Examination of Sex Differences in Energy Availability, Disordered Eating, and Compulsive Exercise Among Male and Female Adolescent Athletes

Abstract

Objective:

The primary aim of this study was to examine sex differences in energy availability (EA) and its relationships with disordered eating, compulsive exercise, and body mass index (BMI) among adolescent athletes.

Design:

Cross-sectional study.

Setting:

University hospital pediatric sports medicine center.

Participants:

Sixty-four participants (61% female) of ages 13 to 18 years, actively participating in at least 1 organized sport.

Main Independent Variable:

Participant sex.

Main Outcome Measures:

Average 7-day EA (kcal/kg FFM/d; calculated using participant-recorded dietary intake and exercise expenditure from a wrist-worn heart rate/activity monitor), Eating Disorder Examination Questionnaire (EDE-Q) score (range 0–6), Compulsive Exercise Test (CET) score (range 0–25), and age- and sex-adjusted BMI percentile.

Results:

There were no significant sex differences in EA (females: 40.37 ± 12.17 kcal/kg FFM/d; males: 35.99 ± 12.43 kcal/kg FFM/d; P = 0.29), EDE-Q (females: 0.68 ± 0.70; males: 0.68 ± 0.83; P = 0.99), or CET scores (females: 11.07 ± 0.44; males: 10.73 ± 0.63; P = 0.66). There were low and insignificant negative correlations between EA and EDE-Q and CET scores for female athletes (EDE-Q: r = −0.22, P = 0.18; CET: r = −0.21, P = 0.09) and male athletes (EDE-Q: r = −0.09, P = 0.66; CET: r = −0.35, P = 0.08). EA and BMI-for-age percentile were inversely correlated in both male (r = −0.451, P = 0.009) and female (r = −0.37, P = 0.02) participants.

Conclusions:

In our sample of adolescent athletes, lower EA occurred in the absence of notable disordered eating or compulsive exercise behaviors, suggesting unintentional underfueling (and/or underreporting of energy intake), without significant sex differences. Low BMI can be an imperfect surrogate marker for low EA. These findings inform risk factors and screening practices for low EA among adolescent athletes.

INTRODUCTION

Relative energy deficiency in sport (REDs) is a syndrome encompassing the physiologic, psychological, and performance impairments that can occur in the setting of low energy availability (EA).¹ Low EA results from a mismatch in dietary energy intake and exercise energy expenditure, leading to inadequate energy to support normal physiologic functioning.¹ The reported prevalence of low EA among different adolescent athlete groups ranges from 6% to 100% with variability between sexes and across different sport types.^2,3 Low EA is classified as either “adaptable” or “problematic”; adaptable low EA is short-term and has little impact on long-term health and performance, while problematic low EA results in more severe and potentially persistent disruption of multiple body systems.¹ Problematic low EA among adolescent athletes can lead to hormonal alterations (eg menstrual dysfunction in female athletes), impaired growth and development, or low bone mineral density, each of which can have long-lasting negative consequences.¹

Low EA can occur in a variety of contexts, including among those with a clinical eating disorder (eg anorexia nervosa), with disordered eating (ie abnormal eating behaviors that do not meet clinical criteria for an eating disorder), or unintentional underfueling (ie lack of nutritional knowledge, limited resources, food insecurity).^1,4 Although disordered eating is more common among female athletes, it is likely underrecognized and underreported in male athletes.^2,3,5 Compulsive exercise—exercise that is aimed at preventing or reducing distress and that the patient feels driven to perform despite interference in daily routines, the presence of medical injury, or lack of enjoyment—can also contribute to low EA.^1,6 Compulsive exercise is common among adolescents with eating disorders, but these data are limited primarily to female patients and are often not specific to the athlete population.⁷ The few studies examining the relationships of EA, disordered eating, and compulsive exercise in athletes are limited to adult samples.^8,9 Although low EA is associated with disordered eating and compulsive exercise, it can occur in the absence of eating or exercise psychopathology, which is an important consideration regarding screening/risk assessment for low EA.¹

In addition to examining eating and exercise behaviors, low body mass index (BMI) is often considered a risk factor for low EA. However, athletes with low EA may present with a BMI in the normal/healthy range, indicating this approach can be limited and may vary by sex.¹⁰ Identification of low EA and its associations with disordered eating, compulsive exercise, and BMI, as well as sex differences, are important concepts to understand to build risk profiles and inform screening practices for REDs.

The primary aim of this study was to examine sex differences in EA and the relationship between EA and disordered eating, compulsive exercise, and BMI, respectively, among adolescent athletes. We hypothesized that female athletes would demonstrate lower EA and more disordered eating and compulsive exercise behaviors than the male athletes. In addition, we hypothesized that lower EA would be associated with lower BMI and more severe disordered eating and compulsive exercise behaviors in both groups, with female athletes demonstrating stronger correlations.

METHODS

Study Participants

We conducted a cross-sectional study of adolescent athletes recruited from multiple clinics affiliated with the same sports medicine center, local high schools, club sport organizations, and community events. Local institutional review board approval was obtained before study commencement. Eligible participants included males and females (based on self-reported sex) of ages 13 to 18 years, who were participating in at least 1 organized sports team, including a high school team, club team, or recreation center league at the time of data collection. Exclusion criteria included those with an active or history of a clinical eating disorder diagnosed by a healthcare professional, athletes currently under the care of a dietitian, or athletes who were injured and not actively participating in their sport. Written informed assent/consent was obtained from participants and their legal guardian if younger than 18 years.

Study Procedures

At the initial visit, all participants completed a series of questionnaires including a health history (inquiring about sport participation, injury history, and menstrual history in females), the Eating Disorder Examination Questionnaire (EDE-Q),¹¹ and the Compulsive Exercise Test (CET).¹² Menstrual history included age of menarche, number of menstrual periods in the prior 12 months, and current use of hormonal contraceptives. Sex was determined based on participant self-report as “male” or “female.” Gender identity information was not collected. Height and weight were obtained, and BMI was calculated. Sex-specific BMI-for-age percentile for each participant was obtained using the US Centers for Disease Control and Prevention BMI calculator.¹³ All participants underwent dual-energy x-ray absorptiometry (DXA) scan of the total body to obtain body composition data. On completion of the assessment, a trained research professional provided participants with a wrist-worn heart rate/activity monitor (Fitbit Inspire 2 device, FitBit, San Francisco, CA) and instructions on how to record each exercise session for a 1-week monitoring period. In addition, the research professional instructed participants on how to download and set up a nutrition tracking application on their smartphone (MyFitnessPal, Inc.) to record dietary intake during the same 1-week period (Figure 1). Before leaving the initial visit, the research professional verified that physical activity and dietary intake applications were downloaded and functioning properly. Participants were provided with a paper log to record dietary intake and exercise in the event that the activity monitor and/or nutrition phone application did not work or was not available for data collection, or per participant preference.

Figure 1.

Study protocol.

Over the course of 7 days after the in-person assessment, each participant recorded their exercise sessions and dietary intake. The research professional monitored the applications to ensure participants were recording activity and food intake and sent reminders if data were missing. After this 1-week period, the participant and a parent met for a virtual interview with either the research professional or the consulting sports dietitian to review the exercise and nutrition data and ensure accuracy of the records. After verification of all food and exercise entries, daily caloric intake and exercise expenditure were analyzed and subsequently used in energy availability calculations (Figure 1). Heart rate data from the wrist-worn activity monitor were used to determine physical activity intensity, with light activity as 50% to 69% of age-predicted maximum heart rate, and moderate-to-vigorous activity as age-predicted maximum heart rate of 70% or higher, a method previously used.^14,15

Outcomes Measures

Disordered Eating

The EDE-Q is a 36-item self-report questionnaire derived from the Eating Disorder Examination, an interview-based evaluation tool used to diagnose eating disorders.¹¹ It assesses the frequency of eating disorder attitudes in the prior 28 days on a 7-point scale from “No days” to “Every day,” and the frequency of eating disorder behaviors (such as binging and purging) by self-reported number of episodes in the prior 28 days. There are 4 subscales: restraint, eating concern, shape concern, and weight concern. The global score is calculated by summing the subscale scores and dividing by 4, and higher scores are indicative of more severe disordered eating behaviors (range 0–6).¹⁶ There are published normative scores for adolescent girls and boys; subscale and global scores of 4 or higher indicate higher risk of clinical severity of eating disorder behaviors.^17,18

Compulsive Exercise

The CET is a 24-item self-report questionnaire developed to assess the core factors of excessive exercise, with 5 subscales including avoidance and rule-driven behavior, weight-control exercise, mood improvement, lack of exercise enjoyment, and exercise rigidity.^12,19 It asks participants to report how much a statement relates to them on a 6-point scale from “Never” to “Always true.” Subscale scores are obtained by taking the average of the sum of points within each subscale, and the total score is calculated as the sum of the mean score for each subscale, with higher scores reflecting more severe compulsive exercise behaviors (range 0–25).¹²

Energy Availability

Seven-day EA was calculated for each participant using the equation:

$$EA=\frac{\text{energy intake (kcal)}-\text{exercise energy expenditure (kcal)}}{\text{fat}-\text{free mass (kg)}}$$

Energy intake was derived from food intake recorded in the nutrition tracking phone application and verified by the paper log and the follow-up interview with the research team. The research team interviewed participants and/or their parents to improve accuracy of the food logs using the United States Department of Agriculture Five-Step Multiple Pass method and portion size guides to validate any unmeasured portions.²⁰ Exercise energy expenditure was quantified using the exercise output data from the wrist-worn heart rate/activity monitor. For exercise sessions not captured by the wearable device (ie the participant was required to remove the device for competition), metabolic equivalents were calculated based on exercise sessions recorded in the paper log. Fat-free mass (FFM) was obtained from the total body DXA scan output. Daily EA was calculated and averaged across the 7-day monitoring period to obtain average EA values during the week. EA was used as a continuous variable in subsequent analyses, and categorized using proposed clinical cutoffs: low (<30 kcal/kg FFM), subclinical (30–40 kcal/kg FFM/day in males, 30–45 kcal/kg FFM/day in females), and optimal (>40 kcal/kg FFM/day in males, > 45 kcal/kg FFM/day in females).^21,22 It should be noted that while these cutoffs are widely used in the literature, there is individual variability in EA thresholds at which physiologic impairments occur (see Discussion section).

Grouping Variables

Participant Sex

Participants were grouped by male or female based on selfreported sex on the intake questionnaire.

Statistical Analysis

Data are presented as mean (SD) for continuous variables and the number within group (corresponding percentage) for categorical variables. We assessed normality distributions for our outcome variables using Shapiro–Wilks test. We confirmed data normality, and the use of parametric statistics was, therefore, appropriate. We compared continuous and categorical demographic characteristic variables between female and male participants using independent samples t-tests and χ² or Fisher exact tests, respectively. For hypothesis 1, we compared EA, EDE-Q scores, and CET scores between female and male participants using independent samples t-tests and Cohen d effect size calculations. For hypothesis 2, we assessed the association between EA with BMI-for-age percentile, EDE-Q, and CET scores using Pearson R correlations, both for the overall sample and separately between female and male participants. Correlations were considered low <0.39, moderate (±) 0.4 to 0.59, moderately high (±) 0.60 to 0.79, and ≥(±) 0.80 high. Statistical significance was set at α = 0.05. All analyses were 2-sided and performed using Stata Statistical Software: Version 18 (StataCorp, LLC, College Station, TX).

Ethical Considerations

The participant and parents/guardians of any participant younger than 18 years were notified by either the study principal investigator or the study dietitian if any of the following criteria were met on the EDE-Q: (1) > 80th percentile for published normative subscale or global scores, (2) 4 or more recurrent episodes of binge eating, self-induced vomiting, laxative use, or compulsive exercise in the prior 28 days. They were provided with resources and the recommendation to seek evaluation with their primary care physician because of increased risk for an eating disorder/disordered eating.

RESULTS

A total of 65 participants enrolled in the study, of whom 64 completed each protocol element (98% retention rate; 60% female participants, average age = 15.5 ± 1.5 years). The participants were active in a variety of different primary sports (Table 1). Male participants were taller and heavier than female participants, but there were no observed differences between sexes in absolute BMI, BMI-for-age percentile, or history of bony stress injuries (Table 1). Table 1 outlines the exercise data (daily step count, minutes of light and moderate physical activity, average daily exercise energy expenditure), and nutrition and body composition data (average daily energy intake, body fat percentage, fat-free mass). Among the female participants in the study, N = 3 (8%) reported menarche at age 15 years or older, N = 11 (28%) reported 9 or fewer periods in the prior 12 months, and N = 8 (21%) reported currently taking hormonal contraception.

TABLE 1.

Demographics, Health History, Exercise, Nutrition, and Body Composition Characteristics of Male and Female Athlete Groups

Variable	Male Participants (n = 26)*	Female Participants (n = 39)*	P
Age (yr)	15.3 (1.3)	15.6 (1.6)	0.53
Race
American Indian or Alaska Native Asian	2 (8%)	0 (0%)	0.16
Black or African American	1 (4%)	1 (3%)	>0.99
Native Hawaiian or Pacific Islander	0 (0%)	1 (3%)	>0.99
White	24 (92%)	34 (87%)	0.69
Unknown or not reported	0 (0%)	4 (10%)	0.14
Ethnicity
Hispanic or Latino/a	4 (15%)	9 (23%)	0.77
Not Hispanic or Latino/a	21 (81%)	27 (93%)
Unknown or not reported	1 (4%)	3 (8%)
Height (cm)	172.6 (9.0)	164.4 (6.5)	<0.001†
Weight (kg)	61.2 (13.7)	54.9 (8.8)	0.03†
BMI (kg/m2)	20.7 (3.7)	20.4 (2.9)	0.70
Sex-specific BMI-for-age percentile	47.0 (27.2)	45.7 (23.5)	0.84
Primary sport type	Soccer: 9 (35%) Football: 3 (12%) Swimming/diving: 3 (12%) Baseball: 3 (12%) Basketball: 2 (8%) Lacrosse: 2 (8%) Track and field: 2 (8%) Cross country: 1 (4%) None listed: 1 (4%)	Soccer: 16 (41%) Volleyball: 6 (15%) Cross country: 6 (15%) Dance: 3 (8%) Track and field: 2 (5%) None listed: 2 (5%) Swimming/diving: 1 (3%) Softball: 1 (3%) Tennis: 1 (3%) Ice hockey: 1 (3%)	—
History of bone stress injury	4 (15%)	8 (21%)	0.75
Moderate-to-vigorous physical activity (min/d)	71.4 (35.2)	61.8 (37.4)	0.32
Light physical activity (min/d)	227.7 (60.9)	263.5 (51.6)	0.02†
Step count (mean steps/d)	10 429 (3694)	10 631 (3799)	0.84
Mean daily exercise energy expenditure (kcal/d)	562.9 (296.1)	435.0 (228.2)	0.06
Body fat percentage	20.6 (6.8)%	28.0 (6.2)%	<0.001
Fat-free mass (kg)	46.5 (13.9)	38.7 (5.4)	0.002
Mean daily energy intake (mean calories consumed/d)	2277 (437)	2001 (405)	0.01

Continuous variables are represented as mean (SD), and categorical variables are presented as n (%).

^*N = 65 participants completed all questionnaires, but only 64 participants (25 males, 39 females) completed EA data collection.

^†P < 0.05.

There were no significant differences between female and male participants for average EA during the week-long monitoring period (females: 40.37 ± 12.17; males: 35.99 ± 12.43; Figure 2A), EDE-Q scores (females: 0.68 ± 0.70; males: 0.68 ± 0.83; Figure 2B), or CET scores (females: 11.07 ± 0.44; males: 10.73 ± 0.63; Figure 2C). When EA was categorized based on clinical cutoffs, there were no proportional differences between female and male participants for those who were deemed to have low (21% vs 27%), subclinical (49% vs 23%), or optimal (31% vs 50%; P = 0.11) EA.

Figure 2.

Box and whisker plots describing the comparison between female and male participants on measures of energy availability, disordered eating (EDE-Q score), and compulsive exercise behaviors (CET score). The median is the solid line within the box, the box maximum and minimum represent the interquartile range, the whiskers represent the range, and dots represent outliers.

Among the entire study sample, there was a moderate and significant inverse relationship between EA with BMI-for-age percentile (Figure 3A), a low/nonsignificant relationship between EA and EDE-Q scores (Figure 3B), and a low but significant relationship between EA and CET scores (Figure 3C). When stratified by sex, male participants demonstrated a moderate and significant inverse correlation, and females demonstrated a low and significant inverse correlation, between EA and BMI-for-age percentile (Figure 3D). There were low and nonsignificant relationships between EA with EDE-Q (Figure 3E) and CET (Figure 3F) scores for both female and male participants.

Figure 3.

The association between energy availability with BMI-for-age percentile, EDE-Q scores, and CET scores, for the overall study sample (A–C) and by sex (D–F).

DISCUSSION

The data demonstrated that 27% of the male athletes and 21% of the female athletes in our sample had low EA. This falls within the wide range of reported prevalence of low EA (using the threshold of < 30 kcal/kg FFM/day) from 6% to 100% in various adolescent and young adult athlete groups.^2,3 A major challenge with EA assessment in the field is the concern for underreporting of dietary/energy intake, which could lead to inaccurately low calculated EA.²³ It is important to consider the possibility of energy intake misreporting when interpreting the results of this study. Calculated estimated energy requirements using the National Academies’ estimated energy requirements equations²⁴ are 3125 kcal/d for male adolescents and 2420 kcal/d for female adolescents in our study. Comparing this with the reported average 2277 kcal/d for the male participants and 2001 kcal/d for the female participants in our study, these findings suggest that our sample was underreporting their energy intake. It is also possible that our sample was underfueling for their energy needs, and we put multiple measures in place to optimize and verify accuracy of the dietary records, as well as the exercise data. Our results should be interpreted with acknowledgement of the limitations of measuring EA outside of the laboratory.

Historically, EA < 30 kcal/kg FFM/day is considered the threshold at which metabolic and hormonal impairments (particularly menstrual dysfunction) occur in adult women, with normal physiologic functioning achieved at 45 kcal/kg FFM/day.¹ However, it has become apparent that a single EA threshold of 30 kcal/kg FFM/day is not applicable to all female athletes because there is individual variability in the EA level at which physiologic impairments occur.²¹ In addition, the more limited studies on male athletes demonstrate that they may be able to tolerate a lower threshold of EA than female athletes, with the recent Male Athlete Triad Coalition Consensus Statement proposing low EA defined as <15 kcal/kg FFM/day.²² These thresholds are derived from studies in adults and may not be applicable to adolescent athletes who have unique developmental energy demands as they grow and progress through puberty.²⁵ Therefore, our findings as categorized into low, subclinical, and optimal EA should be interpreted with the understanding that there is individualized variability in EA thresholds, while still providing useful insight into the distribution of EA in our study sample. Although there are likely individualized and “sliding-scale” targets of EA up to optimal thresholds, it is concerning that most participants (62%) were below accepted optimal EA levels (low or subclinical).

In our sample, low/subclinical EA seemed to occur without notable disordered eating, with no significant differences observed between the male and female athletes. This is an important consideration because while disordered eating is a risk factor for low EA, low EA can occur in the absence of disordered eating because of unintentional underfueling.^1,2 When compared with normative values for EDE-Q scores in adolescent girls,¹⁷ the average global score for the female athletes in our sample was close to the 35th percentile rank. For the male athletes, their average global score was close to the 70th percentile rank when compared with normative values published for adolescent boys.¹⁸ Although studies show female athletes are at greater risk for disordered eating compared to male athletes, disordered eating behaviors in male athletes are less studied and may be underreported.²⁶ A recent scoping review reported a 5% to 50% prevalence rate of disordered eating behaviors in elite male athletes of all ages, with higher prevalence in weight-sensitive sports.⁵ Using the reported normative values for adolescent boys, we observed higher than average global EDE-Q scores in our sample of male athletes, with the majority playing team sports that would be considered less weight sensitive. Regardless, there was not a strong correlation between eating disorder behaviors and EA in our study sample, suggesting that disordered eating behavior was not a primary driver of lower EA.

We did observe a significant but low correlation between EA and compulsive exercise behaviors among the entire sample, but significance did not persist when stratified by sex. Much of the research to date on compulsive exercise in adolescents focuses on the eating disorder population, because compulsive exercise has been linked to eating psychopathology.^7,12,27 The relationship between EA and compulsive exercise, particularly in adolescent athletes who do not otherwise have an underlying eating disorder diagnosis, is not well described.⁴ In a recent study examining adult athletes and disordered eating, exercise dependence (as measured by the Exercise Dependence Scale), and risk of low EA (determined by self-report questionnaires), exercise dependence only increased the risk of low EA when it co-occurred with disordered eating in both male and female athletes.⁸ In a sample of adult male endurance athletes, exercise dependence was significantly associated with severe negative energy balance (derived from EA calculations), and with more disordered eating behaviors.⁹ The differences in findings from our data compared with these studies may be related to differences in methodology and study population (eg adult vs adolescent athletes, different sport types). The major takeaway from our data is that similar to the lack of a significant relationship between EA and disordered eating behaviors, low/subclinical EA seemed to occur without notable compulsive exercise behaviors in both the male and female athletes, with no significant differences in compulsive exercise behaviors between sexes.

BMI and age- and sex-adjusted BMI percentiles (in the adolescent population) are used to assess for risk of low EA.^28,29 We found that lower EA was associated with higher BMI-for-age percentiles. There are inherent limitations to using BMI as a measure of health, because it does not distinguish between lean and fat mass.³⁰ A young athlete with more lean mass may have a higher BMI and be incorrectly classified as overweight or obese using BMI-for-age percentile criteria.³¹ As it pertains to our study, those with higher BMI could include those with more lean mass and less body fat associated with lower EA. Another consideration is that those with higher BMI, if reflective of more adiposity, may be engaging in behaviors to limit their food intake in an attempt to control weight, thus leading to lower EA. We did find that the average EDE-Q score for the male athletes was higher than average normative values; however, the lack of a significant correlation between EA and EDE-Q in the male group makes this theory less likely. Nonetheless, it is important to recognize that low EA can still occur without low BMI, and BMI should not be used as the sole screener for consideration of risk of low EA.

Limitations

The first limitation is our smaller sample size, and thus, the possibility that we were underpowered to detect significant differences between groups. Second, our data were collected through self-report questionnaires, which are subject to several biases including recall, social desirability, and measurement error bias.³² In addition, the EDE-Q is not specific to the athlete population but it has been widely used and is generally accepted as an adequate measure of eating psychopathology in athletic populations.³³ One benefit of using the EDE-Q, which guided the decision to use it in this study, is that there are established normative values in both male and female adolescents.^17,18 The CET is also limited in that it is not specific to the athlete population; the 5-factor structure was found to have moderately good fit in an athlete sample, but an alternative 3-factor structure may be a better fit.³⁴

Another important limitation to address is the potential inaccuracies of the methodology used to measure EA through self-recorded dietary intake and by wearable devices for exercise data collection.³⁵ We attempted to address these possible inaccuracies by having participants complete both electronic and paper logs, sending reminders to participants to record their data as monitored by the research team, including a consulting sports dietitian on the research team to verify EA data, and conducting follow-up interviews with participants and their parents to ensure completeness and optimize accuracy of the data. The EA results should also be interpreted in the context of the measurements occurring in a single, 1-week snapshot.³⁵ Finally, our study participants were from a single geographic area, so the results may not be generalizable to other populations. There are several strengths of our study, including the success in obtaining 7 consecutive days of dietary and exercise data in an adolescent population, inclusive of both male and female athletes, from a variety of sports, and the assessment of concurrent disordered eating and compulsive exercise behaviors, all of which are not often included in prior studies of low EA in athletes.^2,3

In summary, we identified a large proportion of both male and female athletes in our adolescent sample demonstrated low or subclinical EA, but there were no significant sex differences in measured EA. Lower EA seemed to occur in the absence of notable disordered eating or compulsive exercise behaviors, suggesting unintentional underfueling, among both athlete groups. Finally, low BMI is an imperfect surrogate marker for low EA, for which we found higher BMI-or-age percentile to be associated with lower EA in our sample. These findings inform risk factors and screening practices for low EA among adolescent athletes. Specifically, clinicians should not rule out the possibility of low EA based on the absence of disordered eating, compulsive exercise behaviors, or low BMI, and should screen male and female athletes. Screening tools such as the Low Energy Availability in Females Questionnaire (LEAF-Q)³⁶ and Low Energy Availability in Males Questionnaire (LEAM-Q)³⁷ could be used, which assess for signs and symptoms of low EA including injury history, gastrointestinal symptoms, and reproductive hormonal health.