OUP user menu

Contribution of overweight and obesity to adverse pregnancy outcomes among immigrant and non-immigrant women in Berlin, Germany

Katharina Reiss, Jürgen Breckenkamp, Theda Borde, Silke Brenne, Matthias David, Oliver Razum
DOI: http://dx.doi.org/10.1093/eurpub/ckv072 ckv072 First published online: 13 April 2015

Abstract

Background: Maternal excessive weight and smoking are associated with an increased risk of pregnancy complications and adverse pregnancy outcomes. In Germany, immigrant women have a higher prevalence of pre-pregnancy overweight/obesity compared with autochthonous women. We compared the contribution of pre-pregnancy overweight/obesity to adverse pregnancy outcomes among immigrant and autochthonous women in Berlin/Germany. Methods: Data from 2586 immigrant women (from Turkey, Lebanon, other countries of origin) and 2676 autochthonous women delivering in three maternity hospitals of Berlin within 12 months (2011/2012) was used. Cox regression models were applied to estimate the association between overweight/obesity and smoking with the outcomes large-for-gestational-age (LGA), small-for-gestational-age (SGA), preterm birth (PTB) and extreme preterm-birth (E-PTB). Population attributive fractions (PAF) were calculated to quantify the proportion of the outcomes attributable to overweight/obesity and smoking, respectively. Results: Prevalence of overweight and obesity was 33.4% among autochthonous and 53.6% among Turkish women. Prevalence risk ratios of excessive weight were highest for LGA infants among immigrant and autochthonous women. The PAFs were −11.8% (SGA), +16.3% (LGA), +3.6% (PTB) and +16.5% (E-PTB) for the total study population. Conclusions: Overweight/obesity is strongly associated with an increased risk of delivering an LGA infant among both immigrant and autochthonous women. Compared with autochthonous women, the contribution of excessive weight to LGA is even higher among immigrant women, in whom PAFs of overweight/obesity even exceed those of smoking for some outcomes.

Introduction

According to the World Health Organization (WHO), the worldwide prevalence of overweight and obesity in adults is increasing. In 2008, 35% of the adults were overweight and 12% were obese.1 In Germany, e.g. 67% of the men and 53% of the women reported to be overweight. The prevalence of overweight among adult women of reproductive age was 38% in Germany.2

Overweight/obesity has negative effects on health; moreover, pre-pregnancy overweight/obesity does not only affect the mother but also poses a health risk to the (unborn) child: pregnancy complications such as preeclampsia or gestational diabetes; and adverse outcomes such as macrosomic neonates, a low Apgar score, preterm birth (PTB) or stillbirth.3–5 The proportion of adverse outcomes attributable to pre-pregnancy overweight/obesity is substantial—for some conditions, the impact may be higher than that of smoking.6–9

Women who migrated from low- and middle-income to high-income countries often have even higher prevalences of overweight/obesity than women from the majority populations10–12 and thus experience higher pregnancy-related risks. This may be of substantial public health importance in Germany, which had 11 million immigrants (13.3% of the total population) in 2012. The largest groups originate from Turkey and from Eastern European countries.13 Women of Turkish origin have a higher prevalence of obesity and above-average weight gain during pregnancy than autochthonous women.14 However, the contribution of pre-pregnancy overweight/obesity to adverse pregnancy outcomes in Germany has not yet been quantified.

In Amsterdam, The Netherlands, Djelantik et al.15 compared the contribution of excessive weight to the occurrence of adverse pregnancy outcomes among Dutch and immigrant women. The outcome variables were small-for-gestational-age infants (SGA), large-for-gestational-age infants (LGA), PTBs and extreme preterm births (E-PTBs). The prevalence of overweight and obesity was particularly high among Turkish, Moroccan and African women. Overweight and obese women experienced an increased risk of delivering an LGA, PTB and E-PTB infant and a decreased risk of delivering an SGA infant. The contribution of overweight/obesity to LGA, PTB and E-PTB even exceeded that of smoking. However, the study was restricted to one particular setting and specific immigrant groups15; thus, the public health impact of pre-pregnancy excessive weight needs to be assessed in other settings and immigrant populations.

We here (i) analyse the contribution of pre-pregnancy overweight/obesity to adverse pregnancy outcomes among immigrant and autochthonous women; (ii) quantify the magnitude of this effect and (iii) compare overweight/obesity and smoking in their respective contribution to adverse pregnancy outcomes in the two groups.

Methods

We used data of the study ‘The influence of migration and acculturation on pregnancy and birth’. It was conducted in a 12-month period in 2011/2012 in three maternity hospitals of Berlin. Pregnant women aged 18 years and older with a permanent residence in Germany were interviewed directly after admission to labour ward, but before the onset of labour pain, and again shortly after delivery, provided they gave birth to a living newborn at gestation week 24 onward. Questionnaires comprised socio-demographical and socio-economical questions, questions on antenatal care, migration history, acculturation and smoking during pregnancy; they were available in eight languages and were applied by trained, multilingual study staff. Of 8157 women delivering in the three hospitals, 2.9% (n = 235) did not meet the inclusion criteria. Of the remaining women, 7100 participated in the main study (response rate: 89.6%).16

Outcome variables

The outcome variables were LGA, SGA, PTB and E-PTB. The cut-off for E-PTB was a gestational age of 25 to <32 weeks and for PTB in general of <37 weeks (thus including E-PTB). LGA was defined as birth weight above the 90th percentile by gestational age and sex of the child; SGA was defined as birth weight below the 10th percentile by gestational age and sex of the child (classifications based on Voigt et al. 2006).17

Determinants

Pre-pregnancy body mass index (BMI) was based on height and weight measurements taken and documented in the course of the first antenatal care visit offered by a medical doctor. On average, it took place during the 9th/10th gestational week. Weight gain in the whole first trimester is usually 0.5–2 kg at most18 and thus not substantially different from normal fluctuations in body weight. Doctors used the instruments available in their clinic for weight and height measurement. BMI was classified according to WHO recommendations: underweight (<18.5 kg/m2), normal weight (18.5 to <25 kg/m2), overweight (25 to <30 kg/m2) and obesity (≥30 kg/m2). Information on maternal weight and height was missing in 7.2% and 13.7% of the participants, respectively. An imputation procedure using the average of five iterations based on linear regression analyses was applied with IVEware.19 Imputation of maternal height was based on country of birth and age; imputation of maternal weight was based on country of birth, height, parity and age. Smoking was categorized into three categories: non-smoking, smoking regularly and smoking occasionally.

Immigrant women were identified by their country of birth being different from Germany and were grouped as follows: Turkish, Lebanese, other Western and other non-Western. Turkish and Lebanese are the numerically largest immigrant groups in Berlin. Women from other Western and other non-Western countries were classified as in the Amsterdam study. The comparison group comprised autochthonous women who themselves and whose parents were born in Germany. Women born in Germany, but with parents born outside the country, were not included in the analysis.

Educational achievement was measured by highest graduation level and ranged from no graduation to tertiary education. Age of the pregnant women (grouped as 18–29, 30–39, ≥40 years) and parity (0, 1+) were also included as covariates.

Statistical analyses

To estimate the influence of overweight/obesity and smoking on the occurrence of SGA, LGA, PTB and E-PTB, Cox regression models with constant risk periods were applied. The results of the Cox regression models were very similar to those of Poisson regression analyses (not shown). Uni- and multivariate prevalence risk ratios (PRRs) were estimated. PRRs of overweight/obesity and smoking for all defined outcome variables were calculated for the total study population (immigrant and autochthonous women). PRRs of overweight/obesity on LGA and PTB were also calculated stratified by immigrant group. Linear regression models were used to check for multicollinearity of variables selected for Cox regression analyses. No evidence for multicollinearity was found. Additionally, interaction analyses between country of birth and overweight/obesity as well as overweight/obesity and smoking were performed.

Population attributive fractions (PAFs) were estimated to quantify the proportion of adverse outcomes attributable to overweight/obesity and smoking, respectively. For this purpose, we calculated ‘average PAFs’20 based on logistic regression analyses (available as SAS macro20). The Levin formula (as used in the Amsterdam study) may result in an overestimation as the fractions of risk factors can add to more than the possible maximum of 100%. Average PAFs were calculated for the total study population (immigrant and autochthonous women) and stratified by immigrant group. Normal-weight and non-smoking women served as reference groups for the PRRs. Underweight women were excluded before calculating PRRs and PAFs.

As the distribution by country of birth of all pregnant women in Berlin is unknown for any given point of time, we could not estimate the PAFs and the absolute number of affected children per year for the city as a whole.

Results

Complete data from 5262 immigrant and autochthonous women was available for the analyses. As pre-pregnancy weights were based on measurements taken at the first antenatal visit (usually around week 10), we conducted a sensitivity analysis by excluding women who booked after week 13. This hardly affected mean and median weights, and had only small effects (0.1–1 percentage points) on prevalences of excess weight (not shown).

Of the 5262 pregnant women, 25.5% were overweight and 12.8% were obese. Overweight and obesity were least prevalent in autochthonous women (33.4%) and women originating from other Western countries (33.1%); prevalences were highest in women from Turkey (53.6%) and Lebanon (52.6%; both p < 0.001; reference: autochthonous women). In the total study population, the prevalences of regular and occasional smoking were 14.0% and 5.0%. Smoking prevalence was lower in women from other non-Western countries, higher in women from other Western countries and similar between autochthonous women and women from Turkey and Lebanon. The proportion of SGA infants varied between 20.5% in Lebanese and 10.5% in Turkish women. The proportion of LGA infants was lowest in Lebanese and highest in autochthonous women. The highest proportion of PTB was also found in autochthonous women (7.9%), the lowest was observed among Turkish women (3.1%) (table 1).

View this table:
Table 1

Characteristics of the study population

Total study populationCountry of birth
Germany (autochthonous)TurkeyLebanonOther, non-WesternOther, Western
Number of women526226766473465541039
Outcome variables
    SGA %14.914.910.520.515.015.8
    LGA %5.96.86.34.34.94.5
    PTB %
        <37 weeks6.57.93.14.66.16.2
        <32 weeks1.81.90.51.52.91.7
Determinant variables
    Maternal BMI %a
        Underweight (<18.5)5.05.22.84.35.85.6
        Normal weight (18.5 to <25)56.761.343.643.149.661.3
        Overweight (25 to <30)25.521.537.734.727.823.7
        Obesity (≥30)12.811.915.917.916.89.4
    Smoking %
        Non-smoking81.081.079.380.195.374.6
        Smokes occasionally5.05.15.93.82.06.3
        Smokes regularly14.013.914.816.22.719.2
  • a: Numbers in bold print indicate significant (<0.001) results based on chi-square test in 2*2 contingency tables (normal weight vs. overweight/obesity; autochthonous women vs. single immigrant group)

Table 2 lists the PRRs for the total study population. Obese women have a significantly higher risk of delivering an LGA infant. Correspondingly, an inverse significant association was found for delivering an SGA infant among overweight and obese women. Maternal obesity was also associated with an increased risk of delivery before 37 weeks of gestational age (PTB), as well as before 32 weeks (E-PTB) with only the latter being statistically significant. Smoking during pregnancy was associated with a significantly higher risk for delivering an SGA infant and a significantly lower risk for delivering an LGA infant among regular smokers. The risk of PTB was slightly elevated in smokers, but not the risk of E-PTB. However, these associations were not statistically significant.

View this table:
Table 2

PRRs for BMI and smoking during pregnancy on pregnancy outcomes

SGALGAPTBE-PTB
Univar. PRRMultivar. PRR (95% CI)aUnivar. PRRMultivar. PRR (95% CI)aUnivar. PRRMultivar. PRR (95% CI)aUnivar. PRRMultivar. PRR (95% CI)a
BMI
    18.5 to <25Ref.Ref.Ref.Ref.Ref.Ref.Ref.Ref.
    25 to <300.800.77 (0.64–0.92)1.251.30 (0.98–1.72)0.961.00 (0.77–1.31)1.061.20 (0.70–2.03)
    ≥300.750.70 (0.55–0.89)2.432.53 (1.90–3.36)1.241.24 (0.91–1.77)2.112.41 (1.40–4.12)
Smoking
    Non-smokingRef.Ref.Ref.Ref.Ref.Ref.Ref.Ref.
    Smokes occasionally1.531.43 (1.05–2.32)0.670.69 (0.38–1.27)1.211.23 (0.76–1.99)0.700.83 (0.26–2.67)
    Smokes regularly2.031.93 (1.60–2.32)0.430.40 (0.25–0.63)1.261.22 (0.89–1.67)0.891.04 (0.53–2.03)
  • univar., univariate; multivar., multivariate.

  • a: Adjusted for educational level, maternal age, country of birth, parity and, respectively, smoking or BMI.

Table 3 presents the PRRs of autochthonous and immigrant women with pre-pregnancy excessive weight to deliver an LGA infant or having a PTB. Among both autochthonous and immigrant women (irrespective of region or country of origin) excessive weight is positively associated with LGA. The risk is slightly elevated among overweight women but substantially higher among obese women. Moreover, the association between obesity and LGA is statistically significant in all comparison groups, except for women from other non-Western and other Western countries. The PTB risk is also elevated in obese autochthonous women and in women from Turkey, although results are not significant. Among the other immigrant women, no significant results were observed, too. Because of the very small number of cases especially among Lebanese women with PTB, we combined overweight and obesity into one category. For the same reason, some other models were slightly modified (see footnotes of table 3).

View this table:
Table 3

Multivariable PRRsa (95% CI) for BMI on LGA and PTB, by country of birth

Country of birth
Germany (autochthonous)TurkeyLebanonOther, non-WesternOther, Western
LGA2533626331520974
    18.5 to <25Ref.Ref.Ref.Ref.bRef.
    25 to <301.22 (0.84–1.77)1.19 (0.55–2.58)2.68 (0.48–14.78)1.12 (0.43–2.91)1.56 (0.78–3.12)
    ≥302.27 (1.55–3.33)2.96 (1.33–6.61)10.45 (2.20–49.71)1.67 (0.64–4.36)2.28 (0.99–5.27)
PTB
    18.5 to <25Ref.Ref.Ref.cRef.dRef.
    25 to <301.06 (0.74–1.51)1.13 (0.40–3.21)0.690.82 (0.35–1.91)0.78 (0.42–1.46)
    ≥301.40 (0.93–2.09)1.53 (0.46–5.13)(0.26–1.89)1.13 (0.44–2.94)1.15 (0.53–2.48)
  • a: Adjusted for education level, maternal age, parity and smoking.

  • b: Variable smoking removed from regression model.

  • c: BMI-groups (overweight and obesity) combined.

  • d: Smoking dichotomized (yes/no).

The contribution of overweight and obesity to the outcome variables were −11.8% (SGA), +16.3% (LGA), +3.6% (PTB) and +16.5% (E-PTB) for the total study population. The highest contribution of excessive maternal weight to LGA was found in Lebanese, Turkish and other Western women with +57.8%, +14.5% and +19.4%, respectively. This is a higher contribution than that among autochthonous women, where the PAF is +10%. The PAFs of smoking were +17.5% (SGA), −7.8% (LGA), +5.2% (PTB) and −0.9% (E-PTB) for the total study population. The contribution of smoking to SGA was lowest in Lebanese and highest in autochthonous women. The results for PTB and E-PTB were inconsistent between the comparison groups. The PAFs of overweight/obesity and smoking on PTBs and E-PTBs varied, taking negative or positive values among autochthonous and immigrant women (table 4). For the calculation of the PAFs, age (18–29; 30+ years) and education (no graduation/primary education; other) were included in the models as binary variables due to the small number of events in some strata, especially with respect to E-PTB.

View this table:
Table 4

Multivariate PAFa (%) of BMI and smoking during pregnancy on pregnancy outcomes, by country of birth (method: average PAF)

Total study populationCountry of birth
Germany (autochthonous)TurkeyLebanonOther, non-WesternOther, Western
SGA49842533626331520974
    BMI ≥ 25–11.8–10.5–35.8–28.6–14.1–0.3
    Smoking17.520.018.09.49.719.4
LGA
    BMI ≥ 2516.310.014.557.812.019.4
    Smoking–7.8–7.33.03.4–3.5–23.8
PTB
    BMI ≥ 253.65.311.2–16.7–2.0–1.0
    Smoking5.22.510.25.3–1.612.2
E-PTB
    BMI ≥ 2516.514.7–8.812.1–6.19.1
    Smoking–0.9–1.07.1–1.31.93.1
  • a: Adjusted for education level, maternal age, parity and, respectively, smoking or BMI.

Discussion

This study has four main findings. First, overweight and obesity are more prevalent in pregnant immigrant women and here especially in women from Turkey and Lebanon, when compared with autochthonous women. Second, overweight and obesity are associated with an increased risk of delivering an LGA infant both among autochthonous and immigrant women. Third, the contribution of overweight and obesity to LGA is even higher among immigrant than among autochthonous women. Finally, the PAF of excessive weight exceeds that of smoking for some important neonatal outcomes, in particular among immigrant women.

Although the ethnic composition varies between the Berlin study16 and the Amsterdam study,15 we observed partially similar trends. Maternal overweight and obesity were more prevalent among immigrant women compared with autochthonous women in Germany/the Netherlands. In both studies, excessive pre-pregnancy weight was related to a decreased risk of delivering an SGA infant and to an increased risk of delivering an LGA infant in both the autochthonous and the immigrant groups. Unlike the Amsterdam study, we did not find an association between overweight and PTB/E-PTB; in obese women, there was a significant association only with E-PTB. Among pregnant immigrant women, excessive maternal weight was associated with delivering an LGA infant in both studies. Contrary to the Amsterdam study, we did not observe a consistent association between maternal overweight/obesity and PTB among all immigrant groups. In terms of the PAFs among autochthonous women in Germany/the Netherlands, the highest contribution of excessive maternal weight was observed for LGA and E-PTB. However, in the Amsterdam study, the PAF values were higher among immigrant women than among Dutch women for LGA and PTB, whereas this was only observed for LGA in the Berlin study. In sum, there was conformity between both studies in terms of (i) the higher prevalence of overweight and obesity among pregnant immigrant women compared with autochthonous women in Germany/the Netherlands, (ii) the increased risk for overweight/obese women of delivering an LGA infant—both among autochthonous women in Germany/the Netherlands and immigrant women and (iii) a higher contribution of maternal excessive weight to LGA in immigrant women compared with autochthonous women in Germany/the Netherlands. The difference in the associations of maternal excessive weight and PTB might be due to the different composition of the immigrant groups, the small number of E-PTB cases in the Berlin study or the differences in the calculation of the regression models or the PAFs. Other studies also present ambiguous evidence regarding the relationship between overweight/obesity and PTB; some observe an increased risk of PTB,21 whereas others show a reduced risk.22

A large body of evidence links maternal overweight/obesity to adverse pregnancy outcomes, in particular to delivering an LGA infant.6,23,24 A systematic review with a meta-analysis showed that pre-pregnancy overweight/obesity increases the risk of being LGA, macrosomic and having a high birth weight.25 A higher risk among pregnant immigrant women compared with autochthonous women—as found in the Berlin and Amsterdam studies—was not consistently observed in other countries: some studies showed no or few differences between ethnic groups.26–28 However, the studies differ in terms of data sources, methodological approaches and definition of the respective ethnic groups. In the few studies which calculated PAFs for different ethnic groups, the contribution of excessive pre-pregnancy weight to adverse pregnancy outcomes was higher among immigrant compared with autochthonous women,29,30 as in the Berlin and Amsterdam studies. According to Oteng-Ntim et al.,31 e.g. the PAF for maternal obesity on macrosomia was 13.2% for black Americans and 5.2% for white Americans.

Strengths and limitations

The strength of this study is the direct analysis of the contribution of pre-pregnancy overweight/obesity to adverse pregnancy outcomes in Berlin/Germany by means of PAFs. Moreover, this association could be studied in non-immigrant and different immigrant groups. There are two studies from Germany that analysed the effect of maternal BMI on perinatal outcomes32 and the change in maternal characteristics within a time frame of 16 years.33 Neither one calculated PAFs or focused on different immigrant groups. In addition, our study achieved a response rate of almost 90%. A limitation is the small number of cases of adverse pregnancy outcomes in the stratified analyses by immigrant groups. Besides, height and weight of the mother are based on measurements in a large number of physician’s clinics. If, e.g. a woman books late in pregnancy, or does not undress for weighing, her weight might be overestimated and hence also her BMI (albeit less so because of the squared term in the denominator). As a consequence, PAFs would also be overestimated. Thus, the actual contribution of maternal excessive weight on adverse pregnancy outcomes might be somewhat smaller than estimated. Our sensitivity analysis shows, however, that women booking late do not materially change the findings. Moreover, measured height/weight at 9/10 weeks is presumably more valid than self-reported height/weight recalled from an unspecified time before the onset of pregnancy. Finally, pregnant women residing in Berlin may differ from those in towns or rural areas. Although our findings might apply to pregnant women residing in other German metropolitan areas, the transferability to all pregnant women residing in Germany is debatable. However, we do not consider this aspect as crucial since our analysis focuses on a biological association between pre-pregnancy excessive weight and adverse pregnancy outcomes which is independent of e.g. place of residence or maternity hospital.

Conclusion

Both the calculation of PRRs as well as PAFs demonstrate the considerable impact of maternal overweight and obesity on the health of the newborn. This constitutes an important public health challenge, given that in high-income countries the prevalence of overweight and obesity among women is high and projected to increase even further. Besides the reported immediate adverse pregnancy outcomes, many studies also report subsequent childhood overweight/obesity, also in different ethnic groups.34,35 Thus, interventions that contribute towards optimizing the BMI of women by incorporating healthy dietary habits and offering opportunities to exercise before and during pregnancy should be implemented.36,37 The finding that overweight and obesity play an important role should of course not divert from efforts to help smoking women stop their habit, ideally well before they plan a pregnancy.

Future research should compare differences in gestational weight gain between groups of different origin, in addition to pre-pregnancy maternal weight, as the former also appears to be associated with adverse pregnancy outcomes.38,39 In addition, unfavourable outcomes not only among the newborns but also among their mothers should be studied—e.g. in terms of gestational diabetes, pregnancy-induced hypertension or increased caesarean delivery rate.40 Beyond that, the relationship between maternal overweight/obesity and adverse pregnancy outcomes needs to be further investigated among women of different national and ethnic origin to underpin the associations found here.

Funding

This study was supported by Deutsche Forschungsgemeinschaft (DFG), reference: DA 1199/2-1.

Conflicts of interest: None declared.

Key points

  • Maternal pre-pregnancy excessive weight poses a substantial health risk to mother and child.

  • The contribution of pre-pregnancy overweight and obesity to adverse pregnancy outcomes has rarely been studied in immigrant populations.

  • Overweight and obesity are more prevalent in pregnant immigrant women compared with autochthonous women in Germany.

  • Overweight and obesity are strongly associated with an increased risk of delivering a large-for-gestational-age (LGA) infant; the contribution of overweight and obesity to LGA births is even higher among immigrant women compared with autochthonous women.

  • Preventive measures need to optimize the body mass index of women (irrespective of immigration status) by incorporating healthy dietary habits and opportunities to exercise, ideally well before pregnancy.

Acknowledgements

Data protection regulations were observed in the survey and in the linkage to hospital data. The Charité Ethics Committee, Berlin, approved the study on 18 February 2009, reference EA1/235/08. The authors wish to thank patients and staff of the participating hospitals as well as the data collection team. Moreover, the authors wish to express their sincere thanks to Dr. Tanja G. M. Vrijkotte, Project leader ABCD study, Department Public Health, Academic Medical Center (AMC) Amsterdam, for her support.

References

View Abstract