Body composition of infants at 6 months of age using a 3-compartment model

Background Two compartment (2C) models of body composition, including Air Displacement Plethysmography (ADP) and Deuterium Dilution (DD), assume constant composition of fat-free mass (FFM), while 3-compartment (3C) model overcomes some of these assumptions; studies are limited in infants. Objective: To compare 3C estimates of body composition in 6-mo. old infants from Australia, India and South Africa, including FFM density and hydration, compare with published literature and to evaluate agreement of body composition estimates from ADP and DD. Methods: Body volume and water were measured in 176 healthy infants using ADP and DD. 3C-model estimates of fat mass (FM), FFM and its composition were calculated, compared between countries (age and sex adjusted) and with published literature. Agreement between estimates from ADP and DD were compared by Bland-Altman and correlation analyses. Results:


Introduction
Body composition measurements in infants provide a critical baseline for understanding associations between growth in early life and the risk of disease throughout later life (1).However, the accurate assessment of body composition in infancy is challenging, with errors due to unsuitable prediction equations, movement artefacts during imaging methods, and the unsuitability of many instruments for small infants.An optimal infant body composition measurement method needs to be practical, accurate, precise with minimal assumptions.Sexual dimorphism in relation to adiposity is established in early infancy and may be of relevance for later-life diseases (2)(3).
The 2C-model divides the body into FM and FFM, and 2C techniques such as densitometry (body volume measurements by ADP) and isotope dilution (hydrometry, DD), have commonly been used to assess body composition in infants and children (4)(5).While these techniques are practical and relatively simple, they assume values of FFM hydration or density, primarily based on studies conducted on infants from western countries by Fomon et al. (6) and Butte et al. (7) ; these assumed age-and sex-speci c FFM density or hydration may not be appropriate universally, since many external factors can impact the process of growth and maturation.(8-9).These measurements are limited in infants from diverse countries and ethnicities.
Multi-component models of body composition, combining measurements from different techniques, can be used to directly assess the FFM composition (water and mineral content) and thus explore the variations in FFM hydration and density.In the criterion four compartment (4C) model, (10,11), total body water (TBW), bone mineral content (BMC) and body density (from body volume and weight), are assessed using hydrometry, dualenergy X-ray absorptiometry (DXA) and ADP, respectively.However, since DXA imaging involves exposure to a low dose of radiation and is affected by movement artefacts, it may not be appropriate for new-borns and young infants (2,12).On the other hand, the non-imaging 3-compartment (3C) model (13), which combines measures of weight, body volume by ADP, and total body water by DD, avoids the assumption of water content of FFM between individuals and has been used safely in infants to yield estimates of FM, TBW and the anhydrous FFM (mineral, protein and DNA/glycogen) (14).The 3C model can be used in research settings to explore the assumed constants of density and hydration of FFM (6, 7) and provide important understanding and implications while applying 2C models.Evaluation of the FFM density of Fomon et al.( 6) against 3C-model estimates in a small number of Ethiopian infants aged 1.5-6 mo.(13), found them to be similar.However, to our knowledge, no study has simultaneously estimated body composition using the 3C-model in infants from countries with varied income and environmental settings, which could affect the FFM hydration or density.
This study rstly aimed to measure and compare the anthropometry and body composition of 6-mo.old infants from Australia, India and South Africa, using the 3C-model.Secondly, the study aimed to investigate the FFM density and hydration of infants and compare the values between countries and to published values(6, 7) Thirdly, the study aimed to evaluate the agreement of FM and FFM estimates between ADP and DD.

Methods
This study included 6-mo.old infants from three countries, who were part of the larger multi-country Multi-center Infant Body Composition Reference Study (MIBCRS), which aimed to accurately assess body composition in healthy infants from birth to 2 y.The eligibility criteria for recruitment into MIBCRS were consistent with the WHO Multicentre Growth Reference Study (MGRS) (15), and full-term healthy infants (≥ 37 to ≤ 42 wk of gestation), weighing ≥ 2500 g, from singleton pregnancies of healthy non-smoker mothers planning to breastfeed for at least up to 6 mo., were recruited into the study.Infants with signi cant morbidity and congenital abnormalities were excluded.This multinational study ful lled International Ethical Guidelines for Biomedical Research involving human participants, and each country site obtained approval from their local ethics review committee.Informed consent form was obtained from the mothers.The sample size of 44 infants in each country was estimated to observe a mean difference of 3.0% in %FM between different countries, with a SD of 3.9% and 4.2% (16), at 5% level of signi cance and 80% power, after correcting for the multiple comparison between 3 countries.

Infant anthropometry
Infant weight on the day of measurement (at ~ 6 mo.) was measured using a pediatric electronic scale accurate to the nearest 0.01 kg (SECA 376, Hamburg, Germany, for Australian and South African infants; Salter 914, Delhi, India, for Indian infants).Length was measured to the nearest 0.1 cm using an infantometer (SECA 417, Hamburg, Germany).Standardized protocols were developed based on the WHO MGRS protocol (17), measurements were performed by staff who were initially trained at all sites by an international expert and then retrained every three months.

Assessment of body density by ADP
Body volume, and thereby body density, was measured by ADP (PEA POD, Software version 3.5.0,201, COSMED, USA), with a mean precision of 0.07%.Standard protocol was used at all sites (18).The ratio of weight (kg) and the measured body volume (L) was used to calculate total body density (kg/L), from which the proportions of FM and FFM were calculated using assumed densities (6, 19) for each.FM and FFM were expressed as a percentage of body weight (% FM, % FFM), and in relation to height, as FMI (kg/m 2 ) and FFMI (kg/m 2 ).

Assessment of TBW by DD
Infant TBW was determined by DD, where a dose of 1 g deuterium oxide (D 2 O; 99.8 atom %, Sigma-Aldrich, Canada) was orally administered to the infant using a syringe, following a baseline saliva sample, collected at least 15 min prior to their last feed.Saliva was sampled at 2.5 h and 3h, allowing su cient time for the isotope to equilibrate in the body water.The 3-hr point was taken for the calculation (20).Liquids were not provided to the infants between dosing and saliva sample collections.The collected saliva samples were stored in -20°C freezer until analysis.The deuterium abundance in the thawed saliva samples was analyzed in duplicate by Fourier transformed infrared spectrometry (FTIR, Agilent 4500 Series, USA) in separate laboratories.Similar SOPs were followed in all countries and efforts were made to harmonize the equipment and protocol.Quality control of TBW analysis was assured by the regular site visits by IAEA technical expert and the interlab studies.TBW was calculated as the ratio of the administered dose of deuterium to its steady state enrichment in the body water, corrected for deuterium exchange into protein using a correction factor of 1.044 (21).FFM was calculated from TBW assuming an age-and sex-speci c hydration value of FFM (6, 22).FM was calculated as the difference in body weight and FFM.

Estimation of FM, FFM, and density and hydration of FFM, using the 3C model
The 3C-model estimates of FM and FFM were derived using body weight (BW), body volume (BV, measured by ADP) and TBW (estimated from DD), assuming the density of fat (D FM ) to be 0.9007 kg/L (19), and the density of water (D W ) to be 0.9937 kg/L at body temperature (19).The density (D DB ) and composition of the 'dry' body (FM and the anhydrous FFM) were calculated using the Archimedes principle for binary mixtures with the proportions and assumed densities of fat (P F and D F ) and the density of anhydrous FFM (D p+m ), rearranging to calculate the unknown fat proportion P F as shown below.
The density of the protein and mineral (D p+m ) component was derived as 1.4959 kg/L, using sex-speci c proportions and densities of protein (P p , D p ) and mineral (P m , D m ) in a 6-mo infant (19).
FFM density (D FFM ) was calculated from the ratio of FFM and FFM volume as below FFM hydration (HF) was calculated as ratio of measured TBW, and 3C-model estimate of FFM.

Statistical analysis
Primary measurements were tested for normality assumptions using Q-Q plot and Shapiro Wilk test (p > 0.05).
Continuous variables are presented as mean ± SD, while categorical variables are presented as percentage (%).
Maternal and infant characteristics were compared between countries using Pearson's Chi-square test and analysis of variance (ANOVA), as appropriate and multiple comparison was performed by 2x2 Chi-square (adjusted for p value) and Bonferroni correction method respectively.The main effect of country and interaction effect of country x sex were analysed using Analysis of covariance (ANCOVA) to compare the infant anthropometry and body composition estimates from 3C model, between the three countries adjusted for age and sex, followed by Bonferroni correction for multiple comparisons.The agreement of FM and FFM estimates obtained from ADP vs. DD were assessed using intraclass correlation (ICC) and the Bland-Altman method (23).
The one sample t-test was used to determine if the mean difference of FM estimates from ADP and DD were signi cantly different from zero.The estimated D FFM and HF were compared with published data (6, 7) using the one sample t-test.The mean precision of the estimated FM (kg) using the 3C model from the three countries, obtained from the propagation of error method was ± 0.03 kg (1.3% of FM), calculated using the known error in measuring TBW from DD (0.02 kg), body weight (0.01 kg) and body volume (0.002 L).SPSS statistical package version 26.0 (SPSS Inc.Chicago, Illinois) was used.P < 0.05 was considered statistically signi cant.

Results
The data were normally distributed; maternal and infant characteristics are presented in Table 1.Australian mothers were signi cantly older than mothers from India and South Africa.Majority (95%) of the Indian infants were exclusively breastfed (EBF) at 3 mo., while 61% and 21% of infants were EBF in Australia and South Africa, respectively.About 91% of Australian infants were Caucasian, while all infants from India were of Asian ethnicity and South African infants belonged to Black African ethnicity.There were no between-country body weight and length differences in infants at 6 mo. of age.

3C-model estimates of body composition
Body composition estimates obtained from the 3C-model for infants are presented in Table 2. Australian infants had signi cantly higher TBW (kg) and FFM (kg) when compared to Indian and South African infants, but %TBW (59.4 ± 4.1%) was only signi cantly higher than infants from South Africa (56.1 ± 4.6%).The %FFM of Australian infants (74.7 ± 4.4%) was signi cantly higher than infants from South African (71.4 ± 5.0%), while the %FM was signi cantly lower in Australian infants (25.3 ± 4.4%), when compared to South African (28.6 ± 5.0%).The FM (kg) was similar among the three countries.. Infants from Australia had a signi cantly higher FFMI (12.7 ± 0.8 kg/m 2 ) compared to infants from India (11.9 ± 1.0 kg/m 2 ) and South Africa (12.3 ± 1.2 kg/m 2 ).Pooled sex-speci c estimates of the body composition showed that males had signi cantly higher TBW (kg), %TBW, FFM (kg), %FFM and FFMI, while females had signi cantly higher %FM.These results along with the within country sex-speci c results are summarized in Supplementary Table 1. 1 D FFM -Density of fat-free mass; HF -Hydration factor; FMI-Fat Mass Index; FFMI-Fat-free mass index; TBW-Total body water. 2 Values are mean ± SD; P values for main effects of country x sex using ANCOVA.
3 * P < 0.05, Within a row, superscripts denote signi cance between countries -a Australia vs. India, b Australia vs. South Africa; c India vs.South Africa.

Density and hydration of FFM
The D FFM of South African infants (1.071 ± 0.008 kg/L) was signi cantly higher than Australian infants (1.067 ± 0.006 kg/L, P value = 0.047), but not Indian infants (1.067 ± 0.006 kg/L, P value = 0.087).The HF of South African infants (0.785 ± 0.020) was lower compared to Australian infants (0.795 ± 0.020), however not signi cant.(P value = 0.050).These results are reported in Table 2. Sex-speci c analysis between countries (Supplementary Table 1) showed that South African females were signi cantly different from Indian females, but not Australian females in both D FFM and HF.
When comparing the sex-speci c D FFM and HF estimates from the present study with those of Fomon et al. ( 6), only females from South Africa had signi cantly higher D FFM and lower HF, while comparison of D FFM and HF data from the present study with estimates from Butte et al. (7) showed that they were signi cantly different in infants from all three countries (Table 3).

Discussion
This paper presents, for the rst time, 3C-model body composition estimates of 6 mo.infants recruited from three different countries in a simultaneous multicenter study.The rank order of FFMI was Australia > South Africa > India, while for FMI, it was South Africa > India > Australia.Infants from South Africa had signi cantly higher density of FFM when compared to infants from Australia.Sexual dimorphism existed in South African infants, females had a signi cantly higher D FFM and lower HF when compared to published estimates (6).Pooled FM (kg) and FFM (kg) estimates from the three countries, measured by ADP and DD, had good reliability and relatively low differences.
Measurement of body composition at 6 mo.can provide insights on the effect of feeding patterns on body composition.A systematic review summarized the effect of infant feeding (breast feeding vs formula feeding) on infant (0-12 mo.) body composition and observed that formula-fed infants had lower FM when compared to breast fed infants at 3-4 mo. and 6 mo., but there was a trend toward higher FM in formula-fed infants at 12 mo.
of age (24).Similar results were observed in a large, mother-baby dyad group in USA, where breastfed babies had higher percent body fat at 6 mo. of age (25).Breast fed infants were observed to have higher circulating leptin in comparison to formula-fed infants at ages < 4 mo., but not in later infancy (26, 27), which could be a reason for the observed higher FM in breast fed infants during early infancy.Similarly, breastfeeding was associated with lower FFM in infants between 3-6 mo.(24,28) and at 6 mo., when compared to formula feeding (24).The higher protein content of formula feeds compared to breast milk(25) may explain the higher accrual of FFM in early life, however, these ndings need to be prospectively con rmed in longitudinal studies.
In the present study, sex-based differences in D FFM and HF were only observed in South African infants, however, since males only comprised 34% (n = 15), results need to be considered with caution.Earlier studies (7,29,30) have not observed any signi cant sex-speci c differences and it has been suggested that attempts to identify sexbased differences in the composition of FFM in infants may not be desirable unless the methodology was perfect (31).
Most body composition reference data for infants have been published using (18,(32)(33)(34)(35)  Estimates of body composition measured by ADP and DD in the present study showed good reliability, with a low mean difference, concurring to results observed in infants aged 0.5 to 6 mo., where no signi cant differences were observed between percent FM estimates obtained from both methods (37).While DD is accurate for estimating body composition in infancy, it is not practical in infants under 3 mo.due to di culties in dosing or collecting saliva samples.ADP is a safe and rapid measurement, validated using the 4C-model in term infants (38, 39), but being practical only for infants weighing up to 8 kg.The results from the two methods should however not be used interchangeably while measuring longitudinal body composition in infants.
Limitations of the present study were the single DD measurement and the lack of feeding data at 6 mo.Factors such as socio-economic status, feeding patterns, ethnicity and climate, which may have contributed, were not considered in the analysis.In conclusion, accurate estimates of body composition in infants from three countries, using 3C model suggests some signi cant between-country variability at 6 mo. of age, however in view of the signi cant demographic, social, and environmental differences among the countries, larger studies are needed to investigate the true causes of the differences.Estimates using ADP and DD correlated well with relatively low bias, however the wide 95% CI suggests that these techniques may be more reliable for population estimates.The D FFM of South African infants differed from Australian infants and from estimates from Fomon et al. ( 6), while all countries differed from estimates of Butte et al. (7).This highlights the need for further investigation of the assumed FFM variables in infancy, which may have important implications for the different 2C models used in infant research.

Abbreviations
M (kg/L) = Bodyweight (kg) − F M (kg) Bodyvolume (L) − (F M /F M density) (L) authors acknowledge the study team at the different sites and are grateful to all the mothers and their infants who participated in the study.We thank Jonathan CK Wells for his guidance with the 3C model.Author Contributions: RK, APH, NMB, AJMA, LN and SN designed the research (project conception, development of overall research plan, and study oversight); RK, RP, and AVK performed the statistical analyses and wrote the drafts of the paper.RK has primary responsibility for the nal content.All authors have read and approved the nal manuscript.The authors declare no competing nancial interests Funding: This work was supported, in part, by the International Atomic Energy Agency (CRP E43028) and the Bill & Melinda Gates Foundation [OPP1143641].Under the grant conditions of the Foundation, a Creative Commons Attribution 4.0 Generic License has already been assigned to the Author Accepted Manuscript version that might arise from this submission.Bill & Melinda Gates Foundation did not participate in the design, management, analysis, interpretation, or preparation of the manuscript.Ethical approval: In each country, the study protocol was cleared by the Ethics Research Committee of the institution where the study was conducted.All women signed an informed consent form agreeing to participate in the study.

Figures
Figures

Table 1
Maternal and infant characteristics of the study population 2Values are mean ± SD; P values using ANOVA, Pearson's chi-square, or ANCOVA test as appropriate.3* P < 0.05, Within a row, superscripts denote signi cance between countries -a Australia vs. India, b Australia vs. South Africa; c India vs.South Africa.

Table 2
Body composition estimates using the 3C-model in infants from the three countries at 6 mo. of age

Table 3
(7) 2C model, which assumes constant values for the composition of FFM (6, 7).The values from Fomon et al. (6), were based on data from different published reports, modelled with many assumptions to derive reference data for D FFM and HF.Butte et al. values(7)were based on longitudinal data from infants of US aged birth to 2 y.The D FFM of infants from all three countries in the present study were signi cantly higher, while HF estimates were signi cantly lower than those from Butte et al.(7).Similarly, lower body FM estimates at 12 wk of age (36), was observed when using the D FFM value from Butte et al.(7) in comparison to Fomon et al. (6), being particularly evident at 1 wk of age, which could have implications for estimates of FM and FM accretion in early life.