Maternal and Child Health Journal

, Volume 15, Issue 6, pp 772–781

The Association Between Pre-pregnancy BMI and Preterm Delivery in a Diverse Southern California Population of Working Women

  • Jessica Lang Kosa
  • Sylvia Guendelman
  • Michelle Pearl
  • Steve Graham
  • Barbara Abrams
  • Martin Kharrazi
Open AccessArticle

DOI: 10.1007/s10995-010-0633-4

Cite this article as:
Kosa, J.L., Guendelman, S., Pearl, M. et al. Matern Child Health J (2011) 15: 772. doi:10.1007/s10995-010-0633-4

Abstract

Whereas preterm birth has consistently been associated with low maternal pre-pregnancy weight, the relationship with high pre-pregnancy weight has been inconsistent. We quantified the pre-pregnancy BMI—preterm delivery (PTD) relationship using traditional BMI categories (underweight, normal weight, overweight and obese) as well as continuous BMI. Eligible women participated in California’s statewide prenatal screening program, worked during pregnancy, and delivered a live singleton birth in Southern California in 2002–2003. The final analytic sample included 354 cases delivering at <37 weeks, as identified by clinical estimate of gestational age from screening records, and 710 term normal-birthweight controls. Multivariable logistic regression models using categorical BMI levels and continuous BMI were compared. In categorical analyses, PTD was significantly associated with pre-pregnancy underweight only. Nonparametric local regression revealed a V-shaped relationship between continuous BMI and PTD, with minimum risk at the high end of normal, around 24 kg/m2. The odds ratio (OR) for PTD associated with low BMI within the normal range (19 kg/m2) was 2.84 (95%CI = 1.61–5.01); ORs for higher BMI in the overweight (29 kg/m2) and obese (34 kg/m2) ranges were 1.42 (95%CI = 1.10–1.84) and 2.01 (95% CI = 1.20–3.39) respectively, relative to 24 kg/m2). BMI categories obscured the preterm delivery risk associated with low-normal, overweight, and obese BMI. We found that higher BMI up to around 24 kg/m2 is increasingly protective of preterm delivery, beyond which a higher body mass index becomes detrimental. Current NHLBI/WHO BMI categories may be inadequate for identifying women at higher risk for PTD.

Keywords

ObesityBMIPreterm deliveryPre-conception

Introduction

The weight at which women begin their pregnancies can affect their outcomes and those of their infants. In the US, 26% of non-pregnant women 20–39 years of age are overweight (defined as a body mass index [BMI] of 25–29.9 kg/m2), 29% are obese (BMI 30–39.9 kg/m2) and an additional 8% are morbidly obese (BMI ≥ 40 kg/m2). Such high prevalence of overweight and obesity among childbearing women is a major public health concern given its contribution to pregnancy complications including gestational diabetes, gestational hypertension, pre-eclampsia, fetal macrosomia, and cesarean delivery.

Preterm delivery (PTD), defined as birth at less than 37 weeks gestation, has long been a known consequence of maternal underweight [1], but less than 3% of Americans aged 20–39 are underweight (BMI < 18.5). Pre-pregnancy obesity has also been associated with PTD in some but not all studies [27]. Cnattingius [4] found an increased risk specifically among nulliparous obese women. Weiss et al. [6] and Callaway et al. reported a significantly elevated risk of PTD among the morbidly obese only. Baeten et al. [12] found that obesity had a stronger association with early preterm birth (<32 weeks) than overall preterm birth. By contrast, Sebire [7], and Kumari [8] did not find a significantly increased risk of PTD associated with maternal obesity in their patient populations, and Khashan identified a protective effect of overweight and obesity against PTD [1].

Differences in study populations and varying definitions of both PTD and obesity likely contribute to inconsistent findings. All of the abovementioned studies categorized BMI, most using the NHLBI/WHO definitions of underweight, normal weight, overweight, and obese [9, 10] adopted in 1995 based on mortality studies [11]. Treating pre-pregnancy BMI in this fashion implicitly models the increase in PTD risk as a stepwise function, with risk constant within each category, and then increasing abruptly at the category boundary. This assumption may not adequately approximate the actual shape of the risk distribution. Of particular concern is that the “normal” category used as a reference group represents a wide range of weight-for-height, and any heterogeneity in risk within the reference group can mask an association of pre-pregnancy BMI and outcome.

Two recent studies examined continuous pre-pregnancy BMI and found increased PTD risk with decreasing BMI below 22–26 kg/m2, suggesting differences in PTD risk within the category traditionally defined as “normal weight”[20, 21]. Neither study found a significant increase in PTD associated with higher BMI. However, one study limited preterm births to spontaneous deliveries before 36 weeks, and the other relied upon last menstrual period to determine gestational age and included very few PTD cases with BMI > 25.

Given rising rates of PTD and obesity in the United States, even a small increase in risk would affect a large number of women. Therefore, it is important to understand whether pre-pregnancy overweight and obesity contribute to PTD, and if so, to identify the BMI level at which the risk of PTD begins to increase. This paper investigates these associations in a diverse population of working women who delivered preterm and term, normal-weight infants. We compare results using the NHLBI/WHO categories to models using continuous BMI to determine whether BMI categories are adequately characterizing the exposure–response relationship.

Materials and Methods

Participants were from a nested, population-based case–control study, Juggling Work and Life During Pregnancy, designed to examine the relationship between maternity leave, stress and pregnancy outcomes [22]. Eligible women participated in the California Department of Public Health’s mid-trimester Prenatal Screening Program (PNS) in three Southern California counties (Orange, Imperial and San Diego); delivered live births between July 2002 and December 2003; were at least 18 years old; had a singleton birth without congenital anomalies, and had a US mailing address, yielding 38,278 women with linked data (Fig. 1). Sampled women included all women delivering either preterm (PTD) or low-birthweight (LBW) infants (n = 3,357) according to last menstrual period and birthweight from birth records registered between July 2002 and August 2003; a random sample of controls delivering normal weight at term (≥2,500 g and ≥37 weeks gestation) frequency matched on race and month-of-birth (n = 3,367); and an additional 504 PTD/LBW cases delivering between September and December of 2003, as described previously [12, 13]. Sampled potential participants were mailed an introductory letter and subsequently prescreened by telephone to ascertain that they had worked 20 h or more per week during the first two trimesters of pregnancy or through the date of PNS. Details of the 45-min telephone interview have been described elsewhere [13]. Overall, 1,214 women who worked at least 20 h/week through the date of prenatal screening completed interviews. Because most refusals occurred before work eligibility could be determined, the number of refusals among eligible women was estimated by multiplying the number of refusals (n = 740) by the eligibility rate (45%). Incomplete interviews (n = 109) were treated as refusals to eligible women, yielding an imputed refusal rate of 27% among eligible women contacted for the study ((.45 × 740 + 109)/(.45 × 740 + 109 + 1,214)). Eligibility and refusal rates were the same for women delivering cases and controls. Mean and median interview time after birth was 4.5 months in cases and controls. Bilingual Spanish–English interviewers used CATI (Computer Assisted Telephone Interviewing) software to enter the responses into a database and offered $10 gift cards to participants in return for a completed interview.
https://static-content.springer.com/image/art%3A10.1007%2Fs10995-010-0633-4/MediaObjects/10995_2010_633_Fig1_HTML.gif
Fig. 1

Flow chart of exclusion criteria

For analysis, gestational age at delivery was derived from the best clinical estimate recorded in PNS records. Gestational age estimates from PNS records have been shown to be superior to LMP-based gestational age from birth records [14, 15]. Sixty-two percent of study records had gestational age based on first trimester ultrasound, 34.5% on LMP dates, and 3.5% on physical exam. Women delivering term LBW infants (n = 82) were excluded from this analysis of preterm delivery. Women reporting births with congenital anomalies during the interview (n = 38), or for whom BMI was not available (n = 30) were also excluded, leaving a final sample of 354 PTD cases and 710 term, non-LBW controls.

The study protocol was approved by the Committees for the Protection of Human Subjects at the University of California, Berkeley (No. 2003-5-115) and at the California Health and Human Services Agency (No. 02-10-18).

Measures and Data Collection Instruments

The key exposure was pre-pregnancy BMI, calculated from height and pre-pregnancy weight as reported in the survey. It was categorized as underweight (<18.5 kg/m2), normal (BMI = 18.5–<25 kg/m2), overweight (BMI = 25–<30 kg/m2) and obese (BMI ≥ 30 kg/m2) following NHLBI/WHO guidelines [9, 10], and also as a continuous variable. Self-reported pre-pregnancy weight was highly correlated with clinically measured weight at the time of prenatal screening in the second trimester (Pearson’s r = 0.91 among controls, 0.93 among cases).

Several other variables obtained from the telephone interview were considered. Psychosocial stress refers to distress from acute life events during pregnancy, measured by the Life Events Inventory modified for use with pregnant populations [16] Health variables during pregnancy included self-reported maternal smoking, diabetes (gestational and pre-existing), and averaging less than 6 h of sleep per night. Weight gain during pregnancy was calculated from self-reported pre-pregnancy weight and weight at delivery. Because shorter pregnancies provide less opportunity for weight gain, rate of gain (kg/week) was calculated as self-reported weight gained during pregnancy, divided by weeks gestation (LMP gestational age–2 weeks), and classified as low (<0.27 kg/week), adequate (0.27–0.52 kg/week), or high (>0.52 kg/week). This represents an average rate of gain over the duration of the pregnancy, and has been previously shown to predict PTD [17]. Socio-demographic variables were: maternal age, race/ethnicity, parity and highest educational attainment at time of delivery; annual household income; and marital/cohabiting status.

Data Analysis

All analyses were performed using SAS version 9.2 (SAS Institute, Inc., North Carolina). Cases and controls were compared on their socio-demographic and health characteristics, and P-values for comparisons were obtained from logistic regression models adjusting for frequency matching on race and birth month. Multiple logistic regression models were used to test whether BMI, treated as a categorical variable, predicted PTD after adjustment for potential confounders. To avoid excessive influence by outliers, the top and bottom 1% of BMI values were excluded from the models (n = 20). All models were also tested with these outliers included. Covariates were included in the model if associated with the outcome at P ≤ 0.10 in bivariate analyses. Diabetes was not included as a covariate due to the likelihood of its being on the causal pathway. Instead, odds ratios were calculated including and excluding women with diabetes. Height was tested as a covariate in each model but did not alter odds ratios by 10%. To account for frequency matching in the case–control design, odds ratios along with 95% confidence intervals (CI) were adjusted for race and month of birth.

The shape of the BMI–PTD relationship was investigated using BMI as a continuous variable. Locally weighted scatter plot smoothing (LOESS) was used to explore and plot the shape of the BMI–PTD relationship without parametric assumptions [18]. LOESS curves were generated with and without adjustment for race, separately for the two largest ethnic groups (White and Latina), and stratified by parity.

Based on the shape of the LOESS plot, BMI was modeled as two linear functions, one equal to or below and one above the inflexion point (BMI = 24 kg/m2). The slopes from these two logistic models were used to calculate odds ratios and 95% CIs for selected BMI values, with 24 kg/m2 as the reference point. Potential confounders and adjustments for study design were included as described above.

Results

Characteristics Associated with Pre-Term Delivery in the Study Population

Women delivering preterm were more likely to be primiparous, single, to sleep on average less than 6 h/night during pregnancy, to have high (>0.52 kg/week) or low (<0.27 kg/week) weight gain during pregnancy, and to report diabetes (Table 1). They were also more likely than controls to be obese or underweight prior to pregnancy, though the association with obesity was marginally significant (P = 0.06).
Table 1

Characteristics of the study population, by case and control status

 

PTD

Controls

Pa

n

Col%

n

Col%

Sociodemographic

Totals

354

100

710

100

 

Age

     

 18–25

81

23

173

24

0.83

 26–33

201

57

390

55

 

 34+

70

20

145

20

 

Parity

     

 Primiparous

191

54

311

44

0.008

 Multiparous

163

46

399

56

 

Race/Ethnicity

     

 White

159

45

297

42

NAb

 Latina

118

33

288

41

 

 Other

77

22

125

18

 

Marital status

     

 Single

39

11

60

8

0.08

 Married/Cohabiting

315

89

650

92

 

Income

     

 Low

64

18

142

20

0.70

 Middle

105

30

213

30

 

 High

183

52

352

50

 

Education

     

 No College

102

29

239

34

0.95

 Some college or beyond

251

71

471

66

 

Health

Any smoking during pregnancy

9

3

26

4

0.47

Diabetes/High blood sugar

56

16

62

9

0.002

<6 h sleep

44

12

64

9

0.05

Psychosocial stress

     

 None

78

22

174

25

0.25

 Low

89

25

195

28

 

 Moderate

87

25

164

23

 

 High

99

28

175

25

 

Height (tertiles)

     

 <1.58 M

118

33

227

32

0.44

 1.58–1.66 M

133

38

249

35

 

 >1.66 M

103

29

234

33

 

Weight gain during pregnancyc

     

 Low

88

25

145

21

0.004

 Adequate

179

51

425

60

 

 High

86

24

136

19

 

BMId

 Underweight

22

6

24

3

0.05

 Normal

205

58

437

62

 

 Overweight

72

20

158

22

 

 Obese

55

16

91

13

 

aP-value derived from logistic regression model, adjusted for frequency matching by race and birth month

bP-value not valid, as race was a frequency-matching variable

cWeight gain = Total kg gained/Weeks Gestation Low ≤ 0.27 kg/week, Adequate = 0.27–0.52 kg/week, High ≥ 0.52 kg/week) [17]

dUnderweight BMI < 18.5, Normal BMI = 18.5 to <25, Overweight BMI = 25 to <30, Obese BMI ≥ 30

The Association of Pre-pregnancy BMI Categories and PTD

Compared to women with a normal pre-pregnancy BMI, women who were underweight had over twice the odds of PTD (AOR = 2.11 (95%CI = 1.03–4.32) after adjusting for race, birth month, parity, marital status, nightly sleep hours and weight gain during pregnancy (Table 2). Neither overweight nor obesity was significantly associated with increased odds of PTD after adjustment for potential confounders. Inclusion of subjects with BMI in the top 1% (>43 kg/m2) and bottom 1% (<17 kg/m2) of values did not alter any of these odds ratios by 10% or more.
Table 2

Odds ratios and 95% confidence intervals for PTD as a function of categorized and continuous pre-pregnancy BMI

 

Design adjusted only

Fully adjusted

na

ORb

95%CI

n

AORb,d

95%CI

BMI

Categorical

 Underweight versus Normal

1,042

1.99

0.99–4.01

1,037

2.11

1.03–4.32

 Overweight versus Normal

 

1.08

0.78–1.51

 

1.07

0.76–1.51

 Obese versus Normal

 

1.42

0.96–2.11

 

1.31

0.86–2.00

Continuousc

 17 versus 24

580

4.02

1.83–8.80

578

4.31

1.94–9.54

 19 versus 24

 

2.70

1.54–4.73

 

2.84

1.61–5.01

 29 versus 24

462

1.39

1.10–1.76

459

1.42

1.10–1.84

 34 versus 24

 

1.95

1.22–3.11

 

2.01

1.20–3.39

aHighest and lowest 1% of BMI values excluded

bAdjusted for frequency matching by race and birth month

cOdds ratios calculated from beta values of logistic models with a continuous linear BMI term

dAdjusted for parity, marital status, sleep, and weight gain during prgenancy

The Association of Continuous BMI and PTD

As shown in Fig. 2, the log odds of PTD as a function of continuous BMI from the LOESS smooth was V-shaped. Minimum risk of PTD occurred near the border of normal and overweight BMI categories, around 24 kg/m2. Including subjects with BMI in the top 1% (>43 kg/m2) and bottom 1% (<17 kg/m2) of values in the LOESS analysis did not change the shape of the plot, but continued the linear increase in log odds as BMI increased above or decreased below 24 kg/m2. The shape of the plot and position of the inflexion point were not altered by any of the following: exclusion of women who reported diabetes; adjustment for race; modeling White and Latina women separately; stratifying on parity; or stratifying PTDs into early preterm deliveries (n = 144; prior to 35 weeks gestation, range 21–34.5 weeks) and late preterm deliveries (n = 200; 35–36 weeks gestation) (data not shown).
https://static-content.springer.com/image/art%3A10.1007%2Fs10995-010-0633-4/MediaObjects/10995_2010_633_Fig2_HTML.gif
Fig. 2

Loess plot of log odds of PTD as a function of prepregnancy BMI. Equivalentoddsratios are indicated on vertical axis. Verticallines indicate borders of IOM BMI categories as labeled

For women with BMI ≤ 24 kg/m2, each unit increase in BMI corresponded to reduced odds of PTD (β = −0.20, 95%CI = −0.31 to −0.09; AOR = 0.81 95%CI = 0.73–0.91), after adjustment for confounders (data not shown). Exclusion of subjects with diabetes or inclusion of subjects with BMI < 17 did not alter the OR by >10%. As shown in Table 2, the model predicts that relative to a woman with a BMI of 24 kg/m2, a woman with a BMI of 17 kg/m2 (underweight) would have more than a fourfold increased odds of PTD (AOR = 4.31 95%CI = 1.83–8.80), and a woman with a BMI of 19 kg/m2 (low normal) would have almost threefold elevated odds of PTD (AOR = 2.84 95%CI = 1.61–5.01).

Conversely, for women with BMI > 24 kg/m2, the odds of PTD increased with increasing BMI (β = 0.07, 95%CI = 0.02–0.12; AOR = 1.07, 95%CI = 1.02–1.13). The model predicts that an overweight woman with a BMI of 29 kg/m2, relative to a woman with a BMI of 24 kg/m2, would have a 40% increased odds of PTD (AOR = 1.42 95%CI = 1.10–1.84)(Table 2). For an obese woman with a BMI of 34 kg/m2, the odds relative to BMI of 24 kg/m2 would be AOR = 2.01 (95%CI = 1.20–3.39). Excluding women who reported diabetes during pregnancy did not alter the odds ratios by >10% (data not shown). Inclusion of outlier BMI values >43 weakened the relationship, but the ORs at all BMI values remained significant (data not shown).

Comparison of Categorical and Continuous BMI Models

Figure 3 illustrates the predictions of the categorical and continuous models of BMI and shows that a more pronounced effect of both low and high BMI on PTD is apparent when using continuous BMI data. The steep increase in odds on the left side of the inflexion point in the continuous model demonstrates heterogeneity within the normal weight category. The categorical model uses the normal weight category as reference, with odds defined as 1.00 for this BMI range. Grouping together the very different odds within the normal weight range category (18.5–25 kg/m2) inflates the odds of the reference group, resulting in a model that underestimates the increase in odds associated with both high and low BMI.
https://static-content.springer.com/image/art%3A10.1007%2Fs10995-010-0633-4/MediaObjects/10995_2010_633_Fig3_HTML.gif
Fig. 3

Log odds of PTD predicted by categorical (darkline) and continuous (grayline) models of BMI

Discussion

Low and High Pre-pregnancy BMI Predict Increased Preterm Delivery

Our findings from a population of working women in Southern California indicate that PTD is independently associated with both low and high pre-pregnancy BMI. In addition to corroborating previous studies that found increased PTD risk among underweight women [1921], we demonstrated a significant increase in PTD risk associated with high BMI. Our study found a protective effect of increased body mass up to 24 kg/m2 and a detrimental effect beyond that point. In agreement with Simhan [22] and Gilboa [23], minimum PTD risk in this sample is associated with a pre-pregnancy BMI in the high normal/low overweight range. Risk increases more steeply as BMI drops below 24 kg/m2 than as it rises above 24 kg/m2, so that a low normal BMI of 19 kg/m2 is associated with a nearly threefold increase in PTD risk compared to high normal BMI of 24 kg/m2, a higher risk than that associated with an obese BMI of 34 kg/m2.

To our knowledge, this study is the first to find a significant PTD risk increasing continuously with rising BMI. The previous study by Simhan et al. was restricted to spontaneous labors, which may account for the absence of a significant effect of high BMI in their results. In a study that examined spontaneous and indicated deliveries separately, Smith et al. concluded that overweight/obesity was protective with respect to spontaneous PTD, while increasing the risk of medically indicated PTD [24]. Although we cannot distinguish between spontaneous and indicated preterm births in our sample, it is plausible that the increased risk we observed with high BMI is due to medically indicated deliveries. Gilboa et al. also did not find elevated risk with high BMI, which may be due to small numbers of PTD cases in this range, estimation of gestational age based on LMP, or differences in the study population, which was restricted to White and Black women [23]. Our population was predominantly White and Latina, and while we saw similar results in White and Latina women, Simhan et al. found differences among Blacks, Latinas, and Whites. The question of whether and how ethnicity modifies the BMI-PTD relationship requires further study.

The observed shape of the BMI-PTD relationship was very similar for early and late PTD, indicating that the association is not limited to PTD at a specific gestational age. Furthermore, we found that the higher risk associated with overweight/obesity persisted after adjustment for confounders, and across all race/ethnicities. The risk associated with pre-pregnancy BMI was independent of weight gain during pregnancy. These findings suggest a relationship between pre-pregnancy BMI and PTD that is robust.

The risk associated with high pre-pregnancy BMI was largely independent of diabetes, a common complication of pregnancy associated with obesity. Similar findings were reported by Baeten et al. [5] in a population-based cohort study of 96,801 birth certificates in Washington state in the 1990s, in so far as the increased risk of PTD associated with pre-pregnancy obesity remained after excluding women with pre-gestational or gestational diabetes or hypertension. This independence raises questions for future studies of whether the BMI-PTD relationship is mediated by unmeasured disease processes, or by endocrine changes in obese women during or prior to pregnancy.

PTD has been linked to inflammation [25, 26], which is increasingly seen as an important feature of obesity-related metabolic syndrome [27, 28]. Gene expression studies of non-pregnant humans have found increased levels of pro-inflammatory cytokines produced by fatty tissues of obese individuals [2931]. These signaling molecules are also elevated in reproductive tissues just prior to term and preterm delivery, and are believed to play an important role in parturition [3234]. Moreover, increased expression of pro-inflammatory cytokines and accumulation of macrophages has been found in the placenta of obese pregnant women [35]. Taken together, these observations suggest that further studies of pro-inflammatory cytokines during pregnancy might shed light on the association of obesity and PTD.

Continuous Modeling of BMI Offers more Sensitive Risk Assessment

We compared the results of two modeling approaches, one using the NHLBI/WHO categories, and the other treating BMI as a continuous function. The categorical model implicitly assumes that risk is constant across the normal weight category used as a reference; we found that this assumption was inaccurate, as in fact there was a nearly threefold increase in risk between the lower versus upper values of the normal BMI range. Grouping the normal BMI range together inflates the risk of the reference group thereby underestimating the increase in risk associated with both high and low BMI.

From a research standpoint, it is important to note that treating BMI according to NHLBI/WHO categories masked both the significant relationship of PTD to high BMI and the risk associated with low BMI within the “normal” range, and underestimated the risk associated with underweight. From a clinical perspective, our results suggest that the NHLBI/WHO categories, which were recently endorsed by the Institute of Medicine [36], may not be appropriate for identifying women at higher risk for PTD.

Use of a continuous model had several advantages over BMI categories. First, it required no assumption of constant risk within categories. The reference point was derived empirically from the data. Second, statistical power was increased [37]. Third, it closely approximated the V-shape of the risk function. This avoided creating the inaccurate impression of a threshold effect. Finally, it allowed comparison of risks at various points along the BMI continuum.

Our study has several limitations. We could not distinguish between spontaneous and medically indicated deliveries, and the etiology of these two types of PTD may be different. Our data on pre-pregnancy height, weight and weight gain were self-reported and therefore subject to error. However, self-reported pre-pregnancy weight has been shown to produce accurate BMI categories for most women [38], and correlated well with clinically measured weight at mid-pregnancy. We also had insufficient numbers of morbidly obese (BMI > 40) women (n = 23) to analyze the effects of morbid obesity separately. We had only self-reported data on diabetes. In addition, we lacked data on medication use. Furthermore, the sample may not be demographically representative of the general obstetric population, since not all pregnant women work and only 45% of working women that were contacted met the work eligibility criteria. We sampled primarily Latina and White women; very few Black women were included, so our findings may not apply to this very high risk group. Future studies with larger populations are needed to address the effects of race, ethnicity, and induced versus spontaneous deliveries on the relationship between BMI and PTD.

Despite these caveats, the strengths of this study include a diverse three-county regional population, well-validated gestational age data, and an array of socio-demographic and health information, including weight gain during pregnancy. Our findings from continuous BMI models demonstrate a strong relationship between pre-pregnancy overweight/obesity and PTD, as well as low normal BMI and PTD.

Pre-pregnancy BMI is a potentially modifiable risk factor affecting many pregnant women. This study does not form a basis for clinical recommendations to individual women; future studies considering multiple complications and outcomes of pregnancy are needed to address what advice should be given to women, particularly those whose BMI is low-normal, or low-overweight. Most importantly, interventions aimed at helping women achieve a healthy pre-pregnancy BMI must occur before pregnancy. Recent evidence suggests that people who received advice from a health care professional to lose weight are more likely to attempt weight loss [39], but few obese adults receive such advice during routine checkups [40, 41]. Workplaces, universities, urban planners, and policymakers also need to encourage good nutrition and exercise for women of reproductive age. Moreover, obesity often begins well before a woman enters her childbearing years. Therefore, an important implication of this work is that public health measures promoting nutrition and exercise among children and adolescents are likely to improve future obstetric outcomes.

Acknowledgments

This work was funded in part by the Maternal and Child Health Bureau, division of research, training, and education, grant #R40MV00305-01 to SG. This study was also partially funded by NIH grant #R2MH081797, “Socioeconomic and Neuroendocrine Determinants of Perinatal Complications.” We thank Maureen Lahiff and David Lein for helpful discussions of statistical analyses.

Open Access

This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.

Copyright information

© The Author(s) 2010

Authors and Affiliations

  • Jessica Lang Kosa
    • 1
  • Sylvia Guendelman
    • 1
  • Michelle Pearl
    • 2
  • Steve Graham
    • 2
  • Barbara Abrams
    • 1
  • Martin Kharrazi
    • 3
  1. 1.Maternal and Child Health Program, School of Public HealthUniversity of CaliforniaBerkeleyUSA
  2. 2.Sequoia FoundationRichmondUSA
  3. 3.California Department of Public HealthRichmondUSA