Study Design

The data utilized in this cross-sectional study originate from the fourth edition of the National School-Based Health Survey (PeNSE) conducted in 2019 in Brazil. PeNSE is an initiative supported by the Ministry of Education (MEC) included in the Brazilian Surveillance of Risky and Protective Factors for Chronic Diseases. The first edition of PeNSE was carried out in 2009 through a collaboration between the Ministry of Health and the Brazilian Institute of Geography and Statistics (IBGE). All data are anonymized and publicly accessible. PeNSE 2019 adopted a comprehensive sample design covering the entire national territory, including students aged 13 to 17 years from both public and private schools in urban and rural areas across the country. The data collection occurred within the schools, where students directly responded to an electronic questionnaire made available on a personal digital assistant device without the presence of a researcher. Further specific details regarding the sample design can be found in official sources (Instituto Brasileiro de Geografia y Estatística, 2021).

Study Variables

Considering that both the relationship between violence victimization and the consumption of ultra-processed foods, as well as the mediation through psychoactive substances, are complex and, in some cases, bidirectional, the initial framework was described in a theoretical causal diagram built based on previous knowledge available in the scientific literature (Supplemental material, Fig. S1). We used the online tool DAGitty (Textor et al., 2016) to construct a directed acyclic graph (DAG) (Tennant et al., 2021). In addition to providing transparency to the process of selecting relevant main variables and covariables for this study, this method allows the identification of the minimum sufficient adjustment set, which includes all the covariables described above.

Dependent Variable

Ultra-processed food consumption: To identify foods as ultra-processed, the NOVA classification system was used (Monteiro et al., 2018). In the PeNSE 2019, adolescents were administered a questionnaire that inquired about the types of food they had consumed in the past 24 h. From the responses, the following 13 food items were identified as UPF: soft drink, industrialized fruit juice, powdered soft drink, chocolate drink, flavored yogurt, salty snacks (e.g., packaged chips or crackers), sweet snacks (e.g., sweet cookie, cream cookie, or packet cake), industrialized desserts (e.g., chocolate, ice cream, gelatin, flan), processed meat products (e.g., sausage, mortadella or ham), industrialized bread (e.g., flatbread, hot dog bun or hamburger bun), margarine, industrialized sauces (e.g., mayonnaise, ketchup), and industrialized ready meals (e.g., instant noodles, packaged soups, frozen lasagna). The number of UPFs consumed in the previous 24 h ranged from 0 to 13 and was considered the dependent variable in the present analyses. For descriptive purposes, the number of UPFs consumed was divided according to the distribution in the sample by quartiles, generating the following categories: 1st quartile (0 to 2 UPFs), 2nd quartile (3 or 4 UPFs), 3rd quartile (5 or 6 UPFs) and 4th quartile (7 to 13 UPFs).

Independent Variable

Violence victimization: The questions about violence were distributed in different sections of the PeNSE 2019. Bullying victimization was located in the section "Situations at home and at school" and asked, "In the last 30 days, how many times have any of your schoolmates teased, mocked, taunted, intimidated, or made fun of you to the point that you felt hurt, bothered, upset, offended, or humiliated?". The response options were "none in the last 30 days", "once", or "2 or more times". Additionally, in the section "Safety", two other questions inquired about the following violent experiences: "In the last 12 months, how many times were you physically assaulted by your mother, father, or guardian?" and "In the last 12 months, how many times were you physically assaulted by someone other than your mother, father, or guardian?". The response options for both questions were "none in the last 12 months", "once", "2 to 5 times", and "6 or more times". For the present analysis, the last two response options were combined to ensure an adequate sample size and simplify result interpretation.

Mediators

Use of psychoactive substances: Adolescents reported their use of three psychoactive substances, namely, alcohol, tobacco, and illicit drugs, in response to two questions for each substance. Regarding alcohol consumption, the first question asked, "Have you ever had a glass or a drink of alcohol in your life?". If the answer was affirmative, they were asked "In the last 30 days, how many days have you had at least one drink or one dose of alcoholic beverage?". The response options included "no day in the last 30 days", "1 or 2 days", "3 to 5 days", "6 to 9 days", "10 to 19 days", "20 to 29 days" or "all days". Based on the responses to both questions, the adolescents were classified into the following three categories: "never (has drunk)", "previously consumed, but not in the last month", and "consumed in the last month". Similarly, data on tobacco smoking and illicit drug use were collected through two sequential questions. For tobacco smoking, the adolescents were first asked, "Have you ever smoked a cigarette, even one or two puffs in your life?", and then "In the last 30 days, how many days did you smoke cigarettes?". Concerning illicit drug use, the first question was "Have you ever used a drug such as marijuana (cannabis), cocaine, crack, solvent-based glue, general ether-based inhalants, popper, ecstasy, oxy, MD, skank, and others in your life?", and the second was "In the last 30 days, how many days did you use any drug?". Using the same method described for classifying adolescents according to alcoholic beverage consumption, the responses to tobacco smoking and illicit drug use questions also generated three categories, i.e., "never consumed", "previously consumed, but not in the last month", and "consumed in the last month".

Covariates

Information on the following covariates was considered: sex (male, female); age (16 to 17 vs. 13 to 15 years); daily consumption of fruits (yes vs. no); daily consumption of other vegetables (yes vs. no); number of close friends (none, 1 or 2, 3 or more); sedentary behavior (0 to 2, > 2 to 5, > 5 h/day); and total physical activity (≤ 1, > 1 to 6, > 6 h/week). The socioeconomic condition was evaluated according to the combination of maternal schooling, possession of material goods (e.g., cell phone, computer, car), and availability of services (e.g., internet, maid) in the household. Principal component analysis was applied to this set of items, and a weighted score was calculated considering the specific load of each item. Finally, the socioeconomic level score was categorized into quartiles. Adolescents were also asked to report the following mental health symptoms that occurred in the last 30 days: a) "How often have you felt very concerned about the ordinary things in your daily life such as school activities, sports competitions, homework, etc.? "; b) "How often have you felt irritated, nervous or bad-tempered by anything?", c) "How often have you felt that no one cares about you?"; d) "How often have you felt sad?", and e) "How often have you felt that life is not worth living?".

Statistical Analysis

First, the absolute (n) and relative frequency (%) of the categorical variables were calculated for the total number of participants. Then, generalized linear models were used to analyze the association between the number of UPFs consumed (dependent variable, continuous, ranging from 0 to 13 UPFs) and the three types of violence victimization (independent variables recoded in three categories ["none", "once", "two or more times"]: bullying victimization; physical aggression from parents and from others). Initially, unadjusted models were used to estimate the ß-coefficient and the related 95% confidence interval (CI) for experiencing victimization “once” and “two or more times” compared to not experiencing it (reference category) during the time frame considered for each type of violence (Model 1). Next, the other three models were progressively adjusted as follows: Model 2: adjusted for sex, age group, socioeconomic status, number of close friends, daily fruit consumption, daily consumption of other vegetables, sedentary behavior and total physical activity; Model 3: model 2 adjusted for alcohol consumption, tobacco smoking, and illicit drug use; and Model 4: model 3 adjusted for the number of mental health symptoms.

To explore the potential mediating effect of psychoactive substance use on the relationship between violence victimization and UPF consumption, simple mediation models were employed. Nine models, three for each type of violence, were constructed. For instance, in the association between bullying and UPF consumption, separate analyses were conducted to examine mediation through alcohol, tobacco, and illicit drugs. Next, all three possible mediators were examined in the associations between physical aggression by parents and by others and the number of UPF consumed.

All statistical operations were conducted with STATA software version 15.0 (Stata Corporation, College Station, TX, USA) and considered the parameters of a complex survey design (svy commands in Stata).