The environmental profile of a community’s health: a cross-sectional study on tobacco marketing in 16 countries

Abstract Objective To examine and compare tobacco marketing in 16 countries while the Framework Convention on Tobacco Control requires parties to implement a comprehensive ban on such marketing. Methods Between 2009 and 2012, a kilometre-long walk was completed by trained investigators in 462 communities across 16 countries to collect data on tobacco marketing. We interviewed community members about their exposure to traditional and non-traditional marketing in the previous six months. To examine differences in marketing between urban and rural communities and between high-, middle- and low-income countries, we used multilevel regression models controlling for potential confounders. Findings Compared with high-income countries, the number of tobacco advertisements observed was 81 times higher in low-income countries (incidence rate ratio, IRR: 80.98; 95% confidence interval, CI: 4.15–1578.42) and the number of tobacco outlets was 2.5 times higher in both low- and lower-middle-income countries (IRR: 2.58; 95% CI: 1.17–5.67 and IRR: 2.52; CI: 1.23–5.17, respectively). Of the 11 842 interviewees, 1184 (10%) reported seeing at least five types of tobacco marketing. Self-reported exposure to at least one type of traditional marketing was 10 times higher in low-income countries than in high-income countries (odds ratio, OR: 9.77; 95% CI: 1.24–76.77). For almost all measures, marketing exposure was significantly lower in the rural communities than in the urban communities. Conclusion Despite global legislation to limit tobacco marketing, it appears ubiquitous. The frequency and type of tobacco marketing varies on the national level by income group and by community type, appearing to be greatest in low-income countries and urban communities.


Introduction
Tobacco is a leading cause of morbidity and mortality, responsible for an estimated 18%, 11% and 4% of deaths in high-, middle-and low-income countries, respectively. 1 Since the prevalence of smoking is falling in high-income countries but increasing in many middle-and low-income countries, the global burden of disease caused by tobacco use is expected to shift increasingly from high-income countries to countries with lower incomes.
As marketing by the tobacco industry plays a substantial role in smoking initiation, [2][3][4] complete bans on such marketing can be an effective means of reducing tobacco use. 5,6 In 2005, the World Health Organization's (WHO's) Framework Convention on Tobacco Control (FCTC) called for a comprehensive ban on all tobacco marketing. 7 However, the lack of relevant capacity and/or political will in many countries and the insidious influence of the tobacco industry have meant that the implementation of some of the FCTC's recommendations has been slow. 8 In this paper, we assess the global tobacco marketing environment by examining and comparing the extent and nature of tobacco marketing in 462 communities spread across 16 low-, middle-and high-income countries.

Data source
All of the data we analysed were collected as part of the Environmental Profile of a Community's Health study, which has already been described in detail. [9][10][11] This study is a component of the Prospective Urban Rural Epidemiology study -a large cohort study that is designed to examine the relationship between lifestyle factors and cardiovascular disease in adults aged 35-70 years. 10,11 The Environmental Profile of a Community's Health study includes an objective environmental audit in which trained investigators walk a predefined kilometre-long route within a study community. During each such walk, the investigators visit stores and systematically record physical aspects of the environment -e.g. the number of tobacco advertisements that they see. The second part of the Environmental Profile of a Community's Health study involves an interviewer-administered questionnaire that captures individuals' perceptions of their community -including whether the interviewees recall seeing certain types of tobacco marketing within the previous six months. 9 This questionnaire was administered to a subsample of the participants of the Prospective Urban Rural Epidemiology study.
We investigated data collected, between 2009 and 2012, in 16 countries. According to the World Bank's 2006 classification, 11 three of the countries -Canada, Sweden and the United Arab Emirates -were high-income, seven -Argentina, Brazil, Chile, Malaysia, Poland, South Africa and Turkey -were upper-middleincome, three -China, Colombia and the Islamic Republic of Iran -were lowermiddle-income -and three -India, Pakistan and Zimbabwe -were low-income. Although Bangladesh is included in the Prospective Urban Rural Epidemiology study, 10 we excluded Bangladeshi data on tobacco marketing from our analyses because they were relatively incomplete.

Measures of marketing
The Environmental Profile of a Community's Health study records both push and pull marketing. Push marketing, which aims to increase product availability, 12,13 was measured by trained researchers who recorded the number of tobacco outlets -e.g. vendors, street stands and general stores -seen during the audit walk and whether a tobaccoselling store visited during the walk sold single cigarettes. Pull marketing, which encourages customers to seek out a product through advertising and promotion, 12,13 was measured using both direct observation -i.e. the number of tobacco advertisements counted during the audit walk and whether the tobaccoselling store visited during the walk had point-of-sale tobacco advertising -and via self-report in interviews -i.e. whether an interviewee recalled seeing various forms of tobacco advertising in the previous six months. Almost all of the tobacco marketing measures that we examined reflected those covered by the  All  462  235  227  11 842  5809  6033  High-income  Canada  46  31  15  1145  807  338  Sweden  23  20  3  580  496  84  United Arab Emirates  3  1  2  89  26  63  Total  72  52  20  1814  1329  485  Upper-middle-income  Argentina  20  6  14  544  171  373  Brazil  14  7  7  387  202  185  Chile  5  2  3  127  51  76  Malaysia  33  18  15  1168  591  577  Poland  4  1  3  89  26  63  South Africa  6  3  3  194  99  95  Turkey  38  25  13  1207  795  412  Total  120  62  58  3716  1935  1781  Lower-middle-income  China  101  39  62  3131  1224  1907  Colombia  54  31  23  278  151  127  Iran (Islamic Republic of )  20  11  9  593  321  272  Total  175  81  94  4002  1696  2306  Low-income  India  88  37  51  2118  766  1352  Pakistan  4  2  2  111  57  54  Zimbabwe  3  1  2  81  26  55  Total  95  40  55  2310  849 14 However, we also assessed tobacco outlet density as this has been shown to play an important role in smoking prevalence among adults and adolescents. 15,16 Observed data For each country and country income group, the mean numbers of tobacco outlets and advertisements observed per community, the percentage of visited stores that sold single cigarettes and the percentage of visited stores that had point-of-sale tobacco advertising, were calculated -separately for the urban and rural communities.
As statistical tests showed that our outcome data were highly overdispersed, we used negative binomial multilevel regression models to examine differences in the number of observed tobacco outlets and tobacco advertisements between urban and rural communities and between country income groups. In these models, the number of outlets or advertisements was used as the outcome variable. Country income group and community type -i.e. rural or urban -were used as the categorical explana-tory variables, and a random effect was included for the country. Incidence rate ratios (IRRs) were obtained by exponentiation of the regression coefficient and reported with the corresponding 95% confidence intervals (CIs). As data on the sale of single cigarettes and point-ofsale advertising were based on only one tobacco-selling store per communityand it is not possible to know whether the selected store was representative of all tobacco-selling stores within the community -such data were not included in the regression analyses.

Self-reported data
To examine differences in self-reported marketing levels between community types and across country income groups, we considered 13 binary outcome variables. These included whether or not individuals reported seeing tobacco marketing of any of six traditional types of media -i.e. posters, signage, television, radio, print and cinema -and five non-traditional types -i.e. sponsorship, marketing on other products, marketing on the internet, free samples and vouchers. We also combined all the traditional types and all the non-traditional types of marketing into two separate binary variables.
We applied a logistic multilevel regression model to each of the binary outcome measures and again included categorical explanatory variables for country income group and community type. We also included random effects for country and community. Each model was adjusted for potential confoundersi.e. sex, age, education, smoking status, having close friends who smoke, access to the internet, television ownership and radio ownership. 2,[17][18][19] The resulting odds ratios (ORs) are reported with corresponding 95% CIs.
All of the models were fitted using the glmmadmb and glmer functions from the glmmADMB and lme4 packages of R version 3.0.2 (R Foundation, Vienna, Austria).

Results
We analysed data from 235 urban and 227 rural communities, across 16 countries (Table 1). Overall, 11 842 individuals who resided in the observed communitiesi.e. 5809 in the urban and 6033 in the rural communities -were interviewed and included in the final analyses.

Push marketing
There were marked differences in outlet type and density between countries and country income group ( Fig. 1 and Table 2 available at: http://www.who.int/ bulletin/volumes/93/12/15-155846). The mean number of tobacco-selling outlets observed in each community increased with decreasing country income, from 1.7 in the high-income countries to 3.4 in the upper-middleincome countries and over 5.0 in the lower-middle-income and low-income countries. This trend was driven largely by the relatively high numbers of vendors and street stands observed -a mean of almost two per community -in low-income countries. No such outlets were observed in high-income countries and, on average, only 0.2 and 0.7 were observed per community in the upper-middle-income and lowermiddle-income countries, respectively. The mean number of general stores observed per community did not follow the same pattern -1.7, 3.2, 4.6 and 3.4 in the high-, upper-middle-, lowermiddle-and low-income countries, respectively.
Combining data from all 16 countries, more vendors and/or street stands were observed in the urban communities than in the rural -means of 0.9 and 0.5 per community, respectively -and the urban communities also had a higher mean number of general stores selling tobacco -3.7, compared with 3.3 per rural community. However, these urban/ rural differences were not consistent across all four country income groups ( Fig. 1 and Table 2).
The sale of single cigarettes was not observed in any of the communities in eight of the countries (Table 2). However, overall, outlets selling single cigarettes became increasingly common with declining country income ( Fig. 2 and Table 2). Although the urban/rural differences in the sale of single cigarettes varied by country income group, the sale of single cigarettes was more common in urban than rural communities in both lower-middle-and low-income countries.

Pull marketing
Tobacco advertisements were much more common in low-income countries than in the other countries. Very few tobacco advertisements were seen in high-income countries. In middle-and low-income countries, means of approximately 1 and 3 observed advertisements per community were recorded, respectively ( Fig. 3 and Table 2). Combining data from all countries, tobacco advertisements were more common in the urban than rural communities, with means of 1.7 and 0.9 observed per community, respectively.
After controlling for community type and country income group, the middle-income countries had similar   Table 3).

Self-reported data
Of the 11 842 interviewees, 5349 (45%; range: 4-100%) reported exposure to at least one type of tobacco marketing over the previous six months and 1184 (10%; range: 0-56%) reported exposure to at least five types of marketing over the same period (available from the corresponding author).

Traditional
Interviewees in high-income countries were least likely to report exposure to all forms of traditional marketing except print media, although differences between other country income groups varied by the type of marketing ( Fig. 5; further details available from corresponding author). Overall, television marketing -seen by 3501 (30%) of interviewees in the previous six months -was the most common form of traditional marketing, followed by posters (2334; 20%), print media (1949; 16%), signage (1934; 16%), radio (1465; 12%) and cinema marketing (567; 5%). All forms of traditional marketing except television marketing -and exposure to at least one form of traditional marketing -were less common in rural communities than urban ones (Table 4 available at: http://www.who.int/bulwletin/volumes/93/12/15-155846).
The likelihood that interviewees from low-income countries reported exposure to at least one form of traditional marketing was almost 10 times higher (OR: 9.77; 95% CI: 1.24-76.77) than in high-income countries. Specifically, the likelihood of exposure to radio (OR: 46.05; 95% CI: 1.29-1642.57), signage (OR: 11.02; 95% CI: 1.07-113.60), television (OR: 9.42; 95% CI: 1.21-73.20) and cinema marketing of tobacco (OR: 3.08; 95% CI: 1.46-6.49) were significantly higher in low-income than in highincome countries (Table 5). Compared with the interviewees from urban communities, the likelihood that interviewees from rural communities reported exposure to traditional marketing was either significantly lower -posters, signage, print and cinema marketing -or not significantly different -television and radio marketing ( Table 5). Tobacco marketing in 16 countries Emily Savell et al.

Non-traditional
Non-traditional marketing was reported less frequently than traditional marketing (Table 4). Although tobacco marketing on other products -e.g. umbrellas -was the most commonly reported form of non-traditional marketing, only 1468 (12%) of the interviewees reported seeing such marketing in the previous six months ( Fig. 6 and Table 4). Country income group appeared to have little impact on exposure to non-traditional marketing but overall exposure and exposure to each form of non-traditional marketing appeared more common in the urban communities than in the rural. After controlling for confounders, the likelihood of exposure to non-traditional tobacco marketing in the low-and middle-income countries appeared similar to that in the high-income countries (Table 6). However, compared with their urban counterparts, the likelihood that rural interviewees reported exposure to one or more forms of non-traditional marketing was significantly lower (OR:

Discussion
Our study has three important findings in relation to tobacco marketing. First, we identified high levels of ongoing exposure to tobacco marketing -despite 14 of the study countries having ratified the FCTC at the time the data were collected; by December 2014, Argentina had signed but not ratified the FCTC and Zimbabwe had only acceded to it. Although ratification requires countries to implement comprehensive marketing bans, 10% of the interviewees reported seeing at least five types of tobacco marketing in the six months before interview and 45% reported seeing at least one type of tobacco marketing over the same period. Second, we detected substantially higher levels of tobacco marketing in the lower-income countries we investigated than in the higherincome. This result is consistent with the tobacco industry specifically targeting low-and middle-income countries, 20,21 Table 5. The likelihood that interviewees reported seeing traditional types of tobacco marketing within the previous six months, 16   which could be due to large youth populations in lower-income countries and to high-income countries having more established policies on tobacco control. 22 Third, for 13 of 15 marketing measures, exposure was significantly lower in the rural communities than in the urban ones.
High levels of tobacco marketing may reflect failure to enact legislation and/or to enforce compliance. 23 Yet many of our interviewees -even those from countries with highly regarded tobacco control measures such as Brazil, Canada and Sweden 24-26 -reported substantial exposure to tobacco marketing. This indicates that the tobacco industry may still be finding ways to market its products. Given that we recorded 10 times greater exposure to traditional marketing in the low-income countries than in the high-income countriesbut similar levels of exposure to nontraditional marketing across all country income groups -it appears that legislation may have been relatively successful in controlling traditional marketing in high-income countries. This success may have resulted in the tobacco industry using newer, less regulated forms of marketing. Therefore, enforcement may need to be stronger and legislation continuously adapted to the changing marketing practices of the tobacco industry. Data on the tobacco industry's marketing expenditure would also be useful, but such data are available for very few countries 27 and not for any of our study countries.
Our observation of more intense tobacco marketing in urban communities than in rural communities is consistent with evidence that the tobacco industry focuses its marketing and distribution on areas with the greatest potential impact -i.e. areas with dense populations 28,29 that can be easily reached at relatively low cost. 30 Our study had several limitations. First, although diverse, 11 the countries studied are not necessarily representative of low-, middle-and high-income countries globally and the communities investigated within each country are not necessarily representative of all communities. 10 Although this means that the results cannot reliably be extrapolated to all communities within a country, the demographic characteristics of our interviewees do appear to match those of adults in the corresponding national populations. 11 We also note that the main tobacco company in two of the three lower-middle-income countries -i.e. China and the Islamic Republic of Iran -is state-owned. 31 Countries with state-owned monopolies traditionally do not market their products aggressively because the lack of competition renders this unnecessary. 32 Our findings, especially those on self-reported marketing, indicate that the tobacco marketing environment may well be affected by state ownership of the local tobacco industry. In the Islamic Republic of Iran, for example, exposure to most forms of marketing appeared to be less intense than in other lowermiddle-income countries. Our results appear to be consistent with data from WHO's Global Adult Tobacco Survey 33 that was conducted in 16 countries, including six of our study countries -Brazil, China, India, Pakistan, Poland and Turkey. Although the WHO's survey did not include statistical comparisons, it did show relatively high self-reported exposure to tobacco marketing in lowerincome countries -with the exception of the Russian Federation -and in urban communities. 33 Our findings also seem similar to those from the International Tobacco Control Policy Evaluation Project, 34 which has collected data from 22 countries, including five of our study countries -Brazil, Canada, China, India and Malaysia.
Second, the sample size varied markedly by country -both for the number of communities and number of interviewees. We would expect more uncertainty in an estimate for a country in which only a few communities are sampled. Additionally, the number of countries per country income group and the small number of communities surveyed in two of the three low-income countries may explain the wide CIs seen in some significant comparisons between low-and high-income countries. Third, although the methods used have been shown to be reliable, 9 only one tobacco-selling store was visited per community during the walk -and it is not possible to know whether the selected store was representative of all stores within the community. Fourth, our study was limited by difficulties in estimating the tobacco industry's marketing expenditure in each study country and by exposure of many individuals to cross-border marketing -including internet marketing. Finally, the study used data collected between 2009 and 2012 and some of the countries have since taken further steps to strengthen their tobacco marketing regulations.
Our study also has strengths. The Environmental Profile of a Community's  Health study takes a comprehensive approach to data collection, using both direct observation and self-reported data to assess the level and nature of diverse forms of tobacco marketing at both community and individual level; an approach shown to be reliable. 9 The countries included in our analysis are very diverse in terms of both economics and culture. Additionally, although differences in self-reported exposure to marketing will reflect access to certain types of media, we were able to control for internet access and television and radio ownership in the individual-level models.
This study indicates that tobacco marketing remains ubiquitous even in countries that have ratified the FCTC. Given the strength of the link between marketing by the tobacco industry and the prevalence of smoking, 2-4 there is an urgent need for countries either to implement comprehensive controls on tobacco marketing or to enforce such controls more effectively. ■