Additional behavioural support as an adjunct to pharmacotherapy for smoking cessation
Abstract
Background
Effective pharmacotherapies are available to help people who are trying to stop smoking, but quitting can still be difficult and providing higher levels of behavioural support may increase success rates further.
Objectives
To evaluate the effect of increasing the intensity of behavioural support for people using smoking cessation medications, and to assess whether there are different effects depending on the type of pharmacotherapy, or the amount of support in each condition.
Search methods
We searched the Cochrane Tobacco Addiction Group Specialised Register in May 2015 for records with any mention of pharmacotherapy, including any type of nicotine replacement therapy (NRT), bupropion, nortriptyline or varenicline that evaluated the addition of personal support or compared two or more intensities of behavioural support.
Selection criteria
Randomized or quasi‐randomized controlled trials in which all participants received pharmacotherapy for smoking cessation and conditions differed by the amount of behavioural support. The intervention condition had to involve person‐to‐person contact. The control condition could receive less intensive personal contact, or just written information. We did not include studies that used a contact‐matched control to evaluate differences between types or components of support. We excluded trials recruiting only pregnant women, trials recruiting only adolescents, and trials with less than six months follow‐up.
Data collection and analysis
One author prescreened search results and two authors agreed inclusion or exclusion of potentially relevant trials. One author extracted data and another checked them.
The main outcome measure was abstinence from smoking after at least six months of follow‐up. We used the most rigorous definition of abstinence for each trial, and biochemically‐validated rates if available. We calculated the risk ratio (RR) and 95% confidence interval (CI) for each study. Where appropriate, we performed meta‐analysis using a Mantel‐Haenszel fixed‐effect model.
Main results
Forty‐seven studies met the inclusion criteria with over 18,000 participants in the relevant arms. There was little evidence of statistical heterogeneity (I² = 18%) so we pooled all studies in the main analysis. There was evidence of a small but statistically significant benefit from more intensive support (RR 1.17, 95% CI 1.11 to 1.24) for abstinence at longest follow‐up. All but four of the included studies provided four or more sessions of support to the intervention group. Most trials used NRT. We did not detect significant effects for studies where the pharmacotherapy was nortriptyline (two trials) or varenicline (one trial), but this reflects the absence of evidence.
In subgroup analyses, studies that provided at least four sessions of personal contact for the intervention and no personal contact for the control had slightly larger estimated effects (RR 1.25, 95% CI 1.08 to 1.45; 6 trials, 3762 participants), although a formal test for subgroup differences was not significant. Studies where all intervention counselling was via telephone (RR 1.28, 95% CI 1.17 to 1.41; 6 trials, 5311 participants) also had slightly larger effects, and the test for subgroup differences was significant, but this subgroup analysis was not prespecified. In this update, the benefit of providing additional behavioural support was similar for the subgroup of trials in which all participants, including controls, had at least 30 minutes of personal contact (RR 1.18, 95% CI 1.06 to 1.32; 21 trials, 5166 participants); previously the evidence of benefit in this subgroup had been weaker. This subgroup was not prespecified and a test for subgroup differences was not significant. We judged the quality of the evidence to be high, using the GRADE approach. We judged a small number of trials to be at high risk of bias on one or more domains, but findings were not sensitive to their exclusion.
Authors' conclusions
Providing behavioural support in person or via telephone for people using pharmacotherapy to stop smoking has a small but important effect. Increasing the amount of behavioural support is likely to increase the chance of success by about 10% to 25%, based on a pooled estimate from 47 trials. Subgroup analysis suggests that the incremental benefit from more support is similar over a range of levels of baseline support.
PICOs
Plain language summary
Does more support increase success amongst people using medications to quit smoking?
Background
Medications (including all types of nicotine replacement therapy, bupropion and varenicline) have been shown to help people quit smoking, and people seeking help to quit smoking will frequently be offered medication (pharmacotherapy). Behavioural support also helps people to quit. Behavioural support may include brief advice or more intensive counselling, and may be provided face‐to‐face individually or in groups, or by telephone, including 'quitlines'. It has been unclear how much additional benefit is gained from adding support, or providing more intensive support for people who are using medication for a quit attempt.
Study characteristics
We looked for trials that enrolled smokers and provided or offered medication to all participants, then randomized them to receive different amounts of behavioural support. People in the control group could get just written information, or some personal support. The trials had to test the addition of more intensive support, involving additional face‐to‐face or by‐telephone contact, or longer sessions. To assess whether the intervention helped people to quit, the trials had to assess the number of people not smoking after at least six months. We did not include studies that only recruited pregnant women, or adolescents.
Key results
We searched for trial reports in May 2015. Forty‐seven trials were eligible, with over 18,000 participants. Participants usually wanted to make a quit attempt, although a small number of studies offered support to people who had not indicated that they were trying to quit. Combining results from 47 trials suggests that increasing the amount of behavioural support for people using a smoking cessation medication increases the chances of quitting smoking for the long term by about 10% to 25%. Providing some support via personal contact, face‐to‐face or by telephone, is beneficial, and people making a quit attempt using medication will increase their chances of success if they also have access to behavioural support.
Quality of the evidence
We judged the overall quality of evidence to be high, with further research unlikely to change the results.
Authors' conclusions
Summary of findings
| Behavioural interventions as adjuncts to pharmacotherapy for smoking cessation | ||||||
| Patient or population: People using smoking cessation pharmacotherapy | ||||||
| Outcomes | Illustrative absolute effects* (95% CI) | Relative effect | No of Participants | Quality of the evidence | Comments | |
| Assumed successful quitters without intervention | Estimated quitters with intervention | |||||
| Pharmacotherapy (with variable level of behavioural support) | Additional behavioural support (in addition to pharmacotherapy) | |||||
| Smoking cessation at longest follow‐up | Study population1 | RR 1.17 | 18,682 | ⊕⊕⊕⊕ | Updating the review in 2015 with an additional 9 studies (3338 participants) has not materially altered the estimated effect or CI; previous RR 1.16 (95% CI 1.09 to 1.24). | |
| 180 per 1000 | 210 per 1000 | |||||
| The estimated rate of quitting with behavioural intervention (and its 95% confidence interval) is based on the assumed quit rate in the control group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| 1Based on the control group crude average of 1705 quitters out of 9487 participants, which is also the median control group rate, 18% 2Few studies were rated at high risk of bias for any domain. Sensitivity analysis limited to studies at low risk of bias on all domains tended to increase the effect estimate 3Little overall evidence of statistical heterogeneity (I² = 18%), or of differences between the subgroups defined by pharmacotherapy | ||||||
Background
Giving up smoking is the most effective way for people who smoke to reduce their risk of premature death and disability. People who smoke need to quit as soon as possible by using the evidence‐based aids that increase their chances of success. These aids include behavioural support and pharmacotherapies. Behavioural support interventions range from written materials containing advice on quitting to multisession group therapy programmes or repeated individual counselling in person or by telephone. Providing standard self‐help materials alone seems to have at best a small effect on success, but there is good evidence of a benefit of individually tailored self‐help materials or more intensive advice or counselling (Lancaster 2005; Hartmann‐Boyce 2014). There is also good evidence that nicotine replacement therapy products (NRT), varenicline, bupropion and nortriptyline all increase the long‐term success of quit attempts (Cahill 2012; Stead 2012b; Hughes 2014). Clinical practice guidelines recommend that healthcare providers offer people who are prepared to make a quit attempt both pharmacotherapy and behavioural support. The two types of treatment are believed to have complementary modes of action, and to independently improve the chances of maintaining long‐term abstinence (Hughes 1995; Cofta‐Woerpel 2007). Although guidelines recommend intensive support to improve abstinence rates, it is also recognized that many people will not attend multiple sessions. NRT products are available over the counter without a prescription in many countries, and people who purchase them may not access any specific behavioural support. People who obtain prescriptions for pharmacotherapies are more likely to receive some support, but this may be focused on explaining the proper use of the drug and not on counselling.
Other Cochrane Tobacco Addiction reviews have evaluated the evidence on behavioural and pharmaceutical interventions individually. These reviews restrict inclusion to trials where interventions are unconfounded. Trials of pharmacotherapies must provide the same amount of behavioural support (materials, advice, counselling contacts) to all participants, whether they receive active treatment, or a placebo or no medication. Likewise, when behavioural interventions are evaluated there should be no systematic difference in the offer of medications between the active and control arms of the trial. Only reviews that evaluate interventions by specific providers (e.g. nurses, Rice 2013), or in specific settings (e.g. hospitals, Rigotti 2012), may include trials of interventions that combine behavioural therapies and various medications (e.g. NRT, bupropion, varenicline).
This review is one of two that systematically identify trials of interventions that combine effective pharmacotherapies (NRT, varenicline, bupropion, nortriptyline) with behavioural support (tailored materials, brief advice, in‐person or telephone counselling). This review evaluates trials that compare different levels of behavioural intervention for people receiving any pharmacotherapy for smoking cessation, to provide an estimate of the effectiveness of intensifying behavioural support as an adjunct to pharmacotherapy. The companion review (Stead 2012a) includes trials in which an intervention combining pharmacotherapy and behavioural support is compared to standard care or a brief behavioural intervention without pharmacotherapy.
Objectives
To evaluate the effect of increasing the intensity of behavioural support for people using smoking cessation medications, and to assess whether there are different effects depending on the type of pharmacotherapy, or the amount of support in each condition.
Methods
Criteria for considering studies for this review
Types of studies
Randomized or quasi‐randomized controlled trials.
Types of participants
We included trials that recruited people who smoke in any setting, with the exception of trials which only recruited pregnant women or adolescents. These populations are considered in specific reviews (Grimshaw 2006; Coleman 2012). Trial participants did not need to be selected according to their interest in quitting, or their suitability for pharmacotherapy, but since pharmacotherapy was offered or provided, participants were expected to be relatively motivated and prepared to use medication as part of their quit attempt.
Types of interventions
We included trials of smoking cessation interventions where all participants had access to a smoking cessation pharmacotherapy (including NRT, varenicline, bupropion and nortriptyline, or a combination or choice of these) and in which one or more intervention conditions received more intensive behavioural support than the control condition. Control group participants could be offered any level of support from minimal (e.g. written information provided as part of the medication prescription) to multisession counselling, but support must have been of a lower intensity (based on number or length of sessions) than that given to intervention participants. In addition to being more intensive, the intervention could use different or additional types of therapy content (e.g. cognitive behaviour therapy, motivational interviewing).The additional support had to involve person‐to‐person contact which could be face‐to‐face or by telephone. We did not include trials testing specific behavioural components that used a control matched for contract frequency and duration. There are separate Cochrane reviews of trials specifically evaluating motivational interviewing (Lindson‐Hawley 2015) and stage‐based interventions (Cahill 2010) which would also include studies of these behavioural interventions without pharmacotherapy.
Types of outcome measures
Following the standard methodology of the Tobacco Addiction Group, the primary outcome was smoking cessation at the longest follow‐up using the strictest definition of abstinence, i.e. preferring sustained over point prevalence abstinence and using biochemically‐validated rates where available. In addition we noted any other abstinence outcomes reported, and conducted sensitivity analyses if the choice of outcome in a study might have altered the results of a meta‐analysis. We excluded trials reporting less than six months follow‐up from the start of intervention.
Search methods for identification of studies
We identified trials from the Cochrane Tobacco Addiction Specialised Register (the Register). This is generated from regular searches of the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE, EMBASE, and PsycINFO for trials of smoking cessation or prevention interventions. The most recent search of the Register was in May 2015. At the time of the search the Register included the results of searches of the CENTRAL to issue 4, 2015; MEDLINE (via OVID) from 1946 to update 20150501; EMBASE (via OVID) from 1974 to week 201519 and PsycINFO (via OVID) from 1967 to update 20150506. See the Tobacco Addiction Group Module in the Cochrane Library for full search strategies and a list of other resources searched.
We used the same search strategy that we used in the companion review that compared interventions that combined pharmacotherapy and behavioural therapy with controls that offered neither (Stead 2012a). Records were screened for relevance to either review. We searched the Register for records with any mention of pharmacotherapy, including any type of NRT, bupropion, nortriptyline or varenicline in title, abstract or indexing terms (see Appendix 1 for the final search strategy). We checked titles and abstracts to identify trials of interventions for smoking cessation that combined pharmacotherapy with behavioural support. We also checked the excluded study lists of reviews of behavioural therapies and pharmacotherapy for trials excluded because pharmacotherapy was confounded with additional behavioural support compared to the control group. We also considered for inclusion trials with a factorial design that varied both pharmacotherapy and behavioural conditions. For the first version of this review we also tested an additional MEDLINE search using the smoking‐related terms and design limits used in the standard Register search and the MeSH terms ‘combined modality therapy’ or (Drug Therapy and (exp Behavior therapy or exp Counseling)). This search retrieved a subset of records already screened for inclusion in the Register, and was used to assess whether it might retrieve studies where there was no mention of a specific cessation pharmacotherapy in the title, abstract or indexing. We did not find any additional studies from this approach.
Data collection and analysis
Selection of studies
LS (who is also the Trials Search Co‐ordinator for the Cochrane Tobacco Addiction Group) identified potentially relevant trials according to the criteria above and obtained the full text of papers reporting primary outcomes of all potentially relevant trials. Where there was uncertainty about inclusion or exclusion, she discussed trials with PK or TL.
Data extraction and management
LS extracted data and categorized trials for subgroup analysis, with data extraction checked by PK or TL. We extracted the following information:
-
Country and setting of trial
-
Study design
-
Method of recruitment, including any selection by motivation to quit
-
Method of sequence generation
-
Method of allocation concealment
-
Characteristics of participants including gender, age, smoking rate
-
Characteristics of intervention deliverer
-
Common components: type, dose and duration of pharmacotherapy
-
Intervention components: type and duration of behavioural support
-
Control group components: type and duration of behavioural support
-
Outcomes: primary outcome length of follow‐up and definition of abstinence, other follow‐ups and abstinence definitions, use of biochemical validation, adverse events
-
Sources of funding & potential conflicts of interest
Assessment of risk of bias in included studies
We evaluated studies on the basis of the randomization procedure, allocation concealment, incomplete outcome data assessment and any other bias (Schulz 2002a; Schulz 2002b; Cochrane Handbook 2011).
Measures of treatment effect
We expressed trial effects as a risk ratio calculated as: (quitters in treatment group/total randomized to treatment group)/(quitters in control group/total randomized to control group). A risk ratio greater than 1 indicates a better outcome in the intervention group than in the control condition.
Dealing with missing data
We reported numbers lost to follow‐up by group in the 'Risk of bias' table. Following standard Cochrane Tobacco Addiction Group methods, we assumed people lost to follow‐up to be smoking and included them in the denominators for calculating the risk ratio. We have reported any exceptions to this assumption the 'Risk of bias' table. We noted separately any deaths during follow‐up and excluded them from denominators.
Assessment of heterogeneity
We considered pooling all trials comparing more and less intensive behavioural support if statistical heterogeneity as assessed by the I² statistic (Higgins 2003) was less than 50%.
Data synthesis
For groups of trials where we judged meta‐analysis appropriate, we pooled risk ratios using a Mantel‐Haenszel fixed‐effect model, and reported a pooled estimate with a 95% confidence interval.
If trials had more than one intervention condition, we compared the most intensive combination of behavioural support and pharmacotherapy to the control in the main analysis.
We categorize the intensity of behavioural support in both intervention and control conditions based on two of the categories used in the US Guidelines (Fiore 2008): ‘Total amount of contact time’ (Categories: 0, 1 to 30*, 31 to 90, 91 to 300, > 300 minutes (*guideline categories 1 to 3 and 4 ‐ 30 combined)) and ‘Number of person‐to‐person sessions’ (Categories: 0*, 1 to 3*, 4 to 8, > 8 (*guideline used 0 to 1, 2 to 3)).
Subgroup analysis and investigation of heterogeneity
We used the difference in intensity of support (number or duration of contacts) between intervention and control conditions as the main potential feature to explain any heterogeneity. We expected trials testing larger differences in intensity (i.e., using longer, more frequent sessions in the intervention condition and brief or no contact in the control) to show larger treatment effects. We also considered whether the definition and duration of follow‐up, or the use of biochemical validation of cessation, had any impact on treatment effect.
Sensitivity analysis
We considered whether the main results were sensitive to the exclusion of studies at high or unclear risk of bias in any domain.
Results
Description of studies
Results of the search
The Register search retrieved approximately 2940 records. We excluded most of them as not relevant based on title and abstract. Of the records that did relate to trials of interventions for smoking cessation most were not relevant because they were placebo‐controlled trials of pharmacotherapies, in which the behavioural support was the same for intervention and control conditions. We identified 47 studies for inclusion and listed 61 as excluded. We identified 15 ongoing studies. Further studies of combined pharmacotherapy and behavioural support that did not offer pharmacotherapy to the control group are included in Stead 2012a. Some studies had multiple study arms and contributed to both Stead 2012a and to this review. All reports related to a study are listed in the reference section with the primary report used for data extraction identified. Study identifiers are based on the first author and year of publication of the main study report.
Included studies
We identified 47 studies as relevant for inclusion, of which nine were new for the 2015 update. Over 18,000 participants are now included in relevant arms of these studies. Details of each study are given in the Characteristics of included studies table, and a summary of intervention and control group characteristics in Table 1.
| Study | Control | Intervention | Pharmacotherapy | ||||||
| Sessions | Duration (minutes) | Sessions | Duration (minutes) | ||||||
| Number | Category | Total | Category | Number | Category | Total | Category | ||
| No sessions of personal contact in control condition | |||||||||
| 0 | 0 | 0 | 0 ‐ 3 | up to 9 (av 5 for acceptors) | 4 ‐ 8 | est 10 min/sess | 31 ‐ 90 | NRT or bupropion | |
| 0 | 0 | 0 | 0 ‐ 3 | 5 | 4 ‐ 8 | 60 mins total | 31 ‐ 90 | Nicotine patch (2w) | |
| 0 | 0 | 0 | 0 ‐ 3 | 5 | 4 ‐ 8 | 15 ‐ 25 min/sess | 91 ‐ 300 | Nicotine patch | |
| 0 | 0 | 0 | 0 ‐ 3 | 6 | 4 ‐ 8 | est 15 ‐ 20 min/sess | 91 ‐ 300 | NRT or bupropion | |
| 0 | 0 | 0 | 0 ‐ 3 | up to 12, av 7 | > 8 | average 9 min/sess | 31 ‐ 90 | Nicotine patch | |
| 0 | 0 | 0 | 0 ‐ 3 | average 8 so target higher | > 8 | av 10 min/sess | 91 ‐ 300 | Nicotine patch | |
| 1‐3 sessions of personal contact in control condition | |||||||||
| 2 | 1 ‐ 3 | ≤5 mins/sess | 4 ‐ 30 | 2 | 1 ‐ 3 | 60 ‐ 120 min/sess | 31 ‐ 90 | Nicotine patch | |
| 1 ± 2 | 1 ‐ 3 | 15 ‐ 25 min | 4 ‐ 30 | 1 ± 2 | 1 ‐ 3 | 55 ‐ 75 min | 31 ‐ 90 | Nicotine patch | |
| 1 | 1 ‐ 3 | brief | 4 ‐ 30 | 1 + 1 ‐ 2 | 1 ‐ 3 | est 60 mins | 31 ‐ 90 | Nicotine patch | |
| 1 ‐ 2 | 1 ‐ 3 | brief | 4 ‐ 30 | 2 ‐ 3 | 1 ‐ 3 | est 60 ‐ 90 mins | 31 ‐ 90 | Nicotine patch | |
| 1 | 1 ‐ 3 | 20 mins (mean) | 4 ‐ 30 | 1 + up to 4 | 4 ‐ 8 | average 47 or 60 tot | 31 ‐ 90 | Nicotine patch | |
| 1 | 1 ‐ 3 | brief | 4 ‐ 30 | 1 + 3 | 4 ‐ 8 | > 45 min | 31 ‐ 90 | Nicotine patch | |
| 2 | 1 ‐ 3 | 10 + 5 mins | 4 ‐ 30 | 2 + 3 | 4 ‐ 8 | 15 + 30 | 31 ‐ 90 | Nicotine gum | |
| 1 | 1 ‐ 3 | brief | 4 ‐ 30 | 4 | 4 ‐ 8 | 'brief calls' | 31 ‐ 90 | Bupropion | |
| 1 | 1 ‐ 3 | 5 ‐ 10 mins | 4 ‐ 30 | 5 | 4 ‐ 8 | av. tot 60 min | 31 ‐ 90 | Varenicline | |
| 1 | 1 ‐ 3 | 10 mins | 4 ‐ 30 | 6 | 4 ‐ 8 | est 210 mins | 91 ‐ 300 | Nicotine patch | |
| 1 | 1 ‐ 3 | brief | 4 ‐ 30 | 7 | 4 ‐ 8 | 285 min | 91 ‐ 300 | Nicotine patch | |
| 1 | 1 ‐ 3 | 20 mins | 4 ‐ 30 | 3 ‐ 4 | 4 ‐ 8 | 60 min/sess | 91 ‐ 300 | Nicotine patch | |
| 2 | 1 ‐ 3 | est < 30 | 4 ‐ 30 | 2 + 5 | 4 ‐ 8 | est 30 ‐ 60 mins | 91 ‐ 300 | Nicotine gum | |
| 2 | 1 ‐ 3 | est 30 | 4 ‐ 30 | 5 | 4 ‐ 8 | 120 mins + phone | 91 ‐ 300 | Nicotine patch | |
| 1 | 1 ‐ 3 | 10 ‐ 15 min | 4 ‐ 30 | 6 | 4 ‐ 8 | 90 ‐ 120 min | 91 ‐ 300 | Nicotine patch | |
| 1 | 1 ‐ 3 | brief | 4 ‐ 30 | 3 or 8 (not incl base) | 4 ‐ 8 | ? Or 1 hr ss | > 300 | Nicotine patch | |
| 1 | 1 ‐ 3 | brief | 4 ‐ 30 | 6 | 4 ‐ 8 | 240 ‐ 360 min | > 300 | Nicotine patch or gum | |
| 3 | 1 ‐ 3 | 45 mins | 31 ‐ 90 | 3 + 3 | 4 ‐ 8 | 45 + ?30 | 31 ‐ 90 | Nicotine patch | |
| 1 | 1 ‐ 3 | 90 mins incl 'orientation' | 31 ‐ 90 | 1 + 4 | 4 ‐ 8 | 10 ‐ 15 min/sess | 91 ‐ 300 | Nicotine patch | |
| 1 | 1 ‐ 3 | 30 mins | 4 ‐ 30 | 1 + 3 + 12 | > 8 | >300 | > 300 | Nicotine patch | |
| 1 | 1 ‐ 3 | 15 mins | 4 ‐ 30 | 1 + 9 | > 8 | >500 | > 300 | Bupropion | |
| 3 | 1 ‐ 3 | 5 ‐ 10 min/sess | 4 ‐ 30 | 3 + 6 | > 8 | 90 min/sess plus control | > 300 | Nicotine patch | |
| 1 | 1 ‐ 3 | 1 hour 'medication instruction group' | 31 ‐ 90 | 20 | > 8 | > 300 | > 300 | Bupropion | |
| 4‐8 sessions of personal contact in control condition | |||||||||
| 4 | 4 ‐ 8 | ˜80 mins | 31 ‐ 90 | 7 | 4 ‐ 8 | ˜140 | 91 ‐ 300 | Nicotine patch | |
| 4 | 4 ‐ 8 | 60 min/sess (health education) | 91 ‐ 300 | 4 | 4 ‐ 8 | 60 min/sess | 91 ‐ 300 | Nicotine patch | |
| 5 | 4 ‐ 8 | 2½ hrs | 91 ‐ 300 | 7 | 4 ‐ 8 | 210 + 50 | 91 ‐ 300 | Nicotine sublingual tablet | |
| 4 | 4 ‐ 8 | 60 min/sess | 91 ‐ 300 | 8 | 4 ‐ 8 | 60 min/sess | > 300 | NRT | |
| 5 | 4 ‐ 8 | 45 min/sess | > 300 | 5 | 4 ‐ 8 | 90 min/sess plus control | > 300 | Nicotine gum | |
| 4 | 4 ‐ 8 | 5 ‐ 10 min/sess | 31 ‐ 90 | 10 | > 8 | 10 ‐ 15 *10 | 91 ‐ 300 | NRT or bup | |
| 4 | 4 ‐ 8 | 25 ‐ 35 mins total | 31 ‐ 90 | 4 + 5 | > 8 | 90 min/sess plus control | > 300 | Bup or Nort | |
| 4 | 4 ‐ 8 | 45 mins | 91 ‐ 300 | 14 | > 8 | 75 min/sess | > 300 | Nicotine gum | |
| 4 | 4 ‐ 8 | est 30 min/sess | 91 ‐ 300 | 4 + 16 | > 8 | 45 min/sess | > 300 | Nicotine patch | |
| 5 | 4 ‐ 8 | 60 min/sess | 91 ‐ 300 | 14 | > 8 | 75 min/sess | > 300 | Nicotine gum | |
| 5 | 4 ‐ 8 | 60 min/sess | 91 ‐ 300 | 11 | > 8 | 20 ‐ 40 min/sess | > 300 | Bupropion & gum | |
| 6 | 4 ‐ 8 | 30 min/sess | 91 ‐ 300 | 6 + 4 | > 8 | 30 min/sess | > 300 | Bupropion & patch | |
| 5 | 4 ‐ 8 | 90 min/sess | > 300 | 10 | > 8 | 120 min/sess | > 300 | Nicotine gum | |
| 5 | 4 ‐ 8 | 90 min/sess | > 300 | 10 | > 8 | 120 min/sess | > 300 | Nortriptyline | |
| 7 | 4 ‐ 8 | 560 min | > 300 | 15 | > 8 | 840 min | > 300 | Nicotine patch | |
| Over 8 sessions of personal contact in control condition | |||||||||
| 13 | > 8 | est 10 min/sess | 91 ‐ 300 | 13 incl 8 counselling | > 8 | extra 80 mins | > 300 | Bupropion | |
| 10 | > 8 | 150 min | 91 ‐ 300 | 17 | > 8 | 770 min | > 300 | NRT(gum/patch), bupropion, or varenicline | |
| 9 | > 8 | 20 min/sess | > 300 | 24 | > 8 | 45 min/sess | > 300 | Nicotine patch | |
Study setting, participant recruitment and motivation
Fifteen studies recruited people in a healthcare setting (excluding smoking cessation clinics); this included five studies in primary care (Ockene 1991; Fiore 2004; Aveyard 2007; Ellerbeck 2009; Bock 2014), one in a chest clinic (Tonnesen 2006), one in a cardiovascular disease outpatient clinic (Wiggers 2006), two in HIV clinics (Lloyd‐Richardson 2009; Humfleet 2013), one in mental health clinics (Williams 2010), three in substance abuse clinics (Lifrak 1997; Stein 2006; Rohsenow 2014), one in a Veterans Administration hospital (Simon 2003), and one in cardiac wards (Berndt 2014). Since the intervention included the provision of pharmacotherapy, many of the studies recruiting in a healthcare setting recruited volunteers who were interested in making a quit attempt, but motivation to quit was not always an explicit eligibility criterion. Wiggers 2006 used a motivational interviewing approach and participants did not all make quit attempts. Ockene 1991 offered nicotine replacement therapy (NRT) so participants were not initially selected by motivation, and Ellerbeck 2009 included a small proportion of people in the precontemplation stage. A further four studies recruited members of health maintenance organizations (HMOs) (Lando 1997; Swan 2003; Boyle 2007; Swan 2010). Boyle 2007 proactively recruited HMO members who had filled a prescription for smoking cessation medication, while the others sought volunteers by advertising to HMO members. The remaining 28 studies recruited community volunteers interested in quitting, including three which recruited people who were attending cessation clinics (Alterman 2001; Rovina 2009; Yalcin 2014).
Characteristics of intervention and control conditions
Pharmacotherapy
NRT was offered in the majority of studies, with 25 providing nicotine patch only. While most of these provided a supply for between eight and 12 weeks, one study offered only a two‐week supply (MacLeod 2003). Six studies used nicotine gum (Hall 1985; Hall 1987; Ockene 1991; Ginsberg 1992; Hall 1994; Huber 2003), one used sublingual tablets (Tonnesen 2006), and one did not specify the type (Bushnell 1997). One study offered patch or gum (Humfleet 2013). Four studies provided bupropion alone (Swan 2003; McCarthy 2008; Rovina 2009; Gifford 2011), one provided nortriptyline alone (Hall 1998) and one provided varenicline alone (Swan 2010). Three studies offered a choice of pharmacotherapy; NRT or bupropion (Boyle 2007; Ellerbeck 2009), or NRT, bupropion, or varenicline (Yalcin 2014). Gariti 2009 randomized participants to NRT or bupropion using a double‐dummy design. Hall 2002 randomized participants to either bupropion or nortriptyline (placebo arms not used in this review). Two studies provided combination therapy of both NRT and bupropion (Killen 2008; Hall 2009).
Behavioural support
The intensity of the behavioural support, in both the number of sessions and their duration, was very varied for both intervention and control conditions. In six trials, there was no counselling contact for the controls: in five, participants received pharmacotherapy by mail (Solomon 2000; MacLeod 2003; Solomon 2005; Boyle 2007; Ellerbeck 2009), and in Fiore 2004 there was no counselling or advice for the control group although there was face‐to‐face contact with study staff. In 24 studies, the control arms had between one and three contacts (which could be face‐to‐face or by telephone) and most of these had a total contact duration of between four and 30 minutes, although three had between 31 and 90 minutes contact scheduled (Lando 1997; Reid 1999; Gifford 2011). In 15 studies the control group was scheduled to receive between four and eight contacts, with all except three (Aveyard 2007; Gariti 2009; Wu 2009) involving a total contact duration of over 90 minutes. Three studies scheduled over eight contacts for the controls (McCarthy 2008; Williams 2010; Yalcin 2014).
Typically, the intervention involved only a little more contact than the control, so that the least intensive control conditions were in trials with only moderate‐intensity interventions. In four trials the intervention consisted of between one and three sessions, with a total duration of 31 to 90 minutes (Stein 2006; Wiggers 2006; Calabro 2012; Rohsenow 2014), although Calabro 2012 also provided access to a tailored internet programme. Twenty‐four studies tested interventions of between four and eight sessions, about half of which were in the 91 to 300 minutes‐duration category. The remaining 19 studies offered more than eight sessions, typically providing over 300 minutes of counselling in total. The number of contacts planned was not always delivered, but generally using the average number delivered would not have changed the coding category. In a few cases where the number of contacts was either not specified or open‐ended, we coded the average number delivered.
In Analysis 1.2, we grouped trials by the number of intervention and control contacts. In 12 trials, the intervention and control condition fell into the same coding category for number of contacts (one to three contacts; Stein 2006; Wiggers 2006; Calabro 2012; Rohsenow 2014; four to eight contacts: Bushnell 1997; Huber 2003; Tonnesen 2006; Aveyard 2007; Wu 2009; more than eight contacts: McCarthy 2008; Williams 2010; Yalcin 2014). A summary of the number of sessions and duration for intervention and control conditions for each trial is provided in Table 1.
Length of follow‐up and definitions of abstinence
Most studies followed participants for a year from the target quit date, or entry into the study. Exceptions were Hall 2009 and Ellerbeck 2009 which each had a two‐year follow‐up. The design of the Ellerbeck study, in which participants were repeatedly offered support to quit, means that participants who were quit at the end of follow‐up would not necessarily have been quit for as long as two years. Thirteen studies only followed participants for six months (Ockene 1991; Jorenby 1995; Bushnell 1997; MacLeod 2003; Solomon 2005; Stein 2006; Lloyd‐Richardson 2009; Wu 2009; Swan 2010; Brown 2013; Okuyemi 2013; Berndt 2014; Yalcin 2014).
The majority of studies reported abstinence as a prevalence measure, rather than requiring reported sustained abstinence, or abstinence at multiple follow‐up points. Nine studies did not attempt any biochemical verification of self‐reported abstinence (Ockene 1991; MacLeod 2003; Swan 2003; Solomon 2005; Otero 2006; Boyle 2007; Hollis 2007; Swan 2010; Berndt 2014). Only two of these studies (Otero 2006; Berndt 2014) included face‐to‐face contact between study staff and participants; all the others involved only telephone contact, so validation was not felt to be feasible.
Excluded studies
We list 61 studies as excluded, the majority of which were because they provided less than six months follow‐up or because both arms received the same duration and length of behavioural support. Studies in which the intervention group received both pharmacotherapy and behavioural support and the control group received neither (or just brief behavioural support) were eligible for the companion review and are included or excluded there (Stead 2012a). We excluded three studies where the additional behavioural support did not include any person‐to‐person contact. One of these involved automated telephone calls (Velicer 2006) and one provided access to an internet site (Japuntich 2006). The third compared internet and print‐based materials (Emmons 2013). In Ahluwalia 2006; Nollen 2007; Asfar 2010; Macpherson 2010; Bricker 2014, and Klesges 2015, the delivery method or type of support was the component being assessed, so they offered the same intensity of counselling between arms and were therefore excluded. A full list of excluded studies along with reasons for their exclusion can be found in the Characteristics of excluded studies table.
Risk of bias in included studies
Allocation
We judged 16 studies to be at low risk of bias, based on the reported method of random sequence generation and allocation concealment. We judged three studies to be at high risk of selection bias, due to the method of sequence generation (Yalcin 2014), or allocation concealment (Brown 2013; Berndt 2014; Yalcin 2014). The remaining studies did not given enough detail on one or both of these aspects so we rated the risk of bias as unclear.
Blinding
We did not formally assign a risk of performance bias for each trial. Because of the nature of the intervention, people providing the behavioural support could not be blind. The varying designs of the studies means that some providers of care would have had contact with all conditions whilst some would only have had contact with intervention condition participants. Some care providers would have been providing the same support for all trial participants, some would have been providing more intensive support to one or more intervention groups, and in some cases the treatment was highly individualized. Whilst performance bias cannot be ruled out, we do not think it is a major threat to the validity of this group of trials or that individual trials can be identified as being at higher risk than others in this domain.
Ideally all follow‐up data would be collected by people who remained blind to participant status. Only a few trials reported this explicitly, but others described standardized methods for follow‐up, and attempts to contact non‐responders which would reduce detection bias. The use of biochemical validation would also reduce the risk of detection bias; we assess the impact of biochemical validation in subgroup analysis.
Incomplete outcome data
Loss to follow‐up is often relatively high in smoking cessation trials. If trials lost fewer than 20% of participants at longest follow‐up we judged the risk of bias to be low in this domain. In most of the included trials, the proportion lost to follow‐up was more than 20% but losses were balanced across groups and less than 40%; for these we also classified the risk of bias as low. We rated two studies which reported overall losses to follow‐up of greater than 20% but did not provide a breakdown by treatment arm at unclear risk of bias (Bushnell 1997; Tonnesen 2006), as were a further three studies that did not report level of attrition (Hall 1994; Otero 2006; Smith 2001). We judged three studies (Gifford 2011; Calabro 2012; Bock 2014) to be at high risk because there was an overall loss to follow‐up around 50%.
Other potential sources of bias
We found no studies to be at risk of other potential sources of bias.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Effects of interventions
There was little evidence of statistical heterogeneity (I² = 18%) or of differences between the subgroups defined by pharmacotherapy, so we pooled all 47 studies, including a total of over 18,000 participants. Hall 2002 contributes separate data to two subgroups in the primary meta‐analysis. Fifteen of the studies had point estimates below 1, that is, with higher quit rates in the less intensive condition, but all these had wide confidence intervals (CIs). Only seven studies detected benefits of intervention with confidence intervals that excluded 1. The estimated risk ratio (RR) was 1.17, with 95% CI 1.11 to 1.24 (47 studies, 18,682 participants). This suggests that increasing the intensity of behavioural support for people making a cessation attempt with the aid of pharmacotherapy typically leads to a relatively small increase in the proportion who are quit at six to 12 months (Figure 2; Analysis 1.1; summary of findings Table for the main comparison).
Effect of increasing behavioural support. Abstinence at longest follow‐up.
Subgroups by type of pharmacotherapy
The effect size was similar and also reached statistical significance in the subgroups using NRT (RR 1.14, 95% CI 1.06 to 1.23; 34 trials, 12,295 participants) bupropion (RR 1.27, 95% CI 1.10 to 1.46; 5 trials, 2298 participants); choice of pharmacotherapy (RR 1.28, 95% CI 1.07 to 1.53; 4 trials, 2427 participants). The effect in this subgroup was sensitive to the effect of Yalcin 2014, without which the effect was not significant. Subgroup CIs did not exclude no effect for nortriptyline (RR 0.98, 95% CI 0.59 to 1.63; 2 trials, 172 participants), for varenicline (RR 1.11, 95% CI 0.89 to 1.37; 1 trial, 800 participants) or NRT and bupropion combined (RR 1.22, 95% CI 0.98 to 1.52; 2 trials, 690 participants). This is likely to reflect the smaller number of studies and lower precision rather than a true difference in effect. A test for differences between subgroups was not significant (P = 0.61).
Subgroups by difference in intensity
Analysis 1.2 categorizes trials based on the number of contacts, with the subgroups with the largest contrast in intensities listed first and studies where the intensity of intervention and control fell into the same category shown last. There was little evidence of any dose‐response, although the point estimate was highest for the subgroup in which controls did not have any personal contact (RR 1.25, 95% CI 1.08 to 1.45; 6 trials, 3762 participants). All interventions in this subgroup of trials involved at least four contacts. Since there was little heterogeneity overall it was not surprising that subgroups did not consistently reduce heterogeneity. We did not repeat this approach for duration of intervention categories, as inspection suggested that the number of studies falling into different categories was small and that further subgroup analysis could be misleading.
At the suggestion of a peer reviewer, we conducted two additional subgroup analyses. In Analysis 1.3 we categorized by the level of control group contact to investigate whether there might be a difference between trials where the control could be categorized as a brief intervention (up to 30 minutes) and trials which might be characterized as testing a dose‐response for behavioural support, which we defined as being where the controls received more than 30 minutes of behavioural support. The six trials (Solomon 2000; MacLeod 2003; Fiore 2004; Solomon 2005; Boyle 2007; Ellerbeck 2009) where controls had no advice or contact formed a third subgroup. Twenty trials and just over half the participants were in the 'brief intervention' subgroup, and 21 trials with just over a quarter of participants were in the 'dose‐response' category. There was almost no difference between the subgroup estimates; 'brief intervention' RR 1.15, (95% CI 1.06 to 1.24; 20 trials, 9754 participants), 'dose response' RR 1.18, (95% CI 1.06 to 1.32; 21 trials, 5166 participants), and 'no contact' RR 1.25, (95% CI 1.08 to 1.45; 6 trials, 3762 participants) . There was a marginal fall in estimates on exclusion of studies in which the intervention and control fell into the same category for number of contacts (data not shown).
Subgroup by modality of intervention contact
In the second non‐prespecified analysis we categorized studies according to whether there was some face‐to‐face contact as part of the intervention, or whether all support was given by telephone (Analysis 1.4). In the subgroup of six large studies using telephone counselling (which had some overlap with studies where there was no personal contact for the control), the point estimate was a little higher, although the CI was wide (RR 1.28, 95% CI 1.17 to 1.41, 5311 participants). In the remaining 41 studies where all intervention and most control conditions had face‐to‐face support, there was also evidence of benefit in this update, although the estimate was slightly smaller (RR 1.12, 95% CI 1.04 to 1.20; 13,371 participants). A formal test for difference between subgroups was significant (P = 0.03).
Sensitivity analyses
Inclusion of medium‐intensity intervention from studies with multiple intervention conditions
Eight studies (Jorenby 1995; Alterman 2001; Smith 2001; Fiore 2004; Hollis 2007; Ellerbeck 2009; Swan 2010; Humfleet 2013) included an intervention condition intermediate in intensity between the highest intensity and the control. We have not included these arms in the primary analysis in case they reduced the contrast between intervention and control. In a sensitivity analysis we added in these arms. This had almost no impact on the estimated effect (RR 1.16, 95% CI 1.10 to 1.23; Analysis 2.1), tending to support the finding that there is not a clear dose‐response relationship with the amount of support.
Definition of abstinence & use of validation
We considered whether the way in which abstinence was defined was related to the effect size, and also to absolute quit rates. We did not find any difference in relative effect between studies that reported point prevalence rather than sustained abstinence at 12 months (Analysis 2.2). Some studies that reported sustained outcomes also reported point prevalence rates, but substituting the less stringent definition did not change the overall findings. However, studies with point prevalence outcomes had, on average, higher quit rates in both intervention and control arms. A study comparing outcomes based on different abstinence definitions reported within studies found that, for pharmacotherapy studies, point prevalence and sustained abstinence outcomes were strongly related, with sustained abstinence averaging around 74% of point prevalence rates (Hughes 2010). In this review there was a bigger difference, and the average quit rate in studies that reported a sustained outcome was 50% to 60% of the average for studies reporting point prevalence.
Most of the studies used biochemical validation of self‐reported cessation. When we excluded the nine studies that did not, the estimated effect fell slightly (RR 1.14, 95% CI 1.05 to 1.23; 11,168 participants) whilst the effect in the excluded subgroup without validation was slightly larger (RR 1.21, 95% CI 1.12 to 1.31; 7572 participants). The importance of this is unclear; whilst it is possible that the absence of validation was a source of bias, it is also possible that these studies had genuinely larger effects; six involved telephone counselling, and two had no personal contact in the control groups, potentially leading to a larger contrast between intervention and control conditions.
Discussion
Summary of main results
A meta‐analysis pooling 47 studies with a total of over 18,000 participants suggests that providing more intense behavioural support for people making a cessation attempt with the aid of pharmacotherapy will typically increase the success rates by about 10% to 25%. With the addition of nine trials to this update, the number of participants has increased by 20% and yet the risk ratio has remained almost the same, changing from 1.16 to 1.17, with a small narrowing in the CI. This increases confidence that there is a small benefit. There continues to be little evidence of statistical heterogeneity overall, despite the variability in the amount and nature of the behavioural support tested. Sensitivity analyses suggest that this estimate is quite robust. Although the relative effect is small, it is important to put the effect in the context of control conditions that were offering effective pharmacotherapy and, typically, some behavioural support, i.e. a level of support consistent with guideline best practice. Quit rates in the control groups reflected this, with a median quit rate across trials of around 18% (interquartile range 12% to 26%). So the estimated relative increase translates into an absolute increase of around two to four percentage points. Given the importance of smoking cessation for future health outcomes, this is a clinically relevant difference (West 2007).
There were potentially important differences between trials in the relative difference in the support given to the intervention and control groups. Despite the lack of statistical heterogeneity, we undertook a number of subgroup analyses including some that were not prespecified. In response to a concern that we were combining tests of behavioural support versus no support with tests of a dose‐response to intensity of support, we divided trials into those where the control did not involve personal contract (six trials); where the control group provided a brief intervention, operationalized as under 30 minutes contact; and those where the control condition was more intensive (Analysis 1.3). There was no evidence of a difference in the relative effect between these three subgroups. Although the point estimate was highest in the no‐contact control subgroup, confidence intervals overlapped. In the first version of this review, although the point estimates were similar between subgroups, the CIs in the dose‐response subgroup did not exclude 1. In this update the confidence interval for this subgroup narrowed (RR 1.18, 95% CI 1.06 to 1.32), although this was sensitive to excluding one study at high risk of bias. Results continue to suggest that the dose‐response curve is shallow for behavioural support. We drew similar conclusions in a companion review to this, which compared combined pharmacotherapy and behavioural support to minimal support; indirect comparisons between trials using more and less intensive behavioural interventions also failed to detect large differences (Stead 2012a). The present review also detected a clearer benefit of more support in studies where all contact was delivered by telephone, but this too was not prespecified and may reflect the larger size of trials done in quitline settings, or possibly that most of these studies did not use biochemical validation of abstinence.
One explanation for the relatively small impact of providing more behavioural support is that it is not provided at the time when it could be most effective. Relapse after initial success is the norm for quit attempts, and by the time people are getting additional calls they may already have relapsed. Various study authors commented on this (e.g. Reid 1999; Smith 2001). Although these studies are not typically characterized as being about 'relapse prevention' there is a small overlap between this review and the Cochrane review of relapse prevention interventions (Hajek 2009), which concluded that there was no evidence of a benefit of additional behavioural support to prevent relapse. On the other hand, in some cases an initial benefit of the intervention disappeared once treatment ended, and authors suggested that further extended support might have made a difference (e.g. Solomon 2000; Killen 2008), although replication of one of these studies with more extended support (Solomon 2005) still showed the same pattern of late relapse. We did not include two studies by Hall and colleagues which tested behavioural support and pharmacotherapy extended for a year from quit date in a factorial design (Hall 2004; Hall 2011). Including these studies would not have affected the findings. One trial that was excluded because the total duration of support did not differ (Garvey 2012a) compared six sessions of counselling delivered over two weeks with six sessions over six weeks; this trial did detect a benefit for continuous abstinence, supporting the theory that the timing of support may merit more investigation.
An earlier Cochrane review of individual, face‐to‐face counselling (Lancaster 2005) did not detect a benefit of more intensive over less intensive support. Four of the five studies in that comparison offered pharmacotherapy and so also contribute to this review, so the most likely explanation for the difference is the greater amount of evidence here, including large studies using telephone counselling.
A potential limitation of the review is that we focused on the amount of behavioural support rather than the specific components, or the quality of delivery. If shorter interventions happened to include more effective components, an impact of more intensive interventions could be obscured. However, there is little evidence about the effect of specific components beyond the intra‐treatment social support common to all these interventions. Michie and West have developed a taxonomy of behaviour change techniques (Michie 2011) and also shown a significant association between the use of some of these techniques and short‐term outcomes in the English Stop Smoking Services (West 2010). The size of associations with individual techniques was relatively small. Given the large number of potential techniques, and rather small range of effect sizes in the trials in the review, we think it unlikely that any single technique, or group of techniques, could be identified that would critically affect outcomes.
Overall completeness and applicability of evidence
The studies identified for this review have all been conducted in the USA or Europe. It is possible that we have failed to find relevant studies conducted in other places. Participants were typically moderate to heavy smokers and were interested in quitting. Most studies recruited participants who had already tried to quit a number of times. One exception was Wu 2009, which recruited Chinese Americans, many of whom did not report previous quit attempts. This study showed one of the largest effect sizes, suggesting that the additional behavioural support could be at least as effective for people who have not already tried many times. It is possible that the explanation for the relatively small effect of the behavioural adjunct overall was that many participants in the trials were already familiar with the approaches being used.
Quality of the evidence
While we judged most of the trials to be at low or unclear risk of bias, we rated six studies at high risk of bias due to inadequate sequence generation, allocation concealment, or loss to follow‐up (Gifford 2011; Calabro 2012; Brown 2013; Berndt 2014; Bock 2014; Yalcin 2014) as discussed above. Sensitivity analysis excluding studies at high risk of bias did not change the overall effect, and neither did also excluding studies for which risk of bias was unclear for any item, although the confidence intervals widened. The quality of the trials was typical of smoking cessation research in general. We did not formally evaluate whether there was a risk of performance bias due to a lack of blinding of providers or participants. Blinding of providers would not have been possible, and it was difficult to determine whether participants knew how their treatment compared to the other options offered. All participants were getting an active pharmacotherapy and would have been aware of this (apart from a small proportion in placebo‐controlled factorial studies). Expectancy effects for the behavioural components would probably have been small, and we do not think the small effect of the interventions could be attributed entirely to higher expectancies in intervention conditions.
The absence of biochemical validation of self‐reported abstinence in some studies did complicate the interpretation of the results in the first version of the review, where a sensitivity analysis excluding them was no longer significant. In this update the effect estimate was a little smaller for the studies with validation, but the CI still excluded 1. A requirement for validation will almost always reduce the absolute success rates, but absence of validation will only lead to an overestimate of relative effects if intervention participants are more likely to misreport abstinence. All trial participants were receiving an active treatment but some of the trials without validation had control groups with no personal contact, whilst intervention groups had telephone contact (seven trials) or face‐to‐face contact (one trial). It is theoretically possible that this could make people in the intervention condition more liable to misreport their status. It is generally considered that biochemical validation is not needed in adult population studies if there is no face‐to‐face contact. In these settings it is argued that the proportion of people misreporting abstinence is low and similar for intervention and control conditions, whilst difficulties in collecting samples by post may lead to underestimates of quitting success (Hollis 2007).
A funnel plot (not shown) did not suggest the presence of publication bias. Given the large number of included studies and the degree of homogeneity between them, it is unlikely that smaller unpublished studies showing no effect, if they existed, would significantly alter our results.
Potential biases in the review process
We used the Cochrane Tobacco Addiction Specialised Register to identify studies. The Register includes reports of trials identified from the major bibliographic databases. There is no straightforward term for the type of intervention we were interested in but we screened any trial report that mentioned a pharmacotherapy. It is possible that the Register does not include all relevant trial reports or that we failed to identify some. Our methods for data extraction and analysis are those used for other Cochrane reviews. The practice of imputing missing data as smoking has been traditionally used for primary and secondary research in smoking cessation and has the advantage that absolute cessation rates are not inflated by ignoring loss to follow‐up. Bias in the relative effect will only be introduced if misclassification differs for people who are lost from the intervention condition compared to the control. If proportionately more of those who are lost in the control group are assumed to be smokers but have in fact quit, then the treatment effect would be overestimated.
Agreements and disagreements with other studies or reviews
The major source of systematic data about the dose‐response to behavioural support is the US Public Heath Service Clinical Practice Guideline, last updated in 2008 (Fiore 2008). This includes meta‐analyses (last updated in 2000) for different levels of support and contact time. The analyses included trials of different levels of support versus control. These showed trends towards increasing effects in trials that had more sessions and more contact time, compared to minimal conditions. For example, estimated effects compared to minimal contact differed between trials with four to 30 minutes of contact time (OR 1.9, 95% CI 1.5 to 2.3) and trials with 91 to 300 minutes (OR 3.2, 95% CI 2.3 to 4.6) (Fiore 2008 Table 6.9) and between two to three treatment sessions (OR 1.4, 95% CI 1.1 to 1.7) and over eight sessions (OR 2.3, 95% CI 2.1 to 3.0) compared to 0 ‐ 1 sessions (Fiore 2008 Table 6.10). These analyses are not limited to direct (within trial) comparisons of treatment intensity. They also do not distinguish between studies with and without pharmacotherapy, and the majority of studies in our analysis were published after 2000 so would not have been included. Our review is likely to give a more precise estimate of the effect of additional support alongside pharmacotherapy, based on the analysis of trials directly comparing different levels of support.
There is observational evidence that access to more behavioural support is associated with greater success in quitting. For example, a study of English Stop Smoking Services, in which there is a high use of pharmacotherapy, found a positive association between the number of scheduled sessions and short‐term quit rates (West 2010). A study of NRT users calling the California quitline found that people who received multiple sessions of counselling had higher quit rates after one year (Zhu 2000).
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Effect of increasing behavioural support. Abstinence at longest follow‐up.
Subgroups by type of pharmacotherapy
Comparison 1 Effect of increasing behavioural support. Abstinence at longest follow‐up, Outcome 1 Subgroups by type of pharmacotherapy.
Comparison 1 Effect of increasing behavioural support. Abstinence at longest follow‐up, Outcome 2 Subgroups by contrast in number of contacts between intervention & control.
Comparison 1 Effect of increasing behavioural support. Abstinence at longest follow‐up, Outcome 3 Subgroups by duration of contact in control condition (not prespecified).
Comparison 1 Effect of increasing behavioural support. Abstinence at longest follow‐up, Outcome 4 Subgroup by modality of intervention contact (not prespecified).
Comparison 2 Sensitivity analyses, Outcome 1 Sensitivity analysis including intermediate intensity conditions. Adjunct behavioural support versus pharmacotherapy alone.
Comparison 2 Sensitivity analyses, Outcome 2 By outcome definition.
| Behavioural interventions as adjuncts to pharmacotherapy for smoking cessation | ||||||
| Patient or population: People using smoking cessation pharmacotherapy | ||||||
| Outcomes | Illustrative absolute effects* (95% CI) | Relative effect | No of Participants | Quality of the evidence | Comments | |
| Assumed successful quitters without intervention | Estimated quitters with intervention | |||||
| Pharmacotherapy (with variable level of behavioural support) | Additional behavioural support (in addition to pharmacotherapy) | |||||
| Smoking cessation at longest follow‐up | Study population1 | RR 1.17 | 18,682 | ⊕⊕⊕⊕ | Updating the review in 2015 with an additional 9 studies (3338 participants) has not materially altered the estimated effect or CI; previous RR 1.16 (95% CI 1.09 to 1.24). | |
| 180 per 1000 | 210 per 1000 | |||||
| The estimated rate of quitting with behavioural intervention (and its 95% confidence interval) is based on the assumed quit rate in the control group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| 1Based on the control group crude average of 1705 quitters out of 9487 participants, which is also the median control group rate, 18% 2Few studies were rated at high risk of bias for any domain. Sensitivity analysis limited to studies at low risk of bias on all domains tended to increase the effect estimate 3Little overall evidence of statistical heterogeneity (I² = 18%), or of differences between the subgroups defined by pharmacotherapy | ||||||
| Study | Control | Intervention | Pharmacotherapy | ||||||
| Sessions | Duration (minutes) | Sessions | Duration (minutes) | ||||||
| Number | Category | Total | Category | Number | Category | Total | Category | ||
| No sessions of personal contact in control condition | |||||||||
| 0 | 0 | 0 | 0 ‐ 3 | up to 9 (av 5 for acceptors) | 4 ‐ 8 | est 10 min/sess | 31 ‐ 90 | NRT or bupropion | |
| 0 | 0 | 0 | 0 ‐ 3 | 5 | 4 ‐ 8 | 60 mins total | 31 ‐ 90 | Nicotine patch (2w) | |
| 0 | 0 | 0 | 0 ‐ 3 | 5 | 4 ‐ 8 | 15 ‐ 25 min/sess | 91 ‐ 300 | Nicotine patch | |
| 0 | 0 | 0 | 0 ‐ 3 | 6 | 4 ‐ 8 | est 15 ‐ 20 min/sess | 91 ‐ 300 | NRT or bupropion | |
| 0 | 0 | 0 | 0 ‐ 3 | up to 12, av 7 | > 8 | average 9 min/sess | 31 ‐ 90 | Nicotine patch | |
| 0 | 0 | 0 | 0 ‐ 3 | average 8 so target higher | > 8 | av 10 min/sess | 91 ‐ 300 | Nicotine patch | |
| 1‐3 sessions of personal contact in control condition | |||||||||
| 2 | 1 ‐ 3 | ≤5 mins/sess | 4 ‐ 30 | 2 | 1 ‐ 3 | 60 ‐ 120 min/sess | 31 ‐ 90 | Nicotine patch | |
| 1 ± 2 | 1 ‐ 3 | 15 ‐ 25 min | 4 ‐ 30 | 1 ± 2 | 1 ‐ 3 | 55 ‐ 75 min | 31 ‐ 90 | Nicotine patch | |
| 1 | 1 ‐ 3 | brief | 4 ‐ 30 | 1 + 1 ‐ 2 | 1 ‐ 3 | est 60 mins | 31 ‐ 90 | Nicotine patch | |
| 1 ‐ 2 | 1 ‐ 3 | brief | 4 ‐ 30 | 2 ‐ 3 | 1 ‐ 3 | est 60 ‐ 90 mins | 31 ‐ 90 | Nicotine patch | |
| 1 | 1 ‐ 3 | 20 mins (mean) | 4 ‐ 30 | 1 + up to 4 | 4 ‐ 8 | average 47 or 60 tot | 31 ‐ 90 | Nicotine patch | |
| 1 | 1 ‐ 3 | brief | 4 ‐ 30 | 1 + 3 | 4 ‐ 8 | > 45 min | 31 ‐ 90 | Nicotine patch | |
| 2 | 1 ‐ 3 | 10 + 5 mins | 4 ‐ 30 | 2 + 3 | 4 ‐ 8 | 15 + 30 | 31 ‐ 90 | Nicotine gum | |
| 1 | 1 ‐ 3 | brief | 4 ‐ 30 | 4 | 4 ‐ 8 | 'brief calls' | 31 ‐ 90 | Bupropion | |
| 1 | 1 ‐ 3 | 5 ‐ 10 mins | 4 ‐ 30 | 5 | 4 ‐ 8 | av. tot 60 min | 31 ‐ 90 | Varenicline | |
| 1 | 1 ‐ 3 | 10 mins | 4 ‐ 30 | 6 | 4 ‐ 8 | est 210 mins | 91 ‐ 300 | Nicotine patch | |
| 1 | 1 ‐ 3 | brief | 4 ‐ 30 | 7 | 4 ‐ 8 | 285 min | 91 ‐ 300 | Nicotine patch | |
| 1 | 1 ‐ 3 | 20 mins | 4 ‐ 30 | 3 ‐ 4 | 4 ‐ 8 | 60 min/sess | 91 ‐ 300 | Nicotine patch | |
| 2 | 1 ‐ 3 | est < 30 | 4 ‐ 30 | 2 + 5 | 4 ‐ 8 | est 30 ‐ 60 mins | 91 ‐ 300 | Nicotine gum | |
| 2 | 1 ‐ 3 | est 30 | 4 ‐ 30 | 5 | 4 ‐ 8 | 120 mins + phone | 91 ‐ 300 | Nicotine patch | |
| 1 | 1 ‐ 3 | 10 ‐ 15 min | 4 ‐ 30 | 6 | 4 ‐ 8 | 90 ‐ 120 min | 91 ‐ 300 | Nicotine patch | |
| 1 | 1 ‐ 3 | brief | 4 ‐ 30 | 3 or 8 (not incl base) | 4 ‐ 8 | ? Or 1 hr ss | > 300 | Nicotine patch | |
| 1 | 1 ‐ 3 | brief | 4 ‐ 30 | 6 | 4 ‐ 8 | 240 ‐ 360 min | > 300 | Nicotine patch or gum | |
| 3 | 1 ‐ 3 | 45 mins | 31 ‐ 90 | 3 + 3 | 4 ‐ 8 | 45 + ?30 | 31 ‐ 90 | Nicotine patch | |
| 1 | 1 ‐ 3 | 90 mins incl 'orientation' | 31 ‐ 90 | 1 + 4 | 4 ‐ 8 | 10 ‐ 15 min/sess | 91 ‐ 300 | Nicotine patch | |
| 1 | 1 ‐ 3 | 30 mins | 4 ‐ 30 | 1 + 3 + 12 | > 8 | >300 | > 300 | Nicotine patch | |
| 1 | 1 ‐ 3 | 15 mins | 4 ‐ 30 | 1 + 9 | > 8 | >500 | > 300 | Bupropion | |
| 3 | 1 ‐ 3 | 5 ‐ 10 min/sess | 4 ‐ 30 | 3 + 6 | > 8 | 90 min/sess plus control | > 300 | Nicotine patch | |
| 1 | 1 ‐ 3 | 1 hour 'medication instruction group' | 31 ‐ 90 | 20 | > 8 | > 300 | > 300 | Bupropion | |
| 4‐8 sessions of personal contact in control condition | |||||||||
| 4 | 4 ‐ 8 | ˜80 mins | 31 ‐ 90 | 7 | 4 ‐ 8 | ˜140 | 91 ‐ 300 | Nicotine patch | |
| 4 | 4 ‐ 8 | 60 min/sess (health education) | 91 ‐ 300 | 4 | 4 ‐ 8 | 60 min/sess | 91 ‐ 300 | Nicotine patch | |
| 5 | 4 ‐ 8 | 2½ hrs | 91 ‐ 300 | 7 | 4 ‐ 8 | 210 + 50 | 91 ‐ 300 | Nicotine sublingual tablet | |
| 4 | 4 ‐ 8 | 60 min/sess | 91 ‐ 300 | 8 | 4 ‐ 8 | 60 min/sess | > 300 | NRT | |
| 5 | 4 ‐ 8 | 45 min/sess | > 300 | 5 | 4 ‐ 8 | 90 min/sess plus control | > 300 | Nicotine gum | |
| 4 | 4 ‐ 8 | 5 ‐ 10 min/sess | 31 ‐ 90 | 10 | > 8 | 10 ‐ 15 *10 | 91 ‐ 300 | NRT or bup | |
| 4 | 4 ‐ 8 | 25 ‐ 35 mins total | 31 ‐ 90 | 4 + 5 | > 8 | 90 min/sess plus control | > 300 | Bup or Nort | |
| 4 | 4 ‐ 8 | 45 mins | 91 ‐ 300 | 14 | > 8 | 75 min/sess | > 300 | Nicotine gum | |
| 4 | 4 ‐ 8 | est 30 min/sess | 91 ‐ 300 | 4 + 16 | > 8 | 45 min/sess | > 300 | Nicotine patch | |
| 5 | 4 ‐ 8 | 60 min/sess | 91 ‐ 300 | 14 | > 8 | 75 min/sess | > 300 | Nicotine gum | |
| 5 | 4 ‐ 8 | 60 min/sess | 91 ‐ 300 | 11 | > 8 | 20 ‐ 40 min/sess | > 300 | Bupropion & gum | |
| 6 | 4 ‐ 8 | 30 min/sess | 91 ‐ 300 | 6 + 4 | > 8 | 30 min/sess | > 300 | Bupropion & patch | |
| 5 | 4 ‐ 8 | 90 min/sess | > 300 | 10 | > 8 | 120 min/sess | > 300 | Nicotine gum | |
| 5 | 4 ‐ 8 | 90 min/sess | > 300 | 10 | > 8 | 120 min/sess | > 300 | Nortriptyline | |
| 7 | 4 ‐ 8 | 560 min | > 300 | 15 | > 8 | 840 min | > 300 | Nicotine patch | |
| Over 8 sessions of personal contact in control condition | |||||||||
| 13 | > 8 | est 10 min/sess | 91 ‐ 300 | 13 incl 8 counselling | > 8 | extra 80 mins | > 300 | Bupropion | |
| 10 | > 8 | 150 min | 91 ‐ 300 | 17 | > 8 | 770 min | > 300 | NRT(gum/patch), bupropion, or varenicline | |
| 9 | > 8 | 20 min/sess | > 300 | 24 | > 8 | 45 min/sess | > 300 | Nicotine patch | |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Subgroups by type of pharmacotherapy Show forest plot | 47 | 18682 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.17 [1.11, 1.24] |
| 1.1 NRT (Nicotine patch unless specified in Footnote) | 34 | 12295 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.14 [1.06, 1.23] |
| 1.2 Bupropion | 5 | 2298 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.27 [1.10, 1.46] |
| 1.3 Nortriptyline | 2 | 172 | Risk Ratio (M‐H, Fixed, 95% CI) | 0.98 [0.59, 1.63] |
| 1.4 Varenicline | 1 | 800 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.11 [0.89, 1.37] |
| 1.5 NRT & bupropion | 2 | 690 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.22 [0.98, 1.52] |
| 1.6 Choice of pharmacotherapy | 4 | 2427 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.28 [1.07, 1.53] |
| 2 Subgroups by contrast in number of contacts between intervention & control Show forest plot | 47 | 18682 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.17 [1.11, 1.24] |
| 2.1 4 to 8 or > 8 contacts versus no contact | 6 | 3762 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.25 [1.08, 1.45] |
| 2.2 More than 8 contacts versus 1 to 3 contacts | 4 | 1063 | Risk Ratio (M‐H, Fixed, 95% CI) | 0.98 [0.76, 1.26] |
| 2.3 4 to 8 contacts versus 1 to 3 contacts | 15 | 8466 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.13 [1.05, 1.23] |
| 2.4 More than 8 contacts versus 4‐8 contacts | 10 | 1617 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.15 [0.99, 1.35] |
| 2.5 Intervention & control in same contact category | 12 | 3774 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.32 [1.14, 1.53] |
| 3 Subgroups by duration of contact in control condition (not prespecified) Show forest plot | 47 | 18682 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.17 [1.11, 1.24] |
| 3.1 No contact for control | 6 | 3762 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.25 [1.08, 1.45] |
| 3.2 'Brief intervention' for control | 20 | 9754 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.15 [1.06, 1.24] |
| 3.3 'Dose response', over 30 minutes contact for control | 21 | 5166 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.18 [1.06, 1.32] |
| 4 Subgroup by modality of intervention contact (not prespecified) Show forest plot | 47 | 18682 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.17 [1.11, 1.24] |
| 4.1 Intervention delivered by telephone | 6 | 5311 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.28 [1.17, 1.41] |
| 4.2 Intervention included face‐to‐face contact | 41 | 13371 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.12 [1.04, 1.20] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Sensitivity analysis including intermediate intensity conditions. Adjunct behavioural support versus pharmacotherapy alone Show forest plot | 47 | 21038 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.16 [1.10, 1.23] |
| 1.1 NRT | 34 | 14011 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.12 [1.05, 1.21] |
| 1.2 Bupropion | 5 | 2298 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.27 [1.10, 1.46] |
| 1.3 Nortriptyline | 2 | 172 | Risk Ratio (M‐H, Fixed, 95% CI) | 0.98 [0.59, 1.63] |
| 1.4 Varenicline | 1 | 1202 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.11 [0.92, 1.34] |
| 1.5 NRT & bupropion | 2 | 690 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.22 [0.98, 1.52] |
| 1.6 Choice of pharmacotherapy | 4 | 2665 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.27 [1.07, 1.51] |
| 2 By outcome definition Show forest plot | 47 | 18740 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.17 [1.11, 1.24] |
| 2.1 12m validation PP outcomes only | 19 | 5234 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.05 [0.94, 1.18] |
| 2.2 12m validated sustained outcomes | 7 | 2322 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.09 [0.87, 1.36] |
| 2.3 Not 12m, but validated | 12 | 3612 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.29 [1.13, 1.47] |
| 2.4 No validation at all | 9 | 7572 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.21 [1.12, 1.31] |
