Participants
Eighteen female teenage runners (15−19 years old) from the same high school team were evaluated from 2015 to 2020. One runner had graduated at the time of measurement but continued to train under the instruction of the same team coach. All participants were Japanese, competitive middle- or long-distance (800−5000 m) runners, with a personal best record of 4 min 43 s ± 14 s and 9 min 58 s ± 31 s for 1500 m and 3000 m races, respectively. During the five-year study period, the team consistently ranked high, won the Prefectural Championship Road Running Relay race three times, and advanced to the relevant national championship. Runners from high schools representing the 47 prefectures of Japan participate in the national championship, and they are considered as the best high school runners. The team had been ranked average in the race.
All physical training was organized and supervised by the same coach during the study period, while the diet was left to the discretion of each runner. Some participants lived in the same dormitory; in which case their daily lifestyles were strictly controlled. Training sessions were usually performed twice a day: early in the morning and in the evening. Training comprised jogging, tempo running, and high-intensity interval training. Resistance and agility training was similarly performed almost daily.
Training details and diet were regularly recorded by the athletes in their diaries. The maximal running mileage was approximately 100 km per week, and the runners went on intensive training camps for up to three weeks, four to five times a year, where the training intensity, daily training hours, and mileage were substantially increased. No significant abnormalities were observed in the participants following mandatory medical check-ups performed in the high school.
Study protocol
The study was conducted without interrupting school events, classes, or scheduled training programs, and without modifying the diet of the participants. Particularly, the days requiring an overnight stay at the laboratory were carefully chosen, so as not to affect daily exercise training. Consequently, the day of RMR measurement was set for a previously planned rest day, free from exercise training, which was mostly Sunday, and the last exercise training was scheduled on the morning of the day before the RMR measurement. This was done to secure at least 17 h of abstinence from exercise training for the measurement of RMR. This met the American Dietetic Association’s recommendation on best practice methods to measure RMR [24].
The study was scheduled for a total of eight consecutive days (Fig. 1), and the participants were asked to spend their time as usual without changing exercise training and diet. The diets of the participants were recorded with training logs for seven days, and the participants visited our laboratory on the evening of the seventh day after refraining from training since noon. The time of dinner was unchanged and served as desired. After dinner, the participant spent the night in the whole room calorimeter of the laboratory, and RMR was measured early the next morning. The time of sleeping and waking was unchanged for each individual. Immediately after RMR measurement, body composition was evaluated by dual-energy X-ray absorptiometry (DXA), and a submaximal exercise test was performed on a treadmill using an indirect calorimeter.
EI
Runners were instructed to record every meal, including drinks and nutritional supplements, with photographs, and report the meals to Registered Dietitian Nutritionists (K.I.T. and Y.Y.) via daily e-mails, which included details regarding the type of food, ingredients, seasoning, and preparation. These details were obtained from and confirmed by the participants’ parents who served the meals. A ruler supplied by the nutritionists was incorporated into each photo, taken before and after each meal, to estimate the volume of food consumed. As the participants consumed steamed rice with almost every meal, the weight of the rice was measured each time using a digital scale (KD-192, TANITA, Tokyo, Japan).
On the seventh or eighth day of the study, the nutritionists conducted face-to-face interviews with each subject to confirm the accuracy of all records. Daily EI was calculated using the Standard Tables of Food Composition in Japan - 2015 (Seventh Revised Version) [25]. The mean EI for all seven days was calculated and used for analysis.
EEE
EEE was calculated for running and other types of training (resistance training and agility drills) separately, using individual prediction equations for the total oxygen consumption (VO2) of each runner. We used running speed to estimate EEE for the running training, whereas heart rate (HR) was used for resistance training and agility drills. A timekeeper checked and recorded lap times (each 200 m for track training and, at most, 1 km for road running) with a stopwatch for each subject. Similarly, the participants recorded their own lap times with a sports wristwatch and attempted to run at the speed set by the coach. Each lap time was recorded in the training logs, and the running speed was calculated by the lap time for each distance. HR was recorded every 5 s by a wearable HR monitor (Polar Loop, Polar Electro Oy, Kempele, Finland) and uploaded to the POLAR FLOW website [26]. The HR data was exported in .csv format and was analyzed.
The runners were assessed on a treadmill at 1 % slope with breath-by-breath indirect calorimetry (Quark CPET, COSMED, Rome, Italy) to yield submaximal steady-state VO2 (ml⋅kg−1⋅min−1) for three or four sets at 10, 12, 14, and 15 km⋅hr−1. The duration of each stage was set at 5 min and separated with more than 5 min rest. The submaximal steady-state VO2 was plotted according to the velocity and HR, and equations to estimate VO2 were determined using linear regression analysis. The mean R2 of the regression lines was 0.986 (0.920−1.000) and 0.992 (0.967−1.000), respectively.
The participants reported that they had occasionally engaged in sports activities, such as volleyball or swimming, apart from those in their regular training program. Although the amount of energy expenditure in these activities was relatively minor compared to regular training, we included them in calculating total EEE, using a youth compendium of physical activities [27]. Finally, total EEE (kcal) was calculated using the sum of training and other activities with a caloric equivalent of 5 kcal⋅L−1 of O2 and subtracting RMR (kcal) of the equivalent duration of exercise. The mean EEE for all seven days was analyzed.
EA
Daily EA was calculated by subtracting EEE from EI and adjusting for fat free mass (FFM). The means of the seven days were calculated and used for analysis. As the EI on day 1 for one subject, and on days 1 and 2 for another subject, were not available, we calculated these means for six and five days, respectively, for EA as well as EI and EEE.
EA below the threshold value of 30 kcal⋅kg−1 FFM⋅d−1 was defined as low EA according to the American College of Sports Medicine [1]. Similarly, we used a cut-off value of 45 kcal⋅kg−1 FFM⋅d−1 to identify optimal EA [1]. Using t-tests, we compared the means of the values above and below the threshold of 30 (non-low EA vs. low EA, respectively) as well as above and below the cut-off value of 45 (optimal EA vs. non-optimal EA, respectively) for each continuous variable.
RMR
RMR was measured using a whole room calorimeter (Fuji Medical Science, Chiba, Japan) early in the morning after overnight sleep. The room measured 3.7 (W) × 2.7 (D) × 2.5 (H) m with an internal volume of 25.2 m3. The room had a bed with a thick and comfortable mattress (Sealy Corporation, Trinity, NC, USA), a neck pillow made of a pressure-relieving material (Tempur-Pedic International, Inc., Lexington, KY, USA), a TV set, washbasin, and a toilet. The airflow was kept at 50 L⋅min−1 with temperature and relative humidity set to 25.0° C and 50.0 %, respectively. The air was analyzed by mass spectrometry (Thermo Scientific TM Prima PRO Process Mass Spectrometer, Thermo Fisher Scientific, Cheshire, UK). Calibration was performed periodically. VO2 and carbon dioxide production were calculated according to Henning et al. [28], and energy expenditure (kcal) was estimated by Weir’s equation [29]. The accuracy was verified by an alcohol combustion test before each experiment. Test-retest reliability for energy consumption at rest was calculated for five volunteers, and the coefficients of variation (CV) were 4.4 ± 1.3 %.
The participants were familiarized with the procedure by visiting our laboratory and observing the facility prior to the study to reduce emotional stress. They entered the room after finishing dinner (18:00 − 20:00) the day before RMR measurement (Fig. 1). The participants were instructed to stay quietly in the room without any exercise until they went to bed as per their routine (21:00 − 23:00) and stay motionless when they woke up the next morning.
The time of waking was determined beforehand (05:00 − 06:00) as per their routine, whereupon calm, classical music (Mozart: Klarinettenkonzert KV 622; Adagio) was played through a wireless communication device (TY-WSA10, Toshiba Corporation, Tokyo, Japan) for 3 min, and the room was dimly lit using indirect lighting. After confirming that the subject was fully awake using the communication device, they were asked to keep calm and stay motionless. If necessary, they could turn over and stretch gently, lying on the bed for a few minutes, or use the toilet, before RMR measurement.
For RMR measurement, the participants were instructed not to move on the bed and remain supine for 40 to 60 min (50 ± 12 min). The time was announced every 5 min in a gentle and quiet voice during the measurement using the wireless device. We ensured that the participants remembered every announcement after finishing the RMR measurement and that they remained awake during measurement.
The steady-state of calculated energy expenditure was visually assessed by a PC monitor after obtaining approximately 20 min of data, a duration which was adopted to determine RMR (Fig. 2). The duration for the calculation was 21 ± 4 min, and CV of energy expenditure for the duration were 3.7 ± 1.9 %. RMR was measured between 05:00 − 07:00, after 7.7 ± 0.7 h of sleep, 17−19 h since the last training session, and after abstaining from food for 10.7 ± 0.8 h. A wrist-worn accelerometer (wGT3X-BT, Actigraph, Pensacola, FL, USA) was mounted on the wrist of the participants, and the data was analyzed using ActiLife 6.11.6 software (ActiGraph, Pensacola, FL, USA), retrospectively, to confirm that the participants had not moved during RMR measurement. The ratio of measured RMR to predicted RMR (RMR ratio) was calculated. Considering the influence of race, age, and sex on RMR, we used the following formulae to determine a predicted RMR (kcal⋅d−1) for Japanese females (J1): < 18 years old (n = 15), 7.64 × body mass (kg) + 4.22 × height (cm) – 22.5 × age (years) + 526 [30]; ≥ 18 years old (n = 3), (0.0481 × body mass (kg) + 0.0234 × height (cm) – 0.0138 × age (years) – 0.9708) × 1000/4.186 [31]. Since accounting for FFM is supposedly appropriate for RMR prediction in athletes [9], we used Cunningham’s equation (C): 500 + 22 × FFM (kg) [32], and the equation developed for adult female Japanese athletes (J2): 36 + 26.9 × FFM (kg) [33] to predict RMR.
Anthropometry and body composition
Height and body mass were measured using an automatic digital scale (AD-6228, A&D, Tokyo, Japan). Areal bone mineral density (BMD) of total body less head (BMDTBLH) (g⋅cm−2), with Z-score and body composition, was measured by DXA (Prodigy, GE healthcare, Madison, WI, USA), early in the morning directly after RMR measurement. A quality assurance test was performed before each scan using a phantom provided by the manufacturer. We assessed test-retest reliability using 10 girl runners from the same team as the participants. The CV for BMD, fat mass, and lean soft tissue mass were 0.4 %, 1.8 %, and 0.3 %, respectively. The same investigator (N.K.) conducted all scans, and enCORE software versions 13 and 16 (GE Healthcare, Madison, WI, USA) were used for data processing. Body mass index (BMI) was calculated as body mass (kg) divided by height squared (m2). FFM was calculated by adding bone mineral content to lean soft tissue mass.
Menstrual history
Menstrual history was investigated using a questionnaire alongside a personal interview. If the frequency of menstruation in the last 12 months was ≤ 3, or the three most recent consecutive menstrual cycles were absent, it was classified as amenorrhea [34]. If the menstruation frequency in the last 12 months was three times, but the menstruation mostly occurred over the recent 3 months, we regarded it as resumed menstruation and did not classify it as amenorrhea.
Statistical analysis
All statistical analyses were conducted under the guidance of a statistical expert (K.O.) with SPSS® version 25.0 J for Windows (IBM Japan, Ltd., Tokyo, Japan). T-tests were performed to define the differences between two EA groups with Levene’s test for equality of variances. The Fisher’s exact test was used to assess the difference in frequency of amenorrhea between the two groups. Pearson’s correlation analysis was used to evaluate the association between two continuous variates. We primarily tested the hypothesis that RMR reduces linearly with a decrease in EA. Further, the relationship was evaluated for each group of the EA level. A P value < 0.05 was considered statistically significant. Given the exploratory nature and the small number of study participants, effect sizes were reported as Hedges’ g with a correction for small sample bias for t-tests [35] and ϕ for comparing the frequency of amenorrhea. Effect sizes were calculated using R version 4.0.3, and values were presented as mean ± standard deviation.