Predictors of Maternal Serum Concentrations for Selected Persistent Organic Pollutants (POPs) in Pregnant Women and Associations with Birth Outcomes: A Cross-Sectional Study from Southern Malawi

Population exposure to persistent organic pollutants (POPs) may result in detrimental health effects, especially to pregnant women, developing foetuses and young children. We are reporting the findings of a cross-sectional study of 605 mothers in their late pregnancy, recruited between August 2020 and July 2021 in southern Malawi, and their offspring. The aim was to measure the concentrations of selected POPs in their maternal serum and indicate associations with social demographic characteristics and birth outcomes. A high level of education was the main predictor of p,p′-DDE (p = 0.008), p,p′-DDT (p < 0.001), cis-NC (p = 0.014), o,p′-DDT (p = 0.019) and o,p′-DDE (p = 0.019) concentrations in maternal serum. Multiparity was negatively associated with o,p′-DDE (p = 0.021) concentrations. Maternal age was also positively associated (p,p′-DDE (p = 0.013), o,p′-DDT (p = 0.017) and o,p′-DDE (p = 0.045) concentrations. Living in rural areas was inversely associated with high maternal serum concentrations of p,p′-DDT (p < 0.001). Gestational age was positively associated with p,p′-DDE (p = 0.031), p,p′-DDT (p = 0.010) and o,p′-DDT (p = 0.022) concentrations. Lastly, an inverse association was observed between head circumference and t-NC (p = 0.044), Oxychlordane (p = 0.01) and cis-NC (p = 0.048). These results highlight the need to continue monitoring levels of POPs among vulnerable populations in the southern hemisphere.


Introduction
Persistent organic pollutants (POPs) are carbon-based, man-made synthetic chemical substances produced mainly for commercial purposes, such as for pests and disease control, crop production and industrial use [1]. These compounds have very long half-lives when released into the environment. They accumulate in the environment and adversely affect the food chain and living organisms, including humans, because of their lipid solubility

Study Design, Population and Sites
Details on the study design, population and sites were described in detail in the previous publication of our study [16]. In brief, this cross-sectional study of pregnant women and their offspring was conducted in the southern region of Malawi, and study participants were recruited between August 2020 and July 2021.
The inclusion criteria included all pregnant women who were in the late stages of their pregnancy; 16 years of age and above; permanent residents of the Blantyre, Chiradzulu and Thyolo districts; and willing to voluntarily sign an informed consent statement. All women with serious medical conditions present at the time of enrolment, high risk pregnancies with prognoses to be referred to a tertiary health care level, or were also participating in another study were excluded. In this regard, a total of 771 pregnant women were assessed for eligibility and 605 were recruited for the study. However, only 564 women-neonate pairs were included in the final study population for statistical analysis on the associations between POP concentrations, as 41 either declined to have their biological sample collected, or the volume of the sample was insufficient. Figure 1 describes the details of the number of participants assessed for inclusion and recruited.
women-neonate pairs were included in the final study population for statistical analysis on the associations between POP concentrations, as 41 either declined to have their biological sample collected, or the volume of the sample was insufficient. Figure 1 describes the details of the number of participants assessed for inclusion and recruited.

Study Questionnaire
The questionnaire that was used for the collection of information about mothers and neonates was already described in the first publication of our study [16]. Briefly, pregnant women´s personal characteristics, reproductive history, socioeconomic status, lifestyle, environmental characteristics, diet and neonatal data were collected just before delivery using a provider-administered questionnaire.

Sample Collection and Preliminary Analysis
Methods for the whole blood sample collection and preliminary analysis used in the present study were adapted from the international quality control system QA/QC established by the Centre de Toxicologie du Quebec [17]. The detailed process of the sample collection and preliminary analysis were described in our first publication for the study [16].

Study Questionnaire
The questionnaire that was used for the collection of information about mothers and neonates was already described in the first publication of our study [16]. Briefly, pregnant women s personal characteristics, reproductive history, socioeconomic status, lifestyle, environmental characteristics, diet and neonatal data were collected just before delivery using a provider-administered questionnaire.

Sample Collection and Preliminary Analysis
Methods for the whole blood sample collection and preliminary analysis used in the present study were adapted from the international quality control system QA/QC established by the Centre de Toxicologie du Quebec [17]. The detailed process of the sample collection and preliminary analysis were described in our first publication for the study [16].

Serum Sample Analysis
Serum samples were analysed at the Environmental Pollutant Laboratory at the University Hospital of North Norway according to the previously published method by Huber et al. [18]. In short, a Tecan Freedom Evo 200 (Männedorf, Switzerland) liquid-handling workstation was applied for automated sample preparation. First, 150 µL aliquots of serum samples were diluted prior to extraction on reversed phase 96-well plates, which was followed by clean-up on normal phase extraction columns. Gas chromatography atmospheric pressure ionisation coupled to tandem mass spectrometers (GC-API-MS/MS; Waters, Milford, MA, USA) were used for instrumental analysis. The API was conducted in positive mode under charge transfer conditions. For detection on the mass spectrometer, the multiple reaction monitoring mode was applied with two specific transitions for the individual analytes. Quantification was performed using the Masslynx and Targetlynx software (Version 4.1 and 4.2, Waters, Milford, MA, USA) and achieved by the internal-standard method with isotope-labelled compounds.

Measurement of Birth Outcomes
The following birth outcome variables were measured soon after delivery: gestational age (weeks), birth weight (kg), birth length (cm) and head circumference (cm) [16]. Gestational age was estimated at delivery by referring to the antenatal health records and the ponderal index was used to estimate the nutritional status of the newborns and calculated using the following formula [16,19]:

Statistical Analysis
Data analysis was performed using the statistical software Stata for Mac (SE standard version 17; College Station, TX, USA). Our analysis focused on compounds with the detection frequency of ≥50%. With this criterion, the compounds that qualified for the analyses were: HCB, p,p -DDE, t-NC, p,p -DDT, Oxychlordane, cis-NC, o,p -DDT and o,p -DDT (Table 1). To control for missing data, concentrations of selected compounds below the limit of detection (LOD) were replaced by LOD/2. LODs were automatically calculated by the Masslynx software for each individual analyte and sample. Detailed information on the LODs is given in Supplementary Table S1. Descriptive statistics including arithmetic means, standard deviation (SD), median and minimum and maximum or proportion (%) were computed from the data set (Table 1). A significance level of p < 0.05 (two-tailed) was set for all analyses. The data were fitted into a multivariable regression model with maternal characteristics as the response variable and the significant POP concentrations as predictors. All POP concentrations were log-transformed before inclusion in the multivariable regression models due to non-normal distribution of the concentrations among the participants.

Maternal Socio-Demographic and Neonates' Anthropogenic Data
The socio-demographic and neonates' anthropogenic data are described in our previously published article from this Malawi study [16]. The age of women recruited in the study ranged from 16 to 45 years, with a mean (SD) age of 24.8 (6.2) years. Just over half of the women (308) were recruited from the urban setting, and 297 were from rural areas. Women recruited from the urban setting were almost two years older than their counter-parts recruited from rural areas. In this regard, the mean ages (SD) for pregnant women recruited from urban and rural areas were 25.6 (6.7) and 23.9 (5.7) years, respectively.
The data showed differences in gravidity, parity and educational levels between the urban and rural pregnant women. On this, nulliparity was significantly high in rural areas. In contrast, Para 1 and multiparity were statistically high among women recruited from urban areas, as compared to their counterparts from rural sites.
A small (38.6%) proportion of women recruited from urban areas either did not attend any formal school or were educated just up to the primary level. However, the majority (61.4%) attained an education up to the secondary or tertiary level. In contrast, the majority of pregnant women from rural areas did not attain higher educational levels (69.7% attained up to the primary level versus only 30.3% attaining the secondary/tertiary level). The use of tap water as the source of drinking water was very common (96.8%) among urban study participants. As for the rural setting, shallow wells and boreholes were collectively the common source of drinking water, representing 44.1% and 45.5%, respectively.
In total, 572 neonates (51.8% boys) were recruited in the present study. The mean birth weight, length and head circumference for the neonates recruited were 3.09 kg, 45.28 cm and 33.15 cm, respectively, in the overall sample. Detailed information about maternal sociodemographics and the neonates anthropogenic data according to their area of residence (urban versus rural) is also outlined in the Supplementary Table S2.

Maternal POP Serum Concentrations
Eight out of the 19 POPs with detection frequencies (DF) ≥50% were included in the statistical analysis. The eight compounds included in the statistical analysis were Hexachloroben- Table 1 provides descriptive statistics for the selected eight POPs in pg/mL (as wet weight concentration of the POP/unit serum). The two POPs with the highest median concentrations observed in this study were 478 pg/mL and 102 pg/mL for p,p -DDE and HCB with detection frequencies of 99.3% and 99.8%, respectively. On the other end, the lowest median concentration assessed was 0.34 pg/mL for o,p -DDE. Detailed information of the POPs that were included in the in the statistical analysis is presented in Table 1.

Concentrations of POPs in Maternal Serum and Associations with Maternal Characteristics
The POP concentrations followed an approximately normal distribution after logarithmic transformation. Maternal characteristics, namely age and parity, were automatically considered to be included in the multivariate analysis, as they are already known to be associated with POP concentrations in maternal serum based on previous studies. Furthermore, univariable linear regression analysis was used to determine the covariates that needed to be included in the multivariable linear regression models. In this regard, the following factors were selected to be included in the multivariable analysis: maternal educational level, previous breast-feeding and beef and goat meat consumption frequencies. Tables 2 and 3 show the detailed results of the univariate and multivariable analyses, respectively, between maternal characteristics and different POP concentrations in maternal serum.
Multivariable linear regression models of the aforementioned variables provided an overall description of the main factors influencing the concentrations of the selected POPs. Comprehensive descriptions of the main maternal risk factors related to the serum POP levels are given in Tables 2 and 3.  Values shown were simple linear regression analyses coefficients. a All POP concentrations were log 10 transformed before analysis. b Para 0 is the reference category. c None and primary educational level are the reference category. d Urban is the reference category. e Tap water is the reference category. f No previous breast-feeding is the reference category. g,h Less than twice a week is the reference category. * p < 0.05; ** p < 0.001. Significant findings are printed in bold.
After adjusting for maternal age, parity, area of residence (urban versus rural), source of drinking water and beef and goat meat consumption, participants educated up to either secondary or tertiary levels had significantly higher concentrations of p,p -DDE (p = 0.008), Multiparity was significantly associated with decreased maternal serum concentrations of some POPs. This was observed for o,p -DDT (β = −0.927; 95% CI: −1.713 to −0.141; p = 0.021). Living in rural areas was inversely associated with maternal serum concentrations of p,p -DDT (β = −1.302; 95% CI: −1.871 to −0.733; p < 0.001).

Maternal Dietary Habits and Level of Education
Human exposure to POPs is mainly through the ingestion of foods of animal origin, such as meat, fish, dairy items and eggs [20]. In this regard, data on diet and other maternal characteristics, such as the maternal educational level, were also explored. Univariate analysis results showed that the consumption of beef more than twice a week was statistically significantly associated with higher maternal serum concentrations of HCB (β = 0.900; 95% CI: 0.035 to 0.145; p < 0.001), p,p -DDT (β = 0.544; 95% CI: 0.247 to 0.841; p < 0.001) and o,p -DDE (β = 0.209; 95% CI: 0.038 to 0.379; p = 0.017). A similar trend of associations was also observed for the consumption of goat meat as follows: HCB (β = 0.076; 95% CI: 0.024 to 0.129; p = 0.004), p,p -DDT (β = 0.351; 95% CI: 0.066 to 0.636; p = 0.016) and o,p -DDE (β = 0.206; 95% CI: 0047 to 0.369; p = 0.013). However, these associations were not observed in the multivariable linear regression model.

POP Concentrations in Maternal Serum and Their Associations with Birth Outcomes
In order to investigate the possible associations between the selected POPs and birth outcomes, multivariable linear regression models were applied and adjusted for maternal age, area of residence (urban versus rural), maternal educational level, parity and source of drinking water. As shown in Table 4

Discussion
Maternal level of education, age and parity were the main predictors of concentrations of POPs in the maternal serum. In this regard, increased maternal serum POP concentrations were found to be associated with increased levels of maternal education, age and parity. Living in urban settings was associated with increased maternal serum concentrations of HCB and p,p -DDT. Some POPs (p,p -DDE, p,p -DDT and o,p -DDT) were negatively associated with gestational age. Furthermore, t-NC, cis-NC and Oxychlordane were inversely associated with head circumference. Notably, o,p -DDT was positively associated with gestational age.
The maternal serum POP concentrations observed in this study were generally low, as compared to other published studies conducted elsewhere [15,21,22] but comparable to the study conducted by Steinholt et al. in Cambodia [19]. In this regard, the median (min-max) maternal serum p,p -DDT concentrations observed in these two studies were almost consistent (35.5 (0.43-1700) pg/mL ww for the current study and 33.0 (3-519) pg/mL ww for the Cambodian study). It was also noted that the median HCB concentration in the maternal serum observed in the present study was high (102 (0.57-203) pg/mL) compared to the Cambodian study (44.0 (10-196) pg/mL). In contrast, the median (min-max) concentrations of p,p -DDE and o,p -DDT observed in the present study were on the lower side (448 (0.10-23,600) vs. 720.0 (20-8047) pg/mL, and 5.10 (0.11-163) vs. 20.0 (20-20) pg/mL, respectively) and o,p -DDE was on the higher side (8.00 (8-10) vs. 0.34 (0.04-8.9) pg/mL) compared to the Cambodian study.
For DDT and its metabolites, p,p -DDE was the most abundant compound detected in the maternal serum. These findings are similar to previous studies from Cambodia and Vietnam [19,23]. This is because p,p -DDE is the most common and stable metabolite of p,p -DDT; hence, it is the most likely to be detected in both environmental and human samples. Furthermore, various studies worldwide have shown that the detection frequency of p,p -DDE is 100% [24][25][26] despite restrictions on its production and use over 40 years ago. In addition, our dataset showed no indication of a fresh DDT-exposure source, since the concentrations of p,p -DDT were rather low, as compared to p,p -DDE.
The level of education was the main predictor of POP concentrations in the maternal serum. Generally, high educational levels were significantly associated with increased maternal serum concentrations of most of the POPs. This was the case with p,p -DDE (p = 0.008), p,p -DDT (p < 0.001), cis-NC (p = 0.014), o,p -DDT (p = 0.007) and o,p -DDE (p = 0.019). The positive statistically significant associations between the compounds p,p -DDE and p,p -DDT and educational level were consistent with the results found by Steinholt et al. in another study conducted in Cambodia [19]. This could be explained by differences in diets between the two groups. Human exposure to POPs is mainly through ingestion of foods of animal origin, such as meat, fish, dairy items and eggs [20]. Although the univariate analysis showed statistically significant positive associations between consumption of beef and goat and the concentrations a number of POPs (HBC, p,p -DDT and o,p -DDE), similar associations were not observed in the multivariable linear regression analysis. Just as with maternal educational level, higher maternal age was also statistically significantly associated with higher maternal serum concentrations of p,p -DDT, o,p -DDT and o,p -DDE. Concentrations of the above POPs in their serum were higher in older women than in younger women. This was expected, as high levels of most persistent organic compounds in the maternal serum increase with age, as previously reported in published literature [23,[27][28][29][30]. It was also observed that living in urban areas was significantly associated with high concentrations of p,p -DDT in the maternal serum. This could be attributed to different lifestyles, especially related to diet. For instance, pregnant women living in urban setting are more likely to consume meat-based foods frequently, as compared to rural residents. Furthermore, women residing in urban areas have high chances of consuming food that is contaminated with DDT and its metabolites, as most food stuffs in supermarkets are imported from South Africa, where DDT is still used for malaria prevention. However, there is a need for a large study on the effect of diet on the concentrations of maternal POPs to ascertain the above hypotheses.
Multiparity was associated with decreased maternal serum concentrations of o,p -DDT. These results are similar to other studies that have also revealed that parity is negatively associated with maternal serum concentrations of some POPs [23,30] due to the fact that these compounds are permeable through the placenta and are transferred to the growing foetus [18,29,[31][32][33]. Thus, a negative association between p,p -DDT and parity is attributed to mother-to-foetus transfer of these compounds through the placenta. Furthermore, a large proportion of women (58.9%) recruited in this study reported previous breast-feeding. Since POPs are lipophilic in nature, they tend to be secreted into breast milk [34][35][36][37][38], and the inverse association observed can also be explained by depletion of these compounds in the maternal serum due to the transfer from mother to the baby through breast-feeding.
In the present study, positive associations were observed between gestational age and each of the following POPs: p,p -DDE, p,p -DDT and o,p -DDT. These findings concur with results from a study by Jusko et al., in which a positive association was observed between p,p -DDT and o,p -DDT with gestational age [39]. However, the findings in the present study are in contrast to the results in a study by B J Wojtyniac et al. in the Kharkiv cohort, which found that lower gestational ages were associated with an increase in log p,p -DDE exposure [40]. The difference in results between these two studies could be due to the sample size, regional distribution, data analysis methods and selection of confounding factors. In addition, we also observed a statistically significant inverse association between some POPs (t-NC, Oxychlordane and cis-NC) and the head circumferences of the neonates, as reported in other studies. There are limited published data on studies on the association between t-NC, Oxychlordane and cis-NC with head circumference. However, the previ-ously listed substances are endocrine disruptors; hence, they are more likely to affect foetal growth, which may result in smaller head circumferences of the neonates.
The main limitation of this study is its cross-sectional design that prevents causality associations and, therefore, causality associations cannot be shown. However, to our knowledge, this is the first study to explore concentrations of selected POPs in the maternal serum of pregnant women in Malawi and their associated pregnancy outcomes. The study revealed predictors of POP levels in maternal serum and their associations with birth outcomes. Another strength is that samples were drawn from different settings (urban and rural), which makes the results more representative of Malawian women.

Conclusions
Our study made a comprehensive effort to examine the determinants of selected persistent organic pollutant concentrations in maternal serum during pregnancy and their associations with social demographic characteristics and birth outcomes. This study gives a foundation upon which knowledge of the main predictors for POP levels in maternal serum among pregnant women at the delivery stage and from low-middle income African settings can be built. In this regard, maternal level of education, age and parity were the main predictors of the concentrations of POPs in the maternal serum. Although the concentrations of the POPs observed were generally low, there is still a need for more research to periodically monitor the concentrations of these compounds and the associated lifestyles during pregnancy. In addition, the results from the present study may be used to facilitate the identification of specific lifestyles that are associated with high concentrations of POPs in the maternal serum of women of child-bearing age.
Supplementary Materials: The following supporting information can be downloaded at https: //www.mdpi.com/article/10.3390/ijerph20075289/s1: Table S1: Method detection limits (MDL) concentrations (wet weight pg/mL) for all POPs analysed at the laboratory; Table S2: Socio-demographic characteristics of recruited women and their neonates. Informed Consent Statement: Written informed consent was obtained from all participants involved in this study.
Data Availability Statement: Data will be made available upon reasonable request from the corresponding author.
Jamester Chilunjika, for assisting in sample collection and storage; Cythia Amin for data entry and management; and the laboratory technicians at the University Hospital of North Norway, Merete Linchausen Skar, Christina Ripman Hansen and Arntraut Götsch, for POP analysis. We also thank the University Hospital of North Norway, Department of Laboratory Medicine and the Northern Norway Regional Health Authority (Helse Nord) for their financial support of the POP analysis.