WHO 2010 guidelines for prevention of mother-to-child HIV transmission in Zimbabwe: Modeling clinical outcomes in infants and mothers

Background: The Zimbabwean national prevention of mother-to-child HIV transmission (PMTCT) program provided primarily single-dose nevirapine (sdNVP) from 2002-2009 and is currently replacing sdNVP with more effective antiretroviral (ARV) regimens. Methods: Published HIV and PMTCT models, with local trial and programmatic data, were used to simulate a cohort of HIV-infected, pregnant/breastfeeding women in Zimbabwe (mean age 24.0 years, mean CD4 451 cells/μL). We compared five PMTCT regimens at a fixed level of PMTCT medication uptake: 1) no antenatal ARVs (comparator); 2) sdNVP; 3) WHO 2010 guidelines using “Option A” (zidovudine during pregnancy/infant NVP during breastfeeding for women without advanced HIV disease; lifelong 3-drug antiretroviral therapy (ART) for women with advanced disease); 4) WHO “Option B” (ART during pregnancy/breastfeeding without advanced disease; lifelong ART with advanced disease); and 5) “Option B+:” lifelong ART for all pregnant/breastfeeding, HIV-infected women. Pediatric (4-6 week and 18-month infection risk, 2-year survival) and maternal (2- and 5-year survival, life expectancy from delivery) outcomes were projected. Results: Eighteen-month pediatric infection risks ranged from 25.8% (no antenatal ARVs) to 10.9% (Options B/B+). Although maternal short-term outcomes (2- and 5-year survival) varied only slightly by regimen, maternal life expectancy was reduced after receipt of sdNVP (13.8 years) or Option B (13.9 years) compared to no antenatal ARVs (14.0 years), Option A (14.0 years), or Option B+ (14.5 years). Conclusions: Replacement of sdNVP with currently recommended regimens for PMTCT (WHO Options A, B, or B+) is necessary to reduce infant HIV infection risk in Zimbabwe. The planned transition to Option A may also improve both pediatric and maternal outcomes. We linked two published computer simulation models to project clinical outcomes of five PMTCT strategies in Zimbabwe. First, a model of mother-to-child transmission (MTCT) during pregnancy and delivery [24] was modified to incorporate each step of the “cascade” of PMTCT-related care, from first presentation at antenatal care (ANC) through 18 months postpartum (Figure 1). Second, the Cost-effectiveness of Preventing AIDS Complications (CEPAC)-International model of adult HIV infection [25], [26] was used to project clinical outcomes for women following pregnancy, and was expanded to simulate infant outcomes from birth through the first two years of life. The models were linked by using CEPAC results as MTCT model inputs (Text S1). Outcomes of the linked models included risk of infant HIV infection at 4–6 weeks and 18 months of age and 2-year pediatric survival, as well as maternal 2-year survival, maternal 5-year survival, and maternal life expectancy after delivery. Additional details of model structure, data inputs, sensitivity analyses, and results are presented in the Appendix (Text S1). Opportunities to maximize the effectiveness of PMTCT interventions may be lost at each step in the pathway. ANC: antenatal care, ARVs: antiretroviral drugs, ART: antiretroviral therapy, sdNVP: single-dose nevirapine. We simulated a population of pregnant and breastfeeding women in Zimbabwe who were HIV-infected at the time of conception [27]. For women identified as HIV-infected during ANC, five PMTCT strategies were evaluated (Text S1): 1) A “no antenatal ARVs” strategy, for reference comparison; 2) a single-dose of nevirapine (sdNVP) administered to laboring mothers and infants within 3 days of birth, reflecting the 2002–2009 national Zimbabwe PMTCT program and 2006 WHO “minimum” guidelines for resource-constrained settings [14]; 3) WHO 2010 “Option A:” short-course zidovudine during pregnancy and extended infant nevirapine during breastfeeding for women with CD4 >350/µL and no evidence of WHO stage 3–4 disease, with lifelong ART for women with advanced disease [11]; 4) WHO 2010 “Option B:” ART through pregnancy and breastfeeding regardless of CD4 or disease stage, with continuation after weaning for women with advanced disease [11]; and 5) the “Option B+” under consideration in select locations: lifelong ART for all pregnant, HIV-infected women, for comparison [12], [28]. ART-eligible women who linked to HIV-related healthcare after delivery were assumed to receive ART for their own health in all strategies (Text S1). Mothers were assumed to breastfeed their infants for a median duration of 18 months, based on Zimbabwean data [21]. Throughout the manuscript, the term “ARV” refers to any single or dual antiretroviral drug regimen used for PMTCT, while “ART” refers only to three-drug combination therapy (regardless if used for PMTCT or for therapy of maternal disease). The MTCT model is a previously-published decision-analytic simulation of a cohort of pregnant women from conception through delivery [24] (TreeAgePro 2010 software, Williamstown, MA). The model structure was expanded to include key steps in antenatal care, as well as linkage to postnatal maternal and pediatric care (Figure 1 and Text S1). Probabilities of HIV transmission and maternal and infant death before and during delivery were stratified by the severity of maternal HIV infection (ART “eligible,” defined as WHO stage 3–4 disease or CD4 350/<µL [11]; “not eligible;” or deceased) and by maternal receipt of postnatal HIV care and ART. The CEPAC-International model is a first-order Monte Carlo simulation of HIV infection, in which patients are simulated individually from model entry through death. Details of model structure and validation have been published previously [25], [26], [29] and are further described in the Appendix (Text S1). In brief, disease progression is characterized by a sequence of monthly transitions between health states; these include acute opportunistic and other infections prevalent in southern Africa, chronic HIV infection, and death. The model records all clinical events during each patient's lifetime. A cohort of ten million women is simulated to produce stable estimates of outcomes (Text S1). In the CEPAC adult model, current CD4 count, opportunistic infection prophylaxis, and history or absence of previous opportunistic infections determine the monthly risk of opportunistic infections and HIV-related death (Table 1). HIV RNA suppression with effective ART leads CD4 counts to increase, reducing the risks for opportunistic infections and death. Before initiation of ART or after virologic failure on ART, CD4 counts decline at a rate determined by current RNA level; this is accompanied by increased risks of opportunistic infections and death. After planned interruption of suppressive ART at the time of weaning (Option B only), CD4 counts decline more rapidly based on data from ART interruption and PMTCT trials [30]–[32]. A first-order, Monte Carlo simulation model of infant HIV infection and survival was added to the CEPAC model. Modeled infants enter the postnatal model at birth. Based on events occurring before and during delivery in the MTCT model, infants are assigned one of three HIV categories (HIV-unexposed; HIV-exposed but uninfected; or HIV-infected) and three maternal disease categories (HIV-uninfected; HIV-infected and “ART eligible;” or HIV-infected and “ART non-eligible”). Over a two-year horizon, infants face a monthly probability of four key clinical events: 1) maternal HIV infection, if mother was previously uninfected, causing infants to transition from “unexposed” to “exposed-uninfected;” 2) maternal death, with risks derived from the adult CEPAC model as described above, after which infants are no longer at risk for HIV infection but are at higher risk of death due to orphanhood [33], [34]; 3) infant HIV infection through breastfeeding, if infant was previously uninfected; and 4) infant death from any cause. Risks of maternal death and of postnatal HIV transmission are stratified by maternal disease stage and ARV regimen, and risks of infant death are stratified by infant HIV exposure/infection status and receipt of ART if infected. For HIV-infected infants, a one-time probability of HIV diagnosis, linkage to HIV care, and ART initiation was modeled [13]. For mothers, care and ART were modeled in greater detail. A one-time probability of linkage to maternal HIV care was incorporated; this parameter reflects the probability that a post-partum mother will present to an HIV clinic for her own healthcare by six weeks postpartum. Women not linking to care within this period were assumed to present to HIV care upon later development of a severe opportunistic infection. Once in postnatal care for her own health, ART eligibility was assumed to be assessed through both CD4 testing and clinical evaluation. For women identified as ART-eligible, both ART and trimethoprim/sulfamethoxazole prophylaxis were initiated [13], [35]. For women identified as not yet ART-eligible in postnatal HIV care, medications were administered depending on the modeled breastfeeding prophylaxis regimen: no medications were dispensed for the “no antenatal ARVs” and “sdNVP” regimens; infant nevirapine syrup was dispensed for the “Option A” regimen; and maternal ART was dispensed for the “Option B” and “Option B+” regimens. Specific components of 3-drug ART regimens were simulated to reflect 2009 Zimbabwean guidelines and common current practice in Zimbabwe (Text S1 and Table 1) [35]. ART monitoring and switching strategies are also detailed in the Appendix (Text S1). During modeled ANC, women could be lost to follow-up (LTFU) at any stage between first presentation (booking) and delivery; if LTFU, no antenatal ARVs were received, but the opportunity to access HIV testing and sdNVP in labor remained. Women could also be LTFU between delivery and six weeks postpartum [36], or after linkage to postnatal HIV care [37], [38]. In the absence of specific maternal or pediatric data regarding monthly risks of LTFU and cessation of prophylactic ARVs during breastfeeding, the impact of such events was incorporated in sensitivity analyses via the highest published postnatal transmission estimates for Options A, B and B+. For HIV-infected infants, the impacts of loss to follow-up after established pediatric HIV care were included in cohort-based pediatric survival estimates. Baseline maternal characteristics reflected cohorts of pregnant women in Zimbabwe (Table 1 and Text S1). At first ANC visit, mean age was 24.0 [27] and HIV prevalence was 16% [39]; HIV incidence during pregnancy and breastfeeding was 0.96%/year [39]. Among HIV-infected women, mean CD4 was 275 cells/µL if ART-eligible [21], 550 cells/µL if not ART-eligible [21], and 664 cells/µL if incidently infected during pregnancy [40]. Because detailed clinical data to inform HIV disease progression in the absence of ART were not available from Zimbabwe, these natural history model inputs to the CEPAC model were derived from a clinical cohort in South Africa (Text S1) [41]. For women remaining HIV-negative, life expectancy was projected using UNAIDS cause-deleted mortality rates [42]. Modeled rates of access to ANC, HIV and CD4 tests in ANC, ARV drugs for PMTCT, and postnatal care reflected Zimbabwean national estimates whenever available (Table 1). In the sdNVP strategy, women were modeled to undergo clinical ART-eligibility assessment, but not CD4 testing, and to initiate lifelong ART if WHO stage 3–4 disease was identified [11]. The sensitivity of clinical assessment for ART eligibility was 36% [43]. In the base case analysis, to isolate the benefits of each regimen, ARV medications were assumed to be available for, accepted by, and adhered to by all women diagnosed as HIV-infected in ANC, and CD4 testing (but not result return) was assumed for all women under Option A. In addition, sdNVP was modeled to be provided to 80% of women newly HIV-diagnosed in labor [23], [39]. The base-case rate of linkage to postnatal maternal HIV care reflected the median of published values; in sensitivity analyses, linkage rates ranged from 51–100%, depending on antenatal care and PMTCT regimen received [36], [44]–[48]. For women in HIV care and receiving ART, the 24-week “efficacy” of first-line, NNRTI-based maternal ART in suppressing HIV RNA to <400 copies/ml was 85% for women with previous sdNVP exposure [49] and 90% (assumed difference of 5%) for women without exposure to sdNVP [50]. In the six months following ART interruption at weaning (Option B), CD4 count was modeled to decline by 139 cells [30], [31]. Other modeled ART effects are described in Table 1. Mother-to-child HIV transmission risks were derived from randomized clinical trials in several African settings (Table 2) [1], [4], [5], [8]–[10], [15], [17], [21], [51]–[61]. MTCT risks during the intrauterine and intrapartum period (by 4–6 weeks of age) and postpartum period (6 weeks-18 months) were stratified by maternal HIV stage and by PMTCT regimen received. Postpartum transmission risks were additionally stratified by whether breastfeeding during the first six months of life was exclusive (EBF) or mixed (MBF, including any non-breastmilk liquid or solid) [62]. Mortality estimates for HIV-unexposed children were derived from UNAIDS HIV-deleted mortality estimates [42] and mortality rates for exposed-uninfected infants were from the ZVITAMBO study in Zimbabwe [63] (Table 2). Mortality rates for untreated HIV-infected children, stratified by timing of HIV infection, were derived from a patient-level analysis of 1,930 children from 12 African PMTCT trials, after removing non-HIV-related causes of death [6]. ART was assumed available for 36% of HIV-infected children [64]. Mortality rates for children treated with ART were from a systematic review of pediatric ART in Africa [65] and a pooled patient-level analysis from 16 African cohorts [66]. For infants of any HIV infection or exposure status, maternal death was assumed to increase mortality 2-fold [33]. Model results were validated by comparison to published values (Text S1). Sensitivity analyses, described in detail in the Appendix (Text S1), examined the impact of variations in key maternal clinical and demographic characteristics, highest and lowest published MTCT risks for each regimen, access to ANC and HIV testing in ANC, linkage to postnatal HIV care, ART efficacy and survival following sdNVP exposure, CD4 cell decline following ART interruption, and pediatric and maternal non-HIV-related mortality rates. Parameters leading to substantial changes in model results were identified; a substantial change was defined as 1) a change in the relative order of the outcomes of the PMTCT regimens, or 2) a >10% relative change in the difference between projected outcomes for each ARV regimen (example provided in Text S1). In addition, we projected the impact of each PMTCT regimen on infant infection risk for an annual cohort of women becoming pregnant in Zimbabwe, including both HIV-infected and HIV-uninfected women at conception (Text S1). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. This analysis was deemed “not human subjects research” by the Partners Healthcare IRB, and was approved as an exempt analysis.