Model Overview and Modeling Approach

A static, multi-cohort Markov model was developed to evaluate the impact of the adjuvanted RSVPreF3 vaccine on symptomatic RSV-ARI cases and related morbidity and mortality compared with no vaccination in US adults aged ≥ 60 years. The modeled population was aligned to the current indication for the adjuvanted RSVPreF3 vaccine [14] and was stratified into seven age-specific cohorts (60–64, 65–69, 70–74, 75–79, 80–84, 85–89, and ≥ 90 years) to account for differential risks of key outcomes (e.g., RSV-related hospitalizations or deaths). The youngest age in a given cohort was used for the full cohort at the start of the simulation. The model was developed in Microsoft Excel (Microsoft Corporation; Redmond, WA, USA) and included a 3-year time horizon with adjuvanted RSVPreF3 vaccination occurring at the start of the simulation. The modeled time horizon was selected based on results from the adjuvanted RSVPreF3 vaccine’s phase 3 clinical trial. Specifically, data through two full RSV seasons suggest that while adjuvanted RSVPreF3 vaccine efficacy against RSV is greatest within the first year, some degree of protection is projected to persist for up to 3 years, accounting for waning [16, 20]. An overview of the model’s underlying Markov framework, as applied in the present analysis, is presented in Fig. 1.

Fig. 1

Markov model structure. The multi-cohort Markov model follows US adults aged ≥ 60 years over 3 years, with a 1-month cycle length. The model includes seven age-specific cohorts and evaluates outcomes for scenarios with and without adjuvanted RSVPreF3 vaccination. The underlying model structure for RSV reinfections is the same as for initial RSV infections. ARI acute respiratory illness; LRTD lower respiratory tract disease; RSV respiratory syncytial virus; URTD upper respiratory tract disease; US United States

The Markov model and its inputs account for US population characteristics (size, age distribution, and all-cause mortality), RSV epidemiological data, HCRU, vaccine efficacy (VE) and waning, and vaccination coverage in older adults. The model follows older adults over 3 years, with a 1-month cycle length to account for seasonal variation in RSV epidemiology and for waning of VE over time. Use of the underlying Markov structure allows individuals to start in the “No RSV” health state and transition between different health states and events over the modeled time horizon (e.g., from the “No RSV” health state to the disease event “Symptomatic RSV-ARI” [classified as either “RSV-URTD” or “RSV-LRTD”] and then to the recovery health state [“Post-RSV”], also allowing for individuals to be reinfected). Additional details are provided in Table S1 in the Supplementary Material. These health states each represent important model outcomes for estimating the burden of RSV at an individual and population level. Symptomatic RSV-ARI cases provide a measure of the core disease burden of RSV infection within the older adult population, with those considered as RSV-LRTD more likely to have more severe RSV outcomes (e.g., ED visits, hospitalizations, X-ray confirmed pneumonia, RSV-related mortality). The model also accounts for all-cause mortality and allows for RSV reinfection.

Model Input Parameters

Model inputs were based on data from publicly available US sources, the published literature, and the pivotal Adult Respiratory Syncytial Virus (AReSVi-006) phase 3 clinical study (including data through two full RSV seasons) [16, 20] (Table 1). The model structure, data inputs, and underlying assumptions were informed by a targeted review of the literature and validated throughout the model development process by a number of RSV expert clinicians, epidemiologists, and health economists. For symptomatic RSV-ARI, core incidence data were based on data gathered over four consecutive seasons from Falsey et al. [2] (see below). The percentage of cases classified as RSV-LRTD was drawn from data across two seasons from the AReSVi-006 phase 3 study, with vaccine efficacy also based upon these longer-term data [16, 20] (and data on file). A more detailed description of these parameters and specific values are presented below.

Table 1 Epidemiological and healthcare resource use model inputsPopulation Characteristics

In line with the indication for the adjuvanted RSVPreF3 vaccine, the modeled population included all US adults aged ≥ 60 years (n = 82,862,258), based on US Census Bureau population projections for 2023 (Table 1) [21]. All-cause mortality was derived from age-specific US 2020 annualized values for probability of death [22], converted to monthly probabilities (Table S2 in Supplementary Material).

Epidemiological Input Parameters

Epidemiological model input parameters are presented in Table 1. Incidence of symptomatic RSV-ARI was derived from data reported by Falsey et al. [2], calculated as the mean weighted average incidence among non-high-risk older adults (aged ≥ 65 years) and high-risk adults (aged ≥ 21 years with congestive heart failure or chronic lung conditions) across four winter seasons, after removing asymptomatic RSV cases. The derivation used weighting based on the proportions of non-high-risk and high-risk adults enrolled in a large influenza vaccine trial conducted in the US by DiazGranados et al. [23]. As Falsey et al. used multiple RSV testing methodologies [2], no further adjustment for under ascertainment was made. The estimated incidence of symptomatic RSV-ARI per person-year was 0.0465 (0.0272–0.0627); lower and upper bounds used in sensitivity analyses were calculated as the minimum and maximum incidence of symptomatic RSV-ARI across the four consecutive winter seasons evaluated by Falsey et al. [2].

RSV infection rates show substantial seasonality [24]. In our analysis, monthly RSV incidence estimates were adjusted for seasonality using 2018–2019 RSV surveillance data from the National Respiratory and Enteric Virus Surveillance System (NREVSS) [25]. These NREVSS data were prior to the Coronavirus Disease 2019 (COVID-19) pandemic, providing information on RSV seasonality in a typical year. The seasonality adjustment factors were derived from the NREVSS data by dividing the total number of polymerase chain reaction (PCR)-confirmed RSV cases in each month by the average monthly number of PCR-confirmed RSV cases (Table S3 in Supplementary Material). RSV reinfection rates were assumed to be the same as for initial infection. The percentages of symptomatic RSV-ARI cases characterized as RSV-URTD and RSV-LRTD were based on data from the AReSVi-006 phase 3 study; over two full seasons, 47.6% of symptomatic RSV-ARI cases in the placebo arm were categorized as RSV-LRTD [20] (and data on file).

Healthcare Resource Use and Complications

RSV-related HCRU inputs (Table 1) were primarily based on the age-specific rates of medically attended RSV reported from a systematic literature review and meta-analysis by McLaughlin et al. [6]. In the base-case analysis, rates of RSV-related outpatient visits, ED visits, and hospitalizations were adjusted for RSV underdetection (Table 1). A calibration process was used within the model to determine the input values needed in order for the model to estimate the adjusted rates reported by McLaughlin et al. [6]. Specific inputs included within this calibration process included the percentage of symptomatic RSV-ARI cases that are medically attended (where medically attended symptomatic RSV-ARI cases were assumed to have one outpatient visit each), the percentage of RSV-LRTD cases that experience ED visits, and the percentage of RSV-LRTD cases that experience hospitalizations. The model assumed that RSV-LRTD cases are approximately twice as likely to be medically attended compared with RSV-URTD cases based on the previous decision analytic model from Herring et al. [26]. All ED visits and hospitalizations were assumed to only occur among RSV-LRTD cases, and RSV-related deaths were assumed to occur only among hospitalized RSV-LRTD cases.

Model inputs for antibiotic use and X-ray confirmed pneumonia were obtained from Belongia et al. [27]. We assumed that pneumonia could develop only in those individuals with RSV-LRTD. Belongia et al. [27] provides the number of x-ray-confirmed pneumonia cases out of the total number of medically attended moderate-to-severe LRTD (msLRTD) cases. Because these data only include outcomes for medically attended RSV cases, we adjusted this percentage by the percentage of RSV-LRTD cases that are medically attended (estimated from McLaughlin et al. and Herring et al. [6, 26]). Mortality rates were estimated based on mortality following RSV-LRTD hospitalization from the study by Tseng et al. [7]. Although 30-day mortality data were only reported for the overall population aged ≥ 60 years (8.6%), Tseng et al. also reported an overall in-hospital mortality rate of 5.6%, with age-specific rates of 4.6% in adults aged 60–74 years and 6.1% in adults aged ≥ 75 years [7]. By applying the same age distribution to the reported 30-day mortality estimate among adults aged ≥ 60 years, we derived age-specific 30-day mortality inputs for use in the model (60–74 years: 7.1%; ≥ 75 years: 9.4%). These rates were then applied to the age-specific percentages of RSV-LRTD cases resulting in hospitalization to generate overall RSV-LRTD mortality estimates.

Vaccine Characteristics and Vaccine-Specific Parameters

VE inputs for the adjuvanted RSVPreF3 vaccine were based on results from the AReSVi-006 phase 3 clinical study with a median follow-up time of 18 months [16, 20] (and data on file). The model accounted for waning of VE against RSV-ARI and against RSV-LRTD across the 3-year study horizon. Specifically, weighted linear regression models were fitted on the trial data to estimate the monthly VE for RSV-ARI and RSV-LRTD during and beyond the clinical trial follow-up period (Fig. 2).

Fig. 2

Adjuvanted RSVPreF3 vaccine efficacy against RSV-ARI and RSV-LRTD over modeled 3-year time horizon. In the month of first vaccination (i.e., 1st cycle of the model), 50% of peak VE is considered. Waning of the peak VE starts in the second month following vaccination. A linear decrease is applied to VE against RSV-ARI and against RSV-LRTD. The waning rates are applied each month as an absolute percentage point decrease in VE. Peak VE and waning rates were estimated based on the AReSVi-006 phase 3 clinical trial [16, 20] (and data on file). Solid lines represent data from the AReSVi-006 phase 3 clinical trial [16, 20] (and data on file), while dashed lines are extrapolations based on the weighted linear regression analysis. VE point estimates are shown over a 4-year period, with the model limited to a 3-year time horizon. ARI acute respiratory illness, LB lower bound, LRTD lower respiratory tract disease, RSV respiratory syncytial virus, UB upper bound, VE vaccine efficacy

Using this approach, peak VE against RSV-ARI and RSV-LRTD were estimated to be 74.2 and 88.0%, respectively. For the first month following vaccination, it was assumed that VE would be 50% of peak values to allow for an immune build-up period. Waning of the peak VE was assumed to start in the second month following vaccination, and to continue linearly in each subsequent month. Waning rates were applied each month as an absolute percentage point decrease in VE (monthly waning of 2.3% for VE against RSV-ARI and 2.1% for VE against RSV-LRTD). Further details about the VE calculations are described in the Supplementary Material, including Table S4 and Supplementary Fig. 1. It was conservatively assumed that RSV vaccination has no impact on the risk of reinfection following a breakthrough RSV case.

The model assumed vaccination in October to align with typical timing of influenza vaccinations. Vaccination coverage was based on coverage for influenza vaccines among US older adults during the 2021–2022 season (the most recent data that were available at the time of the analysis) [28]. This assumption has been used previously in models evaluating potential benefits of RSV vaccination [26, 29], in part because influenza and RSV follow similar seasonal patterns and influenza vaccination rates represent a potentially achievable target of older adults who are willing to be vaccinated against seasonal respiratory infections. However, we evaluated a wide range of coverage estimates in scenario analyses. As such, the base-case analysis assumes that 52.4% of adults aged 60–64 years and 73.9% of adults aged ≥ 65 years receive the adjuvanted RSVPreF3 vaccine once, at the start of the modeled 3-year time horizon [28].

Outcomes

Key health outcomes of interest focused on the number of symptomatic RSV-ARI cases (overall and by RSV-URTD and RSV-LRTD cases), and their associated HCRU (number of hospitalizations, ED visits, outpatient visits, and antibiotic prescriptions), complications (number of cases of X-ray confirmed pneumonia), and deaths. Outcomes over the 3-year time horizon were calculated for each strategy (vaccination with the adjuvanted RSVPreF3 vaccine and no vaccination), with the model also calculating the incremental differences.

Numbers of individuals needed to vaccinate (NNV) to avoid specific outcomes were also calculated based on the modeled results. NNV estimates were calculated by dividing the number of older adults who were vaccinated by the number of each outcome avoided as a result of vaccination.

Sensitivity and Scenario Analyses

Sensitivity and scenario analyses were conducted to assess the robustness of the results to uncertainties around key input parameters. Because of the wide range of RSV burden reported in the literature, we examined the impact of different input values for symptomatic RSV-ARI incidence (based on the minimum and maximum seasonal incidence from Falsey et al. [2, 23]), the percentage of symptomatic RSV-ARI cases that are RSV-LRTD (based on 95% confidence intervals), RSV-related hospitalization rates (based on data reported by Branche et al. [30] and Herring et al. [26]), and RSV-related mortality (based on 95% confidence intervals). Analyses were also conducted assuming alternate values for VE inputs and vaccination coverage (considering 25, 50, 75, and 125% of influenza vaccination coverage estimates, as well as considering 100% RSV vaccination coverage).

Statement of Ethics Compliance

This analysis is based on previously conducted studies and does not contain any new studies with human participants or animals performed by any of the authors. Ethical approval for the development of the model was not required.