This parallel randomized controlled trial (RCT), allocation ratio 1:1, was approved by the University of South Australia Human Research Ethics Committee (protocol number: 0000033766). Data collection took place in Australia from January 2016 to August 2016. Data analysis occurred from August 2016 to November 2016. Participants provided informed consent online before commencing the study. The study was designed and is reported following the Consolidated Standards of Reporting Trials guidelines [22] and is registered with the Australian and New Zealand Clinical Trials Registry, protocol number: ACTRN12616001500448.

Participants were recruited through a variety of advertising methods including online (via Facebook advertising) and via print media. Participants were eligible to take part in the study if they met the following criteria: (1) aged between 18 and 50 years, (2) Australian residents, (3) current Facebook users, (4) able to read and understand English, and (5) not participating in a regular running program. Individuals were excluded if they had a medical condition that would prevent them from participating in a running program and if they were pregnant or planning to become pregnant within the next 5 months.

Interested participants were directed to the UniSA Run Free Facebook page that provided study information and invited potential participants to register their interest by privately texting their contact details. An online survey was used to confirm eligibility (participants were asked questions related to the eligibility criteria), collect written informed consent, and perform baseline assessments. Participants were formally enrolled once they completed baseline surveys. Upon enrollment, they were randomly allocated (by the primary researcher, JL) to the intervention (UniSA Run Free program) or control condition (self-directed running program) using a computer-generated random number sequence with allocation concealment (opaque envelopes were used for allocation concealment).

Details regarding the honorarium for this study were provided in the participant information sheet and consent form. All participants who enrolled in the study and who completed all three assessments were placed in a prize draw for an Aus $200 gift voucher.

There were three assessment time points: baseline, 2 months (coinciding with the last week of the running program), and 5 months (3 months post program). All assessments were self-administered and completed remotely (online). Blinding of participants was not possible because of the nature of the intervention. Blinding of assessors was not applicable, as assessments were self-administered.

Participants’ baseline characteristics were analyzed descriptively. Changes in primary and secondary outcomes from baseline to the 2- and 5-month assessments were analyzed using random effects mixed modeling. Analyses were conducted using generalized linear mixed models (GLMMs) in Statistical Package for the Social Sciences (SPSS) version 21 (IBM Corp), with the individual entered as a random effect and group, time, and group x time interaction entered as fixed effects. Data analysis was completed on an intention-to-treat basis, where all participants randomized at the commencement of the trial were retained for analysis regardless of compliance [36]. The GLMM function is able to handle missing data; therefore data imputation was not needed. Effect size differences between groups at 2 and 5 months were calculated using Cohen d [37].

Compliance and engagement with the UniSA Run Free program were described descriptively. A subgroup analysis using GLMM was undertaken within the UniSA Run Free group to determine if change in MVPA was related to key sociodemographic characteristics (age and sex), baseline characteristics (BMI, fitness, and MVPA), as well as compliance with the running program and engagement with the Facebook group (ie, liking or commenting on the facilitators posts). For these analyses, the predictor variables were dichotomized into high and low categories based on the median splits. Specifically, compliance was categorized into high (≥70% of running sessions completed) and low (<70%), fitness categorized into high (<100 beats per minute, BPM) and low (≥100 BPM), engagement into high (≥15 interactions) and low (<15 interactions), BMI into high (≥25 kg/m²) and low (<25 kg/m²), and age into older (≥35 years) and younger (<35 years). Baseline PA was dichotomized on the basis of meeting the PA guidelines (≥150 weekly minutes) and failing to meet guidelines (<150 weekly minutes).

A priori sample size calculations suggested that a sample of 114 participants would be sufficient to detect a moderate effect size (d=0.4), assuming a two-group design with three repeated measures, 80% power, and an alpha of .05. Because the study was conducted in the context of a student research project, there were time constraints on participant recruitment. A total of 89 participants were recruited with post-hoc power analyses, suggesting that this sample had 64% power to detect effect size differences of d=0.4.

A total of 210 potential participants registered their interest in the study; however, only 89 met the participant criteria and completed baseline assessment and were therefore formally enrolled. Of these 89 participants, 41 (46%) were randomized to the intervention group and 48 (53%) to the control group (based on the computer generated number sequence). Three-quarters (78% [69/89]) completed the 2-month assessments, whereas two-thirds (65% [58/89]) completed the 5-month assessments.

Twelve participants formally withdrew from the study for various reasons, as listed in Figure 3.

Participants’ demographic and baseline characteristics are given in Table 1. Of the 89 participants, 71 (80%) were female, and the mean age was 35.2 years (SD 10.9). Overall, 37 participants (42% [37/89]) were within the normal BMI range [35], 28 were overweight (32% [28/89]), and 23 were obese (26 [23/89]). A total of 65 participants (73% [65/89]) were currently undertaking or had completed a university degree or higher. Participants reported getting a mean of 318 min (SD 278) of MVPA per week.

The results for the primary and secondary outcome measures are shown in Table 2.

Subgroup analyses was undertaken to determine whether, within the intervention group, changes in MVPA were related to age, sex, highest level of education, percentage of running sessions completed, baseline BMI or fitness, engagement with the Facebook group, and baseline PA. Results showed that participants with high overall program compliance increased their MVPA significantly more than the participants with a low compliance (<70%; P=.03). In addition, participants who failed to meet PA guidelines at baseline (<150 weekly minutes) increased their MVPA significantly more (P=.04). Changes in MVPA appeared unrelated to age (P=.90), sex (P=.07), education (P=.95), baseline BMI (P=.89), and baseline cardiorespiratory fitness (P=.94).

Throughout the 8-week intervention, 5 participants reported adverse events in the form of musculoskeletal lower limb injuries (intervention group n=2 and control group n=3). Three of these participants reported an exacerbation of a pre-existing condition, one participant sustained a knee injury from participating in another activity (nonstudy related), and the remaining participant sustained a new knee injury during a running session. All adverse effects were reported to the University of South Australia Human Ethics Committee.

The key findings of this study was that a beginners’ running program delivered via a Facebook group produced sizable and sustained changes in weekly MVPA compared with the same running program delivered in a self-administered format. Both groups reported a significant improvement in social support for PA, and there was a trend for both groups to improve their cardiorespiratory fitness, though this did not reach statistical significance. The running program delivered via Facebook achieved strong engagement, high compliance, and favorable feedback from participants.

Although some previous studies utilizing Facebook groups to deliver PA interventions have reported significant improvements in MVPA over time, to the best of our knowledge, no significant group by time effects have been noted [15,18]. The positive group by time effects for MVPA for this study may be related to a number of factors. First, the Facebook group was used to deliver all of the intervention materials; rather than being a component of a more complex intervention. Second, facilitator posts that varied in content and style were provided daily to encourage participant engagement. Third, the type of posts made, that is, posts containing questions, photos, and humor were guided by previous research, suggesting these post types are associated with higher engagement [38,39].

There was a nonsignificant trend for both groups to improve their fitness across the study period (time effect P=.12). Given that the study was underpowered and at risk of type 2 errors, the trend may in fact represent true improvement in fitness. Conversely, it is possible that an 8-week program was not long enough to see significant changes in this outcome. Wenger and Bell [40] suggest that it takes 10 to 11 weeks to improve cardiorespiratory fitness. The YMCA step test was selected, as it is easy for participants to self-administer, relatively safe (being a submaximal test), and can be completed with minimal special equipment. However, it is acknowledged that this is less reliable than laboratory-based cardiorespiratory fitness tests [41].

Relative to other online social networking interventions, participant engagement in UniSA Run Free was high; as indicated by a mean number of 35 (SD 41) total Facebook interactions, 19% of participants visiting the Facebook group daily, and 78% visiting between 1 and 6 times per week. In comparison, Wójcicki and colleagues [42] who investigated the feasibility of an 8-week Facebook group–delivered PA intervention in adolescents, reported low levels of group engagement, with only 27% of participants interacting with the Facebook group. Similarly, Napolitano and colleagues [20] reported only 24% of participants interacting with the Facebook group during an 8-week weight loss intervention. These differences in engagement may be because of the Facebook group being implemented differently in this study. Most other studies have used a Facebook group as a supplement to other intervention materials. In contrast, in our study, the Facebook group was central to the intervention and included all key intervention materials.

Retention rates for this study were also reasonably high at 78% at 2 months and 65% at 5 months. Similarly, high retention rates have been noted in previous Facebook-delivered PA interventions, ranging from 77% [43] to 96% [20]. The high retention rates for this delivery method may be related to its real life design, whereby participants can complete the intervention and assessments at home with minimal contact from research personnel, making it less intrusive and easier to fit around daily routines.

This study had a number of methodological strengths. First, it was a parallel RCT, including intention-to-treat analysis, which is the gold standard in clinical trial design. Second, the comparison group were provided with an alternative intervention (hard copy of the running program), which allowed comparison of the two intervention delivery methods (hard copy vs Facebook group delivery) [44].

High levels of engagement and retention were also strengths and provide evidence of the feasibility of a Facebook group for intervention delivery. Furthermore, this delivery method enabled the intervention to be available to participants living all across Australia, thus, further demonstrating that online social networks have minimal geographic boundaries. Finally, results of the subgroup analyses found that the intervention was more effective in participants who failed to meet the minimum PA guidelines at baseline (150 min per week), indicating that this intervention was successful in assisting those most at risk of physical inactivity.

Key limitations of this study should also be addressed. Due to time restrictions with recruitment, this study was underpowered. Future work is needed to determine the effectiveness of this intervention with a larger sample, particularly for cardiovascular fitness and social support, which showed trends for improvement. Higher baseline MVPA of the control group (90.6 min/week higher than the intervention group) must also be acknowledged because of potential ceiling effect.

For practicality, all outcome measures were self-reported or self-administered. Self-reported measures are typically more prone to social desirability bias [45], and the self-administered nature of the step test reduces the ability to standardize test conditions and accuracy of heart rate measurement. In addition, determining participants’ social networking use (ie, frequency) at baseline would be beneficial as this may influence engagement and health behavior outcomes.

Finally, the somewhat homogenous nature of this sample (female and highly educated), which is typical of volunteer research studies, is also acknowledged [46,47]. This, along with the high baseline MVPA levels (for both groups), warrants caution in generalizing results. Further research is required to investigate the effectiveness of this intervention with other population subgroups (eg, teenagers and individuals over 50 years).

Previous research has found modest evidence supporting Facebook groups as a delivery method for PA interventions. Many previous studies have offered additional intervention materials and resources, making it difficult to disentangle the effectiveness of the online social networking component. Therefore, this study addressed the effectiveness of an 8-week beginners’ running program delivered entirely through a Facebook group in improving PA and cardiorespiratory fitness. Significant improvements were found in both groups at 2 months in MVPA; this increase was considerably larger in the intervention group (P=.04). Engagement with the Facebook group was relatively high compared with other online interventions. Further research is warranted to investigate the effectiveness of this delivery method in other health-related behaviors and with other population groups. The ease of use, low cost, and accessibility of a Facebook group make it a promising method for delivering socially supportive health and behavioral programs on a mass scale.