## INTRODUCTION

According to the 2017 World Health Organization report1-5, tobacco smoking remains a major worldwide public health threat, with >7 million deaths directly related to tobacco. Many studies have recognized smoking as a risk factor for chronic diseases, such as chronic respiratory diseases (asthma, chronic obstructive pulmonary disease), hypertension, cardiovascular disease, atherosclerosis, diabetes, cancer, and microbial infections (respiratory infections, bacterial meningitis)6,7. Smoking also burdens healthcare systems and increases social costs. Given the impact of cigarette smoking, the development of effective interventions to address tobacco addiction is a major public health need. There are 350 million smokers in China, which accounts for one-third of the world’s smokers8. Unfortunately, smoking cessation services and counseling are at an early stage of development in China. Moreover, healthcare workers in China do not exert much effort in helping smokers to quit tobacco use9. Therefore, the effectiveness of existing smoking cessation interventions and services are largely unknown in China.

Numerous prior studies have examined individuallevel predictors of successful and unsuccessful cessation attempts2,10,11, including socio-economic status groups, increasing tobacco prices, both peer and family smoking groups, and use of smoking cessation medications, but researchers have come to differing conclusions. In addition, another study shows that it is important to understand the differential roles that pre-quitting and post-quitting experiences play in smoking cessation and to provide help to smokers for not resuming cigarette smoking12. It is important to understand the characteristics of smokers and identify factors predicting successful smoking cessation to improve the efficacy of interventions. Some smokers are unwilling (i.e. not ready, not motivated, or not able) to attempt quitting in the near future, so identification of predictors and determinants of success in smoking cessation is a key component in smoking cessation programs. Therefore, we assessed the outcomes of smokers in our smoking cessation clinic and investigated factors predictive of successful smoking cessation treatment. The primary objective of this study was to develop a valid but simple prediction tool by using only characteristics easily determined at the beginning of treatment to assess the factors associated with successful smoking cessation, with the goal of enabling physicians in smoking cessation clinics to provide individualized treatment strategies.

## METHODS

### Study participants

A total of 278 smokers treated at the smoking cessation clinic of Ningbo First Hospital from March 2016 to December 2018 were enrolled in the study. The inclusion criteria were current smokers (smoked daily for ≥12 months at the time of the survey), aged ≥18 years, motivated to quit, and willing to participate in the follow-up visits. Additionally, smokers whose intention to quit was not clear at the first visit but who had a desire to quit after smoking cessation counseling were also included in the study group. The exclusion criteria were smokers who did not want to participate in a cessation program even after smoking cessation counseling and smokers who were unwilling or unable to receive regular follow-up. The present study was performed with the informed consent of each subject and with the approval of the local Ethics Committee of Ningbo First Hospital (Ningbo, China).

### Feature selection

Based on 278 patients in the cohort, 25 features were reduced to four potential predictors (Figures 1A and B), and the coefficients were non-zero in the LASSO regression model: reason for quitting smoking, number of other smokers in the household, number of visits to the outpatient department, and varenicline use (Table 2).

##### Table 2

Factors predictive of successful smoking cessation

βOR (95% CI)p
Intercept1.26933.558 (1.164–11.132)0.027
Reason for quitting smoking
Mobilization of others vs prevention and treatment of own diseases−2.10060.122 (0.056–0.255)<0.001
Others vs prevention and treatment of own diseases−2.20110.111 (0.049–0.239)<0.001
Living with a smoker or being exposed to workplace smoking
yes vs no
−0.82200.439 (0.219–0.878)0.020
Number of outpatient department visits
≥2 vs 1
1.05262.865 (1.439–5.844)0.003
Varenicline use
yes vs no
0.69432.002 (0.825–4.999)0.128

[i] OR: odds ratio. β: regression coefficient.

##### Figure 1

Demographic and clinical feature selection using the LASSO binary logistic regression model

### Generation of an individualized prediction model

Based on the multivariate analysis results, predictive factors, including the reason for quitting smoking, living with a smoker or being exposed to workplace smoking, number of visits to the outpatient department, and varenicline use, were incorporated into the nomogram; these characteristics are shown in Table 2. A model containing the above independent predictors was established and is represented as a nomogram in Figure 2.

##### Figure 2

Nomogram to predict the probability of quitting smoking

### Apparent performance of the nomogram for prediction of successful smoking cessation

The calibration curve of the nomogram for prediction of successful smoking cessation showed good consistency (Figure 3). The C-index of the predictive nomogram was 0.816 (95% CI: 0.761– 0.871) for this cohort and was confirmed by internal validation as 0.804, indicating that this model had good discriminatory ability. The nomogram had good power for predicting success in quitting smoking.

##### Figure 3

Calibration curves of nomogram prediction of successful smoking cessation in the cohort

### Clinical value of the model

The DCA for predicting the success of quitting smoking is shown in Figure 4. The decision curve shows that the use of this nomogram increased the ability to predict successful smoking cessation when the patient and physician threshold probabilities were 19% and 92%, respectively. In this range, according to the successful smoking cessation nomogram, the net benefit was comparable to several overlaps.

##### Figure 4

Decision curve analysis for the nomogram predicting successful smoking cessation

## DISCUSSION

In the analysis of predictors of quitting smoking and reasons for quitting, living with a smoker or being exposed to workplace smoking, number of outpatient department visits, and varenicline use were associated with successful cessation rate. This nomogram suggested that treatment with varenicline, quitting for health-related reasons, more visits, and not coexisting with other smokers may be key factors that determine the success of smoking cessation.

The efficacy data in our study showed that varenicline was more effective than bupropion for smoking cessation. However, there is strong evidence from multiple randomized clinical trials that both bupropion and varenicline increased smoking cessation rates when used in a quitting attempt17,18. One potential explanation of this discrepancy between our results and those of the trials is that the low-usage rate of bupropion may have resulted in underestimation of its potential efficacy.

To our surprise, the number of outpatient department visits was the most influential factor affecting smoking cessation. In addition, we found that the reason for quitting smoking could predict the success of the attempt. These results indicate that individual motivation, especially intrinsic motivation, was predictive of the smoking cessation result. These results are in accordance with those of many reports from the medical literature12,19-21, which suggest that smokers subjectively recognize the harm of smoking and the benefits of quitting smoking and that taking the initiative to quit smoking is very important for success. From a pooled estimate of 65 trials, Hartmann–Boyce et al.22 concluded that increasing the amount of behavioral support is likely to increase the chance of success by approximately 10% to 20%. Smokers with more outpatient visits may be able to obtain more behavioral and psychosocial support, thus achieving better smoking cessation results.

As the results show, not co-existing with other smokers also was a predictor of smoking cessation success. This finding is consistent with the results of previous reports that living with a smoker or being exposed to workplace smoking made individuals less likely to quit23-26. It is possible that exposure to other people smoking decreases quitting rates and increases the risk of starting to smoke23. Smoking is not only a personal behavior in China, which has a high smoking rate, but also deeply influenced by social factors. Smokers who are often surrounded by other smokers perceive higher approval and acceptance of smoking behavior, thus further strengthening smoking behavior. From this point of view, we should pay more attention to decreasing passive smoking. It should be noted that age, sex, education, occupation, and health status were not predictive factors for the success of smoking cessation in our study. Although the investigators observed that some of these factors were independent predictors21,27,28, the findings of our study appeared to be inconsistent with some published evidence and could not confirm all the previous findings. This inconsistency may be because of differences between countries and regions. Some studies have found that measuring exhaled CO levels were a useful biomarker for predicting successful smoking cessation3-5. However, in our study, exhaled CO levels at the first visit were not associated with success in quitting. In addition, neither the average number of cigarettes smoked daily nor the FTCD scores related to nicotine dependence were associated with the success of smoking cessation. This finding differs from the results in previous studies19, 29,30. In their study, Huang et al.5 found that smokers with lower FTCD scores, those with lower exhaled CO concentrations, and those who smoked <20 cigarettes per day on average, had higher success rates. These differences in conclusions may be because our follow-up time was limited, the self-reports may have underestimated cigarette consumption, and only the first measurements of CO levels were compared.