Contribution of overweight and obesity to the occurrence of adverse pregnancy outcomes in a multi-ethnic cohort: population attributive fractions for Amsterdam

Authors


Dr TGM Vrijkotte, Department of Public Health, Academic Medical Centre, P.O. Box 22660, 1100 DD Amsterdam, the Netherlands. Email T.Vrijkotte@amc.uva.nl

Abstract

Please cite this paper as: Djelantik A, Kunst A, van der Wal M, Smit H, Vrijkotte T. Contribution of overweight and obesity to the occurrence of adverse pregnancy outcomes in a multi-ethnic cohort: population attributive fractions for Amsterdam. BJOG 2011; DOI: 10.1111/j.1471-0528.2011.03205.x.

Objective  To estimate the contribution of pre-pregnancy excessive weight to the occurrence of adverse pregnancy outcomes and to detect the differences in these contributions between different ethnic groups.

Design  Prospective multi-ethnic community-based cohort study.

Setting  The prevalence of excessive weight is increasing and in general higher in immigrant groups in many industrialised countries. Maternal excessive weight, like smoking during pregnancy, is an important risk factor for adverse pregnancy outcomes.

Population  A total of 8266 pregnant women, living in the Netherlands, were included in the ABCD study between January 2003 and March 2004.

Methods  After applying the exclusion criteria, the analysis included 7871 pregnancies. Binomial log-linear regression analyses were performed to estimate relative risks (RRs) expressing the association between overweight/obesity and small-for-gestational-age (SGA), large-for-gestational-age (LGA), preterm birth (PTB; <37 weeks of gestation) and extreme PTB (<32 weeks of gestation), controlling for parity, maternal age, education level and smoking. Next, the RRs were used to estimate population attributive fractions (PAF) for Amsterdam and separately for several ethnic groups.

Main outcome measures  The RRs and PAFs.

Results  The PAFs for overweight/obesity were: SGA −4.9%, LGA 15.3%, PTB 6.6% and extreme PTB 22.0%. In absolute terms, this corresponds to −47 SGA infants, 126 LGA infants, 35 PTB and 20 extreme PTB per year in Amsterdam. Except for SGA, these PAFs were higher than those for smoking (6.2%, −3.9%, 5.5% and 10.6%, respectively). The contribution of overweight/obesity to LGA and PTB was higher in non-Western immigrant groups.

Conclusions  Overweight/obesity has become an important contributor to the occurrence of adverse pregnancy outcomes in Amsterdam. For most outcomes, these contributions are larger than those for smoking. Development of special obesity prevention programmes for young women is required, especially focused on immigrant groups.

Introduction

The increasing prevalence of overweight and obesity among young women in most countries1–5 underlines its large potential impact on adverse pregnancy outcome. During pregnancy, maternal overweight/obesity increases the risk of serious complications, such as gestational diabetes, hypertension and pre-eclampsia, which lead to increased risk for preterm birth (PTB), reduced or augmented fetal growth and stillbirth.6–8 As the obesity epidemic evolves, it is important to monitor the extent to which overweight/obesity contributes to the occurrence of these adverse pregnancy outcomes.

Population attributive fraction (PAF) calculations have been used to measure the proportion of the adverse outcome that is attributable to the occurrence of the risk factor in the population. When applied to smoking, the PAF revealed a substantial impact of smoking on the occurrence of adverse pregnancy outcomes in most Western populations.9–11 Recognition of this large impact has led to increasing awareness among the medical profession and pregnant women for the need to prevent smoking during pregnancy. In 2002, Cnattingius et al.12 warned that overweight and obesity, because of their increasing prevalence, could parallel smoking as a contributor to the occurrence of adverse pregnancy outcomes. However, no quantitative comparisons have yet been published.

The contribution of maternal overweight/obesity to adverse pregnancy outcomes has been examined in very few populations. Three studies have focused on the contribution of maternal obesity to the occurrence of caesarean section, pregnancy-induced hypertension, and large-for-gestational-age (LGA) infants.2,13,14 Lu et al.2 showed that in Alabama (USA) the PAF for LGA infants increased from 6.5% in 1980–84 to 19.1% in 1995–99, illustrating the rising public health impact of obesity for the health of newborns. No comparable estimates have been published for Western European countries.

The prevalence of overweight/obesity in Europe is higher in immigrant groups than in native groups in general,15–17 especially among pregnant women.3 Moreover, the effect of maternal overweight/obesity on adverse pregnancy outcomes is suggested to be stronger in some non-Western immigrant groups.18,19 Hence, obesity/overweight might contribute particularly to the occurrence of adverse pregnancy outcomes among newborn children in these groups. To our knowledge, no study has investigated ethnic differences in the contribution of pre-pregnancy overweight and obesity to major pregnancy outcomes.

The aim of this study is to evaluate the contribution of women’s pre-pregnancy overweight/obesity to adverse pregnancy outcomes (small-for-gestational-age [SGA] infants, LGA infants, PTB and extreme PTB) in a multi-ethnic population. The PAF approach is applied to different ethnic groups, taking into account ethnic differences in overweight/obesity prevalence and in the associations between overweight/obesity and pregnancy outcomes. The second aim of this study is to assess whether the magnitude of these contributions varies between ethnic groups. Finally, the PAFs for overweight/obesity are compared with those of the well-documented risk factor ‘smoking’. All estimates are based on a large prospective community-based birth cohort in Amsterdam.

Methods

Study population and design

The present study is part of the Amsterdam Born Children and their Development (ABCD) study (http://www.abcd-study.nl). The ABCD study is a prospective, representative and community-based cohort study, which examines the relationship between maternal lifestyle and psychosocial conditions during pregnancy and the child’s health at birth and in later life, with specific focus on ethnic differences. Details of the study design have been described earlier.20

In short, between January 2003 and March 2004, all pregnant women living in Amsterdam were invited to enrol in the ABCD study at their first prenatal visit (median 13 weeks; interquartile range 12–14 weeks of gestation) to an obstetric care provider. Two weeks after their first prenatal visit, a pregnancy questionnaire was mailed to all approached women (n = 12 373), to be returned by prepaid mail. The questionnaire covered socio-demographic data, lifestyle and (psychosocial) health and was available in Dutch as well as in English, Turkish or Arabic depending on the woman’s country of birth. Reminders were sent 2 weeks after the initial mailing to improve response. Date of birth, birthweight, infant sex and gestational age (based on ultrasound or, if unavailable [<10%], on the first day of the last menstrual period), as recorded by the obstetric care provider, were obtained from the Youth Health Department at the Public Health Service in Amsterdam.

The questionnaire was filled out by 8266 women (response rate 67%). For this study, only women with a pregnancy duration >24 weeks and singleton births were included (n = 7871). Approval of the study was obtained from the Central Committee on Research involving Human Subjects in the Netherlands, the Medical Committees of participating hospitals, and from the Registration Committee of the Municipality of Amsterdam. All participants gave written informed consent.

Outcome variables

The primary outcome variables were SGA infants, LGA infants, PTB (gestational age <37 weeks) and extreme PTB (gestational age <32 weeks). SGA and LGA were defined as a birthweight below the 10th percentile (SGA) or above the 90th percentile (LGA) for gestational age on the basis of sex-specific and parity-specific standards of the Netherlands Perinatal Register.21

Determinants

Information on the determinants and covariates was obtained from the questionnaire. Maternal pre-pregnancy body mass index (pBMI) was based on self-reported height and weight and categorised in four classes according to the WHO standards: underweight: <18.5, normal weight: 18.5–24.9, overweight: 25–29.9, obesity: ≥30 kg/m2. A random imputation procedure using linear regression analysis was used to complete missing data on maternal weight (9.9% missing) or maternal height (3.8% missing).22 Smoking during pregnancy was categorised into: none or stopped smoking, <1, 1–5, >5 cigarettes a day.

Covariates

We used the educational level of the mother (<5, 5–10, >10 years of education after primary school), age of the mother (<17, 18–29, 30–39, >40 years) and parity (0, ≥1) as covariables.

Ethnicity was based on the country of birth of the pregnant women and included the following categories, based on the Amsterdam main ethnic populations: Dutch, African descent (Antillean/Aruban, Ghanaian and Surinamese of African descent), Turkish, Moroccan, Other non-Western and Other Western countries.

Diabetes mellitus and hypertensive status were not considered as confounders because these complications are mediators in the association between maternal overweight/obesity and the pregnancy outcomes. Because the present study aimed to quantify the full impact of maternal overweight/obesity, adjusting for these intermediate factors would result in overcorrection.

Statistical analyses

Binomial log-linear regression analysis was performed to estimate the relative risks (RR) of overweight/obesity and smoking during pregnancy for SGA infants, LGA infants, PTB and extreme PTB in the total study population, adjusted for maternal age, parity, education level of the mother, ethnicity, smoking during pregnancy (in the overweight/obesity analyses) and overweight/obesity (in the smoking analyses). To test for possible ethnic differences in the association between overweight/obesity and LGA and PTB, the analyses were stratified by ethnicity and formally tested for interaction by adding the ethnicity-to-overweight/obesity interaction term to the model.

The PAF of smoking and overweight/obesity for all defined outcome measures was estimated for the total study population, and for each ethnic group. The following formula was used:

image

where Pi is the proportion of women in the different categories (denoted by i) of overweight/obesity or smoking during pregnancy and RRi is the relative risk of the birth outcomes associated with the exposure category (i). The normal pre-pregnancy weight group (BMI 18.5–25 kg/m2) and the non-smoking group were the reference groups for the RR values.

To calculate the PAFs for Amsterdam and the absolute number of affected children per year, a correction was made for the response differences between the ethnic groups.23 The ethnic distribution of all pregnant women in the year 2003 (singleton, pregnancy duration > 24 weeks; n = 11 697) was obtained from the Youth Health Care registration.

Results

In the total study population, the prevalence of overweight and obesity was 17.3% and 6.3%, respectively (Table 1). Obesity was much more common among Turkish women (P < 0.001), Moroccan women (P < 0.001) and women of African descent (P < 0.001). On the other hand, smoking during pregnancy was less frequent in immigrant groups, except for Turkish women, of whom 16.9% continued to smoke during pregnancy.

Table 1.   Characteristics of the study population in the ABCD study
 Study populationEthnicity
DutchAfrican descentTurkishMoroccanOther non-WesternOther Western
Number of women78714900495328585904659
Outcome variables
SGA, %9.98.217.511.010.614.39.6
LGA, %9.09.86.18.07.98.47.7
PTB, %
 <37 weeks5.55.111.74.93.95.95.3
 <32 weeks0.90.82.40.90.71.10.6
Determinant variables
Maternal pBMI, %
 Underweight (<18.5)5.04.73.44.62.67.47.0
 Normal weight (18.5–25)71.477.447.761.751.463.478.5
 Overweight (25–30)17.313.730.922.733.421.311.1
 Obesity (>30)6.34.118.011.012.77.93.5
Smoking, %
Non smoking90.588.893.383.197.495.192.9
 <1 cigarette/day2.53.12.04.30.31.71.2
 1–5 cigarettes/day3.64.32.26.21.22.02.3
 >5 cigarettes/day3.33.82.46.51.01.23.6

RRs in the total study population

Table 2 shows that overweight and obesity were associated with an increased risk of delivering an LGA infant, but with a decreased risk of delivering an SGA infant. Obesity was also associated with an increased risk of PTB as well as extreme PTB. Levels of smoking during pregnancy were associated with an elevated risk of delivering an SGA infant in a dose–response manner. Smoking (>5 cigarettes per day) was associated with an increased risk of PTB and extreme PTB, but with a decreased risk of delivering an LGA infant.

Table 2.   Relative risks for pBMI and smoking during pregnancy on pregnancy outcomes: SGA infants, LGA infants, PTB and extreme PTB
 SGALGAPTBExtreme PTB
Univar. RRMultivar. RR (95% CI)*Univar. RRMultivar. RR (95% CI)*Univar. RRMultivar. RR (95% CI)*Univar. RRMultivar. RR (95% CI)*
  1. *Adjusted for education level, maternal age, ethnicity, parity and, respectively, smoking or pBMI.

  2. Univar., univariable; Multivar. multivariable.

pBMI
pBMI 18.5–25Ref.Ref.Ref.Ref.Ref.Ref.Ref.Ref.
pBMI 25–300.990.83 (0.69–1.00)1.401.55 (1.30–1.84)1.171.17 (0.91–1.49)1.741.82 (1.00–3.32)
pBMI > 301.090.81 (0.61–1.07)1.712.03 (1.60–2.59)1.631.51 (1.09–2.09)2.692.71 (1.24–5.78)
Smoking
NonsmokingRef.Ref.Ref.Ref.Ref.Ref.Ref.Ref.
1 cigarette/day1.811.78 (1.30–2.45)0.800.82 (0.50–1.34)1.541.51 (0.93–2.45)1.871.87 (0.58–5.99)
1–5 cigarette/day1.671.48 (1.10–1.97)0.570.57 (0.35–0.94)1.291.30 (0.83–2.04)2.212.11 (0.83–5.36)
>5 cigarette/day2.141.95 (1.50–2.54)0.420.43 (0.23–0.79)2.152.10 (1.45–3.03)2.882.84 (1.19–6.77)

RRs in the ethnic groups

Table 3 shows the RRs for overweight and obesity on LGA and PTB for each ethnic group. Although the stratified analysis showed differences in the RRs between the ethnic groups, formal testing showed no significant interaction (LGA: P = 0.541, PTB: P = 0.668). Hence, the PAF estimations given below for each ethnic group are based on the RRs for the total study population and only take into account ethnic differences in the prevalence of overweight and obesity.

Table 3.   Ethnic-specific multivariable relative risk* (95% CI) for pBMI on LGA and PTB
 DutchAfrican descentTurkishMoroccanOther non-WesternOther Western
  1. *Adjusted for education level, maternal age, parity and, respectively, smoking or pBMI.

LGA
pBMI 18.5–25Ref.Ref.Ref.Ref.Ref.Ref.
pBMI 25–301.46 (1.22–1.89)2.57 (1.06–5.23)2.47 (0.87–4.62)1.53 (0.78–2.89)1.20 (0.69–2.09)1.76 (0.91–3.61)
pBMI > 301.78 (1.43–2.77)1.75 (0.62–4.60)2.61 (0.51–4.79)2.61 (1.26–5.30)2.65 (1.52–4.61)0.61 (0.09–4.52)
PTB
pBMI 18.5–25Ref.Ref.Ref.Ref.Ref.Ref.
pBMI 25–301.07 (0.75–1.53)1.33 (0.73–2.43)1.57 (0.45–5.46)1.32 (0.53–3.29)0.76 (0.37–1.56)2.11 (0.89–5.00)
pBMI > 301.33 (0.78–2.27)1.93 (1.01–3.70)1.84 (0.37–9.12)1.24 (0.35–4.36)1.00 (0.36–2.73)3.84 (1.19–12.33)

Population attributive fractions

The PAFs of excessive weight for the total study population were –4.3%, 13.7%, 5.7% and 19.7% for SGA infants, LGA infants, PTB and extreme PTB, respectively. The contribution of excessive weight to LGA and PTB was higher in non-Western immigrant groups than in the Dutch group. For example, in the Dutch group the PAF for extreme PTB was 15.4%, but in women of African descent and in the Turkish and Moroccan groups, the PAFs were 35.7%, 27.2% and 32.9%, respectively.

The PAFs of smoking during pregnancy for the total study population were 6.4%, –4.0%, 5.7% and 11.0%, for SGA infants, LGA infants, PTB and extreme PTB, respectively. For LGA infants and extreme PTB the PAFs of excessive weight were higher than the PAFs of smoking during pregnancy.

When examining differences between the ethnic groups, the Turkish women had the highest PAF for maternal pregnancy smoking, up to 11.1% for SGA infants (Table 4).

Table 4.   Multivariable PAF (%)* of pBMI> 25 and smoking during pregnancy on pregnancy outcomes in different ethnic groups, using the relative risks of the study population
 PAF in ethnic groups, %
Total study populationDutchAfrican descentTurkishMoroccanOther non-WesternOther Western
  1. *Adjusted for education level, maternal age, ethnicity, parity and, respectively, smoking or pBMI.

SGA
pBMI > 25−4.3−3.2−9.6−6.3−8.7−5.3−2.7
Smoking6.47.44.711.11.73.45.2
LGA
pBMI > 2513.710.526.419.223.816.59.0
Smoking−4.0−4.7−2.9−7.6−1.1−1.9−3.4
PTB
pBMI > 255.74.212.48.710.77.13.5
Smoking5.76.64.410.11.62.75.0
Extreme PTB
pBMI > 2519.915.435.727.232.923.713.1
Smoking11.012.68.018.43.35.69.3

For extreme PTB, the PAF values were overall higher for maternal excessive weight (19.9%) compared with smoking (11.0%) with even larger differences in women of African descent, Moroccan women and other non-Western women.

PAF and absolute numbers of affected children in Amsterdam

Table 5 shows the PAF values of overweight/obesity for Amsterdam: SGA −4.9%, LGA 15.3%, PTB 6.6% and extreme PTB 22.0%. With the exception of SGA, these contributions were higher than the PAF estimated for smoking (6.2%, −3.9%, 5.5% and 10.6%). These PAF values imply that annually in Amsterdam overweight/obesity contributes to 126 LGA infants (out of 822 LGA infants), and 35 PTB (out of 536 PTB) and 20 extreme PTB (out of 91 extreme PTB), on the other hand obesity/overweight protects 47 children from being born SGA (on top of the number of 966 SGA infants born annually). Smoking contributes annually to 60 SGA infants, 29 PTB and 10 extreme PTB.

Table 5.   Multivariable PAF (%)* of Amsterdam on pregnancy outcomes and number of infants affected per year in Amsterdam
 PAF, %Infants, n
  1. *Adjusted for education level, maternal age, ethnicity, parity and, respectively, smoking or pBMI.

SGA
pBMI > 25−4.9−47
Smoking6.260
LGA
pBMI > 2515.3126
Smoking−3.9−32
PTB
pBMI > 256.635
Smoking5.529
Extreme PTB
pBMI > 2522.020
Smoking10.610

Discussion

The present study shows that pre-pregnancy maternal overweight/obesity is an important contributor to the occurrence of LGA infants, PTB and extreme PTB in Amsterdam. Obesity tends to be more common in non-Western immigrant women and therefore these groups are especially affected. As a result of the high prevalence of pre-pregnancy maternal overweight/obesity in Amsterdam, its contribution to the annual number of PTB is now higher than that of maternal pregnancy smoking.

Comparison with other studies

Our RR estimates expressing the magnitude of the effect of maternal overweight/obesity on LGA and SGA infants and extreme PTB are comparable to those of other studies.13,24–31 With the exception of the study by Hulsey et al.,13 all other studies show a positive association between obesity and LGA infants and a negative association with SGA infants.24–31 Some studies report that extremely obese women (BMI > 40 kg/m2) have an increased risk for SGA infants.25 In the present study we could not assess this relationship because of the low prevalence of extremely obese women. With respect to PTB, most studies report an increased risk for obese women,24,25,27–31 although one large study in the UK found a small reduction in PTB (odds ratio 0.90, 95% CI 0.84–0.97) compared with those with normal weight.26

Our results indicate ethnic differences in the relation between overweight/obesity and LGA and PTB, albeit these differences were not significant. Obese women of African descent have the highest risk for an extreme PTB. This result confirms the results of earlier studies.20,32–34 It is known that obesity leads to more serious hypertensive disorders in this ethnic population.35–37 Among others, obesity could have a larger effect on placenta insufficiency, leading to a greater degree of intrauterine growth restriction and more (extreme) PTB.

A few studies have calculated the PAF of overweight/obesity on adverse pregnancy outcomes in the general population. Hulsey et al.13 investigated the influence of overweight and obesity on low birthweight in south Colorado (USA), and reported PAF values of 3.5% and 7%, respectively. In contrast, the present study shows a negative association between SGA infants and overweight/obesity. Atalah and Castro14 found a contribution of 25% of overweight to the occurrence of macrosomia among infants in Chile. Lu et al.2 found a PAF of 19.1% for LGA infants in the period 1995–99 in Alabama, USA. According to our estimates, in Amsterdam in 2003/04 the PAF was still lower (15.3%), but in women of African descent and Moroccans the PAFs were as high as in the Alabama population (20.3% and 18.2%, respectively).

Evaluation of data

Our PAF estimations for overweight/obesity are probably too conservative. First, our pre-pregnancy height and weight values were self-reported. Participants are likely to overestimate their height and underestimate their weight, resulting in an underestimation of the prevalence of BMI, and therefore an underestimation of the PAF.38

Second, in our analyses, variations in the RRs between ethnic groups were not taken into account because these differences were not significant, although the observed variations were often in the expected direction (Table 3). By using the general RRs and ignoring the possibility that RRs may have been higher in most ethnic groups, our PAF calculations for these groups may be underestimated. Finally, the number of non-respondents was higher in the ethnic groups. The response could be selective for obesity, which could have resulted in lower estimates of the prevalence of obesity in our study population. To conclude, for several methodological reasons, we suspect that our PAF estimations are on the conservative side, and that the actual contribution of overweight/obesity may be larger than estimated in this study.

Interpretation of results

It is uncertain whether the results for Amsterdam can be generalised to other European cities, even though Amsterdam is comparable to many cities in Europe in terms of prevalence rates of obesity and in having a multi-ethnic population. In London (UK), for example, the prevalence rates of obesity are higher in locally born residents and comparable in immigrant groups, suggesting the same or higher PAF values.39

Our results show a positive association between overweight/obesity and LGA infants, PTB and extreme PTB outcomes, and an opposite association for SGA infants. Estimates of PAF assume that there is a causal relationship between those variables. Physiological evidence on the relationship between overweight/obesity and birthweight is mostly derived from animal studies. For example, obese rats showed an altered placental transfer of nutrients leading to higher birthweight in their offspring.40–42 Observational evidence from humans rests on studies addressing obesity in relation to high glucose and insulin levels leading to augmented fetal growth.43 The relationship between maternal obesity and PTB is probably caused by the higher chance for obese mothers to develop hypertensive and inflammatory disorders like pre-eclampsia during pregnancy. These disorders increase subsequently the risk of elective of induced preterm delivery.44

The higher risk for extreme PTB (<32 weeks) compared with PTB (<37 weeks) has been attributed to placental insufficiency as the main intermediary mechanism.25,30,31,44

Conclusion

Overweight/obesity is an important risk factor for major adverse pregnancy outcomes. Each year, in the city of Amsterdam, it leads to at least 126 LGA infants, 35 PTB and 20 extreme PTB. The absolute numbers may seem modest at first sight. However, LGA and SGA infants and PTB are associated with a high risk of perinatal morbidity and mortality.45–49 Infant mortality from PTB contributes to about 75% of the overall perinatal mortality.47 Preterm infants who survive their first year have an increased risk for cerebral palsy, respiratory and gastrointestinal complications, and long-term disability.46,47 Both LGA and SGA infants have an increased risk of obesity, metabolic disorders and cardiovascular diseases later in life.50–52 Furthermore, epidemiological studies have shown important effects of maternal obesity on later childhood obesity, increasing the risk three to six times even after adjustment for important confounding factors.53–55

The prevalence of overweight/obesity has rapidly increased in recent years among young people in the Netherlands. It already parallels smoking in its impact on the occurrence of adverse pregnancy outcomes. As smoking is an established risk factor for pregnancy outcomes, multiple prevention and education programmes have been organised, leading to a decrease of its prevalence among pregnant women. In the future, similar efforts are needed to develop and implement effective methods for the prevention of overweight and obesity among young women. Special attention should be paid to non-Western immigrant groups in whom the prevalence of obesity is much higher.

Disclosure of interest

The authors declare no conflict of interest.

Contribution to authorship

AD analysed the data and led the writing of the article. AK and TV assisted in the data analysis. MW contributed to the study design and led the data collection. AK, TV, MW, and HS performed the revision of the interpretation of the results. All authors contributed to and have approved the final manuscript.

Details of ethics approval

Approval of the study was obtained from the Central Committee on Research involving Human Subjects in the Netherlands, the Medical Committees of the participating hospitals, and from the Registration Committee of the Municipality of Amsterdam.

Funding

This study was funded by ZonMw.

Acknowledgements

The authors thank all the participating pregnant women for their co-operation, as well as all participating hospitals, obstetric clinics and general practitioners for their assistance in the implementation and continuation of the ABCD study.

Ancillary