Cost-effectiveness of infant respiratory syncytial virus preventive interventions in Mali: A modeling study to inform policy and investment decisions

Highlights • New RSV prevention products can substantially reduce disease burden.• Longer-acting monoclonal antibodies, priced affordably, are likely cost-effective.• Maternal vaccines meeting preferred product characteristics would be cost-effective.• RSV prevention products can provide good value in low-income countries.


Introduction
Respiratory syncytial virus (RSV) is the leading cause of acute lower respiratory tract infections (LRTI) in infants globally [1]. An estimated 93% of RSV-LRTI cases and 99% of RSV-LRTI deaths occur in low-and middle-income countries (LMICs) [2]. Children under six months of age account for approximately 45% of severe RSV cases [2].
The only product licensed for RSV-LRTI prevention is palivizumab-a humanized monoclonal antibody administered monthly during the RSV season to children at elevated risk for severe disease [3]. The palivizumab price point renders it cost-prohibitive for most LMICs. New monoclonal antibodies and vaccines against RSV-LRTI are in development. One dose of an extended half-life monoclonal antibody candidate, nirsevimab, may protect healthy preterm infants against RSV-LRTI for five months, covering the duration of a typical RSV season [4]. While an RSV fusion protein nanoparticle (RSV-F) vaccine candidate for pregnant women failed to meet its primary endpoint in a phase III clinical trial (subsequently referred to as RSV-F vaccine trial), secondary endpoints and subgroup analyses indicated that a single dose administered to pregnant women in their third trimester protected infants from clinically important RSV-LRTI outcomes over the first three months of life [5]. Given the advanced development of RSV-LRTI prevention candidates, it is important to consider the potential economic and health impacts these products could have in LMICs.
We have identified only one other cost-effectiveness analysis of RSV-LRTI prevention in LMICs [6]. This analysis relied on inputs from systematic reviews of global data and modelling data for disease incidence and severity. The Centre pour le Développement des Vaccins in Bamako, Mali has collected high quality RSV surveillance and cost of care data which allowed us the unique opportunity to conduct a country-specific cost-effectiveness analysis to inform policy and investment expectations for future RSV preventive interventions.

Model design
We modeled the costs and benefits of RSV-LRTI preventive interventions in Mali using a probability-based outcome tree with twelve simulated monthly birth cohorts followed through the first six months of life (Table 1, Supplement Fig. 1). Integrating parameter uncertainty, we estimated distributions of expected RSV-associated health and economic outcomes under four scenarios: a) status quo without intervention, b) intra-seasonal infant prophylaxis with monthly doses of shortacting monoclonal antibody (short-acting mAb), c) pre-seasonal infant prophylaxis with a single birth dose of long-acting monoclo nal antibody (long-acting mAb), and d) year-round, single dose maternal vaccination. With these outcome sets, we calculated the incremental cost-effectiveness ratio for each product at specific price points and the maximum product cost given a specific costeffectiveness threshold.

Heath outcomes
The health outcomes of interest were the expected number of RSV-LRTI cases, hospitalizations, and deaths for each birth cohort during the first six months of life. We derived the incidence of RSV-LRTI under status quo from a community-based study of RSV incidence in Bamako, Mali [7], which was nested within a maternal influenza vaccine clinical trial [8]. We calculated probabilities of RSV illness dependent on calendar month and infant age (Supplement Fig. 2). Given laboratory-confirmed RSV illness, probabilities of RSV-LRTI and hospitalization were based on observations in the community-based study (Table 1). RSV-LRTI was defined as clinical pneumonia using World Health Organization (WHO) criteria [8]. Due to the low frequency of RSV-associated deaths observed in the influenza vaccine trial, we obtained Malispecific, RSV-attributable mortality estimates from the Pneumonia Etiology Research for Child Health (PERCH) study among hospitalized children (C. Prosperi, personal communication, September 7, 2020) [1].
We measured health impact in disability-adjusted life-years (DALYs), a metric that combines life-years lost to premature death with productive life-years lost to illness and disability (Online Supplement) [9]. Years of life lost were discounted at 3% annually [10].

Interventions
We parameterized intervention efficacy and durability based on characteristics of tested products ( Table 1). The efficacy and duration of protection for the products were 78% over one month for short-acting mAb [3], 70% over five months for long-acting mAb [4], and 56% over three months for maternal vaccine [5]. We applied maternal vaccine characteristics specific to a South African subgroup from the RSV-F vaccine trial [5]. Although the trial included multiple sites, South Africa was the only site in Africa and had the closest RSV-LRTI incidence to Mali. We applied product-specific administration schedules (Supplement Fig. 3). Both short-acting mAb and long-acting mAb were administered only when the duration of protection coincided with the RSV season (Online Supplement). Pregnant mothers in their third trimester were eligible for a single dose of maternal vaccine at any time of year. We projected coverage using 2018 Mali-specific data for routine immunization or antenatal care (Table 1, Online Supplement). We also explored the product health impact at a range of alternative coverage estimates spanning 0 to 100%.

Economic outcomes
We derived medical and hospitalization costs from a 2013 costeffectiveness study conducted as part of the Mali maternal influenza vaccine clinical trial (Table 1, Online Supplement) [11]. For both mAbs and maternal vaccine, we assumed an administration cost per dose of $1.35, reflecting the incremental cost of adding one product to the Mali national immunization schedule [12].

Budget impact and cost-effectiveness
We delineated four perspectives for evaluating budget impact and cost-effectiveness: household, government, donor, and societal. Households bear the economic cost of medical care for RSV illness but would not bear any cost of intervention. The Malian government would bear the costs of delivery and product administration [13]. We assumed Gavi, the Vaccine Alliance, would be the donor organization bearing costs of vaccine procurement. Like other LMICs in sub-Saharan Africa, Mali receives support from Gavi for much of its national immunization program. The Malian government would co-finance $0.20 per product dose [14], and the remaining price would be paid by Gavi. Therefore, the government perspective includes administration costs plus a consistent cofinancing contribution regardless of product price, while the donor perspective is conditional on price. The societal perspective considers all costs regardless of payer.
We calculated the budget impact for each intervention as the expected change in spending due to medical or intervention costs compared to status quo over a single year and delineated by per- All infants with pneumonia episodes occurring between October 2012 and May 2013 were selected to be tested for RSV in the Mali incidence study, whereas only 48.9% of the infants with influenza-like-illness without pneumonia were tested [7].
To account for the oversampling among infants with pneumonia, we calculate the probability of LRTI given RSV as the proportion of RSV cases with pneumonia reported in the trial adjusted to match the proportion of infants with influenzalike-illness but without pneumonia who were tested for RSV. Probability of inpatient care given RSV-LRTI 0.29 (0.20, 0.37) Mali Among the proportion of infants less than six months with confirmed RSVpneumonia in the RSV incidence study in Mali, 29% required inpatient care [7]. All infants with pneumonia who did not receive inpatient care received outpatient care. The probability of outpatient care given RSV-LRTI is calculated as 1probability of inpatient care given RSV-LRTI. Case fatality rate among those who received inpatient care given RSV-LRTI Average cost of medical care for infants less than six months with confirmed RSV illness who received inpatient care. Costs are inclusive of outpatient services also acquired by this group [11]. Outpatient care costs (USD) 6.56 (5.44, 7.66) Mali Average cost of medical care for infants less than six months with confirmed RSV illness who received outpatient care services only [11]. Intervention delivery cost per dose (USD)

Low-income countries
Average cost of adding one product to existing immunization programs in lowincome countries in 2019 US dollars [12]. RSV Intervention parameters Short-acting mAb efficacy 0.78 (0.60, 0.90) North America, United Kingdom Based on the multinational phase III Impact-RSV prevention trial, monthly prophylaxis with palivizumab results in a 78% reduction of RSV hospitalizations among pre-term infants without bronchopulmonary dysplasia [3]. Without access to the same life-saving postnatal interventions that are available in high-income countries (such as supplemental oxygen and mechanical ventilation), infants born with bronchopulmonary dysplasia in low-income countries do not often survive past the neonatal period, and therefore we used efficacy data for palivizumab which excluded this high-risk group. Although palivizumab has never been evaluated in healthy term infants, a similar monoclonal antibody, motavizumab, was shown to reduce RSV hospitalizations in this group [28]. spective. The incremental cost-effectiveness ratio (ICER) for each intervention compared to the status quo was calculated as Dc/ De, where Dc is the difference in economic costs between intervention and status quo, and De is the change in health outcomes. We did not conduct a head-to-head comparison of products, as the final assortment of choices available to countries is not yet certain.
We integrated parameter uncertainty into our analysis using a Monte Carlo approach with 10,000 independent trials. For each trial, one value was randomly sampled from every distribution of probabilities and costs. We then calculated the health and economic outcomes associated with status quo and each intervention, holding parameters constant across each arm. The 95% confidence intervals for any given outcome were then defined as the range encompassing 95% of the values produced across all trials. The point estimate reported for each outcome is the value produced by executing the simulation using the point estimate specified for each input parameter.
For evaluating the probability that an intervention would be cost-effective across a range of values for willingness-to-pay (WTP) for DALYs, we applied a net health benefits framework [15]. WTP indicates the maximum spending acceptable to avert one DALY. Net health benefits are calculated as: De -(Dc/WTP). Under this approach, any intervention that results in positive net health benefits is considered cost-effective [15]. We evaluated the interventions across a range of WTP values spanning from $0 to $20,000 per DALY. At any specific WTP, we calculated the probability that an intervention would be cost-effective as the proportion of trials with positive net health benefits. For illustrative purposes, we use a WTP of $891 per DALY [10], equivalent to the 2019 per-capita Gross Domestic Product in Mali [16].

Sensitivity analysis
We performed a series of univariate analyses to identify which parameters would have the greatest influence on the results. We varied each individual parameter between its lower and upper 95% confidence limits, with all other parameters held at their point estimate. We then recorded the incremental cost-effectiveness ratio for each product at each limit.

Secondary analyses
We conducted six secondary analyses to assess the sensitivity of our conclusions to changes in model structure and assumptions (Table 2). For each, we altered the relevant feature while retaining all other model elements, including the Monte Carlo sampling. We evaluated the cost-effectiveness of the following: 1) maternal vaccine meeting WHO preferences for efficacy and duration [17]; 2) maternal vaccine with the overall efficacy from the RSV vaccine trial rather than from South Africa subset data; 3) providing a long-acting mAb as a birth dose and within routine immunization schedules compared to birth dose alone; 4) pre-seasonal maternal vaccine administration, instead of year-round; 5) base case interventions assuming they prevent RSV Upper Respiratory Tract Infections (URTI) in addition to RSV-LRTI; and 6) base case interventions assuming all infants receive appropriate medical care.
From the societal perspective, the ICER associated with each intervention would be $4280 (95% CI $1892 to $122,434), $1656 (95% CI $734 to $9091), and $8020 (95% CI $3501 to $47,047) per DALY averted, respectively (Fig. 1B). The cost per death averted for short-acting mAb, long-acting mAb, and maternal vaccine would be $116,342, $4,5104, and $224,593 for these interventions, respectively ( Fig. 2A). All ICER values exist in quadrant 1 of the cost-effectiveness plane. We evaluated the relationship between WTP for a DALY and intervention cost-effectiveness from the donor and government perspectives independently. At a product price of $3 per dose, long-acting mAb is preferable with greater than 50% likelihood of being cost-effective from the donor perspective at WTP above $1521 per DALY (Fig. 1C). For short-acting mAb and maternal vaccine, this mark is achieved at WTP values of $3638 and $7118, respectively. At a product price of $9 per dose, long-acting mAb and short-acting mAb are preferable from the donor perspective at WTP above $4781, $11,435 per DALY, respectively. Maternal vaccine is less than 50% likely to be cost-effective at a product price of $9 and WTP of $20,000. From the government perspective, where vaccine program costs are not influenced by product price, long-acting mAb is preferable at WTP above $841 per DALY (Fig. 1D). For short-acting mAb and maternal vaccine, these WTP values are $2011 and $3934, respectively.
We conducted a series of one-way sensitivity analyses to identify the parameters whose variance had the largest influence on the ICER. The most influential parameter across interventions was the inpatient case fatality rate, capable of modifying the ICER by up to 315% over the point estimate (Fig. 2B). Parameters with less influence-where the ICER remained within less than a 60% change from the point estimate-included the probability of receiving inpatient care, probability of LRTI given RSV, age-based RSV attack rates, intervention product efficacy, and inpatient care costs. For all other parameters, the ICER remained within less than a one percent change from the point estimate.

Secondary analyses
We conducted six secondary analyses to assess changes in model structure and assumptions (Fig. 3 Table 2 Secondary analyses and rationale.
Secondary analyses Rationale

To assess the cost effectiveness of a hypothetical RSV maternal vaccine product meeting WHO recommendations for efficacy and duration
There is a robust pipeline of candidate RSV preventive interventions in development. This analysis uses WHO preferences for product efficacy inputs rather than published efficacy data for limited RSV maternal vaccine products to date. The WHO preferred product characteristics for RSV maternal vaccines are at least 70% efficacy from birth to age four months [17]. 2. To assess cost-effectiveness of providing mAb as a birth dose and within routine immunization schedules compared to birth dose alone Since coverage of birth dose mAb is not expected to be perfect, this analysis explores the potential product cost-effectiveness of an alternative delivery strategy which provides catchup mAb to children during routine immunization visits during the first year of life. Guided by age-specific RSV attack rates estimated for infants from birth to 12 months of age in low-income countries [2], we projected attack rates among infants aged six to 12 months in Mali by applying a linear decline over this period, starting with the attack rate among children aged six months and ending at the attack rate among children aged three months. We then identified the routine pediatric immunization visit at which an older infant should receive the long-acting mAb, based on birth month. 3. To assess the cost-effectiveness of a pre-seasonal RSV vaccination strategy Pre-seasonal maternal RSV vaccination would decrease the product costs as compared to year-round strategies. This analysis explores the cost-effectiveness of pre-seasonal vaccination. 4. To assess cost-effectiveness of maternal vaccine using overall efficacy of RSV-F vaccine rather than subset data from South Africa The maternal vaccine efficacy estimate used in the base case is from a subset analysis of a clinical trial. This analysis uses the overall efficacy estimate against medically attended RSV-LRTI from the clinical trial, 39.4% (95% CI 5.3% to 61.2%) as a model input instead [23].

To assess cost-effectiveness of interventions assuming they prevent RSV Upper Respiratory Tract Infections (URTI) in addition to LRTI
Future RSV preventive interventions may have activity against RSV-URTI in addition to prevention of RSV-LRTI. This analysis assesses the cost effectiveness of interventions that prevent both outcomes. RSV-URTI was captured in the community-based incidence study in Mali through active surveillance for febrile acute respiratory infection, defined in the parent trial as a child presenting with fever in combination with any of the following: runny nose, nasal congestion, cough, difficulty breathing, purulent drainage from ear, or wheezing [8]. Among infants meeting testing criteria and with confirmed RSV illness, we assumed all those without RSV-LRTI developed RSV-URTI (Supplement Figure 5). For infants with RSV-URTI in the study, outpatient care was provided to 93.6%, and all others received no medical care [7]. Disability weights used to calculate DALYs for infants with RSV-URTI were 0.05 (95% CI 0.04 to 0.07) for those who received outpatient care and 0.006 (95% CI 0.003 to 0.010) for infants who did not receive care [26]. We evaluated the costeffectiveness of each intervention presuming any benefit to occur at the point of infection. 6. To assess cost-effectiveness of interventions assuming all infants with RSV-LRTI receive appropriate care Among young children in LMICs, an estimated 53% of severe RSV-LRTI episodes do not receive inpatient care, and 49% of RSV-LRTI deaths occur outside the hospital [2]. This secondary analysis assumes all infants with RSV-LRTI who require inpatient care receive it, and that all deaths due to RSV-LRTI occur in inpatient care settings.

Discussion
We estimated that if short-acting mAb (intra-seasonal), longacting mAb (pre-seasonal birth dose), or maternal vaccine (yearround) programs were implemented in Mali, the incremental cost to society would be $4280, $1656, and $8020 per DALY averted, respectively. This indicates that long-acting mAb may provide better value than either short-acting mAb or maternal vaccine in Mali and similar low-income countries. We evaluated the probability of each intervention being cost-effective across a broad range of willingness-to-pay values from government and donor perspectives. From the government perspective, implementing longacting mAb is likely to be cost-effective at willingness-to-pay values approaching the per capita GDP of Mali. The per capita GDP has been a commonly used threshold for health intervention costeffectiveness [10]. Recent studies examining country-level health spending have suggested that the actual willingness or ability to pay for health in Mali may be much lower, in the range of $14 to $311 per DALY [18,19]. Mali is one of the lowest-income countries in the world, and therefore donors considering a global portfolio may elect to support RSV prevention due to more favorable economic considerations in other countries. Therefore, if these interventions are supported by global donors, Mali may need Fig. 1. A) Disability-adjusted life-years (DALYs) averted as intervention coverage increases for a theoretical birth cohort of infants born in Mali followed for the first six months of life. B) Incremental cost-effectiveness ratio (ICER) as cost per full administration increases for each intervention, from the societal perspective. C) Probability that each intervention would be considered cost-effective from a donor perspective at a given willingness-to-pay for disability DALYs, and at price points of $3 or $9 per dose in 2019 United States dollars (USD). D) Probability that each intervention would be considered cost effective at a given willingness-to-pay for DALYs, from a government perspective. The gray dotted lines indicate the willingness-to-pay (WTP) thresholds, equal to the per capita gross domestic product (1xGDP) for Mali, 0.5xGDP, and 0.25xGDP, as labeled. An ICER less than the per capita GDP could be considered ''very cost-effective" by former WHO standards. subsidies to offset the administrative costs of RSV prevention as well as the product price. Ultimately, the decision about whether the benefits of investment outweigh the costs lies with the Malian people and their government.
In addition to providing sufficient value for money, new health interventions must be affordable [20]. The annual budget impact of adding long-acting mAb to the current immunization program in Mali would be a 0.21% increase of the overall 2017 health budget [16]. Gavi is the likely donor for RSV prevention products in Mali and other LMICs. Gavi has indicated support for such products, contingent on regulatory approvals and value for money commensurate with its investment case [21]. Our estimation of cost per death averted by long-acting mAb at $3 per dose, similar to the price negotiated by Gavi for a single dose of 13-valent pneumococcal conjugate vaccine [22], falls within the range for the Gavi investment case. While many of our parameters differ from those used in the investment case, the lower Mali-specific RSV-LRTI case fatality rate used in our analysis is balanced by the high observed attack rate.
In our secondary analyses, we assessed multiple alternative delivery and product performance scenarios. Our base case findings were generally conservative, and the alternatives led to lower incremental cost-effectiveness ratios for all interventions. In practical terms, if one or more of these alternative scenarios were true, then the RSV preventive interventions would more likely be costeffective. As the exception, replacing the South African sitespecific vaccine efficacy results from the RSV-F vaccine trial with overall efficacy results would raise the incremental costeffectiveness ratio. In that trial, the overall 39.4% efficacy (95% CI 5.3 to 61.2%) against the primary endpoint, medically significant RSV-LRTI up to 90 days of life, did not achieve statistical significance [23]. There were significant reductions in a secondary end- point, RSV-LRTI with severe hypoxia, as well as RSV-LRTI exploratory endpoints over the same duration of follow up, establishing proof-of-principle for maternal RSV vaccination to prevent medically important outcomes [23]. There is a robust pipeline of maternal RSV vaccines under development [17]. Our analysis indicates that a maternal vaccine with 70% efficacy, meeting WHO preferred product characteristics, would have a more favorable ICER. In addition to the challenges of product development and licensure, maternal immunization platforms in Mali and most LMICs require strengthening [24]. Donor support may therefore be required to implement a maternal vaccine program, but such investment could strengthen antenatal care health systems, providing broader benefits to maternal and child health overall.
The greatest strength of our study is use of country-specific RSV epidemiological and cost information. We used RSV incidence estimates from household surveillance standardizing identification of community cases. Nevertheless, our RSV incidence estimates came from a single full year of surveillance in urban Bamako, which were high compared to incidence estimates elsewhere [2]. We do not know whether the burden of RSV disease measured in Bamako reflected typically high disease burden in all of Mali or if it reflected an anomalous year at the study site. A previous costeffectiveness analysis which assigned a lower burden of disease to Mali found that RSV preventive interventions would not be cost-effective unless the willingness-to-pay were as high as $2500 per DALY, despite using more favorable characteristics of the maternal vaccine [6], underscoring the influence of disease burden on cost-effectiveness. Another limitation is that our analysis ends at six months of age. To the extent that RSV disease may be deferred but not averted by short-acting products, we may overestimate the value of prevention. However, as the most severe disease is concentrated in the first six months of life [2], a focus on this period is warranted. Finally, neither long-acting mAb nor maternal vaccine products have achieved licensure. As new clinical data emerge for these and similar products, their cost-effectiveness should be updated.
While RSV-LRTI is a high burden disease, the only licensed RSV-LRTI preventive intervention is cost-prohibitive for most LMICs. Future long-acting mAb and maternal vaccines have the potential to address this unmet global health need [4,23]. We used Malispecific epidemiology and cost data to conduct cost-effectiveness analyses of three potential RSV interventions, short-acting mAb, long-acting mAb, and maternal vaccines. The long-acting mAb product currently in development is likely to be cost-effective at prices near to what Gavi pays for similar interventions, and maternal vaccines which meet WHO preferred product characteristics could also be cost-effective. Ultimately, global health will benefit from the availability of multiple RSV preventive interventions. While the licensed product performance characteristics, prices, and delivery costs will drive policy and investment decisions in LMICs, our work highlights the potential benefit that RSV preventive interventions can have in Mali and similar countries.

Disclaimer
The authors alone are responsible for the views expressed in this publication, and they do not necessarily represent the decisions, policy, or views of their institutions.

Funding
This work was financially supported by grant 1,088,264 from the Bill and Melinda Gates Foundation. The funding source had no role in the study design; in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the article for publication.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.