Cost-effectiveness of dengue vaccination in Puerto Rico

An effective and widely used vaccine could reduce the burden of dengue virus (DENV) around the world. DENV is endemic in Puerto Rico, where the dengue vaccine CYD-TDV is currently under consideration as a control measure. CYD-TDV has demonstrated efficacy in clinical trials in vaccinees who had prior dengue virus infection. However, in vaccinees who had no prior dengue virus infection, the vaccine had a modestly elevated risk of hospitalization and severe disease. The WHO therefore recommended a strategy of pre-vaccination screening and vaccination of seropositive persons. To estimate the cost-effectiveness and benefits of this intervention (i.e., screening and vaccination of seropositive persons) in Puerto Rico, we simulated 10 years of the intervention in 9-year-olds using an agent-based model. Across the entire population, we found that 5.5% (4.6%-6.3%) of dengue hospitalizations could be averted. However, we also found that 0.057 (0.045–0.073) additional hospitalizations could occur for every 1,000 people in Puerto Rico due to DENV-naïve children who were vaccinated following a false-positive test results for prior exposure. The ratio of the averted hospitalizations among all vaccinees to additional hospitalizations among DENV-naïve vaccinees was estimated to be 19 (13–24). At a base case cost of vaccination of 382 USD, we found an incremental cost-effectiveness ratio of 122,000 USD per QALY gained. Our estimates can provide information for considerations to introduce the CYD-TDV vaccine in Puerto Rico.

affect the number of false positive test results and because vaccinating a child with a false positive screeening test result can put the child at risk for more severe dengue illness if infected. It is important therefore to estimate the potential health and economic benefits and risks of using CYD-TDV when paired with a pre-vaccination screening strategy.
Several studies have suggested that pre-vaccination screening with CYD-TDV could be cost-effective in some settings [11][12][13][14], but high specificity of screening is required to minimize individual risk, as well as high sensitivity to maximize population benefits [15]. Pre-vaccination screening is also recommended in Latin America [16]. The health benefits and costeffectiveness of pre-vaccination screening for the Philippines and Brazil have been investigated in a recent study [12]. That analysis showed likely reduced disease incidence with limited adverse events at acceptable costs from this intervention with pre-vaccination screening and subsequent vaccination with CYD-TDV in areas with moderate to high-transmission intensity. Dengue is also endemic to several tropical and sub-tropical jurisdictions associated with the United States, including Puerto Rico [17]. A program of pre-vaccination screening coupled with CYD-TDV given to seropositive individuals is currently being considered to reduce the burden of dengue in Puerto Rico [18]. In this study, we evaluate the epidemiological and economic impacts of prevaccination screening and subsequent vaccination with CYD-TDV in Puerto Rico.

METHODS
We estimated the epidemiological benefits and cost-effectiveness of the intervention over a 10-year time-frame. We based this study on a previous analysis of the impact of this intervention in various transmission settings [12,19], and we adjusted transmission parameters and costs to represent Puerto Rico.
for use under a CC0 license. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Agent-Based Model
To evaluate the impact of a pre-vaccination screening intervention in Puerto Rico, we modified an agent-based model of dengue transmission with humans and mosquitoes represented as agents. This model has been described in previous publications [12,19,20]. While our model has been calibrated to demographic and geographic data from Iquitos, Peru [20], the model can represent DENV transmission in a generic setting [19] and has been modified in this study to simulate DENV transmission dynamics in Puerto Rico. Our model compares two strategies, an intervention strategy and the status quo. The intervention strategy is the routine pre-vaccination screening and subsequent vaccination of seropositive of 9-year-olds in Puerto Rico. Nine-year-olds is the lowest age approved for vaccination, and has been used as a default age of vaccination in other studies [12,19]. For each of the strategies, the model population was followed for 10 years, keeping track of dengue-related health events defined as dengue infections, hospitalizations, and deaths.
Model parameterization for Puerto Rico ( 9 ) We modified the transmission parameters of the model to approximate DENV transmission dynamics of DENV in Puerto Rico. We adjusted the transmission intensity of DENV in our model to achieve the expected age-specific dengue antibody prevalence levels ( 9 ) in Puerto Rico. We estimated the 9 for Puerto Rico based on two serological studies. Coudeville et al. [21] estimated 50% seroprevalence in 9-year-olds in areas where the CYD-TDV phase-3 trials were conducted. Argüello et al. found that 49.8% (95% CI = 43.6-56.0%) of participants between 10-18 years of age had a positive IgG anti-DENV antibodies [22]. We assumed a baseline of 9 = 0.5. Given the uncertainty on these estimates, we explored a lower value of 9 = 0.3, and an upper value of 9 = 0.6 to assess the sensitivity of our estimates to changes in the intensity of transmission in Puerto Rico.
for use under a CC0 license. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted October 9, 2020. . https://doi.org/10.1101/2020.10.07.20208512 doi: medRxiv preprint Individuals can be infected with DENV many times over the course of their life. Evidence suggests the second infection can cause a disproportionate level of disease severity, relative to the first infection or to the third or later infections. In the model, the effect of a vaccination on an individual is similar to a natural infection without symptoms (silent infection), which is consistent with assumptions in other models [13,14,19,23]. If the individual is DENV-naïve, then a vaccination serves as the individual's first (a silent) infection, and a subsequent infection may be more severe. By contrast, if the individual is DENV-exposed, then a vaccination serves as the individual's second or later (silent) infection, which does not result in additional risk for severe disease. These assumptions are supported by empirical evidence that seronegative individuals vaccinated with CYD-TDV have a higher risk of hospitalization after a natural infection, compared to those not vaccinated [8]. Vaccine effectiveness parameters were calibrated to the most recent dengue vaccine trial results [8], we calibrated model parameters characterizing vaccine profile to vaccine trial data using a particle filtering approach, which is explained in more detail elsewhere ( [12], Appendix S2). The results from the calibration step are shown in table 1 with the upper and lower bounds of the 95% confidence interval.
Given the stochastic nature of our model, we simulated 3,000 paired replicates over the parameter ranges, and reported the smoothed output of these simulations using a generalized additive model (GAM) in R. The uncertainty on the model parameters estimated in Table 1 was taken into account by simulating the model over the upper and lower bound of the estimates of the vaccine profile parameters.

Epidemiological outcomes
Population-level benefits were defined as the proportion of symptomatic and hospitalized cases averted for the total population. Individual-level benefits were defined as the relative risk for symptomatic and hospitalized dengue of an individual child who undergoes screening and possibly vaccination for use under a CC0 license. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted October 9, 2020. . https://doi.org/10.1101/2020.10.07.20208512 doi: medRxiv preprint compared to a child who is not given a screening or any vaccination. Since misclassification by a serologic test could result in DENV-naïve individuals receiving a vaccination that could lead to an episode of hospitalization [8], In our simulations, we assumed that coverage of the intervention (i.e., serological screening and vaccination in the event of a positive result) was given to 80% of the target population, but evaluated an alternative scenario of lower coverage (50%) in the sensitivity analysis. We assumed that 100% of children with a positive screening result were vaccinated. The baseline values of sensitivity (0.8) and specificity (0.95) were based on a recent review of rapid diagnostic tests for determination of serostatus [24].
However, the actual properties of serological screening may turn out to be different from this baseline. Assuming that increasing sensitivity would result in lower specificity, and vice-versa, We assumed three additional scenarios: 1) high sensitivity (0.95) and low specificity (0.76), 2) low sensitivity (0.64) and high specificity (0.95), 3) high sensitivity (0.95) and high specificity (0.95). In a set of sensitivity analyses, we also simulated pre-vaccination screening over the full range of values of sensitivity and specificity (0-1) to find the minimum values required to achieve positive proportion of dengue cases averted.
for use under a CC0 license.
This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Cost-effectiveness analysis
We evaluated the incremental cost-effectiveness ratio (ICER) of the intervention over a time horizon of 10 years. We used a public health perspective, as it has been used in previous economic analyses of the potential impact of routine vaccination with CYD-TDV [12,19]. We calculated the ICER as shown in equation 1, which compares an intervention strategy to a no-intervention strategy. In the denominator, three different outcome measures are used to quantify the effectiveness of each strategy ( v e nev v and v − v e nev v ): QALYs, symptomatic cases with medical care, and hospitalizations.
To evaluate the ICER, we identified a set of cost parameters. The for use under a CC0 license.
This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted October 9, 2020. . https://doi.org/10.1101/2020.10.07.20208512 doi: medRxiv preprint For the cost-effectiveness results, our primary health outcome was the quality-adjusted life-years (QALYs), which is a health-related quality of life index that ranges from 0.0 to 1.0 with 0.0 representing a state of death and 1.0 representing perfect health for one year. Following approaches used in previous studies of dengue [19,[26][27][28], we estimated the QALYs gained from the intervention based on lost quality of life due to dengue fever and severe dengue. We used disutility parameters (D), or QALY decrements, that were taken from a study that estimated the quality of life associated with dengue fever and hospitalizations, conducted by Zeng et al. [29]. These values are listed in Table 2. Assuming a discount rate (r) of 3%, we estimated the QALYs, where i is each individual in the model, j is the disease state (dengue fever, hospitalized dengue, or death), N is the total population in the model, and T is the number of years of the intervention (10 years).
To account for utility loss associated with premature death caused by This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted October 9, 2020. . https://doi.org/10.1101/2020.10.07.20208512 doi: medRxiv preprint where the terms not already defined in the discussion of equation 2 are defined as follows: Q is the future value of a QALY for all individuals each year who do not die from dengue, assumed to be 1.0 in all years beyond the simulation time horizon, L is the expected lifespan of an individual in the model, and is the age of individual i at the end of the model horizon.
for use under a CC0 license.
This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted October 9, 2020. . https://doi.org/10.1101/2020.10.07.20208512 doi: medRxiv preprint A reduction of symptomatic cases was achieved under any value of specificity and sensitivity for the pre-vaccination screening laboratory test ( Fig. 2). With regard to hospitalizations, our simulations over the whole range of sensitivity and specificity values suggest that even with perfect sensitivity, the minimum value of specificity to avert hospitalized cases is 0.6 for 9 of 30% (Fig. 2, right panel). Holding the baseline level of specificity (0.95) constant, we found that sensitivity values above 0.2 result in positive cases averted. These specificity values (0.2 9 = 50%, 0.6 9 = 30%) represent the absolute minimum test parameters needed in the model to avoid an increase in hospitalizations.

Individual-level benefits
The relative risk of becoming a symptomatic case was slightly reduced among vaccinated seropositive individuals, compared to individuals of the same age who were not given the intervention (Fig. 3, left panel). In the baseline scenario of sensitivity and specificity of laboratory test screening, the risk was around 0.85 (0.82 -1.0) for 9 = 50% and around 0.9 (0.87 -1.0) for 9 = 30%. The relative risk of a hospitalization was also reduced to around 0.63 (0.3 -1.0) for 9 = 50% and to 0.73 (0.73 -1.0) for 9 = 30% (Fig. 3, right panel).

Magnitude of naïve children at increased risk of hospitalization due to vaccination
We estimated the number of hospitalizations due to false positives of DENVnaïve children who were therefore vaccinated. We estimated these hospitalizations to be around 1. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted October 9, 2020. . https://doi.org/10.1101/2020.10.07.20208512 doi: medRxiv preprint half the baseline number. We also found that for every one additional hospitalization in the group of vaccinated DENV-naïve individuals there were about 19 (13 -24) ( 9 = 50%) hospitalizations averted overall in the baseline scenario (Fig. 4, right panel). In the lower transmission setting, the ratio of averted hospitalizations among DENV-exposed individuals to additional hospitalizations among DENV-naïve individuals declined to around 7.7 (5.5 -9.0).

Cost-effectiveness of the pre-vaccination screening intervention
In the baseline scenario of costs, the incremental cost-effectiveness ratio (ICER) of the intervention was around 122,000 USD per QALY gained (74,000 -182,000) (Fig. 5, left panel). The ICER was 240,000 USD at a lower transmission intensity scenario ( 9 = 30%). At the minimum vaccine price of 32 USD, we found an ICER of 15,000 USD and 57,000 USD for the moderate and low transmission scenarios, respectively. In sensitivity analyses on vaccine cost, we find that for each additional dollar of vaccine cost, the ICER increased by 305 USD ( 9 = 50%) and 520 USD ( 9 = 30%), depending on transmission intensity. At a higher cost of serological screening of 60 USD, we found that the ICER increased to 143,000 USD ( 9 = 50%) and 297,000 USD ( 9 = 30%). In terms on the ICER per averted symptomatic case, we estimated that the intervention costs around 11,000 USD to avert a symptomatic case at moderate transmission (Fig. 5, middle panel), and around 21,000 USD at a lower transmission setting. Finally, the cost to avert a hospitalized case was around 16,000 USD for a moderate transmission scenario and around 32,000 USD at a lower transmission scenario (Fig. 5, right panel).

DISCUSSION
Using an agent-based model of dengue virus transmission, we simulated the impact of a routine program of pre-vaccination screening of 9-year-olds with subsequent vaccination of seropositive individuals, at dengue transmission for use under a CC0 license. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted October 9, 2020. . https://doi.org/10.1101/2020.10.07.20208512 doi: medRxiv preprint levels that were calibrated to Puerto Rico. Assuming a moderate and low transmission intensity (PE 9 = [50%, 30%]) in Puerto Rico, we found that this intervention could be beneficial at the population-level and less so at the individual-level, as long as the serological pre-vaccination screening test has at least moderate values of sensitivity and high specificity. In sensitivity analyses, we found that a minimum specificity of 0.6 for serological screening was required to ensure that the intervention would not result in an increase of hospitalizations due to vaccination of DENV-naïve individuals We focused on a scenario where 80% of 9-year-olds were routinely given pre-vaccination screening and subsequent vaccination of seropositives.
for use under a CC0 license. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
There are many possible reasons that widespread use of pre-vaccination screening and any subsequent vaccinations may not be achieved. A number of challenges, common to many vaccines, may potentially reduce intervention coverage, such as: (1) [13]. One important difference is that our study used a higher cost of vaccination in the baseline scenario. At the lower bound of vaccination cost assumed in our study (32 USD for three doses), the cost per QALY gained was between 15,000 USD and 57,000 USD. Similar results were found by Zeng et al., although they omitted the cost of screening [30]. Our study differs in the assumption of the costs of clinical care. We focused on the average cost paid by the government (1,615 USD) [17], the two studies mentioned above [13,30] used the overall direct cost of hospitalization (4,135 USD), which includes the cost paid by insurance, households, and for use under a CC0 license. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted October 9, 2020. . https://doi.org/10.1101/2020.10.07.20208512 doi: medRxiv preprint employers. While these could account for some differences between the results of our studies, our sensitivity analyses suggest that changes to health care costs would only modestly affect the ICER estimates. At the base case of vaccine cost of 382 USD, dengue vaccination was estimated to cost 122,000 USD per QALY gained. Our estimate for the cost-effectiveness of dengue screening combined with vaccination falls within the range of economic values that have been estimated for some of the other adolescent vaccines in the United States [31]. For example, the cost per QALY gained for influenza vaccination of adolescents 12-17 years who are not at high risk was 119,000 in 2006 USD [31].
Although our model has been carefully parameterized our approach has some limitations. First, we did not explicitly simulate demographic characteristics of Puerto Rico. Instead, we parameterized our model to represent generic settings of transmission intensity with a demographic structure appropriate to Iquitos, Peru. The assumption of similar agestructures can result in some differences in transmission patterns.
Nonetheless, in previous analyses with this model, our approach showed similar projections on the impact of dengue vaccination to seven other models [19]. Another limitation of our model is that we are simulating homogeneous circulation of DENV serotypes. Although this assumption is unrealistic, given that dengue outbreaks are characterized by the dominance of one of the serotypes, it would be unfeasible to make projections of the serotype-specific DENV importations for the next ten years. In our baseline parameter for the cost of a vaccinated child, we did not explicitly include some factors such as transportation costs for the child and a caregiver, and storage and distribution of vaccine materials. These additional factors would likely serve to increase the cost per capita of a fully vaccinated child relative to our baseline assumption, which is the reason for the extensive sensitivity analyses included around vaccine costs and for the relative broad ranges applied to this parameter.
for use under a CC0 license.
This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted October 9, 2020. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted October 9, 2020. . https://doi.org/10.1101/2020.10.07.20208512 doi: medRxiv preprint Figure 1. Proportion of total dengue cases averted in Puerto Rico with the prevaccination screening strategy in 9-year-olds for different values of the sensitivity and specificity of pre-vaccination screening test. Left panel refers to symptomatic cases and right panel to hospitalizations. The x-axis shows different assumptions on the specificity and sensitivity of the pre-vaccination screening test. The simulations were performed for 80% intervention coverage of routine pre-vaccination screening and subsequent vaccination in the event of positive result for 9 year-olds over 10 years.

FIGURES
for use under a CC0 license. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted October 9, 2020. . https://doi.org/10.1101/2020.10.07.20208512 doi: medRxiv preprint  This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted October 9, 2020. . https://doi.org/10.1101/2020.10.07.20208512 doi: medRxiv preprint risk of hospitalizations. Red lines show an assumed intensity of transmission of 9 = 50%, and black lines represent 9 = 30%. The simulations were performed for 80% intervention coverage of routine pre-vaccination screening and subsequent vaccination in the event of positive result for 9 year-olds over 10 year time horizon.  This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted October 9, 2020. . https://doi.org/10.1101/2020.10.07.20208512 doi: medRxiv preprint This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.