Maternal pre-pregnancy underweight as a risk factor for mother and child: analysis of the birth cohort Survey of Neonates in Pomerania I

Background: The prevalence of pre-pregnancy underweight (BMI (body mass index) <18.5 kg/m²) varies between 3% in Europe and 17% in Asia. Illness and low socioeconomic status may be associated with pre-pregnancy underweight which then may result in adverse pregnancy and neonatal outcome. Methods: We analysed pre-pregnancy BMI categories using data from the population-based birth cohort Survey of Neonates in Pomerania 2002-2008 (SNiP) in north-eastern Germany. Multivariable regression analyses adjusted for confounding were used to investigate a) socio-economic risk factors for underweight before pregnancy and b) associations of underweight with pregnancy and neonatal outcomes. Results: Chronic pre-pregnancy diseases were not more frequent in underweight pregnant women. Bivariate analyses showed that underweight women were younger by three years (p<0.001), smoked twice as often (p<0.001), and less likely graduated from high school (p<0.001) compared to women with normal weight. The former were more often unemployed (p<0.001) and had lower available net-income (p<0.001). However, multivariable regression analysis adjusted for confounders revealed that only a younger age (0.90 (0.88-0.92) and smoking (RR 2.52 (1.89-3.41) remained associated with underweight. Compared to women with normal pre-pregnancy BMI, underweight women had an increased risk of premature labour (OR = 1.74; 95% CI: 1.29-2.36) and a reduced placental weight. No association was found between pregnancy-induced hypertension, gestational diabetes, preeclampsia, mode of birth and pre-pregnancy underweight. The offspring of underweight women had an increased risk of late preterm birth (OR = 1.79; 95% CI: 1.17-2.73) or birth with birth weight <2500g (OR = 1.99; 95% CI: 1.24-3.21). Conclusions: Smoking during pregnancy and a younger age were identied as risk factors for maternal pre-pregnancy underweight which then was associated with preterm birth and low birth weight. Targeted measures like smoking cessation programs absolute and/or percentages. Wilcoxon test and two-tailed χ2 test were used to calculate p-values for continuous and categorical variables, respectively. First, associations of socioeconomic status (educational level, income, employment status), and adverse habits (smoking and alcohol consumption) with mothers’ low prepregnancy BMI were analysed by linear, logistic and multinomial logistic regressions adjusted for confounders. Second, associations of mothers’ low pre)


Educational level
The strati cation pattern for educational level followed the already published pattern [16]. Persons without a school diploma, being still at school or with ve years, or less, of secondary school were pooled together and were referred to as having a low educational level. Persons with 6 years of secondary school (German Realsch|c| ̲ hluss´were ∈ cluded ∈ thesecond ≤ vel, referred → asthem ≤ educational ≤ vel. Thethird ≤ vel ∈ cludedpersonswith8yearsofse Fachhochschulreife´ or `Abitur´) and was referred to as the mid-high educational level. The highest educational level was assigned to persons who graduated and was referred to as a high educational level.
De nitions of smoking and alcohol use In this paper, we did not analyse the dose effect of tobacco and alcohol consumption on pregnancy and neonatal outcomes. Therefore, we did not differentiate the cohort by the amount of alcohol consumed or tobacco smoked. Instead, we used a simple dichotomous classi cation: smoke r n on -smoker´and drinker/non-drinker´. A woman was classi ed as a smoker if she declared to smoke during the last four weeks before delivery.
Similarly, a woman was classi ed into the group of drinkers if she continued to drink alcohol during pregnancy, irrespective of the amount and time period of consumption.
De nition of small-for-gestational-age and large-for-gestational age Small-for-gestational-age (SGA) was de ned as birthweight below the 10 th percentile for their estimated gestational age. Large-for-gestational-age (LGA) was de ned as a birth weight greater than the 90 th percentile adjusted for gestational age [17].
The gestational age was based upon the date of the mother's last menstrual period according to records in maternity card.

Diagnosis of neonatal hypoglycaemia
Neonatal hypoglycaemia was diagnosed using biochemical parameters according to national guidelines. A plasma glucose concentration of 45 mg/ml (2.5 mmol/l) within the rst 24 hours after birth was used to diagnose neonatal hypoglycaemia. It was a routine policy to screen babies of mothers with gestational diabetes, preterm babies, babies with low birth weight <2,500g, as well as all SGA-and LGA-babies for hypoglycaemia.
De nition of neonatal asphyxia Neonatal (birth) asphyxia was de ned according to medical diagnosis and included all ICD-10 codes P21.-, which included asphyxia of any grade (severe asphyxia, P21.0, mild and moderate birth asphyxia, P21.1, as well as unspeci ed birth asphyxia, P21.9). Medical diagnosis was taken from child's medical records.

Conditions for admission to neonatal care
According to the institutional policy and the national guideline, the personnel attach great importance to maintaining mother-child contact even in the case of pathology. Babies and their mothers were left at the maternity ward as long as the conditions allowed such a situation. The national guideline clearly de nes when the neonate should be transferred to the neonatal ward, particularly when intravenous administration of glucose was necessary or in case of severe symptoms. For the purposes of the study, `admission to neonatal care´ included both neonatal intensive care and special care with respect to the newborn's condition and needs.

De nition of monthly available equivalent income
The need for housing space, electricity, and other essentials does not increase proportionally with the higher number of members in the household. To account for this phenomenon, we have used equivalence scales, based on the OECD-modi ed scale [16,18].

Potential mediators and confounders
We have considered the following factors as potential mediators in the pathway between maternal low BMI before pregnancy and adverse pregnancy and neonatal outcomes: tobacco smoking and alcohol consumption during pregnancy, maternal age, available monthly equivalent income, and parity. These variables were assessed by the self-administered questionnaire. Ethnicity is another potential confounder; however, this factor could not be analysed because less than 2% of the population were not Caucasian.

Statistical analyses
All data were stored using a Microsoft Access 2002 (Microsoft Corporation, Redmond, WA, USA) database.
Continuous data are reported as the medians with the 25 th and 75 th percentiles; categorical data are expressed as the absolute numbers and/or percentages.
Wilcoxon test and two-tailed χ2 test were used to calculate p-values for continuous and categorical variables, respectively. First, associations of socioeconomic status (educational level, income, employment status), and adverse habits (smoking and alcohol consumption) with mothers' low prepregnancy BMI were analysed by linear, logistic and multinomial logistic regressions adjusted for confounders. Second, associations of mothers' low pre-) pregnancy BMI with adverse pregnancy and neonatal outcomes, such as birth weight, gestational age, hypoglycaemia, admission to the neonatal care unit, mode of delivery were analysed. The respective confounders used in the multivariate analyses are mentioned in the legends of each table. In all analyses, p < 0.05 was considered statistically signi cant. All statistical analyses were carried out using Stata 16.0 (Stata Corporation, College Station, TX, USA).

Results
Socioeconomic characteristics of the studied subpopulation Out of 5,801 mother-child-dyads participating in the baseline SNiP-I study, pre-pregnancy BMI was known for 4,667 mother-child-dyads, which were included into the current analysis (Table 1). Of these, 322 (6.9%) women were underweight, and 3,085 (66.1%) had normal weight. Bivariate analyses showed that underweight women were three years younger in median (p<0.0001), and smoked more than twice as often as normal weight women during pregnancy (p<0.0001). They had lower available equivalent income (p<0.0001), and lower socioeconomic status (employment status and/or educational level, p<0.001 and p<0.0001, respectively) compared with women with normal weight.
Chronic diseases as origin of the low pre-pregnancy BMI Selected chronic diseases were analysed for their association with the low pre-pregnancy BMI. The results of this analysis are shown in the Supplementary   File1, since the prevalence gures were generally too low to include them into a profounder statistical analysis.
Associations of socioeconomic factors with pre-pregnancy BMI <18.5 Increasing maternal age was associated with decreasing risk of being underweight (RR = 0.90; 95% CI, 0.88-0.92) ( Table 2). Low educational level showed signi cant associations with maternal BMI <18.5 but not when confounders (see legend of Table 2  Parity, alcohol consumption during pregnancy, middle or mid-high educational level were not associated with the BMI <18.5. We did not perform the analysis of the available equivalent income and employment status due to low number of available data for these two variables. Less than 50% of participants included into current analysis provided data on their available income or employment status.
Association of maternal outcomes with BMI<18.5 With reference to pregnancy outcome, we observed lower placenta weight (520g vs. 560g, p<0.0001) and more frequent premature labour (32.8% vs. 20.6%, p<0.0001) in the group of underweight women compared to women with normal weight before pregnancy (Table 3). We did not observe signi cant differences regarding the prevalence of pregnancy-induced hypertension, GDM, preeclampsia and delivery mode between underweight and normal weight women.
Regression analysis has con rmed the association between the low pre-pregnancy BMI and lower placenta weight (adjusted ß-coe cient = -36.3 g; 95% CI, -58.4g to -13.5g) and premature labour (OR = 1.74; 95% CI, 1.29-2.36) in the group of underweight women. Similar to bivariate analysis, we did not observe any association between BMI<18.5 and mode of birth, preeclampsia, GDM or pregnancy-induced hypertension ( Table 3).
Regression analyses revealed that a BMI<18.5 was a risk factor for babies to be born lighter, by 209g for male infants (adjusted ß-coe cient, 95% CI, -295g to -124g) and by 247g for female infants (adjusted ß-coe cient, 95% CI, -336g to -157g LGA was the only neonatal risk factor that was lower in the group of underweight women (OR = 0.51; 95% CI, 0.29-0.90).

Discussion
Recent literature suggests that there is an association between low socioeconomic status (BMI <18.5 kg/m²) and low pre-pregnancy BMI , leading to adverse pregnancy and neonatal outcomes. We have investigated this hypothesis using data from the population-based rural-shaped birth cohort study `Survey of Neonates in Pomerania´ (SNiP-I) and analysed whether low socioeconomic status and adverse habits like smoking and alcohol consumption was associated with low pre-pregnancy BMI. Furthermore, the aim was to investigate how low pre-pregnancy BMI affected pregnancy and neonatal outcomes.
The prevalence of underweight in the SNiP I population was 6.9% and within the range reported for Europe and USA in a recent meta-analysis comprising more than 1.3 Mio pregnant women [3]. In Asia, particularly in Japan numbers were higher with 17 and 19% [3,19]. This can be partly explained by applying WHO percentiles for BMI rather than ethnic-speci c percentiles [3]. Body image in young women is suspected to contribute to the high prevalence of pre-pregnancy underweight and low gestational weight gain in Japan [13,20]. In our study, chronic diseases did not contribute to pre-pregnancy underweight. Different cohort studies reported an association of low educational level with underweight in pregnant women [11,21,22]. However, in the SNiP-cohort low educational level was not associated with pre-pregnancy underweight after adjustment for parity and age. This may be partly explained the high proportion of female students in the young age group of women at child-bearing age in the university town of Greifswald.
The most prominent health behaviour associated with underweight in the SNiP-cohort was smoking. The prevalence of smoking within underweight women (40.5%) was twice as high as in the whole SNiP-cohort and higher as the recently reported estimated prevalence of smoking during pregnancy (8.1%, 95% CI 4.0-12.2) for the European Region [23]. In a recent study from Italy (24), only 6.2% of underweight women still smoked during pregnancy [24]. Smoking is associated with lower education [11,25].
Bakker et al reported smoking rates during pregnancy between 10.6% and 25.3% with the highest rate in the youngest women of the Generation R cohort [26].
In our cohort, a younger age was associated with pre-pregnancy underweight. This is in accordance with the literature [27,28]. Age and body weight and constitute main determinants of body image in women. Particularly in Japan, weight control among young women is discussed to contribute to the high prevalence of pre-pregnancy underweight [13,20]. Therefore, age may be an indirectly potentially modi able risk factor by counselling young women.
The risk for complications during pregnancy like pregnancy-induced hypertension, gestational diabetes mellitus and preeclampsia was not increased in underweight women in our cohort. Other reported even a lower risk for these complications in women with pre-pregnancy underweight [29][30][31].
Concerning neonatal outcome, women with pre-pregnancy underweight showed increased risk for low placental weight, late preterm births and low birth weight. This is in line with previous reports and meta-analyses [26,[30][31][32][33][34]. The association between pre-pregnancy underweight and low birth weight and preterm birth might be explained directly by a lack of nutrients resulting in diminished fetal growth or duration of gestation or indirectly through other associated factors such as smoking, poor diet or medical illness. Particularly maternal smoking was very frequent in underweight women in the SNiP-cohort.
Recently, it was shown that beside known toxic effects smoking during pregnancy showed genome-wide methylation differences in cord blood [35].
The strengths of our analysis are the high population coverage of SNiP-I, the large number of participants, homogeneous ethnic compositions, geographically de ned study region and a comprehensive dataset including medical and socioeconomic factors. While other cohort-based studies, such as LIFE in Leipzig or Generation R, investigate urban populations with highly inhomogeneous ethnic compositions, the SNiP I was conducted in a rural area on a population with high prevalence of obesity and low socioeconomic status compared to the populations in other regions of Germany [36][37][38][39].
A limitation of the study was that we calculated the pre-pregnancy BMI using mothers' self-reported data, which might be a source of error. However, a large systematic review showed that reporting error did not bias associations between pregnancy-related weight and birth outcomes [40].The missing analysis of data on the available monthly equivalent income may be seen as the second limitation of the study. However, data on income were available only for about half of participants and, if income is included in any regression analysis, obtained results will not be conclusive. We did not analyse gestational weight gain. Control of gestational weight gain within recommended ranges might improve neonatal outcome [3]. However, results from a recent multi-cohort analysis suggest that pre-pregnancy weight might be a more important target for interventions than gestational weight gain [28].
In conclusion, smoking during pregnancy and a younger age were identi ed as risk factors for maternal pre-pregnancy underweight which then was associated with preterm birth and low birth weight. Targeted measures like smoking cessation programs and counselling young women about the risk of staying too slim particularly for the offspring may improve outcome. Research based on patient-related data and human DNA is strictly regulated by German law. The collection of detailed personal data, combined with the sampling of biomaterials, demands strict con dentiality. Our study complies with international guidelines of ethical research based on the Declaration of Helsinki. The study design was reviewed and approved by the Ethics Committee of the Board of Physicians Mecklenburg-Western Pomerania at the University of Greifswald (Reg.-Nr. III UV 20/00). Eligible women were asked for written informed consent; in cases of legally minor mothers, i.e., aged < 18 years, the additional signatures of the newborn's and legally minor mother's legal caregivers were required. The enrolment procedure was described in details by Ebner et al [14]. The written informed consent form included data assessment in face-to-face interviews, self-administered questionnaires and patient records. It also covered biosamples of blood. Furthermore, informed consent covered the storage of pseudonymized data as well as their analyses and publication.
The current study is a part of a population-based birth cohort `Survey of Neonates in Pomerania´ (SNiP) examining the health and socioeconomic information from 5801 mothers and their children. SNiP was designed as a multipurpose birth cohort to serve as a platform for studies of pregnancy complications, maternal and child health from pregnancy through to adulthood after completion of follow up. Therefore, all analyses from this database have to be regarded as secondary analyses.

Consent for publication
Not applicable.

Availability of data and materials
The paper is based on the data collected during the study 'Survey of Neonates in Pomerania', conducted at the University Medicine Greifswald, Greifswald, Germany, between 2002 and 2008. Data from the SNIP are available via https://www.fvcm.med.uni-greifswald.de/dd_service/ data_use_intro.php?lang=ger.
The repository is managed by the Research Cooperation Community Medicine (RCC) of the University of Greifswald, Germany. This data repository allows any researcher to register and apply for access. It provides a data dictionary and online application tools for accessing the data. Upon application by the registered users, the RCC determines whether to grant access to the data, based on scienti c guidelines.

Competing interests
The authors declare that they have no competing interests.

Funding
The The current analysis presented in this paper was performed without further external support. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Authors' contribution GD and MH were responsible for writing the manuscript, and AEL contributed to acquisition of data and writing the manuscript. GD, AEL, HA and JF analysed and interpreted the data. TI and HA conducted statistical analyses and contributed to their interpretation. MZ has made substantial contribution to conception and design of the obstetrical issues of the study, as well as to acquisition of data. MH, AEL and GD originally conceived of the study. MH is the principle investigator of SNiP and has made substantial contribution to analysis and interpretation of data and writing the manuscript. All authors were involved in revising the manuscript critically for important intellectual content. All authors have read the nal version, given nal approval of the version to be published, and agreed to be accountable for all aspects of the work. a -only raw value is shown as maternal age cannot be affected by any other factor; b -the highest educational level was a reference group, adjusted for maternal age and parity; c -adjusted for maternal age, parity, educational level, and alcohol use during the pregnancy; d -adjusted for smoking during pregnancy, maternal age, parity, and educational level; e -women with rst pregnancy were reference group, adjusted for educational level and maternal age  Flow diagramm showing selection procedure apply to data from SNiP I Study