The contribution of intimate partner violence to vertical HIV transmission: a modelling analysis of 46 African countries

Summary

Background:

Addressing gender inequities could be key to eliminate mother-to-child transmission of HIV (MTCT). Women experiencing intimate partner violence (IPV) may be at an increased risk of MTCT due to their vulnerability to HIV acquisition and barriers to access and retention in care. Sub-Saharan Africa, where IPV burden is among the highest globally, accounts for most new pediatric infections. We aimed to examine the proportion of excess MTCT attributable to IPV in this region.

Methods:

We created a probability tree model of vertical HIV transmission among women (15–49 years) in 46 sub-Saharan African countries. We estimated the proportion of MTCT attributable to past-year physical or sexual IPV (or both), as an age-standardized population attributable fraction (PAF) and MTCT risk per 1,000 births. We incorporated perinatal and postnatal MTCT among women who acquired HIV before pregnancy, during pregnancy, and during breastfeeding. Fertility, HIV prevalence, HIV incidence, ART uptake, and ART retention varied by women’s IPV experience. The model was parametrized using UNAIDS’ 2023 Spectrum model, WHO’s Global Database on Violence Against Women, and the peer-reviewed literature. We calculated uncertainty intervals (95%UI) through 1,000 Monte Carlo simulations.

Findings:

Across 46 countries 13% (95%UI:6–23%) of pediatric infections were attributed to IPV in 2022, corresponding to over 90,000 pediatric infections. PAF ranged from 4% (95%UI:2–7%) in Niger to 28% (95%UI: 13–43%) in Uganda. PAF was highest among 15–19-year-old women (20%; 95%UI:8–33%) and lowest among 45–49-year-olds (6%; 95%UI:3–9%). In Southern Africa, where women’s HIV prevalence is highest (23%), ending IPV could avert 11 (95%UI:5–20) infections per 1,000 births.

Interpretation:

IPV may be responsible for one in eight pediatric infections in sub-Saharan Africa. Ending IPV could accelerate MTCT elimination, especially among young women who bear the highest burden of violence.

Funding:

Canadian Institutes of Health Research, Canada Research Chair, and Fonds de recherche du Québec-Santé.

Introduction

New pediatric infections from mother-to-child transmission (MTCT) of HIV have declined by 58% since 2010.¹ Still, 130 000 children acquired HIV in 2022 globally.¹ Most (85%) of these infections occurred in sub-Saharan Africa.¹ Reductions in MTCT of HIV are largely attributed to increased coverage of HIV testing and antiretroviral treatment (ART) among women living with HIV (WLHIV).¹ However, ART coverage among pregnant WLHIV has recently plateaued at a little over 80%.¹ The 2022 Global Alliance to End AIDS in Children aims to close the prevention and treatment gaps to eliminate MTCT by 2030. It recognizes that structural drivers of HIV are key to achieving this goal. The United Nations Political Declaration on HIV and AIDS further identifies gender-based violence, including intimate partner violence (IPV), among these drivers and commits to reducing its global burden from 27%² to less than 10% by 2025.³ Shedding light on relationships between IPV and vertical HIV transmission is key to inform MTCT elimination strategies.

Sub-Saharan Africa has among the highest IPV prevalence globally, with over 1 in 5 women having experienced IPV in the past year.² IPV could contribute to increasing MTCT risk in several ways. Women subjected to IPV are more likely to acquire HIV,⁴ mainly through indirect pathways driven by interpersonal and societal gender inequities.⁵ In the context of prevention of mother-to-child transmission (PMTCT): women experiencing IPV have lower rates of HIV testing and antenatal care (ANC) engagement⁶, lower uptake of PMTCT⁷, and poorer viral suppression.⁴ Some forms of IPV, such as forced sex and reproductive coercion, may also contribute to increases in pregnancies.⁸ Adverse effects of IPV may compound the hormonal and immunological drivers of the heightened risk of HIV acquisition among pregnant compared to non-pregnant women.⁹ Women who acquire HIV during pregnancy or breastfeeding may have a higher rate of MTCT due to the initial high viral load following seroconversion.¹⁰ Finally, adolescent girls and young women may be at higher MTCT risk, since they are the most vulnerable to IPV² and have lower rates of viral suppression than older women.¹¹

A comprehensive analysis of the contribution of IPV to MTCT, along the full prevention and treatment cascade has not been undertaken. Empirical studies exploring the adverse impact of IPV on the PMTCT cascade have either been inconclusive, focused on a single setting,¹² were qualitative,¹³ or only studied one component of the care continuum.¹⁴ PMTCT scale-up has reduced the number of pediatric HIV infections, making it challenging to empirically estimate vertical transmission.¹ Finding a common effect size for IPV-MTCT relationships is further complicated by the time and setting-dependent variability in PMTCT coverage and uptake, which lie on the pathway between IPV and vertical HIV transmission.

The goal of this study was to estimate the contribution of past-year physical or sexual IPV (or both) on MTCT of HIV. We aimed to quantify the annual proportion of excess risk of MTCT attributable to women’s experience of past-year physical or sexual IPV by women’s age in sub-Saharan Africa. To achieve this, we developed a probability tree model, parameterized through literature reviews and PMTCT program data.

Methods

Study design

A probability tree model for women (15–49 years), stratified by five-year age groups, was developed for the period 2014–2022 (Figure 1).¹⁵ The model was based on the MTCT module of the Spectrum AIDS Impact Model (AIM) (v6.28),¹⁵ used by the Joint United Nations Programme on HIV/AIDS (UNAIDS) to estimate HIV trends from surveillance and survey data.¹⁵

Figure 1. Flowchart of the structure of the probability tree model of the impact of intimate partner violence (IPV) on mother-to-child-transmission of HIV.

The model includes women who acquire HIV before pregnancy, during pregnancy and during breastfeeding. Parameters in red boxes are impacted by IPV. ART regimens that women can be enrolled on during pregnancy include: single dose nevirapine, dual prophylaxis, Option A, Option B, Option B+ (>4 weeks or <4 weeks before delivery). Further detail on these regimens can be found in Table S6 (pp 26). Perinatal ART retention, representing the proportion of women retained in ART at delivery, is incorporated for women on Option B+ (>4 or <4 weeks before delivery) and women on ART before pregnancy. Postnatal ART retention, representing monthly postnatal dropout rate, is incorporated for women on Option A, B, B+ (>4 weeks or <4 weeks before delivery). ANC = antenatal care; ART = antiretroviral treatment; HIV+ = living with HIV; HIV− = not living with HIV; IPV+ = experiencing past-year physical or sexual intimate partner violence; IPV− = not experiencing past-year physical or sexual intimate partner violence; MTCT = mother-to-child HIV transmission

Our model considers MTCT during the perinatal and postnatal periods. It assumed that women’s fertility rate varies by IPV, HIV status, ART uptake, and CD4 cell counts. Women not already living with HIV before conception can acquire it during pregnancy or breastfeeding, considering the additional risk of HIV acquisition during these periods (compared to non-pregnancy or non-breastfeeding). This departs slightly from Spectrum’s assumption that women have the same incidence regardless of pregnancy status. While HIV acquisition risk is higher per-condomless-coital act during pregnancy than non-pregnancy, reduction in sexual activity perinatally might mitigate this risk.¹⁶ Given the heterogeneity in sexual activity patterns, we adhered to the assumption of higher risk by 2.16⁹ and 1.16⁹ for pregnancy and postpartum periods, respectively.¹⁶

Women acquiring HIV during pregnancy/breastfeeding will not be diagnosed and enrolled on ART.¹⁷ For WLHIV before conception, the model incorporates HIV testing at ANC and ART regimens for pregnant women. Women may not receive ART either by not testing for HIV at (or attending) ANC or by testing but not enrolling in care. The country-specific proportion of breastfeeding WLHIV reduces over time, up to 36 months. The probability of MTCT varies by CD4 cell counts, ART regimen, and perinatal/postnatal transmission period.

Procedures

Demographic parameters, PMTCT program data, and HIV projection outputs

Model parameters relied on country-reported PMTCT program data and demographic projections from publicly available 2023 Spectrum projection files.¹⁷ For Djibouti, Mauritius and Nigeria, the 2023 files were unavailable and their 2022 Spectrum files (2014–2021) were used instead. ANC testing data in South Africa was extracted from the proportion of pregnant women tested for HIV at ANC used in Thembisa 4.7.¹⁸ Demographic parameters include age-, year- and country-specific fertility rate, as well as rate ratios accounting for the impact of HIV and ART uptake on fertility. Annual, country-specific PMTCT program data were extracted from Spectrum files: proportion tested for HIV at ANC, proportion on ART by regimen (“ART uptake” hereon), proportion breastfeeding and duration (up to 36 months postnatally), and proportion retained on ART peri- and postnatally. HIV transmission probabilities varied by ART regimen and transmission period (perinatal or postnatal), and CD4 cell counts (<200, 200–350, >350 cells per µL). Annual, country-specific HIV prevalence and cumulative HIV incidence over one year were extracted for women by five-year age group (Table S2; pp 15–16).

Prevalence of past-year physical or sexual (or both) intimate partner violence

Estimates of past-year physical or sexual IPV (or both; subsequently referred to as physical or sexual IPV) prevalence in 2018 were obtained for each country by five-year age group from the Global Database on the Prevalence of Violence Against Women.² We restricted our analysis to four years before and after 2018 (2014–2022) to ensure the validity of the IPV prevalence estimates from 2018. This prevalence was assumed to be constant over time.² Experience of past-year IPV, as opposed to lifetime, was the preferred exposure since recent IPV experiences have a more direct causal link with model parameters. We excluded psychological violence from the IPV definition, due to a lack of agreement on how to universally define and quantify it cross-culturally.¹⁹ The model conservatively assumes that pregnancy does not affect the risk of experiencing IPV.

Impact of intimate partner violence on mother-to-child transmission of HIV

Fertility rate, ART uptake, ART retention, cumulative HIV incidence and HIV prevalence varied by past-year IPV experience, using estimates from meta-analyses of nationally representative surveys (Table 1). Hazard ratio for IPV’s impact on fertility was based on a meta-analysis of 29 population-representative surveys in low-and-middle-income countries (Table 1).⁹ Prevalence ratios for the effect of past-year IPV on cumulative HIV incidence, ART uptake and ART retention among WLHIV were informed by a meta-analysis of six nationally representative surveys in sub-Saharan Africa.⁴ The odds ratio for the relationship between lifetime IPV and HIV prevalence was obtained from a meta-analysis of 12 Demographic and Health Surveys (DHS).²⁰

Table 1.

Effect size estimates for the relationship between past-year physical or sexual intimate partner violence and model parameters relevant to mother-to-child transmission of HIV.

Model parameters affected by intimate partner violence	Adjusted effect estimate (95% CI)*	Source
Effect of intimate partner violence on model parameters
ART uptake before and during pregnancy (%)§	aPR = 0·96 (0·90–1·02)	Kuchukhidze et al. 20234
ART retention peri- and postnatally among WLHIV on ART‡ (%)	aPR = 0·95 (0·90–1·00)	Kuchukhidze et al. 20234
Cumulative HIV incidence over one year†	aPR = 3·22 (1·51–6·85)	Kuchukhidze et al. 20234
HIV prevalence (%)¶	aOR = 1·10 (1·01–1·21)	Durevall et al. 201020
Fertility rate**	aHR = 1·13 (1·07–1·20)	Maxwell et al. 20178

^*A full version of this table, including adjustment variables for the effect estimates is available in Table S3 (pp 19–20).

^§We assume that ART uptake for WLHIV on each ART regimen during pregnancy and before pregnancy is the same as ART uptake among all WLHIV.

^‡aPR for viral suppression among WLHIV by IPV status is used as a proxy estimate for the effect of IPV on perinatal and postnatal ART retention among pregnant WLHIV. Since postnatal ART retention is reported as monthly postnatal ART dropout rate in Spectrum, we operationalized this aPR as an aHR.

^†aPR for recent HIV infection by IPV status is used as a proxy estimate for the effect of IPV on cumulative HIV incidence over one year. Measurement of recent infection is based on a Lag-avidity assay.

^¶aOR represents the effect of lifetime IPV on HIV prevalence to account for the fact that women might have seroconverted prior to experiencing past-year IPV.

^**aHR represents the effect of any IPV on the probability of incident pregnancy.

ART= antiretroviral therapy; CI = confidence interval; aHR = adjusted hazard ratio; IPV = intimate partner violence; aPR = adjusted prevalence ratio; aOR = adjusted odds ratio; WLHIV = women living with HIV

To understand the impact of past-year IPV on HIV testing at the ANC, we analysed 28 DHS surveys with information on IPV, HIV testing and HIV biomarkers among WLHIV who gave birth in the past year. We did not find evidence that IPV impacted ANC testing, consistent with previous work.⁴ Therefore, HIV testing at ANC does not vary by IPV in our model (Table S3 (pp 19–20)).

Outcomes

Our primary outcome was the population attributable fraction (PAF) or fraction of all MTCT caused by IPV. We also estimated risk difference (RD) as the number of MTCT cases that would be averted per 1,000 births if the effect of IPV was eliminated.

The usual denominator for MTCT rate calculation is births among WLHIV. However, women with incident HIV have an added risk of MTCT due to IPV, and the increased HIV acquisition risk during pregnancy and postpartum. This added risk does not apply to WLHIV before conception. To address this, we used all births as the denominator.

To account for confounding of the IPV-MTCT relationship by age, we calculated age-standardized RD and PAF. The standard population was all births (for RD) and vertical transmissions (for PAF) by country and year, from Spectrum’s demographic projections and the probability tree, respectively.²¹

We calculated PAF and RDs for each country and year (2014–2022). We also calculated PAF and RD aggregating the numbers of MTCT and births from 2022 across the four subregions and overall. In Nigeria, Djibouti, and Mauritius we carried over the most recent available data to 2022. This was 2020 for Nigeria and 2021 for the latter two countries. We present PAF stratified by perinatal versus postpartum period and by women with prevalent versus incident HIV.

Uncertainty estimation

We incorporated two sources of uncertainty. First, uncertainty related to the effect size estimates for the impact of IPV on model parameters (Table 1). Second, uncertainty related to perinatal and postnatal MTCT transmission probabilities (Table S5, pp 26). We estimated 95% uncertainty intervals (UI) via 1,000 Monte Carlo simulations where effect size estimates were resampled from lognormal distributions and transmission probabilities from logit-normal distributions.

Sensitivity analyses

We conducted a sensitivity analysis for the impact of parameters on PAF (Table 1). Adjusting for IPV, we calculated partial correlation coefficients (r) between PAF and ART uptake, ART retention, cumulative HIV incidence, HIV prevalence and fertility rate. Further, we conducted a scenario analysis where we set the effect estimates for the relationship between IPV and the model parameters (Table 1) to null (e.g., PR=1) one at a time and assessed the change in PAF. Finally, we assumed the absence of added risk of HIV acquisition during pregnancy and breastfeeding.

Role of the funding source

The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

Results

A total of 46 countries in sub-Saharan Africa had available data. Regionally aggregated results are only presented for calendar years when data were available for at least 50% of the total population in the region (i.e., 2014 onwards for Southern Africa, 2016 for Central Africa, and 2017 onwards for Eastern, Western and all sub-Saharan Africa).

Aggregating data from 46 countries in 2022, 13% (95%UI: 6–21%) of pediatric infections may have been attributable to IPV. PAF ranged from 4% (95%UI: 2–7%) in Niger to 28% (95%UI: 13–43%) in Uganda (Figure 2A). The lowest PAF in Niger was consistent with a combination of low IPV prevalence (13%) and low ART uptake before pregnancy (12%) in 2022 (Figure 2B–2C).

Figure 2.

A) Age-standardized population attributable fraction (PAF) of past-year physical or sexual intimate partner violence (IPV) in sub-Saharan Africa in 2022 in each country. B) Overall (non-IPV stratified) antiretroviral treatment uptake prior to pregnancy among women in sub-Saharan Africa in 2022 in each country. C) Past-year physical or sexual intimate partner violence prevalence in sub-Saharan Africa in 2018 as reported in the Global Database on Violence Against Women. In Nigeria, Djibouti, and Mauritius we carried over the most recent available data (obtained from Spectrum 2022 projection files) to 2022. This was 2020 for Nigeria and 2021 for the latter two countries. ART=antiretroviral treatment; IPV=intimate partner violence; PAF= population attributable fraction.

In 2022, PAF were largest in Eastern (19%; 95%UI:9–29%) and Southern Africa (18%; 95%UI: 8–30%) where almost one fifth of all MTCT was attributable to IPV (Figure S1A). The highest PAF in Southern and Eastern Africa was consistent with the highest ART uptake (73% and 65% respectively), in addition to the high prevalence (24%) of past-year IPV² in Eastern Africa (Figure S1B–S1C, pp 2). This could suggest that a portion of IPV’s effect on MTCT acts through ART uptake: in regions with the highest ART uptake, IPV elimination would be expected to prevent the largest percentage difference in failure to uptake and remain on ART, with subsequent impacts on MTCT. In contrast, regions with lower ART uptake would experience a relatively smaller impact from IPV elimination.

In 2022, PAF was the lowest in Western Africa with 8% (95%UI:3–13%) of all MTCT due to IPV (Figure S1A, pp 2). IPV prevalence (15%; 95%UI: 11–19%)² and ART uptake prior to pregnancy (30%) were also low in this region (Figure S1B–S1C, pp 2).

We did not observe major temporal trends in PAF (Figure S2, pp 3), likely due to the plateau in ART uptake among pregnant WLHIV in recent years. Southern Africa was the exception and, for instance, PAF increased from 10% (95%UI:5–16%) in 2014 to 21% (95%UI:10–32%) in 2022 for Botswana (Figure S2, pp 3). These changes parallel the increased ART uptake before pregnancy in Southern Africa (Figure S3, pp 4). Between 2014 and 2022, ART uptake increased from 43% to 88% in Botswana.

Pooling data from 2022 in sub-Saharan Africa, the RD (rate of MTCT that would be averted through elimination of IPV) was 2·4 (95%UI:1·1–3·9) pediatric infections per 1,000 births. Averted MTCT vary proportional to HIV prevalence (r=0.9). In Eswatini, where HIV prevalence was 35% in 2022, 14·7 (95%UI:6·9–24·4) infections per 1,000 births could be averted by ending IPV compared to only 0·02 (95%UI:0·01–0·03) in Comoros, where HIV prevalence is less than 1% (Figure 3). In 2022, the greatest number of MTCT cases would be averted by elimination of IPV in Southern Africa (11·1, 95%UI:4·8–19·5) per 1,000 births – the region with the highest HIV prevalence (23%) (Figure S4, pp 5).

Figure 3.

A) Age-standardized risk difference (RD) for the effect of past-year physical or sexual intimate partner violence (IPV) on mother-to-child transmission of HIV in sub-Saharan Africa in 2022. B) Overall (non-IPV stratified) HIV prevalence among women in sub-Saharan Africa in 2022. In Nigeria, Djibouti, and Mauritius we carried over the most recent available data (obtained from Spectrum 2022 projection files) to 2022. This was 2020 for Nigeria and 2021 for the latter two countries. IPV=intimate partner violence; RD=risk difference.

Across all subregions the highest PAFs were estimated for adolescent girls and young women who experienced the highest levels of past-year IPV² (Figure 4). In sub-Saharan Africa in 2022, PAF was highest among 15–19-year-old women (20%; 95%UI:8–33%) and lowest among 45–49-year-olds (6%; 95%UI:3–9%). Country-specific analysis confirms the highest PAF among the youngest age groups (Figure S5, pp 6).

Figure 4.

Age-stratified population attributable fraction overall (sub-Saharan Africa), by subregion and age group (years) in 2022. The shaded area refers to the 95% uncertainty intervals. PAF=population attributable fraction.

Stratified PAF showed that twice as much MTCT was attributable to IPV during the postnatal (18%, 95%UI:7–29%) than during the perinatal period (10%, 95%UI:5–14%) (Figure 5). Smaller fractions of women can transmit HIV perinatally than postnatally because the former occurs among WLHIV before pregnancy and during pregnancy. Meanwhile postnatal transmission occurs among both groups, plus women who acquire HIV during breastfeeding. The added risk of HIV acquisition and vertical transmission due to IPV from women with incident HIV contributes to the high PAF postnatally. PAF is not impacted by the nine-month and 36-month time horizons for pregnancy and breastfeeding respectively since these periods do not vary by IPV in our model.

Figure 5.

Pooled proportion of mother-to-child HIV transmission attributable to past-year physical or sexual intimate partner violence (IPV) stratified by perinatal versus breastfeeding periods and by the timing of HIV acquisition in women, in 2022. The error bars refer to the 95% uncertainty interval. In Nigeria, Djibouti, and Mauritius we carried over the most recent available data (obtained from Spectrum 2022 projection files) to 2022. This was 2020 for Nigeria and 2021 for the latter two countries. IPV= intimate partner violence; PAF=population attributable fraction.

Women’s higher risk of HIV acquisition due to IPV during pregnancy and breastfeeding compared to non-pregnancy and non-breastfeeding could explain that, among women with prevalent HIV, only 3% (95%UI:1–5%) of pediatric infections were due to IPV, while among those with incident HIV 32% (95%UI:11–54%) were attributable to IPV (Figure 5). Country-specific analyses confirm these observations (Figure S6–S9, pp 7–10).

Controlling for IPV prevalence, ART uptake prior to pregnancy was most correlated with PAF (r=0·8), followed by ART uptake during pregnancy (r=0·6) (Figure S10, pp11). When setting the effect estimates for IPV’s impact on model parameters to a null value one by one (compared to no change in the parameter), the strength of association between IPV and HIV incidence in women had the largest impact (a 10% reduction in PAF) (Figure S11, pp12). Setting the effect estimate for the relationship between pregnancy and HIV acquisition risk to a null decreased PAF by 2%.

Discussion

Using data from 46 African countries, our probability tree model estimated that over 1 in 8 pediatric HIV infections would have been averted through elimination of IPV in 2022. This corresponds to over 90,000 pediatric infections averted if IPV was eliminated. The proportion of MTCT attributable to IPV varied widely. IPV has the greatest impact among adolescent girls and young women.

The high PAF for IPV in Eastern Africa is driven by the high prevalence of past-year IPV.² This was similar in age stratified analysis: adolescent girls and young women have the highest PAF across all sub-regions which is due to the high burden of past-year IPV in the youngest age groups. More than one in six girls aged 15–19 years have experienced IPV in the past year.²

Southern Africa has the lowest IPV burden with 15% of women experiencing past-year IPV², but the second highest PAF in our study. Two pathways through which IPV affects MTCT can explain this finding. First, via reducing ART uptake before pregnancy among WLHIV. Second, via HIV acquisition among pregnant and breastfeeding women. In regions with high ART uptake, IPV could lead to a larger absolute reduction in ART uptake, and a subsequent rise in MTCT. Conversely, where ART uptake is low, the added benefit of eliminating IPV in preventing MTCT would be relatively smaller. Our sensitivity analyses confirm the importance of ART uptake prior to pregnancy in explaining country variations in PAF.

In high ART uptake settings such as Eastern and Southern sub-Saharan Africa, HIV incidence is also higher, which can affect the second pathway between IPV and MTCT. High incidence contributes to PAF by amplifying the role of IPV in women’s risk of HIV acquisition and vertical transmission during pregnancy and breastfeeding. Indeed, stratifying PAF by the timing of women’s HIV acquisition shows that the proportion of pediatric infections from IPV is much larger among women with incident HIV compared to those already living with HIV at conception. Modelling studies corroborate that MTCT among women who seroconvert during pregnancy accounts for a big portion of all HIV transmissions, despite representing only a small proportion of all pregnant WLHIV.²² Combined effect of the high initial viral load after seroconversion and increased risk of HIV acquisition during pregnancy could contribute to the higher MTCT risk among women with incident HIV. This risk is especially pronounced in young women, who in sub-Saharan Africa account for almost four in five new acquisitions in youth.¹

Other pathways between IPV and MTCT could also play a role, though the correlation between ART uptake during pregnancy and PAF is weaker than the one for ART uptake before pregnancy. This is consistent with evidence suggesting that women who begin their treatment early have the lowest rates of MTCT, due to achieving viral suppression sooner.²³

Our study has several limitations. First, PAF assumes a causal relationship between the IPV and MTCT. The estimates of IPV’s impact on model parameters are derived from observational studies whose methods might still lead to residual biases. Thus, we explored the impact of key parameters in sensitivity analyses. However, our model outcomes are likely conservative, since they do not capture averted MTCT with the elimination of lifetime experience of IPV. Second, PAF interpretation relies on the complete elimination of the exposure. While there are no silver bullets to fully eliminate IPV, several interventions to tackle gender-based violence have been effective.²⁴ It is imperative for the global advocacy and research agenda to be guided by IPV elimination goals. IPV is a fundamental human rights violation, with tolerant and condoning attitudes standing out as major risk factors. Third, our model may be subject to structural misspecification and may not capture all features of MTCT. For example, we assume that women who acquire HIV during pregnancy are not engaged in PMTCT. Although some women might be identified and enrolled in PMTCT, this assumption is supported by existing literature on low rates of HIV retesting at ANC.²⁵ Further, we did not incorporate the impact of IPV on breastfeeding initiation and duration because previous conflicting evidence from population-based surveys.²⁶ Fourth, the most recent estimates of IPV prevalence date from 2018, thus not accounting for longitudinal trends in IPV, including COVID-19. We present data for years proximate to 2018, and IPV prevalence declined by a small average annual rate of 0.2% between 2000–2021 in low-and-middle income countries.²⁷ Finally, our model used estimates of past-year IPV among all women and not specifically pregnant women. However, estimates of IPV prevalence during pregnancy vary widely²⁸, and the evidence is sparce on variation in levels of IPV before, during and after pregnancy.²⁹

Strengths of our study include our novel application of probability tree models to account for the temporal relationships between IPV and vertical HIV transmission. We used country-reported HIV program data and estimates of key HIV indicators from the UNAIDS-supported Spectrum model. These were complemented with meta-analyses, secondary analyses of population-representative surveys, and community-based cohort studies to parametrize our model. We conducted multiple sensitivity analyses to understand the mechanisms through which IPV effects vertical HIV transmission.

Our results have important policy implications for achieving MTCT elimination. Improving ART coverage among pregnant WLHIV which is still lagging in high HIV burden settings, should be prioritized. Repeated HIV testing in late pregnancy or breastfeeding would identify recently infected women and expedite their enrollment in PMTCT. Concomitantly, identifying women experiencing IPV and supporting them to remain in care is important. Differentiated service delivery models could help fill treatment gaps.³⁰ In settings where ART uptake is already high, reduction in IPV could be an important, final hurdle to accelerate MTCT elimination. Interventions could have the largest population-level impact on MTCT by focusing on younger age groups, given that adolescent girls and young women carry a disproportionate burden of IPV and HIV acquisition.

Progress in reducing new HIV infections in sub-Saharan Africa must be accompanied with the corresponding reduction in IPV to achieve MTCT elimination. Reaching this goal requires addressing structural vulnerabilities affecting women beyond IPV, such as poverty and educational attainment. Repercussions of the overlap between IPV and HIV have long-lasting effects on hundreds of thousands of infants globally.

Research in Context.

Evidence before this study

We searched PubMed for empirical and modelling studies (November 23, 2023), without language restrictions using the terms: (vertical HIV transmission OR MTCT OR mother-to-child HIV transmission) AND women AND (violence OR intimate partner OR domestic violence OR GBV OR IPV OR marital violence) AND (Africa* OR sub-Sahara*).

Most existing studies are qualitative and focus on the impact of IPV on prevention of mother-to-child HIV transmission (PMTCT). Empirical studies from Ethiopia, Tanzania, and Mozambique have shown that women who experience IPV have lower rates of HIV testing and antenatal care engagement compared to those who do not. Systematic reviews of the adverse effects of IPV on pregnant women living with HIV (WLHIV) in sub-Saharan Africa demonstrate poor uptake of and adherence barriers to PMTCT interventions among women experiencing IPV. A South African study suggests IPV’s association with elevated viral load postpartum.

A 2023 meta-analysis of six population-based surveys in sub-Saharan Africa found that women experiencing IPV are at an increased risk of HIV acquisition. This, combined with a cohort study in Uganda showing an added risk of HIV acquisition among pregnant compared to non-pregnant women suggests that IPV may exacerbate the risk of HIV acquisition, and subsequent vertical transmission among pregnant women.

Despite this evidence on pathways linking IPV and pediatric HIV, a comprehensive analysis of the contribution of IPV to vertical HIV transmission rates incorporating the full PMTCT cascade has not been undertaken. Empirical estimation of this phenomenon is methodologically difficult due to the relative rarity of vertical transmission (i.e., low power), as well as time- and setting-dependent variability in PMTCT program coverage.

Added value of this study

To our knowledge, our study provides the first comprehensive analysis of past-year physical or sexual (or both) IPV’s contribution to vertical HIV transmission in sub-Saharan Africa, along the full HIV prevention and treatment cascade. We used country-reported PMTCT program data and estimates of key HIV indicators from 46 countries, as well as meta-analyses of population-representative surveys, and community-based cohort studies to parametrize our model. Our custom application of a detailed probability tree model accounts for the temporal relationships between IPV and vertical transmission of HIV. We found that in sub-Saharan Africa, one out of eight new pediatric HIV acquisitions could have been averted through elimination of IPV, with the greatest impact on adolescent girls and young women.

The implications of all the available evidence

The 2022 Global Alliance to End AIDS in Children stakeholders commit to eliminating mother-to-child transmission of HIV by 2030, with a directed focus on gender inequities and structural drivers of HIV. Experience of IPV could exacerbate risks of vertical HIV transmission, especially in adolescent girls and young women where the IPV burden and HIV incidence is the highest. Progress in reducing new pediatric HIV acquisitions must be paired with reductions in IPV to accelerate MTCT elimination goals.