Effects of Web- and Computer-Based Smoking Cessation Programs:  Meta-analysis of Randomized Controlled Trials FREE

Smoking is the single greatest cause of preventable disease and premature death. It was estimated that 4.9 million premature deaths due to smoking occurred worldwide in 2000.¹ To date, recommended smoking cessation strategies include brief tobacco-dependence treatment; individual, group, and telephone counseling; numerous effective medications; and telephone quitline counseling, according to an updated “Clinical Practice Guideline” for treating tobacco use and dependence published by the US Public Health Service in 2008.²

Besides conventional approaches, some studies have found that computer- or Web (Internet)-based smoking cessation programs are effective in randomized controlled trials (RCTs). However, the findings from these trials are inconsistent.^3- 24 Thirteen RCTs reported that an intervention group consisting of a Web- or computer-based smoking cessation program did not yield a significantly higher abstinence rate compared with a control group,^{3- 4,8- 10,13,15,17- 20,23- 24} whereas 9 RCTs found a significant effect on abstinence up to 2.5 times higher than in the control group.^{5- 7,11- 12,14,16,21- 22} Even though review articles^25- 26 have been published, to our knowledge no meta-analysis has been conducted to date. We examined the effects of Web- and computer-based smoking cessation programs in RCTs via a meta-analytic approach.

We conducted electronic searches in MEDLINE (PubMed) (1968 to August 2008), EMBASE (1977 to August 2008), and the Cochrane Central Register of Controlled Trials (CENTRAL) in the Cochrane Library (1953 to August 2008) by using common keywords related to Web- or computer-based smoking cessation programs in RCTs. We also searched the bibliographies of relevant articles to identify additional studies. The keywords used for the literature search were as follows: smoking cessation, quit smoking, or tobacco cessation; Internet program, computer program, online program, or Web program; and trial or intervention. The language of publication was not restricted.

We included RCTs that reported the effects of a Web- or computer-based smoking cessation program for current smokers if they included at least 3 months of follow-up data. Trials involving smokeless tobacco (ST) users and quasi-experimental trials were excluded from this study. The principal outcome measures included point-prevalence abstinence, sustained abstinence, prolonged abstinence, and continuous abstinence. Biochemical validation was not required in the current study (see the limitations in the “Comment” section).

Two of us (S.-K.M. and D.D.M.), independently assessed the eligibility of all studies retrieved from the databases. Disagreements were resolved by discussion. From the studies included in the final analysis we extracted the following data: the study name (along with the name of first author and the year of publication), the journal name, country, the recruitment period and follow-up period (in years), the number of participants by condition, the content of intervention and control conditions, the outcome measures, abstinence rates, relative risk (RR) with 95% confidence intervals (CIs) calculated from the numbers of the 4 cells of the 2 × 2 tables of each of the studies, and the number of quitters compared with the number of participants and percentage lost to follow-up in both intervention and control groups.

We assessed the methodological quality of the trials based on a validated scale for RCTs developed by Jadad et al.²⁷ This 5-point quality scale includes points for randomization (described as randomized, 1 point; table of random numbers or computer-generated randomization, additional 1 point), double-blind (described as double-blind, 1 point; use of masking such as identical placebo, additional 1 point), and follow-up (the numbers and reasons for withdrawal in each group are stated; 1 point) in the report of an RCT. In general, scores of 2 or less are considered as low quality, and scores of 3 to 5 are considered as high quality. In this study, however, we considered a score of 2 or more as high quality and a score of 1 or less as low quality because none of the studies were double-blinded, which is difficult to do with this type of research using a Web-based intervention.

Subgroup analyses were performed to explore heterogeneity. We evaluated the effects of smoking cessation programs according to the type of intervention used, that is, Web- or computer-based program. Web-based programs were those delivered over the Web, and computer-based programs were computerized, computer-tailored, computer-assisted, or computer-generated programs. The effects of those programs were further evaluated according to the use of supplemental intervention (stand-alone or supplemental intervention such as nicotine replacement therapy, bupropion hydrochloride, and individual counseling by a counselor), age group (adolescents or adults), follow-up period (3 months, 6-10 months, or 12 months), type of abstinence measure (point prevalence or sustained), and rate of loss to follow-up (high vs low based on a median split).

To investigate the influence of each individual study on the overall summary estimate, we conducted an influence analysis. In this analysis, the meta-analysis estimates were computed, omitting 1 study in each turn. We also conducted a meta-regression analysis to evaluate associations between the intervention effect and study characteristics such as type of intervention, abstinence measure, follow-up duration, age group, and follow-up frequency during the study.

Statistical analyses were conducted based on 2 × 2 tables from individual study results on the basis of an intention-to-treat analysis using the number of participants and quitters in each intervention and control group. If there was no event (ie, quitters) in one of the intervention or control groups (ie, a “zero cell” in the 2 × 2 table), 0.5 was added to each cell of the table so that the estimated RR would not be zero or infinity and the standard error could be calculated. To estimate heterogeneity across studies, we used Higgins I², which measures the percentage of total variation across studies due to heterogeneity. We calculated I² as follows:

I² = 100% × (Q − df)/Q,

where Q is the Cochran heterogeneity statistic and df is the degrees of freedom. Negative values of I² are set to zero so that I² lies between 0% (no observed heterogeneity) and 100% (maximal heterogeneity). An I² value greater than 50% was considered substantial heterogeneity.

We estimated the pooled RR with 95% CI on the basis of both the fixed- and random-effects models. Because we observed heterogeneity in most analyses, we reported the pooled RR with 95% CI on the basis of the random-effects models. In the random-effects models, the DerSimonian and Laird²⁸ method for calculating summary measures was used.

Publication bias was evaluated using the Begg funnel plot and Egger test, summarized elsewhere.²⁹ If publication bias exists, the funnel plot is asymmetrical or the P value is found to be less than .05 by Egger test. We used Stata SE software (version 10.0; StataCorp, College Station, Texas) for statistical analysis.

Figure 1 shows a flow diagram of how we selected relevant studies. A total of 287 articles were identified after searching 3 databases and hand-searching relevant bibliographies. After excluding 117 duplicated articles and 120 articles that did not satisfy the selection criteria described in the “Methods” section, we reviewed the full texts of 50 articles. Among those, 28 articles^30- 57 were excluded because they lacked sufficient data (n = 3),^30- 32 did not satisfy the follow-up period requirement (n = 5),^33- 37 were quasi-experimental trials (n = 3),^38- 40 were not RCTs (n = 3),^41- 43 reported data that were redundant with another study (n = 3),^44- 46 involved ST users (n = 2),^47- 48 or did not fulfill other selection criteria (n = 9).^49- 57 We included 22 RCTs^3- 24 in the final analysis.

The 22 RCTs included a total of 29 549 participants: 16 050 randomized to a Web- or computer-based smoking cessation program and 13 499 to one of the control groups. In the studies in which age and sex were reported, the mean age was 38 years (range, 11-75 years), and 58.9% of participants were women.

Table 1 shows the general characteristics of the 22 RCTs included in the analysis. The selected trials were reported from 1989 through 2008, spanning 19 years. The countries in which the studies were conducted were as follows: the United States (n = 13),^{3,5,7- 9,13- 16,18,20- 21,23} the United Kingdom (n = 4; 1 trial involved the Republic of Ireland),^4,12,17,19 Australia (n = 2),^10- 11 Germany (n = 1),²⁴ Norway (n = 1),²² and Switzerland (n = 1).⁶ The range of follow-up periods was 12 weeks to 12 months. Study participants included medical center employees; students in school year 9, high schools, or colleges; purchasers of nicotine gum or nicotine patch; people from a general population or community; callers to a quitline service; employees at a worksite; patients in practices or clinics; visitors to Web sites; and participants in health surveys.

Of the 22 trials, 10 used supplemental interventions for smoking cessation, such as counseling, classroom lessons, nicotine replacement therapy (gum or patch), bupropion medication, or quitlines,^{4- 5,9,12- 14,19- 20,22- 23}whereas 12 trials only used the Web- or computer-based smoking cessation program.^{3,6- 8,10- 11,15- 18,21,24} Nine used a Web-based smoking cessation program,^{12- 16,18,20- 22} and 13 used a computer-based smoking cessation program.^{3- 11,17,19,23- 24} Eight reported point-prevalence abstinence rates,^{3- 4,13,16,18,20,22- 23} and 14 reported sustained abstinence rates.^{5- 12,14- 15,17,19,21,24} Seven reported abstinence at short-term follow-up (3 months),^{5,9,12,14,16- 18} 8 reported abstinence at midterm follow-up (6-10 months),^{3,6,13,15,19- 21,23} and 7 reported abstinence at long-term follow-up (12 months).^{4,7- 8,10- 11,22,24} Nineteen had adult participants,^{3,5- 13,16- 24} and 3 had adolescents.^4,14- 15 The rates of loss to follow-up ranged from 4.6% to 58%.

In the intervention group using Web- or computer-based smoking cessation programs, the abstinence rate was significantly greater compared with the control group in both a fixed-effects (RR, 1.35; 95% CI, 1.27-1.44) and random-effects (RR, 1.44; 95% CI, 1.27-1.64) (Figure 2) model for all 22 trials with significant heterogeneity (I² = 60.7%). There was no publication bias in the selected studies (the Begg funnel plot was symmetrical; for the Egger test, P for bias = .21). In the pooled analyses, the smoking cessation rate was 14.8% (95% CI, 14.1%-15.6%) in the intervention group and 14.3% (95% CI, 13.4%-15.2%) in the control group at short-term follow-up (3 months) (z test; P = .42); 11.7% (95% CI, 10.5%-12.9%) and 7.0% (95% CI, 6.1%-8.0%) at midterm follow-up (6-10 months) (z test; P < .001); and 9.9% (95% CI, 8.9%-10.9%) and 5.7% (95% CI, 5.1%-6.3%) at long-term follow-up (12 months) (z test, P < .001), respectively (Table 2).

As shown in Figure 3, the abstinence rate in the intervention group was significantly greater than in the control group, both in trials using a Web-based (RR, 1.40; 95% CI, 1.13-1.72; I² = 62.7%; n = 9) and those using a computer-based (RR, 1.48; 95% CI, 1.25-1.76; I² = 60.1; n = 13) smoking cessation program. Table 3 shows the effects of Web- and computer-based programs in subgroup analyses by methodological quality, use of supplemental intervention, age group, follow-up period, type of abstinence rate, and rate of loss to follow-up.

The effect of the Web- or computer-based intervention was statistically significant in both the high-quality (RR, 1.48; 95% CI, 1.18-1.85; I² = 67.9%; n = 10) and low-quality trials (RR, 1.42; 95% CI, 1.20-1.68; I² = 57.0%; n = 12). Stand-alone programs demonstrated a significant positive effect on smoking cessation (RR, 1.59; 95% CI, 1.23-2.05; n = 12), as did programs with supplemental interventions (RR, 1.31; 95% CI, 1.16-1.47; n = 10). However, a large difference of I² values from the test for heterogeneity was observed between these analyses (70.0% for stand-alone and 35.4% for supplemental intervention).

Regarding age group, the Web- or computer-based smoking cessation programs obtained a significantly greater abstinence rate for adults (RR, 1.49; 95% CI, 1.31-1.70; I² = 58.2%; n = 19) but not for adolescents (RR, 1.08; 95% CI, 0.59-1.98; I² = 65.3%; n = 3).

Also, subgroup analyses showed significant effects of the Web- and computer-based interventions on smoking cessation regardless of the follow-up period, the type of abstinence rate, or the rate of loss to follow-up (see Table 3 for corresponding summary RRs and 95% CIs). An influence analysis showed that omission of any individual study made no difference in summary RRs and 95% CIs when its graph was visually interpreted (the graph is not shown owing to space limitations). Also, a metaregression analysis showed no significant findings (P>.05 for all of the included study characteristics, eg, type of intervention, abstinence measure, follow-up duration, age group, and follow-up frequency during the study).

This meta-analysis of 22 RCTs found that Web- or computer-based smoking cessation programs yielded an abstinence rate about 1.5 times higher than the control group (RR, 1.44; 95% CI, 1.27%-1.64%). When data were pooled, the abstinence rate at the 12-month follow-up was significantly higher in the intervention group (9.9%; 95% CI, 8.9%-10.9%) than in the control group (5.7%; 95% CI, 5.1%-6.3%), as well as at 6-month follow-up (see Table 2 for P values). The stand-alone interventions had a significant effect on smoking cessation as well as on those that had supplemental interventions. However, compared with adults, these programs did not significantly increase the abstinence rate in adolescent populations (RR, 1.08; 95% CI, 0.59-1.98).

Our findings imply that there is sufficient evidence to support the use of a Web- or computer-based smoking cessation program for adult smokers. As of June 2008, 1.4 billion people (21.9% of the world's population) were estimated to use the Web.⁵⁸ Moreover, Web usage grew more than 300% from 2000 to 2008.⁵⁸ As global Web users continue to increase, Web-based smoking cessation programs could become a promising new strategy that is easily accessible for smokers worldwide.²⁷ According to an updated guideline for treating tobacco use and dependence published in 2008,² individually tailored Web-based materials such as interactive Web sites are recommended (ie, “Strength of Evidence = B”²) for treating tobacco dependence; those interventions are not clearly mentioned in the “Ten key guideline recommendations” for treating tobacco dependence.² This meta-analysis could provide important evidence to recommend the use of Web- or computer-based interventions for smoking cessation as one of the key guideline recommendations.

The effect of Web- or computer-based interventions was similar to that of counseling interventions. For trials reporting smoking cessation for follow-up durations of 6 months or longer, the odds ratio (OR) for successful smoking cessation of Web- or computer-based smoking cessation programs in our meta-analysis of 15 RCTs was 1.53 (95% CI, 1.28-1.82; I² = 51.8%; not shown in the “Results” section); the OR of counseling interventions in a published meta-analysis of 17 RCTs was reported to be 1.56 (95% CI, 1.32-1.84).⁵⁹ In addition, the smoking cessation rates after a follow-up of 6 months or longer for Web- or computer-based interventions were also similar to that of intensive counseling interventions from specialists reported in a previous review: 10.7% (95% CI, 9.9%-11.4%) for the Web- or computer-based interventions and 10% (the 95% CI was not reported) for the intensive counseling interventions.⁶⁰

We found that Web- or computer-based smoking cessation programs had no effect on smoking cessation for adolescent smokers. Because our analysis included only 3 RCTs for this age group, more trials are needed. However, our findings were consistent with a meta-analysis by Grimshaw and Stanton⁶¹ of tobacco cessation interventions for adolescents and young adults published in 2006. That study reported that neither the meta-analysis of pharmacological intervention trials nor cognitive behavior therapy intervention trials achieved statistically significant results, although subgroup analyses suggested that a few programs may be effective. The nonsignificant effect of the smoking cessation interventions for adolescent or young adult smokers might be the result of high attrition or low statistical power. High attrition would contribute to the nonsignificant effect because participants lost to follow-up are considered as smokers in an intention-to-treat analysis. Also, wide CIs in both the current meta-analysis and the meta-analysis by Grimshaw and Stanton⁶¹ contribute to the nonsignificant effect. Therefore, more research with large sample sizes is needed to confirm the effectiveness of the smoking cessation interventions among adolescent or young adult smokers. The effects of the stand-alone Web- or computer-based interventions were heterogeneous, whereas those offered with supplemental, especially pharmacologic, interventions were relatively homogeneous.

We could not evaluate overall socioeconomic status of the participants owing to poor information. When we excluded studies for adolescents (n = 3) and college students (n = 3) from the 22 studies in our study, of the remaining 16 studies, 5 studies^5- 9 reported the mean number of years that participants attended school, ranging from 12.7 to 14.2 years, and 4 studies^10,13,18,22 reported the percentage of at least some college or technical school ranging from 28% to 75%; the remaining 6 studies^{3,11- 12,16- 17,19} did not report the educational status of the participants. Only 2 studies^5,9 reported the annual income of participants ($34 000 and $39 000, respectively).

Our study has several possible limitations. First, our findings could not be applied to ST users. We did not evaluate Web- or computer-based programs for ST users because the literature was limited, and the purpose of our study was to evaluate the effects of Web- and computer-based smoking cessation programs. Also, because the program effects were mainly evaluated in the general population, further research is needed to investigate program effects for patients with specific conditions. Second, despite the overall positive effects of Web-based smoking cessation programs, some study authors discussed disadvantages of online health behavior change programs, including the potential for privacy and security breaches, low quality of information, and problems with computer system performance.²⁵ Standardized, high-quality Web-based smoking cessation programs should be developed in the future. Further, because smokers who are elderly, less educated, or living in developing countries may not be able to use a computer or access the Web, their access to these programs would be problematic, and the potential effectiveness of these programs for such populations is unknown. Third, we did not consider biochemical validation techniques such as urinary cotinine concentration or expired carbon monoxide for the estimation of abstinence rates as selection criteria, although in most settings, biochemical verification provides additional assurance that the participant's self-reports are accurate.⁶² However, it has been reported that biochemical validation is considered not to alter the conclusions of low-intensity interventions trials.⁶³ Also, the Society for Research on Nicotine and Tobacco Subcommittee on Biochemical Verification mentioned that the added precision gained by biochemical verification is offset in such a way that its use is not required and may not be desirable in studies in which the optimal data collection methods are through the mail, telephone, or Web.⁶² Furthermore, most of the smoking cessation program trials using the Web or computer did not use biochemical verification. Last, although we may have missed some articles because we did not search a psychiatry-specific database, we believe that this is unlikely because we reviewed relevant bibliographies to supplement the database searches we did.

In sum, we found a notable effect on smoking cessation for Web- and computer-based smoking cessation programs in our meta-analysis of RCTs published since 1989. The programs increase the smoking cessation rate about 1.5 times more than in the control group and obtain an abstinence rate at 12-month follow-up of 9.9%. To our knowledge, this is the first quantitative meta-analysis that evaluates the effect of these programs; we found sufficient clinical evidence to support the use of Web- and computer-based smoking cessation programs for adult smokers.

Correspondence: Seung-Kwon Myung, MD, MS, 111 Jungbalsan-ro, Ilsandong-gu, Goyang, Gyeonggi-do 410-769, South Korea (msk@ncc.re.kr).

Accepted for Publication: February 8, 2009.

Author Contributions: Dr Myung, as principal investigator, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Myung. Acquisition of data: Myung and McDonnell. Analysis and interpretation of data: Myung, McDonnell, Kazinets, Seo, and Moskowitz. Drafting of the manuscript: Myung, McDonnell, and Seo. Critical revision of the manuscript for important intellectual content: Myung, McDonnell, and Moskowitz. Statistical analysis: Myung and Kazinets. Obtained funding: Moskowitz. Administrative, technical, and material support: Myung and McDonnell. Study supervision: Myung, Seo, and Moskowitz.

Financial Disclosure: None reported.

Funding/Support: Drs Myung, McDonnell, Kazinets, and Moskowitz receiving funding from the Centers for Disease Control and Prevention through Cooperative Agreement U48/DP000033.

Disclaimer: This article's contents are solely the responsibility of the authors and do not necessarily represent the official views of the Centers for Disease Control and Prevention.

This article was corrected online for error in data on 5/25/2009, prior to publication of the correction in print.