Weight Gain During Pregnancy: Reexamining the Guidelines (2009)

As part of its approach, the committee commissioned analyses from consultants to aid in decision making by providing information not readily available in current literature. Dr. Ellen Nohr from Aarhus University, Denmark, provided analyses from the Danish National Birth Cohort on low and very high categories of gestational weight gain (GWG), as well as data for obese class I, II and III women. Additionally, she provided information on subgroups pregnant women, such as primiparous, short and young women, and smokers (see Part I). Dr. Amy Herring, University of North Carolina, provided data from the 1988 National Maternal and Infant Health Survey (NMIHS) on the association between GWG and pregnancy outcomes by race. She provided additional analyses on the association between GWG and postpartum weight retention by linking the 1988 NMIHS to its 1991 follow-up (see Part II). Dr. Cheryl Stein, Mount Sanai School of Medicine, provided data on adverse outcomes associated with GWG stratified by racial/ethnic group using births data from 1995-2003 in New York City (see Part III). Dr. James Hammitt, Harvard University, conducted a quantitative analysis of risk trade-offs between maternal and child health outcomes associated with GWG (see Part IV).

Ellen Aagaard Nohr, PhD

Associate Professor of Epidemiology

University of Aarhus, Denmark

The combined associations of prepregnancy body mass index (BMI) and gestational weight gain on pregnancy outcomes have until recent years mostly focused on birth weight. Large data collections with detailed information about maternal characteristics and pregnancy outcomes are now available which makes it possible to investigate these associations in a broader range of maternal and neonatal outcomes while adjusting for important maternal life style factors. Such a study based on the Danish National Birth Cohort (DNBC) (Nohr et al., 2008) was presented to the Committee to Reexamine IOM Pregnancy Weight Guidelines in June 2008 along with a number of analyses that focused on the BMI-specific association between GWG and all outcomes included in the study. These supplementary analyses are presented in the following in the “First DNBC Report.” At the meeting in June, the IOM committee requested new analyses for some outcomes where very low and very high categories of GWG as well as obese class I and obese class II + III were included. This work is presented in the “Second DNBC Report.” In August 2008, additional analyses were presented for the IOM committee that provided information in subgroups of women defined by parity, height, smoking and young age. These results are presented in the “Third DNBC Report.”

The Danish National Birth Cohort (DNBC) is a nationwide study of 100,419 pregnancies among 92,274 women recruited 1996-2002. More detailed descriptions of the study methods and the recruitment were previously published (Olsen et al., 2001; Nohr et al., 2006; Danish National Birth Cohort homepage, available online: http://www.ssi.dk/sw9314.asp [accessed February 2009]). Briefly, data were collected during two telephone interviews during pregnancy at approximately 16 and 30 weeks of

gestation, and two telephone interviews after birth when the child was approximately 6 and 18 months old. The women included in the cohort were mostly Caucasians as only 4 percent were born outside Scandinavia.

This study used information about 60,892 liveborn, full-term singleton (≥ 37 wk of gestation) infants whose mothers had participated in the first pregnancy and the first postpartum interview and provided information about prepregnancy BMI, GWG and postpartum weight retention 6 months after birth. In the following, the data and methods of the study will be shortly presented. A more detailed description has been published (Nohr et al., 2008).

The main exposures were prepregnancy BMI and GWG. In the first pregnancy interview, the women reported their prepregnancy weight and height, which was used to calculate their prepregnancy BMI and categorize them according to the World’s Health Organization’s definitions as underweight (BMI < 18.5 kg/m²), normal weight (18.5 ≤ BMI < 25 kg/m²), overweight (25 ≤ BMI < 30 kg/m²), and obese (BMI ≥ 30 kg/m²) (WHO, 2000). Gestational weight gain was based on information from the telephone interview 6 months after birth. At this time, the woman was asked “How much (in kg) was your total gain in pregnancy?” Her response was divided into four categories: low (< 10 kg) medium (10-15 kg), high (16-19 kg, and very high (≥ 20 kg). The medium category, which has been associated with minimum infant mortality in other populations (IOM, 1990) was used as reference.

From the first pregnancy interview, we also used information about the mother’s age at conception, parity, smoking, alcohol intake and physical exercise during pregnancy, and social status defined by education and occupation. Information about duration of breastfeeding was reported by the women in the first postpartum interview. The categorization of these variables is described in greater detail elsewhere (Nohr et al., 2008).

Pregnancy outcomes during late pregnancy included preeclampsia/eclampsia, chronic/gestational hypertension and gestational diabetes and were identified through linkage to the National Hospital Discharge Register. Because we suspected some underreporting of gestational diabetes, we added self-reported information about this disease from the pregnancy interviews.

Birth complications were also identified in the National Hospital Discharge Register and included instrumental deliveries, which in nearly all

cases covered vacuum extraction, and planned and emergency cesarean deliveries. The latter type covered cesarean section carried out when the woman was in labor.

Postpartum weight retention was calculated as the difference between the woman’s prepregnancy weight and her weight 6 months postpartum as reported in the first postpartum interview. Postpartum weight retention was summarized by two variables defined as postpartum weight loss (loss ≥ 2 kg) and postpartum weight retention (gain of ≥ 5 kg) relative to a woman’s prepregnancy weight. In the same way, postpartum weight retention at 18 months was calculated for those women in the study population who participated in the second postpartum interview, who had not given birth again and who were not pregnant again (39,776 women).

Neonatal outcomes were identified in the National Birth Register and included birth weight, length, gestational age as recorded at birth, and Apgar score after 5 minutes. Birth weight was standardized by gestational age according to the reference curve of Marsal et al. (1996). Standardized birth weight was dicotomized into either a small-for-gestational age (SGA) infant (z-score < 10th percentile) or a large-for-gestational age (LGA) infant (z-score > 90th percentile). Additionally, results for SGA defined as a z-score < 2.5th percentile and for birth weight > 4000 gram were presented.

To estimate the relative fat tissue of the infant, we calculated ponderal index of the newborn (birth weight in grams divided by the birth length in cm cubed). We defined low ponderal index as values < 10th percentile and high ponderal index as values > 90th percentile. Low Apgar score was defined as a value < 8 after 5 min.

A BMI- and GWG-specific variable was generated by cross-classifying BMI group (four categories) and GWG group (four categories). In multiple logistic regression models, the associations between this variable and pregnancy outcomes were estimated. This corresponds to the full model with an interaction term between the original BMI and GWG variables. Normal weight women with medium GWG (10-15 kg) were used as reference. These models were adjusted for a number of maternal characteristics and lifestyle factors and for gestational age at birth. In the analyses of birth complications, neonatal complications, and postpartum weight retention, women with preeclampsia and gestational diabetes were excluded (n = 1,787). In the analyses of emergency cesarean deliveries, women with a planned cesarean were excluded, and in the analyses of instrumental de-

liveries, all women with cesarean deliveries were excluded. In all adjusted models, Wald’s test with nine degrees of freedom and a significance level of 0.05 (two sided p-value) was used to assess the hypothesis that there was no effect modification by BMI group of the association between GWG and pregnancy outcomes.

Because we observed that background risks of most pregnancy outcomes increased with increasing BMI groups in a way that was not well reflected in a multiplicative model, we also used an additive approach to the data. Thus, we used the calculated odds ratios from the above models to compute 16 absolute adjusted risks for each pregnancy outcome according to each category within the BMI- and GWG-specific variable for a woman with a given set of confounder categories: She was primiparous, 25-29 years old, 1.60-1.69 m tall, reported no smoking, no alcohol intake and no exercise during pregnancy, was of high social status and gave birth after 280 days of conception. For postpartum weight retention, she breastfed < 14 weeks.

Figures G-1 through G-18 (and corresponding tables, G-1 through G-18) in this report are supplementary to the study by Nohr et al. (2008). The first 17 figures display odds ratios and adjusted absolute risks for different outcomes. In Figure G-18, the absolute risks for four important outcomes are stratified on BMI group and combined to evaluate the “trade-off” between mother and infant according to GWG:

• Figures G-1A/G-1B (Tables G-1A/G-1B): Preeclampsia

• Figures G-2A/G-2B (Tables G-2A/G-2B): Other hypertensive disorders

• Figures G-3A/G-3B (Tables G-3A/G-3B): Gestational diabetes

• Figures G-4A/G-4B (Tables G-4A/G-4B): SGA infant (< 2.5th percentile)

• Figures G-5A/G-5B (Tables G-5A/G-5B): SGA infant (< 10th percentile)

• Figures G-6A/G-6B (Tables G-6A/G-6B): LGA infant (> 90th percentile)

• Figures G-7A/G-7B (Tables G-7A/G-7B): Birth weight > 4000 g

• Figures G-8A/G-8B (Tables G-8A/G-8B): High ponderal index (> 90th percentile)

• Figures G-9A/G-9B (Tables G-9A/G-9B): Low ponderal index (< 10th percentile)

• Figures G-10A/G-10B (Tables G-10A/G-10B): Caesarean delivery before labor (planned)

• Figures G-11A/G-11B (Tables G-11A/G-11B): Caesarean delivery during labor (emergency)

• Figures G-12A/G-12B (Tables G-12A/G-12B): Instrumental deliveries

• Figures G-13A/G-13B (Tables G-13A/G-13B): Low Apgar score (< 8 after 5 minutes)

• Figures G-14A/G-14B (Tables G-14A/G-14B): Postpartum weight retention ≥ 5 kg at 6 months

• Figures G-15A/G-15B (Tables G-15A/G-15B): Postpartum weight loss ≥ 2 kg at 6 months

• Figures G-16A/G-16B (Tables G-16A/G-16B): Postpartum weight retention ≥ 5 kg at 18 months

• Figures G-17A/G-17B (Tables G-17A/G-17B): Postpartum weight loss ≥ 2 kg at 18 months

• Figure G-18A (Tables G-18A through G-18D): GWG-specific absolute risks for SGA, LGA, emergency caesarean delivery and postpartum weight retention within each BMI group

Odds ratios, displayed in the upper part of all figures, showed that:

• Except for birth weight and postpartum weight retention, pre-pregnancy BMI was by far the strongest predictor of the outcomes under study.

• There was little evidence of interaction between BMI and GWG in a multiplicative model. It was only present for birth weight and postpartum weight retention (p < 0.01), and although statistical important, it was judged to be of minor clinical importance.

In the lower part of all figures, BMI- and GWG-specific adjusted absolute risks for all included pregnancy outcomes showed that:

• Across BMI groups, background risks varied highly, which led to highly varying risk differences when moving from low to high GWG. Especially the risk of SGA and LGA were related to both increasing BMI and increasing GWG (Figures G-4 through G-7). In contrast, the absolute risk of postpartum weight retention was highly responsive to GWG, but not to BMI.

• These observations support the idea of BMI-specific recommendations. According to figure G-18, especially underweight women may benefit from very high GWG to prevent having a small infant while heavier women may benefit from avoiding high and very high GWG which only brings a slight increase of growth restriction for the infant.

At the IOM workshop in Washington, DC, in June 2008, the IOM committee found the additive approach with presentation of absolute adjusted risks across BMI groups useful and informative. They asked for additional analyses of some of the most important outcomes where one more obese class and two more GWG groups were included.

The study population and the methods for deriving the adjusted odds ratios and absolute risks were the same as for the First DNBC Report.

The BMI categories were expanded by dividing the obese group into obese class I (30 ≤ BMI > 35) and obese class II and III (BMI ≥ 35) (WHO, 2000). These categories are denoted obese and extremely obese in the figures. The GWG categories were expanded with two groups and included now the six following categories: two low categories (< 5 kg, and 5-9 kg), one medium category (10-15 kg) and three high categories (16-19 kg, 20-24 kg and ≥ 25 kg). The analyses were carried out for the following four outcomes:

• Figures G-19A/G-19B (Tables G-19A/G-19B): SGA infant (< 10th percentile)

• Figures G-20A/G-20B (Tables G-20A/G-20B): LGA infant (> 90th percentile)

• Figures G-21A/G-21B (Tables G-21A/G-21B): Emergency cesarean delivery

• Figures G-22A/G-22B (Tables G-22A/G-22B): Postpartum weight retention of ≥ 5 kg at 6 months

Finally, the results were stratified on BMI group and combined in one figure to evaluate the “trade-off” between mother and infant:

• Figure G-23: GWG-specific absolute risks for SGA, LGA, emergency cesarean delivery and postpartum weight retention within each BMI group.

• In all BMI groups, risk of SGA responded to increasing GWG throughout the entire spectrum of gain categories. The same was seen for LGA except for extremely obese women with GWG ≥ 25 kg, which did not increase risk of LGA further.

• Only for underweight, normal weight and overweight women was GWG < 5 kg associated with substantial risk of SGA.

• Extremely obese women had risks similar to obese women except for emergency cesarean delivery. Here, data indicated high and increasing risk with increasing GWG.

• The data did not suggest deleterious consequences of GWG < 5 kg in obese and extremely obese women.

Because the 1990 IOM Guidelines did not provide sufficient data about GWG in subpopulations of interest, the committee requested additional information about subgroups of pregnant women, defined by parity, height, age, and smoking. Also, the committee asked for analyses of the association between GWG and emergency cesarean delivery and postpartum weight retention with and without adjustment for birth weight. These results are presented in the Third DNBC Report. The methods and analyses are presented below. More details are available in Nohr et al. (2009).

The initial study population was similar to the one used in the First and Second DNBC Report. However, in this study, women < 18 y old (n = 71) were included and women with diagnosed preeclampsia (n = 1,118) and gestational diabetes (n = 690) were excluded. As a result, there were 59,147 women in the final study population.

The main exposures, self-reported prepregnancy BMI and GWG, were defined in the same way as in the previous reports. For this report, BMI was categorized into four categories: underweight (BMI < 18.5 kg/m²), normal weight (18.5 ≤ BMI < 25 kg/m²), overweight (25 ≤ BMI < 30 kg/m²), and obese (BMI ≥ 30 kg/m²). Gestational weight gain was divided into six categories: two low categories (< 5 kg, and 5-9 kg), one medium category (10-15 kg) and three high categories (16-19 kg, 20-24 kg and ≥ 25 kg). Other covariates in the models were similar to those used in previous analyses.

Pregnancy outcomes used in these analyses were:

• Small-for-gestational-age infant (< 10th percentile)

• Large-for-gestational-age infant (> 90th percentile)

• Emergency cesarean delivery

• Postpartum weight retention of ≥ 5 kg at 6 months

Information about the definition of these outcomes can be found in the First DNBC report.

A BMI- and GWG-specific variable was generated by cross-classifying BMI group (four categories) and GWG group (six categories). Few underweight women reported a low GWG, so for this group the two lowest categories were combined into one category, which was defined as a gain < 10 kg. Thus, this BMI- and GWG-specific variable consisted of 23 categories.

We divided the study population into primiparous (n = 27,030) and multiparous (n = 32,117) women to investigate the associations among BMI, GWG and selected pregnancy outcomes within each of these strata. In multiple logistic regression models, the BMI- and GWG-specific variable and the covariates of age, height and smoking (yes/no) were mutually adjusted to estimate their independent associations with the pregnancy outcomes of interest within each of these subpopulations. The reference category was defined as normal weight women with a medium GWG (10-15 kg), 25-29 years old at conception, height of 1.60-1.69 m who did not smoke during pregnancy. In these models, we also adjusted for alcohol consumption and exercise in pregnancy, social status, and gestational age at delivery in days. In the analysis of postpartum weight retention, duration of breastfeeding was added to the model. In the analysis of emergency cesarean delivery, we excluded women who had a cesarean section before labor (1,485 primiparous and 2,429 multiparous women).

Within the groups of primiparous and multiparous women, we used the calculated odds ratios from these models to compute absolute adjusted risks for pregnancy outcomes according to each category within the BMI- and GWG-specific variable (which produced 23 different absolute risks for each pregnancy outcome). This was done for four different sets of characteristics among primiparous women and three different sets among multiparous women, which created a total of seven different types of women. In each of these models, “a reference woman” was 25-29 years old, 1.60-1.69 m tall and did not smoke or consume alcohol during pregnancy. This woman, which will be denoted “an unexposed woman” in the following, performed a moderate amount of exercise during pregnancy, was of high social status and had a gestational length of 280 days. For postpartum weight reten-

tion, she breastfed < 14 weeks. The same characteristics applied for “a short woman,” only she was < 1.60 m tall. “A smoking woman” was also defined as a reference woman, only she was a smoker. Among primiparous women, we also defined “a young woman,” who was similar to the reference woman, only was she < 20 years old.

First, the absolute risks are presented in seven figures, one for each subtype of woman, to evaluate if the “trade-off” between mother and infant differed across different types of women. Every figure is accompanied with a table with estimates and 95% confidence intervals corresponding to all points in the figure:

• Figure G-24 (Table G-23): Unexposed primiparae, GWG-specific risks of pregnancy outcomes

• Figure G-25 (Table G-24): Short primiparae, GWG-specific risks of pregnancy outcomes

• Figure G-26 (Table G-25): Smoking primiparae, GWG-specific risks of pregnancy outcomes

• Figure G-27 (Table G-26): Young primiparae, GWG-specific risks of pregnancy outcomes

• Figure G-28 (Table G-27): Unexposed multiparae, GWG-specific risks of pregnancy outcomes

• Figure G-29 (Table G-28): Short multiparae: GWG-specific risks of pregnancy outcomes

• Figure G-30 (Table G-29): Smoking multiparae: GWG-specific risks of pregnancy outcomes

To evaluate the differences between subtypes of women within each BMI group, these results were also combined in four new figures, one for each BMI group:

• Figure G-31: GWG-specific risk of pregnancy outcomes in subtypes of underweight women

• Figure G-32: GWG-specific risk of pregnancy outcomes in subtypes of normal weight women

• Figure G-33: GWG-specific risk of pregnancy outcomes in subtypes of overweight women

• Figure G-34: GWG-specific risk of pregnancy outcomes in subtypes of obese women