Design
The study was a pragmatic remote two-arm parallel double-blind randomized controlled trial (RCT) in the UK, with 1:1 automatic randomization to the intervention and control versions based on random numbers function embedded within the registration process within the app, and with 8-week follow-up. The study received ethical approval from UCL Research Ethics Committee (ID: 5398/001), and was prospectively registered (ISRCTN33423896). The reporting follows CONSORT (Ruano-Ravina, Figueiras, Montes-Martinez, & Barros-Dios, 2003) (see Supplementary Material 1) and TIDieR guidelines (FDA, 2018). Two changes to the protocol were made after the trial initiated: 7-month follow-up was suspended, and participants using NRT on prescription were included (see Supplementary Materials 2 for details). Additionally, due to very slow recruitment, the trial was terminated after 18 months.
Participants
Participant recruitment
Recruitment was through self-identification and self-selection, was conducted remotely, with no contact with the researcher (Eysenbach & Group, 2011). The recruitment campaign lasted between 23rd March 2015 and 15th September 2016. Recruitment materials were delivered to around 300 UK community pharmacies, mostly through their central managerial offices, with instructions to display and distribute them among smokers who purchase NRT (see Supplementary Materials 3 for recruitment materials). The materials directed potential participants to the study website with a detailed study information sheet, information about data processing, End User Licence Agreement, and links to download the app for free (www.nrt2quit.co.uk). The app could also be found through online searches and on iTunes.
Recruitment via community pharmacies
The majority (n=250) of the pharmacies belonged to one large pharmacy chain and were identified through the central managerial office who was supportive of the study. Additional 50 pharmacies were recruited from other major pharmacy chains by communicating with their communications teams and by directly approaching several independent pharmacies. However, no training or direct communication between the researchers and the pharmacy staff were planned (to limit staff burden and to ensure the context of recruitment of smokers remains as ecological as possible) nor possible.
The study promotion could take place only outside of the busy periods, such as Christmas and New Year’s. Only leaflets, rather than larger posters, could be distributed in the participating pharmacies. The leaflets delivery was preceded by internal email communication and accompanied by a printed letter for the head pharmacists instructing them to place the leaflets near the counters and NRT displays, and to provide leaflets to customers purchasing OTC NRT. No professional company was involved in developing the recruitment campaign, and it was not possible to trial the recruitment materials or procedures. Some of the pharmacies in London were visited by the first author to identify ways of improving recruitment, but no further changes to recruitment were possible.
Eligibility criteria
Only iPhone users (with iOS 8+) could participate. Eligibility for the trial was assessed based on the information provided during registration via the app: (a) UK-based, (b) aged ≥18 years, (c) daily smoking ≥10 cigarettes/day, (d) use at least one NRT product, (e) downloaded the app to quit, (f) completed registration process, including providing plausible and complete contact details (these were assessed manually by the researcher), and (g) provided consent to participate that also implied no contraindications for NRT use.
Sample size
The target sample size was calculated a priori to be 1186 participants (with alpha=0.05, two-tailed) to have 80% power to detect an expected effect size of OR=1.7, translating to 5% difference in self-reported abstinence rates at 8-week follow-up (8% in the control and 13% in the intervention. The expected cessation rates for intention-to-treat were low as it was expected that attrition from the study will be as high as 50% from each group (Eysenbach, 2005). The expected effects were small, but potentially cost-effective (West, 2007a).
NRT2Quit platform and intervention and control arms
NRT2Quit intervention and control app versions were delivered through a single NRT2Quit app platform that could be used offline except for changing the quit date or NRT use to ensure data were synchronised with the server. Both versions of NRT2Quit were developed to be automated and standalone interventions. The advice offered was tailored to the type of NRT product used and the quit date (control and intervention), and to dependence level (intervention only, see 2.3.2). The support was offered for up to two weeks before the quit date and eight weeks post-quit date. Detailed information about the NRT2Quit functionality, the different behaviour change techniques (BCTs) delivered within the app (25 BCTs targeting adherence to NRT, and 27 targeting smoking cessation in general), screenshots, and user journeys of the intervention and control are provided in Supplementary Materials 4. The app was not modified during the trial.
NRT2Quit – Control (minimal) version
The control version of the app provided only minimal support with quitting and NRT use: (1) setting of a quit date in the next 2 weeks, (2) very brief advice on the use of selected NRT, (3) brief advice on quitting and managing nicotine withdrawal, (4) progress monitoring (days to and since the quit date), (5) a calendar that displays the quit date and the 8-week questionnaire. It also included (6) brief information about the study and the app. Users could (7) change the quit date and the NRT used.
NRT2Quit – Intervention (full) version
The intervention version of NRT2Quit offered the same support as the control version, and in addition provided (1) more comprehensive information about NRT in general and about each of products (e.g. detailed instructions on use, short articles about key misconceptions, e.g. overdosing), (2) an interactive dashboard for monitoring and feedback on NRT use, (3) daily diary on smoking and NRT use followed by tailored feedback, (4) a more detailed advice on quitting, (5) daily tips, (6) additional information about the study team and study rationale, and (7) daily reminders to engage with the app. Feedback and advice on NRT use were minimally tailored to dependence levels (heavy smokers were all those who were smoking 11-19 CPD and smoking the first cigarette within 5 min since waking, or all those smoking ≥20 CPD; moderate smokers: everyone else). The app was designed to encourage daily use (e.g. through app reminders and new daily tips). However, in anticipation of high attrition from the app (Eysenbach, 2005) and given the different preferences for usage of digital interventions among smokers (Herbec et al., 2014), the core content and behaviour change techniques were delivered immediately following the registration.
After downloading the app, participants were guided through a tunnelled registration process that included a summary of study information sheet and links to study website with detailed information, provided informed consent and contact details, completed baseline assessment, entered data on the NRT purchased, and set their quit date (see Supplementary Material 5). After registering participants were automatically randomized to the intervention or control versions of the app and were assigned a unique ID. Participants received an email confirming registration with a link to study website and contact details to the researchers. Duplicate registrations were excluded following a manual check.
The follow-up took place 8 weeks after the registration (18 May 2015-22 November 2016) through an online survey as opposed to within the app in anticipation of participants deleting the app or switching off notifications. The links to the survey were distributed through an e-emails (up to 3 reminders) that were personalized (Jamie Brown et al., 2014). Participants failing to complete the survey were contacted over the phone (up to three calls) to assess smoking status only (a longer survey was judged to be not feasible over the phone). Participants self-reporting prolonged abstinence were posted a saliva kit with instructions, a £20 high street gift voucher as reimbursement, and a freepost envelope addressed to the laboratory, and asked to post the samples as soon as possible (Jamie Brown et al., 2014).
Due to slow recruitment it was decided in early August 2016 to prepare for termination of the trial. Bayes factors were calculated on the primary outcome on 18th August 2016 (after 39 eligible participants were recruited), but no hypothesis testing or other analyses were performed. Before NRT2Quit was removed from iTunes on 15th September (the current app users could still access it) two additional participants meeting eligibility criteria joined the study and were included in the analyses reported here. All study procedures, including the follow-up for all participants, were conducted blind to study arm allocation.
Measurements
Baseline assessment
The baseline questionnaire assessed socio-demographic characteristics (age, gender, having post-16 years of age education vs. not), smoking and quitting history (items from the Heaviness of Smoking Index (Etter, Duc, & Perneger, 1999); when the last quit attempt was made, past use of cessation aids) and reasons for joining the study (to quit smoking/other). Participants also provided information about the NRT type purchased (NRT patch/fast acting NRT/combination), how they obtained NRT (OTC/on prescription/both); and whether they received any support with NRT use from HCPs (yes/no).
Primary outcome
The primary outcome was self-reported 4-week prolonged abstinence assessed at 8-week follow-up, and verified by saliva cotinine levels of <15ng/mL (West, Hajek, Stead, & Stapleton, 2005) or, among participants reporting using NRT or e-cigarettes: anabasine levels of <1ng/mL (Benowitz et al., 2002; Brown et al., 2014). The pre-registered salivary anabasine cut-off value was based on discussions with the processing lab and the information available at the time of trial set up (2011-2014). However, as the lab has conducted more studies since, it now recommends a lower cut-off value for salivary anabasine of <0.2ng/mL. Results for the lower cut-off are reported in the footnote of Table 2. Participants lost to follow-up were assumed to have resumed smoking, as per intention-to-treat principle (ITT).
Secondary outcomes
Secondary outcomes were: (1) the follow-up parameters: follow-up rate, the re-contact channel (survey/phone), proportion of saliva samples returned. The online survey at 8-week follow-up assessed: (2) total number of cigarettes smoked in the past 4 weeks (none/<5/≥5); (3) adherence to NRT: (i) use of NRT on the follow-up day (yes/no), (ii) weeks NRT was used (<5/≥5 weeks), (iii) number of days in those weeks NRT was used (every day/not every day); (4) use of other cessation support, e.g. other medications, behavioural support, or self-help support (yes/no); (5) satisfaction: how helpful was NRT2Quit app for (i) quitting smoking and (ii) using NRT (1=not at all, 5=extremely helpful), (iii) whether would recommend the app to others wanting to quit (yes/no). Additionally, (5) data on app usage: (i) number of logins, and (ii) number of days users logged in on. Due to the structure of app database, data on time spent using the app or on accessing individual app features were not saved.
Data Analysis
The primary outcome was analysed using Fishers’ exact test. Additionally, unadjusted logistic regressions were conducted for the dichotomised cessation outcomes, and odds ratios (ORs) and 95% confidence intervals (CIs) were calculated. In exploratory sensitivity analyses participants who reporting using only Rx NRT (n=14) or for with that data missing (n=3) were excluded. All other analyses were pre-planned. For smoking outcomes, participants with missing data were assumed to be smoking.
Bayes Factors were calculated for the smoking outcomes as they can distinguish between the likelihood of both the null and alternative hypotheses, and assess whether the data provide an insensitive test of the hypotheses (Brown, Michie, Walmsley, & West, 2016; Dienes, 2008, 2014; West, 2016). Bayes Factors were calculated using an online calculator that is available for free at http://www.lifesci.sussex.ac.uk/home/Zoltan_Dienes/inference/Bayes.htm. We used a uniform H1 distribution with a possible expected effect size between OR=1 and OR=3 versus an H0 of OR=1. In sensitivity analyses, we used a conservative H1 with a half-normal distribution with the mean of the log OR of 0, and the standard deviation corresponding to expected effect sizes of OR=1.2, OR=1.7, and OR=2.5 (Rigotti et al., 2017; West, 2016). This distribution means that plausible values have been represented between zero and twice the effect size, with smaller values more likely.
Descriptive statistics are presented for baseline and all secondary outcomes. Categorical variables were compared using Fisher’s exact test, and chi-square test and Linear-by-Linear association for ordered categories, and continuous data using independent t-test or Mann-Whitney U-test for data that were not normally distributed. Data on app usage were not normally distributed, but both medians (IQR) and means (SDs) are reported to enable comparison with other studies. All tests were 2-sided, and alpha was set to 5%.