Freedman DS, Khan LK, Serdula MK, Galuska DA, Dietz WH. Trends and Correlates of Class 3 Obesity in the United States From 1990 Through 2000. JAMA. 2002;288(14):1758–1761. doi:10.1001/jama.288.14.1758
Author Affiliations: Division of Nutrition and Physical Activity, Centers for Disease Control and Prevention, Atlanta, Ga.
Context Although the prevalence of obesity has markedly increased among US adults,
health risks vary according to the severity of obesity. Persons with class
3 obesity (body mass index [BMI] ≥40) are at greatest risk, but there is
little information about this subgroup.
Objective To examine correlates of class 3 obesity and secular trends.
Design, Setting, and Participants Adults (aged ≥18 years) in the United States who participated in
the Behavioral Risk Factor Surveillance System telephone survey between 1990
(75 600 persons) and 2000 (164 250 persons).
Main Outcome Measure Body mass index calculated from self-reported weight and height.
Results The prevalence of class 3 obesity increased from 0.78% (1990) to 2.2%
(2000). In 2000, class 3 obesity was highest among black women (6.0%), persons
who had not completed high school (3.4%), and persons who are short. During
the 11-year period, the median BMI level increased by 1.2 units and the 95th
percentile increased by 3.2 units.
Conclusion The prevalence of class 3 obesity is increasing rapidly among adults.
Because these extreme BMI levels are associated with the most severe health
complications, the incidence of various diseases will increase substantially
in the future.
The prevalence of obesity, defined as a body mass index (BMI) of 30
or more, has markedly increased during the last 3 decades in the United States1- 3 and other countries.4- 6 Between 1976-1980 and
1999, for example, the prevalence of obesity increased from 13% to 27% among
There are few epidemiologic investigations, however, of more extreme
BMIs. Based on differences in treatment and health risks, obesity has been
categorized as class 1 (BMI, 30-34.9), class 2 (BMI, 35-39.9), and class 3
(BMI, ≥40).8- 11 Persons
with class 3 obesity, also termed morbid or extreme obesity, are potential
candidates for antiobesity surgery12 and have
a 2-fold higher risk for all-cause mortality than persons with BMIs of 30
The prevalence of class 3 obesity increased from 1% to 3% between 1960-1962
and 1988-1994 and is highest among black women.1,2 The
objectives of the current study are to determine if trends have continued
through 2000 and to examine various correlates of class 3 obesity.
We analyzed data from the Behavioral Risk Factor Surveillance System
(BRFSS), a multistage survey that uses random-digit dialing to obtain a representative
sample of adults (≥18 years old) in each state.14 Several
steps were taken to ensure quality control in BRFSS, and the current analyses
are based on data collected from 1990 (44 states) through 2000 (50 states).
The questionnaire focuses on behaviors associated with disease risks,15 and the data have been used to examine secular trends
in overweight and obesity (cutpoints of 27-32).3,16 Because
height was recorded in inches and weight in pounds, these units are used throughout
the text; BMI was calculated as the weight in kilograms divided by the squared
height in meters.
More than 1 million persons participated in the BRFSS between 1990 and
2000. Approximately 4% of participants did not report weight or height, and
the current analyses are limited to non-Hispanic whites, non-Hispanic blacks,
and Hispanics. Other race/ethnic groups (approximately 4%) are excluded, as
are pregnant women (approximately 2%) and persons (n = 364) with extreme values
of weight or height (eg, weight >560 lb). Yearly totals for the current analyses
ranged from 75 600 (1990) to 164 250 persons (2000).
All analyses accounted for the unequal selection probabilities.17 Statistical testing is not emphasized in this large
sample, but SEs were calculated using SUDAAN to account for the design.17
Trends in various BMI categories are examined according to sex, race/ethnicity,
age, educational achievement, and height. The independent relation of these
characteristics to class 3 obesity (and possible interactions) was examined
in logistic regression analyses.17 Differences
in the BMI distributions in 1990-1991 vs 2000 were examined using a percentile
The mean BMI increased from 24.9 (1990) to 26.5 (2000), and Figure 1 illustrates trends in 5 BMI categories.
Although the prevalence of all BMI categories greater than 25 increased during
the study period, the most striking increases were in the extreme BMI categories,
with class 3 obesity increasing from 0.78% to 2.2%.
The prevalence of class 3 obesity according to various characteristics
is examined in Table 1. (Because
of the smaller number of persons interviewed in 1990, data for 1990 and 1991
have been combined.) In each year, the prevalence of class 3 obesity was approximately
2-fold higher among women than men, with the highest prevalence among black
women (6% in 2000). Among men, there was little difference in class 3 obesity
by race/ethnicity in 1990-1991, but the prevalence in 2000 was highest among
blacks (2.4%). Although secular increases were seen in all age groups, the
largest proportional increase (0.4% to 1.2%) was among 18- to 29-year-olds.
The prevalence of class 3 obesity also increased within all categories of
educational achievement, with the prevalence highest among persons who did
not complete high school.
Additional analyses indicated that although mean levels of weight increased
by 9 to 12 lb during the 11 years, these weight increases varied only slightly
by height. We therefore examined the prevalence of class 3 obesity within
various height categories (Table 1,
bottom). Among both men and women, the prevalence of class 3 obesity was about
2-fold higher among short persons than among taller persons (eg, <67 vs
≥74 inches among men).
Logistic regression analyses indicated that each characteristic in Table 1, including year of study, was independently
associated with the prevalence of class 3 obesity. For example, educational
achievement was inversely associated with class 3 obesity among 30- to 69-year-olds
in each race-sex group, with the prevalence highest (12% in 2000) among black
women who did not complete high school (Table 2). Additional analyses indicated that the association with
educational achievement was stronger among women than men.
All percentiles of BMI were higher in 2000 than in 1990-1991, but the
increases were larger at high BMI levels (Figure 2). For example, although the 10th percentile of BMI increased
by 0.6 units, the 95th percentile increased by 3.2 units (36.9 − 33.7
Most studies of secular trends in obesity have focused on mean BMIs
or on BMIs of 30 or higher, but our results indicate that the prevalence of
BMIs of 40 or higher increased almost 3-fold between 1990 and 2000. These
trends will greatly increase the risk for various diseases and premature mortality.13,19 Approximately 75% of adults with
class 3 obesity have at least one comorbid condition, such as high blood pressure
or diabetes mellitus.20 Furthermore, a BMI
of 40 or higher is associated with a 2-fold higher risk for all-cause mortality
than are BMIs of 30 to 31.9.13
Despite these consequences, relatively little is known about the distribution
of class 3 obesity. We found the highest prevalence among black women and
among persons with low levels of educational achievement. Although an inverse
association between social class and less extreme obesity has consistently
been found among women, a review21 of articles
before 1990 concluded that the association among men was inconsistent. Our
observations suggest that the (inverse) association among men may have become
stronger during the last decade. In agreement with this possibility, several
recent studies22- 24 have
found that obesity (BMI >30) is inversely associated with social class among
men. However, the sex differences before 1990 may also reflect the greater
stigmatization of obesity among women.
An unexpected finding was the high prevalence of class 3 obesity among
shorter adults. Although it is generally assumed that height and BMI are uncorrelated,
an inverse association (r = −0.10) has been
reported,25 and others26,27 have
found a relatively high prevalence of obesity among short adults. It is possible
that the association between BMI and height may be influenced by characteristics,
such as dietary intake and physical activity, that vary only slightly by height.
In agreement with the trends observed between 1966-1970 and 1988-199428 is our finding that increases in BMI were most striking
at high BMIs. Although all BMI percentiles increased between 1990 and 2000,
the median increased by 1.2, whereas the 95th percentile increased by 3.2.
We also found that 18- to 29-year-olds showed a large proportional increase
in class 3 obesity, possibly reflecting the increases in childhood BMI (particularly
at the upper percentiles) that have occurred since 1975.28,29
An important limitation of the current study is the use of self-reported
rather than measured weight and height. The BMIs based on self-reported and
measured data are highly correlated (r>0.95), and
self-reported data have been used in cohort studies13,30,31 and
in studies of secular trends.3,5,16 However,
because height is overreported and weight is underreported, BMIs based on
self-reported data are biased downward.32- 35 This
bias increases at higher BMIs,32- 34 and
the sensitivity of self-reported data to detect a BMI of 30 or higher ranges
from 63% to 74%.32,33,36 This
underreporting likely accounts for the approximately 50% lower prevalence
of class 3 obesity that we observed in 1995 (1% for men and 1.6% for women)
than the estimates of 2% to 4% reported by others.1,2 However,
if these biases remained fairly constant during the study, our observed trends
and subgroup differences would parallel those calculated from measured data.
The proportional increase in class 3 obesity that we observed between 1990
and 2000 is similar to those reported (based on measured data) from 1976-1980
to 1988-19941 and from 1985 to 1995.2
Although the optimum BMI remains uncertain,37 the
trends in class 3 obesity will result in substantial increases in morbidity
and premature mortality. Additional studies are needed to elucidate future
trends and to identify other characteristics that may be associated with class
3 obesity, such as repeated weight increases during pregnancy among women
with more than 3 children.38,39 Because
weight loss is difficult to maintain, the prevention of obesity should be