Selection of anti-tobacco PSAs
PSA selection and study design were described previously (Strasser et al. 2009). Briefly, 569 cigarette smoking PSAs were acquired from several state and national health authorities. Three trained raters viewed each PSA for content and identified a subset of 99 PSAs that (1) promoted seeking smoking cessation treatment or portrayed the negative consequences of continuing to smoke, (2) targeted adults and (3) were 30 s in duration. These PSAs were rated for MSV features using a scoring template (visual range = 0–10, audio range = 0–5 and content range = 0–5) based on work by Morgan et al. (2003).
To classify PSAs by AS, trained raters viewed the PSAs to generate a single statement reflecting the central argument (or arguments) of each PSA (e.g. ‘If the health harms of smoking are not enough to get you to quit, consider quitting for your children and those you love’, and ‘Although you may think smoking helps you cope, if you don't soon quit you will eventually die’). Next, we conducted a shopping mall intercept survey of 300 current smokers to collect ratings of the transcribed central arguments from which an overall AS score was created for each PSA by taking the mean of the 36–38 individual scores recorded for each PSA (Zhao et al. 2011).
Four groups of PSAs were then created from the existing collection of 99 PSAs: (1) high MSV–high AS, (2) high MSV–low AS, (3) low MSV–high AS and (4) low MSV–low AS. PSAs exceeding 1 SD from the mean on each of the two dimensions were selected for use in the present study; 16 PSAs met this criterion (four in each group).
Smokers responding to recruitment flyers and advertisements participated in an initial telephone contact at which eligibility was determined. A total of 199 eligible individuals completed a single 90-min session. After giving informed consent, participants provided an exhaled breath carbon monoxide sample (Vitalograph, Lenexa, KS, USA) for biochemical verification of smoking status and a saliva sample for genotyping. Of these 199 participants, 122 self-identified as European American [EA, selected to reduce potential bias because of population stratification (Palmatier et al. 1999; Petryshen et al. 2010)]. Of these, genotype data (BDNF rs6265 and COMT rs4680) were collected from 120 participants. Genotypes were classified as Val/Val vs. */Met for both BDNF rs6265 and COMT rs4680 based on previous research showing significant cognitive differences between Val homozygotes and Met carriers for both genes (Colzato et al. 2010; Hariri et al. 2003; Loughead et al. 2009; Schofield et al. 2009).
Standard questionnaires (Lerman et al. 1997) were administered at the beginning of the session to assess demographics, smoking history and nicotine dependence (Fagerström test for nicotine dependence, which includes current smoking rate; Heatherton et al. 1991). Four PSAs were then presented through a 17-in computer monitor using MediaLab Research Software (Empirisoft, New York, NY, USA) with the participant seated in a comfortable chair approximately 1 m away. Each participant viewed one of the four sets of PSAs classified by MSV and AS. After viewing the PSAs, participants completed measures of cognitive processing, narrative processing and sensory processing of the PSAs (Andrews et al. 1990; Chaudhuri & Buck 1995; Palmgreen et al. 2002); affective response to the PSAs (Batra & Holbrook 1990; Chaudhuri & Buck 1995); perceived effectiveness of the PSA; recognition of content (Everett & Palmgreen 1995) and intentions to quit smoking (Fishbein et al. 2001; Norman et al. 1999; Yzer et al. 2003). Items measuring processing were accompanied by response scales ranging from 1 (not at all) to 7 (very much), and assessed thinking about the central message of the PSA (cognitive processing; five items, e.g. ‘Overall, how much did the PSA make you think about arguments for quitting smoking?’), the story told by the PSA (narrative processing; three items, e g., ‘Overall, how much did you pay attention to the characters in the PSA?’) and the sensory qualities of the PSA (sensory processing; three items, e g., ‘Overall, how much did you pay attention to the PSA's sound tracks?’). Perceived effectiveness was measured by four sets of seven items (one set for each PSA) accompanied by response scales running from 1 (strongly disagree) to 5 (strongly agree). Intentions to quit were measured using two items accompanied by four-point response scales. Affective response was measured using four sets of 12 items (one set for each PSA) assessing emotional response to the PSA (e.g. ‘Did the PSA make you feel sad?’) accompanied by response scales ranging from 1 (not at all) to 7 (extremely). For each construct described above (cognitive processing, narrative processing, sensory processing, perceived effectiveness, intentions to quit and affective response), participant responses were averaged across all items measuring that construct to generate a summary score, which was used for analysis. Recognition was assessed using five items; for purposes of analysis, a dichotomous recognition variable was used (answered all five items correctly vs. answered at least one item incorrectly).
Descriptive statistics were obtained for all variables. One-way anovas (analysis of variances) were used to test for differences across genotype groups in demographics, smoking history, quitting intentions, processing, perceived PSA effectiveness and recognition. Linear regression models of quitting intentions, perceived effectiveness, message processing as well as affective response and a logistic regression model of the dichotomized recognition measure were then performed. The predictors were age, nicotine dependence (continuous), education (college graduate = 1, non-college graduate = 0), MSV (low = 0, high = 1), AS (low = 0, high = 1) and the dichotomous versions of BDNF Val66Met (Val/Val = 0 and */Met = 1) and COMT Val158Met (*/Met = 0 and Val/Val = 1). The two-way interactions of the genotype variables with MSV and AS were also included. All predictors were entered as a block, after which nonsignificant (P > 0.05) interaction terms were allowed to drop out.
We investigated the possibility that the association of BDNF genotype with quitting intentions and perceived effectiveness was mediated by central processing using the method originally proposed by Baron and Kenny (1986) and applied to tobacco prevention research by MacKinnon et al. (2002). Using this method, mediation is shown when (1) the predictor is significantly associated with both the outcome and the mediator and (2) in a regression of outcome on predictor and mediator, the mediator is significantly associated with the outcome (which should reduce, or render nonsignificant, the effect of the predictor). To accomplish this, regression models of quitting intentions and perceived effectiveness were performed using BDNF genotype and central processing as predictors.