The full protocol, alongside sample size calculations, has been published previously [22].
Design and setting
This single-center pragmatic three-arm randomized controlled trial was conducted in Sheffield, UK; recruitment took place between June 2017 and July 2019.
Participant recruitment
Participants were recruited from the community [22]. Eligible were smokers who had smoked ≥ 10 cigarettes per day for the last year, were aged ≥ 18, and willing to make a cessation attempt using a stop-smoking service or e-cigarettes.
Exclusion criteria were (i) inability to walk; (ii) recent (within 6 months) CVD events or cardiac surgery; (iii) insulin-controlled diabetes mellitus; (iv) coexisting skin conditions, leg ulcers, vasculitis, or deep venous occlusion; (v) pregnancy; (vi) major surgery scheduled during the study; (vii) contraindications /unsuitability for NRT; (viii) current daily use of e-cigarettes; and (ix) current cessation attempt supported by a stop-smoking service.
Procedures
Following a telephone pre-screening and information about study procedures, participants were invited to Sheffield Hallam University to provide informed consent and undertake baseline assessments. They were enrolled by a researcher not involved in group allocation, intervention delivery, or assessments. Following baseline, participants were randomized remotely into three groups by an independent statistician using a computer-generated (nQuery Advisor 6.0, Statistical Solutions, Ireland) block-randomization stratified by gender and “pack-years” (number of packs (20 cigarettes per pack) per day times number of years smoked). The study statistician allocated a unique trial number to each participant for the study duration.
Outcome assessors were blinded to group allocation and participants were reminded regularly not to share their group allocation with assessors or those providing behavioral support. The study statistician/health economist was blinded to group allocation. Those delivering the intervention were only blinded in relation to which e-cigarette group the participants belonged to as the NRT group was receiving support through the stop-smoking service. No blinding existed for participants, as NRT was delivered separately and as experienced users of nicotine, the e-cigarette groups were able to determine the presence or absence of nicotine.
During their initial behavioral support sessions, participants set a “quit date” to stop smoking completely. This defined timing of follow-up visits.
Intervention
Nicotine-containing e-cigarettes group and nicotine-free e-cigarettes group
Behavioral support was offered for a 3-month period through the research team. This was necessary as e-cigarettes were not part of the treatments offered by the local stop-smoking services at the time of the study. Both groups received complimentary e-cigarette equipment and refills (Tornado V5, Totally Wicked, Blackburn, UK), together with instructions on correct usage. Participants, on average, received 20 bottles of 10 ml during the intervention period. The nicotine-containing e-cigarette group received liquids with nicotine strength of up to 18 mg/ml, the nicotine-free e-cigarette group received nicotine-free liquids (0 mg/ml). Participants could choose ice menthol or tobacco flavor (Red Label, Totally Wicked, Blackburn, UK).
Prescription NRT group
Following allocation, participants were referred to Sheffield stop-smoking services for behavioral support for three months. They received money or shopping vouchers (depending on personal preference) as reimbursement for NRT prescription charges.
To ensure comparability of behavioral support provided, all groups received the same level and type of behavioral support as currently offered as standard by stop-smoking services, in the form of regular face-to-face or telephone appointments as per relevant guidelines, e.g., minimum of 6 support sessions within the 3-month period [23]. All advisors had completed the same behavioral support training. To support the successful participation of our participants, we used our “six pillars of adherence” framework (based upon “social support,” “education,” “reachability,” “small groups intervention implementation,” “reminders,” and “simplicity”). None of the participants experienced side effects due to their participation in the intervention.
Measures
Age, gender, carbon monoxide (CO—to confirm smoking status), body mass index, blood pressure, number of cigarettes and years smoked, and physical activity measured using the SF-IPAQ [24] were recorded at baseline. Body mass index, CO, blood pressure, and physical activity assessments were repeated at 3 and 6-month follow-up.
Primary outcome
Macrovascular function was assessed using percentage change in flow-mediated dilation. Smoking is associated with reduced FMD. Reduced altered brachial artery FMD is an early marker for endothelial dysfunction, a CVD risk factor [6], and considered a predictor for long-term, adverse cardiovascular events [25]. FMD is a non-invasive, nitric oxide-mediated measure. Baseline scanning to assess resting vessel diameter was recorded over 3 min, following a 10-min resting period, using a Nemio XG (Toshiba, Tokyo, Japan) ultrasound machine, according to the International Brachial Artery Reactivity Task Force guidelines [26]. Fixed anatomic landmarks (side branches) were used to ensure that the same artery portion was assessed on each occasion. The technical error in our lab for FMD is 5% [26]. FMD is presented as change in post-stimulus diameter as a percentage of the baseline diameter (%FMD).
Secondary outcomes
Upper-body microvascular function was assessed using peak cutaneous vascular conductance (CVC) responses to acetylcholine (ACh) and sodium nitroprusside (SNP) as indicators of microvascular endothelial-dependent and endothelial-independent vasodilation, respectively, measured using Laser Doppler Fluximetry and Iontophoresis, using a standardized procedure [27].
Cutaneous red cell flux was measured by placing an iontophoresis laser Doppler probe (PF481-1; Perimed AB), connected to a laser Doppler fluxmeter (PF5001; Perimed AB), in the center of each of two drug delivery electrodes ((PF383; Perimed AB, Jarfalla, Sweden) positioned over healthy-looking skin, approximately 4 cm apart with one containing 100 μL of 1% ACh (Miochol-E, Novartis, Stein – endothelium-dependent vasodilator) and the other 100 μl of 1% SNP (Nitroprussiat, Rottapharm – endothelium-independent vasodilator). Measurements of red cell flux (recorded in arbitrary units, AU) were divided by corresponding MAP values (in mmHg) to give CVC in AU/mmHg. Here we present ACh and SNP peak CVC responses. The technical error of measurement for drug-induced peak flux responses in our laboratory is 15% [28].
Mean arterial pressure (MAP) was calculated using the following formula:
$$\frac{\left(2\times Diastolic\;pressure\right)+Systolic\;pressure}3$$
MAP was not included in the original list of outcomes as a separate outcome. However, it was decided to be analyzed and presented separately, due to its common use as surrogate marker for arterial stiffness [28]. All cardiovascular outcomes (following, micro- and macro-vascular assessments) were assessed by the same researcher, at all time-points.
Additional outcome measures include the following: (i) quality of life and health economics; preferences were calculated using EQ-5D-5L Index values and visual analog scale (VAS) scores, and the five discrete EQ-5D dimensions (mobility, self-care, usual activities, pain/discomfort, anxiety/depression) were assessed [29]; (ii) Q-Risk to assess risk of CVD; [30] and (iii) “Finger prick” test to calculate the total cholesterol (TC) over high-density lipoprotein (HDL) ratio [31].
All outcomes were assessed with participants abstaining from caffeine, main meals, supplements, and e-cigarettes and smoking for at least 3 h prior to the assessments. Premenopausal women were studied during days 1–7 of their menstrual cycle to minimize the influence of cyclical changes in female hormones and strenuous exercise was avoided for at least 12 h prior to the assessments. Outcomes were assessed at baseline and repeated at 3 and 6 months after the quit date. Resource use (e.g., GP practice and clinic visits) and personal out-of-pocket expenditure (e.g., travel for appointments, prescriptions, and products purchased to quit smoking) were collected via participants’ diaries maintained from baseline to the end of intervention (3 months) period using standard procedures.
Sample
In total, 248 participants were randomized (Fig. 1). Data of four were removed following their withdrawal from the study (1 in the nicotine-containing, 1 in the nicotine-free, 2 in the NRT group). The 3-month follow-up was completed by 207 participants, the 6-month follow-up by 202 participants. Participants were included whether they stopped smoking or not.
Statistical analysis
Data were analyzed in SPSS 24 (IBM U.K. Limited, Hampshire, UK) on an intention-to-treat basis, using the “last observation carried forward” method for those participants that had no data for the 3- and 6-month follow-ups, respectively.
We developed a multiple linear regression model for our repeated continuous outcome measures (%FMD, ACh, SNP, MAP), using a generalized estimating equations (GEE) approach. We made this change from the ANOVA specified in the protocol as GEE is more robust to misspecification of the dependency structure, which results from the repeated assessments of individual over time [32]. GEE used age (in completed years), gender and baseline measurements (or 3-month values for change over 6-month analysis) to control for and predict net-changes in these outcome measures over the 3- and 6-month periods, across our three randomized groups. To explore the interaction between our repeated assessments and our treatment groups, we introduced treatment groups dummy variables, which were assessed over each time-point (i.e., 3 and 6 months), using an established methodology as described in Twisk et al. [33]. Our analysis was also varied against our original protocol, in that we decided against including pack years in the model, due to the excellent balance between groups; age was included instead. A separate analysis using mixed effect models repeat measurement (MMRM) was completed after using multiple imputation for missing values for four outcome variables at 3- and 6- month follow-ups; findings did not differ to those obtained through the GEE modeling.
As additional post hoc analyses, separate GEE models were run for the “cessation” sub-group, which included those that stated that they had stopped smoking and had their status biochemically validated by exhaled air measurement of < 10 ppm CO in line with the Russell Standard [34].
Unadjusted mean differences (d) were calculated as the mean change from baseline to a given time point, for the %FMD, ACh, SNP, and MAP.
For health-related quality of life (HRQoL), mean values for the three participant groups were compared, and the net changes from baseline to 3 and 6 months were estimated. Also, the proportion of patients with sub-optimal levels (i.e., levels 2–5) on each of the five dimensions was compared at baseline, 3, and 6 months. Quality-adjusted life-years (QALYs) were calculated by multiplying the duration of time spent in various health states by the HRQoL weight (i.e., utility score) associated with that health state. A similar approach (i.e., paired t-tests used to compare differences in means at two points in time (3 and 6 months over baseline values) across three quit-smoking groups) was chosen for the other secondary outcomes, including Q-Risk and the total cholesterol (TC) over high-density lipoprotein (HDL) ratio.
Health economics were assessed using participants’ resource use diaries alongside costs for the delivery of 12 weeks of behavioral change intervention (researcher’s time, room hire charges, and travel expenses), and costs for the intervention materials (electronic cigarette, adapter, atomizer & liquids). NHS smoking cessation delivery costs were estimated to include clinic nurse time and the number of sessions delivered to NRT participants. Participants’ use of NHS services was priced using 2020 PSSRU costings or Department of Health reference costs [35].
Role of the funding source
The trial was funded by Heart Research UK under a Translational Research Grant (RG2658). The funder had no role in the study’s design, conduct, and reporting.