The case for a universal hepatitis C vaccine to achieve hepatitis C elimination

Background The introduction of highly effective direct-acting antiviral (DAA) therapy for hepatitis C has led to calls to eliminate it as a public health threat through treatment-as-prevention. Recent studies suggest it is possible to develop a vaccine to prevent hepatitis C. Using a mathematical model, we examined the potential impact of a hepatitis C vaccine on the feasibility and cost of achieving the global WHO elimination target of an 80% reduction in incidence by 2030 in the era of DAA treatment. Methods The model was calibrated to 167 countries and included two population groups (people who inject drugs (PWID) and the general community), features of the care cascade, and the coverage of health systems to deliver services. Projections were made for 2018–2030. Results The optimal incidence reduction strategy was to implement test and treat programmes among PWID, and in settings with high levels of community transmission undertake screening and treatment of the general population. With a vaccine available, the optimal strategy was to include vaccination within test and treat programmes, in addition to vaccinating adolescents in settings with high levels of community transmission. Of the 167 countries modelled, between 0 and 48 could achieve an 80% reduction in incidence without a vaccine. This increased to 15–113 countries if a 75% efficacious vaccine with a 10-year duration of protection were available. If a vaccination course cost US$200, vaccine use reduced the cost of elimination for 66 countries (40%) by an aggregate of US$7.4 (US$6.6–8.2) billion. For a US$50 per course vaccine, this increased to a US$9.8 (US$8.7–10.8) billion cost reduction across 78 countries (47%). Conclusions These findings strongly support the case for hepatitis C vaccine development as an urgent public health need, to ensure hepatitis C elimination is achievable and at substantially reduced costs for a majority of countries. Electronic supplementary material The online version of this article (10.1186/s12916-019-1411-9) contains supplementary material, which is available to authorized users.

: hepatitis C prevalence among people who inject drugs used for model input. Based on Degenhardt et al. 1 For countries without estimates, population-weighted averages were calculated for each WHO region and applied. Grey countries are missing data. Figure S3: hepatitis C prevalence among the general population used for model input. Based on Blach et al. 2 and Gower et al. 3 For countries without estimates, population-weighted averages were calculated for each WHO region and applied. Figure S4: Epidemic type classification used for model input. Epidemics were classified as mixed, meaning that infection was possible among the general population as well as among PWID, if the country was not in a WHO high income classification AND the total number of people living with hepatitis C was >5 times the total number of estimated hepatitis C-infected PWID. This was done as without transmission among the general population, the model was unable to produce the correct number of people living with hepatitis C based on injecting drug use-related transmission alone.  3 For countries without estimates, population-weighted averages were calculated for each WHO region and applied. f Average salary calculated as the per capita gross domestic product (GDP); from World Bank data 8 . Costs assume two hours of provider time for interaction and any laboratory work, and that providers work 7 hours per day, 5 days per week and 45 weeks per year. ^^ In 2018 these countries were identified as being among 12 that were on track to achieve elimination by 2030 4 . Therefore updated prevalence estimates were generated for PWID and the general community by subtracting the 2016 and 2017 treatment numbers 4 from the estimated number of people with hepatitis C (assuming PWID were equally likely to be treated as members of the general community).

Selecting the optimal strategy
The optimal WHO incidence reduction target strategies for each country with and without a vaccine available were chosen as follows, with five examples illustrated in Figures S5-S9.
First, all combinations of interventions without a vaccine were run and considered in the costeffectiveness plane of total cost versus cumulative cases averted, 2018-2030. Specifically, test/treat programs for PWID [testing two-yearly, annually or six-monthly] with and without testing of the general community. Scenarios were considered dominated if they cost more but prevented fewer cumulative cases than another scenario, and dominated scenario were excluded from further consideration. Among the non-dominated scenario, the optimal WHO target strategy was considered to be the one that achieved an 80% reduction in incidence by 2030 in the most costeffective way (as measured by cost per incident case averted). Where no scenario could achieve this level of incidence reduction, the non-dominated scenario with the greatest 2030 incidence reduction was selected.
This process was then repeated with a broader set of interventions that included the availability of a vaccine. Specifically, all combinations of test/treat programs for PWID [testing two-yearly, annually or six-monthly] with and without vaccination included; testing of the general community with and without vaccination included; and an adolescent vaccination program. Again the optimal WHO target strategy was considered to be the non-dominated scenario that was the most cost effective for achieving an 80% reduction in incidence by 2030, or where this was not possible the nondominated scenario that achieved the greatest 2030 incidence reduction.
As Australia, Brazil and USA were classified as having concentrated epidemics, the general community interventions are suppressed from those figures below. For Chine and Egypt, a log10 scale is used on the cost axis. In China, without a vaccine we see that the general community testing (squares) were only slightly more effective but much more expensive than PWID only testing (circles), and dominated by PWID only testing + vaccination (diamonds).

Figure S5
: Selecting the optimal strategy for Australia. Without a vaccine (only circles considered), 6-monthly testing of PWID (yellow circle; 78% incidence reduction by 2030) was selected as the nondominated optimal WHO target strategy. With a vaccine available, the most cost-effective way to reach an 80% reduction in incidence was a two-yearly test/treat/vaccinate program for PWID (blue diamond). Vaccine was assumed to have 75% efficacy, 10-year duration of protection and cost US$200/course. Line connects non-dominated strategies Text = % reduction in annual incidence by 2030 Figure S6: Selecting the optimal strategy for Brazil. With or without a vaccine, annual testing of PWID was the most cost-effective and non-dominated way to achieve an 80% reduction in incidence by 2030 (orange circle). Vaccine was assumed to have 75% efficacy, 10-year duration of protection and cost US$200/course. Figure S7: Selecting the optimal strategy for China. Without a vaccine (only circles and squares considered), 6-monthly testing of PWID + general population screening (yellow square; 74% reduction in incidence by 2030) was selected as the optimal WHO target strategy. With a vaccine available, the most cost-effective way to reach an 80% reduction in incidence was a two-yearly test/treat/vaccinate program for PWID (blue diamond). Vaccine was assumed to have 75% efficacy, 10-year duration of protection and cost US$200/course.  Line connects non-dominated strategies Text = % reduction in annual incidence by 2030 Figure S8: Selecting the optimal strategy for Egypt. Without a vaccine (only circles and squares considered), 6-monthly testing of PWID + general population screening (yellow square; 48% reduction in incidence by 2030) was selected as the optimal WHO target strategy. With a vaccine available, a 64% reduction in incidence by 2030 was the greatest non-dominated strategy, using twoyearly test/treat/vaccinate program for PWID and screening of the general population (without vaccination) (blue star). Vaccine was assumed to have 75% efficacy, 10-year duration of protection and cost US$200/course. Figure S9: Selecting the optimal strategy for USA. Without a vaccine (only circles considered), 6monthly testing of PWID (yellow circle; 70% incidence reduction by 2030) was selected as the optimal WHO target strategy. With a vaccine available, the most cost-effective way to reach an 80% reduction in incidence was a two-yearly test/treat/vaccinate program for PWID (blue diamond). Vaccine was assumed to have 75% efficacy, 10

Impact of a vaccine
Assuming 80% health care coverage, the availability of a 75% efficacious vaccine as a possible intervention for an elimination strategy led to an additional 16%, 13%, 20% and 25% of cases averted between 2018 and 2030 in Australia, China, Egypt and the USA, respectively, but no difference for Brazil as it did not form part of the optimal WHO target strategy in that setting (see Figure S6). With a 75% efficacious vaccine, testing requirements among PWID became more feasible, with testing requirements among PWID reduced from 6-monthly to two-yearly in Australia, China and the USA.
Availability of a vaccine could lead to different optimal strategies for reaching the WHO target. For China, the optimal strategy without a vaccine included screening of the general community, whereas the additional impact of a vaccine (even if only 50% efficacious) meant that it became possible to reduce incidence by 80% through targeted programs for PWID alone, which was considerably cheaper ( Figure S10).
A US$200 per course 75% efficacious vaccine reduced the total costs of the optimal incidence reduction programs in Australia, China and the USA by US$66 million (58%), US$4.0 billon (80%) and US$2.0 billion (62%), respectively. For Egypt, unless the vaccine was under US$2.46 per course, the optimal strategy with a vaccine resulted in higher total costs, due to the large-scale vaccine delivery required through the age-based vaccination program and the general population screening program. Top panel: cumulative cases averted 2018-2030 using the optimal strategy without a vaccine (blue) and with a 50%, 75% or 90% efficacious vaccine (red, yellow and purple, respectively). Bottom panel: total discounted costs of the optimal incidence reduction strategy without a vaccine (left bars) and with a 50%, 75% or 90% efficacious vaccine (right bars). Lightest grey shading represents total vaccination costs at US$50 per vaccine course, with darker shadings representing total costs as the vaccine price per course increases in US$50 increments. Uncertainty bounds represent scenarios with 70% and 90% coverage of testing, treatment and vaccination compared to a base of 80%.

Settings considered
Independent models were run for settings with concentrated epidemics (25%, 50% or 75% hepatitis C prevalence among PWID), generalised epidemics (1%, 2%, 3%, 5%, 10%, 15%, 25% or 30% hepatitis C prevalence among the general community) and mixed epidemics (all combinations of hepatitis C prevalence among PWID and the general community). Parameters used for these settings are shown in Table S3. Global average, based on two hours of provider time for interaction and any laboratory work. Average salary calculated as the population-weighted per capita gross domestic product (GDP) 8 . Assumes providers work 7 hours per day, 5 days per week and 45 weeks per year.

Results
Availability of a 75% efficacious vaccine had the potential to substantially improve the impact of incidence reduction strategies, with the greatest benefits in settings with higher initial prevalence ( Figure S11).
If a 75% efficacious vaccine cost US$200 per course, the vaccine strategies dominated the nonvaccine strategies in medium and high prevalence concentrated epidemic settings (i.e. cost less and prevented more incident cases) and were included within the optimal test/treat strategies for PWID.
In these settings, the optimal incidence reduction strategies with a vaccine available required 32% and 52% fewer treatments than the optimal strategies without the vaccine, respectively (Table S5).
In low, medium and high prevalence concentrated epidemic settings, the vaccine strategies dominated the non-vaccine strategies provided a vaccine course cost under US$77, US$263 or US$236 per course, respectively ( Figure S11).
In generalised epidemic settings, the optimal incidence reduction strategy without a vaccine was screening and treatment of the general community, and the optimal strategy with a vaccine was to deliver vaccination with the general community screening program as well as through an age-based program ( Table S4). The incidence reduction target was not able to be reached in generalised epidemic settings (unless testing levels among the general community were unfeasibly high, such as having annual testing of the entire population rather than having the entire population screened by 2030). However, use of a vaccine was able to approximately double the achievable incidence reduction in these settings.
For generalised epidemic settings with initial prevalence in the general community under 3%, 10%, 20% or 30%, unless the vaccine was under US$2.00, US$2.91, US$4.31, US$5.86 per course, respectively, use of a vaccine resulted in higher total costs. This is because the large-scale vaccine delivery was not entirely offset by the reduced treatment requirements.
In mixed epidemic settings, the optimal strategies included targeted programs for PWID, as well as programs among the general community, as per the concentrated and generalised epidemic settings. Figure S11: Potential impact of a 75% efficacious hepatitis C vaccine on incidence reduction strategies in concentrated and generalised epidemic settings. Left panels: incidence reduction in 2030 under the optimal incidence reduction strategies without a vaccine (blue bars) and with a 75% efficacious vaccine (red bars). Right panels: total discounted costs of the optimal incidence reduction strategies without a vaccine (left bars) and with a 75% efficacious vaccine available (right bars). Lightest grey shading represents total vaccination costs at US$50 per vaccine course, with darker shadings representing total costs as the vaccine price per course increases in US$50 increments. Uncertainty bounds represent scenarios with 70% and 90% coverage of testing, treatment and vaccination compared to a base of 80%. *For concentrated epidemic settings, results are per 1000 PWID. For generalised and mixed epidemic settings, results are per 1000 population. Mixed epidemic settings assume 0.24% of the population injects drugs (see Table S3). ^Uncertainty bounds represent scenarios with 70% and 90% population coverage of testing, treatment and vaccination compared to a base of 80%.

APPENDIX D: SENSITIVITY ANALYSIS
When the vaccine efficacy was halved (37.5%) for individuals following successful treatment, the price per course had to be 2-20% cheaper in order to reduce the cost of elimination, but the optimal strategies were unchanged. This assumption had a much smaller impact on outcomes than the overall efficacy of the vaccine because the majority of vaccinations were delivered to uninfected people. For example, if the vaccine were 90% efficacious rather than 75%, the price point for the vaccine to reduce the costs of elimination was relaxed by 12-45%, whereas if the vaccine were only 50% efficacious, the price per course had to be 20-45% cheaper in order to reduce the cost of elimination (Table S6).
If no staff costs were included, meaning that the human resources associated with testing and treating to achieve eliminating were accounted for elsewhere, then the cost of a vaccine would have to be 20-46% cheaper in concentrated epidemic settings or 39-85% cheaper in generalised epidemic settings to reduce overall costs.
If the vaccine duration of protection was 5 years or 100 years (i.e. lifetime) compared to 10 years, then the price point was lowered by 13-26% (meaning the vaccine needed to be cheaper to save costs) or increased by 20-41% across the different prevalence settings, respectively.
Harm reduction scale-up only altered the optimal strategy in settings with low (<25%) prevalence among PWID: with a 60% scale-up of harm reduction, two-yearly testing rather than annual testing of PWID was sufficient to achieve the incidence reduction target in these settings. This meant that a vaccine did not reduce costs by as much and so had to be cheaper per course to reduce the cost of elimination. Scaling up harm reduction by 20%, 40% or 60% in non-vaccine scenarios meant that in settings with approximately 50% prevalence among PWID, the achievable incidence reduction increased from 73% to 75%, 76%, and 77% respectively, and in settings with approximately 75% prevalence among PWID the achievable incidence reduction increased from 21% to 26%, 30% and 35%.
Variations in the time between antibody test and RNA test, retention in care following a positive antibody test, and the time between RNA tests and treatment commencement had minimal impact of outcomes (Table S6).