Abstract
OBJECTIVE:
Lifestyle interventions are effective, but those delivered via in-person group meetings have poor scalability and reach. Research is needed to establish if remotely delivered lifestyle interventions are non-inferior to in-person delivered lifestyle interventions.
METHODS:
We conducted a randomized non-inferiority trial (N = 329) to compare a lifestyle intervention delivered remotely and asynchronously via an online social network (Get Social condition) to one delivered via in-person groups (Traditional condition). We hypothesized that the Get Social condition would result in a mean percent weight loss at 12 months that was not inferior to the Traditional condition. Additional outcomes included intervention delivery costs per pound lost and acceptability (e.g., convenience, support, modality preferences).
RESULTS:
At 12 months, no significant difference in percent weight change was observed between the Get Social and Traditional conditions (2.7% vs. 3.7%, p = 0.17) however, criteria for non-inferiority were not met. The Get Social condition costs $21.45 per pound lost versus $26.24 for the Traditional condition. A greater percentage of Get Social condition participants rated participation as convenient (65% vs 44%; p = 0.001).
CONCLUSIONS:
Results revealed a remotely-delivered asynchronous lifestyle intervention resulted in slightly less weight loss than an in-person version but may be more economical and convenient.
TRIAL REGISTRATION:
ClinicalTrials.gov NCT02646618; https://clinicaltrials.gov/ct2/show/NCT02646618.
INTRODUCTION
Lifestyle interventions are effective [1, 2], but the traditional model of in-person group meetings is inconvenient for many people [3]. Remote lifestyle interventions are an alternative, with delivery via mobile app, videoconferencing, telephone, text messaging, or some combination [4]. Two systematic reviews revealed that remote lifestyle interventions are effective [5], with those that include human counseling among the most effective [4].
Remote interventions can be delivered synchronously via videoconference or telephone at scheduled times, or asynchronously, via text messaging or in discussion threads on online platforms. A non-inferiority randomized trial compared a synchronous remote lifestyle intervention that involved telephone counseling, a website, and email to a synchronous in-person condition that involved individual and group visits [6]. Results revealed no differences in weight loss between groups at 6 months. Additional evidence for synchronous remote lifestyle interventions comes from a trial in which investigators were forced to shift from in-person delivery to videoconference due to the pandemic and found no weight loss differences within a non-inferiority margin of 2.5% [7]. Further, a systematic review of behavioral interventions revealed no significant differences between videoconference and in-person delivery [8]. Though evidence to date suggests that remote synchronous interventions are non-inferior to in-person synchronous interventions, remote synchronous interventions share a drawback with in-person interventions in that participants must be able to attend meetings for 6–12 months which is not feasible for some people. Little is known about how asynchronous remote interventions compare to synchronous in-person interventions.
Asynchronous lifestyle interventions entail asynchronous exchanges between counselors and a group of participants. Many also provide content for participants to view at their convenience and/or mobile apps to assist participants in self-monitoring. Asynchronous interventions do not require participants to be available to attend meetings and participants can engage in smaller time increments as convenient. A randomized trial compared an asynchronous diabetes prevention program to a control group and observed weight loss that was comparable to what has been reported in trials of synchronous interventions delivered remotely or in person [9]; however head-to-head comparisons are lacking.
Social media platforms have been used to deliver asynchronous lifestyle interventions because many people already use these platforms [10]. About 72% of US adults use social media, with no differences by sex or race/ethnicity [11], indicating that these platforms have a high potential for reach. Commercial social media platforms are also free to use, which reduces the costs of intervention delivery. Some social media platforms allow for the creation of private groups where asynchronous interactions can occur. Systematic reviews of social media-delivered lifestyle interventions reveal they produce modest weight loss [12–17]. Despite the convenience of remote, asynchronous lifestyle interventions, we know little about how they compare to synchronous lifestyle interventions.
The COVID-19 pandemic increased the demand for telehealth, but the vast majority of telehealth occurs synchronously [18]. Given the convenience of asynchronous care, studies are needed to establish evidence. In a randomized non-inferiority trial, we compared an asynchronous lifestyle intervention delivered via a social media platform (Get Social condition) to a synchronous in-person lifestyle intervention (Traditional condition). We hypothesized that while the Traditional condition is likely to outperform the Get Social condition at 6 months, and that by 12 months, weight loss in the Get Social condition would be not appreciably worse than the Traditional condition. We also compared conditions on cost and acceptability. We hypothesized that the Get Social condition would cost less to deliver per pound lost than the Traditional Condition and that Get Social participants would report higher acceptability (e.g., convenience) than Traditional condition participants.
METHODS
Study design
We conducted a non-inferiority randomized controlled trial in adults with overweight and obesity. This study was approved by the Committee for the Protection of Human Subjects in Research at UMass Chan Medical School and the UConn Storrs Institutional Review Board. The study period was from August 23 2017 to July 31 2020.
Study population
We randomized 329 participants to two conditions. Study procedures have been published elsewhere [19]. Inclusion criteria included: body mass index (BMI) between 27.0–45.0 kg/m2, ages 18–65, written clearance from a primary care provider, smartphone ownership, and a social media account with log ins at least 4 days/week on average. People were excluded if they were pregnant, lactating, were taking medication that affects weight, were participating in a formal weight loss program, lost ≥ 5% of their weight in the previous 3 months, had bariatric surgery or planned to have it during the study period, had a medical condition that precludes dietary or physical activity changes, had type 1 diabetes or uncontrolled type 2 diabetes, smoked ≥ 3 cigarettes per day, and/or expressed a preference for one condition.
Participant recruitment
Study participants were recruited in 9 waves of 36–41 participants. In 2017, the study team relocated from the UMass Chan Medical School in Worcester, Massachusetts to the University of Connecticut in Storrs, Connecticut. As such, the first 6 waves were recruited from Central Massachusetts and the last 3 waves were recruited from Central Connecticut. Participants were recruited using Craigslist, Google ads, postings in local Facebook groups, university listserves, newspapers, and flyers in the community. Some recruitment advertisements targeted males or racial/ethnic minorities because previous weight loss trials have underrepresented these groups [20]. Interested participants completed a screening call and those eligible were scheduled for a baseline visit, which included informed consent, physical measurements, and a survey. Participants were provided $30 for completing baseline.
Pre-randomization informational webinar
Prior to randomization, participants attended a webinar that explained the trial, research questions, and the Diabetes Prevention Program (DPP) lifestyle intervention [21].
Randomization
The study statistician randomized participants 1:1 to two conditions in randomly permuted blocks of sizes 4 and 6 using the ralloc program in Stata (Stata Corp). Randomization was stratified by gender and BMI (27.0–34.9 kg/m2 vs. 35.0–45.0 kg/m2).
Study conditions
Participants in both study conditions received a 12-month lifestyle intervention based on the DPP [22].
Get Social condition.
Get Social participants received 12 months of lifestyle intervention via a private Twitter group facilitated by a dietitian. The dietitian posted intervention content twice daily in months 1–2, then once daily in months 3–6, and three times weekly in months 7–12. The frequency and spacing of intervention posts corresponded to the frequency and spacing of in-person visits in the Traditional condition.
Intervention content:
Each of the 22 DPP modules was distilled into 14 tweets designed to meet all learning objectives [10, 23]. Tweets were of the following types: text only, text and an image of an excerpt of the original materials, text with other images, text and a link to an online resource that elaborates on the topic, and polls. The dietitian logged in at least twice daily to field responses to conversation threads, engage participants in problem solving, and tweet participants who hadn’t engaged recently. Every Monday morning participants were asked to share their diet and exercise goals for the week. Every Friday morning participants were asked to report their weight change for the week which approximated the “weigh-ins” that occur at the beginning of each visit in the Traditional condition.
Orientation:
Get Social condition participants attended an individual and a group orientation visit prior to starting the intervention. At the individual orientation, participants learned how to use MyFitnessPal™ and Twitter. At the group orientation, participants met the dietitian and the private group was created by having everyone follow each other with the “protected” privacy setting on their accounts so that all interactions were only viewable to each other.
Traditional condition.
Participants randomized to the Traditional condition received 12 months of lifestyle counseling based on the Diabetes Prevention Program Lifestyle Intervention via 22 group meetings lasting 90 min per meeting [22]. The group met weekly in months 1–2, biweekly in months 3–6, and then monthly in months 7–12.
Orientation:
Participants in the Traditional condition attended a group orientation visit. At this visit, participants met the dietitian and learned about the program and how to use MyFitnessPal™ for calorie tracking.
Baseline and follow-up assessments
At baseline, 6 and 12 months, participants completed a study visit including physical measurements taken by a blinded assessor and were sent a link to a survey. Participants were compensated for completing these assessments.
Measures
Weight.
Weight was taken using a calibrated scale with the participant wearing light clothing and no shoes. If participants were unable to attend follow-up visits, they were contacted to self-report their current weight. This occurred for 23 (7.0%) participants at 6 months and 42 (12.8%) participants at 12 months. Pounds lost and percent weight loss from baseline was calculated at 6 and 12 months.
Cost.
For each condition, we tracked time and costs associated with intervention delivery in practice [24–26]. Costs included printing, software, administrative tasks, interventionist tasks, and travel time as described elsewhere [19].
Acceptability.
Acceptability was operationalized as convenience, modality preferences, and group support. For convenience, participants rated on a 5-point Likert scale how convenient they thought participation in their group was (dichotomized as very convenient/convenient and neutral/inconvenient/very inconvenient), and then they rated on a 5-point Likert scale the extent to which they agreed that participation was time-consuming. Responses were dichotomized as strongly agree/agree and neither agree or disagree/disagree/strongly disagree. Then they were asked which modality they would find more convenient. We calculated the percentage of participants selecting online as more/much more convenient and the percentage of participants selecting in-person as more/much more convenient.
To measure support, participants rated on a 5-point Likert scale the extent to which they agreed with 4 statements regarding how supportive the group was. Responses were dichotomized as strongly agree/agree and neither agree or disagree/disagree/strongly disagree. Finally, to measure modality preferences, participants rated on a 5-point Likert scale the extent to which they would have preferred both an online group and in-person group meetings (dichotomized as extremely/very/moderately/slightly and not at all), and the extent to which they would have preferred the modality they did not get (dichotomized as extremely/very/moderately/slightly and not at all).
Sample size estimation
We powered the study to test non-inferiority between conditions on the primary outcome: percent weight change at 12 months [27]. Denote μGS as the mean of percent weight change at 12 months for get social condition, and μT as that for the traditional condition, δ as the inferiority margin, our main hypothesis can be stated as H0 : μGS — μT ≥ δ vs. Ha : μGS — μT < δ. We set δ = 2%, based on a clinically meaningful difference in average weight change between the two conditions. We estimated a within condition standard deviation (SD) of 5.5% for our primary outcome based on a previous randomized trial [28]. With significance level (α) = 0.05 and δ = 2%, we calculated that we have 90% power to conclude that the Get Social condition is not inferior to the traditional condition with 131 participants per arm. Accounting for 20% attrition, we enrolled 328 participants total (164 per arm). With N = 131 available per arm (N = 262 total) and α = 0.05, we had 80% power to detect differences in mean cost per participant of 0.35 SDs. For example, if the SD for cost is $100, then we have 80% power to detect differences in mean cost per participant of $35. Power calculations for were conducted using PROC POWER in SAS 9.3 (SAS Institute, Cary, NC) [29, 30].
Analytic plan
Reporting and analyses were in compliance with the Reporting of Noninferiority and Equivalence Randomized Trials Extension of the CONSORT 2010 Statement [31]. Analyses for the primary outcome were intent-to-treat. Linear mixed modeling was employed to examine time by condition interactions at 6- and 12-months using SAS PROC MIXED [32]. Multiple imputations were used to impute missing values of weight at 6 and 12 months, with baseline weight, BMI, and other characteristics (see Table 1) used as covariates in the multiple imputation process [33]. We also performed an exploratory per protocol analysis including only participants who provided weight data and attended at least 1 session (Traditional condition) or engaged at least one time (Get Social condition). For acceptability measures, only participants who completed the survey were included. The distributional properties of all variables were examined to make sure statistical assumptions were met before analysis (e.g., the variances of the weight outcomes for the two conditions were statistically compared before subsequent analyses).
Table 1.
Baseline characteristics of randomized participants (n = 329), total and by treatment condition.
| Get Social n = 167 |
Traditional n = 162 |
Total n = 329 |
p-value | |
|---|---|---|---|---|
| n (%) | n (%) | n (%) | ||
| Age – Mean (SD) | 45.3 (11.2) | 45.5 (11.6) | 45.4 (11.4) | 0.87 |
| BMI – Mean (SD) | 34.9 (4.7) | 35.1 (4.4) | 35.0 (4.6) | 0.68 |
| Female | 136 (81.4%) | 132 (81.5%) | 268 (81.5%) | 0.99 |
| Hispanic/Latino ethnicity | 8 (4.8%) | 8 (4.9%) | 16 (4.9%) | 0.61 |
| Race | ||||
| Caucasian; non-Hispanic White | 152 (91.0%) | 150 (92.6%) | 302 (91.8%) | |
| Non-White | 15 (9.0%) | 12 (7.4%) | 27 (8.2%) | 0.60 |
| Marital status | ||||
| Married | 114 (68.3%) | 96 (59.3%) | 219 (63.8%) | |
| Living with a partner to whom I’m not married | 11 (6.6%) | 20 (12.4%) | 31 (9.4%) | |
| Single | 26 (15.6%) | 26 (16.1%) | 52 (15.8%) | 0.21 |
| Widowed/Divorced/Separated | 16 (9.6%) | 20 (12.4%) | 36 (10.9%) | |
| Education | ||||
| High school degree/G.E.D./equivalent | 6 (3.6%) | 8 (4.9%) | 14 (4.3%) | |
| Trade school/Some College/Associates Degree | 53 (31.7%) | 56 (34.6%) | 109 (33.1%) | |
| Bachelor’s degree/ Some graduate school | 47 (28.1%) | 50 (30.9%) | 97 (29.5%) | 0.58 |
| Master’s degree/Doctoral degree/Professional degree | 61 (36.5%) | 48 (29.6%) | 109 (33.1%) | |
| Employment Status | ||||
| Employed full-time | 124 (74.3%) | 126 (77.8%) | 250 (76.0%) | |
| Employed part-time | 28 (16.8%) | 19 (11.7%) | 47 (14.3%) | 0.41 |
| Other employment status | 15 (9.0%) | 17 (10.5%) | 32 (9.7%) |
Weight change.
We modeled weight loss and percent weight loss at 6 and 12 months using a linear regression model, with weight loss or percent weight loss as the dependent variable and study condition as the independent variable. This analytic approach aimed to test whether the Get Social condition is not appreciably worse than the Traditional condition by our a priori inferiority margin of 2%. A 90% two-sided CI for the mean difference in percent weight change between the two conditions at 12 months (μGS – μT ) was calculated (equivalent to 95% one-sided CI) [33]. If the upper bound of the CI is smaller than the pre-specified non-inferiority margin of 2%, non-inferiority would be established.
Other outcomes.
We compared acceptability items across treatment conditions using Chi-squared tests using data from participants who completed the 12-month follow-up survey. We reported total intervention costs, costs per participant, and costs per pound lost by treatment condition.
RESULTS
Baseline characteristics
Participants (N = 329) with overweight or obesity (mean BMI = 35.0, SD = 4.6) were randomized to the Get Social (n = 167) or Traditional (n = 162) condition. Participants were on average 45.4 years old (SD = 11.4) and were predominantly white (91.8%) and female (81.5%). We observed no baseline differences between treatment conditions in demographics or weight (Table 1).
Intervention participation
Traditional condition participants attended a median of 14 out of 22 sessions (IQR: 6–18; range: 0–22). Total engagement in the Get Social condition was defined as the sum of original tweets, replies, retweets, and likes. Get Social condition participants made a median of 227 total engagements (IQR: 87–567, range 0–4169). Specifically, they made a median of 16 original tweets (IQR: 7–34, range: 0–171), a median of 59 replies to conversation threads (IQR: 23–139, range: 0–749), a median of 0 retweets (IQR: 0–0, range: 0–171) and a median of 127 likes (IQR: 37–395, range: 0–3969). In the Traditional condition, 12% of participants did not attend a single session, whereas in the Get Social condition, 2% did not engage at all. In the Traditional condition 48% attended the last visit, and in the Get Social condition 53% engaged in the last module.
Outcomes
Our primary outcome is percent weight loss at 12 months. At 12 months, the Get Social condition had a mean weight change of −2.7% (SD = 7.5) or −6.1 pounds (SD = 17.1) and the Traditional condition had a mean weight change of −3.7% (SD = 6.9) or −7.9 pounds (SD = 14.8). The difference between conditions in percent weight change was not statistically significant (90% CI: −0.2%, 2.4%; P = 0.086 for superiority test), thus the Traditional condition was not superior to the Get Social condition. However, the upper bound of the CI was larger than the non-inferiority margin of 2%, thus non-inferiority was not established (P = 0.125). At 6 months, participants in the Get Social condition had a mean weight change of −2.6% (SD = 5.8) or −5.8 pounds (SD = 13.0) while those in the Traditional condition had a mean weight change of −4.5% (SD = 5.8) or −9.4 pounds (SE = 12.4; Table 2). At 6 months, the Traditional condition lost significantly more weight than the Get Social condition for absolute weight change [90% CI: 1.3, 5.9; P = 0.004] and percent weight change [90% CI: 0.8%, 2.9%; P = 0.011]. In terms of the percent of participants losing clinically significant weight (≥5%), by 12 months, no statistically significant difference between conditions was observed (P = 0.215), such that 31.7% of Get Social participants and 38.3% of Traditional participants achieved clinically significant weight loss. At 6 months, 26.9% of Get Social participants achieved this compared to 39.5% of Traditional participants, a difference that was statistically significant (P = 0.015). (Table 2).
Table 2.
Change in weight at 6 and 12 months by treatment condition*.
| Change in weight (%) | Time point | Get Social n = 167 |
Traditional n = 162 |
Difference Get Social vs Traditional (two-sided 90% CI) | p-value (two sided) |
|---|---|---|---|---|---|
| M (SD) | M (SD) | ||||
| 12 months | −2.7 (7.5) | −3.7 (6.9) | 1.0 (−0.2, 2.4) | 0.171 | |
| 6 months | −2.6 (5.8) | −4.5 (5.8) | 1.9 (0.8, 2.9) | 0.004 | |
| Change in weight (lbs) | M (SD) | M (SD) | |||
| 12 months | −6.1 (17.1) | −7.9 (14.8) | 1.8 (−1.1, 4.7) | 0.307 | |
| 6 months | −5.8 (13.0) | −9.4 (12.4) | 3.6 (1.3, 5.9) | 0.011 | |
| % losing ≥ 5% | % | % | |||
| 12 months | 31.7 | 38.3 | −6.6 (−15.2, 2.1) | 0.215 | |
| 6 months | 26.9 | 39.5 | −12.6 (−21.1, −4.0) | 0.015 |
Variances for each outcome were tested and deemed equal across the two conditions.
The per-protocol analysis for percent weight loss included 289 participants (88% of randomized sample; Table 3). At 12 months, the Get Social condition had a mean weight change of −2.7% (SD = 7.7) or −6.3 pounds (SD = 17.5) and the Traditional condition had a mean weight change of −4.3% (SD = 7.0) or −9.1 pounds (SD = 15.0). The 90% two-sided confidence interval for percent weight loss was greater than 0 which establishes the superiority of the Traditional condition (90% CI: 0.2%, 3.0%; P = 0.065). At 6 months, participants in the Get Social condition had a mean weight change of −2.6% (SD = 6.0) or −5.6 pounds (SD = 13.3) while those in the Traditional condition had a mean weight change of −5.0% (SD = 5.8) or −10.4 pounds (SD = 12.3). Superiority of the Traditional condition was established [90% CI: 1.3%, 3.6%; P < 0.001].
Table 3.
Per-protocol analysis for weight outcomes.
| Change in weight (%) | Time point | Get Social 6 months: n = 151 12 months: n = 156 |
Traditional* 6 months: n = 135 12 months: n = 133 |
Difference get social vs traditional (two-sided 90% CI) | p-value (two-sided) |
|---|---|---|---|---|---|
| M (SD) | M (SD) | ||||
| 12 months | −2.7 (7.7) | −4.3 (7.0) | 1.6 (0.2, 3.0) | 0.065 | |
| 6 months | −2.6 (6.0) | −5.0 (5.8) | 2.4 (1.3, 3.6) | <0.001 | |
| Change in weight (lbs) | M (SD) | M (SD) | |||
| 12 months | −6.3 (17.5) | −9.1 (15.0) | 2.8 (−0.4, 6.0) | 0.148 | |
| 6 months | −5.6 (13.3) | −10.4 (12.3) | 4.7 (2.2, 7.2) | 0.002 | |
| % losing ≥ 5% | % | % | |||
| 12 months | 33.3 | 40.6 | −7.3 (−16.7, 2.1) | 0.203 | |
| 6 months | 27.8 | 43.7 | −15.9 (−25.1, −6.6) | 0.005 |
Two participants who became pregnant during the intervention were censored.
The total cost of the Traditional condition was $34,280.58 and for the Get Social condition, it was $22,566.48, for a difference of $11,714.10 (Table 4). Cost per participant in the Traditional condition was $217.78 compared to $135.12 in the Get Social condition, for a difference of $82.66 per participant. The Traditional condition costs $26.24 per pound lost and the Get Social condition costs $21.45 per pound lost.
Table 4.
Intervention costs by treatment condition.
| Get social | Traditional | |||
|---|---|---|---|---|
| Total cost | Cost per person | Total cost | Cost per person | |
| Interventionist travel | $1738.26* | $10.41 | $5987.34 | $36.96 |
| Software and materials | ||||
| Hootsuite | $484.32 | $2.90 | na | na |
| BeePro | $420.00 | $2.52 | na | na |
| Printing materials | $25.05 | $0.15 | $1715.32 | $10.59 |
| Contest prizes | $90.00 | $0.53 | $64.31 | $0.40 |
| Personnel costs | ||||
| Administrative salary (scheduling posts, printing, emailing newsletters) | $4337.68 | $25.97 | $502.38 | $3.10 |
| Interventionist salary at $42.92 /hour | $14,750.17 | $88.32 | $26,201.23 | $161.74 |
| Total personnel costs | $19,087.85 | $114.30 | $26,793.61 | $165.40 |
| Total costs | $22,566.48 | $135.12 | $35,280.58 | $217.78 |
Travel in the Get Social condition occurred for orientation visits only.
A total of 237 participants (72.0%) completed the 12-month follow-up survey. Convenience. A greater percentage of Get Social participants rated participation as convenient (65% vs 44%; Χ21,237 = 10.54; P = 0.001). A greater percentage of Traditional participants rated participation as time-consuming (42% vs 28%; Χ21,237 = 4.79; P = 0.029; Table 5), and most Get Social participants (73%) and over half of Traditional participants (56%) rated the online approach as more convenient than in-person groups (Χ21,237 = 7.32; P = 0.001), and 16% of Get Social and 26% of Traditional participants rated in-person groups as more convenient than online (Χ21,237 = 3.47; P = 0.063).
Table 5.
Acceptability in the traditional and get social conditions.
| Get Social (n = 113), n (%) | Traditional (n = 124), n (%) | Χ2 (df) | P value | |
|---|---|---|---|---|
| Convenience | ||||
| Participation was convenient/very convenient.a | 73 (65) | 54 (44) | 10.54 (1) | 0.001 |
| Participation was time consuming.a | 32 (28) | 52 (42) | 4.79 (1) | 0.029 |
| Online groups more convenient than in-person groups. | 82 (73) | 69 (56) | 7.32 (1) | 0.007 |
| In-person groups more convenient than online groups. | 18 (16) | 32 (26) | 3.47 (1) | 0.063 |
| Support | ||||
| I did not feel connected to the group.a | 51 (45) | 29 (23) | 12.50 (1) | <0.001 |
| The questions I asked were answered.a | 82 (73) | 100 (81) | 2.17 (1) | 0.141 |
| I did not feel comfortable speaking up in the group.a | 21 (19) | 11 (9) | 4.78 (1) | 0.029 |
| I received support from other group members when I needed it.a | 60 (53) | 78 (63) | 2.34 (1) | 0.126 |
| Modality preferences | ||||
| Would have preferred hybrid (online and in-person). | 69 (61) | 74 (60) | 0.05 (1) | 0.828 |
| Would have preferred the other modality. | 56 (50) | 54 (44) | 0.86 (1) | 0.354 |
Proportion of participants responding with strongly agree or agree versus strongly disagree, disagree, or neutral.
Support.
A greater percentage of Get Social participants said they did not feel connected to the group (45% vs 23%; Χ21,237 = 12.50; P < 0.001) and did not feel comfortable speaking up in the group (19% vs 9%; Χ21,237 = 4.78; P = 0.029). No differences were observed between conditions in the percentage who said they received support from other group members when they needed it (53% in Get Social vs 63% in Traditional, Χ21,237 = 2.34; P = 0.126) and the percentage who said the questions they asked were nswered (73% vs 81%, Χ21,237 = 2.17; P = 0.141).
Modality preferences.
No differences were observed between conditions in the percentage who would have preferred a hybrid approach (both online and in-person groups; 61% vs 60%, Χ21,237 = 0.5; P = 0.828) or in the percentage who would have preferred the modality they did not receive (50% vs 44%, Χ21, 237 = 0.86; P = 0.354).
DISCUSSION
Findings revealed that participants in an asynchronous remote lifestyle intervention lost significantly less weight at 6 months compared to a synchronous in-person lifestyle intervention but differences at 12 months were not statistically significant. However, non-inferiority was not established at 12 months because the upper bound of the CI was not smaller than the pre-specified non-inferiority margin of 2%. This means that at 12 months, the Traditional condition was not superior to the Get Social condition in weight loss, but the Get Social condition did not meet the threshold of non-inferiority either. The weight loss difference between conditions at 12 months was 1.6 pounds which is of debatable clinical significance. Generally, findings reveal that the Traditional condition produces more weight loss initially, the advantage disappears by 1 year, and the Traditional approach came at a slightly higher cost per pound and most participants in both conditions viewed it as less convenient. To be sure, the per protocol analyses revealed a weight loss advantage of the Traditional condition compared to the Get Social condition of 4.7 lbs and 2.8 lbs at 6 and 12 months, respectively.
The greater expense of the Traditional condition was due to both travel and interventionist time. Interventionist time was higher partly because regardless of how many participants show up to a group meeting, the meeting is the same duration; whereas in the Get Social condition, interventionist time is very much affected by how many participants engage each week. Although the Traditional condition cost more, some aspects of the Get Social condition had unique costs such as the staff time required to schedule posts and scheduling software. Many commercial platforms now allow users to schedule posts in advance for free which means future trials will not have software costs as we did.
The use of a commercial social media platform has an economical advantage over remotely-delivered programs that use platforms that have to be built and maintained. One report showed that the cost to develop a health app is approximately $425 K [34]. Maintenance costs are also required for mobile apps to be sustainable. The use of commercial social media platforms that allow for private group discussions is economical and convenient for both program developers and patients. Other advantages of commercial social media platforms include high usability characteristics, high use rates, and engaging platform features (e.g., multimedia posts). However, use of commercial social media platforms comes with serious considerations. First, commercial platforms have access to content posted on their platforms which means protected health information should not be shared. Second, social media companies employ proprietary algorithms that determine the content users see which could affect treatment receipt, and some platforms do not provide data on post views which disallows measurement of treatment receipt. Third, social media companies’ main source of income is targeted advertisements, thus users will be exposed to ads that are outside of the control of investigators. Participants should be informed of these considerations.
Thus far, the Centers for Medicaid and Medicare Services provide reimbursement for synchronous but not asynchronous behavioral health [35]. Clinical trials are needed to build evidence for asynchronous behavioral care and the present study con-tributes to this growing literature. One randomized controlled trial comparing synchronous versus asynchronous telepsychiatry revealed clinically and statistically significant improvement in clinical outcomes in both groups with no differences between groups [35]. Although the literature on asynchronous behavioral health is still in its infancy, asynchronous distance learning has robust evidence in the education literature. A 2004 systematic review revealed that asynchronous distance learning is as or more effective than synchronous distance learning [36]. Behavioral health researchers might benefit from insights provided in the education literature about effective approaches to distance learning.
Although the majority of Get Social participants felt their condition was more convenient than in person groups, they were more likely to say they did not feel connected to the group and that they did not feel comfortable speaking up in the group relative to the in-person condition. Asynchronous online groups which are characterized by brief non-face-to-face interactions may take longer to generate group cohesion. That said, the two conditions did not differ in the proportion of participants who said they received support from other group members when they needed it. Future research should explore ways to facilitate group cohesion in asynchronous remote interventions.
At the end of the program, nearly half of the participants in each condition said they would prefer the modality they didn’t receive. This could be because people had more experience with the flaws of the option they received or because they preferred that modality in the first place. We attempted to screen out people who expressed a strong preference for one modality, but some may still have had a preference or developed a preference once they experienced their condition. Interestingly, the primary reason for ineligibility, accounting for 65.3% of exclusions (Fig. 1), was the inability to participate in or lack of interest in one condition, especially the in-person condition. Specifically, 32.3% of exclusions were due to lack of availability to attend in-person groups, 20% were due to low social media engagement, 8% were due to lack of interest with no reason stated, and 5% were due to a strong preference for one condition with no reason stated. People vary in their preferences and ability to participate in different intervention modalities, likely due to each participant’s unique barriers to participation. Given the number of people who could not participate in this trial due to inability to attend in-person sessions, more online interventions (both synchronous and asynchronous) are needed to accommodate a wide range of lifestyles.
Fig. 1. CONSORT diagram.
Screening, baseline, randomization and follow-up of study participants.
The present trial has limitations that should be considered. Chiefly, the sample was not representative of the general population by being largely white and female. This is somewhat reflective of the communities we recruited from but also the tendency for white women to volunteer for weight loss studies in larger proportions than other groups [20], which may be in part due to societal pressures to be thin [37]. We targeted recruitment ads towards men and individuals from racial and ethnic groups underrepresented in research, though this did not produce a representative sample. We also excluded people who are not regular users of social media. However, 72% of US adults are social media users [11]. Another limitation is that we had no comparable metric of program engagement/attendance by which to compare conditions. Finally, the magnitude of weight loss observed was less than that observed in the original DPP lifestyle intervention (i.e.,14 lbs at 6 months and 10 lbs at 12 months) [2]. The difference may be a result of the original DPP being delivered individually and it included meal replacements, exercise classes/equipment, exercise trainers, food vouchers, and competitions [22]. The difference may also be due to the original DPP having a 3-week run-in period where potential participants had to demonstrate high diet tracking adherence to qualify for the trial [38]. Weight loss trials with run-in periods have been shown to produce 5 pounds more weight loss than others [39], which is consistent with our findings relative to the DPP.
In conclusion, although we did not establish that a counselor-led remote asynchronous lifestyle intervention using a commercial social media platform was non-inferior to an in-person intervention in weight loss at 1 year, the superiority of in-person intervention was also not established, and the remote asynchronous lifestyle intervention cost less money per pound lost to deliver. Future research should examine how best to implement asynchronous remote lifestyle interventions in real-world settings.
FUNDING
National Institute for Diabetes, Digestive, and Kidney Diseases, Grant/Award Number: R01DK103944. National Heart Lung, and Blood Institute, Grant/Award Number: K24HL124366.
Footnotes
COMPETING INTERESTS
The author declares no competing interests.
ETHICS APPROVAL AND CONSENT TO PARTICIPATE
All methods were performed in accordance with best practices for non-inferiority trials CONSORT guidelines, and the UMass Medical School (H00009012) and University of Connecticut (H17–210) Human Studies Review Boards. Informed consent was obtained from all participants.
DATA AVAILABILITY
The measures, data, and code for the present study can be obtained by contacting the first author.
REFERENCES
- 1.Ratner R, Goldberg R, Haffner S, Santica M, Orchard T, Fowler S, et al. Impact of intensive lifestyle and metformin therapy on cardiovascular disease risk factors in the diabetes prevention program. Diabetes Care. 2005;28:888–94. 10.2337/diacare.28.4.888 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Wing RR, Hamman RF, Bray GA, Delahanty L, Edelstein SL, Hill JO, et al. Achieving weight and activity goals among diabetes prevention program lifestyle participants. Obes Res 2004;12:1426–34. 10.1038/oby.2004.179 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Pagoto S The current state of lifestyle intervention implementation research: where do we go next? Transl Behav Med 2011;1:401–5. 10.1007/s13142-011-0071-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Joiner KL, Nam S, Whittemore R. Lifestyle interventions based on the diabetes prevention program delivered via eHealth: a systematic review and meta-analysis. Prev Med 2017;100:194–207. 10.1016/j.ypmed.2017.04.033 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Villegas V, Shah A, Manson JE, Tobias DK. Prevention of type 2 diabetes through remotely administered lifestyle programs: a systematic review. Contemp Clin Trials 2022;119:106817. 10.1016/j.cct.2022.106817 [DOI] [PubMed] [Google Scholar]
- 6.Appel LJ, Clark JM, Yeh HC, Wang N, Coughlin JW, Daumit G, et al. Comparative effectiveness of weight-loss interventions in clinical practice. N. Engl J Med 2011;365:1959–68. 10.1056/NEJMoa1108660 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Ross KM, Carpenter CA, Arroyo KM, Shankar MN, Yi F, Qiu P, et al. Impact of transition from face‐to‐face to telehealth on behavioral obesity treatment during the COVID‐19 pandemic. Obesity. 2022;30:858–63. 10.1002/oby.23383 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Byaruhanga J, Atorkey P, McLaughlin M, Brown A, Byrnes E, Paul C, et al. Effectiveness of individual real-time video counseling on smoking, nutrition, alcohol, physical activity, and obesity health risks: systematic review. J Med Internet Res 2020;22:e18621. 10.2196/18621 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Katula JA, Dressler EV, Kittel CA, Harvin LN, Almeida FA, Wilson KE, et al. Effects of a digital diabetes prevention program: an RCT. Am J Prev Med 2022;62:567–77. 10.1016/j.amepre.2021.10.023 [DOI] [PubMed] [Google Scholar]
- 10.Pagoto S, Waring ME, May CN, Ding EY, Kunz WH, Hayes R, et al. Adapting behavioral interventions for social media delivery. J Med Internet Res 2016;18:e24 10.2196/jmir.5086 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Social Media Fact Sheet. Pew Research Center Published April 7, 2021. Accessed June 28, 2023. https://www.pewresearch.org/internet/fact-sheet/social-media/#who-uses-each-social-media-platform
- 12.Chang T, Chopra V, Zhang C, Woolford SJ. The role of social media in online weight management: systematic review. J Med Internet Res 2013;15:e262. 10.2196/jmir.2852 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Maher CA, Lewis LK, Ferrar K, Marshall S, De Bourdeaudhuij I, Vandelanotte C. Are health behavior change interventions that use online social networks effective? A systematic review. J Med Internet Res 2014;16:e40. 10.2196/jmir.2952 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Laranjo L, Arguel A, Neves AL, Gallagher AM, Kaplan R, Mortimer N, et al. The influence of social networking sites on health behavior change: a systematic review and meta-analysis. J Am Med Inform Assoc 2015;22:243–56. 10.1136/amiajnl-2014-002841 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Ashrafian H, Toma T, Harling L, Kerr K, Athanasiou T, Darzi A. Social networking strategies that aim to reduce obesity have achieved significant although modest results. Health Aff 2014;33:1641–7. 10.1377/hlthaff.2014.0370 [DOI] [PubMed] [Google Scholar]
- 16.Willis EA, Szabo-Reed AN, Ptomey LT, Steger FL, Honas JJ, Washburn RA, et al. Do weight management interventions delivered by online social networks effectively improve body weight, body composition, and chronic disease risk factors? A systematic review. J Telemed Telecare. 2017;23:263–72. 10.1177/1357633X16630846 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.An R, Ji M, Zhang S. Effectiveness of social media-based interventions on weight-related behaviors and body weight status: review and meta-analysis. Am J Health Behav 2017;41:670–82. 10.5993/AJHB.41.6.1 [DOI] [PubMed] [Google Scholar]
- 18.Shaver J The state of telehealth before and after the COVID-19 pandemic. Prim Care. 2022;49:517–30. 10.1016/j.pop.2022.04.002 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Wang ML, Waring ME, Jake-Schoffman DE, Oleski JL, Michaels Z, Goetz JM, et al. Clinic versus online social network–delivered lifestyle interventions: protocol for the get social noninferiority randomized controlled trial. JMIR Res Protoc 2017;6:e243. 10.2196/resprot.8068 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Pagoto SL, Schneider KL, Oleski JL, Luciani JM, Bodenlos JS, Whited MC. Male inclusion in randomized controlled trials of lifestyle weight loss interventions. Obesity. 2012;20:1234–9. 10.1038/oby.2011.140 [DOI] [PubMed] [Google Scholar]
- 21.Jake-Schoffman DE, Brown SD, Baiocchi M, Bibeau JL, Daubenmier J, Ferrara A, et al. Methods-motivational interviewing approach for enhanced retention and attendance. Am J Prev Med 2021;61:606–17. 10.1016/j.amepre.2021.04.005 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.The Diabetes Prevention Program Research Group. The Diabetes Prevention Program (DPP): description of lifestyle intervention. Diabetes Care. 2002;25:2165–71. 10.2337/diacare.25.12.2165 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Pagoto SL, Waring ME, Schneider KL, Oleski JL, Olendzki E, Hayes RB, et al. Twitter-delivered behavioral weight-loss interventions: a pilot series. JMIR Res Protoc 2015;4:e123. 10.2196/resprot.4864 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Hernan WH, Brandle M, Zhang P, Williamson D, Matulik MJ, Ratner RE, et al. Costs associated with the primary prevention of type 2 diabetes mellitus in the diabetes prevention program. Diabetes Care. 2003;26:36–47. 10.2337/diacare.26.1.36 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Ritzwoller DP, Sukhanova A, Gaglio B, Glasgow RE. Costing behavioral interventions: a practical guide to enhance translation. Ann Behav Med 2009;37:218–27. 10.1007/s12160-009-9088-5 [DOI] [PubMed] [Google Scholar]
- 26.Ramsey SD, Willke RJ, Glick H, Reed SD, Augustovski F, Jonsson B, et al. Cost-effectiveness analysis alongside clinical trials II-An ISPOR Good Research Practices Task Force report. Value Health. 2015;18:161–72. 10.1016/j.jval.2015.02.001 [DOI] [PubMed] [Google Scholar]
- 27.Walker E, Nowacki AS. Understanding equivalence and noninferiority testing. J Gen Intern Med 2011;26:192–6. 10.1007/s11606-010-1513-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Pagoto S, Schneider KL, Whited MC, Oleski JL, Merriam P, Appelhans B, et al. Randomized controlled trial of behavioral treatment for comorbid obesity and depression in women: the Be Active Trial. Int J Obes 2013;37:1427–34. 10.1038/ijo.2013.25 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Usage Note 48616: Design and analysis of noninferiority studies. SAS. Published February 11, 2016. Accessed July 24, 2023. http://support.sas.com/kb/48/616.html [Google Scholar]
- 30.Power (Sample Size) Calculators. SealedEnvelope. Published 2012. Accessed July 24, 2023. https://www.sealedenvelope.com/power/continuous-noninferior/ [Google Scholar]
- 31.Piaggio G, Elbourne DR, Pocock SJ, Evans SJW, Altman DG. CONSORT group for the reporting of noninferiority and equivalence randomized trials. JAMA 2012;308:2594. 10.1001/jama.2012.87802 [DOI] [PubMed] [Google Scholar]
- 32.Littell R, Milliken G, Stroup W, Wolfinger R SAS system for mixed models. SAS Institute Inc: Cary, NC, USA, 1996. [Google Scholar]
- 33.Elobeid MA, Padilla MA, McVie T, Thomas O, Brock DW, Musser B, et al. Missing data in randomized clinical trials for weight loss: scope of the problem, state of the field, and performance of statistical methods. PLoS One. 2009;4:e6624. 10.1371/journal.pone.0006624 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Muoio D Mobile health app development costs $425,000 on average, likely continuing to rise. MobiHealthNews. Published March 26, 2018. Accessed July 23, 2023. https://www.mobihealthnews.com/content/mobile-health-app-development-costs-425000-average-likely-continuing-rise [Google Scholar]
- 35.Yellowlees PM, Parish MB, Gonzalez AD, Chan SR, Hilty DM, Yoo B, et al. Clinical outcomes of asynchronous versus synchronous telepsychiatry in primary care: randomized controlled trial. J Med Internet Res 2021;23:e24047. 10.2196/24047 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Bernard RM, Abrami PC, Lou Y, Borokhovski E, Wade A, Wozney L, et al. How does distance education compare with classroom instruction? A meta-analysis of the empirical literature. Rev Educ Res 2004;74:379–439. 10.3102/00346543074003379 [DOI] [Google Scholar]
- 37.Frederick DA, Pila E, Malcarne VL, Compte E, Nagata JM, Best CR, et al. Demo-graphic predictors of objectification theory and tripartite influence model constructs: the U.S. Body Project I. Body Image. 2022;40:182–99. 10.1016/j.bodyim.2021.12.012 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.The Diabetes Prevention Program Research Group. The Diabetes Prevention Program. Design and methods for a clinical trial in the prevention of type 2 diabetes. Diabetes Care. 1999;22:623–34. 10.2337/diacare.22.4.623 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.McVay MA, Lavoie HA, Rajoria M, Leong MC, Lou X, McMahon LN, et al. Pre-enrollment steps and run-ins in weight loss trials: a meta-regression. Am J Prev Med 2023;64:910–7. 10.1016/j.amepre.2023.01.034 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The measures, data, and code for the present study can be obtained by contacting the first author.