Introduction

The prevalence of obesity has been increasing rapidly in the United States [1, 2]. The rates of Class I Obese (body mass index [BMI (kg/m2)]: ≥30.0 to <35.0) and Class II Obese (BMI ≥35.0 to <40.0) have doubled in the past 30 years in women of reproductive age in this country, while the rate of Class III Obese (BMI ≥ 40.0) has tripled in women of reproductive age over the same period [3]. Approximately 20 % of women are obese (BMI ≥ 30.0) when they begin pregnancy [4]. Excessive gestational weight gain in obese women has been found to be associated with increased risk of large-for-gestational-age (LGA) infants [5–7], while minimal gestational weight gain in obese women has been found to be associated with increased risk of small-for-gestational-age (SGA) infants [6, 8], among other adverse maternal and fetal outcomes. Large-for-gestational-age infants have greater risk of cesarean delivery, operative delivery, and hypoglycemia, and childhood obesity and metabolic syndrome later in life [9–11]. Small-for-gestational-age infants, particularly those that are born preterm, have greater risk of neurologic dysfunction, diminished intellectual capacity, attention deficit hyperactivity disorder, cardiovascular complications, metabolic syndrome, and shorter height later in life [12–16]. The 1990 Institute of Medicine (IOM) guidelines recommended that obese women gain at least 15 pounds during pregnancy, but did not establish an upper limit of weight gain or provide different recommendations by obesity class [17]. The 2009 IOM guidelines retained the same scientific approach and epidemiologic conventions used previously in the 1990 IOM report, but recommended that obese women gain 11–20 pounds during pregnancy and did not provide specific obesity class recommendations due to lack of supporting data [3]. The 2009 IOM recommendation has been called into question in view of the worsening obesity epidemic [18]. The association of gestational weight gain with birth weight in obese pregnant women has been extensively studied using the 1990 IOM guidelines [7, 19–26] and to a lesser extent using the 2009 IOM guidelines [27–32]. The results have varied due to different study populations, different definitions of SGA infants, inadequate control of confounding, or lack of findings by obesity class. Women with pre-existing and gestational diabetes mellitus are at greater risk for LGA infants and the association of gestational weight gain with birth weight could be different according to diabetic status [33, 34]. The purpose of this study was to estimate the association of gestational weight gain with birth weight in obese pregnant women by obesity class and diabetic status using the revised 2009 IOM guidelines.

Methods

A population-based historical cohort study was conducted using data from the Missouri maternally linked birth and fetal death certificate registry. The study population included all obese pregnant women (BMI ≥ 30.0 kg/m2) residing in Missouri who delivered liveborn, singleton, term (clinical estimate of gestational age ≥39 weeks) infants without congenital abnormalities in 2002–2008. Pre-pregnancy BMI was calculated from self-reported weight and height recorded on the birth certificate. The following classes of obesity were defined according to NIH guidelines: Class I Obese (≥30.0 to <35.0 kg/m2), Class II Obese (≥35.0 to <40.0 kg/m2), and Class III Obese (≥40.0 kg/m2). Gestational weight gain was abstracted from the medical chart or provided by the physician. Five categories of gestational weight gain were examined: ≤2 pounds (includes women who lost weight), 3–10 pounds, 11–20 pounds, 21–35 pounds, and >35 pounds. The 11–20 pound group served as the reference category for all comparisons, as it is the current IOM recommendation for weight gain in pregnancy for obese women. The ≤2 pounds group represented women maintaining pre-pregnancy weight or losing weight during pregnancy, while the 3–10 pound group represented women who gained a minimal amount of weight that was less than the IOM recommendation. The 21–35 pound group and the >35 pound group represented clinically relevant categories of women whose weight gain exceeded the IOM recommendation.

The two outcome variables of this study were SGA infants and LGA infants, defined as birth weight below the 10th percentile and above the 90th percentile, respectively, for gestational age and gender [35]. Potential confounders that could impact the risk of SGA or LGA infants that were adjusted for in the study included maternal age (≤25, 26–35, >35 years), race (non-Hispanic white, black, Hispanic, other), socioeconomic status (using enrollment in Medicaid; the Women, Infants, and Children public health program; or food stamp programs as a proxy, since income was not reported on the birth certificate), smoking, parity (0, 1+ prior live births), cardiac disease, renal disease, chronic hypertension, and preeclampsia. Preeclampsia was defined on the birth certificate as pregnancy-induced hypertension occurring >20 weeks gestation that was characterized by either blood pressure ≥140/90 mmHg or an increase in systolic blood pressure ≥30 mmHg or an increase in diastolic blood pressure ≥15 mmHg over baseline on two measurements taken 6 h apart that was accompanied by generalized and overt proteinuria or edema. Eclampsia was defined as preeclampsia accompanied by convulsions or coma. Women who developed eclampsia were included in the preeclampsia variable.

The relationship of gestational weight gain categories with SGA and LGA infants was examined for diabetic and non-diabetic women within each obesity class. Diabetes was defined on the birth certificate as insulin-dependent diabetes or other diabetes. The latter could be either pre-existing or gestational diabetes mellitus. Women were combined into a single diabetes group for purposes of analysis. Categorical variables were expressed as numbers and percentages. Differences in demographic characteristics and medical/obstetrical history were examined across obesity classes by Chi square test. The incidence of SGA infants by weight gain categories during pregnancy was assessed using Chi square test. The multivariate predictability of weight gain categories during pregnancy for an SGA infant was examined using multiple logistic regression. Multivariate findings reflect adjustment for the above potential confounders in the model. Parallel analyses were performed for LGA infants. A P value of <0.05 was used to denote statistical significance. All analyses were performed using SPSS 18.0 (SPSS, Chicago, IL, USA).

This research did not involve interaction with subjects and the existing publicly available data contained no identifying private information. This study was deemed not to require formal review according to federal regulation 45 CFR 46.102.

Results

The study population of 66,010 women consisted of 36,568 (55.4 %) Class I Obese women, 17,195 (26.0 %) Class II Obese women, and 12,247 (18.6 %) Class III Obese women (Table 1). Approximately 75 % of the women were non-Hispanic white. Over a third of the mothers were nulliparous. Class III Obese women were older; poorer; less likely to be non-Hispanic white or nulliparous; had higher rates of diabetes, chronic hypertension, preeclampsia, and LGA infant; and gained less weight during pregnancy than Class I or Class II Obese women. No significant differences were found for smoking, cardiac disease, renal disease, or SGA infant by obesity class. Approximately 80 % of the women received adequate or adequate plus prenatal care. Only 23.3 % of the women followed the Institute of Medicine’s recommended weight gain during pregnancy of 11–20 pounds. The percentage of women gaining ≤10 pounds during pregnancy was 16.8 %, while 59.9 % of the women gained >20 pounds. The number of women who lost weight in the ≤2 pound group was 1,822/4,447 (41.0 %), with a range of weight loss of 1–92 pounds and a median weight loss of nine pounds. The majority of women with diabetes were not insulin-dependent (78.6 %). Small-for-gestational-age infants occurred in 5.2 % of the births while LGA infants occurred in 16.6 % of the births.

Table 1 Demographic/medical/obstetrical data for 66,010 obese pregnant women in Missouri, 2002–2008
Full size table

The incidence and multivariate findings for SGA infants and LGA infants by weight gain during pregnancy are provided by obesity class and diabetic status in Tables 2, 3 and 4. A significant decrease in incidence of SGA infants with increasing weight gain was found only for non-diabetic women regardless of obesity class (Tables 2, 3, 4). A significant increase in incidence of LGA infants with increasing weight gain was found for all obesity classes regardless of diabetic status (Tables 2, 3, 4). Generally the incidence of LGA infants was much higher than the incidence of SGA infants at a particular level of weight gain within any obesity class/diabetic status combination. Multivariate findings indicated that non-diabetic women who gained ≤2 pounds had a significantly increased risk of an SGA infant compared to non-diabetic women who gained 11–20 pounds in all obesity classes. Diabetic women who gained ≤2 pounds did not have a significantly increased risk of an SGA infant compared to diabetic women who gained 11–20 pounds in any obesity class. Class I Obese non-diabetic women who gained 3–10 pounds had significantly higher odds of an SGA infant than Class I Obese non-diabetic women who gained 11–20 pounds [adjusted odds ratio (OR) 1.28, 95 % confidence interval (CI) 1.07, 1.52, Table 2]. A significantly increased risk of an SGA infant was not found for women who gained 3–10 pounds compared to women who gained 11–20 pounds for any other obesity class/diabetic status combination, despite elevated risk in all groups (Tables 2, 3, 4).

Table 2 Incidence and multivariate risk of SGA and LGA infants by pregnancy weight gain for Class I obesea pregnant women in Missouri, 2002–2008, by diabetic status
Full size table
Table 3 Incidence and multivariate risk of SGA and LGA infants by pregnancy weight gain for Class II obesea pregnant women in Missouri, 2002–2008, by diabetic status
Full size table
Table 4 Incidence and multivariate risk of SGA and LGA infants by pregnancy weight gain for Class III Obesea pregnant women in Missouri, 2002–2008, by diabetic status
Full size table

Weight gains less than 11–20 pounds were significantly associated with a decreased risk of an LGA infant in 5/6 obesity class/diabetic status combinations. A significantly decreased risk of an LGA infant was found for non-diabetic women in all obesity classes who gained ≤2 pounds or 3–10 pounds compared to non-diabetic women who gained 11–20 pounds (Tables 2, 3, 4). A significantly decreased risk of an LGA infant was found for Class II Obese diabetic women who gained ≤2 pounds compared to Class II Obese diabetic women who gained 11–20 pounds (adjusted OR 0.57, 95 % CI 0.33, 0.97, Table 3) and for Class III Obese diabetic women who gained ≤2 pounds or 3–10 pounds compared to Class III Obese diabetic women who gained 11–20 pounds (adjusted OR 0.52, 95 % CI 0.34, 0.81; adjusted OR 0.59, 95 % CI 0.38, 0.90, respectively, Table 4).

Discussion

This study focuses on the association of gestational weight gain with birth weight in obese pregnant women by obesity class and diabetic status using the revised 2009 IOM guidelines. Our findings indicate that women in 5/6 obesity class/diabetic status combinations who gained less weight than the current IOM recommendation of 11–20 pounds were not at significantly increased risk of an SGA infant. This minimal amount of weight gain was 3–10 pounds. These findings are consistent with those of other studies involving obese pregnant women [23, 36]. The exception of Class I Obese non-diabetic women who gained 3–10 pounds having 1.28 times the odds of an SGA infant than women who gained 11–20 pounds was offset by a significantly reduced odds ratio of 0.66 of an LGA infant. While non-diabetic women who gained ≤2 pounds had a significantly increased risk of an SGA infant compared to women who gained 11–20 pounds in any obesity class, diabetic women who gained ≤2 pounds did not have a significantly increased risk of an SGA infant compared to women who gained 11–20 pounds in any obesity class. These findings suggest that in at least Class II and Class III Obese non-diabetic women, a weight gain of 3–10 pounds could be considered, while diabetic women, regardless of obesity class, could be considered for maintenance of pre-pregnancy weight or even weight loss during pregnancy. Such levels await corroboration from prospective studies.

The risk of LGA infants was much higher than the risk of SGA infants in obese women. The incidence of LGA infants generally ranged from 1 to 30 times higher than the incidence of SGA infants at a particular level of weight gain within any obesity class/diabetic status combination with differences becoming more pronounced with increasing obesity and the presence of diabetes. Weight gains less than 11–20 pounds were significantly associated with a decreased risk of an LGA infant in 5/6 obesity class/diabetic status combinations. Our findings largely support the dual benefit that lower gestational weight gain is not significantly associated with an increased risk of an SGA infant but is significantly associated with a decreased risk of an LGA infant.

Our findings are largely consistent with those of other studies that have estimated the association of gestational weight gain with birth weight in obese pregnant women using the revised 2009 IOM guidelines [27–32]. Durie et al. [27] found that lower than IOM recommended rates of second and third trimester weight gain were not associated with a significant increase in risk of an SGA infant, despite elevated risks in all three obesity classes. Weight loss was associated with a significant increase in risk of an SGA infant only in Class I Obese women. Blomberg [28] found that weight gain below the IOM recommended 11–20 pounds was associated with a significant increase in risk of an SGA infant in Class I and Class II Obese women. Weight loss resulted in significantly increased risk of SGA infants in Obese Classes I and III. Vesco et al. [29] found that either weight gain below the IOM recommended 11–20 pounds or weight loss was associated with significantly greater risk of an SGA infant, but provided no findings by obesity class. Park et al. [30], using a state-wide birth certificate registry for Florida from 2004 to 2007, had very similar findings to our study in that weight gain below the IOM recommended 11–20 pounds was associated with a significant increased risk of an SGA infant in Obese Class I women, but not in Obese Class II or Class III women, despite elevated risks in both groups . Women who lost weight had significantly increased risk of an SGA infant in all three obesity classes. Hinkle et al. [31] found that weight gain below the IOM recommended 11–20 pounds was associated with a significant increase in SGA risk in all obesity classes when the traditional definition of birth weight below the 10th percentile for gestational age was used for SGA infants, but was not associated with a significant increase in SGA risk in any obesity class when two standard deviations and below the mean infant birth weight at a particular gestational age was used to define SGA infants. Bodnar et al. [32] found that weight gain below the IOM recommended 11–20 pounds was associated with significantly increased risk of an SGA infant in all three obesity classes.

Differences in the above study findings are partly reflective of different definitions for SGA infants. The traditional definition of birth weight below the 10th percentile for gestational age was used in four studies [27, 29, 30, 32], the more restrictive definition of two standard deviations and below the mean infant birth weight at a particular gestational age was used in one study [28], and both definitions were used and compared in one study [31]. The latter study indicated significant or non-significant findings for weight gain below the IOM recommended 11–20 pounds across all obesity classes depending on which definition of SGA infants was used. Estimation of SGA risk as defined as the 10th percentile and below for gestational age involves many neonates who may not be at increased morbidity risk despite being constitutionally small [37, 38]. Infants whose birth weight is two standard deviations and below the mean infant birth weight at a particular gestational age have the highest risks of morbidity and mortality and estimating their occurrence with reduced gestational weight gain in obese women may be desirable [37, 39]. The low numbers of SGA infants in diabetic women that would result from such a definition did not allow this to be examined and was a limitation of our study.

Strengths of our study include a large cohort that enabled estimation of the association of gestational weight gain with birth weight in obese pregnant women by obesity class and diabetic status, a historical cohort design which permitted calculation of SGA and LGA incidence rates, and the availability of and adjustment for many potential confounding variables that could impact SGA and LGA infant risk. Limitations of our study include those that are inherent with the use of birth certificate records, such as incomplete or inaccurate reporting of data; self-reported information, such as height and weight; and potential under-reporting of medical conditions, such as chronic hypertension, heart disease, and renal disease. Preeclampsia has been found to match very well between birth certificates with a check-box format (such as used in Missouri), and hospital discharge data, with 85–97 % of the former being present in the latter [40]. Self-reported maternal weight, as appears on Missouri birth certificates and on which pre-pregnancy BMI classes were based, also has been found to be highly correlated with and similar to clinically recorded weight [41]. Self-reported BMI has been found to match very well with clinically measured BMI in obese [42] and combined overweight and obese samples [43]. Gestational age was determined by clinical estimate, which has been found to be a more accurate measure of gestational age at delivery than length of pregnancy calculated using the last menstrual period [44]. Most of the diabetic women in our study were not insulin-dependent (78.6 %). Since the birth certificate only listed insulin-dependent diabetes and other diabetes, we could not distinguish the number of women with pre-existing from gestational diabetes mellitus. Our findings in diabetic women need corroboration from prospective studies of diabetic women, both pre-existing and gestational. Another study limitation was the relatively small number of SGA infants in obese diabetic women, particularly in the ≤2 pounds gestational weight gain category. While this reflects that LGA infants are simply far more common than SGA infants in obese diabetic women, the consideration that obese diabetic women could maintain pre-pregnancy weight or even lose weight during pregnancy should be interpreted with caution. Our study did not examine the degree of weight loss on SGA infant risk or provide any lower limit of weight loss above which obese diabetic women would not be at significantly increased risk of an SGA infant. Our study further only examined SGA infants and LGA infants. The impact of gestational weight gain on other adverse short- and long-term maternal and infant outcomes such as cesarean delivery, instrumental delivery, low Apgar scores, infant mortality, childhood obesity, childhood cognitive performance, and post-partum weight retention by obesity class and diabetic status remains to be determined. Our study population consisted of liveborn, singleton, term (≥39 weeks gestation) infants free of congenital abnormalities and our findings may not be generalizable to other obese pregnant populations. The characteristics of our study population may further limit generalizability to other racial groups, socioeconomic levels, and geographical locations. Other unknown confounders, such as diet and physical activity, may play a role in the risk patterns that were observed.

Studies have shown that medical conditions, such as pre-existing diabetes mellitus and gestational diabetes mellitus, may be reported on birth certificates only between 42.1 and 64.3 % of the time when verified against the medical record [45, 46]. This may not bias study findings if the under-reporting is random, but could bias study findings if the under-reporting is differential [47]. The degree of under-reporting of diabetes in our study may be lower due to an obese study population, obese women who receive prenatal care routinely are screened for gestational diabetes mellitus as part of standard of care, and diabetes is a risk factor for an LGA infant. Indeed, the risk patterns for LGA infants in diabetic women (as well as non-diabetic women) showed remarkable consistency in our study, increasing with increasing gestational weight gain and increasing obesity class. Under-reporting of diabetes for SGA infants may be more likely in our study since diabetes may not be viewed as a risk factor for the outcome. Such under-reporting could be accentuated by other factors, such as women of minority races and low socioeconomic status not having accessibility to prenatal care. Validation studies that assess the degree of under-reporting of diabetes on birth certificates compared to the medical record for SGA infants and LGA infants by level of obesity and gestational weight gain, as well as other maternal characteristics, are needed.

Our study supports that gestational weight gain appears to be less important in guarding against an SGA infant in obese diabetic women than in obese non-diabetic women. Potential metabolic mechanisms by which diabetes may lower the risk of an SGA infant and increase the risk of an LGA infant in obese women include fetal overgrowth as a direct result of the maternal hyperglycemia/fetal hyperglycemia/fetal hyperinsulinemia pathway, increased insulin resistance which raises maternal free fatty acids and triglyceride levels resulting in increased fetal fat mass and higher birth weights, and poor glycemic control, particularly in the third trimester [48–50].

Recently, the American College of Obstetricians and Gynecologists has recognized that “For an obese pregnant woman who is gaining less weight than recommended but has an appropriately growing fetus, no evidence exists that encouraging increased weight gain to conform with the updated IOM guidelines will improve maternal or fetal outcomes” [51]. Our study found that many obese women who gained less weight during pregnancy than the current IOM recommendation of 11–20 pounds were not at significantly increased risk of an SGA infant but were at significantly decreased risk of an LGA infant. New weight gain levels for consideration based on our findings are 3–10 pounds for Class II and Class III Obese non-diabetic women and maintenance of pre-pregnancy weight or even weight loss during pregnancy for diabetic women regardless of obesity class. These suggested levels await corroboration from prospective studies involving other obese pregnant populations of diabetic and non-diabetic women that estimate the association of gestational weight gain with birth weight and with other short- and long-term maternal and infant outcomes.