Rubella vaccine introduction in the South African public vaccination schedule: Mathematical modelling for decision making

Background: age structured mathematical models have been used to evaluate the impact of rubella-containing vaccine (RCV) introduction into existing measles vaccination programs in several countries. South Africa has a well-established measles vaccination program and is considering RCV introduction. This study aimed to provide a comparison of different scenarios and their relative costs within the context of congenital rubella syndrome (CRS) reduction or elimination. Methods: we used a previously published age-structured deterministic discrete time rubella transmission model. We obtained estimates of vaccine costs from the South African medicines price registry and the World Health Organization. We simulated RCV introduction and extracted estimates of rubella incidence, CRS incidence and effective reproductive number over 30 years. Results: compared to scenarios without mass campaigns, scenarios including mass campaigns resulted in more rapid elimination of rubella and congenital rubella syndrome (CRS). Routine vaccination at 12 months of age coupled with vaccination of nine-year-old children was associated with the lowest RCV cost per CRS case averted for a similar percentage CRS reduction. Conclusion: At 80% RCV coverage, all vaccine introduction scenarios would achieve rubella and CRS elimination in South Africa. Any RCV introduction strategy should consider a combination of routine vaccination in the primary immunization series and additional vaccination of older children. To explore the impact of introduction of RCV into South Africa, we used a previously published deterministic discrete time age-structured model [16,20] which is characterized by a matrix capturing transitions between epidemiological states (maternally immune (M), susceptible (S), infected (I), recovered (R), and vaccinated (V)) and between age groups (Supplement 1 Figure S1). Individuals in the maternally immune (M) compartment are children born to mothers who are immune to rubella and passively acquire immunity. Susceptible (S) individuals are those who lose maternal immunity or are born susceptible and at risk of becoming infected (I). Infected individuals recover by the next time step moving into the recovered (R) compartment. The vaccinated (V) compartment represents individuals who receive RCV and are successfully immunized. The time step used in the model was ~16 days, as this corresponds to the generation time of rubella. See Supplement 1 for model details. One of the key model inputs is the basic reproductive number (R0), which is the average number of secondary infections resulting from a typical infectious person in a totally susceptible population. The value of R0 used in this model was 7.9 and was obtained from a previously published modelling study estimating R0 for 40 African countries [21]. We proceeded to run simulations with different estimates for R0 in a sensitivity analysis. The highest estimate used was an R0 of 12 which was estimated in Ethiopia [22] and the lowest estimate estimated in Burkina Faso was 3.3 [21]. The nature of interactions between individuals influences transmission of infectious diseases. This was represented in the model as a function for seasonal amplification [23,24,25] and age-specific mixing based on estimated age-dependent social contact [26] and non-modelled heterogeneities [27]. Duration of maternal immunity [28] and vaccine efficacy [29] were estimated from published literature. Demographic data over time for South Africa were obtained from projections by the United Nations (UN) Population Division [30]. See Supplement 1 for model parameter details. Preventing CRS is the main reason for administering rubella vaccination, given the mild nature of infection among children and adults. We therefore assessed the impact of vaccine introduction on both rubella and CRS incidence over time for all scenarios. To estimate the burden of CRS, we combined rubella age-specific incidence generated by the model with an age-related fertility profile for South Africa obtained from the UN Population Division 2015 estimates [30]. We explored vaccine introduction scenarios that reflect options that might be implemented in South Africa (Table 1). The measles vaccine is currently administered at six months and 12 months as part of the EPI schedule in South Africa. Previously, country-wide SIAs were organized every four to five years but in recent years, SIAs are only organized as measles outbreak control measures in affected districts or provinces. On the contrary, RCVs are currently available in South Africa but only in the private health sector, which caters for about 15% of the population. We therefore fixed RCV coverage in our simulations to 15% prior to introducing the vaccine in the EPI schedule. The WHO recommends an initial SIA, targeting a wide age range of individuals, with the concurrent introduction of RCV into the routine EPI schedule [6]. We simulated rubella disease dynamics for 55 years (1995 to 2050) by first simulating endemic rubella disease dynamics (from 1995 to 2019) before initiating vaccine introduction (from 2020 onwards). Analyses covered three time horizons (10 years, 20 years, and 30 years) following RCV introduction to encompass various time frames required for CRS elimination using different RCV introduction strategies [6]. Possible scenarios for rubella-containing vaccine (RCV) introduction in South Africa. Rubella containing vaccines if introduced into the South African EPI program will be in combination with measles vaccine. Estimates of coverage for the second dose of routine measles vaccination [31] according to the South African government differ from those of WHO (79% versus 53% in 2017, 75% versus 50% in 2018). To encompass the emergent properties of a range of potential coverage values and target ages for routine immunization (9 or 12 months), we considered an array of scenarios reflecting different levels of coverage for routine vaccination achieved by 12 months, ranging from 60% to 95%. We also considered one scenario in which a dose of RCV was administered to boys and girls at the same age as the human papillomavirus (HPV) vaccine. The HPV vaccine is administered each year to nine year-old girls in schools in South Africa. It is reasonable to assume that this approach could be considered in an attempt to cover the adolescent population in the absence of SIAs. For all RCV introduction scenarios, including SIAs, we set the coverage of RCV during SIAs at 80% because this is the minimum coverage recommended by WHO [6]. We also set the coverage of RCV at 80% at the time of co-administration with HPV vaccine in order to be consistent with RCV coverage for individuals who are above the age for the primary series of RCV. We evaluated costs relating to introducing the RCV from the perspective of the South African government as additional cost per dose of RCV compared to the current practice of administering measles-only containing vaccine. In the absence of detailed information, we assumed that no additional program costs are associated with introduction of the RCV vaccine, due to a direct substitution of the RCV with measles-only containing vaccine. Thus, for rubella, focusing on additional (undiscounted) costs relative to the measles baseline should be appropriate to guiding the investment case for rubella vaccine introduction. The price per dose of the measles vaccine currently used in South Africa (10 doze vial) in South African Rands (ZAR), is ZAR 29.13 [32]. For RCVs, we estimated price per dose for MR (ZAR 38.00 per dose) and MMR (ZAR 81.00 per doze) based on prices reported for the Pan American Health Organization (PAHO) in the Market Information for Access to Vaccines database [33]. PAHO prices are usually within 10% of vaccine prices in South Africa (personal communication with the national cold chain manager). We assume that a multi-year contract will be signed such that the price of the RCV remains the same for the duration of the simulations. To obtain additional costs of RCV introduction, the difference in price per dose between the RCVs (MR and MMR) and the measles vaccine was multiplied by the total number of persons vaccinated under each scenario. Total numbers of persons vaccinated under each scenario were estimated by applying expected coverage estimates to corresponding target populations obtained from the UN population estimates [30]. The number of CRS cases averted in each scenario was obtained by subtracting the number of CRS cases in that scenario from the number of CRS cases in scenario 1. For each scenario, we calculated the number of RCV doses per CRS case averted by dividing the total number of RCV doses used by the total number of CRS cases averted. The corresponding cost per CRS case averted was obtained by dividing additional RCV costs by CRS cases averted. These estimates were obtained for MR and MMR using 60% coverage representing the worst case scenario, 80% coverage representing the WHO recommended minimum coverage for RCV introduction and the 95% coverage level representing the best case scenario. To assess total undiscounted disability-adjusted life years (DALYs) averted by the introduction of RCV, we estimated the number of CRS cases averted from 2020 to 2050 (as the difference in CRS cases between each RCV scenario and the no RCV scenario) and applied this to undiscounted DALYs lost per CRS case. DALYs lost per CRS case were obtained from an existing study reporting DALYs lost for a range of countries using disability weights from the 1990 and 2010 Global Burden of Disease (GBD) study [34]. For the purposes of this paper, we use estimates of DALYs lost reported for upper middle-income countries (World Bank classification for South Africa) using 2010 GBD disability weights. An important measure of the success of vaccination programs is the degree to which they can sustain elimination and (even transiently) prevent outbreaks [35]. The effective reproductive number (RE) is the average number of secondary cases resulting from the introduction of one infectious person into a population containing some individuals who are not susceptible to the infection [36]. Estimates of RE have been used to inform timing of vaccination interventions for preventing disease outbreaks [37], and to determine the likelihood for disease outbreaks in populations if an infectious case was introduced [38].The endemic nature of rubella in the absence of RCV and the subsequent change in number of susceptible individuals with an introduction of RCV could result in a change in RE over time. A value for RE greater than one implies rubella outbreaks can occur and values less than 1 mean that the infection goes into extinction. Values of RE over time were extracted from the simulations to understand the impact of various scenarios on estimated time to rubella elimination and periods when there was a rebound in RE from values below 1 to values greater than 1. These fluctuations in RE associated with different scenarios will inform vaccination activities which should be implemented even after perceived short-term elimination is achieved to avoid possible rebound in rubella incidence. We extracted and presented values of RE for the entire period during which the simulations were run. The effective reproduction number (RE) was estimated from the model output using the next generation method [39].