Globally each year there are approximately 1.3 million pregnancies in people with HIV (PPHIV) [1]. In 2019, South African HIV prevalence amongst those of child-bearing potential peaked at 40% in women aged 38–40 years, and antenatal HIV prevalence across provinces ranged from 15% to almost 40% [2,3]. As the world pursues the Sustainable Development Goals of reducing global maternal mortality and ending preventable newborn and child deaths, the potential impact of HIV in pregnancy on realizing these goals in HIV high-prevalence countries must be considered [4].

Increasing access to antiretroviral therapy (ART) reduced HIV's contribution to global maternal deaths (any cause related to or aggravated by pregnancy and its management during pregnancy, childbirth and within 42 days of termination of pregnancy), from 2.5% at the peak of the HIV epidemic in 2005 to 1% in 2017 [5,6]. However in South Africa, HIV-related early maternal deaths are still the leading cause of maternal mortality estimated at 25%, ahead of hypertensive disorders of pregnancy at 18% [6,7]. It is unknown whether HIV and hypertensive disorders of pregnancy, each known to independently impact maternal and fetal well being, interact to further increase risk for adverse maternal and birth outcomes than each condition alone. Furthermore, few studies have examined the impact of HIV on late maternal deaths (>42 days but less than one year after termination of pregnancy) during the universal ART era in South Africa or globally.

While ART is keeping more mothers with HIV alive, in-utero ART exposure appears to impact on stillbirth, fetal growth restriction and preterm birth, varying by ART regimen and potentially timing of ART initiation [8–10]. Irrespective of geography, prior to widespread universal ART access, stillbirth and other adverse birth outcomes remained elevated in PPHIV on ART compared to their counterparts without HIV [8,11]. Stillbirth is the most devastating adverse birth outcome, but any adverse birth outcome, including preterm birth, small for gestational age, or low birth weight (LBW; a combination of preterm neonates and growth-restricted term or preterm neonates), can have short, medium, and long-term consequences on the life-course [12–14]. Moreover, a maternal death is catastrophic for young children and greatly increases their mortality risk [15,16]. To achieve optimal outcomes for both PPHIV and their children, we need a better understanding of the factors associated with birth and maternal outcomes for people with HIV.

To this end, we primarily described the association of HIV and ART with adverse birth (stillbirth, LBW, very low birth weight (VLBW)) and maternal (hospitalization and death) outcomes at a population-level in the Western Cape Province of South Africa in a recent cohort prior to dolutegravir-based first-line ART implementation. Secondarily, we evaluated whether the associations of HIV and ART with birth and maternal outcomes are modified or mediated by de novo hypertensive disorders of pregnancy (dnHDP) [17].

The Western Cape Provincial Government provides free maternal, child and HIV care (including universal ART) accessed by approximately 75% of the province's population, including ± 110 000 pregnant people annually, with a provincial antenatal HIV prevalence of approximately 19%. During the 2018–2019 study period, routine repeated antenatal HIV testing was indicated at a minimum at first antenatal visit, 32-weeks gestation and delivery, with immediate initiation of tenofovir-emtricitabine-efavirenz (TDF-FTC-EFV) as first-line ART in those newly diagnosed or still ART naive [18]. South Africa had not yet commenced dolutegravir-based ART and the standard second-line regimen was protease-inhibitor-based lopinavir/ritonavir (LPV/r). All people accessing Western Cape provincial health services are issued with a unique identifier used in multiple electronic platforms across the healthcare system. Unique child identifiers, when issued at birth, are routinely linked to the maternal identifier. The Western Cape Provincial Health Data Centre (WCPHDC) integrates these multiple sources of individual-level electronic data based on the unique identifiers. The WCPHDC structure, data sources and governance have been previously described [19].

This was a province-wide retrospective cohort study of pregnant people delivering from 1 January 2018 to 31 December 2019 at all 54 public sector healthcare facilities that provide obstetric delivery services in the Western Cape. Algorithms used by the WCPHDC to identify pregnancy, the identification of this specific 2018–2019 maternity cohort and classification of dnHDP have been described [20,21]. For this analysis, the cohort was restricted to the first viable (i.e. not miscarriage or termination) singleton pregnancy with delivery during the study period irrespective of maternal parity.

The primary exposure of interest was the combined HIV/ART status during pregnancy [no HIV, HIV – no ART, HIV – preconception ART (first ART start >294 days before delivery_, HIV – pregnancy ART (first ART start ≤294 days before delivery)], without considering ART interruptions in the primary analysis. Other variables considered as risk factors included dnHDP (yes/no) defined by ICD-10 code (O12-O15.9) or antihypertensives prescribed for the first time <140 days before delivery, maternal age at delivery (continuous and categorized), multiparity (determined by number of previously enumerated pregnancies in the WCPHDC) and residential district (categorized as one metropolitan district (City of Cape Town) and four nonmetropolitan districts) as a crude proxy for urban/rural location and socio-economic status. For PPHIV, HIV diagnosis and ART initiation dates, pregnancy CD4+ cell count and HIV viral load (earliest CD4+ cell count or HIV viral load within 294 days predelivery and 90 days postdelivery) and pregnancy ART regimen were also available. Adverse maternal outcomes evaluated included early pregnancy-related death (any death irrespective of cause during pregnancy, childbirth, until 42 days postdelivery; time period chosen to be consistent with definition of early maternal mortality), late pregnancy-related death (any death 43–365 days postdelivery), early or late pregnancy-related hospitalization (according to same time intervals as death; categorized according to timing of first hospitalization; hospitalizations for uncomplicated delivery admissions were excluded from the early hospitalization outcomes and defined as follows: for vaginal deliveries date of admission ≤1 day prior to date of delivery AND date of discharge not >1 day after delivery; for caesarean section deliveries date of admission ≤1 day prior to date of delivery AND date of discharge not >3 days after delivery). For late pregnancy-related outcomes, rate calculations and statistical models excluded people with early pregnancy-related deaths. Adverse birth outcomes evaluated included stillbirth (as reported by delivery facilities), LBW (live birth <2500 g) and VLBW (live birth <1500 g). Implausible birth weights less than 350 g or more than 6000 g were recoded to missing [22]. For LBW and VLBW outcomes, rate calculations and statistical models excluded stillbirths.

No a priori sample size calculation was performed and all available data for the study period were included. Summary statistics were calculated and compared according to HIV/ART status and described as frequencies with percentage for categorical variables or median with interquartile range (IQR) for continuous variables. Frequency of unknown or missing data in adjustment variables was described, and if more than 10%, a separate unknown category was included in the statistical models, without any further statistical handling of missing data. For the primary objective comparing adverse birth and maternal outcomes by HIV/ART status, unadjusted and adjusted prevalence ratios (PRs) were calculated for individual outcomes using log-binomial regression [23]. Reference categories for categorical variables were chosen according to the category with the lowest prevalence of the outcome. Adjustment for maternal factors included age at delivery, multiparity and residential district. We examined potential effect modification by dnHDP through stratified analyses, and assessed mediation by dnHDP through comparison of the effect of HIV/ART in models including versus excluding dnHDP. We conducted analyses restricted to women with HIV on ART, with the same approach, to compare birth and maternal outcomes by ART duration (since first ART initiation date to delivery) and pregnancy ART regimen. We performed sensitivity analyses excluding women with ART interruptions defined as those with first ART initiation preconception, but no evidence of ART dispensed during pregnancy until at least 90 days after estimated conception for an average 40 week pregnancy. Analyses were conducted with Stata 17.1 (StataCorp, College Station, Texas, USA).

Study approval was received from Stellenbosch University and University of Cape Town Health Research Ethics Committees and the Western Cape Provincial Health Research Committee. A waiver of informed consent was granted as the study included routinely collected data that were de-identified before secure transfer from the WCPHDC to the principal investigator.

For the calendar years 2018–2019, 171,960 pregnant people, of whom 32,015 (18.6%) were identified with HIV, and their singleton newborns were included (Supplementary Figure 1, https://links.lww.com/QAD/C978). Amongst PPHIV, 59.8% (N = 19 157) started ART preconception, 29.0% (N = 9276) initiated ART during pregnancy and 11.2% (N = 3582) had no evidence of ART during pregnancy. Those pregnant without HIV were generally younger than PPHIV, and PPHIV on preconception ART were generally older and most often multiparous (Table 1). dnHDP were present in 7.6% (N = 13 006) of the overall cohort, highest in PPHIV not on ART (9.9%; N = 354). The rate of early pregnancy-related deaths overall was 36.1/100 000, highest in PPHIV on pregnancy-initiated ART (97.0/100 000). The rate of late pregnancy-related deaths overall was 64.6/100 000 but substantially higher in all PPHIV (ranging from 224.6 in the preconception ART group to 251.4/100 000 in the no ART group). Early hospitalization occurred in 14.0% (N = 24 081), slightly higher in PPHIV (ranging from 14.4% in the pregnancy ART group to 15.4% in the no ART group). Similarly, late hospitalization occurred in 2.0% (N = 3497) and was only slightly higher in PPHIV (ranging from 2.1 to 2.7%). There were 2670 stillbirths at a rate of 15.5/1000 births in the cohort, highest in PPHIV not on ART (20.9/1000; N = 75). Amongst live born infants 13.8% (N = 23 375) were LBW and 2.2% (N = 3635) VLBW, both having the highest prevalence in PPHIV not on ART (Table 1).

Amongst PPHIV on ART, 93.5% (N = 24 685) were on TDF-FTC-EFV throughout pregnancy and 5.2% (N = 1355) were on a second-or-higher-line regimen, predominantly LPV/r-based (N = 1312). The median (IQR) ART duration at delivery was 212.6 (131.3; 320.0) weeks in PPHIV on preconception ART and 21.1 (14.6; 27.4) weeks in PPHIV on pregnancy-initiated ART. Median (IQR) pregnancy CD4+ cell count, known in 63.3% (N = 20 269), was 425 (275; 594) cells/μl and did not differ substantially by ART groups. Pregnancy HIV viral load was only available in 21.5% (N = 6874) of PPHIV (Table 2).

In unadjusted analyses amongst PPHIV compared to pregnant people without HIV, only PPHIV not on ART had higher rates of stillbirth, however all ART exposures, either preconception or initiated during pregnancy, were associated with LBW (Table 3). All groups of PPHIV experienced higher early and late pregnancy-related deaths, with PRs indicating a strong association for late pregnancy-related deaths ranging from 8.27 (95% CI 5.35; 12.79) in PPHIV on preconception ART to 9.26 (95% CI 4.48; 19.13) in PPHIV not on ART. In adjusted models including both HIV/ART status and dnHDP, adjusted for maternal age, multiparity, and residential district, dnHDP was strongly associated with all adverse birth and maternal outcomes (Table 4). Adjusted PRs (aPR) for stillbirth were significantly higher only for PPHIV not on ART, however LBW and VLBW were both significantly higher among all PPHIV, most strongly in PPHIV not on ART. While only pregnancy-initiated ART was associated with early pregnancy-related death (aPR 3.21; 95% CI 1.55–6.65) all HIV/ART groups were strongly associated with late pregnancy-related death [no ART: aPR 9.01 (95% CI 4.33–19.75); preconception ART: aPR 7.89 (95% CI 4.99; 12.46); pregnancy- ART: aPR 8.91 (95% CI 5.20; 15.25)]. While there was some association between HIV/ART groups for late pregnancy-related hospitalization, this was much weaker than for late pregnancy-related death. Excluding women with ART interruptions in sensitivity analyses (N = 4340) did not alter the associations with preconception and pregnancy-initiated ART (Supplementary Table 1, https://links.lww.com/QAD/C978). Neither stratification by dnHDP (Supplementary Tables 2, https://links.lww.com/QAD/C978 & 3, https://links.lww.com/QAD/C978 stratified unadjusted and adjusted models) nor adjusting for dnHDP (Supplementary Table 4, https://links.lww.com/QAD/C978, adjusted models without dnHDP) substantially altered the association of HIV/ART with adverse maternal or birth outcomes, thus effect modification and mediation effects were not further explored.

In unadjusted analyses restricted to PPHIV on ART, duration on ART at delivery was associated with all adverse birth outcomes, the lowest prevalence group being 20-<40 weeks of ART (Supplementary Table 5, https://links.lww.com/QAD/C978). The shortest duration on ART (<20 weeks) had the highest unadjusted PR (uPR), however longer durations were also associated with increased uPR for stillbirth, LBW and VLBW in reference to 20-<40 weeks of ART (Supplementary Table 5, https://links.lww.com/QAD/C978). In adjusted models including ART duration, pregnancy CD4+ cell count, any second-or-higher-line regimen, maternal age, multiparity, and residential district, ART duration remained associated with adverse birth outcomes, with the highest aPR for ART <20 weeks, but still elevated aPRs for ART durations >40 weeks (Table 5). For adverse maternal outcomes, there was little association with ART duration in unadjusted or adjusted analyses. There was, however, a strong association of CD4+ cell count <350 cells/μl versus ≥500 cells/μl for late pregnancy-related death (aPR 5.07; 95% CI 2.10–12.26) and weaker associations with early (aPR 1.15; 95% CI 1.06–1.25) and late hospitalization (aPR 1.42; 95% CI 1.16–1.75). Additionally, any second-or-higher-line regimen remained associated with LBW, early and late hospitalization and late pregnancy-related death in adjusted models. Further adjusting for dnHDP did not meaningfully alter any of the HIV-specific associations with adverse birth or maternal outcomes, but dnHDP was retained in the final adjusted models as it was strongly associated with all adverse birth and maternal outcomes. Similarly, excluding CD4+ cell count and second-or-higher-line regimen from the models did not meaningfully alter any of the associations with ART duration but they were retained in the final models (Supplementary Table 6, https://links.lww.com/QAD/C978).