Background A major factor contributing to continued high under-five mortality in sub-Saharan Africa (SSA), is maternal HIV infection, which is associated with adverse birth outcomes such as preterm delivery (PTD), small-for-gestational age (SGA) and low birthweight (LBW) infants. Introduction of antiretroviral therapy (ART) during pregnancy has been a successful intervention for promotion of maternal and infant health, however it has also been linked to an increased risk of adverse birth outcomes. Consequently, the association between maternal ART use and these adverse outcomes is an important area of research in SSA which has the majority of pregnant women living with HIVand the highest rates of PTD, SGA and LBW infants. However, the current state of epidemiologic knowledge remains limited because most evidence on this association comes from observational studies, which have previously given inconsistent findings. Accordingly, the overarching aim of this PhD was to reliably quantify the relationship between maternal ART use and adverse birth outcomes, by addressing the role of methodological factors inherent in observational research in this association. Methods This research included pregnant women (aged ≥18 years) seeking antenatal care at a public sector midwife obstetric unit in Gugulethu (GMOU), Cape Town, enrolled into two separate dedicated research cohort studies. Enrolled women were followed-up during pregnancy and postpartum with their infants, with data obtained from study questionnaires, physical examinations and abstraction of clinical and obstetric records. In parallel, routine electronic data, linked across clinics and data sources were obtained for all pregnant women at the GMOU (pregnancy exposure registry (PER)) and across the Western Cape province (PHDC). Findings The incidence of gestational age (GA) based birth outcomes and the association between maternal ART use and these outcomes, was found to be substantially influenced by method of GA assessment used. While GA based on both last menstrual period (LMP) and measurement of symphysis fundal height (SFH) led to under and over estimation (relative to ultrasound), only LMP-based GA gave rise to a biased measure of association. Across data sources used in this research, an overall PTD incidence of 17% was observed which was lower than incidences observed in the pre-universal ART era. In the cohort studies, there was no significantly increased PTD risk in women living with HIV (compared to living without HIV), predominantly on the tenofovir + emtricitabine + efavirenz regimen. However, when assessed in the significantly larger population of pregnant women (PHDC), an increased PTD risk in women living with HIV was observed. There did not appear to be differences in PTD risk by ART status in the cohort studies or PER. However, across the province those initiating ART preconception were at increased PTD risk compared to those initiating during pregnancy. Blood pressure, particularly when assessed longitudinally played an important role in the association between maternal ART use and PTD, high normal and abnormal blood pressure trajectories associated with increased PTD risk. There did not appear to be any effect modification in the trajectory groups by HIV status for PTD. In the cohort study the overall incidence of SGA infants was 9%, with an increased SGA risk observed in women living with HIV compared to living without HIV. While no differences were observed in SGA risk by ART status, the highest risk was observed among women initiating in the second trimester. An overall LBW incidence of 13% was observed in the cohort study, with no differences observed in risk by HIV status or ART status. Blood pressure also played a role in the occurrence of LBW infants, with abnormal trajectories associated with increased LBW infant risk. Additionally, effect modification among women with abnormal trajectory groups, with women living without HIV at increased risk of LBW infants compared to women living HIV was observed. Finally, investigating this association using both cohort studies and population based electronic health care data proved to be valuable. The three data sources gave similar effect estimates, with varying levels of precision, and each with distinct but complementary benefits. The cohort studies and PER included smaller select groups of women and provided detailed investigation of risk factors that could impact the overall association. In contrast the provincial dataset, had limited risk factor data, but provided overall associations with the ability to detect subtle differences. Conclusion Reassuringly, the magnitude of difference in PTD risk by HIV status under policies of universal maternal ART use, appears to be smaller than in the past. Of concern, however, was the finding of increased SGA risk. Taken together, these findings highlight the need to improve mechanistic understanding of ART-mediated adverse birth outcomes, in order to ensure optimal maternal and infant outcomes. The methodological findings underscore the importance of considering the potential for bias related to selection and measurement when designing and/or evaluating findings from studies investigating this association and by extension other medications in pregnancy.