Journal Logo

BASIC SCIENCES

A New Approach to Improve the Validity of Doubly Labeled Water to Assess CO2 Production during High-Energy Turnover

VAN HOOREN, BAS1; MOST, JASPER1,2; COLLOMBON, ELINE1; NIEMINEN, HENRIETTA1; PLASQUI, GUY1

Author Information
Medicine & Science in Sports & Exercise: June 2022 - Volume 54 - Issue 6 - p 965-973
doi: 10.1249/MSS.0000000000002865

Abstract

Doubly labeled water (DLW) is considered the gold-standard method for determining energy expenditure over longer time periods in free-living conditions (1–4). The accuracy of energy expenditure estimates by DLW have been shown to be ±1%–6% when compared with whole-room indirect calorimetry when measured over time periods of >3 d with low to moderate physical activity levels (PAL) (1–3,5–10) and high PAL (8). Importantly, more recent advancements in isotope equations yield accuracies closer to 1% (10).

The accuracy of the estimated energy expenditure with DLW depends on several methodological decisions. Most importantly, total body water (TBW) can be calculated by the plateau-method or the slope-intercept method (11–13) (schematic in Fig. 1). Because both methods have their limitations, there is no consensus on which method to use in which situation. Indeed, the plateau method neglects the loss of isotopes during the equilibration period until the first postdosing measurement, potentially resulting in an overestimation of TBW and consequently CO2 production and energy expenditure. In contrast, the intercept method potentially introduces an error because the decline in isotopes during the equilibration period may be smaller when routinely measured overnight than the decline over the entire observation period. This could therefore result in an underestimation of TBW and therefore underestimation of CO2 production and energy expenditure. This underestimation would be most pronounced in situations of greater slopes, that is, high physical activity and hence high-energy turnover (11).

F1
FIGURE 1:
Left: Schematic representation of the 1) slope-intercept (blue), 2) plateau (red), and 3) overnight-slope (yellow) method during periods of high physical activity. Shaded areas indicate night, whereas white areas indicate day. Right: Natural log of the O-18 concentration during a 5-d period of one representative subject. Note: The slope-intercept method is also often referred to as the two-point or multipoint method, whereby two (usually first and last) or multiple time points are used to determine the slope, respectively.

To overcome the limitations of both methods, in this study, we offer a novel, alternative approach to calculate TBW, which we will refer to as the “overnight-slope” method. Instead of the slope over the entire observation interval, we will only use the night slopes to calculate the dilution space (Fig. 1). It is expected that this method will provide a more accurate indication of TBW, in particular in situations of high-energy turnover because the overnight slope likely better reflects the slope during the equilibration period when the dose is provided in the evening with equilibration overnight (Fig. 1). The primary aim of this study is therefore to examine whether the overnight-slope technique produces a more accurate estimation of energy expenditure in subjects with a high-energy turnover as compared with the plateau or slope-intercept technique. As an objective reference, energy expenditure will be measured simultaneously by whole-room indirect calorimetry. In addition, because the accuracy of energy expenditure estimated using DLW is substantially reduced over shorter time periods (2,6), our second aim was to investigate the accuracy of DLW-derived energy expenditure on a day-to-day basis.

METHODS

Study design

This is a one-arm intervention study, in which participants were instructed to maintain high PAL (>2.5). EE was measured simultaneously by whole-room calorimetry and DLW for a period of 5 d and nights.

Institutional approval and ethics

The study was approved by the medical ethics committee of Maastricht University (METC19-004) and was conducted according to the declaration of Helsinki. All participants received written and oral information about the study and signed informed consent before the measurements. The study was registered prospectively in the public trial registry www.ccmo.nl (NL68773.068.19).

Participants

A total of six healthy participants (three males and three females; mean ± SD age, 23.3 ± 2.3 yr; body mass, 69.5 ± 9.9 kg; height, 1.72 ± 0.07 m; and body mass index, 23.5 ± 2.1 kg·m−2) were recruited via advertisements at Maastricht University Medical Center+, Maastricht, the Netherlands. Inclusion and exclusion criteria were checked using a medical questionnaire. Only individuals participating in endurance exercise at least three times per week were included. Participants who suffered from claustrophobia or had any known medical condition affecting their capacity to perform long-term endurance exercise were excluded from the study.

Experimental design

All participants resided in the controlled environment of a respiration chamber (2.5 × 3 × 2.4 m; 18 m3) for 5 consecutive days (Fig. 2). Subjects entered the chamber in the evening of day 0 at 20:45 h, and the data from 2100 h onward were used for further analyses. To reach a daily PAL of >2.5, the participants cycled approximately 4 h·d−1 separated in a 2-h morning and 2-h afternoon session with a cycling intensity of approximately 70%–80% of the individuals’ maximal heart rate (as estimated using 220 − age) inside the respiration chamber. PAL was reviewed daily, to ensure a PAL of >2.5 was achieved. Breakfast (0800 h), lunch (1300 h), dinner (1800 h), snacks (1145, 1500, 1945h), and sports drink were provided to ensure energy balance. Time-stamped urine samples were collected every morning and evening. Participants were allowed to leave the chamber in the evening of days 1–4 at 2030 h for 20–25 min to take a shower in the facilities approximately 15 m from the chamber. The calorimetry data during this period were imputed using the average values of the 30 min before showering. Participants left the chamber on day 5 at 2100 h. A schematic representation of the study protocol is depicted in Figure 2.

F2
FIGURE 2:
Schematic of the experimental design.

Respiration chamber

The respiration chamber is a 18-m3 room equipped with a bed, desk, sink, freeze toilet, computer, and TV and specifically for this study with the subject’s own bicycle and a Tacx neo 2 system (Tacx, Wassenaar, the Netherlands). The respiration chamber was ventilated with fresh air at a rate of ~120 L·min−1. O2 and CO2 concentrations were measured continuously, and data were calculated over 5-s intervals using an Omnical total-capture indirect calorimeter (Omnical, Maastricht University, Maastricht, the Netherlands). The gas analyzers of the calorimeter were calibrated automatically every 15–30 min as described previously (14). Nitrogen gas (≥99.999%; Linde, Schiedam, the Netherlands) was used to set the zero, and a calibration gas with 18% O2 and 0.8% CO2 certified to 1% volumetric content (i.e., 0.18% O2, 0.008% CO2) was used for calibration (HiQ specialty gas; Linde). The four respiration chambers were validated using in situ methanol combustion and showed a deviation ranging from −0.3% to −3.8% in CO2 production and from 0.1% to 2.1% in O2 consumption at a typical rate of 139–190 mL·min−1 for CO2 production and 211–296 mL·min−1 for O2 consumption.

Energy expenditure was calculated from O2 consumption and CO2 production using the Weir equation (15). Sleeping metabolic rate was defined as the energy expenditure from 2:00 to 5:00 am, which typically corresponds to the lowest energy expenditure (16). PAL was calculated as the total energy expenditure divided by sleeping metabolic rate. Individual energy requirements for day 1 were estimated based on the sleeping metabolic rate from the first night multiplied by a PAL of 2.5, and this was daily reevaluated to ensure adequate food provision to maintain energy balance. The diet was composed of 55% of energy from carbohydrates, 15% from protein, and 30% from fat.

Doubly labeled water

During an initial screening visit, participants’ body mass (fasted, in underwear) and height (no shoes, socks) were determined to prepare the dosage of DLW according to “The Maastricht Protocol” (17). The given dose was calculated based on the subjects’ TBW, which was estimated based on body mass index, age, and gender using the prediction equation of Deurenberg et al. (18). Subjects received a dose of 2.5 g·L−1 TBW of a water mixture containing 9.8% enriched H218O and 6.5% enriched 2H2O, resulting in an initial excess body water enrichment of ~135 ppm for deuterium and ~240 ppm for oxygen-18. Before administration of the labeled water in the evening of day 0, baseline urine samples were taken. The DLW dosage was then consumed by the participants from a small bottle of water (80–160 mL). To ensure that all isotopes were administered, after consumption the bottle was filled with 50–75 mL tap water, which was consumed as well. Urine was collected from the second voiding every morning and every evening between 0700–0800 and 2100–2200 h, respectively. Collection times were recorded by the researchers. Immediately after collection, urine was pipetted into airtight cryotubes and stored at −20°C until further analysis.

Isotope enrichment was measured in urine using isotope ratio mass spectrometry (Thermo Scientific Delta V Advantage; Thermo Fisher Scientific, BREDA, the Netherlands) as previously described (17). Dilution spaces of 2H and 18O were calculated using the plateau, intercept, and “overnight-slope” methods (Fig. 1) and an updated standardized equation (10) after correction for isotopic exchange with other body pools (1% for deuterium and hydrogen dilution space/oxygen dilution space (Nd/No) ratio assumed to be 1.036) (19). Deuterium (kD) and oxygen (kO) turnover rates were calculated by linear regression of the logarithm of isotope enrichment as a function of time. kD and kO were calculated as multipoint slope over all 10 time points. TBW was calculated as the average of both TBW estimates (ND and NO). For the plateau-method, the second morning-urine sample was used, which was ~9–10 h after dosing. For the overnight-slope method, we extrapolated each overnight slope onto the following day, and recalculated the intercept of the “overnight slope.” CO2 production was calculated per day per participant using an updated equation of CO2 production (10). The respiratory exchange ratio used to determine energy expenditure was based on the indirect calorimetry data and is the mean respiratory exchange ratio over all 5 d for all participants.

Statistical analysis

For the primary aim, the differences (as compared with indirect calorimetry, in liters per day, megajoules per day and percent) of three different methods to measure CO2 and energy expenditure (plateau, slope-intercept, overnight-slope) were compared using Bland–Altman analysis (20). In addition, a repeated-measured ANOVA with Bonferroni post hoc correction was used to compare all methods. For the secondary aim, we determined the difference between room calorimetry and DLW CO2 for each day per individual using the overnight-slope method. We then computed the mean ± SD difference per day over all subjects and averaged this over all 5 d. All statistical analyses were performed using SPSS (version 25; IBM, Armonk, NY). P < 0.05 was considered statistically significant. Energy expenditure was expressed in megajoules per day.

RESULTS

Energy balance, PAL, and respiratory exchange ratio

Mean ± SD energy expenditure as determined from indirect calorimetry data was 19.7 ± 3.16 MJ·d−1, whereas energy intake over the 5-d period was 18.4 ± 2.32 MJ·d−1, indicating that the participants were in a state of slight negative energy balance during the study period (Appendix, Supplemental Digital Content, https://links.lww.com/MSS/C496). The average sleeping metabolic rate over the 5-d period was 7.06 ± 1.37 MJ·d−1, whereas the average PAL determined based on indirect calorimetry data was 2.8 ± 0.1. The exercise intensity expressed as percentage of maximum estimated heart rate during the exercise-only periods was on average 72%. On average, this corresponded with an exercise MET value of 9.1, and in total, 10.3 MJ spent on exercise over the 2 × 2 h of exercise sessions. The mean ± SD respiratory exchange ratio determined from the indirect calorimetry over the 5-d period was 0.85 ± 0.04, whereas the average respiratory exchange ratio during the cycling periods was 0.89 ± 0.03.

CO2 production and energy expenditure over the 5-d period

TBW values were 41.9 ± 6.1, 38.4 ± 5.7, and 40.4 ± 5.8 L for the plateau, slope-intercept, and overnight-slope methods, respectively. Nd/No ratios are reported in Table 1. CO2 production over the 5-d period for both indirect calorimetry and the three TBW-calculation methods as well as the difference between them are reported in Table 1. Briefly, the overnight-slope method resulted in the smallest mean difference when compared with indirect calorimetry (Fig. 3) and also had the smallest variability of this difference. In contrast, the plateau method significantly overestimated CO2 production, whereas the slope-intercept method underestimated CO2 production, although this difference was not statistically significant. Comparable findings were observed for energy expenditure when compared with indirect calorimetry.

TABLE 1 - Mean ± SD CO2 and energy expenditure over 5 d for indirect calorimetry and the three DLW methods.
Indirect Calorimetry DLW Plateau Method DLW Slope-Intercept Method DLW Overnight-Slope Method
TBW (L) 41.9 38.4 40.4
Dilution ratio (Nd/No) 1.027 ± 0.004 a,b 1.046 ± 0.001 a,c 1.036 ± 0.006 b,c
CO2 (L·d−1) 835 ± 123 d 873 ± 122 d 806 ± 114 a 841 ± 114 b,c
Difference with IC CO2 (L·d−1) 37.9 ± 18.2 a,b −29.6 ± 19.9 a,c 6.15 ± 11.5 b,c
Percentage difference with IC CO2 (%) 4.67 ± 2.6 b,c −3.43 ± 2.27 c 0.92 ± 1.60 b,c
95% LoA with IC CO2 (L·d−1) 2.18 to 73.5 −68.6 to 9.44 −16.3 to 28.6
Energy expenditure (MJ·d−1) 19.7 ± 3.14 20.8 ± 2.90 a,b 19.2 ± 2.72 a,c 20.0 ± 2.70 b,c
Difference with IC (MJ·d−1) 1.02 ± 0.74 a,b −0.62 ± 0.79 a,c 0.27 ± 0.65 b,c
Percentage difference IC (%) 5.45 ± 3.62 a,b −2.68 ± 3.23 a,c 1.70 ± 3.18 b,c
95% LoA with IC CO2 (MJ·d−1) −0.38 to 2.42 −2.14 to 0.97 −1.04 to 1.58
All P values are corrected for multiple comparisons using the Bonferroni procedure (i.e., P value is multiplied by the number of comparisons).
aSignificant difference between the plateau and slope-intercept method.
bSignificant difference between the plateau and overnight-slope method.
cSignificant difference between the slope-intercept and overnight-slope method.
dSignificant difference between indirect calorimetry and the plateau method.
IC = indirect calorimetry; LoA = limits of agreement.

F3
FIGURE 3:
Mean percentage difference in CO2 over 5 d of each DLW method compared with indirect calorimetry. Data points represent individual subjects, and the horizontal line depicts the mean difference. P values are adjusted using the Bonferroni correction.

Using the overnight-slope method, there was no proportional bias, that is, the error between overnight-slope and indirect calorimetry was independent of the absolute CO2 production (Bland–Altman plot Fig. 4).

F4
FIGURE 4:
Bland–Altman plot comparing the CO2 production as measured using the overnight-slope method and indirect calorimetry measured over the 5-d interval for each participant.

CO2 production and energy expenditure on a day-to-day basis

The overnight-slope method resulted in the most accurate estimation of CO2 production over the 5-d period and was therefore used to assess the accuracy of CO2 production from day-to-day. There is substantial within-subject variability in the day-to-day CO2 production, as assessed by indirect calorimetry, and in the accuracy of CO2 production assessment by DLW (Fig. 5, Table 2). On average, daily deviation in CO2 production measurement by DLW from IC was 66 L, which corresponds to ~9% of total CO2 production.

F5
FIGURE 5:
Percentage difference in day-to-day CO2 production between DLW and indirect calorimetry per participant. Positive values indicate an overestimation of DLW compared with indirect calorimetry. Each circle depicts 1 d, and the horizontal line represents the mean difference per participant.
TABLE 2 - Mean ± SD CO2 for each participant and the average difference over all day-to-day comparisons for indirect calorimetry and DLW.
Indirect Calorimetry DLW Mean Difference ± SD (%); P Value
Subject 1 CO2 (L·d−1) 691 ± 188 760 ± 263 16.0 ± 45.8; 0.657
Subject 2 CO2 (L·d−1) 919 ± 88 1033 ± 114 13.3 ± 17.7; 0.170
Subject 3 CO2 (L·d−1) 848 ± 35 890 ± 81 5.1 ± 10.0; 0.319
Subject 4 CO2 (L·d−1) 755 ± 34 794 ± 47 5.4 ± 7.7; 0.187
Subject 5 CO2 (L·d−1) 930 ± 82 1019 ± 160 9.4 ± 11.4; 0.149
Subject 6 CO2 (L·d−1) 664 ± 31 711 ± 31 7.1 ± 2.4; 0.003
Mean ± SD CO2 over all days (L·d−1) 801 ± 115 868 ± 136 66.6 ± 134; 0.044
Mean ± SD percentage difference with IC CO2 (%) 9.4 ± 4.5
Mean ± SD energy expenditure over all days (MJ·d−1) 1.23 ± 2.48
Values in bold reflect significant differences (P < .05) between indirect calorimetry and doubly labeled water.

DISCUSSION

The primary aim of this study was to compare CO2 production estimated with the overnight-slope method to the classical plateau and slope-intercept methods during a 5-d period of high-energy turnover. We show that the overnight-slope method results in the most accurate estimation of CO2 production, objectively assessed by gold-standard whole-room calorimetry, with a deviation of only ~0.9% over a 5-d period. In contrast, the plateau and slope-intercept methods resulted in an overestimation and underestimation of ~4.7% and ~3.4%, respectively. When CO2 production was assessed per day using the overnight-slope method, the average difference was 9.4% ± 4.5% relative to indirect calorimetry.

The more accurate determination of CO2 production using the overnight-slope method when compared with the plateau and slope-intercept method can be explained by their differences in estimating the slope of isotope loss before the first postdose sample. Importantly, in many protocols, including ours, this unknown slope is an overnight slope and therefore determined by resting/sleeping metabolic rate. The plateau-method neglects the loss of isotopes in the period until the first postdosing measurement, hereby underestimating isotope concentration at t0, and overestimating TBW, CO2 production, and energy expenditure (Figs. 1, 3), as we have demonstrated in this study. In contrast, the intercept method extrapolates the slope of the entire observation period, which includes resting metabolic rate and physical activity, onto the first overnight period, hereby overestimating t0 isotope enrichment, and hence underestimating TBW, CO2 production and energy expenditure. This underestimation would be most pronounced in situations of high physical activity because a higher overall slope is extrapolated onto the unknown, first overnight slope (8,11). In the overnight-slope method only, the night slope is used to represent the decline in isotopes during the isotope equilibration period, resulting in a more accurate determination of the dilution space. Importantly, the overnight-slope method also resulted in the smallest variability in the difference between indirect calorimetry and DLW (Table 1, Fig. 3), suggesting that this method may also provide a more accurate estimation of CO2 production per subject.

This observed accuracy of the new method is an important improvement for the utility of DLW during high-energy turnover, because classical calculation, that is, the plateau and slope-intercept methods, provided significantly poorer estimates of CO2 production. Using this new calculation method, DLW achieves an accuracy in line with what previously has been reported over time periods of >3 d with low to moderate PAL (1–3,5–10) and higher PAL (8). Interestingly, the accuracy of both the plateau and slope-intercept methods in our study was somewhat lower than reported in previous studies with an error of 4.7% and −3.4%, respectively. For example, Klein et al. (2) compared CO2 production estimated with the slope-intercept method to indirect calorimetry in one subject over a 5-d period during which the subject performed four 30-min cycling periods per day. Over this 5-d period, the difference in estimated and observed CO2 production was approximately 1%. Similarly, another study compared total CO2 production estimated with different variations of the slope-intercept method to indirect calorimetry in nine subjects over a 7-d period, with the subjects performing two 30-min cycling periods per day (5). The mean difference in CO2 production between the two methods over the 7-d period ranged from ~1.15% to ~1.76% depending on the method used to determine isotope kinetics. Although both studies did not report PAL, it is likely that PAL in these studies was lower than in our study because of the shorter total exercise time (2 and 1 vs 4 h·d−1), which could potentially explain the differences in the accuracy for the slope-intercept method in these studies versus our study as discussed previously. Another study showed an average difference in CO2 production of only 1.0% compared with indirect calorimetry in four subjects with a high PAL (2.6), but the individual difference between both methods in that study ranged between −12% and +12% (8), which is much larger than in the current study. One could argue that in a daytime protocol, with dosing in the morning and equilibration during the day where subjects are active and eating and drinking, the loss of isotope during equilibration might better resemble the loss of isotope over the entire observation interval, and hence, in that case, the slope-intercept method would be more appropriate. However, in studies with athletes, this would be more burdensome and also the isotope loss during equilibration would by definition be larger, and the question remains whether this would improve the calculation using the traditional slope-intercept method. The key message here is that when performing studies in athletes with high-energy turnover, it is important to consider the impact of the chosen protocol and calculation on the results.

In practice, DLW is often used to assess energy expenditure, in order to inform on optimal energy intake in various situations (11,21–27). Energy expenditure is calculated from CO2 production using an (estimated) respiratory quotient or measured food quotient (4,10). In line with the findings for CO2 production, the overnight-slope method showed the smallest difference in estimated energy expenditure with the respiratory quotient measured using indirect calorimetry, followed by the slope-intercept and the plateau method (1.70%, −2.68%, and 5.45%, respectively; Table 1). Overall, these findings therefore also support that the overnight-slope method results in the most accurate estimation of energy expenditure, with the difference over a 5-d period corresponding to only ~0.3 MJ·d−1 (64.5 kcal·d−1), as compared with −0.6 to 1.0 MJ·d−1 (−140 to 240 kcal·d−1) for the slope-intercept and plateau method, respectively. These findings can also have important implications for the limits of human energy expenditure. Specifically, human energy expenditure under extreme conditions (e.g., multiday cycling or running competitions, Artic trekking) has often been determined using the slope-intercept method (28,29). The findings of the current study suggest that energy cost and hence the limits of human energy expenditure under these extreme conditions may be underestimated by ~3%. This would mean that the energy limit that can be maintained for prolonged periods (e.g., >250 d) of 2.5 times the basal metabolic rate as suggested by Thurber et al. (29) could be more around 2.6 times the basal metabolic rate. Similarly, for shorter periods of high physical activity such as 11-h triathlons or 25-h ultramarathons, the energy limits could be 9.7 and 8.8 times the basal metabolic rate instead of 9.4 and 8.5 (30), respectively. The data of the current study showed that 5 d of high physical activity lead to a (minor) negative energy balance with a PAL of 2.8. The reason may well be that, even though participants were reasonably trained, cycling for 4 h·d−1 for 5 d in a row is an amount of exercise they were not used to perform in daily life. It is conceivable that subjects need more time to adjust energy intake to such high energy expenditure.

The second aim of our study was to determine the accuracy of DLW-derived CO2-production and energy expenditure for a single day. Over 5 d, the measurement error of the overnight method was <1%, and the mean error for a single 24-h measurement was 9.4%. The utility of such measurements and its inherent error depends on the research purposes. Although the accuracy is higher compared with other methods for assessing energy requirements such as diet recalls or accelerometry, it also carries higher costs of analysis. Indeed, when expressed as a difference in megajoules per day (Table 2), the average difference is 1.23 MJ (294 kcal), and this is lower than/in line with the error with self-reporting of energy intake using questionnaires (31,32) or slightly higher/in line with using standard prediction equations in athletic populations (33,34). For example, the difference in energy intake as determined using a questionnaire and energy expenditure using DLW has been reported to be 1.9 MJ·d−1 for males (31). Similarly, ten Haaf and Weijs (34) showed several prediction equations (e.g., Harris–Benedicts, Schofield, Mifflin) had an average error of >1 or even >1.5 MJ·d−1 when applied to recreational athletes. In addition, the difference is substantially lower than previously reported for day-to-day comparisons. For example, a study on two subjects found differences in CO2 production of up to 9% for the subject with a low physical activity and up to 27% for the subject with high physical activity when measured on a day-to-day basis (6). Another study among one subject also reported large differences during the initial 2 d and therefore proposed that energy expenditure must be determined over a minimum period of 2 to 3 d when using DLW, as shorter time periods may not result in sufficient differences between the dilution of the two isotopes for accurate results (2). Coward and Prentice (35) even suggested a minimum of 5 d should be used when determining energy expenditure with DLW, although they only reported comparisons of the DLW estimated energy and indirect calorimetry for 5 and 12 d and not for shorter periods. In contrast to these suggestions, we show that day-to-day comparisons during periods of high-energy turnover can be reasonably accurate, with a difference that is only slightly larger than the difference reported between different methods to determine isotope kinetics over a 5-d period (i.e., 9% for day-to-day vs 4.7% with different DLW methods over 5 d, respectively). It is important to emphasize that the difference is likely larger during periods of lower energy expenditure because the turnover rate is lower, hereby reducing the accuracy to detect the isotope decline over such a short time period. In addition, on an individual level, the accuracy can also vary substantially (Fig. 5).

Limitations

There are several important considerations when interpreting the findings of this study. First, the sample size of this study is relatively small with only six subjects. However, both room calorimetry and DLW are expensive techniques, and the burden for subjects was very high, which limited our study to a small sample size. Nevertheless, even with this small sample size, we could observe clear differences in the validity of different methods to determine isotope kinetics over the 5-d period. The cost of DLW can be reduced if less samples are required. Therefore, we also explored the possibility of estimating the overnight CO2 production by means of standard equations (i.e., Oxford equations (36)) instead of measuring it (Appendix, Supplemental Digital Content, https://links.lww.com/MSS/C496). The obtained estimate however substantially underestimated CO2 production by −17%, making it unsuitable as substitute for sampling. The underestimate in high-energy-turnover conditions may be due to an increased resting energy expenditure following the intense cycling protocol.

Second, the Weir equation used here to determine energy expenditure is based on resting conditions and may underestimate energy expenditure during exercise (37,38). However, we do not expect this to have a substantial effect on our findings because the majority of time was spent not exercising. Indeed, for one subject, we determined the energy expenditure during exercise using Jeukendrup’s equation for moderate- to high-intensity exercise (37), while using Weir’s equation for the nonexercise periods, and the difference in energy expenditure was only 0.08 MJ·d−1, or ~0.35%. Third, although urine sampling is generally the most convenient method, differences may occur in comparison to saliva sampling because of urine first being collected in the bladder and therefore presenting the isotope concentration in body water over several hours, whereas saliva sampling may better represent the exact time point of the sampling and the isotope concentration at that time. However, measurement of the oxygen isotope in saliva is technically more prone to error. Fourth, the first dose of DLW was consumed in the evening with overnight equilibration, and different results may be observed when the first dose is provided at different time points and different equilibration times, for example, with a day protocol. However, we deliberately chose for an overnight protocol because participants are at rest and refrain from eating and drinking during the equilibration period, which can cause more variability in the estimation of the dilution space. Also, with shorter equilibration time, there is a possibility of incomplete equilibration in subjects with larger water compartments (39). However, we acknowledge that the time point of sampling after an overnight fast could possibly contribute to the underestimate of maximal isotope concentration using the plateau method.

Fifth, the energy expenditure was slightly higher (7%) than energy intake during the 5-d period, indicating that subjects were in a negative energy balance. However, DLW has been shown to be accurate during periods of negative energy balance (40), suggesting that this has minimal impact on our findings. Furthermore, we estimated the PAL as the total energy expenditure divided by sleeping metabolic rate, whereas this is typically compared with resting metabolic rate measured while individuals are awake. The sleeping metabolic rate is slightly lower than resting metabolic rate (e.g., (41)), which therefore results in a small overestimation of the PAL. However, when we estimated the resting metabolic rate by multiplying the sleeping metabolic rate with a previously determined factor (SMR/RMR = 0.9743; Ref. [41]), the difference in PAL was only 0.07, corresponding to a negligible 2.6% overestimation of the PAL value (Appendix, Supplemental Digital Content, https://links.lww.com/MSS/C496). Finally, we did not assess body composition with an additional independent technique to validate the TBW estimates.

CONCLUSIONS

In conclusion, during high-energy turnover, the overnight-slope method resulted in a more accurate estimation of CO2 production and energy expenditure compared with the plateau or slope-intercept method over a duration of 5 d. Furthermore, day-to-day determination of energy expenditure can also be done with a reasonable accuracy using this method, provided that the turnover rate is high enough to accurately detect the isotope decline over such a short period of time. This makes this approach most applicable for studies in athletes or other circumstances where a high-energy turnover is expected.

No external funding was received for this manuscript. The authors disclose no conflict of interest. The results of the study do not constitute endorsement by the American College of Sports Medicine. All results are presented clearly, honestly, and without fabrication, falsification, or inappropriate data manipulation.

REFERENCES

1. Schoeller DA, Webb P. Five-day comparison of the doubly labeled water method with respiratory gas exchange. Am J Clin Nutr. 1984;40(1):153–8.
2. Klein PD, James WP, Wong WW, et al. Calorimetric validation of the doubly-labelled water method for determination of energy expenditure in man. Hum Nutr Clin Nutr. 1984;38(2):95–106.
3. Melanson EL, Swibas T, Kohrt WM, et al. Validation of the doubly labeled water method using off-axis integrated cavity output spectroscopy and isotope ratio mass spectrometry. Am J Physiol Endocrinol Metab. 2018;314(2):E124–30.
4. Speakman JR. The history and theory of the doubly labeled water technique. Am J Clin Nutr. 1998;68(4):932S–8.
5. Seale JL, Conway JM, Canary JJ. Seven-day validation of doubly labeled water method using indirect room calorimetry. J Appl Physiol. 1993;74(1):402–9.
6. Westerterp K, De Boer J, Saris W, Schoffelen P, Ten Hoor F. Measurement of energy expenditure using doubly labelled water. Int J Sports Med. 1984;5(1 Suppl):S74–S5.
7. Schoeller DA, Ravussin E, Schutz Y, Acheson KJ, Baertschi P, Jequier E. Energy expenditure by doubly labeled water: validation in humans and proposed calculation. Am J Physiol. 1986;250(5 Pt 2):R823–30.
8. Westerterp KR, Brouns F, Saris WH, ten Hoor F. Comparison of doubly labeled water with respirometry at low- and high-activity levels. J Appl Physiol (1985). 1988;65(1):53–6.
9. Ravussin E, Harper IT, Rising R, Bogardus C. Energy expenditure by doubly labeled water: validation in lean and obese subjects. Am J Physiol. 1991;261(3 Pt 1):E402–9.
10. Speakman JR, Yamada Y, Sagayama H, et al. A standard calculation methodology for human doubly labeled water studies. Cell Rep Med. 2021;2(2):100203.
11. Plasqui G, Rietjens G, Lambriks L, Wouters L, Saris WHM. Energy expenditure during extreme endurance exercise: the Giro d’Italia. Med Sci Sports Exerc. 2019;51(3):568–74.
12. Schoeller D, Taylor P. Precision of the doubly labelled water method using the two-point calculation. Hum Nutr Clin Nutr. 1987;41(3):215–23.
13. Cole TJ, Coward WA. Precision and accuracy of doubly labeled water energy expenditure by multipoint and two-point methods. Am J Physiol. 1992;263(5 Pt 1):E965–73.
14. Schoffelen PF, Westerterp KR, Saris WH, Ten Hoor F. A dual-respiration chamber system with automated calibration. J Appl Physiol (1985). 1997;83(6):2064–72.
15. WEIR JB. New methods for calculating metabolic rate with special reference to protein metabolism. J Physiol. 1949;109(1–2):1–9.
16. Chen KY, Smith S, Ravussin E, et al. Room Indirect Calorimetry Operating and Reporting Standards (RICORS 1.0): a guide to conducting and reporting human whole-room calorimeter studies. Obesity (Silver Spring). 2020;28(9):1613–25.
17. Westerterp KR, Wouters L, van Marken Lichtenbelt WD. The Maastricht protocol for the measurement of body composition and energy expenditure with labeled water. Obes Res. 1995;3(1 Suppl):49–57.
18. Deurenberg P, Weststrate JA, Seidell JC. Body mass index as a measure of body fatness: age- and sex-specific prediction formulas. Br J Nutr. 1991;65(2):105–14.
19. Sagayama H, Yamada Y, Racine NM, Shriver TC, Schoeller DA; DLW Study Group. Dilution space ratio of 2H and 18O of doubly labeled water method in humans. J Appl Physiol (1985). 2016;120(11):1349–54.
20. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1(8476):307–10.
21. Hannon MP, Parker LJF, Carney DJ, et al. Energy requirements of male academy soccer players from the English Premier League. Med Sci Sports Exerc. 2021;53(1):200–10.
22. Brinkmans NYJ, Iedema N, Plasqui G, et al. Energy expenditure and dietary intake in professional football players in the Dutch Premier League: implications for nutritional counselling. J Sports Sci. 2019;37(24):2759–67.
23. Sjödin AM, Andersson AB, Högberg JM, Westerterp KR. Energy balance in cross-country skiers: a study using doubly labeled water. Med Sci Sports Exerc. 1994;26(6):720–4.
24. Morehen JC, Bradley WJ, Clarke J, et al. The assessment of total energy expenditure during a 14-day in-season period of professional Rugby league players using the doubly labelled water method. Int J Sport Nutr Exerc Metab. 2016;26(5):464–72.
25. Stroud MA, Coward WA, Sawyer MB. Measurements of energy expenditure using isotope-labelled water (2H2(18)O) during an Arctic expedition. Eur J Appl Physiol Occup Physiol. 1993;67(4):375–9.
26. Westerterp KR, Meijer EP, Rubbens M, Robach P, Richalet JP. Operation Everest III: energy and water balance. Pflugers Arch. 2000;439(4):483–8.
27. Ellis DG, Speakman J, Hambly C, et al. Energy expenditure of a male and female tennis player during Association of Tennis Professionals/Women’s Tennis Association and grand slam events measured by doubly labeled water. Med Sci Sports Exerc. 2021;53(12):2628–34.
28. Rosenkilde M, Morville T, Andersen PR, et al. Inability to match energy intake with energy expenditure at sustained near-maximal rates of energy expenditure in older men during a 14-d cycling expedition. Am J Clin Nutr. 2015;102(6):1398–405.
29. Thurber C, Dugas LR, Ocobock C, Carlson B, Speakman JR, Pontzer H. Extreme events reveal an alimentary limit on sustained maximal human energy expenditure. Sci Adv. 2019;5(6):eaaw0341.
30. Ruby B, Cuddy J, Hailes W, et al. Extreme endurance and the metabolic range of sustained activity is uniquely available for every human not just the elite few. Compar Exerc Physiol. 2015;11(1):1–7.
31. Okubo H, Sasaki S, Rafamantanantsoa HH, Ishikawa-Takata K, Okazaki H, Tabata I. Validation of self-reported energy intake by a self-administered diet history questionnaire using the doubly labeled water method in 140 Japanese adults. Eur J Clin Nutr. 2008;62(11):1343–50.
32. Andersen LF, Tomten H, Haggarty P, Lovo A, Hustvedt BE. Validation of energy intake estimated from a food frequency questionnaire: a doubly labelled water study. Eur J Clin Nutr. 2003;57(2):279–84.
33. Frings-Meuthen P, Henkel S, Boschmann M, et al. Resting energy expenditure of master athletes: accuracy of predictive equations and primary determinants. Front Physiol. 2021;12:641455.
34. ten Haaf T, Weijs PJ. Resting energy expenditure prediction in recreational athletes of 18–35 years: confirmation of Cunningham equation and an improved weight-based alternative. PLoS One. 2014;9(9):e108460.
35. Coward WA, Prentice AM. Isotope method for the measurement of carbon dioxide production rate in man. Am J Clin Nutr. 1985;41(3):659–63.
36. Henry CJ. Basal metabolic rate studies in humans: measurement and development of new equations. Public Health Nutr. 2005;8(7A):1133–52.
37. Jeukendrup AE, Wallis GA. Measurement of substrate oxidation during exercise by means of gas exchange measurements. Int J Sports Med. 2005;26(1 Suppl):S28–37.
38. Kipp S, Byrnes WC, Kram R. Calculating metabolic energy expenditure across a wide range of exercise intensities: the equation matters. Appl Physiol Nutr Metab. 2018;43(6):639–42.
39. van Marken Lichtenbelt WD, Westerterp KR, Wouters L. Deuterium dilution as a method for determining total body water: effect of test protocol and sampling time. Br J Nutr. 1994;72(4):491–7.
40. de Jonge L, DeLany JP, Nguyen T, et al. Validation study of energy expenditure and intake during calorie restriction using doubly labeled water and changes in body composition. Am J Clin Nutr. 2007;85(1):73–9.
41. Zhang K, Sun M, Werner P, et al. Sleeping metabolic rate in relation to body mass index and body composition. Int J Obes Relat Metab Disord. 2002;26(3):376–83.
Keywords:

ENERGY EXPENDITURE; ROOM INDIRECT CALORIMETRY; DOUBLY LABELED WATER; HIGH-ENERGY TURNOVER; VALIDITY

Supplemental Digital Content

Copyright © 2022 The Author(s). Published by Wolters Kluwer Health, Inc. on behalf of the American College of Sports Medicine.