For alcohol, tobacco and drugs (ATD) interventions, a search was conducted for good quality systematic reviews, and the included reference lists for these reviews formed the sample frame for potentially eligible social marketing intervention studies. This strategy had two advantages: it reduced the search process to one that was manageable in the timeframe, and it ensured that all the studies subsequently included had already been judged of sufficient methodological quality, by previous reviewers, to yield reliable evidence.

The initial search for systematic reviews of ATD interventions was limited to reviews designed, conducted and the studies analysed in such a way that biases were minimised and therefore deemed to be of good quality (as defined by Khan et al., 2001). The key sources were taken to be: the Cochrane Database of Systematic Reviews; the Centre for Reviews and Dissemination's databases (Database of Abstracts of Reviews of Effects and Health Technology Assessment Database); the EPPI-Centre (the Evidence for Policy and Practice Information and Co-ordinating Centre's) publications list; and NICE (the National Institute for Health and Clinical Excellence's) publications database. No limits were set on the types of interventions; as social marketing interventions can use many different methods and be implemented in many different settings, it was not desirable to exclude any intervention types at this stage. This yielded 35 systematic reviews, covering a diverse range of interventions (for example, health promotion in schools, interventions for preventing tobacco sales to minors, therapeutic drug treatment programmes, workplace interventions). Unless there was enough information in the reviews on individual interventions to assess them against the six social marketing benchmarks, full text studies were retrieved. Supplementary papers were often required to provide information on, for example, a programme's development. From the 35 reviews, 310 individual studies were retrieved and assessed in full text against Andreasen's six criteria. A total of 35 studies met all six of Andreasen's criteria for a social marketing intervention, and were included in the review. By “a study”, we mean all the published papers reporting on a single evaluation of a specific programme (for example, where a programme was followed up over several years within one evaluation, all the outcome papers are included as one study). Sometimes the same programme or variants of it are evaluated in different settings with different participants; in this case, each evaluation is counted as a separate study, again each potentially comprising more than one paper.

Initially a search was conducted for any existing systematic reviews, non-systematic reviews and individual intervention studies in which the label “social marketing” had been applied to programmes. A series of electronic databases was searched using combinations of the search terms physical activity, exercise and social marketing: the Cochrane Library; PsycINFO; PubMed; the Arts and Humanities Citation, Social Science Citation and Science Citation Indices; the Centre for Reviews and Dissemination's databases; and NICE (the National Institute for Health and Clinical Excellence's) publications database.

Although the database searches described above were for papers including the term “social marketing”, as the search process continued, interventions were included for assessment without the term. This yielded 17 reviews (systematic and non-systematic) and 48 articles in total, covering a range of dedicated physical activity interventions or interventions that included a physical activity component. From the 17 reviews (systematic and non-systematic) and 48 articles retrieved initially, a further 62 articles were generated and retrieved, totalling 110 articles assessed against Andreasen's six criteria. A total of 22 interventions met all six of Andreasen's criteria for a social marketing intervention, and were included in the review.

In total, 21 of the interventions were school-based programmes. Of the ATD programmes, seven primarily targeted a single substance, and nine adopted a generic approach to substances. A total of seven school-based programmes targeted physical activity. Classroom education often took place alongside other activities such as mass media campaigns, school-wide and community events.

Of the interventions 22 were multi-component community interventions (for the purpose of this review, multi-component and community interventions have been discussed together, as they share many features). Of these, ten were targeted at the general adult population; four addressed smoking cessation only, or smoking cessation within the context of cardiovascular risk disease behaviours, and eight addressed physical activity. In total four comprised school-based ATD prevention curricula reinforced by extensive community components (Pentz et al., 1989; Perry, 1992, 1996; Vartiainen et al., 1998), and four comprised community and policy activities designed to reduce youth access to substances as well as activities targeted at youth themselves (Biglan et al., 2000; Pawtucket, 1995; Puska, 2002; Perry et al., 1996). Four targeted youth physical activity (Baranowski et al., 2003; Beech et al., 2003; Resnicow et al., 2000; Story et al., 2003), one targeted physical activity in older people (Reger et al., 2002) and one targeted newly married couples (Burke et al., 2002).

Five interventions were primarily mass media-based. Three were concerned with adult smoking cessation (Egger et al., 1983; McAlister et al., 1992; McPhee et al., 1995), one with youth smoking prevention (Flynn et al., 1994) and one with physical activity (Huhman et al., 2005). All included activities delivered through other channels in addition to the media element (e.g. educational materials, telephone helpline, support in the community). If no channels were used other than mass media, an intervention was judged not to have utilised the marketing mix fully, and was not included in the review.

Two interventions were designed to restrict youth access to substances through increasing retailer/server compliance with existing laws, greater enforcement of the law, or adoption of new policies and legislation (Wildey et al., 1995; Forster et al., 1998). The San Diego intervention mostly comprised retailer education backed up with media and community events, while TPOP adopted a direct action, community organisation approach directed both at retailers and local legislators. In addition, three of the multi-component community interventions listed above included community and policy activities designed to reduce youth access to substances as well as activities targeted at youth themselves (Biglan et al., 2000; Perry et al., 1996; Wagenaar et al., 2000).

One smoking cessation intervention was delivered through church coalitions and primarily targeted at African-Americans (Schorling et al., 1997). Two interventions were delivered in workplaces, one concerned with smoking cessation (Windsor et al., 1988) and one with promoting physical activity (Neiger et al., 2001). A family- and child-training programme designed to reduce drug and alcohol use sought to strengthen family protective factors (Spoth et al., 2001).

As noted above, social marketing is not a theory in itself but rather draws from many bodies of knowledge to understand how to influence people's behaviour (Kotler and Zaltman, 1971). This was reflected in the wide range of theories and models adopted in the programmes (where information was actually provided or could be inferred; not all studies describe the theoretical basis of the programme). Several of the school-based programmes were informed by social influences theory; this emphasises “the importance of social and psychological factors in promoting the onset of drug use” and comprises three major components, “psychological inoculation, normative education and resistance skills training” (Botvin et al., 2001). The majority of the interventions comprised theory-driven interactive classroom curricula adopting a social influences approach, usually involving practice of resistance skills and other activities designed to address direct and indirect pressures to use substances.

Several of the interventions were “multi-component”. In the context of drug prevention, “multi-component” refers to interventions which, using multiple channels and activities, target not only the individual but also their immediate family and peer group and the wider environment, which shapes drug use norms. A typical multi-component programme such as Project STAR (Pentz et al., 1989) comprises a school curriculum, media, parent activities, community organisation and policy activity.

Many studies used social cognitive or social learning theory in their development (e.g. Baranowski et al., 2003; Beech et al., 2003; Story et al., 2003). Social cognitive theory explains how people acquire and maintain certain behavioural patterns, while also providing the basis for intervention strategies. (Bandura, 1997). The theory emphasizes the interactions between a person's cognitions, on the one hand, and his/her behaviour on the other, through processes such as self-efficacy and outcome expectancies (or response efficacy). The trans-theoretical model, which was used in several interventions (e.g. Matsudo et al., 2002; Prochaska and Sallis, 2004; Reger et al., 2002) hypothesises that behavioural change unfolds through a series of stages. Therefore in this model it is critical to understand and identify the stage an individual is in before a successful change intervention can be designed and applied.

Community interventions sometimes claimed to be underpinned by theories of community organisation and community participation, (e.g. Pawtucket, 1995; MHHP, 1994) or the planned approach to community health model (Matsudo et al., 2002; Brownson et al., 1996; Goodman et al., 1995).

Media advocacy was used in some interventions, particularly those concerned with policy and environmental change; for example, in Project TRUST (Wildey et al., 1995), retailers who complied with the law on underage access to tobacco were “rewarded” with positive newspaper coverage, while those who did not were “named and shamed”. Finally, two family-focused alcohol interventions (Spoth et al., 2001; Perry et al., 1996) drew on models of problem behaviour protective and risk factors to foster factors which would protect against the development of substance use, such as family cohesion and managing emotions and conflict.

Use of several other theories and models guided the development of the included interventions including the stage theory of innovation (Brownson et al., 1996), behaviour change theory (O'Loughlin et al., 1999), and organisational change theory (CATCH et al., 1996).

All but 46 of the programmes were implemented in North America; two were Dutch (De Vries et al., 1994; Cuijpers et al., 2002), three were Australian (Burke et al., 2002; Egger et al., 1983; McBride et al., 2000), two were Finnish studies part of a larger programme (Vartiainen et al., 1998; Puska, 2002), and one was Brazilian (Matsudo et al., 2002).

Most of the school-based programmes targeted the early secondary school years (11-14); one targeted a younger age group (CATCH et al., 1996), and one targeted an older age group (Wagenaar et al., 2000). Although the majority of the school-based programmes were “universal”, that is, designed for the whole student population in specific school years, some were deliberately tailored for high risk or disadvantaged populations. For example two trials of Project Towards No Drug Abuse were targeted at pupils in Californian “continuation high schools”, schools for students who are not able to complete formal high school education because of behavioural or other problems, including drug use (Sussman, 1998, 2002).

A total of 14 interventions were targeted at minority ethnic groups. This varied from Native Americans (Caballero et al., 2003), African-American schoolgirls (Story et al., 2003), Vietnamese men in the USA (McPhee et al., 1995) to Hispanic adults in the Texas-Mexican Border area (McAlister et al., 1992). Two interventions included components targeted at older people (Matsudo et al., 2002; Reger et al., 2002) and another two were specifically targeted at low-income groups (O'Loughlin et al., 1999; Brownson et al., 1996). One intervention was targeted at people with low levels of literacy (Pawtucket, 1995).

The vast majority of the interventions were evaluated using a randomised controlled trial or quasi-experimental design. The period of follow-up ranged widely in the studies from a month after implementation to several years (Vartiainen et al., 1998 was unusual in having a 15 year follow-up). The majority of the studies involved one to two-year follow up.

The main outcomes examined in the ATD programmes included overall prevalence of substance use, onset, experimentation, daily/weekly/monthly use, progression from one stage of use to another, and cessation. Some also took attitudinal, knowledge and skills measures. One alcohol-focused intervention also measured alcohol-related behaviours and incidents (drink-driving, drink-related crashes) (Wagenaar et al., 2000). The main outcomes measured in the youth access interventions were illegal sales to underage minors (usually assessed through test purchasing) and sometimes, other measures of retailer behaviour and attitudes, such as frequency of checking ID and perceptions of the risk of prosecution.

In the physical activity interventions the main behavioural outcome examined was level of physical activity. Levels of physical activity were measured in a variety of different ways across the interventions – by frequency, total kcal expended, minutes spent on physical activity, or distance covered – and several studies used more than one measure. Psychosocial measures such as self-efficacy for physical activity or social support for physical activity, and physiological outcomes such as blood pressure, cholesterol level and Body Mass Index, were also reported in several of the included studies.

Smoking prevention. A total of 21 studies (14 school-based interventions, five multi-component community interventions, one media-based programme and one access intervention) examined whether interventions were effective in preventing smoking among young people. Many measured only short-term impact, although a few took medium and longer-term follow-ups. Results are discussed below by length to follow-up.

In total 18 of these 21 studies examined short-term impact (up to one year), of which 13 reported significant positive effects. Several of the short-term effective interventions were school-based programmes adopting the “social influences” approach (Maibach and Parrott, 1995). For example, students who received the Project SMART social influences programme had significantly lower smoking onset than control students, equivalent to a 38 percent reduction, at the end of one year (Hansen et al., 1988), while a 12-session version of Project Towards No Drug Abuse found lower monthly smoking at one-year follow-up for students who received the programme led by a health educator, compared with students who learned it through self-instruction and control group students (Sussman et al., 2002). Four of the short-term effective interventions were multi-component community programmes. For example, Project SixTeen (Biglan et al., 2000), designed to reduce both illegal sales of tobacco and youth tobacco use, was associated with lower smoking prevalence (smoking in the past week) at one-year follow-up.

In total 11 studies examined medium term impact on smoking prevention (up to two years) and seven of these reported significant positive effects. Four of these were multi-component community interventions. Project STAR, or the Midwestern Prevention Project, resulted in a significantly lower increase in smoking in intervention schools compared with controls at both one and two year follow-ups (Pentz et al., 1989). The other multi-component community interventions with medium term effects were the North Karelia project (Vartiainen et al., 1998), the Minnesota Heart Health “Class of 89” study (Perry, 1992) and Project SixTeen. Again, effects were also found for some school-based interventions adopting a social influences approach (e.g. Sussman, 1993, 2002). One intervention, which was effective in both short and medium term, combined mass media smoking prevention television programming with a school curriculum (Flay et al., 1995).

Five studies examined longer-term impact (over two years), and two of these reported sustained significant positive effects on prevalence. Among students who received the youth component of the North Karelia project, a major community intervention to reduce cardiovascular disease, at 15-year follow-up, when participants were aged 28, meantime lifetime consumption was 22 percent lower in the intervention community compared with the control area. In the “Class of 89” study (Perry, 1992) students who received a three-year school-programme within the five-year Minnesota Heart Health Program, had significantly lower weekly smoking prevalence than control community students at all follow-ups up to five years.

A total of 15 studies (ten school-based interventions, four multi-component community interventions, one delivered in a family setting) examined whether interventions were effective in preventing alcohol use or reducing the harm associated with alcohol use (e.g. drink-driving). Again, most took only short-term follow-up measures, although some also measured medium and longer-term impact.

A total of 13 of the 15 studies examined short-term impact on alcohol use (up to 12 months), of which eight reported some significant positive effects. Two were trials of the school-based Life Skills Training programme with minority ethnic inner city student populations (Botvin et al., 1997, 2001), while an Australian school-based alcohol harm minimisation programme found a lower increase in alcohol consumption one month after the intervention (McBride et al., 2000).

Seven studies examined medium-term impact (one to two years) on alcohol use, and four of these reported some significant effects. One of these was a trial of a three-year school-based programme (Cuijpers et al., 2002), which found a significant impact on daily alcohol use at immediate post-test and at two years after the start of the intervention. There was also a significant decrease in number of drinks consumed per occasion at both follow-ups.

Finally, two out of four interventions, which examined effects in the longer term, over two years, reported some significant impacts. Project Northland (Perry et al., 1996), which comprised a three-year school curriculum, peer and parent activities and community taskforces, found a significant impact on past month and past week alcohol use (p < 0.05 for each) in the intervention group compared with the control group at 2.5 years, although the effect had dissipated at four years. Project Northland also found a significant reduction in “proneness to alcohol, drug and family problems” after three years of the programme, suggesting that the family-focused 6th grade component of the programme was effective in influencing wider precursors of problem behaviour. A family-focused intervention (Spoth et al., 2001), targeting both parents and children, found lower alcohol use onset in the intervention group compared with the control at both one and two year follow-ups; the increase in “ever use” and “ever been drunk” was lower in the intervention group than the control group at every follow-up up to four years, with increasing effect sizes, suggesting that the intervention intensified in impact over time.

A total of 13 studies (ten school-based interventions, three multi-component community interventions) examined whether interventions were effective in preventing or reducing illicit drug use.

In total 12 of the studies examined short-term impact (up to 12 months), and eight of these reported some significant positive impact on illicit drug use. Nearly all were school-based programmes, including a trial of Life Skills Training with minority ethnic inner city students which found less frequent marijuana use and lower polydrug at three month follow-up (Botvin et al., 1997), two versions of Project ALERT, both of which found short-term reductions in marijuana use (Ellickson, 1990, 2003) and a Dutch school programme which found significant reductions in marijuana use immediately after the intervention (Cuijpers et al., 2002).

Of the six studies, which examined medium-term impact (one to two years), two reported some significant effects. These were a version of Project Towards No Drug Abuse (Sussman et al., 2002) tested with minority ethnic inner city students, which found significantly lower hard drug use at two years among students who received a health-educator led version of the programme, and the multi-component programme Project STAR (Pentz et al., 1989) which found significant reductions in last month marijuana use in intervention students compared with controls both in the short term and at two year follow-up.

Neither of the two studies which examined longer-term impacts on drug use (over two years) found sustained effects (Ellickson and Bell, 1990, Project Northland).

Nine studies examined whether interventions had an impact on smoking cessation (four multi-component community interventions, three media-based interventions, a church-based programme and a worksite programme.

Two of the interventions, both two-year mass media-based programmes (the North Coast Quit for Life programme in NSW, Australia, and a programme for Vietnamese American men in Texas (McPhee et al., 1995), had significant effects. A further five interventions had weak effects. For example, the Alliance of Black Churches Project (Schorling et al., 1997), which adopted a community organisation approach and included the formation of church coalitions as well as individually-focussed activities and community activities, found a higher quit rate in the intervention community than the comparison community, but the differences were not significant, the trend suggested a possible intervention effect. Two interventions had no effect.

A total of 21 interventions aimed to increase level of physical activity as defined by total time spent or frequency of activity (14 community based interventions, seven school-based interventions, one mass media intervention and one work-based intervention). Of these, eight of the 21 reported a significant positive effect on behaviour. For example, the Social Marketing for Public Health Employees Intervention (Neiger et al., 2001), a workplace-based programme comprising communications and promotions, ongoing activities, one-off events and environmental changes, resulted in pre-test to post-test increases in the primary treatment group on three levels of physical activity. Similarly, the Wheeling Walks intervention (Reger et al., 2002), a community based campaign to promote walking among sedentary 50 to 65 year old adults underpinned by the Theory of Planned Behaviour, had a positive effect on physical activity levels in terms of total time spent and frequency of walking.

A total of 11 studies examined changes in physical activity psychosocial variables such as self-efficacy or stage of change for physical activity. Of these, six reported a positive effect for at least one variable. Three of these six were school-based interventions, two were community based and one was based in a workplace setting. Pathways (Caballero et al., 2003), a school based intervention targeting 3rd and 5th grade American-Indian children and comprising a physical activity programme supplemented by a classroom curriculum and family component, significantly increased self efficacy for physical activity in the treatment group. It also found a positive effect on stage of change to physical activity. The New Moves intervention (Neumark-Sztainer et al., 2003), incorporating a specially designed curriculum, physical activity classes, instruction and education packs, found a mean increase in stage of change for physical activity post-intervention, and at one year follow-up, with around twice as many girls progressing in stage than regressing.

In total 14 studies measured physiological outcomes following physical activity. Of 11 studies using Body Mass Index (BMI) as a measure, two found a positive effect. These were a community based study targeted at newly co-habiting couples (Brownson et al., 1996) and major seven year community based intervention targeting a general adult population. (Pawtucket, 1995).

Of six studies using cholesterol levels as a measure three found a positive effect, all of which were community-based interventions (Burke et al., 2002; Goodman et al., 1995; Puska, 2002).

Two interventions used CVD as a measure with both resulting in a reduction in CVD rates (Puska, 2002; Pawtucket, 1995). Finally five of the 14 interventions used blood pressure as a measure but only one of these five interventions showed a positive effect (Burke et al., 2002).

Several of the interventions sought also to bring about changes at an environmental and/or policy level.

Four studies reported changes in retailers' behaviour in relation to selling tobacco and alcohol illegally to underage youth. In two studies this was the main focus of the programme (Forster et al., 1998; Wildey et al., 1995), while in the others, the work with retailers was part of a multi-component community intervention also involving school- and youth-targeted activities (Biglan et al., 2000; Wagenaar et al., 2000). Project Trust, a one-year retailer education campaign which included personal visits and positive media coverage for law-abiding retailers, reduced illegal sales from 70 to 32 percent in four out of six intervention communities, compared with a comparison area; the effect was sustained at six-month follow-up (Wildey et al., 1995). TPOP (Tobacco Policy Options for Prevention) adopted a direct action community organising approach to encourage adoption of tobacco ordinances and deter illegal underage sales in Minnesota (Forster et al., 1998). Illegal sales fell from 36.7 to 3.1 percent in intervention communities, but also by a similar amount in control communities; the difference was not significant.

A multi-component community intervention in Oregon, comprising media advocacy, youth anti-tobacco activities, community mobilisation, and giving retailers feedback on their response during test purchase attempts, showed a reduction in the mean level of illegal sales from 57 to 22 percent (Biglan et al., 2000). There was no comparison community in this study, which limited the conclusions, which could be drawn. The fourth intervention addressed alcohol and included community mobilisation, media advocacy and merchant education (Wagenaar et al., 2000). Retailers reported more frequent checking of age ID and a higher perceived risk of prosecution following the intervention, while test purchase attempts found lower incidence of selling under-age, but none of the differences were significant.

Three interventions specifically sought to promote the adoption of substance-related policies. One goal of the TPOP (1996) intervention was to “change local ordinances to more effectively restrict youth access to tobacco”. At the end of the intervention period, all seven intervention communities had adopted a comprehensive ordinance, with varying ingredients. Over the same period, three of the seven control communities also adopted modifications to their tobacco ordinances, but the study authors described the control community ordinances as “weaker and much less comprehensive” than the intervention community ordinances.

The CATCH intervention sought to promote the adoption of formal tobacco-free policies in its 56 intervention schools, spread across three states (CATCH et al., 1996). The number of schools adopting policies increased markedly over the three years of the study from a baseline figure of 49.7 to 76.8 percent. However, because of other events and trends at the time of the intervention, it was difficult to attribute the policy adoption process to the intervention. In Project Northland, the authors note that five alcohol-related ordinances were successfully passed in the first year of the intervention, including requirement of responsible beverage service training to prevent illegal sales to underage youth and intoxicated patrons (Perry et al., 1996). Although school alcohol education programmes used in the control communities were monitored in the control communities during the intervention, policy developments in the same period were not recorded, and therefore it was not clear to what extent the alcohol policy changes in Project Northland communities were attributable to the intervention.

There is somewhat limited evidence of the effectiveness of interventions aimed at changing policy or the environment in relation to physical activity. In the M-Span intervention (Sallis et al., 2003), key school personnel met regularly with project staff to select and implement policy changes to create healthier school environments, such as allowing students to use activity areas after school and hiring aides to lead activity programmes. Nonetheless it was noted that project support for the groups was probably inadequate to yield more meaningful policy changes.

In the Bootheel Heart Health Project (Brownson et al., 1996), an evaluation of environmental factors was conducted and walking paths were constructed in low-income communities where cost was a barrier to other forms of physical activity. This led to the adoption of a policy to construct a network of such paths throughout the community. In the North Karelia Project (Puska, 2002) policy changes were effected which led to the provision of dedicated walking space and recreational areas assisting in the delivery of the intervention. The Pawtucket Heart Health Program (Pawtucket, 1995) aimed to influence the environment in which risk factor behaviours occurred, leading to efforts to effect policy change within the community. These efforts led to the construction of a series of leisure facilities and a multiple station exercise course. However, in each of these interventions policy developments in the control communities were not recorded and therefore it is unclear to what extent policy changes in the experimental communities were attributable to the interventions.