Relationship between anthropometric variables and lipid ... - Nature

International Journal of Obesity (1998) 22, 66±72 ß 1998 Stockton Press All rights reserved 0307±0565/98 $12.00

Relationship between anthropometric variables and lipid levels among school children: The Taipei Children Heart Study N-F Chu1,2,3, EB Rimm3,4, D-J Wang3, H-S Liou1 and S-M Shieh2 Department of 1 Public Health and 2 Internal Medicine, National Defense Medical Center, Taipei, Taiwan, ROC and Departments of 3 Epidemiology and 4 Nutrition, Harvard School of Public Health, Boston, MA, USA

OBJECTIVES: To evaluate the association between anthropometric parameters and lipid levels among Taiwanese school children. DESIGN AND METHODS: Using a probability-proportional-to size sampling and multi-stages sampling procedure, we sampled 1500 school children from 10 schools in Taipei city. Anthropometric parameters including body weight, body height, waist circumference, hip circumference and skinfolds were measured. Serum total cholesterol (CHOL), triglycerides (TG), high density lipoprotein-cholesterol (HDL-C), apolipoprotein A1 and B (ApoA1 and ApoB) were measured by standard methods, low density lipoprotein-cholesterol (LDL-C) and CHOL=HDL-C ratio were calculated by formula. RESULTS: We included in our analyses 1366 children (681 boys and 685 girls) with a mean age of 13.3 y (from 12 to 16 y) and with valid anthropometric and biochemical parameters. The boys had higher body height (P < 0.001) and larger body weight (P < 0.05), waist circumference (P < 0.01) and waist=hip ratio (WHR, P < 0.001) than the girls. However, the girls had larger skinfolds than the boys. After adjusting for age, girls had higher total CHOL, TG, HDL-C, LDL-C, ApoA1 and ApoB concentrations than boys. In general, TG was positively associated with most anthropometric parameters (except body height); a similar negative association between HDL-C and anthropometric variables was noted. After controlling, for age, cigarette smoking, alcohol drinking and puberty development, shorter body height was the strongest predictor of total CHOL, LDL-C and ApoB concentrations among boys. Although body mass index (BMI) was a signi®cant positive predictor (P < 0.01) of the CHOL=HDL-C ratio; skinfold measurements were the strongest anthropometric predictors of most lipid concentrations among boys. Among girls, we found WHR and BMI to be the strongest positive predictors of TG and ApoB level respectively (both P < 0.001), but skinfold measurements were best for predicting HDL-C, LDL-C, ApoA1 and the CHOL=HDL-C ratio. CONCLUSIONS: From this large study of school-age children from Taiwan, we found anthropometric parameters, such as body height, BMI or WHR, are adequate predictors of blood lipid levels; however, skinfold measurements are generally more strongly associated with lipid levels in both genders. Keywords: school children; anthropometric parameters; lipid levels

Introduction Over the past twenty years, the prevalence of obesity has increased in most Western countries and the Taiwan area. Obesity is associated with an increased risk of hypertension, diabetes mellitus, dyslipidaemia, atherosclerotic disease and coronary heart disease.1,2 Several studies have shown that obesity and fat distribution are associated with lipid or glucose related metabolic disorders.3±6 Even among children and adolescents, obesity is associated with high blood pressure, abnormal lipid and glucose metabolism, and with the development of hypertension, dyslipidaemia and diabetes mellitus later in life.7±10 Correspondence: Dr Nain-Feng Chu, Dept. of Public Health and Internal Medicine, National Defense Medical Center, Taipei, Taiwan, ROC Received 25 March 1997; revised 12 September 1997; accepted 23 September 1997

For epidemiological studies and in clinical practice, non-invasive anthropometric measurements are more practical for large-scale screening of children. Obesity and fat distribution are assessed by anthropometric parameters such as body mass index (BMI, weight=height2), waist=hip ratio (WHR) and skinfold measurements. Although circumference measurements are useful and valid tools to study central obesity and cardiovascular disease risk factors in younger and older adults11,12 the best anthropometric predictor among children is still controversial. In several,10,13±15 but not all studies5,16±18 of children and adolescents, skinfold indices are more strongly associated with cardiovascular risk factors than conventional anthropometric measurements such as BMI or WHR. These associations have not been thoroughly examined in Taiwan, a country with increasing rates of coronary heart disease and childhood obesity.12,19 Therefore, we examined 1366 school children from Taiwan, to evaluate the association between lipid levels and anthropometric measurements to ®nd the

Association between anthropometric variables and lipid levels N±F Chu et al

most predictive anthropometric parameters of blood lipid levels among school children.

Methods Study design and sampling method

We conducted a cross-sectional survey of children attending junior high school in Taipei city. This survey included the collection of demographic and lifestyle characteristics in addition to a 20 ml blood specimen. Using a probability-proportional-to size sampling method and multi-stage sampling procedure, we collected information on a sample of children from the 101,000 students enrolled in the 47 schools with more than 40 classes and from the 28,000 students enrolled in the 38 schools which have less than 40 classes in Taipei city. We ®rst sampled study schools, according to size; seven large and three small schools were selected. Secondly, we randomly chose six classes from each selected school. Finally, we sampled the study children from the selected class; 28 children from each large school class and 18 children from each small school class. The 10 schools (seven large and three small) and 60 classes (42 from large schools and 18 from small schools) sampled were proportional to the number of students in Taipei city. A total of 1500 school children (1176 from large schools and 324 from small schools) were sampled for this survey. Data collection

General information. All participants completed a structured questionnaire with the help of a welltrained research technician. The questionnaire included personal and family history of disease and general lifestyle characteristics including cigarette smoking, alcohol consumption, puberty development, usual physical activity and dietary intake. Anthropometric variables measurement. Anthropometric parameters were collected by two well-trained research technicians. Body weight was recorded to the nearest 0.1 kg using a standard beam balance scale with subjects barefoot and wearing light indoor clothing. Technicians recorded body height to the nearest 0.5 cm using a ruler attached to the scale. We calculated BMI as the ratio of body weight to body height squared and expressed as kg=m2. Waist circumference was measured at the distal third of the line from the xyphoid process to the umbilicus. Hip circumference was measured 4 cm below the anterior superior process of the iliac spine. We calculated WHR as the ratio of waist circumference divided by the hip circumference. We recorded arm circumference at the midpoint between the acromial and olecranon processes of the

scapula and the ulna, with the arm hanging relaxed at the subject's side. Skinfolds were measured to the nearest 0.1 mm by one well-trained research technician using the Lange skinfold caliper. The triceps skinfold was measured on the posterior aspect of the right arm, over the triceps muscle, midway between the lateral projection of the acromion process of the scapular and the inferior margin of the olecranon process of the ulna. The biceps skinfold was measured directly above the center of the cubital fossa, at the same level as the triceps skinfold and arm circumference. The subscapular skinfold was measured 1 cm below the lowest angle of the scapula and long axis of the skinfold on a 45 angle directed down and to the right side. Finally, the suprailiac skinfold was measured just above the iliac crest at the midaxillary line and the long axis following the natural cleavage lines of the skin and runs diagonally. For each skinfold, we recorded two measurements and used the average. Lipids and lipoproteins measurement. To reduce extraneous between-person variation, we conducted measurements and collected fasting blood from students who had consumed their usual dietary pattern during the previous three days. Children who had recently attended a holiday or family celebration were re-contacted several weeks later during the technician's follow-up visit to each classroom. Venous blood was collected and stored at 4 C and analyzed within two weeks. We measured serum total cholesterol (CHOL) using an esterase-oxidase method,20 triglycerides (TG) using an enzymatic procedure,21 and high density lipoprotein-cholesterol (HDL-C) by an enzymatic method involving magnesium precipitation22 with the Synchron CX5 analyzer (Beckman Instruments, Palo Alto, CA). Apolipoprotein A1 (ApoA1) and apolipoprotein B (ApoB) concentrations determined by nephelometric assay (Beckman Instruments, Palo Alto, CA).23 Each of these commercial assays was calibrated with the dedicated standard material offered by the manufacturer. Because no TG concentration was greater than 400 mg=dl and all samples were collected after a 12 h fast, we used the Friedewald's formula,24 to calculate low density lipoprotein-cholesterol (LDL-C) LDL-C [(total CHOL)7(HDLC)7(TG=5)]. We calculated the CHOL=HDL-C ratio (TCHR, total CHOL divided by HDL-C) as a marker of atherogenicity. Statistic analysis

Mean anthropometric values were age-adjusted and strati®ed by gender to avoid maturation differences due to aging or gender. We calculated Student t-test to evaluate the differences between boys and girls. We calculated the gender-speci®c Pearson's correlation coef®cient, to evaluate the relationship between anthropometric parameters and lipid concentrations

67


68

Table 1 Age-adjusted anthropometric characteristics of school children in Taipei City

Body height (cm) Body weight (kg) BMI Waist circumference (cm) Hip circumference (cm) WHR Arm circumference (cm) Triceps skinfold (mm) Biceps skinfold (mm) Supra-iliac skinfold (mm) Subscapluar skinfold (mm) Two SFs measurement (mm) Sum skinfolds (mm)

All

Male

Female

(n 1366)

(n 681)

(n 685)

Mean

Range

Mean

Range

Mean

Range

159.2 53.3 20.9 65.8 88.6 0.74 23.5 15.8 8.7 13.5 13.4 29.2 51.5

132.5±186.5 27±113 13.9±40.8 30.5±107.6 66.4±126.3 0.36±1.02 9.8±40.0 4.0±37.8 2.1±34.6 2.8±45.3 2.4±42.5 9.4±76.0 15.0±150.8

161.8 55.5 21.1 68.3 87.9 0.77 24.0 13.5 7.0 11.6 11.7 25.2 44.5

132.5±186.5 30±113 13.9±40.8 30.5±107.6 69.3±126.3 0.36±1.02 9.8±40.0 4.0±37.8 2.1±29.4 2.8±40.5 2.4±36.5 9.4±72.5 15.0±137.3

156.2*** 50.7* 20.7 63.0** 89.1 0.71*** 22.9 18.1*** 9.8* 15.5** 15.0** 33.0*** 58.3***

136.5±173.0 27±103 14.1±38.7 49.0±106.0 66.4±120.5 0.62±0.95 16.0±36.0 6.2±37.0 2.7±34.6 4.8±45.3 5.5±42.5 12.4±76.0 20.1±150.8

BMI body mass index; WHR waist=hip ratio. Two SFs Measurement triceps skinfold biceps skinfold. *P < 0.05; **P < 0.01; ***P < 0.001 when boys compared with girls.

in study subjects. As there is no clear de®nition of obesity,25,26 we conducted multivariate regression analyses combining the obese and non-obese children. Multiple regression analysis was used to evaluate the anthropometric parameters most predictive of the lipids and lipoproteins. We conducted modi®ed twostage stepwise regression analysis to select the most adequate anthropometric parameters for each model. In the ®rst stage, we divided the anthropometric parameters into three different groups. The ®rst group consisted of general obesity measures (body height, body weight and BMI), the second group included measures related to fat distribution (waist circumference, hip circumference and WHR) and the ®nal group included arm circumference and skinfold measurements. We regressed biochemical measures on each group of parameters. The P-value criterion for entry into these stepwise regression models was 0.1. The second stage of the analysis procedure was to include the variables from each of the three groups that remained in the model after the ®rst stage. Using the backward elimination regression method, the ®nal model included those anthropometric parameters still signi®cant at the P < 0.05 level. All regression analyses were adjusted for age, cigarette smoking, alco-

hol drinking and puberty development.

Results General characteristics of study children

Among the 1500 children sampled in this survey, we excluded 134 subjects who refused the survey protocol or had missing or incomplete data. The ®nal sample for analysis included 1366 children (681 boys and 685 girls) with a mean age of 13.3 y (from 12 to 16 y). The age-adjusted mean value and range of anthropometric, lipid and lipoprotein factors are shown in Table 1 and Table 2. In general, boys were taller, heavier, had larger waist circumferences and WHR and girls had larger skinfolds. We also found that girls had higher concentrations of lipids and lipoproteins than boys. Correlation between anthropometric and lipids variables

The Pearson correlation coef®cients between anthropometric, lipids and lipoproteins in boys and girls are

Table 2 Age-adjusted lipids and lipoproteins characteristics of school children in Taipei City

Cholesterol (mg=dl) Triglyceride (mg=dl) HDL-C (mg=dl) LDL-C (mg=dl) Apo-A1 (mg=dl) Apo-B (mg=dl) TCHR

All

Male

Female

(n 1366)

(n 681)

(n 685)

Mean

Range

Mean

Range

Mean

Range

156.4 73.5 54.4 87.3 125.2 83.4 3.01

78±324 24±348 15±130 4.8±229.6 46.6±216.0 37.1±212.0 1.14±7.54

151.5 70.3 53.6 83.9 121.9 79.8 2.98

79±263 24±292 15±115 4.8±208.2 52.2±185.0 37.1±192.0 1.14±7.54

161.4*** 77.1*** 55.3* 90.7*** 128.4*** 86.8*** 3.03

78±324 25±348 25±130 10.4±229.6 46.6±216.0 42.7±212.0 1.23±6.22

HDL-C high density lipoprotein-cholesterol; LDL-C low density lipoprotein-cholesterol; Apo-A1 apolipoprotein A1; Apo-B apolipoprotein B; TCHR total cholesterol to high density lipoproteincholesterol ratio. *P < 0.05; **P < 0.01; ***P < 0.001 when boys compared with girls.


69

Table 3 Age-adjusted Pearson correlation coef®cients between anthropometric and lipids variables in schoolboys (n 681) Chol Body Height Body Weight BMI Arm Circumference Triceps Skinfold Biceps Skinfold Supra-Iliac Skinfold Sub-Scapular Skinfold Two Skinfolds Sum of Skinfolds Waist Circumference Hip Circumference WHR

70.175*** 70.069 0.005 70.008 0.073 0.073 0.048 0.059 0.068 0.065 0.009 70.047 0.077*

TG

HDL-C

70.007 0.229*** 0.282*** 0.250*** 0.297*** 0.314*** 0.334*** 0.345*** 0.329*** 0.337*** 0.299*** 0.235*** 0.258***

70.113** 70.287*** 70.295*** 70.261*** 70.243*** 70.220*** 70.281*** 70.278*** 70.267*** 70.269*** 70.249*** 70.267*** 70.124**

LDL-C

Apo-A1

70.130*** 0.020 0.090* 0.065 0.131*** 0.114** 0.115** 0.122** 0.130*** 0.126*** 0.065 0.031 0.082*

Apo-B

70.039 70.144*** 70.155*** 70.138*** 70.163*** 70.116** 70.127*** 70.126*** 70.150*** 70.140*** 70.114** 70.149*** 70.026

70.106** 0.133*** 0.215*** 0.202*** 0.241*** 0.222*** 0.248*** 0.241*** 0.248*** 0.250*** 0.202*** 0.152*** 0.186***

TCHR 0.002 0.257*** 0.317*** 0.258*** 0.326*** 0.294*** 0.341*** 0.338*** 0.341*** 0.341*** 0.263*** 0.241*** 0.181***

Chol total cholesterol; TG triglyceride; HDL-C high density lipoprotein-cholesterol; LDL-C low density lipoprotein-cholesterol; Apo-A1 apolipoprotein A1; Apo-B apolipoprotein B; TCHR total cholesterol to high density lipoprotein-cholesterol ratio; BMI body mass index; WHR waist=hip ratio. *P < 0.05; **P < 0.01; ***P < 0.001. Table 4 Age-adjusted Pearson correlation coef®cients between anthropometric and lipids variables in schoolgirls (n 685) Chol Body Height Body Wight BMI Arm Circumference Triceps Skinfold Biceps Skinfold Supra-Iliac Skinfold Sub-Scapular Skinfold Two Skinfolds Sum of Skinfolds Waist Circumference Hip Circumference WHR

70.026 0.046 0.037 0.040 0.027 0.053 0.017 0.023 0.026 0.030 0.041 0.018 0.049

TG 70.035 0.072 0.094* 0.104** 0.075 0.107** 0.110** 0.115** 0.100** 0.109** 0.127*** 0.055 0.150***

HDL-C

LDL-C

Apo-A1

Apo-B

TCHR

0.012 70.172*** 70.201*** 70.196*** 70.215*** 70.175*** 70.221*** 70.211*** 70.222*** 70.223*** 70.207*** 70.168*** 70.160***

0.032 0.115** 0.114** 0.113** 0.116** 0.116** 0.098* 0.098* 0.111** 0.114** 0.113** 0.087* 0.093*

0.045 70.045 70.069 70.066 70.079* 70.024 70.066 70.069 70.077* 70.066 70.037 70.030 70.024

70.030 0.194*** 0.230*** 0.214*** 0.203*** 0.197*** 0.200*** 0.217*** 0.219*** 0.220*** 0.210*** 0.168*** 0.164***

0.007 0.203*** 0.225*** 0.216*** 0.216*** 0.207*** 0.227*** 0.224*** 0.230*** 0.236*** 0.237*** 0.173*** 0.203***

Chol total cholesterol; TG triglyceride; HDL-C high density lipoprotein-cholesterol; LDL-C low density lipoprotein-cholesterol; Apo-A1 apolipoprotein A1; Apo-B apolipoprotein B; TCHR total cholesterol to high density lipoprotein-cholesterol ratio; BMI body mass index; WHR waist=hip ratio. *P < 0.05; **P < 0.01; ***P < 0.001.

presented in Table 3 and Table 4. In general, anthropometric parameters were positively correlated with TG concentrations (except body height) and negatively correlated with HDL-C. The skinfold measurements were better correlated with most of the lipid or lipoprotein variables (except CHOL in boys and Apo-A1 in girls) than body weight or circumferences. For example, among boys, the correlation between triceps skinfold (r 0.13) and LDL-C was stronger than body weight (r 0.02) or hip circumference (r 0.03); among girls, the correlation between triceps skinfold (r 70.22) and HDL-C was stronger than body weight (r 70.17) or hip circumference (r 70.17). Multivariates regression model to predict lipids and lipoproteins

Table 5 and Table 6 show the gender-speci®c multiple regression analyses of the relationship between anthropometric parameters and lipids, and lipoproteins concentrations after controlling for age, smoking status, alcohol drinking and puberty development. Among boys (Table 5), shorter body height is the strongest negative predictor of blood total cholesterol,

LDL-C and ApoB levels. Among boys, BMI was a signi®cant positive predictor of the CHOL=HDL-C ratio (P < 0.01) and waist circumference was a signi®cant negative predictor of the CHOL=HDL-C ratio (P < 0.05) after controlling for each other. With the exception of total cholesterol, the different site skinfold measurements were signi®cant predictors of all blood lipid and lipoprotein variables. Among girls (Table 6), none of the anthropometric parameters signi®cantly predicted blood total cholesterol levels. The WHR and BMI were the best positive predictors of TG and ApoB concentrations respectively (both P < 0.001). Similar to the boys, skinfold measurements were signi®cant predictors of ApoA1 (P < 0.001 for triceps and P < 0.05 for biceps), LDL-C (P < 0.01) and CHOL=HDL-C ratio (P < 0.001), and signi®cant negative predictors of HDL-C (P < 0.001).

Discussion Among this random sample of 1366 school age Taiwanese children, the mean body heights are


70

Table 5 Multiple regression analysis of relationship between anthropometric variables with lipids concentration in schoolboys (n 681) Dependenta variable Chol (mg=dl) TG (mg=dl) HDL-C (mg=dl) LDL-C (mg=dl) Apo-A1 (mg=dl) Apo-B (mg=dl) Chol=HDL-C ratio

Explanatory variables Body Height (cm) Subscapular Skinfold (mm) Supra-iliac Skinfold (mm) Body Height (cm) Triceps Skinfold (mm) Triceps Skinfold (mm) Body Height (cm) Sum Skinfolds (mm) BMI Waist Circumference (cm) Arm Circumference (cm) Sum Skinfolds (mm)

Partial Regression Coefficient (b) 70.73 1.89 70.51 70.688 0.550 70.590 70.413 0.210 0.070 70.022 70.046 0.015

P-value

Adjusted Model R-square

< 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.01 < 0.05 < 0.05 < 0.001

0.024 0.115 0.088 0.038 0.023 0.083 0.154

a All models were adjusted by age, cigarette smoking, alcohol drinking and sexual maturity. Chol total cholesterol; TG triglyceride; HDL-C high density lipoprotein-cholesterol; LDL-C low density lipoprotein-cholesterol; Apo-A1 apolipoprotein A1; Apo-B apolipoprotein B; BMI body mass index.

Table 6 Multiple regression analysis of relationship between anthropometric variables with lipids concentration in schoolgirls (n 685) Dependent a variable Chol (mg=dl)b TG (mg=dl) HDL-C (mg=dl) LDL-C (mg=dl) Apo-A1 (mg=dl) Apo-B (mg=dl) Chol=HDL-C ratio

Explanatory variables ± WHR Sum Skinfolds (mm) Biceps Skinfold (mm) Triceps Skinfold (mm) Biceps Skinfold (mm) BMI Sum Skinfolds (mm)

Partial regression coefficient (b) ± 115.72 70.143 0.73 70.901 0.714 1.49 0.009

P-value ± < 0.001 < 0.001 < 0.01 < 0.001 < 0.05 < 0.001 < 0.001

Adjusted Model R-square ± 0.026 0.040 0.010 0.011 0.042 0.052

a

All models were adjusted by age, cigarette smoking, alcohol drinking and sexual maturity. None of the anthropometric parameter was signi®cant to predict total cholesterol level in schoolgirls. Chol total cholesterol; TG triglyceride; HDL-C high density lipoprotein-cholesterol; LDL-C low density lipoprotein-cholesterol; Apo-A1 apolipoprotein A1; Apo-B apolipoprotein B; WHR waist=hip ratio; BMI body mass index.

b

156 cm and 161 cm, body weights are 51 kg and 57 kg, mean cholesterol levels are 161 mg=dl and 152 mg=dl and mean TGs are 77 mg=dl and 70 mg=dl for boys and girls, respectively. In this study, almost all anthropometric parameters were positively correlated with TG and CHOL=HDL-C ratio and negatively correlated with HDL-C and ApoA1 concentrations. Body height is negatively correlated with lipid and lipoprotein concentrations. No single anthropometric parameter could predict all lipid concentrations in children. Although BMI was a positive predictor of the CHOL=HDL-C ratio (in boys, after controlling for waist circumference, arm circumference and sum skinfolds) and ApoB (in girls) and WHR was a positive predictor of TG (in girls), skinfold measurements were the most consistent predictors of lipid or lipoprotein concentrations among school children. The limitations of a cross-sectional survey design, and the inherent measurement error in assessing anthropometric parameters and lipid concentrations, may have somewhat attenuated our results. However, children of this age group are rarely on special diet or exercise regimens and generally do not receive treatment for obesity or hyperlipidaemia. In our results,

although anthropometric variables were signi®cant predictors of lipids or lipoproteins concentrations in children, the variance explained by these anthropometric variables was small (less than 16%), which suggests that other genetic or lifestyle parameters are important in the determination of lipids and lipoproteins concentrations among children. During the last 20 years, intake of total energy, fat and cholesterol have increased steadily, concurrent with a more sedentary lifestyle in the Taiwan area.19 These trends toward adverse lifestyles and dietary patterns, may explain the increase in the prevalence of obesity and hyperlipidaemia in Taiwan, as it has in western countries. We found little difference in the distribution of anthropometric and lipid concentrations among school children in Taiwan, compared with those in western countries.26±30 Overweight and hyperlipidaemia have become important public health issues for children and the general population in Taiwan. Obesity and hyperlipidaemia are important risk factors for hypertension, diabetes mellitus and atherosclerotic heart disease. Because weight gain during the adult lifetime increases the risk of cardiovascular


disease,31,32 early detection and prevention of obesity may be the best treatment. Routine physical check-ups which include biochemical studies are indicated for the population with high cardiovascular risk (for example, obesity, hypertension, cigarette smoking, physical inactivity and family history of diabetes mellitus or premature coronary heart disease). However, for epidemiological studies, large-scale invasive procedures are dif®cult and costly to conduct, especially among children. The National Cholesterol Education Program (NCEP) Expert Panel on Blood Cholesterol Levels in Children and Adolescents33 and the American Academy of Pediatrics (AAP) Committee on Nutrition,34 provide guidelines for the selective screening of blood lipids only among children or adolescents with a family history of premature coronary heart disease or atherosclerotic disease. The cost-effectiveness of such selective screening has been challenged on several grounds and the guidelines are still controversial.24 The results of this study suggest that several non-invasive anthropometric parameters may also be ef®cient for screening school children for hyperlipidaemia. In this study, most anthropometric parameters correlated with lipids variables among children (except total cholesterol). Although BMI, WHR and skinfolds each measure a distinct component of obesity or body fat distribution, the skinfolds were most consistently the best predictors of blood lipid concentrations; a result similar to that found among US10,13 and Italian14 children. Body height was an important predictor for lower total cholesterol, LDL-C and ApoB levels among boys in this and other studies.35,36 Height may be a surrogate for adequate nutrition in infancy and childhood development. Regardless of the underlying physiological effects of height, these data suggest that the reduced risk of coronary heart disease among taller adult men and women,31,37 may be traceable to the association between height and lipids among adolescents. Although BMI was not as strong a predictor as the sum of skinfolds, it was a stronger predictor of lipid concentrations (for example, the CHOL=HDL-C ratio and ApoB) than WHR. In Zwiauer's study17, measures of obesity (for example, BMI, skinfolds and percentage of body fat), but not WHR correlated with lipids and lipoprotein concentrations in both genders. In our study, the negative association between waist circumference and the CHOL=HDL-C ratio, after controlling for BMI among boys, is quite com-plicated. Table 3 and Table 4 only illustrate the crude associations between anthropometric variables and lipids before multivariate adjustment. The Pearson correlation coef®cient between waist circumference and TCHR was 0.26 and 0.24 for boys and girls, respectively. In univariate regression analyses, waist was positively associated with the CHOL=HDL-C ratio, with the regression coef®cient of 0.023 (data not shown). After controlling for BMI, arm circumference and sum skinfolds, the waist circumference

was negatively associated with TCHR, with the regression coef®cient of 70.022. After controlling for other anthropometric variables, the waist circumference may be a proxy for lean body mass or body height but not for abdominal obesity. In a study of adolescent girls, Caprio et al5 measured fat stores using magnetic resonance imaging and reported that visceral fat, and not WHR or subcutaneous fat, correlated with insulin resistance and lipids levels. However, several other studies3,10 have found skinfold thickness (as markers of subcutaneous fat) are signi®cantly associated with several measures of blood lipids. Although WHR is also associated with lipid concentrations and not all skinfold measurements may be equally informative, subscapular skinfold measures are generally stronger and more consistently predict lipid levels.3,4,10,36 In the Bogalusa Heart Study15, truncal skinfolds such as subscapular and suprailiac measurements were more strongly correlated with total cholesterol and HDL-C than peripheral skinfolds. Others have reported that several skinfold measurements and ratios of these indices (subscapular=triceps ratio) are positively correlated with TG and ApoB concentrations and negatively correlated with HDL-C in children and adults.14,38 We found that the sum of four skinfold measurements was one of the strongest predictors of most lipid measures, but no single skinfold measure could signi®cantly predict all adverse lipid pro®les. In summary, results from most previous studies, suggest that truncal and peripheral fat distributions are better predictors of lipid concentrations than BMI or WHR. We corroborated these ®ndings in a Taiwanese population of school children. Among boys, truncal and peripheral subcutaneous fat skinfold measurements were better predictors than BMI or WHR of TG, HDL-C, LDL-C, ApoA1, ApoB and the CHOL=HDLC ratio. Among girls, the skinfold measurements were better predictors of HDL-C, LDL-C, ApoA1 and the CHOL=HDL-C ratio than BMI or WHR. Acknowledgements

This study was supported by the Department of Health, Executive Yuan, Taiwan, ROC. The authors acknowledge Dr. Gerald S. Berenson for his valuable guidance and comment on the early proposal and conduct of this study. References

1 Kannel WB, Wilson PWF. An update on coronary risk factors. Med Clin North Am 1995; 79: 951±971. 2 Gidding SS. A perspective on obesity. Am J Med Sci 1995; 310 (suppl) S68±S71. 3 Folosm AR, Burke GL, Ballew C, Jacobs DR, Haskell WL, Donahue RP, Liu KA, Hilner JE. Relation of body fatness and its distribution to cardiovascular risk factors in young blacks and whites. The role of insulin. Am J Epidemiol 1989; 130: 911± 924. 4 Foster CJ, Weinsier RL, Birch R, Norris DJ, Bernstein R, Wang J, Pierson RN, Vanitallie TB. Obesity and serum lipids: An evaluation of the relative contribution of body fat and fat distribution to lipid levels. Int J Obes 1987; 11: 151±161.

71


72

5 Caprio S, Hyman LD, Limb C, McCarthy S, Lange R, Sherwin RS, Shulman G, Tamborlane WV. Central obesity and its metabolic correlates in obese adolescent girls. Am J Physiol 1995; 269: E118±E126. 6 Berenson GS, Radhakrishnamurthy B, Bao W, Srinivasan SR. Does adult-onset diabetes mellitus begin in childhood?: The Bogalusa Heart Study. Am J Med Sci 1995; 310 (suppl 1): S77±S82. 7 Bouchard C, Bray GA, Hubbard VS. Basic and clinical aspects of regional fat distribution. Am J Clin Nutr 1991; 52: 946±950. 8 Rocchini AP. Adolescent obesity and hypertension. Ped Clin North Am 1993; 40: 81±92. 9 Raitakari OT, Porkka KVK, Viikari JSA, Ronnemaa T, Akerblom HK. Clustering of risk factors for coronary heart disease in children and adolescents. The Cardiovascular Risk in Young Finns Study. Acta Pediatr 1994; 83: 935±940. 10 Wattigney WA, Harsha DW, Srinivasan SR, Webber LS, Berenson GS. Increasing impact of obesity on serum lipids and lipoproteins in young adults. The Bogalusa Heart Study. Arch Intern Med 1991; 151: 2017±2022. 11 Muller WH, Wear ML, Hanis CL. Body circumferences as alternative to skinfold measurements of body fat distribution in Mexican American. Int J Obes 1987; 11: 309±318. 12 Ding YA, Chu NF, Wang TW, Lin CC. Anthropometric and lipoprotein-related characteristics of young adult males in Taiwan. Int J Obes 1995; 19: 392±396. 13 Sangi H, Mueller WH. Which measure of body fat distribution is best for epidemiologic research among adolescents? Am J Epidemiol 1991; 133: 870±883. 14 Bellu R, Ortisi MT, Scaglioni S, Agostoni C, Salanitri VS, Riva E, Giovannini M. Lipid and apoprotein A-1 and B levels in obese school-age children: Results of a study in the Milan area. J Ped Gastrol Nutr 1993; 16: 446±450. 15 Freeman DS, Srinivasan SR, Harsha DW, Webber LS, Berenson GS. Relation of body fat patterning to lipid and lipoprotein concentrations in children and adolescents: The Bogalusa Heart Study. Am J Clin Nutr 1989; 50: 930±939. 16 Kempen HJM, de Knijff P, Boomsma DI, van der Voort HA, Leuven G, Havekes L. Plasma levels of Lathosterol and Phytosterols in relation to age, sex, anthropometric parameters, plasma lipids and apolipoprotein E phenotype, in 160 Dutch families. Metabolism 1991; 40: 604±611. 17 Zwiauer KF, Pakosta R, Mueller T, Widhalm K. Cardiovascular risk factors in obese children in relation to weight and body fat distribution. J Am Coll Nutr 1992; 11 (suppl): 41S± 50S. 18 Foldmaker CE, Sveger T, Nilsson-Ehle P Waist measurement correlated to a potentially atherogenic lipoprotein pro®le in obese 12±14-year-old children. Acta Pediatr 1994; 83: 941± 945. 19 Department of Health, The Executive Yuan, ROC. Public Health in Taiwan area, Republic of China, 1992. 20 Richmond W. Preparation and properties of a cholesterol oxidase from Nocardia sp. and its application to the enzymatic assay of total cholesterol in serum. Clin Chem 1973; 19: 1350±1356. 21 Stavropoulos WS, Crouch RD. A new colorimetric procedure for the determination of serum triglyceride. Clin Chem 1974; 20: 857. 22 Warnick GR. Benderson J, Albers JJ. Dextran sulfate-Mg precipitation procedure for quanti®cation of high density

lipoprotein-cholesterol. Clin Chem 1982; 28: 1379±1388. 23 Marcovina SM, Albers JJ, Dati F, Ledue TB, Ritchie RF. International federation of clinical chemistry standardization projects for measurements of apolipoprotein A-1 and B. Clin Chem 1991; 37: 1676±1682. 24 Friedewald WT, Levy R, Fredrickson DS. Estimation of the concentration of low density lipoprotein-cholesterol in plasma without use of the preparative ultracentrifuge. Clin Chem 1972; 18: 499±502. 25 Obarzanek E. Methodological issues in estimating the prevalence of obesity in childhood. Ann NY Acad Sci 1993; 699: 278±279. 26 Troiano RP, Flegal KM, Kuczmarski RJ, Campbell SM, Johnson CL. Overweight prevalence and trends for children and adolescents. The National Health and Nutritional Examination Survey, 1963±1991. Arch Pediatr Adolesc Med 1995; 149: 1085±1091. 27 Rifai N, Neufeld E, Ahlstrom P, Rimm E, D'Angelo L, Hicks JM. Failure of current guidelines for cholesterol screening in urban African-American adolescents. Pediatrics 1996; 98: 383±388. 28 Michel U, Riechers B. Cardiovascular risk factors in schoolchildren. J Am Coll Nutr 1992; 11 (suppl): 36S±40S. 29 Webber LS, Wattigney WA, Srinivasan SR, Berenson GS. Obesity studies in Bogalusa. Am J Med Sci 1995; 310 (suppl 1): S53±S61. 30 Kunze D. Cholesterol blood levels in children: Comparison of a Munich screening to worldwide studies form 1980 to 1990. J Am Coll Nutr 1992; 11: 23S±27S. 31 Rimm EB, Stampfer MJ, Giovannucci E, Ascherio A, Spiegelman D, Colditz GA, Willett WC. Body size and fat distribution as predictors of coronary heart disease among middle-aged and older US men. Am J Epidemiol 1995; 111: 1117±1127. 32 Willett WC, Manson JE, Stampfer MJ, Colditz GA, Rosner B, Speizeer FE, Hennekens CH. Weight, weight change, and coronary heart disease in women. Risk within the `normal' weight range. JAMA 1995; 273: 461±465. 33 NCEP Expert Panel on Blood Cholesterol Levels in Children and Adolescents. National Cholesterol Education Program (NCEP): highlights of the report of the Expert Panel on Blood Cholesterol Levels in Children and Adolescents. Pediatrics 1992; 89: 495±501. 34 American Academy of Pediatrics, Committee on Nutrition. Statement on cholesterol. Pediatrics 1992; 90: 469±473. 35 Forrester TE, Wilks RJ, Bennett FI, Simeon D, Osmond C, Allen M, Chung AP, Scott P. Fetal growth and cardiovascular risk factors in Jamaican school children. BMJ 1996; 312: 156± 160. 36 Whincup PH, Cook DG, Adshead F, Taylor S, Papacoota O, Walker M, Wilson V. Cardiovascular risk factors in British children from towns with widely different adult cardiovascular mortality. BMJ 1996; 313: 79±84. 37 Rich-Edwards JW, Manson JE, Stampfer MJ, Colditz GA, Willett WC, Rosner B, Speizer FE, Hennekens CH. Height and the risk of cardiovascular disease in women. Am J Epidemiol 1995; 142: 909±917. 38 Faber M, Spinnler-Benade AJ, Daubitzer A. Dietary intake, anthropometric measurements and plasma lipid levels in throwing ®eld athletes. Int J Sports Med 1990; 11: 140±145.