Effect of Smoking Reduction Therapy on Smoking Cessation for Smokers without an Intention to Quit: An Updated Systematic Review and Meta-Analysis of Randomized Controlled Trials

Objective: Effective strategies are needed to encourage smoking cessation for smokers without an intention to quit. We systematically reviewed the literature to investigate whether smoking reduction therapy can increase the long-term cessation rates of smokers without an intention to quit. Methods: PubMed, Embase, and CENTRAL (Cochrane Central Register of Controlled Trials) were searched for randomized controlled trials (RCTs) on the effect of smoking reduction therapy on long-term smoking cessation in smokers without an intention to quit. The primary outcome was the cessation rate at the longest follow-up period. A random effects model was used to calculate pooled relative risks (RRs) and 95% confidence intervals (CIs). Results: Fourteen trials with a total of 7981 smokers were included. The pooled analysis suggested that reduction support plus medication significantly increased the long-term cessation of smokers without an intention to quit compared to reduction support plus placebo (RR, 1.97; 95% CI, 1.44–2.7; I2, 52%) or no intervention (RR, 1.93; 95% CI, 1.41–2.64; I2, 46%). In a subgroup of smokers who received varenicline or nicotine replacement therapy (NRT), the differences were also statistically significant. This suggests the safety of using NRT. The percentage of smokers with serious adverse events who discontinued because of these events in the non-NRT group was slightly significantly different than in the control group. Insufficient evidence is available to test the efficacy of reduction behavioural support in promoting long-term cessation among this population. Conclusions: The present meta-analysis indicated the efficacy of NRT- and varenicline-assisted reduction to achieve complete cessation among smokers without an intention to quit. Further evidence is needed to assess the efficacy and safety of reduction behavioural support and bupropion.


Introduction
Smoking is still one of the largest preventable causes of death in the world. Although smoking cessation can reduce the chance of developing smoking-related diseases, 75.6% of Chinese smokers have no plan to quit smoking [1]. A similar phenomenon has been observed in other developed countries [2,3]. These results reveal the necessity for an effective strategy to encourage cessation in smokers without an intention to quit.
Smoking reduction therapy might increase the likelihood of complete cessation, and it has been recommended as a therapeutic choice for cigarette smokers without an intention to quit [4,5]. One of the biggest uncertainties with this method is the association between short-term reduction of daily cigarette consumption and long-term complete cessation. There has been some concern that future attempts to quit smoking might be undermined by short-term reduction; others have argued that the reduction might be an intermediate step before quitting completely [6].
Several reviews have assessed the efficacy of methods for helping smokers without an intention to quit [7][8][9][10][11]. Differences between the current meta-analysis and previous meta-analyses on the same topic should be noted. A meta-analysis by Hughes et al. included studies with different designs (such as cross-sectional, prospective studies and RCTs) [7]. Due to the heterogeneity of the methods and results, a qualitative review was performed instead of a meta-analysis [7]. Two meta-analyses examined only NRT-aided reduction [8,9]. Recently, several studies on this topic have advocated for the use of varenicline to reduce the daily consumption of cigarettes to achieve the goal of complete cessation. A Cochrane review by Stead et al. included smokers who were willing to quit smoking [10]. In a meta-analysis focusing on the same population (unwilling to quit smoking), Asfar et al. only included self-reported point prevalence of cessation at the end of follow-up. In addition, they did not consider the safety of using smoking cessation medication as an aid in reduction [11].
Considering the accumulating evidence, we conducted an updated systematic review and meta-analysis to assess the efficacy and safety of smoking reduction therapy for smokers without an intention to quit. In contrast with the above-discussed meta-analyses, we grouped samples into four categories according to different types of smoking reduction therapies. We also restricted our analysis to only randomized, controlled clinical trials and used the strictest available criteria (sustained, biochemically validated and long follow-up). Moreover, we widely considered different categories according to different types of smoking reduction therapy.

Literature Search
This is a systematic review and meta-analysis of previously published randomized controlled trials (RCTs). We conducted and reported the current study in adherence with Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [12]. We searched PubMed, Embase, and Cochrane Central Register of Controlled Trials (CENTRAL) for records to report the effect of smoking reduction therapy on smokers with no intention to quit. The search terms included "tobacco reduction", "cigarette * reduction", "reduce smoking", "smoking reduction", "unwilling to", "not willing", "no inten *", "not ready", "not interest *", "uninterest *", and "unmotivated". Details of the search strategy are shown in Supplementary 1. No language and data restriction were imposed. The last search was run on 24 April 2015. We manually searched the reference lists of relevant studies to identify other potentially eligible studies.

Selection Criteria
Two investigators (Lei Wu and Samio Sun) independently performed the initial search. Duplicate records were deleted; the titles and abstracts of each trial were screened. We identified each study as excluded or requiring further assessment.
We included studies that met the following criteria: (1) population: adult smokers who were not ready to quit, were unwilling to quit or had no intention to quit smoking (willing or unwilling to reduce their smoking intensity); (2) intervention: smoking cessation medications to assist with smoking reduction (such as gum, inhalable nicotine replacement therapy, varenicline or bupropion) or behavioural support/the provision of self-help materials to promote reduction; (3) comparison: placebo, no intervention, and other behavioural support (other support for smoking cessation with the exception of reduction support); (4) outcome: abstinence from smoking after at least six months of follow-up; and (5) design: randomized controlled trials (RCTs).

Data Extraction
Lei Wu and Samio Sun independently performed data extraction. The following data were extracted from each study: first author, date and place of publication, patient characteristics, number of patients enrolled (each arm and total), summary of intervention and control conditions, reported outcomes and risk of bias. The extracted data were entered into a standardized Excel (Microsoft Corporation, Seattle, WA, USA) file. Any disagreements were resolved by discussion between the two investigators. If a trial had multiple arms, we reused the control group in each comparison.
The primary outcome was the smoking cessation rate at the longest follow-up period (at least six months from the baseline intervention). In each trial, the strictest available criteria (sustained, biochemically validated, and longest follow-up) were used to define the quit rate. We collected data measured during an intention-to-treat analysis. The secondary outcome included serious and nonserious adverse events that were reported in the included trials.

Quality Assessment
The Cochrane Collaboration tool was used to assess the risk of bias in each trial [13]. A value of "high", "low", or "unclear" risk of bias was assigned according to the following domains: random sequence generation (selection bias), allocation concealment (selection bias), blinding of participants and personnel (performance bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias), selective reporting (reporting bias) and other biases. Disagreements were resolved by discussion with a third author (Yao He).

Data Synthesis
We calculated the relative risks (RRs) with 95% confidence intervals (CIs) for the dichotomous outcome data. Because the sample size, population characteristics, and other confounding factors were not consistent among studies, a random effects model was used to pool the outcome data, regardless of heterogeneity [14]. Heterogeneity was assessed using the I 2 statistic. Studies with an I 2 statistic >50% were indicated to have significant heterogeneity [15]. We further performed subgroup analysis according to the type of smoking cessation medication (nicotine replacement therapy, varenicline and bupropion). The influence of a single study on the overall pooled results was estimated by omitting one study at every turn. Sensitivity analyses were performed to explore the influence of various exclusion criteria on the overall pooled estimate.
The presence of publication bias was evaluated by using the Begg and Egger tests [16,17]. Results were considered as statistically significant for p value <0.05. We used Stata (version 12.0; StataCorp LP, College Station, Texas, USA) and Review Manager Software (version 5.2; The Nordic Cochrane Centre, Copenhagen, Denmark) for the statistical analyses.

Study Identification and Selection
A detailed flow diagram of the trials included in the meta-analysis is shown in Figure 1. A total of 135 records were identified from the initial database search. Of these, 92 records were excluded for duplicates, and 47 records were excluded after reading the titles and abstracts. The remaining 45 full-text articles were assessed for eligibility. An additional study was identified from the references. Finally, 14 studies were included in the present meta-analysis [18][19][20][21][22][23][24][25][26][27][28][29][30][31]. Table 1 shows the main characteristics of the included trials, and Table 2 lists the outcome data of each included trial. These trials were published between 2000 and 2015. Seven of the included trials were conducted in the United States. The sample size ranged from 67 to 1410 (total 7981). The follow-up period ranged from 6 months to 60 months. reported in 10 and four trials, respectively. Among the secondary outcomes, adverse events were reported in 11 trials.

Study Characteristics
The studies were grouped into four comparisons according to the following different types of smoking reduction therapy (Table 2): (a) reduction support (in the current study, reduction support was defined as behavioural interventions or self-help materials to increase reduction; smoking cessation medications were not incorporated in this definition) plus medication versus reduction support plus placebo, (b) reduction support plus medication versus no intervention, (c) reduction support plus medication versus other support plus medication, and (d) reduction support versus no intervention. Four trials had three arms [21][22][23][24]. We reused the above arm matching for each comparison (e.g., a trial by Carpenter, et al. had the following three arms: (1) telephone-based reduction support plus NRT plus brief advice group, (2) motivational advice plus NRT for quit attempt plus brief advice group, and (3) a group with no intervention. For consistency, we analysed the differences in the cessation rates between (1) and (2) and between (1) and (3).)

Quality Assessment
Supplementary Figure S1 summarizes the details of the risk-of-bias assessment. Eight trials had a detailed description of random sequence generation, and five trials reported appropriate allocation concealment. Five trials reported that the participants and personnel were blinded to the nature of the examined products (medication or placebo). All trials were judged to have low risks of incomplete outcome data, reporting bias and other bias.

Reduction Support Plus Medication vs. Reduction Support Plus Placebo
We estimated the pooled effect size of reduction support plus medication versus reduction support plus placebo based on nine trials ( Figure 2). The pooled relative risk (RR) was 1.97, and the 95% confidence interval (CI) was 1.44 to 2.71, with evidence of significant heterogeneity (I 2 = 52%). In the subgroup analysis, two of the nine trials were offered varenicline or placebo for smokers with no intention to quit, and the pooled RR was 2.66 and 95% CI was 2.10 to 3.36 with no evidence of significant heterogeneity (I 2 = 0%). Six of the nine trials offered nicotine replacement treatment (NRT) or placebo, and the pooled RR was 1.94 and 95% CI was 1.26 to 3.00 with no evidence of significant heterogeneity (I 2 = 45%). One study offered bupropion or placebo (RR, 1.27; 95% CI, 0.67-2.40). Two trials offered self-help materials rather than behavioural reduction support to assist in reduction [23,24].

Reduction Support Plus Medication vs. no Intervention
We included five trials to test the efficacy of reduction support combined with smoking cessation medication. As shown in Figure 3, compared with no intervention, those smokers who received reduction support plus medication had significantly increased smoking abstinence (RR, 1.93; 95% CI, 1.41-2.64; I 2 = 46%).  (7) 24 (7) SD, standard deviation.

Reduction Support Plus Medication vs. other Support Plus Medication
We identified two trials that examined comparisons between reduction support plus medication and other support plus medication. In an analysis combining two trials (Figure 4), there was no evidence of benefit from reduction support plus medication (RR, 0.93; 95% CI, 0.44-2.00; I 2 = 40%).

Reduction Support vs. No Intervention
One study evaluated the effect of reduction support alone. Glasgow et al. reported that behavioural reduction support did not significantly increase smoking abstinence compared to no intervention (RR, 1.49; 95% CI, 0.56-3.93) [26].

Secondary Outcomes
The secondary outcome of the current study was to compare adverse events ( Figure 5). Overall, 11 of the included trials reported information on adverse events (one trial only reported information on deaths). Four deaths occurred in those randomized to NRT, and no deaths occurred in those randomized to non-NRT (varenicline and bupropion). There were no significant differences between treatment and control groups (RR, 1.80; 95% CI, 0.35-9.30) in death occurrence. Serious adverse events occurred in fewer than 8% of cases in both groups. No trials reported that serious adverse events were likely to have arisen from treatment. Discontinuation because of adverse events (RR, 1.34; 95% CI, 1.02-1.75) was significantly more common for the non-NRT group, which experienced more serious adverse events (RR, 1.87; 95% CI, 1.08-3.24), compared with the control group.

Sensitivity Analysis
Because of the small numbers of studies for comparisons, sensitivity analyses were performed for two comparisons: reduction support plus medication versus reduction support plus placebo and reduction support plus medication versus no intervention. Further exclusion of any single trial did not significantly alter the overall combined RR, which ranged from 1  In addition, sensitivity analyses were performed to explore the influences of various exclusion criteria (Supplementary Table S1). If we only included the trials that had large sample sizes, were performed at multiple centres, included healthy populations, and reported carbon monoxide-confirmed quit rates, the results did not significantly affect the overall pooled estimate.

Publication Bias
Publication bias was assessed using Egger and Begg tests. For smoking reduction plus medication versus smoking reduction plus placebo, there was no potential publication bias among the nine included trials (Egger's test, p = 0.532; Begg's test, p = 0.490). Publication bias was not assessed for other comparisons, because the low power with less than five trials limited the interpretability of the finding.

Discussion
The present systematic review and meta-analysis identified 14 trials (a total of 7981 smokers) investigating the effects of different types of reduction therapies for assisting smoking reduction to achieve the goal of long-term cessation in smokers without an intention to quit. The results suggested that reduction support in combination with smoking cessation medication (varenicline or NRT) significantly increased the cessation rate among smokers without an intention to quit. Moreover, the percentages of smokers with serious adverse events were not significantly different between the treatment and control (NRT) groups. Measurements of varenicline and bupropion safety should be further explored. Insufficient evidence is available to test the efficacy of bupropion plus reduction support in promoting long-term cessation in this population.
According to the present meta-analysis, the combination of reduction support with varenicline significantly increased long-term and CO-confirmed (confirmed by exhaled carbon monoxide) smoking cessation rates by a factor of 2.66. The results suggested that varenicline is an effective treatment option for smokers without an intention to quit. The percentage of smokers with serious adverse events and discontinuation because of adverse events in the varenicline group was significantly different from the control group. However, because of the limited number of studies included in the current meta-analysis, further safety measurements are needed. Bupropion significantly increased short-term abstinence rates, but the rates were not sustained after bupropion was discontinued. [27] Future studies should evaluate the efficacy and safety of bupropion with a larger sample size.
We found that NRT was the most widely used reduction therapy for smokers with no intention to quit. Compared to placebo or no intervention, NRT significantly increased long-term cessation. Various inclusion criteria did not significantly affect the results. Regarding the effect of self-help reduction materials, we could not make a definitive conclusion because of the limited sample size [23,24]. Further trials should use stricter outcomes (sustained abstinence) rather than specific time points of abstinence in the future. In accordance with the previous meta-analyses, we confirmed that NRT-assisted reduction to stop smoking was an effective and safe strategy [8,9]. In a trial including patients with heart disease, Joseph et al. reported that unserious and serious adverse events were roughly distributed in the examined treatment groups [29]. Because concerns about cardiovascular and neuropsychiatric adverse events are a relatively new issue [32,33], many trials have not reported on cardiovascular and neuropsychiatric outcomes. Mathew et al. suggested that for patients receiving NRT who continued to smoke, the sympathetic nervous system might be stimulated by high nicotine serum concentrations [34]. More evidence is needed to assess the cardiovascular and neuropsychiatric events that occur during pharmacology-assisted reduction among smokers without an intention to quit.
In the current meta-analysis, we found that the RRs for NRT versus placebo with reduction support were similar to that for NRT versus no support. This result suggests that reduction support is not necessary to achieve the effect of NRT. However, due to the limited number trials (only three trials were relative studies) included in the meta-analysis, strong and definitive recommendations cannot be made on the effect of reduction support [20,21,26]. It remains unclear whether the combination of reduction support and medication to increase reduction was better than other support plus medication. Furthermore, we still cannot make a conclusion about the efficacy of behavioural reduction support alone. To the best of our knowledge, there are at least three on-going RCTs on this topic (NCT02337400, NCT02370147, and NCT 01866722). In smokers with no immediate intention to quit, the results of these trials could provide further evidence of the efficacy of reduction support.
Our analysis did not include three pilot randomized trials because of their short follow-up periods (less than 6 months) [35][36][37]. These trials used new-style tobacco products (smokeless tobacco and very low-nicotine content cigarettes) as a substitute for smoking for smokers who were not interested in quitting. It is worth noting that the efficacy and safety of these new-style tobacco products assisted in smoking reduction and future long-term cessation.
The current meta-analysis has limitations. First, various definitions and time frames were used to define an intention to quit in the included trials. Future studies should follow the common definition of a "stage-of-change" model for smokers with no intention to quit [38,39]. Second, the treatment duration (ranging from two months to 18 months) and follow-up time (ranging from six months to 60 months) varied between the trials. The longer treatment duration might achieve more effective outcomes. We used the longest follow-up time of smoking abstinence, as used previously [40,41]. Third, most of the included trials reported the point prevalence rather than the sustained abstinence rates. The effectiveness might be overvalued because of the lack of sustained abstinence. Fourth, several trials did not report the frequency or intensity of baseline and follow-up behavioural counselling. It was reported that a more intensive counselling intervention was more effective than a less intensive one [42]. Further RCTs should report outcomes with details of baseline and follow-up interventions. Finally, 11 of the included trials recruited moderate or heavy smokers (at least 10 cigarettes per day). It remains unclear whether smoking reduction would be effective for lighter smokers.

Conclusions
In summary, the evidence suggests that a combination of reduction support and medication (NRT and varenicline) to increase reduction in achieving complete cessation is effective for smokers without an intention to quit. Further evidence is needed to assess the safety of varenicline and bupropion as well as the efficacy of reduction behavioural support and bupropion in reducing the daily consumption of cigarettes with the eventual goal of quitting smoking.

Author Contributions
Lei Wu and Yao He were responsible for the conception and design. Lei Wu, Samio Sun, Yao He and Jing Zeng took part in the acquisition, analysis and interpretation of data. All authors have read and approved the final manuscript.