Prevalence of Obesity and Severe Obesity in US Children, 1999–2016

We are using methods and analyses that are similar to those of previous studies,^5,6,12 which are detailed in brief as follows.

The data source is the NHANES for years 1999–2016. The NHANES is a stratified, multistage probability sample of the civilian, noninstitutionalized US population. Although the NHANES contains multiple components, we used the in-home interview and physical examination here, including measured height and weight. We included all children aged 2 to 19 years. For the present analysis, we used only deidentified secondary data, so it was therefore deemed exempt from further review by the Duke University Health System Institutional Review Board under federal regulation 45CFR§46.101(b).

Weight status is directly measured by using height and weight measurements gathered from the physical examination component of NHANES to calculate age- and sex-specific BMI. We defined overweight and obesity, hereby referred to as class I obesity, to contrast with class II and class III (more severe forms of obesity) by using CDC criteria, which define overweight as age- and sex-specific BMI ≥85th percentile and class I obesity as BMI ≥95th percentile.¹⁴ We defined class II and class III obesity to be consistent with previous reports,^6,12,15 with class II obesity defined as a BMI >120% of the 95th percentile for age and sex or a BMI of ≥35 (whichever is lower) and class III obesity defined as a BMI ≥140% of the 95th percentile for age and sex or a BMI of ≥40 or greater (whichever is lower). These categories were not mutually exclusive; for example, any children or adolescents meeting the criteria for overweight include all the children and adolescents with a BMI ≥85th percentile (even if they are also in the ≥95th percentile).

For years 1999–2010, the NHANES characterized race and ethnicity as non-Hispanic white, non-Hispanic African American, Hispanic, and other race and/or ethnicity. Beginning in 2011, the NHANES included an Asian American oversample, allowing for a more detailed characterization of this group. We include estimates for Asian Americans in years when this information is available. Before 2011, the inclusion of Asian Americans in the other race category made that category different from the other race category of 2011–2016, when Asian Americans were categorized separately. Therefore, we present other race separately for years when Asian Americans were included in the category and present them as their own category from 2011 to 2016, when they were oversampled.

We report the prevalence estimates of overweight and each obesity definition by 2-year NHANES cycles. To test the trends from 1999 to 2016, we report P values from ordinary least squares regression, with the NHANES year as a continuous variable predicting obesity and severe obesity prevalence. To compare the 2 most recent NHANES cycles, we present P values from adjusted Wald tests to compare differences between the most recent cycles, 2013–2014 and 2015–2016. We adjusted all analyses for the complex survey design of the NHANES, including strata, primary sampling units, and probability weights, by using the survey estimation commands in Stata version 15.0 (StataCorp, College Station, TX).

Readers should use the following information as guidance when interpreting our findings. We present results from multiple significance tests but do not make any adjustments for multiple testing, which reduces the chance of a type II error but increases the chance of a type I error. Readers should consider the chance for both type I and type II errors. To reduce the chance of a type II error (indicating as significant a relationship that does not exist), we present all ad hoc data without choosing only those that are significant. We include P values for reference but encourage readers not to focus on P < .05 and instead to consider the body of data. In the supplemental appendices, we present confidence intervals (CIs) to allow readers to draw their own conclusions about more nuanced comparisons. The chance of a type I error (not identifying a relationship that does exist) should be considered, particularly in our comparisons between the 2013–2014 and 2015–2016 cycles. Although the sample size allows for the identification of relatively small differences in the prevalence of the full sample, subgroup analyses should be considered more carefully. We have provided sample sizes throughout the tables to assist readers in their assessments.

Table 1 presents the prevalence of overweight and all classes of obesity by demographic characteristics in the most recent NHANES cycle, 2015–2016. Non-Hispanic African American and Hispanic children had higher prevalence rates of overweight and all classes of obesity compared with other races. Asian American children had markedly lower rates of overweight and all classes of obesity. The prevalence of overweight and obesity increased with age, with 41.5% of 16- to 19-year-old adolescents having obesity and 4.5% meeting criteria for class III obesity.

Table 2 shows the prevalence of overweight and all classes of obesity by ordinal 2-year cycles (1999–2016) for females, males, and both sexes. A positive linear trend is significant for overweight (P = .003), class I obesity (P = .008), class II obesity (P = .019), and class III obesity (P < .001) for both sexes, with all ages combined. The increasing linear trend from 1999 to 2016 is most apparent among Hispanic females (Table 3). Similar to those of females, there are large increases in overweight and class II obesity among Hispanic males (Table 4). All 95% CIs are included in Supplemental Tables 5–9.

There are few differences in the prevalence of overweight and all classes of obesity since the last NHANES cycle, 2013–2014 and 2016–2016. One exception is a sharp increase in the prevalence of class I obesity among 2- to 5-year-olds, particularly in young males. Another notable increase is for overweight, from 36% to 48%, in among older adolescent females. There were no other significant changes from the 2013–2014 and 2015–2016 cycles for any of the race and/or sex subgroups in any of the obesity categories.

Despite reports that obesity in children and adolescents in the United States has stabilized in recent years,¹ our more nuanced view highlights the continued upward trend for this nationally representative sample (Fig 1). Significant increases in obesity and severe obesity in children aged 2 to 5 years and adolescent females aged 16 to 19 years from 2015 to 2016, compared with previous years, show that obesity is increasing in these subgroups. Whether this year-over-year change represents a trend remains to be seen because shifts per cycle can be large. We recommend that readers consider both the long-term trends as well as changes over 2-year cycles when considering the effects in specific populations.

The prevalence of childhood obesity in the United States remains high, with ∼1 in 5 children having obesity. By applying updated obesity classifications¹⁶ to data starting in the 1999–2000 cycle of the NHANES, there continue to be increases in most categories of obesity across all age groups. By age, adolescents have had a significantly increased prevalence across all obesity categories since the 1999–2000 cycle. Substantial racial-ethnic differences remain, with African Americans and Hispanics having a higher prevalence across nearly all classes of obesity and all years between 1999 and 2016. Notably, Asian Americans have a much lower prevalence of obesity in all age and sex categories. There were few differences in obesity prevalence from the previous cycle (2013–2014), with the exception of Hispanic males, who saw significant increases, and boys ages 2 to 5 years, who have had a 40% increase in prevalence since 2011.

Despite intense clinical and public health focus on obesity and weight-related behaviors during the past decade, obesity prevalence remains high, with scant evidence that these efforts are counteracting the personal and environmental forces that contribute to excess weight gain in children, at least on a national scope. These findings are disappointing in light of reported decreases in obesity prevalence in younger children,^2,4,17,–21 which was the only age group as a whole to see a significant increase in prevalence since the 2013–2014 NHANES cycle. Most disconcerting are the substantial disparities in obesity by race and ethnicity; statistical and clinical differences in prevalence between Hispanics and all other races are astounding, with nearly half of all Hispanic youth having overweight or obesity. Building on our previous work,^5,6,12 we have been able to document the steadily rising levels of severe obesity, modeled on adult criteria of class I, II, and III obesity, with the rise of children with severe obesity having been the most significant.

Public health efforts to address obesity in children have been extensive, from Michelle Obama’s Let's Move campaign to the American Academy of Pediatrics establishing a Section on Obesity in 2013 that is distinct from other groups in the academy as well as countless efforts led by states, hospitals, and communities. Despite these efforts, which may have had greater impact in defined populations, more resources are clearly necessary. The obesity epidemic is becoming endemic, and this decline in Americans’ health is occurring without impactful policy at the national level. Evidence-based efforts focused on policy, family-based change, and health improvement (versus weight loss alone) may take another decade to see positive results; effective prevention and treatment interventions remain undeveloped or have not been effectively disseminated, and more insight is needed into the moderators and mediators of excessive weight gain. Additionally, evidence that behaviors in high-risk groups start at a young age suggests that efforts need to focus early in children’s lives.²² 

There are few long-term studies of obesity development or treatment outcomes because this work is occurring in a Biggest Loser environment, with the focus being on short-term changes in weight that we are only beginning to see as an erroneous pursuit in adult populations.²³ These efforts are hampered by declining research dollars, limited or nonexistent reimbursement for prevention and treatment, and difficulties in changing local and national policies that impact environmental health. Finally, there is some evidence of an association between poverty and obesity,²⁴ undoubtedly influencing the health of children nationwide. Activities with the aim of decreasing the prevalence of childhood obesity should not cease but redouble as an effort to improve the health of children and families and stem the rising costs of health care in the United States.

There are several important limitations to note. First, the NHANES data are repeated cross-sections and do not allow for the examination of within-child changes over time. However, this approach allows for a richer picture of obesity prevalence across the United States. A second limitation is that the sample sizes prevent detailed subgroup analyses. We present prevalence rates by age, sex, and race, but caution should be used when interpreting these results. Readers should consider the body of evidence rather than focusing on individual tests of significance. Finally, the inclusion of Asian Americans in this report highlights questions about the reference ranges that define obesity. The current reference charts were developed by using data from a more homogenous group than what is seen in the United States today. It is not clear if the definitions of obesity represent similar levels of adiposity across racial and ethnic groups or if they confer similar levels of health risk.

Nationally representative data provided by the NHANES demonstrates clearly that childhood obesity continues to be a significant concern for the United States. The past 18 years have seen increases in the levels of severe obesity in all ages and populations despite increased attention and efforts across numerous domains of public health and individual care. Groups that are historically disenfranchised are affected the most by this epidemic, predicting increased morbidity across a lifetime. Previously reported improvements seen in younger children were either an anomaly or transient because national data presented here demonstrate a sharp increase from the last cycle. Present efforts must continue, as must innovation, research, and most importantly at this juncture, collaboration among clinicians, public health leaders, hospitals, and all levels of government.

CDC

Centers for Disease Control and Prevention

CI

confidence interval