Factors associated with birthweight and adverse pregnancy outcomes among children in rural Guinea-Bissau - a prospective observational study

Bandim Health Project (BHP) runs a Health and Demographic Surveillance System (HDSS) in rural Guinea-Bissau, [21]. Mobile data collection teams follow women of fertile age and children below 5 years of age through home visits. Women are registered when they reach fertile age or move into the study area. Newly registered women are interviewed about their age, obstetric history, and whether they have attended school. At all visits, the women are asked whether they are pregnant. When a pregnancy is registered, the woman’s nutritional status is assessed by measurement of her mid-upper-arm circumference (MUAC) and information on socio-economic factors (type of roofing, type of toilet, possession of a mobile phone, radio and generator) are recorded. Information on antenatal care is collected prior to giving birth, and at the first visit after delivery. At all visits, information on pregnancy outcome, and vital status of all children are collected. The present prospective observational study is nested within a cluster-randomised trial assessing the effects of early BCG and oral polio vaccine (OPV) on early infant mortality and morbidity (the BCGR trial). The trial is described in more detail elsewhere [22]. The BCGR trial was conducted in three of the nine rural health regions in Guinea-Bissau: Oio, Biombo and Cacheu and enrolments started 28 July 2016. In these three regions, village visits were conducted every 2 months.

The pregnancy registration at the two-monthly visits was complemented by continuous pregnancy registration conducted by local community key informants. The mobile data collection teams confirmed the pregnancies registered by the community key informants at the two-monthly visits and informed the women about the BCGR trial. All pregnant women were invited to a temporary health post in the village. Here a BHP nurse thoroughly explained the trial. If the woman consented to participate in the trial, the BHP nurse assessed BCG scar status of the woman by examining her arms.

Following a delivery, the community key informant or the woman herself informed BHP about the birth. A nurse was then dispatched to visit the new-born child within 72 h after birth. At this visit, the nurse examined and weighed the undressed child to the nearest 50 g using a digital scale (Charder MS 4400).

We described birthweight by calculating the mean birthweight and standard deviation (SD). We furthermore calculated the proportion of children with a birthweight below 2500 g (LBW). We assessed whether maternal BCG scar status and other available background factors for the mother (region of residence; type of roof; possession of a radio, a generator/solar panel, a mobile phone; type of toilet; years of maternal schooling; maternal age), for the pregnancy (number of prenatal consultations; maternal MUAC; total number of pregnancies), and for the delivery (sex; place of birth and season of birth) were associated with birthweight and with the risk of being LBW. We used linear regression to quantify the mean difference in birthweight and compared the proportions of LBW infants using binomial regressions with a log-link function to estimate relative difference as the prevalence ratio (PR). Since, neonatal weight is known to be very dynamic, [23] we examined whether the weight differed according to age at weighing (within the first 72 h). Five sets of twins were included in the birthweight analyses. We conducted sensitivity analyses including a) only the heaviest and b) including only the lightest twin. Since we hypothesised that maternal BCG scar status might be related to birthweight, we assessed whether the background factors were associated with the mother having a BCG scar using binomial regression with a log-link function.

Birthweight was only assessed for children who were live-born and enrolled in the BCGR trial. For all pregnant women, who consented to participate in the BCGR trial (Fig. 1), we assessed the association between maternal BCG scar status and adverse pregnancy outcomes (miscarriage, stillbirth or early neonatal death prior to weighing and enrolment in the BCGR trial) using binomial regression. A twin pregnancy resulting in both a live and a stillborn child or a child who died before weighing was considered a pregnancy with an adverse pregnancy outcome. We then assessed whether associations between maternal BCG scar and adverse pregnancy outcomes or birthweight, respectively, were affected by other measured background factors by including the variables one by one. We adjusted for factors that changed the estimate by more than 5%.

If maternal BCG affects the chance of children being carried to term, a difference in gestational age could potentially affect birthweight. Time of pregnancy registration is independent of maternal BCG scar status. Thus, if having a maternal BCG scar is not associated with the risk of preterm delivery, intervals between time of pregnancy registration and delivery should be independent of having a BCG scar. We therefore used differences in time from registration to delivery as a proxy for differences in gestational age, and assessed, whether maternal BCG scar status was associated with time between pregnancy registration and delivery, using linear regression.

We used robust standard errors in all analyses to account for clustering of observations within village clusters. Analyses were performed using Stata® version 14.1.

Birthweight was assessed for 628 children, of these 67% were home born. The mean birthweight was 2.89 kg (SD 0.43) and 104 children (17%) were LBW (Table 1). Girls had lower birthweight than boys and twins had lower birthweight than singletons. Children in the region of Oio had the lowest mean birthweight and considerably higher risk of being LBW compared with children from Biombo and Cacheu regions (Table 1). Birthweight differed by maternal age, p = 0.001. Children of the youngest mothers (< 20 years) weighed 2.76 kg on average, while children of the oldest mothers weighed 2.99 kg. Children of mothers with the smallest MUAC (< 240 mm) had a lower mean birthweight than mothers with larger MUAC. This was also reflected in the risk of being LBW; mothers with the smallest MUAC had higher risk of having a LBW child compared with the three quartiles with highest MUAC (Table 1). The risk of having a LBW child was similar in the three groups with highest MUAC.

Birthweight differed by parity (p = 0.001); children of mothers with a parity > 5 had a mean birthweight of 3.03 kg while firstborn children weighed on average 2.75 kg. Mothers with several prior pregnancies had lower risk of having a LBW child compared with primiparae. Children of mothers who had attended prenatal consultations had higher mean birthweight and lower risk of being LBW compared with children of mothers, who had not attended any prenatal consultations (Table 1). Not having a mobile phone was associated with higher risk of having a LBW child and lower birthweight (Table 1).

We did not observe a significant difference in birthweight or risk of being LBW between children of mothers with a BCG scar and children of mothers without a BCG scar (mean birthweight of 2.89 kg in children of mothers with a BCG scar and 2.89 kg in children of mothers without a BCG scar (p = 0.92); Prevalence ratio (PR) for LBW of 1.06 (95%CI: 0.78–1.46)) (Table 1). When including background factors one by one in the analysis only maternal age and twinning affected the estimate by more than 5%. The PR for LBW was 1.00 (95%CI: 0.72–1.37) when adjusted for maternal age while the twinning adjusted PR was 1.00 (95%CI: 0.72–1.39).

Results in Table 1 included both twins. Including only the heaviest or only the lightest twin in the birthweight analyses did not affect the conclusions (Supplementary Tables 1 and 2).

Birthweight was similar at different ages of birthweight assessment within 72 h after birth (p = 0.86); children weighed before 24 h had a mean weight of 2.92 kg, children weighed between 24 and 48 h had a mean weight of 2.85 kg and children weighed between 48 and 72 h had a mean weight of 2.90 kg. (Table 1).

Information on maternal BCG scar status was assessed for 1320 of the 1324 consented women. Among these, 848 (64%) mothers had a BCG scar and 472 (36%) did not have a BCG scar. Younger mothers/mothers with fewer pregnancies were more likely to have a BCG scar (p < 0.001) (Table 2). The distribution of background factors was similar among mothers of enrolled children (Table 2).

One thousand three hundred twenty-four pregnant women consented to enter the BCGR trial. We obtained information on maternal BCG scar status for 1320; 848 (64%) had a BCG scar and 472 (36%) did not have a scar. Women without a BCG scar tended to have a higher overall risk of an adverse pregnancy outcome (miscarriage, stillbirth or early neonatal death prior to weighing and enrolment in the BCGR trial) compared with women with a BCG scar, PR 1.29 (95%CI 0.99–1.86), p = 0.06 (Table 3). Stratifying by type of pregnancy outcome revealed that women without a BCG scar tended to have higher risk of stillbirths (PR 1.55 (95%CI 0.98–2.44)) and early neonatal deaths (PR 1.51 (95%CI 0.81–2.80)) compared with women with a BCG scar, while there was little difference in risk of miscarriage, PR 0.90 (95%CI 0.50–1.63). Excluding pregnancies resulting in twin births from the analysis did not affect conclusions (Supplementary Table 3). Not having a BCG scar was thus associated with more stillbirths and early neonatal deaths, PR = 1.49 (1.07–2.08) for all pregnancies and 1.35 (0.96–1.89) for singleton births.

Adjusting the estimated association between BCG scar and pregnancy outcome for the background factors listed in Table 2, none of the variables affected the result by more than 5%.

The study was conducted within the setup of a large randomised trial conducted in the BHP rural HDSS. The clusters were initially selected using the Expanded Programme of Immunization Strategy, [24] thereby ensuring a representative sample of the population. Data collection in the rural HDSS has been ongoing since 1990. The field workers are experienced and frequently supervised. Pregnant women are followed prospectively throughout their pregnancy allowing for reliable estimates of adverse pregnancy outcomes. Registration of pregnancies was done through both two-monthly visits by the mobile teams and continuous pregnancy registration by local community key informants, thereby capturing as many pregnancies as possible. Scar information was assessed by trained BHP nurses, who were frequently supervised, ensuring high quality of the BCG scar data. A team of 15 nurses working at government health centres were trained to perform home visits shortly after birth. The nurses were trained at seminars and frequently supervised to ensure quality and uniformity. Birthweight was obtained using identical medical scales that were calibrated routinely. Therefore, we consider our weight measurements to be precise.

All children in our study were weighed within 72 h after birth, thereby ensuring data on weight in early life. Neonatal weight is known to be very dynamic, [23] however, on average the weight of children did not differ significantly by age at assessment (within the first 72 h). Due to the lack of gestational age data, we were not able to differentiate between small for gestational age and preterm birth as causes for LBW. Reliable measures of gestational age are difficult to obtain in low-resource settings, where medical equipment is scarce. Estimates of gestational age at the health centres were often based on fundal height measurements at the health centres. Available data on gestational age was based on information from the health cards and maternal recall, both of these methods can be unreliable [25]. We therefore used difference in time from registration of pregnancy to delivery as a proxy for difference in gestational length. Time between registration and delivery was similar for children of mothers with and without maternal BCG scar status.

A large number of women, who gave consent to enter the trial were not enrolled (633 children enrolled/1324 women consented). The trial was conducted in a rural setting, and it was difficult to obtain information about birth within 72 h. Giving birth is a stressful experience, especially in a low-income country where medical support for mothers giving birth at home is lacking, which is probably one of the main reasons for not informing timely about the birth of the child. Poor cell phone connection, lack of electricity to charge a cell phone, lack of a phone all together are other obstacles to timely information. We used community key informants to ensure more timely information, but they had the same phone problems, add to that travel, work or school outside the village. The enrolled cohort was similar to the consent cohort based on most measured background factors (Table 2). Thus, we consider our results to be representative.

We found a mean birthweight of 2.89 kg (SD 0.43) among infants in rural Guinea-Bissau. A large multi-site study spread across eight countries found a mean birthweight of 3.3 kg (SD 0.5) across all sites, [26] thus our results indicate that children in rural Guinea-Bissau on average are 400 g below the international mean according to this study. The multi-site study only measured term babies, whereas we measured birthweight of all live-born children surviving until the visit by the nurse, which may explain the difference. WHO estimates the average percentage of LBW infants in western Africa to be 15.4% with 22% for Guinea-Bissau and 64% of children not weighed [1]. Thus, our results of 17% (22% in the region of Oio) with weight obtained also for children born at home are in line with the estimates. Our results are also in line with a recent systematic review, where the LBW prevalence in Sub-Saharan Africa was 16.4% in 2000, falling to 14.0% in 2015 [5].

We used maternal MUAC during pregnancy as a measure for maternal nutritional status and found that low MUAC was associated with a lower mean birthweight and higher risk of having a LBW child, this is supported by previous studies suggesting that maternal nutritional status is an important risk factor for a child’s risk of being LBW [27‐29]. In our study, we found that higher maternal age was associated with higher mean birthweight. A study of birthweight determinants in Oman from 2015 found the same association [30]. The before-mentioned multi-site study from 2017 also found that the very young women had increased odds of having a LBW child. Contrary to our findings, they found the oldest group also had an increased risk of having a LBW child [27]. In line with our results being primipara has been associated with lower birthweight and higher risk of LBW in prior studies [27, 29, 30].

Birthweight data from low-income countries such as Guinea-Bissau can be difficult to obtain especially in rural areas, where many children are born at home. Estimates of birthweight among children born at home are often based on weight at the first contact with a health facility. Birthweight assessed from health cards often contain little information about when and how the child was weighed, [1] thus the birthweight estimates may be unreliable. In our study we present precise birthweight information measured in a standardised way by supervised health professionals. To our knowledge this is the first birthweight data on home-born children from rural Guinea-Bissau.

As LBW is associated with increased mortality, morbidity and bears with it high social and economic costs for a country, [1, 31‐33] a high prevalence of LBW is a substantial burden for a low-income country. To target effective prevention efforts in a low-income setting, it is important to have reliable data on birthweight.

Our results suggest that women without a BCG scar may have higher risk of adverse pregnancy outcomes (Table 3), indicating that maternal BCG-vaccination might affect pregnancy outcomes. Due to the study type, we cannot conclude causation and results should be interpreted with caution. While none of our measured background factors affected the estimate by more than 5%, there may be unmeasured confounding. The results do however stress the necessity of exploring further the associations between maternal vaccination and pregnancy outcomes, both for BCG and other vaccines, especially as maternal vaccination is increasingly explored as a way of protecting children in early life.