Objective
The purpose of this study was to examine associations between the prepregnancy maternal body mass index (BMI) across the 3 clinical presentations of preterm birth (PTB).
Study Design
We conducted a retrospective cohort study of the records of 11,726 women. The World Health Organization International Classification was used to categorize BMI. The primary outcome of the study was PTB (<37 weeks’ gestation) presenting as spontaneous preterm labor, preterm premature rupture of the membranes, or a medical indication. We used univariable and multivariable logistic regression analysis to analyze the data ( P < .05).
Results
We found (1) a significant increase in the overall incidence of PTB at the extremes of BMI, (2) a higher risk for PTB from spontaneous preterm labor at the lower extremes (low plus moderate thinness) of BMI (adjusted odds ratio [aOR], 2.4; 95% confidence interval [CI], 1.4–4.2; P = .003), (3) a higher risk for preterm premature rupture of the membranes at the upper extremes (obese class II plus III) of BMI (aOR, 1.6; 95% CI, 1.1–2.3; P = .02), and (4) a higher risk for a medically indicated PTB at the lower (aOR, 2.8; 95% CI, 1.4–5.6; P = .004) and upper (aOR, 1.5; 95% CI, 1.1–2.2; P = .02) extreme of BMI.
Conclusion
Women at the extremes of prepregnancy BMI are at risk for PTB.
A pathologic outcome of pregnancy, preterm birth (PTB; <37 weeks’ gestation) is a predictor of long-term medical and neurodevelopmental sequelae for the neonate. Preterm birth is accompanied by high economic and emotional costs. After years of a rising rate of PTB in the United States, it recently has shown a decline to approximately 12%. However, the rate remains disproportionately high (17.5%) among select groups of women, specifically African American women. We need to understand the mechanisms and multiple causal pathways behind this major public health problem to make substantive progress in decreasing the rate.
The maternal prepregnancy body mass index (BMI) is one potential modifiable risk factor for PTB. With 23% of women of reproductive age now estimated to be obese, the focus has been on the relationship between prepregnancy obesity and PTB. However, women who are underweight before pregnancy are also at an elevated risk of a PTB.
The objective of this study was to examine the relationship of prepregnancy BMI with PTB in a recent (October 2005 to October 2010) cohort of >11,000 racially and ethnically diverse women from Colorado with the use of the World Health Organization categorization of BMI categories We hypothesized that, when the data were adjusted for other risk factors, women with extremes of BMI before pregnancy would have the highest risk of PTB compared with women who had a BMI in the normal range before pregnancy and that this relationship would be similar across the 3 clinical presentations of PTB (PTB resulting from preterm labor, preterm premature rupture of the membranes [PPROM] or PTB from a medical indication).
In 2009, the Global Alliance to Prevent Prematurity and Stillbirth convened an International Conference on Prematurity and Stillbirth to examine the need for a comprehensive, consistent, and uniform classification system for PTB. Several manuscripts resulted from this conference that challenged the traditional classification of PTB and proposed that PTB must be recognized as a syndrome with many etiologic factors and phenotypic characteristics, many of which are independent of each other. The authors suggested a new phenotypic classification of PTB. Prompted by this report, we included a meticulous classification of the evidence of initiation of parturition in our study to reflect one of the phenotypic components of this new PTB classification. Moreover, we conducted a comprehensive review of the maternal and fetal indications for a medically indicated PTB. This approach was also driven by the paucity of research that examined links between prepregnancy BMI across these 3 clinical presentations of PTB. An additional distinguishing feature of our study was the inclusion in the data analysis of a comprehensive set of maternal risk factors for PTB. Our purpose was to investigate whether the risk factors differed across the 3 clinical presentations of PTB.
Materials and Methods
The study was a retrospective cohort study conducted with the Department of Obstetrics and Gynecology Perinatal Database after approval from the Colorado Multiple Institutional Review Board. In brief, a comprehensive set of data on maternal risk factors, events during labor and delivery, pregnancy complications and neonatal events were collected daily on every delivery that occurred at the University of Colorado Hospital by our Perinatal Professional Research Assistants. With web-based data entry, the data were transmitted to the Data Coordinating Center in the Division of Biostatistics and Bioinformatics at National Jewish Health (Denver, CO). Quality control on this dataset included procedures for automatic range checks and an annual secondary review of 10-15% of the records. For this study, every record coded as having a delivery at <37 weeks’ gestation received a detailed secondary review (O.C.A. and B.M.F.) to ensure that there was no misclassification of the outcome and to characterize the outcome accurately. Discrepancies were resolved by another chart review by an obstetrician gynecologist. For medically indicated PTB, the primary indication for delivery was categorized as either maternal or fetal and further subcategorized by indication for delivery.
We included in this study the records of 13,683 women who delivered between October 2005 and October 2010 at the University of Colorado Hospital. Records were deleted from the dataset for the following reasons: multiple births (n = 368), deliveries at <20 weeks’ gestation or missing gestational age at delivery (n = 67), terminations of pregnancy (n = 34), delivery records missing (home delivery, no prenatal care; n = 12), missing maternal prepregnancy BMI (n = 864), and women who were transported to labor and delivery (n = 612) at University of Colorado Hospital. Only the first delivery of women with multiple pregnancies during the study period were included. The final analytic dataset contained 11,726 deidentified records.
The primary explanatory variable was the maternal prepregnancy BMI defined as the maternal prepregnancy weight divided by the square of the height (kilograms/square meters). We used the World Health Organization International Classification of BMI. The categories were severe thinness (BMI, <16 kg/m 2 ), moderate thinness (BMI, 16-16.99 kg/m 2 ), mild thinness (BMI, 17-18.49 kg/m 2 ), normal weight (BMI, 18.5-24.99 kg/m 2 ), preobesity (BMI, 25.00-29.99 kg/m 2 ), obese class I (BMI, 30-34.99 kg/m 2 ), obese class II (BMI, 35.00-39.99 kg/m 2 ), and obese class III (≥40.00 kg/m 2 ). The normal prepregnancy BMI group was the referent group. The maternal prepregnancy weight was self-reported by each woman. The maternal height was obtained from the prenatal medical record. We also included the following risk factors in the analysis: race/ethnicity, maternal age, parity (nulliparous vs multiparous), a maternal history of cigarette smoking at conception (yes/no), PTB (yes/no), chronic hypertension (yes/no), diabetes mellitus (type 1, type 2, or gestational diabetes [yes/no]), uterine anomalies (yes/no), cervical incompetence (yes/no), infertility (yes/no), and spontaneous or induced abortion (yes/no).
The primary outcome of the study was PTB (from 20 to <37 completed weeks’ gestation) that resulted from a spontaneous preterm labor (SPTL), PPROM, or a medical indication. In accordance with contemporary practice, the assessment of gestational age was based primarily on clinical assessment at the first visit and on early ultrasound examination in most of the cohort.
We analyzed the data using SAS software (version 9.3; SAS Institute Inc, Cary, NC). We used univariable analysis to generate descriptive statistics for the cohort. Differences between categorical variables were examined with the χ 2 test ( P < .05). The relative risk was used as a measure of association between dichotomous variables. Differences in means of continuous variables were tested with the Wilcoxon rank-sum test. Multivariable logistic regression analysis was used to determine the odds ratio of the main explanatory variables for PTB, adjusted for other covariates. Separate logistic regression models were run for each of the 3 clinical presentations.
Results
In the analytic dataset of 11,726 records, 1251 (10.7%) women had a pregnancy that was complicated by a PTB. Among the PTBs, 485 (39%) resulted from SPTL; 336 (27%) resulted from PPROM, and 430 (34%) resulted from a medical indication. We show in Table 1 that 53% of the women had a normal BMI, and 25% of these women were in the preobesity range. We found that the underweight categories accounted for 5% of the cohort, whereas 17% of the women had a prepregnancy BMI in one of the obese categories. The other risk factors included in the analysis are summarized in Table 1 .
| Variable | Measure |
|---|---|
| Preterm birth: all, n (%) | 1251 (10.7) |
| Clinical presentation of preterm birth, n (%) | |
| Spontaneous preterm labor | 485 (4.1) |
| Preterm premature rupture of membranes | 336 (2.9) |
| Medically indicated | 430 (3.7) |
| Body mass index category, n (%) | |
| Severe thinness (<16.00 kg/m 2 ) | 33 (0.3) |
| Moderate thinness (16.00-16.99 kg/m 2 ) | 94 (0.8) |
| Mild thinness (17.00-18.49 kg/m 2 ) | 454 (3.9) |
| Normal (18.50-24.99 kg/m 2 ) | 6223 (53.1) |
| Preobesity (25.00-29.99 kg/m 2 ) | 2892 (24.6) |
| Obese class I (30.00-34.99 kg/m 2 ) | 1252 (10.7) |
| Obese class II (35.00-39.99 kg/m 2 ) | 464 (3.9) |
| Obese class III (≥40.00 kg/m 2 ) | 314 (2.7) |
| Race/ethnicity, n (%) a | |
| Hispanic | 5181 (44.2) |
| Non-Hispanic white | 4038 (34.4) |
| Other | 265 (2.3) |
| African American | 1746 (14.9) |
| American Indian | 54 (0.5) |
| Asian | 438 (3.7) |
| Maternal age, y b | 26.7 ± 6.3 |
| Nulliparity, n (%) | 5312 (45.3) |
| History, n (%) | |
| Spontaneous abortion | 2629 (22.4) |
| Induced abortion | 1045 (8.9) |
| Preterm birth | 997 (8.5) |
| Cigarette smoking | 1642 (14.0) |
| Cervical incompetence | 67 (0.6) |
| Infertility | 459 (3.9) |
| Uterine anomalies | 363 (3.1) |
| Chronic hypertension | 277 (2.4) |
| Diabetes mellitus | 488 (4.2) |
a Four race/ethnicities were missing
In comparison with women with a normal prepregnancy BMI, we found that women at the extremes of BMI were significantly more likely to have a PTB ( Table 2 ). Women who were African American were also at a significantly elevated risk of a PTB. In addition, women with a history of a PTB, cervical incompetence, or chronic hypertension had the highest risk for a PTB. Of the risk factors that were assessed, the only risk factors that were not associated with PTB were maternal age, parity, and a maternal history of an induced abortion.
| Risk factor | Incidence of preterm birth | Relative risk | P value |
|---|---|---|---|
| n/N | |||
| Body mass index categories | |||
| Severe thinness (<16.00 kg/m 2 ) | 8/33 (24%) | 2.4 | .008 |
| Moderate thinness (16.00-16.99 kg/m 2 ) | 22/94 (23%) | 2.3 | < .0001 |
| Mild thinness (17.00-18.49 kg/m 2 ) | 38/454 (8%) | 0.8 | .2 |
| Normal (18.50-24.99 kg/m 2 ) | 630/6223 (10%) | Reference | Reference |
| Preobesity (25.00-29.99 kg/m 2 ) | 299/2892 (10%) | 1.02 | .8 |
| Obese class I (30.00-34.99 kg/m 2 ) | 134/1252 (11%) | 1.1 | .5 |
| Obese class II (35.00-39.99 kg/m 2 ) | 66/464 (14%) | 1.4 | .005 |
| Obese class III (≥40.00 kg/m 2 ) | 54/314 (17%) | 1.7 | < .0001 |
| Race/ethnicity a | |||
| Hispanic | 525/5181 (10%) | 0.96 | .5 |
| Non-Hispanic white | 426/4038 (11%) | Reference | Reference |
| Other | 24/265 (9%) | 0.9 | .4 |
| African American | 216/1746 (12%) | 1.2 | .04 |
| American Indian | 10/54 (19%) | 1.8 | .06 |
| Asian | 50/438 (11%) | 1.1 | .6 |
| Maternal age, y b | 27.0 ± 6.8 | .2 | |
| Nulliparity | |||
| Yes | 547/5312 (10%) | 0.9 | .2 |
| No | 704/6414 (11%) | Reference | Reference |
| History | |||
| Spontaneous abortion | |||
| Yes | 345/2629 (13%) | 1.3 | < .0001 |
| No | 906/9091 (10%) | Reference | |
| Induced abortion | |||
| Yes | 126/1045 (12%) | 1.1 | .1 |
| No | 1125/10,677 (11%) | Reference | |
| Preterm birth | |||
| Yes | 264/997 (26%) | 2.9 | < .0001 |
| No | 987/10,729 (9%) | Reference | |
| Cigarette smoking | |||
| Yes | 224/1642 (14%) | 1.3 | < .0001 |
| No | 1027/10,080 (10%) | Reference | |
| Cervical incompetence | |||
| Yes | 26/67 (39%) | 3.7 | < .0001 |
| No | 1225/11,658 (11%) | Reference | |
| Infertility | |||
| Yes | 62/459 (14%) | 1.3 | .04 |
| No | 1189/11,265 (11%) | Reference | |
| Uterine anomalies | |||
| Yes | 68/363 (19%) | 1.8 | < .0001 |
| No | 1183/11,346 (10%) | Reference | |
| Chronic hypertension | |||
| Yes | 81/277 (29%) | 2.9 | < .0001 |
| No | 1169/11,447 (10%) | Reference | |
| Diabetes mellitus | |||
| Yes | 100/488 (20%) | 2.0 | < .0001 |
| No | 1151/11,237 (10%) | Reference |
a Four race/ethnicities were missing from the term deliveries
In Table 3 , we present the risk of the different categories of BMI that were adjusted for the other covariates for all PTB and, separately, by the clinical presentation of PTB. Severe and moderate thinness and obesity class III remained significant risk factors for overall PTB. Regarding other risk factors, the association of the other covariates with PTB remained similar to that presented in Table 2 , with the exception of nulliparity, which became a significant risk factor for PTB in the adjusted model.
| Variable | All preterm births (n = 11,691) | Spontaneous preterm labor b (n = 10,926) | Preterm premature rupture of membranes c (n = 10,777) | Medically indicated b (n = 10,870) | ||||
|---|---|---|---|---|---|---|---|---|
| aOR (95% CI) | P value | aOR (95% CI) | P value | aOR (95% CI) | P value | aOR (95% CI) | P value | |
| BMI category | ||||||||
| Severe thinness (<16.00 kg/m 2 ) | 2.6 (1.2–5.9) | .02 | 2.4 (0.8–7.0) | .1 | 2.6 (0.6–11.1) | .2 | 2.4 (0.6–10.3) | .2 |
| Moderate thinness (16.00-16.99 kg/m 2 ) | 2.2 (1.4–3.7) | .002 | 2.4 (1.2–4.7) | .01 | 1.3 (0.4–4.1) | .7 | 2.9 (1.3–6.3) | .009 |
| Mild thinness (17.00-18.49 kg/m 2 ) | 0.8 (0.6–1.1) | .2 | 0.7 (0.4–1.2) | .2 | 0.7 (0.4–1.4) | .3 | 0.9 (0.5–1.6) | .7 |
| Normal (18.50-24.99 kg/m 2 ) | Reference | Reference | Reference | Reference | Reference | Reference | Reference | Reference |
| Preobesity (25.00-29.99 kg/m 2 ) | 1.0 (0.8–1.1) | .5 | 0.9 (0.7–1.1) | .2 | 0.9 (0.7–1.2) | .6 | 1.1 (0.8–1.4) | .5 |
| Obese class I (30.00-34.99 kg/m 2 ) | 0.9 (0.8–1.2) | .5 | 0.7 (0.5–1.0) | .07 | 1.0 (0.7–1.5) | 1.0 | 1.2 (0.8–1.6) | .4 |
| Obese class II (35.00-39.99 kg/m 2 ) | 1.2 (0.9–1.6) | .3 | 0.8 (0.5–1.3) | .3 | 1.7 (1.1–2.7) | .02 | 1.1 (0.7–1.8) | .6 |
| Obese class III (≥40.00 kg/m 2 ) | 1.4 (1.0–1.9) | .04 | 0.8 (0.5–1.5) | .6 | 1.3 (0.7–2.4) | .4 | 2.2 (1.4–3.4) | .0006 |
| Race/ethnicity | ||||||||
| Hispanic | 1.0 (0.9–1.2) | .6 | 1.1 (0.9–1.4) | .5 | 1.3 (1.0–1.7) | .07 | 0.8 (0.7–1.1) | .1 |
| Non-Hispanic white | Reference | Reference | Reference | Reference | Reference | Reference | Reference | Reference |
| Other | 0.8 (0.5–1.2) | .3 | 0.7 (0.4–1.5) | .4 | 1.3 (0.7–2.6) | .5 | 0.5 (0.2–1.2) | .1 |
| African American | 1.1 (0.9–1.3) | .3 | 1.1 (0.8–1.5) | .5 | 1.5 (1.1–2.1) | .01 | 0.8 (0.6–1.1) | .2 |
| American Indian | 1.6 (0.8–3.3) | .2 | 1.3 (0.4–4.2) | .7 | 0.7 (0.1–5.1) | .7 | 2.2 (0.9–5.8) | .09 |
| Asian | 1.2 (0.9–1.3) | .3 | 1.2 (0.7–2.1) | .5 | 1.5 (0.9–2.6) | .1 | 1.0 (0.6–1.7) | .9 |
| Maternal age | 1.0 (0.99–1.01) | 1.0 | 0.96 (0.94–0.97) | < .0001 | 1.03 (1.02–1.06) | .0007 | 1.02 (1.0–1.04) | .02 |
| Nulliparity | 1.4 (1.2–1.6) | < .0001 | 1.0 (0.8–1.3) | .8 | 2.1 (1.6–2.7) | < .0001 | 1.3 (1.0–1.6) | .04 |
| History of spontaneous abortion | 1.2 (1.0–1.3) | .04 | 1.1 (0.9–1.4) | .4 | 1.3 (1.0–1.7) | .05 | 1.1 (0.9–1.4) | .2 |
| History of preterm birth | 3.4 (2.9–4.1) | < .0001 | 3.5 (2.7–4.6) | < .0001 | 3.8 (2.7–5.2) | < .0001 | 3.1 (2.3–4.1) | < .0001 |
| History of cigarette smoking | 1.2 (1.0–1.5) | .01 | 1.4 (1.1–1.8) | .004 | 1.2 (0.9–1.6) | .3 | 1.1 (0.8–1.4) | .7 |
a Also adjusted for history of induced abortion, cervical incompetence, infertility, uterine anomalies, chronic hypertension, and diabetes mellitus
b Cases of preterm birth that resulted from preterm premature rupture of membranes and medically indicated were removed from the model
c Cases of preterm birth that resulted from spontaneous preterm labor and medically indicated were removed from the model.
Also in Table 3 , we present separate logistic regression models for the 3 clinical presentations of PTB, which were adjusted for all covariates. The only BMI category that was associated with PTB from SPTL was moderate thinness. Of note, none of the obese categories were associated with this presentation of PTB. In contrast, obese class II and III were both associated with an elevated risk of PPROM, and this relationship was significant for obese class II. In the last model in Table 3 , we show the fully adjusted model for PTB that resulted from a medical indication. Moderate thinness and obese class III were categories of BMI that were associated significantly with this outcome. Compared with women with a normal prepregnancy BMI, there was no significant relationship between the mild thinness, preobesity, and obese class I categories of BMI with any PTB or with any of the clinical presentations of PTB.
A history of PTB was associated with an elevated risk for all of the clinical presentations of PTB. However, we also observed some differences in the association of risk factors across the 3 PTB presentations ( Table 3 ). We note specifically the following events: (1) Younger maternal age was associated with an elevated risk of PTB that was related to SPTL; however, older women were significantly more likely to have a PPROM or a medically indicated PTB; (2) nulliparity was a highly significant risk factor for PPROM, but this risk factor was not associated with PTB from SPTL; (3) a history of cigarette smoking was associated only significantly with PTB that was the result of SPTL, and (4) as compared with non-Hispanic white women, African American women had a significantly elevated risk of PTB that resulted from PPROM. We also found an elevated risk of PPROM among Hispanic women, but this relationship did not reach statistical significance.
Prompted by lower numbers at the extreme categories of BMI in the stratified data ( Table 4 ), we explored the data further by grouping together severe plus moderate thinness and obese classes II plus obese class III. The results of this analysis are shown in Table 4 . Women at the extremes of BMI had the greatest risk of a medically indicated PTB. Women at the lower extreme of BMI had a significantly higher risk for PTB from SPTL. In contrast, women at the upper extreme of BMI were significantly more likely to have a PPROM.
