Maternal demographic and antenatal factors, low birth weight and preterm birth: findings from the mother and child in the environment (MACE) birth cohort, Durban, South Africa

Pregnant women attending the public sector ante-natal clinics in Durban and surrounding areas (Lamontville, Merebank, Bluff, Austerville, Wentworth, KwaMashu, Newlands East and Newlands West), and who met the inclusion criteria were invited into the cohort. The objective of the MACE study was to determine the risk of environmental pollutant exposure commencing in utero on long term respiratory health of children to 6 years of age and specific outcomes such as asthma.

The selection criteria for subjects to enter this study were as follows:

Women who were asthmatic, hypertensive, diabetic or tested positive for human immunodeficiency virus (HIV), syphilis and tuberculosis during their routine ante-natal testing were included. Participating females with complications of pregnancy such as preeclampsia, placenta previa, were also included. Females with multiple pregnancies were excluded from the study.

We report on the first 1099 recruited pregnant women who had progressed to delivery, experienced a stillbirth or miscarriage, or exited the cohort. The cohort profile (Fig. 1) provides details of the changes from the point of recruitment in April 2013 through to March 2018. The cohort size at each trimesters of pregnancy were 987 (1st trimester), 932 (2nd trimester), 869 (3rd trimester), and 760 (at delivery). Relocation of participants outside the study area and choosing to use clinics closer to their new homes was the single largest reason for the loss to follow-up (n = 257). Participants (n = 82) who experienced miscarriages and stillbirths or terminating their pregnancy were subsequently removed from the cohort.

Trained interviewers conducted standardised validated face-to-face interviews with all mothers at enrolment, second and third trimesters, as well as monthly telephonic follow-up interviews. The questionnaires included demographic information, antenatal history, place of birth and residential history, occupational history, maternal smoking, exposure to passive cigarette smoke, occupational and environmental exposures, use of biomass fuels, dietary history, pre-existing medical conditions, and past medical and obstetric history. Maternal demographic information was based on responses to questions about marital status (married/living together/single/divorced/widowed); maternal education (highest grade of school completed/post-high school education/degree), maternal income (six categories of annual income ranging from no income to US$7200 or greater) and maternal employment (student/apprentice/at home/unemployed/disabled/ choice of seven key sectors of work).

Data was captured at the time of the interview using a mobile telephone system, automatically uploading data onto the study database using wireless technology, Mobile Researcher®.

All questionnaires were made available in English and isiZulu (an indigenous South African language). The isiZulu versions were translated from English and then back translated by a second translator to assure that the instruments are truly equivalent.

A clinical assessment of all enrolled subjects were undertaken in the 3rd trimester of pregnancy. Maternal weight was assessed at enrolment and at delivery. Obesity was evaluated just prior to delivery. In addition, information on routine assessments performed as part of their antenatal follow-up by the public health service was extracted from their records. These included their HIV status and if positive, use of antiretroviral medication, tests for syphilis and diabetes.

Time of birth and delivery type data were collected by the fieldworkers or extracted from the labour records. Anthropometric measurements and gestational aging were done by the hospital nursery staff using World Health Organisation (WHO) standards. The midwife undertook the measurement of birthweight for each child independently while early ultrasound and post-natal scoring measured gestational age. The person undertaking these measurements was blinded to the study outcomes. This data were used to classify cases into those that were PB (< 37 completed weeks of pregnancy) and those with LBW (< 2500 g). Gestational age was estimated by a combination of clinical date of last menstrual period, early antenatal ultrasound and the Ballard scoring system postnatally. A combination of prematurity and weight status per individual child was not done.

PB and LBW were the dependent variables of interest. A variety of independent variables were explored through interviews using the standardised instruments. These variables included: socio-demographic variables (marital status; maternal education, maternal income, maternal employment); maternal general, reproductive and obstetric health; residential variables (residential area, housing type; energy type); behaviour and exposure in the current pregnancy (alcohol and smoking history, environmental tobacco smoke exposure, biomass fuel use). In addition, variables with known association with the outcomes of interest were included in our analyses. These included maternal age, HIV positive status, antiretroviral treatment and weight gain during pregnancy.

Standard exploratory data techniques were employed, and descriptive statistics, with frequencies and ranges for categorical, mean and standard deviation (SD) for continuous variables were used to summarize participants’ characteristics.

The variables listed above, which, in the bivariate analyses showed a p-value < 0.25 (allowing for joint effects) were included in the multivariate logistic regression models. We ran separate models for each of the dependent variables of interest, PB and LBW respectively. We additionally included those variables which were clinically important into the models, such as maternal age. We report the crude and adjusted odds ratios (AORs) from the logistic regression models with their 95% Confidence Interval (CI). A p-value less than 0.05 was considered to be statistically significant. The IBM SPSS statistical software package (version 25.0) was used for all statistical analyses.

The study was approved by the Biomedical Research Ethics Committee (BREC) of the University of KwaZulu-Natal (UKZN) (BF263/12). All the participants in this study gave written informed consent, participated voluntarily, received no financial incentives, and had the right to withdraw at any stage.

The majority of mothers were single (79.0%); 16.7% had received tertiary education, over 84.0% did not work, 44.0% earned a yearly gross income of less than R2 000, with most living in low-income housing communities (67.2%) and informal settlements (14.7%) (Table 1). Of those who worked, almost all (98.2%) had low to middle income jobs (Table 1).

The mean gravida and parity scores were 2.5 (SD: 0.9) and 0.7 (SD: 0.9) respectively, with 44.2% of mothers having their first pregnancy. (Table 1). Within the sample, 11.5 and 24.1% were teenage and middle aged pregnancies. While 39.0% of pregnancies were planned, 22.3% occurred on contraception usage. Over 70% of the cohort were sexually active for a minimum of once per week. Of the 613 participants with a previous obstetric history (n = 965 previous pregnancies), 24.0% had bad obstetric outcomes; with the majority (11.3%) having spontaneous abortions (Table 1).

Almost half (45.3%) of the pregnant mums lived with people that were cigarette smokers and approximately a third (31.2%) of them were exposed to environmental tobacco smoke (ETS) (Table 2). A small subset of the cohort were ever-smokers (7.0%) and only 3.6% actively smoked during the pregnancy. A small proportion (7.4%) were ever consumers of alcohol and only 2.7% actively drank during the pregnancy (Table 2).

A third of the study population were infected with HIV (30.9%) and all were receiving antiretroviral therapy (ART), while 15% tested positive for syphilis. The prevalence of hypertension, gestational diabetes, tuberculosis and asthma were extremely low (Table 2). The mean weight gain for mothers across the first to the third trimester of pregnancy was 6.3 (SD: 6.1 kg) (Table 2).

From the enrolled cohort of 1099 cases, a total of 760 (70.8%) live births resulted. As expected, the incidence of miscarriages were higher in the first trimester compared to the second trimester of pregnancy (4.0% versus 1.1% respectively). There were 2.3% of mothers who experienced stillbirths during the third trimester. The occurrence of neonatal and infants deaths were low at 0.7 and 1.4% respectively (Table 3). Over 16.4% of mothers delivered PB and 13.5% of babies that were delivered were LBW (< 2500 g); 0.6% were < 1500 g. (Table 3).

From the bivariate analyses, variables of household risk (location, type or energy type) and socio-demographic variables (marital status; maternal education, maternal income, maternal employment) did not reach statistical significance, and were therefore not included in the logistic regression models.

Mothers aged > 30 years (AOR: 1.8, 95% CI: 1.1–2.9), current smokers (AOR: 2.7, 95% CI: 1.3–5.8) and caesarean deliveries (AOR: 1.7, 95% CI: 1.1–2.7) had statistically significant increased odds ratios for PB. Although not significant, mothers that were teenagers (15–19 years) (AOR: 1.2, 95% CI: 0.6–2.3), were HIV positive (AOR: 1.4, 95% CI: 1.0–2.2), and with a history of tuberculosis (AOR: 2.0, 95% CI: 0.6–6.6) and sexually transmitted infections (AOR: 2.7, 95% CI: 0.6–12.2) also had increased odds ratios for PB babies. Conversely, overweight mothers (BMI: 25.0–29.9 kg/m²) (AOR: 0.5, 95% CI: 0.3–0.9), obese mothers (BMI: > 30 kg/m²) (AOR: 0.5, 95% CI: 0.3–0.9) and mothers with a low BMI (< 18.5 kg/m²) (AOR: 0.9, 95% CI: 0.4–2.4) did not present with increased odds of having PB babies (Table 4).

Mothers that had caesarean deliveries had an increased odds ratio for LBW babies (AOR: 1.7, 95% CI: 1.1–2.7), when compared to those with normal vaginal deliveries. Although not significant, pregnant mothers who were teenagers (AOR: 1.6, 95% CI: 0.9–3.1) and middle age (AOR: 1.7, 95% CI: 1.0–2.9), current smokers (AOR: 1.9, 95% CI: 0.8–4.3), low BMI (AOR: 1.3, 95% CI: 0.5–3.3), having a history of sexually transmitted infections (AOR: 4.2, CI: 1.0–18.3,), being HIV positive (AOR: 1.2, 95% CI: 0.8–2.0), or having pregnancy induced hypertension (AOR: 1.2, 95% CI: 0.1–10.6) also had increased odds of LBW babies. Conversely, overweight (AOR: 0.5, CI: 0.3–0.8) and obese mothers (AOR: 0.4, CI: 0.2–0.7) and having syphilis did not appear to influence the outcome. (Table 4).

A concern in our study was the high dropout rate of pregnant women from enrolment to delivery mainly due to relocation. However, this was likely to be random, and thus not expected to have had an impact on the generalisation of our findings. Furthermore, we did not classify LBW as per gestational age or classify individual cases in relation to the extent of prematurity and its related birth weight. An evaluation of the antenatal risk factors against this classification may be useful. The main strengths of this study were large sample size, the ability to identify modifiable risk factors for PB and LBW outcomes and use this information to promote health intervention programmes for pregnant women.