Measuring fat-free mass (FFM) in infants and young children is of scientific and public health interest. Recent findings on the relationship between intra-uterine growth retardation and diseases in adults such as hypertension, type 2 diabetes and CHD generated renewed interest in studying the developmental indicators of infants and young children(Reference Barker1–Reference Barker4). Furthermore, it has been shown that more rapid postnatal growth (catch-up growth) in infants born with intra-uterine growth retardation is also a risk factor for these adult diseases(Reference Barker, Gluckman, Godfrey, Harding, Owens and Robinson5–Reference Singhal and Lucas8). In South Asian countries, the rate of low birth weight is very high largely due to intra-uterine growth retardation(Reference de Onis, Blossner and Villar9). Furthermore, the incidence of hypertension, type 2 diabetes and CHD is higher among South Asians compared with the Caucasians(Reference Yajnik, Lubree, Rege, Naik, Deshpande, Deshpande, Joglekar and Yudkin10). We therefore need population-based studies on growth and its components like FFM in infancy and early childhood. The absence of validated equations for measuring FFM in infants in South Asia based on simple measurements such as height/length and weight is a known constraint. Are the existing equations suitable for use in infants and young children in South Asia?
In the light of that mentioned earlier, measurement of body composition such as FFM assumes importance. The practical methods for use in clinics and in the field to measure FFM in infants and children are largely based on anthropometry (such as height/length and weight) and equations for children have been developed mainly in developed countries. We found two such validated equations in literature based on anthropometry(Reference Mellits and Cheek11, Reference Morgenstern, Mahoney and Warady12) for use in children. Before we can use these equations with confidence on a population different from that on which these equations were developed, it is appropriate to validate them before they are applied on a large scale in infants and children in countries in South Asia. To address this question, we evaluated these two published equations for FFM based on anthropometry. We compared the total body water (TBW) and FFM derived by these equations with a reference method based on isotope dilution using 2H2O.
Subjects and methods
Subjects
The present study was conducted in apparently healthy children among the urban poor attending an immunisation clinic of a large charitable government hospital in the city of Kolkata. The hospital service is provided free. The immunisation clinic is held once a week. Each week we interviewed the first four mothers who registered their children in the weekly clinic, and we aimed to recruit up to two children each week. The recruitment was spread over a period of 1 year.
The eligibility criteria for inclusion in the present study were age 6–24 months of either sex, absence of illness during the preceding 1 month, absence of gross congenital anomalies and chronic diseases, and parents' willingness to participate. The socio-economic and demographic features of the families are given in Table 1. These children are from low socio-economic status and records on birth weight and gestation were not available for consideration. Eighty-six children, aged 6–24 months, participated in the present study.
Socio-economic status of the families
Written informed consent was obtained from the parents and the present study was approved by the ethical review committee of the Society for Applied Studies, Kolkata, India.
Sample size
Our plan was to evaluate simple prediction equations of measuring FFM for large-scale use, based on anthropometry method. The reference method is based on 2H2O (a stable isotope) dilution. For an ideal prediction equation, the difference in the measured quantity between the reference method and the method under study should be zero. The sample size will depend on the magnitude of deviation from zero and its CI that will be acceptable for the purpose the method is to be used and therefore can only be indicative. The comparisons will be paired, i.e. the results from the two methods on each subject will be compared. The degree of deviation of the mean of the differences from zero is of interest for the present study. We have calculated the sample size to detect 1·5 % or more deviation from zero difference with a standard deviation of 2. To detect this degree of deviation with 95 % confidence and 90 % power, we need a sample of seventy-one subjects in each group(Reference Pocock13). Since the same subject will provide data for FFM by both the reference and the comparison methods, the sample size remains the same, i.e. seventy-one subjects. We added 12 % to this number to give a sample size of eighty subjects.
Anthropometric measurements
The anthropometric measurements were made using recommended protocols(Reference Lohman, Roche and Martorell14, Reference Cameron, Hauspie, Cameron and Molinari15) and are briefly described. The measurements were taken on the same day the infant underwent 2H2O dilution test.
Length
Recumbent length was measured with a wooden measuring board as described earlier(Reference Shaikh, Mahalanabis, Kurpad and Khaled16). The board was made sufficiently broad to cover the shoulder blades. The reading was taken to the nearest 0·1 cm.
Weight
Weight was measured nude, using an electronic platform balance with a precision of 10 g. The balance was checked regularly for accuracy using standard weights.
2H-labelled water dilution method
2H2O, water with a stable isotope of hydrogen, was used. This measurement took place between 09.00 and 14.00 hours. Because of their age, the infants were not fasted. A vast majority of them were breast-fed (88 %). No food other than breast-feeding was allowed for 3 h before administrating the 2H2O. Breast-feeding was stopped 1 h before administrating the 2H2O. As expected, the infants were normally hydrated at the time of the test. Children's baseline saliva was collected by a disposable syringe. We took a measured volume of 2H2O (99·9 %; Sigma, St Louis, MO) approximately equivalent to 50 mg/kg body weight in a disposable container. 2H2O has a specific gravity of 1·107. We then weighed the 2H2O dose using a precision balance and the weight with an accuracy of 0·1 mg was recorded and used in calculating TBW. It was mixed with 20 ml of distilled water and the child drank it from a disposable syringe. The empty container was rinsed with two consecutive lots of 15 ml of distilled water and the child drank both lots from the same disposable syringe. Time was allowed to equilibrate the 2H2O into the body fluid. No food or fluid including breast milk and water was permitted during the equilibration period of 3 h. At the end of the equilibration period, the second sample of saliva was collected and analysed for 2H using a dual-inlet mass spectrometer (Europe Scientific, Crewe, UK), using the zinc reduction technique. The results of 2H2O concentration in the saliva samples were used for calculating the TBW.
The following equation was used:
TBW in moles was calculated as follows:
where F 1 = (concentration of 2H2O in atom per cent (i.e. 99·9) × 106/100 − 150) ppm of 2H2O, F 2 = ppm after dose − ppm before dose of 2H2O (using concentrations in saliva samples) and N 1 = dose of 2H2O in gram/molecular weight of 2H2O (i.e. 20·0274).
In using 2H for estimating TBW, it should be noted that this isotope undergoes some exchange with non-aqueous hydrogen and a correction factor of 1·041 was used. The FFM was calculated as TBW divided by an age- and sex-specific hydration factor for FFM(Reference Butte, Hopkinson, Wong, Smith and Ellis17).
Anthropometry-based published equations
The prediction equations for TBW using weight and length that we evaluated in the present study are given later.
Anthropometry-based equation AN-1(Reference Mellits and Cheek11):
where H = length in cm, W = weight in kg and when H ≤ 132·7 cm for boys and H ≤ 110·8 cm for girls. These equations were derived from studies on the US subjects aged 1–34 years.
Anthropometry-based equation AN-2(Reference Morgenstern, Mahoney and Warady12):
These equations were derived from studies on the US subjects aged 3 months to 13 years.
The same hydration factors for FFM(Reference Butte, Hopkinson, Wong, Smith and Ellis17) were used for calculating FFM from TBW derived by the anthropometry-based equations and 2H2O dilution technique.
Statistical analysis
Epi Info version 6.0(18) and Stata version 7.0(19) were used for data entry and statistical analysis. For anthropometric data, a software package based on National Center for Health Statistics database as provided with the Epi Info software was used. As is conventionally done, correlation and regression analysis were used to compare the TBW based on anthropometry with the reference method, i.e. 2H2O dilution technique. We plotted the results along the ‘line of equality’ (i.e. the line at a 45° angle) to visually examine the concordance(Reference Bland and Altman20). Furthermore, the difference in TBW for each subject between each of the two anthropometry equations and 2H2O dilution (paired data) was evaluated for their difference from zero by t-test. Mean (sd 2) was calculated to derive the range within which 95 % of the values were expected to lie(Reference Bland and Altman20).
Results
Eighty-six children participated in the present study. In eight of them, the 2H2O procedures were incomplete because of vomiting or being fed during the equilibration period. They were dropped from the study. The study infants were predominantly breast-fed (88 %). Characteristics of the study children are shown in Table 2. Weight-for-age, length-for-age and weight-for-length standard deviation scores suggest some degree of underweight, wasting and stunting. Using the two equations, i.e. AN-1(Reference Mellits and Cheek11), AN-2(Reference Morgenstern, Mahoney and Warady12), and by 2H2O dilution, we derived TBW in kg (TBWkg). Mean TBW based on AN-1 was 1·86 % lower than the reference method (P = 0·002) and TBW derived by AN-2 was 0·75 % lower than the reference method (P = 0·2). Calculated TBW and FFM using 2H2O dilution technique and the two anthropometry equations are shown in Table 3. Mean FFM derived by AN-1 was 2·30 % lower than the reference method (P = 0·002) and FFM derived by AN-2 was 0·79 % lower than 2H2O dilution (P = 0·29). As expected, the mean FFM in boys tended to be higher than that in the girls for all the three methods. The differences however did not reach statistical significance. The mean, sd values of the difference in FFM in kg (FFMkg) and percentage of body weight between those derived by 2H2O dilution and each of the two equations under study and 95 % CI of the difference are shown in Table 4. The results show that the FFMkg derived by the equation AN-2(Reference Morgenstern, Mahoney and Warady12) gives a better agreement with the reference method; 95 % limits of agreement with the reference method (i.e. mean (sd 2)) were − 0·96 to 1·32 and − 1·10 to 1·18 for AN-1 and AN-2, respectively. Distribution of points for TBWkg along the line of identity between 2H2O and AN-1 and 2H2O and AN-2 methods is shown in Fig. 1 (a,b). Similarly, FFMkg are plotted in Fig. 2 (a,b). Bland–Altman plots of the differences in TBWkg in each participant between 2H2O and AN-1 and 2H2O and AN-2 methods against average TBWkg are shown in Fig. 1 (c,d); mean differences were 0·15 and 0·05, respectively. Similarly, for FFMkg, the differences are plotted in Fig. 2 (c,d); mean differences were 0·18 and 0·04, respectively. In Fig. 3, individual data points for percentage of FFM derived by the two anthropometry equations and by the reference method are plotted against age in months. Age- and sex-specific reference percentage of FFM in healthy American infants derived by multicomponent models(Reference Butte, Hopkinson, Wong, Smith and Ellis17) are also plotted for comparison. The percentage of FFM in the study infants derived by all three methods are consistently higher than the reference values in the US infants, indicating that the infants under study are considerably leaner and have lower fat mass percentage.
Characteristics of subjects
* WHZ, WAZ and HAZ are weight-for-height, weight-for-age and height-for-age standard deviation scores, respectively, compared with National Center for Health Statistics reference.
Mean and standard deviation (sd) of TBWkg, FFMkg and as percentage of body weight derived by the two equations under study and by the 2H-labelled water dilution technique (reference method)*
* FFM, fat-free mass; TBW, total body water; FFMkg, FFM in kg; TBWkg, TBW in kg; AN-1, anthropometry equations of Mellits and Cheek; AN-2, anthropometry equations of Morgenstern et al..
Mean, sd of the difference in FFM (of paired values) in kg and percentage of body weight between those derived by 2H-labelled water dilution and by each of the two equations under study and 95 % CI of the difference
FFM; fat-free mass.
a Difference between 2H2O and anthropometry methods (AN-1).
b Difference between 2H2O and anthropometry methods (AN-2).
* The t test evaluates the difference in the paired values from zero.
(a,b) TBWkg: Individual data points comparing values derived by the reference method (2H-labelled water dilution) with each of the anthropometric equations under study are plotted along the ‘line of identity’ (at 45° angle). ANTHTBWkg: TBWkg derived by the two equations based on anthropometry (AN-1 and AN-2), plotted against 2H2O method (R 2 = 0·73 and 0·74, respectively). 2H2OTBWkg: TBWkg derived by 2H-labelled water dilution method(Reference Mellits and Cheek11, Reference Morgenstern, Mahoney and Warady12). (c,d) Bland–Altman plots of the difference (mean (sd 2)) in each participant between TBWkg derived by 2H-labelled water reference method and each of the methods under study plotted against average of the TBWkg by 2H-labelled water method and each of the study methods(Reference Mellits and Cheek11, Reference Morgenstern, Mahoney and Warady12). 2H-labelled water–AN-1: mean difference = 0·15, sd = 0·44, P = 0·005, 2H-labelled water–AN-2: mean difference = 0·05, sd = 0·45, P = 0·372.
(a,b) FFMkg: Individual data points comparing values derived by the reference method (2H-labelled water dilution) with those by the two equations under study are plotted along the ‘line of identity’ (at 45° angle). ANTHFFMkg: FFMkg derived by the two equations based on anthropometry (AN-1 and AN-2), plotted against 2H-labelled water method (R 2 = 0·75 and 0·76, respectively). 2H2OFFM kg: FFMkg derived by 2H-labelled water dilution method(Reference Mellits and Cheek11, Reference Morgenstern, Mahoney and Warady12). (c,d) Bland–Altman plots of the difference (mean (sd 2)) in each participant between FFMkg derived by 2H-labelled water as the reference method and each of the methods under study plotted against average of the FFMkg by 2H-labelled water method and each of the study methods(Reference Mellits and Cheek11, Reference Morgenstern, Mahoney and Warady12). 2H-labelled water–AN-1: mean difference = 0·18, sd = 0·57, P = 0·007, 2H-labelled water–AN-2: mean difference = 0·04, sd = 0·57, P = 0·505.
Individual data points for percentage of FFM derived by (a) 2H-labelled water dilution method and (b,c) the two anthropometry equations(Reference Mellits and Cheek11, Reference Morgenstern, Mahoney and Warady12) are plotted against age in months. Age- and sex-specific reference percentage of FFM in healthy American infants derived by multicomponent models(Reference Butte, Hopkinson, Wong, Smith and Ellis17) are plotted for comparison (data points are connected). Boys (
), girls (▲), boys (
, Butte(Reference Butte, Hopkinson, Wong, Smith and Ellis17)), girls (
, Butte(Reference Butte, Hopkinson, Wong, Smith and Ellis17)).
Discussion
The weight-for-height indices such as Z-scores have not been evaluated for their ability to predict FFM. Weight-for-height indices and recently introduced BMI Z-score for children do not distinguish fat and lean masses. In fact there is a 2-fold range of variation in fatness for a given BMI value in individual children(Reference Wells and Fewtrell21). Based on skin-fold measurement, Yajnik and colleagues(Reference Yajnik, Lubree, Rege, Naik, Deshpande, Deshpande, Joglekar and Yudkin10, Reference Yajnik, Fall, Coyaji, Hirve, Rao, Barker, Joglekar and Kellingray22) have shown that Indian babies are not only small at birth but they also have less muscle mass and relatively more fat mass, the so-called ‘thin–fat baby’ syndrome. They further showed that thin–fat babies grow up to become thin–fat adults with thinner limbs and high waist–hip ratio; they appear to be fetally programmed and predisposed to diabetes. They also showed that the smallness and thinness of the Indian babies is present at birth and an unusual thin–fat body composition is associated with the insulin-resistance syndrome(Reference Yajnik, Lubree, Rege, Naik, Deshpande, Deshpande, Joglekar and Yudkin10, Reference Yajnik23). To further understand and study such phenomena, validated prediction equations for FFM based on anthropometry should be of value. The conventional weight-for-height indices and BMI percentile are likely to be inadequate to understand the growth and development of FFM.
The purpose of the present study was to evaluate predictive equations for the percentage of FFM derived from TBW by the anthropometry method for use on South Asian infants and young children. We found two equations in the literature for measuring TBW based on anthropometry of infants and children aged 6–24 months. Comparing these two equations with the reference method, we note that the TBW by AN-2(Reference Morgenstern, Mahoney and Warady12) gave better agreement than the one by AN-1(Reference Mellits and Cheek11). The study to develop AN-2(Reference Morgenstern, Mahoney and Warady12) used a larger sample and includes data from several studies including those of Mellits and Cheek(Reference Mellits and Cheek11). However, due to relatively wide limits of agreement (i.e. large sd) with the reference method, the equation is less useful for individual infants. However, given that the mean difference of the paired values of FFM is nearly zero and the 95 % CI is within a narrow range, the method of Morgenstern et al. should be eminently suitable for comparing groups. Butte et al. (Reference Butte, Hopkinson, Wong, Smith and Ellis17) have recently provided reference data on the body composition of children aged 0·5–24 months from the USA. Fig. 3 shows that the percentage of FFM is consistently higher in the study infants compared with those in the USA. This is so, by both the anthropometry and reference methods based on 2H2O dilution. These findings are consistent with the degree of undernutrition present in these infants (Table 2).
Some comment on the prediction equations we have evaluated is pertinent. Widely used anthropometry equations developed by Mellits & Cheek were derived on children with a wide age range of 3 months to 9 years (estimated). Their data include twenty-three boys and nine girls under 1 year of age. Morgenstern et al. used data on 167 children of a wide age range of 3 months to 13 years derived from several studies and included the data of Mellits & Cheek and it appears from the data plots that they have had a reasonable number under 2 years of age (though not stated).
Techniques for validation such as underwater weighing are not feasible in infants and young children. Use of the methods that involve any degree of radiation (e.g. dual-energy X-ray absorptiometry, radioactive tracers) is also not desirable for infants. We therefore used a stable isotope dilution technique as the reference standard. Variation exists among laboratories in the way TBW is calculated from 2H2O dilution measurements. One of us (S. S.) has had the opportunity to review the isotope dilution methods with Dr Tom Preston of the Stable Isotope Biochemistry Laboratory of the Scottish Universities Environmental Research Centre, and based on this consultation, we used a unified approach for calculating TBW using 2H2O dilution as described by Schoeller & Jones(Reference Schoeller, Jones, Ellis, Yasumura and Morgan24). Isotope dilution methods use a two-component model to measure FFM and fat mass and are generally safe, reliable, accurate and feasible in infants and children. To derive FFM, one has to use age- and sex-specific hydration factors derived by multicomponent models(Reference Butte, Hopkinson, Wong, Smith and Ellis17, Reference Ziegler, O'Donnell, Nelson and Fomon25). While the hydration of FFM changes with age and maturation, use of age- and sex-specific hydration factor largely minimises errors associated with maturation. This however may not be true in some clinical situations such as dialysis patients. For this age group, the hydration factor for FFM ranged from 80·7 to 77·0 for boys and 80·7 to 78·0 for girls(Reference Butte, Hopkinson, Wong, Smith and Ellis17). The hydration factors for FFM proposed earlier for 1- to 5-year-old children give closely similar results(Reference Ziegler, O'Donnell, Nelson and Fomon25, Reference Fomon, Haschke, Ziegler and Nelson26). Measurement of TBW using 2H2O dilution methods has been used as a reference method in many of the classic studies of body composition in children(Reference Fomon, Haschke, Ziegler and Nelson27–Reference Fomon and Nelson30).
Recently, Tennefors & Forsum have evaluated the usefulness of measuring body fatness derived from skin-fold thickness and BMI using TBW by 2H2O dilution as the reference method. Both these approaches were found to be inadequate in placing 9- and 14-month-old children in the correct body fatness group. It may be noted that TBW is better used to measure FFM, and evaluation of fatness by this method is indirectReference Tennefors and Forsum31.
To conclude, among the presently available prediction equations based on anthropometry for measuring FFM in infants and young children, the one by Morgenstern and colleagues is suitable for studying infants in India for prospective cohort studies and intervention studies such as the impact of breast-feeding programmes on chronic illness in later life(Reference Lucas and Sampson32).
Acknowledgements
We thank Md Jakir Hossain for statistical assistance. We thank Dr Sumita Jana, R. M. O. of ID & B. G. Hospital, and nursing staff of this hospital for their support and assistance in the conduct of the present study. The study was supported by the Society for Applied Studies Trust. B. S., a research fellow, carried out the study procedures, and took part in the analysis and in preparing the manuscript. D. M. conceived the study design and experiment, trained B. S. in anthropometry and 2H2O dilution test and statistical analysis, supervised her work and took part in data analysis and interpretation and in preparing the manuscript. S. S. assisted B. S. in the standardisation of procedures for anthropometry and 2H2O dilution. He also took part in the study design, analysis and interpretation of the findings. A. V. K. carried out the 2H2O analysis in the laboratory, took part in the interpretation of the findings and in critically reviewing the manuscript. K. B. took part in the study design, secondary supervision, interpretation and preparing the manuscript. Authors do not have any conflict of interest to declare.
