Must A, Spadano J, Coakley EH, Field AE, Colditz G, Dietz WH. The Disease Burden Associated With Overweight and Obesity. JAMA. 1999;282(16):1523-1529. doi:10.1001/jama.282.16.1523
Author Affiliations: Department of Family Medicine and Community Health, Tufts University School of Medicine (Dr Must); US Department of Agriculture, Human Nutrition Research Center on Aging at Tufts University (Dr Must and Ms Spadano); Departments of Nutrition (Ms Coakley) and Epidemiology (Ms Coakley and Dr Colditz), Harvard School of Public Health; Channing Laboratory (Drs Field and Colditz), Brigham and Women's Hospital and Harvard Medical School; and Department of Pediatric Gastroenterology and Nutrition, New England Medical Center (Dr Dietz), Boston, Mass; and Division of Nutrition and Physical Activity, Center for Disease Prevention and Health Promotion, Centers for Disease Control and Prevention, Atlanta, Ga (Dr Dietz).
Context Overweight and obesity are increasing dramatically in the United States
and most likely contribute substantially to the burden of chronic health conditions.
Objective To describe the relationship between weight status and prevalence of
health conditions by severity of overweight and obesity in the US population.
Design and Setting Nationally representative cross-sectional survey using data from the
Third National Health and Nutrition Examination Survey (NHANES III), which
was conducted in 2 phases from 1988 to 1994.
Participants A total of 16,884 adults, 25 years and older, classified as overweight
and obese (body mass index [BMI] ≥25 kg/m2) based on National
Institutes of Health recommended guidelines.
Main Outcome Measures Prevalence of type 2 diabetes mellitus, gallbladder disease, coronary
heart disease, high blood cholesterol level, high blood pressure, or osteoarthritis.
Results Sixty-three percent of men and 55% of women had a body mass index of
25 kg/m2 or greater. A graded increase in the prevalence ratio
(PR) was observed with increasing severity of overweight and obesity for all
of the health outcomes except for coronary heart disease in men and high blood
cholesterol level in both men and women. With normal-weight individuals as
the reference, for individuals with BMIs of at least 40 kg/m2 and
who were younger than 55 years, PRs were highest for type 2 diabetes for men
(PR, 18.1; 95% confidence interval [CI], 6.7-46.8) and women (PR, 12.9; 95%
CI, 5.7-28.1) and gallbladder disease for men (PR, 21.1; 95% CI, 4.1-84.2)
and women (PR, 5.2; 95% CI, 2.9-8.9). Prevalence ratios generally were greater
in younger than in older adults. The prevalence of having 2 or more health
conditions increased with weight status category across all racial and ethnic
Conclusions Based on these results, more than half of all US adults are considered
overweight or obese. The prevalence of obesity-related comorbidities emphasizes
the need for concerted efforts to prevent and treat obesity rather than just
its associated comorbidities.
Overweight and obesity are common health conditions and their prevalence
is increasing globally.1- 3
Recent estimates suggest that 1 in 2 adults in the United States is overweight
or obese, defined by a body mass index (BMI) of higher than 25, an increase
of more than 25% over the past 3 decades.4
These dramatic increases have occurred among the 3 major racial and ethnic
groups and include both sexes.4
Excess weight is associated with an increased incidence of cardiovascular
disease, type 2 diabetes mellitus (DM), hypertension, stroke, dyslipidemia,
osteoarthritis, and some cancers.5 Associations
of excess weight with overweight- and obesity-related mortality may differ
among racial and ethnic groups.6- 10
In 1998, the National Heart, Lung, and Blood Institute of the National
Institutes of Health published evidence-based clinical guidelines for the
identification, evaluation, and treatment of overweight and obesity in adults.11 This classification resembles the categories used
by the World Health Organization.1 Based on
BMI, calculated as weight in kilograms divided by the square of height in
meters, both approaches use 4 classes of increasing severity, consistent with
the notion of graded risk in the US Department of Agriculture's Dietary Guidelines for Americans.12,13
Several recent articles that rely on the Third National Health and Nutrition
Examination Survey (NHANES III) data14- 18
largely focus on single-health outcomes15,16
and population subgroups.17,18
This study was undertaken to provide estimates of the prevalence of morbid
conditions associated with obesity by severity, race and ethnicity, and age,
as well as by the frequency of multiple obesity-related comorbidities in the
Conducted by the National Center for Health Statistics of the Centers
for Disease Control and Prevention, NHANES III was designed to provide nationally
representative data to estimate the prevalence of major diseases, nutritional
disorders, and potential risk factors. In NHANES III, 2 waves of data were
collected in 2 phases from 1988 to 1994. The total sample included 33,199
persons; 16,884 were at least 25 years old. The sampling plan followed a complex,
stratified, multistage, probability cluster design to produce estimates representative
of the noninstitutionalized civilian US population. To improve the reliability
of estimates for non-Hispanic blacks and Mexican Americans as well as young
children and the elderly, NHANES III oversampled these groups. Further details
of the design and operation of the survey are available elsewhere.19,20
The survey protocol included a home interview and a standardized physical
examination in a mobile examination center or a limited examination in the
subject's home for subjects unable to travel. In both settings, body weight
and height were measured by trained technicians with standardized equipment
We used the National Heart, Lung, and Blood Institute's definitions
for the cutoff points between overweight and obesity and between obese and
its class levels.11 Using a reference category
of BMI 18.5 to 24.9 kg/m2 (termed normal weight), the 4 classes as shown in Table 1, which also lists the World Health Organization's cutoff
points.1 Both organizations use the same categories
with the exception of the BMI range of 25.0 to 29.9 kg/m2
category: WHO calls it preobese and the National
Heart, Lung, and Blood Institute calls it overweight.
Both organizations consider persons with a BMI of between 25.0 and 29.9 kg/m2 to be overweight.
Among the 16,884 subjects at least 25 years old, we excluded pregnant
women (n = 164) and persons who did not have height or weight measurements
(n = 1719).
On the basis of previous research, we selected for study health conditions
for which excess weight is an established risk factor and for which sample
sizes were adequate. These conditions included high blood pressure, type 2
DM, high blood cholesterol level, coronary heart disease (CHD), and gallbladder
disease. In addition, for analyses that considered the number of comorbidities,
osteoarthritis was included.
High blood pressure was deemed present if subjects reported that a physician
had ever told them that they had hypertension or high blood pressure, or if
the mean of at least 3 blood pressure readings (measured by NHANES technicians)
exceeded 140 mm Hg systolic or 90 mm Hg diastolic. The specific details of
the blood pressure measurements are published elsewhere.20
The average of all available blood pressure measurements was used.
Ideally, type 2 DM should be defined based on self-report of physician
diagnosis and fasting plasma glucose levels. However, fasting plasma glucose
levels were unavailable for 44% of the sample. Therefore, type 2 DM was defined
only by self-report of diagnosed disease. Because the structured interview
item for physician-diagnosed DM did not distinguish type 1 from type 2 DM,
we identified subjects with type 2 DM on the basis of the information provided
in the questionnaire. We started with all those who responded affirmatively
to the question: "Has a physician ever told you have diabetes?" We then excluded
those who reported having DM only during pregnancy. Of those who remained,
we excluded those with type 1 DM (diagnosis before age 30 years and number
of years of insulin use equal to within 1 year of duration of disease). Those
who remained were considered to have type 2 DM. This approach may misclassify
a small number of persons with type 2 DM as having type 1 DM, and an even
smaller number of persons with late-onset type 1 DM as having type 2 DM.
Subjects were classified as having high blood cholesterol levels if
they reported that a physician had told them such or if measured serum cholesterol
levels exceeded 6.2 mmol/L (240 mg/dL). Levels were determined by contract
laboratories using reference analytical methods.21
A diagnosis of CHD was based on a history of having had a "heart attack,"
congestive heart failure, or angina. Heart attack and congestive heart failure
were based on self-report of physician diagnosis. The location of chest pain
as described by self-report was used for classification of angina based on
Gallbladder disease or osteoarthritis were considered present for subjects
who had ever been told by their physician that they had either condition.
All analyses were conducted using specialized software that adjusts
for complex sample designs.23 Sample weights
were applied to produce nationally representative estimates. Prevalence estimates
were calculated by cross-tabulation. Race- and ethnicity-specific estimates
were calculated for non-Hispanic whites, non-Hispanic blacks, and Mexican
Americans. Other racial and ethnic groups were not represented in adequate
numbers for reliable estimates for these subgroups but were included in population-based
estimates. All analyses were stratified by sex.
Multivariate logistic models were developed to estimate prevalence odds
ratios (PORs) by obesity class (compared with the normal weight [BMI, 18.5-24.9
kg/m2]), adjusted for smoking status, age,
and race and ethnicity. Age was modeled as a continuous variable. Smoking
status was defined by indicator variables for never, former, and current smoking;
never smoking was the referent category. To test for interactions between
race and ethnicity and the categories of overweight and obesity, we evaluated
the significance of interaction terms for 3 racial and ethnic groups and 5
weight classes (normal, overweight, and obesity classes 1, 2, and 3) by a
likelihood ratio test comparing the logistic model with and without the interaction
terms. The race and ethnicity interaction was not significant. Similarly,
we tested age and weight class interaction terms. This interaction was significant,
so age-specific analyses, stratified into 2 broad categories of younger than
55 years and 55 years or older, are presented. To test for linear trend, we
included BMI as a continuous variable in the logistic models and tested for
its statistical significance. For common outcomes, the POR does not well approximate
the prevalence ratio (PR). We corrected the PORs using the method of Zhang
and Yu.24 Using our logistic regression models,
we estimated the baseline prevalence of each condition for normal weight individuals,
of modal race or ethnicity and modal smoking status. These baseline-adjusted
prevalences were used to convert the PORs to adjusted PRs and are presented
with the PRs to aid in the interpretation of the ratios.25
Statistical significance was set at P<.05
and the stability of the estimates reflected by 95% confidence intervals(CIs).
In describing patterns, we considered PRs to be similar when the CI for one
estimate included the point estimate of the other and when the trend tests
for a pair of models were both significant or both nonsignificant.
To evaluate multiple morbidities, we used cross-tabulation to evaluate
the proportion of persons who had 0, 1, 2, or more comorbidities by weight
status category within strata of sex and racial and ethnic group. To be conservative
for these analyses, we calculated crude prevalences only for type 2 DM, gallbladder
disease, high cholesterol levels, high blood pressure, and osteoarthritis
as obesity-related comorbidities. Coronary heart disease was not included
in these analyses because hypertension, high cholesterol levels, and type
2 DM may be in the causal pathway for heart disease. If these conditions are
intermediate events in the development of CHD, it would not be correct to
count them as 2 separate health conditions.
Based on the NHANES III sample, approximately 63% of men and 55% of
women aged 25 years or older in the US population were overweight or obese (Table 2). Specifically, 42% of men and
28% of women were overweight, and 21% of men and 27% of women were obese.
The crude prevalence estimates showed a generally similar magnitude and pattern
across racial and ethnic groups by sex with 2 exceptions: among non-Hispanic
white women the prevalence of BMI of 25 kg/m2
or higher (overweight or obesity) was lower and among Mexican American men
prevalence was higher compared with other racial or ethnic sex groups.
For both men and women, high blood pressure was the most common overweight-
and obesity-related health condition and its prevalence showed a strong increase
with increasing weight status category (Table 3). The prevalence of type 2 DM, gallbladder disease, and
osteoarthritis increased sharply among both overweight and obese men and women
corresponding with the increasing weight classes. High blood cholesterol level
was very prevalent in both sexes but showed no increase in prevalence with
increasing weight category. However, men and women with BMIs of 25 kg/m2 or more, were more likely than persons of normal weight
to have high blood cholesterol levels. Because these estimates are not adjusted
for age or smoking status, they reflect the presence of these health conditions
in the US population.
Sex-specific PRs by weight class for the health conditions considered
showed no evidence of race and ethnicity–weight class interactions for
either sex. (Racial- and ethnic-specific PRs are available from A.M.) Significant
interactions between weight status category and age group (<55 years, ≥55
years) were observed for all health outcomes examined, except for CHD (Table 4).
Type 2 DM showed a strong increase in prevalence with increasing overweight
class among both younger and older subjects. The PR associated with elevated
weight was 3- to 4-fold greater among younger overweight men and women. Among
those in the most obese group, the PR for younger men was 18.1 (95% CI, 6.7-46.8)
and 12.9 (95% CI, 5.7-28.1) for younger women. Among older men and women,
PRs were more moderate but still substantially elevated, with a PR of 3.4
(95% CI, 1.1-8.3) for the most obese men, and 5.8 (95% CI, 4.2-7.4) for the
most obese women.
The PRs for gallbladder disease differed substantially by sex. Men younger
than 55 years exhibited a marked increase in PRs for gallbladder disease with
increasing weight status, with a PR of 21.1 (95% CI, 4.1-84.2) observed in
the highest obesity class. Among older men, a weaker gradient of PRs was observed.
For women, for whom the condition is more than twice as prevalent as it is
among men, a steady increase in the PRs for gallbladder disease with increasing
weight status category was observed in both age groups, with a PR of 1.9 (95%
CI, 1.3-3.0) for overweight women younger than 55 years.
Unlike the other comorbidities examined, the PRs for CHD did not differ
by age group. Prevalence ratios were not elevated among the overweight but
were significantly elevated in obesity class 1 for men and in all 3 obesity
classes for women. Among these obese classes, PRs ranged from 1.6 among obesity
class 1 women (95% CI, 1.2-2.1) to 3.0 among obesity class 3 women (95% CI,
2.1-4.2). The linear trend of BMI was statistically significant for both sexes.
The PRs for high cholesterol level exhibited a different pattern than
that observed for the other overweight- and obesity-related comorbidities.
Among younger men and for women, PRs for most classes of overweight and obesity
were elevated compared with the reference group (BMI, 18.5-24.9 kg/m2), but there was no evidence of a gradient of PRs with
increasing weight status category. Among older subjects, PRs were significantly
increased only for overweight individuals.
A steeply graded association between weight class and high blood pressure
was observed in men and women younger than 55 years, with elevations observed
even among those in the overweight class, for which PRs were almost doubled
compared with normal weight individuals. Among older men and women, statistically
significant elevations in this highly prevalent condition were observed, beginning
with the overweight category and increasing across weight categories.
Figure 1 shows the number
of comorbidities by overweight class for each race and ethnicity–sex
group (not age adjusted). From left (normal weight) to right (greater degree
of overweight and obesity) the percentage of persons with at least 2 comorbidities
increases. A comparison of race and ethnicity–sex subgroups suggests
that these associations were consistently and directly related across racial
and ethnic groups. An appreciable increase in prevalence was evident even
for the overweight class in every racial and ethnic–sex group.
In this study, we estimated the cross-sectional relationship between
overweight and obesity class levels and morbidity in a contemporary, nationally
representative sample of adults. We observed a substantial prevalence of chronic
health conditions in association with elevated BMI for both age groups and
across racial and ethnic groups. Associations of weight status and health
outcomes did not differ between the 3 major racial and ethnic groups. The
PRs generally increased with increasing severity and, for many comorbidities,
the PRs were significantly elevated even for the overweight class (BMI, 25.0-29.9
kg/m2). Because the sample is cross-sectional,
the data reflect the burden of disease associated with overweight and obesity
in the US population aged 25 years and older from 1988 to 1994.
We observed particularly strong cross-sectional associations for overweight
and obesity with type 2 DM and hypertension, consistent with the findings
of several large cohort studies based on nonrepresentative populations.26,27 Furthermore, we found a significant
increase in PRs of both of these conditions even among persons in the overweight
class. This finding is striking given that individuals with the mildest degree
of overweight comprise more than 42% of men and 28% of women in the United
States. For the majority of health conditions studied, based on overweight
status and age, PRs are increased. Our analyses incorporate the newly adopted
definitions of overweight and obesity.1,11
The previous BMI cutoff points of 27.8 (men) and 27.3 (women) kg/m2
were based on a purely statistical definition (85th percentile from NHANES
II). The new cutoff point of 25 kg/m2 is based on research evidence
that links an elevated BMI with adverse health consequences,13
including type 2 DM, hypertension, cardiovascular disease,28- 30
and death.31 The prevalence estimates based
on these definitions differ slightly from those published by Flegal and colleagues.4 Their report was based on adults 20 years or older,
whereas our report includes adults aged 25 years or older.
Although the relationship of BMI with body fatness may differ by race
and ethnicity,32- 34
whether the relationship between weight status (based on BMI) and adverse
health outcomes differs by race and ethnicity is less clear. The pattern of
the PRs in relation to weight class was consistent across all 3 racial and
ethnic groups for all of the health outcomes we examined, despite differences
in the distribution of BMI by race and ethnicity.
The influence of age on the relationship of BMI to morbidity and mortality
has been the subject of some debate, especially for prospective studies of
For both hypertension28 and CHD,38
the relative risk associated with overweight declines with age. We observed
that the cross-sectional relationship of obesity class to the comorbidities
studied was generally strongest among the younger age groups. Nonetheless,
the PRs were significantly elevated in the older age group except for gallbladder
disease in men and high cholesterol level in both sexes.
Our analysis of multiple comorbidities did not include CHD, because
of the potential for "double-counting" of CHD with known cardiovascular risk
factors such as hypertension, type 2 DM, and high blood cholesterol level.
Our consideration of only the most common obesity-related comorbidities and
avoidance of double counting suggests that we have likely underestimated the
disease burden associated with overweight in the population. The burden to
the individual is increased for those with more severe obesity because they
are more likely to develop a second or even a third morbidity.
Our approach has several limitations. Foremost, a cross-sectional approach
to evaluate the relationship of obesity to the morbidities that we examined
does not provide evidence for causality. In some individuals, weight loss
may accompany some of these conditions, particularly CHD. Because persons
in the higher obesity classes were at increased risk of death, we may have
underestimated the impact of weight on adverse health, particularly among
older individuals. Reliance on self-report of physician-diagnosed disease
underestimates disease burden. For example, a definition based on self-report
of diagnosed disease underestimates DM prevalence by more than 50%.15 The known association between overweight and chronic
health conditions increases the likelihood of diagnosis in heavier people
and represents an additional source of bias. In addition, using BMI as a weight
measure provides an indirect measure of fatness and does not reflect fat distribution,
which may affect the risk of comorbidity independent of BMI. Moreover, BMI
does not distinguish between fat mass and lean tissue mass and may underestimate
fatness in older adults who have greater amount of body fat at a given BMI
than younger ones, due to age-related declines in muscle mass.39
Furthermore, we did not adjust our cross-sectional statistical models for
variables often included in models that estimate incidence (rather than prevalence)
of obesity-related comorbidities such as fat distribution, diet, or exercise,
because the temporal sequence in cross-sectional data is not established.
Our results might have changed if these variables had been included. However,
the consistency of our cross-sectional findings with incident chronic disease
in several large prospective studies26- 29
suggests that the cross-sectional associations we observed are likely to reflect
true associations. The aforementioned misclassification and biases would likely
have a minor impact on the magnitude of these estimates.
In conclusion, these national data suggest that clinicians are likely
to encounter morbidity more frequently among their patients with elevated
BMI, even those patients in the overweight category. A general pattern of
increasing prevalence with increasing severity of overweight and obesity is
consistent across racial and ethnic groups for all of the health conditions
considered, with the exception of high blood cholesterol level. Without concerted
initiatives to prevent and treat overweight in adults, the health care system
will increasingly be overwhelmed with individuals who require treatment for
obesity-related health conditions.