overnight fast at MD Anderson. Their heights and weights were measured, and their fasting blood glucose levels were obtained using a commercially available glucometer. Individuals were deemed eligible for the study after their BMI status was confirmed and if their fasting blood glucose was <125 mg/dL. The recruitment goal for this study was to enroll 20 participants. This sample size was chosen because of the pilot nature of the study and on the basis of previous feasibility and acceptability studies that tested technology-based tools (31, 32).

Upon study enrollment, participants first completed a set of surveys that assessed demographic information and exercise motivations. They were then given a Fitbit Alta HR (Fitbit Inc.) wristband to wear, which continuously tracks steps and activity levels with heart rate monitoring. Study staff assisted participants with downloading the Fitbit application to their mobile phones and setting up the device. Participants were instructed to wear the Fitbit at all times, including while showering and sleeping, and to regularly sync it with the phone application. After the introduction to Fitbit, study staff went over a PA education session with the participant, which is detailed in the next section.

Following the education session, participants were given a Freestyle Libre (Abbott Laboratories) CGM system, which consists of a sensor (measures at 5 mm height and 35 mm diameter, and weighs 5 grams) and a scanner (measures at 95 mm × 60 mm × 16 mm and weights 65 g). Study staff helped participants to insert the sensor into the back of their upper arm. Upon activation, the sensor records interstitial glucose data every 15 minutes continuously for 10 days without the need for finger-stick calibration. Therefore, there was no need to replace the sensor during the 10-day self-monitoring period. The sensor was designed to be water-proofed (i.e., can withstand immersion into 3 ft of water for up to 30 minutes). To ensure the sensor stayed in place, participants were advised to avoid bathing, swimming, or participating in other activities that would submerge the sensor under water for a long period of time. Study staff demonstrated how to access real-time glucose information using the reader.

The 10-day self-monitoring period started the following day as detailed in the next section. Participants came back for another in-person visit at MD Anderson on day 11 to return the study equipment, review their Fitbit and CGM data over the past 10 days with the study staff, discuss their self-monitoring experience, and complete another set of surveys and an exit interview. Participants received up to $100 for completing the study.

This one-on-one session combined standard PA education materials (e.g., providing informational handouts and setting exercise goals) and CGM-based PA counseling that had originally been used in studies with diabetic patients (28, 29). Study staff first provided an overview of the health benefits of PA, different types of PA, and general tips to become more active in daily life (i.e., the handout component of the education session). The short-term (e.g., decrease stress and improve energy), medium-term (e.g., decrease body weight and improve insulin sensitivity), and long-term health benefits (e.g., reduce the risk of certain cancers and improve quality of life) of PA were highlighted. Participants were also presented with a Web-based glucose simulator (see Supplementary Fig. S1; ref. 33) that demonstrated acute PA (i.e., walking)-related glucose reduction. Next, study staff worked with the participants to calculate their targeted heart rate zone (sex and age adjusted) for light-, moderate-, and high-intensity PA. Finally, the staff worked with the participants to develop individualized exercise plans to help them to meet the goal of accumulating 150 minutes per week of moderate-intensity PA. The entire education session took about 30 minutes to complete.

The 10-day self-monitoring period started the day after the one-on-one PA education session. During this period, participants could check their accumulative steps and activity levels at any time through the Fitbit application. Participants were reminded about checking their heart rate reading from the Fitbit wristband while exercising to make sure that the activity reached moderate intensity. Participants' real-time Fitbit data and syncing information were captured by Fitabase (Small Steps Labs LLC), a Web-based platform that processes Fitbit data and generates reports.

Participants were also instructed to carry the CGM reader with them during most waking hours to access their glucose information in real time. Upon scanning the reader with the sensor, the reader displays the current glucose reading and a graph showing 8 hours of glucose history, with an arrow indicating the directional trend. Participants were instructed to obtain a glucose reading each morning when they woke up, each night before they went to sleep, and at least two other times throughout the day. Participants received four reminders each day through text messages for the scanning. The CGM reader records the date and time when the sensor is scanned. Participants were encouraged to observe how their daily glucose patterns were influenced by their behaviors (e.g., eating and exercising).

For the PA education session, participants rated each component (i.e., handout, Web-based glucose simulator, targeted heart rate calculation, and exercise plan) in regards to knowledge (i.e., helped me better understand the health benefits of PA), motivation (i.e., increased my motivation to be more active), and relevance (i.e., the information was relevant to me). The response options were rated using a 5-point Likert scale, ranging from “strongly disagree” to “strongly agree.” For the self-monitoring period, a 10-item survey was used to assess the acceptability of wearing the Fitbit and the CGM, as in previous studies (31). The survey items focused on addressing the barriers and facilitators to the use of mHealth tools, such as convenience, value, and relevance. The response options were rated using a 5-point Likert scale, ranging from “strongly disagree” to “strongly agree.”

The Behavioral Regulation in Exercise Questionnaire-2 (BREQ-2) and the Exercise Stages of Change – Continuous Measure (URICA-E2) were used to assess exercise motivation. The BREQ-2 captured the stages of the self-determination continuum, including amotivation, external regulation, introjected regulation, identified regulation, and intrinsic regulation, reflecting constructs in the self-determination theory (34). The responses were rated on a 5-point Likert scale, ranging from “not true” to “very true.” The URICA-E2 captured readiness to change, as described in the transtheoretical model, in six stages: precontemplation (nonbeliever), precontemplation (believer), contemplation, preparation, action, and maintenance, with the assumption that PA behavior is typically a dynamic process that moves between different stages (35). The responses were rated using a 5-point Likert scale, ranging from “strongly disagree” to “strongly agree.”

Descriptive statistics were generated for all variables, including the means and standard deviations for continuous variables and percentages for categorical variables. A summary score of acceptability was created for Fitbit and CGM by calculating the mean of the 10 survey items, as in previous studies (31). Two-tailed paired-sample t tests were used to examine changes in exercise motivations before and after the intervention. A P value of 0.05 or less was considered statistically significant. All statistical analyses were conducted using SPSS software (IBM Corp.) version 24.0.

Figure 1 summarizes the flow of participants from recruitment to study completion. A total of 79 individuals expressed an interest in participating in the study. Eleven were no longer interested after reading more information about the study. Among these, six felt uncomfortable with the CGM aspect of the study (i.e., did not want to get poked with a sensor, did not want to wear something on their body), two thought they were too active, three were due to other reasons (e.g., scheduling issue). Of the remaining 68 individuals, eight could not be reached after three contact attempts. Thus, the eligibility screener was sent to 60 individuals. Five individuals did not complete the screener, with no response after three reminder attempts; 21 were eligible for the study on the basis of their screener answers. Of these, 19 completed the study and were included in the analysis.

Table 1 shows the demographic characteristics of the 19 participants. Participants' average age was 42 years (SD = 8; range 26–55). Eighty-four percent (16/19) of the participants were female, 79% (15/19) were from a racial minority group, and 63% (12/19) were obese. Seventy-four percent (14/19) had an elevated fasting glucose (100–125 mg/dL) at their first visit.

Overall, most participants rated each of the education session components favorably. For the informational handout component, 78.9% (15/19) agreed that it improved their knowledge, and 84.2% (16/19) agreed that it was motivating and personally relevant. For the Web-based glucose simulator component, all agreed that it improved their knowledge, and 89.5% (17/19) agreed that it was motivating and personally relevant. For the heart rate zone calculation component, 84.2% (16/19) agreed that it improved their knowledge, 94.7% (18/19) agreed that it was motivating, and 89.5% (17/19) agreed that it was personally relevant. For the exercise plan component, 89.5% (17/19) agreed that it improved their knowledge, 78.9% (15/19) agreed that it was motivating and personally relevant. Figure 2 shows the mean rating score for each education session component.

Participants did not report any problems wearing the Fitbit or the CGM device during the self-monitoring period. All participants wore the Fitbit for the entire 10-day period and kept the device synced with the server through the Fitbit application (on the basis of information provided by Fitabase). Sixty-eight percent (13/19) of the participants reported checking their Fitbit application at least once a day. For the first two participants in the study, the CGM sensor fell off within the first 24 hours. New sensors were then placed with extra adhesive tape to secure them in place. All subsequent sensors stayed in place and collected data for the entire 10-day period. Participants, on average, scanned the sensor with the reader six times a day (SD = 2.4; range = 3–10). Participants followed the specified scanning schedule on average 8 days of the 10-day period (SD = 2.2; range = 1–10).

For the Fitbit device, all participants agreed with statements regarding usability, convenience, value, motivating, confidence, recommendability, and likability. Ninety-five percent (18/19) agreed with the statement regarding relevance, and 32% (6/19) expressed concerns about privacy. For the CGM, all participants agreed with statements regarding usability and confidence; 95% (18/19) with convenience, value, relevance, and recommendability; 84% (16/19) with motivating; 90% (17/19) with likability; and 26% (5/19) expressed concern about privacy. Table 2 shows the average score for each acceptability item. Overall, the Fitbit had an acceptability score of 4.51 (SD = 0.43), and the CGM had an acceptability score of 4.46 (SD = 0.53).

Table 3 shows the changes in exercise motivation before and after the study period. Overall, there was no significant change in self-determination motivation. However, there was a trend towards a decrease in amotivation and an increase in intrinsic regulation (P ≤ 0.10). In general, participants scored higher for the contemplation and preparation stages compared with the other stages, suggesting that most of them were considering a change in their physical activity behavior. After the study period, the score decreased significantly for the two precontemplation stages (P < 0.01) and increased significantly for the action stage (P = 0.01).

This study represents the first PA intervention to incorporate the use of CGM in sedentary overweight and obese adults. We found that there was high acceptability of a PA education module with CGM, which includes a one-on-one counseling session followed by a 10-day free-living self-monitoring period. For the one-on-one counseling session, the Web-based glucose simulator, which introduced the acute impact of PA on glucose patterns, was well received by the participants, suggesting that individuals without diabetes can understand and find value in glucose-based biological information. For the self-monitoring period, the acceptability rating of the CGM was comparable to that of the Fitbit wristband and was higher than that found in a previous observational study that used CGM in healthy individuals (31). One reason for this higher acceptability could be that the CGM model used in this study did not require daily finger-stick calibration, as the older model did. During the recruitment phase, six individuals (8% of all those who expressed an interest in the study) declined to participate in the study because they felt uncomfortable with the CGM device. As CGM technology advances and the idea of biosensors becomes more acceptable in society, we expect that the acceptability of CGM will only further improve. For example, the latest model of the Freestyle Libre no longer requires a separate reader to scan the sensor; users have the option of using their smartphone to retrieve the information.

Although the overall acceptability of using the CGM and Fitbit wearable devices was high, the rating for privacy was considerably lower than was that of all other acceptability survey items. Twenty-six percent (5/19) of the participants rated all items favorably (i.e., either agree or strongly agree) except for the privacy question for both CGM and Fitbit. To use the Fitbit application, all users are required to sign up for a Fitbit account. To ensure privacy, we created a study e-mail account for each participant. No personal information was associated with this account. The CGM reader does not require users to input any personal information. However, if a smartphone application is used to replace the reader in future studies, a user account will be required. In that case, we recommend the use of a study e-mail account instead of participants' personal e-mail accounts. Future studies should also further explore the participants' privacy about wearable devices, such as whether they worry about privacy issues with the device manufacturers or with the research team, and whether they worry about their personal information being leaked or about the inappropriate use of the data collected by the devices. Identifying these issues will help the research team to better address them in future studies.

Results from this study show promising effects on participants' exercise motivation, especially in helping them transition from the pre-contemplation stage of behavioral change (i.e., not intending to make changes) to the action stage (i.e., actively engaging in the new behavior). In the self-determination continuum, there was a trend towards decreasing amotivation and increasing intrinsic motivation to exercise. Research on exercise motivation shows the importance of autonomous (identified and intrinsic) regulation in promoting PA; the identified regulation is predictive of initial or short-term adoption of exercise, and the intrinsic motivation is predictive of long-term exercise adherence (36). Future intervention studies could find ways to package the information from CGM (i.e., the immediate impact of PA on glucose patterns) with other constructs of behavioral change strategies (e.g., building self-efficacy in exercising, overcoming perceived barriers in exercising) to develop a more comprehensive approach to adopting and maintaining an active lifestyle.

In this pilot study, we tested the CGM-featured education session and self-monitoring period in a nonclinical setting. Moving forward, this type of intervention has high potential for delivery in clinical settings. For example, future studies could test the feasibility of implementing the intervention in primary care clinics, targeting the prediabetic population—a group that might be particularly motivated to change their behavior on the basis of their glucose information. Most (74%) of the participants in the current study thought that the 10-day monitoring period was about the right duration for them to learn the acute benefits of PA, and 79% of the participants expressed the desire to repeat wearing the devices (in 6–12 months) to determine whether their glucose patterns had changed. Therefore, finding ways to incorporate the intermittent use of CGM along with other wearable devices such as Fitbit into regular clinical encounters (e.g., annual physical examinations) will enable participants to recognize the implications of their behavior changes for long-term health outcomes.

This study has several limitations, given its pilot structure. First, since it was a one-group study design, it is unclear how many of the positive changes in exercise motivation were attributable to the CGM-related component. Second, given the short duration of the study, it is unclear how long the increase in motivation will last or whether it will lead to actual changes in behavior. Third, most of the participants were female. It is unclear whether the results would be generalizable to males. Nevertheless, our pilot study included a diverse sample in terms of race/ethnicity (79% from a racial minority group), education level (53% without a bachelor's degree), and weight status (42% at moderate-to-high risk obesity; BMI ≥ 30). Our data provide preliminary evidence that CGM could be a viable tool in different subgroups and has the potentials to be integrated into tailored interventions targeting some of these groups. Further, although we did not collect information about participants' family history of diabetes, several participants mentioned that their immediate family members had diabetes during the exit interview. Future CGM-featured studies could consider targeting this specific group as they might be more motivated by glucose-related information.

In summary, the current study demonstrated a novel strategy to motivate sedentary overweight and obese adults to participate in PA using a biological sensor—CGM. CGM represents one type of wearable sensors that frequently assesses a biomarker, which is the foundation for providing just-in-time feedback (37). Since a sedentary lifestyle and excess weight are putting an increasing number of individuals at greater risk for various types of cancer, it is imperative to find innovative solutions and develop scalable strategies to promote regular PA in the overweight and obese population and thus reduce the overall burden of cancer.

E. Hawk is a vice president/division head at The University of Texas MD Anderson Cancer Center and reports receiving speakers bureau honoraria from Albert Einstein Cancer Center, NCI, Prevent Cancer Foundation Global Grants Program, Roswell Park Cancer Institute, Rutgers Cancer Institute of NJ, Simmons Comprehensive Cancer Center, University of California, Irvine, University of Nebraska Medical Center, VCU Massey Cancer Center, American Cancer Society, City of Hope, Fred Hutchinson Cancer Research Center, Huntsman Cancer Institute, Johns Hopkins Hospital, Kansas University Medical Center, and Mayo Clinic Cancer Center. No potential conflicts of interest were disclosed by the other authors.

Conception and design: Y. Liao, K.M. Basen-Engquist, D.L. Urbauer, E. Hawk, S.M. Schembre

Development of methodology: Y. Liao, D.L. Urbauer, S.M. Schembre

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T.B. Bevers, S.M. Schembre

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Liao, K.M. Basen-Engquist, D.L. Urbauer, T.B. Bevers, E. Hawk

Writing, review, and/or revision of the manuscript: Y. Liao, K.M. Basen-Engquist, D.L. Urbauer, T.B. Bevers, E. Hawk, S.M. Schembre

Study supervision: Y. Liao, K.M. Basen-Engquist, S.M. Schembre

Y. Liao was supported by a faculty fellowship from the Duncan Family Institute for Cancer Prevention and Risk Assessment at The University of Texas MD Anderson Cancer Center. K.M. Basen-Engquist and S.M. Schembre were supported by the NCI of the NIH under award number R21CA215415. This research was also supported by the NIH through MD Anderson's Cancer Center Support Grant (NCI Grant No. P30 CA016672). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

The authors acknowledge additional support for this project from the Assessment, Intervention and Measurement (AIM) Shared Resource and the Center for Energy Balance in Cancer Prevention and Survivorship at the University of Texas MD Anderson Cancer Center. The authors also thank the Department of Scientific Publications at MD Anderson for editing this manuscript.