PAPER The contribution of fat and fat-free tissue to body mass index in contemporary children and the reference child

(2002) 26, 1323 1328 ß 2002 Nature Publishing Group All rights reserved 0307 0565/02 $25.00 www.nature.com/ijo PAPER The contribution of fat and fat-free tissue to body mass index in contemporary children and the reference child JCK Wells 1 *, WA Coward 2, TJ Cole 3 and PSW Davies 4 1 MRC Childhood Nutrition Research Centre, Institute of Child Health, London, UK; 2 MRC Human Nutrition Research, Elsie Widdowson Laboratory, Cambridge, UK; 3 Centre for Paediatric Epidemiology and Biostatistics, Institute of Child Health, London, UK; and 4 Children s Nutrition Research Centre, Department of Paediatrics and Child Health, University of Queensland, Royal Children s Hospital, Brisbane, Australia BACKGROUND: Body mass index (BMI) is widely used to assess the prevalence of childhood obesity in populations, and to infer risk of subsequent obesity-related disease. However, BMI does not measure fat directly, and its relationship with body fatness is not necessarily stable over time. OBJECTIVE: To test the hypothesis that contemporary children have different fatness for a given BMI value compared to the reference child of two decades ago. DESIGN: Comparison of children from Cambridge, UK with the reference child of Fomon and colleagues (Am J Clin Nutr 1982; 35: 1169 1175). SUBJECTS: A total of 212 children aged 1 10.99 y. MEASUREMENTS: Body composition was assessed by deuterium dilution. Fat-free mass and fat mass were both adjusted for height to give fat-free mass index and fat mass index. RESULTS: Contemporary Cambridge children have similar mean BMI values to the reference child. However, both boys and girls have significantly greater mean fatness and significantly lower mean fat-free mass than the reference child after taking height into account. Contemporary Cambridge children have greater fatness for a given BMI value than the reference child. CONCLUSION: BMI-based assessments of nutritional status may be under-estimating the increase in children s fatness. Any change over time in the relationship between BMI and body fatness will create a mismatch between (1) current estimates of childhood obesity and (2) predicted risk of future adult illness, calculated on the basis of longitudinal cohorts recruited in childhood several decades ago. However, an alternative interpretation is that the reference data are inappropriate. Caution should therefore be used in generalizing from this study, and further investigations of the issue are required. (2002) 26, 1323 1328. doi:10.1038=sj.ijo.0802077 Keywords: body composition; fat-free mass; fat mass; obesity Introduction Western populations are experiencing an increasing prevalence of childhood obesity, 1,2 following a similar trend in adults. It is now established that increased fatness has serious implications for child health, 3 and is a risk factor for adult obesity and hence later disease. 4 Although obesity is defined as an excess of body fat, it is usually classified on the basis of body weight relative to *Correspondence: JCK Wells, MRC Childhood Nutrition Research Centre, Institute of Child Health, 30 Guilford St, London WC1N 1EH, UK Received 5 November 2001; revised 20 March 2002; accepted 22 April 2002 height. In 1985, Garrow and Webster published body mass index (BMI) cut-off values for adults distinguishing underweight, normal weight, overweight and obese individuals. 5 Childhood obesity is generally now defined likewise on the basis of BMI, although other definitions of childhood obesity have also been proposed. 6 BMI varies with age in children, so data are expressed as standard deviation (s.d.) scores (Zscores) for a given age and sex. 7,8 The ease with which weight and height can be measured has ensured that BMI is widely used both by clinicians to assess obesity status, and by epidemiologists researching the paediatric aetiology of the disease. The statistical rationale for BMI is that it is an index of weight that is minimally correlated with height. Weight

1324 (WT) can be adjusted for height (HT) by calculating an index WT=HT n, in which n is chosen such that the correlation of the index and height is zero. The value of n varies moderately with age, tending to increase during puberty. 9 However, it is generally accepted that BMI, where the value of n is 2, shows zero or negligible correlation with height from early infancy onwards, and BMI has become accepted as the primary index of relative weight. 10 In spite of the statistical validity of BMI in children, the additional assumption that greater BMI values are equivalent to greater body fatness in this age group is less well supported by evidence. Although studies have reported high correlations between BMI and percentage body fat, 11,12 neither of these indices is an independent measure of fatness. Weight comprises both the fat-free mass (FFM) and the fat mass (FM), and both of these components can vary between individuals. We have demonstrated previously that between-subject variability in FFM is an important source of variability in BMI in children. 13 Large-scale BMI surveys have been conducted in order to assess the changing prevalence of childhood obesity. 1,2 The validity of such an approach is dependent on there being a stable relationship over time between a given BMI value and a given level of body fatness in any population. This issue has not been investigated in any detail. We report here the body composition of 212 children aged 1 10.99 y from Cambridge, UK, measured during the 1990s by an established technique. BMI and height-normalized body composition were compared with those of the reference child, 14 which was based on the 50th percentile of the NCHS growth reference data collected between 1963 and 1975. Methods Data on anthropometry and body composition were collected for a number of studies conducted at the MRC Dunn Nutrition Unit, Cambridge, UK over the period 1989 1999. Ethical approval for all studies was granted by Cambridge Local Research Ethics Committee. Measurements on all subjects included weight (kg), height (m) and total body water (TBW; kg) by deuterium or 18- oxygen dilution. Isotope dilution space was corrected for non-aqueous isotope exchange. Details of the methodology are described elsewhere. 15,16 TBW was converted to FFM by dividing by the water content of fat-free tissue, using ageand sex-specific values 14 supported by recent studies. 16 FM was calculated as the difference between FFM and weight. Weight, height and BMI s.d. scores were calculated using the 1990 UK growth reference data. 7,17 The reference child There is a dearth of reliable data on children s body composition in earlier decades, due to the difficulty of measuring younger age groups with techniques that were both accurate and acceptable. The reference child represents an idealized child, constructed using data from several different sources collected during the period 1963 1975. 14 These data included measurements of TBW and total body potassium, which were used to estimate FFM. Actual measurements were made at birth, 6 months and 9 y, with these values being extrapolated to ages in between these time points. However, additional skinfold measures of fatness throughout childhood were incorporated into the modelling. All data were then smoothed onto the 50th percentile weight and height values of the US National Center for Health Statistics growth reference. Although it cannot be proved that the reference child is an accurate index of children s body composition in previous decades, it represents the most comprehensive data available, and includes reliable measured values at certain time points. Data expression BMI represents an index of weight that has been normalized for height. Once weight has been normalized in this way, it can be divided into fat-free and fat components: BMI ¼ WT=HT 2 ¼ FFM=HT 2 þ FM=HT 2 These two indices have been termed the FFM index (FFMI) and FM index (FMI) respectively. 18 Hence both FFM and FM, having been normalized for height, can be expressed against age in the same format as BMI 7 for comparative purposes. We have previously demonstrated the validity of this approach for comparisons of populations where mean height is similar. 19 All data were therefore expressed as BMI, FFMI and FMI. Boys and girls were considered separately in recognition of known sex differences in relative fat and lean deposition in childhood. 13,14 For each sex, values from the reference child were plotted as a line representing FFMI or FMI against age, with the Cambridge data superimposed as individual data points. Differences in mean height, BMI, FFMI and FMI between Cambridge children were assessed using paired t-tests for the difference between observed values and the age- and sexspecific reference values. This analysis was restricted to the 107 boys and 105 girls aged between 1 and 10 y, as the reference data do not extend beyond this age. Results Characteristics of the subjects are given in Table 1. All subjects were Caucasian. Mean weight, height and BMI s.d. scores were close to zero in both sexes, with standard deviations close to 1 for all parameters. These values indicate that the sample was representative of the UK population according to the 1990 growth reference values. 7,17 There was no significant trend with age for BMI s.d. values (boys: r ¼ 0.11, NS; girls: r ¼ 0.04, NS).

Table 1 Characteristics of the sample 1325 Boys Girls Mean s.d. Mean s.d. Weight s.d. score 0.30 0.90 0.16 0.94 Height s.d. score 0.18 0.92 0.11 1.08 BMI s.d. score 0.23 0.98 0.12 1.03 The Cambridge children had similar mean height to the reference child. Cambridge boys were an average of 1.0 cm taller for their age (P < 0.01), equivalent to an increase of < 1%, while Cambridge girls were on average 0.5 cm taller (P ¼ 0.32, NS). BMI of the subjects and the reference child are given in Figures 1 and 2. Cambridge boys were not significantly different in mean BMI from the reference child (D ¼þ0.06 kg=m 2, P ¼ 0.78 (NS)). However, Cambridge girls had significantly higher BMI values (D ¼þ0.61 kg=m 2, P < 0.001). FFMI and FMI of Cambridge children compared with the reference child are given for each sex in Figures 3 6. For both sexes, Cambridge children show reduced FFMI and greater FMI compared with the reference data. Comparison with Fomon s age- and sex-specific predicted values, by paired t-test, showed that these differences achieved statistical significance in both sexes (Table 2). These values indicate that the greater BMI of Cambridge girls compared to the reference child is entirely due to greater body fatness. After removing the effects of this extra weight, the Cambridge girls still showed an increase, relative to the reference child, in the ratio of FM to FFM for a given BMI value, as did the boys. Difference between current body composition and the reference value was related to age for FMI (boys: r ¼ 0.36, P < 0.001; girls: r ¼ 0.19, P < 0.001) but not for FFMI (boys: r ¼ 0.06, N.S.; girls: r ¼ 0.07, NS). Thus there was a significant age-related trend towards greater fatness compared with the Figure 2 Body mass index (weight=height 2 ) of Cambridge girls compared Figure 3 Fat-free mass index (fat-free mass=height 2 ) of Cambridge boys compared Figure 1 Body mass index (weight=height 2 ) of Cambridge boys compared Figure 4 Fat-free mass index (fat-free mass=height 2 ) of Cambridge girls compared

1326 Figure 5 Fat mass index (fat mass=height 2 ) of Cambridge boys compared Figure 6 Fat mass index (fat mass=height 2 ) of Cambridge girls compared Table 2 Bias between contemporary Cambridge children and the reference child for fat-free mass index and fat mass index Boys Girls Bias (kg=m 2 ) P-value Bias (kg=m 2 ) P-value FFMI 7 0.42 < 0.002 7 0.46 < 0.0001 FMI 0.63 < 0.0004 1.07 < 0.0001 Bias calculated as Cambridge value 7 Fomon value. reference child, but there was no similar age-related trend towards reduced lean mass. Discussion Several studies have indicated that children have become fatter in recent decades. However, the magnitude of this increase in fatness is difficult to assess: BMI and raw skinfold thickness data are proxy measures of body composition, and provide relative rather than absolute measures of fatness. Although raw skinfold data can be converted into percentage fat and hence fat mass using published prediction equations, such equations have been shown to be significantly biased in groups and, after adjusting for this bias, inaccurate in individuals. 16 In addition to their inability to quantify absolute fatness, BMI and skinfolds also provide no information on relative lean size, assessed in this study by FFMI. The strong correlation between BMI and percentage body fat 11,12 gives the impression that BMI is a good index of body fatness. Nevertheless, such a correlation does not represent the best approach for assessing this proposition. Neither BMI nor percentage fat is independent of relative lean size. We have shown previously that between-child variability in FFMI is two thirds that in FMI, such that a significant proportion of BMI variability is not due to variability in fatness. 13 Ignoring the contribution of FFMI to BMI overestimates the extent to which BMI represents fatness. However, in the present study our findings suggest paradoxically that near-stable BMI values may be concealing the extent of the increase in children s body fatness over time. On the basis of BMI values alone, Cambridge children appear to be negligibly fatter than the reference child of two to three decades ago. The boys show no significant increase in mean BMI, while the girls show an increase of 0.61 kg=m 2, equivalent to an increase of 3.8%. However, these similar BMI values mask significantly higher FMI values and significantly lower FFMI values. Cambridge children are considerably fatter for their height compared to the reference child, with the trend already apparent in late infancy. These increases in fatness, averaging 23% in boys and 35% in girls, represent the first such data to be obtained for UK children. There is an alternative to the explanation that these statistically significant findings represent real changes in body composition over recent decades, namely that the lack of agreement between our measurements and the reference child might indicate that the calculations of Fomon and colleagues are inappropriate, and have led to erroneous values for FFM and FM. The reference child was derived from measurements made in infancy and at 9 10 y, with intermediate values being extrapolated. The measurements themselves, based on multiple measurements of TBW (infancy), and TBW or total body potassium (childhood), may be considered relatively reliable, and it is the intervening values which may be flawed due to inappropriate assumptions. In the present study, differences from the reference values in FFMI and FMI were evident throughout the age range considered. Thus, at 9 10 y, we have greater confidence that the difference in body composition is real, whereas at younger age groups it is more plausible that the assumptions incorporated in the reference child might have played a role in artificially generating apparent differences. Given the limitations of the reference child, we are unable to

resolve this issue, but the hypothesis that the differences are real is consistent with previous findings in adolescents as described below. A further limitation of our study is that we collected no data on socioeconomic status, and therefore cannot demonstrate that our sample is representative of the UK population. However, our sample was representative in terms of anthropometry, as shown by the s.d. scores presented in Table 1. It is possible that the observed differences between Cambridge children and the reference child would not be reproduced elsewhere in the UK population. Potential confounding factors include variation in dietary intake patterns, physical activity patterns and other social factors, although we have no evidence to support such hypothetical differences. However, the main implication of our study is that similarity in anthropometry, in this case BMI, may nevertheless conceal differences in body composition. Our approach is therefore appropriate for all comparisons between groups, whether defined in terms of ethnicity, socioeconomic status, geographic location or historical time of assessment. Our interpretation of changes in fatness for a given BMI value is consistent with a previous analysis of US adolescents, demonstrating stability of BMI but an increase in subcutaneous fatness over the period 1966 1980. 20 In this study, triceps skinfold thickness was higher in the last survey than in the first for each level of BMI, in both sexes and for all age and ethnic groups. Our study suggests a similar scenario, but for whole-body rather than for subcutaneous fatness alone. Our findings also show consistency with a recent comparison of Spanish children studied in 1980 and 1995. 21 This study showed a trend towards increasing central fat distribution in prepubertal children, independent of changes in BMI over the same period. Change in absolute fatness was not evaluated in the study, and hence no assessment of possible changes in relative lean deposition is possible. However, it represents a further example of how changes in dimensions of fatness may exceed changes in BMI in a population, in a relatively short period of time. Lower FFMI values can be attributed to reduced muscle deposition, with the trend apparent from pre-school age. Although quantitative data are scarce, changes in lifestyle imply that UK children are substantially less active now. Compared to previous decades, fewer children walk to school, 22 fewer weekly hours of exercise are provided by schools, 23 and a greater proportion of children s leisure time is spent in sedentary activities such as watching television. 24 This reduction in activity is the most likely cause of a reduction in FFM, since exercise stimulates muscle growth. 25 A previous study of weight gain in Scottish children has likewise suggested that lean deposition is reduced and fat deposition increased, compared to the reference child, 26 and our cross-sectional data are consistent with this finding. The data used in this analysis were combined from several studies, conducted over a 10 y period. Even within this period, children s fatness may have been increasing, such that differences in the older children (measured more recently) would be predicted to be greater than in the younger subjects measured in the early 1990s. This pattern might help explain why the increase in fatness itself grows with age of the child. Nonetheless, the children are representative of weight and height of UK children according to the 1990 reference data and therefore are a suitable population in which to consider whether the ratio of FFM to FM within BMI has changed. We therefore believe our findings are not adversely affected by the nature of the sample of children. Obesity is defined as excess fat but is classified according to excess weight. According to the new international cut-off points for childhood overweight, 8 11% of the Cambridge boys and 11% of the girls are overweight. If all the children had the expected sex-specific FFMI of the reference child for their age, and their observed FMI value, then 17% of the boys and 22% of the girls would qualify as overweight. Thus the classification of obesity on the basis of BMI will underestimate the prevalence of the disease if children have lower FFM than expected for their height. Measurement of BMI is easy to obtain in children of all ages, and the index is at present the best option for screening for the disease. However, while BMI categorizes few nonobese children as obese, it has poor sensitivity and fails to identify large numbers of obese children. 27 These shortcomings, in addition to those highlighted in our analysis, indicate that additional measurements will be required if the aetiology and prevalence of obesity are to be researched with accuracy and confidence. The relationship between childhood obesity and later disease has been assessed using cohort studies, where childhood BMI has been linked firstly to the incidence of adult obesity 28 30 and diabetes 31 and secondly to the adult incidence of cardiovascular disease. 32,33 The use of such studies to infer the likely health consequences of the current obesity prevalence may be flawed if BMI does not retain the same association with body fatness over time. Therefore, further research is required to investigate this issue. Conclusion In summary, our study addresses a relatively unexplored issue, namely the extent to which the body composition underlying a given BMI value is stable over time in children. One interpretation of our findings is that there has been an increase in fat, and a decrease in FFM, for a given BMI value. Such a scenario implies that the widespread use of BMI to assess fatness in children may be providing under-estimates of the current obesity epidemic. Our analysis also provides the first data that assess the increase in children s fatness independently of changes in relative FFM deposition, suggesting that boys are on average 23%, and girls on average 35%, fatter. These findings are consistent with previous studies of BMI and skinfold data. However, an alternative interpretation is that the reference child is inappropriate as a reference, due its derivation using assumptions as well as measure- 1327

1328 ments. Caution should therefore be used in generalizing from our study, and further studies are required to confirm or refute our findings in other populations. References 1 Troiano RP, Flegal KM, Kuczmarski RJ, Campbell SM, Johnson CL. Overweight prevalence and trends for children and adolescents: the National Health and Nutrition Examination Surveys, 1963 to 1991. Arch Pediatr Adolesc Med 1995; 149: 1085 1091. 2 Reilly JJ, Dorosty AR. Epidemic of obesity in UK children. Lancet 1999; 354: 1874 1875. 3 Sokol RJ. The chronic disease of childhood obesity: the sleeping giant has awakened. J Pediatr 2000; 136: 711 713. 4 Must A, Strauss RS. Risks and consequences of childhood and adolescent obesity. Int J Obes Relat Metab Disord 1999; 23 (Suppl 2): S2 11. 5 Garrow JS, Webster J. Quetelet s index (W=H 2 ) as a measure of fatness. Int J Obes 1985; 9: 147 153. 6 Sardinha LB, Going SB, Teixeira PJ, Lohman TG. Receiver operating characteristic analysis of body mass index, triceps skinfold thickness, and arm girth for obesity screening in children and adolescents. Am J Clin Nutr 1999; 70: 1090 1095. 7 Cole TJ, Freeman JV, Preece MA. Body mass index reference curves for the UK, 1990. Arch Dis Child 1995; 73: 25 29. 8 Cole TJ, Bellizzi MC, Flegal KM, Dietz WH. Establishing a standard definition for child overweight and obesity worldwide: international survey. Br Med J 2000; 320: 1240 1243. 9 Cole TJ. Weight=height p compared to weight=height 2 for assessing adiposity in childhood: influence of age and bone age on P during puberty. Ann Hum Biol 1986; 13: 433 451. 10 Prentice AM. Body mass index standards for children. Are useful for clinicians but not yet for epidemiologists. Br Med J 1998; 317: 1401 1402. 11 Sarria A, Garcia-Llop LA, Moreno LA, Fleta J, Morellon MP, Bueno M. Skinfold thickness measurements are better predictors of body fat percentage than body mass index in male Spanish children and adolescents. Eur J Clin Nutr 1998; 52: 573 576. 12 Pietrobelli A, Faith MS, Allison DB, Gallagher D, Chiumello G, Heymsfield SB. Body mass index as a measure of adiposity among children and adolescents: a validation study. J Pediatr 1998; 132: 204 210. 13 Wells JCK. A Hattori chart analysis of body mass index in infants and children. Int J Obes Relat Metab Disord 2000; 24: 325 329. 14 Fomon SJ, Haschke F, Ziegler EE, Nelson SE. Body composition of reference children from birth to age 10 y. Am J Clin Nutr 1982; 35: 1169 1175. 15 Davies PSW, Wells JCK. Calculation of total body water in infancy. Eur J Clin Nutr 1994; 48: 490 495. 16 Wells JCK, Fuller NJ, Dewit O, Fewtrell MS, Elia M, Cole TJ. Fourcomponent model of body composition in children: density and hydration of fat free mass and comparison with simpler models. Am J Clin Nutr 1999; 69: 904 912. 17 Freeman JV, Cole TJ, Chinn S, Jones PR, White EM, Preece MA. Cross sectional stature and weight reference curves for the UK, 1990. Arch Dis Child 1995; 73: 17 24. 18 Van Itallie TB, Yang M, Heymsfield SB, Funk RC, Boileau RA. Height-normalised indices of the body s fat-free and fat mass: potentially useful indicators of nutritional status. Am J Clin Nutr 1990; 52: 953 959. 19 Wells JCK, Cole TJ, ALSPAC study team. Adjustment of fat-free mass and fat mass for height in children aged 8 y. Int J Obes Relat Metab Disord 2002 (in press). 20 Flegal KM. Defining obesity in children and adolescents: epidemiologic approaches. Crit Rev Food Sci Nutr 1993; 33: 307 312. 21 Moreno LA, Fleta J, Sarria A, Rodriguez G, Gil C, Bueno M. Secular changes in body fat patterning in children and adolescents of Zaragoza (Spain), 1980 1995. Int J Obes Relat Metab Disord 2001; 25: 1656 1660. 22 Department of Transport. National travel survey: 1992=94. HMSO: London, 1995. 23 Mason V. Young people and sport in England, 1994. The Sports Council: London, 1995. 24 Office of Population Censuses and Surveys. General household survey. HMSO: London, 1994. 25 Torun B, Viteri FE. Influence of exercise on linear growth. Eur J Clin Nutr 1994; 48 (Suppl 1): S186 189. 26 Ruxton CH, Reilly JJ, Kirk TR. Body composition of healthy 7- and 8-year-old children and a comparison with the reference child. Int J Obes Relat Metab Disord 1999; 23: 1276 1281. 27 Reilly JJ, Savage SAH, Ruxton CHS, Kirk TR. Assessment of obesity in a community sample of prepubertal children. Int J Obes Relat Metab Disord 1999; 23: 217 219. 28 Guo SS, Roche AF, Chumlea WC, Gardner JD, Siervogel RM. The predictive value of childhood body mass index values for overweight at age 35 y. Am J Clin Nutr 1994; 59: 810 819. 29 Valdez R, Greenlund KJ, Wattigney WA, Bao W, Berenson GS. Use of weight-for-height indices in children to predict adult overweight: the Bogalusa Heart Study. Int J Obes Relat Metab Disord 1996; 20: 715 721. 30 He Q, Karlberg J. Prediction of adult overweight during the pediatric years. Pediatr Res 1999; 46: 697 703. 31 Charles MA. Actualites epidemiologiques de l obesite et du diabete de type 2 en France. Diabetes Metab 2000; 26 (Suppl 3): 17 20. 32 Gunnell DJ, Frankel SJ, Nanchahal K, Peters TJ, Davey-Smith G. Childhood obesity and adult cardiovascular mortality: a 57-y follow-up study based on the Boyd Orr cohort. Am J Clin Nutr 1998; 67: 1111 1118. 33 Sinaiko AR, Donahue RP, Jacobs DR Jr, Prineas RJ. Relation of weight and rate of increase in weight during childhood and adolescence to body size, blood pressure, fasting insulin, and lipids in young adults. The Minneapolis Children s Blood Pressure Study. Circulation 1999; 99: 1471 1476.