Tracking Pediatric Obesity: Title and subTitle BreakAn Index of Uncertainty?

Body mass index (BMI), calculated as weight in kilograms divided by height in meters squared, is used extensively to characterize excessive body weight in adults and children. The Centers for Disease Control and Prevention (CDC) tracks changes in the national prevalence of obesity with BMI, and a report by Ogden and colleagues¹ in this issue of JAMA presents the latest data involving children. Recently, a consensus statement prepared by an expert committee, comprising professionals from 15 health care organizations, described BMI as the best available clinical tool to screen for childhood obesity and monitor progress with treatment.² The statement encourages primary care clinicians to assess obesity risk at all well-child visits using BMI-for-age percentiles. This assessment paradigm has extended into schools, and several states now mandate use of BMI to identify overweight or obese children and evaluate the effectiveness of healthful lifestyle initiatives. However, despite its widespread use, BMI continues to cause confusion among patients and practitioners, raising several fundamental questions.

First, if pediatric obesity is defined as a BMI for age exceeding the 95th percentile, why do prevalence estimates differ from 5%? In 2000, the CDC released sex-specific BMI-for-age growth charts that established the basis for defining pediatric obesity.³ Aiming to develop nationally representative references, demographers calculated percentiles for boys and girls aged 2 to 20 years using data from several national surveys, including 2 cycles of the National Health Examination Survey (NHES II, 1963-1965; NHES III, 1966-1970) and 3 National Health and Nutrition Examination Surveys (NHANES I, 1971-1974; NHANES II, 1976-1980; NHANES III, 1988-1994). Because of the rapid increase in body weight beginning in the 1980s, data from NHANES III for children aged 6 years and older, and all data from later surveys, were excluded. Thus, percentiles were based on data collected at a time when BMI for age remained relatively stable and excess adiposity in children and adolescents was not considered a serious threat to public health. As such, the growth charts do not describe current distributions of BMI for age. During the last 3 decades, mean BMI in the pediatric age range increased markedly, and this increase was greatest among the heaviest individuals.⁴

Second, how were BMI-for-age cut points established to define overweight and obesity? Ideally, cut points describing relative degrees of excess weight would be determined from the relationship of BMI to morbidity and mortality, as demonstrated in prospective, nationally representative cohort studies beginning in childhood and continuing into old age. In the absence of such data, an expert committee in the early 1990s assigned sex-specific cut points for the categories at risk for overweight (BMI ≥ 85th percentile but < 95th percentile) and overweight (BMI ≥ 95th percentile) according to population-based estimates of adiposity and concurrent comorbidities.⁵ The committee recognized that these categories would underestimate true prevalence but aimed to maximize specificity at the expense of sensitivity and thereby avoid potential harm from misclassifying children who do not have excessive adiposity. In 2007, a new expert committee recommended maintaining the same cut points but changing terminology by replacing at risk for overweight with overweight and overweight with obese.² An additional cut point was proposed to define severe obesity as BMI at or above the 99th percentile.

Third, how well do BMI-for-age percentiles predict risk for obesity-related disease? Childhood BMI is correlated with body fat,⁶ concurrent risk factors for chronic diseases,⁷ and future morbidity and mortality.^{8 - 9} However, BMI provides no direct information about body composition (fat vs fat-free mass) or fat distribution (eg, central vs peripheral), which both comprise major predictors of disease. Body composition and central adiposity in children vary greatly at any given BMI.^{6 ,10} Moreover, waist-to-hip ratio, a measure of body fat distribution, may better identify cardiovascular disease risk factors than BMI in children.¹¹

Numerous variables confound the associations of BMI with adiposity and adiposity with risk. Clearly, physical activity can increase the ratio of lean to fat mass and decrease risk for cardiovascular disease without a change in body weight. Diet quality profoundly affects chronic disease risk at any BMI, and recent research suggests that nutritional factors also affect body composition when controlling for body weight.¹² Remarkably, psychological stress may promote central fat deposition in adolescents independent of BMI,¹³ a possibility with sobering implications in light of the increasingly hectic lifestyles of youth today. In addition, genetic and perinatal¹⁴ factors and pubertal maturation influence body habitus and risk for obesity-related disease.

Racial/ethnic differences in health outcomes may arise, in part, from the complex interplay among these variables. Freedman et al¹⁵ found that the sensitivity of the BMI-for-age 95th percentile in identifying excess adiposity ranged from only 50% for Asian girls to more than 80% for Hispanic and black children. Consistent with these findings, risk for obesity-related disease increases at a lower BMI in Asian populations compared with other racial/ethnic groups.⁴ In a comparison of black and white adolescents, Bacha et al¹⁶ observed that white adolescents have greater visceral adiposity and a more atherogenic risk profile at any given BMI, whereas black adolescents have more insulin resistance and less insulin secretion. Thus, although BMI in childhood correlates well with body fatness and health outcomes on a population level, racial/ethnic differences and other sociodemographic factors importantly influence these relationships.

Historical cohort studies document an association between childhood BMI and chronic disease in adulthood,^{8 - 9} but optimal levels of BMI for long-term health are not known. Among Danish children born between 1930 and 1976, Baker et al⁹ observed a linear relationship between BMI and incidence of coronary heart disease in adulthood across the entire BMI distribution, questioning the significance of norm-referenced BMI-for-age cut points. Additional uncertainty surrounding established cut points arises because the curve relating BMI to disease risk seen in historical cohorts has likely shifted to the left over time with lifestyle changes. Physical activity level and diet quality have decreased, and psychological stress has arguably increased among children in the last half century. These secular trends suggest that adiposity at any given BMI has increased, as has disease risk at any level of adiposity, and conventional cut points may seriously underestimate the prevalence of risk among US children today.

Fourth, what is the role of BMI-for-age percentiles in pediatric clinical practice? Expert consensus committees have concluded, and we agree, that BMI-for-age cut points can be useful to identify children at increasing categories of risk for the acute and long-term complications of excessive body weight. In addition, a formal classification system, if discussed in a sensitive fashion, may serve as an educational tool to raise awareness among patients and their parents. Furthermore, anthropometric definitions may help the clinician obtain reimbursement for office-based prevention and treatment services. In essence, however, BMI is no more than a screening tool, the relevance of which to any patient must be considered in light of the medical history, physical examination, and presence of comorbidities. Depending on racial/ethnic group, family history, perinatal factors, diet quality, and physical fitness, the long-term risk for cardiovascular disease and other life-threatening complications may begin to rise well below the 85th percentile. A BMI above the 95th percentile describes children and adolescents at substantially increased risk for immediate complications, such as type 2 diabetes. Within the normal range, identifying an upward trend over time in BMI percentile provides an opportunity for primary prevention.

Finally, do current CDC data suggest that the end of the pediatric obesity epidemic is in sight? After years of unremittingly bad news about increasing rates of pediatric obesity, Ogden et al¹ report no increase in prevalence between 1999-2000 and 2005-2006. Perhaps recent public health campaigns aimed at raising awareness of childhood obesity and improving the quality of school food have begun to pay off. However, it is too early to know whether these data reflect a true plateau or a statistical aberration in an inexorable epidemic, and pre-existing racial/ethnic disparities show no sign of abating. On one point there is no uncertainty: without substantial declines in prevalence, the public health toll of childhood obesity will continue to mount, because it can take many years for an obese child to develop life-threatening complications.¹⁷