E-cigarette support for smoking cessation: Identifying the effectiveness of intervention components in an online randomised optimisation trial

This is an online study of 1214 UK smokers which aimed to determine which combination(s) of five intervention components can help smokers to stop smoking by using an e-cigarette. A balanced five-factor randomised factorial design was used, guided by the Multiphase Optimisation Strategy (MOST). The five intervention components were: i) tailored advice on which e-cigarette device to purchase; ii) tailored advice on which e-liquid nicotine strength to purchase; iii) tailored advice on which flavour to purchase; iv) brief information on relative harms of e-cigarettes vs smoking; v) text message support. Participants completed a baseline survey online and were then randomised to either receive or not receive each intervention component (resulting in 32 possible combinations). Tailored advice on device, nicotine strength and flavour were based on responses to baseline questions and were displayed at the end of the survey (for participants randomised to receive that condition). Follow up data was collected via a second online survey after 12 weeks to collect information on abstinence rates and adherence to intervention components. Logistic regressions were used to model the main effects and two-way interactions on the primary outcome (4-weeks abstinence) and secondary outcomes (7-day point prevalence and ≥50% smoking reduction). The full dataset and syntax (as SPSS files) are available.

Link to protocol: https://www.qeios.com/read/9RDLJA.3

Link to trial registry: https://www.isrctn.com/ISRCTN54776958

Link to published paper: To follow (currently under review with Addiction)

Baseline data was collected from 1455 smokers (N=1214 after removing 241 duplicates and bots – see data processing below) between April and July 2020. Between July and October 2020, 529 of the participants completed the 12-week follow up questionnaire in full and a further 107 provided information on the primary outcome variable only via text/email. Five intervention components were tested with each participant randomised to receive each component (ON) or not (OFF) resulting in 32 experimental conditions. Participants were eligible for inclusion if they were aged 18 or over, a daily smoker, resident in the UK, fluent in English, interested in quitting and using an e-cigarette, had access to a mobile phone and able to make an online purchase. Participants were provided with a voucher (to the value of £50) for making a purchase at the online store. Information collected at baseline included: demographic, smoking and vaping-related information; questions around preferences and nicotine dependence to inform tailoring of advice around device, flavour and nicotine strength; motivation and confidence in quitting; identity questions; and e-cigarette harm perceptions. At the 12-week follow up, data was collected on: smoking cessation and reduction outcome variables (continuous 4-weeks abstinence, 7-day point prevalence abstinence, smoking reduction from baseline); use of the product; adherence to recommendations and suitability of the advice, text messages and written information; identity questions and covid questions (given that the study commenced at the beginning of the first covid-19 lockdown). All variables are clearly labelled in SPSS and a description of the variables is provided below.

Our inclusion criteria specified that an individual could only take part once. However, despite adding reCAPTCHA and blocking repeat entries from the same IP address, e-mail and phone number, our automated randomisation in Qualtrics failed to detect all duplicates and bots so some individuals were erroneously randomised.  Duplicates and bots are protocol violations and were defined as multiple completions from the same individual. These include participants who completed the survey in its entirety multiple times (often in quick succession) thus were randomised to more than one condition. Offending participants managed to circumvent our systems by supplying different email addresses or providing fake telephone numbers.  Polite emails were also sent out to repeat offenders to ask them to stop taking the survey and the survey was halted (for a few days) on two occasions to allow us to implement further measures (e.g. blocking repeat post-codes and updating inclusion criteria to only one entry per household). On every instance, only the first entry from that individual was included in the intention-to-treat analysis. All subsequent entries from that same individual were removed. Violations were systematically identified as follows:
Screening for repeated orders from the retailer where payment was made with the study voucher code (data repository from the retailer) 
Matching these entries across with Qualtrics datafile  
In the Qualtrics file looking for duplicate postal, email and IP addresses and/or phone number  
Reconciling email and IP addresses with the study voucher codes  
To remove violators, use the syntax or variable 100.

Data cleaning
Any missing data is coded as -99
-98 is used for ‘not applicable’
Age – Re-coded into Age_Cleaned
CPD: where a range was provided, the mean was taken for ‘CPD-Cleaned’
CPD_7days: Converted to ‘CPD_7days_Cleaned’ where a number was indicated followed by + or ‘plus’ – only the numerical value was taken. Where the number was indicated in ‘packs’ or ‘grams’ or ‘pouches’ – data was coded as missing. Where a range was given, the mean was taken. Extreme values (i.e. > 2000 a week) were considered to be typographical errors and were re-coded as missing.
Years_Smoking: Converted to ‘Years_Smoked_Cleaned’ - Where a range was provided, the mean was taken. Where years smoked exceeded current age, data was coded as missing
QuitAttempt_Number: Converted to ‘QuitAttempt_N_Cleaned’ – numbers written in words were converted to numerals. Vague responses (e.g. lots, several) were coded as missing.

Variable List (Data dictionary)
Variables 1 – 46 were collected at baseline. Variables 47 to 116 were collected at the 12-week follow up
Baseline Variables:
V.1: ID – participant ID
V.2: RecordedDate – Date the baseline questionnaire was completed
V.3: Baseline_violation – (see description above). Violator = a person who completed the survey more than once; Violation = any entries subsequent to the first entry from a violator. Coding was used so that a violation takes precedence over a violator as this was their first entry. Use the syntax to exclude only violations (coded 1).
V.5 – 11: Baseline demographic information
V.12 – V.22: Baseline Smoking-related Characteristics
V.23 – V.24: Tailoring question for advice on flavour: Those who reported that they smoked mostly menthol cigarettes were assigned to menthol flavour. Those who reported not smoking menthol were asked if they want something that tastes like smoking of a complete change. If they stated the former, they were assigned tobacco flavour. If the latter, they were assigned fruit flavour.
V.25 (FTC): - Tailoring question for advice on nicotine strength based on time to first cigarette. Participants who reported smoking within 5 minutes or within 6-30 minutes were assigned to the 18mg/mL nicotine strength. Those who reported they smoked between 31-60 minutes were assigned to the 14mg/mL nicotine strength. Those who reported they smoked after 60 minutes were assigned the 10mg/mL nicotine strength.
V.26 – V. 27: Motivation and confidence to quit
V.28 - V.31: Baseline dual-user, smoker, vaper and non-smoker self-identification measured using the single-item measure of social identification (SISI) by Postmes et al. (2013) – 7-point Likert-type scale with 1 strongly disagree to 7 strongly agree
V.32 - V.33: Previous e-cigarette use
V.34 - V.35: Reasons for stopping previous e-cigarette use or not trying e-cigarettes
V.37 – V.39: Tailoring questions for advice on e-cigarette device. Tailoring_Score_Device: Tailored advice on device was based on responses to the questions ‘ECMustbeSmall’ ‘ECLotsofVapour’ and ‘ECTechnicalitiesPutMeOff’. Participants scoring between 3-7 on ‘Tailoring_Score_Device’ variable were assigned to purchase the Tank system E-cig device (Arc 5). Those scoring between 8-11 were assigned the Tank system pen-like device (Tornado EX2). Those scoring 12 or above were assigned the Pod system (Scope P).
V.40: Perceptions of e-cigarette harm compared to smoking (taken from the ASH surveys in GB)
V.41 – 43: Recommendations to participants (if they had that component ON) regarding which device, nicotine strength and flavour to purchase
V.44: Score from the tailoring question to inform tailored advice on device (see Vs. 37-39 above).
V.45: Automated randomisation function generated in Qualtrics
V.46: Randomised condition: 1 – 32 (everyone ON through to everything OFF as per Table S1)
12-week Follow-up Variables
V.47: Primary Outcome: have you smoked in the last 4-weeks? (NOTE: V.109 below for intention to treat)
V. 48: Secondary Outcome: 7-day point prevalence abstinence (NOTE: V.110 below for intention to treat)
V. 49: Number of cigarettes smoked in the last 7 days.
V.50: How soon did you receive your e-cigarette
V.51: If not received, text box for why
V. 52: How soon after receiving the e-cigarette did you first use it?
V.53-54: Reasons for not using the e-cigarette
V.55: Time to stop smoking after first using the e-cigarette
V.56-57: Ease of setting up, ease of use
V.58: Did you make a quit attempt?
V.59: Are you still using the e-cigarette?
V.60: List of other products purchased
V.61 – 63: Adherence and suitability of the written information (CRUK infographic)
V.64 – 69: Adherence and suitability of the text messages
V.70 – 73: Follow up dual-user, smoker, vaper and non-smoker self-identification measured using the SISI (Postmes et al., 2013) ( 5-point Likert-type scale from strongly disagree to strongly agree) as Baseline measures was used)
V.74 – 82: Covid questions about influence on smoking and vaping and impact of media stories.
V.83 – 87: Adherence questions regarding tailored advice on device, nicotine strength and flavour
V.88 – 90: Ratings of suitability of the device, nicotine strength and flavour
V.91 – 96: Ratings of positive effects of the e-cigarette: satisfaction, hit, craving reduction, enjoyment, tastes & feels like smoking
V.97-99: Ratings of negative effects: throat irritation, dry mouth, other negative effects
V.100: Filter to exclude baseline violations (see data processing description above).
V.101: Permutation_Actual – same as V.46 (copied over from Qualtrics)
V.102 – 106: Lists whether each of the five intervention components were ON or OFF for each participant
V.107: Computed Variable: Percentage reduction in cigarettes per day from baseline to follow up (baseline minus follow-up; higher percentages = greater reduction).
V.108: Computed Variable for secondary outcome (>50% reduction – yes or no) based on V.107.
V.109 – 110: Primary and secondary outcome intention to treat variables (4-weeks abstinence, 7-days abstinence) including non-responders as smoking.
V.111: Did the participant consent for their personal data to be shared with other researchers?
V.112: Did the participant respond to the primary outcome (4-weeks abstinence) question either via the survey or via text.
V.113-114: Re-coded occupation and ethnicity variables
V.115-116: Computed variables for secondary analysis to include those who quit or reduced vs. those who did not.