Exercise interventions for smoking cessation
Abstract
Background
Taking regular exercise, whether cardiovascular‐type exercise or resistance exercise, may help people to give up smoking, particularly by reducing cigarette withdrawal symptoms and cravings, and by helping to manage weight gain.
Objectives
To determine the effectiveness of exercise‐based interventions alone, or combined with a smoking cessation programme, for achieving long‐term smoking cessation, compared with a smoking cessation intervention alone or other non‐exercise intervention.
Search methods
We searched the Cochrane Tobacco Addiction Group Specialised Register for studies, using the term 'exercise' or 'physical activity' in the title, abstract or keywords. The date of the most recent search was May 2019.
Selection criteria
We included randomised controlled trials that compared an exercise programme alone, or an exercise programme as an adjunct to a cessation programme, with a cessation programme alone or another non‐exercise control group. Trials were required to recruit smokers wishing to quit or recent quitters, to assess abstinence as an outcome and have follow‐up of at least six months.
Data collection and analysis
We followed standard Cochrane methods. Smoking cessation was measured after at least six months, using the most rigorous definition available, on an intention‐to‐treat basis. We calculated risk ratios (RRs) and 95% confidence intervals (CIs) for smoking cessation for each study, where possible. We grouped eligible studies according to the type of comparison, as either smoking cessation or relapse prevention. We carried out meta‐analyses where appropriate, using Mantel‐Haenszel random‐effects models.
Main results
We identified 24 eligible trials with a total of 7279 adult participants randomised. Two studies focused on relapse prevention among smokers who had recently stopped smoking, and the remaining 22 studies were concerned with smoking cessation for smokers who wished to quit. Eleven studies were with women only and one with men only. Most studies recruited fairly inactive people. Most of the trials employed supervised, group‐based cardiovascular‐type exercise supplemented by a home‐based exercise programme and combined with a multi‐session cognitive behavioural smoking cessation programme. The comparator in most cases was a multi‐session cognitive behavioural smoking cessation programme alone. Overall, we judged two studies to be at low risk of bias, 11 at high risk of bias, and 11 at unclear risk of bias.
Among the 21 studies analysed, we found low‐certainty evidence, limited by potential publication bias and by imprecision, comparing the effect of exercise plus smoking cessation support with smoking cessation support alone on smoking cessation outcomes (RR 1.08, 95% CI 0.96 to 1.22; I2 = 0%; 6607 participants). We excluded one study from this analysis as smoking abstinence rates for the study groups were not reported. There was no evidence of subgroup differences according to the type of exercise promoted; the subgroups considered were: cardiovascular‐type exercise alone (17 studies), resistance training alone (one study), combined cardiovascular‐type and resistance exercise (one study) and type of exercise not specified (two studies). The results were not significantly altered when we excluded trials with high risk of bias, or those with special populations, or those where smoking cessation intervention support was not matched between the intervention and control arms. Among the two relapse prevention studies, we found very low‐certainty evidence, limited by risk of bias and imprecision, that adding exercise to relapse prevention did not improve long‐term abstinence compared with relapse prevention alone (RR 0.98, 95% CI 0.65 to 1.47; I2 = 0%; 453 participants).
Authors' conclusions
There is no evidence that adding exercise to smoking cessation support improves abstinence compared with support alone, but the evidence is insufficient to assess whether there is a modest benefit. Estimates of treatment effect were of low or very low certainty, because of concerns about bias in the trials, imprecision and publication bias. Consequently, future trials may change these conclusions.
PICOs
Plain language summary
Can exercise help people quit smoking?
Background
We reviewed the evidence about whether exercise helps people who want to quit smoking, or have recently stopped smoking, to stop smoking for at least six months. Taking regular exercise may help people give up smoking by helping with cigarette withdrawal and cravings, and by helping them to manage weight gain, which can be a concern among people trying to quit.
Study characteristics
We found 24 studies with a total of 7279 people. Two studies focused on helping those who had recently stopped smoking and the rest of the studies included current smokers who wished to quit. All the studies were conducted with adults. Eleven studies were with women only and one with men only. Most studies recruited fairly inactive people. Most studies offered supervised and group‐based, aerobic‐type exercise. The evidence is up‐to‐date to May 2019.
Key results
When we combined the results of 21 studies (6607 participants) which compared exercise and smoking‐cessation programmes to smoking cessation programmes alone, there was no evidence that exercise increased quit rates at six months or longer. There was no evidence that the effect was different for different types of exercise. When we combined results from two studies (453 participants), there was no evidence that exercise helped people who had recently quit to stay quit.
Quality of evidence
We judged the quality of evidence for whether exercise programmes help people quit smoking as low certainty, suggesting that future research could change these results. The low certainty is because we cannot rule out chance as an explanation for the suggested slight benefit. It could be that exercise may not help at all, or it could be that supporting people to do exercise modestly increases quit rates. We do not know which of these is true. We also consider that a good number of the trials may be biased. We have concerns that small studies which found smaller effects were less likely to be published than small studies which found bigger effects, making the average result misleading. We judged the evidence from two studies examining whether exercise helps people to avoid relapse to smoking to be of very low certainty, again suggesting that more research is needed. This is due to imprecision of the estimated effects and a high risk of bias in the methods used by one of the studies.
Authors' conclusions
Summary of findings
| Exercise interventions for smoking cessation | ||||||
| Patient or population: People who smoke or who have recently quit | ||||||
| Outcomes | Anticipated absolute effects* (95% CI) | Relative effect | № of participants | Certainty of the evidence | Comments | |
| Risk with smoking cessation support only | Risk with Exercise programme and smoking cessation support or exercise programme alone | |||||
| Smoking abstinence at longest follow‐up | Study population | RR 1.08 | 6607 | ⊕⊕⊝⊝ | Results were not sensitive to the removal of 2 studies where cessation support was not matched between arms (e.g. potential risk of confounding), nor were they sensitive to the removal of the 6 studies in special population groups. In one of these studies, the intervention group was not provided with smoking cessation support | |
| 126 per 1000 | 136 per 1000 | |||||
| Relapse prevention at longest follow‐up | Study population | RR 0.98 | 453 | ⊕⊝⊝⊝ | ‐ | |
| 164 per 1000 | 160 per 1000 | |||||
| *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). The assumed risk in the comparison group is a weighted average of the quit rates of the control arms in the included studies. | ||||||
| GRADE Working Group grades of evidence | ||||||
| aNot downgraded on risk of bias. Sensitivity analysis excluding 10 studies judged to be at high risk of bias was consistent with overall effect, although point estimate showed an increase in favour of intervention (RR 1.25, 95% CI 0.99 to 1.58). | ||||||
Background
Description of the condition
Tobacco use is a leading cause of preventable illness and death worldwide, accounting for over seven million deaths annually (GBD 2015 Risk Factors Collaborators 2016). Based on current smoking trends, there will be approximately 400 million tobacco‐related deaths between 2010 and 2050, mostly among current smokers (Jha 2011). Most smokers would like to stop (CDC 2017); however, quitting is difficult and there is a need to develop more effective interventions.
Description of the intervention
In this review, the exercise interventions focus on more formal, structured activities, such as using a stationary cycle, although some of the interventions promote 'lifestyle' activities, such as walking. Mode of exercise tends to be described as either predominantly cardiovascular (e.g. walking, stationary cycling), where there is an emphasis on improving cardiovascular fitness, or resistance‐based (e.g. weight training), where the emphasis is on developing strength. Some interventions combine cardiovascular and resistance training, while others focus on one or the other.
Besides having the potential to help individuals to stop smoking and to avoid relapse, exercise interventions have the bonus that, if regular exercise is maintained, they have many general health benefits. These benefits have been observed for the general population as well as for people who quit smoking (Albrecht 1998; Korhonen 2011; Shinton 1997) and for people who continue to smoke (Colbert 2001; Hedblad 1997; Senti 2001), and exercise meets the principles of a tobacco harm‐reduction strategy (De Ruiter 2006). For example, physical activity has been negatively associated with lung carcinoma among current and former smokers (Leitzmann 2009) and has been found to reduce oxidative stress in smokers (Koubaa 2015). Also, smokers who adhere to physical activity guidelines show a significant reduction in mortality (Siahpush 2019). In the general population, researchers have shown that regular exercise has benefits for mental health, such as reducing symptoms of depression (Cooney 2013). Smokers who exercise more have been found to have less depression (Vickers 2003; Williams 2008) and exercise has been found to moderate the association between nicotine dependence and depression (Loprinzi 2014).
Smokers trying to quit are attracted to a more physically active lifestyle (Doherty 1998; King 1996), although most smokers are unlikely to spontaneously increase their levels of physical activity after quitting (Allen 2004; Hall 1989; Vander Weg 2001). Being physically active has been positively associated with intention to quit (Frith 2017), initiating a quit attempt (Deruiter 2008; Gauthier 2012; Haddock 2000), with confidence in maintaining smoking abstinence (King 1996) and with success at stopping smoking (Abrantes 2009; Derby 1994; Loprinzi 2016; Paavola 2001; Sedgwick 1988). Among smokers who are not ready to quit, participation in regular physical activity has been associated with reduced cigarette cravings (Haasova 2016). Other work shows a positive trend between avoiding relapse to smoking and physical health and fitness (Metheny 1998), and among those who are more physically active there is evidence for a reduced risk of smoking relapse (López‐Torrecillas 2014; McDermot 2009).
How the intervention might work
Most of the evidence exploring mechanisms for how exercise might aid smoking cessation comes from experimental studies, examining the acute effects of exercise, although it is argued that most of these mechanisms could also apply to long‐term exercise, involving regular bouts of exercise. In experimental studies, temporarily abstinent smokers have consistently been shown to have reduced psychological withdrawal symptoms and desire to smoke following a bout of cardiovascular‐type exercise (Haasova 2013; Haasova 2014; Roberts 2012; Taylor 2007b). Studies with longer bouts of exercise tended to show a more sustained effect on reducing cravings and withdrawal, and further research is needed to understand how exercise dose impacts on the duration of acute effects. However, even brief bouts of exercise, with a brief effect, may work to help abstinent smokers cope with a temporary spike in cravings. Besides exercise potentially helping smokers by reducing self‐reported withdrawal and cravings, several studies have shown that a bout of exercise delays smoking or favourably influences smoking topography, possibly mediated through reduced cravings and withdrawal (De Jesus 2018b; Faulkner 2010; Hatzigeorgiadis 2016; Katomeri 2007; Kurti 2014a; Mikhail 1983; Reeser 1983; Taylor 2007a; Thayer 1993).
The mechanisms underlying these acute beneficial effects of exercise are not clear. Exercise has some similarities to smoking in its effects on stimulating the central nervous system (Russell 1983) and on neurobiological processes (Dishman 2009), including increasing beta‐endorphin levels in smokers (Leelarungrayub 2010), and may provide an alternative reinforcer to smoking; there is evidence that this may also depend on the reinforcing value of physical activity versus smoking for particular individuals (Audrain‐McGovern 2015). It is plausible that attention to somatic cues (e.g. bodily sensations and movements) during exercise could distract smokers from cravings, although one study showed distraction is unlikely to play a major role (Daniel 2006). There is also some evidence that exercise might work by reducing attentional bias to smoking‐related cues that trigger cravings (Janse van Rensburg 2009a; Oh 2014).
Haasova 2014 used an individual participant data meta‐analysis to examine a range of demographic, smoking and other characteristics as potential moderators of the acute effect of exercise on desire to smoke, and examined change in affect as a mediator. None of the characteristics examined were shown to moderate or mediate the effects of exercise. However, it is worth considering the studies in this review, and other studies, which have examined the mediators and moderators of the effects of exercise on smoking‐related outcomes. For example, some studies have examined whether expectancy about the likely effects of exercise influence the effects of exercise; in one study exercise expectancy was modestly associated with psychological symptoms, but not with cigarette craving (Harper 2013), and in another study expectancy did not explain any of the effects of exercise (Daniel 2007). As regards physiological mechanisms, studies have investigated whether exercise might work by overcoming lowered cortisol levels experienced during smoking abstinence (Steptoe 2006); in four studies changes in cortisol concentration were unrelated to changes in cravings (De Jesus 2018a; Janse van Rensburg 2013; Roberts 2015; Scerbo 2010), suggesting that cortisol changes do not mediate the effects of exercise on cravings. Additionally, one study showed that changes in plasma noradrenaline may mediate the effect of exercise on cravings, although heart rate variability was not found to mediate this effect (Roberts 2015). It has been hypothesised that fitness improvements may be beneficial among those attempting to stop smoking (Marcus 1991); however, exercise has the potential to aid cessation without changes in physical capacity, as shown by the above literature on the acute effects of brief bouts of exercise on reducing cigarette cravings.
Taylor 2006a observed that acute reductions in urges to smoke in response to exercise were mediated by reductions in tension. Three studies involving functional magnetic resonance imaging showed that parts of the brain that are typically activated by smoking cues (images) were less activated following a bout of moderate‐intensity exercise (Janse van Rensburg 2009b; Janse van Rensburg 2010; Janse van Rensburg 2012). There is also evidence to suggest that an exercise intervention might work to benefit smokers who are trying to stop smoking by reducing post‐smoking cessation weight gain (Farley 2012), and by reducing cravings for sweet foods (Oh 2014; Teo 2014). Other evidence suggests that regular exercise may facilitate smoking cessation through exercise‐induced increases in smoking‐specific self‐efficacy (Loprinzi 2015) or through fostering a physically active identity (Glowaski 2018; Verkooijen 2008; Taylor 2014). Finally, one methodologically rigorous study has shown that an exercise intervention may be effective for helping smokers to reduce their cigarette consumption (Taylor 2014), although several less rigorous studies did not find any benefit for exercise in reducing cigarette consumption (Bernard 2013; Gorini 2012; Kovelis 2012; Leelarungrayub 2010; McClure 2011; Whiteley 2007; Ybarra 2013).
Why it is important to do this review
There is evidence from large cross‐sectional surveys that levels of physical activity are inversely related to smoking rates, both among adults (Boutelle 2000; Boyle 2000; Hu 2002; Picavet 2010; Schuman 2001; Swan 2018; Takemura 2000) and among adolescents (Aaron 1995; Ali 2015; Coulson 1997; Escobedo 1993; Pate 1996; Peretti‐Watel 2002; Rodriguez 2004; Rodriguez 2008; Verkooijen 2008; Ward 2003). A review of studies examining associations between smoking and physical activity has been published by Kaczynski 2008. However, a review of randomised controlled trials is needed to establish whether these associations are causal. This review of exercise interventions for smoking cessation was first published in 2000, in Addiction (Ussher 2000a), and was converted into a Cochrane Review the same year (Ussher 2000b). Since our last update in 2014, we have become aware of new trials that needed to be considered for inclusion. Furthermore, the Cochrane Tobacco Addiction Group editorial team suggested incorporating a meta‐analysis and other improvements. The aim of this review is therefore to update the evidence in this area, using improved methods.
Objectives
To determine the effectiveness of exercise‐based interventions alone, or combined with a smoking cessation programme, for achieving long‐term smoking cessation, compared with a smoking cessation intervention alone or other non‐exercise intervention.
Methods
Criteria for considering studies for this review
Types of studies
Randomised controlled trials (RCTs) and cluster‐RCTs.
Types of participants
Tobacco smokers wishing to quit, or recent quitters.
Types of interventions
Interventions aimed at increasing exercise, either alone or as an adjunct to a smoking cessation intervention, compared with a smoking cessation programme alone or another type of non‐exercise control group. We excluded interventions which included exercise in a multiple‐component programme, since the specific effects of exercise on smoking abstinence could not be addressed. We therefore excluded yoga‐based interventions, which involved a combination of exercise, meditation and breathing exercises.
Types of outcome measures
Our primary outcome was smoking cessation at the longest follow‐up reported. We excluded trials with less than six months' follow‐up . Where multiple measures of cessation were reported, we preferred continuous/prolonged cessation over point prevalence cessation, and biochemically validated over self‐reported cessation.
As discussed in the Background section, it is postulated that exercise might aid smoking cessation through a range of mechanisms, including effects on tobacco cravings, withdrawal symptoms, or other psychological symptoms, or effects on fitness or physiological or cognitive processes, as well as effects on weight/body mass index (BMI) and reductions in cigarette consumption; we therefore considered any potential mechanisms of action that were examined in the studies.
Search methods for identification of studies
We searched the Cochrane Tobacco Addiction Group Specialised Register in May 2019 for reports of studies including the terms 'exercise' or 'physical activity' in the title, abstract or keywords. The Register has been developed from electronic searching of the Cochrane Central Register of Controlled trials (CENTRAL), MEDLINE, Embase and PsycINFO, together with handsearching of specialist journals, conference proceedings and reference lists of previous trials and overviews. For details of the searches used to create the Specialised Register see the Cochrane Tobacco Addiction Group Website. At the time of the Register search, results from the following databases were included:
• Cochrane Central Register of Controlled trials (CENTRAL), issue 1, 2018;
• MEDLINE (via OVID) to update 20190409;
• Embase (via OVID) to week 201915;
• PsycINFO (via OVID) to update 20190401.
We also searched CINAHL, and Web of Science Indices (SCI‐EXPANDED, SSCI, A&HCI, CPCI‐S, CPCI‐SSH, BKCI‐S, BKCI‐SSH, ESCI) using the terms ‘smoking’, ‘smoking cessation’, ‘exercise’, ‘physical activity’, and ‘intervention’ (searches completed 26 April 2019), and carried out a handsearch of reference lists, conducted searches on key authors, and contacted key authors. In addition, we searched the following online trial registries to identify unpublished studies: ClinicalTrials.gov and the International Clinical Trials Registry Platform (ICTRP).
Data collection and analysis
Selection of studies
Two review authors (from MU, AT, GF, KA) independently screened the title and abstract of each record returned for eligibility. Where there was uncertainty, we put the record forward to the next round of screening. We then retrieved the full‐text reports of any trials considered potentially relevant. Two review authors (from MU, AT, GF, KA) independently assessed the full texts for inclusion, and referred any disagreements to a third review author.
Data extraction and management
Two review authors (from MU, AT, GF, KA) independently extracted the following information about each eligible trial, where available:
• Details of study design, including methods of randomisation and recruitment;
• Participant characteristics, e.g. demographic descriptors (age, sex, ethnicity), cigarette consumption, exercise levels at entry, pre‐existing conditions;
• Description of the exercise intervention(s), including the nature, frequency and duration of exercise;
• Description of comparator(s)/control(s), including the nature, frequency and duration of the smoking cessation programme;
• Rates of exercise adherence and use of techniques to support exercise adherence;
• Primary outcome measures: definition of smoking cessation used for primary outcome, timing of longest follow‐up, any biochemical validation;
• Secondary outcome measures: cigarette consumption, craving and withdrawal symptoms and other psychological symptoms, fitness, BMI and body weight;
• Loss to follow‐up;
• Funding source;
• Declarations of interest.
We then compared and amalgamated extraction for each study, with any disagreements referred to a third review author.
Assessment of risk of bias in included studies
In accordance with the Cochrane guidelines for clinical trials (Higgins 2011; Higgins 2017) and using the Cochrane 'Risk of bias' tool, we assessed studies for risk of selection bias (random sequence generation and allocation sequence concealment), detection bias (blinding of outcome assessment), attrition bias (incomplete outcome data), reporting bias (selective outcome reporting) and other potential sources of bias. Following the standard methods of the Cochrane Tobacco Addiction Group, we rated studies at high risk of detection bias if smoking cessation was not biochemically validated (as the nature of these studies precludes blinding of participants, and in the case of self‐report the participant is the outcome assessor), and at low risk if biochemical validation was used. We did not assess the performance bias domain (blinding of participants and personnel), as blinding of participants and personnel is not feasible due to the nature of the intervention; this domain would not allow us to discriminate between how well the studies were conducted. Two review authors (from MU, AT, GF, KA) independently rated each domain as being at high, low or unclear risk of bias, for each study. We resolved any disagreements through discussion with a third review author. Overall, we considered studies to be at high risk of bias if we rated them at high risk in one or more domains, at low risk if we judged all domains to be at low risk, and at unclear risk otherwise.
Measures of treatment effect
For our primary outcome, we extracted the most stringent definition of smoking cessation for each study (i.e. longest follow‐up, continuous/prolonged versus point prevalence, and biochemically validated versus self‐report). Where possible, we expressed trial effects as a risk ratio (RR), calculated as: (quitters in treatment group/total randomised to treatment group)/(quitters in control group/total randomised to control group), with a 95% confidence interval (CI). A risk ratio greater than 1 indicates a potentially better outcome in the intervention group than in the control group. Potential mechanisms of action (cigarette consumption, craving and withdrawal symptoms and other psychological symptoms, fitness, BMI and body weight) were discussed narratively.
Unit of analysis issues
We considered both individually‐ and cluster‐randomised trials. We deemed no cluster‐randomised trials to be eligible for inclusion.
Dealing with missing data
We conducted our analyses on an intention‐to‐treat basis, i.e. using all participants randomised to their original groups as denominators where data were available, and assuming that those lost to follow‐up were continuing to smoke (West 2005). We extracted numbers lost to follow‐up from study reports and used these to assess the risk of attrition bias. Where any required primary outcome data were not available in study reports, we contacted the authors in an attempt to obtain them.
Assessment of heterogeneity
Before pooling studies, we considered both methodological and clinical variance between studies. Where pooling was deemed appropriate, we investigated statistical heterogeneity using the I2 statistic (Higgins 2003). This describes the percentage of the variability in effect estimates that is due to heterogeneity rather than to sampling error (chance).
Assessment of reporting biases
We used a funnel plot to assess small‐study effects and investigate the possibility of publication bias for the 'exercise versus other smoking cessation treatment’ comparisons.There were not enough studies (fewer than 10) included in the analysis for relapse prevention studies to create a funnel plot.
Data synthesis
For our primary outcome of smoking cessation, we synthesised groups of studies using Mantel‐Haenszel random‐effects models to estimate separate pooled treatment effects (as RRs and 95% CIs), for two types of comparison:
• Effects of exercise versus no exercise intervention, on smoking cessation outcomes (comparison 1, i.e. aimed at current smokers wishing to quit smoking);
• Effects of exercise versus no exercise intervention, on relapse prevention outcomes (comparison 2, i.e. aimed at recent quitters).
Subgroup analysis and investigation of heterogeneity
In view of possible heterogeneity between studies, where relevant and if there were sufficient studies we analysed the trials in subgroups stratified by the type of exercise.
Sensitivity analysis
We carried out the following sensitivity analyses to see if the pooled results of analyses were sensitive to the removal of:
-
Studies judged to be at high risk of bias;
-
Studies with special populations;
-
Studies where the smoking cessation support was not matched between the intervention and control groups.
'Summary of findings' table
Following standard Cochrane methodology (Schünemann 2017), we created a 'Summary of findings' table for the comparison of the effects of exercise versus no exercise intervention on outcomes for smoking cessation and for relapse prevention.
We used the five GRADE considerations (study limitations, consistency of effect, imprecision, indirectness and publication bias) to assess the certainty of the body of evidence for exercise aiding smoking cessation or relapse prevention, and to draw conclusions about the certainty of the evidence within the text of the review.
Results
Description of studies
See Characteristics of included studies; Characteristics of excluded studies; and Characteristics of ongoing studies tables for details of studies.
Results of the search
Our most recent search produced 1957 records. After duplicates were removed, we screened 1772 records for title and abstract. At this stage, we excluded 1705 records, and screened the full text for 67 records. We identified six completed studies and four ongoing studies, and excluded 57 studies. See Figure 1 for the PRISMA flow diagram.
Study flow diagram for 2019 update
Included studies
This review includes 24 individually‐randomised RCTs. We have added six studies since the last version of this review (Bernard 2015; Hassandra 2017; Patten 2017; Prapavessis 2016; Smits 2016; Ussher 2015). We now exclude two studies from the last version of the review, as we decided that they did not meet the inclusion criteria: one was a multi‐component yoga intervention, in which the independent effects of exercise could not be examined (Bock 2012), while the other included some participants who did not wish to stop smoking (Horn 2011). Fourteen studies had more than one associated publication or abstract and these are listed under their study identifier in the reference section.
Two trials were conducted in France (Bernard 2015; Bize 2010), two in New Zealand (Maddison 2014; Prapavessis 2007), two in the UK (Ussher 2003; Ussher 2015), one in Finland (Hassandra 2017), 15 in the USA (Abrantes 2014; Ciccolo 2011; Hill 1993; Kinnunen 2008; Marcus 1991; Marcus 1995; Marcus 1999; Marcus 2005; Martin 1997; McKay 2008; Patten 2017; Russell 1988; Smits 2016; Taylor 1988; Whiteley 2012) and two in Canada (Hill 1985; Prapavessis 2016).
Participants
There was a total of 7279 participants randomised in the included studies, the largest study being an Internet trial with 2318 participants (McKay 2008). Nine trials had fewer than 30 individuals in each treatment arm (Ciccolo 2011; Hassandra 2017; Hill 1985; Hill 1993; Marcus 1991; Marcus 1995; Patten 2017; Russell 1988; Taylor 1988) and three of these were described as pilot or feasibility studies (Ciccolo 2011; Hassandra 2017; Patten 2017). Only nine studies had a sufficiently large sample size to have a good prospect of detecting treatment effects (Bize 2010; Maddison 2014; Marcus 1999; Marcus 2005; Martin 1997; McKay 2008; Prapavessis 2016; Ussher 2003; Ussher 2015).Two studies focused on relapse prevention and recruited those who had recently attempted to stop smoking (Hassandra 2017; Prapavessis 2016); the remaining studies randomised current smokers who wished to quit. Six studies recruited from special populations: one trial recruited people with post‐acute myocardial infarction (AMI) (Taylor 1988); another targeted pregnant smokers (Ussher 2015); two recruited individuals with symptoms reflecting moderate to severe depression (Bernard 2015; Patten 2017); one study involved those with high levels of anxiety sensitivity (Smits 2016); and one was among those recovering from alcohol dependence (Martin 1997). The remaining trials recruited from the general population of smokers.
Eleven trials were limited to women (Kinnunen 2008; Marcus 1991; Marcus 1995; Marcus 1999; Marcus 2005; Patten 2017; Prapavessis 2007; Prapavessis 2016; Russell 1988; Ussher 2015; Whiteley 2012) and one was restricted to men (Taylor 1988). Sixteen studies recorded ethnic status, and all reported a predominantly white sample. Six studies did not present the participants' level of exercise at baseline (Abrantes 2014; Ciccolo 2011; Hill 1985; McKay 2008; Russell 1988; Taylor 1988). All the remaining studies, except Hassandra 2017 and Ussher 2015, reported that they had recruited fairly sedentary participants.
Exercise interventions
Most of the trials used supervised, group‐based cardiovascular‐type exercise supplemented by a home‐based programme, but with some deviation from this formula. Five studies did not provide a home programme (Ciccolo 2011; Marcus 1991; Marcus 1995; Marcus 1999; Smits 2016). Ciccolo 2011 focused exclusively on an individual programme of resistance exercise (i.e. weight training). Whiteley 2012 offered both supervised cardiovascular exercise and resistance exercise. One study used only brief one‐to‐one counselling towards pursuing home‐based exercise, with the type of exercise not specified (Ussher 2003); one focused on telephone‐based counselling towards cardiovascular‐type exercise (Maddison 2014); one provided a web‐based programme designed to encourage engagement in a personalised fitness programme (McKay 2008), although a description of the type of exercise promoted was not provided; and another study focused on providing an app, with brief face‐to‐face instruction, which encouraged short bouts of various types of exercise to manage cigarette cravings (Hassandra 2017).
Only five of the studies promoted exercise as a coping strategy for managing cigarette cravings (Bernard 2015; Hassandra 2017; Patten 2017; Ussher 2003; Ussher 2015). One study promoted exercise as an "opportunity to reestablish a sense of safety around intense bodily sensations" (Smits 2016).
Sixteen studies began the exercise programme before the quit date (Abrantes 2014; Bernard 2015; Bize 2010; Hill 1993; Kinnunen 2008; Marcus 1991; Marcus 1995; Marcus 1999; Marcus 2005; Patten 2017; Prapavessis 2007; Prapavessis 2016; Smits 2016; Ussher 2003; Ussher 2015; Whiteley 2012), three on the quit date (Ciccolo 2011; Hill 1985; Martin 1997), and four after the quit date (Hassandra 2017; Maddison 2014; Russell 1988; Taylor 1988). One study did not state the timing of the exercise programme relative to quit date (McKay 2008).
In Marcus 2005, among participants in the exercise group, those with higher adherence to the exercise prescription were significantly more likely to achieve smoking cessation at the end of treatment than were participants reporting lower adherence to exercise. During the treatment period, researchers used a range of behaviour change techniques to improve adherence to the exercise programme. All the studies included goal‐setting, 14 used self‐monitoring (Abrantes 2014; Bernard 2015; Hassandra 2017; Hill 1985; Kinnunen 2008; Maddison 2014; Martin 1997; Patten 2017; Prapavessis 2016; Russell 1988; Taylor 1988; Whiteley 2012; Ussher 2003; Ussher 2015), six included some element of reinforcement (Bernard 2015; Martin 1997; Patten 2017; Prapavessis 2016; Ussher 2003; Ussher 2015), Hill 1993 used telephone follow‐up in the case of non‐attendance, Prapavessis 2016 offered a telephone maintenance programme, and Taylor 1988 used remote monitoring of heart rate. To assist self‐monitoring, two studies provided pedometers (Maddison 2014; Ussher 2015) and one supplied Kinetic Activity Monitors (Prapavessis 2016). Six trials used comprehensive exercise counselling, including a broad range of behaviour change techniques (Bernard 2015; Maddison 2014; Patten 2017; Prapavessis 2016; Ussher 2003; Ussher 2015).
Where supervised exercise was offered, attendance at these sessions was high, except in two studies where fewer than 50% of sessions were attended (Smits 2016; Ussher 2015). Where the emphasis was on independent/home‐based exercise (Bize 2010; Maddison 2014; Marcus 2005; McKay 2008; Ussher 2003), only a minority of the participants achieved the target level of exercise, except where an exercise app was promoted, in which case exercise levels were significantly higher for the exercise group compared with controls at six‐month follow‐up, despite low uptake of the app (Hassandra 2017). In Patten 2017, only two of 15 participants attended the YMCA post‐intervention to use a free six‐month membership. One study reported greater attrition for the exercise group compared with the controls (Marcus 1999 ‐ see Borrelli 2002). Three studies reported lower attendance for the exercise intervention compared with the control intervention (Kinnunen 2008; Smits 2016; Whiteley 2012).
Besides reporting on attendance at supervised exercise sessions, most studies assessed changes in self‐reported physical activity levels (Abrantes 2014; Bernard 2015; Bize 2010; Ciccolo 2011; Hassandra 2017; Hill 1985; Kinnunen 2008; Maddison 2014; McKay 2008; Patten 2017; Prapavessis 2016; Ussher 2003; Ussher 2015; Whiteley 2012). Of the 11 studies comparing activity levels for exercise versus control groups (Abrantes 2014; Bernard 2015; Bize 2010; Hassandra 2017; Maddison 2014; McKay 2008; Patten 2017; Prapavessis 2016; Ussher 2003; Ussher 2015; Whiteley 2012), seven reported significantly higher activity levels for exercise versus control participants at one follow‐up at least (Abrantes 2014; Bize 2010; Hassandra 2017; Maddison 2014; Ussher 2003; Ussher 2015; Whiteley 2012). Only two studies objectively assessed changes in physical activity (Bernard 2015; Ussher 2015), both using accelerometers, and one of the studies only tested a sub‐sample of participants (Ussher 2015); neither study showed a significant difference in changes in objectively‐assessed activity levels between the study groups.
Smoking cessation programmes/Comparator
Twenty‐two studies included smoking cessation support as the comparator, and two studies had relapse prevention support as the comparator (Hassandra 2017; Prapavessis 2016), with 23 studies offering this support for both exercise and control groups; one study provided this support only for the control group (McKay 2008). Twenty‐one of the studies provided a multisession cognitive behavioural smoking cessation or relapse prevention programme for both intervention and control groups, with one study delivering the programme by telephone (Abrantes 2014), another by face‐to‐face and telephone contact (Maddison 2014), and the rest were all face‐to‐face. Among the remaining three studies, one provided a single cessation session for all participants (Taylor 1988); in another study the exercise group received cessation counselling on alternate weeks whereas the control group only received one brief session of cessation counselling (Bernard 2015), which confounds the effects of exercise; the third study delivered a smoking cessation programme through the Internet, only for the non‐exercise condition (McKay 2008). Twelve studies provided pharmacotherapy for smoking cessation or relapse prevention, which in all cases was matched between intervention and control: eight studies included nicotine patches (Abrantes 2014; Ciccolo 2011; Marcus 2005; Patten 2017; Prapavessis 2007; Prapavessis 2016; Smits 2016; Ussher 2003); one study provided nicotine gum (Kinnunen 2008); two offered several types of nicotine replacement therapy (Bize 2010; Maddison 2014); and one offered patches, gum or varenicline (Bernard 2015).
Twenty of the 22 studies recruiting current smokers set a quit date, and one set a quit date for the non‐exercise condition but did not specify whether or not the exercise group set a quit date (McKay 2008). In fifteen studies the cessation programme began prior to the quit day (Abrantes 2014; Bernard 2015; Hassandra 2017; Hill 1993; Kinnunen 2008; Maddison 2014; Marcus 1999; Marcus 2005; Patten 2017; Prapavessis 2007; Prapavessis 2016; Smits 2016; Ussher 2003; Ussher 2015; Whiteley 2012).
In summary, 22 studies compared exercise plus smoking cessation support to this support alone, and provided an unconfounded estimate of the effect of exercise on smoking cessation. Neither of the two relapse prevention studies provided an unconfounded estimate of exercise plus smoking relapse prevention support compared with this support alone.
Outcomes for smoking abstinence
The strictest measure of abstinence was continuous in 11 studies, prolonged abstinence in three, point prevalence in eight, and was not specified in two. Smoking abstinence was validated objectively in all but three studies (Hassandra 2017; Maddison 2014; McKay 2008). In six studies the method of validation was both measurement of carbon monoxide (CO) in expired air and saliva cotinine (Kinnunen 2008; Marcus 1999; Marcus 2005; Prapavessis 2007; Smits 2016; Ussher 2015), 10 studies used expired CO alone (Abrantes 2014; Bernard 2015; Bize 2010; Ciccolo 2011; Hill 1985; Hill 1993; Martin 1997; Prapavessis 2016; Russell 1988; Ussher 2003), four studies used cotinine alone (Marcus 1991; Marcus 1995; Patten 2017; Whiteley 2012) and one used plasma thiocyanate (Taylor 1988).
Excluded studies
We list 108 studies that were potentially relevant but excluded, with detailed reasons, in the Characteristics of excluded studies table. We summarise reasons for excluding studies at full‐text stage in Figure 1. The reason for excluding most studies at full‐text screening stage was because the study had an ineligible design, and in most cases this was because the study was examining the acute effects of exercise rather than examining effects on smoking cessation outcomes. We have added 56 new excluded studies since the previous version of the review. We also classify seven studies as ongoing (see Characteristics of ongoing studies table), which are likely to be relevant for inclusion once reported.
Risk of bias in included studies
Full details of ’Risk of bias’ assessments are given for each trial within the Characteristics of included studies table. Overall, we judged two studies to be at low risk of bias (low risk of bias across all domains), 11 at high risk of bias (high risk of bias in at least one domain), and the remaining 11 at unclear risk of bias. A summary illustration of the 'Risk of bias’ profile across trials is shown in Figure 2.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Allocation
We assessed selection bias through investigating methods of random sequence generation and allocation concealment for each study. We rated 15 studies as having low risk for random sequence generation, and the remaining nine as having unclear risk. We judged nine studies to be at low risk for allocation concealment, 14 at unclear risk, and one study at high risk (Ussher 2003). We rated Ussher 2003 at high risk as study personnel used a list of random numbers, and therefore knew which condition a person was in before they were randomised. We judged studies as having unclear risk of selection bias when authors provided insufficient information about methods used.
Outcome assessment (detection bias)
We rated 19 studies at low risk for detection bias, as smoking cessation was biochemically validated. We judged two studies to be at unclear risk of detection bias, as we were unsure whether abstinence rates were biochemically verified. We judged three studies to be at high risk of detection bias (Hill 1985; Maddison 2014; McKay 2008), as these studies relied on self‐report alone.
Incomplete outcome data
We judged studies to be at a low risk of attrition bias where the numbers of participants lost to follow‐up were clearly reported, the overall number lost to follow‐up was not more than 50%, and the difference in loss to follow‐up between groups was no greater than 20%. This is consistent with ’Risk of bias’ guidance produced by the Cochrane Tobacco Addiction Group for assessing smoking cessation studies. We judged 13 studies to be at low risk of bias, three at unclear risk and eight at high risk. We judged these eight studies to be at high risk because overall loss to follow‐up was more than 50% (Bernard 2015; Bize 2010; Kinnunen 2008; Marcus 2005; McKay 2008; Prapavessis 2016; Smits 2016), or because the groups had a difference in loss to follow‐up of more than 20% (Prapavessis 2007). We assigned judgements of unclear risk because information on follow‐up was not reported (Martin 1997; Russell 1988; Taylor 1988).
Selective reporting
We rated all the studies at a low risk of reporting bias, as all studies reported the outcomes for smoking cessation as stated in their Methods.
Other potential sources of bias
We did not identify any other sources of bias.
Effects of interventions
See: Summary of findings for the main comparison Exercise interventions for smoking cessation
Smoking abstinence
See: summary of findings Table for the main comparison; Analysis 1.1; and Analysis 1.2.
We included 23 of the 24 studies in analyses; we dropped one study as it did not provide separate abstinence data for the experimental and control groups, although it was reported that no significant difference was found between the groups (Russell 1988).
For the first analysis (see Analysis 1.1), we pooled 21 studies comparing exercise versus no exercise intervention, on smoking cessation outcomes (i.e. aimed at current smokers wishing to quit smoking). This resulted in a pooled risk ratio (RR) of 1.08 (95% CI 0.96 to 1.22; 6607 participants), with no statistical heterogeneity (I2 = 0%). This provides no evidence for an effect of the exercise intervention on smoking cessation. We then analysed the trials in subgroups, stratified by type of exercise. The exercise types/subgroups were: cardiovascular‐type exercise (17 studies, 3635 participants), resistance training (1 study, 25 participants), cardiovascular‐type exercise and resistance training combined (1 study, 330 participants), or exercise type not specified (2 studies, 2617 participants). There was no evidence of subgroup differences (P = 0.72, I2 = 0%).
We then conducted a sensitivity analysis to see if the overall results were sensitive to exclusion of the six studies with special populations; i.e. those with mental health issues (Bernard 2015; Martin 1997; Patten 2017), with other specific psychological symptoms (Smits 2016), with a pregnant population (Ussher 2015), and with a physical health condition (Taylor 1988). Excluding these studies did not affect the interpretation of the results (RR 1.06, 95% CI 0.92 to 1.22; I2 = 0%; 5376 participants). We conducted a further sensitivity analysis removing the 10 studies judged to be at high risk of bias (see Figure 2); the pooled estimate was higher than in the main analysis (RR 1.25, 95% CI 0.99 to 1.58; I2 = 0%; 1802 participants), but the interpretation remained the same. A post hoc sensitivity analysis removing the two studies where smoking cessation support was not matched between the intervention and control arms (Bernard 2015; McKay 2008) also did not affect the results (RR 1.09, 95% CI 0.95 to 1.25; I2 = 0%; 4219 participants).
Next we conducted an analysis pooling the two relapse prevention studies (i.e. aimed at smokers who had recently quit) (Hassandra 2017; Prapavessis 2016). This resulted in a pooled RR of 0.98 (95% CI 0.65 to 1.47; I2 = 0%; 453 participants), providing no evidence for exercise interventions aiding relapse prevention (see Analysis 1.2). We judged one of these studies (Prapavessis 2016) as having high risk of bias.
Potential mechanisms of action
As none of the included studies detected a statistically significant benefit in favour of the intervention we did not assess mechanisms of action as they related to individual study findings. We summarise below the findings related to the all the different mechanisms examined in the studies.
Smoking reduction
Six of the included studies assessed changes in levels of cigarette consumption as a secondary outcome (Bernard 2015; Hill 1985; Taylor 1988; Prapavessis 2007; Maddison 2014; Ussher 2015). None of these studies reported a significant smoking reduction for the exercise versus the control group. Two studies observed significantly lower absolute levels of smoking for an exercise group versus control at 23 week post‐treatment (Taylor 1988) and 24 weeks after quit day (Maddison 2014), although they did not analyse the changes in smoking relative to baseline. One trial reported significantly lower smoking levels for a cessation programme versus an exercise programme (Prapavessis 2007).
Cigarette cravings, withdrawal symptoms and other psychological measures
Nine studies assessed effects on cigarette withdrawal symptoms or cravings or both, and there was little evidence for the exercise interventions having a beneficial effect. Abrantes 2014 reported significantly lower somatic withdrawal symptoms and sleep disturbance for the exercise versus control group during the treatment period; there were no group differences for craving. Ussher 2003 observed a reduction in some withdrawal symptoms for exercise versus controls up to three weeks post‐cessation. Other studies assessing withdrawal symptoms or cravings or both observed no overall effect of the intervention on these outcomes (Bernard 2015; Bize 2010; Hassandra 2017; Kinnunen 2008; Maddison 2014; Marcus 1999; Ussher 2015), although Marcus 1999 observed a beneficial acute reduction in withdrawal and cravings for the exercise versus control group (see Bock 1999).
Nine studies examined symptoms of depression, mood or anxiety. When assessing mood/depression, five studies observed no significant effect of the intervention (Abrantes 2014; Bernard 2015; Bize 2010; Martin 1997; Patten 2017); two of these focused on those with symptoms of moderate to severe depression (Bernard 2015; Patten 2017) and a further study recruited those recovering from alcohol dependence (Martin 1997). Marcus 2005 observed that those who increased their fitness were more likely to report decreases in depressive symptoms (see Williams 2008). Among pregnant smokers, Ussher 2015 observed a significant increase in self‐reports of depression symptoms for the exercise group compared with the control group at the end of pregnancy (see Daley 2018). Based on qualitative work (Giatras 2017), the authors concluded that having to cope with changing two health behaviours simultaneously (i.e. smoking and exercise), while also coping with being pregnant and attending multiple treatment sessions, may have negatively affected participants' mood. Bernard 2015 found no effect of the intervention on reports of anxiety, but Russell 1988 detected a significant increase in tension‐anxiety scores for the active group compared with the controls. Among those with high levels of anxiety sensitivity, Smits 2016 showed that the intervention reduced anxiety sensitivity.
Fitness measures
Thirteen studies examined changes in fitness levels (Abrantes 2014; Bernard 2015; Ciccolo 2011; Kinnunen 2008; Marcus 1991; Marcus 1995; Marcus 1999; Marcus 2005; Patten 2017; Prapavessis 2007; Russell 1988; Taylor 1988; Whiteley 2012 (sub‐sample only)) and five of these showed significant gains in cardiovascular fitness for an exercise versus control group at end of treatment (Bernard 2015; Patten 2017; Prapavessis 2007; Taylor 1988; Whiteley 2012). A further four studies reported significant gains in cardiovascular fitness at end of treatment within an exercise group but not within a control group, although they did not compare the groups statistically (Marcus 1991; Marcus 1995; Marcus 1999 (see also Albrecht 1998); Marcus 2005).
Weight or body mass index
Twelve studies examined weight changes, and one study assessed BMI. Marcus 1999 reported a significantly smaller weight gain for those in the exercise condition compared with the controls at the end of treatment; however, those in the exercise condition weighed more than the controls at baseline, and this difference was not controlled for, which makes interpretation of the finding problematic. This study did not find any significant differences in weight change between the study groups at the three‐month or 12‐month follow‐ups. Prapavessis 2007 observed no difference in weight gain at end of treatment when comparing cognitive‐behavioural support plus nicotine patches with exercise plus nicotine patches; however, at end of treatment those in the exercise‐only condition gained significantly less weight than those receiving only cognitive‐behavioural support. Other studies found no difference in weight gain for the exercise versus controls at end of treatment (Bernard 2015; Marcus 1991; Marcus 1995; Marcus 2005; Ussher 2003; Whiteley 2012), at three‐ and six‐month follow‐ups (Ciccolo 2011), at 12 months post‐cessation (Bize 2010; Ussher 2003) or at end of pregnancy (Ussher 2015). Patten 2017 observed no group difference for BMI at end of treatment.
Discussion
Summary of main results
This update contributed six new studies evaluating exercise programmes for smoking cessation. The review includes 24 trials in total. Twenty‐one of these trials compared a combined exercise and smoking cessation intervention with smoking cessation support alone, one compared exercise alone to smoking cessation support (McKay 2008), and two compared a combined exercise and relapse prevention intervention with relapse prevention alone (Hassandra 2017; Prapavessis 2016). One study could not be included in meta‐analysis (Russell 1988).
A meta‐analysis, pooling all 21 available studies comparing exercise to smoking cessation treatment indicated no evidence of a benefit for the exercise intervention on smoking cessation. We judged this estimate to be of low certainty, as it was imprecise and there was suspected publication bias (see Quality of the evidence; summary of findings Table for the main comparison). There was no evidence of different effects by the type of exercise, and the estimate remained the same when we removed six studies with special populations, and when we removed two studies in which smoking cessation support was not matched between the intervention and control arms. When we removed 10 studies at high risk of bias, the point estimate changed to be more in favour of exercise, but the interpretation was unchanged.
Pooling the two relapse prevention studies provided no evidence for the exercise interventions aiding relapse prevention. Due to one of these studies being rated at high risk of bias and to a lack of precision around the CI, we judged the estimate as being of very low certainty.
Overall completeness and applicability of evidence
There were marked variations between the studies in the length, type, and timing of the exercise intervention, and in the design of the control condition and cessation programme. Despite this variation, the observed effects were very consistent, with no evidence of any statistical heterogeneity. Although pooled analyses did not show evidence of an effect, there are a number of limitations to the evidence base which could potentially have impacted effectiveness and which need to be considered when planning further research in this area.
The studies were mainly conducted in the USA, with others taking place in other high‐income countries; studies are needed in low‐ and middle‐income countries where population smoking rates are higher. Most of the studies were conducted among the general population of smokers, with only six studies among various special populations. These special populations might find it harder to stop smoking and to increase exercise levels, so we conducted a sensitivity analysis excluding the six studies. Reassuringly, results were not sensitive to the exclusion of these studies. Trials are needed among other special populations of smokers who might especially benefit from an exercise intervention. There is a high prevalence of smoking among those with serious mental illness, and those with such disorders are likely to be receptive to exercise as an aid to cessation (Arbour‐Nicitopoulos 2011; Arbour‐Nicitopoulos 2011b; Faulkner 2007); trials are therefore needed in this population. Studies are also needed to examine the effect of exercise interventions on smoking cessation in younger smokers; an excluded study showed that young smokers may benefit from exercise (Horn 2011). People with overweight or obesity who quit smoking may have a particular need for weight control interventions, such as exercise (Lycett 2011), and we have yet to see a trial of exercise focusing on this population, although there was little evidence in this review to suggest that exercise interventions are likely to have an impact on weight change during smoking cessation. However, among the included studies not observing any effect on weight, several were too small to have a realistic chance of detecting a treatment effect (Bernard 2015; Ciccolo 2011; Marcus 1991; Marcus 1995; Patten 2017; Whiteley 2012). Moreover, the studies by Bernard 2015, Bize 2010, Marcus 2005, Patten 2017, Ussher 2003, and Whiteley 2012 included NRT, and post‐cessation weight gain is likely to be less pronounced when using NRT (Farley 2012). The potential for exercise to moderate weight gain was therefore reduced. It is possible that exercise provides a role in weight management once an individual has stopped using NRT, but this has yet to be explored. Only two of the studies recruited reasonably active smokers (Hassandra 2017; Ussher 2015). A substantial proportion of smokers may be physically active (Deruiter 2008; Emmons 1994; Scioli 2009; Ward 2003) and there is some evidence that regular exercisers may be more successful at quitting (Abrantes 2009; Derby 1994; Paavola 2001), but it is not clear whether exercise interventions are effective as an aid to smoking cessation in more active populations.
It has been recommended that a smoking cessation programme should start before the quit date and continue into the period of abstinence (Raw 1998). However, almost half of the trials did not do this (Ciccolo 2011; Hill 1985; Maddison 2014; Marcus 1991; Marcus 1995; Martin 1997; McKay 2008; Russell 1988; Taylor 1988) and one study, testing an app, only provided a single session of support after the quit day (Hassandra 2017).
There is some evidence that exercise is effective for reducing cigarette cravings even when a nicotine lozenge is used (Tritter 2015). Further studies are needed to establish whether exercise offers additional benefits to those provided by NRT and other smoking cessation medications. It is feasible that exercise could address psychosocial and physical needs that are not currently met by NRT‐based programmes. Alternatively, exercise may have limited impact on cigarette cravings if cravings have been reduced by NRT; it is worth noting that most acute exercise studies have shown significantly reduced cravings following experimentally elevated cravings (Haasova 2013).
For those studies beginning exercise either on or after the quit date (Ciccolo 2011; Hassandra 2017; Hill 1985; Maddison 2014; Martin 1997; Russell 1988; Taylor 1988) success rates may have been hampered by the demand to cope simultaneously with two major changes in health behaviour (Emmons 1994; Hyman 2007; King 1996; Patten 2003). In studies where the exercise programme started after a period of smoking abstinence, the potential for exercise to moderate withdrawal symptoms during this period was lost (Haasova 2013; Roberts 2012). A review of the comparative efficacy of simultaneous versus sequential multiple health behaviour change interventions concluded that the approaches should be considered equally efficacious (James 2016). In practice, when the exercise programme begins may depend on individual capabilities and preferences (Everson‐Hock 2010b).
More attention may need to be given to strategies for increasing exercise adherence. In the two studies with exercise programmes lasting for less than six weeks (Hill 1985; Martin 1997), the intervention may have been of insufficient length to encourage long‐term exercise adherence. Where a home/unsupervised exercise programme was not provided it is possible that the participants' high level of dependence on supervised exercise reduced their level of post‐intervention activity. Most of the studies involved cardiovascular‐type exercise and more studies are needed with non‐cardiovascular exercise. For example, isometric exercise has been shown to reduce tobacco cravings and urges to smoke (Ussher 2006; Ussher 2009), and has been successfully piloted (Al‐Chalabi 2008).
Only five studies promoted exercise as a coping strategy for managing cigarette cravings (Bernard 2015; Hassandra 2017; Patten 2017; Ussher 2003; Ussher 2015). The effect of the interventions may have been stronger if exercise were actively presented as such a coping strategy (Taylor 2010). The evidence consistently demonstrates the benefits of an acute bout of exercise on alleviating cravings and withdrawal symptoms under optimum conditions for observing such an effect (i.e. with experimentally manipulated increased baseline cravings – through temporary abstinence, and in some cases in the presence of smoking‐related cues, prior to exercising). As regards relapse prevention, exercise could be presented as a strategy which increases self‐esteem and pride in one's health, and reinforces an identity as a non‐smoker and as a physically active person (Verkooijen 2008; Taylor 2014) in such a way that being a smoker is incompatible with these perceptions. Critically, if exercise aids smoking cessation, it is likely that exercise needs to be maintained for it to continue to support smoking cessation.
Among the included studies, only two offered digital support towards pursuing independent/home‐based exercise (Hassandra 2017; McKay 2008), providing Internet and app‐based support respectively. Further studies need to consider offering these types of support, as well as other common digital support such as text/SMS messaging. Patten 2018 has recently shown that smokers are receptive to robotic‐assisted exercise coaching, and other novel digital interventions need to be explored for supporting independent exercise among smokers who are trying to quit. Only two studies objectively assessed changes in physical activity (Bernard 2015; Ussher 2015); future studies need to assess exercise dosage through including both self‐reported and objective measures of physical activity, as well as recording supervised exercise attendance.
Those adequately‐powered trials not showing a consistent effect of exercise on smoking abstinence (Bize 2010; Maddison 2014; Marcus 2005; McKay 2008; Prapavessis 2016; Ussher 2003; Ussher 2015) promoted moderate‐ or moderate to vigorous‐intensity exercise, rather than vigorous‐intensity exercise. One of these studies relied solely on brief exercise counselling (Ussher 2003), and another focused on telephone‐based counselling (Maddison 2014). In two other studies supervised exercise was only provided once a week (Bize 2010; Marcus 2005), and another study relied on a web‐based programme (McKay 2008). In these studies the exercise intervention may have been insufficiently intense to benefit smoking abstinence. Future studies should consider providing more intensive interventions. Intensity here refers to both the exercise intensity itself (i.e. light, moderate or vigorous) and the extensiveness of the support being provided (e.g. the number of supervised exercise sessions). The findings from Marcus 2005 suggest that abstaining smokers may need to accumulate at least 110 minutes of activity a week to maintain abstinence (at least during the intervention period), and supervised exercise on two or three days a week may be necessary to achieve this. A pilot study showed promising findings for an intervention involving moderate‐intensity exercise supervised on three days a week over eight weeks (Williams 2010); this needs to be tested in a larger trial.
Only five of the studies provided any post‐intervention exercise programming (Hassandra 2017; Hill 1993; Maddison 2014; Prapavessis 2016; Ussher 2003), and this may have increased post‐intervention exercise adherence. Hassandra 2017 made an exercise‐promoting app available for use through to six‐month follow‐up, and observed that self‐reported physical activity levels were higher for the exercise versus control group at six months. Prapavessis 2016 offered telephone counselling up to 12 months post‐treatment and found no difference in activity levels for the study groups between baseline and six or 12 months after treatment. For the other studies providing post‐intervention exercise programming it is not possible to draw any conclusions about whether the intervention affected levels of exercise adherence after the formal supervised programme ended, because none of these studies reported rates of adherence for this period.
Quality of the evidence
As described above and in summary of findings Table for the main comparison, we rated the evidence in this review to be of low certainty for smoking cessation. Although 21 studies contributed to the meta‐analysis of smoking cessation, many of them were small, and imprecision remained an issue, with confidence intervals spanning both no effect and a clinically significant benefit. A funnel plot (Figure 3) was asymmetrical, suggesting the presence of publication bias, with small studies appearing more likely to be published if they detected a greater effect. This sort of asymmetry risks inflating the estimated effect. In addition, we judged 10 of the 21 studies contributing to the main analysis to be at high risk of bias. However, when we removed these studies findings were consistent with the overall analysis, although the point estimate more clearly favoured the intervention.
Funnel plot of comparison: 1 Exercise component versus smoking cessation programme only, outcome: 1.1 Smoking abstinence at longest follow‐up, subgroup by exercise type.
Only two studies contributed to the analysis of relapse prevention; here quality was very low, due to serious imprecision (confidence intervals spanning both a clinically significant benefit and clinically significant harm) and to risk of bias (we judged the larger of the two studies contributing to the analysis to be at high risk of bias).
Overall, our judgement is that the true effect of exercise interventions for smoking cessation may be substantially different from the effect estimate observed when pooling the available data.
Potential biases in the review process
We followed standard Cochrane methods and are unaware of any introduced bias. Three review authors were authors of included studies, but these studies were assessed independently by other members of the author team, again following standard Cochrane methods. Our search strategy included the Cochrane Tobacco Addiction Group Specialised Register and we also searched trial registries and contacted key authors in an attempt to capture unpublished and ongoing studies. There may be unpublished data that our searches did not reveal, and our funnel plot indicates that this may bias results for the comparison for smoking cessation.
Agreements and disagreements with other studies or reviews
The findings are consistent with previous versions of this review and with a previous meta‐analysis of exercise interventions for smoking cessation (Klinsophon 2017). The findings are not consistent with a multitude of 'laboratory' studies showing that brief bouts of exercise can acutely reduce withdrawal and cravings (Haasova 2013; Haasova 2014; Roberts 2012), which presents probably the most plausible explanation of how an exercise intervention might aid smoking cessation or relapse prevention. This inconsistency may be because the acute studies involved temporarily abstinent smokers rather than those attempting to stop smoking completely. Moreover, as this evidence comes from single bouts of acute exercise it is possible that for the laboratory results to be replicated during quit attempts individuals will need to take part in multiple bouts of exercise throughout the day. It is also plausible that the effect of exercise does not last very long and it is not feasible for people to exercise with sufficient frequency through the day. The intervention effect is therefore short‐lived and not that effective. We observed little evidence for exercise having an impact on weight, but, when pooling the results from three studies in this review (Bize 2010; Marcus 1999; Ussher 2003), Farley 2012 found no evidence for exercise moderating weight gain at end of treatment, but reported a benefit at 12‐month follow‐up, concluding that "More studies are needed to clarify whether this is an effect of treatment or a chance finding". Additionally, Spring 2009 conducted a meta‐analysis with 10 studies of weight management interventions during smoking cessation, including five of the studies included in our review (Marcus 1991; Marcus 1995; Marcus 1999; Marcus 2005; Ussher 2003), and observed a significant benefit for the intervention in the short term (less than three months), but not in the long term (more than six months).
Study flow diagram for 2019 update
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Funnel plot of comparison: 1 Exercise component versus smoking cessation programme only, outcome: 1.1 Smoking abstinence at longest follow‐up, subgroup by exercise type.
Comparison 1 Exercise component versus smoking cessation programme only, Outcome 1 Smoking abstinence at longest follow‐up, subgroup by exercise type.
Comparison 1 Exercise component versus smoking cessation programme only, Outcome 2 Relapse prevention at longest follow‐up.
| Exercise interventions for smoking cessation | ||||||
| Patient or population: People who smoke or who have recently quit | ||||||
| Outcomes | Anticipated absolute effects* (95% CI) | Relative effect | № of participants | Certainty of the evidence | Comments | |
| Risk with smoking cessation support only | Risk with Exercise programme and smoking cessation support or exercise programme alone | |||||
| Smoking abstinence at longest follow‐up | Study population | RR 1.08 | 6607 | ⊕⊕⊝⊝ | Results were not sensitive to the removal of 2 studies where cessation support was not matched between arms (e.g. potential risk of confounding), nor were they sensitive to the removal of the 6 studies in special population groups. In one of these studies, the intervention group was not provided with smoking cessation support | |
| 126 per 1000 | 136 per 1000 | |||||
| Relapse prevention at longest follow‐up | Study population | RR 0.98 | 453 | ⊕⊝⊝⊝ | ‐ | |
| 164 per 1000 | 160 per 1000 | |||||
| *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). The assumed risk in the comparison group is a weighted average of the quit rates of the control arms in the included studies. | ||||||
| GRADE Working Group grades of evidence | ||||||
| aNot downgraded on risk of bias. Sensitivity analysis excluding 10 studies judged to be at high risk of bias was consistent with overall effect, although point estimate showed an increase in favour of intervention (RR 1.25, 95% CI 0.99 to 1.58). | ||||||
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Smoking abstinence at longest follow‐up, subgroup by exercise type Show forest plot | 21 | 6607 | Risk Ratio (M‐H, Random, 95% CI) | 1.08 [0.96, 1.22] |
| 1.1 Cardiovascular exercise | 17 | 3635 | Risk Ratio (M‐H, Random, 95% CI) | 1.08 [0.94, 1.24] |
| 1.2 Resistance training | 1 | 25 | Risk Ratio (M‐H, Random, 95% CI) | 1.85 [0.19, 17.84] |
| 1.3 Cardiovascular and resistance | 1 | 330 | Risk Ratio (M‐H, Random, 95% CI) | 1.81 [0.69, 4.78] |
| 1.4 Not specified | 2 | 2617 | Risk Ratio (M‐H, Random, 95% CI) | 1.05 [0.84, 1.32] |
| 2 Relapse prevention at longest follow‐up Show forest plot | 2 | 453 | Risk Ratio (M‐H, Random, 95% CI) | 0.98 [0.65, 1.47] |
