Association of Lipid Abnormalities With Measures and Severity of Adiposity and Insulin Resistance Among Overweight Children and Adolescents

Abstract

Obesity and lipid abnormalities in children may increase premature cardiovascular disease risk, but the relationship of dyslipidemia with adiposity among obese children is not well defined. The authors performed a cross‐sectional analysis of children and adolescents (N=698) in 3 age groups (3–8 years, 9–11 years, and 12–18 years; 53% female, 81% African American, and 16% Hispanic) attending an obesity treatment program. More than 50% of the sample had abnormal levels of triglycerides (TG) or high‐density lipoprotein (HDL) cholesterol or both. Only HDL cholesterol and TG were significantly associated with adiposity measures and insulin resistance (measured by homeostasis model assessment [HOMA]) and only in adolescents. All measures of adiposity, adjusted for age and sex, among adolescents were modest predictors of abnormal TG and HDL cholesterol, but these associations were attenuated when adjusting for HOMA. Despite the high prevalence of dyslipidemia in overweight children and adolescents, severity of adiposity appears to be a poor predictor of lipid values except among adolescents. Insulin resistance may in part mediate the relationship of adiposity and dyslipidemia among obese adolescents.

Atherosclerotic cardiovascular disease is one of the important causes of morbidity and mortality in adulthood that lead to an immense public health burden. ¹ Clinical manifestations of atherosclerotic cardiovascular disease such as angina, ischemia, and hypertension mostly begin to manifest in middle age. The atherosclerotic process begins, however, in early life as shown by postmortem studies, ² , ³ , ⁴ , ⁵ and the severity of lesions is related to antemortem lipoprotein levels. ⁶

There is an increasing prevalence of obesity in children that is accompanied by increasing insulin resistance (IR), dyslipidemia, and hypertension. ⁷ The prevalence of the so‐called metabolic syndrome is reported to be up to 50% in severely obese children. ⁸ Just recently, the American Academy of Pediatrics released revised guidelines that include screening overweight children with a fasting lipid profile and treating patients with significantly elevated low‐density lipoprotein (LDL) cholesterol. ⁹ However, universally accepted cutpoints for children based on age, sex, and race are not available to implement the new guidelines. ⁹

Key features of obesity‐related dyslipidemia include raised levels of triglycerides (TG), reduced high‐density lipoprotein (HDL) cholesterol values, and increased numbers of small, dense LDL particles. ⁹ This adverse combination can be associated with a normal total cholesterol (TC) value when HDL cholesterol levels are low; thus, it can be missed if TC alone is measured. Although obesity is associated with dyslipidemia, not all obese children have dyslipidemia, and factors influencing abnormal lipid levels among obese children are not well defined.

The objective of our study was to assess the age‐specific associations of lipid abnormalities with different measures and severity of adiposity in a large group of inner‐city overweight, nonwhite children and adolescents and to identify how IR is related to these associations.

Design and Methods

Data were collected on overweight children and adolescents (3–19 years) with a body mass index (BMI) >95th percentile for age, who attended a pediatric obesity program at an inner‐city university hospital in Brooklyn, New York, from January 1, 2003, through December 31, 2006. Overall, 698 patients with complete data for anthropometric and metabolic risk factors constituted the study sample. The study was approved by the internal review board, and informed consent to participate in the community‐based obesity program and to be included in data analyses was obtained from the parent or guardian with assent from the children when first enrolled. Fasting plasma concentrations of glucose, insulin, TG, and HDL cholesterol were assessed using standard hospital laboratory methods. Weight and height were measured to the nearest 0.1 kg and 0.5 cm, respectively. BMI was calculated as weight in kilograms divided by height in meters squared and converted to BMI‐z score (BMIZ) standard units according to US Centers for Disease Control age/sex tables. ¹⁰ Waist circumference was measured at the midpoint between the lateral iliac crest and the lowest rib in centimeters during expiration and defined as abnormal if >90th percentile for age, sex, and race. ¹¹ Adiposity was assessed alternately by waist circumference, waist‐to‐height ratio, and BMIZ. Lipid values included fasting TC, HDL cholesterol, LDL cholesterol, and TG. TC/HDL cholesterol and TG/HDL cholesterol were also calculated as lipid measures. Abnormal lipid levels were defined as fasting TG values >90th percentile for age and sex, HDL cholesterol levels <40 mg/dL, and LDL cholesterol values >110 mg/dL. ¹² Dyslipidemia was considered as abnormal values of TG (≥90th percentile) or HDL cholesterol (<40 mg/dL) or both for age. An abnormal glucose value was defined as ≥100 mg/dL. Homeostasis model assessment (HOMA; fasting serum insulin [μU/mL] × fasting plasma glucose [mmol/L]/22.5) was used as a measure of IR. Blood pressure was measured with an appropriate‐sized cuff with a standardized automated dynamap in the right arm in the sitting position, and the average of 3 readings was recorded for analysis. Elevated systolic and diastolic blood pressures were defined as values >90th percentile for age and sex. ¹³ The prevalence of the metabolic syndrome defined by modified Cook criteria ¹⁴ (abnormal values for ≥3 of waist circumference, HDL cholesterol, TG, blood pressure, and glucose) was assessed for all 3 groups.

Statistical Analyses

Participants were divided into 3 age categories: 3 to 8 years (n=143), 9 to 11 years (n=202), and 12 to 19 years (n=353), similar to our previous study, ¹⁵ to assess the effect of age and puberty on lipid profiles. Bivariate associations of patient characteristics with age group were assessed with chi‐square for categoric variables and analysis of variance for continuous variables. Bivariate associations of HOMA and adiposity measures with lipid values were assessed with Pearson product‐moment correlation. Associations of abnormal TG, HDL cholesterol, or both were tested for linear trend across quartiles of waist circumference and HOMA within age group. Multivariate linear and logistic regression models were constructed to assess the age‐ and sex‐adjusted associations of waist circumference with lipid measures both with and without adjusting for HOMA. Statistical analyses were performed with SPSS (SPSS Inc, Chicago, IL) and Stata (StataCorp LP, College Station, TX). Statistical significance was denoted as P< 2‐tailed alpha of .05.

Results

Of the 698 participants, 143 were aged 3 to 8 years (group I), 202 were aged 9 to 11 years (group II), and 353 were aged 12 to 19 years (group III). In total, 81% were black and 16.2% were Hispanic. Table I describes sample characteristics and metabolic risk factors according to age group. Metabolic risk factors as continuous variables tended to get significantly worse with increasing age except for values of LDL cholesterol and TC. The prevalence of abnormal HDL cholesterol levels increased with age (36%, 38%, and 45% for groups I, II, and III, respectively), but LDL cholesterol did not follow that pattern (45%, 42%, and 43%), nor did TG (33%, 34%, and 24%). Dyslipidemia (abnormal levels of HDL cholesterol, TG, or both) was present in 52%, 53%, and 53% of groups I, II, and III, respectively. The prevalence of the metabolic syndrome overall by age group was 42%, 37%, and 37%, respectively, while elevated TC declined with age group (55%, 48%, and 45% of groups I, II, and III, respectively).

Table I.

 Demographic and Metabolic Syndrome Characteristics by Age Group

Characteristic, %	Ages 3–8 (n=143)	Ages 9–11 (n=202)	Ages 12–19 (n=353)	Total (N=698)	P Valuea
Female	54.5	51.0	54.4	53.4	.71
Race
African American	79.7	81.2	81.3	80.9	.98
Hispanic	17.5	16.3	15.6	16.2
Other	2.8	2.5	3.1	2.9
Continuous variables, mean±SD
Insulin, μU/mL	13.5±14.4	18.5±13.3	20.5±14.2	18.5±14.2	<.001
HOMA	2.8±3.3	3.9±3.1	4.4±3.3	3.9±3.3	<.001
Q Index	0.35±0.04	0.33±0.03	0.32±0.03	0.33±0.04	<.001
Glucose, mg/dL	80.0±10.1	84.0±10.4	85.0±18.8	84.0±15.3	.01
Waist circumference, cm	81.0±11.6	95.0±13.2	108.0±15.8	99.0±17.9	<.001
Hip circumference, cm	87.0±12.4	104.0±12.4	120.0±13.5	109.0±18.3	<.001
Waist‐to‐hip ratio	0.92±0.09	0.91±0.08	0.90±0.08	0.91±0.08	.01
Waist‐to‐height ratio	0.64±0.08	0.63±0.09	0.66±0.09	0.65±0.09	.008
Height, cm	127.0±11.9	150.0±9.3	165.0±10.0	153.0±18.0	<.001
Weight, kg	44.0±13.5	71.0±17.2	102.0±23.4	81.0±30.4	<.001
BMI, kg/m2	26.9±5.0	31.2±5.8	37.2±7.3	33.3±7.7	<.001
BMIZb	2.8±0.68	2.4±0.40	2.4±0.36	2.5±0.48	<.001
HDL cholesterol, mg/dL	45.0±11.5	44.0±10.4	42.0±10.1	43.0±10.6	.002
Triglycerides, mg/dL	83.0±33.7	94.0±42.1	96.0±51.8	93.0±46.1	.02
LDL cholesterol, mg/dL	111.0±28.4	106.0±29.0	107.0±31.6	108.0±30.2	.39
TC, mg/dL	172.0±32.7	170.0±31.4	169.0±36.5	170.0±34.3	.75
TC/HDL cholesterol ratio	3.9±.98	4.1±1.4	4.2±1.2	4.1±1.2	.06
TG/HDL cholesterol ratio	2.0±1.0	2.3±1.4	2.5±1.8	2.4±1.6	.002
Blood pressure, mm Hg
Systolic	113±10	118±9	124±11	120±11	<.001
Diastolic	65±8	69±8	73±9	70±9	<.001
Categoric variables, %
Glucose ≥100 mg/dL	2.8	4.0	4.8	4.2	.59
Increased waist circumferencec	95.1	96.0	87.0	91.3	<.001
HDL cholesterol <40 mg/dL	35.7	38.1	44.8	41.0	.11
Elevated triglyceridesc	32.9	34.2	24.1	28.8	.02
Hypertensionc	63.6	54.5	57.2	57.7	.23
Metabolic syndromed	42.0	37.6	37.4	38.4	.62
LDL cholesterol >110 mg/dL	44.8	42.1	42.5	42.8	.87
TC >170 mg/dL	54.5	48.0	44.5	47.6	.13

Abbreviations: BMI, body mass index; BMIZ, BMI‐z score; HDL, high‐density lipoprotein; HOMA, homeostasis model assessment; LDL, low‐density lipoprotein; TC, total cholesterol. ^aComparing age groups by chi‐square (categoric variables) or analysis of variance (continuous variables). ^bStandard units for age and sex. ^c≥90th percentile for age and sex. ^dBy modified National Cholesterol Education Program Adult Treatment Panel III criteria: abnormal values for ≥3 of waist circumference, HDL cholesterol, triglycerides, blood pressure, and glucose.

HOMA, waist circumference, and waist‐to‐height ratio were significantly correlated with HDL cholesterol (inversely), TG, TC/HDL cholesterol, and TG/HDL cholesterol (all P≤.02) but not with TC or LDL cholesterol among 12‐ to 19‐year‐olds. Similarly, BMIZ was significantly correlated with these, and LDL cholesterol as well, in the 12 to 19 age group. No such significant correlations were observed in the 3 to 8 or 9 to 11 age groups, except that TG had a borderline correlation (P=.04) with HOMA among 9‐ to 11‐year‐olds and LDL cholesterol was correlated with HOMA (P=.01) in the 3‐ to 8‐year‐olds. Since only 2 out of a possible 48 bivariate correlations were statistically significant in these younger (3–8 and 9–11 years) age groups, subsequent multivariable analyses were performed only for the 12‐ to 19‐year‐olds.

There were clear linear trends among 12‐ to 19‐year‐olds of proportions who had abnormal values of TG or HDL cholesterol and increasing quartiles of waist circumference (Figure, panel A) and HOMA (Figure, panel B). No such linear trends were observed in the 3‐ to 8‐year‐olds, and there were only weak trends among 9‐ to 11‐year‐olds (data not shown).

Figure.

 Data for participants aged 12 to 19 years (A) by age‐ and sex‐specific quartiles of waist circumference ( P values for trend: triglycerides, P=.16; high‐density lipoprotein cholesterol, P<.001; either, P=.001) and (B) by age‐ and sex‐specific quartiles of homeostasis model assessment ( P values for trend: triglycerides, P<.001; high‐density lipoprotein cholesterol, P=.02; either, P=.01). TG indicates abnormal trigylcerides; HDL, abnormal high‐density lipoprotein cholesterol.

When multivariable linear regression models were constructed for 12‐ to 19‐year‐olds, simultaneously adjusting for sex, age, and race, each of the 3 measures of adiposity (waist circumference, BMIZ, and waist‐to‐height ratio) were all similarly significant predictors of TG and HDL cholesterol (data not shown). For subsequent modeling, only waist circumference, as the most direct measure of central adiposity, was used. Waist circumference was a significant predictor of HDL cholesterol (P=.002) and a borderline significant predictor of TG (P=.06) (Table II). However, when HOMA was added to the models for TG, the regression coefficient for waist circumference was reduced by >60%. For HDL cholesterol, this attenuation of the regression coefficient was about 30%. HOMA was a highly significant predictor (P≤.001) of TG and HDL, even with waist circumference in the model. When elevated TG levels and abnormal HDL cholesterol values were used as outcomes in binary logistic models (Table III), the addition of HOMA to the model led to some attenuation of the association with waist circumference, but not nearly as much as for the linear models in which TG and HDL cholesterol values were used as continuous variables.

Table II.

 Linear Associations of Waist Circumference and HOMA With Dyslipidemic Outcomes Among 12‐ to 19‐Year‐Olds

Outcome	Model 1 Without HOMA	Model 2 With HOMA	HOMA
	Waist Circumference a	Waist Circumference a
Triglycerides, mg/dL	3.26 (−0.11 to 6.62) P=.06	1.11 (−2.25 to 4.47) P=.52	3.86 (2.34 to 5.37) P<.001
High‐density lipoprotein cholesterol, mg/dL	−1.07 (−1.75 to −0.39) P=.002	−0.77 (−1.46 to −0.078) P=.03	−0.54 (−0.85 to −0.23) P=.001

Abbreviation: HOMA, homeostasis model assessment. Regression coefficients (95% confidence intervals) estimated in linear regression models adjusting for age, sex, and race. ^aPer 10 cm.

Table III.

 Adjusted Odds Ratios for Dyslipidemic Outcomes Among 12‐ to 19‐Year‐Olds

Outcome	Model 1 Without HOMA	Model 2 With HOMA	HOMA
	Waist Circumference a	Waist Circumference a
Elevated triglycerides	1.12 (0.95 to 1.31) P=.18	1.04 (0.87 to 1.23) P=.68	1.14 (1.06 to 1.23) P<.001
Abnormal high‐density lipoprotein cholesterol	1.27 (1.10 to 1.47) P=.001	1.23 (1.06 to 1.42) P=.01	1.07 (1.00 to 1.15) P=.05

Abbreviation: HOMA, homeostasis model assessment. Odds ratios (95% confidence intervals) estimated in logistic regression models adjusting for age, sex, and race. ^aPer 10 cm.

Discussion

These data suggest that among a predominantly nonwhite population of overweight children and adolescents, there is a high prevalence of atherogenic dyslipidemia: over half of the patients had an elevated TG value and/or low HDL cholesterol level, which is typical of the metabolic syndrome. Statistically significant associations with adiposity measures and IR were evident for HDL cholesterol and TG only in the adolescent age group. Notably, although there was a substantial prevalence of dyslipidemia and metabolic syndrome in the 2 younger age groups, they were not associated with severity of adiposity measures or IR among these overweight children within the age groups younger than 12. For adolescents aged 12 to 19, among measures of adiposity in multivariate regression analysis adjusted for age and sex, all measures of adiposity were modest predictors of abnormal levels of TG and HDL cholesterol. These associations were attenuated when IR was included as a covariate. In fact, HOMA as a measure of IR appeared to be a better predictor of both HDL cholesterol and TG as continuous variables in the older age group, but again none of these associations were present for LDL cholesterol or TC. More than 40% of participants in all age groups, however, had elevated LDL cholesterol levels.

Current cholesterol screening guidelines in children are based on reported evidence that high plasma cholesterol and LDL cholesterol concentrations are associated with an increased risk of coronary heart disease (CHD) in adults. ¹⁶ However, as shown in the Framingham Heart Study, ¹⁷ measuring LDL cholesterol levels along with other traditional risk factors, even in adults, such as hypertension, cigarette smoking, diabetes, and family history of CHD identifies only approximately 50% of the population in whom the disease will eventually develop. Studies have also shown that there is a considerable overlap in plasma cholesterol concentrations among patients with and without CHD, with about 50% of patients with the disease showing evidence of normal plasma TC. ¹⁸ Thus, the ability of TC alone to accurately discern persons at risk for CHD is relatively weak.

Beyond traditional risk factors and LDL cholesterol levels, a collection of other, nontraditional risk factors associated with obesity have emerged including TG and HDL cholesterol levels, homocysteine levels, prothrombotic factors, and IR that are associated with increased risk of CHD. It is now well established that this clustering of nontraditional risk factors is linked to obesity, especially abdominal obesity, which reflects the metabolically active visceral adipose tissue. ¹⁹ However, some obese individuals could maintain normal metabolic function. ²⁰ Variations in the nature and magnitude of obesity‐related dyslipidemia are likely due to the interaction of genetic factors with environmental influences, most notably diet and physical activity. ²¹ Our study also supports the finding that a high TG/low HDL cholesterol phenotype is associated with overweight status and IR. This is distinctly different from LDL cholesterol elevations in children, which may be familial and although present may not be directly related to obesity. While it has been shown that among obese patients, LDL particle size is considerably more important than the overall LDL cholesterol level (with small, dense particles being more atherogenic ²² ), assessing particle size is currently not practical or available in clinical settings. We found a high prevalence of elevated LDL cholesterol values in all age groups; however, we were not able to measure the proportion of small, dense particles.

The most important data showing the association of the high TG/low HDL cholesterol phenotype with IR and obesity came from the Bogalusa Heart Study. ²³ , ²⁴ A series of cross‐sectional and longitudinal analyses involving children with several decades of follow‐up elucidated these typical associations now called the metabolic syndrome. They concluded that overweight in childhood tends to persist into adulthood; that independent cardiovascular risk factors such as elevated blood pressure levels and lipid abnormalities are associated with the presence of overweight or elevated insulin; that the multiple associated risk factors also persist into adulthood; and that elevated insulin values as a child are associated with multiple risk factors later in life. The obesity epidemic has now made this association a common finding in regular clinical practice among overweight children and adolescents, as also shown in our study. Even among the 2 Bogalusa Heart study cohorts, those recruited in 1984 gained significantly more weight during the 8‐year follow‐up compared to the group recruited in 1972. The weight gain in the 1984 group was associated with higher TG and lower HDL cholesterol levels compared to the 1972 cohort. ²⁵ The Minneapolis Children’s Blood Pressure Study ²⁶ also showed that future dyslipidemia and IR are associated with both overweight status at initial evaluation and weight change. Although other studies have shown an increasing prevalence of this dyslipidemia with age, ²⁷ in this study, which was cross‐sectional and based on currently accepted cutpoints, even the youngest group had a very high prevalence of dyslipidemia that was similar to the older groups. This has not been previously demonstrated, and our study only looked at overweight children. However, in these younger age groups there were no associations between the severity of different measures of obesity or IR, which suggests that the cutpoints used in younger children to identify abnormal levels may not be optimal or that the associations with IR may not be evident or detected by currently used methods. Further longitudinal studies are needed to track progression of risk over time in these young overweight children with dyslipidemia compared to overweight children who are metabolically normal by currently available cutpoints. Because our study was cross‐sectional and patients were not followed over time, we could not assess whether lipid levels in the younger age groups were predictive of subsequent outcomes. Alternately, once overweight is present, severity of different adiposity measures appears to be a poor predictor of lipid values, except in the older age group in which a modest association was observed. Given that >90% of the patients had increased waist circumference, it was not possible to determine which measure of obesity confers a greater risk of dyslipidemia among overweight children and adolescents. In other words, in this study, BMIZ, waist circumference, and waist‐to‐height ratio all predicted dyslipidemia risk reasonably well in the obese adolescents, and any one measure may be a useful screening tool for assessing severity of adiposity in a population of overweight persons. HOMA added to the ability of multivariate models to predict HDL cholesterol, TG, and TC/HDL cholesterol ratio among adolescents, but among preadolescents, neither adiposity nor HOMA were observed to be linearly related to lipid values.

The role of obesity and IR in relation to pediatric dyslipidemia is currently an important issue for further research. In our study, the associations of severity of adiposity and HOMA with lipid levels were modest (adolescents) or hardly noticeable (preadolescents). Nevertheless, substantial proportions (>50%) in each age group had lipid or TG values that could be considered abnormal, at least by population‐wide norms adjusted for age and sex. This means that among the obese, lower or higher measures of obesity cannot be reliably used as an indicator of dyslipidemia status. On the contrary, direct measurement of lipid values seems warranted.

Because there is evidence linking abnormal lipoprotein levels in children and adolescents with preclinical atherosclerosis, it may be appropriate to extrapolate the adult risk of dyslipidemia to younger ages, although the validity of such extrapolation needs to be studied. Given that lifestyle and pharmacologic interventions at a young age are effective in modifying lipoprotein levels and markers of atherosclerosis, ²⁸ , ²⁹ , ³⁰ it is important to identify those at risk and assess whether modification at an early age will improve clinical as well as intermediate outcomes.

Appropriate lipoprotein cutoffs in children that would predict adult disease need to be clearly defined and standardized. Recently, Magnussen and associates ³¹ compared the National Health and Nutrition Examination Survey (NHANES) lipoprotein classifications for children and adolescents with those of the National Cholesterol Education Program (NCEP) in predicting abnormal adult lipoprotein levels. The study used data from 3 longitudinal cohort studies (Childhood Determinants of Adult Health study, Cardiovascular Risk in Young Finns, and the Bogalusa Heart Study) and showed that NHANES cutpoints were more strongly predictive of low adult values of HDL cholesterol than were NCEP criteria. However, NCEP cutpoints predicted elevated TC, LDL cholesterol, and TG levels in adulthood. The NHANES cutpoints may be more reflective of the lipid profile of today’s adolescent, as these data were collected during the obesity epidemic. Based on the study by Magnussen and colleagues, ³¹ neither universal screening nor selective screening for HDL cholesterol and TG were efficient in identifying adolescents who had abnormal lipid values as adults. In identifying those with elevated TC and LDL cholesterol levels that persisted from childhood into adulthood, these methods similarly showed high rates of false‐positives. In fact, the Magnussen study showed that obesity/overweight is a better predictor of adult HDL cholesterol levels than are childhood cholesterol levels. Thus, obesity may be the primary risk factor among adolescents that is associated with adult cardiovascular risk. The revised guidelines on the evaluation and treatment of dyslipidemia address this issue, including lifestyle modifications for those with elevated TG and low HDL cholesterol values related to obesity and pharmacologic therapy for children with elevated LDL cholesterol levels. ⁹ Our data also suggest that obesity by itself is an important risk factor for screening for dyslipidemia given the high prevalence in all overweight children and adolescents, although the association with the degree and type of obesity or IR is not evident until adolescence. Because the association among obesity, IR, and elevated lipid levels seems to develop with age in these overweight children, it may be prudent to infer that obesity should be a target for early intervention irrespective of lipid levels and should be intensified in children with existing dyslipidemia and IR.