How to make HIV vertical transmission prevention good value for money in settings with very low HIV prevalence? Using economic evaluation guiding policy in the Philippines

Background: Preventing mother-to-child transmission of human immunodeciency virus is important due to the impact of the disease to the women and their children’s health. Established guidelines have recommended strengthening the ability to detect and treat HIV as early as possible. This study attempts to explore cost-effective PMTCT interventions in a low incidence, low-middle income setting such as the Philippines that can be replicated in other similar country settings. Methods: The study utilized a model-based cost-effectiveness analysis through a decision tree analysis. The decision tree reected the rst ANC visit of a pregnant woman in the Philippines. Vertical transmission program was explored as ten possible PMTCT policy strategies including Do-nothing approach; status quo approach, which had 9% HIV testing, provision of tenofovir-based ART and neonatal prophylaxis, and automatic cesarean section; Eight universal screening policy strategies, which had 100% HIV testing and counselling, provision of tenofovir-based ART and neonatal prophylaxis with inclusion or exclusion of raltegravir for aggressive late antenatal care (ANC) HIV treatment on top of the ART given, breastfeeding or provision of substitute feeding, and normal delivery or cesarean section delivery Results: Base case analysis revealed that policy strategy of universal HIV screening, with additional provision of raltegravir on those receiving late antiretroviral therapy, neonatal prophylaxis, and substitute feeding on normal delivery (urndnpsf) had the lowest ICER values among all policy strategies compared to status quo (₱291,170.26/QALY) and do nothing (₱291,710.26/QALY). Through universal screening coverage, at least 91% of the HIV cases in newborn may be averted. Cost of HIV test must be reduced by at least 45% to have a cost-effective PMTCT program. Alternatively, by performing group pre-test HIV counselling of at least three persons per session, the program will become cost-effective when compared to the unocial Philippine threshold of ₱150,000/QALY. Conclusions: Model design on HIV testing among pregnant women allows exploration of costs and outcomes of PMTCT interventions that focused on a low prevalence, low-middle income setting. From the study, HIV testing can be a

necessary treatment in order to prevent the transmission of HIV to the newborn as demonstrated by Thailand and Malaysia, which have recently achieved elimination of mother-to-child transmission of HIV last 2016 and 2018 respectively [12], [13]. This study attempts to design cost-effective PMTCT interventions in a low incidence, low-middle income setting such as the Philippines that can be replicated in other similar country setting.

Modelling approach
The study utilized a model-based cost-effectiveness analysis (CEA) from the perspective of the government healthcare system. A CEA is de ned as a comparative analysis of alternatives, in terms of both their costs and outcomes. Government healthcare system perspective explored the interventions that should be paid by the government in order to provide continuum of care to the HIV-con rmed patients. Cost of each PMTCT policy strategy are estimated, relative to their bene ts or outcomes, which are identi ed as quality- The decision tree developed as shown in Fig. 1 started with a cohort of HIV test naive pregnant women strati ed to their de nitive HIV status. Pregnant women may either have their rst ANC in the early (≥ 6 weeks before delivery) or late (< 6 weeks) period. As part of the routine panel of test for their rst ANC visit, an HIV testing and counselling may be performed to which they may accept or reject. For those who accepted the HIV test and tested positive, a vertical transmission program (VTP) which included PMTCT interventions was offered to which they may accept or decline. Pregnant women who received late ANC had the same branches as early ANC with varying probability parameters.
Ten PMTCT policy strategies being considered in the study that represented the VTP in the model as shown in Table 1. Do-nothing approach simulates the delivery of a pregnant woman, regardless of her HIV status. This would mean no HIV testing, no provision of tenofovir-based antiretroviral therapy (ART) among HIV-positive, normal delivery without neonatal prophylaxis and substitute feeding.
Status quo or ad hoc approach refers to the current Philippine efforts on PMTCT. This includes less than 10% HIV testing and counselling among pregnant women, provision of tenofovir-based ART to all those tested positive and received subsequent neonatal prophylaxis to their newborn; automatic cesarean section for all HIV-positive pregnant women regardless of their duration of ART is performed; nally, substitute feeding was also not provided to these women.

Model Parameters
Input parameters used in the analysis are listed in Table 2.  country-speci c HIV surveillance, national surveys, case reports, and vital statistics [15], [16]. Distribution of rst ANC visit of the pregnant women was retrieved from the most recent published Philippines National Demographic Health Survey [17]. As for the status quo screening coverage, HIV/AIDS & ART Registry of the Philippines (HARP), which was deemed the best available data, provided this parameter [10].

Intervention effects
Clinical outcomes for each PMTCT intervention were sourced through most relevant published systematic review, randomized clinical trial, or observational study [18]- [24]. Baseline HIV transmission was computed by the summation of the probability of HIV transmission during pregnancy and postnatal period [18], [19]. Due to limited studies that compared the e cacy of tenofovir (TDF)based regimen with placebo on MTCT outcomes, the study by Hoffman et al. [20], which strati ed the transmission rate depending on the duration of highly active antiretroviral therapy (HAART) intake during pregnancy period was used to determine e cacy of the ART regimen, regardless of the components of HAART regimen. To incorporate the e cacy of intensi cation of raltegravir for HIV detected during late pregnancy, values were computed through the data from published study conducted in Thailand on late-presenting HIVpositive high-risk pregnant women [21]. A study on HIV-positive pregnant women taking HAART from 2000-2011, that was strati ed the outcomes based on the mode of delivery, determined the HIV transmission on greater than 4 weeks on ART with vaginal delivery [22].
Relative risk of HIV transmission of elective cesarean section delivery versus vagina delivery on HIV-positive pregnant women was determined through the systematic review by Kennedy et al. [23]. Also, it was also necessary to demonstrate the bene t of receiving ART regimen less than 4 weeks before delivery in a breastfeeding population versus no treatment [24]. An assumption of 100% sensitivity and speci city for the HIV test was inputted as advised by expert panel.
HIV transmission rates for each scenario of the policy strategies were calculated as shown in Table 3. Scenario with the least possible HIV transmission rate, which was receiving early antiretroviral therapy with normal delivery on a breastfeeding population, resulted to 0.15% MTCT rate, whereas the highest HIV transmission rate of 35.12%, were those who did not receive any PMTCT interventions.

Utility values
Assumption that utility values will be exclusive to HIV-positive or HIV-negative newborns was applied to the model. Utility value of HIVpositive newborn was adapted from the analysis of Sanders et al. [25]. The utility value for HIV-negative newborns, assumed to be healthy Filipino individuals, used the Philippine value set of EQ-5D-5L for 11111 [26]. A simple lifetime Markov model was simulated to determine the QALY of HIV-positive and HIV-negative newborns using Philippine life table and assigning those alive with utility value of 1 and dead as zero.

Cost parameters
Cost parameters utilized for the model included direct medical costs associated in providing PMTCT interventions to the pregnant women until delivery and lifetime horizon for HIV-positive newborns. All costs were presented in Philippine peso and adjusted to 2019 values adopting the methods previously discussed by Turner et al. [27].
Government procurement rates, whenever available, were obtained from the annual procurement report from the Department of Health [28]. Costs associated with providing HIV test for either HIV-positive or HIV-negative pregnant women were computed using activitybased costing validated with local HIV experts and hospital staff from the national HIV referral center. As for the lifetime pediatric HIV costs, the authors have calculated the value through a separate Markov model with discount rate of 3% [29]. Providing substitute feeding for 6 months to compensate the time period for exclusive breastfeeding was based on a local study on the economic burden of infant formula [30]. PhilHealth coverage was applied for cesarean section delivery while the cost of raltegravir, being not presently registered in the Philippine Food and Drug Administration, was adopted from a Thailand report [31], [32].

Analyses
First, the total program costs for each PMTCT policy strategy was compared with status quo and do nothing approach. Moreover, the incremental costs and incremental QALY compared with status quo and do nothing approach were calculated. Strati cation of the total program costs into screening cost, PMTCT costs, and lifetime pediatric HIV costs were determined.
Second, the cost-effectiveness of each PMTCT policy strategy was assessed compared to status quo and to do nothing approach with an acceptance to the vertical transmission program strategies of 95%. The difference between the costs of each policy strategy with the do nothing approach or status quo over the difference between their utilities estimated the incremental cost-effectiveness ratios (ICERs).
ICERs were compared to the uno cial set threshold of ₱150,000/QALY gained, country. This value is referenced in the previous guidelines of the Philippine National Formulary [33].
Third, univariate and probabilistic sensitivity analyses, through Bayesian framework, were carried out for all applicable parameters.
Determining the major drivers for the change in the ICER values was performed through univariate sensitivity analysis on 95% con dence interval. A tornado diagram was generated to illustrate the parameters that elicit the most change to the ICER value. Next, probabilistic sensitivity analysis utilized Monte Carlo simulation using the parameters' appropriate probability distributions. Probability parameters bounded between the range of zero and one applied the beta distribution. Cost parameters, which yields only positive values, used the gamma distribution. Moreover, the population of pregnant women which is expected to have an additive natural growth processes was put under the log-normal distribution. All of the applicable parameters were sampled through the joint distribution and used to yield the costs and outcomes associated with each proposed policy option with 10,000 iterations. ICER values for each of the pair of cost and outcome results were calculated, then the cost-effectiveness acceptability curves and cost-effectiveness acceptability frontiers were generated.
Lastly, a post hoc two-way sensitivity analysis was performed by selecting two sensitive parameters from univariate sensitivity analysis that is predicted to change within the next ve years. A head-to-head two-way sensitivity analysis was performed to determine the range of values of each parameter that will result to an ICER value of less than ₱150,000/QALY gained.

Results
Program costs consisted of cost of HIV screening, cost of providing tenofovir-based ART to the pregnant women, cost of nevirapine for the newborn for 6 weeks, cost of delivery (normal delivery, cesarean section), cost of providing substitute feeding for 6 months, and cost of lifetime pediatric HIV.
Majority of calculated program costs for the proposed policy strategies were dedicated for HIV screening costs, which was computed by the sum of costs of the HIV test kits and costs health worker performing pre-and post-test counseling. HIV screening costs for status quo screening was 45% of the total program cost whereas, HIV screening costs were 99% of the program cost proposed policy strategies. For the lifetime HIV consequences of the newborn, 45% of the program costs were dedicated for lifetime pediatric HIV treatment compared to 1% in universal screening policy strategies.
Tables 4 and 5. presented the cost-effectiveness analysis of the policy strategies in terms of ICERs in Philippine peso (₱) per QALY gained versus versus status quo screening and do nothing approach. The policy strategy of universal HIV screening, with the provision of raltegravir on those receiving late antiretroviral therapy, neonatal prophylaxis, and substitute feeding on normal delivery had the lowest ICER values among all proposed policy strategies (urndnpsf). Furthermore, it yielded the least program costs with highest QALY gained among the proposed policy strategies. However, ICER values generated were higher than the uno cial PH threshold set at ₱150,000/QALY gained.   The probabilistic sensitivity analysis on the ICER of the proposed policy strategy with the lowest ICER versus do nothing approach or status quo screening suggests that the strategy was 42% cost-effective when compared to the PH threshold as shown in Fig. 3. Willingness-to-pay (WTP) of at least ₱180,000/QALY gained will make the proposed policy strategy more value for money compared to status quo screening.
Post-hoc two-way sensitivity analysis on parameters of prevalence of HIV-positive pregnant women and cost of HIV screening among HIV-negative pregnant women were purposely selected as the most dynamic sensitive parameters based on the univariate sensitivity analysis. Adjustments on the parameters by introducing the concept of group HIV pre-test counselling, which will result to lower labor costs, were performed. The cost for each pre-test HIV counselling for HIV-pregnant women was calculated based on the number of persons per group counselling session (1-20 people), while the range of prevalence of pregnant women was estimated from Spectrum-AEM model. Figure 4. shows the optimum combination of these two parameters to have cost-effective results depending on the threshold value being compared. This shows that with increasing prevalence of pregnant women and decreasing the cost of HIV test, the lower the ICER will become. Incorporating the group pre-test HIV counselling from 1-20 persons per session approach, optimal number of persons per counselling session can be determined depending on the value of prevalence of pregnant women applied. In the case of the Philippine setting, with 422 pregnant women needing PMTCT, at least 3 persons per group HIV pre-test counselling will guarantee the program to become cost-effective.

Discussion
The study comprehensively evaluated a cost-effectiveness analysis of different program components in PMTCT which spanned from HIV counseling and testing to providing substitute feeding in a low incidence, low-middle income setting using Philippines as the case study. In addition, this is the rst economic evaluation that analyzed combination of PMTCT interventions including raltegravir intensi cation for late ART and the rst PMTCT economic evaluation done in the country. For countries intending to cover or update their coverage on the combination of PMTCT interventions, the study may assist them by illustrating a model template for HIV screening program among pregnant women.
From the country's perspective, HIV testing can be a major cost constraint for PMTCT. Reducing the cost of HIV test for HIV-negative pregnant women by 46% will be the most feasible option for the policy strategy to be cost-effective. Also, due to the expected rise in the prevalence of pregnant women needing PMTCT, the probability of the policy strategy to become good value for money will increase. It should also be taken consideration that the threshold used to compare the computed ICER values was around one Gross Domestic Product (GDP) per capita of the country, which was used in previous economic evaluations conducted in Philippine context as suggested in the Philippine National Formulary guidelines.
Twenty-one cost-effectiveness analyses included in the systematic review published by Bert et al. in 2018 had agreed on the costeffectiveness of universal antenatal screening [7]. However, due to heterogenous characteristics of the countries and methodologies applied to the studies, several factors may have heavy in uence on the results of the studies. These include perspective of the study, interventions used for the policy strategy, HIV prevalence per country, and the income level of country setting, which may determine the threshold used to compare the ICER value.
Perspectives used by similar studies conducted may be simply classi ed either societal, which included the PMTCT costs along with lifetime costs for the HIV-positive women and the perinatally infected newborns, or government payer, where the focus was mostly providing PMTCT and treatment only to the infected newborn [7]. Government payer perspective may underestimate the bene t of providing the universal HIV test due to several reasons. First, the detection of HIV among pregnant women may prompt additional detection among their respective partners as well as prevent further transmission of HIV on their succeeding pregnancies. In fact, according to the study of Kendall [34], the most promising opportunity of women to determine their HIV status is during their pregnancy. When the pregnant women were not diagnosed during their ANC, the most probable next event that would trigger seeking relevant care will be during their won experiences AIDS-related complications or those of a close family member's. Time period for providing appropriate care for HIV is critical to prevent HIV perinatal transmission. Therefore, HIV diagnosis must be performed as early as possible, wherein ANC visit may serve as a good entry point for any country setting [5]. Prevention of addressing more progressed stages of HIV that would incur additional costs and decrease their overall QALYs were not accounted in government payer perspective.
Second, the consequences on the women after delivery due to cesarean section was not assessed. These include higher rates of infectious complications and surgical traumas, longer hospital stay and in-hospital deaths [35]. Although it is proven on systematic reviews that cesarean section provides less risk of maternal HIV transmission, the consequences of maternal injuries were not included in the payer's perspective. Third, by including only the lifetime treatment costs of infected newborns, the productivity contribution of these population due to increased lifetime was also not considered.
The interventions used in the PMTCT policy strategies of the studies included in the systematic review contained a screening strategy (universal or targeted), with ART and prophylaxis to the newborn with or without considering the health of the woman after her labor [7].
Although most of the studies included considered retesting during late pregnancies, which was not considered in this study, none had investigated the addition of raltegravir as an additional intervention during HIV detection in late pregnancies. The published study by Puthanakit et al. conducted in Thailand showed that providing raltegravir on top of the standard of care for those diagnosed with HIV on late pregnancies drastically reduces the risk of perinatal HIV transmission [21]. This nding was deemed signi cant in the development of the optimal PMTCT policy strategy. Furthermore, the conducted study was novel in terms of considering how to design an appropriate and feasible PMTCT policy strategy for any country setting who will start to publicly cover PMTCT interventions.
The national prevalence from passive surveillance of HIV-positive pregnant women in the Philippines is around 0.0248% in 2019, which is considered as low prevalence according to UNAIDS [9], [36]. This level of prevalence is comparable to Australia, New Zealand, and Hong Kong but has lower prevalence compared to India, Thailand, Uganda. and South Africa [4], [37]- [42]. All of these countries included suggested that PMTCT interventions including universal HIV screening among pregnant women is cost-effective in their respective country setting [7]. As determined in the univariate sensitivity analysis, prevalence was identi ed parameters that has great in uence on the ICER value. Higher prevalence, with similar income levels and threshold level utilized, would easily make the PMTCT policy strategy good value for money, especially since the prevalence in those countries is higher than the Philippines, with ranges from 1.63-19.5% [38], [39], [41], [43]. However, those countries with comparable prevalence with the Philippines were high income countries and had a much greater threshold used to compare with their ICER values [44]. Compared to the PH threshold of ₱150,000/QALY gained, which is around US$3,000, high-prevalence countries used the estimate values US$17,600, US$24,000, US$10,100 for New Zealand, Australia, and Hong Kong respectively as thresholds for good value of money [37], [40], [41].
Among the sensitive parameters in the univariate sensitivity analysis, the cost of HIV screening can be adjusted by the government payer. One method that was performed in the study is group HIV pre-test counseling, in order to reduce the total amount of time spent of the health worker for counselling, as well as to address the foreseen insu ciency on the number of health workers available. According to the WHO Guidance on testing and counselling for HIV, in order to improve access, a combination of group health information talks to be followed by individual risk based assessments tailored to the patient's HIV status is a valid method to nd a balance between complicated and overdrawn counselling sessions and rapid, intense education/information transfer [45]. As demonstrated in the study, for the case of Philippines, a group HIV pre-testing counselling of at least three persons per session will make the PMTCT policy strategy cost-effective. To add, communities with limited number of health workers may bene t the most of this technique by organizing health information sessions attended by all the people in the community [46]. Also, the local HIV patient group in the Philippines have stated that their organization have trained their members to become quali ed peer HIV counselors in order to supplement the lack of health workers pro cient in HIV counselling in the country.
The study has its limitations whenever it will be used as a model template for countries that will consider coverage of HIV test among pregnant women. First, several assumptions on the input parameters were made due to unavailability of speci c data needed. By assuming equal e cacy of treatment regimens described in Hoffman et al, with the HAART regimens listed current WHO, guidelines may underestimate the bene ts gained by reducing the rate of MTCT due to HAART alone [13], [20]. Adaptation and adjustments of foreign data on the utility values of HIV-infected newborns, cost of raltegravir, and lifetime costs of providing treatment for HIV-positive newborns were due to absence of local studies. Second, only direct medical costs were considered in the study. Accounting for productivity costs (e.g. cost of sick leaves and hospitalizations due to HIV complications) and the direct non-medical costs (e.g. travel cost) along with inclusion of costs of adverse events due to treatment regiments in the course of pregnancy, treatment costs of long term side effects to the women and children receiving PMTCT interventions and ARV resistant women and children may provide truer estimate of costs and bene ts. Third, the study also did not account for the additional bene ts due to early HIV detection and treatment including the prevention of MTCT on succeeding pregnancies, as well as slowing down disease progression of women and their partners. Fourth, any cultural or social factors that may indirectly contribute to the costs and bene ts of the program was not explored.
To illustrate, published studies from Uganda and England, revealed that HIV patients travel signi cantly farther to access healthcare than those without HIV. Factors may include lack of necessary specialty services, supply of ART, and possibly to protect con dentiality [47], [48]. Therefore, local feasibility studies must supplement before actual program implementation to determine other technical, political, economic or practical barriers that may impede the operation of the HIV testing program.
HIV/AIDS is not just a health problem, the spread of this infectious disease is highly correlated with socioeconomic, environmental and ecological factors such as population growth, environmental and land-use changes, changing human behaviors and political reorganizations [1]. The research conducted illustrated novel ways how to design cost-effective PMTCT policy strategies in a low income and low prevalence setting. By introducing adjustment methods of the parameters, these imply that for countries with low prevalence, PMTCT strategies coverage may be implemented accordingly. The recognition of maternal HIV transmission as a public health concern in the country must trigger the implementation of PMTCT policies. In such cases where there are underreporting of HIV cases along with the uncontrollable rise in the number of in the Philippines as mentioned by local expert consultations, the consideration for a full-scale implementation of PMTCT must be realized. Furthermore, although the research used scenarios based on actual global practices, outdated practice such as performing cesarean section on HIV-positive pregnant women, regardless of their viral load levels should be discontinued [49]. Knowledge on the recent clinical protocols on the criteria for cesarean section must be enforced.

Conclusions
The model design on HIV testing among pregnant women allowed exploration of costs and outcomes of PMTCT interventions that focused on a low prevalence, low-middle income setting. From the study, HIV testing can be a major cost constraint for PMTCT. Furthermore, strategies that will affect sensitive parameters like by performing group pre-test HIV counselling instead of per individual may result to a cost-effective PMTCT program, even at low prevalence, low income setting. One-way sensitivity analysis (top 11 most sensitive parameters)