Racial differences in cigarette brand recognition and impact on youth smoking

Paper-and-pencil surveys were administered to 6-8th graders in the Vallejo school district’s four middle schools in the spring of 2006. Two follow-up surveys were conducted 12 months and 24 months after initial baseline for 7-9th graders and 8-10th graders, respectively. All students were eligible to participate in the first two years, therefore the 2007 survey functioned as both a follow-up for the 2006 participants and a baseline survey for new participants. Only students who had completed at least one baseline survey (either in 2006 or 2007) were eligible to participate in the 2008 follow-up and no new participants were recruited.

Surveys were conducted in courses that were required by all students, but these courses varied by school. Active parental consent and student assent were obtained using a protocol approved by Stanford University’s Administrative Panel on Human Subjects. In the first year, students were offered a $2 incentive to return a parental consent form, regardless of positive or negative permission. Of the 3,482 eligible students in the district, 48% of students returned a consent form. Although the incentive was increased to $5 in the second year, the return rate was similar (50%).

This paper reports analyses for both a cross-sectional sample and a longitudinal cohort (retention rate: 68%). The cross-sectional analysis (n = 2,589) combines all participants with valid baseline data for either 2006 (n = 1,647) or 2007 (n = 942). The longitudinal analysis includes participants who were never smokers at their respective baseline and reported valid data for smoking status one year later. Only 12-month follow-up data (2007) were used for those students who participated in all three surveys. Of the 2,589 students who completed a baseline survey, 1,905 reported having never tried smoking, and 1,179 of those students reported their smoking behavior at follow-up. Of the 726 never smokers that were lost to 12-month follow-up, 39% moved out of the district, 56% did not return a parental consent form for the follow-up survey, and 5% were absent on the survey day. This includes 141 students who had a 24-month follow-up but did not have a 12-month follow-up. Attrition analyses used chi-square tests for categorical variables and independent sample t-tests for numeric variables to compare all baseline covariates for the longitudinal sample (n = 1,179) and the never smokers who were lost to follow up (n = 726).

Tobacco retailers in the study community were identified by state licensing records, and trained coders completed observations of cigarette advertising in a census of stores (n = 102, response rate = 94%) in 2006. Following procedures described in our previous studies [14, 28], advertisements were counted and categorized by flavor (menthol, non-menthol, or both), by brand (Camel, Marlboro, Newport, or other), and by location (exterior, interior). For each store, we computed menthol share of voice, defined as the proportion of all cigarette advertisements in a store that featured any menthol brands.

The primary outcome for the cross-sectional analysis was brand recognition. To develop a measure of brand recognition, digital photographs of retail tobacco advertisements from the study community were altered. All brand names were removed but slogans and pack imagery were retained. Students saw the three cigarette advertisements in a fixed order: Newport, Camel, then Marlboro. These brands were selected because of their popularity among youth [29]. The advertisements showed the menthol varieties of Newport and Camel and the non-menthol variety of Marlboro. To assess advertising awareness students were first asked if they had ever seen the advertisement. They were then asked to write the name of the brand as free text. Recognition was assessed by coding these responses and misspellings were credited if they contained at least one correct syllable or approximated the phonetic pronunciation.

The primary outcome for the longitudinal analysis was progression from never to ever smoking. Smoking behavior was assessed at baseline and follow-up using the item “Have you ever tried smoking a cigarette, even one or two puffs?” We did not examine current smoking because the incidence of smoking (in the previous 30 days) was 5%.

Baseline characteristics that could confound the relationships between race, brand recognition, and smoking initiation have been described elsewhere [16]. In brief, students answered three questions about how many times they visited convenience, small market, and liquor stores each week, and these responses were combined as an overall shopping frequency measure (sum of visits per week for the three store types). Other risk factors for smoking were self-reported grades in school, recoded into 0.0 to 4.0 grade-point average (GPA), unsupervised days per week after school, and risk-taking propensity [30]. Social influences to smoke were assessed by asking about how many smokers live at home and how many of four best friends smoke; both were recoded into any versus none. Socio-demographic items included gender, race, ethnicity, and grade level. Race was coded to compare African-American students versus all other races and ethnicity was coded to compare Hispanic students to non-Hispanic students regardless of race. All analyses included an indicator for whether the baseline data came from 2006 or 2007.

Tobacco retailers contained an average of 3.6 exterior ads (SD = 5.1) and 22.2 interior ads (SD = 25.9) for cigarettes. On average, the proportion of cigarette ads that featured a menthol brand (menthol share of voice) was 36.9% (SD = 18.3), which was higher than the average for the state (25.7% (SD = 26.1) [12]. The proportion of stores with any exterior advertisements for the three brands examined in the school-based surveys was 5.9% for Camel menthol, 38.2% for Marlboro non-menthol, and 33.3% for Newport. The proportion of stores with any interior ad for these brands was 22.5% for Camel menthol, 93.1% for Marlboro non-menthol, and 55.9% for Newport.

Table 1 describes the cross-sectional sample and longitudinal cohort. The racial/ethnic distribution of the sample reflects the student population in the school district at the time of data collection [31]. At baseline, the prevalence of ever smoking for the entire sample was 26%, which was higher than prevalence estimates for California 7th graders in statewide school-based surveys [27]. Among African-American students, the prevalence of ever smoking was 28% and among other students it was 25%, which was not significantly different (p = 0.08). On average, students reported visiting stores almost four times per week (SD = 4.0) and African-American students reported more frequent visits compared to non-African-Americans (p < 0.01). Smoking at home also differed by race and ethnicity. The prevalence of at least one smoker at home was higher among African-American students than among others (46.8% vs. 40.6%, p < 0.01); the prevalence of home smoking was lower among Hispanic students than among non-Hispanics (37.5% vs 45.2%, p < 0.01).

At baseline, the proportion of students who reported having seen the ads was 81% for Camel, 65% for Marlboro, and 66% for Newport. Fewer students could provide the brand name: 52% for Camel, 36% for Marlboro, and 32% for Newport. As shown in Figure 1, a significantly greater proportion of African Americans recognized the Newport brand than other students. Conversely, a significantly smaller proportion of African American students recognized Marlboro than other students. After adjusting for shopping frequency, other risk factors for smoking, and socio-demographics, the association of race and brand recognition persisted. The odds of recognizing the Newport brand was three times higher for African-American students than other students (OR = 3.03, 95% CI = 2.45, 3.74, p < 0.01) while African-American students were significantly less likely than others to recognize the Marlboro brand (OR = 0.60, 95% CI = 0.48, 0.73, p < 0.01). There were no significant racial differences in recognition of the Camel brand. Hispanic students were less likely than others to recognize Newport (OR = 0.73, 95% CI = 0.58, 0.90, p < 0.01). Other significant predictors of recognition for all three brands were living with a smoker and risk-taking propensity. In addition, older students and those who had ever smoked were more likely to recognize Newport and Camel (data not shown).

Never smokers who were lost to follow-up were more likely to be boys (50.3% vs 40.5%; p < 0.01), Hispanic (36.6% vs 26.9%; p < 0.01), younger (grade level 7.2 vs 7.4; p < 0.01), and to report lower grades (GPA 2.5 vs 2.7; p < 0.01) and more store visits (3.9 vs 3.3; p < 0.01). African American youth were not more likely than other races to be lost to follow-up. No differences were observed for the other covariates, including household smoking, peer smoking, unsupervised days after school, or risk-taking propensity. In addition, never smokers who were lost to follow-up did not differ from the analysis sample on brand recognition measures (Camel p = 0.40, Marlboro p = 0.22, Newport p = 0.20).

The incidence of smoking initiation at follow-up was 17% and a greater proportion of African-American students initiated smoking than other students (22% vs 15%; p < 0.01). In an unadjusted HGLM, recognition of Newport predicted smoking initiation (p < 0.01) but neither Camel nor Marlboro recognition had a significant relationship with smoking initiation (p = 0.10 and p = 0.18, respectively; data not shown).

Table 2 presents the odds ratios and confidence intervals from three HGLMs predicting smoking initiation. Each model has a brand-specific predictor for recognition and is adjusted for all variables listed in the table. The odds of smoking initiation increased by 49% for students who recognized the Newport brand at baseline. Brand recognition for Camel or Marlboro did not predict smoking initiation at follow-up after adjusting for shopping frequency and other risk factors. Other significant risk factors for smoking initiation were race, ethnicity, risk-taking propensity, and exposure to household and peer smoking. Being African American was a significant risk factor for smoking initiation in two of the adjusted models; however, Newport brand recognition explained some of the variance attributed to that racial difference. Tests of an interaction examined whether the impact of brand recognition on smoking initiation was greater for African-American students but it was not statistically significant. In all three adjusted models, being Hispanic was also a risk factor for smoking initiation, but tests of an interaction for ethnicity with brand recognition were not statistically significant.

An ancillary analysis imputed missing data for the subset of baseline never smokers (n = 141) who had follow-up data at 24 months but not at 12 months. When self-reported smoking at 24-month follow-up was substituted for missing data at 12-month follow-up, the association of Newport brand recognition with smoking initiation persisted (data not shown).