New England Journal Medicine: The of
New England Journal Medicine: The of
New England Journal Medicine: The of
The
journal of medicine
established in 1812 january 30, 2014 vol. 370 no. 5
A BS T R AC T
Background
Although the increased prevalence of childhood obesity in the United States has From the Hubert Department of Global
been documented, little is known about its incidence. We report here on the na- Health (S.A.C., K.M.V.N.) and the Depart-
ment of Epidemiology (M.R.K.), Emory
tional incidence of obesity among elementary-school children. University, Atlanta. Address reprint re-
quests to Dr. Cunningham at the Hubert
Methods Department of Global Health, Emory
University, 1518 Clifton Rd., Atlanta, GA
We evaluated data from the Early Childhood Longitudinal Study, Kindergarten 30322, or at [email protected].
Class of 19981999, a representative prospective cohort of 7738 participants who
N Engl J Med 2014;370:403-11.
were in kindergarten in 1998 in the United States. Weight and height were measured DOI: 10.1056/NEJMoa1309753
seven times between 1998 and 2007. Of the 7738 participants, 6807 were not obese Copyright 2014 Massachusetts Medical Society.
at baseline; these participants were followed for 50,396 person-years. We used stan-
dard thresholds from the Centers for Disease Control and Prevention to define
overweight and obese categories. We estimated the annual incidence of obesity,
the cumulative incidence over 9 years, and the incidence density (cases per person-
years) overall and according to sex, socioeconomic status, race or ethnic group, birth
weight, and kindergarten weight.
Results
When the children entered kindergarten (mean age, 5.6 years), 12.4% were obese
and another 14.9% were overweight; in eighth grade (mean age, 14.1 years), 20.8%
were obese and 17.0% were overweight. The annual incidence of obesity decreased
from 5.4% during kindergarten to 1.7% between fifth and eighth grade. Overweight
5-year-olds were four times as likely as normal-weight children to become obese
(9-year cumulative incidence, 31.8% vs. 7.9%), with rates of 91.5 versus 17.2 per
1000 person-years. Among children who became obese between the ages of 5 and
14 years, nearly half had been overweight and 75% had been above the 70th percen-
tile for body-mass index at baseline.
Conclusions
Incident obesity between the ages of 5 and 14 years was more likely to have oc-
curred at younger ages, primarily among children who had entered kindergarten
overweight. (Funded by the Eunice Kennedy Shriver National Institute of Child Health
and Human Development.)
C
hildhood obesity is a major health in the fall semester of 1998 (mean age, 5.6 years)
problem in the United States.1 The preva- and followed 9358 children through sequential
lence of a body-mass index (BMI; the phases of data collection, in 1999 (spring semester
weight in kilograms divided by the square of the of kindergarten; mean age, 6.1 years), 2000 (first
height in meters) at the 95th percentile or higher grade; mean age, 7.1 years), 2002 (third grade;
among children between the ages of 6 and 11 mean age, 9.1 years), 2004 (fifth grade; mean age,
years increased from 4.2% in 19631965 to 11.1 years), and 2007 (eighth grade; mean age,
15.3% in 199920002,3 and may have plateaued 14.1 years). The NCES also collected data from a
during the first decade of the 21st century.4,5 Al- representative subsample of one third of the chil-
though trends in the prevalence of obesity are dren in 1999 (mean age, 6.6 years; fall semester
documented, surprisingly little is known about of first grade). With appropriate survey adjust-
the incidence of childhood obesity. Examining ments, this longitudinal sample is representative
incidence may provide insights into the nature of of all children enrolled in kindergarten in 1998
the epidemic, the critically vulnerable ages, and and 1999 in the United States (approximately
the groups at greatest risk for obesity. 3.8million children).
National data on the incidence of pediatric The survey included extensive data collection
obesity to date have pertained only to adoles- from caregivers, school staff, teachers, and chil-
cents transitioning to adulthood. A study that dren, along with direct measurements, as described
was based on data from the National Longitudi- previously.8 Trained assessors measured childrens
nal Study of Adolescent Health showed that the height in inches (to the nearest 0.25in.) with the
5-year cumulative incidence of obesity among use of a Shorr board and recorded weight in
persons who were 13 to 20 years of age in 1996 pounds with the use of a digital scale. For this
and 19 to 26 years of age in 2001 was 12.7%, analysis, in order to track the incidence of obe-
ranging from 6.5% among Asian girls to 18.4% sity, we selected variables height, weight, and
among non-Hispanic black girls.6 However, since parent-reported age, sex, race or ethnic group,
many of the processes leading to obesity start socioeconomic status, and birth weight from
early in life,7 data with respect to incidence be- among the variables in the restricted-use data
fore adolescence are needed. set for children from kindergarten through
We report here the incidence of obesity ac- eighth grade. The analytic sample consists of the
cording to data from a large, nationally represen- 7738 children with data on these variables across
tative longitudinal study of children who were phases of data collection.
followed from entry into kindergarten to the end
of eighth grade (ages 5 to 14 years); the study Evaluation of Data
included direct anthropometric measurements at We used the 2000 Centers for Disease Control
seven points between 1998 and 2007. and Prevention (CDC) Growth Charts to calcu-
late each childs BMI, standardized to the refer-
Me thods ence population for the childs age and sex.9 We
determined cutoffs for normal weight, over-
Study Population weight, and obesity using the CDCs standard
We analyzed data from the Early Childhood Longi- thresholds of the 85th percentile for overweight
tudinal Study, Kindergarten Class of 19981999, and 95th percentile for obesity. The use of alter-
which was designed and conducted by the Na- native specifications with cutoffs set by the Child
tional Center for Education Statistics (NCES) of Obesity Working Group of the International Obe-
the Department of Education. The NCES selected sity Task Force10 showed consistent results.
a nationally representative cohort using multi- We calculated the prevalence of obesity as the
stage probability sampling, in which the primary proportion of all children in each age group who
sampling units were counties or groups of coun- were obese. Incidence was defined as the occur-
ties, the second-stage units were schools within rence of a new case of obesity in a child who was
the sampled units, and the third-stage units not previously obese. We calculated the inci-
were students within schools.8 The study enrolled dence of obesity on the basis of the follow-up
21,260 children who were starting kindergarten data for 6807 children who were not already
obese in kindergarten and thus were at risk for Education Statistics to make nationally represen-
incident obesity. We also calculated incidence tative inferences. All analyses were performed
proportions by dividing the number of newly with the use of SUDAAN software, version 10.1
obese children by the number of children at risk (Research Triangle Institute).
during the follow-up period. Because the inter-
vals between the study phases varied, we calcu- R e sult s
lated the annual incidence by dividing the inci-
dence by the length of the interval between the Prevalence of Obesity
study phases in years. Cumulative incidence When children were entering kindergarten, at a
shows the 9-year risk of obesity. mean age of 5.6 years, 14.9% were overweight
In prespecified alternative analyses, we calcu- (Table S1 in the Supplementary Appendix, avail-
lated incidence density rates, which better account able with the full text of this article at NEJM.org),
for the unequal intervals between study phases and 12.4% were obese (Table 1 and Fig. 1A and 1B,
and nonconstant incidence according to age. We left panels). The prevalence of obesity increased
divided the number of new obesity cases by the at subsequent ages, reaching 20.8% by eighth
number of person-years of follow-up, which was grade (mean age, 14.1 years). There were no sig-
expressed as a rate per 1000 person-years. We mea- nificant increases in prevalence between the ages
sured person-years at risk using age in months, of 11 and 14 years.
beginning at baseline and continuing either to the The prevalence of obesity was higher among
midpoint between the last study phase in which Hispanic children than among non-Hispanic
the child was not obese and the study phase in white children at all ages (Table 1). Starting in
which incident obesity was noted or to the end third grade, non-Hispanic black children also had
of follow-up in eighth grade whichever oc- a significantly higher prevalence of obesity than
curred first.11 non-Hispanic white children. Among all chil-
dren during the follow-up period, the greatest
Statistical Analysis increase in the prevalence of obesity was between
To assess the incidence of obesity in the major first and third grades, when the prevalence in-
population groups, we stratified the estimates creased from 13.0% to 18.6%. Between kindergar-
of prevalence and incidence according to sex, ten and eighth grade, the prevalence of obesity
quintile of the kindergartners household socio- increased by 65% among non-Hispanic white
economic status, and race or ethnic group (non- children, 50% among Hispanic children, nearly
Hispanic white, non-Hispanic black, Hispanic, 120% among non-Hispanic black children, and
or other). To understand the importance of more than 40% among children of other races
weight early in life, we also stratified the data (Asian, Pacific Islander, Native American, and
according to birth weight (<2500 g, 2500 to multiracial children).
3999 g, and 4000 g) and baseline weight in kin- Children from the wealthiest 20% of families
dergarten (normal weight vs. overweight but not had a lower prevalence of obesity in kindergarten
obese). To compare the risk of obesity between than did those in all the other socioeconomic
normal-weight and overweight children, we cal- quintiles (7.4%, vs. 13.8% and 16.5% among chil-
culated risk ratios for the incidence of obesity in dren in the two poorest quintiles, respectively);
overweight kindergartners divided by the inci- these differences increased through eighth grade.
dence in normal-weight kindergartners. Finally, we At all ages, the prevalence of obesity was highest
used logistic regression to determine clinically among children in the next-to-poorest quintile,
relevant predictive risks by calculating the mar- reaching 25.8% by eighth grade.
ginal predicted probabilities of being obese in There were no significant differences in the
eighth grade as a function of the percentile of prevalence of obesity between kindergartners with
BMI and z score at younger ages. a low birth weight (<2500 g) and those with an
We used variance estimates for constructing average birth weight (2500 to 3999 g) (9.3% and
95% confidence intervals with Taylor series linear- 11.2%, respectively), but there was a significantly
ization to account for the complex sample design.12 higher prevalence at all ages among children who
We used longitudinal weights and survey adjust- had a high birth weight (4000 g) than among
ments constructed by the National Center for children in the other two birth-weight groups.
Kindergarten, Kindergarten, First Grade, First Grade, Third Grade, Fifth Grade, Eighth Grade,
Fall Semester: Spring Semester: Fall Semester: Spring Semester: Spring Semester: Spring Semester: Spring Semester:
Mean Age, 5.6 Yr Mean Age, 6.1 Yr Mean Age, 6.6 Yr Mean Age, 7.1 Yr Mean Age, 9.1 Yr Mean Age, 11.1 Yr Mean Age, 14.1 Yr
All children 7738 12.4 (11.213.7) 12.2 (11.013.5) 12.6 (10.415.1) 13.0 (11.914.3) 18.6 (17.319.9) 21.9 (20.423.4) 20.8 (19.122.5)
Boys 3865 13.4 (11.615.5) 13.3 (11.515.3) 14.1 (10.618.3) 14.1 (12.216.2) 19.4 (17.621.4) 23.6 (21.525.9) 23.5 (21.325.9)
The
Girls 3873 11.2 (9.712.9) 11.0 (9.612.7) 11.1 (8.714.1) 11.9 (10.313.7) 17.6 (15.919.5) 20.0 (17.922.2) 17.8 (15.620.2)
Socioeconomic quintile
1 1025 13.8 (11.516.5) 12.6 (10.515.1) 15.6 (10.821.9) 14.7 (12.117.7) 20.1 (17.423.0) 25.2 (21.828.9) 24.1 (19.928.9)
2 1278 16.5 (12.821.0) 17.4 (14.121.3) 18.1 (12.625.4) 15.7 (12.519.6) 23.5 (20.027.3) 25.5 (21.829.7) 25.8 (21.430.6)
3 1484 12.0 (9.315.3) 13.4 (10.516.9) 12.3 (8.118.2) 16.2 (13.020.0) 20.7 (17.624.3) 25.7 (22.229.5) 24.2 (20.528.3)
4 1594 12.2 (9.315.8) 11.5 (8.815.0) 12.8 (8.618.5) 11.4 (8.914.5) 17.8 (14.521.6) 20.9 (17.424.9) 20.5 (16.824.8)
5 2047 7.4 (5.69.7) 6.6 (5.08.6) 5.2 (3.28.4) 7.0 (5.49.1) 10.8 (8.813.3) 12.3 (10.214.9) 11.4 (9.114.2)
Race or ethnic group
Non-Hispanic white 4822 10.3 (8.912.0) 9.3 (7.911.0) 9.7 (7.113.3) 10.6 (9.212.3) 14.6 (13.116.2) 17.9 (16.219.8) 17.0 (15.318.9)
n e w e ng l a n d j o u r na l
Hispanic 1301 17.8 (15.020.9) 18.8 (15.722.4) 19.5 (14.326.0) 18.6 (15.622.0) 27.2 (23.730.9) 29.2 (25.333.5) 26.6 (23.130.5)
Other 853 14.5 (10.619.5) 15.3 (11.420.2) 13.2 (8.220.7) 16.1 (11.821.5) 19.5 (15.224.8) 23.0 (17.829.1) 20.8 (15.327.7)
<2500 g 711 9.3 (6.114.0) 10.4 (7.015.1) 9.8 (5.516.8) 12.4 (9.416.2) 15.2 (11.320.2) 19.0 (14.824.2) 19.5 (15.224.7)
25003999 g 6035 11.2 (10.012.5) 11.1 (10.012.4) 12.0 (9.914.5) 11.8 (10.513.3) 17.8 (16.419.3) 20.9 (19.222.8) 19.4 (17.721.2)
4000 g 915 22.5 (18.127.5) 21.1 (16.326.8) 21.7 (13.533.1) 21.3 (17.026.4) 26.3 (21.731.4) 30.9 (25.536.8) 31.2 (25.837.2)
* Data are from the Early Childhood Longitudinal Study, Kindergarten Class of 19981999.8 CI denotes confidence interval.
Downloaded from nejm.org on September 12, 2017. For personal use only. No other uses without permission.
The fall semester of first grade was a random subsample of the entire cohort consisting of 2277 children.
Quintiles range from the lowest fifth (1) to the highest fifth (5). Data on socioeconomic status were missing for 310 children, so the numbers in the subcategories do not total 7738.
Race or ethnic group was reported by parents of the children or collected from school records. The category designated as other includes Asian, Pacific Islander, Native American,
and multiracial background.
Data on birth weight were missing for 77 children, so the numbers in the subcategories do not total 7738.
Incidence of Childhood Obesity in the United States
A Boys
30 30
Overweight Obese Overweight at baseline
Normal weight at baseline
20 20
Prevalence (%)
15 15
10 10
5 5
0 0
5 6 7 8 9 10 11 12 13 14 5 6 7 8 9 10 11 12 13 14
Age (yr) Age (yr)
B Girls
30 30 Overweight at baseline
Overweight Obese
Normal weight at baseline
Annual Incidence of Obesity (%)
25 25 Total cohort
20 20
Prevalence (%)
15 15
10 10
5 5
0 0
5 6 7 8 9 10 11 12 13 14 5 6 7 8 9 10 11 12 13 14
Age (yr) Age (yr)
Figure 1. Prevalence and Incidence of Obesity between Kindergarten and Eighth Grade.
Shown are the age-specific prevalence of overweight and obesity (left graph on each panel) and annual incidence of obesity according to
the weight status at baseline (right graph on each panel) among boys (Panel A) and girls (Panel B). The black vertical lines and I bars
represent 95% confidence intervals.
Boys 13.7 (11.915.5) Reference 9.1 (7.510.6) Reference 36.6 (29.743.4) Reference 4.03 (3.145.18)
Girls 10.1 (8.212.0) 0.006 6.6 (4.98.3) 0.03 26.9 (20.333.4) 0.04 4.07 (2.945.65)
Socioeconomic quintile
1 13.7 (10.017.4) 0.004 9.3 (5.912.6) 0.02 31.6 (21.441.9) 0.34 3.41 (2.115.50)
2 13.7 (10.117.3) 0.006 9.9 (6.413.3) 0.03 31.7 (19.643.8) 0.37 3.21 (1.935.34)
3 15.4 (12.218.6) <0.001 10.0 (6.813.3) 0.005 38.3 (29.846.8) 0.04 3.82 (2.595.63)
4 11.5 (8.114.8) 0.04 6.8 (4.09.7) 0.24 34.3 (21.047.6) 0.22 5.02 (2.858.83)
5 7.4 (5.29.6) Reference 4.9 (3.06.8) Reference 24.3 (13.734.8) Reference 4.99 (2.778.99)
n e w e ng l a n d j o u r na l
Other 10.1 (5.714.5) 0.97 6.6 (2.810.5) 0.98 27.0 (11.142.8) 0.81 4.07 (1.849.02)
Birth weight
<2500 g 10.7 (7.913.5) 0.56 8.6 (5.811.3) 0.58 25.8 (10.840.8) 0.54 3.40 (1.666.95)
25003999 g 11.6 (10.113.1) Reference 7.7 (6.59.0) Reference 30.7 (25.336.2) Reference 3.97 (3.105.09)
Downloaded from nejm.org on September 12, 2017. For personal use only. No other uses without permission.
4000 g 16.4 (11.321.4) 0.07 8.1 (4.411.8) 0.86 41.2 (28.054.4) 0.14 5.11 (2.928.94)
* Data are from the Early Childhood Longitudinal Study, Kindergarten Class of 19981999.8
Incidence of Childhood Obesity in the United States
Incidence of Obesity According to Weight factor of 2.8. The largest differences in risk were
in Kindergarten among children who had a birth weight of more
A total of 45.3% of incident obesity cases between than 4000 g and had become overweight by the
kindergarten and eighth grade occurred among age of 5 years. These children were 5.1 times as
the 14.9% of children who were overweight when likely to become obese during the subsequent 9
they entered kindergarten (Table S4 in the Sup- years as were children with the same high birth
plementary Appendix). The annual incidence of weight whose growth trajectories led to a normal
obesity during kindergarten among these chil- weight at the age of 5 years.
dren was 19.7%, as compared with 2.4% among
children who entered kindergarten with normal Quantifying Weight Trajectories
weight (Fig. 1A and 1B, right panels, and Table S2 Children at the 50th percentile of body-mass in-
in the Supplementary Appendix). Consistent with dex at the age of 5 years had a 6% probability of
these data, incidence density rates were 91.5 vs. 17.2 being obese at the age of 14 years (Table 3). This
per 1000 person-years for overweight and normal- probability increased to 25% among 5-year-olds at
weight kindergartners, respectively (Table S3 in the the 85th percentile and to 47% among those at the
Supplementary Appendix). 95th percentile. Among children who were at the
The high incidence of obesity among children 99th percentile in kindergarten, 72% could expect
who were overweight in kindergarten fell with to still be obese as they finished eighth grade.
increasing age, so that between the ages of 11 and
14 years, the annual incidence was 3.7% (4.8% for Discussion
boys and 2.6% for girls) (Table S2 in the Supple-
mentary Appendix). A total of 31.8% of the The incidence of obesity between the ages of 5 and
children who were overweight at kindergarten 14 years was 4 times as high among children who
entry had become obese by the age of 14 years, had been overweight at the age of 5 years as among
as compared with 7.9% of their normal-weight children who had a normal weight at that age.
kindergarten classmates (Table 2). Even among Consequently, 45% of incident obesity between the
kindergartners from families with the highest ages of 5 and 14 years (kindergarten and eighth
socioeconomic status, the incidence was much grade) occurred among the 14.9% of children
higher among those who had been overweight who were overweight (85th to 95th percentile for
rather than normal weight in kindergarten. There age- and sex-specific BMI) at the age of 5 years.
were no significant differences in incidence among Furthermore, 87% of obese eighth graders had
children of various races or ethnic groups who had a BMI above the 50th percentile in kinder-
were already overweight in kindergarten. garten, and 75% had been above the 70th percen-
Overweight kindergartners had four times the tile; only 13% of children who were normal
risk of becoming obese by the age of 14 years as weight in eighth grade had been overweight in
normal-weight kindergartners (Table 2). The rela- kindergarten.
tive risks of obesity among overweight kindergart- The annualized incidence of obesity was fairly
ners, as compared with normal-weight kinder- constant among normal-weight kindergartners
gartners, were highest among children from the but fell with increasing age from high levels
two highest socioeconomic groups. Thus, over- among children who were overweight at kinder-
weight children from the two highest socioeco- garten entry. The results are consistent with in-
nomic groups had five times the risk of becom- cident obesity occurring largely among the mi-
ing obese as normal-weight children of similar nority of children who become overweight at
socioeconomic status, whereas an overweight young ages, with incidence tapering off as this
child from the lowest socioeconomic group had susceptible pool is exhausted.
only 3.4 times the risk of obesity as a normal- Our estimates are consistent with nationally
weight child of similar socioeconomic status. representative data, which showed the prevalence
Non-Hispanic white and black kindergartners of obesity at 16.9% among all children and 18.0%
who were overweight had higher incidences of among elementary-school children between the
obesity (by factors of 4.4 and 4.3, respectively) ages of 6 and 11 years in 2009 and 2010.4 The in-
than did normal-weight children; among His- cidence of obesity between adolescence and adult-
panic children, the incidence was higher by a hood in the United States was estimated at 2.5%
Table 3. Probability of Obesity in Eighth Grade, Spring Semester (Mean Age, 14.1 Years), According to z Score and Percentile of Body-Mass
Index at Earlier Ages.*
Percentile of
Weight Category Body-Mass
and z Score Index Probability of Obesity in Eighth Grade, Spring Semester
Kindergarten, Kindergarten, First Grade, First Grade, Third Grade, Fifth Grade,
Fall Semester: Spring Semester: Fall Semester: Spring Semester: Spring Semester: Spring Semester:
Mean Age, Mean Age, Mean Age, Mean Age Mean Age, Mean Age,
5.6 Yr 6.1 Yr 6.6 Yr 7.1 Yr 9.1 Yr 11.1 Yr
percent
Normal weight
0.00 50 6 6 5 5 2 <1
0.25 60 9 9 8 8 3 1
0.52 70 13 13 12 12 5 1
0.84 80 19 20 19 19 11 4
Overweight
1.04 85 25 25 25 24 16 7
1.28 90 33 34 33 33 25 16
Obese
1.64 95 47 49 48 48 44 39
2.33 99 72 75 75 76 80 87
* Data are from the Early Childhood Longitudinal Study, Kindergarten Class of 19981999.8
annually from 1995 through 2000.6 In a study of normal weight, the incidence of obesity was low
386 children between the ages of 5 and 7 years and constant between the ages of 5 and 14 years.
attending Philadelphia health care centers from Emerging from the finding that a substantial
1996 through 2003, the incidence of obesity was component of childhood obesity is established
2% annually among normal-weight children and by the age of 5 years are questions about how
14% among overweight children.13 early the trajectory to obesity begins and about
Although prevalence estimates provide infor- the relative roles of early-life home and pre-
mation on the burden of obesity, understanding school environments, intrauterine factors, and
incidence is key to understanding risk over a genetic predisposition. Although these questions
lifetime and identifying potential ages for inter- are beyond the scope of our study, we have shown
vention. We uncovered several important points some evidence that factors that are established
by examining incidence. First, a component of before birth (indicated by birth weight) and those
the course to obesity is already established by that occur during the first 5 years of life (indi-
the age of 5 years: half of childhood obesity oc- cated by weight at kindergarten entry) are im-
curred among children who had become over- portant. Even though high-birth-weight children
weight during the preschool years, even after the made up 12% of the population, they represented
exclusion of the 12.4% of children who were al- more than 36% of those who were obese at the
ready obese at the age of 5 years. There is evi- age of 14 years. Thus, more than one third of
dence that body weight and eating patterns early high-birth-weight children became obese adoles-
in life are strongly related to subsequent obesity cents, as did almost half the children (45.3%)
risks.7 Second, obesity incidence among over- who entered kindergarten overweight.
weight children tended to occur early in elemen- This study has certain limitations. First, to
tary school. This pattern is consistent with ex- maintain the sample size for stratified analyses,
haustion of the population of persons who are we grouped into an other race category Asian,
highly susceptible to becoming obese.14 In con- Pacific Islander, Native American, and multiracial
trast, among children who entered school at a children. Second, we did not have information
on weight between birth and kindergarten or patterns from these methods are consistent, as
after eighth grade, so we cannot map the entire are results from sensitivity analyses separating
trajectory of incidence or identify the age at children who reversed weight trajectories from
which children who entered kindergarten over- those who remained obese through the end of
weight or obese had become overweight or follow-up.
obese. Lacking data before and after a period of By the time they enter kindergarten, 12.4% of
observation, called left and right censoring, is American children are already obese, and 14.9%
common in studies of disease incidence.15 Third, are overweight. Almost half the obesity incidence
the cohort is representative of children who were from kindergarten through eighth grade occurs
in kindergarten in 1998 and 1999 and may not among children who were overweight as kinder-
reflect the experiences of earlier or later cohorts. gartners. Furthermore, 36% of incident obesity
Still, this cohort is of particular interest because between the ages of 5 and 14 years occurred
they were growing up during the 1990s and among children who were large at birth. These
2000s, when obesity became a major health con- findings highlight the importance of further re-
cern. Finally, given the focus on documenting search to understand the factors associated with
obesity incidence, it was beyond the scope of this the development of overweight during the first
study to model the factors associated with the years of life. We speculate that obesity-prevention
development of obesity. efforts that are focused on children who are over-
A question regarding statistical analysis is weight by the age of 5 years may be a way to target
how to treat data for children who are obese at the children who are most susceptible to becom-
one point but subsequently lose weight and be- ing obese during later childhood and adolescence.
come overweight or normal weight. In the analy-
sis of incidence, we considered everyone who was The views expressed in this article are those of the authors
and do not necessarily represent the official views of the Eunice
not obese at a given study phase to be at risk for Kennedy Shriver National Institute of Child Health and Human
becoming obese by the next study phase, regard- Development or the National Institutes of Health.
less of whether they had previously been obese. Supported by a grant from the Eunice Kennedy Shriver Na-
tional Institute of Child Health and Human Development
In alternative models of incidence density rates, (R03HD060602).
we reported cumulative obesity risks, consider- Disclosure forms provided by the authors are available with
ing as incident cases only children who became the full text of this article at NEJM.org.
We thank Patricia Cheung, Lisa Matz, and Mark Hutcheson for
obese during the period of observation and re- their assistance in literature searches and in the preparation of
mained obese through the end of follow-up. The earlier versions of the figures.
References
1. Glickman D, Parker L, Sim LJ, Cook dence in the transition period between for child overweight and obesity world-
HDV, Miller EA, eds. Accelerating prog- adolescence and adulthood: the National wide: international survey. BMJ 2000;320:
ress in obesity prevention: solving the Longitudinal Study of Adolescent Health. 1240-3.
weight of the nation. Washington, DC: Am J Clin Nutr 2004;80:569-75. 11. Szklo M. Nieto J. Epidemiology: be-
National Academies Press, 2012. 7. Baidal JA, Taveras EM. Childhood obe- yond the basics. Sudbury, MA: Jones and
2. Jolliffe D. Extent of overweight among sity: shifting the focus to early prevention. Bartlett, 2006.
US children and adolescents from 1971 to Arch Pediatr Adolesc Med 2012;166:1179- 12. Williams RL. Taylor Series Lineariza-
2000. Int J Obes Relat Metab Disord 81. tion (TSL). In: Lavrakas PJ, ed. Encyclope-
2004;28:4-9. 8. Tourangeau K, Nord C, L T, Sorongon dia of survey research methods. Thousand
3. Lakshman R, Elks CE, Ong KK. AG, Najarian M. Early Childhood Longitu- Oaks, CA: Sage, 2008.
Childhood obesity. Circulation 2012;126: dinal Study, Kindergarten Class of 199899 13. Robbins JM, Khan KS, Lisi LM, Rob-
1770-9. (ECLS-K): combined users manual for the bins SW, Michel SH, Torcato BR. Over-
4. Ogden CL, Carroll MD, Kit BK, Flegal ECLS-K Eighth-Grade and K8 full sam- weight among young children in the
KM. Prevalence of obesity and trends in ple data files and electronic codebooks Philadelphia health care centers: inci-
body mass index among US children and (NCES 2009004). Washington, DC: Na- dence and prevalence. Arch Pediatr Ado-
adolescents, 1999-2010. JAMA 2012;307: tional Center for Education Statistics, In- lesc Med 2007;161:17-20.
483-90. stitute of Education Sciences, 2009. 14. Vaupel JW, Yashin AI. Heterogeneitys
5. Singh GK, Kogan MD, van Dyck PC. 9. Vidmar S, Caelin J, Hesketh K, Cole ruses: some surprising effects of selection
Changes in state-specific childhood obe- TJ. Standardizing anthropometric mea- on population dynamics. Am Stat 1985;
sity and overweight prevalence in the sures in children and adolescents with 39:176-85.
United States from 2003 to 2007. Arch new functions for egen. Stata J 2004;4: 15. Kleinbaum D, Klein M. Survival analy-
Pediatr Adolesc Med 2010;164:598-607. 50-5. sis: a self-learning text. 2nd ed. New York:
6. Gordon-Larsen P, Adair LS, Nelson 10. Cole TJ, Bellizzi MC, Flegal KM, Dietz Springer, 2005.
MC, Popkin BM. Five-year obesity inci- WH. Establishing a standard definition Copyright 2014 Massachusetts Medical Society.