Data were collected using a digitized questionnaire and a pre-test was carried out on other targets in order to correct any errors and clarify ambiguous questions.

The dependent variable in this study was the risk of at least one eating disorder. It was assessed using several scales: the “EAT: Eating Attitudes Test-26”, the “Bulimic Investigatory Test Edinburgh: BITE” and the “HAD: Hospital Anxiety and Depression scale”.

Ø Eating Attitudes Test-26.

The EAT-26 questionnaire was developed in 1979 by Garner et al. It is a screening tool for measuring the frequency of attitudes and behaviours characteristic of anorexia nervosa, with a short administration time of around two minutes. Initially in English, it was translated into French before validation by Leichner et al. in 1994. The questionnaire comprises 26 items. Each item was rated in terms of frequency of behavior or thought, from “always” to “never”. A total score of 20 or more indicates the presence of anorexia nervosa attitudes [12] .

Ø Bulimic Investigatory Test Edinburgh: BITE.

The BITE questionnaire was created in 1987 by Henderson and Freeman. It’s a 30-item self-administered questionnaire for screening for bulimia nervosa and binge eating disorder. Only the original English version is validated. Nevertheless, it has been translated into French and used by many authors in their studies. It is short and simple to use, and has good reliability and validity for assessing bulimic behaviors, including patients with both bulimia nervosa and binge eating disorder [13] .

- Score < 10: Eating behavior within the norm.

- Score between 10 and 19: Eating frenzy to be investigated.

- Score ≥ 20: high probability of bulimia [13] .

In the absence of a pica screening scale, a mini-questionnaire consisting of 04 questions, including two main questions with binary responses, was drawn up on the basis of the DSM-V diagnostic criteria for pica. A “yes” answer to any of these questions would indicate a positive screen for pica.

Ø Hospital Anxiety and Depression scale: HAD scale

This is a screening instrument for anxiety and depressive disorders. It was developed by Phillip Snaith and Anthony Zigmond in 1983 and translated in 1985 by JP Lépine et al. It comprises 14 items rated from 0 to 3. It takes between 2 and 6 minutes to complete. Seven questions relate to anxiety (total A) and seven to depression (total D), giving two scores (maximum score for each = 21). To screen for anxiety and depressive symptoms, the following interpretation was adopted for each of the scores (A and D) [14] .

- 7 or less: no symptoms;

- 8 to 10: doubtful symptomatology;

- 11 and over: definite symptomatology [14] .

Data analysis was performed using SPSS version 25 software. Qualitative variables were described using percentages with their confidence intervals, and quantitative variables using average and standard deviation or median and interquartile ranges, depending on the type of data. Frequency comparisons were made using Chi2 tests (Pearson, Yates or Fisher as appropriate), and a p-value < 0.05 was considered significant.

The extremes of age were 16 and 34 years. The average age was 21.21 ± 2.24 years in the general student population and 21.15 ± 2.79 years in the population of students at risk of eating disorders.

The age group most represented in both the general population and the population of students at risk of eating disorders was the 18 - 24 age group, with proportions of 84% and 85.45% respectively (Table 1).

In the general population, the majority of respondents (55.52%) were male compared with 44.42% who were female. Meanwhile, in the population of participants at risk of eating disorders, 54.49% were male versus 45.51% female, giving a sex ratio of 1.97 (Table 1).

Among the general population of respondents, 20 (3.29%) had suffered from a mental illness at least once and 28 (4.61%) had undergone surgery at least once. Slightly more than a third 222 (36.57%) of respondents used psychoactive substances: alcohol was consumed by 213 students (35.09%), tobacco by 7 (1.15%) and only 2 (0.33%) claimed to use drugs (cannabis and others). In terms of criminal record, 03 students (0.49%) claimed to have had trouble with the Law (Table 2).

Of the students at risk of eating disorders, 12 (3.72%) had a psychiatric history and only 3 of them (0.93%) had received psychiatric or psychological treatment. Similarly, 12 (3.72%) of the subjects had undergone surgery at least once. One person (0.31%) reported having had problems with the Law. As for the use of psychoactive substances, there were 119 of them, i.e., more than a third (36.84%) (Table 2).

Of those surveyed, 43 (7.08%) had at least one relative suffering from a mental pathology. In the group of students at risk of eating disorders, a slightly higher proportion (8.98%) had a family history of mental illness.

In the general population, 206 students (33.94%) used strategies to control their weight, the most common being sport (25.54%). Among those at risk of eating disorders, 120 (37.15%) used weight-control methods, with sport also the main method.

In terms of body mass index (BMI), the average BMI in the general population was 21.56 ± 2.53 kg/m², with extremes of 14.69 and 40.17 kg/m². Subjects at risk of at least one binge-eating disorder had an average BMI of 21.44 ± 3.72 kg/m², with extremes of 15.24 and 40.17 kg/m² (Table 4).

According to the HAD scale, 22.90% had doubtful symptoms of anxiety disorders, and 14.17% had definite symptoms. Depressive symptomatology was doubtful for 12.32% of students and certain for 4.45%.

Among students at risk of at least one eating disorder, 26.01% had doubtful anxiety symptomatology and 16.72% had definite symptomatology. Similarly, 14.55% had questionable symptomatology for depression and 4.64% had definite symptomatology (Figure 1).

In this Figure 1: the blue bar represents students with no symptoms of anxiety or depressive disorders; the red bar represents students with doubtful symptomatology for anxiety and depressive disorders; the green bar represents students with definite symptoms of anxiety and depressive disorders.

Among the cohort of 607 students surveyed, 323 individuals with at least one eating disorder were identified. The prevalence of eating disorders among these students was 53.21%.

Among these 323 at-risk students, data on the numbers and frequencies of each disorder investigated were as follows: 234 (86.37%) presented with anorexia nervosa, 94 (29.1%) with binge eating disorder, 77 (23.83%) with pica, and 5 (1.54%) with bulimia nervosa (Figure 2).

In relation to the total population surveyed (N = 607), the proportions of each of these disorders were as follows: anorexia nervosa (45.96%), bulimia nervosa (15.48%), pica (12.68%), and bulimia nervosa (0.82%) (Figure 3).

In bivariate analysis, the following variables showed a statistically significant association with the risk of at least one eating disorder: place of residence (p = 0.004), religion (p = 0.018), type of university attended (p = 0.024), body mass index BMI (p = 0.005), weight control strategies (p = 0.046), anxiety (p = 0.008) and parents’ educational level (father’s educational level: p = 0.017; mother’s educational level: p = 0.001).

In multivariate analysis, the factors associated with the existence of at least one eating disorder were: urban residence (OR (95% CI) = 5.059 (1.75 - 14.65); p = 0.003); year of study (OR (95% CI) = 0.47 (0.28 - 0.79); p = 0.035); type of university attended (private university: OR (95% CI) = 1.63 (1.08 - 2.44); p = 0.019);

parents’ marital status (parents in couple or not) (OR (95% CI) = 1.50 (1.01 - 2.24); p = 0.046); father’s level of education: secondary (OR (CI 95%) = 3.85 (1.96 - 7.54); p < 0.001)/higher (OR (CI 95%) = 2.83 (1.36 - 5.86); p = 0.005); mother’s: secondary (OR (CI 95%) = 0.30 (0.18 - 0.51); p < 0.001)/superior (OR (CI 95%) = 0.31 (0.16 - 0.61); p = 0.001); the presence of doubtful (OR (CI95% = 1.69 (1.11 - 2.58); p = 0.009) or definite (OR (CI 95%) = 1.89 (1.13 - 3.18); p = 0.009) anxiety symptomatology (Table 5).

Age

The average age of the students was 21.21 ± 2.24 years. This average is comparable to the 20.73 years found by Chan et al. in Malaysia in 2020 [15] and the 21.38 years reported by Aina et al. in Nigeria in 2016 [8] , and higher than the 18.7 years in the study conducted by De Matos et al. in Brazil in 2020 [16] , as well as the 16.05 years of Sabry et al. in Egypt in 2020 [17] . On the other hand, Loni et al. in Saudi Arabia and dos Reis et al. in Brazil in 2014 [18] found a higher median age (23.5 years ± 1.8 years and 23.4 years ± 6.13 years respectively). These results may be explained by the disparity between the study populations.

Indeed, De Matos et al. conducted a study exclusively among 1st-year students, while Sabry et al. targeted secondary school pupils.

Gender

The male gender was predominant in this work. Indeed, 55.52% of respondents were male, compared with 44.42% who were female. This male-dominated distribution reflects the unequal gender distribution within Benin’s universities. In fact, during the 2020-2021 academic year, only 30.2% of students were female, compared with almost 70% male. A similar distribution was found by Fadipe et al. in Nigeria in 2017 (55.6%) [19] , and Pengpid et al. in Bangladesh in 2015 (56.6%) [6] .

On the other hand, Tavolacci et al. in France in 2020 (36.6% men) [5] , Tayfur et al. in Turkey in 2020 (26.6% men) and dos Reis et al. in Brazil in 2014 [18] (23.0% men), found a preponderance of women. These differences can be explained by the specific demographic characteristics of each of these areas.

Around one in thirty of the students surveyed had a personal history of mental illness, and 4.61% had a history of surgery. On the other hand, very few revealed any criminal record. The fact that mental illness is attributed to supernatural forces and that the stigma associated with it is a reason for low attendance at treatment centers, could explain this rate. According to a study conducted by Tognon et al. in Benin in 2020, 54.22% of adults in northern Benin had a mental disorder [20] .

Regarding the use of psychoactive substances, this study revealed that alcohol was the most widely consumed (35.09%), followed by tobacco (1.15%) and drugs (0.33%). According to the results of the WHO’s “STEPSwise” surveys, carried out in Benin in 2015, 30% of men aged 18 to 29 consumed alcohol. As for drug use, 1.97 of students surveyed had used cannabis and 19.1% reported having taken methamphetamines or amphetamines at least once in their lives according to Tee et al. in Benin in 2022 [21] . These proportions are much lower than those reported by De Matos et al. in Brazil in 2020 [16] . According to this study, 13.8% of students smoked and 59% drank alcohol [16] . On the other hand, drug use was found among 3.1% of students in the survey carried out by Tavolacci et al. in France in 2020 [5] . This discrepancy in results reflects the high prevalence of psychoactive substance use among Western students.

Most of the respondents’ parents had attended school up to secondary level. However, there were more educated fathers than mothers (10.71% versus 27.68%). These figures are significantly higher than those found by Alwosaifer et al. in Saudi Arabia in 2018 (1.1% vs. 2.9%) [22] . This difference reflects the literacy rate in Benin, which according to the UNESCO Institute for Statistics will stand at just 46% in 2021, much lower than the 99% found by the same institution in Saudi Arabia in 2020.

The majority of those surveyed came from monogamous families (61.12%). Indeed, according to the fifth Benin Demographic and Health Survey (EDSB-V), 62% of women aged 15 to 49 were in monogamous unions and 38% were in polygamous unions. A similar trend was found by Aina et al. in Nigeria in 2016, but with a higher proportion 84% [8] .

Very few respondents in this study showed symptoms of depression (4.45%). Anxiety was found in 14.17% of participants. In a study carried out in a rural community in northern Benin, Tognon et al. reported a prevalence of 32.34% for major depressive episode and 9.62% for generalized anxiety.

Other authors have found prevalences similar to those in this study. These include Tavolacci et al. in France in 2020, who reported a prevalence of depression and anxiety of 3.3% and 15.5% respectively [5] . However, higher values were noted by De Matos et al. in Brazil in 2020 (35.2% and 46.8% respectively for depression and anxiety) [16] . The use of different diagnostic scales may explain this wide disparity in results. In the present study, the HAD scale was used to diagnose depression and anxiety, while Tavolacci et al. used the Perceived Stress Scale (PSS-10) and the Beck Depression Inventory to diagnose anxiety and depression respectively.

In this study, 33.94% of students used weight control methods. Of these, the most commonly used was sport, followed by fasting. Very few used laxatives, appetite suppressants or other medications for this purpose. This proportion is higher than that reported by Attouche et al. in Morocco in 2021 (24.7%), and Eya et al. in Tunisia in 2022 (18%) [23] , but much lower than the 65% found by Charfi et al. in Tunisia in 2015 [24] . The predominant means varied from adopting a diet (Attouche et al.; Eya et al.) to voluntarily skipping meals (Charfi et al.).

With regard to Body Mass Index (BMI), in the general population, students had an average BMI of 21.56 ± 2.53 kg/m², with extremes of 14.69 and 40.17 kg/m². A normal BMI was found in 78.91% of students, 10.21% were overweight, 9.55% were thin, while only 1.31% were obese. According to the results of surveys on risk factors for non-communicable diseases using the WHO’s “STEPSwise” approach, carried out in Benin in 2015, 75.5% of subjects aged between 18 and 29 had a normal BMI. A similar average was found by Aina et al. in Nigeria in 2016 [8] . Similarly, they report a similar predominance of students with a normal BMI in the same proportions (72.5%). Other authors, including Chan et al. in Malaysia in 2020 (46.7%), and El-Bagoury et al. in Egypt in 2017 (50.8%) [25] , noted this predominance, but with lower proportions. This could be explained by the different geographical and socio-cultural contexts of the samples studied.

At the end of the present study, 53.21% of students had tested positive for at least one eating disorder. This relatively high prevalence may be explained by the number of disorders studied at the same time. The prevalences found by the various authors vary. This great variability may be attributable to the psychometric instruments used, the methodology, and the characteristics of the sample studied.

After multivariate analysis, the factors associated with the risk of presenting at least one eating disorder among students in the town of Parakou were: urban residence, type of university attended, parents’ marital status, father’s level of education, mother’s level of education, presence of doubtful or definite anxiety symptoms.

Several authors have reported significant associations between living environment and eating disorders. In this study, urban living environment showed an association with the presence of eating disorders. A similar association was found by Bilali et al. in Greece in 2010 and by Kilani in Jordan in 2017 [26] , whose results suggested that students in urban areas were at significantly higher risk of developing an eating disorder than students in rural areas.

In Italy and the Netherlands, Preti et al. [27] and Hoek et al. reported that bulimia nervosa was more frequently observed in urban populations than in rural or suburban populations [28] . Similarly, in Japan, Kuboki et al. noted that subjects living in urban areas had a higher prevalence of anorexia nervosa [29] . These disparities could be explained by the distinct lifestyles observed within urban populations, as well as the diverse social, environmental, economic and cultural contexts that characterize them.

The marital status of the respondents’ parents was found to be a risk factor for eating disorders. Indeed, the risk of developing at least one eating disorder is multiplied by 1.5 in students whose parents live together, compared with those whose parents live apart. Other authors have established this relationship. Doksat et al. in Turkey in 2017 had reported a relationship between adolescent girls at risk of eating disorders and parental marital status [30] . However, in contrast to the results of this work, the parents’ divorced status was the one incriminated.

Fadipe et al. in Nigeria in 2017 found a link between anxiety and the risk of eating disorders [19] . Safi et al. in Tunisia in 2022 [23] , and Charfi et al. in Tunisia in 2015 [24] , reported a similar association. This association could be explained by the fact that the experience of anxiety makes the individual more vulnerable to the development of eating disorders, on the one hand, and by the coping strategy or even the protective and defensive mechanism against anxiety that eating disorders could represent, on the other. Furthermore, as with anxiety, some authors (Attouche et al., Tavolacci et al. [5] ) have shown a significant association between depression and eating disorders, but in the present series, only anxiety showed such an association.

In this study, the non-probability sampling method was used for convenience. The sample size was calculated using Schwartz’s formula, based on the average prevalence found in Nigeria, a sub-Saharan African country bordering Benin Republic. The minimum sample size was 219, but in the end 607 subjects were surveyed. Recognized and validated scales: the Eating Attitudes Test-26 and Bulimia Inventory Test Edinburgh were used to screen for anorexia nervosa, bulimic hyperphagia and bulimia nervosa. For the screening of pica, a questionnaire based on the DSM-V diagnostic criteria was used.

Despite the strengths of this study, some of its limitations and biases should be noted. The study was based primarily on the subjective statements of the subjects. These were a function of the subjects’ understanding of the theme and the various questions on the scales used. This aspect of the work raises the problem of the possible over- or under-estimation of questions by certain targets, even though the explanations provided by the interviewers helped to reduce this bias.

The other aspect is that of the scales themselves. Developed for the most part by English-speaking authors, then translated into French, they all originate from Western civilizations. None of them have been validated in an African context, and are therefore not adapted to the continent’s socio-cultural realities. Nor can these scales be used to establish a precise medical diagnosis, but rather as initial screening methods. Studies should be carried out to provide African authors with screening scales specific to African realities.