点击显示 收起
1 From the Unité du métabolisme Protéino-énergétique, National Institute for Agronomic Research/University of Auvergne, Human Nutrition Research Center Auvergne, the Centre hospitalier et Universitaire, Clermont-Ferrand, France (HD-B, LM, and YB) and and the Service de pédiatrie B, University Hospital Center, Clermont-Ferrand, France (MM)
2 Address reprint requests and correspondence to Y Boirie, Laboratoire de nutrition Humaine, BP 321, 58, rue Montalembert, 63009 Clermont-Ferrand Cedex 1, France. E-mail: yves.boirie{at}sancy.clermont.inra.fr.
ABSTRACT
Background: Recommendations for energy intake in obese children rely on accurate methods for measuring energy expenditure that cannot be assessed systematically.
Objective: The aim was to establish and validate new equations for predicting resting energy expenditure (REE), specifically in obese children.
Design: REE (indirect calorimetry) and body composition (bioelectrical impedance analysis) were measured in 752 obese subjects aged 3–18 y. The first cohort (n = 471) was used to establish predictive equations, the second (and independent) cohort (n = 211) was used to validate these equations, and the third cohort, a follow-up group of children who lost weight (n = 70), was used to examine predictive REE in the postobese period. REE values predicted with the use of various published equations and the new established equation were compared with measured REE by using the Bland-Altman method and Student’s t tests.
Results: In cohort 1, significant determinants of the new prediction equations were fat-free mass in boys (model R2 = 0.79) and age and fat-free mass in girls (model R2 = 0.76). External validation conducted by using the Bland-Altman method and Student’s t tests, in cohort 2, showed no significant difference between measured REE and predicted REE with the new equation. When already published equations were applied, systematical bias appeared with all published equations except for that of the World Health Organization. In cohort 3, the children who lost weight, almost all equations significantly underestimated REE.
Conclusions: These new predictive equations allow clinicians to estimate REE in an obese pediatric population with sufficient and acceptable accuracy. This estimation may be a strong basis for energy recommendations in childhood obesity.
Key Words: Resting energy expenditure • body composition • obesity • fat-free mass • predictive equation • children
INTRODUCTION
The prevalence of overweight and obesity in children and adolescents is rapidly increasing in Western societies (1, 2). This alarming situation is a serious public health problem because the obese adolescent has the highest risk among adolescents of becoming an obese adult (3), regardless of parental obesity (4). Recommendations for energy intake should rely on accurate methods for assessing energy expenditure (EE). However, because not all offices have the correct apparatus so that REE can be measured in everyday situations, accurate equations to predict EE are required. It is also necessary to measure changes in lean body mass and EE during weight loss so that weight-reduction programs can be adapted to preserve fat-free mass (FFM) and growth rate. Bioelectrical impedance analysis (BIA) is an appropriate and simple method for assessment of body composition and could be used for the ultimate prediction of EE in humans (5). Predictive equations of resting EE (REE) are usually based on sex, age, and body weight (6–8). This information is particularly relevant because REE contributes 60–70% of daily EE in individuals, although the contribution of physical activity should also be considered (9).
Equations to predict REE in obese subjects, especially in children, have already been established (10). Although FFM explains interindividual variations in REE better than does body weight (11, 12), body composition is rarely considered, particularly in children. Moreover, the applicability of these equations and their accuracy have not been tested in the same population after changes in body composition, eg, weight loss.
Therefore, the 3 aims of the first part of this study were to establish new predictive equations using body composition, on a large sample of French obese children, to validate the equations on a second independent population of obese children, and to evaluate their accuracy in a longitudinal follow-up survey in which body composition and REE were simultaneously measured. Finally, boys and girls with a large age range, 3–18 y old, were included to allow examination of a period of life that is characterized by various changes in growth rate and metabolic changes. In a second part, the accuracy of the newly established equations was compared with that of published equations: a World Health Organization (WHO) equation (6), the Harris-Benedict equation (7), the Schofield weight and height equations (8), and the equation of Tverskaya et al (Tverskaya equation; 10).
SUBJECTS AND METHODS
Subjects
Inclusion and exclusion criteria for the first and second cohorts were the same except for date of inclusion. To take part in this study, children aged 3–18 y should have a body mass index (BMI; in kg/m2) z score 2 and must be visiting a pediatric nutritionist for the first time. Children and parents must agree to nutritional follow-up with weight control, nutritional and exercise advice, and REE measurement. For cohort 1, the inclusion period was the years 1993 through 1999, and, for cohort 2, it was the period from 1 January 2001 to 1 March 2002.
REE was measured at the Human Nutrition Laboratory (Clermont-Ferrand, France) after exclusion of any evolving disease. A total of 471 children constituted the cohort 1. New equations were established in this population. We validated these equations in cohort 2 (n = 211), which was constituted of other obese children who were referred between 1 January 2001 and 1 March 2002 for an REE measurement.
Some of the cohort 1 children were measured again after the first investigation; 70 children had lost weight (
Methods
Data were compiled by using EXCEL software (version 98; Microsoft, Redmond, WA), and analyses were conducted by using SAS statistical software (version 8.0; SAS Institute, Cary, NC).
The SD score for BMI (z score) was determined by the following formula:
RESULTS
Descriptive analysis of the 3 cohorts
The main characteristics of the 471 children included in cohort 1 are shown in Table 2. Children were aged 3–18 y with a Gaussian distribution (20%, 35%, 31%, and 14% for age 3–9, 9–12, 12–15, and 15–18 y, respectively). Sex ratio (M/F) was 0.68, and no significant age differences existed between boys and girls. Girls had a lower waist-hip (W-H) z score, and their body composition was different in that they had more fat mass and less FFM than did the boys (Table 2). Measured REE was higher in boys than in girls, and this variation increased with age (P < 0.001). When adjusted for FFM, the difference between the sexes decreased but remained significant (P = 0.02). A second and independent population (cohort 2; n = 211) was composed of children who were significantly older (P < 0.01) and heavier (P < 0.001) and who had a significantly (P < 0.001) higher FFM than did the first population (Table 2). The age distribution in cohort 2 was 15%, 15%, 45%, and 25% for ages 3–9, 9–12, 12–15, and 15–18 y old, respectively. Finally, 70 of the children in cohort 1 were measured again 6.22 ± 1.44 mo after the first measurement; this group made up cohort 3. During the period between measurements, they had lost 12% of their initial weight and 5% of their initial fat mass. Characteristics of this cohort are summarized in Table 2. This cohort was older than both other cohorts. There was no age significant difference between boys and girls, but boys lost more weight and fat mass than did girls.
View this table:
TABLE 2. Clinical characteristics of all subjects1
Analysis of the first part of the study
Establishment of new equations
FFM and weight were major determinants of measured REE, and they explained 86% and 82%, respectively, of the variance in simple linear regression. Correlation coefficients of height, fat mass, BMI, and age varied from 0.63 to 0.79. In a multivariate analysis, we tested the model by entering only age, weight, height, and BMI. When age and BMI were entered in the model, R2 was 0.67 in boys and 0.58 in girls. When age, weight, and height were entered in the model, R2 was 0.76 in boys and 0.74 in girls. When the body-composition variables (ie, FM and FFM) were entered in the model, REE was mainly explained by FFM (79% of variation in REE in boys and 72% of variation in REE in girls; Table 3). In girls, age and FFM explained 76% of REE variation (Table 3). Consequently, new prediction equations adapted for an obese pediatric population with age ranging from 3 to 18 y are as follows:
DISCUSSION
In the current study, predictive equations were developed in both girls and boys to predict REE in a large population of obese children (mean W-H z score = 3.73 ± 0.97) ranging from 3 to 18 y of age. In the process of building the equation, we considered, first, a model with sex as the binary variable. FFM, age, and sex were significant, but the accuracy was <79% and <76% in boys and girls, respectively, with the use of the newly established equation (Table 3). We considered a second model by age group (3–9, 9–12, 12–15, and 15–18 y old), in which FFM and sex were significant factors in the youngest group, and BMI was a significant factor in the oldest group. R2 in this model varied between 0.5 and 0.66. According to selection criteria adopted in this study (R2, SE, and P value), we did not select these models. Sex was a variable that was always significant in various models, and, therefore, a model by sex finally was adopted. These equations were validated by using an independent sample with the same recruitment in the first part of the study and a longitudinally surveyed subsample in the second part. Differences in REE are known to be related to FFM (17, 18), but genetics also could explain the difference in REE between populations (19). For this reason, predictive equations for REE based on body-composition formulas are generally population specific and thus should be more appropriate.
In the new equations, as expected, FFM was the major determinant of REE; it explained 79% and 76% of REE in boys and girls, respectively,. These results are in agreement with other reports (12). In boys, the new equation was the most accurate for use in the obese population, but, in girls, age was also a significant determinant of REE. This finding suggests a greater influence of hormonal factors in girls: our sample included children aged 3–18 y, which included those in the pubertal period. This period is associated with rapid anatomical and physiologic changes, including variations in metabolic rate and energy requirements. REE in obese children and adolescents was significantly higher than that in normal-weight children (20). When REE was adjusted for body-composition differences, no differences persisted between obese and nonobese populations in some reports (21), but they did persist in others (22).
When an external validation was conducted in cohort 2, no significant difference between measured and predicted REE with the newly established equation was found. REE was well predicted by the new equation. The Bland-Altman method allowed us to consider the new equation as a good tool for predicting REE. When we tested the simple linear regression (Figure 1), the slope was significantly different from 1, and the intercept was significantly different from 0. There was effectively a negative bias, which was strongly linked to the z score, and there was more inaccuracy in a high z score. Therefore, we rebuilt the model with consideration of the z score, but R2 was not improved (0.75). This problem was observed recently in extremely obese adult women and reported (23). Indeed, the authors of that report stated that, in studies in the morbidly obese, predictive equations developed for nonobese populations were more accurate (3%) than were the obese-specific equations, because REE can be 40% overpredicted or 21% underpredicted by the equations developed for use in moderately obese populations. Therefore, we recommend caution when using predictive equations in extremely obese children. When a weight loss occurred in these subjects, both new and published equations, except the Tverskaya equation, significantly overestimated REE. With multiple linear regression, relative FFM loss explained 6% of measured REE variability in boys. Thus, we believe that all these equations are to be used only during the weight-stable period and not during the weight-changing period. The equations should be applied in obese children before any weight loss or after weight stabilization in the postobesity state. Maintenance of a reduced body weight is associated with compensatory changes in EE that oppose the maintenance of a body weight that is different from the usual weight (24, 25). Predictive equations must be established in a weight-stable period when energy metabolism has adapted.
All predicted equations used in this study led to significant miscalculations of REE. Individual estimations were markedly underestimated (31%) or overestimated (57%). Previous studies established various equations to measure REE by using BIA. With restriction to the analysis of obese children, the WHO equation overestimated measured REE in 4 studies (10, 26–28) but not in the fifth study (29). These reports also concluded that differences existed according to sex (26, 27). The question now is to justify the choice of one formula rather than another. Individual underestimations or overestimations might have a deleterious effect on diet prescription and may limit weight loss. These calculations misestimated REE by 3.76 MJ/d, which is not acceptable in clinical practice. This underestimation could lead to a miscalculation in dietary recommendations. The populations used for validation can explain these large errors. Harris and Benedict derived their equations by using data from healthy nonobese children (7), so these formulas were not specific to obese children. In agreement with literature, all predictive equations underestimated or overestimated REE according to the population (10). Predictive equations were population specific and constituted a population-based REE estimation, which is less valid on an individual scale. So the question remains as to the extent to which such a variation between individual REE measurements and REE from predictive equation should be tolerated. Predictive equations permit a first estimation of REE. When a failure of dietetic intervention occurred, REE measurement was appropriate.
Body composition was estimated by BIA. This is a useful technique for body-composition analysis in healthy persons who are overweight or mildly to moderately obese (30, 31). Among numerous studies, Utter et al (31) compared body composition evaluated by leg-to-leg BIA or underwater weighing before and after a weight-reduction program in obese and normal-weight control subjects. No significant difference was found between underwater weighing and BIA in estimating the FFM at the baseline and after weight loss. However, BIA values are affected by numerous variables including body position, hydratation status, ambient air, skin temperature, and recent physical activity (32). Equations used by various BIA apparatuses are different and not always known with precision. BIA validity has not really been confirmed, particularly in severely obese children (33, 9). A well-defined procedure specifically for performing routine BIA measurements and the equations used for BIA are necessary (30). In this study, external validation in an independent population of obese children confirmed the validity of the new equation. Relative FFM loss must be taken into account when obese children lose weight, but metabolic adaptation during rapid weight modifications should also be considered (25). In conclusion, new predictive equations for REE calculations in obese children have been validated in a large population with an accuracy sufficient to allow clinicians to better estimate energy balance in pediatric subjects. These equations have been compared with other equations not specifically dedicated to obese subjects of all ages or developed for pediatric populations, particularly obese and formerly obese children. Because body-weight changes are associated with modifications in the relation between lean mass and metabolic rate, it is recommended that these equations are used in a several-week period of body-weight stability. When REE is measured, these equations may be helpful in detecting specific alterations in the regulation of energy balance that lead to severe obesity in children.
ACKNOWLEDGMENTS
We thank the children for their valuable contribution to the study.
HDB contributed to the design of the study, data collection, data analysis and interpretation, and manuscript writing. MM enrolled the children in the study and contributed to the manuscript preparation. LM was responsible for all measurements and data collection. YB was responsible for study conception, data interpretation, and manuscript preparation. None of the authors had any financial or personal conflicts of interest.
REFERENCES