INTRODUCTION
Tobacco use remains a major global public health problem, causing approximately 8 million deaths each year and disproportionately affecting low- and middle-income countries, including Morocco1,2. Tobacco use often begins during adolescence, a developmental period marked by heightened sensitivity to social influence and experimentation with risk behaviors3,4. Early initiation is associated with a greater likelihood of nicotine dependence and continued use into adulthood; therefore, identifying factors that shape adolescents’ susceptibility to tobacco products remains a priority for prevention and tobacco-control research3-6.
Tobacco marketing is one such factor. International evidence indicates that exposure to tobacco advertising, promotion, and sponsorship is associated with tobacco initiation and progression among young people. Marketing contact occurs through multiple channels, including point-of-sale (POS) displays, branded items, direct promotional offers, entertainment media depictions, and increasingly digital environments5-10.
Importantly, these indicators do not all measure the same construct: portrayals of tobacco use in television, videos, or movies capture exposure to tobacco imagery rather than direct advertising7, whereas owning tobacco-branded items or being offered free products reflects promotional engagement and direct marketing contact. Likewise, willingness to use or wear branded items may reflect receptivity to marketing rather than exposure alone11,12. Considering these measures jointly as tobacco marketing exposure and engagement may therefore better reflect adolescents’ real-world marketing contact than single-channel indicators7,11,12.
To the best of our knowledge, evidence remains limited on how different forms of tobacco marketing exposure and engagement cluster among Moroccan adolescents. Existing studies have commonly assessed exposure at the level of individual channels using binary measures or basic scoring indices, which may overlook the complexity of real-world marketing contact, where adolescents experience overlapping patterns of exposure, receptivity, and engagement across channels and in varying combinations7,13,14. A person-centered approach, such as latent class analysis (LCA), can help identify subgroups of adolescents with similar response patterns across multiple indicators15-17, thereby characterizing distinct marketing exposure and engagement profiles and examining how these profiles relate to cigarette use, other smoked tobacco use, and smokeless tobacco use13-15,18.
Accordingly, this study aimed to identify latent classes of tobacco marketing exposure and engagement among Moroccan adolescents using data from the 2016 Morocco Global Youth Tobacco Survey (GYTS) and to examine sociodemographic differences across classes. We further assessed associations between latent class membership and current use of cigarettes, other smoked tobacco products, and smokeless tobacco after adjusting for potential confounders. Finally, the study identifies multidimensional patterns of tobacco marketing exposure and engagement, including media portrayals, POS exposure, receptivity, and promotional engagement, using LCA, and examines how these patterns are associated with tobacco use among adolescents.
METHODS
Data source
This was a secondary dataset analysis of the 2016 Morocco GYTS, a nationally representative, school-based survey conducted in Morocco among school-going adolescents. The present analysis included students with complete data on the latent class indicators, resulting in an analytic sample of 3614 participants. The GYTS is conducted within the Global Tobacco Surveillance System (GTSS)19 and applies a standardized protocol for questionnaire design, sampling, and data collection to generate comparable estimates of tobacco-related indicators among school-going adolescents. The survey uses a two-stage cluster sampling design (schools, then classes) and an anonymous, self-administered questionnaire. Although the GYTS primarily targets students aged 13–15 years, the class-based sampling approach may include some respondents outside this age range; therefore, age distribution is reported accordingly19. Although a more recent GYTS was conducted in Morocco in 2023, the corresponding individual-level microdata were not publicly available at the time of this study; therefore, the 2016 GYTS remains the most recent publicly accessible individual-level dataset suitable for this analysis.
Study variables and measurement
Indicator variables
Five GYTS items were used as latent class indicators of tobacco marketing exposure and engagement. The first item assessed exposure to tobacco use in media: ‘During the past 30 days, did you see any people using tobacco on TV, in videos, or in movies?’ with responses coded as 1 (Yes) and 2 (No). The second item assessed POS exposure: ‘During the past 30 days, did you see any advertisements or promotions for tobacco products at points of sale, such as stores, shops, or kiosks?’ with responses coded as 1 (Yes) and 2 (No). The third item assessed receptivity to branded items: ‘Would you ever use or wear something that has a tobacco company or tobacco product name or picture on it, such as a lighter, t-shirt, hat, or sunglasses?’ with responses coded as 1 (Yes), 2 (Maybe), and 3 (No). The fourth item assessed ownership of tobacco-branded items: ‘Do you have something, for example a t-shirt, pen, or backpack, with a tobacco product brand logo on it?’ with responses coded as 1 (Yes) and 2 (No). The fifth item assessed direct promotional contact: ‘Has a person working for a tobacco company ever offered you a free tobacco product?’ with responses coded as 1 (Yes) and 2 (No).
Sociodemographic and tobacco use variables
The analysis included sociodemographic and socioeconomic covariates as well as tobacco use outcomes. Covariates comprised sex, age group, area of residence, parents’ employment, father’s and mother’s education levels, father’s and mother’s jobs, parental tobacco smoking status, and weekly spending money. The three tobacco use outcomes were past-30-day cigarette use, past-30-day use of smoked tobacco products other than cigarettes, and past-30-day use of smokeless tobacco products, each coded as a binary variable. Detailed coding and categories of these variables are provided in Supplementary file Table S1.
Statistical analysis
Descriptive statistics were used to summarize the analytic sample. All variables summarized in the descriptive analysis were categorical, including sociodemographic characteristics and tobacco use outcomes, and were presented as frequencies and percentages. Denominators varied because of item-level non-response.
LCA was used to identify distinct profiles of tobacco marketing exposure and engagement based on adolescents’ responses to the five indicator items. LCA classifies individuals into mutually exclusive latent classes that summarize common response patterns across multiple indicators. Models with increasing numbers of classes, from two to seven, were estimated. Model selection was informed by a combination of fit indices, including AIC, BIC, CAIC, and ABIC, as well as entropy, goodness-of-fit statistics (G2 and Pearson χ²), class sizes, and interpretability. Given the large sample size, greater weight was placed on BIC, CAIC, ABIC, and classification quality, as assessed by entropy, when choosing the final solution. To reduce the risk of local maxima, models were re-estimated, and the solution with the highest log-likelihood was retained. The LCA output provided posterior class-membership probabilities and a most-likely class assignment.
Differences in sociodemographic characteristics and tobacco use across latent classes were assessed using chi-squared tests. Associations between latent class membership and each tobacco use outcome – cigarette use, smoked tobacco products other than cigarettes, and smokeless tobacco use – were examined using binary logistic regression. For multivariable models, candidate covariates were identified from bivariate analyses; variables with p≤0.20 were retained for adjustment. Area was forced into the models regardless of its bivariate p-value to improve comparability with prior literature and to control for potential contextual confounding. To address potential overadjustment, we reported two models: Model A, adjusted for sociodemographic and socioeconomic covariates, and Model B, additionally adjusted for peer influence, a psychosocial factor that may lie on the causal pathway between marketing exposure and engagement and tobacco use. All analyses were performed using Jamovi version 2.3.28. Statistical significance was set at p<0.05.
Ethics
This study relied on data from the GYTS, a national survey conducted by Morocco’s Ministry of Health. As the anonymized dataset is publicly available for academic or research use, no additional ethical review was required19.
RESULTS
Descriptive results
Of the 3915 participants in the 2016 Morocco GYTS, 3614 were included after excluding respondents with missing data on at least one latent class indicator. The sample was nearly balanced by sex (51.18% female) and predominantly urban (82.16%). Although the GYTS targets adolescents aged 13–15 years, the class-based sampling approach may include some respondents outside this age range; accordingly, 13.09% of the analytic sample were aged ≥16 years. Additional sociodemographic variables are detailed in Table 1.
Table 1
Sociodemographic characteristics and tobacco use among a national sample of Moroccan school-going adolescents, Global Youth Tobacco Survey 2016 (N=3614)
Selection of model and latent class description
Model-fit statistics for latent class solutions with two to seven classes are presented in Table 2. Competing models were compared using information criteria (AIC, BIC, CAIC, ABIC), entropy, goodness-of-fit statistics (G2 and Pearson χ²), class sizes, and interpretability. As shown in Table 2, BIC (16379), CAIC (16399), and ABIC (16316) reached their minimum values for the three-class solution, which also yielded the highest entropy (0.81), indicating the best classification quality. Although AIC was marginally lower for the 4- and 6-class solutions, these models showed substantially lower entropy and less clearly separated class profiles. Considering parsimony, interpretability, and classification performance, the three-class model was retained as the most appropriate representation of tobacco marketing exposure and engagement patterns in this sample.
Table 2
Model fit statistics for two- to seven-class latent class solutions among a national sample of Moroccan school-going adolescents, Global Youth Tobacco Survey 2016 (N=3614)
| Number of classes | LL | AIC | BIC | CAIC | ABIC | Entropy | G2* |
|---|---|---|---|---|---|---|---|
| 2 | -8190 | 16406 | 16486 | 16499 | 16445 | 0.65 | 227.70 |
| 3 | -8108 | 16255 | 16379 | 16399 | 16316 | 0.81 | 63.25 |
| 4 | -8099 | 16253 | 16420 | 16447 | 16334 | 0.44 | 46.90 |
| 5 | -8095 | 16258 | 16468 | 16502 | 16360 | 0.49 | 37.82 |
| 6 | -8086 | 16253 | 16507 | 16548 | 16377 | 0.56 | 19.17 |
| 7 | -8083 | 16261 | 16559 | 16607 | 16406 | 0.53 | 13.30 |
The distribution of probabilities of latent classes (LC) in Table 3 highlighted a substantial predominance of LC1 (66.60%), followed by LC2 (24.71%), while LC3 (8.69%) was the least represented in the sample.
Table 3
Item-response probabilities conditional on latent class membership for the three-class model of tobacco marketing exposure and engagement among a national sample of Moroccan school-going adolescents, Global Youth Tobacco Survey 2016 (N=3614)
LC1 includes adolescents with moderate exposure to tobacco depictions in TV/videos/movies [probability, Pr(Yes)=0.61] and near-zero POS exposure [Pr(Yes)=0.01], with consistently low probabilities for promotional engagement indicators [willingness: Pr(Yes)=0.02; ownership: Pr(Yes)=0.02; free offer: Pr(Yes)=0.02]. This profile reflects a predominantly media-based exposure pattern with minimal retail or promotional contact. Based on this response pattern, we labeled this group as ‘media depiction exposure’.
LC2 includes adolescents with high exposure to tobacco depictions in TV/videos/movies (Pr(Yes)=0.84) and very high exposure to tobacco marketing at POS [Pr(Yes)=0.98], while probabilities of promotional engagement remained low [willingness: Pr(Yes)=0.02; ownership: Pr(Yes)=0.04; free offer: Pr(Yes)=0.04]. This pattern reflects primarily passive exposure through media and retail environments with limited direct promotional contact. We therefore labeled this class ‘media and POS exposure’.
LC3 includes adolescents with higher probabilities across multiple indicators, including exposure to tobacco depictions in TV/videos/movies [Pr(Yes)=0.71], exposure at POS [Pr(Yes)=0.45], and notably higher promotional engagement [willingness: Pr(Yes)=0.53; ownership: Pr(Yes)=0.67; free offer: Pr(Yes)=0.31]. This pattern indicates multi-channel marketing contact combined with active engagement with tobacco branding and direct promotional contact. Accordingly, this class was labeled ‘multimodal exposure and engagement’.
Sociodemographic and tobacco use disparities across latent classes
Latent classes differed significantly in several sociodemographic characteristics and tobacco use outcomes (Table 4). Male adolescents were more represented in the multimodal exposure and engagement class (LC3: 63.78%) than in the media and POS exposure class (LC2: 54.71%) and the media depiction exposure class (LC1: 44.69%; p<0.001). Adolescents aged ≥16 years were also more frequent in LC3 (20.38%) than in LC1 (12.88%) and LC2 (11.11%; p=0.001). Parental tobacco smoking was more common in LC3, particularly having one smoking parent (26.37%) or both parents smoking (7.53%), compared with LC1 (15.63% and 0.97%) and LC2 (17.65% and 2.00%; p<0.001). Weekly spending money above 100 Moroccan dirhams (approximately US$11) was also highest in LC3 (10.58%) compared with LC1 (4.96%) and LC2 (7.31%; p<0.001). Tobacco use was consistently more prevalent in LC3: cigarette use was 10.76% in LC3 versus 1.27% in LC1 and 1.93% in LC2; other smoked tobacco use was 9.03% versus 2.11% and 1.82%; and smokeless tobacco use was 10.03% versus 2.08% and 3.08%, respectively (all p<0.001).
Table 4
Sociodemographic characteristics and tobacco use by latent class membership, based on chi-squared tests, among a national sample of Moroccan school-going adolescents, Global Youth Tobacco Survey 2016 (N=3614)
Association between tobacco use and latent class membership
Latent class membership was associated with tobacco use outcomes in adjusted analyses (Table 5). Using media depiction exposure (LC1) as the reference, adolescents in the multimodal exposure and engagement class (LC3) had higher odds of use across all products in Model A: cigarette use (AOR=6.56; 95% CI: 3.49–12.35), other smoked tobacco use (AOR=3.61; 95% CI: 1.98–6.58), and smokeless tobacco use (AOR=3.37; 95% CI: 1.89–6.00). After additionally adjusting for peer influence in Model B, these associations were attenuated but remained statistically significant (cigarette use: AOR=3.70; 95% CI: 1.86–7.38; other smoked tobacco use: AOR=2.14; 95% CI: 1.11–4.13; smokeless tobacco use: AOR=1.95; 95% CI: 1.04–3.67), suggesting that peer-related susceptibility may partially account for the observed differences between classes. In contrast, the media and POS exposure class (LC2) did not differ significantly from LC1 across outcomes in either model.
Table 5
Adjusted associations between latent class membership and tobacco use outcomes using binary logistic regression, Model A and Model B, among a national sample of Moroccan school-going adolescents, Global Youth Tobacco Survey 2016 (N=3614)Research Paper|Research Paper|
| Characteristics | Cigarette use | Smoked tobacco other than cigarettes | Smokeless tobacco | |||
|---|---|---|---|---|---|---|
| Model A | Model B | Model A | Model B | Model A | Model B | |
| AOR (95% CI) | AOR (95% CI) | AOR (95% CI) | AOR (95% CI) | AOR (95% CI) | AOR (95% CI) | |
| Sex (N=3605) | ||||||
| Female (ref.) | 1 | 1 | 1 | 1 | 1 | 1 |
| Male | 2.37 (1.28–4.40)** | 2.58 (1.33–4.99)** | 1.92 (1.12–3.29)* | 1.95 (1.11–3.41)* | 1.89 (1.15–3.11)* | 1.97 (1.18–3.30)** |
| Residence (N=3588) | ||||||
| Rural (ref.) | 1 | 1 | 1 | 1 | 1 | 1 |
| Urban | 1.52 (0.69–3.33) | 1.48 (0.64–3.43) | 1.63 (0.83–3.21) | 1.74 (0.86–3.53) | 1.15 (0.61–2.17) | 1.25 (0.64–2.45) |
| Age (years) (N=3605) | ||||||
| ≤13 (ref.) | 1 | 1 | 1 | 1 | 1 | 1 |
| 14–15 | 2.47 (1.09–5.59)* | 1.93 (0.82–4.51) | 2.40 (1.16–4.93)* | 2.39 (1.11–5.18)* | 1.40 (0.79–2.48) | 1.27 (0.70–2.29) |
| ≥16 | 7.04 (2.90–17.10)*** | 4.52 (1.78–11.52)** | 4.15(1.87–9.23)*** | 3.38 (1.43–8.00)** | 2.24 (1.13–4.46)* | 1.90 (0.92–3.92) |
| Father’s education level (N=2995) | ||||||
| No formal schooling (ref.) | 1 | 1 | 1 | 1 | 1 | 1 |
| Primary school | 0.99 (0.45–2.20) | 0.86 (0.36–2.06) | 1.51 (0.74–3.06) | 1.47 (0.70–3.07) | 0.56 (0.27–1.14) | 0.54 (0.26–1.13) |
| Secondary school | 0.49 (0.19–1.27) | 0.39 (0.14–1.10) | 0.84 (0.36–1.97) | 0.78 (0.32–1.91) | 0.59 (0.27–1.28) | 0.56 (0.25–1.25) |
| Post-secondary | 0.81 (0.29–2.24) | 0.61 (0.20–1.80) | 0.65 (0.24–1.81) | 0.58 (0.20–1.67) | 0.35 (0.14–0.88)* | 0.30 (0.12–0.79)* |
| Mother’s education level (N=3201) | ||||||
| No formal schooling (ref.) | 1 | 1 | 1 | 1 | 1 | 1 |
| Primary school | 1.25 (0.57–2.78) | 1.04 (0.44–2.48) | 1.09 (0.57–2.11) | 0.89 (0.45–1.77) | 1.85 (0.93–3.69) | 1.73 (0.84–3.55) |
| Secondary school | 1.81 (0.80–4.11) | 1.46 (0.59–3.58) | 0.88 (0.41–1.87) | 0.60 (0.27–1.36) | 1.59 (0.74–3.43) | 1.44 (0.64–3.21) |
| Post-secondary | 0.78 (0.25–2.48) | 0.71 (0.21–2.36) | 0.58 (0.20–1.72) | 0.51 (0.17–1.56) | 2.01 (0.78–5.18) | 2.08 (0.79–5.47) |
| Weekly spending money (Dirhams) (N=3598) | ||||||
| No spending money (ref.) | 1 | 1 | 1 | 1 | 1 | 1 |
| ≤100 | 2.30 (0.79–6.68) | 2.01 (0.67–6.04) | 1.30 (0.60–2.83) | 1.09 (0.49–2.41) | 1.07 (0.55–2.10) | 0.93 (0.47–1.86) |
| >100 | 6.01 (1.72–21.07)** | 6.44 (1.71–24.21)** | 4.62 (1.72–12.40)** | 4.34 (1.55–12.17)** | 2.16 (0.85–5.48) | 1.95 (0.75–5.08) |
| Parental tobacco smoking (N=3420) | ||||||
| None (ref.) | 1 | 1 | 1 | 1 | 1 | 1 |
| Both | 3.30 (0.86–12.69) | 1.99 (0.48–8.34) | 1.42 (0.31–6.53) | 0.91 (0.19–4.35) | 3.97 (1.40–11.24)** | 2.73 (0.90–8.27) |
| One | 1.75 (0.97–3.16) | 1.41 (0.73–2.70) | 1.26 (0.71–2.23) | 0.98 (0.52–1.83) | 1.48 (0.86–2.54) | 1.29 (0.73–2.28) |
| Peer influence to use tobacco if offered by best friend (N=3600) | ||||||
| Definitely not (ref.) | NC | 1 | NC | 1 | NC | 1 |
| Probably not | NC | 6.60 (2.95–14.76)*** | NC | 3.97 (1.73–9.11)** | NC | 2.77 (1.26–6.08)* |
| Probably yes | NC | 15.95 (7.45–34.13)*** | NC | 12.28 (6.03–24.98)*** | NC | 7.57 (3.76–15.20)*** |
| Definitely yes | NC | 23.68 (8.06–69.58)*** | NC | 17.03 (6.18–46.92)*** | NC | 18.86 (7.25–49.09)*** |
| Latent class membership (N=3614) | ||||||
| Media depiction exposure (LC1) (ref.) | 1 | 1 | 1 | 1 | 1 | 1 |
| Media and POS exposure (LC2) | 1.44 (0.72–2.87) | 1.11 (0.53–2.34) | 0.93 (0.49–1.77) | 0.81 (0.41–1.58) | 1.27 (0.72–2.23) | 1.04 (0.58–1.88) |
| Multimodal exposure and engagement (LC3) | 6.56 (3.49–12.35)*** | 3.70 (1.86–7.38)*** | 3.61 (1.98–6.58)*** | 2.14 (1.11–4.13)* | 3.37 (1.89–6.00)*** | 1.95 (1.04–3.67)* |
DISCUSSION
To the best of our knowledge, this is the first study in Morocco to apply LCA to characterize profiles of tobacco marketing exposure and engagement among adolescents using GYTS data. Three key findings emerged. First, we identified three distinct profiles that reflect heterogeneous patterns of marketing-related contact: media depiction exposure, media and POS exposure, and multimodal exposure and engagement. Second, class membership differed across sociodemographic and family-context characteristics. Third, and most importantly, adolescents in the multimodal exposure and engagement class showed consistently higher odds of tobacco use across product types compared with the reference class, even after adjustment for sociodemographic and socioeconomic covariates.
This study demonstrated a significant association between the multimodal exposure and engagement class and tobacco use across products, with adolescents in this class showing consistently higher odds of use than those in the reference class, even after adjustment for sociodemographic and family-context covariates. These results are consistent with prior cross-sectional and longitudinal studies5,9,11,12,20-23. In addition, a meta-analysis by Wellman et al.6 reported that exposure to pro-tobacco marketing and media was associated with higher odds of pro-tobacco attitudes among youth (OR=1.51) and more than doubled the odds of tobacco use initiation (OR=2.23). Importantly, associations attenuated after adding peer influence, suggesting that peerrelated susceptibility may partially account for the observed differences between profiles, in line with previous evidence on the role of peer influence in adolescent smoking. However, given the cross-sectional design, this should be interpreted as potential explanatory overlap rather than mediation24-26.
In response to this issue, Morocco enacted Law 15-91 in 1995, which prohibits tobacco advertising and promotion across multiple channels, including broadcast media and films, print media, and promotional displays at or around tobacco-selling locations, and restricts promotional activities linked to sports and public events27-30. However, despite this legal framework, enforcement has been reported as inconsistent, and tobacco marketing can still reach adolescents through POS promotions, brand stretching via merchandise, and indirect or digital promotion (e.g. content in television/movies and social media)2,18,31-34. In light of the observed association between the multimodal exposure and engagement class and tobacco use, continued monitoring of adolescents’ exposure to tobacco marketing remains relevant, particularly at retail points of sale and through indirect promotional practices. However, because of the cross-sectional design, this study does not evaluate the effectiveness of enforcement measures or specific tobacco-control policies. Therefore, these implications should be interpreted cautiously and considered in light of the WHO FCTC Article 13 guidance2,28,35,36.
Strengths and limitations
The study has some strengths. It drew on a large, nationally representative, school-based dataset, supporting population-level inference among Moroccan school-going adolescents. In addition, LCA provided a person-centered approach to capture co-occurring patterns of tobacco marketing exposure and engagement across multiple channels rather than relying on single-item measures. The standardized GYTS methodology further supports cross-country comparability and population-level inference.
The study has several limitations. First, the measures relied on self-reported responses, which may be affected by recall or reporting bias, including potentially different recall of marketing-related contact among tobacco users versus non-users. Second, the cross-sectional design precludes establishing temporality, so findings should be interpreted as associations rather than causal effects. Third, this analysis used the 2016 GYTS dataset. Although this was the most recent publicly available individual-level GYTS dataset for Morocco at the time of the study, tobacco marketing strategies and adolescents’ media environments have evolved since then, particularly with the expansion of digital and social media marketing. Consequently, the identified exposure and engagement profiles may not fully reflect current marketing practices or current adolescent exposure patterns. Finally, the indicator set does not fully capture contemporary digital marketing tactics and platform-specific advertising formats; future studies should incorporate more recent individual-level data, more detailed measures of online exposure and engagement, and, where feasible, longitudinal designs.
CONCLUSIONS
Three exposure and engagement classes were identified among Moroccan adolescents. Membership in the multimodal exposure and engagement class was associated with higher odds of cigarette, other smoked tobacco, and smokeless tobacco use. These associations weakened but remained significant after accounting for peer influence, suggesting that peer-related factors may partly overlap with marketing exposure and engagement. These findings suggest that continued attention to adolescents’ exposure to tobacco marketing remains relevant, particularly at points of sale and through indirect promotional practices. Given the cross-sectional design, these results should be interpreted as associations rather than evidence of the effectiveness of specific interventions or tobacco-control policies. Future prevention efforts may benefit from addressing media literacy, peer influence, and refusal skills within school-based programs.
