Our findings show that the number of lives saved in a head-to-head comparison of a single-age cohort of infants and pre-adolescent girls, with the rotavirus and HPV vaccines, respectively, in the GAVI countries are very similar even though the two vaccines target different populations in terms of size and gender, and even though the diseases the vaccines target have different epidemiological features. The target population for rotavirus vaccines is approximately twice as large as the target population for HPV vaccines (~76 million versus ~32 million for a single cohort) since the former includes both genders while the latter only includes girls. However, the absolute number of deaths averted within each age cohort over their lifetime is slightly lower for rotavirus vaccines than for HPV vaccines (~274,000 versus ~286,000); partly, this is due to the higher disease-specific mortality (i.e., case fatality rate) of cervical cancer compared to that of rotavirus gastroenteritis.
We found 5.2 lives were saved per 1,000 children vaccinated against rotavirus, and 12.6 lives were saved per 1000 pre-adolescent girls vaccinated against HPV 16,18. The timing of the lives saved, however, drastically differs since the deaths attributable to rotavirus occur in close proximity to the acute infection, while deaths attributable to HPV-related cervical cancer occur in the girl's adult life largely after the age of 30.
Given this time difference of lives saved, when we translated the health outcomes into DALYs averted applying a 3% discount rate, the number of DALYs averted was much higher with rotavirus vaccines compared to HPV vaccines (~7.15 versus ~1.30 million). Recently, arguments to use 'slow discounting' are made on the basis of this large time effect when the health outcome of an intervention is far in the future from the time of the actual intervention [64
When we used WHO's cost-effectiveness threshold based on GDP per capita, in most of the GAVI countries (68 for rotavirus and 66 for HPV, at the cost of I$25 per vaccinated individual) the incremental cost per DALY averted was lower than each country's GDP per capita - implying both vaccines would meet the criterion for being "very cost-effective" [44
We emphasize that our analysis was intended to provide a broad comparative overview of the potential impact of the rotavirus and HPV vaccines in the GAVI-eligible countries, rather than to provide precise estimates of the cost-effectiveness profiles of the two vaccines. The comparison is not based on empiric data obtained from a head-to-head clinical trial where randomization can minimize the risk of introducing bias caused by differences in the two vaccination programs. Such clinical trials are incredibly difficult to design when the comparison is based on different populations and for different conditions. Our cost-effectiveness results should be considered preliminary since the composite costs for vaccine purchase and delivery were assumed to be the same for both vaccines and uniform across all GAVI-eligible countries, and thus the differences in the cost-effectiveness results primarily reflect the relative disease burden within and across each of the GAVI countries. We therefore caution that our base-case cost-effectiveness results should be interpreted carefully. We suggest that, for priority setting decisions at a country level, more rigorous comparisons should be performed as country-specific data become available. We hope our analysis would facilitate a more specific discussion about important issues (as will be discussed later in this section) that should be considered in setting priorities among different vaccination programs competing for limited resources.
Other limitations of our analysis include the following: First, our static models are not designed to capture potential herd immunity effects that might result from either vaccine intervention, thus likely underestimating the true impact of the two vaccines. From a comparative standpoint, our results do not reflect potential differences in level of herd immunity between the two vaccines, which in turn are dependent on multiple complex factors (e.g., the nature of disease transmission and contact pattern in a given setting) for each pathogen. Second, when country-specific data were severely lacking, our analysis relied on standardized assumptions that may skew our estimates for both absolute and relative vaccine benefits in different ways for each country. For example, similar to the case of the composite vaccination program costs described previously, we relied on standardized assumptions on medical utilization practices (e.g., one-time outpatient visit plus a 3-day admission for treating severe rotavirus gastroenteritis requiring hospitalization) when calculating the costs to treat diseases caused by the two pathogens in each country. Third, due to the scarcity of relevant data on rotavirus disease in older individuals, our rotavirus model focuses on the disease burden among children under age 5 with a time horizon of 5 years. While severe rotavirus diseases occur primarily in younger children and adult cases are usually not severe, it is known that individuals older than 5 years may also develop symptomatic rotavirus diseases. Our study therefore presents a likely, but moderate, underestimate of the health benefits of rotavirus vaccination. Finally, we did not take into account the potential impact of current supplementary or alternative interventions (e.g., cervical cancer screening) on vaccine cost-effectiveness or the technological changes that may affect the cost-effectiveness profiles of the vaccines in the long-term.
Despite the limitations, the present comparative analysis, coupled with our previously published analyses, highlight some key factors related to the comparative impact, affordability, cost-effectiveness, and distributional equity that decision-makers must consider when introducing new interventions in resource-limited settings. Moreover, the analysis provides insight into several uncertainties that should be considered when assessing these vaccines.
(a) Uncertainty about natural history, epidemiology, and vaccine efficacy
Many aspects of the natural history of rotavirus and HPV infection are still unknown, and there is still uncertainty about the vaccines' clinical effectiveness in different settings. For example, for rotavirus, additional research is needed to explore the nature of natural immunity, the serotype distribution of rotavirus in local settings, and long-term vaccine efficacy [16
]. For HPV, in addition to uncertainty surrounding serotype distribution by setting (and over time) as well as the magnitude of cross protection, there are also uncertainties about duration of immunity and age-specific efficacy once sexual debut has occurred [4
]. Vaccine performance, as well as natural history of HPV, in boys is less certain than in girls although information is being collected. Thus, whether an HPV vaccination program should also include boys has been a subject of much discussion [32
]. These issues may affect the effectiveness, cost-effectiveness, and budgets of programs vaccinating against rotavirus or HPV. Accordingly, as new clinical evidence from local settings becomes available, re-evaluation of the vaccines in real-world settings will be needed.
(b) Potential effects of different program costs
As previously discussed, it should be noted that our comparative evaluation standardizes vaccination cost per individual using a composite cost approach (i.e., set at either I$10 or I$25 for all GAVI countries) [6
]; this is partly due to uncertainty about vaccine prices and country-specific program delivery costs associated with the introduction of either vaccine and partly for ease of comparison. Accordingly, the differences in the cost-effectiveness profiles of rotavirus versus HPV vaccines across the 72 GAVI countries stem mainly from the differences in estimated disease burden and vaccine efficacy adjusted for serotype distributions. It is of course possible that the unit prices of the two different vaccines may settle at different levels. In addition, because it is likely to be harder to reach adolescent girls than infants, the net costs for delivering the two vaccines may not be similar within some countries; on the one hand, the cost for delivering HPV vaccines could be higher, and on the other delivery may be packaged with other health programs, directed towards adolescents, a country is prioritizing [4
]. Both of these factors could influence the prioritization of the two vaccines within a country or across all GAVI countries. Given that our sensitivity analyses show that the cost-effectiveness results of the two vaccines were most sensitive to the cost per vaccinated child or girl, it would be crucial to re-evaluate the vaccines' economic impact as more local data on delivery costs become available through country-specific data collection efforts [74
(c) Herd immunity effects
A limitation of any static companion model used to assess vaccine programs is the omission of indirect herd immunity effects that may be realized after a large-scale introduction of either rotavirus or HPV vaccination program [37
]. Yet it is notable that the extent of herd immunity effects can vary between vaccines and across settings, depending on the scale of interventions and multiple other biological, epidemiological, and behavioral factors. A recent study in the United States has reported that a rotavirus vaccination program with 70% coverage would reduce rotavirus infection prevalence by an additional 15-25% due to herd immunity effects [57
], but another recent study in Kyrgyzstan has suggested that the contribution of herd immunity effects to the overall severe rotavirus disease burden reduction would be minimal (~1%). Other studies have reported the potential herd immunity effects due to HPV vaccination would be more prominent [32
]. Given that there have been no negative
indirect effects (e.g., as shown in the case of rubella vaccine due to a shift in the age of onset of the disease) reported from either of the vaccines, the explicit inclusion of any possible herd immunity effects would presumably lead to more favorable cost-effectiveness results for both rotavirus and HPV vaccines (though the extent may vary between the two vaccines and across the countries).
As others have discussed, a constant discount rate may underestimate the benefits of a health intervention in which benefits are realized in the far distant future when compared to an intervention where benefits occur relatively soon after implementation [4
]; our comparative model-based evaluation of rotavirus versus HPV vaccines illuminates this fact clearly. Our findings show that, although the total number of averted deaths following a single cohort would be very similar between the two vaccines (~274,000 for rotavirus versus ~286,000 for HPV vaccines) under the base-case assumptions, the numbers of undiscounted DALYs averted substantially differ (~15.77 million for rotavirus versus ~6.19 million for HPV vaccines) as a majority of the deaths averted due to HPV vaccination would occur at much older ages (after age 30). This discrepancy in DALYs averted was even greater when we assumed a 3% discount rate; DALYs averted were significantly reduced to ~7.15 million for rotavirus and ~1.30 million for HPV vaccines, leading to a greater than 5-fold difference. Applying an alternative discounting technique such as a slow discounting [64
] to the evaluation of HPV vaccines may lead to a more favorable cost-effectiveness profile (of an already favorable profile) of HPV vaccines. As we will discuss later, the choice of discount rate may also be related to some ethical concerns regarding whether and to what extent some lives should be considered to have more value. Accordingly, this study highlights the need for the use of alternative discount rates even in a secondary analysis, as some previous studies [11
(e) Population dynamics
In the standardized simulation for the GAVI countries, we used population data projected by the UN Population Prospects (The 2006 Revision) [38
]. Among alternative projections, we chose to use the data that are based on a medium fertility scenario, incorporate background mortality, and are interpolated at 1-year intervals by age and sex. The Prospects predicts various changes in population structure over the next few decades. Most relevantly, it assumes a rapidly increasing population growth rate among adolescents and adults due to decreased overall mortality paired with relatively slow growth in new births in most developing countries (with some exceptions) [38
]. For example, if we follow the trajectory of a cohort of 9-year-old girls during their lifespan, the sum of the numbers of women alive at each age between ages 40-79 years--the age band where a majority of cervical cancer deaths occur--would be ~951 million in total for the 72 GAVI countries while the corresponding figure from the snapshot in 2010 is ~372 million. In contrast, the corresponding figures for the population size of young children aged 0-4 years--the relevant age band for rotavirus vaccines--are ~371 million following the 2010 birth cohort and ~366 million using the snapshot in the same year. The remarkable increase of the age band 40-79 years explains in part the gap between the recently observed total number of cervical cancer deaths (~160,000; from a snapshot) and the model projected figure (~560,000; from a trajectory incorporating population dynamics) without HPV vaccination in the GAVI countries. This suggests that, for both vaccines, the magnitudes of the actual financial requirements and reduction in disease burden might be different from the values projected using the companion model if actual population dynamics are different from those projected by the version of population prospects that we used.
In addition to efficiency issues and uncertainties in disease burden and program costs, there are ethical as well as political considerations, that may be relevant in prioritizing new vaccine introduction in resource-poor settings. In terms of potential ethical considerations, in addition to distributional equity, our comparison of the age-distribution of the health outcomes between rotavirus and HPV vaccines suggests that, at a local level, a country's decision to place priority on one vaccine over the other under limited resources may cause public concerns of inter-generational equity. Further, at the global level, given the regional distribution of the health outcomes is different between the two vaccines (i.e., rotavirus vaccines would avert a majority of deaths in African settings while HPV vaccines would do so in South Asian settings), prioritization of one vaccine over the other may raise some inter-regional equity concerns. Finally, since rotavirus vaccination provides direct benefits to girls and boys, and right now HPV 16,18 vaccination would provide direct benefit to girls only (boys would only benefit indirectly), there could be concerns about gender equity that might be relevant to decision-making. One purposeful goal of an exercise such as this is to generate information that will stimulate a deliberative dialogue among both decision makers and stakeholders about just these very issues.
Political considerations will also be important to consider in priority setting [78
]. Although a full analysis and discussion of the criteria is beyond the scope of this article, it is crucial to recognize potential policy implications of taking into account different dimensions of criteria when making decisions on which vaccines to prioritize, not only for a comparison of rotavirus versus HPV vaccines as illustrated here, but for a more comprehensive comparison including other new vaccines such as pneumococcus. Recently there has been a renewed commitment to help local policy makers with priority setting and resource allocation, arguing for the need for multi-criteria decision analysis [12
]. Furthermore, when multiple new vaccines are assigned priority in a country, it would be also important to consider the implications of different timing or sequences of vaccine introduction. For example, a new vaccine that reduces early childhood mortality (e.g., rotavirus vaccine) may increase the impact of a subsequent vaccine targeting adolescents (e.g., HPV vaccine) by increasing the size of the target population for the latter vaccine. This suggests the necessity of developing a more flexible model that can capture population dynamics as well as transmission dynamics for multiple vaccines of interest.