Preterm infant body composition cannot be accurately determined by weight and length

Abstract

BACKGROUND: Body composition is a key metric for assessing nutrition in preterm infants. In many neonatal intensive care units body composition is estimated using anthropometric indices which mathematically combine body weight and length. However, the accuracy of these indices is unknown in preterm infants. In contrast, air-displacement plethysmography (ADP) has been shown to accurately measure neonatal fat mass, but it is not widely available.

OBJECTIVE: The aim was to determine which anthropometric index is most correlated to infant fat mass, as determined by ADP.

DESIGN: We performed a retrospective observational study, comparing ADP-determined percent body fat at 366 time points for 239 preterm infants (born <32 weeks), with simultaneous weight and length measurements. Non-linear regression was performed to determine the best fit anthropometric index to the body fat percentage as determined by ADP. Our non-linear regression model, % fat = Axwt^αxL^β, is the generalization of the most common anthropometric indices (BMI, ponderal index, etc.).

RESULTS: The best-fit regression formula most closely matched the formula for BMI. However, the regression explained only 51% of variability seen in body fat percentage at post-menstrual age <50 weeks, and 16% of variation seen at 50 weeks or greater.

CONCLUSION: Even optimal formulas relating weight and length to body fat percentage predict only a fraction of the variation seen in body composition, especially beyond 50 weeks. BMI was the anthropometric index most predictive of body fat percentage, but still has limited accuracy.

Keywords

Body composition measurement high risk newborn prematurity growth & development

Abbreviations

ADP

air displacement plethysmography

BMI

body mass index

PMA

post-menstrual age

VLBW

very-low birth weight

BSA

body surface area

1 Introduction

Body composition is highly variable among infants, especially those born preterm. The amount of body fat is an important factor in the growth of these infants, and is crucial for energy supply to growing tissues, especially the brain. It has been established that both adequate nutritional support and adequate weight gain are linked to improved clinical outcomes in the vulnerable preterm infant population [1 –3]. Given the heightened awareness of the crucial importance of nutrition and growth, interest in the quantification of infant growth is resurging, including two major revisions to preterm growth charts published in the last four years [4, 5]. While these charts have brought increasing sophistication to the statistical description of infant weight, length and head circumference, there is no direct link between these measurements and an estimation of actual body composition. Studies have shown “that the prematurity” and the associated medical morbidities can fundamentally alter the adiposity pattern at term age [6]. Other studies have indicated that increased growth early in life may alter eventual propensity towards obesity or cognitive impairment [7]. These outcomes have been linked to differences in anthropometric indices (BMI), but not to direct measurement of adiposity.

Previously described methods have attempted to estimate a proxy for body composition from a mathematical combination of weight and length. These estimates include ponderal index, body mass index (BMI), weight-to-length ratio, and others (see Table 1) [8 –14]. There has been no validation of any of these measures as an estimate for the body composition of a preterm infant.

Recently, direct measurement of body fat percentage has become feasible with air-displacement plethysmography (ADP). ADP determines body density from simultaneous measurements of body mass and body volume. The composition of the body in terms of fat mass and fat-free mass is calculated from the body density and body weight. A piglet model has demonstrated that ADP accurately assesses body composition in low-birth weight infants [15].

ADP technology is not universally available in neonatal intensive care units, and many patients are not medically stable enough to undergo ADP. Therefore, having a proxy for body composition which may be derived from the weight and length is still a useful and desirable tool for neonatologists. Decisions regarding macronutrient composition of enteral and parenteral nutrition could be guided by a reasonable understanding of the individual patient’s body composition if such an anthropometric correlate to body composition was known to exist. Our objective was to use ADP to determine whether any of the anthropometric indices used historically or currently accurately predict body composition.

2 Patients and methods

2.1 Study design and participants

We performed a secondary analysis of data collected for a study relating neurodevelopmental outcomes to body composition. For the initial study, approval was obtained from our hospital institutional review board, and informed consent was obtained from parents. Patients born with very-low birth weight (VLBW), defined as less than 1500 grams, were recruited for body composition measurement during and after hospitalization. Exclusion criteria for the study were grade III/IV intraventricular hemorrhage, chromosomal abnormality, necrotizing enterocolitis, or hypoxic ischemic encephalopathy. At our institution, all VLBW patients are scheduled with follow-up developmental screening at a dedicated hospital-based follow-up clinic within 2 to 3 months of discharge. Infants for this study were either recruited during their initial hospitalization or at their initial outpatient developmental appointment. For this secondary analysis we limited our data to infants born at less than or equal to 32 weeks gestational age. We divided our data set into two groups based on post-menstrual age (PMA) at the time of measurement. We chose to separate the data by those collected at less than 50 weeks PMA and those collected at greater than or equal to 50 weeks. This cut point was chosen to approximately reflect growth during the hospitalization compared to growth after hospitalization.

2.2 Measurements

Each infant had body composition measured at least once by ADP, using a PeaPod unit (Cosmed, California USA). The commercially available unit used in this study utilizes two air filled chambers, one of which is for reference and one which is occupied by the patient. An oscillating diaphragm connecting the two chambers produces small pressure perturbations, and calculates the relative volume of each chamber using Poisson’s gas law [16, 17]. Body density is calculated from the patient body volume, and a two-compartment model gives the proportion of fat and fat-free mass.

Infants recruited during their initial hospitalization had a first body composition measurement done with ADP immediately prior to discharge. The weight was measured simultaneously, and the length was measured per standard nursery procedure in the neonatal intensive care unit or level 2 nursery. At each outpatient clinic visit, patients underwent a body composition measurement with ADP. Weight was measured at the time of ADP. All length measurements, inpatient and outpatient, were done using a standardized length board. Subsequent appointments for the developmental clinic were scheduled based on the needs of the individual patients.

We chose body fat percentage as the primary outcome of interest as it is already normalized to infant weight (as opposed to total fat mass). Lean body mass percentage is the complementary value to body fat percentage, and would yield the same, but inverse, results in the analysis below.

2.3 Analysis

All of the anthropometric indices listed in Table 1 are specific cases of a more generalized formula, weight^α x length^β, with different choices for the exponents α and β. We performed a generalized non-linear regression to determine best fit parameters relating body fat percentage to a generalized anthropometric formula:

%fat = A x weight^α x length^β

The parameters of the non-linear regression are an overall constant, A, and the exponents on the weight and length, α and β, respectively. This method simultaneous tests all of the anthropometric indices listed in Table 1 at once, as well as all other possible combinations of exponents on the weight and length. The best fit for the exponents will determine which index is most closely correlated to the body fat percentage, and the goodness of fit (R²) represents the variation in body fat percentage which can be determined by this optimal anthropometric index.

For this model, we used weight measurements in kilograms, length measurements in centimeters, and fat content is expressed as a percentage of body mass. We performed the non-linear regression for the two predetermined age ranges: infants whose post-menstrual age was less than 50 weeks, and for those whose age was greater than or equal to 50 weeks. Our division of 50 weeks was chosen as the data was naturally separated at this age, and this cutoff roughly separated the data into growth during initial hospitalization and growth after discharge.

We used a non-linear regression model to determine the best fit parameters for each group. The non-linear fit was performed with SPSS version 21 (IBM, New York USA), using a modified Levenberg-Marquardt algorithm. This non-linear fit calculated best estimates of the three parameters of the model, 95% asymptotic confidence intervals for each parameter, and an ANOVA table, which was used to calculate the R² value.

3 Results

A total of 239 infants met inclusion criteria for the original study. These infants had a total of 366 separate body composition measurements. Of these measurements, 157 were done at less than 50 weeks PMA, and 209 were done at or beyond 50 weeks PMA. Seventy-three patients had measurements done during both time periods, with the remainder having measurements done in only one of the post-menstrual age ranges. Demographic information for the two groups is shown in Table 2.

The non-linear regression performed on the two post-menstrual age groups yielded the best-fit parameters listed in Table 3. Also reported are the R² values for each non-linear regression. The calculated value of R² for each fit represents the amount of variation in body fat percentage that is explained by the model given the best-fit parameters. In the group of measurements taken at PMA less than 50 weeks the value of R² was 0.507, and the best fit parameters most closely matched BMI, among the commonly used indices listed in Table 1. The group of measurements taken at PMA greater or equal to 50 weeks had R² equal to 0.161, and again the best fit exponents most closely matched BMI. Figure 1 plots the correlation between measured fat percentage and the BMI for patients in the two age ranges.

To explore the magnitude of the discrepancy between the best fit model for body fat percentage and the actual body fat percentage we calculated the difference between measured and predicted body fat percentage, and calculated the standard deviation of this difference. The standard deviation of the difference between predicted and measured body fat percentage was 3.24% for measurements done at PMA less than 50 weeks and 5.20% for measurements done at greater than or equal to 50 weeks. Of the infants in the older group, 16 of 207 have a deviation between their predicted and actual body fat percentage of more than 10%.

4 Discussion

As neonatal nutrition becomes ever more sophisticated, the need for accurate assessment of body composition in preterm infants becomes more relevant to the nutritional management of this at-risk population. In most neonatal intensive care units, the choice of which anthropometric index is utilized as a proxy for body composition is largely based on tradition and convenience. Most units do not have access to new modalities, such as ADP, and many patients do not have the stable status required to undergo this testing. Therefore, determining which body composition index most closely represents actual body composition is required for optimal monitoring of preterm infant growth.

In the two decades since the advent of commercially available ADP devices, several studies have shown the accuracy of ADP in the determination of fat mass. Recent studies have looked specifically at the use of ADP in the infant population. Comparison of ADP and dual-energy X-ray absorptiometry in full term infants has shown good correlation between the methods in the determination of fat percentage [18]. Using ADP, intrauterine growth restriction in preterm infants has been shown to correlate with lower neonatal fat mass [19, 20]. In another study, preterm and full term infants had fat mass calculated by both ADP and labeled water dilution measurement [21]. Analysis showed a high degree of correlation between the modalities, and no intrinsic bias in the ADP measurements. More direct confirmation of the accuracy of ADP was undertaken with a study of live piglets, in a weight class commensurate with VLBW infants. ADP measurements of the live piglets were found to accurately predict the fat content as determined by chemical analysis after the animals were euthanized [15]. These studies demonstrate that ADP is an accurate and reliable method of body fat percentage determination for infants.

The attempts to describe infant body composition by use of anthropometric indices date back more than 50 years. In 1966, Lubchenco published his second analysis of fetal growth as determined by weights of preterm infants, including percentiles for the ponderal (Rohrer) index, calculated by the weight divided by the third power of the length [13, 22]. These graphs are still in use in many neonatal units. Interest persists in improving the description of body composition with anthropometric indices. A recent study found very poor statistical correlation between weight-for-age and weight-for-length measures, which the authors use to suggest the importance of including a measure of body proportionality in the description of infant growth [23]. No previous study has sought to compare the utility of all anthropometric indices to an accurately determined body composition measurement in preterm infants.

The results of our study demonstrate that, in both age groups tested, the traditional anthropometric index which most closely predicts the body fat percentage is the BMI. However, even though the optimal fit falls closest to BMI, the variation in body fat percentage is only partially explained by the mathematical model. In the group of infants measured at less than 50 weeks PMA, 51% of the variation in body fat percentage was predicted by the regression model. In the infants measured at greater or equal to 50 weeks PMA, the predictive ability of the model falls to 16%.

Furthermore, the deviation between predicted and actual body fat percentages is larger in the older group. These deviations highlight the pitfalls of relying on anthropometric indices in clinical decision-making. Our regression determined the most accurate formulas for predicting the body fat percentages from weight and length measurements in this age group. Yet, the predictions of these optimal formulas very often differ from the actual fat content by clinically significant amounts.

The strength of this study is that we employed a novel approach to interrogate the efficacy of all previously proposed anthropometric indices simultaneously in our regression analysis. The best fit for this analysis represents the best possible performance of any anthropometric index of this form in this group of patients. Though our analysis was retrospective, we had a large sample size at both ages tested. Limitations of the study are that a small fraction of patients had more than one measurement in a given time period, and that the timing of the measurements was not exactly the same for each patient, depending on clinic schedule and discharge age. A prospective study could set more uniform ages for measurement and could be designed for a mixed model analysis which would control for repeated measurements.

5 Conclusion

This study demonstrates the inherent inaccuracy in using anthropometric indices to predict body composition. Our unique approach “allowed us to test all” the utility of all possible anthropometric indices in this population of preterm infants. By our method we identified the best fit formula and determined the strength of the correlation of that formula with actual body composition. The optimal fit for infants of both age groups tested most closely matched the formula for BMI. However, the predictive power of even these optimal formulas is low, especially in infants beyond 50 weeks PMA. This reiterates the need for a nuanced, multifactorial determination of growth and nutritional status in these infants. In addition, these results suggest that, when needed, accurate determination of body composition requires methods beyond the calculation of traditional anthropometric indices.

Financial disclosure

This work was supported by a grant from Cosmed USA (Concord, California). The company that supplied the grant had no input into the design of the study, data collection, interpretation, or manuscript preparation. The authors declare no conflicts ofinterest.

Human research statement

All research described here was conducted in accordance with the ethical standards of the relevant national and institutional committees, and conforms to the Helsinki Declaration. The study was performed under the approval of the Institutional Review Board of the Medical University of South Carolina.

Footnotes

Acknowledgments

We would like to acknowledge Pam Smith, Betty Bivens, Myla Ebeling, and Amy Ruddy for assisting in the enrollment and execution of this study.

References

Ehrenkranz

. Early nutritional support and outcomes in ELBW infants. Early Hum Dev 2010;86:s21–5.

Ehrenkranz

, Das

, Wrage

, et al. Early nutrition mediates the influence of severity of illness on extremely LBW infants. Pediatr Res 2011;69:522–9.

Ehrenkranz

, Dusick

, Vohr

, Wright

, Wrage

, Poole

. Growth in the neonatal intensive care unit influences neurodevelopmental and growth outcomes of extremely low birth weight infants. Pediatrics 2006;117:1253–61.

Olsen

, Groveman

, Lawson

, Clark

, Zemel

. New intrauterine growth curves based on United States data. Pediatrics 2010;125:e214–24.

Fenton

, Kim

. A systematic review and meta-analysis to revise the Fenton growth chart for preterm infants. BMC Pediatr 2013;13:59.

Uthaya

, Thomas

, Hamilton

, Dore

, Bell

, Modi

. Altered adiposity after extremely preterm birth. Pediatr Res 2005;57:211–5.

Belfort

, Gillman

, Buka

, Casey

, McCormick

. Preterm infant linear growth and adiposity gain: Trade-offs for later weight status and intelligence quotient. J Pediatr 2013;163:1564–9.

Ahn

, Garruto

. Estimations of body surface area in newborns. Acta Paediatr 2008;97:366–70.

Benn

Some mathematical properties of weight-for-height indices used as measures of adiposity. Br J Prev Soc Med 1971;25:42–50.

10.

Billewicz

, Kemsley

, Thomson

. Indices of adiposity. Br J Prev Soc Med 1962;16:183–8.

11.

Burton

. Why is the body mass index calculated as mass/height2, not as mass/height3? Ann Hum Biol 2007;34:656–63.

12.

Fok

, Hon

, Ng

, et al. Use of anthropometric indices to reveal nutritional status: Normative data from 10,226 Chinese neonates. Neonatology 2009;95:23–32.

13.

Lubchenco

, Hansman

, Boyd

. Intrauterine growth in length and head circumference as estimated from live births at gestational ages from 26 to 42 weeks. Pediatrics 1966;37:403–8.

14.

Raynor

, Rudolf

. Anthropometric indices of failure to thrive. Arch Dis Child 2000;82:364–5.

15.

Frondas-Chauty

, Louveau

, Le Huerou-Luron

, Roze

, Darmaun

. Air-displacement plethysmography for determining body composition in neonates: Validation using live piglets. Pediatr Res 2012;72:26–31.

16.

Dempster

, Aitkens

. A new air displacement method for the determination of human body composition. Med Sci Sports Exerc 1995;27:1692–7.

17.

Fields

, Goran

, McCrory

. Body-composition assessment via air-displacement plethysmography in adults and children: A review. Am J Clin Nutr 2002;75:453–67.

18.

Fields

, Demerath

, Pietrobelli

, Chandler-Laney

. Body composition at 6 months of life: Comparison of air displacement plethysmography and dual-energy X-ray absorptiometry. Obesity (Silver Spring) 2012;20:2302–6.

19.

Law

, Katikaneni

, Taylor

, et al. Customized versus population-based growth curves: Prediction of low body fat percent at term corrected gestational age following preterm birth. J Matern Fetal Neonatal Med 2012;25:1142–7.

20.

Law

, Korte

, Katikaneni

, Wagner

, Ebeling

, Newman

. Ultrasound assessment of intrauterine growth restriction: Relationship to neonatal body composition. Am J Obstet Gynecol 2011;205:255e1–6.

21.

Roggero

, Gianni

, Amato

, et al. Evaluation of air-displacement plethysmography for body composition assessment in preterm infants. Pediatr Res 2012;72:316–20.

22.

Lubchenco

, Hansman

, Dressler

, Boyd

. Intrauterine growth as estimated from liveborn birth-weight data at 24 to 42 weeks of gestation. Pediatrics 1963;32:793–800.

23.

Olsen

, Lawson

, Meinzen-Derr

, et al. Use of a body proportionality index for growth assessment of preterm infants. J Pediatr 2009;154:486–91.