Effects of interventions on smoking cessation: A systematic review and network meta‐analysis

Abstract

A network meta‐analysis (NMA) including randomized controlled trials (RCTs) was conducted to evaluate the effects of different interventions on smoking cessation. Studies were collected from online databases including PubMed, EMBASE, Cochrane Library, and Web of Science based on inclusion and exclusion criteria. Eligible studies were further examined in the NMA to compare the effect of 14 interventions on smoking cessation. Thirty‐four studies were examined in the NMA, including a total of 14 interventions and 28 733 participants. The results showed that health education (HE; odds ratio ([OR] = 200.29, 95% CI [1.62, 24 794.61])), other interventions (OI; OR = 29.79, 95% CI [1.07, 882.17]) and multimodal interventions (MUIs; OR = 100.16, 95% CI [2.06, 4867.24]) were better than self‐help material (SHM). HE (OR = 243.31, 95% CI [1.39, 42531.33]), MUI (OR = 121.67, 95% CI [1.64, 9004.86]) and financial incentive (FI; OR = 14.09, 95% CI [1.21, 164.31]) had positive effects on smoking cessation rate than smoking cessation or quitting APP (QA). Ranking results showed that HE (83.6%) and motivation interviewing (MI; 69.6%) had better short‐term effects on smoking cessation. HE and MUI provided more smoking cessation benefits than SHM and QA. FI was more effective at quitting smoking than QA. Also, HE and MI were more likely to be optimal smoking cessation interventions.

1. INTRODUCTION

Smoking is a severe global public health problem. The World Health Organization's (WHO) report on the global tobacco epidemic, released on 27 July 2021, indicates that the global smoking rate among people over the age of 15 in 2019 is 17.51%. As of July 2021, an estimated of 847 million adult males, 153 million adult females and 24 million adolescents between the ages of 3 and 15 smoke. ¹ The number of deaths due to smoking is expected to rise to 8.3 million by 2030. ² This public health problem shortens life expectancy by an average of 10–11 years for those who smoke throughout their lives and causes more than 7 million deaths yearly. ³ , ⁴ Smoking (i.e., tobacco use) is a significant risk factor for diseases, such as malignant tumours (e.g., lung cancer), cardiovascular diseases (e.g., heart disease), respiratory diseases (e.g., chronic obstructive pulmonary disease), tuberculosis and stroke. ⁵ , ⁶ In addition, smoking places an enormous economic burden on societies. The global economic costs associated with tobacco use are estimated to exceed $1 trillion annually. In the United States, it costs nearly $300 billion annually in lost health care and productivity. ⁷ Fortunately, smoking cessation can reverse much of the damage, which dramatically improves life expectancy, reduces morbidity and decreases the medical costs associated with treating smoking‐related diseases. ⁸ Moreover, most smoking populations have a willing to quit smoking. ⁹ However, this is a challenge for them especially for those who are addicted to cigarettes, and most smokers make several attempts to quit before they succeed. ¹⁰ Currently, the smoking cessation interventions include drug‐assisted and non‐drug‐assisted ones. Studies have found that drug‐assisted intervention causes palpitations, chest pain, nausea and vomiting, insomnia and other adverse reactions. ¹¹ Non‐smoking cessation drug intervention includes financial incentives (FIs), short message service (SMS), and telephone counselling (TC). Because of the different content, intensity, approaches, and implementation methods across interventions, the effects of various interventions on smoking cessation also differ.

Heckman conducted a conventional meta‐analysis study and revealed that motivational interviewing (MI) was effective for smoking cessation in adolescents and adults. ¹² In another conventional meta‐analysis study, Scott‐Sheldon found that smokers who received SMS interventions were more likely to quit smoking (7‐day smoking rate) than the control groups. ¹³ Li et al, however, found that SMS intervention did not increase the possibility of smoking cessation (7‐day smoking rate) when compared with conventional intervention groups, ¹⁴ indicating a controversial effect of SMS on smoking cessation. However, these conventional meta‐analyses were conducted by collecting studies that evaluated the same treatment, defining treatment and control groups, and evaluating the effect of interventions. ¹⁵ One limitation of this approach is that they are not able to evaluate the contribution of each intervention to smoking cessation. ¹⁶ Unlike conventional meta‐analyses, network meta‐analysis (NMA) encompasses the synthesis of direct and indirect evidence to simultaneously compare and rank different interventions within a coherent treatment network. ¹⁷ Currently, most NMAs focus on the effects of drug‐assisted interventions on smoking cessation. ¹⁸ , ¹⁹ , ²⁰ , ²¹ Hartmann conducted an NMA evaluating the effect of non‐drug‐assisted interventions on smoking cessation, ²² but they included studies on populations with specific diseases and did not apply an uniform outcome indicator for smoking cessation. ²³ Therefore, NMA was used in the current study to unify outcome indicators and compare the effect of non‐smoking cessation drug intervention.

2. MATERIALS AND METHODS

This NMA was designed according to the guidelines for Preferred Reporting Items of Systems Review and Network Meta‐Analysis (PRISMA‐NMA). ²⁴

2.1. Guidance and search strategy

We searched PubMed, Web of Science, Embase, and Cochrane Library for randomized controlled trials (RCTs) evaluating smoking cessation interventions published before 10 November 2022. This search strategy is shown in Additional Document 1: Appendix S1.

2.2. Study selection

Two independent reviewers screened the titles and abstracts of publications retrieved by this search strategy to identify eligible studies. The full texts of the screened studies were evaluated according to the inclusion and exclusion criteria stated below. Disagreements between reviewers were resolved through discussions. The software Note Express was used to help manage the study selection.

2.3. Inclusion criteria

Type of participant: The smoking population.

Type of design: RCTs.

Types of interventions: Interventions include at least two or more of the following interventions: telephone counselling (TC), self‐help material (SHM), short message service (SMS), quitting email (QE), video messages (VMs), brief quitting advice (BQA), quitting app (QA), health education (HE), FIs, usual care (UC), video counselling (VC), MI, other intervention (OI) and multimodal intervention (MUI). We repeatedly mention some abbreviations here for readers' review. Each intervention is defined in Additional Document 1: Appendix S2.

Result type: 7‐day point quit rate (quitters who had not smoked in the 7 days before the follow‐up assessment) and a biochemical test reporting zero smoking with exhaled carbon monoxide concentrations of less than 10 parts per billion were used as indicators of smoking cessation.

2.4. Data extraction

A couple of reviewers independently extracted the following data: first author, year of publication, country, sample size, cigarettes per day, years of smoking, age, type of intervention and number of male and female participants.

2.5. Risk of bias assessment

Risk of bias was assessed by two reviewers independently and adjudicated by a third reviewer using the Cochrane Collaboration's tools, ²⁵ including sequence generation, assignment hiding, blinding, incomplete results data, non‐selective results reporting and other sources of bias. Each criterion was judged to have a low, unclear or high risk of bias.

2.6. Statistical analysis

Review Manager 5.3 was used to plot risk bias and perform the NMA. Stata 15.0 was used for grid element analysis under the consistency model. Binary variables were used as outcome variables and statistical significance was determined using odds ratios (ORs) and 95% confidence intervals (CIs). The ‘network plot’ function was used to generate network diagrams to present different forms of intervention mode. We used nodes to represent various interventions and edges to represent comparisons between interventions. Node splitting tests assessed local inconsistencies between direct and indirect evidence. The difference between the direct and indirect coefficients, via p values, was used to detect inconsistencies. The p < 0.05 indicated a local inconsistency, in which case non‐transitivity was suspected and potential influencing factors were examined. ²⁶ The effect of different interventions on smoking cessation was estimated based on the surface under the cumulative ranking (SUCRA) curve. Stata was used to sort the curative effect and draw the cumulative probability sorting chart to obtain the SUCRA. ²⁷

3. RESULTS

3.1. Literature selection

We initially searched 4974 relevant studies. After the deletion of 1991 studies based on their titles and abstracts, 2983 studies were selected. After reviewing the full text, 2848 studies were excluded with 135 remaining. Premised on inclusion and exclusion criteria, 101 of studies were excluded, and finally, 34 studies were included in our study. ²⁸ , ²⁹ , ³⁰ , ³¹ , ³² , ³³ , ³⁴ , ³⁵ , ³⁶ , ³⁷ , ³⁸ , ³⁹ , ⁴⁰ , ⁴¹ , ⁴² , ⁴³ , ⁴⁴ , ⁴⁵ , ⁴⁶ , ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵⁰ , ⁵¹ , ⁵² , ⁵³ , ⁵⁴ , ⁵⁵ , ⁵⁶ , ⁵⁷ , ⁵⁸ , ⁵⁹ , ⁶⁰ , ⁶¹ The research flow chart is shown in Figure 1.

FIGURE 1.

Flow chart.

3.2. Description of included studies

A total of 28 733 participants and 14 different interventions were included in these studies (Figure 2). The average ages of participants ranged from 16.2 to 47.2 years. Twenty‐three studies reported an average of more than 10 cigarettes per day in smoking population. Sixteen studies were from the United States, three from Australia, four from China, one from India, one from Mexico, three from the United Kingdom, two from the Netherlands, one from Switzerland, two from Norway and one from Spain. The characteristics of the studies and the participants are shown in Table 1. We assessed the quality of the included studies using the Cochrane risk of bias tool. Risk factors were classified into ‘high risk’, ‘low risk’ or ‘unclear’. The risk of bias is presented in Additional Document 1: Appendix S3.1–S3.2.

FIGURE 2.

The network constructed for all interventions.

TABLE 1.

The characteristics of the studies and the participants.

Study	Country	Year	Age (years E/C)	Exercise/control	Sample size	Cigarettes per day	Follow‐up time
Danaher et al. 28	USA	2015	37.91 ± 8.2	OI/TC/MUI/SHE	1683	1.85 ± 1.3 pack	3 months
Krishnan et al. 29	USA	2013	NA	FI/OI/MUI	73	NA	1 months
Kumar et al. 30	India	2012	20–40	MUI/SHE	400	NA	2 months
Danaher et al. 31	USA	2019	45.6 ± 12.3/44.2 ± 12.9	SMS/QE	1271	NA	3 months
Stanczyk et al. 32	Netherlands	2014	45.7 ± 12.8	VM/SMS/BQA	2551	17.0 ± 7.8	1 months
Severson et al. 33	USA	2009	30.4 ± 7.6	MUI/BQA	785	NA	3 months
Pbert et al. 34	USA	2021	16.9 ± 1.1	QA/SHE	121	5	6 months
Catley et al. 35	USA	2016	45.8 ± 10.9	MI/HE/BQA	256	17.1 ± 8.9	6 months
Abroms et al. 36	USA	2014	35.7 ± 10.7	SMS/SHE	1745	17.29 ± 8.08	6 months
Colby et al. 37	USA	2012	16.2 ± 1.3	MI/BQA	162	NA	1 months
Chan et al. 38	China	2016	35.7 ± 6.9/35.4 ± 7.1	TC/BQA	1158	14.1 ± 8.7/13.0 ± 7.1	6 months
Etter and Schmid 39	Switzerland	2016	32 ± 11/32 ± 11	FI/SHM	805	16 ± 9	3 months
Scheffers‐van Schayck 40	Netherlands	2021	39.2 ± 7.20	TC/SHM	83	15.5 ± 6.67	3 months
Toll et al. 41	USA	2010	47.2 ± 13.4 46.4 ± 13.9	TC/UC	1282	20.1 ± 11.2/ 20.1 ± 11.0	3 months
Chan et al. 42	China	2015	NA	TC/SMS/SHM	1003	NA	6 months
Sood et al. 43	Mexico	2009	42.84 ± 13.46/42.48 ± 13.29	TC/SHM	990	23.19 ± 12.58/22.01 ± 12.61	3 months
Buller et al. 44	Australia	2014	24.9	QA/SMS	102	17	6 weeks
Abdullah et al. 45	China	2005	NA	MUI/SHE	952	14.5 ± 8.9	6 months
Byaruhanga et al. 46	Australia	2021	43.7 ± 11.8	VC/TC/SHM	655	NA	3 months
Wangberg et al. 47	Norway	2011	37.3/36.9	OI/QE	2072	16.1/16.2	3 months
Gram et al. 48	Norway	2019	39	SMS/QE	4335	NA	6 months
Tzelepis et al. 49	Australia	2011	45.4 ± 12.7/ 44.4 ± 13.8	TC/SHE	1532	19.9 ± 9.6/ 18.9 ± 9.9	4 months
Lipkus et al. 50	USA	2004	NA	TC/SHM	302	10 ± 8	4 months
Míguez and Becoña 51	Spain	2008	37.1 ± 9.6	TC/SHM	228	27.1 ± 10.5	3 months
Klemperer et al. 52	USA	2017	51 ± 11	MI/OI	560	>10	6 months
Smith et al. 53	USA	2001	41.7 ± 12.0/42.1 ± 12.1/41.1 ± 9.9	OI/MI/BQA	370	21.4 ± 9.6	6 months
Hokanson et al. 54	USA	2006	54 ± 9/53 ± 9	MI/UC	114	22.0 ± 9.8/ 19.7 ± 12.0	6 months
McClure et al. 55	USA	2005	32.7 ± 11.7	TC/UC	275	14.3 ± 6.8	6 months
Glasgow et al. 56	USA	2000	24 ± 5	MI/BQA	1033	12 ± 7	6 months
Tappin et al. 57	UK	2005	26.5 ± 5.8 26.9 ± 6.6	MI/HE	762	27/27.5	NA
Colby et al. 58	USA	2005	16.3 ± 1.5	MI/BQA	85	10.5 ± 7.3	6 months
Shi et al. 59	China	2013	17.6 ± 1.1/16.9 ± 0.7	SMS/SHM	179	NA	12 weeks
Naughton et al. 60	UK	2014	41.8 ± 13.0	UC/SMS	602	12.8	8 weeks
Naughton et al. 61	UK	2012	NA	SMS/BHM	207	NA	3 months

Abbreviations: BQA, brief quitting advice; FI, financial incentive; HE, health education; MI; motivational interviewing; MUI, multimodal intervention; OI, other intervention; QA, quitting app; QE, quitting email; SHM, self‐help material; SMS, short message service; TC, telephone counselling; UC, usual care; VC, video counselling; VM, video message.

3.3. Inconsistency test

A global inconsistency test was performed, and no difference was found (p = 0.69). In addition, to better explore the local inconsistencies, a node‐splitting test was conducted. A total of 30 paired comparisons involving loops were analysed, and the results showed statistically significant inconsistencies between FI and MUI (p < 0.001) as well as FI and SHM (p = 0.001; Table 2). In addition, we made pairwise comparisons of forest plots (Additional Document 1: Appendix S4) and loop inconsistency (Additional Document 1: Appendix S5).

TABLE 2.

Local inconsistency test based on side‐split.

Side	Direct		Indirect		Difference		P > |z|
	Coef.	Std. Err	Coef.	Std. Err	Coef.	Std. Err
BQA‐HE	2.152	1.373	0.301	0.775	1.851	1.667	0.267
BQA‐MI	0.499	0.366	1.176	0.804	−0.677	0.874	0.439
BQA‐MUI	1.338	0.686	0.972	0.512	0.366	0.857	0.699
BQA‐OI	0.121	0.526	0.679	0.528	−5.558	0.746	0.454
BQA‐SMS	0.361	0.689	0.358	0.473	0.002	0.835	0.998
BQA‐TC	0.460	0.714	0.430	0.443	0.030	0.840	0.971
BQA‐VM	0.271	0.690	0.267	1.519	0.005	1.670	0.661
FI‐MUI	3.225	1.167	−1.689	0.629	4.913	1.287	0.000
FI‐OI	0.163	1.400	−1.496	0.771	1.659	1.562	0.288
FI‐SHM	−2.451	0.594	1.979	1.219	−4.430	1.357	0.001
HE‐MI	−0.116	0.570	−2.299	3.195	2.182	3.263	0.504
MI‐OI	−0.056	0.526	−0.474	0.677	0.418	0.857	0.626
MI‐UC	−0.302	0.762	−0.042	0.620	−0.260	0.983	0.791
MUI‐OI	−0.767	0.589	−0.635	0.618	−1.132	0.857	0.877
MUI‐SHM	−0.616	0.394	−1.997	0.620	1.381	0.734	0.060
MUI‐TC	0.113	0.647	−0.994	0.433	1.107	0.778	0.155
OI‐QE	−0.413	0.677	−0.154	0.666	−0.259	0.949	0.785
OI‐SHM	−0.294	0.683	−0.358	0.468	0.064	0.828	0.939
OI‐TC	0.066	0.682	0.025	0.485	0.041	0.836	0.961
QA‐SHM	0.195	0.836	−0.335	0.658	0.530	1.064	0.618
QA‐SMS	0.003	0.586	0.533	0.888	−0.530	1.064	0.618
QE‐SMS	0.176	0.475	0.435	0.822	−0.259	0.950	0.785
SHM‐SMS	0.283	0.369	0.318	0.470	−0.035	0.597	0.659
SHM‐TC	0.363	0.259	0.452	0.613	−0.890	0.665	0.660
SHM‐VC	0.847	0.808	0.897	1.475	−0.050	1.723	0.655
SMS‐TC	0.094	0.688	0.076	0.369	0.018	0.780	0.662
SMS‐UC	−1.999	0.674	0.299	0.530	−0.450	0.857	0.560
SMS‐VM	−0.889	0.686	−0.840	1.524	−0.005	1.670	0.998
TC‐UC	0.264	0.486	−0.320	0.596	0.584	0.770	0.448
TC‐VC	0.494	0.781	0.444	1.518	0.051	1.724	0.655

Abbreviations: BQA, brief quitting advice; FI, financial incentive; HE, health education; MI; motivational interviewing; MUI, multimodal intervention; OI, other intervention; QA, quitting app; QE, quitting email; SHM, self‐help material; SMS, short message service; TC, telephone counselling; UC, usual care; VC, video counselling; VM, video message.

3.4. Network meta‐analysis

Figure 3 shows an estimated effect from the NMA of the overall quit rates for each smoking cessation intervention. HE (OR = 200.29, 95% CI [1.62, 24 794.61]), OI (OR = 29.79, 95% CI [1.07, 828.17]) and MUI (OR = 100.16, 95% CI [2.06, 4867.24]) showed a better pain improvement than SHM. HE (OR = 243.31, 95% CI [1.39, 42 531.33]), MUI (OR = 121.67, 95% CI [1.64, 9004.86]) and FI (OR = 14.09, 95% CI [1.21, 164.31]) showed a better pain improvement than QA.

FIGURE 3.

The results of network meta‐analysis for all pairwise comparisons.

3.5. Probability ranking

Figure 4 shows the SUCRA for all interventions. For self‐reported 7‐day smoking cessation rate, SUCRA was able to predict the likelihood of different interventions and acted as the reference for selecting the best treatment. The results showed that HE was most likely to be the best intervention (83.6%). For OIs, the effectiveness was ranked as follows: MI (69.6%), VC (67.1%), UC (62.1%), SMS (56.1%), VM (53.0%), TC (51.8%), OI (49.7%), MUI (48.1%), BQA (44.1%), SHM (41.0%), QE (38.1%), FI (31.4%) and QA (4.3%).

FIGURE 4.

Self‐reported 7‐day point prevalence abstinence of total effective rate.

3.6. Publication bias

As shown in Figure 5, the examined research was roughly symmetrical on both sides of the median line, indicating that the small sample effect was unlikely to exist.

FIGURE 5.

Funnel chart comparing the effectiveness of 14 interventions on smoking cessation. A = BQA, B = FI, C = HE, D = MI, E = MUI, F = OI, G = QA, H = QE, I = SHM, J = SMS, K = TC, L = UC, M = VC, N = VM.

4. DISCUSSION

This NMA reported the effect of 14 types of interventions on smoking cessation, including BQA, FI, HE, MI, MUI, OI, QA, QE, SHM, SMS, TC, UC, VC and VM, which indicated that both HE and MUI yielded more smoking cessation benefits than SHM and QA. MI is a patient‐centred form of communication designed to reinforce motivation and commitment to behaviour change so as to avoid confrontation or persuasion. ⁶² In addition, MI can target critical elements in behaviour change by focusing on cognitive processes and shifting the perceived costs, thereby contributing to the changes of unhealthy behaviours. ⁶³ , ⁶⁴ The clinical practice guidelines for the treatment of tobacco dependence recommend MIs for smokers who express low motivation to quit after receiving advices. ⁶⁵ HE adopts flexible means and approaches of health publicity to improve the smokers' awareness of the harm of smoking so as to encourage them to quit smoking. Interestingly, we found FI to be more effective than QA. One study found that FIs of $750 and $800 increased quit smoking rates among educated and relatively affluent employees of large American companies and remained unchanged for 6 months after the final incentives. ⁶⁶ , ⁶⁷ While FIs were effective, most previous studies used small incentives, and it remained unclear whether they have a lasting effect. ⁶⁸

In addition, the results of this study showed that VC, UC, SMS and VM had certain advantages in smoking cessation, and HE and MI tended to benefit smoking cessation. In the NMA, interventions such as smoking cessation counselling, FIs and the use of computer equipment to quit smoking showed better effects on smoking cessation. Still, the impact of professional HE and MIs remains unknown, which required additional investigation.

5. LIMITATIONS OF THE STUDY

The limitations of the current study need noting. Specifically, NMA was conducted only for the 7‐day smoking cessation rate, and the effect of intervention parameters was not evaluated, which might affect the robustness of the results. ⁶⁹ Moreover, the included RCTs varied in the durations of the follow‐up assessments and intervention per se. In general, most RCTs reported a follow‐up period of more than 6 months, but for smoking cessation, the follow‐up period should be extended longer. Also, different intervention durations would affect the results and introduce bias to the NMA results. ⁷⁰ In addition, the network formed by all interventions shows that most interventions were indirectly compared in these studies, which indicates that the number of direct comparisons between interventions may not be sufficient. ⁷¹ , ⁷² At the same time, there is a high risk of bias in the design and implementation of several included studies, which may affect the robustness of our findings. Furthermore, our study included some recently published studies, in which the long‐term smoking cessation effects might not be analysed. Because our study adopted one outcome indicator, other indicators of smoking cessation should be considered and evaluated in the future.

6. CONCLUSION

This NMA showed that HE and MUI provided more smoking cessation benefits than SHM and QA. We also found that FI was more effective at quitting smoking than QA. Confirmed evidence suggested that HE and MI tended to be the best interventions for smoking cessation.

AUTHOR CONTRIBUTIONS

All authors contribute concepts, search terms and methods of review. Li Ying came up with the idea. Ying Li and Jianhua Wang designed the study. Ying Li and Lei Gao completed the database search and study selection. Tianci Qin and Lei Gao completed an assessment of bias in included studies. Ying Li, Lingyu Hou and Jianhua Wang extracted data from the included studies. Tianci Qin and Ying Li conducted the meta‐analysis. Ying Li wrote the first draft. Xiaoan Chen completed the critical revision of the manuscript. Yaqing Chao and Xiaohua Zhou provided technical guidance and participated in the revision process of the article. In addition, they provided financial support. All authors contributed to the writing or correction of the final manuscript. All authors have read and approved the final version submitted to this newspaper.

CONFLICT OF INTEREST STATEMENT

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. All authors have no conflict of interest disclosures.

Supporting information

Appendix S1. Supporting Information.

Appendix S2. Supporting Information.

Appendix S3.1. Supporting Information.

Appendix S3.2. Supporting Information.

Appendix S4. Supporting Information.

Appendix S5. Supporting Information.

Li Y, Gao L, Chao Y, et al. Effects of interventions on smoking cessation: A systematic review and network meta‐analysis. Addiction Biology. 2024;29(3):e13376. doi: 10.1111/adb.13376

DATA AVAILABILITY STATEMENT

Data sharing is not applicable to this article as no datasets were generated or analysed during the current study.