Prevalence and Association of Exercise Dependence and Eating Disorder Risk in Collegiate Student-Athletes

The study was part of a larger investigation with a cross-sectional design. The data included were descriptive in nature and based on a web-based survey developed from previously validated instruments for quantitative analysis examining EXD and the risk for EDs. This research was approved by the institutional review board. All recruits completed an online informed consent form with the opportunity to decline participation and withdraw from the study at any time.

Student-athletes from National Collegiate Athletic Association (NCAA) Divisions I and II were invited to complete the survey. A total of 1885 (age = 19.8 ± 1.4 years; females = 69.6%, n = 1312; males = 30.4%, n = 573) individuals from 40 institutions participated in the study. Volunteers were included if they were at least 18 years of age and an active member of a team during the completion of the survey. No exclusion criteria were used.

Basic personal and demographic information, including age, sex, race or ethnicity, self-reported height and weight (current, lowest, highest, and ideal), and academic status were obtained. We classified academic status as freshman (first-year students), sophomore (second-year students), junior (third-year students), senior (fourth-year students), or fifth-year or graduate student. Student-athletes reported their primary sport, and these data were further organized by sport category using the Sundgot-Borgen²¹  classification: endurance (eg, cross- country, swimming, and track middle and long distance), aesthetic (eg, cheerleading, dance, diving, and equestrian), power (eg, football, track and field, and track sprints), nonlean events (discus, hammer, and shot put), ball and team (eg, baseball, basketball, softball, soccer, volleyball, and beach volleyball), and technical sports (eg, golf, tennis, and track and field lean events [high jump and javelin]).

The Exercise Dependence Scale-21 (EDS-21) was used to assess EXD. Originally, the EDS was developed with 36 items and then revised to 27 with an internal consistency of 0.93.²²  The instrument was based on the Diagnostic and Statistical Manual of Mental Disorders-IV²³  criteria for substance dependence. It consists of the following 7 subscales and was reduced to 21 items: tolerance (α = .78), withdrawal (α = .93), continuance (α = .89), lack of control (α = .82), reduction in other activities (α = .67), time (α = .88), and intention effects (α = .92). The 21 items were scored on a 6-point Likert scale (1 = never to 6 = always), and scores were calculated as the sum of all items. Higher scores indicate more characteristics of EXD, and the tool further differentiates among (1) the risk for EXD, (2) nondependent symptomatic, and (3) nondependent asymptomatic. The EDS-21 specifies if participants present with evidence of (1) physiological dependence (ie, evidence of tolerance and withdrawal) or (2) no physiological dependence (ie, no evidence of tolerance or withdrawal).

Student-athletes completed the Eating Attitudes Test-26 (EAT-26), a self-administered screening tool that allows clinicians to identify characteristics and behaviors associated with EDs. Originally a 40-item tool, the EAT-26 now contains 26 items scored on a Likert scale (always = 3, usually = 2, often = 1, sometimes = 0, rarely = 0, never = 0) with 1 item that is reverse scored; the intercorrelations suggest the EAT-26 is highly predictive of the total EAT-40 (r = 0.98).²⁴  Five items focus on pathogenic behaviors (ie, binge eating; vomiting; use of laxatives, diet pills or diuretics; excessive exercise; or loss of ≥9.07 kg [20 lbs] in the last 6 months) to control or lose weight, 4 items are scored on a Likert scale based on time frames, and the final question is answered either yes or no. Scores can range from 0 to 78 across the following 3 subscales: dieting, bulimia, and food preoccupation or oral control. Participants with scores of ≥20 who are engaging in ≥1 pathogenic behavior to control or lose weight or both were defined as at risk for EDs, following the guidelines provided by Garner et al.²⁴ 

After receiving approval of the study from the institutional review board, we used a snowball sampling method to recruit volunteers. Athletic trainers at NCAA Division I and II institutions were contacted via email and asked to forward the invitation email and web-based survey link to their student-athletes. Student-athletes were able to access the web-based survey via SurveyMonkey for 1 month, and the athletic trainers received 3 reminder emails at 10-day intervals asking them to disseminate the survey to their student-athletes.

All statistical analyses were carried out using SPSS (version 28; IBM Corp) with a significance level of P < .05. We carried out 2 a priori power analyses using G*Power statistical software (version 3.1.9.4; Heinrich-Heine Universität Dusseldorf). The first analysis was for sex; applying χ² tests, α of .05, and a moderate effect size (0.3), our power calculation indicated a sample of 440 was needed (females = 220, males= 220) for estimated power of 0.95. For the sport category, using χ² tests and a large effect size (0.4), our power calculation indicated a sample size of 132 participants (females = 66, males= 66) was needed for estimated power of 0.90. All demographic information (eg, age, sex, height, weight, body mass index, and academic status), EDS-21 scores, EAT-26 scores and subscale scores were examined using basic descriptive statistics, including means, SDs, and frequencies. We calculated a 1-way analysis of variance to compare the EDS-21 total score, EAT-26 total score, and subscale scores with sex and sport category. If assumptions for the homogeneity of variances were not met, the Welch F test was computed, and Games-Howell post hoc tests were used to determine the mean differences within each group. Cross-tabulations and χ² analyses were performed to examine the proportion of participants classified as at risk for EXD using the EDS-21 and for EDs using the EAT-26 by sex and sport category; all assumptions for χ² analyses were met. A multiple logistic regression was conducted to identify EXD as a predictor of risk for EDs using the combined total and stratified by sex. The variables for EXD as a predictor were (1) at risk, (2) nondependent symptomatic, and (3) nondependent asymptomatic.

A total of 1950 collegiate student-athletes from 40 NCAA Division I and II institutions began the study and 1885 completed it, yielding a rate of 96.7% (females = 69.6%, n = 1312, age = 19.8 ± 1.4 years; males = 30.4%, n = 573, age = 19.6 ± 1.4 years) and ultimately meeting the estimated power calculation. Student-athletes were engaged in a total of 23 sports, which we categorized via the Sundgot-Borgen guidelines²¹ : endurance (37.3%, n = 704), aesthetic (18.6%, n = 351), power (9.2%, n = 174), ball and team (25.4%, n = 478), and technical (9.4%, n = 178). Academic status was as follows: 29.1% (n = 548) freshmen (first-year students), 26.1% (n = 492) sophomores (second-year students), 23.8% (n = 448) juniors (third-year students), and 21.1% (n = 397) seniors (fourth-year students) and graduate students. Self-reported physical measures are displayed in Table 1.

According to the EDS-21, 4.9% (n = 92; females = 63, males = 29) of student-athletes were at risk for EXD, with differences by sex and sport category ([all: χ²_8,1885 values = 99.1, P < .001). The proportions of those at risk for EXD are presented in Table 2. Raw scores for the EDS-21 are given in Table 3. Differences were observed for sport category and EDS-21 total scores (Welch F_1,1880 = 35.6, P < .001). The Games-Howell post hoc test revealed differences between endurance (54.4 ± 21.4) and aesthetic (40.7 ± 16.0; P < .001), ball and team (50.7 ± 19.0; P = .019), and technical (47.7 ± 18.5; P < .001) and between aesthetic (40.7 ± 16.0) and power (50.4 ± 18.2; P < .001), ball and team (50.7 ± 19.0; P < .001), and technical (47.7 ± 18.5; P < .001) sports.

The ED risk across student-athletes is provided in Table 4. Overall, 22.7% (n = 428) of student-athletes presented with a combined risk for EDs (EAT-26 score or use of pathogenic behaviors or both), with differences by sex and sport category [χ²_4,1885 = 10.1, P = .039]. Looking at females specifically, we found differences by sport category (χ²_4,1312 = 10.1, P = .038). Raw scores for the EAT-26 and subscales are shown in Table 5. Differences were demonstrated for sex and the EAT-26 total score (Welch F_1,1885 = 55.2; P < .001), the dieting subscale (Welch F_1,1885 = 97.3; P < .001), and the bulimia subscale (Welch F_1,1885 = 15.4; P < .001). Most sport categories except for power displayed differences by sex and EAT-26 scores individually, including endurance (Welch F_1,1885 = 19.7; P < .001), aesthetic (Welch F_1,1885 = 26.7; P < .001), ball and team (Welch F_1,1885 = 18.9; P < .001), and technical (Welch F_1,1885 = 9.0; P = .003). Moreover, we noted differences between sport category and the dieting subscale (Welch F_1,1885 = 32.5; P = .039) and the bulimia subscale (Welch F_1,1885 = 3.2; P = .012). The Games-Howell post hoc test revealed differences in the dieting subscale and sport category for aesthetic and ball and team (3.5 ± 5.1 versus 2.5 ± 4.5; P = .025) sports and in the bulimia subscale and sport category for power versus aesthetic (1.9 ± 2.6 versus 1.2 ± 2.0; P = .037) and power versus ball and team (1.9 ± 2.6 versus 1.2 ± 1.7; P = .014) sports.

Multiple logistic regressions (Table 6) indicated a significant association between EXD and the ED risk for all athletes. Among student-athletes at risk for EXD, the odds of being at risk for EDs were approximately 5 times higher than among those who were nondependent asymptomatic (odds ratio [OR] = 5.1; 95% CI = 3.237, 8.046), more than 2 times higher than among nondependent-symptomatic athletes (OR = 2.4; 95% CI = 1.5618, 3.7089), and similarly 2 times higher among nondependent-symptomatic versus nondependent-asymptomatic athletes (OR = 2.12, 95% CI = 1.664, 2.702). Among females, the odds of being at risk for ED were 7 times higher for those at risk for EXD than those who were nondependent asymptomatic (OR = 7.2; 95% CI = 4.142, 12.418) or nondependent symptomatic (OR = 3.0; 95% CI = 1.8017, 5.1429); the odds were also higher among those who were nondependent symptomatic than those who were nondependent asymptomatic (OR = 2.4; 95% CI = 1.781, 3.118). Significant associations were found for nondependent-symptomatic versus nondependent-asymptomatic student-athletes (OR = 1.7; 95% CI = 1.013, 2.701).

The health benefits of exercise are important; however, EXD can increase the risk to those who have a risk for EDs. The motives behind increasing exercise loads can be intrinsic or extrinsic: for many athletes, they can be related to their physical appearance, body image, or the belief that a thinner body will improve their performance. The purposes of our study were to assess the prevalence of the risks for EXD and EDs among collegiate student-athletes by sex and sport category and to determine if there was an association between EXD and EDs. In accordance with our hypothesis, the prevalence of EXD and EDs was higher in females than males. Furthermore, participants at risk for EXD had increased odds of being at risk for EDs.

We used the EDS-21 to examine EXD in a sample of collegiate student-athletes across the United States. Our main result was that 4.9% of student-athletes presented with EXD across 5 sport categories and 16 sports, as follows: endurance (eg, cross-country, swimming, and track middle and long distance), aesthetic (eg, cheerleading, dance, diving, and equestrian), power (eg, football, track and field, and track sprints), ball and team (eg, baseball, basketball, soccer, softball, volleyball, and beach volleyball), and technical sports (eg, golf, tennis, and track and field lean events [high jump, javelin]). Moreover, males had a higher prevalence (5.1%) than females (4.8%) but the difference was not statistically significant. Our findings were consistent with those of previous studies (1.44%–18.3%)^25–29  in which researchers assessed EXD in athletes using the EDS-21. However, they differed from the outcomes of more recent studies in which the authors suggested that females engaged in compulsive exercise and presented with a greater EXD prevalence than males, perhaps because females in general felt more comfortable disclosing their problems.^30,31  Furthermore, the mean total score on the EDS-21 (54.4) for our sample size was similar to the scores of other male endurance athletes (54.7)³²  and female inpatients with EDs (55.8).¹¹ 

Comparing sport categories, we noted that our endurance-sport athletes had the highest prevalence of EXD at 8.0% and accounted for 60.9% of all athletes with EXD (n = 56/92). Earlier examinations of endurance athletes typically focused on long-distance runners and ultramarathoners, with prevalences of 1.44% and 3.2%, respectively.^26,27  However, Maselli et al³³  also used the EDS-21 and reported rates as high at 12.9% across endurance athletes (cycling, running, swimming, and triathlon). The higher prevalence may be associated with triathlons, as this competitive sport requires time-intensive training across 3 disciplines, namely, swimming, cycling, and running. Moreover, the second highest prevalence of EXD was in ball and team sports at 4.4%, which was lower than that in previous studies (8.1%–18.3%).^8,28,33  Differences between our findings and those of others can be attributed to the study populations; we looked specifically at collegiate student-athletes who had structured training schedules, whereas other researchers evaluated members of fitness clubs. Additionally, the earlier investigations were conducted in Italy, and training and expectations may have differed. Here, we categorized football, track and field, and track nonlean events (discus, hammer, and shot put) as power sports categories. Yet other authors who explored EXD in power sports looked primarily at bodybuilders and weightlifters. We demonstrated that EXD in our power sport category was 4%, which was significantly lower than that of bodybuilders and weightlifters, which ranged from 13.5% to 15.1%.^25,34  An additional novel finding of our work was the increased EXD prevalence of nondependent-symptomatic athletes. More than half of the student-athletes (53.9%) displayed symptoms of EXD without dependence, which was similar to the result among Hungarian runners in 2019.³⁵  Student-athletes are an important population for implementing EXD prevention strategies as symptoms arise and are recognized.

Using the EAT-26 or pathogenic behavior or both to determine the risk for EDs, we observed an overall prevalence of 22.7%. Previous researchers characterized the ED risk across physically active populations as ranging from 11% to 45%, which is consistent with our findings.^18,36–38  By sex, typically, we detect higher prevalence rates in females than in males; our results were similar, with 25.5% of females at risk compared with 16.4% of males. Males are often understudied in regard to EDs, and a possible explanation is that before the diagnostic criteria were established in the Diagnostic and Statistical Manual of Mental Disorders, loss of menses was a criterion for an ED diagnosis.⁴  Investigators^18,19,37,39  who examined EDs in male athletes identified similar rates ranging from 3.2% to 19.2%.

Moreover, differences for ED risk were evident across sport categories: 23.7%, endurance; 24.8%, aesthetic; 25.3%, power; 17.6%, ball and team; and 25.8%, technical. Interestingly, technical sports (golf, tennis, and track and field lean events [high jump and javelin]) had the highest prevalence. Typically, lean sport categories, such as endurance and aesthetic sports, have an increased risk for EDs. Aesthetic sports emphasize leanness primarily because athletes are being graded or judged on various factors; on the other hand, a common belief is that a lean figure will improve performance in endurance sports. Researchers³⁹  have suggested that athletes in these sports are susceptible to body image dissatisfaction and EDs. However, we should not negate the idea that athletes participating in ball and team, power, and technical sports are also at elevated risk for EDs. Technical sports are commonly individual sports; although they are part of a team, these athletes usually compete individually in their sport. Individual athletes scored higher on the EAT-26 subscales for bulimia and food preoccupation.⁴⁰ 

Furthermore, these athletes were also cognizant of their appearance and weight as compared with team athletes.⁴⁰  Performing individually and receiving all the attention from spectators, including judges and other coaches, may offer valid reasons as to why they may be at risk for EDs or body image dissatisfaction.

Exercise continues to be one of the behaviors often associated with controlling or losing weight. A primary finding by Dalle Grave et al,¹⁰  who assessed inpatients with EDs, was that nearly 50% reported compulsive exercising to control their weight or shape. In our study, individuals with EXD had significantly increased odds for being at risk for EDs compared with both nondependent-asymptomatic and nondependent-symptomatic athletes. Researchers⁴¹  proposed that EXD mediates the relationship between exercise and EDs by 14.3% and that intervening in psychological factors such as EXD may be as beneficial as interventions for patients with EDs who exercise excessively. A positive correlation (r = 0.41) has been established between scores on the EDS-21 and other ED risk assessment tools.³²  We should additionally acknowledge that 53.9% of student-athletes were nondependent symptomatic based on the EDS-21, meaning that although they were not at risk, symptoms were present. This population also had increased odds (5 times higher) of being at risk for EDs compared with those without risk or symptoms of EXD, which was nearly identical to the finding of 5 times higher odds among those at risk for EXD versus nondependent-asymptomatic athletes. Perhaps our focus should also be on this population when observing and preventing behaviors associated with both EXD and EDs.

We must acknowledge several limitations to our work. Both the EAT-26 and EDS-21 have been validated, but they are self-reported measures. Therefore, participants may not be truthful and honest when responding to all questions. The clinical diagnosis of EDs is performed through medical assessment and interviews by mental health providers and physicians. The EAT-26 is a risk assessment tool detailing eating attitudes and behaviors and should not be used to diagnose patients with clinical EDs. Also, our sample was predominately female, which could have altered the prevalence of EXD and the ED risk. Student-athletes are an at-risk population for EXD and EDs; thus, future authors should address the implementation of prevention programs across various levels of collegiate athletics.

The risk of EDs continues to be prevalent across student-athletes, and EXD may be associated with this increased risk, specifically across individuals in endurance sports. Although physical activity has become a public health intervention to improve overall health in populations, excessive exercise can pose additional risk factors to the athletic population. Eating disorders can be affected by EXD, as exercise can be a compensatory behavior for weight loss. However, as athletic trainers working with an at-risk population, we should consider the potential effects of EXD on injuries, especially chronic and bone stress injuries. The literature supports the increased risk for EDs among athletes in sports with a leanness focus, yet we noted that athletes in nonlean sports were also at risk. Particular attention should be given to student-athletes and to females who display signs of EXD; additional education and awareness are needed to minimize the overall risk of EDs in student-athletes. Initiatives associated with nutrition interventions and education programs should be provided to student-athletes as prevention tools. A reduction in stigma, a healthy relationship with food, and a positive image should be the foci of educational programs. Also, we must educate patients on the potential health and performance consequences, specifically illnesses and long-term injuries, associated with increased exercise and abnormal eating behaviors or EDs. Lastly, allied health care professionals, such as athletic trainers, dietitians, and mental health providers, should be supplied with comprehensive and continuing education regarding EXD and EDs so that they can assist athletes, regardless of their sex or sport category. Athletic trainers should consider screening all athletes for disordered eating behaviors and their risk for EDs during preparticipation examinations. Furthermore, those working specifically with endurance athletes (cross-country, swimming, and track and field) can implement screening tools for EXD and educate athletes on adequate recovery and nutrition practices as well as other leisure activities that will not increase the training load already implemented by coaches.