Prevalence of the Female Athlete Triad Syndrome Among High School Athletes FREE

Nearly 15 years have passed since the female athlete triad was first recognized and reported as a syndrome of 3 interrelated serious health disorders (disordered eating, menstrual dysfunction, and low bone mass).¹ Although there have been numerous reports on prevalence estimates of individual components of the triad, only 2 studies^2- 3 have reported the prevalence or directly examined the interrelationships of the triad components concurrently. Lauder et al² examined the simultaneous occurrence of all 3 components in active-duty female soldiers and reported that no soldier exhibited the full female athlete triad. Cobb et al³ provided evidence of the triad's existence among young adult competitive runners by demonstrating interrelationships among disordered eating, menstrual irregularity, and low bone mineral density (BMD) levels. However, they did not report whether any runner met criteria for all 3 components concurrently.

Most reports^3- 8 on the individual components of the triad have been limited to collegiate, young adult, or elite athletes. The reported prevalence estimates of eating disorders or disordered eating among female collegiate and elite athletes range from 15% to 62%.^3- 8 The large variation in prevalence estimates for athletes may be due to different assessment methods and criteria for classifying disordered eating, sport populations studied, training volume, or competition-level differences. In the general adolescent female population, approximately 13% to 20% have reported disordered eating attitudes and 0.4% to 4.0% have reported the practice of pathogenic behaviors, including self-induced vomiting and laxative or diuretic use.⁹

Menstrual dysfunction, including primary or secondary amenorrhea or oligomenorrhea, has been reported more often in athletes, primarily among collegiate athletes.^10- 11 Menstrual dysfunction may be caused by weight fluctuations, insufficient energy availability relative to energy expenditure, and/or stress associated with intense or a high volume of training in the absence of energy imbalance.^10,12 Prevalence estimates of menstrual dysfunction among athletes have ranged from 3.4% to 66.0%.^8,10,13 The wide range in estimates may be due in part to varying criteria used to define amenorrhea and specific sport populations sampled. To our knowledge, the prevalence of menstrual dysfunction among high school athletes has not been reported.

Data on the prevalence of low bone mass in adolescent female athletes are scarce, yet adolescence is a critical time for bone acquisition. Peak bone mass accrual in girls has been linked closely to menarche.¹⁴ Recent literature has demonstrated the positive effect of impact exercise on bone mass in children and adolescents.^15- 16 However, several studies^17- 20 have demonstrated that the benefits on bone conferred by sports participation may be negated in young athletes with menstrual dysfunction. Therefore, adolescents with menstrual dysfunction may miss a key developmental period to maximize bone mineralization.

According to the National Federation of High School Associations, approximately 2.9 million girls competed in high school sports in the United States during the 2003-2004 school year, representing a 3.5-fold increase in participation since Title IX was implemented in 1972, which prohibited sex discrimination in all public education programs. Increased participation and competitiveness, greater pressure by coaches, and a preoccupation with body image that normally exists during adolescence may increase the risk of developing the triad in this population.^21- 22 Currently, limited evidence exists regarding the prevalence of the female athlete triad in high school athletes in the United States. Thus, the primary objectives of our study were to estimate the prevalence of the female athlete triad as a whole and to examine the interrelationships among disordered eating, menstrual irregularity, and low bone density in female high school athletes.

The subject sample consisted of female athletes competing on interscholastic teams from 6 high schools in southern California. Participants were 13 to 18 years old who had had initial onset of menarche or, if not menstruating, were 16 to 18 years old. Data from athletes who reported taking any medications known to affect bone mass were excluded from analysis. Subsequently, only 1 athlete was excluded because of long-term use of a steroid inhaler for asthma. A total of 589 athletes (79.6% response rate) completed questionnaires regarding menstrual history and eating behaviors and were measured, without shoes, for height and weight. One research assistant per 6 to 8 athletes reviewed the Eating Disorder Examination Questionnaire (EDE-Q)²³ response scale and questions deemed more difficult to interpret, defined terms (eg, binge episode), and then remained in the room to assist athletes who requested further clarification as they completed the questionnaires individually. After stratifying by eating behavior (normal vs elevated EDE-Q score) and menstrual status (eumenorrhea vs oligomenorrhea or amenorrhea), approximately 30% of the total sample was selected to receive a BMD scan. Athletes were randomly selected by sport using a weighted selection process according to the number of participants in each sport. The final sample size was 170, including track and field (n = 60), cross-country running (n = 33), soccer (n = 21), softball (n = 16), swimming (n = 13), volleyball (n = 11), tennis (n = 10), and lacrosse (n = 6). The study was approved by San Diego State University's institutional review board.

At the end of the first month of each sport season, each athlete completed the EDE-Q to identify disordered eating attitudes and behaviors.^9,23 The EDE-Q, which is a shorter version of the Eating Disorder Examination²⁴ and designed for self-report, is composed of 4 subscales, weight concern, shape concern, eating concern, and dietary restraint, and a global score, which is the composite mean score of the 4 subscales. Scores ranging from 0 to 6 on a Likert scale correspond to the number of days during the past 4 weeks the respondent has experienced a specific attitude, feeling, or behavior. The EDE-Q has high internal consistency²⁵ and moderate to high concurrent and criterion validity.^9,23,26- 27 In addition to its psychometric properties, we chose the EDE-Q over other widely used questionnaires because it addresses a specific time frame and specifically assesses the frequency of eating behaviors. We evaluated EDE-Q scores as continuous and dichotomous variables. Athletes were classified as having disordered eating if they had a mean score of 4.0 or higher on the weight concern or shape concern subscales or had a mean global score of 4.0 or higher. A cutoff score of 4, which indicates that a specific attitude or behavior was reported on more than half of the past 28 days, was used to define disordered eating because it was shown to be predictive of eating disorders.^9,28 Also, scores of 4 or higher correspond to the 80th to 95th percentiles from reference norms for adolescent girls.⁹ Based on this scoring rubric, athletes were divided into 2 groups: disordered eating (elevated EDE-Q score) or normal EDE-Q score.

Following administration of the EDE-Q, the athletes completed a menstrual status and history questionnaire, which was derived from an athletic preparticipation medical history form developed to screen for the presence of female athlete triad components.²⁹ The criteria for classifying athletes with menstrual irregularity were as follows: primary amenorrhea (no onset of menses by the age of 16 years), secondary amenorrhea (cessation of menstrual cycles for ≥3 consecutive months in the past year), or oligomenorrhea (menstrual cycles occurring at intervals >35 days after onset of menses by the age of 16 years).²² For analyses purposes, girls who met any of these criteria were combined into a single (oligomenorrheic/amenorrheic) group and compared with girls with normal menses (eumenorrheic).

At 2 to 4 weeks after the completion of the questionnaires, bone mineral density at the spine (L1-L4), proximal femur, and total body and body composition (percentage of fat and lean tissue mass) were assessed by dual-energy x-ray absorptiometry (DXA) using a densitometer (Lunar DPX-NT; Lunar/GE Corp, Madison, Wis). Quality assurance tests were performed each morning of testing. The coefficient of variation in BMD in our laboratory was 0.6% for the total hip, 1.2% for the spine (L1-L4), and 0.99% for total body.

Currently, no consensus exists for determining osteopenia or osteoporosis in children and adolescents. Therefore, we used the World Health Organization (WHO)³⁰ and International Society for Clinical Densitometry (ISCD)³¹ criteria to define low bone mass. Girls were categorized as having low bone mass for their age if their values for the spine or total body were 1 SD (WHO) or 2 SDs (ISCD) or more below the age-matched, sex-specific reference data from the Lunar/GE Corp pediatric database (Z-score of ≤−1 or ≤−2). The Z-scores are currently not available for the hip in children.

Outcome measures included disordered eating, menstrual status, and BMD of the spine (L1-L4), proximal femur (total hip, femoral neck, and trochanter), and total body. We reported menstrual irregularity as a dichotomous variable (eumenorrhea vs oligomenorrhea/amenorrhea), whereas eating attitudes or behaviors and BMD were reported as continuous and dichotomous variables. We used independent t tests to compare demographic and physical characteristics of athletes in the final sample (ie, those who received a DXA scan [n = 170] compared with the girls who completed the initial study questionnaires but were not selected to receive a DXA scan [n = 419]). Independent t tests were also used to compare physical characteristics between eumenorrheic and oligomenorrheic/amenorrheic athletes, and between athletes with normal vs elevated EDE-Q scores. We conducted χ² analysis to determine proportions of oligomenorrheic athletes as a function of years post menarche. Univariate analysis was used to determine the relationship among menarcheal age, menstrual status, and BMD groups. Because body mass has been reported to influence scores on eating attitudes questionnaires,³² we used analysis of covariance to control for body mass index (BMI) (calculated as weight in kilograms divided by the square of height in meters) when comparing EDE-Q scores of eumenorrheic vs oligomenorrheic/amenorrheic athletes. Analysis of covariance was also used to control for chronological age, age of menarche, BMI, ethnicity, and sport type according to mechanical loading characteristics (swimmers, runners, and high-impact sports) when comparing the BMD of eumenorrheic vs oligomenorrheic/amenorrheic athletes and athletes with disordered eating who did or did not report pathogenic behaviors.

Test-retest intraclass correlation coefficients for the EDE-Q global score, subscales, and pathogenic behaviors, excluding binge eating, ranged from 0.84 to 0.92. The intraclass correlation coefficient for binge eating was 0.37. Interrater reliability coefficients for agreement among interviewers were 0.92 to 0.96. Test-retest reliability for the menstrual status questions ranged from 0.87 (number of consecutively missed cycles) to 1.00 (age at menarche), and the intraclass correlation coefficient for interrater reliability ranged from 0.99 to 1.00.

The athletes' self-reported race/ethnicity was as follows: 55% white, 21% Latina, 14% African American, 7% Asian, and 3% other. Athletes' age, height, weight, and BMI (mean ± SD) were 15.7 ± 1.3 years, 165.2 ± 7.0 cm, 59.4 ± 8.9 kg, and 21.8 ± 2.9 kg/m², respectively. Body fat percentage, lean tissue mass, and age of menarche were 26.6% ± 7.0%, 39.5 ± 4.6 kg, and 12.4 ± 1.2 years, respectively. No differences in ethnicity (P = .32), age (P = .67), height (P = .92), weight (P = .61), BMI (P = .74), or age at menarche (P = .88) were found between the athletes selected (n = 170) and not selected (n = 419). Since body composition was assessed by DXA, we do not know whether body fatness and lean tissue mass were different between the girls who received or did not receive a DXA scan.

Of the 170 athletes in the sample, 31 (18.2%) met the criteria for disordered eating, 40 (23.5%) had menstrual irregularity (oligomenorrhea/amenorrhea), and 37 (21.8%) had low BMD for their age (WHO criteria), whereas 7 (4.1%) met the ISCD criteria for low bone mass. Ten athletes (5.9%) met criteria for any 2 triad components, whereas 2 other athletes (1.2%) met the criteria for all 3 components of the triad. Among the 419 athletes who were not selected for a BMD scan, 82 (19.6%) and 85 (20.2%) met the criteria for disordered eating and menstrual irregularity, respectively; thus, athletes within the subsample of 170 were similar to those within the larger sample.

The mean EDE-Q scores on all subscales for athletes who met the criteria for disordered eating were significantly higher than those of athletes with normal EDE-Q scores (Table 1), thus indicating that the EDE-Q discriminated well between those with normal eating attitudes and behaviors vs disordered eating (P<.001). For comparison purposes, Table 1 also gives the EDE-Q scores from previously published data for elite female runners with eating disorders,³³ as well as normative data from young adolescent girls.⁹

Ten percent of the sample reported the use of pathogenic behaviors to control body weight on 2 or more of the past 28 days. The most common pathogenic behavior reported was vomiting (7%), followed by binging (6.5%), laxative use (1.8%), and diuretic use (1.2%). All athletes who reported pathogenic behaviors also met the criteria for disordered eating based on their scores for weight concern, shape concern, or global score.

Menstrual irregularity was reported by 23.5% of the athletes. Of these, oligomenorrhea (17.1%) was the most common menstrual irregularity reported followed by secondary amenorrhea (5.3%) and primary amenorrhea (1.2%). Eumenorrheic and oligomenorrheic/amenorrheic athletes were similar in age, weight, BMI, and percentage of fat. Among the athletes who were less than 5 years post menarche, 17.3% reported oligomenorrhea; among those who were 5 years or more post menarche, 16.3% reported oligomenorrhea (P = .88). Thus, oligomenorrhea was not reported more frequently in athletes whose reproductive age was younger.

After adjusting for BMI, oligomenorrheic/amenorrheic athletes reported significantly higher dietary restraint and EDE-Q global scores than eumenorrheic athletes (Table 2). After adjusting for chronological age, ethnicity, BMI, and percentage of body fat, menarcheal age was significantly older in oligomenorrheic/amenorrheic athletes compared with eumenorrheic athletes and in athletes with BMD Z-scores of −1 or less (Table 3). After adjusting for age, age of menarche, BMI, race/ethnicity, and sport type, oligomenorrheic/amenorrheic athletes had significantly lower BMD values for the trochanter compared with eumenorrheic athletes (Table 4). Among athletes with disordered eating, those who reported pathogenic behaviors (n = 18) had lower BMD for all bone sites compared with those who did not report pathogenic behaviors (n = 13), but these differences were not statistically significant (P = .17 to .44).

In contrast to most studies that have focused on collegiate or elite athletes, the present study included athletes from multiple high school sports. Although the percentage of athletes who met criteria for all 3 components of the triad was low, our data demonstrate its existence among high school athletes. Because 5.9% of the athletes met criteria for 2 components and approximately 20% met criteria for any 1 component of the triad, a substantial number of these young athletes may be at increased risk for the full triad over time.

Eighteen percent of the athletes were classified with disordered eating. Beals and Manore⁴ found a similar percentage of athletes in multiple sports to be at risk for disordered eating (15.2%), whereas Cobb et al³ reported elevated Eating Disorder Inventory scores in 25% of a sample of female long-distance runners. Comparisons to their findings are difficult because they studied collegiate or postcollegiate athletes and used different questionnaires to assess disordered eating. Although it is not uncommon for adolescents to report dissatisfaction with body shape and weight in the absence of pathogenic weight control behaviors, the mean scores on all EDE-Q subscales for the athletes with disordered eating were similar to those from a sample of older elite runners diagnosed as having frank eating disorders.³³ Furthermore, the EDE-Q scores of our athletes with normal eating behaviors were, on average, lower than those from a reference sample of younger adolescent girls.⁹ These similarities support the scoring rubric we used to define disordered eating and demonstrate that a substantial percentage of the athletes in our study may be at risk for developing pathogenic weight control behaviors.

The percentage of athletes who reported pathogenic eating behaviors was lower than that reported previously for collegiate athletes.^4,6 Although the actual number of girls engaging in pathogenic behaviors was relatively small, 24% of those who engaged in these behaviors reported using 2 or more different pathogenic behaviors in the past 28 days. Moreover, 47% reported a frequency of 4 or more episodes in the past 28 days. Given their young age, these athletes may be at a higher risk for developing serious short- and long-term health problems associated with disordered eating, including nutrient deficiencies, cardiac disturbances, and osteoporosis.

Previous reports on the prevalence of menstrual irregularity in athletes have focused on elite athletes and/or collegiate populations. Cobb et al³ reported menstrual dysfunction in 36% of distance runners, whereas Beals and Manore⁴ reported menstrual irregularity in 31% of collegiate athletes from multiple sports. In our study, most of the athletes with menstrual irregularity reported oligomenorrhea. This finding should be interpreted cautiously because cycles more than 35 days apart may occur during the first 5 years after the onset of menarche, although 89% of girls attain normal cycles within 2 years of menarche.³⁴ In our sample, 75% were less than 5 years post menarche; thus, we are uncertain whether the oligomenorrheic athletes were actually experiencing menstrual dysfunction. However, the reported percentage of oligomenorrhea was similar among girls who were less than 5 or 5 or more years post menarche. The clinical significance of oligomenorrhea during the first few years post menarche needs further examination to determine whether it is predictive of low peak bone mass in young adults. Only 1.2% of our high school sample reported primary amenorrhea, whereas 7.4% of collegiate athletes have reported primary amenorrhea.⁴

Menarcheal age in our sample was similar to estimates reported for US women³⁵ indicating that on average, the athletes did not have delayed menarche, which had been reported previously in athletes from a variety of sports.^34,36 Consistent with previous reports,^3,20 we found a positive association between menarcheal age and menstrual irregularity. Menarcheal age of our amenorrheic athletes was approximately 6 months older than that of eumenorrheic athletes, despite no differences in chronological age, body weight, percentage of fat, or BMI. Thus, other factors predictive of late onset of menarche did not explain the differences within menstrual status groups. Moreover, the age at menarche, which was a significant predictor of BMD for all bone sites measured, was approximately 7 months older in the athletes with BMD Z-scores of −1 or less and a full year older in those with BMD Z-scores of −2 or less, although their chronological age and body fatness were similar to those of girls with normal BMD. Therefore, some young athletes may be missing a key bone development period during which participation in sports cannot compensate for other factors associated with bone mineral accrual. Clearly, adolescent athletes with persistent menstrual dysfunction are at high risk of premature osteoporosis. Moreover, bone loss or suboptimal bone mineral accrual associated with amenorrhea may not be completely reversible with the resumption of menses.^19,37

According to WHO criteria for adult women, we found that 21.8% of the athletes had low BMD values, whereas 4.1% were classified as having low BMD using ISCD criteria. Although the WHO criteria are based on T-scores and are not directly applicable to adolescents, a BMD T-score of −1 SD corresponds to approximately 10% lower bone mass for age.³⁸ We support the WHO criterion to detect girls at risk earlier than when pharmacologic treatment would normally be prescribed. We believe that behavioral intervention is important for girls with Z-scores of −1 to −2. Since 98% of skeletal mass is acquired by the approximate age of 20 years, girls in their late teens are rapidly approaching their peak BMD. Thus, high school athletes with Z-scores between −1 and −2 are likely to be classified as having osteopenia or osteoporosis within a few years unless the underlying cause of their low bone mass is recognized and addressed.

Somewhat inconsistent with several reports of lower bone mass in amenorrheic vs eumenorrheic athletes,^3,18- 19 BMD was significantly lower only for the trochanter in our oligomenorrheic/amenorrheic athletes. However, BMD values for all other bone sites tended to be lower in oligomenorrheic/amenorrheic compared with eumenorrheic athletes. It is possible that the younger age of our athletes and/or their relatively brief history of menstrual irregularity may have precluded results similar to those reported previously, since it has been shown that the amount of bone loss associated with amenorrhea depends on the duration of menstrual dysfunction.³⁹ The results of the present study underscore the need for longitudinal studies in this population.

We recognize the limitations of self-report and the ability of adolescents to accurately recall past events; however, we believe that we were able to minimize inaccurate reporting by using a semistructured interview procedure and standardizing the procedures for administering the questionnaires. The high reliability coefficients we obtained increased our confidence in the athletes' ability to recall their menstrual history and eating habits and attitudes. By explicitly discussing confidentiality issues with the athletes, we believe that we minimized underreporting of disordered eating and irregular menses, particularly by girls who feared that their results would be shared with their coach or parents. Another limitation was the lack of laboratory data on menstrual function. Assessment of reproductive hormones might have permitted better interpretation of menstrual data, especially in girls who met the criteria for oligomenorrhea/amenorrhea.

In conclusion, this study provides data to inform health care professionals and coaches of the existence of triad components among high school athletes. The prevalence estimates and relationships found between the individual components indicate a need for education and behavioral interventions in this age group. We believe that screening for disordered eating and menstrual irregularity is potentially more important for high school than for college athletes as a first step in preventing comorbidities associated with the triad, particularly because adolescence is a critical period for optimizing bone mineral accrual.

Correspondence: Jeanne F. Nichols, PhD, Department of Exercise and Nutritional Sciences, San Diego State University, San Diego, CA 92182-7251 (jeannebernhard@cox.net).

Accepted for Publication: August 10, 2005.

Funding/Support: This study was funded by grant 903GGP005 from the Research and Education Foundation of the National Athletic Trainers Association (Dallas, Tex) and by a grant from the Graduate Division and Research Affairs, San Diego State University, San Diego, Calif.

Acknowledgment: We thank the high school coaches, athletic directors, and athletes for their cooperation and enthusiasm for Project SPIRIT (Sports Injury Research in Teens), as well as the graduate student assistants associated with Project SPIRIT. We also thank Caroline A. Macera, PhD, MPH, FACSM, for reviewing the manuscript.