Summary of findings
Tobacco-related deaths and disabilities are on the increase worldwide because of continued use of tobacco (mainly cigarettes). Tobacco use has reached epidemic proportions in many developing countries, while steady use continues in industrialized nations like the United States (The Tobacco Atlas 2012; CDC 2012). The following two factors may help to reduce the prevalence of cigarette smoking. According to the Centers for Disease Control, 68.8% of adult smokers in the USA want to quit and millions have attempted to quit (CDC 2011), with 70% of smokers visiting a healthcare professional each year (AHRQ 2008). Nurses, representing the largest number of healthcare providers worldwide, are involved in the majority of these visits and therefore, have the potential for a profound effect on the reduction of tobacco use (Youdan 2005).
Systematic reviews (e.g. Stead 2013) have confirmed the effectiveness of advice from physicians to stop smoking. The Agency for Health Care Research and Quality Clinical Practice Guideline (AHRQ 2008) lists nurses as one of the many providers from whom advice to stop smoking could increase quit rates, but identifies the effectiveness of advice to quit smoking given by clinicians other than physicians (including nurses) as an area requiring further research. The American Nurses Association (ANA 2012) wrote that nurses have tremendous potential to implement smoking cessation interventions effectively and advance tobacco use reduction goals proposed by Healthy People 2010, and noted that nurses must be equipped to assist with smoking cessation, prevent tobacco use, and promote strategies to decrease exposure to second-hand smoke. The American Nurses Association/American Nurses Foundation promotes the mission of Tobacco-Free Nurses to the nation’s registered nurses through its constituent associations, members, and organizational affiliates (ANA 2012).
A review of nursing's specific role in smoking cessation is essential if the profession is to endorse the International Council of Nurses (ICN) call to encourage nurses to "...integrate tobacco use prevention and cessation ... as part of their regular nursing practice" (ICN 2012).
The aim of this review is to examine and summarize randomized clinical trials where nurses provided smoking cessation interventions. The review therefore focuses on the nurse as the intervention provider, rather than on a particular type of intervention. Smoking cessation interventions targeting pregnant women are not included here, because of the particular circumstances and motivations among this population. Interventions for pregnant smokers have been reviewed elsewhere (Lumley 2009; Coleman 2012).
The primary objective of this review was to determine the effectiveness of nursing-delivered interventions on smoking behaviour in adults. A priori study hypotheses were that nursing-delivered smoking cessation interventions:
(i) are more effective than no intervention;
(ii) are more effective if the intervention is more intensive;
(iii) differ in effectiveness with health state and setting of the participants;
(iv) are more effective if they include follow-ups;
(v) are more effective if they include aids that demonstrate the pathophysiological effect of smoking.
Criteria for considering studies for this review
Types of studies
Inclusion criteria for studies were:
(i) they had to have at least two treatment groups;
(ii) allocation to treatment groups must have been stated to be 'random'.
Studies that used historical controls were excluded.
Types of participants
Participants were adult smokers, 18 years and older, of either gender and recruited in any type of healthcare or other setting. The only exceptions were studies that had exclusively recruited pregnant women. Trials in which 'recent quitters' were classified as smokers were included, but sensitivity analyses were performed to determine whether they differed from trials that excluded such individuals.
Types of interventions
Nursing intervention was defined as the provision of advice, counselling, and/or strategies to help people quit smoking. The review includes cessation studies that compared usual care with an intervention, brief advice with a more intensive smoking cessation intervention or different types of interventions. Studies of smoking cessation interventions as a part of multifactorial lifestyle counselling or rehabilitation were included only if it was possible to discern the specific nature and timing of the intervention, and to extract data on the outcomes for those who were smokers at baseline. Advice was defined as verbal instructions from the nurse to 'stop smoking' whether or not information was provided about the harmful effects of smoking. Interventions were grouped into low and high intensity for comparison. Low intensity was defined as trials where advice was provided (with or without a leaflet) during a single consultation lasting 10 minutes or less with up to one follow-up visit. High intensity was defined as trials where the initial contact lasted more than 10 minutes, there were additional materials (e.g. manuals) and/or strategies other than simple leaflets, and usually participants had more than one follow-up contact. Studies where participants were randomized to receive advice versus advice plus some form of nicotine replacement therapy (NRT) were excluded, since these were primarily comparisons of the effectiveness of NRT rather than nursing interventions. These are covered in a separate review (Stead 2012).
Types of outcome measures
The principal outcome was smoking cessation rather than a reduction in withdrawal symptoms or a reduction in the number of cigarettes smoked. Trials had to report follow-up of at least six months for inclusion in the review. We excluded trials which did not include data on smoking cessation rates. We used the strictest available criteria to define abstinence in each study, e.g. sustained cessation rather than point prevalence. Where biochemical validation was used, only participants meeting the biochemical criteria for cessation were regarded as abstainers. Participants lost to follow-up were regarded as continuing smokers (in intention-to-treat analyses).
Search methods for identification of studies
We searched the Tobacco Addiction Review Group Specialized Register for trials (most recent search June 2013). This Register includes trials located from systematic searches of MEDLINE, EMBASE and PsycINFO and handsearching of specialist journals, conference proceedings, and reference lists of previous trials and overviews. At the time of the search the Register included the results of searches of the Cochrane Central Register of Controlled trials (CENTRAL), 2013, Issue 6 ; MEDLINE (via OVID) to update 20130607; EMBASE (via OVID) to week 201324; PsycINFO (via OVID) to update 20130610. See the Tobbaco Addiction Group module in The Cochrane Library for full search strategies and list of other resources searched. We checked all trials with 'nurse*' or 'nursing' or 'health visitor' in the title, abstract, or keywords for relevance. See Appendix 1 for the search strategy. We also searched the Cumulative Index to Nursing and Allied Health Literature (CINAHL) on OVID for 'nursing' and 'smoking cessation' from 1983 to June 2013.
Data collection and analysis
Selection of studies
One review author screened titles and abstracts. Where there was uncertainty, we requested the full text. Two review authors checked the full text of articles flagged for inclusion, with discrepancies resolved via discussion or by referral to a third review author.
Data extraction and management
The authors extracted data from the published reports independently. Disagreements were resolved by referral to a third person. For each trial, the following data were extracted: (i) author(s) and year; (ii) country of origin, study setting, and design; (iii) number and characteristics of participants and definition of 'smoker'; (iv) description of the intervention and designation of its intensity (high or low); and (v) outcomes and biochemical validation.
Assessment of risk of bias in included studies
We used the Cochrane 'Risk of bias' tool to assess bias in four domains:
- random sequence generation (a potential source of selection bias);
- allocation concealment (also a potential source of selection bias);
- incomplete outcome data (attrition bias);
- other biases.
We did not judge the trials on the basis of blinding, as we tested behavioural interventions where blinding of participants and providers is not possible.
We judged each included study to be at high, unclear, or low risk of bias in each of the above domains according to the guidelines in the Cochrane Handbook.
Measures of treatment effect
We use the risk ratio (RR) for summarizing individual trial outcomes and for the estimate of the pooled effect. Where we judged a group of studies to be sufficiently clinically and statistically homogeneous we used the Mantel-Haenszel fixed-effect method (Greenland 1985) to calculate a weighted average of the risk ratios of the individual trials, with 95% confidence intervals.
Dealing with missing data
In trials where the details of the methodology were unclear or where the results were expressed in a form that did not allow for extraction of key data, we approached the original investigators for additional information. We treated participants lost to follow-up as continuing smokers. We excluded from totals only those participants who died before follow-up or were known to have moved to an untraceable address.
Assessment of heterogeneity
To assess statistical heterogeneity between trials we used the I² statistic (Higgins 2003). This measures the percentage of total variation across studies due to heterogeneity rather than to chance. Values of I² over 75% indicate a considerable level of heterogeneity (Chapter 8, Cochrane Handbook).
Description of studies
Forty-nine trials met the inclusion criteria, of which seven were new for this update (Duffy 2006; Aveyard 2007; Jiang 2007; Wood 2008; Meysman 2010; Cossette 2011; Chan 2012). Longer-term follow-up was reported for two trials (Hilberink 2005; Hanssen 2007) and the data they contributed to the meta-analysis were updated. Trials were of nursing interventions for smoking cessation for adults who used tobacco (primarily cigarettes), published between 1987 and 2012. One trial (Sanders 1989a; Sanders 1989b) had two parts with randomization at each stage, so is treated here as two separate studies, making a total of 50 studies in the Characteristics of included studies table. Thirty-five studies contributed to the primary meta-analysis that compared a nursing intervention to a usual care or minimal intervention control. Eleven studies included a comparison between two nursing interventions, involving different components or different numbers of contacts. Six studies did not contribute to a meta-analysis and their results are described separately. Sample sizes of studies contributing to a meta-analysis ranged from 25 to 2700 but were typically between 150 and 500.
Seventeen trials took place in the USA, ten in the UK, four in Canada, and two each in Australia, China, Denmark, Japan, The Netherlands, Norway and Spain. One trial was reported from Belgium, South Korea and Sweden. One multicenter study was conducted in multiple European countries.
Twenty trials intervened with hospitalized participants (Taylor 1990; Rigotti 1994; DeBusk 1994; Allen 1996; Carlsson 1997; Miller 1997; Lewis 1998; Feeney 2001; Bolman 2002; Hajek 2002, Quist-Paulsen 2003; Froelicher 2004; Hasuo 2004; Chouinard 2005; Hennrikus 2005; Nagle 2005; Hanssen 2007; Wood 2008; Meysman 2010; Cossette 2011). One trial (Rice 1994) recruited hospitalized participants, but with follow-up after discharge. Twenty-four studies recruited from primary care or outpatient clinics (Janz 1987; Sanders 1989a/Sanders 1989b; Risser 1990; Vetter 1990; Nebot 1992; Hollis 1993; OXCHECK 1994; Family Heart 1994; Tonnesen 1996; Campbell 1998; Lancaster 1999; Steptoe 1999; Canga 2000; Aveyard 2003; Ratner 2004; Tonnesen 2006; Kim 2005; Hilberink 2005; Duffy 2006; Sanz-Pozo 2006; Aveyard 2007; Jiang 2007; Wood 2008; Chan 2012). In some trials, the recruitment took place during a clinic visit whilst in others the invitation to enrol was made by letter. One study (Terazawa 2001) recruited employees during a workplace health check, two enrolled community-based adults motivated to make a quit attempt (Davies 1992; Alterman 2001), one recruited mothers taking their child to a pediatric clinic (Curry 2003) and one recruited people being visited by a home healthcare nurse (Borrelli 2005).
Fifteen of the studies focused on adults with diagnosed cardiovascular health problems (Taylor 1990; DeBusk 1994; Family Heart 1994; Rice 1994; Rigotti 1994; Allen 1996; Carlsson 1997; Miller 1997; Campbell 1998; Feeney 2001; Bolman 2002; Hajek 2002; Jiang 2007; Cossette 2011; Chan 2012 (subgroup with cardiovascular disease)); two studies were with participants with respiratory diseases (Tonnesen 1996; Tonnesen 2006) and one with participants with diabetes (Canga 2000). One study recruited participants either with diagnosed cardiovascular health problems or judged to be at high risk of developing heart disease (Wood 2008). Two studies recruited surgical patients: Ratner 2004 recruited people attending a surgical pre-admission clinic and Meysman 2010 recruited people admitted to surgical wards. One study recruited head and neck cancer patients at four medical centres (Duffy 2006).
All studies included adults 18 years and older who used some form of tobacco. Allen 1996, Curry 2003 and Froelicher 2004 studied women only, and Terazawa 2001 men only. The definition of tobacco use varied and in some cases included recent quitters.
Seven of the studies examined a smoking cessation intervention as a component of multiple risk factor reduction interventions in adults with cardiovascular disease (DeBusk 1994; Allen 1996; Carlsson 1997; Campbell 1998; Hanssen 2007; Jiang 2007; Wood 2008). In four studies, the smoking cessation component was clearly defined, of high intensity, and independently measurable (DeBusk 1994; Allen 1996; Carlsson 1997; Jiang 2007), whereas in the remaining three the smoking component was less clearly specified (Campbell 1998; Hanssen 2007; Wood 2008).
Thirty-five studies with a total of over 17,000 participants contributed to the main comparison of nursing intervention versus control. We classified 28 as high intensity on the basis of the planned intervention, although in some cases implementation may have been incomplete. In seven, we classified the intervention as low intensity (Janz 1987; Vetter 1990; Davies 1992; Nebot 1992; Tonnesen 1996; Aveyard 2003; Nagle 2005). All of these were conducted in outpatient, primary care or community settings. One further study (Hajek 2002) may be considered as a comparison between a low intensity intervention and usual care. Participants in the usual care control group received systematic brief advice and self-help materials from the same nurses who provided the intervention. Unlike the other trials in the low intensity subgroup, this trial was conducted amongst inpatients with cardiovascular disease. Since the control group received a form of nursing intervention, we primarily classified the trial as a comparison of two intensities of nursing intervention. But since other studies had usual care groups that may have received advice from other healthcare professionals, we also report the sensitivity of the main analysis results to including it there as a low intensity nursing intervention compared to usual care control.
Hajek 2002 and ten other studies contributed to a second group comparing two interventions involving a nursing intervention. Three of these tested additional components as part of a session: demonstration of carbon monoxide (CO) levels to increase motivation to quit (Sanders 1989b); CO and spirometry feedback (Risser 1990); and CO feedback plus additional materials and an offer to find a support buddy (Hajek 2002). Five involved additional counselling sessions from a nurse (Alterman 2001; Feeney 2001; Tonnesen 2006; Aveyard 2007; Jiang 2007). One other study compared two interventions with a usual-care control (Miller 1997). The minimal intervention condition included a counselling session and one telephone call after discharge from hospital. In the intensive condition, participants received three additional telephone calls, and those who relapsed were offered further face-to-face meetings, and nicotine replacement therapy if needed. We classified both interventions as intensive in the main meta-analysis, but compared the intensive and minimal conditions in a separate analysis of the effect of additional follow-up. Chouinard 2005 also assessed the effect of additional telephone support as an adjunct to an inpatient counselling session, so is pooled in a subgroup with Miller 1997. We included in the same subgroup a study that tested additional telephone follow-up as a relapse prevention intervention for people who had inpatient counselling (Hasuo 2004).
Five studies (Family Heart 1994; OXCHECK 1994; Campbell 1998; Steptoe 1999; Wood 2008) were not included in any meta-analysis and do not have results displayed graphically because their designs did not allow appropriate outcome data to be extracted. The first part of a two-stage intervention study is also included here (Sanders 1989a); the second part (Sanders 1989b) is included in one of the meta-analyses. These six studies are discussed separately in the results.
We determined whether the nurses delivering the intervention were providing it alongside clinical duties that were not smoking-related, were working in health promotion roles, or were employed specifically as project nurses. Of the high intensity intervention studies, 12 used nurses for whom the intervention was a core component of their role (Hollis 1993; DeBusk 1994; Allen 1996; Carlsson 1997; Terazawa 2001; Quist-Paulsen 2003; Froelicher 2004; Duffy 2006; Aveyard 2007; Meysman 2010; Cossette 2011; Chan 2012). In nine studies the intervention was delivered by a nurse specifically employed by the project (Taylor 1990; Rice 1994; Rigotti 1994; Miller 1997; Lewis 1998; Canga 2000; Hennrikus 2005; Hanssen 2007; Jiang 2007). In four of these, the same nurse provided all the interventions (Rigotti 1994; Lewis 1998; Canga 2000; Jiang 2007). One study (Kim 2005) employed retired nurses who were trained to provide a brief intervention using the '5 As' framework. In only four studies were intensive interventions intended to be delivered by nurses for whom it was not a core task (Lancaster 1999; Bolman 2002; Curry 2003; Sanz-Pozo 2006). Most of the low intensity interventions were delivered by primary care or outpatient clinic nurses. One low-intensity inpatient intervention was delivered by a clinical nurse specialist (Nagle 2005).
Follow-up periods for reinforcement and outcome measurements varied across studies, with a tendency for limited reinforcement and shorter follow-up periods in the older studies. All trials had some contact with participants in the first three months of follow-up for restatement of the intervention and/or point prevalence data collection. Five of the studies had less than one year final outcome data collection (Janz 1987; Vetter 1990; Davies 1992; Lewis 1998; Canga 2000). The rest had follow-up at one year or beyond. Outcome used for the meta-analysis was the longest follow-up (six months and beyond), with the exception of Hanssen 2007 in which 12-month data were used in preference over 18-month data. The outcome in this study was point prevalence abstinence and the 18-month data were judged to be too conservative due to a rise in abstinent participants in the control group. There was no evidence from a subgroup analysis that the differences in length of follow-up explained any of the heterogeneity in study results.
A brief description of the main components of each intervention is provided in the 'Characteristics of included studies' table.
Fifty-four studies that we had identified as potentially relevant based on title and abstract were excluded upon screening the full text. These are listed in the Characteristics of excluded studies table along with the reason for exclusion for each. The most common reasons for exclusion were: study design (not a randomized clinical trial); less than six months follow-up; multicomponent studies with insufficient detail on smoking intervention/outcome; and studies in which the impact of the nursing intervention was confounded by additional pharmacological or behavioural treatment that was not provided to the control arm.
Risk of bias in included studies
As seen in Figure 1, the majority of studies were judged to be at low or unclear risk of selection bias (random sequence generation and allocation concealment) and attrition bias (loss to follow-up).
|Figure 1. Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.|
Twenty-four studies provided details of a method of random sequence generation judged to be at low risk of bias, and a further 20 studies did not report how the sequence was generated and were hence rated as unclear in this domain. Five studies were judged to be at high risk based on their reported methods of random sequence generation: Bolman 2002 was a cluster-randomized study in which some hospitals picked their allocation; in Curry 2003 participants drew a coloured ball from a bag; Davies 1992 allocated based on order of attendance; Hollis 1993 randomized participants based on health record number; and Sanders 1989a/Sanders 1989b randomized participants based on day of attendance. In addition to these five studies, three further studies in which providers rather than participants were randomized were judged to be at high risk of selection bias: Hilberink 2005 reported that self selection at practice level may have affected the results; in Janz 1987 allocation was determined by clinic session; and in Nebot 1992 the providers were also responsible for allocating participants, rendering allocation concealment impossible. A sensitivity analysis including only the results of studies judged to be at low risk of selection bias did not alter the main conclusions.
Incomplete outcome data
Twenty-seven studies that reported minimal to moderate loss to follow-up and accounted for all participants in their reporting were judged to be at low risk of attrition bias. A further 19 studies did not provide sufficient detail with which to judge likelihood of attrition bias and hence were rated as 'unclear' in this domain. Four studies were judged to be at high risk of attrition bias: in Feeney 2001, 79% of usual-care participants were not followed up; OXCHECK 1994 stated that their methods of accounting for missing participants may have overestimated the effect; in Sanders 1989a/Sanders 1989b only a subsample of participants from the control group was followed up; and in Steptoe 1999 overall drop-out rates were high and varied between intervention and control groups.
Other potential sources of bias
Definitions of abstinence ranged from single point prevalence to sustained abstinence (multiple point prevalence with self report of no slips or relapses). In one study (Miller 1997) we used validated abstinence at one year rather than continuous self-reported abstinence because only the former outcome was reported for disease diagnosis subgroups. Validation of smoking behaviour using biochemical analysis of body fluids (e.g. cotinine or thiocyanate) was reported in 15 (47%) of the 31 studies in the primary meta-analysis. Expired CO was used for validation in another six (24%) of the trials. One study tested CO levels only amongst people followed up in person (Curry 2003). Five others used some validation but did not report rates based on biochemical validation of every self-reported quitter (Nebot 1992; Rice 1994; Miller 1997; Froelicher 2004; Borrelli 2005). Six studies (23%) did not use any biochemical validation and relied on self-reported smoking cessation at a single follow-up (Janz 1987; Allen 1996; Carlsson 1997; Bolman 2002; Hilberink 2005; Hanssen 2007), though two of these warned participants that samples might be requested for testing (i.e. 'bogus pipeline'). Where both self-reported and validated quitting were reported, the level of misreporting or failure to provide a sample is typically similar across intervention and control groups. One study, however, reported differential validation failure rates so that the significant differences based on self report were not found for validated abstinence (Hennrikus 2005); this study was judged to be at high risk of other bias.
Almost all the trials used convenience rather than randomly selected samples. Only one of the studies (Vetter 1990) did not let participants know initially that they were going to be part of a smoking cessation study. In most of the research, the basis for sample size was not specified a priori, nor was a retrospective power analysis conducted. Most studies did not report 'refusal to participate' rates.
Effects of interventions
Effects of intervention versus control/usual care
Smokers offered advice by a nursing professional had an increased likelihood of quitting compared to smokers without intervention, with evidence of only moderate statistical heterogeneity between the results of the 35 studies contributing to this comparison (I² = 50%). Heterogeneity was marginally more apparent in the subgroup of 28 high-intensity trials (I² = 54%). There was one trial with a significant negative effect for treatment (Rice 1994) and two with particularly large positive effects (Canga 2000; Terazawa 2001). Pooling all 35 studies using a fixed-effect model gave a risk ratio (RR) of 1.29 with a 95% confidence interval (CI) 1.20 to 1.39 at the longest follow-up (Figure 2, Analysis 1.1). Because of the heterogeneity we tested the sensitivity to pooling the studies using a random-effects model. This did not materially alter the estimated effect size or greatly widen the confidence intervals (RR 1.32, 95% CI 1.17 to 1.49). A sensitivity analysis excluding the three outlying trials widened the CIs but did not alter the point estimate whilst greatly reducing statistical heterogeneity not attributable to chance in the high intensity subgroup (I² = 7%).
|Figure 2. Trials of nursing intervention versus control grouped by intensity of intervention. Outcome: Smoking cessation at longest follow-up.|
We also tested the sensitivity of these results to excluding studies that did not validate all reports of abstinence, limiting the analysis to studies judged to be at low risk of selection bias and excluding studies with less than 12 months follow-up. None of these altered the estimates to any great extent, although confidence intervals became wider due to the smaller number of studies. Excluding one study (Bolman 2002) for which we were not able to enter the numbers of quitters directly did not alter the results.
Some participants in Taylor 1990 had been encouraged to use nicotine replacement therapy (NRT). Exclusion of these people did not alter the significant effect of the intervention in this study. In Miller 1997 more people in the intervention conditions than the control used NRT (44% of intensive and 39% of minimal intervention versus 29% of control). People who were prescribed NRT had lower quit rates than those who were not, but the relative differences in quit rates between the usual-care and intervention groups were similar for the subgroups that did and did not use NRT. However, because of the different rates of use of NRT, it is probable that the increased use of NRT contributed to the effects of the nursing intervention. Use of NRT was also encouraged as part of the Canga 2000 intervention, with 17% of the intervention group accepting a prescription, and as part of the Duffy 2006 intervention, although at six months similar percentages in the intervention and control group had used NRT over the course of the study.
Effect of intervention intensity
We detected no evidence from our indirect comparison between subgroups that the trials we classified as using higher intensity interventions had larger treatment effects. In this update of the review the point estimate for the pooled effect of the seven lower-intensity trials is effectively the same as for the 28 of higher intensity, although for the low-intensity group the confidence interval does not exclude 1 (high-intensity subgroup RR 1.26, 95% CI 1.17 to 1.36; low-intensity subgroup RR 1.27, 95% CI 0.99 to 1.62). In a sensitivity analysis we included Hajek 2002, a study for which we were uncertain over the classification of the control group (as noted above in the Description of studies section), in the low-intensity subgroup. Including this study in the low-intensity subgroup further reduces the point estimate and there was no evidence of a treatment effect (RR 1.09, 95% CI 0.92 to 1.29). Compared to the other trials in the low-intensity subgroup, the Hajek trial was conducted amongst hospitalized participants with cardiovascular disease and the overall quit rates were high. The large number of events gave this trial a high weight in the meta-analysis.
The distinction between low- and high-intensity subgroups was based on our categorization of the intended intervention. Low levels of implementation were particularly noted in the trial reports for Lancaster 1999, Bolman 2002 and Curry 2003, so we tested the effect of moving them from the high- to the low-intensity subgroup. This reduced the point estimate of effect in the low-intensity subgroup and increased it in the high-intensity one. If these three studies and Hajek 2002 are included in the low-intensity subgroup, the pooled estimate of effect is small and non-significant (RR 1.09, 95% CI 0.96 to 1.25, Analysis 4.1). We also assessed the sensitivity of the results to using additional participants in the control group for Aveyard 2003 (see Characteristics of included studies for details). This reduced the size of the effect in the low-intensity subgroup but did not alter our conclusions.
Effects of differing health states and client settings
Trials in hospitals recruited participants with health problems, but some trials specifically recruited those with cardiovascular disease, and amongst these, some interventions addressed multiple risks whilst most only addressed smoking. Trials in primary care generally did not select participants with a particular health problem. We combined setting and disease diagnosis in one set of subgroups ( Analysis 2.1).
Five trials that included a smoking cessation intervention from a nurse as part of cardiac rehabilitation showed a significant pooled effect on smoking (RR 1.35, 95% CI 1.14 to 1.59). Four of these (Allen 1996; Carlsson 1997; Hanssen 2007; Jiang 2007) did not use biochemical validation of quitting, and in the fifth (DeBusk 1994) we were unable to confirm the proportion of drop-outs with the study authors.
There was moderate heterogeneity (I² = 50%) amongst seven trials in hospitalized smokers with cardiovascular disease, due to the strong intervention effect in one of the seven trials (Taylor 1990). The estimated RR was 1.29 (95% CI 1.14 to 1.45) and the effect remained significant if Taylor 1990 was excluded or if a random-effects model was used. A sensitivity analysis of the effect of including Hajek 2002 in this category increased the heterogeneity (I² = 60%), and the pooled effect was just significant whether a fixed-effect or a random-effects model was used ( Analysis 5.1). Excluding Taylor 1990 again removed heterogeneity but the pooled effect was then small and not significant (RR 1.1, 95% CI 0.99 to 1.26, analysis not shown).
Amongst the six trials in non-cardiac hospitalized smokers the risk ratio was small and the confidence interval did not exclude no effect (RR 1.09, 95% CI 0.94 to 1.28). We included in this subgroup one trial that began the intervention in a pre-admission clinic for elective surgery patients (Ratner 2004).
Heterogeneity was high (I² = 94%) between two trials of interventions delivered to non-hospitalized adults with cardiovascular disease,(Rice 1994; Chan 2012;). Subgroup analysis in Rice 1994, however, suggested that smokers who had experienced cardiovascular bypass surgery were more likely to quit, and these participants were over-represented in the control group who received advice to quit but no structured intervention.
Pooling 15 trials of cessation interventions for other non-hospitalized adults showed an increase in the success rates (RR 1.81, 95% CI 1.48 to 2.22). A sensitivity analysis testing the effect of excluding those trials (Janz 1987; Vetter 1990; Curry 2003; Hilberink 2005) where a combination of a nursing intervention and advice from a physician was used did not substantially alter this.
Higher versus lower intensity interventions
Effects of physiological feedback
Effects of other components at a single contact
One trial in hospitalized smokers with cardiovascular disease (Hajek 2002) failed to detect a significant benefit of additional support from a nurse giving additional written materials, a written quiz, an offer of a support buddy, and carbon monoxide measurement compared to controls receiving brief advice and a self-help booklet (RR 0.91, 95% CI 0.73 to 1.13, Analysis 3.1.2).
Effects of additional telephone support
There was weak evidence from pooling three trials that additional telephone support increased cessation, as the lower limit of the confidence interval was at the boundary of no effect (RR 1.25, 95% CI 1.00 to 1.56; Analysis 3.2.1).
Effects of additional face-to-face sessions
One trial of additional support from an alcohol and drug assessment unit nurse for people admitted to a coronary care unit (Feeney 2001) showed a very significant benefit for the intervention. The cessation rate among the controls, however, was very low (1/97), and there were a large number of drop-outs, particularly from the control group. This could have underestimated the control group quit rate. In another trial (Alterman 2001), offering four nurse sessions rather than one as an adjunct to nicotine patch showed no benefit, with the control group having a significantly higher quit rate (RR 0.43, 95% CI 0.21 to 0.89, Analysis 3.2.3). No explanation was offered for the lower than expected quit rates in the intervention group.
Effects of additional face-to-face sessions and telephone support
Pooled results from three trials did not show an effect of providing additional clinic sessions and telephone support to participants (RR 0.92, 95% CI 0.65 to 1.31, Analysis 3.2.4).
Results of studies not included in the meta-analysis
We identified six studies (Sanders 1989a; Family Heart 1994; OXCHECK 1994; Campbell 1998; Steptoe 1999; Wood 2008) in which nurses intervened with primary care patients. All except Sanders 1989a addressed multiple cardiovascular risk factors, and all except Campbell 1998 and Wood 2008 targeted healthy people. Campbell 1998 recruited participants with coronary heart disease. Wood 2008 recruited general practice patients deemed to be at high risk of cardiovascular disease, and also recruited hospitalized participants with established coronary heart disease. Although they met the main inclusion criteria, in five of the trials the design did not allow for data extraction for meta-analysis in a comparable format to other studies. In the other (Sanders 1989a) only a random sample of the control group was followed up. We therefore discuss these trials separately.
Sanders 1989a, in which smokers visiting their family doctor were asked to make an appointment for cardiovascular health screening, reported that only 25.9% of the patients made and kept such an appointment. The percentage that had quit at one month and at one year and reported last smoking before the one-month follow-up was higher both in the attenders (4.7%) and the non-attenders (3.3%) than in the usual-care controls (0.9%). This suggests that the invitation to make an appointment for health screening could have been an anti-smoking intervention in itself, and that the additional effect of the structured nursing intervention was small.
We do not have comparable data for OXCHECK 1994, which used similar health checks, because the households had been randomized to be offered the health check in different years. The authors compared the proportions of smokers in the intervention group who claimed to have stopped smoking in the previous year to patients attending for their one-year follow-up, and to controls attending for their first health check. They found no difference in the proportions that reported stopping smoking in the previous year.
The Family Heart 1994 study offered nurse-led cardiovascular screening for men aged 40 to 59 and for their partners, with smoking cessation as one of the recommended lifestyle changes. Cigarette smokers were invited to attend up to three further visits. Smoking prevalence was lower amongst those who returned for the one-year follow up than amongst the control group screened at one year. This difference was reduced if non-returners were assumed to have continued to smoke, and if CO-validated quitting was used. In that case there was a reduction of only about one percentage point, with weak evidence of a true reduction.
Campbell 1998 invited people with a diagnosis of coronary heart disease to nurse-run clinics promoting medical and lifestyle aspects of secondary prevention. There was no significant effect on smoking cessation. At one year the decline in smoking prevalence was greater in the control group than in the intervention group. Four-year follow-up did not alter the effect of a lack of benefit.
Steptoe 1999 recruited people at increased risk of coronary heart disease for a multi-component intervention. The quit rate amongst smokers followed up after one year was not significantly higher in the intervention group (9.4%, 95% CI -9.6 to 28.3), and there was greater loss to follow-up of smokers in the intervention group.
Wood 2008 recruited people with established or increased risk of coronary heart disease for a multicomponent lifestyle intervention, coordinated by nurses. The authors report results separately for those participants recruited in hospital and those recruited in general practice. For coronary patients recruited in hospital who had smoked within one month at baseline, abstinence at one year favoured the intervention group (58% versus 47%), but the difference was not significant (P = 0.06). For participants at high risk of coronary heart disease recruited in general practice the prevalence of smoking fell from baseline but did not differ between conditions.
Summary of main results
The results of this meta-analysis support a modest but positive effect for smoking cessation intervention by nurses, but with caution about the effects that can be expected if interventions are very brief or cannot be consistently delivered (see Summary of findings for the main comparison). A structured smoking cessation intervention delivered by a nurse was more effective than usual care on smoking abstinence at six months or longer from the start of treatment. The direction of effect was consistent in different intensities of intervention, in different settings, and in smokers with and without tobacco-related illnesses. In a subgroup of low-intensity studies the confidence interval did not exclude no effect, but the point estimate was effectively the same as that in the larger group of high-intensity studies. There was also some evidence of statistical heterogeneity, although this was attributable to a very small number of outlying studies. In the one study (Rice 1994) that showed a statistically significantly higher quit rate in a control group, participants had been advised to quit and the control group included a significantly larger proportion of people who had had coronary artery bypass graft surgery. A multivariate analysis of one-year follow-up data in this study revealed that a quitter was significantly more likely to be less than 48 years, male, to have had individualized versus group or no cessation instruction, and to have had a high degree of perceived threat relative to their health state.
Overall completeness and applicability of evidence
Overall, these meta-analysis findings need to be interpreted carefully in light of the methodological limitations of both the review and the clinical trials. In terms of the review, it is possible that there was a publication selection bias due to using only tabulated data derived from published works (Stewart 1993). Data from the unpublished and/or missed studies could have shown more or less favourable results, though a funnel plot for the main comparison did not suggest the presence of reporting bias. Secondly, the results of a meta-analysis (based on the findings of many small trials) should be viewed with caution even when the combined effect is statistically significant (LeLorier 1997). In this analysis one study (Miller 1997) contributes 21% of the weight to the primary analysis, while the next largest contributes 11% of the weight. Finding statistical heterogeneity between the relative incidence of cessation in different studies limits any assumption that interventions in any clinical setting and with any type of participant are equally effective.
A difference among the studies that may have contributed to the differences in outcome was baseline cigarette use. There was an inverse relationship between number of cigarettes smoked per day and success in quitting; the more addicted the individuals, the more difficult it was for them to quit. Studies that recruited a higher proportion of lighter smokers or that included recent quitters could have achieved better results. Interestingly, the studies in the meta-analysis that reported the highest cigarette use rates had the weakest effect for the intervention (Davies 1992; Rice 1994). Although some trials included recent quitters in their recruitment, there was no evidence that these trials had different results.
The findings of this review, and in particular the estimated size of the treatment effect, have remained remarkably stable since its initial publication. In 1999. fifteen studies contributed to the main analysis, with a pooled risk ratio of 1.30 (95% CI 1.16 to 1.44). Further studies have more than doubled the number of participants and thus narrowed the CIs but have had little impact on the point estimate
Effectiveness by intervention characteristics and population
The effect estimates are similar for high- and low-intensity smoking cessation interventions by nurses, as was found in a review of physicians' advice (Stead 2013). Presumably, the more components added to the intervention the more intensive the intervention; however, assessing the contribution of factors such as total contact time, number of contacts, and content of the intervention was difficult. Our distinction between high and low intensity based on the length of initial contact and number of planned follow-ups may not have accurately distinguished among the key elements that could have contributed to greater efficacy. We found that the nature of the smoking cessation interventions differed from advice alone, to more intensive interventions with multiple components, and that the description of what constituted 'advice only' varied. In most trials, advice was given with an emphasis on 'stopping smoking' because of some existing health problem. To make most interventions more intensive, verbal advice was supplemented with a variety of counselling messages, including benefits and barriers to cessation (e.g. Taylor 1990) and effective coping strategies (e.g. Allen 1996). Manuals and printed self-help materials were also added to many interventions along with repeated follow-up (Hollis 1993; Miller 1997). In some studies, the proposed intervention was not delivered consistently to all participants. In recent updates the evidence for the benefit of a low-intensity intervention has become weaker than that for a more intensive intervention, and the estimated effect is sensitive to the inclusion of one additional study (Hajek 2002) and to the classification of intensity of three studies. Almost all the intensive interventions were delivered by either dedicated project staff or nurses with a health promotion role. Most studies in which the intensive intervention was intended to be delivered by a nurse with other roles reported problems in delivering the intervention consistently. None showed a statistically significant benefit for the intervention. We found no studies of brief opportunistic advice that were directly analogous to the low-intensity interventions used in physician advice trials (Stead 2013).
In two studies in the low-intensity category (Janz 1987; Vetter 1990), advice from a physician was also part of the intervention and this almost certainly contributed to the overall effect. The most highly weighted study in the high-intensity subgroup (Miller 1997) produced only relatively modest results. This was due in part to the effect of the minimal treatment condition that had just one follow-up telephone call. However, using just the high-intensity condition in the analysis did not materially alter the pooled estimate.
One study (Miller 1997) provided data on the effect of the same intervention in smokers with different types of illness and showed a greater effect in cardiovascular patients. In these individuals the intervention increased the 12-month quit rate from 24% to 31%, which just reached statistical significance. In other types of patients, the rates were increased from 18.5% to 21%, an effect that did not reach statistical significance. In this study participants were eligible if they had smoked any tobacco in the month prior to hospitalization, but were excluded if they had no intention of quitting (although they were also excluded if they wanted to quit on their own). These criteria may have contributed to the relatively high quit rates achieved. Also, a higher proportion of participants in the intensive treatment arm than in the minimal or usual care intervention arms were prescribed nicotine replacement therapy (NRT). However, the intervention was also effective in those not prescribed NRT. Those given NRT were heavier smokers (with higher levels of addiction) who achieved lower cessation rates than those who did not use NRT.
This suggests that nursing professionals may have an important 'window of opportunity' to intervene with patients in the hospital setting, or at least to introduce the notion of not resuming tobacco use on hospital discharge. The size of the effect may be dependent on the reason for hospitalization. The additional telephone support, with the possibility of another counselling session for people who relapsed after discharge, seemed to contribute to more favourable outcomes in the intensive intervention used by Miller 1997, although pooled results from three studies testing the addition of telephone counselling and further face-to-face contact did not detect an effect. A separate Cochrane review of the efficacy of interventions for hospitalized patients has been recently updated (Rigotti 2012), and this supports the efficacy of interventions for this patient group, but only when the interventions included post-discharge support for at least one month.
Providing additional physiological feedback in the form of spirometry and demonstrated carbon monoxide level as an adjunct to nursing intervention did not appear to have an effect. Three studies in primary care or outpatient settings used this approach (Sanders 1989b; Risser 1990; Hollis 1993). It was also used as part of the enhanced intervention in a study with hospitalized patients (Hajek 2002).
The identification of an effect for a nurse-mediated intervention in smokers who were not hospitalized is based on 15 studies. The largest study (Hollis 1993) increased the quit rate from 2% in those who received only advice from a physician to 4% when a nurse delivered one of three additional interventions, including a video, written materials, and a follow-up telephone call. Control group quit rates were less than 10% in almost all these studies, and more typically between 4% and 8%. The risk ratio in this group of studies (1.8) was a little higher than in some subgroups, but because of the low background quit rate the proportion of participants likely to become long-term quitters as a result of a nursing intervention in these settings is likely to be small. However, because of the large number of people who could be reached by nursing, the effect would be important.
The evidence is not strong for an effect of nurse counselling about smoking cessation when it is provided as part of a health check. It may be unrealistic to expect a benefit from this type of intervention. Two studies that invited smokers to make an appointment with a nurse for counselling (Lancaster 1999; Aveyard 2003) also had relatively poor results. In both cases the uptake of the intervention was reported to be poor, with participants reluctant to schedule visits.
Combined efforts of many types of healthcare professionals are likely to be required. The US Public Health Service clinical practice guideline 'Treating Tobacco Use and Dependence' (AHRQ 2008) used logistic regression to estimate efficacy for interventions delivered by different types of providers. Their analysis did not distinguish among the non-physician medical healthcare providers, so that dentists, health counsellors, and pharmacists were included with nurses. The guideline concluded that these providers were effective (Table 15, OR 1.7, 95% CI 1.3 to 2.1). They also concluded that interventions by multiple clinician types were more effective (Table 16, OR 2.5, 95% CI 1.9 to 23.4). Although it was recognized that there could be confounding between the number of providers and the overall intensity of the intervention, the findings confirmed that a nursing intervention that reinforces or complements advice from physicians and/or other healthcare providers is likely to be an important component in helping smokers to quit.
Implications for practice
The results of this review indicate the potential benefits of interventions given by nurses to their patients. The challenge remains to incorporate smoking cessation interventions as part of standard practice so that all patients are given an opportunity to be asked about their tobacco use and to be given advice to quit along with reinforcement and follow-up. Nicotine replacement therapy has been shown to improve quit rates when used in conjunction with counselling for behavioural change and should be considered an important adjunct, but not a replacement for nursing interventions (Stead 2012). The evidence suggests that brief interventions from nurses who combine smoking cessation work with other duties are less effective than longer interventions with multiple contacts, delivered by nurses with a role in health promotion or cardiac rehabilitation.
Implications for research
Further studies of nursing interventions are warranted, with more careful consideration of sample size, participant selection, refusals, drop-outs, long-term follow-up, and biochemical verification. Additionally, controlled studies are needed that carefully examine the effects of 'brief advice by nursing' as this type of professional counselling may more accurately reflect the current standard of care. Work is now required to systematize interventions so that more rigorous comparisons can be made between studies. None of the trials reviewed was a replication study; this is a very important method to strengthen the science, and should be encouraged.
Nicky Cullum and Tim Coleman for their helpful peer review comments on the original version of this review. Hitomi Kobayasha, a doctoral student, for assistance with Japanese translation of a study.
Data and analyses
- Top of page
- Summary of findings [Explanations]
- Authors' conclusions
- Data and analyses
- What's new
- Contributions of authors
- Declarations of interest
- Sources of support
- Index terms
Appendix 1. Register search strategy
Run using Cochrane Register of Studies (CRS) software
#1 (nurse* or nursing):TI,AB,XKY,MH,EMT,KY
#2 (health visitor*):TI,AB,XKY,MH,EMT,KY
#3 #1 OR #2
XKY, MH, EMT, KY are keyword fields. XKY field includes indexing terms added for the use of the tobacco addiction group.
Appendix 2. Glossary of terms
Last assessed as up-to-date: 27 June 2013.
Protocol first published: Issue 3, 1998
Review first published: Issue 3, 1999
Contributions of authors
VHR extracted data and wrote the review. LS conducted searches, extracted data and assisted in drafting the review. Both authors contribute to review updates. JHB contributed to the 2013 update, extracting data and assisting in updating the text.
Declarations of interest
V.H. Rice was the principal investigator in one of the studies included in this review.
Sources of support
- Wayne State University College of Nursing, Adult Health & Administration, USA.
- Department of Primary Health Care, Oxford University, UK.
- American Heart Association, USA.
- NHS Research & Development Programme, UK.
Medical Subject Headings (MeSH)
MeSH check words
Adult; Female; Humans; Male
* Indicates the major publication for the study