We conducted a 2-arm, double-blind, RCT in five cities in China (Beijing, Dezhou, Baotou, Yakeshi, and Linzi). Participants were randomized to an intervention group or a control group between April 2021 and July 2021. Daily or weekly smokers 18 years or older who smoked more than 5 cigarettes per week were eligible for inclusion if they owned a mobile phone and used WeChat. We advertised the trial to smokers through paper advertisements (leaflets), digital advertisements (WeChat), and staff (teachers and leaders). Potential participants contacted the local Center for Disease Control and Prevention to register. All eligible smokers were told they needed to come to a specific place on a fixed date to finalize the recruitment process. Participants’ eligibility was double-checked, and they signed an informed consent form during the first face-to-face contact.

The trial was approved by the Ethics Committee of Peking University Health Science Center (IRB00001052-30063). All the participants signed informed consent forms before randomization and knew that they could withdraw from the study at any time. All patients’ information was accessible only to the personnel participating in the study. We did not provide money to the participants, but we provided gifts (a towel, an umbrella, or a cup) if the participants completed one follow-up visit. The clinical trial registration number is ChiCTR2100041942.

The first stage was to develop the intervention message bank. Messages were developed by Peking University, School of Public Health, with the input of smokers and smoking cessation professionals. The messages had a three-layer framework. The first layer was divided based on time and consisted of the prequit message (1-7 days), quit day message (8 days), withdrawal symptom management message (9-18 days), early quit message (19-36 days), and late period message (37-90 days). The second layer was divided based on the TTM. Before the quit day, messages were classified as strong quitting intention or weak quitting intention. After the quit day, messages were classified as maintained abstinence or relapsed. The third layer was divided based on the PMT. Messages were classified as increased severity and susceptibility, decreased response cost and intrinsic and extrinsic rewards, and increased self-efficacy and response efficacy. The core motivational messages consisted of 14 subgroups with a total of 200 SMS text messages. There were also approximately 200 contact messages. The participants received partial information, and the information they received was different depending on their status. The WeChat app evaluated smokers’ status on day 0, day 19, day 36, day 45, day 60, and day 75 to decide what messages to send them.

All messages were sent through the WeChat app (the most popular Chinese mobile chatting app) [4]. The WeChat app evaluated the PMT construct score by asking questions and recording information; then, it automatically calculated the lower score of the subcontent that needed to be strengthened. Specifically, the scale comprised 21 items using a 7-point Likert-type scale with responses ranging from 1 (definitely disagree) to 7 (definitely agree). Each construct subscale included three items, and we computed the mean as the subscale score. We have published the details of this scale and evaluation process elsewhere [15-18]. To evaluate smokers’ quitting intention, we used a 5-point scale ranging from not at all likely (1) to very likely (5) regarding the likelihood that they would try to quit in the next 6 months (1-3: weak intention; 4-5: strong intention) [19]. A more detailed framework of the text and intervention based on the TTM and PMT is reported separately [5,10].

The last stage was to deploy the message library on the WeChat platform by using IT. Our IT team completed the development process with several important considerations. First, the app needed to ensure the confidentiality of the data. Second, the system needed to be user-friendly with ease for quick data entry. Third, there needed to be a back-end server to store the data. Fourth, the information needed to be presented in a way that could be easily read and interpreted.

After recruitment, the participants were required to complete the baseline questionnaire and register through WeChat. A randomized block design was used, and the score of the Fagerström Test for Nicotine Dependence was treated as a stratified factor. The WeChat system automatically generated two blocks by the score of the test: (1) low or moderate nicotine dependence group (0-6 points) and (2) high dependence group (≥7 points) [20]. Eligible participants were assigned to the intervention group or the control group within each block by simple randomization. Independent IT personnel created a randomization sequence but were not involved in the research. The WeChat system was also used to balance demographic characteristics. Randomization was fully computerized and automated with equal allocation. The researchers and participants were all blinded.

All participants were informed that the eighth day after randomization would be their quit day. Participants who were allocated to the intervention group received the program of interventions. As described above, this consisted of 1-2 personalized SMS text messages per day for 3 months.

Control group participants received a nonpersonalized SMS text message smoking cessation intervention developed by the National Cancer Institute. It was based on well-established cognitive behavioral cessation approaches and contained 91 messages. Participants received approximately 1 message per day for 3 months. The SMS text messages provided encouragement, practical advice to help maintain cessation, and information on the health effects of smoking. The details of the control group can be found elsewhere [7].

All participants in the two groups were instructed to attend face-to-face follow-ups with research staff 1 month, 3 months, and 6 months after randomization. The outcomes were the change in PMT construct scores and the change in TTM stages. TTM stages of change were measured by smoking status using the question “Are you going to quit smoking?”: “No, not thinking of quitting” (precontemplation; individual is not ready or thinking about making a change); “Yes, within the next 6 months” (contemplation; individual thinking about making a change but not in the immediate future); “Yes, within the next 30 days” (preparation; individual ready to change, intends to try to change in the immediate future, and might be making small preparatory changes); and action, the biochemically verified patient has not smoked since the last follow-up [21,22].

This study engages in the debate in recent years over the effects of mobile cessation interventions. It provides a more comprehensive picture of the impact of personalized mobile cessation interventions from psychological change perspectives using Chinese data. It successfully shows that the intervention group had lower scores for the subconstructs of intrinsic rewards of smoking, extrinsic rewards of smoking, and response cost of quitting, and was more likely to promote smokers’ movement from the preparation stage to the action stage, confirming that personalized interventions can provide more positive psychological changes than a nonpersonalized SMS text message intervention. We believe that this finding is worth noting, as we extend the findings from target behavior to psychological change (first hypothesis).

Regarding the second hypothesis, although we found that the psychological determinants varied at different stages to facilitate smokers’ movement from precontemplation to the action period, self-efficacy was the most significant variable associated with forward movement in each stage of change. Self-efficacy is the belief in one’s competence to manage adversity in specific demanding situations. A branch of the literature has suggested that self-efficacy plays an important role in determining health behavior; for example, a German study found that self-efficacy was the strongest factor that significantly predicted subsequent smoking-related behavioral intention [23]. This finding was also confirmed by other health behavioral studies. A meta-analysis reviewed 65 studies and suggested that coping appraisal variables and especially self-efficacy were the strongest predictors of protection motivation and behavior [24].

Another finding of our study is the identification of the psychological determinants at different stages to facilitate smokers moving forward to the next stage. The comparison between groups showed that the personalized intervention decreased the intrinsic rewards of smoking; this outcome is worth noting since it is a determinant of the movement from the preparation stage to the action stage. It also provides some indications of why a personalized intervention is more likely to promote smokers from the preparation stage to the action stage than a nonpersonalized intervention. Conversely, we found that our intervention did not lead to a significant change among smokers in the determinants that promoted smokers in moving forward from the precontemplation stage (self-efficacy and response efficacy) or contemplation stage (self-efficacy) to the next stage compared with the control group. This finding explains why our intervention had no better effect on promoting smokers to move forward from the precontemplation or contemplation stage, providing further direction for scaling up our intervention.

As stated, this RCT is the first to use personalized SMS text messages for a mobile cessation intervention with a positive control group design in China. We believe that our study has policy and theoretical implications. First, we identified psychological determinants of forward movement in the stage of change, which could be an important reference for further intervention. Second, this study demonstrated a clear framework for the intervention through the systematic and transparent application of the PMT and the TTM. It not only allows other researchers to take advantage of our experience when designing mobile interventions but also provides a framework for exploring why such an intervention is effective.

This study still has some limitations that could be addressed by further studies. First, although efforts were made to ensure that both the researchers and participants remained masked to the allocation, a risk of breaking the blinding was present. Second, some researchers believe that the definitions used for the TTM stages are arbitrary and that the categories are not qualitatively distinct [25,26]. Third, the scale that we used to evaluate PMT subconstructs was the same at each follow-up time; although there were intervals of a few months, some participants might have remembered some of the questions, providing inaccurate answers.

This study reported clearly on the development of a mobile cessation intervention. GEE analysis identified the psychological determinants of forward movement in the stage of change and confirmed that a personalized mobile cessation intervention was more likely to promote smokers from the preparation stage to the action stage. The results of this empirical analysis could not only be equally applicable to the development of interventions targeting other health behaviors but also could provide a framework to explore why such an intervention is effective, thereby adding to earlier research on this topic.