Studies were eligible for inclusion if they met the following criteria: participants: athletes of both sexes, all disciplines, and advanced or elite level; outcomes: evaluation of the type, variety, number, and frequency of individual markers used in REDs diagnosis, as well as compliance with the IOC recommendation. Studies were ineligible if the outcomes of interest were not measured or the results were not described. Literature reviews, guidelines, letters to the editor, conference abstracts, dissertation thesis, and non-English language articles were also excluded. Considering that REDs was officially defined in 2014, the search was performed with a data range from March 2014 to. The search was conducted on Medline (via PubMed), Web of Science, and SPORTDiscus (via EBSCOhost) in April 2023. The search terms followed the descriptors from categories #1 and #2 related to “athlete” AND “relative energy deficiency” using the entry terms and derivative words (available in the Table 1).

The articles were imported into Rayyan systematic review software to proceed with the selection process. This process was performed as follows: (1) a researcher (KD) uploaded the articles from each database and then (2) excluded the review articles, letters to the editor, duplicates, and articles in non-English languages (identified by the software); (3) two independent researchers (DP and AW) screened the articles’ titles and abstracts, and a third checked those excluded (KD); and (4) finally, two independent researchers (KD and AW) screened the full text of the articles for final inclusion. Any disagreements between reviewers were resolved by a third reviewer (AP). A prior pilot selection, with the first 25 articles, was performed to test the researchers’ understanding, demonstrating an agreement of 88% between the two reviewers (DP and AW). 

Data related to the sample characteristics (e.g., sex, sport modality, age, and size), the presence of REDs, biomarkers used in REDs diagnosis [e.g., hormones, resting metabolic rate, bone mineral density, blood glucose, body mass index (BMI), and cholesterol], and any potentially relevant outcomes were extracted from included studies by two researchers (KD and AW).

The assessment of methodological quality for the articles with a descriptive approach was performed using the STROBE tool (29) and for those with an intervention approach was performed by ROBINS-I (30). Three researchers participated in this phase (KD, AW, and AP). The STROBE checklist assesses the quality of cohort, case–control, and cross-sectional studies. It contains 22 items assessing risk factors for bias. Response options are a score of 0 if the articular checklist item is not fulfilled, 1 if the articular checklist item is fulfilled, and NA if the checklist item does not apply to the specific publication. Based on the sum of the total score and the percentage gain of the possible maximum, the quality of the study is then evaluated as follows: ≥85 = excellent, 70 to <85 = good, 50 to <70 = fair, and <50 = poor, as used previously. The ROBINS-I rating system is based on seven domains, each consisting of a subset of questions focusing on possible areas of systematic error. The domains include confounding, participants, classification of interventions, deviations from intended interventions, missing data, measurement of outcomes, and selection of the reported results. In this review, we used only domains 2–7 for evaluation; more details on this process are provided in the Discussion section. The response options are “Yes,” “Probably yes,” “Probably no,” “No,” and “No information.” Based on the continuous responses, each domain is then evaluated as a whole, and the rating of all the domains is reflected in the labeling of the study as “Low risk,” “Moderate risk,” “Serious risk,” and “Critical risk” of bias.

The biomarkers used in the included studies were compared with the IOC REDs CAT2 (14), an improved version derived from the original IOC REDs Clinical Assessment Tool (CAT) introduced in 2015 (31). The development of the IOC REDs CAT2 involved internal validation through expert voting statements and external validation via clinical cross-agreement assessments by experts. The assessment protocol of IOC REDs CAT2 comprises three sequential steps: I.

Initial screening using population-specific REDs questionnaires or clinical interviews, with individuals deemed at risk moving on.

In the review process, markers identified in the included studies were compared to those outlined in the REDs CAT2. Given the focus on objective measurement methods, subjective markers obtained through interviews or questionnaires were omitted from this comparison. Subsequently, reviewer KD computed agreement rates between each study and the REDs CAT2 tool for scored, potential, and overall indicators. A second independent reviewer (AW) checked this process to ensure reliability.

In total, 595 articles were found in the databases matching the combination of keywords entered. After excluding articles that were duplicates (n = 96) and for other reasons, such as those written in a foreign language (non-English) (n = 10) and with no access (n = 1), 488 articles were evaluated during the title and abstract screening. Of these articles, 155 were excluded through the review method, and 463 did not meet the eligibility criteria. For 25 articles, the full text was assessed; of these, 12 studies were excluded due to non-compatibility. Therefore, 13 studies were included in the final process (Figure 1).

The main characteristics and the categories of the REDs markers used in the included studies (bone mineral density, resting metabolic rate, blood markers, anthropometric parameters, nutritional intake, and performance) are presented in Table 1. Most of these studies focus on female athletes (7 out of 13); the most investigated disciplines were endurance sports, team sports, ballet, climbing, or a mix of disciplines or para-athletics disciplines. Athletes competed at the performance levels of well-trained, competitive, elite, national, and Olympic levels. Two studies also included a control group.

Among the 13 papers, 12 presented a descriptive study design and 1 presented an intervention design. For the descriptive ones, the methodological quality, assessed by the STROBE tool, demonstrated a range from good to excellent quality. Specifically, six studies were rated as excellent (17, 18, 20, 23–25) and six were rated as good (16, 21, 22, 26–28) (see the Table 2). One paper was designed as an intervention and demonstrated a moderated risk of bias based on the ROBINS-I tool.

We found 79 biomarkers used to determine the presence of REDs in the 13 included studies. Table 2 presents the complexity and diversity of the biomarkers used within the included studies. The biomarkers were categorized into five groups (bone mineral density biomarkers, resting metabolic rate biomarkers, blood biomarkers, anthropometric parameters, nutritional intake parameters, and performance parameters). All 13 studies used at least two (or more) categories of markers to determine the presence of REDs.

Table 3 presents the quantification of biomarkers used to assess REDs and complements Table 2. It shows a comprehensive overview of the frequency of their use in the included studies. The biomarkers most often used were BMD, BMI, BM (body mass), FM (fat mass), and T3 (triiodothyronine) blood concentrations, which were involved in 10 of the 13 studies (76.9%). Nine studies (69.2%) used the measurement of RMR (resting metabolic rate), while 8 studies (61.5%) used total testosterone level and EI (energy intake). Seven studies (53.9%) used nutritional parameters such as carbohydrate, protein, and fat intake EA (energy availability) was used in six studies (46.2%).

Anthropometric parameters, such as BMI and body composition, are widely used in medical practice. According to the Centers for Disease Control and Prevention's recommendations for general practitioners (32), BMI is a simple, inexpensive, and non-invasive method of estimating body fat and health risk, requiring no special equipment. However, several studies have pointed to the inaccuracy of BMI, particularly among patients with different ethnic backgrounds or an inability to distinguish body weight between body fat and muscle mass (33, 34). Thus, although this calculation can provide valuable information in the REDs diagnostic process, as with other markers, it cannot be evaluated in isolation (35). According to REDs CAT2, BMI is considered a potential indicator in assessing REDs risk, underscoring that the need for further research to quantify the parameters and cutoffs more accurately (14).

To accurately determine body composition, it is necessary to use valid methods that contribute to an objective assessment of the athlete’s overall condition. Body composition and adipose tissue thicknesses can be accessed via skinfold measurement. However, B-mode ultrasound is a more reliable and preferred method, which can provide good results even in lean individuals (36). Despite its costliness, the dual-energy x-ray absorptiometry (DXA) measurement is also the recommended method of choice as the gold standard for assessing body composition (37).

The authors of the IOC Consensus statement also pointed out that too much focus on anthropometric parameters and body composition can intensify the pressure placed on athletes, especially on adolescents under the age of 18 years (13, 38). It is, therefore, essential to identify valid and reliable methods and develop guidelines for interpreting, managing, and communicating with athletes (39).

Biomarkers assessing bone health are among the most used, as shown by the results of this review. Impaired bone health has been associated with low energy availability from its onset. It was also included in the original definition of the female athlete triad (34), from which the concept of REDs was developed (1). Low energy availability affects bone health through reduced levels of hormones such as estrogen, leptin, and T3 associated with insulin-like growth factor 1 (IGF-1) secretion (40–43). In addition, inadequate intake of essential nutrients, including protein, calcium, or vitamin D, has been linked to the low energy intake associated with REDs (44, 45).

DXA is the most used method for measuring bone mineral density (46) and is also noted as a “preferred method” in the 2023 IOC Consensus statement (13). According to REDs CAT2, the authors recommend the following as a positive finding: •

Premenopausal women and men aged <50 years: BMD Z-score <−1 at the lumbar spine, total hip, or femoral neck or decreased BMD Z-score from previous testing.

Moreover, markers reporting bone mineral density status should continually be assessed in the context of supplementary information, considering the specificity of each sport discipline. For example, the bone density of weightlifters generally reaches higher values than the reference range (50), and average values may indicate reduced BMD in these athletes.

RMR was assessed in nine of the 13 included studies, accounting for 69.2% (see Table 3). RMR represents the energy necessary to maintain homeostasis while at rest. Unlike basal metabolic rate, which necessitates strict conditions such as a 12-h fasting period and a thermoneutral environment, RMR can be measured throughout the day (51). The suppressed RMR associated with LEA may be explained by adaptive responses aimed at conserving energy (52).

Various methodologies are employed in studies to determine RMR. Indirect calorimetry is often called the gold standard but requires specialized equipment (53). Consequently, researchers usually resort to estimating RMR using predictive equations, such as those proposed by Cunningham (54), Harris and Benedict (55), or Owen et al. (56). Another approach is the RMR ratio, defined as the ratio between measured RMR and predicted RMR. Some studies suggest that the RMR ratio serves as a valid indicator of LEA (57, 58). However, it is advisable to evaluate RMR in conjunction with other markers due to variations in the degree of metabolic suppression among athletes. These variations are influenced by factors such as the severity of LEA (58).

The 2023 IOC Consensus statement recognizes RMR testing as a “used and recommended” method for identifying impaired energy metabolism. Specifically, the endorsed procedures include indirect or room calorimetry measurements (13). In addition, REDs CAT2 identifies RMR as a potential indicator, with a reduced or low RMR [<30 kcal/kg fat-free mass (FFM)/day] or an RMR ratio (<0.90) considered indicative of the condition (14). However, Sterringer and Larson-Meyer (59) pointed out that a threshold of 0.9 may not be appropriate for all cases. In particular, for studies using the Cunningham equation from 1991 (60) or DEXA measurement, a threshold of 0.9 may lead to an underestimation of the prevalence of LEA.

One of the most used markers in the included studies (76.9%) was T3. It is one of the hormonal agents released by the thyroid gland and is indispensable in energy metabolism and growth (61). T3 is also involved in the reproductive process (62) and bone tissue metabolism through the local production of IGF-1 (63). Although its concentration is strongly associated with metabolic functions, this marker still needs to be evaluated in the context of other methods. This is because its concentration may be affected by many conditions, such as circadian rhythms, thyroid disease, alterations in serum binding proteins, or other associated medical conditions (64, 65). Clinically or subclinically low total or free T3 is also considered one of the primary REDs indicators listed in REDs CAT2 (14), and clinically or subclinically low IGF-1 is included in the list of potential indicators.

Testosterone concentration is also a frequently used marker in the studies (61, 5%). Subclinically low total or free testosterone is listed in REDs CAT2 primary indicators; clinically low total or free testosterone is considered a severe primary indicator (counted as two primary indicators) (14). While disturbances in the menstrual cycle may affect the hypothalamic–pituitary–gonadal axis in women, this condition may not be detected as early in male athletes. Thus, for male athletes, in addition to testosterone levels, it is often necessary to consider self-reported data, such as the presence of low libido or decreased frequency of morning erections, in the diagnosis of REDs. Thus, as already mentioned, a combination of diagnostic methods is required. In women, low energy availability disrupts luteinizing hormone (LH) pulsatility, which further affects the hypothalamic–pituitary–gonadal axis, including levels of follicle-stimulating hormone (FSH), estrogens, and progesterone (66, 67). Two studies tested estradiol levels (17, 26), while one study tested levels of LH, FSH (17), and progesterone (26). However, the REDs CAT2 tool does not directly use these hormones as female reproductive cycle function indicators. Instead, it uses self-reported data on the presence of primary amenorrhea, secondary amenorrhea, or oligomenorrhea (14). LEA also affects other endocrine pathways such as cortisol, leptin, growth hormone, IGF-1 axis, sympathetic and parasympathetic tone, or thyroid hormones (66).

The calculation of energy availability has been used in 46.2% of the included studies (6 out of 13). Given that low energy availability is a direct etiological factor in developing REDs (1), its inclusion in diagnostic methods appears logical. The variables required for the calculation of energy availability can also be obtained in a non-invasive and non-burdensome way. The prevalence of low energy availability was detected in four of the thirteen included studies: 67% (16), 23% (18), 22% (24), and 46% (28). In another study, a prevalence of 28% was assessed through the SEAQ-I questionnaire (21). In conclusion, the mean observed prevalence of LEA across the studies is 39.5%.

However, previous studies have indicated that calculating energy availability carries a high risk of error (9). Sources of this inaccuracy can include energy intake, while data obtained through nutritional recall may underestimate actual intake by 10%–20% (68, 69), and even cases of an underestimation of 50% are not uncommon (70). The measurement of energy expenditure also needs to be evaluated cautiously. Various methods of assessing energy expenditure are used across studies, such as doubly labeled water technique, direct calorimetry, indirect calorimetry, accelerometry, heart rate monitoring, or pedometry (71). Nevertheless, using more accurate methods is often complicated by the high cost of these devices in research settings. Therefore, epidemiological studies frequently rely on self-reported methods, which can lead to significant inaccuracies in the observed outcomes (72). Thus, calculating energy availability may serve as a valuable complementary method for diagnosing REDs and could also be beneficial in determining the optimal therapeutic approach (73). However, like other markers, it should not be evaluated in isolation.

Although there is still no uniform and standardized methodological approach for the diagnosis of REDs, the new IOC Consensus statement of 2023 provides a comprehensive overview of: (I)

Preferred methods;

The authors of the REDs CAT2 emphasize that this advanced tool should not be used in isolation but in combination with clinical consideration and other tools, such as screening questionnaires. In addition, they warn that the tool's reliability decreases if all the included indicators cannot be assessed and that REDs CAT2 is not a substitute for professional clinical diagnosis, advice, and/or treatment (13). Nevertheless, REDs CAT2 represents a scientifically supported system for evaluating LEA indicators and was selected as a reference tool to assess the quality of the markers used in the included studies.

As this review primarily focuses on objective methods of practical measurement, some subjective indicators obtained through interviews or questionnaires were excluded from this comparison. However, as previously mentioned, objective and subjective methods cannot be entirely separated, and combining them is desirable. After excluding methods that are not objectively measurable, 15 markers were identified in the reference tool. Five of these markers are scored as primary or secondary indicators; 10 potential markers are not scored but are considered emerging. An overview of the included and excluded indicators and the results of the agreement can be found in Table 4. The highest agreement with the CAT2 REDs was achieved in the study of Stenqvist et al. (22) using 80% of the scored indicators. Six studies used 60% (19–21, 25, 26, 74), two used 40% (18, 28), three used 20% (16, 23, 24) of the scored indicators, while one study did not include any of these scored markers but only the potential ones (27).

The main limitation, not only of this review but to the entire field of REDs, is that no single marker or group of markers can reliably indicate the presence of REDs in athletes at this time. Therefore, we can only determine athletes’ risk levels as “low/moderate/high” rather than diagnosing the presence or absence of REDs. REDs cannot be diagnosed based on a single variable. Instead, several factors must be considered. Thus, this review can only provide an overview of the markers used in REDs diagnosis in current studies and highlight their frequency of use. The most commonly used markers were also analyzed with respect to the REDs CAT2 tool. Another potential source of error is the assessment of study quality and the risk of bias. Although three researchers performed these tasks independently, evaluating individual questions and the overall evaluation of the included categories might be influenced by subjective perceptions or interpretations of the questions related to the REDs topic.

The process of diagnosing REDs is currently fragmented, with studies employing various methods and a broad range of markers in their methodologies, as evidenced by the findings of this review. In addition, determining the presence or absence of REDs is challenging. In response, it is crucial to identify reliable markers suitable for diagnosing REDs, establish diagnostic cutoffs, and develop guidelines for their evaluation (13). It is essential to approach this condition holistically, considering factors that may influence the final diagnosis, such as the age of the athletes, their overall nutritional status, or the type and intensity of their training schedule. Furthermore, the importance of interdisciplinary and multidisciplinary collaboration in diagnosing, treating, and preventing this syndrome cannot be overstated, as it is necessary to improve the future approach to REDs. The fragmentation of complex conditions like REDs can lead to erroneous conclusions and flawed therapeutic strategies (75). The prevention of REDs should not rely solely on the sports physician. Coaches, physiotherapists, nutritional therapists, psychiatrists, the athletes themselves and, when appropriate, their parents should all be involved in every part of this process–primary, secondary, and tertiary REDs prevention (73, 76).