Cholesterol and triglyceride concentrations, birthweight and central obesity in pre-school children

I Cowin*, P Emmett and the ALSPAC Study Team

University of Bristol, Institute of Child Health, Unit of Paediatric and Perinatal Epidemiology, Bristol, BS8 1TQ, UK

International Journal of Obesity (2000) 24, 330339 2000 Macmillan Publishers Ltd All rights reserved 03070565/00 $15.00 www.nature.com/ijo

OBJECTIVE: To investigate the relationship between blood cholesterol and triglyceride, birthweight and central obesity in pre-school children, after controlling for height and body mass index. METHODS: This was a longitudinal population-based study in south-west England. Research clinics were held when the children were 31 and 43 months of age, where anthropometric measurements were made and a non-fasting blood sample was taken and analysed for triglyceride, total cholesterol and high-density lipoprotein (HDL) cholesterol. Lowdensity lipoprotein (LDL) cholesterol values were calculated using the Friedewald equation. Central obesity was estimated using the ratio of waist circumference:arm circumference (WC:AC). RESULTS: Complete blood lipid and anthropometric data were available for 385 children at 31 months and 470 children at 43 months. Height was negatively associated with the concentration of triglyceride, and total and LDL cholesterol. There was little evidence for a relationship between body mass index (BMI) and blood lipids at either 31 or 43 months. The only signicant relationship between birthweight and blood lipids was a negative association with HDL (and consequently a positive association with the ratio of total:HDL cholesterol) in boys at 43 months. Adjustment for current height and BMI had little effect on the associations between birthweight and blood lipid concentrations. WC:AC was positively associated with triglycerides and negatively associated with HDL values in boys, and had a quadratic relationship with LDL concentrations among girls. These relationships were unchanged or became stronger on adjustment for current height and body mass index. CONCLUSION: In the pre-school child, central obesity has a relationship with triglyceride and HDL concentrations that is independent of current height and BMI. We have found no evidence that increasing birthweight is associated with a more favourable blood lipid prole at 31 and 43 months. International Journal of Obesity (2000) 24, 330339 Keywords: pre-school children; central obesity; triglycerides; HDL cholesterol

Introduction
Britain has one of the highest rates of coronary heart disease in the world.1 This is partly due to the high mean cholesterol levels in this country,2 as high cholesterol is related to a greater degree of atherosclerosis3 and to increased coronary heart disease morbidity and mortality.4 Strong relationships between blood lipid levels and anthropometry have been found in adults5 7 and school age children.8 11 Two particularly consistent relationships are the association of increased height with a more favourable blood lipid prole among children, and of obesity with a less favourable blood lipid prole among both adults and children. Most studies on adults have found central obesity to be particularly strongly related to blood lipid levels,12 14 but the importance of fat distribution in pre-adolescent children is unclear, with studies giving conicting results.8,11,15 18 In
*Correspondence I Cowin, University of Bristol, Institute of Child Health, Unit of Paediatric and Perinatal Epidemiology, 24 Tyndall Avenue, Bristol BS8 1TQ, UK. E-mail: imogen.cowin@bris.ac.uk Received 29 March 1999; revised 27 July 1999; accepted 8 October 1999

addition, historical cohort studies have suggested that intrauterine factors represented by birthweight may have a life-long effect on blood lipid metabolism.19,20 However, it is uncertain that the relationships between blood lipid levels and birthweight in children born in the 1990s will be the same as among those born in the rst half of the century, when the dietary and environmental conditions experienced by pregnant women were very different to those today. It appears that the process of atherosclerosis begins very early in life,21,22 thus the maximum chance of preventing coronary heart disease developing would be given by reducing cholesterol levels in children at as young an age as possible. However, there is very little information on the determinants of blood lipid levels of children below school age. The aim of this study was to examine the relationships between blood lipid levels, birthweight and central obesity in a group of pre-school children, after controlling for current height and body mass index (BMI).

Methods
Study design

This study forms part of the Avon Longitudinal Study of Pregnancy and Childhood, a geographically based

Blood lipids in pre-school children I Cowin et al

cohort study investigating factors inuencing the health and development of children. All pregnant women resident within that part of Avon which was within the area of the South Western Regional Health Authority with an expected date of delivery between April 1991 and December 1992 inclusive were eligible. The study has been described in more detail elsewhere.23 Information in the ALSPAC study is collected both from medical records and via a series of self-completion postal questionnaires sent to the mother. Between 80% and 90% of eligible mothers enrolled for the study, resulting in 14,893 pregnancies. A proportion of the children born in the last 6 months of the study (equivalent to 10% of the whole cohort) were selected at random to take part in a sub-study known as Children in Focus (CIF). These children attended research clinics at which a variety of physical and developmental measures were taken approximately every 6 months. The blood lipid data and anthropometric measurements used in this study were obtained at the CIF clinics. Ethical approval for the study was obtained from the ALSPAC Ethics and Law Sub-Committee and from the three Medical Ethics Committees in the Avon study area.
Blood collection and lipid analysis

Foundation, and recorded to the nearest 0.1 cm. Waist circumference (WC) was measured using Babytape at 31 months, and a Holtain anthropometric tape at 43 months, and recorded to the nearest 0.1 cm. Two indices were calculated using these measurements. Body mass index (BMI) was calculated from the weight in kilogrammes divided by the height in metres squared, for use as a measure of obesity. WC:AC was calculated from the ratio of waist circumference to arm circumference and used as a measure of fat distribution analogous to the waist:hip ratio, on the assumption that a higher ratio of waist to arm circumference is an indication of a more central pattern of obesity. We used this measure in preference to waist hip ratio. This was because a study which determined peripheral and abdominal adipose tissue content in children using magnetic resonance imaging found that WC:AC correlated positively with intraabdominal adipose tissue (r 0.63, P 0.001) while waist:hip ratio did not (r 0.11).25
Other variables

331

A non-fasting blood sample for lipid analysis was collected when the children were aged 31 months (between January and June 1995) and when the children were aged 43 months (between January and June 1996). The blood sample was obtained using venepuncture after the application of a topical anaesthetic cream (EMLA). The resulting plasma samples were stored at 20 C and analysed using a Technicon RA500 which was calibrated regularly. Total cholesterol concentrations were determined using an enzymatic colorimetric test with cholesterol esterase. High-density lipoprotein (HDL) concentrations were also assessed by a colorimetric test using cholesterol esterase, after precipitation of low-density lipoproteins (LDL) by dextran sulphate and magnesium. Triglyceride concentrations were determined using an enzymatic colorimetric test. LDL levels were calculated by applying the Freidewald equation24 this equation is not applicable when triglyceride values are greater than 2 mmol=l. The method of analysis was validated by comparing our results with those obtained by a hospital laboratory using the same samples.
Anthropometry

Breast-feeding history was obtained from a questionnaire sent to the mother when the child was 6 months old. Maternal pre-pregnancy BMI was calculated from self-reported pre-pregnancy weight and height collected by questionnaire.
Statistical methods

The anthropometric measurements taken were weight, height, arm circumference and waist circumference. Height was measured to the nearest 0.1 cm using a Leicester height measure. Weight was recorded to the nearest 100 g with the child wearing only underwear, using a Seca scale, model 835. Arm circumference (AC) was measured midway between scapula and elbow using Babytape, made by the Child Growth

The CIF sample included a mixture of ethnic groups and both twin and singleton births. A number of studies have found ethnic group to inuence both anthropometry and blood lipid levels.8,16,26,27 As the number of non-white children in the sample was too small to analyse separately they have been excluded from the statistical analyses in order to prevent confounding. Twins have also been excluded from statistical analyses as the two members of a twin pair do not represent truly independent points, and unlike the rest of the CIF sample they were not a randomly selected group. In addition, some children were excluded from the analyses where we judged that there had been errors in recording anthropometric measurements, for example where there was a large change in waist or arm circumference between 31 and 43 months without a similarly large change in weight this led to the exclusion of 48 children. The relationship between blood lipid concentrations, anthropometry and birthweight was investigated using regression analysis. The blood lipid variables investigated were triglyceride, total cholesterol, HDL cholesterol, LDL cholesterol and the ratio of total:HDL cholesterol (T:HDLC). Triglycerides and T:HDLC were transformed to the natural logarithm to reduce skewness in the distribution. Breastfeeding history was included in some multivariate analyses as a possible confounder, as research has suggested that it may have long-term effects on blood lipid levels,20,28 and the growth patterns of breast-fed and
International Journal of Obesity

Blood lipids in pre-school children I Cowin et al

332

non-breast-fed children have been shown to differ.29 Birthweight, WC:AC, height and BMI were tted as continuous variables. Breast-feeding history was tted as a dichotomous variable, with a breast-fed child being dened as one who was breast-fed for a month or more. Initially all the anthropometric variables and birthweight were regressed singly against the blood lipid variables in order to examine the t of the linear and quadratic models. The quadratic model was used if it increased the level of variance explained as compared to the linear model, and if the P-value of both the linear and quadratic terms in the quadratic model was < 0.05. Then a second set of analyses was carried out, in which birthweight and WC:AC were adjusted for the effects of current height and BMI to determine how this affected their relationship with blood lipid levels. The third set of analyses examined the relationship between birthweight and WC:AC and blood lipid levels when all the variables (birthweight, WC:AC, height, BMI and breastfeeding) were entered into a regression analysis simultaneously. We also examined the effect of adding maternal BMI to the nal model, as this is associated with both birthweight and childhood BMI. All analyses were carried out separately for boys and girls. All analyses were carried out using the SPSS statistical package.

had been withdrawn from the study by their parents between the 31 and 43 month clinics). A total of 878 mothers gave consent for blood to be taken from their children and 742 blood samples were obtained. Total cholesterol concentrations were measurable in 616 of these samples (57.8% of attendees), triglyceride in 613 and HDL in 615, allowing calculation of the T:HDL cholesterol ratio for 603 children. The triglyceride concentration was less than 2 mmol=l in 549 samples, allowing calculation of the LDL values in 543 children. Only white singletons with a complete set of valid measures were used in the analyses relating lipid concentrations to anthropometric measurements. This reduced the numbers to 371 for triglyceride analysis at 31 months (35.9% of white singletons attending) and 451 for triglyceride analysis at 43 months (46.1% of white singletons attending).
Anthropometry

Results
Response rate

In all, 1305 children were invited to the 31 month clinic, of whom 1135 (87%) attended. At the 31 month clinic the mothers of 926 children gave consent for a blood sample to be taken. However, due to the practical difculties with taking blood from children of this age blood samples were only actually obtained from 680 children, and not all of these samples were large enough for a full lipid analysis. Total cholesterol concentrations were successfully measured in 496 of the samples (43.7% of attendees), triglyceride in 496 and HDL in 372 samples, allowing calculation of the T:HDL cholesterol ratio in 368 children. The triglyceride concentration was less than 2 mmol=l in 410 samples, allowing calculation of the LDL values for 310 children. A total of 1248 children were invited to the 43 month clinic, of whom 1065 attended (67 children
Table 1

The anthropometric measurements were taken by a total of six measurers at 31 months and nine at 43 months. Reliability of the measurers had been assessed at clinics held when the children were aged 25 months and 49 months. Comparison of the results obtained from different pairs of measurers measuring the same child revealed no systematic biases between them. Therefore, we felt it was unnecessary to adjust the anthropometric results according to the measurer. Although all parents gave consent for their children to be measured, due to the practical difculties in measuring children of this age not all the anthropometric measurements were obtained for all the children. The mean anthropometric measurements at 31 and 43 months are shown in Table 1.
Representativeness of the sample

The weight and length measurements of the CIF sample at birth, 1 y and 2 y have been compared to published national standards30 and found to be very similar.31 In order to check that those children from whom we obtained blood samples were representative of the CIF sample we compared the anthropometric measurements of children who did and did not have blood taken. At 31 months boys who had a blood sample taken were marginally heavier (14.2 vs 13.9 kg, P 0.063) than those who did not, and had a slighter higher BMI (16.7 vs 16.4 kg m2, P 0.057) and waist circumference (50.4 vs 49.8 cm, P 0.048). However, in girls at 43 months those children who had

Mean (s.d.) anthropometric measurements at 31 and 43 months in boys and girls Weight (kg) Height (cm) 92.1(3.3) 90.9(3.2) 99.8(3.6) 98.6(3.7) Waist circumference (cm) 50.5(3.1) 50.4(3.1) 51.7(3.0) 51.7(3.6) Arm circumference (cm) 16.8(1.3) 16.8(1.2) 17.1(1.2) 17.2(1.3)

Boys, 31 months (n 613) Girls, 31 months (n 510) Boys, 43 months (n 585) Girls, 43 months (n 471) International Journal of Obesity

14.2(1.7) 13.7(1.6) 16.5(1.8) 16.0(2.0)

Blood lipids in pre-school children I Cowin et al

a blood sample taken were marginally lighter (16.0 vs 16.8 kg, P 0.056), with a lower BMI (16.4 vs 17.1 kg m2, P 0.042) and arm circumference (17.2 vs 17.7 cm, P 0.064). There were no anthropometric differences between those who did and did not have blood taken among girls at 31 months or boys at 43 months. Compared to the 1991 National Census data of mothers with infants under 1 y who were resident in Avon, the ALSPAC population demonstrated a slight shortfall in those living in rented accommodation, those without a car, single parent families and unmarried cohabiting couples. The mothers of the CIF subsample were, on average, slightly older than the ALSPAC population in general, and had a higher rate of owner occupation. We speculated that within CIF there might be socio-economic differences between children who did and did not have a blood sample taken. However, we could nd no association between presence or absence of a blood sample and the housing tenure or highest educational achievement of the mother (data not shown).

Tables 3 6 show the associations of height, BMI, birthweight and WC:AC with triglycerides, HDL cholesterol and LDL cholesterol. The associations with total cholesterol and T:HDLC are not shown in most of the tables, as there were very few signicant associations with total cholesterol concentrations, and the signicant relationships with T:HDLC tended to be the inverse of those with HDL. The nal models, including anthropometric variables, birthweight and breastfeeding history generally explained less than 5% of the variance in blood lipid values.
Height and blood lipid concentrations (Table 3)

333

Blood lipids

The mean blood lipid concentrations at the two time points are shown in Table 2. There was a signicant fall in the total cholesterol and triglyceride and a signicant increase in the HDL cholesterol concentrations between 31 and 43 months (P < 0.001 in all cases). The between-batch CV was less than 5% for all blood lipid analyses.

Height was negatively linearly associated with triglyceride concentrations in girls at both 31 and 43 months. Among boys there was a positive quadratic relationship between triglycerides and height at 31 months, but no relationship at 43 months. There was a negative linear relationship between height and total cholesterol at 31 months, which was signicant in boys and marginally signicant in girls. By 43 months the association between height and total cholesterol had disappeared. There were no signicant associations between height and HDL cholesterol concentrations or T:HDLC. The only signicant relationship between height and LDL cholesterol was a negative linear association among boys at 31 months.
Childhood BMI and blood lipids (Table 4)

There was no apparent relationship between BMI and triglyceride, total cholesterol, HDL cholesterol or

Table 2 Mean non-fasting blood cholesterol and triglyceride concentrations (mmol1) in the Children in Focus (CIF) cohort of pre-school children at 31 and 43 months (white singletons only) T otal cholesterol 31 months 43 months
a

HDL cholesterol 0.85(0.21) (n 343) 1.02(0.25) (n 560)

T:HDL cholesterol a 4.98 (95% CI: 4.80, 5.16) (n 339) 3.43 (95% CI: 3.34, 3.54) (n 557)

Triglyceride a 1.31 (95% CI: 1.26, 1.37) (n 455) 1.10 (95% CI: 1.05, 1.15) (n 559)

4.14(0.71) (n 455) 3.51(0.84) (n 562)

The data for triglyceride and T:HDL cholesterol ratio in the CIF survey was transformed, therefore the geometric mean with 95% condence intervals is shown.

Table 3

Regression coefcients of height against blood lipid variables among white singletons at two age points Boys, 31months Girls, 31months Percentage change (CI) P Boys, 43 months Percentage change (CI) P Girls, 43 months Percentage change (CI) P 0.010

Ln triglycerides Height (cm) Height2 (cm2)

Percentage change (CI) P 47 (72, 0) 0.4 (0.4, 0.4) (n 207)

T otal cholesterol Height (cm) 0.028(0.01) (n 207) B (s.e) HDL cholesterol Height (cm) 0.0056(0.01) (n 160) LDL cholesterol Height (cm) 0.033(0.01) (n 135)

0.047 2 (4, 0) 0.044 (n 173) 0.047 0.029(0.02) (n 175) P B (s.e.) 0.256 0.009(0.01) (n 129) 0.014 0.00029(0.02) (n 106)

0.016 0 (2, 2) (n 253) 0.091 0.00052(0.02) (n 253) P B (s.e.) 0.076 0.00044(0.01) (n 251) 0.988 0.0007(0.02) (n 217)

0.983 2 (4, 0) (n 215) 0.974 0.016(0.02) (n 217) P B (s.e.) 0.927 0.0067(0.00) (n 217) 0.962 0.0089(0.02) (n 190)

0.315 P 0.112 0.584

International Journal of Obesity

Blood lipids in pre-school children I Cowin et al

334

Table 4

Regression coefcients of BMI against blood lipid variables among white singletons at two age points Boys, 31months Girls, 31months Percentage change (CI) P Boys, 43 months Percentage change (CI) P Girls, 43 months Percentage change (CI) P 0.924 P 0.701 0.031 0.021

Ln triglycerides BMI (kgm2)

Percentage change (CI) P 1 (3, 5) (n 207)

HDL cholesterol BMI (kgm2) B(s.e.) 0.0004(0.01) (n 160) LDL cholesterol 2 BMI (kg/m ) 0.019(0.04) BMI (kg2/m4) (n 135)

0.697 3 (3, 9) (n 175) P B(s.e.) 0.976 0.019(0.01) (n 129) 0.586 0.048(0.05) (n 106)

0.330 4 (2, 10) (n 253) P B(s.e.) 0.167 0.00052(0.01) (n 251) 0.362 0.079(0.04) (n 217)

0.210 0(6, 6) (n 215) P B(s.e.) 0.970 0.0047(0.01) (n 217) 0.049 1.43(0.66) 0.045(0.02) (n 190)

T:HDLC. At 43 months only there was a signicant negative linear relationship between BMI and LDL cholesterol in boys, while in girls there was a negative quadratic relationship.
Birthweight and blood lipids (Table 5)

There was no apparent relationship between birthweight and triglyceride or total cholesterol concentrations, either in unadjusted or adjusted analyses. Among boys there was a negative association between HDL cholesterol values and birthweight at both 31 and 43 months. At 31 months this association was only marginally signicant, and became less signicant on adjusting for height and BMI, and other factors in the model. At 43 months the unadjusted association between birthweight and HDL was signicant. It was little affected by adjustment for height and BMI, but became stronger on further adjustment for WC:AC and breastfeeding history. There was no apparent relationship between birthweight and HDL among girls. There was a relationship between birthweight and T:HDLC in boys at 43 months which mirrored that with HDL concentrations. The association was positive, and became stronger on adjustment for height
Table 5

and BMI and all factors in the model. In an unadjusted analysis there was a marginally signicant negative association between T:HDLC at 31 months and birthweight in girls this was weakened on adjustment for height, BMI, and other factors in the model. There was no apparent association between birthweight and T:HDLC in boys at 31 months or in girls at 43 months. There was no association between birthweight and LDL cholesterol concentrations in either univariate or multivariate analyses.

Waist circumference:arm circumference and blood lipids (Table 6)

WC:AC was signicantly positively associated with triglyceride values in boys at both 31 and 43 months this relationship was strengthened on adjustment for current height and BMI. The mean triglyceride concentrations in boys at 43 months by quintile of WC:AC are shown in Figure 1. There was no association between WC:AC and triglyceride in girls. There was no apparent relationship between WC:AC and total cholesterol concentrations.

Regression coefcients of birthweight against blood lipid variables among white singletons at two age points Boys, 31months Girls, 31months Percentage change (CI) P Boys, 43 months Percentage change (CI) P Girls, 43 months Percentage change (CI) P 0.646 0.374 0.201 P 0.336 0.473 0.598 0.290 0.206 0.087

Ln triglycerides Birthweight (kg) Birthweighta Birthweightb

Percentage change (CI) P 1 (8, 12) 2 (13, 11)d 3 (14, 9)e (n 201 210) B(s.e.)

0.801 6 (18, 8) 0.741 3 (15, 11) 0.530 5 (19, 11) (n 170 175) P B(s.e.) 0.069 0.044(0.04) 0.127 0.011(0.04) 0.109 0.023(0.04) (n 126 129) 0.086 0.058(0.15) 0.493 0.11(0.16) 0.458 0.092(0.16) (n 104 106)

0.379 8 (4, 21) 0.686 7 (7, 23) 0.494 8 (6, 24) (n 244 254) P B(s.e.) 0.258 0.073(0.03) 0.791 0.079(0.03) 0.596 0.093(0.03) (n 242 252) 0.696 0.052(0.10) 0.488 0.096(0.10) 0.556 0.13(0.11) (n 209 218)

0.234 4 (11, 21) 0.322 8 (8, 26) 0.258 12 (6, 33) (n 207 216) P B(s.e.) 0.024 0.036(0.04) 0.022 0.028(0.04) 0.007 0.022(0.04) (n 209 218) 0.598 0.15(0.14) 0.355 0.18(0.14)c 0.240 0.25(0.15)d (n 182 191)

HDL cholesterol Birthweight 0.051(0.03) 0.047(0.03) Birthweighta b 0.05(0.03) Birthweight (n 157 161) LDL cholesterol Birthweight 0.14(0.08) Birthweighta 0.060(0.09) 0.066(0.09) Birthweightb (n 133 136)
a b

Adjusted for height and BMI. Adjusted for all factors in the model (height, BMI, WC:AC and breastfeeding history).

International Journal of Obesity

Blood lipids in pre-school children I Cowin et al Table 6 Regression coefcients of waist:arm circumference against blood lipid variables among white singletons at two age points Boys, 31months Ln triglycerides WC:AC WC:ACa WC:ACb Percentage change (CI) P 43 (3, 100) 62 (16, 126)c 63 (17, 128)d (n 201 209) B (s.e.) Girls, 31months Percentage change (CI) P Boys, 43 months Percentage change (CI) P Girls, 43 months Percentage change (CI) P 0.427 0.409 0.469 P 0.060 0.070 0.072 0.304 0.329 0.421

335

HDL cholesterol WC:AC 0.20(0.10) 0.22(0.10) WC:ACa WC:ACb 0.22(0.10) (n 157 161) LDL cholesterol WC:AC 0.49(0.26) WC:AC2 WC:ACa 0.55(0.27) WC:AC2a b WC:AC 0.52(0.27) WC:AC2b (n 133 136)
a

0.031 14 (26, 75) 0.006 25 (19, 92) 0.005 22 (21, 88) (n 170 175) P B (s.e.) 0.043 0.072(0.13) 0.026 0.081(0.13) 0.028 0.053(0.13) (n 126 129) 0.063 37.76(13.21) 6.06(2.19) 0.039 36.78(13.43) 5.90(2.22) 0.053 37.65(13.69) 6.05(2.27) (n 104 106)

0.536 79 (21, 164) 0.318 101 (33, 204) 0.352 101 (31, 210) (n 244 254) P B (s.e.) 0.568 0.27(0.10) 0.525 0.32(0.11) 0.684 0.35(0.11) (n 241 252) 0.005 0.30(0.31) 0.007 0.007 0.50(0.33) 0.009 0.007 0.51(0.34) 0.009 (n 209 218)

0.005 19 (22, 79) 0.001 19 (22, 79) 0.002 16 (23, 75) (n 207 216) P B (s.e.) 0.009 0.18(0.10) 0.003 0.18(0.10) 0.002 0.18(0.10) (n 209 218) 0.345 0.36(0.35) 0.125 0.34(0.35) 0.129 0.29(0.36) (n 182 191)

WC:AC, waist circumference:arm circumference. Adjusted for height and BMI. Adjusted for all factors in the model (height, BMI, birthweight and breastfeeding history).

HDL cholesterol concentrations were linearly negatively associated with WC:AC in boys at both ages (Table 2). This association became slightly stronger on adjustment for current height and BMI. A scatterplot of HDL concentrations against WC:AC in boys at 43 months is shown in Figure 2. In girls at 43 months there was also a negative association between HDL and WC:AC this was marginally signicant and was little affected by adjustment for other factors in the model. There was no apparent relationship between WC:AC and HDL concentrations in girls at 31 months. There was no association between T:HDLC and WC:AC in adjusted or unadjusted analyses. At 31 months there was a quadratic relationship between WC:AC and LDL concentrations in girls

only, which was largely unaffected by adjustment for current height and BMI. In boys there was a marginal negative linear association with LDL, which became signicant on adjustment for height and BMI. There was no association between LDL concentrations and WC:AC at 43 months in either sex.
Effect of including maternal pre-pregnancy BMI in the model

In univariate analyses the only signicant association between maternal pre-pregnancy BMI and blood lipids was with cholesterol concentrations in girls, which were signicantly negatively associated with maternal BMI at both 31 and 43 months (r 0.176, P 0.024 and r 0.151, P 0.030, respectively).

Figure 1 Mean unadjusted triglyceride levels in white singleton boys at 43 months by quintile of waist circumference : arm circumference (WC=AC).

Figure 2 Waist circumference : arm circumference at 43 months against unadjusted HDL cholesterol levels in white singleton boys. International Journal of Obesity

Blood lipids in pre-school children I Cowin et al

336

Adding maternal BMI to the multivariate models had very little effect on the association between blood lipids and the other predictor variables, except in the models predicting total cholesterol at 43 months, where the addition of maternal BMI resulted in the association between birthweight and total cholesterol concentrations reaching signicance in both boys and girls (B 0.253, P 0.033 and B 0.302, P 0.0485, respectively).

Discussion
As the children we were studying were so young, it was not possible to obtain fasting blood samples. This means that the triglyceride values we have obtained must be interpreted with caution. However, using nonfasting blood has a negligible effect on measures of total and HDL cholesterol concentrations. We have found a signicant change in the mean blood lipid concentrations between 31 and 43 months, with the combination of a fall in total cholesterol and an increase in HDL cholesterol meaning that the blood lipid prole is considerably less atherogenic than at 43 months. The reasons for these changes are unclear, in other surveys of pre-school children blood lipid concentrations are very stable over this age range.26,32,33 We chose to analyse the data from boys and girls separately. This was because we found gender differences in the mean blood lipid values at 43 months boys had a signicantly higher mean HDL concentration and a lower T:HDLC (data not shown). Gender differences in blood lipid concentrations have been observed previously in surveys of pre-school children.26,27 Data on 8 18 y olds in `Project HeartBeat!' suggest that the relationship between blood lipids and anthropometry differs between the sexes.34 When the data from boys and girls in our study were combined it tended to reduce the strength and signicance of any associations between blood lipid concentrations and anthropometry, despite the considerable increase in the sample size (data not shown). The relationships we have found between anthropometry, birthweight and blood lipid concentrations are a little hard to interpret. The associations are fairly weak, and the models can only explain a small amount of the variation in blood lipids. However, there will have been a certain amount of measurement error in our blood lipid measurements due to within person variation the correlation between repeated plasma cholesterol measurements some time apart within individuals is around 0.7.35 Thus the models used are likely to underestimate the true strength of any relationships between anthropometry and birthweight and blood lipids as a result of regression dilution bias.36 While it is sometimes unclear whether the relationships are linear or quadratic in form, and whether the differences between the two time points
International Journal of Obesity

are real or a feature of factors such as changing sample size, a number of patterns do emerge from the data. Increasing height appears to be associated with more favourable blood lipid concentrations in these children, being negatively correlated with triglyceride, total cholesterol and LDL cholesterol. Similar relationships between height and plasma cholesterol have been observed in British 9 y olds8 and school-age Jamaican children.9 We have found very little evidence that a higher BMI has a detrimental effect on blood lipid proles at 31 and 43 months, indeed at 43 months there was some evidence of a negative association between BMI and LDL concentrations. It should be remembered that these LDL values, are a biased sample; as we were using non-fasting blood we were unable to calculate LDL values, for those children with a triglyceride concentration of more than 2 mmol=l, and it is uncertain whether the associations we have found would apply across the whole population. Our results are in contrast to those of a number of studies of children aged between 5 and 16, where, indicators of fatness such as weight-for-height, BMI and skinfold thicknesses have been associated with higher total and LDL cholesterol, lower HDL cholesterol and higher triglyceride concentrations.8 11 However, a number of other studies of pre-school children have also failed to nd an association between BMI and blood lipid concentrations. In a multiracial sample of 7439 1 4 olds from Arizona there was no association between BMI or weight-for-height and blood cholesterol, although very overweight (BMI  20) 3 and 4 y olds tended to have higher cholesterol concentrations than the other children.27 Similarly, in a longitudinal study of 256 children from infancy to age 7, Freedman et al. found no consistent associations between ponderosity and blood lipid values during the rst 4 y of life.26 We thought it was possible that we had not observed an effect of obesity because there was a lack of overweight children in our sample, however, when we compared the CIF sample with the UK reference data (Child Growth Foundation London) we found that we had higher than expected numbers of children over the 95th percentile of BMI (6.1% of boys and 7.7% of girls at 31 months, and 7.1% of boys and 9.0% of girls at 43 months). It may be that the detrimental effects of obesity on blood lipid concentrations are not apparent until later in life. Our data suggest that fat distribution (as measured by the ratio of waist circumference to arm circumference) is a stronger determinant of blood lipid concentrations in pre-school children than BMI. We have found WC:AC to be positively associated with triglyceride concentrations in boys, and there is evidence of a negative association with HDL in both sexes, although it should be noted that WC:AC explains a relatively small amount of the variation. As the blood samples we obtained were non-fasting, the triglyceride concentrations will have been affected

Blood lipids in pre-school children I Cowin et al

by the time since and composition of the last meal. However, this effect is likely to have been random with respect to WC:AC, so is likely to have led to an underestimate of the strength of the association between WC:AC and triglycerides. Any meal eaten by the children would have been at least 1 3 h before the blood sample was taken, due to the amount of time spent in the clinic, which included an hour for the anaesthetic cream to take effect. The quadratic relationship between WC:AC and LDL concentrations in girls is hard to interpret, due to the biased nature of the LDL samples the results of studies in adults would have led us to expect a positive association between WC:AC and LDL.12 Some studies in adults have suggested that waist circumference alone is a better predictor of fat distribution37 and of the metabolic consequences of obesity38 than ratios such as waist:hip or WC:AC. We used waist:arm circumference in preference to waist alone in order to avoid the problem of having two highly correlated variables in a multivariate analysis, ie waist circumference and BMI, making it hard to distinguish between the effects of overweight and fat distribution. In univariate analyses, the association between triglycerides at 31 and 43 months in boys and waist circumference was similar to that with waist:arm circumference. However, while there was a signicant association between HDL and waist:arm circumference at both ages, there was no such association with waist circumference (data not shown). The relative merits of waist circumference and waist: arm circumference as epidemiological tools in young children need to be investigated further. The results of other studies in children relating blood lipid concentrations to fat distribution are mixed. Rona et al found no relation between the ratio of limb:body skinfold thicknesses and cholesterol concentrations in 9 y old British children,8 however, in 9 10 y old American girls this ratio was positively associated with HDL cholesterol values,16 and among 2816 5 17 y old children in the Bogalusa Heart Study the ratio of subscapular:triceps skinfold was positively associated with LDL and triglycerides, and negatively associated with HDL.15 A Swedish study of obese 12 14 y olds found waist circumference to be signicantly negatively related to HDL-cholesterol, and signicantly positively related to triglycerides.17 However, two American studies, one in 8 11 y old girls11 and one in 12 17 y olds,18 found no independent association of waist:hip ratio with blood lipids. Interpretation of these results is complicated by uncertainty about the validity of the different measures of fat distribution, and in some cases by the inuence of stage of puberty. The relationship between fat distribution and metabolic risk factors may also be affected by race.39 Although we adjusted for the effect of breastfeeding we found little to suggest that infant diet is an important determinant of subsequent blood lipid

concentrations. The only signicant association was a 0.28 mmol lower total cholesterol concentration in breast-fed children among girls at 31 months. While animal studies have shown that different infant diets may have permanent effects on blood lipid concentrations and metabolism,28 several studies in humans have failed to nd persistent effects of breastfeeding on cholesterol values.27,40,41 Some studies have suggested that intrauterine factors may be a determinant of subsequent plasma cholesterol concentrations. Analysis of birth records of men born in Shefeld in 1939 1940 found that low abdominal circumference at birth was associated with raised total and LDL cholesterol concentrations.19 The authors suggested that this reected a permanent impairment of cholesterol metabolism resulting from reduced liver growth in the foetus as a response to undernutrition. In addition, lower birthweights have been found to be associated with an increased risk of dying from cardiovascular disease.42 Our results give little support to the theory that the intra-uterine environment may have longterm effects on blood lipid metabolism. The only signicant associations we have found are a negative association between birthweight and HDL concentrations in boys at 43 months, a positive association with T:HDLC in the same group, and a positive association with total cholesterol at 43 months in both sexes (after adjustment for maternal pre-pregnancy BMI). Furthermore these associations are not in the direction which might be expected from the Barker hypothesis, in that if high birthweight is associated with a reduced risk of coronary heart disease one might expect it to be associated with a more favourable blood lipid prole. We have found increasing birthweight to be associated with lower HDL, increased T:HDLC, and increased total cholesterol concentrations, all of which are considered risk factors for coronary heart disease. Several other studies have found increased birthweight to be associated with a less favourable blood lipid prole. Among men born in Hertfordshire at the beginning of the century, serum cholesterol rose with increasing birthweight.20 Similarly, in a survey of middleaged Australians there was a positive correlation between birthweight and cholesterol, despite a negative correlation with current height43 and in the National Study of Health and Growth in Britain there was a marginal positive association between birthweight and total cholesterol concentrations in 9 y-old children.8 In conclusion we have found no evidence that increasing birthweight is associated with a more favourable blood lipid prole, while there is some evidence that a central pattern of fat distribution may have an adverse effect on the blood lipid prole as early as 21 2 y of age. While the pattern of fat distribution is unlikely to be changeable by intervention, it may be of use in identifying those children who are likely to go on to develop heart disease.

337

International Journal of Obesity

Blood lipids in pre-school children I Cowin et al

338

Acknowledgements

We would like to acknowledge the dedicated work of the ALSPAC study team; this includes interviewers, computer technicians, clerical workers, research scientists, volunteers and managers. We would particularly like to thank the staff of the Children in Focus research clinics, the laboratory staff responsible for analysing the blood samples, and the Children in Focus parents and children. We would also like to acknowledge the help and advice of Dr Peter Whincup and Professor Jean Golding, and to thank Dr Charles Pennock for his supervision of the lipid analysis. This study has been supported by the Northern and Yorkshire Regional Health Authority NHS Research and Development Programme on Cardiovascular Disease and Stroke. The ALSPAC study is part of the WHO initiated European Longitudinal Study of Pregnancy and Childhood. The Children in Focus sub-study is, however, unique to ALSPAC.

References

1 Uemura K, Pisa Z. Trends in cardiovascular disease mortality in industrialized countries since 1950. World Health Stat Q 1988; 41: 155 178. 2 Simons LA. Interrelations of lipids and lipoproteins with coronary artery disease mortality in 19 countries. Am J Cardiol 1986; 57: 5G 10G. 3 PDAY Research Group. Relationship of atherosclerosis in young men to serum lipoprotein cholesterol concentrations and smoking. JAMA 1990; 264: 3018 3024. 4 Holme I. An analysis of randomized trials evaluating the effect of cholesterol reduction on total mortality and coronary heart disease incidence. Circulation 1990; 82: 1916 1924. 5 Ko GTC, Chan JCN, Woo J, Lau E, Yeung VTF, Chow G-C, Wai HPS, Li JKY, So W-Y, Cockram CS. Simple anthropometric indexes and cardiovascular risk factors in Chinese. Int J Obes 1997; 21: 995 1001. 6 Rabkin SW, Chen Y, Leiter L, Liu L, Reeder BA, Balram C, Blair L, Butler-Jones D, Cameron R, Collins-Nakai R, Connelly PW, Donovan C, Dyck R, Edwards AC, Gelskey D, Hogan K, Joffres MR, Lessard R, Macdonald SM et al. Risk factor correlates of body mass index. Can Med Assoc J 1997; 157(Suppl 1): S26 S31. 7 Mendleim JM, Freedman DS, Peter DG, Allen B, Percy CA, Ballew C, Mokdad AH, White LL. Risk factors for coronary heart disease among Navajo Indians: Findings from the Navajo Health and Nutrition Survey. J Nutr 1997; 127(Suppl 10): 2099S 2105S. 8 Rona RJ, Qureshi S, Chinn S. Factors related to total cholesterol and blood pressure in British 9 year olds. J Epidemiol Commun Health 1996; 50: 512 518. 9 Forrester TE, Wilks RJ, Bennett FI, Simeon D, Osmond C, Allen M, Chung AP, Scott P. Fetal growth and cardiovascular risk factors in Jamaican shoolchildren. Br Med J 1996; 312: 156 160. 10 DeStefano F, Berg RL, Griese GG. Determinants of serum lipid and lipoprotein concentrations in children. Epidemiology 1995; 6: 446 449. 11 Craig SB, Bandini LG, Lichtenstein AH, Schaefer EJ, Dietz WH. The impact of physical activity on lipids, lipoproteins, and blood pressure in preadolescent girls. Pediatrics 1996; 98: 389 395. 12 Ko GTC, Chan JCN, Lau E, Yeung VTF, Chow C-C, Wai HPS, Li JKY, So W-Y, Cockram CS. Simple anthropometric indexes and cardiovascular risk factors in Chinese. Int J Obes 1997; 21: 995 1001.

13 Hsieh SD, Yoshinaga H. Abdominal fat distribution and coronary heart disease risk factors in men waist=height ratio as a simple and useful predictor. Int J Obes 1995; 19: 585 589. 14 James RW, Brulhart-Meynet M-C, Lehmann T, Golay A. Lipoprotein distribution and composition in obesity: their association with central adiposity. Int J Obes 1997; 21: 1115 1120. 15 Kikuchi DA, Srinivasan SR, Harsha DW, Webber LS, Sellers TA, Berenson GS. Relation of serum lipoprotein lipids and apolipoproteins to obesity in children: the Bogalusa Heart Study. Prev Med 1992; 21: 177 190. 16 Simon JA, Morrison JA, Similo SL et al. Correlates of highdensity lipoprotein cholesterol in black girls and white girls: the NHLBI growth and health study. Am J Public Health 1995; 85: 1698 1701. 17 Flodmark CE, Sveger T, Nilsson-Ehle P. Waist measurement correlates to a potentially atherogenic lipoprotein prole in obese 12 14-year-old children. Acta Paediatr Scand 1994; 83: 941 945. 18 Gillum RF. The association of the ratio of waist to hip girth with blood pressure, serum cholesterol and serum uric acid in children and youths aged 6 17 years. J Chronic Dis 1987; 40: 413 420. 19 Barker DJP, Martyn CM, Osmond C, Hales CN, Fall CHD. Growth in utero and serum cholesterol concentrations in adult life. Br Med J 1993; 307: 1524 1527. 20 Fall CHD, Barker DJP, Osmond C, Winter PD, Clark PMS, Hales CN. Relation of infant feeding to adult serum cholesterol concentration and death from ischaemic heart disease. Br Med J 1992; 304: 801 805. 21 National Heart Lung and Blood Institute. Report of the Task Force on research in epidemiology and prevention of cardiovascular diseases. National Institutes of Health: Bethseda, MD, 1994. 22 Williams CI, Wynder EL. Hyperlipidemia in childhood and the development of atherosclerosis. Ann NY Acad Sci 1991; 623: 1 482. 23 Golding J, ALSPAC. Children of the nineties: a resource for assessing the magnitude of long-term effects of prenatal, perinatal and subsequent events. Obstetrics 1996; 8: 89 92. 24 Friedewald WT, Levy RI, 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 Brambilla P, Manzoni P, Sironi S, Del Maschio A, Di Natale B, Chiumello G. Peripheral and abdominal obesity in childhood obesity. Int J Obes 1994; 18: 795 800. 26 Freedman DS, Srinivasan SR, Cresanta JL, Webber LS, Berenson GS. Serum lipids and lipoproteins. Pediatrics 1987; 80(Suppl): 789 796. 27 Freedman DS, Lee SL, Byers T, Kuester S, Sell KI. Serum cholesterol levels in a multiracial sample of 7,439 preschool children from Arizona. Prev Med 1992; 21: 162 176. 28 McGill HC, Mott GE, Lewis DS, McMahan CA, Jackson EM. Early determinants of adult metabolic regulation: effects of infant nutrition on adult lipid and lipoprotein metabolism. Nutr Rev 1996; 54: S31 S40. 29 Rogers IS, Emmett PM, Golding J. The growth and nutritional status of the breast-fed infant. Early Hum Dev 1997; 49(Suppl): S157 S174. 30 Tanner JM. Growth from birth to two. A critical review. Acta Med Auxol 1994; 27: 7 45. 31 ALSPAC. Children of the nineties. http:==alspac2.ich.bris. ac. uk=alspacext=default.html. 32 Sporik R, Johnstone JH, Cogswell JJ. Longitudinal study of cholesterol values in 68 children from birth to 11 years of age. Arch Dis Childhood 1991; 66(Suppl): 134 137. 33 Gregory JR, Collins DL, Davies PSW, Hughes JM, Clarke PC. The National Diet and Nutrition Survey: children aged 1 1=2 to 4 1=2 years. Volume I. Report of the Diet and Nutrition Survey. HMSO: London, 1995.

International Journal of Obesity

Blood lipids in pre-school children I Cowin et al

34 Labarthe DR, Nichaman MZ, Harrist RB, Grunbaum JA, Dai S. Development of cardiovascular risk factors from ages 8 to 18 in Project HeartBeat! Circulation 1997; 95: 2636 2642. 35 Department of Health Report on Health and Social Subjects. Nutritional aspects of cardiovascular disease. HMSO: London, 1994. 36 Fuller WA. Measurement error models. Wiley: New York, 1987. 37 Taylor RW, Keil D, Gold EJ, Williams SM, Goulding A. Body mass index, waist girth, and waist-to-hip ratio as indexes of total and regional adiposity in women: evaluation using receiver operating characteristic curves. Am J Clin Nutr 1998; 67: 44 49. 38 Seidell JC, Cigolini M, Charzewska J, Ellsinger BM, Despres JP, Cruz A. Fat distribution in European men: a comparison of anthropometric measurements in relation to cardiovascular risk factors. Int J Obes 1992; 16: 17 22.

39 Lovejoy JC, de la Bretonne JA, Klemperer M, Tulley R. Abdominal fat distribution and metabolic risk factors: effects of race. Metabolism 1996; 45: 1119 1124. 40 Huttunen JK, Saarinen UM, Kostiainen E, Siimes MA. Fat consumption of the infant diet does not inuence subsequent serum lipid levels in man. Atherosclerosis 1983; 46: 87 94. 41 Jooste PL, Rossouw LJ, Steenkamp HJ, Rossouw JE, Swanepoel ASP. Effect of breastfeeding on the plasma cholesterol and growth of infants. J Pediatr Gastroenterol Nutr 1991; 13: 139 142. 42 Osmond C, Barker DJP, Winter PD, Fall CHD, Simmonds SJ. Early growth and death from cardiovascular disease in women. Br Med J 1993; 307: 1519 1524. 43 Brown MJ, Hughes M, White R, Ashby-Swain J, Trutwein D, Bailie M, Hopper R, Molton OR. Blood pressure, cholesterol and height. The birth weight hypothesis revisited in 15696 adults. J Hypertens 1992; 10: 1428.

339

International Journal of Obesity