Previous - Next
Abstract
1. Introduction
2. Method2.1. Principle of the doubly-labelled water method
2.2. Validation studies
2.3. Possible sources of error in field applications3. A review of published studies
3.1. Studies in hospitalised patients
3.2. Studies in free-living, but sick, children
3.3. Studies relating to obesity
3.4. Studies in breast-fed and formula-fed infants
3.5. Studies in undernourished children
3.6. Using DLW estimates to establish energy requirements4. Outstanding methodological concerns
5. Future studies
6. Conclusions
References
A.M. PRENTICE, L. VASQUEZ-VELASQUEZ, P.S.W. DAVIES, A. LUCAS and W.A. COWARD *
* Dunn Nutrition Unit, Milton Road,
Cambridge, U.K. and Keneba, The Gambia.
The development of the doubly-labelled water method for use in humans provides nutritionists with the first non-invasive means of assessing total habitual energy expenditure in free-living infants and children. In combination with other more classical techniques of indirect calorimetry and assessment of body composition the new method is offering new insights into the partition of energy between maintenance metabolism, physical activity and growth. This paper describes the technique and considers its advantages and limitations with particular reference to its use in children. The accuracy of the method has been extensively tested in cross-validation studies involving adults, but there are only two reported validation studies in infants and one in premature babies. These suggest that the method is accurate, but there is a need for further validation. There are 12 published reports of DLW studies in childhood which include approximately 500 different measurements. The salient findings from these studies are summarised. Optimal uses of the new technique in studies of energy metabolism in infants and children are discussed. The new method is likely to play an important role in the establishment of new recommendations for energy requirements.
A literature search over the past decade reveals very few papers containing new data on energy expenditure during infancy. This is largely due to the difficulty of performing the necessary measurements in any circumstances other than when children are asleep. Torun discusses the problems in detail elsewhere in this volume and points out the limitations of attempting to predict total energy expenditure of children by the factorial method (TORUN, 1990). Similarly, Spurr describes the use of heart-rate monitoring and its limitations in very young children for whom calibrations cannot be obtained (SPURR, 1990).
The development of the
doubly-labelled water method for use in man has removed these limitations and promises to
revolutionise our concepts of energy metabolism and requirements in these age groups. This
paper provides a description of the new method and a review of the results obtained so
far.
2.1. Principle of the doubly-labelled water method
2.2. Validation studies
2.3. Possible sources of error in field applications
The doubly-labelled water (DLW) method provides investigators with the first non-invasive means of assessing energy expenditure in free-living children. The method simply involves the administration of an oral dose of water labelled with the stable (non-radioactive) isotopes 2H (deuterium) and 480 (oxygen-18), followed by the collection of serial daily samples of urine or saliva for approximately 7 days. Isotopic enrichments are measured by means of an isotope-ratio mass spectrometer.
The principle of the method is illustrated in Figure 1. Deuterium labels the body's water pool, and the rate constant for its disappearance (kD) provides a measure of water turnover (rH2O). Oxygen-18 labels both the water and bicarbonate pools which equilibrate rapidly due to the carbonic anhydrase reaction. The rate constant for oxygen-18 disappearance (kO) therefore measures the sum of water and carbon dioxide turnover (rCO2+H2O) Carbon dioxide production can be calculated from the difference between the two rate constants (kO-kD). The two rate constants are derived as the slopes of mono-exponential disappearance curves plotted on a log scale. Two variants of the technique are currently used: multi-point and two-point. The former establishes the slope by regression of serial daily samples whereas the latter simply uses a single sample at the beginning and the end of the period. Under most conditions the two methods give virtually identical answers, but each has certain strengths and weaknesses in special circumstances.
k = experimentally-determined rate constant;
r = production rate
Examples of isotope disappearance curves are illustrated in Figure 2 for a Gambian and a British subject. Note the steeper lines for the subject living in a tropical environment due to the increased water turnover which increases both kD and kO. Energy expenditure is computed from carbon dioxide production rate using classical respirometry formulae for which an estimate of the mean RQ over the measurement period is required.
The DLW method has been extensively cross-validated in small animals and in adult man (IDECG, 1990). However, there have only been three attempts to check the accuracy of the method in children. In each case the reference technique has been non-continuous respiratory gas exchange measurements. The results of the validations are summarised in Figure 3, and indicate no evidence of significant discrepancies between the two methods. The study by ROBERTS et al. (1986) involved measurements in four premature babies whose gas exchange was monitored for an average of 77% of the DLW period. Agreement between the two methods was excellent with a mean offset of only 0.3 ±1.8%. In the studies by JONES et al. (1987, 1988), gas exchange was only measured "for 2-6 hours on at least two occasions during the 5- to 6-day study". This inevitably results in rather imprecise reference values and probably accounts for the poor matching of some individual estimates, especially in the latter study in which two out of the eight estimates differed by over 25% (Figure 4).
The paucity of cross-validation data
in children results from the extreme difficulty of making comparator measurements, and
this is likely to remain a serious limitation for the foreseeable future. In practice,
this means that much of our confidence in the accuracy of DLW in children is based on
inferences from validation studies performed in small animals and in adult humans. This is
admittedly less than optimal, but there are no theoretical reasons why DLW should perform
less well in children.
There are two main possible sources of error, and a number of minor possible causes which could only introduce significant error under exceptional circumstances. These are as follows:
2.3.1. Fractionation corrections
It is necessary to correct the isotope elimination rates for the effects of physical fractionation which occur as water and CO2 pass from a liquid to a vapour phase. The calculations require a knowledge both of the fractionation factors and of the relative proportion of water turnover to which these factors should be applied.
There is general agreement as to which are the appropriate factors to use, and where there is some uncertainty it is offset by the fact that there is a self-cancelling effect in the calculation which minimises the impact of using inappropriate factors.
There is also broad agreement as to the proportion of water turnover which undergoes fractionation (IDECG, 1990). It is assumed that respiratory and non-sweating transcutaneous water losses are fractionated and that other losses (urine, sweat and feces) are unfractionated since the water is excreted in liquid form and any fractionation will occur outside the body. The DLW calculation therefore requires an estimate of respiratory and transcutaneous losses as a proportion of the total water turnover.
LIFSON and McCLINTOCK (1966) initially assumed that 50% of water was fractionated, but both theoretical and empirical estimates in children suggest that a value of 10-20% is more appropriate (Figure 5). This is fortunate since it reduces the likely extent of error from this source as demonstrated in Figure 6 which shows the consequences of making an incorrect estimate of the fractionated proportion. The value 'R' represents the ratio of kH2O+CO2:kH2O. In children, this is usually in the range 1.15-1.25. The left-hand panel illustrates the consequences of errors around Lifson's initial assumption of 0.5 (i.e., 50%), and the right-hand panel shows the impact of errors around an initial assumption of 20%. A two-fold error in the estimate of the fractionated proportion (e.g., using 0.1 when the true value is 0.2) would only create a 3-4% error in TEE assuming a typical value of R = 1.2. Thus, under normal circumstances in a temperate environment, the error should be small.
2.3.2. Respiratory quotient
In contrast to oxygen, which usually forms the basis of metabolic measurements, the energy equivalence of CO2 varies quite markedly according to the substrate mixture being oxidised. Each 0.01 RQ unit alters the energy value by about 1%. It is therefore necessary to estimate the mean RQ over the entire DLW period in order to convert CO2 production into energy expenditure. BLACK et al. (1986) have proposed a method for deriving this assumption based on the calculation of the dietary food quotient and appropriate corrections for any changes in body composition, for instance during growth. They demonstrate that, even with only limited dietary data, it should be possible to keep the error from this source below 2%.
2.3.3. Other minor possible sources of error
There are a number of other theoretical sources of possible error in the DLW method. These include: (a) failure to allow for expansion of the isotope dilution spaces due to growth, (b) sequestration of isotope during tissue synthesis, and (c) changes in isotopic background during a study. Except under extreme circumstances, which should be apparent to the investigator (e.g., switching from one formula to another during a measurement (JONES et al., 1988)), these errors are predicted to be lower than those outlined above and can be made even smaller by appropriate computional corrections. A full discussion of these is published elsewhere (IDECG, 1990).
2.3.4. Propagation of errors
One of the advantages of the multi-point variant of DLW is that it permits the calculation of confidence limits for each measurement that is made. This is achieved by summing the individual errors for the two isotope dilution spaces and two rate constants, and propagating these through the entire calculation with appropriate allowances for covariance of errors. The initial errors are based on the confidence intervals for the regression line through the multiple data points, hence the calculation cannot be performed using the two-point method.
The propagated error provides a
measure of both random analytical error and the extent to which the subject deviates from
the initial conditions of Lifson's model due to biological day-today variation. It does
not provide any indication of potential error or bias inherent in any of the assumptions
described above, but is nonetheless a very useful index of the confidence which can be
ascribed to individual measurements. In our experience, the average propagated error is
around 3-5% in temperate conditions, but considerably higher in tropical conditions due to
the fact that the large increase in water turnover (kH2O)
reduces the 'R' value (kH2O+CO2kH2O),
and this increases the error. A root-mean square summation of all of the errors outlined
above suggests that DLW should be accurate to better that ±7%.
3.1. Studies in hospitalised patients
3.2. Studies in free-living, but sick, children
3.3. Studies relating to obesity
3.4. Studies in breast-fed and formula-fed infants
3.5. Studies in undernourished children
3.6. Using DLW estimates to establish energy requirements
In spite of the fact that several
laboratories are experiencing hardware difficulties which are causing serious delays to
their DLW research programmes, there is already quite a substantial literature reporting
DLW measurements in children. This will be briefly reviewed below in order to illustrate
the types of opportunity provided by the new method.
The three validation studies outlined in Section 2 all involved measurements in hospitalised infants. ROBERTS et al. (1986) studied 4 premature babies weighing between 1090 and 2575 g at the time of study, and being maintained in incubators. JONES et al. (1987) studied 9 infants (age range = 6-57 days; weight range = 1.9-4.2 kg) following abdominal surgery. In their second validation they studied another 8 infants (age range = 4-57 days; weight range = 1.9-4.2 kg) who were also recovering from abdominal surgery, and had the further complication of changing from parenteral to alternative parenteral and/or oral nutrition.
All of these infants were extremely inactive as a consequence of their clinical condition, and their TEE was very close to their resting metabolic rate measured by respiratory gas exchange (indeed this assumption forms the whole basis for the validation procedure). Under such conditions, DLW has little or no advantage over short-term measurements using indirect calorimetry, and the latter are cheaper and quicker. The rationale for using DLW in the examples cited was simply to perform the cross-validations. It is important for prospective clinical users to make an objective assessment of which method will best suit their particular circumstances, and not to assume that DLW must be best simply because it is novel and fashionable.
Doubly-labelled water may have real
advantages under conditions where patients are hospitalised but not moribund. FJELD et
al. (1989) have made very effective use of the method under such conditions, and have
established a new model for predicting energy requirements of children during catch-up
growth. They studied 22 children in the early recovery phase from malnutrition (mean age =
16 months; mean weight = 7.0 kg) and 19 children in the late recovery phase (mean age = 16
months; mean weight = 8.2 kg). Energy expended for maintenance and activity estimated by
DLW was 97 ±12 and 98 ±12 kcal/kgFFM/d in the two groups. Subtraction of these values
from the metabolisable energy intake (estimated using metabolic balance techniques)
yielded an estimate of energy stored averaging 5-6 kcal per g of weight gain. A regression
analysis of TEE against weight gain indicated that tissue synthesis increased energy
expenditure by 1.0 +0.7 kcal/g gained in both early and late recovery. From these data the
authors derived a mathematical model to predict energy requirements for children during
catch-up growth as a function of initial body composition, and rate and composition of
weight gain. This is an entirely appropriate application of DLW since there is no other
available method for partitioning energy expenditure between the different components of
interest in this study.
Although there are a number of studies which fall into this category currently in progress, only one has so far been published. SHEPHERD et al. (1988) studied free-living TEE in 9 clinically-well children (age range = 9-24 months; weight range = 7-11 kg) suffering from cystic fibrosis (CF) in order to test the hypothesis that an excessive energy requirement may be partly responsible for the failure to thrive which is a feature of this disease. When compared to 16 healthy controls matched for age, weight, or age and weight, the CF children showed a 25% elevation in TEE. The authors suggested that this was additional evidence of a previously proposed basic energy-requiring defect, and that the increased TEE may be a significant contributor to undernutrition in CF children.
In studies such as these it might be
useful to have a body of normative data with which clinical cases could be compared
without the necessity of performing fresh measurements in healthy controls every time a
new study is performed. Davies and Lucas are collecting such data in large groups of
children participating in a semi-longitudinal, semi-cross-sectional design. Data have so
far been published for 41 infants studied at 1.5, 3 and 6 months (DAVIES et al.,
1989; DAVIES, 1990). Results from children aged 1 to 18 years will shortly be available.
The doubly-labelled water method is ideally suited to study energy expenditure in pre-obese or obese children and has already been used to address the hypothesis that obesity results from a depressed level of energy expenditure. ROBERTS et al., (1988) studied 6 children with lean mothers and 12 children with obese mothers. At one year of age, 6 of the latter group had become overweight (> NCHS 90th centile of weight for length), and the other 6 remained lean. This permitted the 18 infants to be divided into three equal groups: lean/lean, lean/obese and obese/obese. Measurements of postprandial metabolic rate by indirect calorimetry and of total energy expenditure by DLW were performed at 3 months of age (i.e., prior to the divergence of body weights). Postprandial metabolic rate (RMR) averaged 346, 336 and 354 kcal/d, or 65, 62 and 60 kcal/kg/d in the L/L, L/O and O/O groups, respectively. Total energy expenditure averaged 77, 77 and 61 kcal/kg/d. The energy available for physical activity (derived as TEE-RMR) was therefore 12, 15 and 1 kcal/kg/d. The authors concluded that low energy expenditure was an important cause of rapid weight gain in the infants who subsequently became overweight. Reduced expenditure on physical activity appeared to be the most important component. This is an important study which represents one of the most elegant applications of DLW published so far. However, the conclusions are based on a very small sample and are in urgent need of independent confirmation. A similar study is in progress in 9- to 10-year-old children and should provide useful additional data (W.H. DIETZ, personal communication).
SCHOELLER and colleagues (1988) have used DLW to investigate potential energy-sparing defects in 10 Prader-Willi patients (age = 16 ±4 years; weight = 60 ±23 kg; height = 141 ±14 cm) and in 10 obese controls (age = 15 ±3 years; weight = 103 ±26 kg; height = 168 ±9 cm). Total energy expenditure of the two groups was 1980 ±580 and 3710 ±810 kcal/d, or 36 ±9 and 38 ±9 kcal/kg/d, respectively. The estimated energy expended on physical activity was 11.5 ±5.9 and 18.3 ±6.6 kcal/kg/d indicating that the Prader-Willi patients were substantially less active.
BANDINI and colleagues (1989) have
studied energy expenditure during 2 weeks of carbohydrate overfeeding in 6 non-obese (age
= 15 ±1 years; weight = 62 ±8 kg; height = 168 ±6 cm) and 7 obese (age = 15 ±1 years;
weight = 90 ±12 kg; height = 165 ±3 cm) adolescents. Energy intake was 1.61 x BMR in the
maintenance period and 2.54 x BMR during overfeeding. BMR increased comparably in both
groups: 7.9 ±1.2 and 8.6 ±1.9%. The thermic effect of eating was similar in the two
groups and did not change during overfeeding: maintenance = 9.4 ±0.6 and 9.8 ±0.4%;
overfeeding = 8.6 ±0.3 and 9.2 ±0.7%. Total energy expenditure assessed by DLW averaged
2839 ±234 and 3418 ±300 kcal/d in the lean and obese groups, respectively. The
difference was not significant, and in 9 subjects, in whom a baseline measurement of TEE
had been performed, the period of overfeeding induced no significant increase in TEE. The
authors conclude that their results do not support the hypothesis that obese adolescents
have a reduced thermogenic response to food or to overfeeding.
In recent years there has been an increasing awareness that there may be profound differences in the nutrition of breast-fed and formula-fed babies. DLW has been contributing to this debate. BUTTE et al. (1988) published the first comparison of energy expenditures in sixteen 4-month-old babies fed breast milk or formula. These initial results have since been recalculated in the light of further experience with DLW (BUTTE et al., 1989) and will shortly be published in full. In summary, they report a higher energy intake (87 versus 72 kcal/kg/d), TEE (73 versus 64 kcal/kg/d) and sleeping metabolic rate (53 versus 47 kcal/kg/d) in the formula-fed group (BUTTE, personal communication, this meeting). Thus, energy intake was 15 kcal/kg/d higher, leaving a residual excess which presumably contributed to their faster growth rate (22 versus 12 g/d).
This study has demonstrated that the substantial differences in energy intake which have been observed between breast- and formula-fed babies are consistent with observed differences in the infants' daily energy budgets. Lucas and colleagues have published data confirming the Houston findings elsewhere in this volume (LUCAS, 1990).
In another study of breast-fed
babies, LUCAS et al. (1987) have used DLW in a novel approach to estimating the
energy density of suckled breast milk. This has been a long-standing problem due to the
high level of variability of breast-milk fat concentrations and the difficulty of devising
a realistic sampling protocol which does not interfere with the natural biology of the
suckling process. They used DLW to assess metabolisable energy intakes at 5 and 11 weeks
(1.81 and 2.22 MJ/d) calculated as the sum of TEE (1.28 and 1.68 MJ/d) and energy
deposited (0.52 and 0.54 MJ/d). The energy deposited was calculated on the basis of
changes in body composition measured by isotope dilution. The isotope data also yielded an
estimate of the volume of milk ingested (767 and 868 mL/d) which permitted energy density
to be calculated as metabolisable energy intake divided by volume. The resultant values of
0.24 and 0.25 MJ/100 mL (57 and 60 kcal/100 mL) are substantially lower than previous
estimates and, if confirmed, have important implications for the calculation of energy
intakes of breast-fed infants. These are discussed further by LUCAS (1990).
We have completed an extensive study of energy expenditure in 137 undernourished Gambian children aged 1-36 months and averaging 79% weight-for-age against the NCHS standard (VASQUEZ-VELASQUEZ, 1988). When expressed in absolute terms and plotted against age, TEE was significantly lower in Gambian children compared to British controls (Figure 7). However, this appears to be entirely attributable to the fact that the Gambian children were smaller, since there are no significant differences between the groups when TEE is expressed per kg body weight (Figure 8) or per kg FFM (not illustrated).
This suggests that the Gambian
infants do not reduce their physical activity in response to their marginal dietary
intake, but that their smaller attained size results in a pro rata reduction in
energy requirements when compared to a British child of the same age. This surprising
result is causing us to re-examine many of our former assumptions concerning the energy
status of the Gambian children and the likely role of simple energy deficiency in growth
retardation. If the infants are able to meet their requirements for maintenance and
activity in full, it seems difficult to envisage why they could not divert towards growth
the 1-2% of total requirements that would be necessary to achieve progress along the NCHS
50th centile. If these results are substantiated in other undernourished populations, they
would force an overall reassessment of several of our current concepts relating to
protein-energy malnutrition.
Until recently, expert committees charged with establishing recommendations for energy intake during infancy and childhood have had to rely on estimates of food intake in groups of healthy, well-nourished people. This approach has a number of acknowledged disadvantages, and the WHO consultative panel stated that future guidelines should be based on measurements of energy expenditure if and when these became available.
At the end of 1988, Prentice and colleagues collated all of the DLW TEE data available at that time (355 individual measurements) in order to compare them against the current WHO/FAO/UNU 1985 recommendations for energy intake (PRENTICE et al., 1988). Figure 9 shows the results plotted as group means from different studies and age ranges. The data are remarkably consistent in view of the fact that they were obtained by a novel technique applied by several laboratories each employing slightly different variants of the method. This consistency lends extra credibility to the results and reduces (though does not exclude) the possibility of a consistent uni-directional bias.
TEE incorporates the cost of basal metabolism, thermogenesis, physical activity and the anabolic processes associated with growth. Therefore, in order to calculate energy requirements it is only necessary to add the energy deposited as new tissue.
Figure 10 illustrates the results of this addition. The lower line represents a smoothed, weighted mean drawn through the data in Figure 9. The hatched area corresponds to an estimate of the amount of energy which would be deposited in order to sustain growth along the NCHS 50th centile. This was calculated using Fomon's data for the body composition of reference children (FOMON et al., 1982).
The new estimates of requirements,
represented by the upper edge of the hatched area in Figure 10, are substantially
lower than the WHO/FAO/UNU figures, particularly during the second and third years of
life. At 12, 24 and 36 months the discrepancies are 23, 22 and 18 kcal/kg/d, respectively.
Future expert committees will have to decide whether they accept the new data as valid,
and if so, whether they adopt them in their entirety or whether they simply use it as an
adjunct to food intake data. In reaching this decision they will need: (a) to be assured
as to the accuracy of the DLW method, and (b) to be confident that a reasonable
explanation exists for the differences between measurements of intake and expenditure.
Lucas has addressed these issues elsewhere in this volume (LUCAS, 1990).
The very limited amount of cross-validation data in children has been highlighted above. It would be preferable to have much more data in this respect, but in reality it is unlikely to be forthcoming due to the difficulty of obtaining comparator measurements of carbon dioxide production or energy expenditure in young children over long periods. Furthermore, it is difficult to validate a technique which the theoretical estimates of error suggest is likely to be more accurate than any possible 'reference' method.
A specific concern that has arisen from the review of the world literature presented above is that many of the studies are yielding very high coefficients of variation for TEE in any given age group. Examples of these are illustrated in Figure 11 in which all CVs refer to data expressed per kg body weight which should remove a large part of the variance arising from differences in body size. The CVs are commonly in excess of 25% implying that there is a three-fold range in TEE between a child at -2 SD and one at +2 SD. Is this a genuine phenomenon, or is it a result of random errors which are expanding the true range? We have taken two approaches to this question. The first is to re-analyse samples from subjects who yield outlying values and to scrutinise their data with particular vigilance in order to search for a priori grounds for excluding the result. This approach sometimes reveals errors in the analysis or unacceptably high propagated errors, but usually fails to do so. Furthermore, the results presented for publication are calculated after any such exclusions (see Vasquez-Velasquez and Davies data sets in Figure 11). A second way of detecting unacceptably low TEE values is to calculate the ratio TEE/SMR since this would never be expected to be below 1.2. VASQUEZ-VELASQUEZ (1989) and DAVIES (unpublished) have performed such an analysis and found very few results lower than 1.2 x SMR and none below 1.0 x SMR. These results fail to reveal any obvious reason to question the high CVs, but they remain intuitively surprising and require further thought as fresh data sets become available.
We are aware of on-going studies
involving DLW measurements in infants and children which involve a total of over 600
subjects. These range from small-scale clinical studies with sample sizes as low as 6,
through hypothesis testing studies with samples of 10-60, to field surveys commissioned by
government departments with samples of 100 and 200 subjects. They offer the exciting
prospect of a wealth of new information which should more than double the existing
data-base within the next few years.
It should be evident from the above review that the doubly-labelled water method has rapidly established itself as the best available technique for assessing total energy expenditure. It is ideally suited to studies of infants and young children, and has already produced a large body of new data which is influencing current concepts in energy metabolism in these age groups. So far the results have withstood critical analysis which has probed for potential flaws by using 'worst case' theoretical assumptions in order to establish the maximum possible limits of error and bias. It is crucial that this level of sceptical scrutiny is maintained, and that investigators do not consider the method to be irrefutable in all instances simply because no other techniques are available for use in young children. IDECG has already played an important role in ensuring the maintenance of high standards by convening an expert workshop whose consensus views are soon to be published (IDECG, 1990). It is to be hoped that this publication will help other laboratories to establish the method soon in order to assist in tackling the many outstanding questions regarding activity, energy expenditure and energy requirements in children.