Reliability of Measurements of Energy Expenditure and Substrate Oxidation Using Whole-Room Indirect Calorimetry

Abstract

Objective:

The objective of the present analysis was to measure the intra-subject reliability - the intraclass correlation coefficient (ICC) – of all the components of daily energy expenditure (EE) (24 hr EE, sleep EE, resting EE, basal EE, and thermic effect of food) over a period of 3 consecutive days in 35 study participants.

Methods:

The components of daily EE and substrate utilization (respiratory exchange ratio; RER) were measured over 3 consecutive days before and after a 3-week 1000 kcal/day caloric restriction/weight loss intervention.

Results:

There was a high degree of reliability for sleep EE (96.8%), 24 hr EE (97.8%), basal EE (90.6%), and resting EE (93.2%) during the run-in period. The ICC for the follow-up period after weight loss (3.67 ± 1.10 kg) remained high for sleep EE (95.6%), 24 hr EE (100%), basal EE (96.1%), and resting EE (92.5%). The minimal detectable differences in EE were reduced by 30% for both 24hr EE and sleep EE when comparing two vs. one day spent in the whole-room indirect calorimeter.

Conclusions:

The reliability of the daily components of EE is very high both prior to and after a weight-loss intervention. We here provide instrumental data for investigators to adequately power studies investigating energy metabolism using whole-room indirect calorimetry.

Introduction

According to the World Health Organization in 2016, 1.9 billion adults were overweight with an additional 650 million being classified as having obesity. In the United States, the prevalence of obesity has been reported to be 42.4 % as of 2017–2018 (1). Obesity is now recognized as a chronic medical condition that increases the risk of hypertension, dyslipidemia, type 2 diabetes mellitus (T2DM), cardiovascular disease, and many forms of cancer (2–4). The medical community has made concerted efforts to develop lifestyle interventions, promote the use of safe and efficacious medications as well as metabolic surgeries to treat obesity with varying degrees of efficacy (5).

Weight gain and eventually obesity is caused by a sustained positive energy balance during which energy intake exceeds energy expenditure (EE). To better understand the causes of weight gain, we need to study energy balance which requires valid and reproducible methods to measure both energy intake (calories consumed) and EE (calories expended). The regulation of energy intake is a complex feedback system involving neurological and humoral signals integrating external and internal stimuli. Several methods exist to measure energy (food) intake ranging from food recalls and questionnaires, to the use of the intake/balance method with doubly labeled water in free-living conditions to highly controlled eating environments setup in laboratories. Similarly, EE can be measured by different methods including direct calorimetry, indirect calorimetry, and doubly labeled water, all of which are considered valid methods but with varying degrees of applicability and feasibility (6).

The use of indirect calorimetry using metabolic carts, while providing a convenient measurement of resting metabolic rate (RMR), does not allow for longer-term measures of EE including all the components of daily EE. The use of whole-room indirect calorimetry is a valuable research tool because it allows for real-time minute-by-minute data collection of oxygen consumption (VO₂) and carbon dioxide production (VCO₂) to measure EE and substrate utilization (6). Indeed, whole-room calorimeters (often called metabolic chambers) can capture the individual components of 24-hour EE (24hr EE) including (1) RMR, (2) thermic effect of food (TEF), and (3) activity thermogenesis (6, 7). The RMR represents approximately 70% of 24hr EE and is determined by fat-free mass, fat mass, sex, age, as well as genetic factors (7, 8). The thermic effect of food (TEF) is the smallest component of 24hr EE (approximately 10%) and refers to EE above fasting EE divided by the ingested energy (9–11). Physical activity is the most variable component of 24hr EE that is also subject to complex interactions between physiological and neurological pathways (12).

To date, very little data exist regarding the reliability (repeated testing) of the components of EE and substrate oxidation using whole-room indirect calorimetry. As treatments for obesity continue to evolve, with now new pharmaceutical agents including thermogenic drugs, the need for information regarding the reliability and reproducibility of the different measures of EE is warranted. The purpose of this analysis was to calculate the intraclass correlation coefficients (ICCs) for 24-hr EE, sleeping EE, and other measurements of substrate oxidation using three consecutive days of whole-room indirect calorimetry. We also investigated the impact of a weight loss intervention (−1000 kcal/day) for 19 days to determine the variability of our measurements in the context of the weight reduced state.

Methods

Participants

We are reporting data from a larger two-site clinical trial using an investigational weight-loss medication compared to a placebo-controlled group with both groups undergoing a caloric restriction period. In this manuscript, we report the baseline “run-in” data from both groups (n=35) and the weight loss data from the placebo group only (n=18). The study is registered with ClinicalTrials.gov, number NCT03376802. Twenty-four males and eleven females participated in the study. The study eligibility criteria included male and female participants with 1) overweight or obesity class I (BMI 28–35 kg/m²), 2) 18–45 years old, 3) without diabetes with a fasting plasma glucose ≤ 125 mg/dl and an HbA1c ≤ 6.5%. Some obesity-associated concomitant conditions (i.e., mild hypertension with SBP <160 mm Hg, mild hypercholesteremia (total cholesterol between 200–239 mg/dl), and hyperlipidemia with fasting triglycerides up to 450 mg/dL) were allowed. Stable medications were allowed except for treatment with statins or anti-hypertensive drugs (expect β-blockers) as well as other agents known to impact energy metabolism. Female participants were required to be either postmenopausal or, if perimenopausal should have a regular menstrual cycle or receive stable treatment with a monophasic oral contraceptive. The study was approved by the Institutional Review Boards at Pennington Biomedical Research Center (PRBC) and AdventHealth Translational Research Institute (TRI). Study volunteers provided written informed consent before participation. Subject characteristics are provided in Table 1.

Table 1.

Study participants characteristics at run-in period

	Run-in (n=35)	Follow-up (n=18)
Age (years)
Mean (SD)	36.5 (7.1)	36.2 (7.1)
Min : Max	22 : 45	23 : 45
Sex [n (%)]
Female	11 (31.4%)	6 (33.3%)
Male	24 (68.6%)	12 (66.7%)
Race [n (%)]
White	14 (41.2%)	8 (47.1%)
Black or African American	16 (47.1%)	7 (41.2%)
Asian	1 (2.9%)	0
Unknown	3 (8.8%)	2 (11.8%)
Ethnicity [n (%)]
Hispanic or Latino	10 (28.6%)	7 (38.9%)
Not Hispanic or Latino	22 (62.9%)	10 (55.6%)
Not Reported	3 (8.6%)	1 (5.6%)
Weight (kg)
Mean (SD)	91.94 (10.48)	90.33 (11.73)
Min : Max	74.4 : 117.5	74.4 : 117.5
BMI (kg/m2) [n (%)]
<30	20 (57.1%)	10 (55.6%)
≥30	15 (42.9%)	8 (44.4%)

BMI, body mass index; SD, standard deviation

Study Design

Eligible participants (n=35) were required to be institutionalized at their respective study site for the duration of the study. All standardized meals were prepared by the two-study site’s metabolic kitchens and consumed at the two different metabolic wards. Participants entered the room calorimeter during the run-in period (days −3 to −1) for 23 hrs on 3 consecutive days. After completing the run-in period, study participants were randomized to an investigational drug group (not included in this analysis) or placebo control group (n=18) at a 1:1 ratio and stratified by study site and biological sex. Both groups underwent an inpatient caloric restriction intervention which consisted of custom diets to achieve a 1000 kcal/day deficit for 19 days to induce a modest weight loss. On days 17–19, all participants completed an additional 3 consecutive days (follow-up) in the room calorimeter to assess the reliability of the energy metabolism measures after weight loss. We provide data only for the participants receiving placebo treatment and completing all study visits (n=17).

Indirect calorimetry

The current study utilized 2 whole-room calorimeters at PBRC and 2 at TRI. The whole-room calorimeters have a total volume of 27 m³ and 30 m³ for the PBRC and TRI site respectively. The whole-room calorimeters at both sites were maintained at 22.5°C and below 40% humidity. The protocol for calibration was standardized across sites. Measurement of O₂ and CO₂ concentration as well as flow rate through the calorimeter were obtained every 10 seconds. EE was calculated from the measurement of VO2 and VCO2(13). Substrate oxidation was then determined via the respiratory exchange ratio (RER=VCO2/VO2) and from the rate of urinary nitrogen excretion (13). According to the study protocol, study participants were required to refrain from drinking alcohol and caffeine-containing substances (i.e. coffee, tea, chocolate, sodas) as well as refraining from any strenuous physical activity for 3 days before the run-in period. During the 3 days preceding the run-in room calorimeter confinement, participants were provided with a weight-maintenance diet to consume at home. The diet composition was 15% protein, 30% fat, and 55% carbohydrates. The energy content of each diet was prescribed according to the body composition model as described by Lam et al (14). The equation was modified with a factor of 1.2 to account for free-living physical activity. 24hr Energy Intake (kcal/day) = 1.2 × [26.2 ([FFM (kg)]) + 5.2 ([FM]) – 2.32 ([age(y)]) – 96 (African American) + 546]

Per study protocol, participants entered the whole-room calorimeter at 0800 after an overnight stay in the inpatient unit. At 0830 participants were instructed to lay in the bed quietly for 25 minutes (0830 to 0855) to measure resting EE (REE) and RER (resting RER). At 0900 participants were served breakfast and allowed 30 minutes to eat it. The TEF was measured in the postprandial period between 1000 to 1125 awake under resting conditions while lying in bed. Participants were allowed free time, which included watching TV, reading, computer work, etc. but excluded activities such as stretching or pushups. Participants were served lunch at 1300 and dinner at 1800 and allowed 30 minutes to eat the meals. At 2200 participants were instructed to void their bladder, collect the urine before turning out the lights. Sleep EE and sleep RER were measured between 0000 and 0500 when radar-measured activity was zero. At 0550 the lights were turned on and participants were awakened and instructed to empty their bladder. Participants returned to lay on the bed quietly for 1 hr (0600 to 0700) to measure basal EE and basal RER. At 0700 participants were instructed to exit the whole-room calorimeter for 50 minutes while the chamber was cleaned, and the instruments calibrated for the next 23-hr measurement period.

Weight loss intervention

After participants completed the run-in baseline measurements, a caloric restriction diet was provided with the goal of a 1000 kcal deficit. The macronutrient content of the diet was between 19–22% protein, 53–54% carbohydrate, and 25–27% fat. The caloric content of the diet was calculated as follows:

$$\text{Energy}\text{intake}=\left[\text{EE}\text{metabolic}\text{chamber}\left(\text{average}\text{of}3\text{days}\right)*1.2\right]-1000\text{kcal}$$

Participants were institutionalized during the entire study and required to eat all meals on the metabolic ward under the supervision of the study dieticians or research staff. No outside food was allowed. All study meal trays were checked after consumption, and any food that was not consumed was weighed and subsequent meals were adjusted for uneaten food. The total calories not consumed per meal were calculated and recorded. Participants completed the final 3-day room calorimeter visit on day 17–19 and were fed the same energy-restricted diet.

Statistical Analysis

All participants (n=35) were included in the run-in analysis, whereas only participants treated with placebo (n=18) were included in the analysis of the follow-up period. The primary analysis of this study used a linear model including treatment and stratification factors of site and sex as fixed effects and the components of EE at run-in period and body composition and treatment period as covariates. Differences between the mean of the run-in and follow-up periods for the placebo group were analyzed using a paired t-test in JMP software (version 15.4). The ICC for the run-in period before the study intervention (day −3, −2, and −1) and the follow-up period (day 17, 18, and 19) were computed separately using linear mixed effects model on raw calorimeter values including no fixed effect apart from the intercept and subject intercept as a random effect. Analyses were done with SAS version 9.4, using the Proc MIXED procedure. Bland-Altman analysis (difference between run-in day vs average) was conducted to assess the agreement between run-in days for sleep and 24hr EE (15). To show the increased measurement sensitivity of using multiple days whole-room indirect calorimeter, a power analysis was conducted. Assuming a power of 0.80 and an alpha = 0.05 (Type 1 error), and using data from the baseline visit, the minimal detectable differences needed for a two-sample t-test to observe statistically significant changes over time were computed for sample sizes between 20 and 200.

Results

Energy Expenditure and Substrate Utilization

Participants assigned to the placebo control group were provided with a caloric restriction diet to induce weight loss. Figure 1 demonstrates participants lost a significant amount of body weight between pre-run-in (Day −4) and pre-follow-up periods (Day 16) (−3.67 ± 1.10 kg, P<0.0001). EE and RER were measured during the 3-day study for the run-in (Day −3 to −1) and follow-up periods (Day 17 to 19). Figure 2A shows 24hr EE for the run-in period (2083 ± 288 kcal/day) which was significantly reduced (P<0.0001) when repeated during the follow-up period (1908 ± 301 kcal/day). The 24hr RER (24hr RER, Figure 2B) was significantly reduced when comparing run-in to follow-up periods (0.89 ± 0.02 to 0.85 ± 0.02; P<0.0001). Similarly, sleeping EE (Figure 2C) and sleep RER (Figure 2D) were significantly (P<0.0001) reduced when comparing the run-in period (1579 ± 257 kcal/day & 0.86 ± 0.02) to the follow-up period (1467 ± 272 kcal/day & 0.81 ± 0.02). Resting EE (1847 ± 331 vs. 1704 ± 327 kcal/day; P<0.0001) and resting RER (0.88 ± 0.04 vs. 0.85 ± 0.03; P=0.01) as well as basal EE (1748 ± 260 vs. kcal/day; P=0.01) and basal RER (0.88 ± 0.03 vs. 0.84 ± 0.03, <0.0001) were all reduced when comparing the run-in versus the post-caloric restriction period. There was no effect of study site when evaluating the changes in sleep EE (p=0.76) and 24hr EE (p=0.84).

Figure 1.

This figure shows the inter-participant variability in body weight of the study participants. Body weight was significantly reduced when comparing the pre-run-in period (day -4) with the pre-follow-up period (day 16) values ***P ≤ 0.001 after a 16-day 1000 kcal/day caloric restriction

Figure 2.

This figure provides data on the reliability of measures of energy expenditure and respiratory exchange ratio (RER) during a run-in eucaloric phase (day −3 to −1) and after a 16-day 1000Kcal/day caloric restriction (day 17–19). (A) 24hr EE was reduced the run-in period (day −3 to −1) versus the follow-up period (day 17–19). (B) 24hr RER Sleep EE, (C) Sleep RER, (D) Resting EE were all significantly reduced when comparing run-in period (day −3 to −1) versus the follow-up period. ****P ≤ 0.0001

We calculated the ICC as a measurement of reliability for the different components of EE and RER over 3 consecutive days for the run-in and follow-up periods, respectively. Table 2 shows that all parameters of EE (sleep EE, 24hr EE, basal EE, and resting EE) for the run-in and follow-up periods are greater than 90% which indicates excellent reliability (16, 17). In contrast, the ICCs for the RER parameters ranged from approximately 50% to 75% during the run-in period which would be considered moderate to good reliability (16). Despite lower overall ICC for RER versus EE, all the ICC values for RER parameters (sleep RER, 24hr RER, basal RER, and resting RER) during the follow-up period were increased when compared to the run-in period (~58% to 87%). A Bland-Altman (15) for the run-in period (Figure 3) reveals good agreement between days (D-3 vs D-2, D-2, vs D-1, & D-3 vs. D-1) in the whole room indirect calorimeter for sleep EE (Figure 3A, C, & D) and 24hr EE (Figure 3B, D, & F).

Table 2.

Measures of variance and intraclass correlation for run-in and follow-up periods

	Run-in Period n = 35				Follow-up Period n = 17
Parameter	Between-participant variance	Within-participant variance	ICC (%)	CV (%)	Between-participant variance	Within-participant variance	ICC (%)	CV (%)
Sleep Energy Expenditure (kcal/day)	63517	2094	96.8%	2.1%	68356	2888	95.9%	2.8%
24hr Energy Expenditure (kcal/day)	80420	1843	97.8%	1.9%	36540855	818	100%	1.5%
Basal Energy Expenditure (kcal/day)	60789	6322	90.6%	3.9%	71704	2872	96.1%	2.4%
Resting Energy Expenditure (kcal/day)	100874	7388	93.2%	3.7%	93879	7557	92.5%	4.6
Thermic Effect of Food (%)	156	18	89.5%	25.6%	117	16	87.9%	21.3%
Sleep Respiratory Exchange Ratio	0.00031	0.00021	60.4%	1.5%	0.00038	0.00013	74.9%	1.1%
24hr Respiratory Exchange Ratio	0.00031	0.00011	74.2%	1.1%	0.00043	0.00006	87.0%	0.7%
Basal Respiratory Exchange Ratio	0.00043	0.00044	49.6%	2.1%	0.00050	0.00036	57.9%	1.9%
Resting Respiratory Exchange Ratio	0.00081	0.00046	63.7%	2.2%	0.00067	0.00025	72.3%	1.5

Dependent variable: raw calorimeter data values collected at Day −3, −2 and −1 for the run-in period and Day 17, 18, and 19 for the follow-up period.

Unconditional means model: the linear mixed model includes no fixed effect apart from the intercept (random intercept by participant).

Intraclass correlation coefficient estimates the proportion of total variation in the dependent variable that lies “between” participants. ICC, intraclass correlation coefficient; CV, coefficient of variation.

Figure 3.

This figure shows Bland-Altman plots to assess the agreement between days during the run-in period in the whole room indirect calorimeter. A, C, & E: Sleep EE run-in days D3 vs D2. D2 vs D1 & D3 vs D1, respectively. B, D, & F: 24hr EE run in days D3 vs D2. D2 vs D1 & D3 vs D1, respectively. Solid line indicates the regression model and dotted lines represent the 95% limits of agreement.

Since the ICC is a relative measurement of reliability, we performed a power analysis for sample size and the change in minimal detectable differences (MDD) in kcal/day between the 3 days in the room calorimeter (Figure 4). Overall, the MDD was reduced by 30% for both 24hr EE and sleep EE when comparing one (D1) to two days (D2) of whole-room calorimeter days. When comparing three days (D3) to one day (D1) of calorimeter measurement, the MDD for 24hr EE and sleep EE were both reduced by 41%. The MDD was only reduced by 17 and 15% when comparing D2 and D3 respectively.

Figure 4.

This figure shows a comparison of sample size and the minimal detectable difference (y axis) as a function of the number of study participants (x axis) for changes over time using either 1, 2 or 3 days in the whole-body room calorimeter for 24hr energy expenditure (24hr EE; A) and sleeping energy expenditure (Sleep EE; B) during the run-in period (n=35) with an 80% power (1-β=.80).

Discussion

The purpose of our analysis was to assess the reliability of EE and RER over 3 consecutive days in whole-room indirect calorimeters. We report for the first-time excellent reliability for both the run-in and follow-up period according to our ICC values for 24hr EE, sleep EE, basal EE, and resting EE (all ≥90%). The ICC values for all RER measurements were lower than for EE across all parameters. While a clear explanation for lower ICC during the run-in period for the RER is not apparent, we noticed a consistent increase in the ICC for all parameters during the follow-up period. During the follow-up period (post-weight loss) the average RER was reduced due to increased fat oxidation triggered by caloric restriction with probably a “floor effect” compared to a less controlled dietary conditions during baseline despite a run-in period. How the change in substrate oxidation alters the reproducibility of this measurement is not known and is outside the scope of the present report. We were most interested in the day-to-day reliability in 24hr EE and sleep EE. Sleep EE is considered an important primary endpoint to study when comparing changes in EE because it is not subject to several sources of variability including physical activity, the TEF, and stress (18, 19). Not surprisingly, our data also demonstrated that weight loss (3.67 ± 1.10 kg) was accompanied by a significant reduction in sleep EE (Figure 2A). However, our measurement of reliability (ICC) remained similar despite the reduction in EE experienced after weight loss.

Importantly, we here provide a retrospective power analysis to help researchers identify MDD in EE regardless of study intervention (Figure 3). Our analysis demonstrates the value of two days of measurements when compared to a single day in the whole-room calorimeter. For example, to detect a 160 kcal/day difference in 24hr EE, a single day whole-room calorimeter measurement would require 100 participants at 80% power. Adding a day (D1 and D2) would reduce the sample size requirement by 50% (n=50). Adding an additional day (D1, D2 and D3) also reduces the sample size, but by only approximately 30%. Therefore, an additional increase in the number of whole-room calorimeter measurement days to a single day significantly reduces the sample size, but with some diminishing response as additional days are added. In one of our previous studies using 6 consecutive days of whole-room calorimetry measurement, a power analysis revealed an MDD of 26.5 kcal/day (20). The subtleness of differences in EE expected after the intervention will ultimately help to determine the measurement period required to power a study using whole-room calorimeters.

Approaches for obesity, which include novel pharmaceuticals, seek to target both sides of the energy balance equation (energy intake and EE) and efficacy might be improved by dampening metabolic adaptation. To test the efficacy of weight loss interventions, which includes pharmaceuticals acting on EE, sophisticated measuring methods are needed. Whole-room indirect calorimetry represents the state-of-the-art method to measure reliably all the components of sedentary EE before and after weight loss and therefore reliably determine the metabolic adaptation. Furthermore, whole-room indirect calorimeters allow the measurement of EE over the course of several hours to days. With a properly constructed study script, the individual components of EE including sleep EE, resting EE, and basal EE can all be measured. Despite these advantages, the widespread use of whole-room indirect calorimetry is still limited probably because of their cost and technical need for maintenance. Unfortunately, studies implementing repeated days of measurement or multi-site trials are scarce. However, according to the recently published Room Indirect Calorimetry Operating and Reporting Standards (RICORS 1.0), more whole-room calorimeters have been constructed and utilized worldwide (7). Therefore, studies documenting the reliability of whole-room calorimeters are now critical to power such trials.

A limitation of our study is that we only investigated study participants with overweight or class I obesity that were relatively young (~36 year of age). Therefore, we are not able to generalize our findings to those in other BMI categories or older adults. While nearly half (~47%) of our participants were classified as black or African American we are also hesitant to generalize our results to other non-white or non-black individuals without a more robust sample size. Finally, the highly structured nature of the study script which included rigorous mealtimes and bedtimes may be different from real-world habitual activity. Studies designed to test the impact of scripted versus voluntary behaviors (eating and sleeping times) would be required to better understand this paradigm.

In conclusion, we have demonstrated that the measurement of EE using whole-room indirect calorimetry is very reliable. Furthermore, the reliability of our measurement is not changed when a weight-loss intervention is implemented, and participants are retested. The growing number of whole-room indirect calorimeters in use suggests more research will interrogate EE as an endpoint for weight loss studies. Each whole-room calorimeter and study site should establish integrated reproducibility to power studies. The result of our current study demonstrates that 24 hr and sleep EE are stable endpoints that can be reliably measured.