Internet addiction, depressive symptoms, and anxiety symptoms are associated with the risk of eating disorders among university students in Bangladesh

The university-aged period (18–26 years), which falls into an emerging adulthood stratum of lifespan, is often considered a key developmental phase of life¹. University students (i.e., young adults) go through this transition era where they leave the family home to pursue higher education, and their way of life changes due to academic obligations, growing flexibility and independence, and reform of peer relationships. Studies have shown that university students are at risk of various mental health conditions (e.g., depression, anxiety, etc.), addictive behaviors (such as internet addiction), unhealthy dietary habits, and eating disorders^2–5.

Eating disorders (EDs) are abnormal eating behaviors that impair social and psychological functioning⁶. EDs have been associated with various long-term physical consequences and a high risk for comorbidity and mortality, with the highest proportion of cause-specific death among psychiatric disorders^7,8. Globally, approximately 42 million people suffer from different forms of EDs (e.g., anorexia and bulimia); however, people with EDs tend to not seek medical help^9,10. Thus, early screening and diagnosis of individuals who are at risk of EDs are crucial to providing treatment and reducing disease burden.

Evidence shows that university or college students have a higher prevalence of EDs compared to the general population¹¹, which can adversely impact students’ physical and mental health, lower academic functioning and functional impairment¹². A recent systematic review and meta-analysis estimated that the prevalence of screen-based disordered eating behavior among university students, globally, was 20%². In recent years, the risk of EDs appears to have increased among university students in Asian countries^13,14, particularly in Bangladesh, with estimates ranging between 20.4 and 37.6%^15,16. Previous studies conducted among students of public universities in Bangladesh reported that the prevalence of ED risk ranges between 20.4 and 23.0%^15,17. A recent study undertaken among female Bangladeshi university students reported that approximately 37% of them were at risk of developing an ED¹⁸. Another previous study of Bangladeshi undergraduate university students revealed that the prevalence of developing an ED risk was 37.6%¹⁶. The increasing trends of disordered eating behaviors in Asian regions may be due to economic expansion, social transformation in gender responsibilities and conventional family constructions, greater exposure to the Western tradition and standards of beauty, and increased access to junk and fast foods¹⁹.

The etiology of EDs is complex with a wide range of sociocultural, psychological, and biological influences²⁰. Previous studies have identified several factors associated with ED risk or disordered eating attitudes in different settings, such as: (i) sociodemographic and cultural factors: age, female sex, married status, exposure to cultural transition, globalization, urbanization and lack of religiosity^21–24; (ii) body-related factors: body mass index (BMI), body shape concern, desire for thinness, and cosmetic surgery^24–26; and (iii) psychiatric and behavioral factors: depression, anxiety and pathological internet use^27–29.

In Bangladesh, a key area of public health concern is common mental health problems including depression and anxiety, as well as internet use/internet addiction among university students^4,5. A meta-analysis by Puccio et al.²⁸ concluded that eating pathology and depression are concurrent risk factors for each other, with other studies showing problematic internet use is a predictor of EDs in students³⁰; however, contradictory findings have also been reported. One study of university students in the Association of Southeast Asian Nations (i.e., ASEAN refers to Indonesia, Malaysia, Myanmar, Thailand and Vietnam) found that depression symptoms and pathological internet use were significantly associated with ED risk¹³. Although students are one of the most susceptible groups to developing ED, in Bangladesh, few studies have investigated ED risk among university students and additional sociodemographic, health/lifestyle, and/or behavioral factors associated with EDs risk^15,16,18.

Since psychiatric disorders and addictive behaviors predominantly exist among Bangladeshi university students, these could make them more vulnerable to disordered eating habits or ED risk. Thus, in this study, we included university students’ socio-demographic characteristics, internet addiction, depression, and anxiety symptoms to observe the estimated adjusted effect of these factors on ED risk. The main objective of the present study was to assess the association between mental health conditions (such as depression and anxiety) and addictive behaviors (e.g., internet addiction) with the risk of ED among university students in Bangladesh. Initial hypotheses were female gender, internet addiction, and depressive and anxiety symptoms would be associated with the risk of ED in Bangladeshi university students.

A cross-sectional study was undertaken among students sampled from two conveniently selected public universities in Bangladesh. The following eligibility criteria were set up to enroll a study participant: (i) being at least 18 years old (i.e., adults), (ii) being a Bangladeshi citizen by birth, and (iii) being a current student. Moreover, participants who had any prior history of a clinically diagnosed ED and medical conditions such as diabetes and hypertension were excluded from the study due to the potential for research outcomes to be overestimated or underestimated. The study was conducted between May and October 2022.

A minimum required sample was estimated using a single population proportion test. The following assumptions were factored in: (i) a 23.0% prevalence of ED risk among Bangladeshi public university students was used (p = 0.23)¹⁷, (ii) 95% confidence (Z = 1.96), and (iii) 5% margin of error (d = 0.05). The calculation formula was as follows:

A 10% non-response rate was also considered, resulting in an optimal sample size of 300. To increase the external validity and generalizability of the study, we targeted to include more participants than our calculated sample size³¹. Thus, a total of 700 participants constituted the final sample size in this study.

A simple random technique was applied to recruit and enroll participants. Three field investigators (i.e., interviewers) randomly visited different areas on the university campuses where students gather together to spend their leisure time such as the cafeteria, tea stall, central field, teacher student center (TSC), etc. There are no restrictions on accessing public university campuses in Bangladesh, thus, the field investigators randomly approached students on the campuses, clarified the study’s purpose to them, and assessed study eligibility criteria. Almost equal proportions of participants were recruited from each of the two universities (i.e., 345 vs. 355 participants).

Prior to data collection study procedures and protocols related to human subjects research was approved by the Research Ethical Committee (REC) of Department of Environmental Sanitation, Patuakhali Science and Technology University, Bangladesh [approval number: ENS:09/04/2022:06]. All methods were carried out in accordance with the relevant guidelines and regulations. Interviewers, who were trained by the principal investigator, collected data through a person-to-person interview using a paper-based questionnaire comprising of the study variables measures. Before beginning the questionnaire, written informed consent was obtained and the study's purpose was explained to the participants, and it was made clear that study participation was voluntary. Participants’ de-identifying information and the confidentiality of their data were strictly maintained. The interviewer read each question on the questionnaire with all possible response options to the participant and completed it in-person.

Data were analyzed by STATA (BE version 16.0, StataCorp, College Station, TX, USA) and SPSS (Windows version 23.0). Enumerative statistics, including frequency, percentage, mean and standard deviation (SD) were computed. A chi-square test was performed to show the distribution of the outcome variable across the independent variables. A bivariate and multivariable logistic regression model was fitted to identify the factors associated with ED risk. The nature of the variables derived from the data met all the key assumptions of a binary logistic regression. First, the dependent variable of this study was dichotomous (i.e., at-risk or no risk for ED). That is, it is either present or absent but never both at once. Second, since all our variables were categorical and binary, we confirmed using the proportional distribution of the categories in terms of their number and distribution across the outcome variable that there were no extreme outliers. Finally, before performing the adjusted regression model, all explanatory variables (16 variables) were tested for multicollinearity using variance inflation factor (VIF), and no multicollinearity was observed [mean VIF = 1.433, minimum VIF = 1.035, Maximum VIF = 3.123]. In the bivariate logistic regression model, the odds ratio with the corresponding confidence interval (CI at 95% level) was computed for all variables. A backward binary logistic regression method was used to construct the best model to explain the data, where the model starts with a full model and gradually drops the least significant variables from the model. The adjusted model was fitted based on the Hosmer and Lemeshow goodness of fit test (Chi-square = 9.466, df = 8, p = 0.305). The threshold for statistical significance was set at < 0.05.

The study included 700 university students between the age range of 18 to 27 years (mean age 23.11 years, SD 1.77), with males making up more than half (52.0%) of the sample. The response rate of this study was 91.02%. Based on the PHQ-9 scale (mean score 9.35, SD 5.71), around 39.1% of the participants were identified as having depressive symptoms. The average score of the GAD-7 scale was 8.42 (SD 4.62) and a quarter (25.9%) of the participants were found to have anxiety symptoms. According to the OIAS scale (mean score 5.10, SD 3.14), nearly two-thirds (66.6%) of the participants had moderate to severe levels of internet addiction (Table 1).

The average score for the EAT-26 was 17.85 (SD 12.80). Based on EAT-26 screener, the prevalence of ED risk among study participants was 30.57%. The chi-square test showed participants’ gender (p = 0.008), age (p < 0.001), study level (p < 0.001), marital status (p = 0.019), monthly family income (p = 0.003), tobacco smoking status (p < 0.001), self-reported BMI status (p < 0.001), depressive symptoms (p < 0.001), anxiety symptoms (p < 0.001), and internet addiction level (p = 0.007) was significantly associated with ED (see Table 1).

Table 2 presents the findings of bivariate logistic regression analysis, showing all variables except participants’ academic performance, residence, family size, physical activity level, and sleep duration (p > 0.05) were significantly associated with the risk of ED.

The adjusted estimated effects of the factors associated with the risk of ED among study participants are shown in Table 3. The adjusted regression analysis demonstrated that female students had a higher likelihood of ED risk compared to male students (adjusted odds ratio, aOR 3.11, 95% CI 2.00 to 4.84, p < 0.001). The higher odds of having at-risk ED were found among students with depressive (aOR 3.04, 95% CI 2.69 to 6.08, p < 0.001) symptoms and anxiety (aOR 2.06, 95% CI 1.33 to 3.17, p = 0.001) symptoms compared to their counterparts. Students who had severe (aOR 3.95, 95% CI 1.92 to 8.11, p < 0.001) and moderate internet addiction (aOR 2.15, 95% CI 1.14 to 4.05, p = 0.019) were at higher odds of having at-risk ED than those who had no internet addiction (see Table 3).

This study had several strengths. It was one of the first studies to explore how mental health conditions and addictive behaviors are associated with the risk of ED among university students in Bangladesh. Being a baseline evidence for Bangladesh, these findings would assist university authorities and policymakers to develop behavioral interventions and coping strategies to reduce the risk of ED. Another strength of this study is its large sample size, validated screening tool, and robust methodological approaches. This study was not without limitations. Causal interference between outcome variable and explanatory variables is limited due to the cross-sectional study design. We did not perform any clinical examination for diagnosing EDs, and an ED cannot be diagnosed reliably with a single screening technique⁴³. Since the EAT-26 represents restrictive eating pathology and does not precisely detect binge eating⁵⁸, our study findings could not estimate the real picture of binge eating in this subject group. Furthermore, the EAT-26 has been questioned for its utility in estimating ED risk in individuals at low-risk for ED (for example; overweight and obese individuals undergoing weight loss treatment)⁵⁹. Lastly, due to the self-report nature of our measures, response bias and social-desirability may be present in our current data.

The datasets and questionnaire used and/or analyzed during the current study will be shared upon request to the corresponding author.