A Cost-Effectiveness Analysis of Vaccination for Prevention of Herpes Zoster and Related Complications: Input for National Recommendations

Abstract

Herpes zoster (HZ) causes substantial illness in older adults, and the high rates of complication increase with age (13). Complications associated with HZ infection include severe pain and discomfort, which may last for months to years (postherpetic neuralgia [PHN]); vision loss; and skin infection (3). Since 2008, the Advisory Committee on Immunization Practices (ACIP) has recommended vaccination with zoster vaccine live (ZVL) to prevent shingles and complications in adults aged 60 years or older (4). A new HZ subunit vaccine, recombinant zoster vaccine (RZV), was approved by the U.S. Food and Drug Administration and recommended by the ACIP in October 2017. This newer vaccine has shown higher efficacy than ZVL in clinical trials (5). Uptake of ZVL has been low, and only 33% of adults aged 60 years or older reported receipt of ZVL by 2016 (5). The Food and Drug Administration has also approved ZVL for persons aged 50 to 59 years, but it is not recommended by the ACIP and has similarly low uptake in this age group. Studies after the approval and recommendation of ZVL showed significant waning of vaccine efficacy over 11 years (6). The objective of this study was to evaluate the cost-effectiveness of vaccinating adults aged 50 years or older with the new adjuvanted RZV compared with ZVL and no vaccination. Our specific goal was to inform the following 3 key questions for ACIP policy recommendations from an economic perspective. First, should vaccination with RZV be recommended for adults aged 50 years or older? Second, should it be recommended for adults aged 50 years or older who previously received ZVL? Finally, should it be preferentially recommended over ZVL? Methods Model Structure An HZ state-transition simulation model was used to project costs and health outcomes for the following 3 strategies: vaccination with RZV, vaccination with ZVL, and no vaccination (Figure 1). Costs and health outcomes were simulated for a hypothetical cohort of immunocompetent U.S. adults, stratified by age (50 to 59, 60 to 69, 70 to 79, 80 to 89, and 90 to 99 years). The simulation began at age of vaccination for each age cohort and continued throughout the lifetime. During each year of the simulation, a person could have an episode of HZ or PHN, other complications, or death. The model included the possibility of repeated reactivations of HZ and adverse events associated with vaccination (injection site reactions, systemic reactions, and serious adverse events). Each health state in the model was associated with costs and health-related quality-of-life weights (health utilities). The model was programmed using TreeAge Pro 2017, version R2.1 (TreeAge Software). Figure 1. Simplified schematic of the HZ vaccination simulation model. Square nodes indicate decisions, and circles indicate chance nodes (or probabilities). Post-HZ indicates a health state after an episode of HZ in which the person has recovered from HZ, PHN, or other complications. HZ= herpes zoster; PHN= postherpetic neuralgia. Model Parameterization Published data sources and primary data were used to derive transition probabilities, costs, and other model inputs; for parameters with no available data, these sources were supplemented by expert panel input and other methods to guide assumptions. The Supplement lists all model parameters, data sources, and assumptions. Studies for use in model parameterization were identified through a systematic literature review done in PubMed between December 2014 and March 2015 and updated in September 2016. For HZ and PHN incidence, we prioritized studies that reported results by age or sex to reflect differences across these characteristics. Clinical trials were used preferentially for vaccine effectiveness estimates. All other parameters were selected by the group of coauthors as representing the best available evidence and were reviewed by an expert panel and the ACIP Zoster Work Group. Supplement. Supplementary Material Transition Probabilities We derived HZ incidence (2, 7, 8); the conditional probability of PHN given HZ (3, 914) and of ocular (15), neurologic (3), and cosmetic (3) complications; and HZ mortality (16) from published literature. Age-adjusted all-cause mortality was also included in the model (17). The probability of recurrent HZ was determined by expert opinion (18, 19) (Supplement Table 1). Effectiveness of RZV over 4 years was derived from 2 randomized controlled trials (20, 21). Waning duration was derived by extrapolating a linear function using published data from the 4 years of the clinical trials (20, 21) (Supplement Table 1 and Supplement Figure 1). The proportion of patients completing the 2-dose series of RZV was assumed to be 100% in the base-case analysis. The base case was designed to correspond with ACIP recommendations for the 2-dose regimen of RZV and to be consistent with previous analyses by the Centers for Disease Control and Prevention (CDC), which assumed 100% completion of a multiple-dose regimen. This assumption was relaxed in sensitivity analyses based on data from existing vaccines with a multiple-dose protocol. Estimates of 1-dose vaccine effectiveness for RZV were based on post hoc analyses from clinical trial data. Initial 1-dose effectiveness was reported at 90% for recipients aged 50 to 69 years and 69% for those aged 70 years or older (20, 21), and waning durations were assumed to be 11 and 4 years, respectively. Effectiveness and waning of ZVL were modeled using data from clinical trials (Supplement Table 1 and Supplement Figure 2). We calculated net adverse events associated with RZV (20, 21) and ZVL (2224) from clinical trial data and accounted for the expected co-occurrence of grade 3 systemic reactions and injection site reactions (21, 23). Health care use for provider visits associated with a grade 3 systemic reaction and the proportion of patients requiring an emergency department visit or hospitalization were based on published data, or on expert opinion if no published data were available (25). Serious adverse events were assumed to reflect a 3-day hospitalization with a full recovery (Supplement Table 1). Costs and Quality-of-Life Adjustments Direct medical costs of uncomplicated HZ and PHN were derived from 2 studies (26, 27), and those of other complications were derived from 1 study (27). Productivity losses per case were obtained from 3 studies for HZ (2830) and 1 for PHN (28). Vaccine dose costs were determined from published prices (31). Vaccine administration costs were sourced from the 2016 Medicare Physician Fee Schedule. The base-case analysis assumed that vaccinations would be delivered through various settings (such as physician offices and pharmacies) and that visits would not be vaccination-specific. The base case did not include patient time costs; patient time required for vaccination was assessed in sensitivity analyses. Estimated costs of health care use for adverse events were based on published estimates and expert opinion. All costs were adjusted to 2016 U.S. dollars (32) (Supplement Table 2). The quality-adjusted life-year (QALY) is the primary measure of health benefits. Health states included in the model are predominantly temporary, and each temporary health state is associated with a loss in QALYs that is subtracted from the total QALYs experienced. The loss in QALYs associated with each health state was derived from a separate model that combined length and severity of HZ illness (using data on health utilities collected from a patient sample using time-tradeoff questions) to predict an average loss in QALYs associated with an episode of HZ or PHN (33) (Supplement). The model also included quality adjustments for vaccination-related adverse events (Supplement Table 3). Outcomes Assessed Cost-Effectiveness Analysis and Secondary Outcomes The primary outcome measure was the incremental cost-effectiveness ratio (ICER), which is measured in dollars per QALY and was calculated for the health care sector and societal perspectives (34) (Supplement Figure 3, shows an impact inventory). The analysis focused mainly on the societal perspective. A 3% annual discount rate was used (34). Secondary outcomes included total numbers of cases of HZ, cases of PHN, and cases of HZ and PHN averted; total costs; and total QALYs per 1000-person cohort. For the base-case analysis, we estimated total costs, total QALYs, and ICERs for all age groups. Cost-Effectiveness of Vaccination With RZV in Persons Who Had Previously Received ZVL We evaluated vaccination with RZV at the following 3 intervals after administration of ZVL: immediately (to approximate an 8-week interval), a 1-year interval, and a 5-year interval. The incremental effectiveness of RZV after ZVL administration was approximated assuming a linear additive relationship between the vaccine effectiveness waning functions for the 2 vaccines (Supplement Figure 4). Uncertainty Analysis We did 1-way sensitivity analyses on parameters using the ranges listed in Supplement Tables 1 to 3. We also did a series of scenario analyses to explore multiway sensitivity analysis or to assess the effect of parameter values outside the ranges identified for the initial sensitivity analyses (Supplement Tables 1 to 3). The scenarios included a range of adherence to the recommended 2-dose regimen from 20% to 100%, and additional costing perspectives that included or excluded time costs of vaccination and productivity losses in addition to costs typically accounted for in the health care sector perspective. We did probabilistic sensitivity analysis and generated cost-effectiveness acceptability curves. For the probabilistic sensitivity analyses, each parameter was defined as a distribution and 10000 simulations were run. In each run, a value for each parameter was drawn from its specific distribution and used to calculate each health and economic outcome, allowing for the generation of distributions for intermediate outcomes of c