Introduction
Respiratory syncytial virus (RSV) is one of the major causes of acute lower respiratory infection (ALRI), resulting frequently in hospitalisation (and sometimes death) in children under the age of 5 years [
1,
2]. A recent systematic review estimated that 33 million RSV-ALRI cases and 118 thousand deaths occur annually in children under 5, of which 22 million cases and 103 thousand deaths are in low-income countries (LICs) and lower-middle-income-countries (LMICs) [
2].
The currently available prophylactic RSV intervention (palivizumab) consists of a monthly injection over the course of the RSV season. It is only recommended for high-risk groups and only used in a few countries, partly because it is expensive [
3]. However, there are multiple maternal RSV vaccine and monoclonal antibody (mAb) candidates under clinical development [
4]. The most advanced maternal vaccine candidate, NCT02624947 (Novavax), has completed the phase 3 trial, but did not meet its primary endpoint of medically significant RSV lower respiratory tract infection (LRTI) [
5,
6]. A new single-dose mAb candidate, MEDI8897 (Sanofi Pasteur / MedImmune), has also completed the phase IIb trial and met its primary endpoint of medically attended LRTI due to RT-PCR-confirmed RSV [
7,
8].
Passive immunisation through maternal immunisation or monoclonal antibodies could potentially prevent RSV disease after birth, by protecting vulnerable infants [
9]. In most LMICs and LICs, the Expanded Programme on Immunization (EPI) is well established, including paediatric and neonatal (at birth) components, and the single-dose RSV mAb, administered at birth, could be integrated into existing prevention programmes. In recent years, initiatives to administer inactive influenza vaccines and pertussis vaccines to mothers during antenatal care (ANC) visits have been successfully implemented in many high-income countries. In LICs and LMICs, the dramatic increase in the antenatal care (ANC) coverage during the past decade renders feasible the implementation of the maternal RSV vaccination programme via the ANC platform [
10].
There are limited resources and funding available in LICs and LMICs, which leads to inequitable access of vaccines in these countries. The investment decisions on vaccines need to be made both at the country and a multi-country level. Gavi, the Vaccine Alliance, is an organisation that invests in vaccines to protect children’s lives and health in LICs and LMICs. Every 5 years, Gavi develops a new vaccine investment strategy (VIS) to prioritise the vaccines made available to countries throughout their vaccine support programme. RSV interventions, including both maternal vaccine and mAb, were considered as one of the six prioritised vaccine programmes as part of Gavi VIS for the 2021–2025 funding period [
11].
The majority of LICs and LMICs have sparse or no data to make decisions on RSV interventions, but studies on RSV epidemiology and outcomes may represent a costly endeavour in resource-strapped settings. Therefore, it is imperative that a cost-effectiveness analysis of RSV interventions carefully evaluates the uncertainty and main drivers of implementation of RSV prophylaxis. This enables identifying the value that additional data can provide to reduce uncertainty. While RSV prevention technologies are still in development, data can be collected on its disease burden and epidemiology to inform decision-making once the technologies are available. In order to be informative at both the multi-country (i.e. Gavi, WHO) and single country levels, this analysis aims to identify the key drivers of the cost-effectiveness of potential maternal vaccination and infant monoclonal antibody strategies against RSV in 72 of the current and former Gavi-eligible countries.
1
Discussion
Health economic evaluations for multiple interventions and multiple countries are invaluable for decision-makers involved in global initiatives. Our model application, MCMARCEL, has been designed specifically to simultaneously and efficiently evaluate the costs and effects of many countries and strategies, fully accounting for parametric uncertainty, with the aim to identify influential parameters. This study evaluated the health and economic effects of potential maternal RSV vaccines and monoclonal antibodies for infants in 72 current and former Gavi-eligible countries. We estimated that the annual disease burden in these 72 countries is substantial, including 20.8 million RSV cases, 1.8 million hospital cases, and 40 thousand deaths, with average total direct costs exceeding 600 million USD. We found that the maternal strategy would be the optimal strategy given a price of 3 USD per dose at a WTP value greater than 3000 USD in most countries (range 1000–3500 USD) per DALY averted, whereas the mAb strategy would be optimal given a price of 6 USD per dose at a WTP value greater than 6000 USD (range 3500–8000 USD) per DALY averted. At these WTP values, the results are surrounded by large uncertainties mainly caused by uncertainty around RSV incidence, mortality, and especially RSV hospitalisation rates. At lower WTP values (below 2500 USD per DALY averted), we are more than 75% certain that no intervention is optimal for the majority of countries. Hence, despite the notable burden of RSV in early life, the short-lived protection of the potential tools indicates that any prophylaxis would have to be competitively priced in order to be considered cost-effective in Gavi-eligible settings.
To date, there is no licensed RSV vaccine (for mothers or children) nor single-dose mAb, and the interventions’ duration of protection is unknown. An additional month of protection was assumed for the mAb versus the maternal vaccination strategy, which is consistent with the assumptions in the Gavi RSV investment case [
31]. If mAb would offer the same efficacy with a shorter duration of protection, but at a higher price than the maternal vaccine, then mAb would simply be dominated by the maternal vaccination strategy, irrespective of willingness to pay. For both the maternal vaccination and mAb strategies, simplified ‘all or nothing’ assumptions on duration of protection were made, where we assumed no waning of protection up to the end of the period of protection, at which point protection drops to zero. We also assume the same duration of illness for both non-vaccinated and vaccinated children. The reason for such assumptions is that the duration of protection is unknown. Hence, sensitivity analyses were performed (Additional file
1). In this analysis, the focus is on estimating the disease burden, given the currently available data, and in identifying the main data gaps to fill (besides information on the intervention tools) in order to reduce decision uncertainty about interventions (defined according to simple characteristics). Given a modelled period of protection, this simplification—without waning immunity—is likely to result in an overestimation of disease averted from these interventions, ceteris paribus.
The most advanced candidate is the maternal RSV vaccine (NCT02624947) developed by Novavax. Unfortunately, its phase 3 randomised controlled trial did not demonstrate efficacy against its primary endpoint (medically significant RSV LRTI), but it showed significant efficacy against one of its secondary endpoints (hospitalisation) [
6]. In order to maximise the information provided by our simulations, we not only modelled an extensive range of scenarios based on efficacy intervention and price ranges (while accounting for all parameterised uncertainty), but we also modelled a scenario using currently available peer-reviewed data on the Novavax maternal vaccine (while accounting for all parameterised uncertainty, including wide uncertainty intervals for efficacy against the primary and secondary endpoints; see Additional file
1) [
6]. However, additional phase 3 trial data of this vaccine were presented at a recent international conference [
32]. These new per-protocol post hoc analyses were interpreted by Novavax as indicating their maternal vaccine is effective against the broad endpoint of all-cause pneumonia (i.e. not just RSV LRTI) among infants. Clearly, if the RSV interventions (maternal vaccine or mAb) would offer such additional protection, its value would greatly increase. This remains as yet a controversial result—especially since the same trial failed to demonstrate significant efficacy against its FDA-prescribed primary endpoint—that needs to be confirmed by thorough peer review. In view of potential confounders influencing the broad endpoint of ‘any medically significant LRTI’, it is important to establish at least whether (a) randomisation accounted for comparable within-household and community vaccination status against influenza, pneumococcus, and
Haemophilus influenzae type b in both arms of the trial; (b) these observations hold in the different country sites of the trial; and (c) to which extent this is observed for outpatient versus inpatients. We therefore opted not to explore the ramifications of these potential benefits in additional scenario analyses in the current paper. Nevertheless, our extensive sensitivity analyses on efficacy, waning, and price improve our understanding of the cost-effectiveness of new RSV interventions. Therefore, this type of analysis can inform vaccine developers’ product portfolio development, as well as decision-makers and their advisors’ planning activities.
When WTP values are less than 1000 USD per DALY averted, there is little uncertainty, given the product characteristics we explored, that current practice is optimal, and the resources are better spent on other interventions in these countries. If WTP values above 1000 USD and 3500 USD are considered acceptable, the maternal and mAB strategies, respectively, are increasingly likely to become the optimal strategies. This last result is however surrounded by much uncertainty. Hence, it would be valuable to obtain more evidence on RSV incidence, mortality rate, and especially hospitalisation probability. Of course, a refined understanding of the features of any RSV prophylaxis would go a long way in reducing the uncertainty around the optimal intervention. However, community-based incidence studies, even those undertaken prior to the implementation of RSV prophylaxis, could also play a role in reducing the uncertainty around the question of cost-effectiveness. According to our analysis, the importance of community-based incidence studies, as opposed to hospital-based incidence studies, cannot be stressed enough; the probability of cases in the community that are hospitalised is one of the most valuable parameters that we could study, and this parameter cannot be estimated from any of the 76 hospital-based incidence studies that are documented in Shi et al. (as they lack any information on the incidence burdening the community in which the studies were set) [
2]. Moreover, the hospital-based case-fatality ratios cannot reflect the overall RSV-associated deaths in LIC and LMIC settings. We calculated the overall RSV-associated death rates based on adjustment factors derived in Shi et al. from only three studies in LMIC, as no country-specific data is available. The adjustment factor of 2.2 (with uncertainty interval of 1.5–2.9) was used for all countries to country-specific hospital case-fatality rates in order to estimate country-specific RSV community deaths. In countries where relatively fewer RSV-associated ALRI cases seek health care, this factor is likely to be higher and vice versa. It is therefore highly likely that community RSV deaths were underestimated for LIC. In our EVPPI analysis, this adjustment factor is one of the top influential factors for this analysis. Hence, in addition to the importance of the burden information per se for policy-makers, studies on RSV-associated community deaths can also provide highly useful information for cost-effectiveness analysis.
To the best of our knowledge, this is the first multi-country, multi-intervention cost-effectiveness analysis for childhood prophylactic interventions against RSV in LIC and LMIC settings. Two modelling studies applied Kenyan data to understand the impact of RSV vaccination in low-income settings. Poletti et al. evaluated several prevention strategies, including maternal, paediatric (at 3 months), and school-age vaccination. They predicted a 30% reduction in RSV infant infection via maternal vaccination, which is very similar to our prediction [
26]. Additionally, Kinyanjui et al. used a dynamic transmission model to investigate the health impact of vaccinating older children when natural maternal immunity has waned (5–10 months) on infants, which is a different research question [
33].
This analysis has several other strengths: firstly, we re-estimated the age-specific disease incidence on a per-month basis using detailed data presented in a published systematic review and we used model-based estimates of the total country-specific burden in all 72 countries. Next, rather than estimating one region-level cost of illness [
34], we estimated country-specific costs using a simple method that leveraged the existing literature on pneumonia costs in combination with the WHO-CHOICE costs to capture the country-to-country variation in costs. Secondly, we accounted for parameter uncertainty in a probabilistic way and performed extensive scenario analyses to assess the impact of uncertainty in parameters that have multiple and complex effects on the outcomes and costs of RSV. We also plotted the overall trends and impact of different assumptions in both current disease trends and potential implementation programmes. Thirdly, this analysis includes a value of perfect information analysis to inform data collection prioritisation both for countries and vaccine developers in order to improve the timeliness of the decision process once the products become available. Moreover, it also reflects the ‘full value of vaccines’ agenda from the WHO that encourages evaluation of vaccine value during the development phase so as to avoid vaccine manufacturers bring unmarketable vaccines into late stage development or onto the market [
35]. Finally, the investment decisions of RSV interventions need to be made both at the country and multi-country level (i.e. through the WHO policy recommendations and Gavi financial support). Since the WHO has explicitly discouraged the use of GDP per capita as a benchmark of the incremental direct costs that a country should be willing to pay to gain a QALY, or avert a DALY [
36], we present and interpret this complex analysis across different WTP levels for various stakeholders. Consequently, the decision of the optimal strategy can be made based on any WTP threshold within the wide range we explored.
Within each country, the coverage is likely to differ between different intervention strategies, whereas we assumed BCG infant coverage for all strategies in the interest of parsimony. Since this is a static model and marginal intervention costs are assumed to be directly proportional to doses administered, the coverage will only impact the overall disease burden averted and not the cost-effectiveness [
37]. Although the maternal tetanus vaccination programme is available in a few countries under the tetanus elimination goal, the maternal vaccination programme is not yet well established in most of the 72 countries, representing an implementation challenge. The maternal vaccine could be administered during an ANC visit. The current WHO recommendation for maternal acellular pertussis vaccination is in the second or third trimester and at least 15 days before birth [
38]. With the ANC visit coverage and frequency (ranging from 4.3% in Somalia to 96% in Armenia, at least four times throughout pregnancy) in LIC and LMIC countries [
10], maternal immunisation coverage is unlikely to achieve the BCG coverage level, implying we may have overestimated the impact of this strategy on disease burden.
The maternal strategy captures only the health benefit to infants in our model; however, the potential benefits to the mothers and household contacts are not considered, as the burden of RSV in childbearing-age females and the RSV vaccine duration of protection in adults are both unknown. From the literature, the highest disease burden of RSV is in the very young age group; therefore, the health benefits to the mothers or contacts are unlikely to have a significant impact on the overall results. Additionally, there may be uncaptured psychosocial benefits to caregivers in improved child survival and health. Future research could consider factors in RSV epidemiology that were beyond the scope of the current model’s structure and the data available. The impact of transmission dynamics could capture the effects of prophylaxis on the contacts of the mother or the child, but the existent data on disease dynamics precludes the development of a dynamic model applicable to all settings in our analysis. Relatedly, our model assumes that cases averted on a certain month are averted altogether, rather than shifted to later ages, potentially favouring the case for RSV prophylaxis. However, RSV incidence peaks at a very young age in most countries and younger cases are more severe than older cases (in terms of the need for hospitalisation and the probability of death); hence, we believe that a slight increase in incidence in children over 6 months of age would not represent a displacement in time of hospitalised/fatal cases, since older cases are less likely to result in hospitalisation or death.
The seasonality of RSV is not considered in this evaluation as the analysis includes countries in several continents with differing seasonal patterns. It is challenging to identify the start and peak of RSV seasons in the absence of good surveillance in these countries. Cromer and colleagues explored a seasonal RSV vaccination programme in the UK, showing that targeting births in particular months of the year might be more efficient than a year-round programme [
12]. The implementation challenges of such a seasonal programme in low- and middle-income countries are currently not well documented. For instance, young infants born before the RSV season would still be vulnerable for severe RSV disease, but they may be difficult to reach with a seasonal programme. One way to interpret our results in light of a seasonal programme is by assuming that this would require fewer doses for the same (or similar) effectiveness. For instance, if only half of pregnant women/newborns would need to be targeted to protect them through an RSV season occurring over part of the year, then all WTP changepoints (where the optimal strategy of choice changes) in our analyses would be approximately halved.
This analysis also has a few additional limitations, but all render our analysis more conservative (we bias against implementation and the potential misuse of existing resources). The indirect effects on disease transmission are ignored, which in the case of the strategies considered here would likely represent an underestimation of the benefits. The fixed implementation costs (i.e. campaign, establishing a new vaccination visit instead of using existing platform) were not considered, although the intervention cost (procurement and delivery costs) per dose was included as a variable in our model. If one strategy would require higher fixed costs than another, it may impact the ICERs and therefore the choice of the optimal strategy. Direct non-medical costs (e.g. transportation) and indirect costs (e.g. productivity losses) are not included in our analyses due to a lack of data. Uncertainty around the proportion of RSV cases not seeking health care was not explored due to insufficient information. Interventions targeting high-risk groups only (i.e. preterm, bronchopulmonary dysplasia, and chronic heart disease) are not considered as these groups are unlikely to be identified easily in these countries. This analysis focuses on the RSV-associated ALRI as defined in the systematic reviews [
1,
2]. Other RSV-associated acute infections are not included (i.e. upper respiratory tract infection, otitis media). Longer-term chronic conditions, such as recurrent wheezing and asthma, are also not considered yet due to limited evidence that these are influenced directly by RSV infections; there is also a paucity of health and cost burden data on these conditions for the countries considered here. Furthermore, a recent study showed that current RSV prophylaxis has a limited impact on the reduction of these conditions [
39]. Evidence about the causal association between RSV and these conditions is still emerging.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.