Cost-effectiveness of pre-exposure HIV prophylaxis during pregnancy and breastfeeding in Sub-Saharan Africa

Introduction: Antiretroviral pre-exposure prophylaxis (PrEP) for the prevention of HIV acquisition is cost-effective when delivered to those at substantial risk. Despite a high incidence of HIV infection among pregnant and breastfeeding women in sub-Saharan Africa (SSA), a theoretical increased risk of preterm birth on PrEP could outweigh the HIV prevention benefit. Methods: We developed a decision analytic model to evaluate a strategy of daily oral PrEP during pregnancy and breastfeeding in SSA. We approached the analysis from a health care system perspective across a lifetime time horizon. Model inputs were derived from existing literature and local sources. The incremental cost-effectiveness ratio (ICER) of PrEP versus no PrEP was calculated in 2015 U.S. dollars per disability-adjusted life year (DALY) averted. We evaluated the effect of uncertainty in baseline estimates through one-way and probabilistic sensitivity analyses. Results: PrEP administered to pregnant and breastfeeding women in SSA was cost-effective. In a base case of 10,000 women, the administration of PrEP averted 381 HIV infections but resulted in 779 more preterm births. PrEP was more costly per person ($450 versus $117), but resulted in fewer disability-adjusted life years (DALYs) (3.15 versus 3.49). The incremental cost-effectiveness ratio of $965/DALY averted was below the recommended regional threshold for cost-effectiveness of $6462/DALY. Probabilistic sensitivity analyses demonstrated robustness of the model. Conclusions: Providing PrEP to pregnant and breastfeeding women in SSA is likely cost-effective, although more data are needed about adherence and safety. For populations at high risk of HIV acquisition, PrEP may be considered as part of a broader combination HIV prevention strategy. We constructed a decision analytic model to compare the costs and effectiveness of the reference strategy of no PrEP against a scenario of PrEP provision to pregnant and breastfeeding women in SSA (TreeAge Pro, Williamstown, MA; Supplemental Fig. 1, http://links.lww.com/QAI/A829). We considered a lifetime time horizon and approached the analysis from the perspective of the health care system. Costs and utilities were discounted at a 3% annual rate.28 The primary outcome measure was the incremental cost-effectiveness ratio (ICER), measured as 2015 U.S. dollars (USD) per disability-adjusted life years (DALYs) averted. Cost-effectiveness was defined according to WHO guidelines as an ICER <3 times the per capita gross domestic product (GDP); a very cost-effective intervention requires an ICER less than the per capita GDP.28 The WHO Afro-E region of SSA (http://www.who.int/healthinfo/statistics/gbdestimatesregionallist.xls) has an overall per capita GDP of $2154, which translates to a cost-effectiveness threshold of $6462/DALY averted.28 Our study population was composed of HIV-negative pregnant and breastfeeding women presenting to antenatal care (ANC).30,31 We searched the medical literature, conference abstracts, and published reports to identify setting-specific estimates for model parameters (Table ​(Table1).1). Where such data were unavailable or of a narrow scope, we drew estimates from our HIV care and treatment program in Zambia47 and expert opinion. Probabilities were calculated from rates using the standard equation P = 1 − e−rt. We derived probabilities of HIV infection from incidence rates in pregnancy (4.7 per 100 person-years) and postpartum (2.9 per 100 person-years) reported by a meta-analysis of studies in SSA.18 The MTCT risk if maternal HIV infection occurs during pregnancy or breastfeeding is 22.7%.18 For women who were infected with HIV in pregnancy but do not transmit HIV to their fetus, the MTCT risk during breastfeeding (assuming a median breastfeeding duration of 18 months) was assumed to be similar to that among women with chronic HIV, or 9%.43 The baseline PTB risk in SSA is 12%;45 HIV infection confers a risk ratio of PTB of 1.5, which is comparable with the PTB risk in our HIV-infected Zambian cohort.46,47 The risk of PTB among women taking PrEP is theoretical and for this analysis was inferred from a randomized trial comparing the efficacy and safety of triple ARV regimens for the prevention of MTCT among HIV-infected women, and was also consistent with data from Zambia.22 Model Parameters We assumed that once-daily oral PrEP medication consisting of TDF-FTC would begin at the first ANC visit with a negative HIV screening test and terminate with cessation of breastfeeding (median 15 months postpartum).51–53 We derived the median gestational age (GA) of entry into ANC as 19 weeks and the median GA at delivery as 39 weeks from an international database.31 We assumed uniform effectiveness of PrEP for each woman in the base-case analysis, based on a median time spent in ANC of 20 weeks (ie, 39-week GA at delivery minus 19-week GA at entry into ANC). Because there has been no observational study of PrEP in pregnancy, we varied widely the estimate for effectiveness of PrEP during pregnancy and breastfeeding to account for variable adherence and exposure duration. We focused on the index pregnancy only (and assumed no subsequent pregnancies occurred) and did not consider the cost or disability of subsequent transmissions beyond mother and child (eg, to sexual partners). For women who were infected with HIV during pregnancy or breastfeeding, we assumed initiation of lifelong antiretroviral therapy (ART) under recent WHO guidelines for a realistic proportion of women (43%).11,17 Although current guidelines recommend that all infants are started on ART as soon as they are diagnosed with HIV, true coverage of pediatric HIV treatment approximates 34%.11,17,32 We assumed this “real world” coverage rate in our model to account for infants who fail to access timely health services and die before diagnosis or treatment. Cost parameters were derived from international economic sources and previous cost analyses (Table ​(Table1).1). Where available, we used relevant purchasing power parity to convert original costs in local currency to international dollars and then inflated to 2015 USD. Given that many aspects of HIV prevention programs are paid for in USD through international funding agencies, if costs were reported in the literature only in USD with no reference to original local currency, we directly inflated these costs to 2015 USD using historical consumer price index data from the National Bureau of Labor Statistics.57 The cost of PrEP medication for the duration of pregnancy and breastfeeding was estimated from the cost of TDF-FTC negotiated by the Clinton Health Access Initiative in its list of ceiling prices.54 The cost of toxicity surveillance, based on the recommended quarterly basic metabolic panel plus HIV and hepatitis B testing,11 was micro-costed from previous economic analyses in SSA.15,55 We estimated additional PrEP program costs to include personnel and facility expenses required for adherence counseling and monitoring activities based on data from previous economic evaluations of voluntary counseling and testing strategies.56 Our composite program cost is higher than that reported by a recent analysis of PrEP for serodiscordant couples in Uganda; however, given likely broad regional variations in cost, we intentionally chose the higher estimates of programmatic costs for our model inputs.58 The annual cost of care for HIV/AIDS, including provision of ARVs, was derived from an analysis of costs of government programs in 45 sites in Zambia35 and is consistent with ranges reported in previous analyses.36,59,60 The average cost of PTB per infant was based on two studies in SSA that reported the cost of care for neonates with low birth weight as a proxy for prematurity given imprecise GA estimates.48,49 The cost of PTB was assumed to be a uniform upfront cost and was not discounted. The primary payoff was calculated in DALY as the sum of years of life lost (YLL) and years lived with disability (YLD). We calculated YLD associated with both mother and infant outcomes using disability estimates from the Global Burden of Disease Study.34 We used the disability weight for HIV over the entire life expectancy of an HIV-positive woman or infant. Disability due to PTB was applied to the lifetime of all infants born prematurely; for preterm HIV-positive infants, we used the more severe disability weight associated with HIV to generate YLD.61 We modeled known differences in neonatal mortality between term and preterm infants.39 We accounted for differences in survival between HIV-negative infants born to infected mothers and those born to uninfected mothers by modeling differential mortality risk at 24 months of age based on HIV exposure.40–42 Data are scarce on long-term survival of HIV-infected children who initiate ART in infancy62; therefore, we modeled childhood mortality based on a previously published simulation of the effect of ART on pediatric HIV disease progression in South Africa63 and subsequently estimated life expectancy of surviving adolescents from a Ugandan cohort.37 For HIV-positive infants not receiving treatment, we modeled life expectancies based on known survival differences between infants infected in utero or peripartum versus those infected postpartum.38 Discounting for DALY payoffs in YLL and YLD for mother and infant was performed using the following standard equations, where N is number of deaths, r is annual discount rate (eg, 3%), I is number of incident cases, DW is disability weight, and L represents years of either standard life expectancy at age of death (in YLL equation) or duration of disability (in YLD equation).44 Baseline values for cost, utility, and expected survival for the PrEP scenario were used for the base-case analysis (Table ​(Table1).1). The ICER, our primary outcome measure, was defined as the incremental cost (in 2015 USD) per number of DALYs between a scenario of PrEP provision in pregnancy and breastfeeding versus the standard scenario of no PrEP. We performed one-way deterministic sensitivity analyses of parameters that influenced the ICER outcome using ranges of best- and worst-case scenarios, derived from previous clinical trials, meta-analyses, or reasonable assumption. In one-way sensitivity analyses, costs were varied by a factor of at least three times the base value to account for uncertainty in estimates and assumed wide regional variations. Probabilities were varied according to ranges from the medical literature or to account for uncertainty in point estimates and regional variation. We performed threshold analyses to understand at which point the value of key uncertain parameters rendered PrEP not cost-effective. We determined the thresholds of PrEP effectiveness and the probability of PTB on PrEP, as the estimates for these variables were most uncertain. We conducted probabilistic sensitivity analysis (PSA) using Monte Carlo simulation to assess the confidence in our ICER outcome by varying all parameters simultaneously over distributions informed by parameter ranges reported in the literature at a minimum.