Cardiometabolic Risks and Severity of Obesity in Children and Young Adults

Data Source and Study Design

We obtained data from the National Health and Nutrition Examination Survey (NHANES, 1999–2012). The NHANES includes a stratified, multistage probability sample of the civilian, noninstitutionalized U.S. population. The data obtained included responses to an in-home interview on a variety of demographic variables and health topics, findings from a physical examination performed at a mobile examination center that included measured height and weight, and laboratory measurements.⁵ The National Center for Health Statistics (NCHS) designed the NHANES and collected the data after obtaining written informed consent from the participants or their parents or guardians and assent from minors; most components of the NHANES, including those used in the current study, are publicly available. The first author designed and performed the analyses, with no input from the NCHS, and assumes responsibility for the analyses. The institutional review board at the University of North Carolina at Chapel Hill, in accordance with the Code of Federal Regulations, deemed this study to be exempt from further review and from the requirement to obtain written informed consent because it used only deidentified secondary data.

Study Sample

Table 1. Table 1. Definitions of Abnormal Values for Risk-Factor Variables.

Our study sample included NHANES participants who were 3 to 19 years of age at the time of examination. We excluded 513 children and young adults with missing BMI values. Given our focus on obesity, we excluded 807 participants who were underweight and 15,469 participants who had a healthy weight (i.e., all participants who were below the 85th percentile for age-specific and sex-specific BMI, as defined below). Thus, the final sample comprised 8579 children and young adults. Because certain NHANES measures have a more targeted sampling frame (e.g., limited to specific ages), specific reported measures had different sample sizes (Table 1).⁵

Weight Status

We classified weight status using height and weight measurements obtained at the time of the physical examination component of the NHANES to calculate BMI and to determine the BMI percentile, which we derived from the Centers for Disease Control and Prevention (CDC) growth charts using the SAS code that was developed for this purpose.⁶ In 2011–2012, the NHANES reported age in years for children and young adults 2 to 19 years of age, rather than age in months as in previous releases; age in months was reported only for children from birth to 2 years of age. To estimate the prevalence of obesity consistently across our study period, we used the midpoints of the whole-year ages of the children in all calculations (e.g., an 11-year-old child was considered to be 11.5 years of age). This method of calculation yielded prevalence rates that were similar to calculations based on age in months, with differences of 0.1 to 0.2 percentage points,¹ and was used for all 2-year cycles of the NHANES to ensure consistency in the determinations of weight status.

The weight categories according to age-specific and sex-specific BMI percentiles that we used in the current study were as follows: overweight (≥85th to <95th percentile), class I obesity (≥95th percentile to <120% of the 95th percentile), class II obesity (≥120% to <140% of the 95th percentile, or BMI ≥35, whichever was lower), and class III obesity (≥140% of the 95th percentile, or BMI ≥40, whichever was lower). The definitions of healthy weight, overweight, and obesity were based on the recommendations of the CDC,⁷ and the definition of class II obesity was based on the recommendation of the American Heart Association.⁴ We used the range of 120% to less than 140% of the 95th percentile to define class II obesity instead of the 99th percentile or higher because the former definition has been shown to have more stability in estimation procedures.⁸ We used 140% of the 95th percentile to define class III obesity because it approximated a BMI of 40 in late adolescence, the same age at which 120% of the 95th percentile approximated a BMI of 35.¹ Among adults, persons with these ranges of BMI are considered to be at a higher risk for earlier death.⁹ The BMI percentiles in our study were not derived from the study sample we evaluated but were defined by the CDC growth charts that used a historical sample; this allowed for a consistent categorization of weight status throughout the entire study sample.

Cardiometabolic Risk Factors

Because we evaluated children and young adults, we examined variables that are associated with known cardiometabolic risk rather than hard end points of cardiovascular events. Total cholesterol and high-density lipoprotein (HDL) cholesterol levels were measured in the full study sample in the targeted age ranges, whereas measurements of low-density lipoprotein (LDL) cholesterol and triglyceride levels were limited to a subgroup of participants who underwent an examination in the morning as part of the NHANES study procedures. Fasting times varied among participants in whom fasting laboratory studies were performed. We used standard cutoff values for levels of total cholesterol (≥200 mg per deciliter [>5.18 mmol per liter]), HDL cholesterol (<35 mg per deciliter [<0.9 mmol per liter]), LDL cholesterol (≥130 mg per deciliter [≥3.4 mmol per liter]), and triglycerides (≥150 mg per deciliter [>1.7 mmol per liter]) to define abnormal values; when the recommendations for cutoff values were inconsistent, we chose the more conservative definition.^10-12

Blood pressure was recorded as the mean value of up to three measurements or as a single measurement (86% of the children and young adults had three measurements, 8% had two, and 6% had one). For children younger than 18 years of age, we used standardized blood pressure tables in which abnormal values were determined according to age, sex, and height; abnormal values were defined as any value that was in at least the 95th percentile in those tables.¹³ For young adults 18 and 19 years of age, we used cutoffs of 140 mm Hg for systolic blood pressure and 90 mm Hg for diastolic blood pressure.¹⁴

Glycated hemoglobin levels were measured in all participants 12 years of age or older as part of the standard NHANES laboratory measurements; a value greater than 5.7% was defined as abnormal. Fasting glucose was measured in the subgroup of NHANES participants who attended the morning session; a value of 100 mg per deciliter or higher (≥5.6 mmol per liter) was defined as abnormal. The definitions of abnormal values for glycated hemoglobin and fasting glucose were based on recommendations by the American Diabetes Association for identifying persons at high risk for diabetes.¹⁵

Detailed information about the collection and measurement procedures of the NHANES can be found in the NHANES Laboratory Procedures Manual.¹⁶ The specific measures obtained and the age group for which they were obtained varied by NHANES cycle. The sampling frame for each laboratory measure is detailed in the NHANES content brochure.⁵ The ages for the sampling frame, the total sample size, and the definitions used for abnormal values are shown in Table 1. We did not limit lipid values to those measured when the participants were fasting because previous research has shown that the differences in values according to fasting status are minimal.¹⁷

Statistical Analysis

The primary results of the bivariate analysis are presented as mean values for cardiometabolic variables and for prevalence of cardiometabolic risk factors, according to weight status. All differences were examined with the use of adjusted Wald tests. We report only the results of joint Wald tests of differences across all four weight categories and include 95% confidence intervals to allow for further interpretation of the differences between specific groups (e.g., class II obesity as compared with class III obesity). We report mean values for all children and young adults, as well as for male and female participants separately. Because the quantification of risk is critical to policy development, we report the results of a further examination of the prevalence of cardiometabolic risk factors by weight category according to subgroups defined by age, sex, and race or ethnic group. Age, sex, and race or ethnic group categories were specified before analysis, and we report all cardiometabolic risk factors for each. We provide sample sizes throughout our results for reference, but all reported mean values, percentages, prevalence values, and results of statistical tests are weighted to represent the U.S. population.

Multivariable analyses were performed with the use of generalized linear models with a logarithmic link. Exponentiation of the coefficients was used to report risk ratios for the effect of obesity severity on cardiometabolic risks. All analyses were adjusted for the strata, primary sampling units, and probability weights used in the complex survey design of the NHANES^18,19 and were performed with the use of the survey estimation routines in Stata software, version 13.1 (StataCorp). We pooled all years and adjusted the probability weight accordingly, as directed by the NCHS.¹⁹ Because our primary interest was the relationship between severe obesity and cardiometabolic risks and because the prevalence of obesity changed minimally during the time frame of our study,¹ we did not perform any time-trend analysis in this study. P values of less than 0.05 were considered to indicate statistical significance, although the differences between the groups are better represented by the confidence intervals, and we believe that the focus should be on the clinical importance of the estimates rather than on statistical significance. Finally, missing data were due primarily to the design of the NHANES, and therefore no additional imputation of data was performed.

Table 2. Table 2. Demographic Variables and Sample Size Weighted for Survey Design.

Among the 8579 children and young adults with a BMI at the 85th percentile or higher, 46.9% were overweight, 36.4% had class I obesity, 11.9% had class II obesity, and 4.8% had class III obesity (Table 2). Table S1 in the Supplementary Appendix (available with the full text of this article at NEJM.org) shows the mean values for each cardiometabolic variable in all participants and separately in male and female participants. With few exceptions, in both male and female participants, the mean values for the cardiometabolic variables were higher with greater severity of obesity; for HDL cholesterol, the mean values were lower with greater severity of obesity.

Table 3. Table 3. Prevalence of Abnormal Values for Risk-Factor Variables by Age and Weight Category.

Figure 1. Figure 1. Prevalence of Cardiometabolic Risk Factors by Weight Status.

The study sample comprised 8579 children and young adults 3 to 19 years of age in the National Health and Nutrition Examination Survey from 1999 through 2012. Weight categories were defined as follows: overweight (≥85th to <95th percentile for age-specific and sex-specific BMI, according to the Centers for Disease Control and Prevention growth charts), class I obesity (≥95th to <120% of the 95th percentile), class II obesity (≥120% to <140% of the 95th percentile), and class III obesity (≥140% of the 95th percentile). HDL denotes high-density lipoprotein, and LDL low-density lipoprotein.

Table 3 and Figure 1 show the prevalence of abnormal values for each cardiometabolic variable. As was the case with mean values for cardiometabolic variables, the prevalence of abnormal values was higher with greater severity of obesity. Table 3 also shows the statistical significance of differences across all weight categories, although the multivariable analyses described below better represent the differences between individual categories. We report our findings for all age groups; however, the sample sizes for children 3 to 5 years of age were very small, and the associated point estimates for those participants must be interpreted with caution. Among participants 12 to 19 years of age, the prevalence of all risk factors, with the exception of LDL cholesterol, increased by weight category, and most risk factors increased by weight category among children 6 to 11 years of age. Prevalence according to sex and race or ethnic group are shown in Tables S2 and S3 in the Supplementary Appendix, respectively.

Table 4. Table 4. Risk Ratios for Cardiovascular Risk Factors by Sex and Weight Category.

Table 4 shows the results from multivariable general linear models (see Table S4 in the Supplementary Appendix for data on all variables). In models that controlled for age, race or ethnic group, and sex, the risks of low HDL cholesterol level, high systolic blood pressure, high diastolic blood pressure, high triglyceride level, and high glycated hemoglobin level were greater among children and young adults with class III obesity than among those with class I obesity. There were few significant differences in these variables according to weight category among female participants, but all the differences remained significant among male participants. As expected, overweight children and young adults had a lower risk of most risk factors than did those with class I obesity.

Severe obesity in children and young adults is associated with a high prevalence of abnormal levels of cardiometabolic variables. A high prevalence of abnormal values for certain variables among children and young adults with class II or class III obesity may provide important information beyond that identified with the use of standard obesity classifications, especially for boys and young men. The greater prevalence among participants 12 to 19 years of age is consistent with a previous report of the overall U.S. population of children and adolescents.²⁰

Determination of the severity of obesity can help identify children and young adults who are at the greatest risk for the negative health effects associated with obesity. Current guidelines for screening do not differentiate among levels of obesity.^7,21 Although the prevalence of abnormal values does not increase with obesity severity in the case of all risk-factor variables, our findings of greater risks of abnormal HDL cholesterol level, systolic blood pressure, and glucose metabolism support the stratification of risk on the basis of the American Heart Association recommendations for classification of a higher level of obesity at 120% of the 95th percentile.⁴ Our findings of the additional risk of abnormal triglyceride and glycated hemoglobin levels provide initial support for further risk stratification according to 140% of the 95th percentile.

The differences between male and female participants in our study are notable; severe obesity is associated with a higher prevalence of abnormal systolic blood pressure, triglyceride level, and glycated hemoglobin level among male participants than among female participants. It is possible that cardiometabolic risk factors develop earlier in boys than in girls. Alternatively, use of the same definitions of obesity for girls and boys may not determine equivalent risk. Our findings differ from those in other reports that used only standard definitions and showed minimal differences between boys and girls.^22-24

Obesity during childhood increases the risk of long-term obesity,^25,26 as well as the risks of substantial complications and death in adulthood.^27,28 Severe obesity during adolescence is associated with significantly higher risks of complications.²⁹ Although prevention remains the primary goal in the management of obesity, focus has shifted to also identifying and treating complications associated with obesity.²¹ Classifying obesity into three categories provides a more fine-tuned approach to identifying patients with the greatest risk of potential complications and death.^28,29 The treatment of childhood obesity is both recommended and effective³⁰; however, intensive therapies, such as medically supervised meal replacement, pharmacologic treatment, and bariatric surgery, confer certain risks and are often costly. In addition, resources are too limited to appropriately service every child with obesity because most treatment clinics and programs are concentrated within tertiary-care pediatric hospitals.³¹ The confluence of risks and limited resources leaves many children with severe obesity and established cardiometabolic risk factors without effective options. The application of a more complex classification of obesity, if it identifies those at highest risk, may permit targeted interventions that could decrease morbidity and mortality and may also be cost-effective.

An accurate classification of children with marked obesity may improve the care of this patient population. The cost of treating complications related to obesity in adults has been estimated to be as high as $147 billion³² and will probably continue to rise.³³ Children with obesity have higher outpatient health care expenditures than those with normal weight, with the increased costs believed to stem from the evaluation and treatment of coexisting conditions³⁴; an improved estimation of risk that is based on obesity severity could lower the costs of evaluation without resulting in missed diagnoses of coexisting conditions. Children with severe obesity already have increased cardiometabolic risk, which may predict the early onset of serious diseases such as hypertension and diabetes.

As older adolescents transition to young adulthood, the recognition that teens with obesity have increased cardiometabolic risk will be important. Current treatment approaches are not as effective at higher levels of obesity than they are at lower levels of obesity.^35-37 The findings in our study, when considered in the context of the negative effect that these risk factors have on subsequent quality of life,³⁸ imply a need for the early identification of abnormal weight gain, primary and secondary prevention of obesity, and systemwide policy changes in the evaluation and management of obesity in health care.³⁹

We examined a range of cardiometabolic risk-factor variables in a large and nationally representative population; however, our study has certain methodologic limitations. First, because this is a cross-sectional study, we could not show causality between obesity and risk factors. Second, although we used standard definitions of abnormal values for risk-factor variables, the cross-sectional design did not allow us to examine the effects of these abnormal values on future morbidity or mortality. Finally, the sample size was small in certain subgroups, particularly in the subgroup of young children, which resulted in wide confidence intervals and point estimates that should be interpreted with caution.

In conclusion, severe obesity in children and young adults is associated with a high prevalence of abnormal levels of cardiometabolic risk-factor variables. The prevalence of these abnormal values among children and young adults appears to be dependent on both age and severity of obesity. The inclusion of class II obesity in the standard obesity classification may assist in the identification of children who could be at the greatest risk for abnormal levels of HDL cholesterol, systolic blood pressure, and glucose, and the inclusion of class III obesity in the standard obesity classification may assist in the identification of those at the greatest risk for abnormal levels of triglycerides, diastolic blood pressure, and glycated hemoglobin.