This study used data from a larger cross-sectional research project on REDs in high-level Icelandic athletes aged ≥15 years. The eligibility criteria and recruitment of participants have been described in detail elsewhere (26). The athletes initially responded to an online questionnaire in 2021 (July–December) consisting of the Low Energy Availability in Females Questionnaire (LEAF-Q) (27) and additional background questions. The measurement phase was between April and September 2022. Of the 56 female athletes that participated in the measurement phase, 48 logged their dietary intake and training via a photo-assisted mobile application. Seven participants provided insufficient registration (<5 days dietary intake recorded and/or no training session registered) for determination of EA. Therefore, 41 athletes were included in the present analysis. For most participants (n = 35), 7 days of registered dietary intake and training were available. Participants with 6 (n = 5) or 5 (n = 1) days were also included. The athletes came from five different sport groups, using definitions suggested by another study (28): ball (39%); endurance (24.4%); aesthetic (17.1%); weight-class (12.2%); and power sports (7.3%).

Fasted serum blood samples were collected and analysed as described earlier (26). Ferritin, iron, and total iron binding capacity (TIBC) were measured for the evaluation of iron status, and 25-hydroxyvitamin D (25-OH-Vitamin D) for vitamin D status. Using the laboratory reference values, low ferritin (adolescents <12 µg/L, adults <15 µg/L), low iron (<10 µmol/L), and high TIBC (>70 µmol/L) served as indicators of iron deficiency. In addition, transferrin saturation (TSAT) was calculated using the following formula: (iron/TIBC) × 100, with TSAT <20% considered low (43). Vitamin D insufficiency is defined as serum 25-OH-Vitamin D concentrations <50 nmol/L, and concentrations <30 nmol/L are a marker of vitamin D deficiency (40). The prevalence of concentrations below 80 nmol/L, an often-used definition of insufficiency in athletes (3, 44), was also evaluated. Other measured markers of nutrition status were calcium (reference range: 2.15–2.6 mmol/L), magnesium (reference range: 0.74–0.99 mmol/L), and vitamin B12 (reference range: 142–725 pmol/L).

Statistical analyses were conducted using IBM SPSS statistics 29.0.1.1, with significance set to α<0.05. Distributions of all data were checked using the Shapiro–Wilk test and visual inspection of Q-Q plots. Continuous variables were summarised as mean ± SD for normally distributed data, and medians with 25th and 75th interquartile ranges (IQR) for non-parametric data. Cross-tabulation and Pearson chi-square statistics were used for the evaluation of categorical data, including the occurrence of LEAF-Q, EDE-QS, EAI, and MDDI scores above cutoff.

Participant training characteristics, dietary intakes, and questionnaire scores were compared with one-way analysis of variance (ANOVA) and Kruskal–Wallis statistics. The independent samples t-test was used to compare differences in nutrient intake between athletes using sport foods and supplements compared to non-users. Body composition and nutrition status were compared based on EA + CHO categorisation (fixed factor) and adjusted for age using analysis of covariance (ANCOVA). When appropriate, Bonferroni post hoc for multiple comparisons was applied. For pairwise comparisons, mean differences (MD) and confidence intervals (95% CI) were reported for parametric data, and effect size (r = Z/√n) for non-parametric outcomes with threshold values set at 0.1 (small effects), 0.3 (moderate), and 0.5 (large) (50).

Participant characteristics, with age adjusted comparisons, are summarised in Table 1. The age range was 15–48 years but did not differ between the EA + CHO groups (Kruskal–Wallis H = 1.227, p = 0.747). The SEA + LCHO group had a higher body weight (BW) compared to the LEA + SCHO [MD = 14.0 kg (2.2–25.8), p = 0.013] and SEA + SCHO groups [MD = 16.4 kg (5.7–27.1), p < 0.001]. Group differences were observed for FFM and FFMI and post hoc analyses revealed that the SEA + LCHO group had a higher FFMI compared to the SEA + SCHO group [MD = 1.2 kg/m² (0.06–2.4), p = 0.034]. The SEA + LCHO group also had the highest body fat percentage and differed significantly from the LEA + SCHO [MD = 10.1% (2.4–17.8), p = 0.005] and SEA + SCHO groups [MD = 9.2% (2.2–16.2), p = 0.005] but not from the LEA + LCHO group. No group differences were observed for whole-body bone mineral density (BMD) Z scores and RMR. One athlete, in the SEA + SCHO group, had a RMR ratio <0.90. Of the athletes aged <45 years, 30% were using hormonal contraceptives, either oral contraceptive pills (n = 9) or other forms: ring, coil, and injections (n = 3). Previous (i.e., not current) use of oral contraceptives was reported by 10 (25%) athletes. Current amenorrhoea was reported by one athlete, and that athlete was categorised as LEA + LCHO and not taking contraceptives. A considerably larger proportion (n = 13) had a history of amenorrhoea and six reported current oligomenorrhea.

EA + CHO categorisation and training characteristics of the five sports groups are shown in Table 2. Most participants were categorised as SEA + SCHO (36.6%), of whom 70% were the endurance and 31.3% were the ball sport athletes. Moreover, half (n = 8/16) of the ball sport athletes had patterns characterised by LEA and that was accompanied by LCHO in four of them. Four out of seven aesthetic athletes but none of the weight-class athletes had LEA patterns.

Between-sport group differences were observed for weekly number of training hours and daily EEE. Multiple comparisons were not significant for EEE after Bonferroni corrections were applied. However, average training hours from the training records were higher in aesthetic athletes than ball sport athletes [MD = 6.4 h (0.95–11.8), p = 0.012]. The weekly number of training hours, derived from the questionnaire, was also higher in aesthetic sports compared to ball [MD = 6.0 h (1.3–10.7), p = 0.005] and weight-class [MD = 6.4 h (0.4–12.4), p = 0.030] sports.

In addition to differences in CHO intake, intakes of protein, fat, and fibre differed by groups (Table 3). More specifically, the LEA + LCHO group had the lowest relative intake of all macronutrients and the SEA + SCHO group had the highest. The average protein intake differed between SEA + SCHO and LEA + LCHO [MD = 0.6 g/kg (0.04–1.1), p = 0.028]. SEA + SCHO also had higher fat intakes compared to SEA + LCHO [MD = 0.4 g/kg (0.1–0.7), p = 0.004] and LEA + LCHO [MD = 0.6 g/kg (0.3–0.9), p < 0.001]. Finally, fibre intake was higher in SEA + SCHO compared to LEA + LCHO [MD = 0.2 g/kg (0.02–0.47), p = 0.028].

A total of 36 (87.8%) athletes used sport foods and/or supplements, of whom, two used vitamins and minerals only. Vitamin D intake was <15 µg in 82.9% of all athletes but 12 (29.3%) took supplements with vitamin D and their total intake averaged at 19.5 µg compared to 5.1 µg in those who did not supplement. Vitamins without D were taken by 5 (12.2%) and minerals or electrolytes by 11 (26.8%). Iron intake was <15 mg in 78% of all athletes, but supplements with iron were taken by 4 (9.8%). Folate intake was <300 µg in 46.3% of all. None had low intakes of vitamin B12.

In total, 32 (78%) athletes used protein supplements and/or protein-enriched products. Of them, 27 (65.9%) used protein-enriched dairy or ready to serve protein drinks, 13 (31.7%) used protein powders, and 21 (51.2%) used protein bars or snacks. The average protein intake of those who used protein products was 1.7 g/kg (range 1.0–3.1) compared to 1.3 g/kg (range 0.8–1.8) for those who did not (p = 0.019). The use of energy drinks and/or pre-workout products was reported by 14 (34.1%) athletes, exogenous CHO, such as sport drinks, gels, and bars, by 10 (24.4%) and ergogenic aids such as creatine by 4 (9.8%). The contribution of sport foods and supplements to the total protein intake ranged between 12% in the LEA + SCHO group and 19.7% in the SEA + LCHO group. Similarly, sport foods and supplements contributed 2.9 (LEA + SCHO) to 8.0% (SEA + LCHO) to the total CHO intake.

Between-group differences were not observed for nutrition status (Table 4). None of the SEA + SCHO athletes had vitamin D deficiency or insufficiency, but one in the LEA + LCHO group was deficient (<30 nmol/L) and three were insufficient (<50 nmol/L). Moreover, three athletes in the LEA + SCHO group and two in the SEA + LCHO group were deficient in vitamin D. Of all participants, 31 (75.6%) had vitamin D concentrations below the frequently used cutoff for insufficiency in athletes, i.e., 80 nmol/L. No apparent differences were found in vitamin D status between those who used vitamin D supplements and those who did not (66.7 ± 21.7 vs. 65.3 ± 16.9 nmol/L, p = 0.838). Ferritin was below reference for age in five (12.2%) participants, and was accompanied by low iron and/or high TIBC in four. In addition, six athletes had low levels of iron only, of which four were normally menstruating. Of those five with low ferritin, three were normally menstruating. TSAT was <20% in 10 athletes, with no apparent group differences. No athlete was deficient in vitamin B12, magnesium, and calcium.

Approximately 60% of participants in this study came from ball and endurance sports, with the latter sport group often referred to as being at high-risk of REDs and disordered eating due to pressure to be thin or beliefs that a lower body weight leads to better performances (4, 51). Interestingly, 70% of the endurance athletes had both sufficient to optimal EA and CHO, compared to one-third of the ball sport athletes. Moreover, 50% of the athletes with LEA + LCHO patterns were ball sport athletes. The lower number of athletes from the three other sport groups challenge further investigation of sport-specific risk. Yet our findings support that occurrence of REDs is not limited to certain types of sports; moreover, that individual characteristics and various external factors appear to have a bigger impact than type or nature of the training per se. What eventually dictates the onset of REDs is insufficient energy intake in relation to training demands (4, 52). Indeed, the training characteristics of participants varied substantially, with the highest number of training hours reported by aesthetic athletes and the lowest by ball and weight-class athletes. Therefore, it appears that the relatively high occurrence of LEA + LCHO in the ball and power sports is not explained by greater training demands or EEE, but rather reasons such as unawareness of energy/nutrition requirements or dietary restrictions (53). In comparison, five out of nine LEA + SCHO athletes were from aesthetic and endurance sports where EEE is often very high. Accordingly, Melin et al. (54) reported a 7-day average EEE of 1,222 kcal and total EE of 3,266 kcal in endurance athletes displaying LEA. Moreover, Brown et al. (55) reported a 7-day average total EE of ∼2,800 kcal and EA of 26 kcal/kg FFM in pre-professional female dancers.

Nutrition periodisation, characterised by adaptable LEA and CHO intakes tailored to the demands of training, is a common practice in endurance and other sports (20, 21). While athletes were not asked if they periodised their nutrition or were following a specific diet, such approaches were likely adhered to by some of the LEA + SCHO athletes. Athletes with high EEE and long training days may also have prioritised CHO intake, while challenged by limited eating opportunities and/or inability to fulfil total energy requirements (51).

Body composition, primarily body fat percentage, differed between the EA + CHO groups. More specifically, the SEA + LCHO group had higher body fat levels compared to the LEA + SCHO and SEA + SCHO groups but not the LEA + LCHO group. However, the cross-sectional nature of this study limits the possibility to investigate potential causal relationships of this finding with REDs. Sport-specific training adaptations and physiological demands, genetics and a plethora of individual factors influence adiposity (56). Theoretically, the observed group differences in body fat levels could be rooted in energy conservation, including reduction in energy metabolism in response to prolonged or repeated LEA and/or LCHO exposure (57). In contrast, no differences were seen for RMR and whole-body BMD.

This study built on the assumption that more frequent exposure to LEA and LCHO could better predict the risk of problematic LEA compared to calculated averages alone (22).

For all participants, the average CHO intake was marginally higher than the 3.0 g/kg BW cutoff used for the categorisation but varied considerably between days for many. Indeed, CHO requirements are subject to change based on training demands and it is of great importance for athletes to be aware of it and ensure that CHO intake is sufficient for the work required (58). Low or restricted intakes of CHO also increase the likelihood of insufficient intakes of other essential nutrients as well. Nutrients work in synergy to support metabolism and other body functions, and any modifications of athletes’ diet must therefore be well considered to avoid nutrient inadequacies or deficiencies (19). Current sport-specific nutrition recommendations for protein are 1.2–2.0 g/kg BW (59, 60). The average protein intake of all athletes in the two SCHO groups exceeded 1.0 g/kg BW while there were a few cases with average intakes <1.0 g/kg BW in both LCHO groups. Thus, exposure to LEA and LCHO appears to modulate protein intake, which may consequentially result in missed opportunities for recovery and training adaptations (59, 60). No group differences were found in energy and macronutrient intakes from sport foods and supplements, which suggests that such products were often used to compensate low dietary intakes or seen as a convenient solution. In accordance, poor diet is among potential reasons for using supplements, with ∼30% of track and field athletes reporting this reason (61). Moreover, dietary restrictions seemed to be primarily focused on food sources but not sport foods and supplements, perhaps due to beliefs that the latter is healthier and/or provides athletic advantages (31). In agreement with the literature (62), vitamin D and iron intakes were below the recommended intakes for the majority of all athletes, and this points towards important room for improvement in favour of bone health and wellbeing.

The IOC has listed adequate vitamin D status (>30 ng/mL/∼80 nmol/L) as critical, especially for athletes at risk and/or those recovering from REDs, for the sake of bone health and reduced risk of bone stress injuries (4). The role of vitamin D in skeletal muscle function and sport performance has also been highlighted (63). Of all participants, only 25% reached adequate intake levels for athletes. Moreover, half of the LEA + LCHO athletes had vitamin D insufficiency or deficiency (<50 nmol/L) compared to none of SEA + SCHO. Measurements in this study were conducted at one time point between April and September, with majority of the females in the larger research project measured in the spring/early summer (April–June). In Iceland and the other Nordic countries, it is recommended to supplement vitamin D, especially in the wintertime (40). It is thus possible that those who took vitamin D supplements in the wintertime had recently stopped or taken a break from supplementation in the springtime. That could indeed explain why those who currently took vitamin D supplements did not have higher serum concentrations compared to those who did not report current use. Moreover, seasonal variation in vitamin D status among elite athletes has been reported in the literature, with highest levels measured in the late summer but lowest in late winter (64). Therefore, and given that many barely exceeded 50 nmol/L, it is possible that the incidence of deficiency for vitamin D would have been different if measurements had been taken at other seasons. Accordingly, major determinants of vitamin D status are sun exposure, supplementation and regular intake of foods high in vitamin D (40). Although between EA + CHO group differences were not significant, low levels in the LEA + LCHO group spark a special worry in terms of long-term bone health. Moreover, based on the presented data, many of the participating athletes would benefit from yearlong vitamin D supplementation.

It has been suggested that suboptimal iron status can be either a cause or consequence of REDs, although this remains to be supported by more robust scientific evidence (65).

The present study found no group differences for any of the iron markers. The design of this study, and the fact that it includes females with variable menstrual function, does not allow for deep evaluation of the interrelations between iron status and occurrence of REDs. However, the findings point out that potential relationships (or lack thereof) of iron metabolism with REDs are likely complicated by menstrual characteristics and/or use of contraceptives in females. Accordingly, low iron and ferritin levels were predominantly found in normally menstruating females. It has been well established that menstruation is a primary non-exercise-related cause of iron loss in females, and therefore ferritin levels are commonly lower in females than males (66). Moreover, exercise-related factors, such as haemolysis, haematuria, and gastrointestinal bleeding, contribute towards increased blood loss, while elevated post-exercise hepcidin levels potentially impair iron absorption from meals consumed soon after exercise (67). Whether iron status is directly linked to REDs or not, it holds a key role in oxygen transport, fuel utilisation, and other key functions, and is therefore extremely important for athletic performance. Moreover, co-occurrence of problematic LEA and iron deficiency can make a bad situation considerably worse (68).

The present study has some limitations. First, due to its explorative nature, it should first and foremost be regarded as a step towards further understanding of the intersection between adaptable and problematic LEA. Therefore, a larger and/or better controlled study (e.g., in terms of sport groups) with other outcomes, such as site-specific bone mineral density (i.e., not only whole body) and the primary REDs indicators recently suggested by the IOC (4, 36), could have resulted in somewhat different or more comprehensive findings. It should be acknowledged that some errors in evaluating dietary intake and energy expenditure are inevitable. However, we strived to limit such errors by using a photo-assisted mobile application that was specially tailored for the convenient reporting of dietary intake and training. Accordingly, it has been suggested that the replacement of traditional approaches with digital applications may reduce chances of dietary intake misreporting and even, importantly, reduce participant burden (74). We had one-on-one discussions with each athlete to ensure they were fully informed about the aims of this reporting, and the importance of not making changes to their routines due to participation in the study. It must also be clearly stated that MET scores do not provide precise individualised estimates of EEE, as they were primarily designed for the standardisation of physical activity intensities (42). Extracting measured but not estimated RMR from total energy expenditure to yield energy cost of exercise alone, as was done for most participants in this study, partly compensates for this limitation (75). Also important is that coding, calculations, and cleansing of nutrition and training data were performed by well-trained experts who each had their separate task in the process. Therefore, any potential errors should apply to estimations for all participants, which allows for reliable comparisons. Seven days is only a snapshot of the individuals’ life and does not provide information on potential variations between weeks, months, or training periods. That, in addition to the study design, does not allow for any conclusions on causality. Finally, selection bias and other considerations for the greater research project have been addressed elsewhere (26).