Description of studies

The combination of the electronic database searches and additional citations found by scanning references in relevant Cochrane Reviews, other meta‐analyses, and journals in 2006, 2009 and 2013 resulted in 2761 records (Appendix 4; see flowchart in Figure 1). After exclusions on the basis of title and abstract, 364 full‐text articles were assessed for inclusion. Of these a further 321 full‐text articles were excluded for reasons detailed in the characteristics of excluded studies table (Characteristics of excluded studies) and under http://www.juergen‐barth.de/en/wp‐content/uploads/2013/09/Reasons%20for%20excluded%20studies%20Initial%20Review%202006.pdf . Four papers are awaiting assessment as we could not access the papers and authors did not respond to PDF requests (Becker 2003; Boulay 2001; Enriquez‐Puga 2001; Puente‐Silva 1989).

Figure 1

---

---

Flow chart of study selection

Forty trials (with short‐term and long‐term results) were included in the review (Allen 1996; Benner 2008; Blasco 2012; Bolman 2002a; Burt 1974; Carlsson 1997; CASIS 1992; Chan 2012; Cossette 2011; Costa e Silva 2008; DeBusk 1994; Dornelas 2000; Froelicher 2004; Gao 2011; Hajek 2002; Han 2011; Hanssen 2007; Hanssen 2009; Heller 1993; Holmes‐Rovner 2008; Jiang 2007; Kubilius 2012; Mildestvedt 2007; Mosca 2010; Naser 2008; Neubeck 2011; Ortigosa 2000; Otterstad 2003; Pedersen 2005; Quist‐Paulsen 2003; Quist‐Paulsen 2005; Quist‐Paulsen 2006a; Reid 2003; Rigotti 1994; Sivarajan 1983; Smith 2009; Taylor 1990; van Elderen (group); van Elderen (phone); Zwisler 2008). Of those 40 trials 33 reported on short‐term results, 4 trials reported both short‐ and long‐term results and 3 trials reported on long‐term results only. In summary seven trials provided data on a long‐term follow up after 12 months (CASIS 1992; Froelicher 2004; Hanssen 2009; Mildestvedt 2007; Naser 2008; Otterstad 2003; Rigotti 1994). As a result of this update we identified 21 new trials, which contributed to this review. Five new trials contributed to long‐term results.

From the studies with short‐term findings (N = 37), sixteen studies were carried out in Europe (1 Sweden, 2 United Kingdom, 3 Netherlands, 5 Norway, 2 Spain, 1 Lithuania, 2 Denmark), ten were from the USA, two from Australia, three from Canada, four from China, one from Brazil and one was a multinational study. The papers were mainly published in English, one was written in Spanish (Ortigosa 2000), one in French (Cossette 2011) and one in Danish (Pedersen 2005). All trials compared a specific smoking cessation intervention with a usual care condition, had comparable groups at study entry, had lower than 50% drop out rates and assessed smoking status before a cardiac event or procedure. 3830 patients were randomised to the usual care group and 3852 received a special psychosocial intervention. As expected, 70 to 90% of the patients were male, mean age was relatively young, 50 to 60 years.The patients suffered predominantly from myocardial infarction or had invasive interventions (bypass surgery, stent). The intervention strategies employed were behavioural therapeutic interventions (20 studies), and self‐help programmes (18 studies). Additional phone support was provided in 26 trials. Seventeen studies reported interventions aimed specifically at smoking cessation, 20 studies employed multi‐risk strategies. Behavioural therapeutic interventions were either provided in a group setting or as individual counselling. The aim was to identify cues related to smoking, or more generally stress reduction and relaxation techniques. Other components included preparation for relapse or specific motivational techniques based on the transtheoretical model (Prochaska 1986) or the strategy of motivational interviewing (Rollnick 1997). Self‐help interventions consisted of information booklets, audio‐ or videotapes. Information booklets which simply described risk factors were not considered self‐help interventions. No studies were available for the comparison of different psychosocial interventions or of psychosocial intervention with different intensity.

In an initial pooling of the short‐term results from 40 included trials we found a RR of 1.24 (95% CI 1.14 to 1.35, n = 7928, N = 40) indicating the efficacy of psychosocial smoking cessation interventions in CHD patients (Analysis 1.1). However, a funnel plot showed outliers with very large effect sizes (Figure 2). These outliers contributed to a large amount of heterogeneity, since we found in the initial analysis an I² of 61%. Therefore a further three studies were excluded using the post hoc exclusion criteria of an unrealistic treatment effect of an RR larger than 5. Feeney 2001 reported a RR of 32.94. Lisspers 1999 reported a RR of 8.25. Mitsibounas 1992 found an effect of RR = 6.00. These trials all had also very substantial methodological flaws. Excluding these three studies had only a small effect on the overall effectiveness of the intervention (RR 1.22, 95% CI 1.13 to 1.32, n = 7682, N = 37) but achieved a more balanced funnel plot (Figure 3).

Figure 2

---

---

Funnel plot of comparison: 1 Initial analysis: Efficacy of psychosocial interventions on abstinence (6 to 12 months; all trials), outcome: 1.1 Abstinence 6 to 12 months (ITT preferred over completer).

Figure 3

---

---

Funnel plot of comparison: 2 Efficacy of psychosocial interventions on abstinence (6 to 12 months; all trials), outcome: 2.1 Abstinence 6 to 12 months (ITT preferred over completer).

Risk of bias in included studies

Out of 37 studies, 18 reported an adequate method for random sequence generation (Allen 1996; Benner 2008; Blasco 2012; Chan 2012; Cossette 2011; Costa e Silva 2008; DeBusk 1994; Dornelas 2000; Froelicher 2004; Hajek 2002; Hanssen 2007; Jiang 2007; Mosca 2010; Neubeck 2011; Reid 2003; Smith 2009; Taylor 1990; Zwisler 2008). Seventeen studies had unclear information concerning sequence generation (Bolman 2002a; Carlsson 1997; CASIS 1992; Gao 2011; Han 2011; Holmes‐Rovner 2008; Kubilius 2012; Ortigosa 2000; Otterstad 2003; Pedersen 2005; Quist‐Paulsen 2003; Quist‐Paulsen 2005; Quist‐Paulsen 2006a; Rigotti 1994; Sivarajan 1983; van Elderen (group); van Elderen (phone)) and only Burt 1974 and Heller 1993 reported an inadequate method. Only 13 studies reported adequate allocation concealment (Benner 2008; Chan 2012; Cossette 2011; DeBusk 1994; Froelicher 2004; Hajek 2002; Hanssen 2007; Otterstad 2003; Pedersen 2005; Quist‐Paulsen 2005; Quist‐Paulsen 2006a; Reid 2003; Zwisler 2008). Sixteen studies reported unclear information (Allen 1996; Blasco 2012; Bolman 2002a; Carlsson 1997; CASIS 1992; Gao 2011; Han 2011; Jiang 2007; Kubilius 2012; Mosca 2010; Ortigosa 2000; Otterstad 2003; Rigotti 1994; Sivarajan 1983; van Elderen (group); van Elderen (phone)) and eight studies were inadequate concerning allocation concealment (Burt 1974; Costa e Silva 2008; Dornelas 2000; Heller 1993; Holmes‐Rovner 2008; Neubeck 2011; Smith 2009; Taylor 1990). Blinding was inadequate in all studies, except in Bolman 2002a where it was adequate since the personnel were not blinded but this was unlikely to introduce bias because the whole hospital was randomised with significant difference between the two groups at the process evaluation. Concerning incomplete outcome data, 26 studies reported adequate information (Blasco 2012; Bolman 2002a; CASIS 1992; Chan 2012; Cossette 2011; Costa e Silva 2008; DeBusk 1994; Dornelas 2000; Froelicher 2004; Hajek 2002; Hanssen 2007; Holmes‐Rovner 2008; Jiang 2007; Mosca 2010; Ortigosa 2000; Otterstad 2003; Quist‐Paulsen 2003; Quist‐Paulsen 2005; Quist‐Paulsen 2006a; Reid 2003; Rigotti 1994; Sivarajan 1983; Smith 2009; Taylor 1990; van Elderen (group); Zwisler 2008). Eleven studies were either unclear or reported inadequate information about incomplete outcome data (Allen 1996; Benner 2008; Burt 1974; Carlsson 1997; Gao 2011; Han 2011; Heller 1993; Kubilius 2012; Neubeck 2011; Pedersen 2005; van Elderen (phone)). In 24 studies, smoking cessation was self‐reported and 13 studies validated self‐reported smoking cessation with measurements of cotinine in urine or saliva. Summary of risk of bias are found in Figure 4; Figure 5.

Figure 4

---

---

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Figure 5

---

---

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Effects of interventions

Psychosocial smoking cessation interventions were effective in achieving smoking abstinence in CHD patients (number of patients = 3852), compared with usual care (number of patients = 3830) (see Table 1). In all trials, patients receiving the specific psychosocial intervention had more than a 20% higher chance of quitting (RR 1.22, 95% CI 1.13 to 1.32, n = 7682, N = 37). There was moderate heterogeneity between studies (chi² 77.98; df 36, P < 0.0001, I² 54%) (Analysis 2.1). Therefore the overall result should be interpreted with some caution. There were considerable differences between trials in the proportion of abstinent patients after the intervention: Kubilius 2012 achieved 100% abstinence in the intervention group, but Chan 2012 reports the lowest abstinence rate with 26.5%. These differences between trials in quit rates might also be responsible for some of the heterogeneity in findings. The pooled RR suggests that psychosocial interventions can increase the chance of quitting compared with usual care, but the heterogeneity in the results need further exploration. The quality of the trials may partly explain this heterogeneity and therefore a subgroup analysis according to risk of bias indicators was used.

Sub‐group analyses according to risk of bias indicators:

None of the quality indicators affected the effect estimates (all p values > 0.20). Trials with adequate sequence generation showed similar effects compared to trials with inadequate or unclear sequence generation. Trials with adequate sequence generation had a pooled RR of 1.21 (95% CI 1.07 to 1.36, n = 5046, N = 18); trials with inadequate or unclear sequence generation had a pooled RR of 1.24 (95% CI 1.12 to 1.37, n = 2636, N = 19) (Analysis 3.1). No reduction in heterogeneity was achieved by this stratification.

We have a similar picture when we compared trials with adequate allocation concealment versus inadequate or unclear allocation concealment. The RR was 1.24 (95% CI 1.11 to 1.38, n = 2898, N = 24) for trials with inadequate or unclear allocation concealment, and 1.21 (95% CI 1.09 to 1.34, n = 4784, N = 13) for trials with adequate allocation concealment (Analysis 3.2). Heterogeneity in the subgroup of studies with adequate allocation concealment was considerably reduced (I² 39% compared to the initial analysis with an I² of 54%)

Trials with an adequate method to analyse incomplete outcome data (ITT) appeared less efficacious compared to trials with inadequate or unclear methods to analyse incomplete outcome data (completer analysis), but this difference was not statistically significant. Trials with completer analysis had a larger effect with a RR of 1.36 (95% CI 1.12 to 1.65, n = 1246, N = 11) (Analysis 3.3). Trials with ITT analysis found a reduced effect of 1.18 (95% CI 1.09 to 1.28, n = 6436, N = 26). Heterogeneity in the subgroup of studies with ITT analysis was slightly reduced (I² 46% compared to the initial analysis with an I² of 54%).

Trials which validated self‐reported smoking status showed similar efficacy to trials with non validated outcome assessment. Trials with validated abstinence reported a RR of 1.22 (95% CI 1.07 to 1.39, n = 2803, N = 13); non‐validated trials had a similar effect with a RR of 1.23 (95% CI 1.12 to 1.35, n = 4879, N = 24) (Analysis 3.4). Heterogeneity was not reduced.

Types of intervention
We found no clear evidence that any treatment strategy was more efficacious than others, but heterogeneity was reduced slightly within the intervention cluster. Behavioral therapeutic interventions showed a significant effect on abstinence (RR 1.23, 95% CI 1.12 to 1.34, n = 5170, N = 20) with lower heterogeneity (I² 40%) (Analysis 4.1). Telephone support was also effective (RR 1.21, 95% CI 1.12 to 1.30, n = 5807, N = 26) and trials more consistent (I² 44%) (Analysis 4.2). However, as most behavioural therapy trials also used telephone support as an intervention strategy, it is difficult to separate the effects of these two types of interventions. Five trials used solely a behavioural therapeutic approach without additional phone contacts (Bolman 2002a; Otterstad 2003; Sivarajan 1983; van Elderen (group); Zwisler 2008). Nine trials used telephone support without behavioural therapeutic techniques (Benner 2008; Blasco 2012; Gao 2011; Han 2011; Hanssen 2007; Jiang 2007; Neubeck 2011; Ortigosa 2000; Quist‐Paulsen 2005). Interventions using self‐help materials showed comparable effectiveness (RR 1.22, 95% CI 1.12 to 1.33m n = 3789, N = 18) (Analysis 4.3). Stratification of trials using self‐help materials reduced heterogeneity (I² 40%). We also considered the specificity of the intervention (smoking cessation alone compared with a multi‐risk factor intervention). No meaningful difference was found between multi‐risk factor interventions (RR 1.19, 95% CI 1.08 to 1.32, n = 2337, N = 20) and specific cessation intervention (RR 1.26, 95% CI 1.11 to 1.42, n = 5345, N = 17) (Analysis 4.4). Specific intervention showed highly heterogenous effects (I² 61%); likewise multi‐risk factor intervention effects differed substantially between trials (I² 52%).

Duration of the intervention
We found clear evidence that brief interventions (i.e. no follow‐up contact or within 4 weeks after initial intervention) were not effective (Chan 2012; Hajek 2002; Heller 1993; Ortigosa 2000; Rigotti 1994) (RR 1.01, 95% CI 0.91 to 1.12, I² 0%, n = 2693, N = 5) (Analysis 5.1). When CHD patients were treated with interventions including follow‐up contacts after the initial period of 1 month, the chance of quitting increased substantially (RR 1.28, 95% CI 1.17 to 1.40, I² 58%, n = 4968, N = 31) (Analysis 5.1). One study did not report on the duration of the intervention (Gao 2011).

Long‐term follow‐up
We found preliminary evidence from seven trials for the efficacy of smoking cessation interventions in the long‐term (number of patients in the intervention group = 382; number of patients with usual care = 359) (see Table 2). Due to high drop‐out rates after 12 months, separate meta‐analyses for completer and ITT effects were done. In the completer analysis smoking cessation interventions were effective (RR 1.16, 95% CI 1.02 to 1.31, n = 741, N = 7) however, this initial finding was not confirmed in the ITT analysis (RR 1.08, 95% CI 0.92 to 1.28, n = 854, N = 5) (Analysis 6.1 and Analysis 6.2, respectively).

Out of seven studies, three reported an adequate method for random sequence generation (Froelicher 2004; Hanssen 2009; Naser 2008) and four had unclear information (CASIS 1992; Mildestvedt 2007; Otterstad 2003; Rigotti 1994). Only three studies reported adequate allocation concealment (Froelicher 2004; Hanssen 2009; Otterstad 2003) and four contained unclear information (CASIS 1992; Mildestvedt 2007; Naser 2008; Rigotti 1994). Blinding was coded as inadequate in all studies, however, blinding of treatment providers is not possible in psychosocial interventions. Concerning incomplete outcome data, five studies reported adequate information (CASIS 1992; Froelicher 2004; Hanssen 2009; Mildestvedt 2007; Naser 2008) and only two were inadequate (Otterstad 2003; Rigotti 1994).

A potential problem with all systematic reviews is publication bias. Our literature search was comprehensive, prepared and partly carried out by the Cochrane Heart Group (UK). Additionally, we investigated publication bias using a funnel plot. The results appear reasonably symmetric which is an indicator of a publication of the studies independent of the study result (Figure 5). There may be a slight tendency for larger trials to show smaller benefits; but larger studies may have interventions with shorter duration and hence smaller effect sizes.