The World Health Organization recommends the inclusion of HPV vaccination in national immunization programs for girls 9 to 13 years old if (1) prevention of cervical cancer is a public health priority, (2) the introduction is programmatically feasible, (3) sustained financing can be secured, and (4) the cost effectiveness is considered . However, without detailed analyses of costs based on observed data collected in countries, it is impossible to reliably estimate cost effectiveness or to assess whether sustained financing will be available for HPV vaccination. In sub-Saharan Africa, where more than 1 million cancer cases could be averted through HPV vaccination of 10 consecutive birth cohorts of girls , empirical information about the delivery costs of vaccines to young adolescent girls was unavailable until very recently [11, 29].
This is one of the first studies to report detailed resource-use-based cost estimates for delivering HPV vaccines to schoolgirls in sub-Saharan Africa. Using data collected alongside a cluster randomized trial, our project cost analysis demonstrated large differences in the costs of delivering HPV vaccines to girls between age-based and class-based delivery. In addition, modeled costs of a scaled-up vaccination program for Mwanza Region in Tanzania showed that three doses of HPV vaccine, when excluding costs of the vaccine itself, could be delivered to girls at an incremental economic cost of US$3.09 per dose and about US$10 per fully-immunized girl. These estimates are similar to the US$3.15 per dose that have been reported by PATH and the national EPI program for a school-based HPV vaccination program in Uganda , but significantly higher than estimated for the costs of delivering new childhood vaccines (also excluding vaccine costs), such as pneumococcal or rotavirus vaccines in Kenya (US$1.06) , Uganda (US$0.34) , or Malawi (US$0.35) .
Our study has a number of limitations. First, despite our high reliance on observed data, the analysis of project costs was based on estimates and extrapolations for certain inputs. Personnel time was allocated to different activities of the intervention on the basis of staff interviews, which are a known source of uncertainty , although prospective time sheets (from the third round of vaccination), which are a better source of information, were used to allocate costs between procurement and vaccination. Vehicle logbooks were used for the allocation of transport costs but detailed records for the allocation between rural and urban schools were only consistently available for vaccination round three. However, these allocations would not be expected to change over time since the same schools were visited in each round. Furthermore, project costs are not representative of how much it would cost to deliver HPV vaccines within a national immunization program: the project delivered HPV vaccines through a vertical system specifically set up for the intervention. Such systems imply high costs for administration/supervision, transport, and a dedicated cold chain. Procurement costs in the project were particularly high for rural schools as vehicles always had to travel from Mwanza City to the rural district. In the scaled-up program, the difference in procurement costs between urban and rural schools was estimated to be much smaller as vaccines would be delivered from the district EPI office (see Additional file 2, Figure S1). In addition, salary costs of international staff accounted for about 18% of project costs (Figure 1). If they were replaced by their local equivalents, this would result in a reduction in total project costs by almost 20%.
Second, our scaled-up cost estimates, which aimed to overcome the problems of the project cost analysis by adjusting the costs for administration/supervision, transport and cold chain, have other limitations related to the adopted ingredients approach for modeling of costs. National level costs (such as administrative and supervision costs) were excluded, although these cost categories can be important; and the ingredients approach may underestimate inefficiencies. For example, for cold storage costs, the incremental costs of HPV vaccines were calculated as the costs of the additional volume required for storing the vaccines. However, if the introduction of HPV vaccines required an increase in the number of refrigerators or cold rooms, which would not be used during non-HPV-vaccination months, the incremental costs would be much higher. For cold storage, such a scenario was tested in the sensitivity analysis, which revealed that increasing the number of refrigerators in Mwanza Region to what district cold chain coordinators deemed necessary for adding HPV vaccination would increase the costs per fully-immunized girl by about 6%, that is, from US$26.41 to US$27.90 (Figure 3).
Third, our model calculated scaled-up costs based on experiences gathered over the course of the research project and adjusted observed costs based on assumptions derived from interviews with EPI staff in Mwanza City and Misungwi District. The model assumed that similarly high coverage rates would be achieved in a regional program as in the research project; that distances travelled from district stores to health facilities would be similar for all rural districts as those in Misungwi District; and that wastage rates of vaccines and materials would be similar to those in existing EPI programs . However, the influence of alternative assumptions was tested in sensitivity analyses (Figure 3), which showed that except for the unlikely case of the full worst case scenario, estimated costs would be relatively robust.
Despite these caveats, our research has major implications for policymakers and researchers. Firstly, in our project cost analysis, we found that costs per fully-immunized schoolgirl of delivering three doses of HPV vaccines using a class-based delivery strategy were about one-third below those of using age-based delivery, although the costs per school reached were not higher for age-based delivery (see Additional file 3, Figure S2). The reason for the lower costs per fully-immunized schoolgirl of class-based delivery in the project was the substantially higher number of girls vaccinated at class-based schools due to more eligible girls being identified and higher vaccine coverage (Table 1 and Watson-Jones et al. ). In a scaled-up regional program, which is likely to target a different group of girls (class 4 instead of class 6), the number of eligible girls per school would be similar under age-based and class-based delivery (Table 1). However, targeting a different age group may have further cost and effectiveness implications, which need to be considered. About 99% of girls in class 4 are in the target group for vaccination (between 9 and 13 years of age)  and are less likely to ever have had sex before [13, 34] than girls in class 6, where almost 20% are 13 years and older. However, as the full duration of protection offered by HPV vaccines still remains uncertain  (even if recent evidence has confirmed vaccine efficacy of 8.4 years  and modeling results suggest 20 years protection ), a trade off might exist between targeting girls early in life (prior to their sexual debut) and ensuring that they are still protected later in life. Furthermore, earlier vaccination implies a requirement for earlier booster doses, which may increase costs.
Secondly, we estimated scaled-up costs of US$26 per girl fully immunized against HPV. Even when excluding vaccine costs (see Figure 3), US$10 is required for delivering vaccines to schoolgirls. It is essential that these costs be adequately budgeted during vaccine introduction, in particular because a previous study found that system costs (social mobilization/IEC, training, cold storage, transport) were not covered by GAVI Alliance introduction grants during the introduction of DTwP-hepatitis B-HiB vaccine in Ethiopia . Furthermore, we found that the costs of a school-based vaccination program, when excluding costs of the vaccine, are mostly driven by the costs of reaching schools, for social mobilization/IEC, procurement and vaccination. Consequently, the marginal costs per fully HPV-immunized girl and social mobilization costs could potentially be reduced if HPV vaccines were delivered together with other school-based health interventions, such as tetanus toxoid (TT) vaccination, vitamin A supplementation or deworming with antihelminthics [29, 39] and/or reproductive health interventions. In addition, if recent findings are confirmed that two-dose HPV vaccine regimens may be effective for cervical cancer prevention , costs per fully-immunized girl can be expected to drop by about 30%.
Thirdly, our results should be incorporated into future cost-effectiveness analyses of HPV vaccines in Eastern Africa. Our model of a scaled-up regional HPV vaccination program estimated that costs per fully-immunized girl would be about US$26 (see Figure 3), when including a vaccine price of US$5 per dose . Results of existing cost-effectiveness studies [11, 41] suggest that HPV vaccination at US$25 per fully-immunized girl would be very cost effective in all countries of Eastern Africa: at US$25 purchasing power parities (PPP) per vaccinated girl, the costs per life-year saved (Campos et al. ) or per disability adjusted life-year (DALY) averted (Kim et al. ) would be below the per capita income in these countries. Of course, if HPV vaccine prices drop below US$5 and vaccines are delivered together with other school-based health interventions (see above), cost effectiveness would further increase. Currently, the GAVI decision to support the introduction of HPV vaccines in developing countries  and the donation offer to the Tanzanian government of 2 million doses of HPV vaccine  present an unprecedented opportunity for the introduction of this life-saving intervention.