Total Energy Expenditure in Healthy Ambulatory Older Adults Aged ≥80 Years: A Doubly Labelled Water Study

Worldwide, 125 million people are aged ≥80 years or older, and it is predicted that by 2050, this will increase to 434 million [1]. In Australia, over half a million people are aged ≥85 years or older, and it is predicted that by 2066, this will increase to 2.2 million [2]. Although the population is living longer, old age may frequently be associated with poor quality of life and frailty. For the promotion of healthy ageing, an accurate assessment of energy requirements is essential.

Fundamental to well-being is the consumption of adequate energy from food to match the body’s requirements. In Australia, almost 70% of older adults are overweight or obese [2], while one-third of those admitted to hospital are malnourished (undernourished) [3]. Overweight and obesity increase the likelihood of developing many chronic conditions, including some cancers, cardiovascular disease, diabetes, and dementia, and the impact on the Australian economy is $8.6 billion per year [4]. Malnutrition is the major contributor to increased morbidity and mortality in older adults, associated with increased frequency and length of admission to hospital (additional 5 days), decreased functional status and quality of life [3]. Accurate determination of total energy expenditure (TEE) helps promote healthy weight and guide dietary advice; this information is crucial in helping tackle both obesity and undernutrition in older adults. Moreover, the assessment of energy requirements, with the greatest possible degree of accuracy, is one of the most important and central decrees of the Food and Agriculture Organization of the United Nations [5].

The Schofield equation is used to estimate resting energy expenditure and together with a physical activity level (PAL) can provide an estimation of TEE [6]. The Schofield equation is commonly used in clinical practice and public nutrition guidelines, including the World Health Organization (WHO) [5] and the NHMRC Nutrient Reference Values for Australia and New Zealand [7]. This equation was, however, published in 1985 and derived from largely younger adults (<60 years old), with disproportionate number of males compared with females and with bias towards certain population groups (e.g., males in the military) [6]. While the Schofield equations are separated into age categories (10–18 years old, 18–30 years old, 30–60 years old), the energy requirements for all individuals aged ≥60 years are calculated with just a sole equation [6] and based on extrapolated data for those >70 years [7]. There is no age categorization for those in their 80th or 90th decade and beyond. This fails to recognize the range of factors that potentially differentiate energy requirements in older age brackets, such as body composition, metabolic rate, and physical activity [8].

The reference method for the measurement of TEE is doubly labelled water (DLW). Briefly, DLW is based on the measurement of the dilution spaces and the elimination rates of the tracers (via spot urine collection) over a period of 7–14 days after ingestion of water labelled with two non-radioactive stable isotopes, deuterium and oxygen-18. The difference in elimination rates is proportional to carbon dioxide (CO₂) production and is thus used to calculate TEE [9, 10]. Our team has previously systematically identified and collated over 800 international participant level (TEE) data measured by DLW in adults aged ≥65 years old [11]. We then compared this large dataset with a series of 17 prediction equations and identified that the Mifflin [12], Ikeda [13], and Livingston [14] equations most accurately estimated measured energy expenditure [15]. Moreover, through our systematic review, we were able to determine the extent to which energy requirements have been obtained using DLW in older adults ≥80 years. The paucity of measurements was marked, with only nine studies identified, dating back to 1989 with no studies from Australia [11]. This is in comparison with 28 studies in the 65–79 year old age range, from which only one study was from Australia (n = 4, 66–69 years old) [16].

Given the pronounced absence of Australian data and limited knowledge on the energy requirements of contemporary older adults aged ≥80 years, the aim of this study was to characterize the TEE measured using DLW of older Australian adults aged ≥80 years and to compare them with established prediction equations. The hypothesis was that the Mifflin, Ikeda, and Livingston equations will more closely estimate energy requirements than the commonly used Schofield equation.

This study was of cross-sectional design. Data were collected prior to the COVID-19 pandemic between September 2018 and September 2019. Participants reported to the Monash University, Department of Nutrition, Dietetics and Food, Be Active Sleep Eat (BASE) Facility, Melbourne (between 07: 30 and 08: 00 h) after an overnight 10 h fast. Participants were asked to avoid alcohol and strenuous physical activity for 24 h prior to each study day. Baseline anthropometric and body composition measures were taken, and participants rested supine for 15 min after which resting metabolic rate (RMR) was measured for 30 min by indirect calorimetry. After a fasting urine sample, a dose of DLW was administered for the measurement of TEE and spot urine, and saliva samples were provided at and between three and 4 h post dose, respectively. Quality of life questionnaires were administered during this period. At the end of the 4 h period before leaving the facility, participants were provided lunch and instructions on the collection of urine samples at home over the subsequent 2-week study period. At the end of the 2 weeks, the urine samples were collected from the participant’s home and stored at the research facility at −20°C until analysis.

Participants were recruited from the east and south eastern suburbs of Melbourne via flyers, social media, newsletters, and notices placed at local libraries, community centres, gyms, and retirement villages. Participants were eligible if they were aged ≥80 years. Interested participants contacted the researcher directly via phone call or email, and screening for eligibility was conducted over the phone. Inclusion criteria were healthy, ambulatory, non-institutionalized adults and able to provide consent. To assess that adults were cognitively adept, a screen for cognition was conducted using the six-item screener [17]. Exclusion criteria were taking medications known to affect energy expenditure; diabetes (on insulin or diabetes medications); inability to fast overnight; requirement for supplemental oxygen; blood transfusions, administration of blood products, or administration of intravenous fluids in excess of 500 mL in the week prior to the study or throughout the 2-week study. The need for travel more than 320 km from home during the week prior to the study and throughout the 2-week study period was also an exclusion criterion, as changes in geographical location can cause baseline isotope shifts and therefore inaccuracies in the calculation of energy expenditure [18]. All participants provided informed consent for participation in this study. This study was conducted according to the guidelines laid down in the Declaration of Helsinki, and all procedures involving human participants were approved by the Eastern Health Human Research Ethics Committee (project code E13-2017) and the Monash University Human Research Ethics Committee (project code 12115).

Anthropometric measures were determined without shoes and in light clothing. Height was measured using a stadiometer (Holtain Ltd., Crymych, UK), weight using an electronic scale (SECA Group, Hamburg, Germany), and umbilical waist circumference and hip circumference using a stretch-resistant tape measure (Figure Finder, Novel Products, Rockton, IL, USA). Blood pressure was measured using an automated blood pressure monitor after a 5 min seated rest period (Welch Allyn, Skaneateles Falls, NY, USA). All measures were taken in duplicate following standardized procedures and recorded to the nearest one decimal place. Hand grip strength was measured using the Jamar hand dynamometer (J.A. Preston Corporation, Clifton, NJ, USA) three times on each hand starting with the right hand and averaged. Body composition was measured using dual-energy X-ray absorptiometry whole body scans (i-DXA, GE Healthcare Lunar, Madison, WI, USA), and calibration was performed daily using a known calibration standard according to manufacturer’s instructions.

The Investigating Choice Experiments CAPability measure for Older people (ICECAP-O) questionnaire was administered on the study day [19]. This comprised five questions assessing capability across the attributes: attachment; role; enjoyment; security; and control. A single index score of between 0 (no capability) and 1 (full capability) was calculated using an established algorithm [19]. The Assessment of Quality of Life (AQoL-8D) questionnaire was also completed [20]. This comprised 35 questions across the dimensions of independent living, happiness, mental health, coping, relationships, self-worth, pain, and senses. A global “utility” score of between 0 (death) and 1 (full health) was calculated using an established algorithm [20].

RMR was measured by indirect calorimetry using the Vyntus CPX canopy system (Care Fusion, Höchberg, Germany) in a thermally controlled environment. On each study morning, volume and gas analyser were calibrated against standard oxygen (O₂) and CO₂ calibration gases (Air Liquide, Sunshine Australia). After a 15 min rest period whereby participants were asked to lie at complete rest and awake in a supine position, energy expenditure (O₂ and CO₂ respiratory exchange measured every minute) was measured for 30 min. Discarding the first 5 min, steady state was determined by at least a subsequent 5 min period with ≤10% coefficient of variation for VO₂ and VCO₂ [21]. Energy expenditure (kJ) was derived from oxygen consumption (VO₂) and carbon dioxide (VCO₂) production using the modified Weir equation {([VO₂ × 3.941]) + ([VCO₂ × 1.11]) × 1,440} × 4.182 [22]. Estimated RMR was calculated using the Ikeda, Livingston, and Mifflin equations.

TEE was measured using the DLW technique following standard protocol from the International Atomic Energy Agency (IAEA) [10]. A baseline fasting urine sample was collected for determination of background isotope enrichment. Participants were then provided with an oral dose of DLW according to the recommended dose based on total body weight. The mixture from which the dose was derived consisted of a 1:15 ratio of deuterium oxide (99.8 atom % excess; Sigma Aldrich, Castle Hill, Australia) to 18-oxygen (10 atom % excess; Taiyo Nippon Sanso, Tokyo, Japan) [10]. The dose bottle was rinsed with 2 × 50 mL tap water to ensure no labelled water remained in the bottle. One hour after dosing, participants were asked to completely void their bladders – this was not analysed. A small snack was offered thereafter. At 3 and 4 h post-dosing, participants provided urine samples, and a saliva sample was collected between the three and 4 h time points. During this period of equilibration, the volume of water consumed by participants was recorded and did not exceed 500 mL and was subtracted from the final total body water (TBW) value. Over the next fortnight, participants collected urine samples on days 1, 2, 3 and days 12, 13, 14 post dose. Urine samples were stored frozen at −20°C for batch analysis. Analysis of isotopic enrichment was determined with an Isoprime Dual Inlet Isotope Ratio Mass Spectrometer (Isoprime, Manchester, UK) coupled in-line with a Multiprep Gilson autosampler. Analysis was undertaken at the Department of Nutrition, Dietetics and Food (Green Chemical Futures Building, Monash University, VIC, Australia). Hydrogen analyses were completed by a 4 h equilibration with H₂ gas at 40°C using Hokko Coils. Oxygen analysis was completed with an overnight equilibration with CO₂ at 40°C. All samples were analysed in duplicate, and laboratory standards were calibrated using the international standards Standard Mean Ocean Water, Standard Light Antarctic Precipitation, and Iso-Analytical RO55. Results were reported in % (delta per mil units) relative to SMOW. TEE and TBW were calculated using the methods as described previously [23]. TBW was then converted to fat-free mass (FFM), assuming that it contains 73% water, and fat mass was then calculated by subtracting FFM from total body weight.

Comparisons of participant characteristics were performed using independent t tests for normally distributed variables or the Mann-Whitney U test for non-normally distributed variables. The Bland-Altman method was used to assess the level of agreement between two measures (i.e., predicted compared with measured RMR by indirect calorimetry and predicted TEE compared with measured TEE by DLW). Predicted TEE was determined by multiplying predicted RMR from the equations (Mifflin, Ikeda, and Livingston and Schofield) by the average PAL for adults ≥65 years old determined from our previous research (PAL = 1.69 n = 320 men; PAL = 1.66, n = 668 women) [15]. Applying a predefined group PAL reflects the process undertaken in clinical practice. A one-sample t test was applied to assess whether the difference between the two measurements varied significantly from zero (p < 0.05). If the variation was not significantly different from zero, a Bland-Altman plot was constructed to assess the agreement between measures [24]. The difference between measured and predicted RMR or TEE (y-axis) was compared with measured RMR or TEE (rather than the mean of the two variables as is traditionally used), given these are reference measures. The upper and lower limits of agreement were calculated as the mean difference ±1.96 × SD. A linear regression was conducted to assess for proportional bias (p < 0.05) to determine whether there was a trend of more data points being above or below the mean difference. Correlations were performed using Pearson for normally distributed variables and Spearman for non-normally distributed variables. The significance was defined at p < 0.05. Data were expressed as mean ± SD unless otherwise stated. Analyses were performed using IBM Statistical Package for Social Sciences (SPSS) version 26.0 (IBM Corp., Armonk, NY, USA). Sample size calculations to determine the number of participants for this study were not conducted as this was an observational pilot study aimed to provide descriptive data on TEE in older Australian adults, where there are currently no existing data available.

As shown in Figure 1, n = 35 participants were screened and of these participants, 14 were excluded. Baseline characteristics of study participants included and analysed (n = 21) are presented in Table 1. Measurements were conducted on twenty-one older adults aged ≥80 years old (n = 7 males, n = 14 females) with a median age of 83.7 years (range 80.7–90.1 years old). One participant was excluded from TEE analysis due to incomplete isotopic equilibration. There were no withdrawals.

RMR measured by indirect calorimetry and TEE measured by DLW are presented in Table 2. Mean RMR was 4,734 ± 912 kJ/day and mean TEE was 8,599 ± 1,162 kJ/day, with males having a significantly greater RMR and TEE than females. Mean RQ for males was significantly lower than females (0.79 ± 0.03 vs. 0.85 ± 0.07, p = 0.030). When adjusted for kg of body weight, fat, and FFM, the difference in RMR and TEE between males and females was not significant. RMR was positively correlated with FFM (r = 0.673, p = 0.001) and PAL (r = 0.805, p < 0.001). TEE was positively correlated with FFM (r = 0.716, p < 0.001), activity energy expenditure (calculated from TEE minus RMR) (r = 0.755, p < 0.001) and right-hand grip strength (r = 0.614, p = 0.004). There were no significant correlations between RMR and TEE with % fat, BMI, left hand grip strength, ICECAP-O, and AQoL-8D.

The performance of the Bland-Altman analyses for RMR is presented in Table 3. All equations overestimated measured RMR. The Mifflin equation demonstrated the closest agreement (least bias), it overestimated RMR by 293 ± 713 kJ/day, met the assumptions of the one-sample t test (p = 0.074), and showed no proportional bias (p = 0.269) (Table 3; Fig. 2). The Ikeda and Livingston equations overestimated RMR by 454 ± 646 kJ/day and 339 ± 682 kJ/day, respectively; did not meet the assumptions of the one-sample t test; and proportional bias was seen in the Ikeda equation. The Schofield equation demonstrated the greatest bias (overestimation by 865 ± 662 kJ/day) among the four equations, violated the assumption of the one-sample t test, and demonstrated proportional bias.

The performance of the Bland-Altman analysis for TEE is presented in Table 3. The Mifflin, Ikeda, and Livingston x PAL equations underestimated measured TEE, met the one-sample t test, and showed no proportional bias. The Ikeda equation demonstrated the closest agreement (least bias) of 37 ± 1,103 kJ/day, followed by the Livingston and Mifflin equations by 251 ± 1,108 kJ/day and 354 ± 1,140 kJ/day, respectively (Table 3; Fig. 3 a–c). The Schofield equation demonstrated the greatest bias (overestimation by 641 ± 1,066 kJ/day), violated the assumption of the one-sample t test, and demonstrated proportional bias.

The mean ICECAP-O group score was 0.87 ± 0.07; the mean score for males was 0.87 ± 0.09, and the mean score for females was 0.88 ± 0.62. The range of scores was from 0.720 to 1.00 (full capability). The mean group ICECAP-O score of 0.87 ± 0.07 was greater than that observed from a group of older Australian adults aged ≥65 years (0.83 ± 0.17), albeit this latter score was derived from a larger sample of n = 789 [25]. The mean AQoL-8D group score was 0.81 ± 0.12; the mean score for males was 0.81 ± 0.01; and the mean score for females was 0.82 ± 0.13. The range of scores was from 0.56 to 1.00 (full health).

The objective of this study was to characterize the TEE measured using DLW of older Australian adults aged ≥80 years and to compare it with TEE derived from established prediction equations. In support of our hypothesis, the Ikea, Livingston, and Mifflin equations provided less bias in the estimation of total energy requirements in comparison to the Schofield equation. Of the three equations, the Ikeda equation demonstrated the smallest bias.

Equations for the estimation of RMR, together with a PAL, provide an estimation of TEE. In this study, the comparison of a set of four equations (Ikea, Livingston, Mifflin, and Schofield) to estimate RMR showed that the Schofield equation overestimated approximately up to double the mean difference of that derived from the three other equations. This greater overestimation compared with the other three equations is consistent with our previous research comparing measured with predicted RMR in adults ≥65 years (n = 988) and also in the subgroups 65–79 years (n = 733) and ≥80 years (n = 255) [15]. The overestimation of RMR using the Schofield equation has not only been demonstrated in older adults but also in a study of n = 128 young Australian adults [26]. This observation is not surprising, as the population from which the Schofield equation was derived comprised over 50% males and predominately Italian men, with a preponderance of physically active occupations such as miners and military service [6]. The Italian group was found to have a higher basal metabolic rate per kg of body weight than other Caucasian groups in the database [27, 28]. The dominance of physically active occupations along with a greater FFM to fat mass ratio may explain the higher basal metabolic rate [29, 30] and therefore why the equations may tend to overestimate in present day populations including older adults.

The mean PAL derived from our sample of older adults was 1.94 for men and 1.85 for women. These PAL values are greater than that derived from our larger dataset of adults ≥65 years for both men (PAL = 1.69, n = 320) and women (PAL = 1.66, n = 668) [15]. The higher PAL obtained in this group of older adults may be reflective of the greater than average ICECAP-O scores, suggesting this small sample of adults was above average in terms of capability and mobility and undertook activity regularly. Volunteer bias also cannot be discounted as older adult volunteers who partake in studies tend to be generally high-functioning individuals, more likely to be women, and less likely to rely on health and human services [31]. The average PAL derived from our larger dataset [15] was thus instead used to calculate TEE in this study, as it was derived from a greater sample and therefore should offer more representation of the population. When this PAL was applied to the RMR equations to estimate TEE, the Schofield equation exhibited the greatest bias (overestimation of 641 ± 1,066 kJ/day) compared with measured TEE using DLW. The equation with the least bias (underestimation) was the Ikeda (37 ± 1,103 kJ/day) followed by the Livingston (251 ± 1,108 kJ/day) and the Mifflin equation (354±1,140 kJ/day). Again, the overestimation of energy by the Schofield equation is not surprising given the population from which it was derived.

The common variables used in current prediction equations for the determination of energy expenditure are age, height, sex, and weight. FFM, however, is the major determinant of RMR, accounting for up to 60–70% of TEE [30]. It has been shown from a recent analyses of a large dataset of participants (aged 8 days to 85 years) with TEE measured using DLW (n = 6,421 subjects) and RMR measured using indirect calorimetry (n = 2,008) that age-related declines in TEE and RMR are attributable to a reduction in FFM [32]; thus, body composition measures should probably be considered integral in future equation development. Indeed, this study demonstrated a significantly positive correlation between FFM and RMR (r = 0.673, p = 0.001) as well as FFM and TEE (r = 0.716, p < 0.001). An example of a practical bedside measure of body composition that could be used is the bioelectrical impedance, which is a quick, economical, and non-invasive method of measuring fat and FFM. The bioelectrical impedance method has been validated against the reference deuterium dilution method in older adults [33]. Additionally, hand grip strength which is a common measure of functional status was positively correlated with TEE (right-hand grip strength; r = 0.614, p = 0.004). This relationship has also been observed in a group of older adults by Lopes De Pontes et al. [34], who not only found a positive association between TEE measured by DLW and greater hand grip strength but also other functional measures such as greater gait speed and the 6 min walk test. This demonstrates that functional measures may be important variables to consider in the estimation of TEE in the older adult population.

This study has derived the first set of TEE measured using the reference method of DLW, along with applying reference methods in the measurement of REE using indirect calorimetry and body composition using dual-energy X-ray absorptiometry in Australian adults ≥80 years old. Currently, there is a lack of contemporary datasets using reference methods that characterize this understudied but increasing age group within the population. These unique data are thus an important contribution to the field of nutrition and metabolism in older adults, and further research is needed to better characterize this group.

This study also demonstrates for the first time the feasibility of the execution of the complex DLW protocol in a sample of older Australian adults. The use of the DLW protocol in older, compared with younger adults, can be complicated by urinary retention [35]; thus, the collection of saliva samples in this study was utilized as a quality control measure to confirm isotopic equilibration. The validity of a study’s result can only be as good as the protocol on which its execution is based; therefore, expert input was important in the development of such an intricate protocol. To facilitate global data sharing, our unique dataset will be integrated into the recently established International Atomic Energy Agency (IAEA) DLW database [36]. Our previous research has indicated the challenges of retrieving DLW for reasons such as changes in technology and inability to contact authors [11]; thus, the deposition of such valuable data into a central depository will facilitate future research.

One limitation of this study is the small sample size. Additionally, the results from this study reflect a group of healthy ambulatory adults (67% female); therefore, the translation of findings to other groups such as those with various chronic diseases and hospitalized adults would require further research and also consideration of different DLW sampling techniques [37]. Strengths of this study were that reference methods for the measurement of RMR and TEE were used. The DLW samples in this study were also analysed using a dual inlet isotope ratio mass spectrometer which has superior precision to a continuous flow isotope ratio mass spectrometer [10].

Future research should focus on the verification of these data in a larger group of older adults, which would be capable of supporting the development of contemporary prediction equations for TEE. The inclusion of body composition, practical strength, and functional measures into future equations should also be considered. Changes to lifestyle, diet, environment, work practices, technology, and increased life expectancy over the last decade mean contemporary energy data are necessary to reflect current trends in energy expenditure. This will assist with the provision of nutrition advice, whereby knowledge of both energy intake and expenditure is required for the modulation of body weight, whether the goal may be for weight gain, maintenance, or weight loss.

This study shows that in a group of healthy, ambulatory older Australian adults ≥80 years, the Ikeda, Mifflin, and Livingston equations, when utilized together with a standard PAL, provide closer estimates of TEE than the Schofield equation, with the Ikeda equation demonstrating the least bias. This research also supports our previous findings of the international dataset that the Schofield equation is not the most accurate for use within the older adult population [15]. As such, population-level energy predictions such as those used in the development of NHMRC Nutrient Reference Values for Australia and New Zealand [7] should consider the utilization of equations that more accurately predict age-specific requirements.

Authors acknowledge the contribution of Dr Kim Baulbys for training and analyses in the dual inlet isotope ratio mass spectrometer. Authors would also like to thank Professor Dale Schoeller for advice on DLW study day protocol; Professor Peter Davies for imparting knowledge on DLW analyses; Timothy Shriver and Natalie Racine for imparting knowledge on mass spectrometry and doubly labelled water practicalities in the laboratory, and Josh Chang for assisting with participant data collection.

All participants provided written informed consent for participation in this study. This study was conducted according to the guidelines laid down in the Declaration of Helsinki, and all procedures involving human participants were approved by the Eastern Health Human Research Ethics Committee (project code E13-2017) and the Monash University Human Research Ethics Committee (project code 12115).

The authors have no conflicts of interest to declare.

This project was funded by an Eastern Health Foundation Research Grant (EHFRG2017_054).

KN: conceptualization, formal analysis, investigation, project administration, and writing – original draft. JP: conceptualization, investigation, writing – review and editing, and supervision. HT: conceptualization, writing – review and editing, and supervision.

All data generated or analysed during this study are included in this article. Further enquiries can be directed to the corresponding author.