Adverse Birth Outcomes in Botswana: Findings from the Tsepamo Study

Abstract

The Botswana-Harvard AIDS Institute Partnership (BHP) was formed in 1996 between the Government of Botswana’s Ministry of Health and the Harvard T.H. Chan School of Public Health. Research conducted under this partnership includes studies to prevent mother-to-child HIV transmission, or vertical transmission; HIV treatment studies, including treatment trials in pregnancy; and HIV pediatric cure initiatives. The Tsepamo Study (PI: R. Shapiro, R01HD080471, R01HD095766), a surveillance study that collects obstetric and birth outcomes data from women who deliver live-born or stillborn infants at 24 weeks gestation or later within government-funded hospitals, has been operating under the direction of BHP since 2014. When the Tsepamo Study was first launched its aims included assessing associations between ART regimens used during pregnancy and adverse birth outcomes, and identifying and exploring neural tube defects among women who used efavirenz-containing ART during the first trimester. Over the past decade, the original Tsepamo aims have expanded to include evaluating new ART regimens in pregnancy, including dolutegravir-based ART; assessing the impact of additional exposures in pregnancy, including antibiotics; tracking policy changes related to ART use in pregnancy and their effect on adverse birth outcomes; and determining the distribution of major congenital abnormalities in Botswana. This dissertation focuses on three major analyses of the Tsepamo dataset: 1) measuring associations between antibiotics used for vaginal discharge syndrome and adverse birth outcomes, 2) determining the impact of the expansion of Botswana’s HIV treatment program to include non-citizens on adverse birth outcomes, and 3) evaluating the distribution of major congenital abnormalities in Botswana over time and across seasons. In Chapter 1, we explore the associations between antibiotic use for treating diagnosed vaginal discharge syndrome (VDS) during pregnancy and adverse birth outcomes. The adverse birth outcomes included preterm and very preterm delivery, small and very small for gestational age, neonatal death, and stillbirth. Due to the detection of immortal time bias when assigning our comparison groups as treated vs. untreated, our analysis was restricted to women who presented to care before 24 weeks of gestation. As a result of the generated log-binomial regression models, we reported a modest association between treated VDS and VPTD and no harmful associations related to antibiotic use 24 weeks gestation and adverse birth outcomes. Our findings remained robust when conducting various sensitivity analyses. Our findings did not provide evidence to alter the current syndromic treatment method for diagnosed VDS in pregnancy. However, due to the limitations of our analysis, mainly the restriction of the study population to pregnant women presenting to care 24 weeks gestation, we acknowledge the limited generalizability of our findings. In Chapter 2, we determine the impact of expanding the Government of Botswana’s HIV Treatment policy to include non-citizens living with HIV in Botswana. We compared the proportion of pregnant women in Tsepamo who had did not have a recorded HIV status, attendance of at least one ANC visit, and specific ART regimens pre- vs. post-ART policy expansion between citizen and non-citizen pregnant women living with HIV. We then generated log-binomial regression models to determine the association between citizenship status and the same adverse birth outcomes as stated in Chapter 1. In the pre-ART expansion period, non-citizen pregnant women had a greater risk of all adverse birth outcomes, with the exception of small and very small for gestational age. However, in the post-ART expansion period, these associations became non-significant. To account for bias due to calendar year, we conducted a similar analysis among citizen and non-citizen pregnant women without HIV, and we did not see a similar result. While causality could not be determined in this study, there is a beneficial impact as a result of reducing barriers to care among non-citizens living with HIV. In Chapter 3, we estimate the prevalence of major congenital abnormalities (MCAs) among women without HIV in Botswana. We evaluated the prevalence of any MCA, and neural tube defects (NTDs) specifically, by estimated calendar year of conception and estimated month and season of conception and identified any potential associations using Chi-square and Benjamini-Hochberg methods. We assessed if there was any meaningful variation of MCAs by delivery site in these analyses. Our findings included that the anatomical region impacted the most by MCAs was the limbs, followed by the nervous system (including neural tube defects), and the musculoskeletal region. We did not identify any meaningful variability of MCAs by calendar year or season of conception. However, there was notable variation of NTDs by calendar year with the lowest prevalence occurring among mothers who conceived in 2019. We also identified seasonal variation of NTDs, with its peak occurring in the winter/early dry season. Lastly, the highest prevalence MCAs were in the tertiary referral centers in Botswana. While our ability to identify MCAs was limited to conditions that could be detected on surface examinations among deliveries that occurred at 24 weeks gestation or later, to our knowledge, this was the first study in Botswana to describe associations between MCAs and seasonality, calendar year, and delivery site.