Feasibility and preliminary efficacy of a novel web-based physical activity intervention in adults with overweight/obesity: a pilot randomized controlled trial

Abstract

Background:

Only 14% of adults with obesity attain federal guidelines for physical activity (PA), but few interventions address obesity-specific barriers to PA. We designed the web-based Physical Activity for The Heart (PATH) intervention to address this gap.

Purpose:

Test the feasibility and preliminary efficacy of PATH for promoting PA and reducing cardiovascular disease (CVD) risk in adults with overweight/obesity.

Methods:

In a 12-week pilot RCT, participants were randomized to PATH (n=41) or wait-list control (n=41) groups. Treatment group received access to PATH and met twice/month with a remote coach. The control group received a self-help PA guide and newsletters on general health. Moderate-to-vigorous PA (MVPA) was assessed via Actigraph-GT3X, steps via Fitbit Charge 2^™, weight via smart scale, blood pressure (BP) via Omron BP device, and lipids/HbAIC via dry blood spot. Linear mixed modeling examined between- and within-group differences in PA and CVD risk.

Results:

The sample (N=82) was on average 55.9±8.2 years old; mean BMI 35.5 ±6.2 kg/m²; 57.3% white and 80.5% female. Recruitment lasted 6-months, and 12-week retention was 96.3%. Treatment group accessed PATH ≥twice/week (92.1%), spent ≥10 minutes/visit (89.5%) and thought the site was culturally appropriate (79%). At 12 wks, the PATH group had greater mean changes in weekly MVPA (+58.9 vs. +0.9 min, p=.024) and daily steps (+1,246.4 vs. −64.2 steps, p=.002) compared to the control group. Also, the PATH group improved in weight, BMI, body fat, waist circumference, and BP (p<.05).

Conclusion:

The PATH intervention is feasible/acceptable and demonstrated preliminary efficacy for promoting PA among adults with overweight/obesity.

Introduction

The United States’ 2018 Physical Activity (PA) Guidelines recommend that adults should attain ≥150 minutes of moderate intensity PA, 75 minutes of vigorous PA, or an equivalent combination of both (MVPA) weekly.^1,2 Yet, adherence to these Guidelines is low, with 14% of adults with obesity attaining the minimum recommended PA levels.^3,4 Low levels of PA are associated with the rising prevalence of obesity and increase the relative risk of cardiovascular disease (CVD).^5–7 Since individuals with obesity are more vulnerable to CVD, weight-loss is recommended⁸ though long-term obesity treatment outcomes remain poor.^9,10 Yet, even without weight-loss, PA can significantly reduce the risk of CVD among individuals with obesity.^11–13 However, individuals with obesity face multifaceted barriers that reduce their long-term engagement in PA.^14,15

In the general population, common barriers to PA include lack of time, amotivation, and low exercise self-efficacy.^16,17 In addition, individuals with obesity face weight-related impediments including reduced mobility, poor fitness, and stigma which contributes to aversion of PA.^15,18,19 These impediments have been observed in both white^14,15 and minoritized samples^20–22 and inform ongoing efforts to develop web-based interventions that are efficacious for a diverse group of insufficiently active adults with obesity. Limitations of existing interventions include lack of human contact, unmet weight-loss expectations, and generic content that fails to address the weight-related impediments.^15,23

A growing body of evidence suggests that insufficiently active individuals with overweight/obesity prefer PA programs that include vicarious experiences^24,25 featuring role models who are diverse and relatable in terms of body size, fitness level, and age.^15,20,26 Yet, few PA interventions are designed to incorporate vicarious experiences tailored to the preferences of insufficiently active individuals with obesity.^27–29 We conducted a 12-week pilot randomized controlled trial (RCT) to examine the feasibility and preliminary efficacy of a web-based PA intervention designed to reduce the risk of CVD in insufficiently active adults with overweight/obesity.

The Feasibility was assessed via a) number of participants screened for eligibility; b) percentage eligible; c) percentage enrolled; d) percentage retained within each randomized group; and e) adherence to self-monitoring (Actigraph wear time) and study protocol (e.g., PATH utilization, coaching sessions attended). Acceptability was evaluated via user ratings/comments on specific workouts included in PATH, and a post-intervention survey developed by study team to appraise the intervention’s usability and perceived efficacy for increasing PA (Appendix 1). Selected cardiometabolic risk factors were considered as secondary outcomes for assessing preliminary efficacy. The study was funded by two intramural grants, with one specifically focusing on increasing the participation of underserved populations in clinical research. Consequently, the sample was recruited in two concurrent cohorts, one focusing on African Americans (n=30), while the other was race/ethnic agnostic (n=52). The study staff and all the study methods and procedures were similar for both cohorts.

Methods

Intervention design

The design and development of the web-based PA for The Heart (PATH) intervention has been described elsewhere.³⁰ Briefly, the PATH intervention leverages openly accessible platforms, such as YouTube, to provide workout videos that match the preferences expressed in our formative studies^26,29 and the extant literature.^14,15,20 The intervention development was guided by the social cognitive theory’s (SCT)²¹ proposition that observing similar others succeed can motivate action and help demonstrate a plan for success.²⁴ The workout videos included in the PATH program provide vicarious experiences by featuring PA role models with diverse body sizes, fitness levels, and age. In addition, the workouts provide a variety of enjoyable PA options that allow participants to build their PA skills using their favorite routines, at their own pace, and in environments where they are less likely to feel embarrassed. These strategies were employed to enhance exercise self-efficacy as envisioned in the SCT.

The intervention development process entailed an iterative process (Figure 1) where our target population was engaged in the selection and rating of the workout videos. Then, highly rated workout videos were vetted by the study team for content relevance/safety. Appropriate videos were then curated on our PATH website in 3 levels of intensity (beginner, intermediate, proficient) to foster gradual progression from low to higher intensity PA. The study design was guided by the Obesity Related Behavioral Intervention Trials (ORBIT) development model.³¹ The vetted workouts are available without ads even at low internet speed.³² We also added backend features that enable a remote health coach to help users set their PA goals and select a PA regimen that is safe for their fitness level. Additionally, the PATH website has a feedback loop facilitated by intuitive tools embedded on each workout (comment box, rating of perceived exertion (RPE) and 5-star favorability rating) to allow users to provide timely feedback on each workout they use.

Figure 1.

The PATH intervention development process, dashboard, and feasibility study overview

The PATH platform has the capability to interface with any wearable sensor offering accessing to its application programming interface (API) and is currently interfaced with Fitbit and Actigraph. PATH has a user-facing self-monitoring dashboard that displays progress towards the weekly PA goal and the recommended workouts to attain the goal (Figure 1). The moderated forum on PATH allows users to share and comment on each other’s experiences, providing a sense of community. Lastly, the PATH Fresher algorithm analyzes PATH website utilization metrics and interacts with the YouTube recommender system³³ to identify new workouts that are comparable to those favorably rated by PATH users. The new recommendations are saved for the study staff to evaluate making the PATH program highly scalable and easy to update. Other innovative features of the PATH website are detailed elsewhere.³⁰

Beta testing the technology and usability of PATH

The PATH platform was beta tested in a convenient sample of 25 community dwelling participants (mean age [SD], 54.3 [9.5] years; 81% female; mean BMI [SD] 35.3 [7.7] kg/m²) who had not used PATH before. The participants were asked to watch ~120 min of workout videos every week for 12 weeks. Technical readiness was evaluated using PATH utilization data collected via automated web analytics and technical logs that included time stamped records of PATH web login patterns, devices used, resources accessed on PATH platform, ratings, and duration of access. PATH usability was assessed via an end of study survey that was developed by the study team to evaluate perceived esthetics, access, navigability, and content appropriateness (see Appendix 1). On average, the participants reported visiting the beta PATH website at least twice per week (61%) where they spent ≥10 minutes per visit (91%). They thought that the beta PATH website was appealing (70%) and easy to navigate (96%) and were very likely to use the workouts for exercise (mean score 83.8±16.7%). Objective tracking data revealed sustained engagement throughout the 12-week period.

Setting, sampling and study population for the pilot RCT

The study was conducted in Western Pennsylvania after approval by the University of Pittsburgh IRB and registration on clinicaltrials.gov: NCT04621045 and NCT04280783). Written informed consent was obtained before enrollment. For secondary outcomes, with a sample of 82 (41 per treatment group), we could detect medium-to-large effect sizes in terms of the differences between the treatment arms in the change in a continuous-type outcome from baseline to 12 weeks as small as d=0.771 at a significance level of .01 (adjusted for multiple testing) with .80 power. To obtain our target sample, we employed multiple recruitment strategies including leveraging research registries of individuals interested in participating in PA studies, and flyers in the community/primary care providers’ offices. The RCT was completed in March 2022.

Screening and eligibility

Individuals who responded to recruitment solicitations were provided a brief online overview of the study and eligibility criteria. The eligibility criteria included: access to the Internet, age 40–70 years, BMI ≥25 kg/m², self-monitoring of PA (≥4 days with ≥10 hours wear time) via waist worn Actigraph at baseline, and self-reported insufficient PA (<150 min of MVPA/week). Exclusion criteria entailed pregnancy, mobility restrictions, any condition requiring supervised PA, and history of CVD or diabetes. Given the small sample size, we introduced age limits to recruit a sample with comparable risk and safety profile. Evidence-based CVD risk assessment guidelines consider ≥40 years of age as an independent risk factor of CVD,³⁴ while the earlier version of the physical readiness questionnaire used in this study recommends that individuals ≥70 years old should engage in unsupervised PA after getting clearance from a clinician.³⁵ Successful self-monitoring of PA before randomization was employed to orient participants with the self-monitoring burden and collect objective baseline PA measures. Those eligible completed REDCap questionnaires on sociodemographic, health history and lifestyle habits. The Consort flow diagram (Figure 2) outlines the screening, randomization, and participant flow throughout the 12-week RCT.

Figure 2.

The PATH CONSORT Participant Flow Diagram

Randomization and study measurements

Eligible participants were randomized with equal allocation (1:1) to either the treatment or wait-list control arm using treatment assignments generated via minimization with pre-randomization stratification based on gender and age (≤55 years, >55 years). After randomization participants were oriented to either the PATH intervention or wait-list control group resources.

The key study assessments were conducted remotely, under Zoom supervision, by the study staff. Participants were instructed to empty their bladder, wear light clothing, and stand barefoot on a smart scale³⁶ to measure their weight and percent body fat. These measures were relayed to the study team via Fitbit API. To measure BP, participants were instructed to apply the Omron Model BP7450 cuff³⁷ on their bare non-dominant arm and relax, sitting upright in a chair with both feet flat on floor and arm resting on a flat surface at heart level. After 5 minutes of rest, they were asked to turn on the BP machine and take 3 measurements at one-minute intervals and show the monitor screen to the study staff who entered the 3 readings in a REDCap data form.

The waist circumference was assessed using the Perfect Tape measure following the self-measurement video guidelines.³⁸ The participant showed the measurement to the study team for recording in REDCap. Blood samples for lipids (LDL, HDL, and total cholesterol) and HbA1C were collected via fingerstick using HemaSpot SE kits³⁹ under Zoom supervision by the study staff and then mailed to the lab for analysis using established protocols.^39,40

The Actigraphs and Fitbits were shipped to participants with other baseline assessment equipment. Eligible participants were asked to wear the Actigraph GT3X on their waist and Fitbit Charge 2^™ tracker on the wrist for 7 consecutive days to monitor baseline MVPA and step count. Actigraph GT3X needed to be worn for ≥10 hours on ≥4 days, while the Fitbit needed to record ≥500 steps on ≥4 days to be eligible for randomization. The Actigraph was only used for 1 week at baseline and end of study and the activity data were used to compute the main outcomes MVPA. Actigraph non-wear time was defined as ≥90-minute periods of zero Counts Per Minute (CPM) allowing for ≤2 min of ≤100 cpm using the Choi wear time validation algorithm.⁴¹ The Freedson cut points were used to classify MVPA (≥1952 cpm),^42,43 The average weekly MVPA was computed as total MVPA recorded for the days of the week with valid wear time (≥10 hours/day) divided by the number of valid days multiplied by 7. Fitbit valid data were defined as having ≥4 days with ≥500 steps/day to represent typical PA and eliminate days when the monitor was likely not worn. This approach was recommended by Thorndike et al.,⁴⁴ and has been employed in other studies.⁴⁵ The average daily step count for each week was computed as total steps count for the days of the week with valid wear time (≥500 steps/day) divided by the number of valid days. We used ≥7,500 steps/day cut-off to identify participants who met what is considered to be equivalent to the 150 mins of MVPA/week.⁴⁶ The Fitbit was used for continuous monitoring of step count throughout the study and the data captured via API with Fitbit.

PATH utilization data was collected via automated web analytics and technical logs that included time stamped records of PATH web login patterns, devices used, specific resources accessed on the PATH platform, and duration of access. User ratings and comments on specific workouts included in the PATH platform were collected using a comment box and 5-star rating system embedded with each workout included in PATH. The post-intervention survey developed and piloted in the Beta testing phase was used to appraise the intervention’s usability and perceived efficacy for increasing PA (Appendix 1). All the baseline assessments were repeated at 12 weeks and at 24 weeks for control group participants who crossed over to use PATH. Individuals randomized to the wait-list control arm were given access to the intervention after completing their control group commitment at 12 weeks. After using the intervention for 12 weeks they repeated the assessments at the 24^th week. All participants wore their Fitbits throughout the period they were involved in the study.

Intervention and Control Group Delivery Procedures

Intervention group

Individuals randomized to the PATH intervention were granted access to one of the 3 PATH levels based on their baseline fitness status (see below for more information). During an initial remote session, a health coach worked with each participant to develop their PA prescription guided by the FITT principle (frequency, intensity, time, type)⁴⁷ with a goal of increasing baseline PA by ≥6000 steps per week by the end of the study. The PA prescription began with the coach identifying the appropriate PATH level for each participant based on their estimated VO₂ max (maximal oxygen uptake during exercise),⁴⁷ calculated via a validated non-exercise prediction model (covariates: sex, age, waist circumference, resting heart rate and PA index).⁴⁸ Participants with estimated VO₂ max below the 35^th percentile were assigned to the Beginner PATH level (light intensity PA), while those between the 35^th and 75^th percentile were assigned to the Intermediate PATH level (moderate intensity PA). No participant was assigned to the Proficient PATH level (vigorous intensity PA) at baseline. The coach guided each participant in selecting their weekly PA goal and workout videos that could help them attain the goal. The coach instructed the participants to use workout videos at an intensity level of 12 to 14 on the RPE Scale to promote safety. A Borg RPE Scale⁴⁹ was embedded with each workout video included on PATH platform. During subsequent remote coaching sessions (twice a month), progress toward PA goals was reviewed and a new weekly goal set. Transition to the next PATH level was allowed by the coach when most workouts within the assigned PATH level were perceived to be “very light” (≤9 on RPE Scale). This approach promoted safety and gradual progression along the PA continuum. Participants were also asked to share their experiences on the PATH community forum moderated by the study staff. Lastly, they received motivational PA reminders (text/email) at the time they indicated they were more likely to engage in PA.

Wait-list control group.

The wait-list control group received an electronic copy of the Be Active Your Way booklet, which guides individuals on pragmatic strategies for integrating PA into their daily lives, and bi-weekly newsletters focusing on general health every two weeks to match the contact frequency.⁵⁰ The control strategy was chosen to reflect usual care where primary care providers encourage their patients to increase PA using self-help handouts. After 12 weeks, the wait-list control group was given access to the full PATH intervention.

Statistical analysis

Recruitment feasibility was evaluated based on the ability to recruit, screen, and enroll the target sample within 6 months, and retaining ≥80% of the sample. Feasibility of PA self-monitoring was indicated by the proportion of the sample with ≥4 days per week of valid wear time. Feasibility of remote testing of HbA1C and lipids was evaluated by examining the proportion of the sample who successfully collected blood sample with valid results. Acceptability of the PATH intervention was indicated by PATH utilization data, collected via web- analytics, and a post-intervention survey focusing on PATH utility. Data were initially screened for any issues (e.g., outliers, missing data, multicollinearity, etc.) that could impact the validity of the results.

Baseline characteristics were summarized as means with standard deviation (SD) for continuous variables or as frequencies with percentages for categorical variables. Treatment groups were compared at baseline using chi-square or Fisher exact tests for categorical descriptors, and two-sample t-tests tests for continuous-type descriptors. Proportions or means with 95% confidence intervals (CI) using small sample approaches were used to estimate parameters related to feasibility and acceptability of the PATH intervention.

Following intention to treat principle, generalized linear mixed-effects modeling (assuming a normal distribution and identity link for normally distributed outcomes and binomial distribution and logit link for binary outcomes) was used to estimate within- and between-group differences in PA and CVD risk factors at the end of the study. As part of the linear mixed modeling, the distribution of the residuals was assessed. In instances where the distribution of the residuals was non-normal, data transformations (e.g., square-root, logarithmic) were applied to the outcome variable to induce a more normal distribution in the residuals. There were also some instances where there were outliers, however, application of data transformation appeared to remedy these extreme values. Sensitivity analyses were also conducted for models based on the outcome data that was not transformed. In general, our results were robust and did not change with application of data transformations or the omission of outlying observations. Consistent with linear mixed modeling, data were assumed to be missing at random, and all participants were included in the analysis unless there were no data on the outcome of interest at either time point. Covariate adjustment was considered secondarily based on the literature, and sensitivity analyses conducted to explore the robustness of parameter estimation in the event of outliers, as well as variability by race and sex. No influential outliers were identified. Data were analyzed using SAS (version 9.4, SAS Institute, Inc., Cary, NC).

Results

The CONSORT diagram below outlines the flow of participants through each stage of the RCT.

Feasibility and acceptability

As illustrated in the CONSORT participant follow diagram (Figure 2), we screened 432 individuals, enrolled 82 over 6 months, and retained 96.3% at 12 weeks (1 dropout (control), 2 withdrawn (treatment) for personal reasons). The sample (N=82) was on average 55.9 ±8.2 years, 57.3% white, 80.5% female, with obesity (mean BMI 35.5 ±6.2 kg/m²). The treatment groups were balanced on baseline characteristics and values for key study outcomes (Table 1).

Table 1.

Baseline Characteristics of PATH Pilot RCT Participants by treatment arm

Characteristic	PATH Treatment n=41	Control n=41	P-value
Race, n (%)
White	24 (58.4)	23 (48.9)
Black	16 (39.0)	17 (41.5)	.975
Others	1 (2.44)	1 (2.44)
Female, n (%)	33 (80.5)	33 (80.5)	1.000
Age (years), mean ± SD	55.9 ± 8.0	55.9 ± 8.5	1.000
Education category
< 4 years of college, n (%)	17 (41.5)	14 (34.2)	.494
≥ 4 years of college, n (%)	24 (58.5)	27 (65.6)
Employment category
Unemployed/retired, n (%)	27 (65.9)	28 (68.3)	.814
Employed/self-employed, n (%)	14 (34.2)	13 (31.7)
BMI (kg/m2), mean ± SD	35.9 ± 6.8	35.2 ± 5.8	.398
Systolic BP, mean ± SD	120.9 ± 17.3	123.8 ± 15.5	.413
Diastolic BP, mean ± SD	79.9 ± 8.9	80.7 ± 8.6	.418
Self-reported weekly MVPA using MAQ, mean ± SD	103.6 (100.1)	97.0 (83.7)	.749
Objective weekly MVPA minutes, mean ± SD	196.4 ± 151.2	208.1 ± 150.1	.727
Adherence to MVPA guidelines (≥150 mins/wk), n (%)	22 (53.7)	19 (46.3)	.508
Objective average daily steps, mean ± SD	4889.1 (3077.9)	4629.8 (2561.1)	.680
Adherence to steps guidelines (≥7500 steps/day), n (%)	8 (19.5)	9 (22.0)	.785

Nearly all intervention participants accessed the PATH website at least twice/week (92.1%), where they spent ≥10 min/visit (89.5%) and thought that the site was navigable (97.4%). Most of the participants also reported that the PATH website was very appealing (68.4%), and culturally appropriate (79%). When asked “What did you like most about the PATH website?” and “Was there anything in particular that you did not like about the PATH website?” the participants provided open ended responses that were coded and summarized as the most favorable and least likable components of PATH as depicted in Figure 3 A and B, respectively.

Figure 3.

Participants views on likeability of various PATH components

In Figure 3B, some of the participants did not like their PATH content due to controlled access where we restricted their access to workout videos based on their baseline fitness levels. Others (7%) reported that their PA trackers did not always sync with their apps leading to discrepancies between the step count displayed on their tracker and what was on their PATH dashboard. Actigraph LLC investigated and acknowledged the glitches in their software and provided technical support to resolve them. About 12% of the participants reported glitches on our PATH website where they would watch a workout video and then their watch time was not captured in the user facing dashboard. Our technology support team promptly addressed these problems. Lastly, 14% of the participants did not like our PATH website layout when viewed on a mobile device because the step count tracking graphics were not displayed.

There was sustained intervention engagement with minimal login decay over the entire study period (Fig. 4). The users provided 1,030 RPE ratings (78% ≤14), 974 favorability ratings (84% ≥3.5/5 stars), and 2,022 comments on the workouts (e.g., “Friendly stretching routine. I appreciate a leader who is not the quintessential body type”). About 96% of the sample attended ≥5 of the 6 coaching sessions. The PATH Fresher algorithm identified 1,781 new workouts that will be vetted for inclusion in the next update of the PATH website. Objective tracking via web analytics revealed sustained engagement throughout the study period with participants in the PATH group visiting the site 5,218 times where they spent an average of 7.5 mins per visit (workout videos range from 5–60 mins).

Fig 4.

Total visits on PATH website by the PATH group

Preliminary Efficacy of the PATH intervention

Using generalized linear mixed effects modeling significant between-group differences were detected for MVPA and daily steps from baseline to 12 weeks. Specifically, at 12 weeks, the PATH group had greater mean changes in objective weekly MVPA (58.9 vs. 0.9 min, p=.024) and daily steps (1,246.4 vs. −64.2 steps, p=.002) compared to the wait-list control group. All PATH effects for MVPA and daily steps were similar by race and sex. No significant between-group differences were detected for adherence to the PA Guidelines over time, but the proportion of the sample that adhered to the guidelines increased by 22% in the PATH group (p=.041), while the waitlist control group showed no improvement (p≥.05). The average daily step count during the initial 12-week period showed improvements in the PATH group vs. no change in control over time (Figure 5). After 12 weeks, the PATH intervention group exited the program, while the wait-list control group crossed over to use the PATH intervention. All but one participant crossed over and used PATH for 12 weeks. Their average daily steps improved consistently until the follow-up was complete at 24 weeks (Fig. 5)

Figure 5.

Average weekly step count over time by treatment arm

Although there were no significant between-group differences for change in CVD risk factors over time (p≥.05), the PATH group showed significant within-group improvements in weight (−3.2±1.5 lbs, p=.043), waist circumference (−1.0±0.4 inches, p=.010), and body fat (−0.9±0.3%, p=.004), and a trend for improvement in systolic BP (−4.2±2.2 mmHg, p=.057). No significant changes in CVD risk factors were observed in control group (all p≥.05). All PATH effects were similar by race and sex for CVD risk factors.

Focusing the logistics of remote sample collection for HbA1C and lipids at baseline and 12 weeks, 24.4% of our sample had invalid lab results mainly due to inadequate sample collection and dehydration. At 12 weeks, the proportion of the sample with invalid lab results declined to 13.4% after we addressed the problems through detailed instructions to participants. There were no major adverse events. Minor events (n=27) such as mild COVID-19 were similar in PATH (48%) vs. control (52%) groups (p=0.06).

Discussion

Findings from this pilot study demonstrate the feasibility of recruiting and retaining a diverse group of insufficiently active individuals with overweight/obesity in a study employing a web-based PA intervention, virtual coaching, and remote assessment of study outcomes. The acceptability data suggest that the PATH intervention is appealing, easy to navigate, includes culturally appropriate content, and promotes high engagement throughout the study period. The findings also suggest that the PATH intervention may have preliminary efficacy in promoting PA and mitigating CVD risk factors in a diverse sample of insufficiently active adults with overweight/obesity.

Previous studies have recognized the critical role that web-based interventions can play in promoting PA among populations at high risk for CVD and other chronic illnesses.²⁷ The ability of these interventions to reach large audiences at low cost and to provide on-demand access to a wide variety of resources that can be tailored to individual preferences is widely appreciated.⁵¹ Yet, web-based PA programs have struggled to reach those most in need of increasing PA⁵¹ and are usually plagued with recruitment challenges, non-usage attrition and high dropout rates (about 30% on average), especially among those with obesity.^52–54

Our findings suggest that an end-user centered intervention development and implementation process can mitigate some of the limitations and enhance feasibility and acceptability in the target population. Factors that promoted the appeal of our intervention included content that reflected end users’ preferences, scheduling PA and virtual coaching around each participant’s convenience, and remote physical and laboratory assessments. The remote assessments were very popular with our participants and the Zoom technology used by the study team to supervise the assessments was intuitive even to first time users.

Our strategy for remote intervention and assessment was not without challenges. Some of the participants struggled to install the three apps (Zoom, Fitbit, Centrepoint) that were needed to participate in the study. We had anticipated this problem and developed a strategy where we sent all participants detailed instructions on what to download followed by a 15-minute phone call to review what was downloaded and guide those who were not able to download the apps. If the 15 minutes phone call was not adequate to help a participant download all the apps, they were given an option to withdraw from the study or stay on the phone longer. All the participants opted to stay longer, with at least one case of an older adults who needed about 2 hours of phone guidance to download and log in the apps.

Once the participant downloaded and logged in the apps, they only needed to open them when using them (Zoom) and when synching the data (Fitbit and Centrepoint). The data were passively collected at captured to the PATH back end through API with Fitbit and Centrepoint. As depicted in Figure 3B, some users reported glitches that interfered with the synching of the devices with the apps, but the issues were resolved for all participants. We were able to promptly deal with these problems because we provided all participants with contact information for reporting (email, phone, text) any problems they encountered to the study team. We also regularly conducted data audits which enabled us to identify participants with devices that were not submitting any data and check in with them. These approaches could help future studies mitigate the problem of data loss with remote interventions.

The main challenge in our remote assessment strategy was the self-collection of blood samples for lipids and HbA1C. At baseline, almost a quarter of all the samples collected could not be processed mainly because the participants did not get an adequate sample due to dehydration or vasoconstriction. In subsequent assessments, we were able to reduce the number of invalid results by almost 50% through brief instructions emphasizing proper hydration and the need to clean hands with warm water before assessments.

Although most web-based PA interventions have not included insufficiently active adults with overweight/obesity,²⁷ several studies have succeeded in promoting PA in this high risk population.^52,55,56 The Active Living Every Day-Internet intervention (N=41) demonstrated increased step count (1,404 steps) compared to no significant change (339 steps) in the control group in a 16 week study.⁵³ Similarly, Morgan et al. (N=110) demonstrated a 35 MET-minutes between group difference in favor of the intervention group (p=0.03) in a 14 week study.⁵² These improvements were modest, and the studies were limited by small sample sizes and relatively high attrition rates (19% and 25%, respectively).^52,53 Our findings collaborate these studies, with higher retention and greater change in PA, and contributes to the growing body of evidence supporting the utility of web-based interventions in promoting PA among insufficiently active individuals with overweight/obesity.

In a larger RCT (N=441) with one year follow-up which employed web-based learning modules to promote PA among men with overweight/obesity,⁵⁵ participants (mean age 43.9±8.0) accessed the study website for less than half of the stipulated 52-week study period. Self-reported PA increased by 16 min/day in the intervention vs. control group, with an overall attrition rate of 30%.⁵⁵ Minorities and those with BMI ≥35.2 kg/m² were more likely to withdraw from the study. ⁵⁵Although the study was limited by relatively low engagement and high attrition rate, the impact on self-reported PA was significant. These data provide further evidence of the potential role that web-based interventions can play in promoting long-tem improvements in PA among high risk populations.

Our findings depict greater user engagement and improvements in objective PA. The objective web-analytics and self-reported data on PATH utilization indicate sustained intervention engagement throughout the study duration with low non-usage attrition and droupout rate (4%). Additionally, there were no racial/ethnic differences in our study outcomes suggesting that the intervention was helpful to a divese group of individuals with overweight/obesity. It is worth noting that our sample was smaller (n=82) and intervention duration shorter (12 weeks) compared to the study by Patrick et al. ⁵⁵ Therefore, we do not know how the PATH intervention engagement and attrition rate would look like in a larger study of longer duration.

These data suggest that strategies to increase engagement with web-based interventions can be effective in promoting retention and longterm increase in PA. This is important especially when targeting minoritized populations with low PA levels, who are less likely to enroll but more likely to drop out from PA studies.⁵⁷ The status quo for web-based PA programs entails providing online access to interactive learning modules and print-based resources that encourage behavior change, goal setting and self-monitoring.²⁷ Although these strategies have had modest success in promoting PA in the general population,^27,58 preliminary studies focusing on minoritized populations highlight the limitations of print based resources. In our study, the diverse workout videos were the most preferred component of the PATH intervention because they provided participants with vicarious experiences along the PA continuum.

Providing authentic vicarious experience as part of a PA intervention can be daunting, mainly due to the logistical challenges of modelling experiences along the PA continuum.²⁹ Several studies have attempted to overcome these barriers by sharing self-monitoring PA data among study participants for peer to peer comparison.⁵⁹ In our study, we were able to provide first-hand vicarious experiences by employing an innovative strategy that allowed us to leverage open access content created by a global community of PA experts. This approach gave us access to a wide variety of resources that we used to model PA for our diverse study participants. Our findings add to the growing body of evidence on the importance of vicarious experiences in promoting PA.

Lastly, our PATH intervention was designed to be used in any home environment with basic internet access, thus helping individuals increase PA even in environments where safety concerns and limited access to fitness amenities limit their ability to engage in traditional PA options.^57,60 This is important because the conditions in which people live, work, learn and play may facilitate or impede their efforts to be more physically active.^29,61,62 The PATH program shifts focus from expert generated print-based resources to user preferred combination of culturally salient workout videos and other types of PA, providing a resource efficient strategy for promoting PA.

The short study length and small sample size limit our ability to test effectiveness; however, as our study was meant to demonstrate feasibility and acceptability, in line with the ORBIT model,³¹ this is appropriate for the formative phase of behavioral intervention development. Another limitation was use of self-reported PA in our eligibility criteria which led to enrollment of individuals who were already adhering to the PA guidelines based on objective PA data. Despite this limitation, the intervention group was still able to increase their PA. Lastly, about 40% of our sample used the wrist-worn Actigraph GT9X for continuous monitoring of step count, while the other 60% used Fitbit Charge 2. This difference was necessitated by a defect in Actigraph GT9X wrist bands without a lasting solution. This change did not affect the primary outcome of MVPA that was measured using the waist-worn Actigraph GT3X.

In summary, once the efficacy trial is successfully completed, we plan to deploy PATH as a convenient, enjoyable, and scalable stand-alone program for promoting PA in high-risk populations such as individuals with overweight/obesity. In addition, PATH could be used to complement traditional fitness programs to foster regimen continuity especially when challenges like time constraints or inclement weather preclude physical attendance. Insights from this study may be used to inform the design of other PA interventions targeting other high risk and underserved communities, contributing to the broader goal of eliminating disparities and creating a society where most people are empowered to adhere to the PA Guidelines.

Conclusions

This pilot RCT demonstrated that the PATH intervention is feasible, acceptable, and preliminarily efficacious in promoting PA among insufficiently active adults with overweight/obesity. The promising intervention could offer a novel strategy for promoting PA in populations that are underserved and at high risk of cardiometabolic diseases. A larger scale RCT designed to test the PATH intervention at a longer duration is currently in progress.