Participants were aged 18-25, English literate, and had smoked ≥100 cigarettes in their lifetime, currently smoked at least 3 days per week (consistent with the definition of smoking used in the National Health Interview Survey) [29], and used Facebook at least 4 days per week. Participants had to have access to technology that could take and send an online picture for verification of tobacco abstinence with a study mailed saliva cotinine test.

Recruitment efforts included a paid Facebook ad campaign conducted between June and August 2013 with details reported previously [30]. Advertisements invited participants to a secure, confidential online survey to evaluate eligibility and, for those eligible, informed consent to participate in the intervention. Online consent questions were used to confirm understanding of study procedures. Consented participants were asked to send proof of identity either by emailing a copy of a photo ID with birth date or by “friending” the study on Facebook to determine age.

After completing an online baseline assessment, all participants were assigned to “secret” Facebook groups tailored to their stage of change for quitting smoking (preceontemplation, contemplation, preparation) [31,32]. As per Facebook privacy options for groups, only study administrators and group members knew the existence of “secret” groups, members of the group, or any member activity. This was distinct from a “public” group, in which anyone could join and actions were public, or a “closed” group, in which anyone could ask to join or be invited and existence/membership was not private. More information about the differences between group types on Facebook is described in Facebook’s Help Center [33]. Secret groups were chosen based on mixed-methods formative work with young adults suggesting that privacy was an important consideration that would likely affect participation in a Facebook smoking cessation intervention [23]. A stage-matched group began 2 weeks after the first participant was assigned so that no group member was kept waiting longer than 2 weeks to start the intervention; thus, group sizes varied. The intervention was delivered daily for 90 days. During the intervention, participants’ stage of change was assessed monthly and if they had advanced, they were invited to join a later stage group. As this was a feasibility evaluation, participants could join the later stage group, remain in their original group, or add a second group and participate in two. We assessed program use and acceptance at intervention end (3-month follow-up), and smoking outcomes at 3, 6, and 12 months. Participants were contacted by email and Facebook to complete online follow-up assessments and compensated for each assessment with US $20 gift cards to their choice of national retailer (eg, Amazon, Best Buy) and an additional US $20 gift card if all four assessments were completed for a total possible compensation of US $100.

All measures were administered online using Qualtrics. Qualtrics is a secure, online, Health Insurance Portability and Accountability Act (HIPAA) compliant survey software that transmits data to and from secure, firewalled data centers using Transport Layer Security encryption, the successor to Secure Sockets Layer encryption. Secure access is available to all faculty and staff at the University of California, San Francisco.

Descriptive statistics characterized the sample at baseline, evaluated the number of completed assessments at 3, 6, and 12 months, and characterized the intervention (number and size of groups, number of participants changing or leaving groups, number of participants opting to attend CBT sessions). Primary and secondary outcomes were assessed at 3, 6, and 12 months. Primary outcomes were 7-day point prevalence abstinence and biochemically verified abstinence. Secondary outcomes of interest were the proportion that made at least one 24-hour quit attempt, calculated for each time period, and for the entire 12-month follow-up period. Use of strategies to quit smoking between baseline and 12-month follow-up was summarized and predictors of strategy use were analyzed with logistic regression analyses. Wilcoxon signed rank tests compared past 7-day cigarettes smoked between baseline and each follow-up. Bowker’s test for change in a repeated categorical measures evaluated stage of change transitions between baseline and 3 months, 3 and 6 months, and 6 and 12 months (3 models). To evaluate engagement, total number of “likes” and comments to Facebook groups over 3 months were tallied. Due to skew, the non-parametric Kruskal-Wallace analysis of variance (ANOVA) evaluated comments to the Facebook groups by baseline stage of change. Mann-Whitney U tests compared comments for those with biochemical-verification of abstinence at 3 months to those non-abstinent. Kruskal-Wallace tests evaluated the effect of incentive condition (personal, altruistic, no incentive) on comments to the Facebook group, in both the full sample and only those who made at least one comment to the group to address highly skewed data. Within the altruism incentive group, Kendall’s tau test indexed trait altruism by comments. Kendall’s tau test also indexed incentive condition by 3-month reported abstinence and biochemically verified abstinence. Usability and satisfaction with the intervention were evaluated by tallying the proportion of users answering “agree” or “disagree” to each item on the usability measure.

The 79 enrolled participants who completed a baseline assessment, on average, were 20.8 years old (SD 2.1) and primarily male (63/79, 80%) and non-Hispanic white (63/79, 80%). Seventeen percent (13/79) identified as sexual minority (lesbian, gay, or bisexual). About one quarter (22/79, 28%) had a household income >US $60,000. Average years of education was 12.4 (SD 2.0); 38% (30/79) were currently enrolled in school and 56% (44/79) were currently employed; and 43% (34/79) lived with their parents. Three quarters (59/79, 75%) were daily smokers, smoking 11.5 cigarettes per day on average (SD 8.3). Only a minority (8/79, 10%) of participants reported an abstinence smoking goal, 60% (47/79) reported a reduction goal, and 30% (24/79) reported no goal. More than half (41/79, 52%) of participants smoked within 30 minutes of waking. Participants smoked their first cigarette at age 14.2 years (SD 3.4), started smoking regularly at age 16.3 (SD 2.8), and had smoked for 2.7 years (SD 0.7) on average. Over half (45/79, 57%) had made at least one smoking quit attempt in the past year.

Follow-up completion was 76% (60/79) at 3 months, 82% (65/79) at 6 months, and 72% (57/79) at 12 months with no difference in 3-month retention by incentive condition (χ² _2,79=.31, P=.855).

Seven Facebook secret groups were created of varying sizes (two “Not Ready to Quit,” three “Thinking About Quitting,” and two “Getting Ready to Quit”; group size range 6-22). During the 90-day intervention period, 5% (4/79) of participants opted to change to a later stage group; 8% (6/79) of participants left their Facebook group completely at some point during the 3-month intervention period (3 precontemplation, 3 contemplation). Several participants (5/79, 6%) entered into CBT treatment, with one group made up of 3 participants, and 2 participants with individual treatments due to lack of other participants for groups. Participants attended 6 of 7 CBT chat sessions on average.

Assuming missing=missing (ie, complete case analysis), reported 7-day point prevalence abstinence was 12% (9/79), 22% (17/79), and 18% (14/79) at 3-, 6-, and 12-months follow-up (see Figure 2). Assuming those lost to follow-up were smoking (ie, missing=smoking), 7-day point prevalence abstinence was 9% (7/79), 18% (14/79), and 13% (10/79) at 3-, 6-, and 12-months follow-up. At each time point, cotinine-verified abstinence was confirmed with approximately half of those reporting abstinence.

In our pilot investigation, a novel Facebook-integrated smoking cessation intervention was found to be likable and feasible to deliver to young adults who smoke cigarettes. Early efficacy data are encouraging and support further investigation. Retention also was good, especially for a digital health intervention, though it dropped off for those in preparation at 12 months. Unknown is whether the loss to follow-up was related to quitting or relapsing. This was the smallest group, hence, the values are less stable and more subject to variability. The low sample size in this group could have led to less engagement with the intervention, subsequently leading to a lower desire to participate in follow-up surveys. However, usability data did not differ across readiness to quit groups. A trial with a sample size to support larger groups of those ready to quit is warranted.

Reported and biochemically verified abstinence increased from 3 months to 6 months and then declined slightly at 12 months, while biochemically verified abstinence was flat between 6 and 12 months. Previous trials testing the efficacy of stage-tailored interventions for smoking cessation and other health risk behaviors have found increased abstinence rates over 24 months with repeated exposure to treatment across 1 year (eg, [50]). Participants had access to Facebook groups and all associated materials over the 12-month trial; however, extending the active intervention phase (daily postings) past 3 months, or having “booster” postings over time, may enhance the efficacy of the intervention at 12 months.

Monetary incentives increased intervention uptake, but only for those who had some level of engagement (ie, commented at least once during the 3-month intervention). Further, engagement was related to abstinence only among those who commented at least once. For those motivated to engage, monetary incentives appear to be an important adjunct to a health behavior change intervention and could translate to behavior change. Monetary incentives have been identified as a main motivator for participation in online mental health interventions [51]. For smoking cessation interventions with young people, incentives have been associated with lower retention [52]. In our study, we tied a single monetary incentive (US $50) to commenting on all daily posts for 90 days. Findings suggest that a personal incentive was more effective than one in which participants provided it to others regardless of trait altruism. A more frequent incentive (eg, weekly) may further increase participation. Future studies should test various incentive schemes for engagement, balance incentives across participants and stages of change, and be powered to test the effects of groups to better determine the strategies most effective to increase engagement and for whom.

Participation in CBT counseling was relatively low, yet attendance among those who elected to receive CBT was high (6 of 7 sessions were attended on average). Conducting counseling sessions through Facebook chat was thought to provide a more personal connection. To maximize engagement, however, these sessions could be offered in a more open format than invited chat sessions, perhaps through the stage tailored groups to which all intervention participants were assigned. This would allow for those motivated to engage to do so and allow others to still view session material (eg, handouts, tip sheets, interactions between other participants and study counselors). Additionally, allowing participants to move between groups in response to an increasing stage of change was not used by many and potentially may undermine the social support generated from participating in a single group; thus, this strategy should likely not be used in the future.

Our feasibility study found that two-thirds of all Facebook intervention participants and one third of participants in the Precontemplation stage of change undertook one or more quit attempts, with each participant making at least one attempt reporting more than four on average. These findings are in line with results from daily assessments [53] and speak to the dynamic nature of the smoking cessation process and to the fact that many smokers need multiple quit attempts in order to attain smoking abstinence. However, although we used a standard definition of quit attempts (24 hours of purposeful abstinence), our inclusion of non-daily smokers could have led to overreporting of non-smoking days as quit attempts, although the assessment made clear that the definition included only purposeful attempts to quit smoking. Trying to quit without assistance other than the Tobacco Status Project intervention was much more popular than using a cessation aid such as medication among young adults. This is congruent with previous studies among young adults [54] and adolescents [55] and highlights the need to incorporate evidence-based treatment into media that are widely accessed by young people such as Facebook. Electronic cigarettes were used more frequently as a cessation aid than nicotine replacement therapy, even though clear evidence for their effectiveness is still lacking [56,57] and their use was not recommended in the intervention. Young adults may have strong expectations for the effectiveness of electronic cigarettes as a smoking cessation strategy [58]. Given the wide marketing of these products online to young adults [59], the high use in our sample is not surprising. Until convincing data on the effectiveness of electronic cigarettes for smoking cessation is presented, their use should not be recommended to young adults trying to quit smoking.

We tested a novel method of collecting biochemical verification of tobacco abstinence by mailing participants saliva cotinine test strips and asking them to send back pictures of test results. Incentives for assessment completion were not given until proof was sent. This strategy was successful for about half of those reporting abstinence at 3-, 6-, and 12-month assessments. Of the 13 cases for which we were unable to obtain biochemical verification of abstinence at any time point, 6 were unreachable, 1 had technical difficulties preventing transmission of verification data (photos), 1 had difficulty completing the test at home and results were seen as inconclusive, and 5 were not sent test kits due to errors in survey programming on the study end. This piloting of methods was valuable for informing quality control in our future efforts. Overall success with completing the biochemical procedure at home once test kits were sent was promising, and our strategy shows that validation measures are possible in an environment where information is increasingly easy to send as data (eg, through a smartphone). Notably, we did not have trouble recruiting young people into the study despite knowledge that the trial would ask for biochemical verification of abstinence, as has been a concern of others [60,61].

With a goal of determining feasibility and initial efficacy, this study was not adequately powered to fully test many of the relationships examined here. A larger, 2-group randomized trial powered to detect effects on primary outcomes and differences among incentive or motivation groups is warranted. Further, Facebook is a dynamic environment where design issues become out-of-date quickly. Potential changes to user agreements mean that researchers must be vigilant about privacy and confidentiality when working in this medium. Formative work with our target population indicated that young adult smokers were ashamed of smoking in many cases and wanted to keep the study involvement private from their larger Facebook social network [23]. This suggested the intervention should be conducted within the context of Facebook secret groups where only study administrators and group members have knowledge of their existence. Investigators then watched Facebook’s privacy practices to ensure that this did not change; if so, we would have considered a different context in which to run groups.

Another limitation is the general lack of sample diversity. In contrast to other online interventions in the United States or Canada that show a majority of female participants [62,63], our sample was primarily male. This was a surprise, given that a larger proportion of American women online use Facebook than men (77% vs 66%) [64]. There was no direct targeting of participants by gender in the Facebook advertising campaign, as one goal of this feasibility study was to determine the characteristics of participants who would enroll. Strategies are needed to recruit more female and ethnic minority participants. Facebook targeting can be used to place ads in locations including states and cities where more ethnic minority smokers reside. Advertising images can be used to target women and non-white smokers. Bull et al [65] recruited a large proportion of non-white participants into a Facebook sexual health intervention using respondent-driven sampling. Using a strategy to recruit “seed” participants and friends in their (real-world) social networks could generate samples that are diverse with respect to gender, ethnicity, and sexual orientation.

Harnessing the popularity of social media to treat tobacco use in young adults holds great potential considering the overwhelming numbers using this medium daily. Focused on young adult smokers, a challenging group to engage, our study’s high retention and usability ratings suggest the Facebook quit smoking intervention is attractive and feasible to deliver. Early efficacy data are encouraging and support further investigation in a larger sample with a randomized design.