Body mass index and percentage fat mass in healthy German ... - Nature

International Journal of Obesity (1998) 22, 461±469 ß 1998 Stockton Press All rights reserved 0307±0565/98 $12.00 http://www.stockton-press.co.uk/ijo

Body mass index and percentage fat mass in healthy German schoolchildren and adolescents F Schaefer, M Georgi, E WuÈhl and K SchaÈrer Division of Pediatric Nephrology, University Children's Hospital, Heidelberg, Germany

OBJECTIVE: To provide reference data for obesity indices in Mid-European schoolchildren and adolescents, to evaluate the usefulness of body mass index (BMI) as an indicator of obesity in children, and to analyse the patterns of fat accumulation during childhood. DESIGN: Cross-sectional observational study in 2554 healthy schoolchildren and adolescents (age, 6±19 y) living in Heidelberg, Germany in 1989=1990. Centile charts for BMI and skinfold-derived percentage body fat mass (PFM) were constructed using Cole's LMS method for normalised growth standards. RESULTS: The BMI centile values of German children ranged higher than French, lower than North American and Italian, and similar to Swedish and British children. While BMI steadily increased with age, PFM was markedly lower in peripubertal than in pre- and postpubertal boys. BMI predicted PFM with reasonable precision in girls (r 0.84), and in obese boys (r 0.58), but not in the leaner two thirds of the male population (r 0.01, NS). The 75th BMI percentile was the most appropriate cutoff value to screen for the 15% most obese patients by PFM (sensitivity: 82%, speci®city: 85%). The pattern of the trunk-to-extremity skinfold ratio across childhood suggested that the typical adult distribution of central and peripheral fat is achieved in mid puberty in girls, but not before the end of adolescence in boys. CONCLUSIONS: The major differences observed between BMI charts obtained in different countries underline the need for population-speci®c reference data. BMI is of limited usefulness in predicting relative fat mass in individual children. The developmental pattern of fat accumulation and distribution during adolescence is highly dynamic and gender-speci®c. Keywords: body mass index; obesity; fat mass; centile charts; children

Introduction An increasing prevalence of childhood obesity has been reported within the last two decades in Europe and North America.1±5 This observation raises serious concern, since childhood obesity may track into adult life,6,7 thus predisposing persistently obese individuals to early atherosclerosis and increased cardiovascular morbidity.8 Despite the growing concern about childhood obesity, remarkably few population-based reference data are available to assess an individual child's body composition in relation to his or her peers. The body mass index (BMI), de®ned by the ratio of body weight and the square of body height, is becoming a popular surrogate marker of obesity in clinical studies. Up to now, BMI reference data have been published only for North American, French, British, Swedish and Italian children.9±14 The well-known marked intercultural differences in nutritional habits and obesity preponderance, but also the age of some studies (for example, French data collected 1956±1979) limit Correspondence: Dr Franz Schaefer, Division for Pediatric Nephrology, University Children's Hospital, Im Neuenheimer Feld 150, 69120 Heidelberg, Germany. Email: [email protected] Received 11 September 1997; revised 18 December 1997; accepted 9 January 1998

the applicability of published reference values in MidEuropean children. Moreover, the BMI is an easily obtained but not very accurate index of obesity in children.15±17 Whenever possible, the anthropometric assessment of body composition should therefore include a direct estimate of body fat. Despite recent technological advances, such as bioelectrical impedance and dual photon X ray absorptiometry, skinfold thickness measurements remain the most widely available tool to quantitate body fat stores. Improved childhood-speci®c prediction equations to estimate percentage body fat (PFM) have been developed and validated.18 In order to provide current reference values, we present BMI and PFM percentile charts obtained cross-sectionally in 1989=1990 in a representative sample of more than 2500 healthy schoolchildren and adolescents living in an urban area in Southwest Germany.19,20

Methods Subjects

Anthropometric data were collected from 2554 healthy children and adolescents (1276 boys, 1278 girls) aged 6±19 y. The age distribution of the subjects is given in Table 1. All subjects were of German

BMI and percentage fat mass in German schoolchildren F Schaefer et al

462 Table 1 Age distribution of the study population Age (y)

Boys

Girls

6±6.99 7±7.99 8±8.99 9±9.99 10±10.99 11±11.99 12±12.99 13±13.99 14±14.99 15±15.99 16±16.99 17±17.99 18±19.99 Total

45 98 107 104 108 105 129 110 97 104 76 70 123 1276

43 108 94 92 108 107 105 125 108 105 97 88 98 1278

origin and lived in or within a 10 km radius of Heidelberg. Measurements were obtained between June 1989 and April 1990 in Heidelberg schools. The study population represented approximately 19% of the total age-matched population living in the Heidelberg area during the study period. While the distribution of school types was representative of the local population, the proportion of adolescents attending a `Gymnasium' (that is the school type attended more commonly by children from middle and upper class families) was slightly higher (50%) than in the state of Baden-WuÈrttemberg (39%) and on the national level (41%).19,20 Study approval was obtained from local and state school authorities. All children, parents and teachers were thoroughly informed about the purposes and contents of the study, and written informed consent was obtained from the parents or the adult probands. Approximately 96% of the subjects addressed eventually participated in the study. Anthropometric measurements

Measurements were performed between 08.00± 12.00 h according to standardized guidelines.21 Stature was determined using a mobile Harpenden anthropometer (Holtain Ltd., Crymych, UK) to the nearest mm, with the subject's head in the Frankfurt plane. Body weight was determined to the nearest 100 g using a digital scale (Seca, Hamburg, Germany). The subjects were weighed in light underwear, which was accounted for by subtraction of 200± 400 g from the measured weight. Skinfolds thicknesses were measured using a Harpenden skinfold caliper at four reference sites: the biceps and triceps skinfold thicknesses were determined midway between the acromion and olecranon processes at the anterior and posterior surface of the right arm, respectively, the subscapular skinfold 1 cm caudally and medially to the right scapular angle, and the suprailiac skinfold thickness 1 cm above the superior anterior rim of the right iliac crest. Pubertal staging was performed by direct visual inspection according to Tanner.22

All anthropometric measurements including the pubertal staging were performed by one of two single observers (medical students), who had been trained jointly by an experienced auxologist (FS) and whose quality of performance was evaluated prior to the study. Their intra- and interobserver coef®cients of variation (CV) were 0.08 and 0.17% for height, 3.18 and 7.43% for triceps, and 4.16 and 9.45% for subscapular skinfold measurements, respectively.19 Statistics

The BMI is de®ned as the ratio of an individual's weight (kg) and height squared (m2). PFM was measured according to the equations developed by Slaughter et al.18 The following equations were used (subscapular (subsc), triceps (tric) skinfold thickness in mm): A: Boys : If subsc tric 35 mm and Tanner G stage 2: PFM 1:21 subsc tric ÿ 0:008 subsc tric2 ÿ 1:7 If subsc tric 35 mm and Tanner G stage 3: PFM 1:21 subsc tric ÿ 0:008 subsc tric2 ÿ 3:4 If subsc tric 35 mm and Tanner G stage 4 : PFM 1:21 subsc tric ÿ 0:008 subsc tric2 ÿ 5:5 If subsc tric > 35 mm : PFM 0:783 subsc tric 1:6 B: Girls If subsc tric 35 mm : PFM 1:33 subsc tric ÿ 0:013 subsc tric2 ÿ 2:5 If subsc tric > 35 mm : PFM 0:546 subsc tric 9:7

At the given degree of precision achieved by the two observers regarding skinfold measurements (see above), an intra-observer CV of 2.4% and an interobserver CV of 5.8% could be expected for repeated PFM estimations. Regional fat distribution was expressed by the trunk-to-extremity skinfold ratio de®ned as the sum of the subscapular and suprailiac divided by the sum of the biceps and triceps skinfolds. Both BMI and PFM tend to be positively skewed, and vary with age during childhood. The LMS method of Cole and Green23 and maximum penalized likelihood were used to construct percentiles for BMI and PFM. In brief, the LMS method describes the distribution of a measurement y by its median (M) the coef®cient of variation (S), and a measure of skewness (L) required to transform the data to normality. These three parameters are dependent on age. After ®tting smooth estimates of L, M and S over age, percentile (Ca) estimates can be calculated as: Ca(t) M(t)[1 L(t) S(t) za]1=L(t) where M(t), L(t), S(t) and Ca(t) indicate the correspond-


ing values of each parameter at age t. Za is the appropriate normal equivalent deviate (for example for a 97%, za 1.88). This equation can be rearranged to convert a child's BMI or PFM value to an exact standard deviation score (SDS) as follows: SDS [(Y=M(t))L(t)71] = (L(t)S(t)) where Y is the child's individual BMI or PFM value, and L(t), M(t) and S(t) are the gender-speci®c values of L, M and S interpolated for the child's age. The normality of data distribution was evaluated by the Shapiro-Wilk test. Descriptive statistics of the normally distributed SDS data were performed by ANOVA followed by Duncan's test for multiple comparisons. Associations between individual parameters were evaluated by Pearson product-moment correlation analysis. Stepwise linear regression analysis was performed in order to identify independent predictors of PFM.

Results Smoothed LMS values for BMI and PFM calculated at six monthly intervals are given in Table 2 and Table 3, respectively. The computed 3rd, 10th, 25th, 50th, 75th, 90th and 97th percentiles are shown plotted against age for BMI in Figure 1 and for PFM in Figure 2. Conversion of the BMI and PFM data to SDS yielded values, which were normally distributed and did not systematically deviate from the standard normal distribution, as indicated by a mean ( s.d.)

BMI SDS of 0.001 0.97 in boys and 0.02 1.0 in girls, and a mean PFM SDS of 70.006 0.97 in boys and 0.01 0.99 in girls. The comparison with BMI reference data published for populations in France,11 the UK,12 Sweden,13 Italy14 and the US9,10 showed relatively small differences regarding the lower half of the distribution. For example, the 50th percentile for boys aged 10 y was 16.4 in French and British, 16.5 in Swedish and German, 16.8 in North American and 17.3 in Italian children. In relation to the German reference data, the differences of the population medians ranged from 70.05 to 0.35 SDS. As shown in Figure 3, larger differences were present with respect to the upper half of the distribution. When all age groups and both genders were considered together, the relative differences compared to the German 85th percentile ranged from 70.28 SDS (France) to 0.36 SDS (Italy). Since the absolute BMI value is in¯uenced by the height of a subject, it should be noted that the Heidelberg children were the tallest of all reference studies under comparison; for example boys aged 10 y were on average between 0.3 (USA) and 0.8 SDS (UK) taller than their peers in the other countries. The distribution of PFM showed a distinct sexual dimorphism. While PFM increased almost steadily with age in girls with an only slightly skewed distribution, boys showed a pre-adolescent increase with age, a sharp decrease in the pubertal age range and a small increase beyond the age of 16 y. As evident from Figure 2 and Table 3, the distribution of PFM was much more skewed in boys than in girls. When

Table 2 Smoothed measure of skewness (L), median (M) and coef®cient of variation (S) values for body mass index (BMI) at six-monthly intervals L Age (y) 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0 10.5 11.0 11.5 12.0 12.5 13.0 13.5 14.0 14.5 15.0 15.5 16.0 16.5 17.0 17.5 18.0 18.5 19.0

M

S

Boys

Girls

Boys

Girls

Boys

Girls

71.94 72.00 72.07 72.12 72.17 72.20 72.22 72.22 72.21 72.18 72.13 72.08 72.03 71.96 71.90 71.84 71.77 71.71 71.66 71.60 71.55 71.51 71.46 71.41 71.36 71.30 71.24

72.19 72.01 71.81 71.61 71.42 71.24 71.11 71.03 70.99 70.98 70.98 70.97 70.97 70.98 71.00 71.08 71.21 71.35 71.45 71.50 71.49 71.45 71.40 71.33 71.25 71.17 71.08

15.17 15.26 15.38 15.54 15.74 15.94 16.12 16.31 16.52 16.76 17.02 17.29 17.58 17.89 18.21 18.55 18.94 19.36 19.77 20.13 20.44 20.71 20.94 21.16 21.35 21.53 21.68

15.12 15.20 15.32 15.48 15.66 15.85 16.05 16.29 16.58 16.92 17.28 17.61 17.90 18.18 18.49 18.87 19.26 19.62 19.91 20.11 20.25 20.34 20.42 20.49 20.56 20.63 20.71

0.104 0.102 0.102 0.102 0.104 0.107 0.110 0.113 0.116 0.119 0.122 0.124 0.125 0.126 0.125 0.123 0.121 0.118 0.116 0.114 0.112 0.112 0.113 0.114 0.117 0.120 0.124

0.098 0.100 0.103 0.106 0.109 0.112 0.117 0.121 0.126 0.130 0.134 0.137 0.139 0.139 0.137 0.133 0.127 0.121 0.116 0.111 0.107 0.105 0.102 0.100 0.098 0.095 0.093

463


464

Table 3 Smoothed measure of skewness (L), median (M) and coef®cient of variation (S) values for percentage body fat mass (PFM) at six-monthly intervals L Age (y) 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0 10.5 11.0 11.5 12.0 12.5 13.0 13.5 14.0 14.5 15.0 15.5 16.0 16.5 17.0 17.5 18.0 18.5 19.0

M

S

Boys

Girls

Boys

Girls

Boys

Girls

71.14 71.08 71.02 70.96 70.89 70.84 70.79 70.74 70.69 70.64 70.60 70.56 70.53 70.50 70.47 70.46 70.44 70.43 70.42 70.42 70.42 70.42 70.42 70.42 70.52 70.42 70.42

70.60 70.56 70.53 70.49 70.45 70.42 70.38 70.34 70.31 70.27 70.23 70.20 70.16 70.12 70.09 70.05 70.01 0.02 0.06 0.10 0.13 0.17 0.21 0.24 0.28 0.32 0.35

12.01 11.76 11.69 11.86 12.33 12.84 13.24 13.70 14.26 14.82 15.30 15.58 15.51 15.01 14.09 13.06 12.28 11.88 11.71 11.64 11.64 11.68 11.80 12.00 12.26 12.53 12.77

14.03 14.00 14.09 14.30 14.62 15.01 15.47 15.98 16.51 17.05 17.55 18.02 18.47 18.90 19.35 19.83 20.32 20.79 21.23 21.61 21.91 22.15 22.34 22.50 22.63 22.72 22.78

0.232 0.243 0.256 0.272 0.288 0.304 0.320 0.334 0.347 0.359 0.370 0.380 0.390 0.399 0.406 0.410 0.412 0.412 0.409 0.406 0.403 0.401 0.400 0.401 0.403 0.406 0.411

0.245 0.255 0.264 0.272 0.279 0.285 0.289 0.292 0.294 0.295 0.295 0.293 0.289 0.282 0.275 0.266 0.256 0.246 0.237 0.228 0.221 0.215 0.210 0.205 0.200 0.194 0.188

Figure 1 Percentiles for body mass index (BMI) derived from 1276 boys (left panel) and 1278 girls.

the development of PFM in the peripubertal period was focused on pubertal stages, boys showed a sharp decrease in PFM from the late prepubertal state (median 14.5 (interquartile range 8.9) %) to a nadir at Tanner Genital stage 4 (10.3 (6.8) %). In the last age group of post-pubertal (G5) boys, median PFM was 12.5 (8.1) %. In the girls, the steady increase of PFM with age in the prepubertal period was transi-

ently decreased around Tanner Breast stage 3 (19.9 (7.6) %) and 4 (20.1 (5.9) %), followed by a continued rise in the fully mature females (B5) to a maximum at age 19 years (23.0 (5.7) %). In order to assess the usefulness of BMI as an index of obesity, we evaluated the relationship between BMI and PFM. While BMI was closely correlated with PFM in girls (r 0.84, P < 0.0001), a rather loose


465

Figure 2 Percentiles for percentage body fat mass (PFM) derived from 1276 boys (left panel) and 1278 girls (right panel) aged 6 ±19 y.

Figure 3 Comparison of 85th body mass index (BMI) percentile estimates in several international reference studies.

association was present in boys (r 0.54, P < 0.0001). The linear relationship between BMI and PFM in boys was restricted to the more obese subjects (PFM > 66th percentile) (r 0.58, P < 0.0001), whereas no association existed in the leaner two thirds of the population (r 0.01, NS). In girls, BMI predicted PFM equally well in individuals below (r 0.81, P < 0.0001) and above (r 0.80, P < 0.0001) the 66th percentile. Stepwise multiple regression analysis revealed that

the prediction of PFM from BMI could be signi®cantly improved in both genders by correcting for pubertal stage. In the boys, 65.8% of the total variance of PFM could be explained by the combination of BMI and Tanner G stage, compared to 29.4% by BMI alone. The best-®tting prediction equation was PFM [%] 2.16BMI [kg=m2]73.09 (G stage) 716.9. In the girls, puberty had only a minor independent


466

confounding effect on the relationship between BMI and PFM. Inclusion of Tanner B stage in the prediction formula PFM [%] 1.78*BMI [kg=m2]70.4* (B stage)712.2 improved R2 marginally albeit signi®cantly from 0.707 (BMI alone) to 0.714. Since BMI is more commonly used as a screening variable to identify subjects at risk for obesity rather than to precisely predict PFM, we evaluated the sensitivity and speci®city of this index in discriminating obese from non-obese subjects as identi®ed by skinfold-derived PFM assessment. Different PFM and BMI cut-off percentile values were arbitrarily chosen to de®ne obesity. The results are shown in Table 4. BMI was of limited sensitivity and speci®city in identifying obese subjects; for example, using the 75th BMI percentile as cutoff value only 82% of the 15% most obese subjects were correctly identi®ed. At this cutoff level, 15% of the non-obese population were classi®ed as obese. The sensitivity and speci®city of PFM prediction did not differ between the genders at any chosen cut off value. Besides differences in the developmental pattern of total body fat acquisition, the genders also diverged in the pubertal period regarding regional fat distribution. As illustrated in Figure 4, the trunk-to-extremity skinfold ratio increased steadily, with similar mean values in either gender, in the prepubertal period. In the peripubertal age range, a steep increase was observed in the boys which was steepest around Tanner G stage 4 but continued in the post-pubertal period. In contrast, the adult ratio was attained already at Tanner B stage 3 in girls. In order to assess possible differences in relative obesity related to the socioeconomic background, calculated SDS values of BMI and PFM were compared between age-matched groups of children attending different types of German secondary schools: the `Hauptschule' (mainly attended by children from lower social strata), the `Realschule' (mixed, predominantly intermediate strata) and the `Gymnasium' (middle to upper social strata). While no differences in the obesity parameters were observed for boys, girls in a `Gymnasium' were signi®cantly less obese as judged by either BMI (70.18 0.84 SDS) or PFM (70.21 0.92 SDS) than girls attending a `Hauptschule' (BMI:

Table 4 Sensitivity and speci®city of body mass index (BMI) percentiles in identifying different degrees of obesity as de®ned by percentage body fat mass (PFM) percentiles. Values indicate sensitivity/speci®city (1 100%) BMI 66 75 85 90 95

PFM 75

PFM 85

PFM 90

PFM 95

0.81/0.81 0.70/0.90 0.49/0.96 0.36/0.98 0.20/0.99

0.91/0.75 0.82/0.85 0.65/0.94 0.51/0.97 0.31/0.99

0.94/0.72 0.88/0.82 0.76/0.91 0.64/0.96 0.42/0.98

0.95/0.69 0.89/0.78 0.81/0.88 0.73/0.93 0.55/0.97

Figure 4 Gender-speci®c patterns of trunk-to-extremity skinfold ratio during childhood. Values are mean s.e.m. Annual means comprise only prepubertal subjects until age 10.5 y and only postpubertal subjects (G5=B5) from age 15.5 y onward. In the peripubertal age range, mean trunk to extremity ratios for groups with Tanner pubertal stages 2, 3, and 4 respectively are plotted against the mean age of the subjects within each group.

0.44 1.15 SDS; P < 0.01; PFM: 0.44 1.18 SDS, P < 0.01) or a `Realschule' (BMI: 0.33 0.99 SDS; P < 0.05; PFM: 0.20 0.93 SDS, P < 0.05).

Discussion This study provides current reference values for BMI and PFM as indices of obesity in an urban population of schoolchildren and adolescents in Germany. Moreover, we evaluated the usefulness of the BMI as a predictor of relative obesity, and assessed the genderspeci®c patterns of regional fat distribution across childhood. Several recent methodological advances were utilized in the analysis of this dataset. First, the use of the LMS method as introduced by Cole and Green23 enabled us to construct smooth percentile curves from our cross-sectional data, and permitted calculation of precise s.d. scores despite the skewed distribution of obesity in the normal population. Secondly, a problem in estimating percentage body fat from skinfold measurements in children arises from the agedependent density of the fat-free body mass (FFM),24 which partially invalidates prediction equations derived from hydrodensitometry assuming a constant FFM density. A possible solution to this problem has been proposed by Slaughter et al18 who developed a set of gender- and age-speci®c PFM prediction equations accounting for the variability of FFM density in childhood by application of a multicomponent model


of body composition. The Slaughter equations have been successfully cross-validated in another pediatric population, with standard errors of estimate in the range of 4%.25 Thus, we are con®dent that the use of the Slaughter equations in this study permitted a suf®ciently accurate estimation of PFM. With respect to the reproducibility of the skinfoldderived PFM measurements, we calculated CVs of 2.4% intraindividually and 5.8% interindividually for the two skilled observers that performed all the measurements in this study. In absolute terms, this means that 95% of repeated measurements with an initial PFM value of 20% could have been expected to fall between 19 and 21% with a single observer, and between 17.7 and 22.3% with a different observer. This level of reproducibility is certainly acceptable for screening studies, but would limit the sensitivity of individual longitudinal monitoring. The sample studied here represented almost 20% of the total population of schoolchildren and adolescents in the Heidelberg area. We were not permitted to obtain any socioeconomic information apart from voluntary statements by the parents, which almost certainly would have introduced a participation bias. Restricting the socioeconomic assessment to the monitoring of the (social strata-related) attended school type, we were able to obtain a 96% participation rate. Since the selection of schools re¯ected the distribution of school types in Heidelberg, the socioeconomic mix in the study population should be largely representative of the local population. In relation to the total population of the state of Baden-WuÈrttemberg or Germany as a whole, Heidelberg is characterized by a slightly higher proportion of upper middle class families with an academic background. Thus, in terms of socioeconomic pro®les, the population studied here may be a better re¯ection of a modern urban population in Mid-Europe than of the German population on a national level. In view of the marked cultural differences and secular trends in the prevalence of obesity in different Western countries,1±4 it was of interest to compare our BMI percentile curves with other published international reference data. Our results suggest that the French BMI percentiles,11 commonly applied in studies of German children, considerably overestimate the current prevalence of obesity in urban areas in Germany. The observed difference, which, with respect to the 85th percentile, averaged 0.28 s.d. units across childhood, may be explained by socioeconomic or ethnocultural differences, a continued secular trend (French data collected in 1956±1979), or by a combination of these factors. In contrast, the North American and Italian reference data9,10,14 suggest a higher prevalence of obesity as compared to the German children. Since the Italian data were collected almost simultaneously with the German study, and the prevalence of childhood obesity has even increased in the North American population in the two decades since the NHANES I study,5 the differences can only

be explained by sociocultural and ethnic differences. The smallest differences were observed for the comparison with the Swedish and British percentiles, countries with socioeconomic conditions similar to Germany.12,13 Interestingly, the socioeconomic background, as far as interpretable from the attended school type, was a determinant of obesity in this study at least in girls. Girls from upper social strata tended to be leaner than their peers from a less favoured background. This observation is in keeping with ®ndings in North American children, in whom a low socioeconomic and educational level were identi®ed as important risk factors for childhood obesity.26 Thus, the classical association of malnutrition with a low socioeconomic status is apparently reversed in Western societies by a trend towards an inverse relationship between socioeconomic strata and relative obesity. While a steady increase in BMI with age was observed in both genders, the accumulation of body fat showed a distinct sexual dimorphism. PFM increased in boys until approximately the age of 13 y, decreased during puberty with a minimum at pubertal stage 4, and increased again slightly in the post-pubertal years. In girls, PFM also increased during the pubertal period, but the rise in relative fat mass was, as far as this can be interpreted from a cross-sectional study, temporarily reduced during mid puberty. Longitudinal studies of skinfold thicknesses have conclusively demonstrated the existence of a pubertal `dip' of fat accumulation,27,28 which is more marked in boys, but also present in girls. The `dip' coincides with the peak of the pubertal growth spurt. As the pubertal growth spurt is associated with a potentiation of sex steroid release and with a transient increase of endogenous growth hormone secretion, it is tempting to speculate that the transient diminution of fat accumulation during puberty re¯ects lipolytic effects of these hormones.29 Of course, it should be kept in mind that this study was strictly cross-sectional so that our charts are of limited value in the longitudinal monitoring of individual patients during puberty. However, the reported gender- and pubertal stage-speci®c PFM data should be helpful when body composition is to be screened in peripubertal populations. Our results have some bearing on the controversial issue of the usefulness of BMI as a measure of obesity. The BMI concept is based on the assumption that height2 is a measure of lean body mass, and the ratio of weight to height2 consequently indicates relative obesity. While this notion has been shown to be roughly correct in adults,30 the relationship between BMI and relative obesity is more complex in children since, as discussed above, the relative contributions of lean and fat mass to body weight are perturbed in a gender-speci®c manner around puberty. During mid to late puberty, PFM transiently decreases in boys while (lean) body mass is continuously growing. Consequently, the close positive asso-

467


468

ciation between BMI and PFM observed in prepubertal boys is disrupted or even reversed in the pubertal age range. We were able to reduce the residual variability in the prediction of PFM by BMI in boys, by more than two thirds by including pubertal status in the prediction equation. In girls, this phenomenon is much less pronounced, and a more constant relationship between BMI and PFM exists. Another confounder of the predictive value of BMI for PFM in boys, at least is the degree of obesity itself. While the degree of obesity was predicted reasonably well by BMI in the upper third of the PFM percentiles, no relationship between BMI and PFM was detected in the leaner two thirds of the male population. Thus, the BMI does not appear to be an appropriate surrogate marker of percentage fat when studying non-obese subjects. Realizing the conceptual limitations of BMI as a predictor of relative obesity in children and adolescents, we sought to evaluate the statistical power of using the BMI percentiles to detect obese subjects in our population. Our results demonstrate a moderate sensitivity and speci®city of BMI to identify subjects with different degrees of relative obesity (Table 4). For example, in order to detect the 15% most obese subjects (PFM greater than 85th percentile) with a sensitivity greater than 90%, the 66th BMI percentile would need to be chosen to de®ne subjects at risk for obesity. At this level of stringency, the false-positive error rate would be as high as 25%. From these results we conclude that due to the low sensitivity of BMI, a low (75th or even 66th) cutoff percentile should be chosen when BMI is to be used as a parameter to screen obesity. The pattern of regional fat distribution, particularly central fat accumulation, is an important independent predictor of cardiovascular risk in adults.31 Our analysis of the trunk-to-extremity skinfold ratio, complementing previous studies of fat distribution in childhood,32,33 suggests that the typical adult fat distribution pattern is attained during early puberty in girls, but not before the end of the second decade of life in boys. This explains why relatively low intraindividual correlations between childhood and adult trunk-to-extremity skinfold ratios have been reported.32

Conclusion The sample of German children and adolescents studied here, showed BMI values higher than published French, similar to British and Swedish, but lower than Italian and North American reference data. The sensitivity and speci®city of BMI to predict relative obesity in children and adolescents is limited. The gender-speci®c pattern of fat accumulation during childhood is characterized by an almost steady

increase in relative adiposity in girls, but a transient decrease in relative fat mass in late pubertal boys. Moreover, body fat is continuously redistributed across childhood from the extremities to the trunk in boys, while the adult pattern of regional fat distribution is attained during early puberty in girls.

Acknowledgements

We are grateful to all local and district school authorities for approval and active organizational support of the study. The help of Mrs Ina Kattwinkel with performing a large part of the anthropometric measurements is appreciated. We also thank Dr T.J. Cole for kindly providing the LMS analysis software.

References

1 Gortmaker SL, Dietz WH, Jr, Sobol AM, Wehler CA. Increasing obesity in the United States. Am J Dis Child 1987; 141: 535±540. 2 Harlan WR, Landis R, Flegal KM, Davis CS, Miller ME. Secular trends in body mass in the United States 1960 ±1980. Am J Epidemiol 1988; 128: 1065±1074. 3 Gulliford MC, Rona RJ, Chinn S. Trends in body mass index in young adults in England and Scotland from 1973 to 1988. J Epidemiol Commun Health 1992; 46: 187±190. 4 Cernerud L. Height and body mass index of seven-year-old Stockholm schoolchildren from 1940 to 1990. Acta Paediatr 1993; 82: 129±135. 5 Freedman DS, Srinivasan SR, Valdez RA, Williamson DF, Berenson GS. Secular increases in relative weight and adiposity among children over two decades: the Bogalusa Heart Study. Pediatrics 1997; 99: 420±426. 6 Power C, Lake JK, Cole TJ. Measurement and long-term health risks of child and adolescent fatness. Int J Obes 1997; 21: 507±526. 7 Must A, Jacques PF, Dallal GE, Bajema CY, Dietz WH. Longterm morbidity and mortality of overweight adolescents. N Engl J Med 1992; 327: 1350±1355. 8 Srinivasan SR, Bao W, Wattigney WA, Berenson GS. Adolescent overweight is associated with adult overweight and related multiple cardiovascular risk factors: the Bogalusa Heart Study. Metabolism 1996; 45: 235±240. 9 Must A, Dallal GE, Dietz WH. Reference data for obesity: 85th and 95th percentiles of body mass index for children and adolescents. Am J Dis Child 1991; 145: 259±263. 10 Must A, Dallal GE, Dietz WH. Reference data for obesity: 85th and 95th percentiles of body mass index (wt=ht2)Ða correction. Am J Clin Nutr 1991; 54: 773. 11 Rolland-Cachera MF, Cole TJ, SempeÂ M, Tichet J, Rossignol C, Charraud A. Body mass index variations: centiles from birth to 87 years. Eur J Clin Nutr 1990; 45: 13±21. 12 Cole TJ, Freeman JV, Preece MA. Body mass index reference curves for the UK. Arch Dis Child 1995; 73: 25±29. 13 Lindgren G, Strandell A, Cole T, Healy M, Tanner J. Swedish population reference standards for height, weight and body mass index at 6 to 16 years (girls) or 19 years (boys). Acta Paediatr 1995; 84: 1019±1028. 14 Luciano A, Bressan F, Zoppi G. Body mass index reference curves for children aged 3±19 years from Verona, Italy. Eur J Clin Nutr 1997; 51: 6±10. 15 Marshall JD, Hazlett CB, Spady DW, Conger PR, Quinney HA. Validity of convenient indicators of obesity. Hum Biol 1991; 63: 137±153. 16 HannanWJ,WrateRM,CowenSJ,Freeman CP.Body mass index as an estimate of body fat. Int J Eat Disord 1995; 18: 91±97.


17 Warner JT, Cowan FJ, Dunstan FD, Gregory JW. The validity of body mass index for the assessment of adiposity in children with disease states. Ann Hum Biol 1997; 24: 209±215. 18 Slaughter MH, Lohman TG, Boileau RA, Horswill CA, Stillman RJ, van Loan MD, Bemben DA. Skinfold equations for estimation of body fatness in children and youth. Hum Biol 1988; 60: 709±723. 19 Georgi M, Kattwinkel I. Normwerte von KoÈrpergroÈûe, KoÈrpergewicht und KoÈrperzusammensetzung bei Kindern und Jugendlichen im Groûraum Heidelberg unter BeruÈcksichtigung der bioelektrischen Impedanz-Analyse. Medical Thesis, University of Heidelberg, Germany, 1994. 20 Georgi M, Schaefer F, WuÈhl E, SchaÈrer K. Body height and weight in healthy schoolchildren and adolescents in the Heidelberg region. Monatsschr Kinderheilkd 1996; 144: 813±824. 21 Cameron N. The measurement of human growth. Croom Helm Ltd: London and Sidney, 1984. 22 Tanner JM. Growth at adolescence (2nd edn.) Blackwell Scienti®c Publications: Oxford, 1962. 23 Cole TJ, Green PJ. Smoothing reference centile curves: the LMS method and penalized likelihood. Stat Med 1992; 11: 1305±1319. 24 Boileau R, Lohman T, Slaughter M, Ball T, Going S, Hendrix M. Hydration of the Fat-free body in children during maturation. Hum Biol 1984; 56: 651±666. 25 Janz KF, Nielsen DH, Cassady SL, Cook JS, Wu YT, Hansen JR. Cross-validation of the Slaughter skinfold equations for children and adolescents. Med Sci Sports Exerc 1993; 25: 1070±1076.

26 Dietz WH, Bandini LG, Gortmaker S. Epidemiologic and metabolic risk factors for childhood obesity. Klin PaÈdiatr 1990; 202: 69±72. 27 Gasser T, Ziegler P, Kneip A, Prader A, Molinari L, Largo RH. The dynamics of growth of weight, circumferences and skinfolds in distance, velocity and acceleration. Ann Hum Biol 1993; 20: 239±259. 28 Gasser T, Kneip A, Ziegler P, Molinari L, Prader A, Largo RH. Development and outcome of indices of obesity in normal children. Ann Hum Biol 1994; 21: 275±286. 29 Rogol AD. Growth at puberty: interaction of androgens and growth hormone. Med Sci Sports Exerc 1994; 26: 767±770. 30 Garrow JS, Batra SW. Quetelet's index (W=H2) as a measure of fatness. Int J Obes 1985; 9: 147±153. 31 Kannel WB, Cupples LA, Ramaswami R, Stokes J. 3rd Regional obesity and risk of cardiovascular disease; the Framingham Study. J Clin Epidemiol 1991; 44: 183±190. 32 Rolland-Cachera MF, Bellisle F, Deheeger M, Pequignot F, SempeÂ M. In¯uence of body fat distribution during childhood on body fat distribution in adulthood: a two-decade follow-up study. Int J Obes 1990; 14: 473±481. 33 Taylor RW, Cannan R, Gold E, Lewis-Barned NJ, Goulding A. Regional body fat distribution in New Zealand girls aged 4±16 years: a cross-sectional study by dual energy X-ray absorptiometry. Int J Obes 1996; 20: 763±767.

469