The Cost-Effectiveness of 13-Valent Pneumococcal Conjugate Vaccine in Seven Chinese Cities

Objective: This study estimates the cost-effectiveness of vaccination with the 13-valent pneumococcal conjugate vaccine (PCV13) among infants in Beijing, Shanghai, Shenzhen, Chengdu, Karamay, Qingdao, and Suzhou. Methods: A previously published cost-effectiveness model comparing vaccination with PCV13 to no vaccination was localized to the included Chinese cities. A systematic literature review was undertaken to identify age-specific incidence rates for pneumococcal bacteremia, pneumococcal meningitis, pneumonia, and otitis media (AOM). Age-specific direct medical costs of treating the included pneumococcal diseases were taken from the Chinese Health Insurance Association database. The base case analysis evaluated vaccine efficacy using direct effect and indirect effects (DE+ IDE). A subsequent scenario analysis evaluated the model outcomes if only DE was considered. A vaccination rate of 70% was used. The model reported outcomes over a one-year period after it was assumed the vaccine effects had reached a steady state (5–7 years after vaccine introduction) to include the direct and indirect effects of vaccination. Health outcomes were discounted at 5% during the steady-state period. Results: Vaccination with PCV13 was cost-effective in the base case analysis for all included cities with the incremental cost-effectiveness ratio (ICER) ranging from 1145 CNY(Shenzhen) to 15,422 CNY (Qingdao) per quality-adjusted life-year (QALY) gained. PCV13 was the dominant strategy in Shanghai with lower incremental costs and higher incremental QALYs. PCV13 remained cost-effective in the DE-only analysis with all ICERs falling below a cost-effectiveness threshold of three times GDP per capita in each city. Conclusions: Vaccination with PCV13 was a cost-effective strategy in the analyzed cities for both the DE-only and DE + IDE analyses. PCV13 became very cost-effective when a vaccination rate was reached where IDE is observed.


Introduction
Invasive pneumococcal diseases (IPD) caused by Streptococcus pneumoniae (S. pneumonia) include meningitis and bacteremia. S. pneumonia can also cause non-invasive pneumococcal infections such as pneumonia and acute otitis media (AOM). The World Health Organization (WHO) has estimated 1 million children die from pneumococcal disease each year, with the majority in developing countries [1]. Patients with pneumococcal diseases also experience decreased quality of life and incur significant direct and indirect costs due to medical healthcare resource utilization, absenteeism, and presenteeism [2].
The burden of pneumococcal diseases in China is significant, with an estimated 12% of global cases occurring in China [3]. In China lower respiratory infections including pneumonia were the 12th leading cause of overall deaths and premature deaths (years of life lost) in 2017 [4]. Pneumococcal diseases pose a significant burden in China where there are an estimated 30,000 deaths per year [3,5]. In 2000, it was estimated there were 22 million The model took a payer perspective for each city, and outcomes were calculated over a one-year time horizon after a steady-state period was reached, which was assumed to take 5-7 years, to include the direct and indirect effects of vaccination. Outcomes were recorded for each arm in the model (vaccination and no vaccination) and included the number of cases and deaths for each health state, the number of quality-adjusted life-years lost due to death, sequelae, and acute disease, vaccination costs, and treatment-related costs of the included health states. Life-years were estimated over a lifetime time horizon, by using the conditional expected remaining life-years according to each age group in the model. Lifetime health outcomes were discounted at a rate of 5%, and costs were expressed in CNY. Per the approved vaccination schedule of Prevenar 13 ® in China, children were to receive a 3 + 1 schedule, which included a total of 4 doses in the base case analysis. A separate analysis explored the impact of a 2 + 1 dosing schedule on the results of the base case analysis. The model took a payer perspective for each city, and outcomes were calculated over a one-year time horizon after a steady-state period was reached, which was assumed to take 5-7 years, to include the direct and indirect effects of vaccination. Outcomes were recorded for each arm in the model (vaccination and no vaccination) and included the number of cases and deaths for each health state, the number of quality-adjusted life-years lost due to death, sequelae, and acute disease, vaccination costs, and treatment-related costs of the included health states. Life-years were estimated over a lifetime time horizon, by using the conditional expected remaining life-years according to each age group in the model. Lifetime health outcomes were discounted at a rate of 5%, and costs were expressed in CNY. Per the approved vaccination schedule of Prevenar 13 ® in China, children were to receive a 3 + 1 schedule, which included a total of 4 doses in the base case analysis. A separate analysis explored the impact of a 2 + 1 dosing schedule on the results of the base case analysis.

Demographic Inputs
Key demographic data was included for each of the analyzed cities in China and included the total population, the percentage of patients under one year of age, and the age distribution [26].

Epidemiologic Inputs
The model considers age-based incidence rates for pneumococcal meningitis, pneumococcal bacteremia, pneumonia (inpatient and outpatient), otitis media (mild and moderate/severe), and sequelae (Table 1). Incidence rates were calculated based on rates retrieved from a systematic literature review and the results from an analysis of the Chinese Health Insurance and Research Association Database (CHIRA) (Supplementary Materials: Table  S1.). The CHIRA database includes data from a sample (2-10%) of Tier 1 municipalities, capital cities of each province (Tier 2), prefecture-level cities (Tier 3), and country-level cities (Tier 4). The methodology used to collect the data for CHIRA has been well described by Shen et al. [19]. To better understand the context of the incidence rates retrieved from the CHIRA database and compare them to others that have been used in China, a systematic literature review (SLR) was undertaken. The objective of the SLR was to identify economic evaluations of PCVs for mainland China between 1/1/2006 and 10/11/2016. Of the 1012 records identified there were seven studies of PCVs focusing on mainland China [17][18][19][20][27][28][29]33].
Each of the studies identified in the SLR was analyzed to extract the age-based incidence rates for bacteremia, meningitis, pneumonia, and otitis media, as well as the source(s) for the incidence rates. The rates and sources are shown in Supplementary Material Table S1.
Within each of the diseases, age-based incidence rates were calculated for each source by taking the average across all studies using the data source. Point estimates for each disease were calculated by then taking an average across all of the sources. This method ensures that sources not weighted by their frequency of use in economic evaluations.
The results of the SLR and subsequent data extraction did find that CHIRA may overestimate the incidence of inpatient pneumonia. The point estimates for meningitis, bacteremia, pneumonia, and AOM developed through an average of all available sources were utilized in the model ( Table 1).

Vaccine Effectiveness
The direct and indirect effects of vaccination are both considered in this model. The derivation of the effectiveness rates for PCV13 was previously described in Shen et al. [14,19,[34][35][36][37][38]. The direct effect of PCV13 refers to the reduction of disease in children that are vaccinated, while indirect effect refers to the herd effect or the reduction in disease for unvaccinated individuals. Herd effect within PCV13 has been well documented in the literature [14,37,38]. It has been estimated that the herd effect is observed with vaccination rates starting at 54% [14].
A scenario analysis examined a 2 + 1 vaccination schedule, which has been previously studied, and is suited for situations with high uptake in a national immunization program [39]; a 3 + 1 schedule is recommended in low uptake settings to maximize individual protection. A recent study comparing 3 + 1 and 2 + 1 vaccination schedules in China reported IgG geometric mean concentrations and OPA geometric mean titers trended numerically higher with 3 + 1 versus 2 + 1 dose schedules after the 3 priming doses versus 2 priming doses, respectively, in the first year of life [40]. However, no significant differences in immunogenicity were observed between the 3 + 1 versus 2 + 1 dose schedules after the toddler dose [40]. PCV13 was well-tolerated across all schedules [40]. Therefore, while direct protection in the first year of life may differ, we assumed 3 + 1 and 2 + 1 vaccination schedules had similar effectiveness when uptake in the infant population was greater than 70%.

Direct Costs
Direct treatment-related costs for pneumococcal meningitis, pneumococcal bacteremia, inpatient and outpatient pneumonia, and mild and moderate/severe otitis media were included in the model ( Table 2). Costs were sourced from the CHIRA database, which provided costs for Tier 1 cities and national costs. In cases where data was not available specifically for Tier 1 cities, the national cost data was used. Treatment-related costs were assumed to include all of the direct medical costs for the treatment of a case and were specific to each age group. A cost of 698 CNY was used for each dose of PCV13 for all included cities. Costs related to the service fee of the vaccination and transportation were only considered if they were direct costs. Service and transportation fees per dose, sourced from provincial governments or based on assumptions, were applied to Shanghai (5.5 CNY) (Assumption), Suzhou (20 CNY) [41], Shenzhen (25 CNY) [42], Beijing (25 CNY) [43], Chengdu (30 CNY) [44], and Qingdao (34 CNY) [45,46].

Quality of Life
Utility values were assigned to patients in the model to calculate the health-related quality of life. The general population of healthy patients was assumed to have a utility of 0.9. Disutility values through QALY decrements were applied to patients with bacteremia (0.008), meningitis (0.023), inpatient pneumonia (0.006), outpatient pneumonia (0.004), and acute otitis media (0.005) [47]. These QALY decrements were applied once for each observed case, which aimed to represent the short-term impact of the disease.
Patients with long-term sequelae were assigned separate utilities, which were applied to life-years to calculate QALYs. Neurological impairment and hearing loss due to pneumococcal meningitis carried utilities of 0.60 and 0.80, respectively [48].

Analysis
The base case analysis considered both DE and IDE and compared the costs and health outcomes of vaccination with PCV13 with no vaccination. For each arm of the model, the total number of QALYs and costs are reported and the incremental values for all outcomes between the PCV13 vaccination and no vaccination arms are calculated. The incremental cost-effectiveness ratio (ICER) is calculated by dividing the incremental total costs by incremental total QALYs to determine the cost per additional QALY for vaccination compared to no vaccination. China has not published an explicit cost-effectiveness threshold for the willingness to pay for an additional QALY. The China Guidelines for Pharmacoeconomic Evaluations, published in 2020, endorse the use of GDP per capita to calculate the willingness to payer threshold [49]. Interventions with ICERs less than GDP per capita income are considered "very cost-effective" and ICERs between 1-3x GDP per capita are considered "cost-effective".
A one-way sensitivity analysis was performed by using the known confidence intervals of parameters or varying the parameters by +/− 10%. The minimum and maximum values of the disease incidence rates were used as the low/high values. Furthermore, an analysis of the incidence rates was undertaken to address the uncertainty within the point estimates of the incidence rates. For all included disease states the minimum and maximum values for each point estimate found either in the CHIRA database or the literature were recorded. A "minimum" and "maximum" scenario analysis were undertaken where all of the minimum values were used simultaneously in the minimum scenario and all of the maximum values were used in the maximum scenario.
Lastly, a scenario analysis evaluated the outcomes of the model when only DE was considered.   Vaccination with PCV13 dominated the no vaccination strategy in Shanghai with lower incremental costs and higher incremental QALYs. In the remaining cities, ICERs ranged from 1145 CNY in Shenzhen to 15,422 CNY in Qingdao. All ICERs were below 1× GDP per capita in each city.

Base Case Results
The base case scenario was further analyzed by the total number of vaccination doses, a 2 + 1 (three total doses) vaccination schedule was evaluated along with the base case 3 + 1 (four total doses) ( Table 4). Vaccination with PCV13 remained cost-effective across both vaccination schedules. One-way sensitivity analysis was performed on the base case (DE + IDE) analysis for each city (Figure 2). The top ten most influential parameters according to the spread between the high and low results, for each comparison and city were recorded. The incidence of inpatient pneumonia in those aged 18 to 34, 2 to 4, and 0 to less than 2 were the three most influential parameters. The total direct vaccine cost and the discount rate, which was varied between 3% and 7%, were the fourth and fifth most influential parameters, respectively. One-way sensitivity analysis was performed on the base case (DE + IDE) analysis for each city (Figure 2). The top ten most influential parameters according to the spread between the high and low results, for each comparison and city were recorded. The incidence of inpatient pneumonia in those aged 18 to 34, 2 to 4, and 0 to less than 2 were the three most influential parameters. The total direct vaccine cost and the discount rate, which was varied between 3% and 7%, were the fourth and fifth most influential parameters, respectively.

Incidence Rate Analysis
The results for the scenario analysis comparing the minimum and maximum pneumococcal disease incidence rates for IPD, pneumonia, and otitis media are shown in Figure 3. Vaccination with PCV13 dominated the no vaccination strategy when the maximum incidence rate values were used and increased to between 47,890 CNY (Shanghai) and

Incidence Rate Analysis
The results for the scenario analysis comparing the minimum and maximum pneumococcal disease incidence rates for IPD, pneumonia, and otitis media are shown in Figure 3. Vaccination with PCV13 dominated the no vaccination strategy when the maximum incidence rate values were used and increased to between 47,890 CNY (Shanghai) and 98,971 CNY (Chengdu) when all the minimum incidence rate values were used, both remaining below the 1× GDP per capita threshold.

DE-Only Scenario
The results of the DE-only scenario analysis are shown in Table 5. The total number of pneumococcal disease cases (thousands) avoided with PCV13 vaccination was 26.03, 23.85, and 11.73 in the cities of Beijing, Shanghai, and Shenzhen, respectively. In the other analyzed cities, the number of pneumococcal disease cases (thousands) avoided was 0.64 in Karamay, 11.01 in Suzhou, 12  98,971 CNY (Chengdu) when all the minimum incidence rate values were used, both remaining below the 1× GDP per capita threshold.

DE-Only Scenario
The results of the DE-only scenario analysis are shown in Table 5. The total number of pneumococcal disease cases (thousands) avoided with PCV13 vaccination was 26.03, 23.85, and 11.73 in the cities of Beijing, Shanghai, and Shenzhen, respectively. In the other analyzed cities, the number of pneumococcal disease cases (thousands) avoided was 0.64 in Karamay, 11.01 in Suzhou, 12

Discussion
Vaccination with PCV13 at a rate of 70% was cost-effective in Beijing, Shanghai, Shenzhen, Suzhou, Chengdu, Karamay, and Qingdao. PCV13 remained cost-effective when a 2 + 1 dosing schedule was considered as well as in the separate sensitivity analysis of the disease incidence rates, when low incidence rate values were utilized, and became the dominant strategy with high incidence rate values. In the DE-only scenario, PCV13 was again cost-effective with all ICER falling below three times GDP per capita for each city. As vaccination with PCV13 was cost-effective in the DE-only and IDE + DE scenarios, it can be assumed to be cost-effective across all rates of vaccination, regardless of the presence of herd effect. While we assumed that a 3 + 1 and 2 + 1 schedule provide equivalent protection, it should be noted that a 3 + 1 schedule remains the preferable schedule where uptake of PCVs remains low [39,50]. Furthermore, while both schedules elicited equivalent immune responses following the booster dose, and we assumed similar protection, it has been hypothesized and modeled in the US that using two priming doses could result in incremental disease cases due to reduced protection in the first year of life, specifically when uptake is low and herd protection cannot be achieved [51].
The results of this study show that vaccination with PCV13 across all cities would avoid 2270 cases of IPD, 167,899 cases of pneumonia, and 23,407 cases of AOM in the base case analysis considering direct and indirect effects of vaccination. In the base case scenario, there were 3853 deaths avoided and an increase of 56,831 QALYs and 287,334,913 CNY in incremental total (vaccination and direct medical) costs.
All four studies evaluating PCV13 vs. no vaccination found PCV13 to be cost-effective. The ICERs for PCV13 vs. Incidence rates were highly influential parameters on the results of this study, and on the results of the cost-effectiveness studies in the literature. Studies using CHIRA data had significantly higher rates of inpatient pneumonia than those using HIRD data, local surveillance data, or other sources. Rather than focus on a single data source this study took an average of the point estimates for each source. Additionally, the low and high values observed in the literature for each point estimate were tested in the incidence rate analysis. While disease incidence did significantly influence the ICER for each city, PCV13 remained cost-effective when the low incidence rates were tested, and became the dominant strategy when the high incidence rates were tested.
The direct costs related to the vaccine and age distributions were the only parameters that varied from city to city, as all other parameters, including disease incidence rates, were consistent. As expected, lower direct vaccine and vaccine-related costs resulted in lower ICERs. Cities with younger populations tended to have lower ICERs as the vaccine's effectiveness was higher in these younger populations. However, age distribution was only influential on the results in the base case scenario, which considered the indirect effects of the vaccine.
This study has several limitations. First, serotype coverage was assumed to be the same across all included cities. Large countries such as China can have considerable epidemiological differences between cities, especially those such as Karamay, with a smaller and demographically unique population. The model also did not consider serotype replacement in the long run. Second, the estimates of disease incidence vary considerably depending on the source for all included health states, especially inpatient pneumonia. These incidence rates are highly influential on the outcomes of the model as made evident by the one-way sensitivity analysis. Third, treatment costs were assumed across all tier 1 and non-tier 1 cities in the analysis. Costs are likely to vary from city to city within tier 1 and non-tier 1 cities. Lastly, the model may underestimate the benefit of vaccination with PCV as the number of patients carrying vaccine-serotype Streptococcus pneumoniae [52] and the prevalence of antimicrobial resistance [53] may decrease.

Conclusions
Vaccination with PCV13 is a cost-effective strategy for both the IDE + DE and DE-only scenarios in Beijing, Shanghai, Shenzhen, Suzhou, Chengdu, Karamay, and Qingdao. At vaccination rate levels where IDE are observed, PCV13 becomes very cost-effective with ICERs falling below GDP per capita in the respective cities. PCV13 remained cost-effective regardless of the dosing schedule and through an extended analysis of disease incidence. We estimate that vaccination with PCV13 will prevent a considerable number of cases of IPD, pneumonia, and AOM, as well as avoid mortality associated with IPD and pneumonia for both the vaccinated and unvaccinated population through the direct and indirect effect of the vaccine. This analysis can be used by policymakers when deciding on the allocation of resources in the studied cities.