Professional Documents
Culture Documents
Obesity
Solveig A. Cunningham, PhD,a,* Shakia T. Hardy, PhD,b,* Rebecca Jones, PhD,a Carmen Ng, PhD,a
Michael R. Kramer, PhD,a,** K.M. Venkat Narayan, MDa,**
a
Hubert Department of Global Health, Emory University, Atlanta, Georgia; andbDepartment of Epidemiology,
University of Alabama at Birmingham, Birmingham, Alabama
2 CUNNINGHAM et al
Downloaded from http://publications.aap.org/pediatrics/article-pdf/150/2/e2021053708/1344063/peds_2021053708.pdf
by guest
incidence of obesity across data socioeconomic patterning of obesity, RESULTS
waves. We calculated obesity we stratified cumulative incidence Table 1 shows children by BMI
prevalence as the percentage of by household socioeconomic status categories in each of the 2 cohorts.
children with obesity at a given (SES) quartiles at baseline (1 being When the birth cohorts of 1992/93
point in time. We then calculated the most disadvantaged households entered Kindergarten in 1998,
annual obesity incidence rates as the and 4 the most socioeconomically 72.9% (95% confidence interval
annualized risk of developing advantaged) and by school location 71.2%–74.6%) had normal BMI,
obesity between data waves, at baseline (city, urban fringe and whereas 15.1% (13.6%–16.5%) had
calculated by dividing the number of large town, small town and rural overweight and 12% already had
new obesity cases in the wave by area). To consider importance of obesity: 9.2% (8.1%–10.2% had
the number of person–years of prenatal growth, we stratified by moderate obesity and 2.9% had
follow-up contributed by children birth weight (low birth weight: severe obesity (2.2%–3.5%). Twelve
when they were nonobese.10 For <2500 g; normal birth weight: years later, when the birth cohorts
each individual, the follow-up period 2500–3999 g; high birth weight:
was defined as half of the time of 2004 to 2005 entered
$4000 g). kindergarten in 2010, the
between each survey wave and the
percentage of children starting
following wave for an incident case, BMI was missing for <5% of
or the entire time between these school with a normal BMI had
observations in every full sample
2 waves if the child remained decreased to 69% (67.6%–70.5%).
wave. However, some survey waves
without obesity. This approach The percentage entering
were designed as representative
makes the implicit assumption that kindergarten with overweight in
subsamples of the full cohorts to
the obesity threshold was crossed at 2010 had not changed substantially,
reduce survey costs, so some
the midpoint between 2 data waves. children were not measured at all at 15.7% (14.6%–16.7%), but the
data waves by design. When an percentage that already had obesity
It could be that having more increased to 15.3%. Thus, unhealthy
observation was missing a BMI
frequent data waves in the more BMI levels shifted to younger ages
value at the first or last observation,
recent cohort could allow us to between children growing up in the
we used the value from the closest
observe more events of incident 1990s and those growing up in the
available wave to avoid making
obesity that turn out to be fleeting, 2000s as evidenced by increase in
estimates outside the available
which would affect estimates of prevalence at kindergarten entry; in
range. For observations with
incident cases. Therefore, we the more-recent cohort, obesity
missing BMI values at other waves,
estimated all incidence rates for the already affected 1 in 6 children
the missing value was linearly
2010–2011 cohort using the same before they entered school.
waves of data as the 1998–1999, interpolated. This is a reasonable
thus, using a balanced panel. The assumption because the time
At kindergarten entry, at average
estimates, including all data waves between measurements was fairly
age 6 years, not only had a higher
available, are shown for reference in short; the alternative method,
proportion of children in the more
Supplemental Fig 4. multiple imputation, has been found
to yield little difference in recent cohort already entered
conclusions for within-person overweight and obesity, but a higher
We calculated cumulative obesity
trajectories.17 Missing data on proportion had already reached
incidence as the proportion of
demographic and socioeconomic severe obesity: 3.9% (3.3%–4.5%)
children who developed obesity by
variables was generally <1%, with in 2010, compared with 2.9%
the end of fifth grade, among
children who did not have obesity at the exception of SES in 1998 (5.9%) (2.2%–3.5%) in 1998. In both
kindergarten entry. To examine the and urbanicity in 2010 (8.3%). We cohorts, the prevalence of severe
risk of obesity for demographic used listwise deletion for each obesity increased with increasing
groups, we stratified cumulative subgroup analysis. Information on age. Additionally, the age-specific
incidence estimates by gender (boys birth weight was missing for 17.2% prevalence of severe obesity was
or girls) and parent-reported race in 1998 and 27.1% in 2010, so we greater at every age point in the
and ethnicity (non-Hispanic White, show associations with weight 2010 as compared with 1998–1999
non-Hispanic Black, Hispanic, and status for those with missing birth cohort. For example, in fifth grade
other race, which includes Asian weight. ( age 11), the prevalence of severe
American, Pacific Islander, American obesity was 5.6% (4.8%–6.4%) in
Indian, and multiracial). To measure Analyses were performed with R the 1998–1999 cohort versus 7.2%
whether there were changes in (version 3.5.1).18 (6.4%–8.0%) in the 2010–2011
cohort. Thus, obesity was more obesity by age 11 among children higher for the 2010–2011
severe in the recent cohort. who entered kindergarten nonobese kindergarten entrants than for new
for the 1998 and 2010 kindergarten kindergartners 12 years earlier for
Fig 1 shows that the prevalence of cohorts. Among all children who did children who already had
obesity during primary school was not already have obesity when they overweight at average age 6 years.
higher in the 2010 cohort at all entered kindergarten, cumulative
ages, with the largest differences incidence of obesity between Table 2 shows 5.5-year cumulative
being at kindergarten entry. kindergarten entry and the end of incidence proportion of obesity
fifth grade (ages 6 and 11 years) stratified by sociodemographic
Fig 2 shows the annualized was 16.2% (15.0–17.3) in 2010; this groups. Boys had higher risk of
incidence rate of obesity during is a 4.5% relative increase over the developing obesity than girls in each
primary school. The annual cumulative incidence of 15.5% for cohort, but only boys experienced
incidence of obesity was highest children over the same age span an increase in risk in 2010 to 2011
during kindergarten for the 1998 (17.7% [16.1%–19.4%]) as
12 years earlier.
cohort, at 5.12 (4.01–6.22) per 100 compared with 1998 to 1999
person–years, and lower thereafter, When further stratified by BMI (16.2% [14.2%–18.3%]), with a
oscillating around 3 per 100 category at kindergarten entry, relative risk of 1.09 (1.08–1.11).
person–years. In the 2010 cohort,
differences emerge. There was no
annual incidence was somewhat
change between the 2 cohorts in the Non-Hispanic Black kindergartners
lower during kindergarten but
risk of developing obesity for were 29% (risk ratio [RR] 1.29, 95%
peaked in first grade at 4.12
children who entered with normal confidence interval 1.25–1.34) more
(3.15–5.09) per 100 person–years
BMI (9.8% cumulative incidence for likely to develop new onset obesity by
and remained higher than for the
1998 cohort thereafter, reaching the both cohorts). However, the risk of fifth grade in the 2010–2011 cohort
largest difference in fifth grade at developing obesity for children who compared with non-Hispanic Black
3.86 (3.44–4.28) per 100 entered kindergarten overweight children in the 1998–1999 cohort.
person–years, compared with 2.83 increased slightly from 42.9% Although in the late 1990s, Hispanic
(2.35–3.31) per 100 person–years in (38.0–47.9) to 44.3% children were experiencing the highest
the 1998 cohort. (41.3%–47.3%) in the later cohort. incidence of obesity during primary
These patterns suggest that the school (19.9% [16.1–23.8]), they were
Table 2 shows the cumulative cumulative risk of developing surpassed by non-Hispanic Black
incidence proportion (eg, “risk”) of obesity during primary school was children in the 2000s. Thus, there
4 CUNNINGHAM et al
Downloaded from http://publications.aap.org/pediatrics/article-pdf/150/2/e2021053708/1344063/peds_2021053708.pdf
by guest
FIGURE 1
Prevalence of obesity between kindergarten and fifth grade for the 1998 and 2010 kindergarten cohorts. Notes: Prevalence of obesity is shown for the co-
horts of US children who entered kindergarten in 1998 and in 2010, followed yearly until the end of fifth grade. Whiskers indicate 95% confidence intervals.
Data: Early Childhood Longitudinal Study–Kindergarten cohorts of 1998 and 2010.
were widening disadvantages in middle of the socioeconomic were in suburban schools. Children
obesity for non-Hispanic Black spectrum across the 2 cohorts, but in suburban schools and large
children compared with all other increased by 15% for children from towns had the largest increases in
groups. The risk of incident obesity the least advantaged households (RR incidence, leading to even more
during primary school plateaued for 1.15 [1.12–1.17]), from 17.7% similarity across environments for
Hispanic and Non-Hispanic White
(14.6%–20.8%) to 20.3% the 2010 cohort (RR 1.13
children, though at a higher level for
Hispanic (19.9%) than for White (18.0–22.6)) and from the most [1.11–1.15]).
children (13.6%). Children in other advantaged households (RR 1.15
race groups, which includes Asian [1.14–1.17]). The socioeconomic To consider intergenerational
American, Pacific Islander, and gradient of obesity incidence was transmission of risk, Table 2
American Indian, had a 16% reduction maintained and the disadvantages in presents obesity incidence in
in risk of developing obesity during terms of unhealthy weight increased relation to birth weight. These
primary school between the 2 cohorts for the most socioeconomically preliminary data indicate that
(RR 0.84 [0.80–0.87]); thus, they
disadvantaged children relative to incidence of obesity in primary
surpassed non-Hispanic White
children as the group with lowest all other SES groups. school increased linearly with birth
incidence in the 2010–2011 cohort weight in the early 2000s but
(13.4% [10.1%–16.7%]). The risks of developing obesity followed a J-shape a decade later
were similarly distributed across because of large relative increases in
The risks of developing obesity urban, suburban, and rural schools risk for children born with low birth
stayed similar for children in the in 1998, though the lowest risks weight and high birth weight.
DISCUSSION children passing through this age race and ethnic groups experienced
Previous studies reporting on the group 12 years earlier. This pattern plateauing or even decreasing
prevalence of childhood obesity suggests earlier onset of elevated incidence of obesity. Recognizing
have shown continued increases in BMI in the more-recent cohort that the obesity risks faced by Black
obesity during the past decade, with occurring during the preschool children are continuing to increase
plateauing of obesity prevalence years. Increases in BMI across the 2 highlights the need to identify
among some age groups.9 This study cohorts and higher proportions factors that may underlie their
contributes by identifying how the reaching severe obesity indicate vulnerability to develop more
incidence of new cases has changed temporal trends to more children successful prevention efforts.19
between recent cohorts. having elevated BMI and severely Although extensive public health
high levels of BMI. efforts have been directed toward
Compared with children born in the childhood obesity since 2010, such
early 1990s, those born in the mid- Across the 2 cohorts, social as the Let’s Move! campaign and the
2000s experienced obesity with disparities in unhealthy weight Healthy, Hunger-Free Kids Act, these
higher incidence, at younger ages, increased. Most markedly, the risk policies have had no impact on
and at more severe levels. In 2010, a of developing obesity during reducing population-level obesity.20
higher proportion of children primary school increased The heterogenous risk of obesity by
arrived at kindergarten with significantly for non-Hispanic Black race–ethnicity, coupled with the lack
moderate or severe obesity, children, surpassing that of Hispanic of impact of interventions, highlights
compared with the cohort of children. During this period, other the need for public health policies to
6 CUNNINGHAM et al
Downloaded from http://publications.aap.org/pediatrics/article-pdf/150/2/e2021053708/1344063/peds_2021053708.pdf
by guest
TABLE 2 Cumulative Incidence Proportion of Obesity From Kindergarten Through Fifth Grade Across Sociodemographic Groups and Risk Ratios
Comparing the US 1998 and 2010 Kindergarten Cohorts
1998 Kindergarten 2010 Kindergarten RR for Obesity,
Cohort Cohort 2010 vs 1998
be tailored to counterbalance of obesity during primary school but proposition directly. To attempt to
obesogenic factors. Notably, we girls did not, suggesting the capture some prenatal factors, we
identified increasing heterogeneity importance of psychosocial, examined associations between
in incidence of obesity across behavioral, and epigenetic factors birth weight and obesity incidence.
schools in urban, suburban, and beyond those originating from race, The proportion of children born
ethnic, and socioeconomic with low and high birth weight was
rural areas because all locations
circumstances. similar across the 2 cohorts (6.3%
converged toward higher obesity
and 6.7% low birth weight and 8.9%
incidence. Among these factors, the and 6.6% high birth weight), but the
intergenerational transmission of incidence of obesity for children
That children from both the highest
obesity may offer insights into who were born small or large was
and lowest socioeconomic
households were more likely to changes in obesity across cohorts. substantially higher in the more-
develop obesity in the 2010s is a Specifically, obesity has been recent cohort. These patterns show
reminder that children of all walks increasing among women of mixed support for the possibility
of life are at risk for obesity. At the reproductive ages,9,21,22 and that childhood obesity incidence is
same time, the socioeconomic increasing because of changing
maternal obesity predisposes
gradient of obesity has persisted maternal and prenatal exposures;
children to develop obesity23–25;
because children from the most because of the high proportion of
consequently, more children in children missing data on birth
disadvantaged households
recent cohorts may be predisposed weight, this hypothesis should be
experienced increases in incidence
higher than other children; thus, to developing obesity. Information examined with other data sets.
their disadvantages in terms of long- on mothers’ health was not collected
term health continue to grow. Boys in the Early Childhood Longitudinal In light of the observed higher
experienced increases in incidence Study, so we cannot assess this incidence of obesity across primary
8 CUNNINGHAM et al
Downloaded from http://publications.aap.org/pediatrics/article-pdf/150/2/e2021053708/1344063/peds_2021053708.pdf
by guest
DOI: https://doi.org/10.1542/peds.2021-053708
Accepted for publication May 3, 2022
Address correspondence to Solveig A. Cunningham, PhD, Hubert Department of Global Health, Emory University, 1518 Clifton Rd, Atlanta, GA 30322. E-mail: sargese@
emory.edu
PEDIATRICS (ISSN Numbers: Print, 0031-4005; Online, 1098-4275).
Copyright © 2022 by the American Academy of Pediatrics
CONFLICT OF INTEREST DISCLAIMER: The authors have indicated they have no conflicts of interest relevant to this article to disclose.
REFERENCES 10. Cunningham SA, Kramer MR, Narayan K. 16. Flegal KM, Wei R, Ogden CL, Freedman
1. Must A, Anderson SE. Effects of obesity Incidence of childhood obesity in the DS, Johnson CL, Curtin LR. Characteriz-
on morbidity in children and adoles- United States. N Engl J Med. ing extreme values of body mass index-
cents. Nutr Clin Care. 2003;6(1):4–12 2014;370:401–409 for-age by using the 2000 Centers for
2. Strauss RS. Childhood obesity and self- 11. Tourangeau K, L^e T, Nord C, Sorongon Disease Control and Prevention growth
esteem. Pediatrics. 2000;105(1):e15 AG. Early Childhood Longitudinal Study, charts. Am J Clin Nutr. 2009;90(5):
Kindergarten Class of 1998–99 (ECLS-K), 1314–1320
3. Friedlander SL, Larkin EK, Rosen CL, Pa-
Eighth-Grade Methodology Report (NCES
lermo TM, Redline S. Decreased quality 17. Ng CD, Elliott MR, Riosmena F, Cunning-
2009–003). Washington, DC: National
of life associated with obesity in school- ham SA. Beyond recent BMI: BMI expo-
Center for Education Statistics, Institute
aged children. Arch Pediatr Adolesc sure metrics and their relationship to
of Education Sciences, U.S. Department
Med. 2003;157(12):1206–1211 health. SSM Popul Health. 2020;11:
of Education; 2009
4. Dietz WH. Health consequences of obesity 100547
12. Tourangeau K, Nord C, L^e T, et al. Early
in youth: childhood predictors of adult dis- 18. R: A Language and Environment for Sta-
Childhood Longitudinal Study, Kinder-
ease. Pediatrics. 1998;101(3 Pt 2):518–525 tistical Computing [computer program].
garten Class of 2010–11 (ECLS-K:2011),
5. Thompson DR, Obarzanek E, Franko DL, User’s Manual for the ECLS-K:2011 Kin- Version 3.5.1. Vienna, Austria; 2018
et al. Childhood overweight and cardio- dergarten Data File and Electronic 19. Singh GK, Kogan MD, Van Dyck PC, Siah-
vascular disease risk factors: the Na- Codebook, Public Version (NCES 2015- push M. Racial/ethnic, socioeconomic,
tional Heart, Lung, and Blood Institute 074). Washington, DC: U.S. Department and behavioral determinants of child-
Growth and Health Study. J Pediatr. of Education; 2015 hood and adolescent obesity in the
2007;150(1):18–25 United States: analyzing independent
13. Vogel M. childsds: data and methods
6. Hales CM, Carroll MD, Fryar CD, Ogden and joint associations. Ann Epidemiol.
around reference values in pediat-
CL. Prevalence of obesity among adults 2008;18(9):682–695
rics. Available at: https://CRAN.
and youth: United States, 2015–2016. 20. Kenney EL, Barrett JL, Bleich SN, Ward
R-project.org/package=childsds. Ac-
NCHS Data Brief. 2017; (288):1–8
cessed November 1, 2019 ZJ, Cradock AL, Gortmaker SL. Impact of
7. Fryar C, Carroll, MD, Afful, J. Prevalence the Healthy, Hunger-Free Kids Act on
14. Kelly AS, Barlow SE, Rao G, et al. Ameri-
of overweight, obesity, and severe obe- obesity trends. Health Aff (Millwood).
can Heart Association Atherosclerosis,
sity among children and adolescents 2020;39(7):1122–1129
aged 2–19 years: United States, Hypertension, and Obesity in the Young
Committee of the Council on Cardiovas- 21. Deputy NP, Dub B, Sharma AJ. Preva-
1963–1965 through 2017–2018. NCHS
cular Disease in the Young, Council on lence and trends in prepregnancy nor-
Health E-Stats. 2020
Nutrition, Physical Activity and Metabo- mal weight – 48 states, New York City,
8. Ogden CL, Carroll MD, Lawman HG, lism, and Council on Clinical Cardiology. and District of Columbia, 2011–2015.
et al. Trends in obesity prevalence MMWR Morb Mortal Wkly Rep. 2018;
Severe obesity in children and adoles-
among children and adolescents in the 66(51-52):1402–1407
cents: identification, associated health
United States, 1988–1994 through
risks, and treatment approaches: a sci- 22. Shah NS, Wang MC, Freaney PM, et al.
2013–2014. JAMA. 2016;315(21):
entific statement from the American Trends in gestational diabetes at first
2292–2299
Heart Association. Circulation. live birth by race and ethnicity in the
9. Ogden CL, Fryar CD, Martin CB, et al. 2013;128(15):1689–1712 US, 2011–2019. JAMA. 2021;326(7):
Trends in obesity prevalence by race and 660–669
Hispanic origin: 1999–2000 to 2017–2018. 15. Gulati AK, Kaplan DW, Daniels SR. Clini-
JAMA. 2020;324(12):1208–1210 cal tracking of severely obese children: 23. Williams S. Overweight at age 21: the
a new growth chart. Pediatrics. association with body mass index in
2012;130(6):1136–1140 childhood and adolescence and
10 CUNNINGHAM et al
Downloaded from http://publications.aap.org/pediatrics/article-pdf/150/2/e2021053708/1344063/peds_2021053708.pdf
by guest