Journal of Clinical Medicine Article Pornography Use in Adolescents and Its Clinical Implications Josep M. Farré 1 , Angel L. Montejo 2 , Miquel Agulló 1 , Roser Granero 3,4 , Carlos Chiclana Actis 5,6,7 , Alejandro Villena 6 , Eudald Maideu 8 , Marta Sánchez 9 , Fernando Fernández-Aranda 4,10,11 , Susana Jiménez-Murcia 4,10,11 and Gemma Mestre-Bach 1,5, * 1 Department of Psychiatry Psychology and Psychosomatics, Dexeus University Hospital, 08028 Barcelona, Spain; psico.dex@quironsalud.es (J.M.F.); agullo.miquel@gmail.com (M.A.) 2 Ital Universitario Psychiatry Department, University of Salamanca Nursing School, 37007 Salamanca, Spain; amontejo@usal.es 3 Departament de Psicobiologia i Metodologia de les Ciències de la Salut, Universitat Autònoma de Barcelona, 08193 Barcelona, Spain; roser.granero@uab.cat 4 Ciber Fisiopatología Obesidad y Nutrición (CIBERObn), Instituto de Salud Carlos III, 28925 Madrid, Spain; ffernandez@bellvitgehospital.cat (F.F.-A.); sjimenez@bellvitgehospital.cat (S.J.-M.) 5 Facultad de Ciencias de la Salud, Universidad Internacional de La Rioja, 26006 La Rioja, Spain; carloschiclana@doctorcarloschiclana.com 6 Unidad de Sexología Clínica y Salud Sexual, Consulta Dr. Carlos Chiclana, 28003 Madrid, Spain; alejandrovillena@doctorcarloschiclana.com 7 Departamento de Psicología, Facultad de Medicina, Universidad CEU-San Pablo, 28925 Madrid, Spain 8 Hospital de Campdevanol, 17530 Girona, Spain; dr.emaideu@gmail.com 9 Consorcio de Eduacación de Barcelona, 08010 Barcelona, Spain; msanc21@xtec.cat 10 Department of Psychiatry, Bellvitge University Hospital-IDIBELL, 08907 Barcelona, Spain 11 Department of Clinical Sciences, School of Medicine, University of Barcelona, 08007 Barcelona, Spain * Correspondence: gemma.mestre.bach@gmail.com; Tel.: +93-4560111  Received: 30 September 2020; Accepted: 4 November 2020; Published: 11 November 2020  Abstract: (1) Background: The Differential Susceptibility to Media Effects Model (DSMM) suggests that pornography use effects are conditional and they depend on dispositional, developmental, and social differential susceptibility variables. This framework also highlights that the differential susceptibility variables act as predictors of pornography use and as moderators of the effect of pornography on criterion variables. (2) Methods: By administering a survey to n = 1500 adolescents, we tested whether these assumptions were met. (3) Results: Pornography use was related to being male and older, having a bisexual or undefined sexual orientation, higher substance use, being non-Muslim, and reporting sexual interest and the use of the media to obtain sexual information. Structural Equation Modeling (SEM) showed that higher levels in the criterion variables were directly related to pornography use, older age, substance use, and being women. Some mediational links also emerged. Pornography use mediated between the age and criterion variables. Moreover, substance use mediated the association between age and gender with the criterion variables. (4) Conclusions: Our findings support the clinical applicability of the theoretical DSMM framework. Knowing adolescent pornography consumers’ profiles and the impact of pornography on this population would allow for the designing of more effective prevention and regulation proposals. Keywords: sexuality; adolescents; pornography; porn; gender; risky behaviors J. Clin. Med. 2020, 9, 3625; doi:10.3390/jcm9113625 www.mdpi.com/journal/jcm J. Clin. Med. 2020, 9, 3625 2 of 19 1. Introduction The presence of sexually explicit materials has increased significantly both in mass media and social media [1,2]. Moreover, with the emergence of the Internet, the use of pornography has become widespread throughout the world [3,4]. In the case of adolescents and young adults, recent rates of pornography use have been reported to be around 43% [5]. This increase in consumption patterns may be partly explained by the “Triple A” theory, which highlights easy access to the Internet, the fact that a large part of the population can afford it, and the anonymity that the Internet guarantees to its consumers [6]. Numerous studies have focused on evaluating the use of pornography in this age group and its association with multiple variables. Some authors have tried to define possible profiles of adolescents and young people who consume pornography. For example, Efrati et al. [7] identified that those adolescents who used pornography were usually boys, low on social intimacy, introverted and neurotic, and more overt narcissists, among other factors. In this line, Brown et al. [8] identified three types of pornography users taking into account variables such as age, pornography acceptance, use, motivations for use and religiosity—porn abstainers, auto-erotic porn users and complex porn users. The Differential Susceptibility to Media Effects Model (DSMM) was designed by Valkenburg and Peter [9] and focuses on microlevel media effects. This model is based on multiple solid theories such as Social Cognitive Theory [10], the Neoassociationist Model [11], the Selective Exposure Theory [12], and the Media Practice Model [13]. The DSMM is structured around four central propositions: (1) Media effects are conditional and depend on dispositional, developmental, and social differential susceptibility variables. (2) Media effects are indirect and cognitive; emotional and excitative media response states mediate the relationship between media use and media effects. (3) The differential susceptibility variables act as predictors of media use and as moderators of the effect of media use on media response states. (4) Media effects are transactional; they influence media use, media response states, and differential susceptibility variables [9]. On the basis of the DSMM framework, Peter and Valkenburg [14] have published a review including studies that have evaluated pornography use in adolescents. In terms of dispositional predictors of pornography use, demographics, personality traits, norm-related variables, sexual interest, and Internet behavior have been explored [14]. It has been suggested that male adolescents are more exposed to pornography than females, although gender differences are smaller the more liberal their country of origin is [15–17]. Moreover, rule-breaking and adolescents who use substances may use pornography more frequently [18,19]; the same goes for adolescents with greater sexual interest [20]. Regarding developmental variables, age, pubertal maturation, and sexual experience have been studied in adolescents. There is controversy about whether pornography use increases with age, and existing studies reported conflicting results [15,16,18]. In studying possible trajectories of adolescent pornography use, however, it has been suggested that early puberty may be linked with earlier exposure to pornography and more frequent pornography use later [21]. The same applies to sexual experience, with some authors associating it with more frequent pornography use, while others associated it with a lower frequency [15,20]. Taking social variables into account, poor family functioning, desire for popularity, peer pressure, and victimization online and offline have been related with higher pornography use in adolescents [18,22–24]. In this vein, Nieh et al. [21] evaluated the influence of factors such as peer behaviors and parenting style on adolescent pornography use trajectories, finding that parental monitoring protected adolescents from pornography use. Relatedly, Efrati et al. [25] highlighted that the impact of loneliness on the frequency of pornography use may be dependent on individuals’ attachment orientations. In terms of victimization, the possible association between the use of pornography and violence and sexual aggression and coercion, as well as the problematic use of pornography, have been particularly studied [26–30]. Finally, with regard to criterion variables, pornography use has been related to more permissive sexual attitudes [31–33]. However, evidence for an association between pornography use and risky sexual behaviors, such as unprotected sex, is mixed [34,35]. J. Clin. Med. 2020, 9, 3625 3 of 19 Therefore, the existing evidence on how these multiple variables interact with each other is contradictory, and to the best of our knowledge, no study has yet evaluated all the variables proposed by the DSMM. Therefore, there is still a lack of systematic data on how the multiple variables of the DSMM model interact with each other. To this end, the present study aimed to assess in an integrated way the nuclear correlates of the use of pornography in adolescents suggested by the DSMM (dispositional, developmental, social and criterion variables). For this purpose, we tested two of the four DSMM propositions: (1) we explored whether dispositional, developmental, and social variables predict pornography use; (2) we evaluated whether dispositional, developmental, and social variables may not only predict pornography use but also moderate the extent to which pornography use predicts criterion variables. We hypothesized that the explored DSMM proposals will be fulfilled. 2. Experimental Section 2.1. Participants and Procedure An e-mail was sent to all the public and private high schools in Catalonia (Spain) that appeared on the list provided by the Catalan government. Special education centers were excluded. Of all the high schools, excluding those that did not answer or refused to participate, 14 schools were finally included, with a total of n = 1500 adolescent students (14–18 years old). It was the principals or boards of education who gave permission to participate in the present study. The 14 high schools belonged to different geographical areas of Catalonia and included participants of different socioeconomic statuses to ensure that the results were representative. The evaluation was carried out during the same academic year. Once the high schools showed interest, our research team went in person to explain the details of the research, resolve doubts, and specify the procedure. All the students from the same high school were evaluated on the same day by a member of the research team, together with a teacher from the high school. Besides supervising the administration of the paper-and-pencil self-administered survey, our research team addressed the potential doubts of the students. There was no financial reward. However, at the end of the sample collection, our research team returned to each high school to explain, to the boards of education, the main results of the research. It is not possible to calculate the refusal rate because some centers chose not to provide us with this information, but we estimate that it was less than 2%. 2.2. Assessment The survey contained 102 items assessing dispositional, developmental, social, criterion, and media use variables. The items included have not been evaluated for their psychometric properties. Due to practical issues of time and adolescent fatigue, we decided to design items to evaluate the variables of interest instead of using validated psychometric instruments, which are much more extensive. 2.2.1. Dispositional Variables Dispositional variables included: sociodemographic, norm-related, and sexual interest variables—Internet behavior variables. The sociodemographic variables assessed in the survey were gender and sexual orientation. Drug use and religion were evaluated in the category of norm-related features. Frequency of drug use was coded into one of four categories: non-consumption, once a month or less, between twice a month and once a week, and more than once a week. 2.2.2. Developmental Variables Developmental variables included age and sexual experience. Sexual experience assessed aspects such as the age of their first sexual experience and the current frequency of sexual intercourse. J. Clin. Med. 2020, 9, 3625 4 of 19 2.2.3. Social Variables Social variables contained family-related factors and victimization. Family-related factors included items related to the adolescent’s nuclear family and the possible presence of siblings. The victimization section evaluated sexual assault, malpractice during sexting, and online victimization. 2.2.4. Criterion Variables Criterion variables assessed the following domains: risky sexual behaviors (such as unprotected sex, and sex after alcohol and substance use), and permissive sexual attitudes (such as infidelity). 2.2.5. Media Use Survey items measured pornography use and related sexual behaviors, sexting, and cybersex behaviors with responses coded dichotomously as “yes/no”. 2.3. Statistical Analysis Statistical analysis was carried out with Stata16 for Windows [36]. A logistic regression fitted predictive models of the pornography media use. Different logistic models were fitted for each of the variables defined as dependent variables (downloading sexual content, use of social nets to send sex content, participation in sexual chats and use of erotic lines). The set of the potential predictors included all the other variables analyzed for this work (dispositional variables (sex, sexual orientation, drug use/abuse, brought up following a religion, religious practitioner, feeling religious, interest in social nets for obtaining sexual content), developmental variables (age, age at first sexual experience and frequency of sexual experiences), and social variables (persons living at home, being abused and being forced to share sex content)). A stepwise method was used to build a final model in which the choice and selection of the significant predictors is carried out by an automatic procedure, adding or removing in subsequent steps the predictors according to pre-specified parameters. This method is particularly useful in studies with a large set of potential independent variables and no underlying empirical hypothesis on which to base the model selection. For categorical independent variables, different contrasts were defined: pairwise comparisons for nonordered variables and polynomial contrasts for ordered variables (polynomial post-hoc tests are particularly useful to determine whether a particular mathematical pattern emerges for the levels of the predictor, such as linear, quadratic, cubic or quartic levels) [37]. Adequate goodness of fit for the final models was considered for nonsignificant results (p > 0.05) in the Hosmer-Lemeshow test. Nagelkerke’s R-squared coefficient (N-R2 ) estimated the global predictive capacity, considering null for N-R2 < 0.02, low-poor for N-R2 > 0.02, mild-moderate for N-R2 > 0.13, and high-good for N-R2 > 0.26 [38]. The area under the receiver operating characteristic (ROC) curve (AUC) measured the discriminative capacity (AUC < 0.65 was interpreted as low-poor, AUC > 0.65 mild-moderate, and AUC > 0.70 high-good [39]). Path analysis was used to describe the underlying mechanisms explaining the pornography use based on the set of variables registered in this work. Path analysis procedures represent a straightforward extension of multiple regression modeling, which allows estimating the magnitude and significance level of associations into a set of variables, including mediational links [40]. This procedure can be used for both exploratory and confirmatory modeling, and therefore it allows theory testing and theory development [41,42]. In this work, and due to the existence of multiple criteria measures, we defined a latent variable defined by the observed indicators contraception, unprotected sex, emergency contraception, practicing sex after alcohol use/abuse, practicing sex after drugs use/abuse and infidelity (the latent variable in this study allowed us to simplify the data structure and therefore facilitated a more parsimonious fitting) [43]. In this study, path analysis was adjusted through Structural Equation Modeling (SEM), using the maximum-likelihood estimation for the parameter estimation, and valuing the goodness of fit through the standard statistical measures: the root mean square error of approximation (RMSEA), Bentler’s Comparative Fit Index (CFI), the Tucker-Lewis J. Clin. Med. 2020, 9, 3625 5 of 19 Index (TLI), and the standardized root mean square residual (SRMR). An adequate fit was considered for models meting the next criteria Barret [44]: RMSEA < 0.08, TLI > 0.90, CFI > 0.90, and SRMR < 0.10. The global predictive capacity of the model was measured by the coefficient of determination (CD), whose interpretation is similar to global R2 in multivariate regression models. 2.4. Ethics The Hospital Ethics Committee (Comité Ético de Investigación Clínica del Grupo Hospitalario Quiron) approved the procedures of this study (REF: 012/107) in December 2014. The present study was carried out in accordance with the latest version of the Declaration of Helsinki. We obtained a permit from the management boards of each school that agreed to participate in our study. Each school provided the parents or legal guardians of underage students with information about the study. Those parents or minors who did not wish to participate informed the school board. It was clarified that participation was voluntary and they could withdraw at any time. The data of n = 1 student were withdrawn from the study after the request of the school board. 3. Results 3.1. Characteristics of the Sample Table 1 includes the distribution for the variables analyzed in the study. Most individuals reported heterosexual orientation (90.5%), while 2.1% indicated that they were homosexual, 3.9% bisexual, and 3.6% not defined. The percentage of individuals brought up Catholic was 36.1%, Muslim 4.9%, and other religions 5.3% (the remaining 53.8% indicated that they were atheist). Only 10.7% described themselves as a religious practitioner, with 17.0% being religious or very religious. Around 20% of the sample reported substance use or abuse. The percentage of adolescents who reported sexual interest and the use of the media to obtain sexual information was 25.6%. The proportion of individuals with sexual experience was around 33%, with 15–16 years old being the most likely age of sexual initiation. The prevalence of adolescents who indicated being victims of sexual abuse was 6.5%, while 17.6% indicated that they had been forced to share sexual content. Regarding media use, 43.6% reported pornography use. Other related behaviors showed lower percentages (between 6.1% for use of erotic telephone lines and 9.5% for downloading sexual content). The criterion variables were distributed as follows: 31.0% used contraception, 17.3% reported unprotected sex, and 8.7% used emergency contraception; sexual behavior after alcohol use was reported by 29.9% of the participants, while sex after substance use was reported by 11.7%. The percentage of adolescents who reported being unfaithful was 15.7%. 3.2. Predictive Models of Pornography Use Table 2 contains the results of the logistic regression, selecting the best predictors of pornography use in the study. This model achieved adequate fitting (p = 0.385 in the Hosmer–Lemeshow test), large predictive capacity (N-R2 = 0.32), and large discriminative capacity (AUC = 0.79). Increases in the odds of pornography use were related to being male, older, bisexual or with undefined sexual orientation, higher substance use, and reporting sexual interest and the use of the media to obtain sexual information; in addition, being Muslim (compared to being atheist) decreased the likelihood of pornography use. J. Clin. Med. 2020, 9, 3625 6 of 19 Table 1. Descriptive variables of the study (n = 1500). Dispositional Variables n % Social Variables n % Female 833 55.5% Both parents 1424 94.9% Gender Living with . . . Male 667 44.5% Other family 40 2.7% Heterosexual 1357 90.5% Other 36 2.4% Sexual orientation Homosexual 31 2.1% 0 Siblings 350 23.3% Bisexual 3.9% 1 sibling 818 54.5% Living with . . . Not defined 54 3.6% 2 siblings 229 15.3% Never 1212 80.8% 3 or more siblings 103 6.9% Substance use/abuse 1×/month or less 126 8.4% No 1403 93.5% Sexual abuse 2×/month or 1×/week 75 5.0% Yes 97 6.5% 3× or more/month, or weekly 87 5.8% No 1236 82.4% Forced to share sexual content Atheist 807 53.8% Yes 264 17.6% Brought up in a religion Catholic 541 36.1% Pornography media use Muslim 73 4.9% Pornography use Yes 654 43.6% Other 79 5.3% Social media to send sex content Yes 112 7.5% No 1339 89.3% Social media to send self-sex. Yes 98 6.5% Religious practitioner Yes 161 10.7% Participation in sexual chats Yes 93 6.2% None 936 62.4% Use of erotic telephone lines Yes 91 6.1% A little religious 308 20.5% Downloading sexual content Yes 143 9.5% Religious feeling Religious 218 14.5% Criterion variables Very religious 38 2.5% Use of contraception Yes 465 31.0% Sexual interest/Internet No 1116 74.4% Unprotected sex Yes 260 17.3% behavior Yes 384 25.6% Use of emergency contraception Yes 130 8.7% Developmental variables Sex after alcohol Yes 449 29.9% 14 years 117 7.8% Sex after other substances Yes 176 11.7% 15 years 340 22.7% Infidelity Yes 236 15.7% Age 16 years 360 24.0% Infidelity: caresses Yes 145 9.7% 17 years 454 30.3% Infidelity: kisses Yes 41 2.7% 18 years 229 15.3% Infidelity: embraces Yes 221 14.7% J. Clin. Med. 2020, 9, 3625 7 of 19 Table 1. Cont. Dispositional Variables n % Social Variables n % Age (years old) Mean—SD 16.23 1.18 Infidelity: oral sex Yes 58 3.9% Never 1008 67.2% Infidelity: masturbation Yes 143 9.5% Under 13 20 1.3% Infidelity: penetration Yes 44 2.9% First sexual experience at 13–14 years 130 8.7% 15–16 years 284 18.9% 17–18 years 58 3.9% Never 1012 67.5% Only 1 time 64 4.3% Frequency of sexual 2–5 times 99 6.6% experience 6–10 times 55 3.7% More than 10 times 270 18.0% SD: standard deviation. J. Clin. Med. 2020, 9, 3625 8 of 19 Table 2. Predictive models of pornography use: stepwise logistic regression (n = 1500). Criterion: Pornography Use. Fitting Indexes: H-L(p-Value) = 0.385; N-R2 = 318; AUC = 0.790 (95%CI: 0.767 to 0.813) Predictor Contrast B SE p OR 95%CI Sex Male vs. Female 2.122 0.128 <0.001 8.349 6.503 10.720 Sexual orientation 0.019 Homosex. vs. Heterosex. 0.221 0.425 0.603 1.247 0.543 2.867 Ambisex. vs. Heterosex. 0.702 0.307 0.022 2.019 1.107 3.682 Non-defined vs. Heterosex. 0.738 0.324 0.023 2.092 1.108 3.949 Drugs use/abuse 0.003 Linear trend 0.413 0.192 0.032 1.511 1.036 2.202 Quadratic trend −0.364 0.214 0.088 0.695 0.457 1.056 Cubic trend −0.113 0.233 0.627 0.893 0.566 1.410 Brought up in religion 0.064 Catholic vs. Atheist 0.028 0.131 0.832 1.028 0.796 1.328 Muslim vs. Atheist −0.771 0.300 0.010 0.463 0.257 0.833 Other vs. Atheist −0.159 0.274 0.562 0.853 0.498 1.460 Sexual interest/Internet Yes vs. No 0.747 0.139 <0.001 2.112 1.608 2.773 behavior Age (years-old) 0.252 0.053 <0.001 1.287 1.159 1.429 H-L: Hosmer–Lemeshow. N-R2 : Nagelkerke’s pseudo-R2 coefficient. AUC: Area under the ROC curve. B: Logistic parameter. SE: Standard error. OR: odds ratio. J. Clin. Med. 2020, 9, 3625 9 of 19 J. Clin. Med. 2020, 9, x FOR PEER REVIEW 9 of 17 Tableuse/abuse Drugs 3 contains the results of the logistic models obtained 0.019 for the other predictors of the pornography use and cybersex behaviors Linear trend analyzed 0.566 in this 0.280work. Downloading 0.043 1.762 sexual1.018content was 3.050 most probable for males, those withtrend Quadratic a bisexual orientation, −0.356 those reporting 0.302 0.238 sexual 0.701 interest 0.388and the use 1.265 of social networks to obtainCubic trend information 0.147 sex and regarding 0.332 earlier0.659 1.158experiences. first sexual 0.604 The 2.218 use Age (years-old) of social media to send sexual content was more −0.234 likely for0.101 males,0.021 those who 0.791 0.649those use drugs, 0.965 with Frequency sexual sexual interest who use social media to obtain information about sex,0.001 and those who had been sexually experiences abused by adults or other adolescents. The use of social media to send sexual content to others was Linear trend −0.215 0.342 0.530 0.807 0.413 1.576 related to bisexual orientation, sexual interest and the use of social media to obtain sexual information, Quadratic trend −0.713 0.315 0.024 0.490 0.264 0.909 earlier first sexual experiences, being Cubic trenda victim of 0.784 sexual abuse, 0.530and being 0.139 forced to share0.774 2.190 sexual content. 6.194 The odds of participation in sexualtrend Quartic chats was higher 0.708 for males, 0.445 those with sexual 0.112 2.031 interest, 0.849 those who 4.860 use social H-L: media to obtain sexual Hosmer‒Lemeshow. N-Rinformation 2: Nagelkerke’sand pseudo-R those who have beenAUC: 2 coefficient. forcedArea to share under sexual content. the ROC Finally, the use of erotic telephone lines was higher for curve. B: Logistic parameter. SE: Standard error. OR: odds ratio.men, participants with higher substance use, younger respondents, and those with a higher frequency of sexual experiences. 3.3. Path Analysis 3.3. Path Analysis Figure 1 includes the path diagram with the standardized coefficients obtained in the SEM, in whichFigure only1significant includes the path diagram parameters were with the (only retained standardized coefficients relationships obtained inlevels with significance the SEM, in p < 0.05 which only significant are plotted). Figure 1parameters were retained uses conventional rules(only relationships for path diagramswith andsignificance SEM schemes; p