The history and theory of the doubly labeled water technique1,2

John R Speakman

ABSTRACT Scientists have been measuring energy jars over water and charted the change in volume of the air inside
expenditure by using gas exchange for the past 200 y. This the jar from the change in the movement of the water. He observed
technique is based on earlier work in the 1660s. Gas exchange in that mice died when they had consumed about one-fourteenth of
respirometers provides accurate and repeatable measures of the air in the bell jar. Mayrow accordingly established the idea that
resting metabolic rate. However, it is impossible to duplicate in the air consists of different parts, only some of which are usable
a respirometry chamber the diversity of human behaviors that for the process of respiration. A significant discovery, for which he
influence energy expenditure. The doubly labeled water is not generally credited, is the invention of a chamber that
technique is an isotope-based method that measures the energy allowed the quantification of the consumed portion; this was the
expenditure of unencumbered subjects from the divergence in first respirometer. The significance of Mayrow’s observations was
enrichments of 2 isotopic labels in body water—1 of hydrogen not widely recognized, mostly because of the prevailing alchemic
and 1 of oxygen. The method was invented in the 1950s and framework within which they were interpreted.
applied to small animals only until the early 1980s, mostly A century after the innovative work of Boyle and Mayrow, the
because of the expense. Since 1982, when the first study in French chemists Lavoisier and Seguin started systematic investi-
humans was published, its use has expanded enormously. gations of respiration as a process analogous to combustion. In
Although there is some debate over the precise calculation the intervening century, important advances had been made
protocols that should be used, the differences between toward the understanding of the chemistry of gases: in 1757
alternative calculations result in relatively minor effects on total Joseph Black had discovered carbon dioxide (“fixed air”), and in
energy expenditure estimates (<6%). Validation studies show 1774 Joseph Priestly had discovered oxygen. The procedures
that for groups of subjects the method works well, but that used by Lavoisier and Seguin mimicked closely those developed
precision is still relatively poor (8–9%) and consequently the by Mayrow, the key difference being the framework within which
method is not yet sufficiently refined to provide estimates of the observations were interpreted. The methods involved confin-
individual energy expenditures. Am J Clin Nutr 1998; ing animals or humans in chambers to quantify their consumption
68(suppl):932S–8S. of oxygen and production of carbon dioxide. The chambers they
used were relatively unsophisticated. Animals were still confined
KEY WORDS Doubly labeled water, gas exchange, in bell jars, and human subjects were confined in varnished silk
respirometry, energy expenditure, carbon dioxide production, bags secured around the nose and lips with pitch.
indirect calorimetry, dilution spaces Lavoisier and Seguin made several important discoveries
about oxygen consumption. First, they found that larger persons
consume more oxygen than smaller persons. Second, they found
INTRODUCTION that people sitting quietly at rest consumed less oxygen than
The knowledge that respiration and ventilation are essential those standing up or moving about. Finally, they discovered that
functions of life dates back to at least biblical times. In the Old after a meal, oxygen consumption was elevated even in a person
Testament, Book of Psalms, for example, it is stated with respect sitting at rest. Perhaps most importantly, the work of Lavoisier
to animals, “When thou takest away their breath they die” (Psalm and Seguin established the methodology of indirect calorimetry
104). Although it was clear to ancient humans that animals had to that has remained the benchmark method for the quantification
breathe to live, the exact physiologic function of breathing was of animal and human energy expenditure to this day.
obscure. For many years it was widely believed that inspired air Since the end of the 18th century, the chambers in which ani-
served to cool the heat generated by a fire burning in the heart (1). mals and humans are confined have become increasingly sophis-
Scientific study of animal respiration only commenced in the ticated. Moreover, sealed systems have been replaced with open
1600s. In 1660 Robert Boyle observed that mice that had been flow systems linked to advanced gas analysis equipment, so
sealed into bell jars expired at the same time that a burning flame
became extinguished. Boyle thus established 2 important princi- 1
From the Department of Zoology, University of Aberdeen, Scotland,
ples: the equivalence of fire and life as combustion processes and United Kingdom.
the requirement of air to support these processes. 2
Reprints not available. Address correspondence to JR Speakman, Depart-
Less well known, but probably of greater significance, was the ment of Zoology, University of Aberdeen, Aberdeen AB24 2TZ, Scotland,
work of John Mayrow. In 1668 Mayrow placed mice in sealed United Kingdom. E-mail: J.Speakman@abdn.ac.uk.

932S Am J Clin Nutr 1998;68(suppl):932S–8S. Printed in USA. © 1998 American Society for Clinical Nutrition

Downloaded from https://academic.oup.com/ajcn/article-abstract/68/4/932S/4648697

by guest
on 23 March 2018
 DOUBLY LABELED WATER TECHNIQUE 933S

subjects no longer have to die to provide a measurement. Never- involved comparisons of estimates of carbon dioxide production
theless, it is clear that no matter how intricately designed such by standard respirometry and the DLW method in 15 laboratory
chambers become they will never be able to reproduce the com- mice, was published in 1955 (8). The technique provided an esti-
plexity of activities in which people are engaged as they go about mate that differed from the standard method by <2–3% on aver-
their routine lives. Yet, our understanding of many of the nutri- age, depending on the assumptions made about the behavior of
tional problems facing humans depends on reliable and accurate isotopes both in the body and during elimination from it. It was
measurements of their energy demands as they go about their immediately clear that the method worked and that it had a
everyday business. The inadequacy of traditional calorimetry tremendous advantage over standard respirometry measure-
apparatus has been recognized for a long time, and there have ments. As Lifson et al (8) put it in the introduction to their 1955
been many attempts to develop methods, such as heart-rate mon- paper, “a measurement is made possible merely by taking 2
itoring, that enable the energy demands associated with free-liv- blood samples to reconstruct the isotope elimination curves.”
ing activities to be monitored (2–4). The doubly labeled water This meant the animal could perform a whole variety of natural
(DLW) technique is a method that allows the measurement of behaviors between the taking of the 2 samples unconstrained by
energy demands of free-living subjects. The success of this confinement in a respirometry chamber. Yet it would be another
method prompted Prentice (5) to remark that its development 9 y before this potential advantage of the method was actually
was as significant an event in the history of animal and human exploited to study metabolism of a wild animal (9). The method
nutrition as the work of Lavoiser and Seguin had been. was used to measure the energy costs of flying in homing
pigeons (Columba livia) (9) in 1964. The author was also based
at Minnesota and associated with Lifson’s group. Indeed, Lif-
HISTORY OF THE DOUBLY LABELED WATER son’s group was the only group in the world to use the method
TECHNIQUE between its original development and 1970, and during that time
The discovery of rare heavy isotopes of oxygen and hydrogen they produced only 9 papers using it, of which 6 were validations
in the late 1920s and the 1930s transformed many areas of biol- (all on small rodents), 2 were applications, and 1 was a complete
ogy. This is because these substances were chemically and theoretical analysis of the method and its underlying assump-
physically almost identical to the common light forms of the tions (10). This theoretical paper has formed the basis of all the
elements. The rare isotopes were thus ideal as tracers for the subsequent theoretical discussions of the method. Few of the
behavior of oxygen and hydrogen, and therefore water, in the advances made since the mid-1960s are not addressed in this
body. Until their discovery, scientists relied on water-soluble paper. By the mid-1990s, 30 y after publication, records from the
dyes to investigate water dynamics. For example, known Science Citation Index (Institute for Scientific Information, Uni-
amounts of the dye Evans blue were injected into subjects to see versity of Aukland, New Zealand) indicate that this paper was
how diluted it became and thus evaluate their plasma blood vol- cited at least 50 times/y.
ume (eg, 6). In the early years after the discovery of isotopes, In the 1970s, several other groups started to use the method.
the major problems with their use related to their expense, lack Studies by these groups exploited the ability of the method to
of availability, and, for stable forms, the lack of equipment measure the free-living energy demands of small animals. The
capable of accurately measuring their abundance. These prob- first measures on a free-ranging mammal (pocket mouse, Perog-
lems were solved principally as a byproduct of the need to nathus formosu) were made in 1970 (11) and the first on a free-
understand more about the behavior of isotopes and to artifi- living reptile (lizard, Sauromalus obesus) in 1972 (12). All the
cially enrich them during the development of atomic weapons at animals on which the method was used between 1955 and 1975
the end of the Second World War. weighed < 1 kg. There were no attempts to apply the method to
By 1949, Lifson et al (7), at the University of Minnesota, had humans to evaluate free-living energy expenditures for 2 rea-
performed several experiments in mice that involved injecting sons. The first reason is that it was prohibitively expensive. In
them with stable isotopes of oxygen in water or forcing them to 1973, for example, it was estimated that to use the method on a
breathe air enriched with 18O. These experiments clearly showed single 70-kg human would cost <US$50 000 (13). For compari-
that the oxygen in body water was in complete isotopic equilibra- son, the most expensive production automobile in 1973 was the
tion with the oxygen in respiratory carbon dioxide. This equili- Rolls Royce Phantom VI that cost £17 817 (<US$30 000 at cur-
bration occurred principally because of isotope-exchange reactions rent exchange rates). The second reason was that clinical prob-
catalyzed by carbonic anhydrase during blood gas transport. The lems involving disorders of energy balance (eg, obesity) were
consequence of these exchange reactions is that a dose of oxygen less prevalent in Western populations at the time (14, 15), and
isotope introduced into the body of an animal is removed by the links between such disorders and other clinical problems were
flow rates of water, inspired oxygen, and expired carbon dioxide not fully understood. Thus, in the early 1970s we had available a
through that body. Although this consequence was not stated by method that would have cost the equivalent of 2 Rolls Royce
Lifson et al (7), its significance was clearly not lost on the automobiles per subject to implement and no immediately appar-
authors. If an isotopic label of oxygen is eliminated from the ent clinical problem of sufficient importance to which the results
body by both carbon dioxide and water, an isotopic label of could be applied.
hydrogen would be eliminated only by the water. Thus, the dif- The cost of 18O declined throughout the 1970s, and develop-
ference in the elimination of the 2 labels simultaneously intro- ments in mass spectrometer technology made it feasible to label
duced into the animal (thus DLW) would provide a measure of subjects at much lower levels and still retain precision and accu-
the carbon dioxide production and hence indirectly the energy racy of analysis. By 1980, the costs of a study of a 70-kg subject
expenditure. had declined to <US$2000. Moreover, the impetus to perform
It took 6 y to develop this simple theory into a working such studies was increasing because the prevalence of morbid
method. The seminal publication on the DLW technique, which obesity increased during the previous decade, and links between

Downloaded from https://academic.oup.com/ajcn/article-abstract/68/4/932S/4648697

by guest
on 23 March 2018
 934S SPEAKMAN

FIGURE 1. Number of studies in peer-reviewed journals (excluding

abstracts) that used the doubly labeled water technique in the years
1981–1997 (through June) from the Science Citation Index (Institute for
Scientific Information, University of Aukland, New Zealand). Since the FIGURE 2. Theoretical time course of enrichments of isotopes of
first study in humans in 1982 the use of the technique has continued to oxygen and hydrogen in body water after administration at time zero.
grow. Over an equilibration period of several hours the isotope enrichments
rise to a peak. If the amount of administered isotope is known, the peak
enrichment can be used to establish the volume of dilutant material in
obesity, heart disease, and other clinical problems were becoming the body (called the dilution space) by using the dilution principle. After
increasingly apparent (eg, 16, 17). It seems plausible that increas- equilibration, the isotopes are washed out of the body along an expo-
ing financial feasibility combined with practical applicability led nential track (which is linear when expressed as the log of the enrich-
to the first use of the method in humans. Indeed, previous reviews ment above background as shown here). The oxygen isotope leaves the
body faster than the hydrogen isotope because it is washed out of the
of the history of the method have painted this picture (eg, 5). The
body by water and carbon dioxide. A measure of carbon dioxide pro-
facts, however, do not accord with this interpretation.
duction and hence energy expenditure is made from the divergence of
The method was apparently independently discovered by acci- the enrichments between time 1 and time 2. During this interval (4–14 d
dent in the early 1980s because of an anomaly that occurred dur- in a typical study in adult humans) the subject is free to engage in usual
ing attempts to crosscheck hydrogen turnover estimates of water activities without the need to be confined in a respirometery chamber.
requirements in humans (18). When the first DLW measurement Note: the relative timings of the equilibration and elimination phases are
was made on a man in 1980, Schoeller had no notion that he was not drawn to scale and the extent of divergence is exaggerated for illus-
using a method that had been developed 25 y previously and by trative purposes.
that time was in relatively widespread use by researchers study-
ing the energy demands of small animals. The first validation
study using the DLW method in humans was published in 1982 range of human subjects of varying body masses, ages, and activ-
(19). Validation studies of the method in humans proliferated in ities; and more bizarre human endeavors, such as climbing Mount
the literature over the next 6 y. These validations included stud- Everest (27) and competitive bicycle racing for 21 d (28). A
ies of infants (20, 21) and adults (22, 23). The first application of review of the measurements made on free-living human subjects
the method in a group of free-living humans was published in was published in 1996 (29).
1985 and addressed the energy expenditure of obese and lean
subjects living in Cambridge, United Kingdom (24).
Since the mid 1980s, use of the method has expanded enor- THEORY OF THE DOUBLY LABELED WATER METHOD
mously (Figure 1) in concert with a proliferation of validation The fundamental basis of the DLW method is that oxygen
work (in both humans and animals). By the mid-1990s the tech- turnover in a body is dominated by the flow of water through the
nique was being used in <70–90 publications/y. Attention has body as well as inspired oxygen and expired carbon dioxide. The
focused on the underlying assumptions on which the method is turnover of body hydrogen, however, is dominated only by the
based, and refinements in the calculation method have been pro- flow of water through the body. Consequently, the difference
posed (eg, 25) and widely adopted. The method has been applied between the turnovers of oxygen and hydrogen provides a meas-
to an ever increasing array of wild animals of an expanding ure of the excess efflux of oxygen that is equivalent to the pro-
range of body sizes. As of June 1997, the largest species to which duction of carbon dioxide (Figure 2).
the method was applied was the polar bear (Ursus maritimus), In practice, the method involves introducing isotopes of both
weighing 250 kg (26), and the smallest was the honeybee (Apis oxygen and hydrogen into the body water to quantify the
mellifera), weighing only 83 mg (TJ Wolf, unpublished observa- turnover of the body pools of the 2 elements. The size of the
tions, 1996). In addition, the method has been applied to meas- pool in which the isotopes are distributed and the rates of their
ure the energy costs associated with several clinical conditions elimination provide direct estimates of the flow of compounds
(paraplegia, parenteral nutrition, and burns); routine activities of a carrying the 2 labels out of the body. The rates of isotope elim-

Downloaded from https://academic.oup.com/ajcn/article-abstract/68/4/932S/4648697

by guest
on 23 March 2018
 DOUBLY LABELED WATER TECHNIQUE 935S

ination are quantified from the exponential decline in isotope rCO 2 =

enrichment in the body water (Figure 2). Given the isotope [(N/2.076) · (k o 2 k d )] 2 0.0246 · 1.05 · N(k o 2 k d ) (4)
enrichment in a sample taken shortly after isotope administra-
tion (EI) and a second sample taken some time later (Ef) the Pool size models
elimination rate (k) is calculated from the extent that these 2 In the original formulation of the model, Lifson et al (8) assumed
body enrichments exceed the background isotope level (Eb). that the oxygen and hydrogen in the body were turning over in the
Generally, for any isotopic label, body water (N). However, both oxygen and hydrogen exist in other
pools in the body, and the sizes of these extra exchanging pools are
k = [Log e (E I 2 E b ) 2 Log e (E f 2 E b )]/t (1) not equal. Generally, the hydrogen pool exceeds the oxygen pool by
3–5%, which in turn exceeds the body water pool by <1%. Kinetic
In its very simplest form, therefore, the DLW method sug- analyses indicate that the true estimate of the flow of materials is the
gests that the elimination rates of oxygen (ko) and hydrogen (kd) product of the elimination rate and the size of the pool in which the
turning over in the body water (N) are related to the rate of car- isotopes are turning over. Thus, rather than multiplying the differ-
bon dioxide production (rCO2) by the formula ence in elimination rates by the body water (N), it was thought that
each elimination constant should perhaps be multiplied by its own
rCO 2 = (N/2) · (k o 2 k d ) (2) dilution space in the body, No and Nd respectively. Therefore,
Schoeller et al (25) modified equation 4 to
where the value 2 is a constant reflecting the fact that 1 mol CO2
removes 2 atoms of O2 and 1 mol H2O leaving the body removes rCO 2 = (1/2.076) · (N o · k o 2 N d · k d ) 2
only 1. 0.0246 · 1.05 · N(k o 2 k d ) (5)
Fractionation In this equation, the first term includes the elimination con-
Unfortunately, rare isotopes do not behave exactly like their stants each multiplied by the respective dilution spaces, but in
common counterparts. In particular, when compounds react or the second term they are still multiplied by the body water (N).
when they are vaporized, the lighter species tend to require less Schoeller (30) modified the equation so that in both terms the
activation energy and thus react or vaporize more readily than elimination constants are multiplied by their respective dilution
the heavier species. This discriminatory effect is termed frac- spaces, as follows:
tionation, and a constant value, called the fractionation factor,
expresses the extent of the discrimination in any particular rCO 2 = (1/2.076) · (N o · k o 2 N d · k d ) 2
process. The fractionation factor represents the ratio of the 0.0246 · 1.05 · (N o k o 2 N d k d ) (6)
enrichment of the rare isotope in the 2 phases. For example, the
fractionation factor for hydrogen in gaseous water compared However, the justification for this modification was ques-
with liquid water at 25 8C is 0.94, meaning the enrichment of the tioned recently (31) on the basis that the value 1.05 in equation
rare isotope (deuterium) in the gaseous phase is only 94% of that 5 was included to account for the difference between N(ko2kd)
in the liquid. The existence of fractionation complicates the sim- and (Noko2Ndkd), and consequently, by including 1.05 and
ple model because fractionated material leaving the body does (Noko2Ndkd), a double correction for the pool size difference is
not take with it the same amount of label as that left behind in made. However, the practical consequences of these differences
the body. The effect is further complicated by the fact there are for a typical study has been shown to be relatively trivial
2 different modes of fractionation (kinetic and equilibrium) and (< 0.5%) (31).
they discriminate against the heavy species to different extents. Lifson et al (8) evaluated the effects of using both pool sizes
The importance of fractionation was recognized by Lifson et rather than a single pool size in their validations of mice and
al (8), who generated a correction based on some simplifying concluded that it did not improve the fit of DLW estimates to
assumptions about the nature of the fractionating process. They those made by indirect calorimetry. However, this may have
assumed that the carbon dioxide leaving the body was all frac- been because the effects of the different pool sizes on the cal-
tionated and that half the water losses were also fractionated. culation were already being accounted for by the erroneous
They then used the equilibrium fractionation factors at 25 8C assumptions concerning fractionation. Nevertheless, recalcula-
measured in vitro for all the fractionating processes to derive tion of all the original validation studies from the 1950s using
the following equation: the more modern equations confirmed that, in small rodents,
the equation that includes only a single pool performs better
rCO 2 = (N/2.08) · (k o 2 k d ) 2 0.015 · k d · N (3) than the 2-pool model (31).
Coward et al (32) and Schoeller et al (25) reevaluated the
In validation, this equation performed much better than the effects of using different pool models in humans and came to the
equation that ignores the fractionating effects (Eq 2 above) (8). conclusion that using the 2-pool model provided a better fit to
However, it is clear that the corrections are based on several calorimetry data than did using the single-pool model. This dif-
dubious assumptions. Schoeller et al (25) overhauled the frac- ference between the appropriateness of the different models in
tionation correction, making more realistic assumptions for the different-sized subjects appears to reflect a real biological dif-
processes involved in humans. Specifically, they used the frac- ference between small rodents and humans.
tionation factors pertinent at 37 8C rather than 25 8C and In fact, in addition to a small extra exchangeable pool, there is
assumed that fractionated water losses would occur in propor- also a small extra hydrogen elimination route. The excess hydro-
tion to the rate of carbon dioxide production. The resultant gen space in adult humans is similar to that in adult small mam-
equation was mals (being 3–4% greater than the oxygen space in both) but the

Downloaded from https://academic.oup.com/ajcn/article-abstract/68/4/932S/4648697

by guest
on 23 March 2018
 936S SPEAKMAN

extra elimination route is not; proportionally, it is much larger in the final estimate of carbon dioxide production (36, 38, 39).
the small mammal (33). Theoretical analyses of the pool struc- I reviewed previously all the data available as of July 1996 on
ture show that the extent of the extra elimination offsets the both animals and humans for this ratio (31). This survey revealed
effect of the extra pool (8, 33, 34). Thus, in small mammals with that across 15 studies in mammals, the mean (± SD) dilution
a high extra elimination route, the single-pool model works best, space ratio was 1.041 ± 0.025 and across 9 studies in birds the
but in larger subjects the 2-pool model works best. Subsequent ratio was 1.0427 ± 0.033. Across 19 studies comprising 590
validations in animals of a range of body sizes confirmed this adult humans the weighted mean was 1.0387 (31). This value
effect and suggest that the single-pool model is probably appro- includes all data, irrespective of whether analyses were per-
priate in animals weighing ≤ 5 kg (35). formed directly on urine or on water distilled from the urine.
The effect of body size on the appropriateness of different Debate over the most appropriate population estimate is certain
pool models raises the interesting question of what might be the to evolve as more data are published. However, a consensus is
most appropriate method to calculate the metabolism of small likely to fall between the limits of 1.03 (25) and 1.0427 (36).
infants (< 3–4 kg). In fact, validation studies in infants clearly
Validation
show that the 2-pool formulation gives a close fit to the calori-
metric data even in preterm infants weighing < 2 kg (20, 21), A summary of all the validation comparisons of DLW to indi-
although calculations using the single-pool model have not been rect calorimetry in human subjects between 1982 and 1996 is
assessed. It seems likely that the 2-pool model works in small provided in Table 1. It is not possible to compare the different
humans because of a combination of 2 factors. First, small formulations of the equations directly on a matched sample of
babies do not necessarily behave physiologically like adult subjects because the data necessary to recalculate early valida-
small animals of the same body mass with respect to the magni- tions were not published in the papers. Overall (Table 2), across
tude of the subsidiary hydrogen elimination. Second, the extent 201 subjects the Schoeller et al (25) equation, with the lowest
of their extra hydrogen space is much smaller than that found in population ratio (1.03), overestimated the simultaneous indirect
adults; consequently, the difference between the single- and 2- calorimetry by 2.1%. In contrast, in 54 subjects the Speakman et
pool model is reduced. Indeed, in very small infants the ratio of al (36) equation, with the highest ratio (1.0427), underestimated
hydrogen to oxygen dilution space does not differ from unity, simultaneous indirect calorimetry by 2.2%. The equation of
thus 2-pool and single-pool models give identical results. Racette et al (38) with an intermediate ratio (1.034) overesti-
mated by on average 1.2% (n = 36). A limited matched sample
Using individual or population-based dilution spaces in the can be compared. In the same 36 individuals reviewed by Racette
2-pool model. et al (38), the Schoeller et al (25) equation overestimated indirect
An issue that has emerged over the past decade is whether to calorimetry results by 3.3% and the Speakman et al (36) equa-
use individual values of No and Nd in the 2-pool model equation tion underestimated these results by 3.1%. Differences between
or to replace these with a population-based average (25). If the equations are consequently relatively minor, but should per-
observed variation in No and Nd is biological, then it would make haps be borne in mind when relatively small differences are com-
sense to use individual values for the dilution spaces. Alterna- pared between studies that might have used different methods, or
tively, if the variation is mostly analytic it would make sense to when comparisons are made between estimates of energy expen-
use a population-based ratio of the 2 pools. Some authorities diture using DLW and other methods for determining energy
continue to use individual ratios. However, it is now widely demands, such as food intake.
accepted that the dominant source of variation in the observed It is important to recognize that the average comparisons
pool size ratio (Nd/No) is analytic. This being the case, a popula- across groups indicate only the efficacy of the method for
tion-based average would be better used in the equation. defining the energy demands of groups. Individual compar-
Schoeller et al (25), on the basis of a limited sample of animal isons of DLW to indirect calorimetry have much greater devi-
studies, suggested that the oxygen pool exceeds N by 1.01 and ations than the group means. Across studies the precision is
the hydrogen pool exceeds N by 1.04, and thus the hydrogen <10% on average (Table 1). Thus, at the extremes, some indi-
pool exceeds the oxygen pool by 1.03. Therefore, the 2-pool viduals may have DLW estimates > 20% divergent from simul-
model equation (6) can be reformulated as taneous indirect calorimetry. There is some evidence that part
of this variation can be attributed to analytic problems in cer-
rCO2 = (N/2.076) · (1.01ko 2 1.04kd) 2 tain laboratories; a multiple laboratory comparison (40)
0.0246 · 1.05 · N(1.01ko 2 1.04kd) (7) showed wide discrepancies in the isotope enrichment determi-
nations of the same samples analyzed by different groups.
where n = (No/1.01 + Nd/1.04) /2. Nevertheless, precision in the best laboratories is still rela-
The appropriateness of the population-based average of 1.03 is tively poor and the method is not yet sufficiently refined to
debated. This value was rounded from a true value of 1.034 (25), make confident estimates of individual energy requirements.
but it was based on the relatively small sample size available at Repeatability of measurements made by the method at
the time. Subsequent reevaluations using larger data sets have <6–10% confirm this (53).
resulted in different population averages, ie, 1.0427 (36), 1.027 The reasons for this lack of precision are not yet entirely clear.
(37), 1.034 (38), and 1.035 (39). The differences in these esti- Precision may improve as the technical instrumentation for mass
mates depend on the inclusiveness of the reviews. Some authors spectrometric determination of isotope abundances improves.
have concluded that analyses of deuterium directly from urine are However, in many circumstances the extents of deviations of
inaccurate and have eliminated these from their compilations of estimates made by DLW and simultaneous indirect calorimetry
data (eg, 38). The consequent differences in the formulation of already exceed the theoretical precision of the DLW measure-
the equation can result in biologically significant differences in ment, given the precisions of the component analyses (54). This

Downloaded from https://academic.oup.com/ajcn/article-abstract/68/4/932S/4648697

by guest
on 23 March 2018
 DOUBLY LABELED WATER TECHNIQUE 937S

TABLE 1
Validation studies of the doubly labeled water method in human subjects in comparison to indirect calorimetry (IC) and food intake/mass balance (FI) or
parenteral nutrition (PN)1
n nobs Mass Com Eqn Abs Prec Arith n1 n2 Min Max Reference
kg % %
4 4 78.1 FI L+M 5.0 5.6 12.1 3 1 25.8 7.1 19
2 2 — IC ?L+M 3.5 3.5 22.5 1 1 26.0 1.0 40
1 1 70.5 IC L+M — — 24.6 0 1 — — 23
5 5 72.9 IC L+M 8.5 6.8 15.9 4 1 26.5 14.1 22
4 4 — IC C — 2.0 12.0 — — — — 31
6 6 68.9 IC L+M 8.0 9.0 15.0 5 1 — — 25
14 14 — IC S — 7.0 11.0 10 4 — — 25*
14 14 — IC L+M — 8.0 14.0 11 3 — — 25*
5 5 53.2 PN S 5.7 5.9 13.3 4 1 — — 41
4 4 1.59 IC C 4.2 4.8 21.4 1 3 24.8 5.8 20
9 10 3.2 IC S — 6.2 20.9 5 5 — — 21
5 5 — IC S 4.3 3.9 11.4 3 2 — — 42
4 8 — IC S 4.8 7.0 21.0 3 5 — — 43
8 8 2.67 IC S 10.7 12.9 28.7 — — 229.3 4.8 44
33 33 — ICFI L+M 6.9 6.6 14.3 26 7 28.7 20.2 30*
33 33 — ICFI S 5.1 6.3 10.6 21 12 212.1 16.7 30*
1 1 66.9 FI S? — — 20.8 0 1 — — 45
4 4 84.6 IC S 2.5 2.5 115.3 4 0 — — 46
12 12 108.0 IC S 6.4 6.9 22.5 5 7 — — 47
27 27 79.8 FI S — 0.7 12.5 17 10 — — 48
8 8 74.1 FI S 9.7 14.4 12.2 5 3 235.0 9.0 49
9 9 71.8 IC S — 4.6 11.8 6 3 24.0 12.0 50
6 6 50.1 IC C 15.0 16.8 17.0 4 2 223.0 32.0 51
18 18 — IC L+M — — 123.1 — — — — 52
18 18 — IC S — 10.7 110.3 — — — — 36*
18 18 — IC SNG — 9.3 20.4 — — — — 36*
36 36 — IC S — 8.9 13.3 — — — — 37*
36 36 — IC SNG — 7.9 23.1 — — — — 37*
36 36 — IC R — 8.5 11.2 — — — — 37*
1
Abs, absolute mean deviation (average percentage error with all signs made positive); Arith, arithmetic percentage error including signs; C, Coward 2-
pool model using individual dilution spaces; Eqn, equation used to calculate the doubly labeled water results; L+M, Lifson and McClintock single-pool model;
Mass, mean mass of subjects in kg; Max, maximum deviation; Min, minimum deviation; n+, number of observations in which the DLW method exceeded the
reference method; nobs, number of comparisons; n +, number of observations in which the DLW method exceeded the reference method; n2, number of obser-
vations in which the DLW method was lower than the reference method; Prec, precision calculated as the SD of individual estimates used to derive Arith; R,
Racette et al, 1994 (2-pool model fixed ratio 1.034); S, Schoeller et al, 1986, equation A6 (2-pool model fixed pool ratio 1.03); SNG, Speakman et al, 1993
(2-pool fixed pool ratio 1.0432); *, review of previously published validations.

suggests that analytic considerations alone do not underpin all of REFERENCES

the observed variability. The ultimate precision of the method 1. Brown JC. A history of chemistry. 2nd ed. London: Churchill, 1920.
may then depend on individual variability in competing physio- 2. Livingstone MBE, Prentice AM, Coward WA, et al. Simultaneous
logic reactions involving the isotopic labels, dilution spaces, measurement of free-living energy expenditure by the doubly
background drift, and fractionation effects, about which we now labeled water method and heart-rate monitoring. Am J Clin Nutr
know relatively little. 1990;52:59–65.
3. Livingstone MB, Coward WA, Prentice AM, et al. Daily energy
expenditure in free-living children: comparison of heart-rate moni-
TABLE 2 toring with the doubly labeled water (2H218O) method. Am J Clin
Summary statistics of data in Table 11 Nutr 1992;56:343–52.
4. Schulz S, Westerterp KR, Bruck K. Comparison of energy expendi-
Equation Mean arith Prec n
ture by the doubly labeled water technique with energy intake, heart
L+M 17.7 7.4 87 rate, and activity recording in man. Am J Clin Nutr 1989;
S 11.8 8.2 201 49:1146–54.
SNG 22.2 9.3 54 5. Prentice AM. The doubly labelled water method for measuring
R 11.2 8.5 36 energy expenditure, technical recommendations for use in humans.
1
Mean arith, weighted mean arithmetic error across all validation stud- Nahres-4. Vienna: International Atomic Energy Agency, 1990.
ies; Prec, weighted mean precision; n, number of individual comparisons 6. Gibson JG, Evelyn KA. Clinical studies of the blood volume IV. Adap-
included in the calculation; L+M, Lifson and McClintock single-pool tation of the method to the photoelectric microcolorimeter.
model; R, Racette et al, 1994, (2-pool model fixed ratio = 1.034); S, J Clin Invest 1938;17:153–9.
Schoeller et al, 1986, equation A6 (2-pool model fixed pool ratio 1.03); 7. Lifson N, Gordon GB, Visscher MB, Nier AO. The fate of utilised
SNG, Speakman et al, 1993 (2-pool fixed pool ratio 1.0432). molecular oxygen and the source of the oxygen of respiratory carbon

Downloaded from https://academic.oup.com/ajcn/article-abstract/68/4/932S/4648697

by guest
on 23 March 2018
 938S SPEAKMAN

dioxide, studied with the aid of heavy oxygen. J Biol Chem and energy expenditure measurements in epidemiological research
1949;180:803–11. based upon direct and indirect calorimetry. The Hague: CIP-gegevens
8. Lifson N, Gordon GB, McClintock R. Measurement of total carbon diox- Koninklijke Bibliotheek, 1985.
ide production by means of D218O. J Appl Physiol 1955;7: 704–10. 33. Speakman JR. Principles, problems and a paradox with the measure-
9. LeFebvre EA. The use of D218O for measuring energy metabolism in ment of energy expenditure of free-living subjects using doubly
Columba livia at rest and in flight. AUK 1964;81:403–16. labelled water. Stats Med 1990;9:1365–80.
10. Lifson N, McClintock R. Theory of use of the turnover rates of body 34. Speakman JR. Calculation of CO2 production in doubly labelled water
water for measuring energy and material balance. J Theor Biol studies. J Theor Biol 1987;126:101–4.
1966;12:46–74. 35. Dykstra CR, Meyer MW, Karasov WH. Validation of the doubly
11. Mullen RK. Respiratory metabolism and body water turnover rates of labeled water method in bald eagles (Haliaeetus leucocephalus) and a
Perognatus formosus in its natural environment. Comp Biochem Phys- comparison of two equations for the calculation of energy expenditure.
iol 1970;32:259–65. Physiol Zool 1997;70:19–26.
12. Nagy KA. Water and electrolyte budgets of a free-living desert lizard, 36. Speakman JR, Nair KS, Goran MI. Revised equations for calculating
Sauromalus obesus. J Comp Physiol 1972;79:39–62. CO2 production from doubly labelled water in humans. Am J Physiol
13. Mullen RK. The D218O method of measuring the energy metabolism of 1993;264:E912–7.
free-living animals. In: Gessaman JA, ed. Ecological energetics of 37. Matthews DE, Gilker CD. Impact of the H-2 and O-18 pool size deter-
homeotherms. Logan, UT: Utah State University Press, 1973; 32–43. minations on the calculation of total energy expenditure. Obes Res
14. Kuczmarski RJ, Flegal KM, Campbell SM, Johnson CL. Increasing 1995;3:S21–9.
prevalence of overweight among US adults—the National Health and 38. Racette SB, Schoeller DA, Luke AH, Shay K, Hnilicka J, Kushner RF.
Nutrition Examination Surveys 1960–1991. JAMA 1994; 272:205–11. Relative dilution spaces of 2H- and 18O-labeled water in humans. Am J
15. Prentice AM, Jebb SA. Obesity in Britain—gluttony or sloth? Br Med Physiol 1994;30:E585–90.
J 1995;311:437–9. 39. Coward WA, Ritz P, Cole TJ. Revision of calculations of the doubly-
16. Keys A. Overweight, obesity, coronary heart disease, and mortality. labeled water method for measurement of energy expenditure in
Nutr Rev 1980;38:297–307. humans. Am J Physiol 1994;30:E805–7.
17. Garrow JS. Treat obesity seriously: a clinical manual. Edinburgh:
40. Westerterp KR, de Boer JO, Saris WHM, Schoffelen PFM, Ten Hoor F.
Churchill Livingstone, 1981.
Measurement of energy expenditure using doubly labeled water. Int J
18. Schoeller DA. Development of the doubly labelled water method for
Sports Med 1984;5:74–5.
measuring energy expenditure. Proceedings of the AO Nier Symposium
41. Schoeller DA, Kushner RF, Jones PJH. Validation of doubly labeled
on Inorganic Mass Spectrometry. Durango Colorado: 1993.
water for measuring energy expenditure during parenteral nutrition. Am
19. Schoeller DA, van Santen E. Measurement of energy expenditure in
J Clin Nutr 1986;44:291–8.
humans by doubly labelled water. J Appl Physiol 1982;53:955–9.
42. Westerterp KR, Brouns F, Saris WHM, Ten Hoor F. Comparison of dou-
20. Roberts SB, Coward WA, Schlingenseipen KH, Nohra V, Lucas A.
bly-labelled water with respirometry at low activity and high activity lev-
Comparison of the doubly labeled water (2H218O) method with indirect
els. J Appl Physiol 1988;65:53–6.
calorimetry and a nutrient balance study for simultaneous determina-
43. Jones PJ, Winthrop AL, Schoeller DA, et al. Evaluation of doubly labeled
tion of energy expenditure, water intake, and metabolizable energy
water for measuring energy expenditure during changing nutrition. Am J
intake in preterm infants. Am J Clin Nutr 1986;44:315–22. Clin Nutr 1988;47:799–804.
21. Jones PJH, Winthrop AL, Schoeller DA, et al. Validation of doubly 44. Seale JL, Miles C, Bodwell CE. Sensitivity of methods for calculating
labelled water for assessing energy expenditure in infants. Pediatr Res energy expenditure by use of doubly labeled water. J Appl Physiol
1987;21:242–6. 1989;66:644–53.
22. Schoeller DA, Webb P. Five-day comparison of the doubly labeled 45. Seale JL, Rumpler WV, Conway JM, Miles CW. Comparison of doubly
water method with respiratory gas exchange. Am J Clin Nutr labeled water, intake balance, and direct- and indirect-calorimetery meth-
1984;40:153–8. ods for measuring energy expenditure in adult men. Am J Clin Nutr
23. Klein PD, James WPT, Wong WW, et al. Calorimetric validation of the
1990;52:66–71.
doubly labelled water method for determination of energy expenditure in
46. Ravussin E, Harper I, Rising R, Bogardus C. Energy expenditure by dou-
man. Hum Nutr Clin Nutr 1984;38C:95–106.
bly labeled water: validation in lean and obese subjects. Am J Physiol
24. Prentice AM, Coward WA, Davies HL, et al. Unexpectedly low levels of
energy expenditure in healthy women. Lancet 1985;1: 1419–22. 1991;261:E402–9.
25. Schoeller DA, Ravussin E, Schultz Y, Acheson KJ, Baertschi P, Jequier 47. Hoyt RW, Jones TE, Stein TP, et al. Doubly labeled water measurement of
E. Energy expenditure by doubly labelled water: validation in humans human energy expenditure during strenuous exercise. J Appl Physiol
and proposed calculation. Am J Physiol 1986;250: R823–30. 1991;71:16–22.
26. Arnould JPY. The energetics of lactation in polar bears (Ursus mar- 48. Jones PJH, Leitch CA. Validation of doubly labeled water for measure-
itimus Phipps). MSc thesis. University of Saskatoon, Saskatchewan, ment of caloric expenditure in collegiate swimmers. J Appl Physiol
Canada, 1990. 1993;74:2909–14.
27. Westerterp KA, Kayser B, Brouns F, Herry JP, Saris WHM. Energy 49. Seale JL, Conway JM, Canary JJ. Seven-day validation of doubly labeled
expenditure climbing Mt Everest. J Appl Physiol 1992;73:1815–9. water method using indirect room calorimetry. J Applied Physiol
28. Westerterp KA, Meijer GAL, Janssen EME, Saris WNM, Ten Hoor F. 1993;74:402–9.
Long-term effect of physical activity on energy balance and body com- 50. Pullicino E, Coward WA, Elia M. Total energy expenditure in intra-
position. Br J Nutr 1992;68:21–30. venously fed patients measured by the doubly labelled water tech-
29. Black AE, Coward WA, Cole TJ, Prentice AM. Human energy expen- nique. Metabolism 1993;42:58–64.
diture in affluent society—an analysis of 574 doubly labelled water 51. Speakman JR. How should we calculate CO2 production in doubly
measurements. Eur J Clin Nutr 1996;50:72–92. labelled water studies of animals? Funct Ecol 1993;7:746–50.
30. Schoeller DA. Measurement of energy expenditure in free-living 52. Roberts SB, Dietz W, Sharp T, Dallal GE, Hill JO. Multiple laboratory com-
humans by using doubly-labeled water. J Nutr 1988;118:1278–89.
parison of the doubly labeled water technique. Obes Res 1995;3:S3–13.
31. Speakman JR. Doubly labelled water: theory and practice. London:
53. Schoeller DA, Hnilicka JM. Reliability of the doubly-labeled water
Chapman and Hall, 1997.
32. Coward WA, Prentice MA, Murgatroyd PR, et al. Measurement of CO2 method for the measurement of total daily energy expenditure in free-
and water production rates in man using 2H, 18O labelled H2O: compar- living subjects. J Nutr 1996;126:S348–54.
isons between calorimeter and isotope values. In: van Es AJH, ed. Euro- 54. Speakman JR. Estimation of precision in DLW studies using the two-
pean Nutrition Report 5: human energy metabolism: physical activity point methodology. Obes Res 1995;3(suppl 1):31–9.

Downloaded from https://academic.oup.com/ajcn/article-abstract/68/4/932S/4648697

by guest
on 23 March 2018