Cost-effectiveness and public health impact of RTS,S/AS01E malaria vaccine in Malawi, using a Markov static model

Background: The RTS,S/AS01 E malaria vaccine is being assessed in Malawi, Ghana and Kenya as part of a large-scale pilot implementation programme. Even if impactful, its incorporation into immunisation programmes will depend on demonstrating cost-effectiveness. We analysed the cost-effectiveness and public health impact of the RTS,S/AS01 E malaria vaccine use in Malawi. Methods: We calculated the Incremental Cost Effectiveness Ratio (ICER) per disability-adjusted life year (DALY) averted by vaccination and compared it to Malawi’s mean per capita Gross Domestic Product. We used a previously validated Markov model, which simulated malaria progression in a 2017 Malawian birth cohort for 15 years. We used a 46% vaccine efficacy, 75% vaccine coverage, USD5 estimated cost per vaccine dose, published local treatment costs for clinical malaria and Malawi specific malaria indicators for interventions such as bed net and antimalarial use. We took a healthcare provider, household and societal perspective. Costs were discounted at 3% per year, no discounting was applied to DALYs. For public health impact, we calculated the DALYs, and malaria events averted. Results: The ICER/DALY averted was USD115 and USD109 for the health system perspective and societal perspective respectively, lower than GDP per capita of USD398.6 for Malawi. Sensitivity analyses exploring the impact of variation in vaccine costs, vaccine coverage rate and coverage of four doses showed vaccine implementation would be cost-effective across a wide range of different outcomes. RTS,S/AS01 was predicted to avert a median of 93,940 (range 20,490-126,540) clinical cases and 394 (127-708) deaths for the three-dose schedule, or 116,480 (31,450-160,410) clinical cases and 484 (189-859) deaths for the four-dose schedule, per 100 000 fully vaccinated children. Conclusions: We predict the introduction of the RTS,S/AS01 vaccine in the Malawian expanded programme of immunisation (EPI) likely to be highly cost effective. An intervention is considered cost-effective if the ICER per disability-adjusted life year (DALY) averted is less than three times the GDP per capita and is highly cost effective if the ICER per DALY averted is less than the per capita GDP 12. We used a Markov static cohort model developed by GSK for the RTS,S vaccine that has been validated for sub-Saharan Africa; the model is described in depth by Sauboin et al. 13. The model simulates a birth cohort followed over 15 years under fixed-exposure levels of malaria transmission, taking into account parameters reflecting healthcare provider and societal perspective to calculate the incremental cost effectiveness ratio per DALY averted (ICER) of the RTS,S vaccine 13. Figure 1 is a diagrammatic representation of the model. The model has compartments susceptible (S), infected (I), clinical disease (C) and severe disease (F) divided into six successive immunity levels following each infection levels. The model assumes two processes for acquisition of immunity, one process that protects against clinical malaria of any severity and a faster process that protects against severe malaria. M = maternal protection; S = susceptible; I = infected (parasites emerging from the liver); C = clinical disease episode; F = severe disease episode. There are six levels of immunity with compartments S, I, C and F divided into six levels. R = resistant; wm = waning of maternal immunity; q = probability of infection; s = susceptibility to infection as a function of age; a = probability of asymptomatic infection; r = recovery rate from clinical disease; w = waning rate of acquired immunity; r imm = probability of developing full immunity. The model assumes initial protection against malaria from maternal antibodies (M) 14. Neonates are considered either protected from (M) or are susceptible to (S1) malaria infection. Initial immunity is presumed to wane exponentially over three months, leaving the child susceptible to infection. An infected (I1) child will have asymptomatic parasitaemia which clears and susceptibility returns (Si), or the child will develop clinical disease (C1). From clinical disease a child may recover (r1) or develop severe disease (F1) where they could either survive returning to a susceptible state or they could die. Immunity is enhanced every level from an asymptomatic state to clinical malaria and to severe disease. The model permits up to six repeated infections to cumulatively increase immunity. Beyond six infections, a fixed proportion of children is assumed to develop a state of resistance (R). The model uses an estimated 2017 annual birth cohort for Malawi and followed for 15 years 15. This birth cohort was the mean of four prior birth cohorts using the United Nations population data 15. The model accounts for heterogeneity of individual level exposure and a fixed probability of infection within each transmission category. The model assumes the vaccine efficacy wanes over time. Malaria transmission intensity in the model was defined categorically as low, medium or high based on Plasmodium falciparum parasite prevalence (PfPR) in children aged 2–10 years old of <5%, 5–40% and >40% was respectively, using the Malaria Atlas Project 16. Table 1 shows the input parameters used in the model. The inputs were point estimates extracted from published literature or reasonable assumptions. The vaccine price was based on previously published assumptions since the product has not yet been priced by GSK. The cost of RTS,S vaccine delivery per dose was assumed equal to DTP3 (given as part of pentavalent) in Malawi 17. Service delivery make up the bulk (63%) of vaccine delivery costs whilst supply chain and logistics constitute the remainder of vaccine delivery costs 18. Vaccine delivery costs mainly comprise of cold chain management, transportation of vaccines to health facilities, waste disposal and additional training for health workers. We sought to calculate cost savings from a healthcare and household perspective. Societal costs are a combination of healthcare and household costs. The Phase III RTS,S/AS01 trial vaccine schedule of 6, 7, 8 and 26 months of age and 18-month follow-up results, following the third dose, were fitted in the model. Vaccine efficacy against clinical and severe malaria in children was 46% (95% CI 42–50%) and 34% (95% CI 15–48%) respectively 5. Third and fourth dose RTS,S coverage were assumed to be 75% and 60% of the DTP3 dose 1 coverage respectively. The fourth dose was assumed to boost the waning efficacy. Access to artemisinin combination therapy (ACT) or private dispensaries was extracted from the 2014 Malawi Malaria Indicator Survey 19. We used published treatment costs for mild-moderate and severe gastroenteritis, respectively 20, 21, since published treatment costs for malaria were outdated or unavailable. These health costs including drugs, laboratory investigations, staff salaries and facility costs. Where these specific costs were unavailable for clinical and severe malaria, we used malaria sequelae costing data from Tanzania 22, as cost data from Malawi were not available. Direct and indirect household costs incurred in care seeking were also based on those locally empirically observed in gastroenteritis 21. Direct household costs included travel, consultation fees, treatment sought before and after health facility visit and the costs of food and shelter for the carer. Indirect costs comprised income lost while caring for the child 21. Bed net use and access to and usage of ACTs and the proportion of those who seek treatment at a private dispensary were derived from the Malawi Malaria Indicator Survey 19. Vaccine price per dose has not yet been set by GSK, so we assessed a range of costs of USD1, USD5 and USD10. RTS,S delivery in the Malawi EPI was taken from the administration cost pentavalent vaccine 7. The cost of delivery includes all the necessary materials and health worker time required to administer a vaccine in the EPI. The mean Malawi GDP per capita from 2010 to 2015, as reported by the World Bank, was used to compare with the ICER per DALY averted 28. Univariate analysis was conducted by running the model through different values of vaccine price and vaccine coverage, as shown in Table 3 and Table 4, whilst the other input parameters were held constant.