Study design, setting and participants

This two-arm pilot randomized controlled trial proactively recruited community smokers from smoking hotspots in Hong Kong, where ashtrays were available and smokers gathered to smoke, including vicinities of public transport stations, housing estates, and outside shopping malls. University students were trained as SC ambassadors to approach and recruit smokers. Using a ‘foot-in-the-door’ approach²⁵, SC ambassadors proactively approached smokers at the smoking hotspots and asked about their smoking behaviors (e.g. daily cigarette consumption, history of quit attempts), assessed their exhaled carbon monoxide level, and invited them to join the study to quit or reduce smoking. Interested smokers were then screened for eligibility for participation. Inclusion criteria were: 1) being a Hong Kong resident aged ≥18 years; 2) having smoked ≥1 cigarette daily in the past 3 months, validated by an exhaled carbon monoxide level of ≥4 ppm²⁶ (using a Smokerlyzer); 3) having a smartphone; and 4) having an instant messaging app (e.g. WhatsApp). Individuals who were using SC medication or other SC services, had any contraindication for NRT (e.g. arrhythmia, myocardial infarction, pregnancy) or were physically or mentally unable to communicate in Cantonese or Putonghua, were excluded. Table 1 shows the detailed schedule of enrolment, interventions, and assessments of the study.

Table 1

The schedule of enrolment, interventions, and assessments

	Enrolment	Allocation	Month 1	Month 2	Month 3	Month 6
Enrolment and baseline intervention
Eligibility screen	×
Informed consent	×
Brief SC advice	×
NRT sampling	×
Active referral to SC service	×
Allocation		×
Interventions
Intervention group (chat-based IM support)
Control group (regular text messages)
Assessment
Sociodemographic characteristics a	×
Smoking behavior	×		×	×	×	×
Self-efficacy on quitting	×				×	×
Intervention satisfaction					×	×
Biochemically validated abstinence					×	×

[i] SC: smoking cessation, NRT: nicotine replacement therapy, a Sociodemographic characteristics include sex, age, marital status, co-living with children, education level, employment status, and monthly household income.

Randomization and blinding

The allocation sequence was generated using an online tool (https://www.sealedenvelope.com/simplerandomiser/v1/lists) with a block size randomized among 2, 4 and 8 in the 1:1 allocation ratio and concealed using a sequentially numbered, opaque and sealed envelope (SNOSE). The co-investigator (WY) prepared the SNOSE for group assignment. All envelops were labelled with serial numbers and SC ambassadors were concealed from the allocation sequence. Group allocation was determined by opening the SNOSE on site once the consent form was signed. Masking of the interventionist is not possible due to the nature of the interventions. But the participants were not informed about the intervention in the other group. Outcome assessors were blinded to the group allocation.

Intervention group

Participants in the intervention group received brief SC advice, 1 week of NRT (gum, patch or lozenge) sampling, and were actively referred to SC services at baseline upon completion of the baseline questionnaire. NRT-S (14/21 mg patch, 2 mg gum, and 1/2 mg lozenges) was provided according to their preferences and daily cigarette consumption. Participants who consume ≤20 cigarettes daily received the 14 mg patch, 2 mg gum, or 1 mg lozenges while participants who consume >20 cigarettes daily received 21 mg patch, 2 mg gum, or 2 mg lozenges²⁴. The brief SC advice was delivered using the AWARD (Ask, Warn, Advise, Refer, Do-it-again) model in which participants were asked about smoking behaviors (Ask); warned about smoking hazard using a health-warning leaflet (Warn); advised to quit or reduce smoking as soon as possible and to set an initiation date (Advice); actively referred to SC services for standard treatment (Refer); and repeat the intervention during the chat-based interaction and follow-up calls (Do-it-again; see below). The AWARD process usually took one to two minutes to complete and has been validated in community smokers^6,7. Among smokers who were willing to try NRT-S²⁷, an NRT use card containing the instructions and potential side effects were given with a brief oral explanation¹⁷. Ambassadors introduced the SC services (free-of-charge) using a 2-sided color printed information card and actively referred the participants to their preferred SC services. With a separate consent form signed, names and telephone contacts of the agreed participants were sent to the service provider for appointment booking of clinical treatment within a week of enrollment (active referral).

Participants received two months of fix scheduled personalized regular messages combined with one-to-one interactive mobile chat support via instant messaging after enrolment. The mobile chatbased intervention was conducted by a trained SC counsellor with two years of experience in smoking cessation research. Messages were sent twice per week in the first four weeks and once per week for the following four weeks. Two additional follow-up reminders were sent before the follow-up call at 3 and 6 months, a total of 14 regular messages. Each message was personalized according to sex, age, and smoking pattern of the participants. The contents of the messages included benefits of quitting, tips for quitting, coping with craving, and the effectiveness of SC services. To initiate and facilitate the interactive conversations, some simple questions related to their smoking behavior were asked after sending the informative message, e.g. ‘How many cigarettes have you smoked today?’. Participants engaged in the interaction received synchronous feedback and personalized behavior support. Counsellors also monitored participants’ quitting progress, advised about the safety and effectiveness of NRT, and checked for NRT side effects via the interaction. All incoming messages were replied as soon as possible during working hours (9:30–18:30) on weekdays. For non-engaged participants, 3 additional prompt messages were sent in the first month to encourage engagement. Messages included asking their reasons for continuing smoking, history of quit attempts and attitudes towards successful quitters. The instant messaging dialogues were recorded and checked, by a research nurse trained in smoking cessation, biweekly to ensure intervention fidelity.

The frequency and content of the messages were informed by our formative qualitative study¹³. The motivational interviewing (MI) behavior change model underpinned the design of the regular messages and the conduct of chat-based support. MI model focused on enabling change through the enhancement of intrinsic motivation and the exploration and resolution of ambivalence²⁸. Consultation strategies adopted in the intervention included identifying discrepancies between participants’ thought and action, supporting their autonomy, and positive encouragement. Two participants who refused to receive instant messaging were contacted via SMS messages or telephone calls to remind them of follow-up calls.

As an extension of the AWARD model, participants were encouraged to use SC services, helped with rebooking appointments and actively referred to SC services (for those who refused at baseline) during the chat-based interaction (Refer and Do-it-again). At the 1 and 2 months follow-up, a booster intervention of brief (one to two minutes) SC advice was delivered through telephone. Upon completion of the followup questionnaire (Ask about smoking behaviors), non-quitters were warned about the health risks of continue smoking (Warn), instructed about the use of NRT, monitored for potential side effects of NRT, and advised to quit again or reduce smoking (Do-itagain).

Control group

Participants in the control group also received brief SC advice, NRT-S and active referral at baseline. During the following 2 months, participants only received general SC messages through SMS with frequency similar to the regular instant messaging received by the intervention group. The SMS contents included brief SC advice, tips for coping with craving and reminders for follow-up calls²⁹. Counsellors did not respond to any messages from participants and no telephone booster was given at follow-up.

Measurements

Outcomes

The primary outcome was biochemically validated abstinence, defined by an exhaled carbon monoxide level of <4 ppm (by Smokerlyzer³¹) and saliva cotinine concentration of <10 μg/L (by NicAlert test strip³²) at 3 and 6 months. Secondary outcomes included self-reported 7-day point prevalence abstinence (PPA), smoking reduction by at least 50% of baseline cigarette consumption (quitters included), quit attempts (abstinence for ≥24 h), NRT-S use and SC service use since the baseline. Participants in both groups received a small cash incentive of HK$200 (about US$25.6) for passing each validation at 3 and 6 months, which was found to have no effect on abstinence in our previous trial³³.

Sample size

In this feasibility study with integrated interventions of brief advice, NRT-S, active referral and sustained chat-based interaction, we aimed to recruit as many participants as resources allowed during December 2017 and March 2018 to generate preliminary estimates for the intervention efficacy. Analyses (post hoc) were conducted using the effect size of the primary outcome to estimate the power of sample we achieved.

Statistical analysis

Intention-to-treat (ITT) analysis was conducted by treating participants with missing outcome measures as having no changes in smoking behaviors from baseline³⁴. Differences in baseline demographic characteristics and smoking behavior between the intervention and control groups were compared, using t-tests, Wilcoxon rank-sum test, chi-squared tests or Fisher’s exact test, and controlled in subsequent analysis. We performed logistic regressions to estimate the intervention effect (adjusted odds ratios, AOR) for primary and secondary outcomes. We tested the differences in NRT-S use, quit attempt, smoking reduction, and self-reported quitting by intervention engagement, defined by having interaction in the IM chat-based intervention (verified by conversation log), controlling for sex and perceived importance of quitting at baseline³⁵. The absolute number in some outcomes was relatively small (i.e. n<5), suggesting the analysis may be subjected to the sparse data issue. We thus conducted post hoc sensitivity analyses by incorporating weakly informative prior to penalize regression estimates and calculating profile likelihood-based CIs. Change-in-estimate procedure [ΔOR = (ORpenalization - ORoriginal)/ORoriginal×100%] was adopted to compare the ORs obtained from the penalization analysis (model 3) with the ORs in adjusted regression model (model 2) (Supplementary file Tables S2 and S3). A value of p<0.05 indicated statistical significance in the 2-tailled analysis. All analyses were conducted in Stata (version 15.1).

Participant characteristics

From December 2017 to March 2018, 164 potential participants were screened for eligibility, 121 were found eligible and provided written consent (Figure 1). Two participants did not complete the baseline questionnaires and 119 were randomized (62 in the intervention group and 57 in the control group). Most (94.2%) participants had secondary or above education, were employed (83.7%) and had a monthly household income of more than HK$20000 (72.4%) (Table 2). Half were married and one-third (33.7%) were living with children. Table 3 shows that more than half of the participants had a low level of nicotine dependence (56.4%), had past quit attempts (61.9%) but were not ready to quit within one month (55.1%), perceived quitting as important but difficult and had median confidence in quitting. Most participants have received the NRT-S at baseline (77.3%). Baseline characteristics were similar between the two study groups except that the intervention group had a greater proportion of men (88.7% vs 77.0%, p=0.02) and a higher score in perceived importance of quitting (median: 8% vs 7%, p=0.01) than the control group. The overall retention rate was 68.9% at 3 and 6 months.

Figure 1

Study flow diagram

SC outcomes

Among the 14 and 13 self-reported quitters at 3 and 6 months, respectively, only 3 (21%) and 1 (7.7%) participated in the face-to-face biochemical validation. The intervention group showed non-significant higher validated abstinence than the control group of 3.2% vs 1.8% (OR=1.87; 95% CI: 0.16–21.16) at 3 months, and 1.6% vs 0% at 6 months. In model 2, we adjusted for unbalanced variables between groups (sex and perceived importance of quitting). The intervention group (vs the control group) reported more quitting at 3 (12.9% vs 10.5%; AOR=1.12; 95% CI: 0.34–3.71) and 6 (16.1% vs 5.3%; AOR= 2.82; 95% CI: 0.70–11.30) months. The smoking reduction and cessation rate was also greater in the intervention group at 6 months prior change-in-estimates (<10%), suggesting that (32.3% vs 19.3%; AOR=1.74; 95% CI: 0.71–4.26).

Quit attempts, use of NRT-S and SC services

Consistently, more participants in the intervention group reported quit attempts (90.3% vs 77.2%; AOR=2.61; 95% CI: 0.88–7.82) and SC service use (3.2% vs 1.8%; AOR=1.33; 95% CI: 0.10–18.19). Significantly more NRT-S use (62.9% vs 38.6%; AOR=2.27; 95% CI: 1.04–4.96) was observed in the intervention group. Results from sensitivity analyses (Supplementary file Tables S2 and S3) show that the observed estimates remained statistically stable and largely unaffected by the weakly informative prior change-in-estimates (<10%), suggesting that estimates are with moderate to high precision and not overly sparse across the comparisons (Table 4).

Intervention engagement

Table 5 shows that among the 62 participants in the intervention group, 21 (33.9%) had interacted with the counsellor via instant messaging (verified by instant messaging dialogue). ‘Too busy’, ‘not interested’ or ‘too boring’ were reported as main reasons for non-usage by participants. Engaged in mobile chat support appeared to be associated with a consistent increase in NRT-S use (76.2% vs 56.1%; AOR=3.21; 95% CI: 0.84–12.28), quit attempt (100% vs 85.4%), smoking reduction (42.9% vs 26.8%; AOR=1.65; 95% CI 0.49–5.57) and self-reported quit rate (23.8% vs 12.2%; AOR=2.00; 95% CI: 0.45–9.01) at 6 months compared to non-engaged participants, although statistically non-significant.