Problematic Internet Use as a Predictor of Eating Disorders in Students: A Systematic Review and Meta-Analysis Study

Problematic Internet use (PIU) has begun to be linked to the development of certain eating disorders. This uncontrolled use of the Internet is mainly found in the student population. The purposes of this paper were to determine PIU-related eating disorders in students from a systematic review of the literature and to analyze the incidence of PIU in eating disorders through a meta-analysis of the literature. We used two electronic databases (Web of Science and Scopus) from inception to June 2019. The systematic literature review was based on fixed inclusion and exclusion criteria. A total of 12 studies were identified (systematic review) and 10 studies for meta-analysis, which included 16,520 students. Different eating disorders were associated with PIU: anorexia nervosa, bulimia nervosa, binge-eating disorder, food preoccupation, loss of control eating, and dieting. Furthermore, meta-analysis confirmed that PIU is a predictor of eating disorders in students. The groups of students with PIU presented a higher rate in the presence of eating disorders, these differences being significant. Finally, this study showed empirical evidence on the link between PIU and eating disorders. The need for prevention in childhood and adolescence is highlighted.


Introduction
Uncontrolled use of technology is a common practice among younger people. The Internet has revolutionized ways of interacting with the world. This fact together with the portability of electronic devices has enabled the proliferation of cases of problematic Internet use (PIU) or Internet addiction.
The PIU has been catalogued within the so-called behavioral addictions [1]. This kind of addiction directly affects the user's behavior, where there is a loss of self-control and an increase in impulsivity. These factors often cause the user to connect unconsciously to the Internet.
On the other hand, the development of smartphones and social networks have considerably increased the level of internet addiction among university students [2][3][4]. This casuistry mainly affects this population group, increasing the anxiety, depression, and stress levels of students [5,6]. Thus, Internet addiction is a cause of disease in the population [7]. This problem has a global perspective, and mainly affects developed countries.
In childhood it also entails different problems such as a decrease in attention to learning [8], and problematic behaviors such as loss of self-control [9]. It is also a factor related to some eating disorders [10]. In this regard, some years ago several works began to include Internet addiction and eating disorders among problematic behaviors in childhood [11]. Therefore, prevention in the child population is fundamental to preventing the development of this problem in adolescence and adulthood.
As for the topic of Internet addiction, previous works of systematic review have been elaborated with meta-analysis that have analyzed different factors linked to the PIU. Liu, Nie, and Wang [28], analyzed 58 papers on Internet addiction in the East Asian region to assess the effects of group counseling programs, cognitive-behavioral therapy, and sports intervention on Internet addiction. Tokunaga [29], reviewed 247 documents to determine correlations between Internet habits and loneliness and depression. Wang et al. [30], analyzed 15 studies to determine the association between Attention Deficit/Hyperactivity Disorder (ADHD) and Internet addiction. Cheng et al. [31], reviewed 25 papers to determine the association between PIU and suicide. Fumero et al. [32], analyzed 28 documents to establish psychological, personal, and social factors related to PIU in adolescents. Lei et al. [33], analyzed 76 documents to determine the relationship between social support and Internet addiction in China. Lei et al. [34], reviewed 77 papers to analyze whether different coping styles correlate with increased Internet addiction. Li et al. [35], included 70 papers in their review on the prevalence of PIU and its associated factors in Chinese university students. Li, Lei, and Tian [36], reviewed 79 studies to explore the relationship between positive and negative indicators of parenting style and Internet addiction. Shao et al. [37], analyzed 26 documents to estimate the prevalence of Internet addiction among university students in China. Su et al. [38], reviewed 101 papers to see if there were significant gender differences in the prevalence of PIU. Lei, Chiu and Li [39], analyzed 70 papers to determine whether the PIU was related to subjective well-being, life satisfaction, positive emotions, or negative emotions.
Although PIU or Internet addiction has begun to link with eating disorders, no previous systematic review study with meta-analysis has addressed these two variables together. In addition, the population at risk of problematic use of the Internet are students [40]. At the same time, eating disorders are more frequent in the student population [41]. All this has repercussions on the incidence of one variable on another. A cause-effect relationship or feedback is generated between these two factors.

Methods
It was proposed as objectives, (i) to determine PIU-related eating disorders in students from the systematic review of the literature and (ii) to analyze the incidence of PIU in eating disorders through meta-analysis of the literature. Based on them, the research questions were: RQ1. How many studies were published over the years? RQ2. Who are the most active authors in the area? RQ3. In which sources appear this kind of studies? RQ4. What is the educational stage and the country of the students? RQ5. What are the eating disorders related to PIU? RQ6. Is there a significant difference in the presence of eating disorders between the control population and those with PIU? For this purpose, a systematic review methodology with meta-analysis was used [29,42,43]. In the methodological process, the quality standards of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) have been applied [44].
The analysis variables were established according to the systematic review methodology [45], and answered the research questions posed. These were: years of publication, authors, source title, characteristics of the students, and eating disorders. On the other hand, the meta-analysis calculated the size of the effect of the set of studies with respect to three assumptions: (i) subjects with PIU have a higher rate of eating disorders; (ii) subjects with non-PIU have a higher rate of eating disorders; (iii) there is no difference between the PIU and non-PIU group in the presence of eating disorders.

Search Strategy
The search was established from the application of different descriptors that defined the topics of the review: ("Problematic Internet use" OR "Internet addiction") AND ("Eating disorder" OR "Disordered eating" OR "Feeding disorders"). The Web of Science (WOS) and Scopus databases were selected because they have internationally recognized impact indices: Journal Citation Reports (JCR) and Scimago Journal & Country Rank (SJR) [46].
The descriptors were applied in the search engine of both databases to proceed to subsequent filtering. To this end, a series of inclusion and exclusion criteria were established [47].
The inclusion criteria were: To avoid bias in study selection, two investigators undertook the systematic review using the same descriptors and inclusion and exclusion criteria. The degree of agreement on the inclusion of articles was 98%. The disagreement was addressed by a third researcher who chose to include 100% of the extracted scientific literature.

Data Collection and Analysis
The literature screening process was carried out in four phases following the standards of the PRISMA declaration [44] (Figure 1). The first phase "Identification" consisted of the initial search of the literature in the WOS (n = 21) and Scopus (n = 79) databases, and in the identification of other documents from the references cited in the articles. All references were then extracted and duplicate citations deleted. In the second phase, 'Screening', the inclusion (IC1, IC2, IC3) and exclusion (EX1, EX2, EX3) criteria were applied. In the third phase, "Eligibility", each title and summary of the articles was analyzed in detail on the basis of the inclusion (IC4, IC5) and exclusion (EX4, EX5) criteria. Finally, in the fourth phase, "Included", the study sample was included for systematic review (n = 12) and meta-analysis (n = 10). Data analysis from the systematic review was performed on the basis of extraction of key information for each analysis variable. This was done by means of a detailed reading of the documents. On the other hand, in the meta-analysis configuration, we included the articles that contained the data referring to the mean and standard deviation of the comparison between a group with PIU and another non-PIU group regarding the presence of eating disorders. Articles that collected dichotomous data regarding the number of eating disorder events for each group (PIU and non-PIU) were also included in the meta-analysis. Only two articles were excluded from metaanalysis because they did not collect this type of data in the text.
Meta-analysis data were analyzed using Review Manager software (Cochrane, London, UK) version 5.3. The confidence interval was set at 95%. The search was conducted on 5 June 2019. All articles published to date were reviewed.

Systematic Review
The studies analyzed were grouped according to the year of publication (  Data analysis from the systematic review was performed on the basis of extraction of key information for each analysis variable. This was done by means of a detailed reading of the documents. On the other hand, in the meta-analysis configuration, we included the articles that contained the data referring to the mean and standard deviation of the comparison between a group with PIU and another non-PIU group regarding the presence of eating disorders. Articles that collected dichotomous data regarding the number of eating disorder events for each group (PIU and non-PIU) were also included in the meta-analysis. Only two articles were excluded from meta-analysis because they did not collect this type of data in the text.
Meta-analysis data were analyzed using Review Manager software (Cochrane, London, UK) version 5.3. The confidence interval was set at 95%. The search was conducted on 5 June 2019. All articles published to date were reviewed.

Systematic Review
The studies analyzed were grouped according to the year of publication ( On the other hand, most of the articles are written by multiple authors. Only one of them is of unique authorship [58]. The topic of eating disorders in students has been addressed primarily by different authors who submit only one manuscript (Table 1). Only four authors appear in two separate articles.

References
Journal h-Index [52] Adicciones 20 [50] Archives of Medicine 7 [49] Cyberpsychology, Behavior, and Social Networking 54 [48,51,58,59] Eating and Weight Disorders 23 [55] International Journal of Eating Disorders 49 [53] International Journal of Environmental Research and Public Health 68 [57] International Journal of Mental Health and Addiction 23 [54] International Journal of Preventive Medicine 35 [56] Nutrición Hospitalaria 35 In terms of student characteristics (Table 3), most studies analyze eating disorders in college students (83.33%). To a lesser extent there are three studies that focus on the High School stage (25%) On the other hand, most of the articles are written by multiple authors. Only one of them is of unique authorship [58]. The topic of eating disorders in students has been addressed primarily by different authors who submit only one manuscript (Table 1). Only four authors appear in two separate articles.

Meta-Analysis
Two meta-analyses were made based on the data presented by the scientific articles. Meta-analysis with continuous data consisted of five studies [49,54,[57][58][59] (Figure 3). In Tao [58] and Tao and Liu [59] Nutrients 2019, 11, 2151 7 of 11 all the comparisons between groups with PIU and groups with non-PIU (differentiated in the forest diagram with the letters A, B and C) were collected.
Specifically, two assumptions were made regarding the effect of groups (PIU and non-PIU) on eating disorders: a majority in favor of the PIU group and a single study that had no effect [49]. Thus, most of the investigations were placed to the right of the no-effect line and one of them was on the no-effect line. The diamond figure at the far right confirmed that the difference between groups is statistically significant in favor of the PIU group. This was corroborated in the value that was obtained in the size of the effect (p < 0.00001). In addition, according to Cohen's d coefficient, the prevalence of eating disorders in the PIU group had a mean effect size (d = 0.63, 95% CI = 0.41, 0.85). As a result, Internet addiction is associated with the presence of eating disorders. On the other hand, meta-analysis with dichotomous data was composed by five other studies [48,50,52,53,56] (Figure 4). The same two assumptions were repeated as in the previous forest plot: studies with effect in favor of the PIU group and a single study with no effect [53]. The figure of the diamond was placed at the far right confirming the significant difference between groups. The effect size obtained a statistically significant value (p < 0.00001). So the second meta-analysis continued to confirm that PIU is related to the presence of eating disorders (OR = 2.03, 95% CI = 1.58, 2.62).

Discussion
The total number of documents analyzed collected a total sample of 16,520 students. In this large sample of students from different countries and continents, the PIU was linked to the presence of eating disorders. Although PIU is associated with different eating disorders, the published literature on these two variables is limited. However, in 2018 scientific production has begun to resume.
The diversity of authors denotes that there is no specialization in the topic. This confirms that empirical research on PIU and eating disorders is incipient. At the same time, the journals that have published on these two variables are different. The only journal with more than one document is Eating and Weight Disorders, which has a total of four.
On the other hand, the educational stage where this phenomenon has been studied the most is university students, coinciding with the fact that they are the population that is most highly at risk [5,6,40]. Thus, attention was focused on that sector of the population, which has the highest rate of PIU. Therefore, the need arises to establish preventive measures in childhood and adolescence to Specifically, two assumptions were made regarding the effect of groups (PIU and non-PIU) on eating disorders: a majority in favor of the PIU group and a single study that had no effect [49]. Thus, most of the investigations were placed to the right of the no-effect line and one of them was on the no-effect line. The diamond figure at the far right confirmed that the difference between groups is statistically significant in favor of the PIU group. This was corroborated in the value that was obtained in the size of the effect (p < 0.00001). In addition, according to Cohen's d coefficient, the prevalence of eating disorders in the PIU group had a mean effect size (d = 0.63, 95% CI = 0.41, 0.85). As a result, Internet addiction is associated with the presence of eating disorders.
On the other hand, meta-analysis with dichotomous data was composed by five other studies [48,50,52,53,56] (Figure 4). The same two assumptions were repeated as in the previous forest plot: studies with effect in favor of the PIU group and a single study with no effect [53]. The figure of the diamond was placed at the far right confirming the significant difference between groups. The effect size obtained a statistically significant value (p < 0.00001). So the second meta-analysis continued to confirm that PIU is related to the presence of eating disorders (OR = 2.03, 95% CI = 1.58, 2.62). Specifically, two assumptions were made regarding the effect of groups (PIU and non-PIU) on eating disorders: a majority in favor of the PIU group and a single study that had no effect [49]. Thus, most of the investigations were placed to the right of the no-effect line and one of them was on the no-effect line. The diamond figure at the far right confirmed that the difference between groups is statistically significant in favor of the PIU group. This was corroborated in the value that was obtained in the size of the effect (p < 0.00001). In addition, according to Cohen's d coefficient, the prevalence of eating disorders in the PIU group had a mean effect size (d = 0.63, 95% CI = 0.41, 0.85). As a result, Internet addiction is associated with the presence of eating disorders. On the other hand, meta-analysis with dichotomous data was composed by five other studies [48,50,52,53,56] (Figure 4). The same two assumptions were repeated as in the previous forest plot: studies with effect in favor of the PIU group and a single study with no effect [53]. The figure of the diamond was placed at the far right confirming the significant difference between groups. The effect size obtained a statistically significant value (p < 0.00001). So the second meta-analysis continued to confirm that PIU is related to the presence of eating disorders (OR = 2.03, 95% CI = 1.58, 2.62).

Discussion
The total number of documents analyzed collected a total sample of 16,520 students. In this large sample of students from different countries and continents, the PIU was linked to the presence of eating disorders. Although PIU is associated with different eating disorders, the published literature on these two variables is limited. However, in 2018 scientific production has begun to resume.
The diversity of authors denotes that there is no specialization in the topic. This confirms that empirical research on PIU and eating disorders is incipient. At the same time, the journals that have published on these two variables are different. The only journal with more than one document is Eating and Weight Disorders, which has a total of four.
On the other hand, the educational stage where this phenomenon has been studied the most is university students, coinciding with the fact that they are the population that is most highly at risk [5,6,40]. Thus, attention was focused on that sector of the population, which has the highest rate of PIU. Therefore, the need arises to establish preventive measures in childhood and adolescence to

Discussion
The total number of documents analyzed collected a total sample of 16,520 students. In this large sample of students from different countries and continents, the PIU was linked to the presence of eating disorders. Although PIU is associated with different eating disorders, the published literature on these two variables is limited. However, in 2018 scientific production has begun to resume.
The diversity of authors denotes that there is no specialization in the topic. This confirms that empirical research on PIU and eating disorders is incipient. At the same time, the journals that have published on these two variables are different. The only journal with more than one document is Eating and Weight Disorders, which has a total of four.
On the other hand, the educational stage where this phenomenon has been studied the most is university students, coinciding with the fact that they are the population that is most highly at risk [5,6,40]. Thus, attention was focused on that sector of the population, which has the highest rate of PIU. Therefore, the need arises to establish preventive measures in childhood and adolescence to avoid this type of behavior in later stages. From the educational and family spheres, specific actions must be taken to educate students in the proper use of technology.
The variability of countries where research data have been collected confirms the global scale of the problem. So the PIU is not an isolated event from a given region, but is occurring in different countries and continents at an incessant pace [2][3][4].
With respect to the type of eating disorders linked to the PIU, the presence of anorexia nervosa, bulimia nervosa, binge-eating disorder, food preoccupation, loss of control eating, and dieting was detected [17][18][19]26,27]. PIU encourages sedentarism because of the user's need to be connected to the Internet as often as possible [12,13]. This leads to neglect of healthy food and high ingestion of junk food, which is easily accessible and quickly processed.
On the other hand, another danger associated with PIU is the constant consultation of social networks. This phenomenon has had a direct impact on the way of life and behavior of young people. Social networks such as Instagram broadcast daily thousands of audiovisual contents that have an impact on different users. One of the dangers is the recurrence to social comparison with this type of users who have a high audience and present a stylized body figure [24,25]. This can affect the development of eating disorders such as anorexia nervosa or bulimia nervosa.
Finally, meta-analyses confirmed that students with PIU are more likely to develop eating disorders. Thus the assumption (i) that subjects with PIU have a higher rate of eating disorders is verified. So the PIU is a predictor of eating disorders in the student population.

Conclusions
The confirmation of the link between PIU as a predictor of eating disorders leads to a new line of research in studies on the development of eating disorders in childhood, adolescence and young adulthood. The population most at risk are university students. Preventive measures should therefore be introduced at lower educational stages.
This paper responded to the objectives of the study insofar as it determined PIU-related eating disorders in students and analyzed the incidence of PIU in eating disorders. In addition, the paper highlights the need for prevention in childhood and adolescence. Therefore, the main practical application is the elaboration of informative material on the risks of suffering a PIU and the implementation of formative sessions in the educational levels of Middle School and High School.
The limitations of the study are highlighted as (i) the limitation of searching scientific documents only in the WOS and Scopus databases, although for a first approximation to the topic of study it is more than sufficient, since the literature with the greatest scientific impact has been analyzed, and (ii) the bias of researchers in selecting articles, which we attempted to resolve with the opinion of a third researcher.
Finally, the data obtained enable the creation of new studies and the continuity of this paper. Confirmation that PIU is a predictor of eating disorders should continue to be investigated in new empirical studies linking these two variables. On the other hand, the analysis of the type of behavior carried out by users on the Internet and how this behavior is linked to specific eating disorders is an interest topic. So that studies emerge that analyze the relationship between the abusive use of social networks with anorexia nervosa and bulimia nervosa. Or the abusive use of video games with dieting and binge-eating disorder.
To sum up, Internet addiction is one of the main challenges facing today's society. The excessive use of technology is causing different problems in human health. One of them is an increase in eating disorders. Researchers and professionals can help to detect these consequences and propose measures to avoid this type of problematic behaviour.