WHAT IS ALREADY KNOWN ON THIS TOPIC
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
Using %fat Z-score ≥−1.0 as a target goal, rather than standard BMI or %EBW thresholds, may be helpful in treating female athletes with REDs-related menstrual irregularities.
Introduction
Relative energy deficiency in sport (REDs) is a syndrome of health and performance consequences resulting from exposure to problematic (prolonged and/or severe) low energy availability (LEA).1 Problematic LEA can occur when athletes fail to consume sufficient calories to offset exercise energy expenditure, leading to insufficient energy for essential physiological processes.1 LEA can cause deleterious health outcomes for numerous body systems, with reproductive dysfunction being one of the most extensively studied and well-understood consequences in females.1 2 The relationship between LEA and menstrual dysfunction was initially identified in the Female Athlete Triad (Triad).2 The Triad describes the interrelationships among LEA, impaired menstrual function and poor bone health, now components of the larger REDs model.1 Research into the Triad and REDs has demonstrated a strong association between LEA and the development of amenorrhoea or oligomenorrhoea in some female athletes,1 which are driven by hypothalamic-pituitary-ovarian (HPO) dysfunction and disrupted hormonal pulsatility of gonadotropin-releasing hormone (GnRH), luteinising hormone (LH) and others.3 When other causes are ruled out, oligo-/amenorrhoea may serve as a surrogate marker of prolonged and/or severe LEA.4
For female athletes with REDs-related oligo-/amenorrhoea, weight restoration and menstrual resumption are often primary treatment goals.5 Historically, in patients with eating disorders (EDs), body mass index (BMI) ≥18.5 kg/m2 (standard of care for patients age >20) and per cent expected body weight (%EBW) ≥90% [BMI/(median BMI for age); standard of care for patients age ≤206] have often been cited as clinical targets to guide weight restoration and menses resumption.7 However, using BMI and %EBW as clinical targets is impractical for athletes, who typically have higher lean mass and lower fat mass than the general population. Oligo-/amenorrhoeic athletes often present with BMI or %EBW within the ‘normal’ range, indicating the limitations of relying solely on these measures as treatment benchmarks.
Prior research examining the relationship between per cent body fat (%fat) and menstrual function has yielded mixed results. Some research has indicated that %fat is a positive predictor of menstrual resumption in functional hypothalamic amenorrhoea and suggested that relative fat reserve may be a useful marker of adequate energy storage to support normal hormonal and reproductive function.8 9 Yet, other work found higher fat mass in athletes with hormonal suppression compared with eumenorrhoeic athletes.10 Studies aiming to define a specific %fat threshold to predict menstrual function have employed various body composition measurement techniques, often without normalising to population-based reference values.5 11–13
Dual-energy X-ray absorptiometry (DXA) is the preferred technique for measuring bone mineral density (BMD)14 15 and is increasingly used to monitor body composition in athletes.16 17 DXA directly measures bone mineral content and, indirectly, fat and lean mass.16 18 When standardised protocols are followed, DXA provides reliable estimates of fat mass and is a recommended method for body composition assessment.1 16 19 Clinicians frequently order DXA for BMD evaluation in athletes with suspected REDs, offering an opportunity to assess %fat measures during routine care.20 Z-scores, a statistical tool used to normalise data measurements and facilitate comparison to a reference population’s mean, are routinely used to interpret DXA-derived BMD in premenopausal females, as they enable age- and ethnicity-matched comparisons, though they are not athlete-specific. Applying this Z-score approach to DXA-derived %fat may enable a more standardised evaluation of body composition values.
Therefore, this study aimed to investigate whether DXA-derived %fat Z-scores could detect menstrual function in female athletes. We hypothesised that %fat Z-score would be a better discriminator of menstrual status than BMI or %EBW, thus providing an additional clinical tool to aid in REDs assessment and treatment.
Methods
Participant selection
We conducted a retrospective chart review to identify eligible patient records for this study. Using our institution’s electronic medical record (EMR) extraction software, we initially reviewed the charts of female patients aged 15–30 years who presented to a large tertiary care hospital for clinical evaluation between 2012 and 2022. Search criteria included the terms “DXA”, “DEXA” and “bone density scan”. We then manually reviewed all resulting charts to determine patient eligibility. Patients were excluded if medical records lacked reported menstrual status or DXA-derived %fat; if they were non-athletes (ie, 21); had chronic conditions (other than dietary- or exercise-induced LEA) that could affect menstrual status; or were on exogenous hormones (eg, transdermal 17β-oestradiol, hormonal contraceptives) in the prior six months. Patients with menstrual irregularity who were within 1 year of menarche were excluded. For patients with multiple DXAs, only the first scan between the ages of 15 and 30 was used for analysis. Figure 1 details the exclusion process.
Figure 1
Flow diagram of inclusion and exclusion of patient charts. †Chronic medical conditions excluded include: cancer; coeliac disease; CHARGE syndrome; Crohn’s disease; cystic fibrosis; diabetes; Ehlers-Danlos; growth hormone deficiency; hyperprolactinaemia; hyperthyroidism; hypothyroidism; Jacob’s syndrome; jejunal atresia; PHACE syndrome; Turner’s syndrome; ulcerative colitis; XY karyotype. DXA, dual-energy X-ray absorptiometry; OCPs, oral contraceptive pills; IUDs, intrauterine devices.
Definitions and criteria
Sport participation, including reported competition and activity level, was collected from clinical records. Patients were classified as ‘athletes’ if they were Tier 2 or higher as defined by the McKay et al framework.21 We categorised athletes by primary sport(s) reported, with a total of 34 sports mentioned, and then grouped into five categories: dance, running, rowing, team sport and other.
Clinician-obtained menstrual history (ie, age of menarche, contraceptive use, number of periods in the past year) was extracted from medical records, and athletes were categorised based on their menstrual status at the time of the DXA scan. Categories were defined as: ‘naturally menstruating’ (NM): >9 menses (cycle length 21–35 days) in the previous year22 23; ‘oligomenorrhoea’ (OLIGO): menstrual cycle length of >35 days or 4–9 menstrual cycles in the previous year22; and ‘amenorrhoea’ (AMEN), capturing ‘primary amenorrhoea’: absence of menses by 15 years of age with secondary sex characteristics present, and ‘secondary amenorrhoea’: no menstruation for >3 months in females with previously normal menstrual cycles or for >6 months in females with existing menstrual irregularity.24
Data collection
All DXAs included in this study were performed on a Horizon DXA System and analysed with Hologic APEX Software. Patients were scanned without jewellery and in hospital gowns, and urinary β-hCG tests were obtained prior to scans to ensure patients were not pregnant; exercise, hydration and nutrition status were not assessed or controlled for prior to DXA scanning as part of the institution’s clinical protocol. %fat, %fat Z-score, total body less head (TBLH) BMD (g/cm2) and TBLH BMD Z-scores were derived from total body (TB) scans. For patients ≥18 years, TB, lumbar spine L1–L4 (LS) and unilateral hip were scanned; for those <18 years, TB and LS were scanned.14 BMD and BMD Z-scores were recorded for all scanned sites.25–31 Height, weight and BMI were measured at the time of scan.
For those ≤20.0 years, %EBW was calculated by dividing BMI by median BMI for age in months.32 Additional variables (eg, demographics, medical and injury history, labs, scan indication) were extracted from the EMR by the study team.
Statistical analyses
We compared the ability of three measures of body composition—%fat Z-score, %EBW and BMI—to discriminate the menstrual status of athletes from three separate groups: NM, OLIGO and AMEN. Initially, we treated the body measures (%fat Z-score, BMI and %EBW) as continuous variables. C-statistics were calculated from binary logistic regression models where the outcome was AMEN versus NM, AMEN versus no AMEN (OLIGO or NM), or OLIGO or AMEN (OLIGO/AMEN) versus NM, and the predictor was the specified body measure (%fat Z-score, BMI or %EBW) treated as a continuous variable. A higher C-statistic indicates better discrimination between NM and irregular menses. We then compared C-statistics for the different body measures using the DeLong method.33 Analyses involving %EBW only include athletes aged ≤20.0 years.
Next, we treated the body measures as binary variables. Athletes were classified as having low %EBW/BMI if they met traditional risk thresholds of <90% EBW (age ≤20.0 years) or BMI <18.5 kg/m2 (age >20.0 years), and low %fat Z-score if Z-score <−1.0 (chosen a priori). Sensitivity and specificity of these binary measures for detecting AMEN and AMEN/OLIGO were calculated using 2×2 tables. We compared the sensitivities and specificities of low %EBW or BMI to low %fat using McNemar’s test.
We also determined the ‘optimal’ BMI, %EBW and %fat Z-score cut-offs for this dataset by finding cut-off values that maximised the C-statistic in an ordinal logistic regression with the ordinal outcome of menstrual status (NM, OLIGO, AMEN) and the binary body measure as the predictor. For each body measure, we tried all possible cut-offs in increments of 0.1 until we found the cut-off for each body measure that yielded the highest C-statistic (ie, best discriminative ability between menstrual groups).34
Analyses were performed with SAS V.9.4 (SAS Institute). Statistical significance was determined with the nominal α=0.05.
Patient and public involvement
Due to the retrospective nature of the study, patients were not directly involved in study design.
Equity, diversity, and inclusion statement
Our team included five women from various disciplines—translational research, sports medicine, endocrinological clinical care and biostatistics—ranging in experience. Our study population included female athletes of different sports, competition levels and ages. However, we note the lack of ethnic and racial diversity in our cohort and the need for broader representation in future related research.
Results
Our initial search returned a total of 5068 charts (figure 1). Of these, there were 961 patients and 1221 DXAs. Of the 961 individuals screened, 388 female athletes were eligible and included in the analysis.
Descriptive characteristics of participants, stratified by menstrual group, are displayed in table 1. Mean age was slightly higher for AMEN (18.7 years) compared with OLIGO (17.8 years) and NM (17.6 years). Most athletes were white or had race unavailable at the time of data collection. AMEN had more Tier 3 and Tier 4 athletes than OLIGO or NM, and AMEN (44.7%) was more likely to have a history of EDs than OLIGO (27.4%) and NM (13.1%). Mean lifetime bone stress injury (BSI) was higher in NM (2.8) than in OLIGO and AMEN athletes (2.1 and 0.9, respectively). The most common reason for DXA referral was BSI history in NM and menstrual irregularity for AMEN and OLIGO.
Table 1
Descriptive statistics by menstrual group, mean (SD) or N (%)
Mean BMI, %EBW and %fat Z-score were highest in NM and lowest in AMEN (table 1). Of note, mean BMI and %EBW in AMEN and OLIGO exceeded the traditional cut-offs of 18.5 and 90%, respectively (table 1, figure 2). Conversely, the mean %fat Z-score in AMEN was −1.38, below our proposed cut-off of −1.0.
Figure 2
Box plots showing the distributions of per cent fat Z scores (A), BMI (B) and percentage of expected body weight (C, ≤20.0 years only) by menstrual group. Horizontal lines indicate typically used cut-offs for risk stratification for BMI (<18.5) and %EBW (<90%) and the proposed cut-off of Z<−1.0 for %fat Z-score. BMI, body mass index; %EBW, per cent expected body weight.
When treating %fat Z-score as a continuous variable, it was superior to both BMI and %EBW in discriminating between menstrual groups. In the full sample, %fat Z-scores had C-statistics between 0.74 and 0.80, indicating moderately strong discriminative ability for AMEN and menstrual irregularity, compared with C=0.66–0.70 for BMI (p<0.01 for all comparisons). In the sample with %EBW values available (age ≤20), %fat Z-scores had C-statistics between 0.74 and 0.81 versus 0.69 and 0.73 for %EBW (table 2); the difference was statistically significant when comparing AMEN to NM and AMEN to OLIGO or NM (OLIGO/NM) but not when comparing OLIGO/AMEN to NM. The ‘optimal’ cut-offs in our dataset for detecting oligo-/amenorrhoea were: <20.7 for BMI, <96% for %EBW and <−0.8 for %fat Z-score.
Table 2
C-statistics from logistic regressions testing the ability of per cent fat Z-score, per cent of expected body weight and body mass index to discriminate between menstrual groups
Of the 243 OLIGO/AMEN, 34.6% (n=84) met the low %EBW or BMI threshold, while 69.1% (n=168) met the low %fat Z-score threshold, indicating that traditional cut-offs fail to capture many athletes with oligomenorrhoea or amenorrhoea. The %EBW or BMI cut-offs also had low sensitivity for distinguishing AMEN (29.3%) and OLIGO/AMEN (25.9%), indicating that they are poor predictors of menstrual status in female athletes. %fat Z-score <−1.0 had significantly greater sensitivity for distinguishing AMEN (68.9%, p<0.0001) and OLIGO/AMEN (57.7%, p<0.0001). There was a slight loss of specificity for %fat Z-score <−1.0 compared with %EBW or BMI cut-offs for AMEN (−8.5%, p=0.0078) as well as OLIGO/AMEN (−5.3%, p=0.14) (figure 3, table 3).
Figure 3
Sensitivity and specificity of per cent fat Z-score and BMI/EBW for AMEN and OLIGO/AMEN. AMEN, amenorrhoeic; BMI, body mass index; EBW, expected body weight; OLIGO, oligomenorrhoeic.
Table 3
Sensitivity and specificity for amenorrhoea and any menstrual irregularity using traditional per cent expected body weight or body mass index risk cut-offs
Discussion
To our knowledge, this is the first study to explore the potential of DXA-derived %fat Z-score as a discriminator of menstrual status in a clinical population of athletes. In our cohort of 388 female athletes, DXA-derived total body %fat Z-score outperformed BMI or %EBW in predicting menstrual status, whether treated as a continuous or binary variable. A %fat Z-score <−1.0 exhibited greater sensitivity for predicting both amenorrhoea and oligomenorrhoea than traditional BMI or %EBW cut-offs. Many female athletes in our cohort experiencing oligo/amenorrhoea had ‘normal’ BMI or %EBW, but %fat Z-score <−1.0.
HPO dysfunction is a recognised consequence of LEA. Loucks et al first demonstrated reproductive hormone suppression in sedentary females with LEA,35 36 and subsequent studies supported this in active females.37–39 Oligo/amenorrhoea is an established marker of hormonal and LEA-related dysfunction and is associated with physiological consequences, including bone impairments,40 41 cardiovascular dysfunction42 and performance decrements.10 Thus, restoring menses remains a primary treatment goal in REDs to mitigate subsequent consequences associated with oestrogen deficiency and LEA.
Previous research has explored the relationships among BMI, %fat and menstrual function in girls and women with anorexia nervosa (AN) and in female athletes cross-sectionally, with mixed results. Pitts et al observed a positive association between DXA-derived %fat, BMI and changes in BMI and resumption of menses in adolescents with AN.11 Similarly, in an adult cohort, Winkler et al found associations among BMI, %fat and menstrual cycle resumption; however, DXA-derived %fat was not a better predictor of menstrual function than BMI.9 Conversely, Golden et al found no significant relationship between skinfolds-derived %fat, body weight or BMI and menstrual cycle resumption in a cohortof adolescents with AN43; Arimura et al also found no significant difference between DXA-derived %fat in adolescents with AN who did and did not resume menses.44 Other studies have generally agreed that %fat is positively associated with normal menstrual function5 12 and found that oligo-/amenorrhoeic athletes (and exercising women) have lower fat mass and %fat than normally menstruating athletes and exercising women (via skinfolds, regression equations, hydrostatic weighing, bioimpedance analysis and DXA).13 45–47
Efforts to identify a minimum %fat threshold required for menstrual function have also yielded inconsistent results.13 48–50 Initial work by Frisch et al assessed %fat via hydrostatic weighing and proposed a minimum 22% body fat to maintain normal menstruation51 52; research has since challenged such a target. Other prior studies have used indirect and/or less precise body composition measurement methodologies, including skinfolds, bioimpedance analysis and regression equations. We employed DXA, which, despite some variability depending on time of day, hydration and feeding status, and other factors, is more precise than other methods and is widely used in clinical practice.18 We also used %fat Z-score, directly comparing individuals to age- and ethnicity-matched groups,28 rather than relying on absolute %fat values. Indeed, when normalising %fat, we found that a DXA-derived %fat Z-score <−1.0 is a better predictor of menstrual dysfunction than either BMI <18.5 or %EBW <90%, providing a benchmark for further investigation and potential integration into clinical practice.
The mechanisms linking low %fat and menstrual dysfunction are multifactorial. The HPO axis is suppressed in LEA through disrupted GnRH and LH pulsatility, contributing to menstrual irregularity.53 Appetite-regulating hormones, such as leptin and ghrelin, are also implicated. Leptin, an adipocytokine, is positively associated with fat mass and LH pulsatility54 55; ghrelin is upregulated with energy deficiency and inversely correlates with %fat and LH secretion.54 In adolescent and young adult athletes, Ackerman et al found oligo-/amenorrhoeic athletes exhibited lower leptin and LH and higher ghrelin54— each associated with lower %fat—suggesting these hormonal shifts may contribute to HPO axis suppression in REDs.54 However, it remains unclear whether fat mass or EA primarily drives these hormonal changes.56 Adipose tissue is an endocrine organ that contributes to sex steroid production and leptin signalling,57 58 yet hormonal responses may vary depending on the cause (eg, exercise- vs diet-induced), duration and frequency of LEA,59 as well as lean mass and training load; thus, disentangling the effects of LEA and low adiposity is challenging and may explain the equivocal findings in prior studies. From an evolutionary perspective, life history theory posits that in energy-deficient states, physiological resources are redistributed to prioritise survival over reproduction.60 In athletes with LEA, reduced %fat (ie, a decrease in energy reserve) may prompt reproductive suppression in an effort to allocate energy to systems necessary for immediate survival. These trade-offs shift across the life cycle, so age and developmental stage may influence the %fat threshold necessary for reproduction. Future research examining the underlying pathophysiology and hormonal contributions driving %fat and menstrual resumption in REDs-affected athletes of various ages and stages of development is therefore warranted.
Importantly, in our cohort, not all athletes with %fat Z-score <−1.0 had menstrual dysfunction, nor were all athletes with a %fat Z-score ≥−1.0 naturally menstruating. Factors such as athlete tier, training age, gynaecological age, training load, genetics, nutrition support and LEA duration/severity likely influence individual tolerance. Similarly, not all athletes respond to LEA with a decrease in %fat10 61; suppressed resting metabolic rate in the context of LEA can contribute to weight gain, and LEA can cause attrition of muscle mass. Thus, %fat Z-score may offer an additional tool in evaluating LEA-related menstrual dysfunction, but individualised assessment remains critical. Sensitivity for detecting oligo-/amenorrhoea was significantly greater for %fat than either BMI or %EBW, although there was a modest reduction in specificity. Practically, this means that more athletes ‘at risk’ of oligo-/amenorrhoea will be caught, which is preferred, but it is important to note that some NM athletes may also fall into the low %fat category. This highlights the need for gathering appropriate clinical context and approaching each athlete with an individualised approach.
Clinical implications
While the current study focuses on restoring menses in OLIGO/AMEN athletes, a large proportion of females use hormonal contraception. DXA may also help assess athletes taking hormonal contraception (eg, oral contraceptive pills, implant, intrauterine device), which can obscure true menstrual status, as some can induce a monthly vaginal withdrawal bleed and others suppress bleeding altogether. In those using hormonal contraception, the suppression of endogenous reproductive hormones and possibly other bone-regulating hormones is masked.62 Prior studies have found that many female athletes use hormonal contraceptives at some point in their lives63–65; thus, developing tools for these athletes to monitor their hormonal profiles is important. As a %fat Z-score <−1.0 is predictive of oligo-/amenorrhoea in female athletes, using this threshold in conjunction with lab monitoring (eg, triiodothyronine (T3)) may be useful for athletes not wanting to discontinue hormonal contraception during their REDs recovery process. Further research is required to continue examining clinically useful benchmarks in the broader population of female athletes.
Our findings support the clinical utility of a %fat Z-score ≥−1.0 when managing female athletes with REDs and menstrual dysfunction. However, to protect athletes’ mental health and well-being, body composition DXA measurements should only be ordered by appropriate providers66 for medical purposes in athletes aged <18, with careful adherence to best practices outlined by the 2023 IOC REDs Consensus Group.1 Focusing on body composition can contribute to ED behaviours in female athletes and have detrimental effects on their mental health.66 67 As such, providers need to approach the subject intentionally and mindfully. Additionally, DXA-derived %fat measurements often differ from those of other modalities, which may be misleading and can be ‘triggering’ for athletes when provided without appropriate context.18 In clinical practice, clinicians might emphasise recovery by sharing DXA images without exact numerical values, emphasising the importance of REDs recovery and the need to increase fat mass, with a ‘health first—performance second’ approach.1
Limitations
Despite the strengths of our study, including the large sample size, there are several limitations. First, we gathered data for menstrual history, medical history and athletic participation retrospectively from medical records. While we only included patients with sufficient menstrual and athletic history to allow categorisation, a degree of misclassification is possible due to incorrectly documented clinical records. Our cohort lacked diversity in race/ethnicity and sport type and came from a clinical setting, where evaluations were prompted by concerns such as an underlying injury, menstrual irregularities or REDs. Therefore, the NM group may not be fully representative of the overall athlete population, as they often presented for a DXA due to a history of BSI. Investigation of %fat Z-score in a more diverse/representative sample of athletes is warranted to increase generalisability, as well. While DXA is a recommended method of assessing body composition in athletes, both technical and biological factors can affect measurement precision.68 DXAs were conducted in a clinical setting without strict control over factors like hydration (including exercise-induced changes in soft-tissue hydration), diet and time of day, which may affect the precision and accuracy of %fat.18 68 There are also inherent technical variations, as reflected by the per cent coefficient of variation; however, proper technician training can limit the influence of machine-related variability. Future research examining %fat Z-scores and menstrual function would benefit from additional standardisation and adherence to gold-standard methodologies for athlete body composition.68 Additionally, the normative database used for Z-scores is not athlete-specific.28 Developing athlete-centred reference data could enhance clinical utility. Finally, the cross-sectional design precludes causal inference. Longitudinal research is needed to assess whether changes in %fat Z-score can predict menstrual recovery. Overall, the %fat Z-score was indeed better than either BMI or %EBW in discriminating between menstrual status in athletes; however, additional research is necessary to test this concept.
Conclusion
Our findings suggest that a %fat Z-score <−1.0 is a better discriminator of menstrual status than BMI or %EBW in female athletes. This simple threshold potentially serves as a useful clinical tool when setting weight restoration targets and supporting menstrual resumption in athletes with REDs. Further prospective research, including serial DXAs, coupled with ovulation testing and additional REDs markers, is needed to refine the clinical utility of %fat Z-score. Despite these future research needs, the improved sensitivity of %fat Z-score makes it a promising alternative to the commonly used BMI or %EBW benchmarks for female athletes experiencing REDs-associated oligo-/amenorrhoea.
Data availability statement
Data are available upon reasonable request.
Ethics statements
Patient consent for publication
Ethics approval
We received institutional review board approval for this project (IRB-P00031951).
Acknowledgments
We acknowledge and thank the Boston Children’s Hospital Bone Health and DXA programmes, as well as our patients included in this project.
