Results

A total of 10 studies met the criteria for inclusion.12 34-42 The majority were clinic based smoking trials using a group format.12 34-40 One of the remaining two studies was based in a general practice clinic42 and the other was a community based minimal intervention using pamphlets including social support instruction.41

The “buddy” support interventions were of two broad types. The majority were directed support interventions,25 using populations who identified buddy support before randomisation.12 34-41 These studies therefore made use of pre-existing support structures. They attempted to improve the quality of support with training, drawing on previous research to indicate what is beneficial. Only one study fell into the second category of the initiation of new ties.42 All the studies involved some level of guidance to buddies and/or smokers regarding how to be supportive. To what extent “support training” was provided varied from intensive group treatments with role playing and rehearsal37 to a simple instruction to call each other regularly.42

Two studies showed a significant effect of the intervention on smoking cessation.37 42 Gruder and colleagues37offered group meetings as an adjunct to a televised self help programme. They randomised “buddies” to either discussion only or support training groups. They found a significant benefit of support training at all time points up to 24 months. West and colleagues42 randomised smokers to individual treatment or buddy support in a smokers clinic in primary care. They found a significant benefit of the buddy intervention maintained to the final follow up one month after quitting (table 1).

Critical evaluation of papers

Many of the studies reviewed had design flaws that would be likely to affect the results.

MEASUREMENT OF OUTCOME

Seven of the studies reviewed had sample sizes below 100 people at baseline. This makes comparison groups very small and reduces the likelihood of detecting any effect.12 34-36 38-40Interestingly two of the three studies with samples above 100 people reported a significant improvement in abstinence rates in the intervention group.37 42

Some studies used not only complete abstinence but also smoking reduction as an outcome measure.12 34 36 38 This is of little clinical value as evidence suggests that smokers quickly revert back to previous levels of smoking and that they compensate for reductions in numbers and reduced tar content of cigarettes by changing their smoking style.43 The latter point also means that accurate measurement of reductions in tar and nicotine concentrations is very difficult. Counting or taking self reports of the number of cigarettes smoked, the brands smoked or the weight of cigarette butt left does not accurately reflect an individuals tar and nicotine concentration. Also all but one42 of the studies reviewed used point prevalence rate of smoking abstinence as their outcome measure. Typically abstinent was defined as no smoking in the previous week for the later follow ups. This means that participants may relapse following the intervention but stop again using an entirely different method at a later date and be counted as a treatment success. As a result some studies reported more people successful at one year than immediately post treatment.35 39 Finally, self reports of smoking can be validated by carbon monoxide measurement and thiocyanate or cotinine analysis. However most studies did not adequately validate self reports12 37 39 40 despite the observation that up to 50% of participants may misreport smoking status.44

Success rates in smoking interventions tend to be low and people who fail often feel disappointed and do not attend for follow up appointments. It is important, therefore, that results are analysed on an intention to treat basis with people lost to follow up counted as failures; this was not done in two cases.34 37 Only one of the other studies42 reported their treatment of such cases.

MEASUREMENT OF SUPPORT

Assessments of smoking related support were typically made using the PIQ.17 The PIQ was originally designed to assess partner support but has since been extended to other contexts. It is a list of positive and negative behaviours that a smoker may experience when trying to stop. Examples of positive behaviour would be “compliment you on not smoking” or “celebrate your quitting with you”; examples of negative behaviours would be “criticise your smoking” and “refuse to clean up your cigarette butts”. Participants are asked to rate frequency and helpfulness of the behaviour and experienced helpfulness scores are calculated. Studies generally found positive correlations between support and abstinence in their populations irrespective of intervention group.12 39 41 One found the ratio of positive to negative scores predicted abstinence37 and another that higher overall support (measured with a single item) was predictive, as was lower negative support measured using PIQ.41

One problem with this scale is that it may confuse the existence of a smoker in an individual's network with how supportive they are—for example, “during the past week my partner smoked in my presence”. In addition some of the negative “support” behaviours are only applicable where a person is still smoking and some positive ones are only applicable when the smoker is abstaining (see above). So it would not be surprising to find that the experience of negative behaviour is associated with continued smoking.36 Finally the PIQ is used in various forms, sometimes with 61 items,25 39sometimes with 45,38 sometimes 20,37 and sometimes in a modified form36 41 making comparisons between studies difficult.

Studies that asked about perceived functional support typically only did so at a single time point, usually asking the participants remaining at end of treatment.37-39 This may be open to reporting biases. Hence people who fail to stop smoking may not remain at the end of treatment and may be more likely to report a lack of social support for stopping. PIQ scores were used in the studies reviewed to determine effectiveness of the intervention and as a predictor for successful abstinence, not to assess baseline differences or changes over time. Two studies35 37 found a significant difference between groups in PIQ scores at end of treatment, suggesting that the intervention may have improved functional support as measured by PIQ. However, this did not translate into improved outcome in one of those studies,35 leading its authors to suggest that the PIQ may not be measuring what is beneficial about support.

Two studies asked about general or smoking related support at baseline.35 41 The studies did not establish if patients actually utilised the support offered, nor did they distinguish between expected support and support that was actually received.

INTERVENTION

The studies were limited by incomplete reporting, making the design of follow up research more difficult. Where reports were made, interventions were frequently multifactorial—for example, the provision of both support and relapse prevention training in the experimental group37 or attempting to influence not only the existence of support but also its quality.35 36 38This makes it difficult to establish the active ingredient of an effective intervention. In addition, in one study the experimental and control groups had different numbers of visits.35 In another, sites and not subjects were randomised37 and, of course, none of the studies were double blind so therapist effects cannot be ruled out.

All but one42 of the studies reviewed attempted to manipulate social support in the smokers' natural environment. In most cases this involved selecting only those patients who had a buddy available to them who was prepared to take part, thereby excluding those who did not have such a buddy. Another implication of this is that patients had already gone through the process of identifying and approaching a buddy, who in some cases would even sign consent before randomisation.35 Identified buddies not allocated to the social support condition have already been given a cue to act as a support and may well continue to do so anyway. A further implication of this is that in most cases researchers were attempting to influence pre-existing supportive relationships, often with a spouse. Other behavioural research suggests that these relationships can be very resistant to change.45

Most research assessed the use of social support within a clinic context using groups as the treatment method.12 34-40 In addition to the cost advantages of treating smokers in this way, groups are postulated to have greater efficacy owing to the social support element inherent in them.31 33 This could obscure any effect of the supportive intervention. In addition, where the timing of the groups was reported the majority of the meetings were before the quit date,36-39 with most smokers quitting around two weeks before the end of treatment. For most people craving will still be quite strong at this time. It is possible that interactions between smoker and their buddy may change once the support of a formal treatment programme is no longer available and that two weeks is too early in the quitting process for that to occur.

Discussion

Although social support is frequently correlated with success at stopping smoking, the results of buddy interventions for smoking are not impressive. The difficulty of translating the benefits of natural resources to effective interventions is not unique to this field.46 However, the poor design of much of the research could be partly responsible for the disappointing results.

Most of the studies occurred within the context of a smokers clinic, offering professional input. The evidence would suggest that in this context the use of buddies may be of some benefit. The positive results were from the only two clinic based studies with samples of over 100. Hence, it may be that other investigations simply lacked the power to detect any effect. However as the studies were so dissimilar no judgments can be made at this point about the form that any intervention should take.

Only one study attempted to influence buddy support in a self-help intervention,41 so conclusions in this area are even more difficult. They found that the provision of a support guide to be given to two allies had no effect either on outcome or on perceptions of support. However, they also found no difference between conditions in terms of the number of people invited to help, with only 58% of people in the support condition reporting handing out their guides. Currently then there is no strong experimental evidence that smokers attempting to stop on their own would benefit from approaching a family member or friend to support them.

It is likely that although all the studies reviewed aimed to influence social support, in practice they were manipulating quite different aspects of a broad construct. Interventions involving new ties and interventions using existing ones can both offer the “buddies” various levels of training. However, the latter involve attempting to develop or change an established relationship. In fact, it is not clear that support for stopping smoking has its effect through specific behaviours or if it does, what those behaviours may be. In a small prospective study Ginsberg and colleagues47 found that the interactions that predicted a successful quit attempt were not the same as interactions that led to a perception of support in smokers. Buddies from an intervention using new ties on the other hand may benefit from a sense of “common adversity”. Hence in the study reviewed they stopped smoking together, playing dual roles of supporter and supported. Equally, as the relationships were new they were unaffected by previous roles and so may have been prepared to make more effort. Therefore, there is a need for further research into the interpersonal process by which support of different types has an effect.

Overall there is enormous scope for research in this area using improved methodologies. Primarily better quality research is required to examine the benefit of different forms of buddy intervention in different contexts of smoking cessation—for example, community interventions, smokers clinics, primary care. Equally, the benefits of support interventions may differ for different populations. Much of the research has required pre-existing support as an inclusion criterion. However, it may be that socially isolated smokers benefit more from such interventions. Equally people with smokers in their social environments may benefit more from a buddy; thus Pirie and colleagues28 reported that having a buddy was particularly important for people with spouses who smoked. Also the influence of the smoking behaviour of the buddy requires examination. For example, Kviz and colleagues29 found that even though engaging a buddy was beneficial overall, the quit rate was near zero among people who chose to buddy up with continuing smokers. Interestingly West and colleagues,42 using new ties, found no evidence that relapse in one member of a pair was associated with relapse in the other. The role of different aspects of buddy support also requires elucidation. For example, the relationship between expected and received support may be of significance. Research in other areas suggests that if a person is disappointed with the levels of support they receive in a crisis that may be associated with a poorer outcome.48 Stopping smoking could be seen as an example of such a crisis. In addition, even if smokers perceive support within their social environment they may not utilise it.

The time course of any influence of social support also needs further examination. Gruder and colleagues37 found the benefit of their intervention was maintained in the long term. West and associates,42 however, did not conduct long term follow up. It is possible that if smokers benefit from the support of a new tie when quitting, they may be expected to have greater relapse rates when the support ends. Whereas pre-existing support may continue its influence over a longer period.

Most treatment packages for smokers recommend the use of social support despite the paucity of the experimental research evidence. Correlational and epidemiological research tends to uphold this recommendation, supported by some encouraging clinic based trials. However, more randomised controlled intervention studies are required to justify its widespread promotion as a smoking cessation aid.