Doubly labelled water assessment of energy expenditure

The doubly labelled water method for the measurement of energy expenditure is based on the difference between the apparent turnover rates of the hydrogen and oxygen of body water as a function of carbon dioxide production⁽ Reference Lifson and McClintock ^{8 )}. In practice, subjects get a measured amount of doubly labelled water (²H₂ ¹⁸O) to increase background enrichment of body water for ¹⁸O of 2000 ppm with about 200 ppm and background enrichment of body water for ²H of 150 ppm with 120 ppm. Subsequently, the difference between the apparent turnover rates of the oxygen and hydrogen of body water is assessed from blood, saliva or urine samples, collected at the start and end of the observation interval of 1–3 weeks. Samples are analysed for ¹⁸O and ²H with isotope ratio MS⁽ Reference Speakman ^{9 )}. The doubly labelled water method is the indicated method to measure TDEE in any environment, especially with regard to AEE, without interference with the behaviour of the subject. We adopted a standard protocol⁽ Reference Westerterp ^{7 ,} Reference Westerterp, Wouters and Van Marken Lichtenbelt ^{10 )} with adaptations for uncommon situations as described later, including changing baseline tracer levels, high-intensity endurance exercise and extreme diets with a large contribution of sport drinks.

Exercise studies sometimes induce changes in background abundances of ¹⁸O and ²H, by changes in diet pattern or environment. To account for differences in background levels, control subjects can be observed, participating in the same activity and not receiving any tracer⁽ Reference Bhutani, Racine and Shriver ^{11 )}. Thus, we observed background isotope levels of 1986·9 (sd 0·9) and 142·4 (sd 0·6) ppm for ¹⁸O and ²H, respectively, in subjects at 4 d after reaching the summit of a 6·542 m mountain in Bolivia. The isotope levels showed a further decrease during a 3-week stay on the summit and were well below the initial values of 1999·0 (sd 1·1) and 151·0 (sd 0·8) ppm while at home in Europe⁽ Reference Westerterp, Kayser and Wouters ^{12 )}. Similarly, we observed a mean difference of 15 ppm for ¹⁸O and 10 ppm for ²H between background levels in subjects drinking lowland water and on Mt. Everest, where all water for consumption was produced by melting snow⁽ Reference Westerterp, Kayser and Brouns ^{2 )}. The difference in background abundances of ¹⁸O and ²H between geographical locations reflects isotope fractionation during evaporation, resulting in higher enrichment of water sources near the equator and lower enrichments near the poles and at high altitude. A recent suggestion to measure simultaneously ¹⁷O as a reflection of background fluctuations in ¹⁸O and ²H requires further study⁽ Reference Berman, Melanson and Swibas ^{13 )}. Water enriched in ¹⁸O, as used to increase background enrichment of body water, is also enriched in ¹⁷O. Until then, the optimal study design for the measurement of energy expenditure in subjects changing diet pattern or environment remains the inclusion of control subjects participating in the same activity.

Exercise studies sometimes require administration of multiple isotope doses. Examples are studies with a baseline measurement followed by an exercise intervention and studies exceeding the optimal observation interval, where final isotope enrichments get too close to background levels for precise measurement. The optimal observation interval for the measurement of energy expenditure with doubly labelled water is one to three times the biological half-life of the isotopes⁽ Reference Lifson and McClintock ^{14 )}. The biological half-life of ¹⁸O and ²H ranges from 3 d in extremely active subjects to about 10 d in very sedentary subjects⁽ Reference Westerterp ^{7 )}. Thus, a typical observation interval is 1 week in endurance athletes, 2 weeks in normally active subjects and 3 weeks in the very sedentary. Multiple dosing rather than increasing the isotope doses can accomplish extension of the observation time beyond the duration mentioned. Larger doses will not significantly increase the observation interval for precise measurements⁽ Reference Schoeller ^{15 )}. For multiple dosing, a second dose can be applied before the tracers from the first dose are completely eliminated from the body⁽ Reference Trabulsi, Troiano and Subar ^{16 )}. One of the first examples of a multiple dose study was the measurement of energy expenditure over 22 d in the Tour de France⁽ Reference Westerterp, Saris and Van Es ^{3 )}. The total dose of 260 g/subject was split up into three portions of 105, 80 and 75 g, given at days 0, 7 and 15 of the 22-d observation interval. Similarly, subjects got three weekly doses to measure energy expenditure the week before departure and throughout a 14-d cycling expedition covering 2706 km from Copenhagen to Nordkapp⁽ Reference Rosenkilde, Morville and Andersen ^{17 )}. In a study in a hypobaric chamber, to simulate the ascent of Mt. Everest, an initial dose was administered on day 1 and a second dose on day 16 of the 31-d observation interval⁽ Reference Westerterp, Meijer and Rubbens ^{18 )}.

The doubly labelled water method allows the assessment of carbon dioxide production as a measure for energy expenditure. Conversion of carbon dioxide production to energy expenditure requires information on substrate utilisation. The energy equivalent of carbon dioxide ranges between 21·1 kJ/l for carbohydrate oxidation and 27·8 kJ/l for fat oxidation⁽ Reference Brouwer ^{19 )}. The energy equivalent is 23·5 kJ/l for subjects consuming a typical Western diet with 55 % carbohydrate, 15 % protein and 30 % fat⁽ Reference Black, Prentice and Coward ^{20 )}. Exercising subjects often consume diets with a large contribution of carbohydrate-rich sport drinks. Sjödin et al. ⁽ Reference Sjödin, Andersson and Högberg ^{21 )} observed a 16 % contribution of sport drinks to total energy intake, resulting in a relatively higher carbohydrate intake, and a corresponding lower figure for the energy equivalent of carbon dioxide of 23·0 kJ/l. In the Tour de France, the contribution of sport drinks was twice as high, resulting in an energy equivalent of 22·5 kJ/l carbon dioxide. Even then, using the real energy equivalent of 22·5 kJ/l instead of the figure of 23·5 kJ/l for a typical Western diet results in an adjustment <5 % of TDEE. Thus, in practice, a change in substrate utilisation does not have a large effect on energy expenditure as calculated from doubly labelled water-assessed carbon dioxide production.

Limits of physical activity level and energy balance

The first review of human energy expenditure measured by the doubly labelled method assessed PAL values. The authors suggested a maximum value of 2·0–2·4 for subjects with strenuous work or high leisure activity⁽ Reference Black, Coward and Cole ^{22 )}. Subsequently, FAO/WHO/United Nations University^{( 5 )} classified the PAL of a subject in three categories: 1·40–1·69 for a sedentary or light active lifestyle, 1·70–1·99 for active or moderately active lifestyles and 2·00–2·40 for a vigorously active lifestyle. Indeed, in the general population, PAL ranges between 1·1 and 2·5 as shown by the frequency distribution of a large sample of measurements performed in Maastricht between the 1980s and the present day (Fig. 1). The sample includes subjects aged at least 18 years, not involved in interventions, and all observed over 2 weeks under free-living conditions.

Fig. 1.

Frequency distribution of the physical activity level of a large sample of adults, observed over 2 weeks under free-living conditions, living in Maastricht and surroundings. Data from Speakman and Westerterp⁽ Reference Speakman and Westerterp ^{24 )} and unpublished studies. Grey bars women (n 416) and open bars men (n 550).

Children spend about 20 % of TDEE in AEE at age 1, when they start to move around independently⁽ Reference Butte, Ekelund and Westerterp ^{23 )}, resulting in an average PAL value of 1·4. The PAL value reaches adult values about age 10. The PAL tends to remain stable in adults until age 50. From then onwards, there is a gradual decline⁽ Reference Speakman and Westerterp ^{24 )}. Combining doubly labelled water-measured PAL data for all ages results in a general model as presented in Fig. 2, showing that PAL reaches a maximum at reproductive age, with slightly higher values for men than for women. This may have an evolutionary significance⁽ Reference Heitmann, Westerterp and Loos ^{25 )}.

Fig. 2.

General model for physical activity level in relation to age for women (continuous line) and men (broken line), derived by combining data from children^{( 5 )} and adults⁽ Reference Speakman and Westerterp ^{24 )}.

In a search of the literature, ten well-controlled studies were identified that utilised the measurement of PAL to assess the effect of exercise before and during an exercise intervention⁽ Reference Blaak, Westerterp and Bar-Or ^{26 –} Reference Hunter, Wetzstein and Fields ^{34 )}. In these studies, the training mode was aerobic training or resistance training, two to five 40–90 min sessions per week, for 4–40 weeks. In all studies, subjects were sedentary or had a light active lifestyle before the intervention started, as shown by a mean PAL value of 1·40–1·69 (Fig. 3). In only five of the ten studies, the exercise intervention induced a significant increase in PAL⁽ Reference Blaak, Westerterp and Bar-Or ^{26 –} Reference Westerterp, Meijer and Janssen ^{30 )}; in four of those studies, subjects reached values of 2·00–2·40 for a vigorously active lifestyle⁽ Reference Blaak, Westerterp and Bar-Or ^{26 –} Reference Van Etten, Westerterp and Verstappen ^{28 ,} Reference Westerterp, Meijer and Janssen ^{30 )}. In the five studies not showing an exercise-induced increase in PAL, training was combined with an energy-restricted diet⁽ Reference Kempen, Saris and Westerterp ^{31 )}, or training was preceded by weight loss⁽ Reference Hunter, Fisher and Neumeier ^{29 )}, or subjects were aged above 50 years⁽ Reference Meijer, Goris and Wouters ^{32 –} Reference Hunter, Wetzstein and Fields ^{34 )}. PAL gradually decreases with increasing age⁽ Reference Speakman and Westerterp ^{24 )}. Thus, exercise training can increase PAL in younger subjects with ad libitum access to food. Surprisingly, none of the exercise interventions employed in the ten studies identified⁽ Reference Blaak, Westerterp and Bar-Or ^{26 –} Reference Hunter, Wetzstein and Fields ^{34 )} increased PAL above the maximum value of 2·0–2·4 for a sustainable lifestyle, i.e. maintenance of energy balance and body weight.

Fig. 3.

Physical activity level before (open bar) and at the end of an exercise training programme (grey bar) for ten studies (A⁽ Reference Blaak, Westerterp and Bar-Or ^{26 )}, B⁽ Reference Bingham, Goldberg and Coward ^{27 )}, C⁽ Reference Van Etten, Westerterp and Verstappen ^{28 )}, D⁽ Reference Hunter, Fisher and Neumeier ^{29 )}, E⁽ Reference Hunter, Fisher and Neumeier ^{29 )}, F⁽ Reference Westerterp, Meijer and Janssen ^{30 )}, G⁽ Reference Kempen, Saris and Westerterp ^{31 )}, H⁽ Reference Meijer, Goris and Wouters ^{32 )}, I⁽ Reference Goran and Poehlman ^{33 )} and J⁽ Reference Hunter, Wetzstein and Fields ^{34 )}), and presented in sequence of mean age of the participants (years) in brackets. The lower broken line denotes the maximum value of the physical activity level of 1·69 for a light active lifestyle, the upper broken line denotes the minimum value of the physical activity level of 2·0 for a vigorously active lifestyle^{( 5 )}. *P < 0·05; **P < 0·01 for difference with before training programme.

Only one of the ten exercise training studies (study F as presented in Fig. 3) reported an exercise-induced change in energy balance, as indicated by a change in body weight and body composition⁽ Reference Westerterp, Meijer and Janssen ^{30 )}. It included the longest and possibly most demanding exercise training. Sixteen women and sixteen men, aged 21–41 years, BMI 19·4–26·4 kg/m², not participating in any sport before the start of the experiment, prepared to run a half-marathon competition after 40 weeks. The training consisted of four sessions per week, increasing running time to 30–90 min per training session, finally running 50 km/week. During the study, nine subjects withdrew being unable to keep up with the training programme. They were all in the higher fat mass range with BMI over 24 kg/m². The remaining subjects showed a mean weight loss of 1·0 (sd 1·8) kg (P < 0·01) over the 40-week period, with more pronounced losses of fat mass (3·7 (sd 2·1) kg, P < 0·001) and gains of fat-free mass (2·6 (sd 1·5) kg, P < 0·001). Assuming an energy equivalent of 38 MJ/kg fat mass and 6·0 MJ/kg fat-free mass⁽ Reference Westerterp, Donkers and Fredrix ^{35 )}, the energy content of the body decreased by 125 MJ/d over 40 weeks, representing an energy mobilisation from body reserves of 0·4 MJ/d. Thus, the main part of the exercise-induced increase in TDEE of 2·3 (sd 1·0) MJ/d must have been covered by an increase in energy intake. Indeed, in the majority of exercise intervention studies, an exercise-induced increase in energy expenditure is compensated by a concomitant increase in energy intake⁽ Reference Thomas, Bouchard and Church ^{36 )}.

There are situations where exercise induces higher PAL, as defined for a vigorously active lifestyle⁽ Reference Piersma ^{37 )}. Cooper et al. ⁽ Reference Cooper, Nguyen and Ruby ^{38 )} reviewed studies with doubly labelled water-assessed PAL measurements in extremely active subjects, including athletes and military personnel. The highest PAL values, in excess of 5·0, were observed in the world's most demanding cycle race, the Tour de France⁽ Reference Westerterp, Saris and Van Es ^{3 )}, and in subjects attempting to reach the North Pole on foot⁽ Reference Stroud, Coward and Sawyer ^{39 )}. The subjects in the Arctic trek walked from the northernmost point of Siberia, pulling each a 135 kg sledge with food, fuel and equipment, in the direction of the North Pole. After 30 d and 450 km, the sledges were abandoned and supplies were switched to rucksacks. After a further 360 km, the expedition was abandoned on the 48th day. They lost more than 1 kg bodyweight per week. In the Tour de France, subjects managed to maintain energy balance as shown by minimal changes in body weight and body composition. Endurance athletes, like Tour de France participants, consume energy-dense carbohydrate-rich foods and liquid formulas in order to compete at the top level. However, in most studies with PAL values higher than 2·00–2·40, subjects are in negative energy balance such as soldiers during field training⁽ Reference Westerterp ^{40 )} and exercise interventions in older subjects⁽ Reference Rosenkilde, Morville and Andersen ^{17 ,} Reference Ainsli, Campbell and Frayn ^{41 )}.

Exercise-induced compensatory changes in energy expenditure

An exercise-induced increase in energy expenditure can be compensated by a reduction in REE, a reduction in non-exercise physical activity and an increase in exercise efficiency. Examples of each of the three options mentioned are described later.

One potential effect of exercise training on energy expenditure is an increase in REE due to increases in fat-free mass⁽ Reference Speakman and Selman ^{42 )}. However, in a study where sedentary subjects were trained to run a half-marathon⁽ Reference Westerterp, Meijer and Janssen ^{30 )}, REE as a function of fat-free mass was lower after 40 weeks training than before training (Fig. 4). REE decreased by 0·3 (sd 0·5) MJ/d, and the decrease was related to a decrease in body mass, possibly as a defence mechanism by the body to maintain body mass⁽ Reference Westerterp, Meijer and Schoffelen ^{43 )}. In a weight loss study maximising fat loss and minimising fat-free mass loss with an intensive 30-week exercise programme, REE decreased by 3·3 (sd 2·0) MJ/d⁽ Reference Johannsen, Knuth and Huizinga ^{44 )}. The decrease in REE was 2 MJ/d greater than accounted for by the change of body weight and composition. Thus, an exercise-induced change in fat-free mass is not necessarily reflected in a proportional change in REE. An exception is elite endurance athletes. Top international-level cross-country skiers showed a 16 % higher REE than sex-matched and fat-free mass-matched sedentary control subjects⁽ Reference Sjödin, Forslund and Westerterp ^{45 )}. Here, a possible explanation is positive energy balance, where a high-energy flux and unchanged eating habits lead to overeating on resting days when REE is measured.

Fig. 4.

Resting energy expenditure as a function of fat-free mass in sedentary subjects before (open dots) and after 40 weeks training (closed dots) to run a half-marathon, with the linear regression lines (data from Westerterp et al. ⁽ Reference Westerterp, Meijer and Janssen ^{30 )}).

An exercise intervention may reduce non-exercise physical activity, reducing the effect of exercise on TDEE. Reviews on the effect of exercise training on non-exercise physical activity concluded that, in healthy adults, no statistically or clinically significant change in non-exercise physical activity occurs during exercise training⁽ Reference Washburn, Lambourne and Szabo ^{46 ,} Reference Fedewa, Hathaway and Williams ^{47 )}. Indeed, two studies combining doubly labelled water assessment of AEE with accelerometer assessment of activity pattern did not detect a change in non-exercise activity⁽ Reference Van Etten, Westerterp and Verstappen ^{28 ,} Reference Meijer, Janssen and Westerterp ^{48 )}. However, in studies not showing an exercise-induced increase in PAL, where training was combined with energy restriction⁽ Reference Kempen, Saris and Westerterp ^{31 )} or subjects were aged above 50 years⁽ Reference Meijer, Goris and Wouters ^{32 –} Reference Hunter, Wetzstein and Fields ^{34 )} (Fig. 3), exercise activity probably was compensated by a reduction in non-exercise physical activity⁽ Reference Melanson, Keadle and Donnelly ^{49 ,} Reference Melanson ^{50 )}. Kempen et al. ⁽ Reference Kempen, Saris and Westerterp ^{31 )} observed no difference in PAL between women on an energy-restricted diet with and without exercise. In older subjects, an exercise training programme had no effect on total daily physical activity. Training activity was compensated for by a decrease in non-training physical activity⁽ Reference Meijer, Westerterp and Verstappen ^{51 )}. Apparently, TDEE is constrained by energy restriction and at higher age.

Finally, an exercise-induced increase in energy expenditure can be compensated by an increase in exercise efficiency. Plotting TDEE as a function of accelerometer-assessed physical activity in a comparative analysis between subjects, TDEE increases with physical activity at low-activity levels but plateaus at higher activity levels⁽ Reference Pontzer, Durazo-Arvizu and Dugas ^{52 )}. In a longitudinal study, training sedentary subjects in preparation to run a half-marathon competition after 40 weeks, we observed an initial increase in PAL but no further increase between 8 and 40 weeks training, while training distance was doubled over the same interval (Fig. 5). The initial increase in PAL was nearly twice as high as the predicted cost of the training and could not be attributed to a change in non-training activity as recorded with an accelerometer⁽ Reference Meijer, Janssen and Westerterp ^{48 )}. After the initial increase in PAL, the novice runners increased the efficiency of the exercise as a result of the training. Training increased exercise economy, and thus reduces an exercise-induced increase in energy expenditure, especially in untrained subjects⁽ Reference Van Etten, Westerterp and Verstappen ^{28 ,} Reference Valenti, Bonomi and Westerterp ^{53 )}.

Fig. 5.

Training distance and physical activity level in subjects before and during a 40-week preparation to run a half-marathon (data from Westerterp et al. ⁽ Reference Westerterp, Meijer and Janssen ^{30 )}).