Edited by: Shaojing Sun, Fudan University, China
Reviewed by: Jake Linardon, Deakin University, Australia; Xiaomeng Xie, University of Macau, Macao, China; Wesley Barnhart, Bowling Green State University, United States
This article was submitted to Eating Behavior, a section of the journal Frontiers in Psychology
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Social Networking Sites (SNSs) are common tools with which modern people share their lives and establish social relationships. However, some studies have found SNSs to be associated with eating disorders, although other have identified no connection between the two. To explore the interaction between SNSs and eating disorder behaviors, this study aimed to comprehensively synthesize previous studies using meta-analysis methods. Based on selection criteria, there were 87 effect sizes from 22 studies. After analysis using a three-level random-effects meta-analysis model, a positive correlation between the use of SNSs and irregular eating behaviors was found,
Eating disorders (EDs) are recognized mental illnesses characterized by irregular eating habits and abnormal concerns about body weight and shapes. Such disorders are chronic, difficult to recover from, prone to relapse and often have serious sequelae (Brownell and Walsh,
Numerous studies have been conducted to investigate the causes of disordered eating behaviors. In this regard, SNSs, as online communication platforms, have become a novel research area of particular interests (Brandtzæg and Heim,
Although no previous meta-analysis has been conducted in this area, there was one systematic review of 20 articles published before 2016 (Holland and Tiggemann,
Since 2018, the related studies have risen sharply in number, and they have begun to explore the interaction between SNSs and EDs in more detail. Some have confirmed the findings of the review mentioned above (Teo and Collinson,
SNSs are defined as websites or applications located on the Internet that provide individuals with platforms for displaying and sharing their personal lives and interacting with others through provided functions such as comments, likes, and reposts (Perloff,
Since the focus of this study is on the duration, frequency, and intensity of SNS use, most of the questionnaires it examines are adapted from questionnaires previously used to measure other network activities such as Internet use and Facebook use (Tiggemann et al.,
Although the way that people participate in SNS activities is completely different from their use of traditional media, research has shown that they might still be as influenced by SNSs as they are by traditional media (Holland and Tiggemann,
Two theoretical frameworks are relevant to explaining the possible relationship between EDs and SNSs. According to sociocultural models, people who interact closely with individuals influence their views about weight and the body (Stice,
Another widely-accepted theoretical basis for the relationship between EDs and SNSs is “self-objectification,” by which individuals commoditize their own value, and believes that the value of their personal identity is derived from the use and consumption of their body and its appearance to others (Fredrickson and Roberts,
In view of the sociocultural models, the objectification theory and the varying results summarized above, certain potential moderators should be considered when analyzing the data, to gain a thorough understanding of the relationship between SNS usage and disordered eating behaviors. These moderators are discussed below.
There are several types of SNS, each with unique designs and functions. Image-centric social media (hereinafter referred to as image-based SNS) are social media whose functions are mainly based on the use of photography and other images. Examples include Instagram, Snapchat and Facebook (Rodgers and Melioli,
Social media have developed rapidly, and new features or entirely new platforms may be appear over a short time (Bowman and Clark-Gordon,
A study pointed out that men and women display different habits in their use of SNSs, and that women tend to post more photos on social platforms, whether selfies, group photos or food photos (Wilksch et al.,
Given the difference in political system between China and the West, the phenomena of social comparison in Chinese and Western societies may also differ, as the socialist environment places less emphasis on the individual (Hofstede,
People of different ages have different life priorities (Erikson,
To the best of our knowledge, no previous meta-analysis has explored the combined effect of SNS usage and disordered eating behaviors. Given that meta-analysis can quantitatively summarize previous findings (e.g., the correlations between variables) with a large sample size, and can further speculate on the factors that might have affected the relationships between variables by moderator analysis, it is expected that the current meta-analysis will fill the research gap and enhance understanding of the literature on the relationship between disordered eating behaviors and SNS usage. First, this study aimed to collect all the relevant and accessible studies that were conducted before 2020. Second, this research analyzed many potential moderators in order to better explain the inconsistencies between previous research findings. A three-level random-effects meta-analysis model was adopted, which is suitable for obtaining a more accurate evaluation of the overall effect size in a large body of research (Van den Noortgate et al.,
It was hypothesized that individuals experiencing a high intensity or duration of SNS usage would be more likely to exhibit disordered eating behaviors. However, since the results of previous studies did not reach a broad consensus on this subject, further hypothesize could not be on the influence of the moderators on the results of the current study.
Related studies were searched and retrieved from five databases (PsychINFO, PubMed, Web of Science, Communication and Mass Media Complete, and ProQuest Dissertations) on July 18th, 2020. The following search keywords for social network usage and disordered eating were chosen: (“social media” OR “social networking sites” OR “SNS” OR “Twitter” OR “Facebook” OR “Weibo” OR “Instagram”) AND (“eating” OR “disordered eating” OR “eating disorder”). In addition, a manual search was performed of the reference list in the identified articles to find any other relevant research. It is worth noting that Google Scholar was selected as the fourth resource because it can span multiple disciplines, so that it can be used as a final check to ensure that all articles that meet the current inclusion criteria are captured. Considering the rise of social media, only documents published after 2010 were selected. In fact, no document exceeding this time limit during the search was found. This meta-analysis also aimed to find out the relationship between SNS and disordered eating behaviors in existing studies.
After the initial search, 480 articles were found. The following criteria were applied to screen the 480 articles:
(a) written in English;
(b) published in journals or dissertations;
(c) reported the correlation between social network usage (intensity or frequency) and disordered eating behaviors (r), which had to be a primary goal of the studies.
Only the articles that met the above three criteria were selected.
The following information was retrieved from the selected studies: (1) name of the first author; (2) publication year; (3) age; (4) Body mass index (BMI); (5) percentage of males; (6) percentage of college degree; (7) percentage of white; (8) publication type (dissertation or journal article); (9) region (Western or Eastern); (10) survey methods (paper-and-pencil or online); (11) sample source (university, children and adolescent, clinical or other); (12) SNS type (image-based, non-image-based or general); (13) SNSs usage (duration: time spent on SNSs; frequency: number of times of SNSs usage in a certain period; intensity: integration of SNSs into daily life); (14) type of disordered eating behavior (combined disordered eating behavior, binge eating, driving for thinness, bulimia or dietary restraint); (15) measure of eating (EAT-26, Project Eat III- Eating Behavior Questions, Eating Disorder Inventory, Dutch Eating Behavior Questionnaire, The EDE-Q or others; (16) correlations (
To establish internal encoder reliability, two independent coders coded three articles randomly selected from 22 articles. After two rounds of coding, all coders achieved acceptable inter-coder reliability (the Cohen Kappa range of all variables in the coding scheme was 0.85–0.87). They then independently coded the remaining 19 articles and reached an absolute consensus of 95%. The coders resolve any differences through discussion to obtain the final coding result.
Quality assessment was performed by Q.L and Y.W independently. Disagreement about scores was resolved through discussion between the two authors. The quality of the studies included in the meta-analysis was retrieved and adapted from the previous studies (He et al.,
To prevent the quality of a selected article from affecting the results, which was analyzed as a categorical moderator in the subsequent analysis (the first category is articles with a quality of 80% and above, and the rest are in the second category). The results show that the quality of the article does not affect the results of this study [
As this was a cross-sectional analysis, all data analyses were performed with the R 4.0.0 (R Core Team,
The outliers were inspected through the
The heterogeneity was assessed by the Cochrane
To avoid dependence problems such as effect sizes, observations, and error terms which are dependent and correlated if they are from the same study, a multilevel meta-analysis was used (Van den Noortgate et al.,
PRISMA flowchart.
All studies included in this meta-analysis were published between 2010 and 2020. There were 29 independent sample sizes and 87 effect sizes in all 22 studies. A total of 13,301 samples were covered. The sample size of males was 5,031 (37.82% of the total), and 8,270 females. The average age of the sample was between 11.19 and 30.53, and the average BMI varied from 18.92 to 24.69. There were 73 effect sizes obtained from Western and 14 from Eastern. Moreover, 87 effect sizes were reported on the relationship between SNS and disordered eating behaviors (51.72% of the total effect size). Please refer to
Samples, research methods, and data processing for the chosen articles were examined (see
As
Baujat plot. Notes: each number represents one study included in this research.
From the 22 studies that examined the correlation between SNS and disordered eating behaviors, 87 effect sizes were observed, ranging from −0.35 to 0.45. Significant heterogeneity existed among the effect sizes [
Forest plot for all samples. Notes: The horizontal lines show 95% confidence interval; the diamond represents the point estimate and confidence interval of the pooled effect size.
Moderator analyses for studies reporting the correlation between SNS and disordered eating behaviors.
Publication year | 27 | 87 | −11.707 (−32.051; 8.638) | 0.006 (−0.004; 0.016) | 1.329 (1, 84) | 0.005 |
0.001 | |
Age | 18 | 36 | 0.126 (−0.023; 0.276) | −0.003 (−0.011; 0.005) | 0.523 (1, 34) | 0.002 |
0.002 | |
BMI | 13 | 37 | 0.746 (0.199; 1.292) |
−0.032(−0.058; −0.006) |
6.080 (1, 34) |
0.001 | 0.002 | |
Percent of male | 11 | 39 | 0.089 (0.057; 0.121) | −0.007 (−0.058;0.045) | 0.065 (1, 84) | 0.005 |
0.001 | |
Percent of college | 14 | 37 | 0.107 (0.071; 0.144) | 0.001 (−0.000; 0.003) | 2.194 (1, 32) | 0.007 |
0.000 | |
Percentage of white | 14 | 23 | 0.049 (−0.049; 0.147) | 0.008(−0.136; 0.152) | 0.014 (1, 23) | 0.002 |
0.003 | |
Publication type | 0.113(1, 84) | 0.005 |
0.001 | |||||
Journal | 21 | 59 | 0.089 (0.057; 0.121) | 0.088 | ||||
Thesis | 6 | 28 | −0.012 (−0.085; 0.060) | −0.0120 | −0.101 | |||
Region | 2.776 (1, 84) |
0.005 |
0.000 | |||||
Western | 22 | 73 | 0.077 (0.050; 0.104) |
0.0768 |
||||
Eastern | 5 | 14 | 0.052 (−0.010; 0.114) | 0.0520 | −0.025 | |||
Survey methods | 5.253 (1, 84) |
0.004 |
0.000 | |||||
Paper–and-pencil | 16 | 43 | 0.114 (0.081; 0.147) |
0.114 |
||||
Online | 11 | 44 | −0.055 (−0.102; −0.007) |
−0.055 |
−0.169 | |||
Sample source | 2.876 (3, 82) |
0.005 |
0.000 | |||||
University sample | 12 | 35 | 0.089 (0.049; 0.129) |
0.089 |
||||
Children and adolescent sample | 7 | 29 | 0.035 (−0.023; 0.092) | 0.035 | −0.054 | |||
Clinical sample | 1 | 4 | −0.171 (−0.379; 0.038) | −0.169 | −0.26 | |||
Other sample | 7 | 19 | −0.039 (−0.099; 0.021) | −0.039 | −0.128 | |||
SNS use measure | 0.105 (3, 82) | 0.005 |
0.001 | |||||
Duration | 13 | 51 | 0.084 (0.040; 0.127) | 0.084 | ||||
Frequency | 10 | 32 | 0.010 (−0.052; 072) | 0.010 | 0.074 | |||
Intensity | 2 | 2 | −0.014 (−0.171; 0.143) | −0.014 | −0.098 | |||
Mixed | 2 | 2 | −0.024 (−0.170; 0.122) | −0.024 | 0.108 | |||
SNS type | 0.005 | 0.002 | ||||||
Image-based | 10 | 34 | 0.070 (0.027; 0.112) |
0.070 | ||||
Non image–based | 13 | 31 | −0.234 (−0.334; −0.134) |
−0.223 | −0.304 | |||
General | 14 | 22 | 0.044(−0.021.108) | 0.044 | −0.026 | |||
Type of disordered eating | 0.712(4, 82) | 0.006 |
0.000 | |||||
Combined disordered eating behavior | 22 | 54 | 0.077 (0.047; 0.106) |
0.077 | ||||
Binge eating | 4 | 18 | 0.028 (−0.033; 0.090) | 0.028 | −0.049 | |||
Driving for thinness | 3 | 5 | 0.005 (−0.086; 0.097) | 0.005 | −0.072 | |||
Bulimia | 2 | 4 | 0.025 (−0.080; 0.131) | 0.025 | −0.052 | |||
Dietary restraint | 5 | 6 | 0.069 (−0.025; 0.164) | 0.069 | −0.008 | |||
Measure of eating | 2.252(5, 80) |
0.005 |
0.000 | |||||
EAT-26 | 13 | 25 | 0.109 (0.068; 0.150) |
0.109 | ||||
Project Eat III–Eating behavior questions | 2 | 10 | −0.005 (−0.099; 0.089) | −0.005 | −0.114 | |||
Eating disorder inventory | 4 | 11 | −0.018 (−0.091; 0.055) | −0.018 | −0.127 | |||
Dutch eating behavior questionnaire | 3 | 3 | 0.073 (−0.033; 0.180) | 0.073 | −0.036 | |||
The EDE-Q | 9 | 28 | −0.058 (−0.109; −0.006) |
−0.058 | −0.167 | |||
Others | 3 | 10 | 0.007 (−0.064; 0.078) | 0.007 | −0.102 |
The Rank Correlation Test for Funnel Plot Asymmetry indicated no publication bias for the correlation between SNS and disordered eating behavior (Kendall's tau = 0.050,
Funnel plot. Notes: Black circles are the studies included in the meta-analysis.
The current study was intended to expand upon previous work by adopting a three-level meta-analysis model to analyze the association between SNSs and disordered eating behaviors. Analysis in this study revealed a weak but significant positive correlation between the use of SNSs and disordered eating behaviors, in line with the results of several previous studies (Mabe et al.,
Moderation models help to understand if other variables explain the strength of relationship between two variables. This study therefore assessed several potential moderators, hoping to gain a clearer understanding of the inconsistency of findings in the literature as to whether and to what extent SNS usage is related to disordered eating behaviors. Moderator analyses showed that the sample source, survey method, and mean BMI of the sample were the significant moderators, which may explain the individual differences of the previous findings.
As for the sample source, university students were the main contributors to the discrepancies in disordered eating behaviors associated with SNS usage, compared with children, adolescents, clinical samples, and other samples. College students are the main contributors to studies investigating social media use (Zhang and Leung,
The survey methods are likely to also adjust the association between disordered eating behaviors and SNSs usages between different subgroups, because sometimes measuring the same variable through different methods (such as the pen-and-paper and online methods mentioned in this article) may yield different results (Moessner et al.,
As far as BMI is concerned, the results indicate that an increase in the average BMI of the sample is usually accompanied by a decrease in the tendency for people to suffer from eating disorders associated with SNS use. BMI has been shown to have a significant impact on an individual's eating behaviors (Burnette et al.,
In conclusion, the present meta-analysis has offered a quantitative synthesis of the current state of knowledge on the relationship between SNS usage and disordered eating behaviors. Based on a three-level meta-analytic model and moderator analysis, the research has demonstrated that SNS use is significantly linked to disordered eating behaviors and attitudes, which might be altered by the sample source, survey method, and mean BMI of the sample. According to sociocultural theory and self-objectification theory, individuals who use SNSs frequently and intensively seems to be more likely to internalize the ideal value of slimness of their social and cultural environment through information in SNSs, and to take part in social comparisons related to appearance. On the other hand, SNS usage might encourage individuals to connect their values with their body shapes. Both these inferences suggest that SNS usage is likely to lead to body dissatisfaction and may indeed play a causal role in the development of disordered eating behaviors.
Some limitations should be considered when interpreting the current research. First, language was set to English when filtering the research. In addition, some gray literature (such as ongoing research) would not appear in the general search process. This may have caused some literature to be lost from analysis of this study. Therefore, future meta-analyses should include richer studies to conduct a more comprehensive and thorough analysis of related issues. Second, in addition to the proposed moderator, there might be other factors that affect the consistency of the research results, such as sexual orientation and the ways of using SNSs (Ryding and Kuss,
Future research needs to explore the other popular and current forms of SNSs (i.e., Twitter, Instagram, and Pinterest) owing to the rapid development of SNS platforms (Duggan et al.,
The current meta-analysis revealed a small, positive correlation between frequent and intensive use of SNSs and disordered eating behaviors. In addition, the BMI of the sample, the source of the sample, and the survey method (paper-and-pencil or online) were identified as the moderators that may explain the inconsistent findings between SNSs usage and disordered eating.
The findings from this meta-analysis have several clinical implications. First, this study found a positive correlation between the use of SNS and disordered eating behaviors. Clinicians may therefore consider evaluating the influence of SNS use on patients' irregular eating behaviors during the intervention process for disordered eating behaviors, and intervene with a view to controlling the length and frequency of SNS use.
The current meta-analysis also draws attention to the importance of the proper usage of SNSs in preventing disordered eating behaviors. According to the sociocultural theory and objectification theories which may explain the underlying principles of SNS usage and their positive associations with disordered eating behaviors, in addition to the need to control the frequency and duration of use of SNSs, the overall aesthetic orientation of these media and the comments made in them by the others are what affect the consequences of using SNSs. Frequent social comparisons on social media may aggravate the conflicts between the glamorous social images that people see displayed on their homepages and their perceptions of themselves. The findings of this meta-analysis may therefore have several media-related implications. Relevant institutions are advised to regularly organize some positive campaigns on SNSs to encourage people to pay attention to personal characteristics other than mere appearance. The promotion of the positive use of SNSs, which has been connected to fewer negative outcomes such as eating disorders and more positive outcomes such as the formation of social bonds, is also recommended (Verduyn et al.,
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
YW and JZ contributed to the research design. YW and QL collected relevant articles, completed the coding, and drafted the manuscript. YW analyzed the data. JZ and CW carefully revised the manuscript. All authors read and approved the final version of the manuscript.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The Reviewer XX declared a shared affiliation with several of the authors, JZ, YW, and QL, to the handling editor at time of review.
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
The Supplementary Material for this article can be found online at:
Quality assessment.
Santarossa ( |
1 | 1 | 1 | 1 | 1 | 1 | 6 | 100% |
Blassingame ( |
1 | 0 | 1 | 1 | 1 | 1 | 5 | 83.3% |
Cohen et al. ( |
0 | 0 | 1 | 1 | 0 | 1 | 4 | 66.7% |
Rodgers et al. ( |
1 | 0 | 1 | 1 | 1 | 1 | 5 | 83.3% |
Suplee ( |
1 | 0 | 1 | 1 | 1 | 1 | 5 | 83.3% |
Ferguson et al. ( |
0 | 0 | 1 | 1 | 1 | 1 | 4 | 66.7% |
Latzer et al. ( |
0 | 1 | 1 | 1 | 1 | 1 | 5 | 83.3% |
Mabe et al. ( |
0 | 0 | 1 | 1 | 1 | 1 | 4 | 66.7% |
Acar et al. ( |
0 | 0 | 1 | 1 | 1 | 1 | 4 | 66.7% |
Pollack ( |
1 | 0 | 1 | 1 | 1 | 1 | 4 | 66.7% |
Walker et al. ( |
1 | 1 | 1 | 1 | 1 | 1 | 6 | 100.0% |
Griffiths et al. ( |
1 | 1 | 1 | 1 | 0 | 1 | 5 | 83.3% |
Teo and Collinson ( |
1 | 1 | 1 | 1 | 1 | 1 | 6 | 100.0% |
Howard et al. ( |
1 | 1 | 1 | 1 | 1 | 1 | 6 | 100.0% |
Slater and Tiggemann ( |
1 | 0 | 1 | 1 | 1 | 1 | 5 | 83.3% |
Zeeni et al. ( |
0 | 0 | 1 | 1 | 1 | 1 | 4 | 66.7% |
Niu et al. ( |
1 | 0 | 1 | 1 | 1 | 1 | 4 | 83.3% |
Aparicio-Martinez et al. ( |
0 | 1 | 1 | 1 | 1 | 1 | 5 | 83.3% |
Griffiths et al. ( |
1 | 0 | 1 | 1 | 0 | 1 | 4 | 66.7% |
Fardouly et al. ( |
1 | 1 | 1 | 1 | 1 | 1 | 6 | 100.0% |
Wilksch et al. ( |
0 | 0 | 1 | 1 | 1 | 1 | 4 | 66.7% |
Schreyer-Hoffman ( |
1 | 0 | 1 | 1 | 1 | 1 | 4 | 83.3% |