Percentage of gestational diabetes mellitus attributable to overweight and obesity.

Abstract

We calculated the percentage of gestational diabetes mellitus (GDM) attributable to overweight and obesity.
We analyzed 2004 through 2006 data from 7 states using the Pregnancy Risk Assessment Monitoring System linked to revised 2003 birth certificate information. We used logistic regression to estimate the magnitude of the association between prepregnancy body mass index (BMI) and GDM and calculated the percentage of GDM attributable to overweight and obesity.
GDM prevalence rates by BMI category were as follows: underweight (13-18.4 kg/m(2)), 0.7%; normal weight (18.5-24.9 kg/m(2)), 2.3%; overweight (25-29.9 kg/m(2)), 4.8%; obese (30-34.9 kg/m(2)), 5.5%; and extremely obese (35-64.9 kg/m(2)), 11.5%. Percentages of GDM attributable to overweight, obesity, and extreme obesity were 15.4% (95% confidence interval [CI] = 8.6, 22.2), 9.7% (95% CI = 5.2, 14.3), and 21.1% (CI = 15.2, 26.9), respectively. The overall population-attributable fraction was 46.2% (95% CI = 36.1, 56.3).
If all overweight and obese women (BMI of 25 kg/m(2) or above) had a GDM risk equal to that of normal-weight women, nearly half of GDM cases could be prevented. Public health efforts to reduce prepregnancy BMI by promoting physical activity and healthy eating among women of reproductive age should be intensified.

Full text

Page 1

PercentageofGestationalDiabetesMellitus
AttributabletoOverweightandObesity
Shin Y. Kim, MPH, Lucinda England, MD, MSPH, Hoyt G. Wilson, PhD, Connie Bish, PhD, MPH, Glen A. Satten, PhD, and Patricia Dietz, DrPH
Gestational diabetes mellitus (GDM) is defined
as carbohydrate intolerance leading to hyper-
glycemia with onset or first recognition during
pregnancy.1GDM affects1% to14% of preg-
nancies, depending on the population studied
and the diagnostic tests used.1,2It has been
associated with maternal, fetal, and infant com-
plications, including infant macrosomia and birth
trauma, infant hypoglycemia, cesarean section,
and increased medical costs.3–7Although some
women withdiagnosed GDMwillhavepersistent
abnormal glycemia, most women will revert to
normal carbohydrate metabolism after delivery.8
However, women with a history of GDM remain
at increased risk of developing type 2 diabetes
mellitus in the future.9GDM and type 2 diabetes
share many common risk factors, including
overweight and obesity, and GDM is considered
by many to be a precursor of type 2 diabetes.10
In addition, evidence suggests that the in-
cidence of GDM increased in the 1990s.1 1,12
This rise, whichwasconcurrentwiththe growing
prevalence of prepregnancy obesity (a 69.3%
increase between1993–1994 and 2002–
2003)13and increases in type 2 diabetes in the
general population (a 48.8% increase from1994
through 2002),14was independent of other risk
factors such as maternal age and parity.13Al-
though GDM risk increases substantially with
increasing prepregnancy body mass index (BMI;
defined as weight in kilograms divided by height
in meters squared),15the percentage of GDM
specifically attributable to overweight and obe-
sity is currently unknown.
Population-based risk estimates are needed
to calculate the percentage of GDM cases that
could potentially be prevented if all women
who are overweight or obese had a GDM risk
equivalent to that of women of normal weight.
We sought to calculate the percentage of
pregnancies affected by GDM and the per-
centage of GDM attributable to overweight and
obesity as a means of better understanding the
potential effects of weight management on
GDM prevalence.
METHODS
We analyzed data from the Pregnancy Risk
Assessment Monitoring System (PRAMS), an
ongoing population-based surveillance system
that collects information on self-reported ma-
ternal characteristics before, during, and after
pregnancy in participating states. Each month,
a stratified systematic sample of approximately
150 mothers is selected from the birth certifi-
cate records of each state. To participate in
PRAMS, women must be state residents who
have recently delivered a live-born infant,
typically in the preceding 3 or 4 months.
A self-administered,14-page questionnaire is
mailed to each eligible mother. If the mother
fails to respond, a second or third questionnaire
is sent to her. If there is no response to these
additional mailings, attempts are made to reach
the mother for a telephone interview. Each
mother’s self-reported survey data are linked
back to her child’s birth certificate record; only
selected birth certificate variables are included
in the final PRAMS data set. Currently, 37
states, New York City, and the Yankton Sioux
Tribe in South Dakota participate in PRAMS.
We selected states that had an annual
weighted PRAMS response rate of 70% or
higher and had implemented the 2003 revised
US birth certificate, the latter because this
version of the birth certificate includes infor-
mation on GDM separate from preexisting
diabetes and does not combine the 2 condi-
tions. Our study sample included 23904
women who were surveyed in 2004, 2005, or
2006 in 7 states: Florida (2004–2005),
Nebraska (2005–2006), New York (excluding
New York City; 2004–2006), Ohio (2006),
South Carolina (2004–2006), Vermont
(2004–2006), and Washington (2004–
2006). We increased the sampling weights of
records from states with fewer years of data so
that the sum of each state’s weights represented
the same number of years.
Maternal Characteristics
We used birth certificate information to
analyze data on maternal characteristics such
Objectives. We calculated the percentage of gestational diabetes mellitus
(GDM) attributable to overweight and obesity.
Methods. We analyzed 2004 through 2006 data from 7 states using the
Pregnancy Risk Assessment Monitoring System linked to revised 2003 birth
certificate information. We used logistic regression to estimate the magnitude of
the association between prepregnancy body mass index (BMI) and GDM and
calculated the percentage of GDM attributable to overweight and obesity.
Results. GDM prevalence rates by BMI category were as follows: underweight
(13–18.4 kg/m2),0.7%; normalweight (18.5–24.9 kg/m2), 2.3%; overweight (25–29.9
kg/m2), 4.8%; obese (30–34.9 kg/m2), 5.5%; and extremely obese (35–64.9 kg/m2),
11.5%. Percentages of GDM attributable to overweight, obesity, and extreme
obesity were 15.4% (95% confidence interval [CI]=8.6, 22.2), 9.7% (95% CI=5.2,
14.3), and 21.1% (CI=15.2, 26.9), respectively. The overall population-attributable
fraction was 46.2% (95% CI=36.1, 56.3).
Conclusions. If all overweight and obese women (BMI of 25 kg/m2or above)
had a GDM risk equal to that of normal-weight women, nearly half of GDM cases
could be prevented. Public health efforts to reduce prepregnancy BMI by
promoting physical activity and healthy eating among women of reproductive
age should be intensified. (Am J Public Health. Published online ahead of print
April 15, 2010: e1–e6. doi:10.2105/AJPH.2009.172890)
RESEARCH AND PRACTICE
Published online ahead of print April 15, 2010 | American Journal of Public Health
Kim et al. | Peer Reviewed | Research and Practice | e1

http://ajph.aphapublications.org/cgi/doi/10.2105/AJPH.2009.172890The latest version is at
Published Ahead of Print on April 15, 2010, as 10.2105/AJPH.2009.172890

Page 2

as age, race/ethnicity, educational level, marital
status, parity, smoking status, prepregnancy
weight and height, and GDM diagnosis. We did
not obtain information on preexisting diabetes
from the birth certificates because this infor-
mation is not part of the final PRAMS data set.
Maternal race/ethnicity was self-reported and
categorized as Hispanic, non-Hispanic White,
non-Hispanic Black, or other (Alaska Natives,
American Indians, Asian Americans, or indi-
viduals of other racial/ethnic backgrounds). In
Vermont, all women with the exception of non-
Hispanic Whites are combined as ‘‘other’’ be-
cause of the small number of women in other
racial/ethnic groups residing in Vermont;
thus, we included only information for non-
Hispanic White women in our analyses of
Vermont data.
Data analyzed from the PRAMS question-
naire included prenatal enrollment in the Spe-
cial Supplemental Nutrition Program for
Women, Infants, and Children (WIC); Medicaid
status; smoking status; prepregnancy weight
and height; and whether the woman had
preexisting diabetes. Women who indicated on
either the birth certificate or the PRAMS
questionnaire that they had smoked during the
final 3 months of their pregnancy were classi-
fied as smokers.
When available, we used prepregnancy
weight and height data from birth certificates
(such data were available in 92% of cases) to
calculate prepregnancy BMI; we used infor-
mation from PRAMS if birth certificate data
were missing. Our reason for using birth
certificate data in these calculations was that, in
PRAMS, prepregnancy height and weight are
collected typically 3 or 4 months after delivery.
We excluded BMI values below the 0.01st
percentile and above the 99.99th percentile
(less than13 kg/m2and more than 64.9 kg/m2,
respectively). Thus, on the basis of our exclu-
sions and the BMI categories defined by the
National Heart, Lung, and Blood Institute, we
defined underweight as 13 to 18.4 kg/m2,
normal weight as 18.5 to 24.9 kg/m2, over-
weight as 25 to 29.9 kg/m2, obese as 30 to
34.9 kg/m2,and extremely obeseas 35 to 64.9
kg/m2.
Data Analysis
We excluded the following women from the
analysis: those who reported on PRAMS that
they had preexisting diabetes (these women by
definition could not develop GDM), those for
whom information about prepregnancy weight
or height or information about GDM was
missing, and those in Vermont who were of
a racial/ethnic group other than non-Hispanic
White. After these exclusions, data on 22767
women (95% of the total sample) were avail-
able for the final analysis.
We estimated the prevalence and standard
errors of various demographic characteristics
by state and calculated the prevalence of GDM
by category for each characteristic. Using sam-
ple-weighted multivariate logistic regression,
we estimated the independent contributions of
BMI to GDM risk via aggregated data from the
7 states in our study. We assessed potential
confounding for each demographic character-
istic and considered a covariate to be potential
confounder if its inclusion in the regression
models changed the unadjusted odds ratio (OR)
by 10% or more. Using the logistic regression
results, we computed relative risks (RRs) and
their confidence intervals (CIs) according to
methods described by Flanders and Rhodes.16
Finally, employing methods described by
Graubard and Fears,17we used the logistic
regression results to estimate the population-
attributable fraction (PAF) and its CI for each
overweight or obese BMI category and for all
overweight and obese categories combined. We
interpreted each PAF estimate to be the reduc-
tion in disease prevalence that would be
expected to occur if all women in the overweight
or obese BMI categories had a GDM risk equiv-
alent to that of women in the normal BMI
category, assuming that the risk for GDM among
those with a low or normal BMI remained
unchanged.18
We also used locally weighted scatterplot
smoothing (LOESS) to estimate the probability
of GDM as a continuous function of BMI. In this
method, the smoothed value of the function at
each data point is computed from a weighted
regression fit to neighboring points. Neighbor-
ing points that are closer to the point at which
the smoothing occurs are weighted more
heavily.19,20
In all of the analyses other than those in-
volving LOESS, the data were weighted to
adjust for survey design, months of data sam-
pling for the state, nonresponse, and the extent
to which some individuals in the target
population were not included in the sampled
population. We used Sudaan version 10.0
(Research Triangle Institute, Research Triangle
Park, NC) to fit the logistic models and compute
ORs and their standard errors, S-Plus version
7.020to perform the LOESS analyses, and SAS
version 9.1(SAS Institute, Cary, NC) for all other
computations.
RESULTS
Demographic characteristics of the PRAMS
population in each of the 7 states are described
in Table 1. The overall GDM prevalence was
4.0% (SE=0.2), with a range from 3.1%
(SE=0.4) in Florida to 5.0% (SE=0.7) in Ohio
(Table 1). For all states combined, GDM prev-
alence estimates by BMI category were as
follows: underweight, 0.7% (SE=0.3); normal
weight, 2.3% (SE=0.3); overweight, 4.8%
(SE=0.5); obese, 5.5% (SE=0.7); and ex-
tremely obese, 11.5% (SE=1.3; Table 2).
In addition, we found that 0.9% (SE=0.4) of
women with gestational diabetes were under-
weight, 28.4% (SE=2.8) were of normal
weight, 28.5% (SE=2.7) were overweight,
16.2% (SE=2.1) were obese, and 26.0%
(SE=2.7) were extremely obese. The proba-
bility of GDM increased with increasing BMI,
although the confidence bands became quite
wide when BMIsexceeded 40 kg/m2(Figure1).
There was no clear BMI threshold below which
a dose–response relationship was not evident.
Because none of the potential confounders
changed ORs by 10% or more, we included in
our adjusted model covariates that have been
found in the literature to be associated with
both the exposure (BMI) and the outcome
(GDM). When the normal-weight BMI category
was used as a reference group, we found that
the unadjusted RRs of developing GDM were
0.3 (95% CI=0.1, 0.7) for underweight
women, 2.1 (95% CI=1.6, 2.9) for overweight
women, 2.4 (95% CI=1.7, 3.4) for obese
women, and 5.0 (95% CI=3.6, 6.9) for ex-
tremely obese women. RRs did not change
after adjustment for maternal age, race/eth-
nicity, marital status, and parity (Table 3).
The overall adjusted PAF due to overweight
and obesity was 46.2% (95% CI=36.1, 56.3;
Table 2). Adjusted percentages of GDM in-
dividually attributable to overweight, obesity,
and extreme obesity were 15.4% (95%
RESEARCH AND PRACTICE
e2 | Research and Practice | Peer Reviewed | Kim et al.
American Journal of Public Health | Published online ahead of print April 15, 2010

Page 3

TABLE 1—Sample-Weighted Demographic Characteristics: Pregnancy Risk Assessment Monitoring System, 7 US States, 2004–2006
Maternal Characteristic
Total
(2004–2006),
% (SE)
Florida
(2004–2005),
% (SE)
Nebraska
(2005–2006),
% (SE)
New Yorka
(2004–2006),
% (SE)
Ohio
(2006),
% (SE)
South Carolina
(2004–2006),
% (SE)
Vermont
(2004–2006),
% (SE)
Washington
(2004–2006),
% (SE)
Age, y
<20 10.4 (0.3)11.0 (0.1)7.4 (0.5)7.5 (0.7) 12.0 (1.2) 13.3 (0.9) 7.1 (0.5)9.0 (0.6)
20–34
‡35
Education, y
74.7 (0.5)74.9 (0.8) 80.2 (0.8) 70.7 (1.1)75.5 (1.5) 76.2 (1.1) 75.7 (0.8)76.2 (0.9)
14.9 (0.4)14.1 (0.8) 12.4 (0.7)21.7 (1.0) 12.5 (1.1)10.5 (0.8)17.1 (0.7) 14.8 (0.7)
<1218.2 (0.5) 20.4 (0.8)14.9 (0.6)16.2 (1.0)15.7 (1.3) 23.0 (1.2) 9.3 (0.6)17.4 (0.7)
1227.9 (0.6) 32.4 (1.0) 21.1 (0.9)23.1 (1.0) 28.4 (1.6)26.0 (1.2) 32.1 (0.9) 24.5 (0.9)
>1253.9 (0.6) 47.2 (1.1)64.1 (0.9)60.7 (1.2)55.9 (1.7) 50.9 (1.3)58.6 (0.9) 58.1 (1.0)
Race/ethnicityb
Hispanic 15.6 (0.4)27.1 (1.0) 14.0 (0.1) 13.1 (0.9)3.3 (0.7) 7.4 (0.7) NA17.5 (0.1)
Non-Hispanic White 63.9 (0.5)48.2 (1.0) 75.5 (0.2) 74.5 (1.1)77.3 (1.0) 58.3 (1.3)100.0 64.9 (0.4)
Non-Hispanic Black15.2 (0.2)20.1 (0.5) 5.5 (0.1)7.9 (0.7)14.8 (0.2) 31.9 (1.3)NA 3.2 (0.1)
Other 5.3 (0.3)4.7 (0.5)4.9 (0.2)4.5 (0.5) 4.6 (0.8)2.4 (0.4) NA14.4 (0.4)
Married
Yes 61.9 (0.6)58.2 (1.0)71.1 (0.9) 67.0 (1.2)60.1 (1.7)56.9 (1.3)68.8 (0.9) 69.6 (0.9)
No38.1 (0.6)41.8 (1.0) 28.9 (0.9) 33.0 (1.2)39.9 (1.7) 43.1 (1.3)31.2 (0.9)30.4 (0.9)
Medicaid recipient
Yes 45.9 (0.6)51.2 (1.1)40.9 (1.0)35.5 (1.2) 41.7 (1.7)58.1 (1.3)42.1 (0.9) 49.2 (1.0)
No 54.1 (0.6)48.8 (1.1) 59.1 (1.0)64.5 (1.2)58.3 (1.7)41.9 (1.3) 57.9 (0.9) 50.8 (1.0)
WIC recipient
Yes55.7 (0.6)52.1 (1.1) 63.3 (0.9) 62.4 (1.2) 58.1 (1.7)45.0 (1.3) 59.2 (0.9)55.9 (1.0)
No 44.3 (0.6)47.9 (1.1)36.7 (0.9)37.6 (1.2)41.9 (1.7)55.0 (1.3)40.8 (0.9) 44.1 (1.0)
Parity
0 42.1 (0.6)43.3 (1.1)36.7 (1.0) 41.6 (1.2) 41.0 (1.7) 42.4 (1.3) 44.8 (0.9)42.9 (1.0)
1–247.9 (0.6) 46.8 (1.1)50.8 (1.1) 49.1 (1.2)47.3 (1.8) 50.3 (1.3)48.2 (0.9)47.0 (1.1)
>210.0 (0.4) 9.8 (0.7)12.5 (0.7) 9.3 (0.7)11.7 (1.1) 7.3 (0.7)6.9 (0.5)10.1 (0.6)
Smoking status
Smokerc
13.4 (0.4)9.1 (0.7)15.8 (0.8)15.5 (0.9) 17.4 (1.3) 15.6 (1.0)18.3 (0.7)11.5 (0.7)
Nonsmoker
BMI category, kg/m2
86.6 (0.4) 90.9 (0.7)84.2 (0.8)84.5 (0.9) 82.6 (1.3) 84.4 (1.0) 81.7 (0.7) 88.5 (0.7)
Underweight (13–18.4) 4.8 (0.3)6.1 (0.5)4.2 (0.4) 3.3 (0.4) 5.1 (0.8)4.7 (0.6) 3.2 (0.3)3.0 (0.3)
Normal weight (18.5–24.9)50.3 (0.6) 53.8 (1.1) 50.3 (1.1)50.4 (1.2) 48.0 (1.7)44.1 (1.3) 50.5 (0.9)49.0 (1.0)
Overweight (25–29.9) 23.8 (0.5) 22.1 (0.9) 24.5 (0.9)24.3 (1.0) 24.2 (1.5)25.3 (1.2)24.1 (0.8) 26.4 (0.9)
Obese (30–34.9) 11.9 (0.4)10.5 (0.7) 13.0 (0.7)12.3 (0.8) 12.3 (1.2)14.7 (0.9) 12.2 (0.6)12.0 (0.7)
Extremely obese (35–64.9) 9.2 (0.4)7.6 (0.6) 7.9 (0.6)9.7 (0.7)10.4 (1.0) 11.2 (0.8)10.1 (0.6)9.5 (0.6)
GDM
Yes4.0 (0.2)3.1 (0.4) 4.0 (0.4) 3.8 (0.4)5.0 (0.7)4.9 (0.6) 3.4 (0.3) 4.8 (0.4)
No 96.0 (0.2) 96.9 (0.4) 96.0 (0.4) 96.2 (0.4)95.0 (0.7)95.1 (0.6)96.6 (0.3)95.2 (0.4)
Note. BMI =body mass index; GDM=gestational diabetes mellitus; NA=not applicable; WIC=Special Supplemental Nutrition Program for Women, Infants, and Children. The total sample size was
N=22767;for Florida, n=4053;for Nebraska,n=3411; for New York, n=2744;for Ohio, n=1497;for South Carolina, n=3848; for Vermont, n=3097;for Washington, n=4117. Overalland state
sample sizes are unweighted.
aExcludes New York City.
bOnly non-Hispanic White women were included in analyses of Vermont data.
cDefined as smoking during the final 3 months of pregnancy (self-reported in PRAMS) or during the third trimester (reported on birth certificates).
RESEARCH AND PRACTICE
Published online ahead of print April 15, 2010 | American Journal of Public Health
Kim et al. | Peer Reviewed | Research and Practice | e3

Page 4

CI=8.6, 22.2), 9.7% (95% CI=5.2,14.3), and
21.1% (95% CI=15.2, 26.9), respectively.
DISCUSSION
Our results show an increased risk of GDM
associated with increasing BMI. The overall
PAFdue to overweight and obesity was 46.2%.
In other words, if all women with BMIs of 25 or
above had a GDM risk equal to that of women
in the normal BMI category, nearly half of
GDM cases potentially could be prevented. In
addition, we found that more than 70% of all
women with GDM had a BMI of 25 or higher,
whereas approximately a quarter had a normal
BMI.
High maternal BMIs have been consistently
associated with an increased risk of GDM in the
literature.15,21In a meta-analysis estimating the
magnitude of GDM risk among women with high
prepregnancy BMIs, Chu et al. found that GDM
risk increases substantially with increasing pre-
pregnancy BMI.15Their results showed that,
relative to women of normal weight, the
unadjusted ORs for developing GDM were 2.14
(95% CI=1.82, 2.53) among overweight
women, 3.56 (95% CI=3.05, 4.21) among
obese women, and 8.56 (95% CI=5.07,16.04)
among extremely obese women.14Moreover,
similar to our findings, a dose–response rela-
tionship between increasing BMI and type 2
diabetes has been described in the general
population, even within the normal BMI cate-
gory.22–24
Although women with GDM are at increased
risk for type 2 diabetes,9evidence strongly
suggests that type 2 diabetes is preventable in
this population. Several randomized trials have
demonstrated that weight loss and increased
physical activity reduce the risk of type 2 di-
abetes in individuals at high risk, including
women with a history of GDM.25–27Similarly,
evidence suggests that GDM risk is reduced in
women who engage in high levels of physical
activity28and consume high-fiber diets.29
Therefore, to the extent that prepregnancy
overweight and obesity cause GDM, reducing
prepregnancy weight in these women should
TABLE 2—Sample-Weighted
Gestational Diabetes Mellitus (GDM)
Prevalence, by Demographic
Characteristics: Pregnancy Risk
Assessment Monitoring System, 7 US
States, 2004–2006
Maternal Characteristic
GDM Prevalence,
% (SE)
Overall4.0 (0.2)
Age, y
<201.0 (0.3)
20–34
‡35
Education, y
3.6 (0.3)
8.4 (0.9)
<122.8 (0.4)
123.9 (0.5)
>12 4.5 (0.3)
Race/ethnicitya
Hispanic4.4 (0.6)
Non-Hispanic White4.0 (0.3)
Non-Hispanic Black3.6 (0.4)
Other 5.4 (1.0)
Married
Yes4.6 (0.3)
No 3.1 (0.3)
Medicaid recipient
Yes3.4 (0.3)
No 4.6 (0.4)
WIC recipient
Yes4.3 (0.4)
No 3.7 (0.3)
Parity
03.0 (0.3)
1–2 4.8 (0.4)
>2 5.0 (0.9)
Smoking status
Smokerb
3.7 (0.6)
Nonsmoker
BMI category, kg/m2
4.1 (0.3)
Underweight (13–18.4)0.7 (0.3)
Normal weight (18.5–24.9) 2.3 (0.3)
Overweight (25–29.9)4.8 (0.5)
Obese (30–34.9) 5.5 (0.7)
Extremely obese (35–64.9)11.5 (1.4)
Note. BMI =body mass index; WIC=Special Supple-
mental Nutrition Program for Women, Infants, and
Children.
aOnly non-Hispanic White women were included in
analyses of Vermont data.
bDefined as smoking during the final 3 months of
pregnancy (self-reported in PRAMS) or during the third
trimester (reported on birth certificates).
Note. BMI is defined as weight in kilograms divided by height in meters squared.
FIGURE 1—Unweighted probability of gestational diabetes mellitus (GDM), by mother’s
prepregnancy body mass index (BMI): Pregnancy Risk Assessment Monitoring System, 7 US
states, 2004–2006.
RESEARCH AND PRACTICE
e4 | Research and Practice | Peer Reviewed | Kim et al.
American Journal of Public Health | Published online ahead of print April 15, 2010

Page 5

reduce diabetes-related adverse pregnancy out-
comes. Sustaining this weight loss beyond preg-
nancy should reduce women’s future risk for
type 2 diabetes.30
Limitations
To our knowledge, our study provides the
first population-based estimates of the contri-
bution of overweight and obesity to GDM.
However, the study involves some limitations.
Prepregnancy weight isself-reported in PRAMS
and is likely to be self-reported on birth
certificates; estimates of obesity prevalence
based on self-reported weight tend to be lower
than those based on measured data.31There-
fore, we may have underestimated the preva-
lence of prepregnancy overweight and obesity,
which could have resulted in an underestimation
of the contribution of overweight and obesity to
the PAF assuming that the BMI misclassification
was nondifferential.
In addition, because PRAMS collects data
only on women who have delivered a live-born
infant, our analysis did not include women
whose pregnancies ended in a miscarriage, fetal
death, or stillbirth. However, GDM typically
develops in the late second or early third
trimester of pregnancy, and only a small pro-
portion of women (6.3 per 1000 women)
experience fetal loss after 20 weeks. Therefore,
our estimates of GDM prevalence should not
have been substantially affected by the re-
striction of our analysis to live births.
Our data also may not be generalizable to
states not included in our analyses. For exam-
ple, GDM prevalence has been shown to vary
by racial/ethnic group. Nineteen percent of all
US live births were represented in our analyses,
and non-Hispanic White women were over-
represented in our study (64% of the mothers
included in our analysis were non-Hispanic
White, as compared with 55% in the total US
population of mothers delivering a live-born
infant).32Moreover,prepregnancyobesityvaries
by state, ranging from13.9% to 25.1% in 26
PRAMS states that represented 47% of all live
births in the United States during 2004–
2005.33This variation, in part, may explain the
differences in the prevalence of GDM across
states. Therefore, a national estimate of the
contribution of prepregnancy weight to GDM
risk may be different from our estimate.
Furthermore, although the association be-
tween BMI and GDM risk appears to vary
according to race/ethnicity, we were not able
to calculate PAFs for specific racial/ethnic
groups as a result of our small poststratification
sample sizes. Large databases are needed to
conduct more in-depth analyses of BMI and
GDM in specific racial/ethnic groups. In addi-
tion, our analysis did not account for potential
confounders such as physical activity, diet, and
genetics because PRAMS does not collect in-
formation on these indicators.
Finally, the quality of the GDM variable on
the revised 2003 birth certificate has not been
validated in the published literature. However,
studies in states that have amended their1989
birth certificate to differentiate between pre-
existing diabetes and GDM have consistently
shown specificities above 98% and sensitivities
ranging from 46% to 83% in identifying
GDM.34With the implementation of the revised
2003 birth certificate and as certifiers have
become more familiar with the separate cate-
gories, accuracy of GDM identification most
likely has improved.35
Conclusions
A large percentage of GDM cases are po-
tentially attributable to overweight and obesity
and could be avoided by preventing these
conditions. Data such as ours can help public
health officials estimate the potential effects of
prevention interventions on GDM prevalence
rates. Lifestyle interventions designed to re-
duce BMIs have the potential to lower GDM
risk. Therefore, public health efforts to promote
recommended levels of physical activity and
healthy eating habits among women of repro-
ductive age should be intensified. j
About the Authors
The authors are with the Division of Reproductive Health,
National Center for Chronic Disease Prevention and Health
Promotion, Centers for Disease Control and Prevention,
Atlanta, GA.
Correspondence should be sent to Shin Y. Kim, MPH,
4770 Buford Hwy NE, MS K-23, Atlanta, GA (e-mail:
skim1@cdc.gov). Reprints can be ordered at http://www.
ajph.org by clicking the ‘‘Reprints/Eprints’’ link.
This article was accepted October 7, 2009.
Note. The findings and conclusions in this article are
those of the authors and do not necessarily represent the
official position of the Centers for Disease Control and
Prevention.
Contributors
S.Y. Kim and L. England originated the study. S.Y. Kim
analyzed the data, led the writing, and supervised all
aspects of study implementation. L. England, C. Bish,
G.A. Satten, and P. Dietz synthesized the analysis. H.G.
Wilson analyzed the data and synthesized the analysis.
All of the authors helped conceptualize ideas, interpret
findings, and review drafts of the article.
Acknowledgments
We thank Brian Morrow for his technical expertise and
consultation. Data from the Pregnancy Risk Assessment
Monitoring System (PRAMS) included in this study were
collected at the state level by the following state working
group collaborators and their staff: Albert Woolbright
(Alabama), KathyPerham-Hester(Alaska), MaryMcGehee
(Arkansas), Alyson Shupe (Colorado), Charlon Kroelinger
(Delaware), Jamie Fairclough (Florida), Carol Hoban
TABLE 3—Sample-Weighted Gestational Diabetes Mellitus (GDM) Risk Data:
Pregnancy Risk Assessment Monitoring System, 7 US States, 2004–2006
RR (95% CI) PAFb(95% CI)
Unadjusted Adjusteda
Unadjusted Adjusteda
Underweight (13–18.4 kg/m2)
Normal weight (18.5–24.9 kg/m2; Ref)
Overweight (25–29.9 kg/m2)
Obese (30–34.9 kg/m2)
Extremely obese (35–64.9 kg/m2)
0.32 (0.14, 0.75) 0.38 (0.16, 0.89)... ...
1.001.00 ......
2.12 (1.55, 2.90) 2.17 (1.58, 2.97) 15.1 (8.3, 21.9)15.4 (8.6, 22.2)
2.41 (1.71, 3.41)2.51 (1.76, 3.56) 9.5 (4.9, 14.0) 9.7 (5.2, 14.3)
5.02 (3.64, 6.93) 5.03 (3.64, 6.95)20.8 (15.0, 26.6)21.1 (15.2, 26.9)
All BMI categories above normal weight...... 45.4 (35.3, 55.5)46.2 (36.1, 56.3)
Note. BMI=body mass index; CI=confidence interval; PAF=population-attributable fraction; RR=relative risk. The sample
size used for unadjusted RR was n=22767; for adjusted RR, n=22200.
aAdjusted models included covariates for maternal age, race/ethnicity, marital status, and parity.
bWe interpreted each PAF estimate to be the reduction in disease prevalence that would be expected to occur if all women in
the overweight or obese BMI categories had a GDM risk equivalent to that of women in the normal BMI category, assuming
that the risk for GDM among those with a low or normal BMI remained unchanged.
RESEARCH AND PRACTICE
Published online ahead of print April 15, 2010 | American Journal of Public Health
Kim et al. | Peer Reviewed | Research and Practice | e5

Page 6

(Georgia), Mark Eshima (Hawaii), Theresa Sandidge
(Illinois), Joan Wightkin (Louisiana), Kim Haggan
(Maine), Diana Cheng (Maryland), Hafsatou Diop
(Massachusetts), Violanda Grigorescu (Michigan), Jan
Jernell (Minnesota), Marilyn Jones (Mississippi), Venkata
Garikapaty (Missouri), JoAnn Dotson (Montana),
Brenda Coufal (Nebraska), Lakota Kruse (New Jersey),
Eirian Coronado (New Mexico), Anne Radigan-Garcia
(New York State), Candace Mulready-Ward (New York
City), Paul Buescher (North Carolina), Sandra Anseth
(North Dakota), Connie Geidenberger (Ohio), Alicia
Lincoln (Oklahoma), Kenneth Rosenberg (Oregon),
Kenneth Huling (Pennsylvania), Sam Viner-Brown
(Rhode Island), Mike Smith (South Carolina), Christine
Rinki (South Dakota), Kate Sullivan (Texas), David Law
(Tennessee), Laurie Baksh (Utah), Peggy Brozicevic
(Vermont), Marilyn Wenner (Virginia), Linda Lohdefinck
(Washington), Melissa Baker (West Virginia), Katherine
Kvale (Wisconsin), and Angi Crotsenberg (Wyoming).
Data were also collected by the Centers for Disease
Control and Prevention PRAMS team.
Human Participant Protection
The Pregnancy Risk Assessment Monitoring System
protocol was approved by the Centers for Disease
Control and Prevention’s institutional review board; the
analysis plan was approved in the participating states.
Informed consent was obtained from all participants via
either mail or telephone.
References
1.American College of Obstetricians and Gynecolo-
gists. Clinical management guidelines for obstetrician-
gynecologists: gestational diabetes. Obstet Gynecol.
2001;98(3):525–538.
2.
diabetes in pregnancy. Obstet Gynecol Clin North Am.
2007;34(2):173–199.
Hunt KJ, Schuller KL. The increasing prevalence of
3.
Pregnancy outcomes in women with gestational diabetes
compared with the general obstetric population. Obstet
Gynecol. 1997;90(6):869–873.
Casey BM, Lucas MJ, Mcintire DD, Leveno KJ.
4.
Gestational diabetes mellitus: prevalence, risk factors,
maternal and infant outcomes. Int J Gynaecol Obstet.
2001;75(3):221–228.
Xiong X, Saunders LD, Wang FL, Demianczuk NN.
5.
Period of gestational diabetes mellitus diagnosis and
maternal and fetal morbidity. Acta Obstet Gynecol Scand.
2005;84(7):622–627.
Barahona MJ, Sucunza N, Garcia-Patterson A, et al.
6.
Westergaard JG, Beck-Nielsen H. Maternal and perinatal
outcomes in 143 Danish women with gestational di-
abetes mellitus and 143 controls with a similar risk
profile. Diabet Med. 2000;17(4):281–286.
Jensen DM, Sorensen B, Feilberg-Jorgensen N,
7.
gestational diabetes mellitus in the United States in 2007.
Popul Health Manag. 2009;12(3):165–174.
Chen Y, Quick WW, Yang W, et al. Cost of
8.
factors, diagnosis, management, and implications. Semin
Perinatol. 2007;31(3):176–184.
Bottalico JN. Recurrent gestational diabetes: risk
9.
and the incidence of type 2 diabetes: a systematic review.
Diabetes Care. 2002;25(10):1862–1868.
Kim C, Newton KM, Knopp RH. Gestational diabetes
10. England LJ, Dietz PM, Njoroge T, et al. Preventing
type 2 diabetes: public health implications for women
with a history of gestational diabetes mellitus. Am J Obstet
Gynecol. 2009;200(4):e1–e8.
11. Lawrence JM, Contreras R, Chen W, Sacks DA.
Trends in the prevalence of preexisting diabetes and
gestational diabetes mellitus among a racially/ethnically
diverse population of pregnant women, 1999–2005.
Diabetes Care. 2008;31(5):899–904.
12. Ferrara A, Kahn HS, Quesenberry CP, Riley C,
Hedderson MM. An increase in the incidence of gesta-
tional diabetes mellitus: northern California, 1991–
2000. Obstet Gynecol. 2004;103(3):526–533.
13. Kim SY, Dietz PM, England L, Morrow B, Callaghan
WM. Trends in pre-pregnancy obesity in 9 states, 1993–
2003. Obesity (Silver Spring). 2007;15(4):986–993.
14. Dabelea D, Snell-Bergeon JK, Hartsfield CL, Bischoff
KJ, Hamman RF, McDuffie RS. Increasing prevalence of
gestational diabetes mellitus (GDM) over time and by
birth cohort: Kaiser Permanente of Colorado GDM
Screening Program. Diabetes Care. 2005;28(3):579–
584.
15. Chu SY, Callaghan WM, Kim SY, et al. Maternal
obesity and risk of gestational diabetes mellitus. Diabetes
Care. 2007;30(8):2070–2076.
16. Flanders WD, Rhodes PH. Large sample confidence
intervals for regression standardized risks, risk ratios, and
risk differences. J Chronic Dis. 1987;40(7):697–704.
17. Graubard BI, Fears TR. Standard errors for attrib-
utable risk for simple and complex sample designs.
Biometrics. 2005;61(3):847–855.
18. Levine BJ. The other causality question: estimating
attributable fractions for obesity as a cause of mortality.
Int J Obes. 2008;32(suppl 3):S4–S7.
19. Hastie T, Tibshirani R. Generalized Additive Models.
London, England: Chapman & Hall; 1990.
20. S-Plus Version 7, Guide to Statistics. Seattle, WA:
Insightful Corp; 2005.
21. Torloni MR, Betran AP, Horta BL, et al. Prepreg-
nancy BMI and the risk of gestational diabetes: a system-
atic review of the literature with meta-analysis. Obes Rev.
2009;10(2):194–203.
22. Rana JS, Li TY, Manson JE, Hu FB. Adiposity
compared with physical inactivity and risk of type 2
diabetes in women. Diabetes Care. 2007;30(1):53–58.
23. Ford ES, Williamson DF, Liu S. Weight change and
diabetes incidence: findings from a national cohort of US
adults. Am J Epidemiol. 1997;146(3):214–222.
24. Shai I, Jiang R, Manson JE, et al. Ethnicity, obesity,
and risk of type 2 diabetes in women: a 20-year follow-up
study. Diabetes Care. 2006;29(7):1585–1590.
25. Knowler WC, Barrett-Connor E, Fowler SE, et al.
Reduction in the incidence of type 2 diabetes with
lifestyle intervention or metformin. N Engl J Med. 2002;
346(6):393–403.
26. Orchard TJ, Temprosa M, Goldberg R, et al. The
effect of metformin and intensive lifestyle intervention on
the metabolic syndrome: the Diabetes Prevention Pro-
gram randomized trial. Ann Intern Med. 2005;
142(8):611–619.
27. Ratner RE. Prevention of type 2 diabetes in women
with previous gestational diabetes. Diabetes Care.
2007;30(suppl 2):S242–S245.
28. Zhang C, Solomon CG, Manson JE, Hu FB. A
prospective study of pregravid physical activity and
sedentary behaviors in relation to the risk for gestational
diabetes mellitus. Arch Intern Med. 2006;166(5):543–
548.
29. Zhang C, Liu S, Solomon CG, Hu FB. Dietary fiber
intake, dietary glycemic load, and the risk for gestational
diabetes mellitus. Diabetes Care. 2006;29(10):2223–
2230.
30. Artal R, Catanzaro RB, Gavard JA, Mostello DJ,
Friganza JC. A lifestyle intervention of weight-gain re-
striction: diet and exercise in obese women with gesta-
tional diabetes mellitus. Appl Physiol Nutr Metab. 2007;
32(3):596–601.
31. Flegal KM, Carroll MD, Ogden CL, Johnson CL.
Prevalence and trends in obesity among US adults,
1999–2000. JAMA. 2002;288(14):1723–1727.
32. Martin JA, HamiltonBE, SuttonPD, et al. Births: final
data for 2005. Natl Vital Stat Rep. 2007;56(6):1–103.
33. Chu SY, Kim SY, Bish CL. Prepregnancy obesity
prevalence in the United States, 2004–2005. Matern
Child Health J. 2009;13(5):614–620.
34. Devlin HM, Desai J, Walaszek A. Reviewing per-
formance of birth certificate and hospital discharge data
to identify births complicated by maternal diabetes.
Matern Child Health J. 2009;13(5):660–666.
35. Menacker F, Martin JA. Expanded health data from
the new birth certificate, 2005. Natl Vital Stat Rep.
2008;56(13):1–24.
RESEARCH AND PRACTICE
e6 | Research and Practice | Peer Reviewed | Kim et al.
American Journal of Public Health | Published online ahead of print April 15, 2010