Higher-valency pneumococcal conjugate vaccines in older adults, taking into account indirect effects from childhood vaccination: a cost-effectiveness study for the Netherlands

We updated a previously published static multi-cohort model [7] to compare costs and quality-adjusted life year (QALY) losses of various pneumococcal vaccination strategies for older adults in the Netherlands. The primary analysis focused on a 65-year-old cohort over a 15-year time horizon. We also explored alternative vaccination ages (60, 70, 75, 80, and 85 years). The following strategies were evaluated:

Strategy 2 aligns with the current Dutch PPV23 program.

The cost-effectiveness of vaccination of older adults was assessed with the continuation of PCV10 in childhood vaccination, or with a switch to PCV13, PCV15, or PCV20 at year 0, resulting in indirect effects on older adults. Notably, PCV21 is not considered for childhood vaccination, as its serotypes have been selected based on the pneumococcal serotype epidemiology in older adults. Moreover, PCV21 is unadjuvanted and, therefore, potentially less immunogenic in children < 2 years of age compared to other PCVs [24]. Indirect effects from childhood vaccination were included by adjusting the incidence of infection in older adults with a correction factor (see the “Indirect effects from childhood vaccination” section). As our analysis is centered on informing decisions regarding the vaccination of older adults, we did not include additional vaccination costs associated with a switch to higher-valency PCVs in childhood vaccination. Furthermore, the analysis does not account for the direct or indirect impact of these higher-valency vaccines on the disease burden in children themselves or in younger adults.

A comprehensive model description and the parameterization details are available in the Supplementary Methods of Additional file 1. Briefly, our model tracked single year of age cohorts aged ≥ 60 years in annual time steps, in which they could develop IPD or hospitalized non-invasive pneumococcal pneumonia (NIPP). Pneumococcal infections requiring primary care were not included, given the uncertainty regarding whether pneumococcal vaccination provides significant protection against this outcome [25]. To avoid the influence of COVID-19 measures on pneumococcal epidemiology [26], we utilized pre-2020 epidemiological data. As the programmatic pneumococcal vaccination for older adults started in the Netherlands in 2020, this concerned data from an unvaccinated population. Consequently, there is no need for back calculation to determine the burden of pneumococcal disease in an unvaccinated population. The average IPD incidence by age during the years 2017–2019 was derived from Dutch national surveillance data. The NIPP incidence by age was estimated from the average all-cause CAP hospitalization incidence during 2012–2014, assuming that 22.1% of these cases were caused by S. pneumoniae [27]. NIPP incidence was adjusted to account for overlap with IPD cases [28] and indexed to 2017–2019 using the IPD time trend. IPD and NIPP cases were categorized by serotype based on data of 2019 from Dutch IPD cases aged ≥ 60-year-olds. Age-specific 30-day case fatality rates for IPD and NIPP were used to estimate pneumococcal-related deaths [29, 30]. Cohort sizes and aging of cohorts were informed by data from Statistics Netherlands [31, 32].

The impact of vaccination was modeled as a reduction in the incidence of vaccine serotypes, involving the vaccine uptake and vaccine-type specific vaccine effectiveness (VE). A vaccination coverage of 70% reflected the current uptake among Dutch older adults [33, 34]. The VE at time of vaccination decreased with increasing age. At age 65 years, the vaccine-type specific VE for PPV23 was estimated at 54% against IPD and 27% against NIPP, using data from multiple observational studies [35‐37]. A revaccination with PPV23 was assumed to have the same VE as a first dose would have had at that age. For PCV, a vaccine-type specific VE of 81% against IPD and 54% against NIPP for vaccination at age 65 years was derived from age-specific VE estimates in a post hoc analysis of Dutch trial data for PCV13 [38]. This VE was adjusted downwards to account for the relatively healthy trial population. The vaccine-type specific VE for PCV13 was extrapolated to PCV15, PCV20, and PCV21 serotypes, and it was also applied to non-vaccine serotype 6C, recognizing evidence on cross-protection via serotype 6A [39]. The decrease of VE over time due to waning immunity was modeled similarly to previous assumptions [7]. For PPV23, the VE remained constant for 2 years and then linearly decreased to 0% at 5 years post-vaccination. For PCV, the VE remained constant for 4 years, and then linearly declined to 0% at 15 years post-vaccination. For combined strategies with PCV and PPV23, the VE of the vaccine with the higher VE was utilized at each time step; no additional efficacy against serotypes covered by both vaccines was assumed.

With continued use of PCV10 in childhood vaccination, we assumed a steady pneumococcal disease incidence and serotype distribution in older adults. With the implementation PCV13, PCV15, or PCV20 in childhood vaccination, we based the time courses of indirect protection and serotype replacement on observations from the multi-country studies PSERENADE and SpIDnet after the introduction of PCV10 and PCV13 childhood programs [40, 41]. The PSERENADE data shows an over 75% decrease in the incidence of IPD from PCV10-included serotypes in individuals aged ≥ 65 years at 8 years after the initiation of the childhood PCV program. Subsequently, there was an observed increase in IPD incidence from non-vaccine serotypes, attributed to serotype replacement. The net impact on IPD incidence varied across countries, with some experiencing no change, while others saw up to a 50% decline. In the Netherlands, despite an over 90% reduction in IPD incidence from PCV10-included serotypes, there was no significant net change in IPD incidence among individuals aged ≥ 60 during the period 2008–2019 (Additional file 1: Figure S2). Taking these observations into account, we used the following assumptions for our main analysis:

Figure 1 illustrates the change in incidence and serotype distribution of IPD at the age of 65 years over time. The new steady state reached 8 years post-switch was maintained for the remaining time horizon. Recognizing the uncertainty associated with these assumptions, we performed a sensitivity analyses by altering the magnitudes of indirect effects. Considering modeling suggestions that an increase in vaccine valency could diminish indirect effects [42], we explored a scenario incorporating 40% indirect protection. Regarding serotype replacement, we investigated a scenario in which 50% of the reduction in pneumococcal disease burden was replaced.

The societal perspective was taken, following Dutch guidelines [43]. We used IPD- and NIPP-related hospitalization costs, patient expenses, and productivity losses for the Netherlands [30, 44‐46], adjusted to the 2021 price year [47] (Additional file 1: Table S4). Vaccine administration costs at the general practitioner in 2021 were €21 per dose [48]. Vaccine prices per dose were acquired from a Dutch pricelist for individual usage: PPV23 at €25.94, PCV15 at €74.73, and PCV20 at €82.17 [49]. In the absence of a price for the in-development PCV21, we assumed it to have the same price as PCV20. This approach ensures that any differences in costs between PCV20 and PCV21 can be attributed solely to the prevented economic burden. This does not imply a prediction that the eventual price of PCV21 will be the same as PCV20. QALYs lost due to either IPD or NIPP was 0.0709, based on utility loss during the acute phase and first month of hospitalized non-fatal CAP cases [46]. QALYs lost from premature death were age-specific, using the life expectancy adjusted for general population utilities (Additional file 1: Table S5 and Figure S5).

Costs and QALY losses for each strategy were accumulated over the time horizon, with annual discount rates of 4% for costs and 1.5% for QALYs [43]. Incremental cost-effectiveness ratios (ICERs) were calculated, with strongly dominated (i.e., presence of a more effective and less costly alternative) and extendedly dominated strategies (presence of a more effective alternative with lower ICER) being removed from the comparison. Probabilistic sensitivity analyses involved 1000 Monte Carlo simulations, with all parameter values drawn from their distributions. The 2.5% and 97.5% percentiles informed 95% confidence intervals (95% CI) of estimated case reductions. Uncertainty of cost-effectiveness outcomes was displayed using cost-effectiveness acceptability frontiers, illustrating the preferred strategy at different willingness-to-pay (WTP) thresholds. A common WTP threshold for preventive interventions in the Netherlands is €20,000/QALY gained [50]. One-way sensitivity analyses assessed the impact of assumptions or alternative sources on the cost-effectiveness outcomes (Additional file 1: Table S6) [51‐53].

Figure 2 shows the relative reductions in IPD cases and NIPP cases for different vaccination strategies in a 65-year-old cohort (228,209 individuals, 70% vaccination coverage) over a 15-year period compared to no vaccination, while varying the childhood vaccine. Absolute numbers of cases prevented are detailed in Additional file 1: Tables S7-S10. With continued use of PCV10 in children, the current Dutch strategy of 3xPPV23 was estimated to reduce the number of IPD cases by 20% (95% CI: 15–24%) and the number of NIPP hospitalizations by 10% (95% CI: 0–19%). This corresponds to 274 prevented IPD cases and 227 prevented NIPP hospitalizations and the avoidance of 69 deaths. PCV15 had a smaller impact, while PCV20 had a greater impact than 3xPPV23, resulting in a 26% (95% CI: 20–29%) reduction. PCV15 + 3xPPV23 had similar impact to PCV20, while PCV20 + 3xPPV23 had most impact, preventing 32% (95% CI: 26–34%) of IPD cases. PCV21, which is in development, prevented 31% (95% CI: 23–34%) of IPD cases. The impact of vaccination on NIPP was relatively lower compared to the impact on IPD, particularly for strategies involving PPV23 (Fig. 2, bottom left). PCV21 prevented most NIPP hospitalizations with a reduction of 19% (95% CI: 11–25%) compared to no vaccination.

Indirect effects from the use of higher-valency PCVs in childhood vaccination reduced the impact of pneumococcal vaccination in older adults, depending on the serotype overlap between the vaccines used for children and older adults (Fig. 2). A switch from PCV10 to PCV13 or to PCV15 in children reduced the impact of PCV15 in older adults substantially, but the impact of PPV23, PCV20, or PCV21 minorly. Use of PCV20 in children reduced the impact of both PCV20 and PPV23 in older adults, although the change in impact was larger for PCV20. In that scenario, PCV20 prevented 15% (95% CI: 12–17%) of IPD cases in the 65-years-old cohort, about equally high as the impact of 3xPPV23. Indirect effects from PCV20 in children did minimally affect the impact of PCV21 in older adults, preventing 29% (95% CI: 22–32%) of IPD cases in older adults.

Figure 3 shows the discounted net costs and QALY gains for different vaccination strategies of a 65-year-old cohort over a 15-year period compared to no vaccination, while varying the childhood vaccine (details provided in Additional file 1: Tables S11-S14). With PCV10 in children, we found PCV20 to dominate 3xPPV23, PCV15, and PCV15 + 3xPPV23 in the 65-year-old cohort (Fig. 3, with higher QALY gains against lower costs). Compared to no vaccination, PCV20 resulted in a gain of 963 QALYs, from which 57% was attributed to prevented IPD cases and 43% to prevented NIPP cases. With net costs of €8.7 million, the ICER was estimated at €9051/QALY gained. Switching to PCV13 or PCV15 in childhood vaccination resulted in a slight increase of the ICER of PCV20 to €10,228/QALY gained and €11,173/QALY gained, respectively. However, the cost-effectiveness of PCV20 in older adults diminished substantially if children were vaccinated with PCV20; the ICER compared to no vaccination increased to €22,550/QALY gained, above the conventional Dutch threshold of €20,000/QALY gained. PCV20 + 3xPPV23 yielded a higher QALY gain in older adults than PCV20, though at an ICER of €81,193–€115,412/QALY gained compared to PCV20, depending on the childhood vaccination scenario. At equal vaccine prices, PCV21 dominated PCV20 for older adults. Indirect effects had minimal impact on the cost-effectiveness of PCV21; the ICER of PCV21 compared to no vaccination in older adults increased from €6352/QALY gained with PCV10 in children to €7876/QALY gained with PCV20 in children.

Focusing on current available vaccines for older adults, we found PCV20 to be the cost-effective strategy at a €20,000/QALY gained threshold (vertical dashed line) in 90% of the simulations if PCV10 was continued in children (Fig. 4, top left). However, with PCV20 in children, no pneumococcal vaccination was the cost-effective strategy for older adults at a €20,000/QALY gained threshold in 74% of the simulations (Fig. 4, bottom right).

We found the results to be sensitive to the magnitude of indirect effects (Fig. 5A). For a scenario with 40% indirect protection from PCV20 in children (instead of 80%), the ICER of PCV20 compared to no vaccination for the 65-year-old cohort was €13,883/QALY gained (instead of €22,550/QALY gained). Reducing serotype replacement from PCV20 in children primarily diminished the cost-effectiveness of PCV21 for older adults. If only 50% of the IPD and NIPP cases prevented by indirect protection were replaced (instead of 100%), the ICER of PCV21 compared to no vaccination was €10,589/QALY gained (instead of €7876/QALY gained).

We tested other parameters and found that the results were most sensitive to the vaccine price, the VE and its waning rate, and the proportion of CAP caused by S. pneumoniae (Additional file 1: Figure S6). With PCV20 in childhood vaccination, PCV20 was cost-effective for a 65-year-old cohort at a €20,000/QALY threshold if the VE did not wane for at least 8 years or if its price was reduced by 13% (Additional file 1: Figure S6D). To achieve an ICER of €10,000/QALY compared to no vaccination, as was found for PCV20 in older adults with PCV10 in children, the price of PCV20 should be reduced by 50%. PPV23 could become cost-effective compared to PCV20 by using the upper bound of the VE of PPV23, the lower bound of the VE of PCV20, assuming no waning of PPV23 within 5 years, or reducing its price by 50%.

Varying the vaccination age, we found the cost-effectiveness results to be consistent within the range of 60–80 years. Vaccination at the age of 70 or 75 years resulted in the lowest ICERs (Additional file 1: Figures S7-S8 and Tables S15-S24). The cost-effectiveness of pneumococcal vaccination diminished substantially at the vaccination age of 85 years.

As indirect effects were modeled to progress gradually over time, its impact on the cost-effectiveness of pneumococcal vaccination in older adults becomes larger for cohorts vaccinated in the years following the switch of the childhood vaccine (Fig. 5B). In a 65-year-old cohort that is vaccinated 8 years after the switch to PCV20 in childhood vaccination, the ICER of PCV20 compared to no vaccination was €45,081/QALY gained (instead of €22,250/QALY gained if vaccinated in the year of the switch). From 3 years after the switch onwards, the ICER of 3xPPV23 compared to no vaccination was lower than that of PCV20, eventually becoming €34,571/QALY gained at 8 years post-switching. The ICER of PCV21 compared to no vaccination remained stable over time.