|Home | About | Journals | Submit | Contact Us | Français|
African Americans currently bear the greatest burden of morbidity and mortality due to smoking, and exposure to pro-tobacco media messages predicts smoking. This study compared the concentration (proportion of media messages that are for tobacco) and density (pro-tobacco media messages per person) of pro-tobacco media messages between African American and Caucasian markets.
We searched Medline (1966 to June 2006), PsychINFO (1974 to June 2006), and CINAHL (1982 to June 2006) for studies from peer-reviewed journals directly comparing the volume of pro-tobacco media messages in African American and Caucasian markets. From each study, we extracted the number of total media messages, the number of tobacco-related messages, and the number of residents living in each market area. We calculated the concentration and density of tobacco advertising in each market.
Out of 131 studies identified, 11 met eligibility criteria, including seven comparing billboard/signage in African American and Caucasian markets and four comparing magazine advertising in African American and Caucasian markets. Meta-analysis estimated a pooled odds ratio of 1.7 (95% confidence interval [CI] 1.1, 2.6) for a given billboard being smoking-related in African American vs. Caucasian market areas (i.e., concentration). The pooled rate ratio of the density of smoking-related billboards was 2.6 (95% CI 1.5, 4.7) in African American vs. Caucasian market areas. Magazine data were insufficient for meta-analysis.
Available data indicated that African Americans are exposed to a higher volume of pro-tobacco advertising in terms of both concentration and density. These findings have important implications for research, policy measures, and educational interventions involving racial disparities due to tobacco.
Smoking remains the leading cause of preventable death and disease in the U.S., causing more than 440,000 deaths per year1 and costing the U.S. more than $150 billion in direct and indirect costs annually.2 African Americans currently bear the greatest burden of morbidity and mortality due to smoking.3 For instance, total mortality from lung cancer is 21% higher among African Americans than among Caucasians,4 and African American mortality from stomach cancer is 127% higher than that of Caucasians.4 Epidemiological analyses have suggested that tobacco-related disparities between African Americans and Caucasians are so profound that reversing them could help eliminate all cancer disparities between these racial groups.5
Exposure to pro-tobacco media messages is now known to be a potent risk factor for tobacco use.6–10 In fact, recent studies suggest that exposure to pro-tobacco media may have a more powerful influence on smoking than other factors more traditionally linked with smoking, such as parental smoking, sibling smoking, sensation seeking, and rebelliousness.7,10,11 Furthermore, the most recent report of the U.S. Surgeon General focusing on racial/ethnic health disparities attributed to tobacco concluded that the top three factors influencing tobacco use among African Americans were all related to tobacco promotion: (1) the tobacco industry's marketing relationship with the African American community, (2) the targeting of minority members by the tobacco industry, and (3) the promotion of tobacco in minority neighborhoods and in publications geared toward African Americans.3
It is unclear, however, whether African Americans are indeed exposed to a higher volume of pro-tobacco advertising than are Caucasians. Although some researchers have found that there are more smoking advertisements in minority neighborhoods and publications,12–15 other work has not confirmed these results.16–19 Furthermore, prior studies have reported different results for the concentration (i.e., proportion of all advertisements per area for tobacco) and density (i.e., cigarette advertisements per resident) of advertising. For example, Hackbarth et al. found that African American neighborhoods in Chicago had a higher density of tobacco advertising per resident, but that Caucasian neighborhoods actually had a higher concentration of tobacco billboards.16 Magazine studies also have reported conflicting results with regard to concentration and density.14,19
It will be useful and instructive to distinguish these different elements of volume of advertising (density vs. concentration). If African Americans are exposed only to a higher density of advertisements but not a higher concentration, then many different types of products, not only tobacco, vie for the attention of African Americans. If African Americans are exposed to both a higher density and a higher concentration of tobacco-related advertisements, however, this would imply that African Americans may be special targets of the tobacco industry. In this latter case, more aggressive policy and educational measures may be necessary to reduce the impact of such advertising exposure in these populations.
The purpose of this study was to conduct a systematic review and meta-analysis of peer-reviewed literature to determine if African Americans are exposed to a higher volume of pro-tobacco mass-media messages, both in terms of density and concentration, and to determine point estimates of any differences. Our a priori hypothesis was that pooled data would show that African Americans are exposed to both a higher density and a higher concentration of pro-smoking advertisements.
We searched Medline (1966 to June 2006), PsychINFO (1974 to June 2006), and CINAHL (1982 to June 2006) for all English-language journal articles involving mass communication, smoking, and African Americans. Articles involving mass communication were identified using expanded searches with subject headings related to media communications and marketing, and subheadings such as, but not limited to, communication, radio, television, mass media, and advertising.
We searched for pro-tobacco messages in all formats, including Internet, point-of-sale, promotions, sponsorships, billboards, magazines, movies, television programs, and newspapers. Articles involving smoking were identified using “smoking” and “tobacco” as expanded subject headings; those with “cigarette” in the title or abstract were also included. Finally, articles involving African Americans were included if the article had “African American” as a keyword anywhere in the title, abstract, or subject heading, or if “blacks” was listed as a subject heading. To identify additional relevant studies, we searched the reference lists of all articles obtained and consulted experts in public health, media research, and health disparities.
Two investigators independently searched the titles, abstracts, and methods sections of the 131 studies identified with our search strategy. Manuscripts were selected for inclusion if they directly compared pro-tobacco media messages in African American and Caucasian markets. Manuscripts were excluded if they did not specifically compare the two racial groups.
One study was eliminated, for example, because although it described in depth the cigarette advertising in three magazines geared toward African Americans (Essence, Jet, and Ebony), it did not provide data from magazines targeted toward Caucasians.20 Articles were also excluded if they were opinion pieces, policy statements, or review articles. Articles were accepted only from peer-reviewed journals. Also, articles were selected only if they involved pro-tobacco messages, as anti-tobacco messages were not the focus of this study. When there was any discrepancy, consensus was easily achieved between the two researchers.
From each eligible manuscript identified, we extracted key information, including the year of data collection, the particular media studied, the geographic region of study, and a summary of the methods. We also extracted the number of total messages (e.g., billboards or magazine articles), the number of tobacco-related messages in that sample, and the number of residents living in each of the markets studied.
These data, which were collected separately in both African American and Caucasian markets, were then used to compute concentration and density. Concentration was defined as the number of tobacco messages divided by the number of total messages. Density was defined as the number of tobacco messages divided by the number of people residing in the market area, expressed as messages per 10,000 people. The extraction process was carried out by two investigators—working independently—who ultimately arrived at identical data.
Data were sufficient to pool two important outcomes available in multiple articles: concentration of tobacco-related billboards/signage in African American vs. Caucasian market areas and the density of tobacco-related billboards/signage in these areas. Data were imported into STATA version 9.0,21 which was used to derive pooled estimates of African American and Caucasian concentrations and their odds ratios, as well as African American and Caucasian densities and their rate ratios. The pooled estimates were calculated using a random effects model with inverse-variance weighting using the DerSimonian and Laird method.22 Prior to implementing the random-effects model, statistical heterogeneity between and within groups was measured using the Q statistic and assessed visually using the Galbraith plot of heterogeneity.23 Because the Q test was statistically significant (p<0.05), we used the random effects method to analyze the data.24
We assessed publication bias for both concentration and density using the Begg rank correlation method and the Egger weighted regression method.25 We also examined the cumulative effect on the pooled estimates by adding studies one at a time ordered by publication date.25 To evaluate the weight of particular studies on the pooled estimate, we performed influence analysis. This method recalculates the pooled prevalence estimate, omitting one study at a time.25 Meta-regression was also used to analyze the relationship between the pooled estimates and publication date.
We defined statistical significance a priori by calculating 95% confidence intervals (CIs) around all estimates. We log-tranformed density prior to conducting the meta-analyses and subsequently transformed density back to the original units for publication. This was done to conduct the meta-analyses using a closed-form expression of the variance.25
Of the 131 studies initially identified using the search strategy, 11 articles met all eligibility criteria. Seven of these studies investigated the medium of billboards and signage, including large point-of-sale advertisements and other placards,12,13,15–17,26,27 and four compared pro-tobacco advertising in magazines14,18,19,28 (Figure 1). Although we exhaustively searched for all types of pro-tobacco media and promotion, only studies involving billboards and magazines were represented in the findings.
Of the billboard/signage studies, six contained sufficient information to compute the concentration of tobacco advertisements (i.e., they measured both the number of tobacco advertisements and the number of total advertisements) and could be used to calculate a pooled estimate for this outcome12,13,15–17,26 (Table, Figure 2). The remaining study counted only tobacco advertisements, not total advertisements.27 The pooled concentration of tobacco advertisements was 22.0% (95% CI 17.9, 26.1) in African American markets and 15.5% (95% CI 8.8, 22.2) in Caucasian markets, with a pooled difference of 6.7% (95% CI 1.5, 11.8) higher in African American markets. The pooled odds ratio for a billboard being smoking-related in an African American area compared to a Caucasian area was 1.7 (95% CI 1.1, 2.6).
We detected no publication bias for either the African American or the Caucasian data on tobacco advertising concentration. Additionally, publication year had no effect on concentration for either race according to two methods (cumulative meta-analysis adding one study at a time ordered by publication date and meta-regression of year on concentration). Also, by removing one study at a time, the data showed that no particular study had undue influence on the pooled concentrations of tobacco-related billboards for either race.
A total of five billboard/signage studies reported sufficient data to compute the density of tobacco advertisements (how many tobacco advertisements were present per resident of the geographic study region)15–17,26,27 (Table, Figure 3). The pooled density of tobacco advertisements was 11.8/10,000 African American inhabitants (95% CI 5.0/10,000, 28.3/10,000) and 4.5/10,000 Caucasian inhabitants (95% CI 1.3/10,000, 15.2/10,000). The pooled rate ratio of African American to Caucasian tobacco-related billboard densities was 2.6 (95% CI 1.5, 4.7), indicating that there were 2.6 times as many tobacco advertisements per person in African American neighborhoods compared with Caucasian neighborhoods.
Again, we detected no publication bias for either the African American or the Caucasian density data. Influence plots showed that no one study had undue influence on the pooled density levels for either race area; however, Pucci reported lower densities for both races.27 Finally, meta-regression showed that publication year was not a significant predictor of density for either race (p=0.87 for African American densities and p=0.97 for Caucasian densities).
Although four studies compared magazine advertisements between African American and Caucasian populations (Figure 1), their methods and outcomes were too varied for their results to be combined with meta-analysis. Only one study compared the concentration of advertisements between African American and Caucasian publications: Cummings et al. found that cigarette advertisements comprised 12.0% of all advertisements in African American magazines, whereas cigarette advertisements comprised only 9.9% of Caucasian magazines (p=0.04).14
Three studies investigated the number of pro-tobacco messages per issue, analogous to the density of advertising. Basil et al. coded 1,171 magazine advertisements and found that African American publications contained about twice as many cigarette advertisements per issue as did Caucasian publications (statistical significance not reported).28 Landrine et al. found that Ebony had 2.25 cigarette ads per issue vs. People's 1.87 ads per issue (p=0.06).18 Finally, Pollay et al. found that Ebony and Life magazines had roughly the same density of cigarette advertisements during the years 1950–1965 (540 ads in Ebony vs. 526 in Life, p=1.0).19
This study comparing tobacco advertising in African American and Caucasian market areas demonstrated that, according to the available studies, tobacco signage was increased in African American markets in terms of both density and concentration. The odds that any given advertisement was smoking-related were 70% higher in African American areas vs. Caucasian areas, and there were 2.6 times as many tobacco advertisements per person in African American areas as compared with Caucasian areas.
The meta-analytic findings were consistent with relevant prior studies. For example, the two studies we identified that did not report sufficient data to be included in the density meta-analysis reported results consistent with our overall findings. Luke et al., who collected data on 1,239 billboards in St. Louis in early 1998, reported a statistically significant positive correlation between the percent of African American residents in a given geographic region and the proportion of tobacco billboards (r=0.15, p=0.004).12 Stoddard et al., who examined signage along the roadsides of African American and Caucasian neighborhoods in Los Angeles from 1990 to 1994, reported a tobacco advertisement density of 2.41/mile in African American neighborhoods compared with 0.46/mile in Caucasian neighborhoods (p<0.001).13
Our findings imply that African Americans may be special targets of the tobacco industry. Policy makers may wish to keep this disproportionate advertising in mind when designing future policies involving tobacco-related media. They would have good reason, for instance, to seek universally applicable limits on the concentration and/or density of tobacco advertising.
Our findings also suggest that this population may require disproportionate public health interventions to counter the effect of the disproportionate pro-tobacco promotion. Programs involving analysis of media messages—also known as media literacy programs—may be particularly effective in this population.29–31 Because this population is highly exposed to and familiar with media messages, techniques to analyze and evaluate these ubiquitous messages may be useful because of the direct relevance they provide. Finally, when addressing African Americans directly, medical practitioners (in both clinical and community settings) can emphasize the known disproportionate exposure of African Americans to pro-tobacco mass media. This knowledge of being targeted may motivate African Americans to refuse to fall prey to industry tactics, helping them to avoid smoking.
Research is necessary to determine more precisely the cause of this disproportionate advertising so that it can be effectively curtailed. Are tobacco marketers more aggressive when approaching minority neighborhoods? Are point-of-sale venues in African American neighborhoods more likely to seek out tobacco advertisements from the industry? Are other social, economic, or political factors involved? Researchers will also need to increase efforts to study this population's susceptibility to advertising, which will be of particular concern as we now know their exposure is disproportionately high. Future research should also continue to characterize specific techniques tobacco makers use to attract African Americans to co-opt those techniques for use in anti-tobacco social marketing programs.
Another important finding of this study was that, although we did search for data involving other media such as films, music, radio, Internet, promotions, or sponsorships, most published data in this area focused on billboards/signage and magazines. Billboards and magazines, however, are not currently the most important carriers of pro-tobacco media.32,33 The lessons these studies provide are still very relevant because the tobacco industry is remarkably adept at transitioning one type of advertising to another while still retaining its overall strategy.32,34 In the future, however, research involving the impact of tobacco advertising on African Americans should focus more on the forms of promotion currently and increasingly utilized by the tobacco industry. In 2003, for instance, 71.4% of tobacco industry expenditures went directly into such promotional activities as the distribution of free cigarettes.33 Future research should also address media such as smoking in movies, because it has now been established that as much as half of adolescent smoking initiation can be linked to watching smoking in movies.9 Do these forms of tobacco promotion target and affect African Americans in particular and/or more frequently? If so, in what particular way do they affect African Americans? These questions will be essential to address in the future to most effectively reduce health disparities related to tobacco.
This study had limitations worth noting. First, the only studies appropriate for meta-analysis involved billboards and signage. Although the tobacco industry is well-known to employ a consistent marketing strategy regardless of the specific medium used,32,34 it should be considered that studies involving other media may have different findings. However, this is an important finding of the study in itself, as it elucidates the need for future research investigating these same issues while focusing on emerging forms of promotion.
Second, the studies identified were relatively heterogeneous according to statistical analysis with a Q statistic. However, we responded appropriately to this issue by conducting the meta-analysis using a random effects model.23
Third, to assure a certain level of quality, we excluded unpublished studies from the analysis. Although this exclusion introduced potential publication bias, we did not detect publication bias using two statistical methods.
A final potential limitation involved the population denominators used to determine advertisement density. It should be considered that these population figures may not accurately reflect the market of a billboard because billboards are often positioned to be viewed by those who live outside of their vicinity. However, we felt these figures were the most appropriate approximations of the market population available.
This systematic review and meta-analysis demonstrates that African Americans are disproportionately exposed to pro-tobacco mass-media messages in terms of both concentration and density. These findings suggest that important policy and intervention techniques should be considered in this population to appropriately reduce tobacco-related health disparities. Also, more research will be required that focuses on other important forms of media, such as tobacco promotions and smoking in films, to fully understand the impact of pro-tobacco promotion in African American communities.
Dr. Primack was supported by a Physician Faculty Scholars Award from the Robert Wood Johnson Foundation and by a career development grant from the National Cancer Institute (K07-CA114315). Dr. Fine was supported in part by a career development award from the National Institute of Allergy and Infectious Diseases (K24-AI01769). No funding agencies were involved in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation, review, or approval of the article.