Body Mass Index reference curves for children aged 3 19 years from Verona, Italy

Transcription

1 European Journal of Clinical Nutrition (1997) 51, 6±10 ß 1997 Stockton Press All rights reserved 0954±3007/97 $12.00 Body Mass Index reference curves for children aged 3 19 years from Verona, Italy A Luciano 1, F Bressan 2 and G Zoppi 1 1 Chair of Paediatrics, University of Verona and Paediatric Division, City Hospital of Verona; and 2 Institute of Statistics, University of Verona Objective: To compile curves for Body Mass Index (BMI) for Italian children and adolescents. Design: Cross sectional study. Setting: All primary and secondary schools of Verona, Italy between October 1986 and January Subjects: males and females children, aged 3 19 y. Methods: Weight and height were measured using Salus balances, and age in days was calculated between the date of measurement and that of birth: centiles of BMI by age were calculated by the LMS method of Cole (1990). Results: The centiles obtained were similar to those obtained in UK by Cole et al, Compared to Cachera s data for France and Hammer s for USA, our BMI values are higher, though closer to the American than the French ones. Sponsor: Regione Veneto, Italy. Descriptors: Body Mass Index; centiles; overweight; underweight; childhood; adolescence Introduction Body composition can be assessed by either elaborate methods or anthropometry. The methods generally used for estimating adiposity either have high costs and high precision such as DEXA, TOBEC, K40, RMN, etc., or low costs and low precision such as anthropometry. Anthropometric measurements are used either directly or as indices or in regression equations. Indirect measures of adiposity, such as skinfold thickness and weight for height indices, are often used as proxy measures of adiposity or as criteria for defining obesity. One such indicator, Quetelet s Index (Wt/ Ht 2 ) (Quetelet, 1871), also known as Body Mass Index (BMI), has been shown to be more valid than Wt/Ht or Wt/ Ht 3 in children (Rolland-Cachera et al, 1982), and has been adopted in adults (Garrow and Webster, 1985). Obesity is defined as an excess of fat, and the obese child runs an increase risk of becoming an obese adult (Dietz, 1986). Many studies have established that obesity is associated with metabolic disturbances, such as hyperinsulinemia, hyperlipemia, cardiovascular disease and cancer (Abraham et al, 1971; Le Marchand et al, 1988). Body composition changes with the stages of growth, adulthood and ageing. As opposed to traditional growth charts, the BMI curve reflects the real changes of the child s body shape and fatness during growth. Recently, Cole (1990) recommended that BMI be used for all ages. In fact he claims that BMI is well correlated with body fat at all ages. It has been observed, however, that the distribution of BMI Correspondence: Dr A Luciano, Divisione di Pediatria, Cattedra de Pediatria dell Universita, Ospedale Civile Maggiore, Piazzale Stolifani 1, Verona, Italia. Received 7 July 1996; revised 26 August 1996; accepted 6 September 1996 varies considerably with age (Rolland-Cacherà et al, 1991). Rolland-Cachera et al in France, and recently Cole (Cole et al, 1995) in the UK have shown that centiles and standard deviation (Z) scores of BMI can be obtained if Cole s LMS method (Cole, 1990) is used, where L, M and S are respectively the age-dependent exponent, mean and coefficient of variation. Sufficient data are therefore required to estimate L, M and S for each age group, say at least 100 of each age group. In this study of our town s population, the number of subjects was high, thus ensuring that the data could obtain valid centiles and related curves through Cole s method, which at the present time is the most suitable for our purposes. Methods The data used to construct the curves were based on measurements obtained from all the primary and secondary schools in the Verona area. The data concerning the ages 3 19 y were obtained in the period October 1986 January A questionnaire to collect personal and family informations was completed for each subject. Weight and height were measured during interviews using Salus (Milan) balances. Weight was measured either in under clothes or adjusted for clothing weight, while height was measured with the head in the Frankfort plane. The number used in this study was ( males and females). The age in days was calculated between the date of measurement and that of birth. They were subdivided into quarter-year age groups, each group containing over 100 subjects (an average of 322 per group). Each computation always took into consideration the correct value of the mean age (expressed in months for case consultation) for subsets of data. Standard deviation was corrected, taking

2 into consideration the effect of different ages. Values of Wt/Ht 2 corresponding to the various centiles were obtained for each age and sex group, using the Cole s LMS method. The distribution of Wt/Ht 2 is very skew, so that the Z- scores and centiles cannot be calculated from the mean and standard deviation assuming a normal distribution. The BMI reference curves for children LMS method uses three quantities, the power (L), median (M) and coefficient of variation (S) calculated for each group; it can determine the value that best approximates the median (50th centile of the distribution) and gives the minimum S. Though the L, M and S values obtained display considerable oscillation, the LMS method allows 7 Table 1 Values of mean and SD, L (Box Cox Power), M (Generalized Mean), S (Coefficient of Variation) for males and females MALES FEMALES Months N Mean DS L M S Months N Mean DS L M S ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % % % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ % ÿ %

3 BMI reference curves for children 8 one to draw smooth curves of L, M and S. The smooth spline of the S-Plus programme (Statistical Science, 1994) was used to determine the age distribution of the average BMI and the relative centiles. The LMS method also converts Wt/Ht 2 values in individuals to Z-scores and centiles. The formula to obtain the Z-scores is given by: Z ˆ Q=M L ÿ1š=ls where Z is the Z-score, Q is the observed Wt/Ht 2, L the power, M the median and S the coefficient of variation for each age and sex group. The Z-scores, once known, can be converted to a centile using normal distribution tables. Results Figures 1 and 2 show the changes in BMI for males and females aged 3 19 y. From the curves we note that BMI gradually increases from around the ages of 6 up to 14, at which point female weight diminishes slightly and then plateaus; that of the males, on the other hand, increases steadily though slowly. At 3 y the mean BMI value for males and females is c. 16; until 12 y it remains approximately equal in both sexes; at 19 y it is 22 for males and 19 for females. With the power (L) of 1 indicating a normal distribution, in Figure 3 we note variations ranging from ÿ0.6 to ÿ2.6 during the pediatric age under study. As for the coefficient of variation (S) of Wt/Ht 2 (Figure 3), the variability ranges from 6.7 to 15%. Table 1 shows the Figure 2 Body mass index centiles for females 3 19 y. subject groups analysed using the data for age (months), number of subjects, the mean, standard deviation and the exact values of L, M and S. Figure 1 Body mass index centiles for males 3 19 y. Discussion This study aims to describe the changes in BMI in the pediatric age. The data obtained give a picture of the situation for the ages ranging between 3 and 19 y. Crosssectional studies can provide a record of the nutritional status for a precise period and for a specific population. In our case we measured children and adolescents of the Verona area through their growth in weight and height. Owing to the important strategic position it occupies (for historical and geographical reasons), Verona has been the frequent target of immigration from the various Italian regions. Recent statistics, drawn up at more or less the same time as the collection of our data, showed that only 55% of the resident population was born in Verona, while the remaining 45% is from various parts of Italy (Annuario Statistico, 1989). Furthermore, many of those born in Verona in their turn come from families that had previously immigrated, so we may reasonably claim that our data are sufficiently representative of the Italian population as a whole. Though children with parents of other nationalities were also used, their number was insufficient to modify the results. Whereas height has a normal distribution, with mean and standard deviations, weight is subject to numerous

4 BMI reference curves for children 9 Figure 3 Curves of L (Box Cox power), M (generalized mean) and S (generalized coefficient of variation) for males and females.

5 10 BMI reference curves for children influences and has no normal distribution. The values obtained therefore needed to be transformed into BMI values, in accordance with the suggestions of Box and Cox (Box and Cox, 1964) and the LMS method perfected by Cole, so as to obtain the centiles and respective curves. In our application of Cole s method we used a large number of subjects and three-month age groups, for it is indispensable that the centiles should give a picture of the real trend of BMI in the pediatric age. In fact, since variations between subjects in the stages of growth are known to be considerable, an accurate representation of BMI requires the groups to be both well populated and of homogeneous age. For though the excellence of Cole s method is unquestionable, its validity depends on the age groups being numerous. Each group should have no fewer than 100 subjects. The data obtained from the measurements of all the children and adolescents attending the schools of our town show both the variations in BMI during infancy and adolescence and the difference between males and females above all in the prepuberal period (with a greater increase in BMI in males). We should also note the decrease in female BMI around the age of 16 y, contrasting with the continued increase in males, a phenomenon particularly linked to the increase of muscular mass. Comparing our values with those of the French study (Cachera), there is initial agreement at the age of 3 y in both males and females. Subsequently the BMI values of our population are higher than the French ones (for example, the 10th, 50th and 90th centiles for the males of 10 y are 14.5, and 18.7 in France and 15.02, and in Italy). For the females until c. 6 y, the 3rd, 10th, 25th and 50th centiles are similar to those of the French population, whereas the other centiles begin to differ noticeably and increase with age right up to the upper age limit of our study. Another difference is the downward trend in our curves for 15 year-old females; in the French curves the trend is upward until 18 y, when a momentary reversal of trend is followed a subsequent increase. As for the males, after substantial agreement at the age of 3, our values gradually become higher, with a 50th centile almost always comparable to the French 75th, and all the other centiles comparable to the higher centile in the corresponding curves. In the period from 6 12 y, the 90th centile is even higher than the 97th centile of the French curves. On the other hand, our data are closer to those of Hammer (USA) (Hammer, 1991), above all for the adolescent period, where the differences vanish from the ages of y. Our data are more similar to those published by Cole (UK). Though dissimilarity in the results of the different studies is inevitable, we did not expect the differences (above all those with the French study) to be so great. The similar in general shape and values of BMI centiles to those for British children than the other BMI centiles may reflect the more recent data and the numbers of subject measurements by age and sex of ours and UK information. The results show some limits of BMI in estimating nutritional status: the greater the diversity of variables influencing body weight (such as race, diet, the period and methods of measurement), the greater the differences in the results. For the centiles and curves to be useful, the data must be collected correctly and be sufficiently abundant and representative of the age under study, particularly when one is dealing with the pediatric age. And the use of the curves requires a thorough awareness of each BMI value, which must be assessed together with other simple methods or indices, such as the measurement of skinfolds and/or circumference of limbs and waist-hips. Only in this way can we obtain a sufficiently close estimate of the real nutritional status of the subject. As regards BMI, the very nature of its data is such that it can be verified and corrected on subsequent occasions through the use of a carefully chosen sample. By this means one can evaluate changes in the population under study, describe nutritional status in well-defined periods of time and assess the factors responsible for change, with a view to taking steps against excess body weight, which, as we well know, is a risk factor in many adult pathologies. A final consideration: BMI curves show the different values of BMI in relation to age, representing one of the simplest methods for approaching an overweight subject. Acknowledgements The authors acknowledge Prof. Hugh Ward-Perkins (Lector of English, University of Verona) for revising the manuscript and thank the following physicians who performed the measurements in a program of health control in the schools of Verona: A. Bariani, A. Bagnani, M. B. Bertolin, M. B. Gattoni, A. Isolan, S. Minghetti, R. Palmeri, M. L. Placucci, C. Pasti, N. Randi, A. Ramini, P. Residori, A. Sacchetto, E. Schiera, E. Tellarini, G. B. Zivelonghi and F. Zenatelli. References Abraham S, Collins G, Nordisieck M (1971). Relationship of childhood weight status to morbidity in adults. HSMHA Health Rep. 86: Annuario StatisticoComune di Verona (1989). Serie IV N o 1 pag Ed Comune di Verona, Servizi Studi e Statistica. Box GEP, Cox DR (1964). An analysis of transformations. J R Statist Soc B. 26: Cole TJ (1990). The LMS method for constructing normalized growth standards. Eur J Clin Nutr. 44: Cole TJ, Freeman JV, Preece MA (1995). Body mass index reference curves for the UK, Arch Dis Child. 73: Dietz WH (1986). Prevention of childhood obesity. Pediatr Clin North Am. 33: Garrow JS, Webster J (1985). Quetelet s index (Wt/Ht 2 ) as a measure of fatness. Int J Obes. 9: Hammer LD et al (1991). Standardized percentile curves of body mass index for children and adolescents. AJDC. 145: Le Marchand I, Kolone LN, Earle ME, Mi M (1988). Body size at different periods of life and breast cancer risk. Am J Epidemiol. 128: Quetelet LA. Anthropometric ou measure des differentes facults de l homme. Brussels: C. Muquardt, Rolland-Cachera MF et al (1982). Adiposity indices in children. Am J Clin Nutr. 36: Rolland Cachera MF et al (1991). Body Mass Index variations: centiles from birth to 87 years. Eur J Clin Nutr. 45: Statistical Science (1994). S plus for Windows Version 3.2 Suppl, Seattle: StatSci, a division of MathSoft, Inc; page 9.52, page 9.45.