Maternal Prepregnancy Body Mass Index and Risk of Spontaneous Preterm Birth

Authors


Abstract

Background

Findings from studies examining risk of preterm birth associated with elevated prepregnancy body mass index (BMI) have been inconsistent.

Methods

Within a large population-based cohort, we explored associations between prepregnancy BMI and spontaneous preterm birth across a spectrum of BMI, gestational age, and racial/ethnic categories. We analysed data for 989 687 singleton births in California, 2007–09. Preterm birth was grouped as 20–23, 24–27, 28–31, or 32–36 weeks gestation (compared with 37–41 weeks). BMI was categorised as <18.5 (underweight); 18.5–24.9 (normal); 25.0–29.9 (overweight); 30.0–34.9 (obese I); 35.0–39.9 (obese II); and ≥40.0 (obese III). We assessed associations between BMI and spontaneous preterm birth of varying severity among non-Hispanic White, Hispanic, and non-Hispanic Black women.

Results

Analyses of mothers without hypertension and diabetes, adjusted for age, education, height, and prenatal care initiation, showed obesity categories I–III to be associated with increased risk of spontaneous preterm birth at 20–23 and 24–27 weeks among those of parity 1 in each race/ethnic group. Relative risks for obese III and preterm birth at 20–23 weeks were 6.29 [95% confidence interval (CI) 3.06, 12.9], 4.34 [95% CI 2.30, 8.16], and 4.45 [95% CI 2.53, 7.82] for non-Hispanic Whites, non-Hispanic Blacks, and Hispanics, respectively. A similar, but lower risk, pattern was observed for women of parity ≥2 and preterm birth at 20–23 weeks. Underweight was associated with modest risks for preterm birth at ≥24 weeks among women in each racial/ethnic group regardless of parity.

Conclusions

The association between women's prepregnancy BMI and risk of spontaneous preterm birth is complex and is influenced by race/ethnicity, gestational age, and parity.

Globally, 13 million babies are born preterm, before 37 weeks gestation, every year with a frequency approximating 10%.[1] Preterm birth in the US occurs in 13% of all births and 11% of singleton births.[2] Preterm birth is substantially more frequent among African American babies, approximating 18%.[2] Despite medical advances, the incidence of preterm birth has increased in the US for decades and only recently have shown a slight decrease.[3] Survival for most babies born preterm has improved considerably, but survivors remain at increased risk for a variety of neurodevelopmental, gastrointestinal, and respiratory complications, many of which extend well beyond the neonatal period and contribute to lifelong challenges for individuals and their families, as well as to burdensome economic costs to society.[4]

Preterm birth is a complex phenomenon and is not well understood as a singular condition defined simply by the dichotomy of birth at or before 37 completed weeks gestation vs. later.[5] Such a classification has been argued as too simplistic for aetiological studies owing to the heterogeneity that has been observed with this outcome.[6] Indeed, a more detailed phenotypic classification such as extremely early (<28 weeks gestation) preterm birth has been suggested.[7]

Factors associated with risk of preterm birth have been reviewed previously with leading candidates including race-ethnicity, infection, smoking, and genetics.[8] Prepregnancy body mass index (BMI) has been investigated as a factor contributing to preterm birth risk in several studies. Many studies have found an increased risk of preterm birth associated with elevated BMI, specifically obesity.[9-11] These studies have substantially varied in their sample sizes and approaches: which categories of gestational ages of preterm birth were considered (e.g. <37 weeks only or more narrower categories such as <28 weeks or 28–32 weeks), how BMI was considered (e.g. obese vs. not obese or the full distribution of BMI to explore potential effects of lower BMI), and how covariates such as race/ethnicity, parity, and education were considered (e.g. primiparity as a distinct analytic group vs. muiltiparity adjusted as a potential confounder). No one US study has approached this line of inquiry with a large population-based cohort, considered the full BMI spectrum, and simultaneously explored population groupings such as race/ethnicity and parity for their contributions to an association between prepregnancy body mass and risk of preterm birth by specific narrowly defined gestational age groups.

In this analysis, we extend the extant literature by employing a large population (over a million births) to explore more narrow gestational age categories of preterm birth to determine whether an association between BMI (particularly elevated BMI) and spontaneous preterm birth severity differs by parity and race/ethnicity.

Methods

Data for this study come from the 2007–09 California birth cohorts. These data link California Vital Statistics birth records with Office of Statewide Health and Planning maternal and infant hospital discharge data. This linked dataset contains information on a range of maternal and pregnancy characteristics found on the birth certificate paired with clinical detail from the delivery hospitalisation for practically all inpatient livebirths and has been well described elsewhere.[12] The linkage algorithm utilised to assemble the birth cohorts is accurate and has previously been described.[13] This study was approved by both the Stanford University Institutional Review Board and the California State Committee for the Protection of Human Subjects.

Demographic data are based on information recorded in birth certificates. Variables considered included parity, race/ethnicity, maternal education, gestational age, maternal age, maternal height and prepregnancy weight, prenatal care, and payer type for delivery. Maternal comorbidities were identified from codes pertinent to the birth hospitalisation in the form of International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) diagnoses. Specifically, we applied previous methods used to assess maternal morbidity in pregnancy as follows: diabetes (250 and 648.0), gestational diabetes (648.8), chronic hypertension (401–405, 642.0, 642.1, 642.2, 642.7, and 642.9), gestational hypertension (642.3), and preeclampsia/eclampsia (642.4, 642.5, and 642.6).[14] Gestational age was the obstetric estimate of gestation at delivery in weeks provided on the birth certificate. Preterm birth was defined as <37 weeks, but analytically defined more narrowly, 20–23 weeks, 24–27 weeks, 28–31 weeks, or 32–36 weeks (compared with 37–41 weeks). Analysis was restricted to spontaneous preterm birth events based on information coded on hospital discharge or birth certificate records. Spontaneous preterm birth was identified as those births <37 weeks with preterm premature rupture of membranes (ICD-9-CM code 658.1 or birth certificate complication of labor/delivery code 10), those with premature labor (ICD-9-CM code 644), or the use of tocolytics (birth certificate complication/procedure of pregnancy code 28). BMI was computed based on height and prepregnancy weight as weight (in kilograms)/height[2] (in metres). BMI was analytically considered as a categorical variable using National Institutes of Health/National Heart, Lung, and Blood Institute guidelines (http://www.nhlbi.nih.gov/health/public/heart/obesity/lose_wt/bmi_dis.htm): <18.5 (underweight); 18.5–24.9 (normal); 25.0–29.9 (overweight); 30.0–34.9 (obese I); 35.0–39.9 (obese II); and ≥40.0 (obese III). Data on height and prepregnancy weight came from birth certificates.

In the 3-year period 2007–09, there were 1 601 404 livebirths linked with maternal and infant hospital discharge summaries representing 97% of all statewide births. From these linked records, 49 583 multiple gestations and 15 903 births with gestational age <20 or >41 weeks were removed, leaving a total 1 535 918 singleton births. Of these, we identified births with exclusion criteria (not mutually exclusive) such as unknown maternal education (n = 50 000), maternal age <13 or >55 years (n = 64), maternal height <1.35 or >1.96 metres (n = 90 009), maternal prepregnancy weight <34.02 or >204.12 kg (n = 129 436), and parity unknown or >10 (n = 1842), and resulting in 1 337 899 (87.1%) births with complete outcome and covariate data. Included for initial inspection were 1 152 888 births to non-Hispanic White, non-Hispanic Black, and Hispanic mothers. Among these, we further excluded those whose mothers had pregestational diabetes, gestational diabetes, gestational hypertension, pre-eclampsia/eclampsia, and chronic hypertension for each gestational age group, except the 20–23-week group. These exclusions were motivated by our goal to determine whether obesity alone, i.e. not mediated by or through these co-morbid conditions, led to associations with spontaneous preterm birth. (Results of analyses without these exclusions have been included as Supporting Information Tables S1 and S2.) For the 20–23 week group, gestational diabetes was not considered an exclusion criterion owing to delivery occurring before the time gestational diabetes is typically diagnosed (24–28 weeks). This resulted in the removal of 145 505 births, leaving 1 007 415 births to mothers without hypertensive or diabetic disease (60 499 preterm, 946 916 term). Lastly, we included only the 42 771 spontaneous preterm births for analysis.

The association between BMI and spontaneous preterm groupings was estimated with relative risks and their 95% confidence intervals [CI] employing Poisson regression models. Covariates considered were: ethnicity (non-Hispanic White, Hispanic, and non-Hispanic Black), educational attainment (some high school or less, high school graduate (or equivalent), some college, and college graduate or more education), prenatal care initiation (≤5 months or after 5 months gestation establishes same temporal opportunity for preterm and term pregnancies), parity (parity 1 or parity ≥2), maternal age (years), and maternal height to reduce further potential residual confounding associated with the BMI algorithm.[15] BMI was categorised as underweight, normal, and overweight, and assessed two ways with regards to obesity: (i) collectively for all classes of obesity and (ii) separately for obese class I (BMI 30.0–34.9), obese class II (BMI 35.0–39.9), and obese class III (BMI ≥40.0). Due to delivery occurring before gestational diabetes could be diagnosed, logistic models for the 20–23 week gestational age group included term (37–41 weeks) births whose mothers had gestational diabetes for more appropriate comparison. The additional term births were (beyond those shown in Table 1): non-Hispanic White (n = 326 487), non-Hispanic Black (n = 60 586), and Hispanic (n = 616 744). All analyses were performed with SAS version 9.3 (SAS Institute, Cary, NC, USA).

Table 1. Characteristics of the study population, spontaneous preterm, and term singleton births to mothers without hypertensive or diabetic disease, California 2007–09, n = 989 687, n (%)
 Non-Hispanic White (n = 325 655)Non-Hispanic Black (n = 62 185)Hispanic (n = 601 847)
Gestational age at delivery (weeks)   
20–23202 (0.06)188 (0.30)615 (0.10)
24–27516 (0.16)345 (0.55)1385 (0.23)
28–31987 (0.30)412 (0.66)2173 (0.36)
32–3611 799 (3.62)3015 (4.85)21 134 (3.51)
37–41312 151 (95.85)58 225 (93.63)576 540 (95.80)
Prepregnancy body mass index   
Underweight14 776 (4.54)2763 (4.44)18 649 (3.10)
Normal194 530 (59.73)28 076 (45.15)276 936 (46.01)
Overweight70 277 (21.58)16 392 (26.36)178 669 (29.69)
Obese I28 274 (8.68)8276 (13.31)84 566 (14.05)
Obese II11 655 (3.58)3909 (6.29)29 000 (4.82)
Obese III6143 (1.89)2769 (4.45)14 027 (2.33)
Maternal age (years)   
<2015 734 (4.83)9441 (15.18)84 261 (14.00)
20–2456 072 (17.22)18 939 (30.46)170 877 (28.39)
25–2989 946 (27.62)16 041 (25.80)165 960 (27.58)
30–3492 494 (28.40)10 807 (17.38)115 263 (19.15)
≥3571 409 (21.93)6957 (11.19)65 486 (10.88)
Maternal education   
Some high school20 895 (6.42)10 951 (17.61)257 435 (42.77)
High school graduate72 848 (22.37)21 917 (35.24)185 766 (30.87)
Some college90 350 (27.74)20 841 (33.51)112 768 (18.74)
≥4-year college degree141 562 (43.47)8476 (13.63)45 878 (7.62)
Prenatal care initiation by fifth month of gestation   
Yes309 966 (95.18)56 618 (91.05)557 597 (92.65)
No15 689 (4.82)5567 (8.95)44 250 (7.35)
Parity   
1143 523 (44.07)25 922 (41.69)214 186 (35.59)
≥2182 132 (55.93)36 263 (58.31)387 661 (64.41)
Payer type for delivery   
Medi-Cal74 022 (22.73)33 123 (53.27)394 093 (65.48)
Private236 612 (72.66)23 554 (37.88)177 060 (29.42)
Uninsured4349 (1.34)1150 (1.85)15 096 (2.51)
Other10 672 (3.28)4358 (7.01)15 598 (2.59)

Results

Displayed in Table 1 are characteristics of the 989 687 singleton births investigated that served as our analytic base. The majority of births were to women of Hispanic ethnicity, and approximately 20% of mothers had a prepregnancy BMI that categorised them as obese (BMI ≥30).

Obesity

Obesity, defined as BMI ≥30, was not associated with increased risk for spontaneous preterm birth broadly defined as <37 weeks gestation, i.e. relative risk 0.99 [95% CI 0.96, 1.01]. However, analyses with more granular definitions of gestational age and analyses stratified by parity showed associations between obesity and preterm birth. Figures 1-3 show these associations for obesity and preterm birth categories for the three studied ethnic groups, with analyses adjusted for maternal age, education, height, and prenatal care initiation. In these analyses, obesity is associated with increased risk of preterm birth at gestational ages 20–23 weeks (relative risks of 3.49, 2.92, and 2.48 among non-Hispanic Whites, Hispanics, and non-Hispanic Blacks, respectively) and 24–27 weeks for each of the racial/ethnic groups among parity 1 women. Observed odds ratios among parity ≥2 women revealed less of an effect for preterm birth at 20–23 or 24–27 weeks in comparison with parity 1.

Figure 1.

Relative risks for spontaneous preterm birth and prepregnancy obesity [body mass index (BMI) ≥30 compared with normal BMI] by parity for non-Hispanic White mothers without hypertensive or diabetic disease, California births 2007–09. Relative risk adjusted for maternal age, education, height, and prenatal care initiation.

Figure 2.

Relative risks for spontaneous preterm birth and prepregnancy obesity [body mass index (BMI) ≥30 compared with normal BMI] by parity for Hispanic mothers without hypertensive or diabetic disease, California births 2007–09. Relative risk adjusted for maternal age, education, height, and prenatal care initiation.

Figure 3.

Relative risks for spontaneous preterm birth and prepregnancy obesity [body mass index (BMI) ≥30 compared with normal BMI] by parity for non-Hispanic Black mothers without hypertensive or diabetic disease, California births 2007–09. Relative risk adjusted for maternal age, education, height, and prenatal care initiation.

Results of analyses considering a broader spectrum of BMI are shown in Tables 2 (parity 1) and 3 (parity ≥2) stratified by race/ethnicity and adjusted for maternal age, education, height, and prenatal care initiation. Each of the three obesity categories (I–III) was individually associated with increased risk of preterm birth at 20–23 weeks for parity 1 women across all race/ethnicity groups, with risk increasing with increasing level of obesity. These findings were similar for preterm birth 24–27 weeks for parity 1, but lower in magnitude of risk. The consistency of the association, its magnitude of risk, and the increasing risk with increasing obesity category were not similarly observed among women of parity ≥2 (Table 3).

Table 2. Relative risksa (RR) [95% confidence intervals] for spontaneous preterm birth and prepregnancy body mass index by race/ethnicity among parity 1 mothers without hypertensive or diabetic disease, California 2007–09
Parity 120–23 weeks24–27 weeks28–31 weeks32–36 weeks
nRR [95% CI]nRR [95% CI]nRR [95% CI]nRR [95% CI]
  1. aReflects estimated effect relative to term births (37–41 weeks) among women of normal body mass index (BMI = 18.5–24.9). Relative risk adjusted for maternal age, education, height, and prenatal care. For relative risk calculation among gestational age group 20–23 weeks, comparison group of term births included the following additional births beyond those shown in Table 1: non-Hispanic White (n = 326 487), non-Hispanic Black (n = 60 586), and Hispanic (n = 616 744).
Non-Hispanic White        
Underweight40.93 [0.34, 2.60]151.26 [0.74, 2.16]301.20 [0.83, 1.75]3511.28 [1.15, 1.43]
Normal481.00 [Reference]1401.00 [Reference]3141.00 [Reference]35021.00 [Reference]
Overweight231.38 [0.84, 2.27]571.23 [0.90, 1.67]950.94 [0.75, 1.19]10690.97 [0.91, 1.04]
Obese I182.69 [1.56, 4.65]221.25 [0.79, 1.96]421.13 [0.81, 1.56]3890.98 [0.88, 1.09]
Obese II114.02 [2.07, 7.78]263.79 [2.48, 5.78]140.99 [0.58, 1.69]1360.91 [0.76, 1.08]
Obese III96.29 [3.06, 12.9]72.06 [0.96, 4.42]81.15 [0.57, 2.32]710.97 [0.76, 1.22]
Non-Hispanic Black        
Underweight10.22 [0.03, 1.63]81.03 [0.50, 2.15]141.70 [0.96, 3.00]861.45 [1.15, 1.82]
Normal401.00 [Reference]711.00 [Reference]811.00 [Reference]5671.00 [Reference]
Overweight231.16 [0.69, 1.94]421.26 [0.86, 1.85]330.86 [0.57, 1.29]2000.76 [0.64, 0.89]
Obese I171.89 [1.07, 3.34]251.70 [1.07, 2.68]231.36 [0.85, 2.16]960.82 [0.66, 1.02]
Obese II112.47 [1.26, 4.84]101.48 [0.76, 2.88]50.65 [0.26, 1.61]480.91 [0.68, 1.22]
Obese III134.34 [2.30, 8.16]173.86 [2.26, 6.58]30.61 [0.19, 1.93]260.77 [0.52, 1.14]
Latina        
Underweight141.72 [0.98, 3.01]361.47 [1.03, 2.08]581.55 [1.17, 2.04]4671.32 [1.20, 1.45]
Normal971.00 [Reference]2801.00 [Reference]4371.00 [Reference]41591.00 [Reference]
Overweight761.66 [1.23, 2.24]1751.38 [1.14, 1.66]2121.06 [0.90, 1.25]17940.95 [0.90, 1.00]
Obese I522.76 [1.97, 3.87]1042.04 [1.63, 2.56]961.20 [0.96, 1.50]7470.99 [0.92, 1.07]
Obese II172.65 [1.58, 4.46]472.84 [2.08, 3.87]431.66 [1.21, 2.27]2130.88 [0.77, 1.01]
Obese III144.45 [2.53, 7.82]232.94 [1.92, 4.51]181.46 [0.91, 2.35]1080.94 [0.78, 1.14]
Table 3. Relative risksa (RR) [95% confidence intervals] for spontaneous preterm birth and prepregnancy body mass index by race/ethnicity among parity ≥2 mothers without hypertensive or diabetic disease, California 2007–09
Parity ≥220–23 weeks24–27 weeks28–31 weeks32–36 weeks
nRR [95% CI]nRR [95% CI]nRR [95% CI]nRR [95% CI]
  1. aReflects estimated effect relative to term births (37–41 weeks) among women of normal body mass index (BMI = 18.5–24.9). Relative risk adjusted for maternal age, education, height, and prenatal care. For relative risk calculation among gestational age group 20–23 weeks, comparison group of term births included the following additional births beyond those shown in Table 1: non-Hispanic White (n = 326 487), non-Hispanic Black (n = 60 586), and Hispanic (n = 616 744).
Non-Hispanic White        
Underweight51.54 [0.61, 3.90]121.36 [0.75, 2.46]422.12 [1.53, 2.94]3761.48 [1.33, 1.65]
Normal431.00 [Reference]1211.00 [Reference]2731.00 [Reference]35471.00 [Reference]
Overweight190.93 [0.54, 1.59]520.96 [0.69, 1.33]1000.81 [0.65, 1.03]13410.89 [0.83, 0.94]
Obese I141.41 [0.76, 2.58]411.65 [1.15, 2.36]390.69 [0.49, 0.96]6370.94 [0.86, 1.02]
Obese II71.53 [0.69, 3.42]151.38 [0.80, 2.37]200.81 [0.51, 1.28]2380.81 [0.71, 0.93]
Obese III10.37 [0.05, 2.66]81.29 [0.63, 2.66]100.71 [0.38, 1.34]1420.86 [0.73, 1.02]
Non-Hispanic Black        
Underweight20.64 [0.15, 2.66]91.50 [0.75, 3.01]151.44 [0.84, 2.48]1111.40 [1.14, 1.70]
Normal341.00 [Reference]721.00 [Reference]1231.00 [Reference]9211.00 [Reference]
Overweight150.62 [0.33, 1.13]480.92 [0.64, 1.32]510.58 [0.42, 0.80]4750.75 [0.67, 0.83]
Obese I221.63 [0.95, 2.79]210.74 [0.45, 1.20]360.75 [0.51, 1.08]2800.80 [0.70, 0.92]
Obese II71.03 [0.46, 2.34]100.72 [0.37, 1.39]200.85 [0.53, 1.36]1230.73 [0.60, 0.88]
Obese III30.59 [0.18, 1.92]121.16 [0.63, 2.14]80.46 [0.22, 0.94]820.65 [0.52, 0.82]
Latina        
Underweight51.00 [0.41, 2.47]171.43 [0.87, 2.34]562.20 [1.67, 2.90]4441.52 [1.38, 1.68]
Normal1121.00 [Reference]2541.00 [Reference]5441.00 [Reference]59491.00 [Reference]
Overweight1011.01 [0.77, 1.32]2241.07 [0.89, 1.28]4140.92 [0.81, 1.05]41370.87 [0.83, 0.90]
Obese I761.43 [1.06, 1.91]1531.43 [1.17, 1.75]1930.84 [0.72, 1.00]20490.85 [0.81, 0.89]
Obese II361.90 [1.30, 2.76]481.30 [0.96, 1.77]710.90 [0.70, 1.15]7140.85 [0.79, 0.92]
Obese III151.59 [0.93, 2.73]241.34 [0.88, 2.04]310.81 [0.56, 1.16]3530.87 [0.78, 0.97]

In contrast to elevated risks observed between obesity and early preterm birth, obesity was observed to be associated with decreased risk for late preterm birth (Figures 1-3 and Table 3) among women of parity ≥2.

Underweight

In contrast to obesity findings, elevated risks for underweight women tended to be associated with later gestational age preterm birth and were similar for parity (Tables 2,3).

Comment

In an analysis of over a million California births, we explored four defined gestational age subgroups of preterm birth among women without hypertension and diabetes to determine whether an association between prepregnancy BMI and spontaneous preterm birth risk differed by certain maternal characteristics, particularly parity and race/ethnicity. Results indicated that relationships between women's prepregnancy BMI and risk of preterm birth are complex. Such relationships appear to depend on the gestational timing of the preterm birth and to be further influenced by race/ethnicity and parity.

We observed prepregnancy obesity to be associated with increased risk of early spontaneous preterm birth (20–23 or 24–27 weeks) regardless of maternal race/ethnicity among women of parity 1. Indeed, these risks tended to increase with increasing level of obesity. In general, obesity was much less consistently associated with later onset preterm birth (32–36 weeks), if at all. Of note, low BMI (underweight) was generally associated with increased risks of preterm birth ≥24 weeks irrespective of race/ethnicity or parity. Our observations appeared to be unrelated to influences of gestational diabetes, pregestational diabetes, chronic hypertension, maternal age, maternal height, and prenatal care timing owing to analyses either excluding individuals or adjusting for these potential confounding or effect modifying variables.

In their recent meta-analysis of 84 studies representing approximately one million patients, McDonald et al.[9] note that an association between prepregnancy overweight or obesity in women and preterm birth is debated. From this body of literature involving sizable design variation across studies, the authors concluded that women with elevated BMI (overweight and obesity) were at higher risk for induced, but not spontaneous, preterm birth defined as <37 weeks gestation. Among 662 711 singleton livebirths in Florida, Salihu and colleagues[16] observed that obesity (BMI ≥30) was a risk factor among first-time mothers for preterm birth <28 weeks (odds ratio = 1.97) and <33 weeks (odds ratio 1.56), relative to term birth. Three recent studies from Norway, Finland, and Sweden are also informative. Khatibi et al.[17] investigated the Norwegian Mother and Child Cohort Study for an association between BMI and preterm birth. This study involved 83 544 pregnancies, much larger than any study conducted previously, but still only 1/12th the size of our study population. The Norwegian study also divided preterm birth into early (22–31 weeks) and late (32–36 weeks). These investigators observed obesity to be associated with both early and late spontaneous preterm birth. Reported odds ratios for early preterm birth were 1.57 [95% CI 1.03, 2.41] for obese group I and 3.29 [95% CI 1.21, 8.99] for obese group III. A large study similar to ours was conducted in Sweden (>1.5 million births) finding similar results.[18] Specifically, elevated risks of 1.2–2.7 were also observed for early (defined as 22–27 weeks) spontaneous preterm birth associated with obesity, with largest risks observed for the most obese women. Most recently, Raisanen and colleagues[19] in a Finnish population of more than a million singleton births observed increased risks for preterm birth <28 weeks (odds ratio 1.48) and <32 weeks (odds ratio 1.46) associated with women whose BMI was ≥30. The Scandinavian results compare well with those in the current study despite differences in definitions (for obesity and very early preterm birth) and underlying populations studied such as prevalence of preterm birth or prevalence of obesity among these countries (Norway, Sweden, and Finland) and the US.

Our observations extend the extant literature on BMI and preterm birth risk in several ways. Like the recent Swedish study,[18] our analytic base was large including nearly one million births. Such a large study population allowed for parsing of preterm birth definitions into narrower gestational week definitions rather than the single cutpoint of <37 weeks and even more narrowly than <28 weeks. The large study population also allowed for examination of the BMI spectrum in such a way to estimate risks for underweight and for several categories of obesity. Most notably, the large study population permitted an investigation of narrow gestational groups and BMI by potentially important covariates, parity, and race/ethnicity.

Simhan and Bodnar[20] also explored whether BMI and preterm birth associations varied by maternal characteristics, specifically race-ethnicity. They observed low BMI, but not higher BMI, to be associated with increased risk of preterm birth defined as <37 weeks. The low BMI association was more pronounced among Hispanic White and Black women. Our study also found low BMI associated with risk of preterm birth groups ≥24 weeks, with similar risk estimates observed across race/ethnic groups.

Although the strengths of the current study included its large sample size and attendant ability to explore narrow gestational timing of spontaneous preterm birth in combination with numerous characteristics, the study is not without its limitations. First, data were derived from birth certificates and discharge databases, not from detailed medical records. Owing to these sources of data, we did not have information on previous history of preterm birth. The lack of this information would not influence our observed results for parity 1 women, but may have influenced those for parity ≥2 women. Second, although complete data were available on >87% of the eligible population, our observations need to be considered in light of a lack of information on approximately 13% of the population. Third, like ours and many other studies that explore BMI based on self-reported information, such studies have some potential for error owing to women not accurately reporting their height and prepregnancy weight. Such errors have been identified to result in biased estimates of risk for preterm birth and other adverse outcomes of pregnancy in investigations of BMI associations.[21] We cannot predict the magnitude of bias that could arise from misclassified BMI measures nor predict the directionality that such bias would have on observed risk estimates in this study. A related issue is the accuracy of height and prepregnancy weight specifically recorded (self-reported) on birth certificates. Birth certificates as a source of information of BMI have been demonstrated to be valid.[22] Although Park et al.[22] observed some underestimation of prepregnancy weight, such underestimation would likely result in underestimating relative risks in the current study because we used cut-offs of BMI that would tend to place the underestimators into the referent population (normal range BMI group). A fourth potential limitation of the current work is the lack of reliable information on gestational weight gain. High weight gain has been associated with medically indicated preterm birth and low weight gain with spontaneous preterm birth specifically among obese women.[23] If such information had been available, it is not likely to have altered our observations between early preterm birth (20–23 weeks) and obesity, but it would have likely sharpened our inferences. For example, it would have been advantageous to explore whether obese women with lower weight gain differed from those with higher weight gain for risk of spontaneous preterm birth at various gestational weeks and whether such relationships were further influenced by race/ethnicity and parity. Another limitation is a lack of information on cigarette smoking – a known risk factor for preterm birth.

Our finding of an increased risk for early spontaneous preterm birth associated with obesity may partially reflect a known underlying association between structural birth defects and early preterm birth,[24] and additionally between obesity and selected birth defects.[25] We did not have reliable information (from birth certificates) on structural birth defects to remove affected infants from our analyses, which would be the preferred approach. However, some pregestational diabetes-related birth defects were likely removed based on our exclusions of births whose mothers had pregestational diabetes. We did attempt to estimate the potential magnitude for a connection among birth defects, obesity, and gestational timing in a smaller, but more granular dataset. Among 838 infants with structural birth defects identified for the California component of the National Birth Defects Prevention Study,[26] we did not observe an excess percentage of such infants in earlier gestational age groups for obese mothers.

Further, our finding an increased risk for very early preterm birth (20–23 weeks) associated with obesity may be partially attributed to the previously described increased risk observed between obesity and stillbirth.[27-29] Although the early preterm births studied here were classified as liveborn, it is notable that the survival for many of these infants under 24 weeks would be poor even with the improved neonatal care currently available. Thus, irrespective of registration issues of liveborn or stillborn at these early gestational ages, prepregnancy obesity appears to be contributing to the risk of fetal/infant mortality.

The biological explanation for a specific association between obesity and early spontaneous preterm delivery is not readily apparent. Obesity represents a complex phenomenon of over and under nutrition and is also associated with altered immune states. Both of these generalised pathways (nutrition and inflammation) have been implicated in preterm birth. Indeed, changes in gut microbiome also have been implicated in substantive changes in a pregnant woman's metabolism and can induce symptoms akin to metabolic syndrome in non-pregnant mice.[30] There is also some evidence that among women with cervical insufficiency (potentially greatest in the 20–23 week group), obesity may be associated with decreased gestational age at delivery.[31]

This large-scale study cannot identify specific mechanisms underlying the association between prepregnant weight (underweight and obesity) and risk of spontaneous preterm birth. Importantly, however, our study does identify segments of the population that may offer clues as to where to initiate mechanistic studies that can improve our understanding and prevention of preterm birth.

Acknowledgements

The March of Dimes Foundation and Stanford University School of Medicine provided grant support for the March of Dimes Prematurity Research Center at Stanford University School of Medicine.

Ancillary