Journal Logo

SPECIAL COMMUNICATIONS: Methodological Advances

Measuring the Exercise Component of Energy Availability during Arduous Training in Women

Gifford, Robert M.1,2; Greeves, Julie P.3,4; Wardle, Sophie L.3; O’Leary, Thomas J.3; Double, Rebecca L.3,4; Venables, Michelle5; Boos, Christopher6; Langford, Joss7,8; Woods, David R.2,6,9; Reynolds, Rebecca M.1

Author Information
Medicine & Science in Sports & Exercise: April 2021 - Volume 53 - Issue 4 - p 860-868
doi: 10.1249/MSS.0000000000002527

Abstract

Low energy availability (EA) (insufficient energy intake [EI] in relation to exercise energy expenditure [EEE]) has important physiological and performance ramifications for male and female athletes (1). The measurement of EA is challenging and varies between studies. Measuring EI is notoriously challenging, being hampered by systematic underreporting (2), whereas EEE is defined and measured inconsistently (3). Some have extrapolated the EEE of total physical activity (EEEtpa) from calorimetry or doubly labeled water (DLW) (4–6), whereas work demonstrating the importance of low EA (7), and subsequent studies (8–10), measured EEE directly from moderate and vigorous physical activity (EEEmvpa). In a prospective study of 35 women, Lieberman et al. (11) demonstrated that low EA measured from purposeful EEEmvpa was linearly related to ovulatory dysfunction. Thus, EEEmvpa is advantageous in that it focuses on specific activity addressed by the relative energy deficiency in sport or female athlete triad paradigms; however, measurement is often hampered by reporting bias, and direct measurement is normally impossible in the field. As the importance of low EA becomes increasingly apparent, there is a pressing need to develop reliable and feasible methods for real-world measurement (3).

The context of basic military training is germane for developing EA measurement because it is characterized with high physical demands and multiple stressors, in free-living but well circumscribed field environment. It is repeatable (12) and provides routine measures of training adaptation (1). Furthermore, because a ban on women joining the infantry has recently been lifted, women may be required to train more arduously, which could put them at increased risk of conditions associated with low EA, like premature osteoporosis, increased cardiovascular risk, or reproductive dysfunction (1,13,14).

Low EA may impair cardiovascular adaptation to exercise (1), regulated by the parasympathetic and sympathetic nervous systems (PNS and SNS). Increased resting PNS activity, considered a beneficial effect of exercise (15), is manifested by higher heart rate variability (HRV). Conversely, when SNS activity predominates, lower HRV is found and may accompany overtraining, psychological stress, and restricted sleep (16,17).

This study aimed to compare EA measured from EEEmvpa, measured using an open-source accelerometry technique, with EA based on EEEtpa using DLW, in women during an 11-month basic military training program. The secondary aim was to determine the relationship of EAmvpa and EAtpa with putative benefits of training, namely, physical adaptations (improved 1.5-mile run time and body composition changes) and autonomic adaptation (increased resting PNS activity), together with evidence of disordered eating behavior. These were assessed, respectively, by fitness test scores and body composition, HRV, and the Brief Eating Disorder in Athletes Questionnaire (BEDA-Q). We hypothesized that EEEmvpa as measured using accelerometry would correlate significantly with EEEtpa, as measured using DLW, and EA calculated from EEEmvpa would lead to a relative overestimation of EA. We hypothesized that EAmvpa and EAtpa would be associated with concordant changes in physical and autonomic adaptation and eating behavior.

METHODS

Participants and setting

Women commencing the British Army Officer Commissioning Course at the Royal Military Academy, Sandhurst, were invited to participate. Participants underwent a routine detailed medical screen, which included a full history, physical examination and an ECG before participation, to meet exacting medical standards mandated before employment in the Army (18). The entry medical included a review of preexisting medical records for a multiplicity of conditions, before enrolment in the Army, including diagnosed thyrotoxicosis, eating disorder, malabsorption, or food intolerance. The study was approved by the UK Ministry of Defence Research Ethics Committee (790/MoDREC/16) and was conducted in accordance with the Declaration of Helsinki. All participants provided written informed consent.

The study followed a repeated-measures design summarized in Figure 1. Height was measured at visit 1 (Seca stadiometer model 217, Birmingham, UK) and weight at every study visit (Seca scales model 874), wearing T-shirt and combat trousers or shorts. Eating behavior and body composition were measured at the beginning and end of each term. Body composition was measured by dual-energy x-ray absorptiometry (GE Lunar iDXA, GE Healthcare Systems, Chalfont St. Giles, UK) at study visits 1 (week 1), 2 (week 14), 4 (week 29), and 6 (week 43), wearing T-shirt and shorts. A self-constructed physical activity and diet questionnaire was completed at visit 1 with reference to the preceding 6 months, comprising 16 questions on exercise and diet (see Table, Supplemental Digital Content 1, Exercise and diet at the commencement of the study, https://links.lww.com/MSS/C158).

F1
FIGURE 1:
Scheme of study visits and EAP. ① Recruitment, ② height, ③ weight, BEDA-Q ④ dual-energy x-ray absorptiometry (DXA) and HRV, ⑤ 10-d EAP. PCBC, precourse briefing course, 6 to 20 wk before start of term 1.

EA assessment phases

Once per term, TEE, EEE, and EI were measured over 10-d EA assessment phases (EAP, denoted 1 to 3). EAP were selected in consort with training staff to be representative of the entire course. Physical activity was a prominent feature of the course, and working days lasted 13.0 ± 4.0 h·d−1 (mean ± SD; see Document, Supplemental Digital Content 2, Description of the Commissioning Course, https://links.lww.com/MSS/C159).

EEE and EA calculation

We measured EEEmvpa using wrist-worn GENEActiv Original triaxial accelerometer (Activinsights, Cambridgeshire, UK). The device was worn for 24 h·d−1 throughout each EAP, sampling at 75 Hz. Data were processed with the GENEAread R package from CRAN (19) using a customized, openly available script (20). The data were calibrated, and days with more than 7 h nonwear were excluded: 1322 d (93%) of valid wear were included (21,22). The mean absolute gravity-subtracted acceleration was calculated for each 1-min epoch within a 24-h period, per participant per EAP. Acceleration accumulated in sedentary activities was separated from acceleration accumulated in MVPA using a cut point of 0.09g (23). Moderate and vigorous activities during each EAP were expressed as metabolic equivalents (METs), based on programmed activities (24). EEE was calculated from the accumulated duration spent undertaking moderate and vigorous activity for each EAP as follows:

EEEmvpa=tmvpa×MET×3.5×0.0049×weight

where tmvpa is the mean daily duration (min) of moderate and vigorous physical activity, MET is the mean daily MET of activity, 3.5 is the assumed oxygen cost for one MET (mL·kg−1⋅min−1), and 0.0049 is the calorific value (kcal) of 1 mL oxygen and weight (kg) before the EAP.

During each EAP, TEE was measured using DLW. In brief, a baseline urine sample was provided, before 174 mg·kg−1 body weight H218O and 70 mg·kg−1 body weight 2H2O were ingested. Ten consecutive daily urine samples were then obtained from which isotopes were measured (see Document, Supplemental Digital Content 3, Doubly labelled water method, https://links.lww.com/MSS/C160). Analytical precision was 0.3 ppm for 2H and 0.5 ppm for 18O, and a method precision of 1.2%. Because of the compressed nature of the course, it was not feasible to conduct indirect calorimetry. Therefore, resting metabolic rate (RMR) was estimated from fat-free mass (FFM) according to the equation of Cunningham et al. (25), and the energy expenditure of all physical activity (EEEtpa) was calculated by subtracting RMR from TEE.

RMR=370+21.6×FFM
EEEtpa=TEERMR

where EEEtpa is the EEE of total physical activity and TEE is total energy expenditure.

Energy intake was measured using a 24-h food diary, aided by interview with researchers at the end of each day to prompt missed items. Meals served in the canteen were weighed in a validatory cohort (see Document, Supplemental Digital Content 4, Energy intake assessment, https://links.lww.com/MSS/C161). All food diary data were entered into the Nutritics database (Dublin, Ireland) by the same member of the research team (RLD) to calculate EI, before EA was calculated as follows:

EAmvpa=EIEEEmvpaFFM
EAtpa=EIEEEtpaFFM

where EAmvpa is the EA after moderate and vigorous physical activity, EI is the energy intake, EEEmvpa is the EEE of moderate and vigorous physical activity, EAtpa is the EA after total physical activity, and EEEtpa is EEE of total physical activity.

Physical and autonomic training adaptation and eating behavior

A best-effort 1.5-mile (2.4 km) run test was undertaken during the same week as, but on a different day to, study visits 1, 2, 3, 5, and 6 (Fig. 1). This test is a good indicator of cardiorespiratory fitness and correlates strongly with maximal rate of oxygen uptake (V˙O2max; r = 0.79, 95% confidence interval = 0.73 to 0.85) (26). HRV was measured at study visits 1, 2, 4, and 6, as described previously (27). A 5-min single-lead ECG was measured using CheckMyHeart™ devices (DailyCare Biomedical, Taoyuan, Taiwan), according to manufacturer’s instructions, yielding time domain, frequency domain, and nonlinear metrics of PNS and SNS balance (see Document, Supplemental Digital Content 5, Heart rate variability measurement, https://links.lww.com/MSS/C162).

Eating attitudes were assessed using the BEDA-Q, a sensitive screen for low EA (28). The binary item “Are you dieting?” was scored at each study visit, and “Have you ever dieted?” was scored at study visit 1 only. Questionnaires were completed on a Web-based application (SmartSurvey, Tewkesbury, UK).

Statistical analysis

Data were analyzed using IBM SPSS for Macintosh (Version 24.0, Armonk, NY). Baseline characteristics of participants who completed the study were compared with those who did not using independent samples t-tests. For each EAP, TEE was compared with EI, EEEtpa with EEEmvpa, and pre- with post-EAP weight using paired samples t-tests; EAP were compared using repeated-measures ANOVA. Partial correlations and linear regression were used to describe the relationship between EEEtpa and EEEmvpa. Systematic bias between EAmvpa and EAtpa was assessed by the methods of Bland and Altman (29). Where EAmpva data were missing because of technical issues (loss of, or failure to deploy, accelerometers), EAmpva was imputed from EAtpa via the regression equation for that EAP (41 EAP exposures, 24%) before determining the relationship with adaptation.

Repeated-measures ANOVA was used to compare weight, fat mass, FFM, 1.5-mile run time, HRV indices, and BEDA-Q scores across visits. Spearman’s rank correlation (rs) was used to assess dieting status over time. Physical adaptations were taken as the difference between post- and pre-EAP measurements: loss of FM, gain of FFM, and improvement of 1.5-mile run time. Partial correlations were used to assess associations between EA measures and physical adaptations, HRV (measured at the study visit after each EAP), and BEDA-Q score (measured at the study visit before each EAP). Point-biserial correlation (rpb) assessed relationships between EA measures and dieting status (whether dieting or not). Independent samples t-tests were used to compare EA measures between participants who reported ever having dieting and participants who did not. Alpha was set at P < 0.05, except for multiple correlations of training adaptations with EA, where Bonferroni adjustment was made (HRV was treated as one adaptation, adjusted alpha P < 0.0083).

RESULTS

Participants

Recruitment and loss to follow-up are illustrated in Figure 2. Fifty-nine women attended study visit 1, and 47 women completed three EAP (mean ± SD; 23.9 ± 2.6 yr, baseline BMI = 23.3 ± 2.1 kg·m−2). Age, height, BMI, and body composition did not differ between participants who the completed study and those who withdrew, although more women who completed the study reported ever dieting than those who withdrew, as described in the table in Supplemental Digital Content 6 (Baseline evaluation of participants at baseline who completed all study measures with those who did not, https://links.lww.com/MSS/C163).

F2
FIGURE 2:
Recruitment and follow-up. EA Phase: energy availability assessment phases where energy requirement was measured using multipoint doubly labeled water and EI and EEE were estimated. At “study visits,” weight, HRV, and body composition were measured, and questionnaires were completed. *Two participants declined and two provided insufficient urine samples. †Two declined and four provided insufficient urine samples.

Participants reported exercising more than most women their own age before the study, particularly running and weight training. Thirty-five participants (51%) reported skipping meals beforehand, but only 2 (3%) reported doing so after starting the course. Six participants (10%) were vegetarian (four lacto-ovo vegetarians and two vegan). Detailed diet and exercise findings are shown in the table in Supplemental Digital Content 1 (Exercise and diet at the commencement of the study, https://links.lww.com/MSS/C158).

EAP

As expected, EEEtpa was higher than EEEmvpa during EAP 1, 2, and 3 (mean ± SD difference = 452 ± 358, 504 ± 544, and 506 ± 413 kcal·d−1, respectively, all P < 0.001) (Table 1). TEE was higher than EI during EAP 1, 2, and 3 (energy balance = −654 ± 558, −1573 ± 578, and −673 ± 663 kcal·d−1, respectively, all P < 0.001).

TABLE 1 - EA measurements.
Phase 1 Phase 2 Phase 3 η p 2 P
Weight before/after, kg 64.4 ± 7.7 65.6 ± 7.7 65.0 ± 7.6 63.7 ± 7.4 65.8 ± 7.7 65.3 ± 7.6
Difference, kg +1.2 ± 1.1, P < 0.001 −1.3 ± 3.0, P = 0.003 −0.3 ± 1.3, P = 0.11 0.471 <0.001
EI, kcal·d−1 2667 ± 696 2320 ± 574* 2358 ± 422* 0.115 0.003
TEE, kcal·d−1 3332 ± 424 3849 ± 363* 3041 ± 286* 0.321 <0.001
EEEtpa, kcal·d−1 2228 ± 355 2811 ± 455* 1963 ± 325* , 0.695 <0.001
EEEmvpa, kcal·d−1 1865 ± 312 2253 ± 536* 1513 ± 336* , 0.752 <0.001
EAtpa, kcal·kg−1·d−1 8 ± 11 −10 ± 11* 9 ± 12* , 0.511 <0.001
EAmvpa kcal·kg−1·d−1 18 ± 13 1 ± 13* 23 ± 15* , 0.419 0.001
Values are presented as mean ± SD. P for repeated-measures ANOVA (main effect of time); ηp2, partial eta squared.
*Significant (P < 0.05) difference vs phase 1.
Significant vs phase 2.
TEE, total energy expenditure; EEE, either from total physical activity (tpa; measured by doubly labeled water) or from measured moderate and vigorous physical activity (mpva; measured by accelerometry); EA for each measure of EEE, expressed as kilocalorie per kilogram FFM per day.

Comparisons of EA from accelerometry with EA from doubly labeled water

Partial correlations between EAmvpa and EAtpa were strong at all EAP (Fig. 3A and B); the linear regression equations are shown in the table in Supplemental Digital Content 7 (Linear regression equations of energy availability measured by accelerometry with energy availability measured by doubly labeled water, https://links.lww.com/MSS/C164). Across the range of measurement, EEEtpa was higher than EEEmvpa by 10.2 (SD ±8.3) kcal·kg−1 FFM·d−1; therefore, EA was lower when using EEEtpa (Fig. 3C).

F3
FIGURE 3:
Comparisons of EAtpa and EAmvpa. A, Scatter plot of all paired EAtpa against paired EAmvpa values with overall linear regression equation. B, EAP plotted separately. C, Bland–Altman Plot, demonstrating difference between EAmvpa and EAtpa at the range of values measured. D and E, Change in weight during 10-d assessment EAP (weight after − weight before) plotted against EAtpa and EAmvpa, respectively. In panels B to E, blue circle represents EAP 1, unfilled red circles EAP 2, and green triangles EAP 3. EAtpa, EA from total physical activity (measurement based on total energy expenditure from doubly labeled water); EAmvpa, EA from moderate and vigorous physical activity (measurement based on accelerometry).

Training adaptation and eating behavior

Participants gained weight during EAP 1, lost a similar amount of weight during EAP 2, and EAP 3 was weight neutral (Table 1). Overall, term 1 was weight neutral, but participants demonstrated modest gains in FFM and loss in FM. These beneficial changes were then reversed: weight and FM were higher than study baseline by the end of term 2 with no change in FFM. During term 3, weight and body composition regressed to baseline levels. Throughout the study, HRV metrics demonstrated beneficial adaptations, in particular from time domain, PNS, and SNS indices and sample entropy. Table 2 shows physical, autonomic, and eating behavior scores. A narrative detailing body composition changes during training can be found in Supplemental Digital Content 8 (Body composition changes, https://links.lww.com/MSS/C165).

TABLE 2 - Physical, autonomic, and eating behavior changes.
Visit 1 Visit 2 Visit 3 Visit 4 Visit 5 Visit 6 η p 2 P
Weight, kg 64.1 ± 7.9 63.7 ± 7.9 64.5 ± 7.9 65.0 ± 7.7* 65.2 ± 7.9* 63.9 ± 7.8 0.073 0.006
FFM, kg 49.6 ± 5.3 50.2 ± 5.4* 49.4 ± 5.0 49.4 ± 5.0 0.072 0.041
Fat mass, kg 15.6 ± 4.0 14.4 ± 3.9** 16.1 ± 4.0* 15.6 ± 4.2 0.245 <0.001
1.5-mile run time, mm:ss 10:41 ± 1:02 10:08 ± 0:55** 10:20 ± 0:57** 10:38 ± 0:58 10:29 ± 1:00* 0.399 <0.001
Heart rate 76.4 ± 11.9 69.7 ± 9.6* 71.2 ± 8.9* 68.2 ± 10.3** 0.122 <0.001
HRV
Time domain
 RMSSD, median [IQR] 35.7 [23.5 to 56.4] 51.4 [33.6 to 66.81]* 45.9 [37.6 to 55.29]* 47.2 [30.6 to 66.6]* 0.075 0.036
 pNN50,%, median [IQR] 14.7 [4.9 to 31.6] 28.4 [11.1 to 43.8]* 24.4 [16 to 35.7]* 27.6 [9.6 to 44.9]* 0.079 0.003
Frequency domain (fast Fourier transformed)
 LF (log) 6.87 ± 0.88 7.09 ± 0.97 6.84 ± 0.77 7.00 ± 1.07 0.002 0.70
 HF (log) 6.25 ± 1.19 6.70 ± 1.17* 6.48 ± 0.93 6.59 ± 1.28 0.028 0.11
 LF:HF, median [IQR] 1.8 [1.2 to 3.3] 1.4 [0.8 to 3.1] 1.8 [1.1 to 2.5] 1.6 [0.9 to 2.5] 0.021 0.17
Nonlinear
 Sample entropy 1.61 ± 0.26 1.69 ± 0.24 1.72 ± 0.25* 1.75 ± 0.26* 0.149 0.003
Autonomic nervous system indices
 PNS index, median [IQR] −0.7 [−1.4 to 0.2] 0.1 [−0.7 to 0.6*] −0.2 [−0.6 to 0.1]* 0.3 [−0.7 to 1.2]** 0.090 0.001
 SNS index 0.86 ± 1.35 0.14 ± 1.09* 0.30 ± 0.89* 0.12 ± 1.26** 0.083 0.002
BEDA-Q score, median [IQR] 3 [1 to 4] 3 [1 to 5] 3 [2 to 5] 4 [2 to 6] 3 [1 to 5] 3 [2 to 5] 0.034 0.70
BEDA-Q dieting (“yes”), n (%) 11 (18.6) 7 (13.0) 5 (15.6) 14 (29.2) 16 (35.6) 16 (30.8) 0.139 a 0.010 a
Data are presented as mean ± SD unless otherwise stated. BEDA-Q score was log-transformed before analysis. For continuous variables, P values refer to repeated-measures ANOVA (main effect of time).
*Pairwise difference with visit 1 (P < 0.05).
**Pairwise difference with visit 1 (P < 0.001).
aSpearman’s correlation for BEDA-Q dieting (dichotomous) with visit week.
IQR, interquartile range; ηp2, partial eta squared; RMSSD, root-mean-square of successive differences; pNN50, percentage of successive normal R-R intervals above 50 ms; LF, low-frequency power; HF, high-frequency power; log, transformed by natural logarithm.

Correlations of EA with training adaptations and eating behavior

As demonstrated in Table 3, physical training adaptations correlated weakly with one another. FM loss and improvement in 1.5-mile run time correlated inversely with BEDA-Q score. There was no association between BEDA-Q scores or physical adaptations and autonomic adaptation. Increasing EA was associated with increased 1.5-mile run performance and fat mass loss. There was no correlation between EA and FFM gain or autonomic adaptation. As expected, EA demonstrated a modest negative association with BEDA-Q score and dieting status. Average EAtpa and EAmvpa were lower among participants who reported ever dieting, compared with those did not (−0.18 ± 11.7 vs 5.5 ± 16.1 kcal·kg−1 FFM·d−1, P = 0.016, and 22.0 ± 13.0 vs 29.0 ± 15.6 kcal·kg−1 FFM·d−1, P = 0.004, respectively).

TABLE 3 - Correlations between EA measures and training adaptation.
EAtpa EAmvpa FM Loss FFM Gain 1.5-mile Run Improvement RMSSD (Log) pNN50 (Log) LF (Log) HF (Log) LF:HF (Log) Sample Entropy PNS Index (Log) SNS Index BEDA-Q Score
FM loss 0.376** 0.373**
FFM gain 0.161 0.213 0.222**
1.5-mile run improvement 0.284** 0.252** 0.482** 0.082
RMSSD (log) 0.203 0.208 0.007 0.154 0.087
pNN50 (log) 0.182 0.193 −0.037 −0.23 −0.165 0.838**
LF (log) −0.118 −0.097 −0.018 −0.005 −0.227 0.730** 0.484**
HF (log) 0.252 0.227 −0.006 0.201 −0.115 0.918** 0.847** 0.678**
LF:HF (log) −0.197 −0.182 0.044 −0.204 −0.067 −0.271 −0.472** 0.276* −0.445**
Sample entropy 0.171 0.204 0.047 0.059 −0.048 0.2938 0.054 −0.302 0.286 −0.391**
PNS index (log) 0.168 0.188 0.068 0.028 −0.042 0.806** 0.823** 0.564** 0.761** −0.348** −0.018
SNS index −0.163 −0.184 −0.035 0.05 0.074 −0.799** −0.762** −0.672** −0.679 0.413** −0.321** −0.880**
BEDA-Q score (log) −0.367** −0.321* −0.285* 0.036 −0.302 −0.318 −0.19 −0.246 −0.299 0.133 0.029 −0.157 0.038
BEDA-Q—“Are you trying to lose weight?”pb −0.211* −0.206* 0.037 −0.057 0.037 0.002 0.007 0.024 −0.010 0.062 0.024 −0.088 0.120 0.343**
Correlations are between EAtpa, EAmvpa measurements during energy assessment phases, and pre- to post-phase training adaptation, or between concurrent training adaptation measures. All are partial correlations (taking account of repeated measures in individuals) except marked “pb” (point-biserial nonparametric correlation). EA is measured by total physical activity (tpa, from total energy expenditure) or moderate and vigorous physical activity (mvpa, from accelerometry), FM and FFM loss, and 1.5-mile run improvement: difference in fat mass, FFM, and 1.5-mile best-effort run time, respectively, from pre to post-EA measurement (pre minus post). HRV and BEDA-Q are measured before each EA measurement. Associations of HRV variables are italicized; these are expected to correlate strongly with one other.
Significant correlations after Bonferroni adjustment: **P < 0.0001; *P < 0.008.
RMSSD, root-mean-square of successive differences; log, transformed by natural logarithm; pNN50, percentage of successive normal R-R intervals above 50 ms; IQR, interquartile range; LF, low-frequency power; HF, high-frequency power.

DISCUSSION

EA based on estimation of moderate and vigorous physical activity using accelerometry demonstrated a strong agreement with EA based on total physical activity from the gold standard technique across the measurement range in this setting of multistressor military training. EAmvpa was higher than EAtpa by 10 kcal·kg−1 FFM·d−1, which likely represents the energy expenditure difference between “total physical activity” and “moderate and vigorous exercise,” i.e., nonexercise activity.

In this study, all EA values were ostensibly well below the purported threshold of low EA for which has been previously been mooted and then refuted (7,11). We found that increased EA was associated with improved run times and body composition during the entire military training. TEE and EEE were commensurate with reports in athletes (2). On average, EI was 26% ± 8% (958 ± 732 kcal·d−1) lower than TEE. Based on an average tissue density of 7000 kcal·kg−1 for adult women (30), such an energy deficit would be associated with an average weight loss of 1.06 kg per EAP. Instead, we observed no significant weight change during EAP on average, implying crude energy balance. Applying the 95% confidence intervals of EI:EE ratio plausibility, derived from the Goldberg cutoff (0.82 to 1.18 [31]), 73% of our measurements fell below this level (mean ratio = 0.72 ± 0.19). We therefore surmise EI was underestimated, possibly because of participant motivation, fatigue during prolonged measurement durations, competing pressures from the course itself, or a combination of these. Such underreporting has been identified widely elsewhere (2,32), including in similar settings of military training (33). Thus, although we have demonstrated the validity of the EEE component of an EA assessment, the usual limitations to EI assessment apply in this population.

Both EAmvpa and EAtpa were positively associated with loss in FM and improved 1.5-mile run time. These findings are particularly noteworthy for the very low measured EA at which they occurred, underlining the linearity of EA’s effects on physical adaptation and performance, rather than a threshold below which its effects are seen (11). The adaptations we observed could be interpreted as relating to “energy compensation” after lower EA experienced during EAP, i.e., compensatory increases in EI and reduction in nonexercise activity 5 to 9 wk between the EAP and the following study visit (34). Military training involves rapid changes in the volume and nature of physical activity, and although our participants reported active lifestyles beforehand, they found the training intensity highly challenging, as reported in a linked manuscript on stress responses in these participants (35). Changing feeding habits takes several years (36), so adjustment of habitual EI after the abrupt and immersive onset of initial military training was likely to have been delayed (37). Studies of controlled exercise protocols have found that interventions were followed by compensatory EI increases (38), increased FM, and reduced FFM (39). A synthesis of two randomized, controlled trials of exercise interventions for obesity showed that such energy compensation was inversely associated with peak oxygen uptake (V˙O2peak) (40). Military training is not an exercise intervention per se, but our study suggests that similar energy compensation took place in response to a multistressor environment, which could be relevant for women undertaking a wide range of physically demanding employment. This context makes it more remarkable that performance improvements were observed overall.

We found positive autonomic adaptations throughout the study, but these were not correlated with EA. Increased parasympathetic and decreased sympathetic activity were observed in time domain measures consistently throughout the course, especially during term 1. Time and frequency domain measures were slightly below means for athletes reported elsewhere (41). HRV has been measured in studies of psychological stress as well as exercise and may decrease after negative effects of psychological stress (17) and increase with improved aerobic capacity (42). Although psychological stress was experienced throughout the course (35), we demonstrated autonomic benefits, independent of improvement in cardiovascular fitness or EA. In a study of six military women undertaking an arduous Antarctic crossing, HRV demonstrated a latent increase in nonlinear, time frequency domains 2 wk after the expedition (27), suggesting that beneficial autonomic adaptation of exercise occurred independently to the marked energy deficit seen during the expedition (43).

The BEDA-Q has received increasing recognition (1,28). We found that BEDA-Q scores were associated inversely with EA, FM loss, and 1.5-mile improvement, and participants reporting ever trying to lose weight demonstrated slightly reduced EA. Our findings support this tool’s potential for ongoing use in stressful real-world settings.

Multiple stressors induced by the course provided a challenging context for the measurement of EA. Assessing EA is often harder in real world than experimental settings, owing to competing interests of data collection with other priorities (37). A key strength of our study was therefore to demonstrate the potential of a novel real-world measure of EEEmvpa, a relevant metric to the EA paradigm in many occupations. Other strengths include its comparatively long duration, use of a gold standard referent of TEE, and measurement of concurrent training adaptations. This tool would warrant further validation alongside concurrent DLW or indirect calorimetry for use in other contexts.

Our study has several limitations. Accelerometry tends to underestimate TEE and EEE in free-living environments and may yield mean negative bias of 8% TEE compared with gold standard techniques, with significant interindividual variation (44). Importantly for our population, currently available accelerometry platforms do not capture the energy cost of load carriage, which would be likely to increase estimates of EA. The wrist site is associated with less negative bias than the hip, although both demonstrate significant interindividual variation (44,45). On the other hand, our approach of calculating EAmvpa from movement above the nonpurposeful activity cutoff (0.09g) differed from that of Loucks and Thuma (7), who added background TEE (from accelerometry) to EEE (from indirect calorimetry) to calculate EEEmvpa. This may have led to modest relative underestimation of EAmvpa during the prolonged bouts of activity we observed. However, overall, EA was grossly underestimated likely because of the underreporting of EI. We sought to overcome this limitation by the use of weighed food analysis in a validatory cohort (46). Underestimation of EI is endemic in self-reports, applies to traditional diaries as much as mobile technology, and varies widely between individuals (32). We were unable to carry out the measurement of V˙O2peak, V˙O2max, or indirect calorimetry because of constraints imposed by the course timetable.

We conclude that the simple accelerometry-based measure of moderate and vigorous physical activity may be recommended for women undertaking complex, multistressor training. Because purposeful exercise activity is a more useful concept for trainers and athletes than total physical activity, EEEmvpa could be specified within EA definitions in future. Yet until the perpetual barrier of EI underreporting is overcome, it is difficult to rely heavily on field measures of EA. Instead, screening tools like BEDA-Q (as in this study) and biomarkers (47) demonstrate promise. A low EA-associated performance decrement is clearly an important concept for men and women in sports and physical occupations; our findings underline the importance of addressing low EA to optimize performance in military training.

This study was funded by the Ministry of Defence, UK. The authors acknowledge the support of the British Heart Foundation Centre for Research Excellence Award III (RE/18/5/34216). None of the funders had any role in study design or completion.

The authors acknowledge the support of the staff at the Royal Military Academy, Sandhurst, for generously accommodating the study and the officer cadets for enthusiastic participation. They are grateful to research assistants Ms. Sally Handford, Ms. Jennifer Wright, Mr. Shaun Chapman, Ms. Rachael Bradley, Dr. Sam Saunders, Dr. Susan Dewhurst, Ms. Mollie Drew, Mr. Fionn Sullivan, Ms. Jessica Dearman, Mr. Alfie Gordon, Ms. Louise Corfield, Ms. Amy Ranford, and Ms. Vicky Edwards. We acknowledge the support of the Wellcome Trust Clinical Research Facility led by Jo Singleton, Finny Patterson, and Steve McSwiggan. The results of the study are presented clearly, honestly, and without fabrication, falsification, or inappropriate data manipulation and do not constitute endorsement by the American College of Sports Medicine.

J. Langford works as a technical director of Activinsights Ltd., which manufactures the GENEActiv accelerometer. None of the other authors has any conflict of interest to declare.

REFERENCES

1. Mountjoy M, Sundgot-Borgen JK, Burke LM, et al. IOC consensus statement on relative energy deficiency in sport (RED-S): 2018 update. Br J Sports Med. 2018;52(11):687–97.
2. Capling L, Beck K, Gifford J, Slater G, Flood V, O’Connor H. Validity of dietary assessment in athletes: a systematic review. Nutrients. 2017;9(12):1313.
3. Burke LM, Lundy B, Fahrenholtz IL, Melin AK. Pitfalls of conducting and interpreting estimates of energy availability in free-living athletes. Int J Sport Nutr Exerc Metab. 2018;28(4):350–63.
4. Brown MA, Howatson G, Quin E, Redding E, Stevenson EJ. Energy intake and energy expenditure of pre-professional female contemporary dancers. PloS one. 2017;12(2):e0171998.
5. Silva AM, Matias CN, Santos DA, et al. Compensatory changes in energy balance regulation over one athletic season. Med Sci Sports Exerc. 2017;49(6):1229–35.
6. Levine JA. Non-exercise activity thermogenesis (NEAT). Best Pract Res Clin Endocrinol Metab. 2002;16(4):679–702.
7. Loucks AB, Thuma JR. Luteinizing hormone pulsatility is disrupted at a threshold of energy availability in regularly menstruating women. J Clin Endocrinol Metab. 2003;88(1):297–311.
8. VanHeest JL, Rodgers CD, Mahoney CE, De Souza MJ. Ovarian suppression impairs sport performance in junior elite female swimmers. Med Sci Sports Exerc. 2014;46(1):156–66.
9. Torstveit MK, Fahrenholtz I, Stenqvist TB, Sylta Ø, Melin A. Within-day energy deficiency and metabolic perturbation in male endurance athletes. Int J Sport Nutr Exerc Metab. 2018;28(4):419–27.
10. Silva M-RG, Silva H-H, Paiva T. Sleep duration, body composition, dietary profile and eating behaviours among children and adolescents: a comparison between Portuguese acrobatic gymnasts. Eur J Pediatr. 2018;177(6):815–25.
11. Lieberman JL, DE Souza MJ, Wagstaff DA, Williams NI. Menstrual disruption with exercise is not linked to an energy availability threshold. Med Sci Sports Exerc. 2018;50(3):551–61.
12. Siddall AG, Powell SD, Needham-Beck SC, et al. Validity of energy expenditure estimation methods during 10 days of military training. Scand J Med Sci Sports. 2019;29(9):1313–21.
13. Friedl KE. Biomedical research on health and performance of military women: accomplishments of the Defense Women’s Health Research Program (DWHRP). J Womens Health (Larchmt). 2005;14(9):764–802.
14. Gifford RM, Reynolds RM, Greeves J, Anderson RA, Woods DR. Reproductive dysfunction and associated pathology in women undergoing military training. J R Army Med Corps. 2017;163(5):301–10.
15. Bellenger CR, Fuller JT, Thomson RL, Davison K, Robertson EY, Buckley JD. Monitoring athletic training status through autonomic heart rate regulation: a systematic review and meta-analysis. Sports Med. 2016;46(10):1461–86.
16. Booth CK, Probert B, Forbes-Ewan C, Coad RA. Australian army recruits in training display symptoms of overtraining. Mil Med. 2006;171(11):1059–64.
17. Kim HG, Cheon EJ, Bai DS, Lee YH, Koo BH. Stress and heart rate variability: a meta-analysis and review of the literature. Psychiatry Investig. 2018;15(3):235–45.
18. Ministry of Defence. Joint Service Manual of Medical Fitness. London (UK): Joint Service Publication 950; 2018, Lft 6-7-7.
19. CRAN [Internet]. Available from: https://cran.r-project.org/. Accessed October 22, 2020.
20. Activinsights [Internet]. Available from: https://open.geneactiv.org/. Accessed October 22, 2020.
21. van Hees VT, Renstrom F, Wright A, et al. Estimation of daily energy expenditure in pregnant and non-pregnant women using a wrist-worn tri-axial accelerometer. PLoS One. 2011;6(7):e22922.
22. van Hees VT, Fang Z, Langford J, et al. Autocalibration of accelerometer data for free-living physical activity assessment using local gravity and temperature: an evaluation on four continents. J Appl Physiol (1985). 2014;117(7):738–44.
23. Esliger DW, Rowlands AV, Hurst TL, Catt M, Murray P, Eston RG. Validation of the GENEA accelerometer. Med Sci Sports Exerc. 2011;43(6):1085–93.
24. Jette M, Sidney K, Blumchen G. Metabolic equivalents (METS) in exercise testing, exercise prescription, and evaluation of functional capacity. Clin Cardiol. 1990;13(8):555–65.
25. Cunningham JJ. Body composition as a determinant of energy expenditure: a synthetic review and a proposed general prediction equation. Am J Clin Nutr. 1991;54(6):963–9.
26. Mayorga-Vega D, Bocanegra-Parrilla R, Ornelas M, Viciana J. Criterion-related validity of the distance- and time-based walk/run field tests for estimating cardiorespiratory fitness: a systematic review and meta-analysis. PLoS One. 2016;11(3):e0151671.
27. Gifford RM, Boos CJ, Reynolds RM, Woods DR. Recovery time and heart rate variability following extreme endurance exercise in healthy women. Physiol Rep. 2018;6(21):e13905.
28. Gordon CM, Ackerman KE, Berga SL, et al. Functional hypothalamic amenorrhea: an Endocrine Society Clinical Practice Guideline. J Clin Endocrinol Metab. 2017;102(5):1413–39.
29. Bland JM, Altman DG. Measuring agreement in method comparison studies. Stat Methods Med Res. 1999;8(2):135–60.
30. Black AE, Prentice AM, Coward WA. Use of food quotients to predict respiratory quotients for the doubly-labelled water method of measuring energy expenditure. Hum Nutr Clin Nutr. 1986;40(5):381–91.
31. Livingstone MB, Black AE. Markers of the validity of reported energy intake. J Nutr. 2003;133(3 Suppl):895s–920s.
32. Burrows TL, Ho YY, Rollo ME, Collins CE. Validity of dietary assessment methods when compared to the method of doubly labeled water: a systematic review in adults. Front Endocrinol (Lausanne). 2019;10:850.
33. Tharion WJ, Lieberman HR, Montain SJ, et al. Energy requirements of military personnel. Appetite. 2005;44(1):47–65.
34. Riou ME, Jomphe-Tremblay S, Lamothe G, Stacey D, Szczotka A, Doucet E. Predictors of energy compensation during exercise interventions: a systematic review. Nutrients. 2015;7(5):3677–704.
35. Gifford RM, O’Leary TJ, Double RL, et al. Positive adaptation of HPA axis function in women during 44 weeks of infantry-based military training. Psychoneuroendocrinology. 2019;110:104432.
36. Blundell JE, King NA. Physical activity and regulation of food intake: current evidence. Med Sci Sports Exerc. 1999;31(11 Suppl):S573–83.
37. Ackerman KE, Holtzman B, Cooper KM, et al. Low energy availability surrogates correlate with health and performance consequences of relative energy deficiency in sport. Br J Sports Med. 2019;53(10):628–33.
38. Whybrow S, Hughes DA, Ritz P, et al. The effect of an incremental increase in exercise on appetite, eating behaviour and energy balance in lean men and women feeding ad libitum. Br J Nutr. 2008;100(5):1109–15.
39. Schubert MM, Palumbo E, Seay RF, Spain KK, Clarke HE. Energy compensation after sprint- and high-intensity interval training. PLoS One. 2017;12(12):e0189590.
40. McNeil J, Brenner DR, Courneya KS, Friedenreich CM. Dose–response effects of aerobic exercise on energy compensation in postmenopausal women: combined results from two randomized controlled trials. Int J Obes (Lond). 2017;41(8):1196–202.
41. Shaffer F, Ginsberg JP. An overview of heart rate variability metrics and norms. Front Public Health. 2017;5:258.
42. Task Force. Heart rate variability. Standards of measurement, physiological interpretation, and clinical use. Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology. Eur Heart J. 1996;17(3):354–81.
43. Gifford RM, O’Leary T, Cobb R, et al. Female reproductive, adrenal, and metabolic changes during an Antarctic traverse. Med Sci Sports Exerc. 2019;51(3):556–67.
44. Murakami H, Kawakami R, Nakae S, et al. Accuracy of wearable devices for estimating total energy expenditure: comparison with metabolic chamber and doubly labeled water method. JAMA Intern Med. 2016;176(5):702–3.
45. Crouter SE, Flynn JI, Bassett DR Jr. Estimating physical activity in youth using a wrist accelerometer. Med Sci Sports Exerc. 2015;47(5):944–51.
46. Gemming L, Utter J, Ni Mhurchu C. Image-assisted dietary assessment: a systematic review of the evidence. J Acad Nutr Diet. 2015;115(1):64–77.
47. Elliott-Sale KJ, Tenforde AS, Parziale AL, Holtzman B, Ackerman KE. Endocrine effects of relative energy deficiency in sport. Int J Sport Nutr Exerc Metab. 2018;28(4):335–49.
Keywords:

RELATIVE ENERGY DEFICIENCY; EXERCISE ENERGY EXPENDITURE; WEARABLE TECHNOLOGY; PHYSICAL ADAPTATION; WOMEN; FEMALE ATHLETE TRIAD; HEART RATE VARIABILITY

Supplemental Digital Content

Copyright © 2020 by the American College of Sports Medicine