Potential serotype-specific effectiveness against IPD of pneumococcal conjugate vaccines V114 and PCV20 in children given a 2+1 dosing regimen

ABSTRACT

Background

Next generation, higher valency pneumococcal conjugate vaccines (PCVs) are assessed and licensed by comparing the immune response across serotypes shared with the PCVs that are standard of care for prevention of pneumococcal disease.

Methods

Using a previously qualified method we predicted the serotype-specific vaccine effectiveness (VE) against invasive pneumococcal disease of V114 and PCV20 for the serotypes shared with PCV13 in an EU, Russian, and Australian pediatric population that is recommended to receive a 2 + 1 dosing regimen.

Results

The estimated protective antibody concentrations ranged from 0.03 (serotype 23F) to 1.49 µg/mL (serotype 19F). Predicted VE values for V114 ranged from 79% (serotype 5) to 100% (serotype 23F). V114 had comparable effectiveness to PCV13 for all but one of shared serotypes, with predicted higher effectiveness (in V114) against serotype 3 (93% vs. 65%). Predicted VE values for PCV20 ranged from 47% (serotype 3) to 91% (serotype 14). PCV20 predicted VE was lower than PCV13’s for serotypes 4, 19F, 23F, 1, 3, 5, 6A, 7F, and 19A.

Conclusions

Predicted serotype-specific VE values suggest that, with a 2 + 1 dosing regimen, V114 will have greater effectiveness than PCV20 against PCV13 serotypes, particularly for the still-prevalent serotype 3. Real-world VE studies will ultimately provide clarity on the effectiveness of novel PCVs and support further confidence in and/or improvements to modeling efforts.

Plain Language Summary

Pediatric pneumococcal conjugate vaccines (PCVs) were first introduced in Europe in the early 2000s and their incorporation into national immunization programs has helped decrease the incidence of invasive pneumococcal disease (IPD) in Europe and globally. However, some IPD persists, due both to the emergence of non-vaccine pneumococcal serotypes and to the persistence of certain vaccine-targeted serotypes. Higher valency vaccines have been developed to help prevent IPD arising from these serotypes. The goal of the present study is to employ a previously developed model to predict the serotype-specific vaccine effectiveness of higher valency PCVs in a pediatric population that is recommended to receive a 2 + 1 dosing schedule.

KEYWORDS:

• Antibodies
• child
• immunization schedule
• modeling and simulation
• pharmacometrics
• pneumococcal vaccines
• protective concentration
• streptococcus pneumoniae

1. Introduction

The impact of pediatric pneumococcal conjugate vaccines (PCVs) is measured in terms of reduced incidence of invasive pneumococcal disease (IPD) in children. The first pediatric PCV (7-valent PCV7) was introduced in Europe in the early 2000s, followed nearly a decade later by the 10-valent PCV10 and 13-valent PCV13 [1,2]. Incorporation of these vaccines into national programs has caused IPD incidence to decrease dramatically in Europe and globally [3]. Nevertheless, some IPD persists [4,5], due both to the emergence of non-vaccine pneumococcal serotypes and to the persistence of certain vaccine-targeted serotypes.

Emergent non-vaccine pneumococcal serotypes prevalent in Europe include 22F, 33F, 8, 10A, and others [6,7]. Higher valency vaccines have been developed to prevent IPD arising from these serotypes. Serotypes 22F and 33F are the targets of the 15-valent PCV, V114 (VAXNEUVANCE, Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, NJ, USA), along with all of the serotypes targeted by PCV13 (4, 6B, 9 V, 14, 18C, 19F, 23F, 1, 3, 5, 6A, 7F, and 19A; Supplementary Table S1). Serotypes 8, 10A, 11A, 12F, and 15B are targeted by the 20-valent PCV, PCV20 (PREVNAR 20, Pfizer, Inc.), in addition to all the serotypes in V114.

The ability of the higher valency PCVs to target emergent serotypes is essential, as is the maintenance of protection against vaccine-type serotypes. In the PCV era, persistence of vaccine-type disease has been attributable primarily to serotypes 3 and 19A [3,8,9]. Furthermore, health and economic modeling analyses have predicted that PCV13-specific serotypes, 3 and 19A in particular, will comprise ≥ 10% of all IPD cases in the post-PCV13 era [1,10]. Thus, determining the serotype-specific vaccine effectiveness (VE) of new broad-coverage PCVs is essential to the success of national immunization programs, and reliable, early prediction of VE could provide valuable insight for designing and implementing such programs.

A modeling method was developed for determining serotype-specific VEs for the serotypes included in higher valency PCVs in common with lower valency PCVs. A previous study used this method to predict the serotype-specific effectiveness of serotypes shared by PCV13 with V114 and PCV20 in the context of a 3 + 1 dosing schedule [11]. However, a 2 + 1 dosing schedule is used in many countries throughout the world [2,12]. The goal of the present study is to employ the same modeling method to predict the serotype-specific VE of V114 and PCV20 with a 2 + 1 dosing schedule.

2. Methods

2.1. Summary of the method

The method for estimating VE has been described in prior publications [11,13] and is illustrated schematically in Supplementary Figure S1. The modeling method is based on the relationship between the antibody responses of vaccinated infants (i.e. immunogenicity) and clinical protection from IPD. The relationship is found by deriving serotype-specific protective antibody concentrations (C_ps) via a modification of the method originally described by Siber et al. [14]. This method was adapted here to predict serotype-specific VE for serotypes common to PCV13 for V114 and PCV20 (Supplementary Table S1). IgG titers from the one month post-primary series timepoint were chosen for analysis, as this is the timepoint used by Siber et al. to derive the currently accepted overall C_p (0.35 µg/mL) [14]. This timepoint reflects the immune response elicited within the first year of life when children are at the highest risk of IPD and has been demonstrated to be predictive of protection before and after the booster dose (see Ryman et al. [11] and others [14–16]).

The population of interest was pediatric participants ages 2 months to 5 years from the European Union, Russia, and Australia (hereinafter collectively referred to as ‘EU/R/A’) who were vaccinated with a 2 + 1 dosing regimen (2 months, 4 months, and 12–15 months; or 3 months, 5 months, and 12–15 months). Briefly, the antibody responses elicited by PCV13 were used in conjunction with publicly available VE values to derive protective concentrations for each serotype in PCV13. These C_ps were then used in conjunction with antibody concentrations elicited by V114 and PCV20 to predict VE values for each serotype that V114 and PCV20 have in common.

2.2. Data sources

2.2.1. Vaccine effectiveness

Source data are listed in Table 1 with study population characteristics presented in Supplementary Table S2. Real-world serotype-specific VE values for PCV13 serotypes were from an observational study in European Union pediatric participants who received a 2 + 1 dosing regimen [2]. This study used an indirect cohort (Broome) design [17] that compared the vaccination status of vaccine-serotype IPD cases to that of non-vaccine-serotype IPD cases. The observation period for IPD cases for PCV13 was from 2012 to 2018. In this study [2], effectiveness estimates were not available for serotypes 4, 18C, 23F, and 5, so the overall estimate for ‘PCV7’ was used for serotypes 4, 18C, and 23F, and the estimate for ‘PCV13non7’ was used for serotype 5 (quotes indicate the labels in the original publication).

Table 1. Summary-level input data and sources.

	Vaccine effectiveness^a	ELISA-normalized geometric mean serotype-specific antibody concentration (µg/mL)
Serotype	PCV13 (95% CI)	Placebo (95% CI)^b	PCV13 (95% CI)^c,d,e	V114 (95% CI)^c	PCV20 (95% CI)^d
4	93% (89-96)^f	0.04 (0.03-0.05)	1.03 (0.98-1.07)	1.22 (1.17-1.28)	0.55 (0.51-0.60)
6B	93% (82-97)	0.09 (0.08-0.09)	0.26 (0.24-0.28)	0.46 (0.42-0.50)	0.12 (0.10-0.13)
9 V	97% (85-100)	0.08 (0.07-0.08)	0.83 (0.80-0.87)	0.86 (0.82-0.89)	0.47 (0.43-0.52)
14	97% (93-99)	0.17 (0.15-0.19)	2.88 (2.69-3.08)	3.95 (3.68-4.23)	1.07 (0.97-1.19)
18C	93% (89-96)^f	0.06 (0.06-0.07)	0.91 (0.87-0.96)	0.85 (0.81-0.88)	0.60 (0.55-0.66)
19F	84% (69-92)	0.24 (0.23-0.25)	3.20 (3.07-3.33)	2.73 (2.62-2.85)	1.98 (1.83-2.14)
23F	93% (89-96)^f	0.10 (0.10-0.11)	0.47 (0.45-0.50)	0.77 (0.73-0.83)	0.17 (0.16-0.19)
1	85% (68-93)	0.05 (0.04-0.05)	1.25 (1.20-1.30)	1.16 (1.11-1.21)	0.62 (0.57-0.68)
3	65% (44-78)	0.07 (0.07-0.08)	0.51 (0.49-0.53)	0.85 (0.82-0.88)	0.38 (0.35-0.41)
5	79% (71-85)^f	0.09 (0.08-0.10)	0.73 (0.70-0.76)	0.65 (0.61-0.68)	0.49 (0.44-0.53)
6A	93% (82-97)	0.13 (0.13-0.14)	1.12 (1.07-1.18)	0.72 (0.68-0.76)	0.50 (0.46-0.53)
7F	91% (82-96)	0.07 (0.06-0.07)	1.65 (1.58-1.71)	1.53 (1.46-1.60)	0.95 (0.88-1.01)
19A	83% (74-89)	0.33 (0.31-0.34)	1.68 (1.61-1.76)	1.48 (1.42-1.55)	0.95 (0.89-1.03)

CI, confidence interval; PCV13, 13-valent pneumococcal conjugate vaccine; PCV20, 20-valent pneumococcal conjugate vaccine; V114, 15-valent pneumococcal conjugate vaccine.

^aSource: Savulescu et al. [2].

^bSource: Clinical trial NCT01340937 [18].

^cSource: Clinical trials PNEU-PED-EU-1 (NCT04031846) [20] and PNEU-PED-EU-2 (NCT04016714) [21].

^dSource: Clinical trial NCT04546425 [22].

^eTrial arms for PCV13 were pooled.

^fIn Savulescu et al., effectiveness estimates were not available for serotypes 4, 18C, 23F, and 5, so the overall estimate for ‘PCV7’ was used for serotypes 4, 18C, and 23F, and the estimate for ‘PCV13non7’ was used for serotype 5 (quotes indicate the labels in the original publication).

2.2.2. Immunogenicity data

Immunogenicity data for placebo, PCV13, V114, and PCV20 (Table 1) were obtained from clinical trials as follows. For placebo, summary-level IgG concentrations were extracted from data reported at the pre-vaccination timepoint from a United States (US)-based phase 3 trial in healthy infants concomitantly administered a hexavalent combination vaccine and PCV13 [18]. The placebo IgG concentrations in this US population were assumed to be the same as those in an EU/R/A pediatric population, and an assumption was made, in line with previous studies [14,19], that pre-vaccination IgG concentrations elicited in 2-month-old pediatric participants reflect post-primary series timepoint placebo concentrations. For PCV13, individual-level IgG concentrations from the post-primary series timepoint of two phase 3 trials of EU/R/A infants receiving a 2 + 1 regimen, PNEU-PED-EU-1 [20] and PNEU-PED-EU-2 [21], were pooled with summary-level IgG concentrations from a separate phase 3 trial of similar design in a similar population, NCT04546425 [22]. For V114, individual-level IgG concentrations were obtained from PNEU-PED-EU-1 [20] and PNEU-PED-EU-2 [21]. For PCV20, summary-level IgG concentrations were obtained from phase 3 trial NCT04546425 [22].

IgG concentrations in these trials were, in some cases, measured by different assays and so were made directly comparable by normalizing them to the appropriate standards. The IgG concentrations measured by enhanced chemiluminescence [23] and Luminex-based 13-plex direct immunoassay [24] were normalized to the World Health Organization’s standard ELISA as previously described [11,13]. Supplementary Table S3 lists the slope and intercept values used in the normalization, and IgG concentrations presented in Table 1 were normalized using these parameters.

2.3. Data analysis

For each serotype/vaccine combination, the post-primary series geometric mean concentration (GMC) of IgG (Supplementary Table S4) and the published VEs and their geometric standard deviations (GSDs; calculated from the published 95% confidence intervals [CIs]), and/or the subject-level IgG concentrations were used as inputs for Monte Carlo simulations (5,000 iterations). Each simulation i) solved for a serotype-specific C_p based on reverse cumulative distribution curves (RCDCs) generated for placebo and PCV13 and the published real-world VE for PCV13, and then ii) predicted serotype-specific VEs of V114 and PCV20 based on the model-estimated C_ps and the respective RCDCs (generated similarly from the clinical data). The RCDCs were calculated based on the assumption (typical for immune titers) of log-normal distribution (RCDC plots are in the Supplementary Material. See Supplementary Figure S2). Due to the PCV13 titers for 6B being very low, and the observed VE for 6B coming from data combined with those for 6A, we were unable to estimate reliably the C_p (and, hence, the VE) for 6B. This unique circumstance applies to 6B only in this analysis.

In cases where summary-level data were used (IgG concentrations for placebo, PCV13 from phase 3 trial NCT04546425 [22], and PCV20), observed VE values were used as input, and each simulation sampled a new GMC and VE using the log-normal distribution derived from the summary-level confidence interval data. Uncertainty in the GSD was accounted for by sampling GSD estimates around the reported values assuming chi-square distributions. When subject-level IgG concentration data were available (for PCV13 and V114 IgG concentrations from the PNEU-PED-EU-1/2 phase 3 trials [20,21]), the IgG concentrations were bootstrapped (i.e. sampled with replacement) for each serotype from individual IgG concentrations post-vaccination. The number of subjects simulated or bootstrapped was the same as the number of subjects in the respective vaccine trial arms. For PCV13 titers, summary-level IgG concentrations [22] were pooled with bootstrapped individual-level IgG concentrations [20,21] to capture the GMC and distribution of post-primary series IgG concentrations for the pediatric population. Pooling these data is necessary to allow the C_ps to be applied to both V114 and PCV20, and appropriate given that the patient demographics from the trials are sufficiently similar (Supplementary Table S2). Furthermore, it increases the power of the analysis by increasing the sample size for PCV13.

Variability in the input data results in some VE estimates being less than zero for single iterations of the sampling. The VE prediction (for each simulation sample) for PCV20 was truncated at 0% since the observed PCV13 VE data show that it is highly improbable that PCV20 has negative effectiveness (i.e. increases the risk of the targeted disease). This truncation was not done for V114 as effectiveness data are not yet available for V114.

Analysis was performed in R version 4.2.1. Simulated protective IgG threshold concentrations (μg/mL) and VE values (%) were summarized as medians with 95% CIs. The absolute difference in predicted VE of V114 and PCV20 (V114 VE minus PCV20 VE) was calculated for each simulation and summarized across the 5,000 simulations as medians with 95% CIs: values above zero favor use of V114, and those below zero favor PCV20.

3. Results

3.1. Protective antibody concentrations

Protective antibody concentrations derived from VE source data for PCV13 are shown in Table 2. Estimated protective antibody concentrations 30 days after a 2-dose primary series, ranged from 0.03 (serotype 23F) to 1.49 µg/mL (serotype 19F). Natural variability in the input data (due to, e.g. between-subject and assay variability, as well as from finite study sizes) resulted in variability (i.e. quantified ‘uncertainty’) in the C_ps and predicted VEs which is captured in the results.

Table 2. Estimated protective antibody concentrations.

Serotype	Protective antibody concentration (median (95% CI) µg/mL)
4	0.27 (0.19-0.36)
6B^a	NA
9 V	0.09 (0.03-0.35)
14	0.09 (0.05-0.22)
18C	0.16 (0.10-0.24)
19F	1.49 (0.95-2.39)
23F	0.03 (0.01-0.06)
1	0.55 (0.33-0.87)
3	0.36 (0.28-0.54)
5	0.32 (0.26-0.42)
6A	0.19 (0.04-0.47)
7F	0.57 (0.39-0.84)
19A	0.73 (0.12-1.02)

CI, confidence interval.

^aC_p for serotype 6B could not be calculated, as described in Methods and Discussion sections.

3.2. Predicted vaccine effectiveness

Table 3 shows the predicted serotype-specific VE values for V114 and PCV20 against IPD with the observed VE for PCV13. Predicted VE values for V114 were similar to the observed values for all serotypes except serotype 3, which was 28% points higher for V114. The median difference between observed PCV13 VE values and predicted V114 VE values was 2% points (with V114’s values being higher). In contrast, except for serotypes 9 V, 14, and 18C, predicted VE values for PCV20 were substantially lower than the previously observed values for PCV13 (Table 3), with a median difference of −23% points. The predicted VE of 0% for PCV20 serotype 23F results from a relatively low GMC value and high variability of PCV20 titers.

Table 3. Observed PCV13 effectiveness versus effectiveness predicted for V114 and PCV20.

		V114		PCV20
Serotype	Observed PCV13 effectiveness, median % (95% CI)^a	Predicted effectiveness, median % (95% CI)	Difference, percentage points^b	Predicted effectiveness, median % (95% CI)	Difference, percentage points^b
4	93% (89-96)^d	97% (93-98)	+4	70% (62-79)	−23
6B^c	93% (82-97)	NA	NA	NA	NA
9 V	97% (85-100)	98% (88-100)	+1	85% (59-95)	−12
14	97% (93-99)	99% (95-100)	+2	91% (81-97)	−6
18C	93% (89-96)^d	97% (93-99)	+4	85% (76-91)	−8
19F	84% (69-92)	81% (58-93)	−3	61% (45-75)	−23
23F	93% (89-96)^d	100% (94-100)	+7	0% (0-15)	−93
1	85% (68-93)	88% (68-96)	+3	52% (37-69)	−33
3	65% (44-78)	93% (78-97)	+28	47% (36-55)	−18
5	79% (71-85)^d	79% (68-85)	0	60% (52-67)	−19
6A	93% (82-97)	88% (70-100)	−5	64% (46-72)	−29
7F	91% (82-96)	91% (81-96)	0	73% (58-84)	−18
19A	83% (74-89)	84% (71-100)	+1	59% (30-68)	−24

CI, confidence interval; PCV13, 13-valent pneumococcal conjugate vaccine; PCV20, 20-valent pneumococcal conjugate vaccine; V114, 15-valent pneumococcal conjugate vaccine.

^aSource: Savulescu et al. [2].

^bDifferences are predicted VE minus observed VE.

^cVE for serotype 6B could not be calculated, as described in Methods and Discussion sections.

^dIn Savulescu et al., effectiveness estimates were not available for serotypes 4, 18C, 23F, and 5, so the overall estimate for ‘PCV7’ was used for serotypes 4, 18C, and 23F, and the estimate for ‘PCV13non7’ was used for serotype 5 (quotes indicate the labels in the original publication).

3.3. Relative vaccine effectiveness

Figure 1 shows the summary results of 5,000 simulations of the absolute difference between the predicted VE values for V114 and PCV20 for the PCV13 serotypes. For all serotypes, the difference was positive, favoring V114. The results are further summarized in the Discussion.

Figure 1. Absolute difference in predicted vaccine effectiveness for V114 and PCV20.

PCV20, 20-valent pneumococcal conjugate vaccine; V114, 15-valent pneumococcal conjugate vaccine.

The solid points are the median values, over 5,000 simulations, of the predicted VEs (%) for V114 minus the (respective) VEs for PCV20. Values greater than zero favor V114. The red dashed line indicates the median difference (20.9). The whiskers are 95% CIs calculated using the empirical 2.5%ile and 97.5%iles of the differences.

*The predicted difference for serotype 23F is partly due to relatively low values and high variability of PCV20 titers, as explained further in the manuscript the VE was bounded at 0. CIs were bounded at 100%.

4. Discussion

In this study, a previously qualified method [11,13] was applied to predict the serotype-specific VE values of V114 (a new 15-valent vaccine) and of PCV20 (a new 20-valent vaccine) in an EU/R/A pediatric population using the post-primary series timepoint. This analysis predicted that V114 will have effectiveness comparable to that of PCV13 against 11 of the 12 shared serotypes VE was predicted for, with improved effectiveness against serotype 3 (93% for V114 vs. 65% for PCV13). PCV20 was predicted to have decreased effectiveness compared to PCV13 for 9 of the 12 shared serotypes VE was predicted for. The predicted VE was higher for V114 than PCV20 for all shared serotypes. (Because this is a post-hoc analysis using a model-based meta-analysis, it did not include formal, prespecified statistical hypothesis testing.) The decreases in PCV20 VE values (relative to PCV13) were driven by PCV20 generating lower IgG concentrations than PCV13, with an average GMC decrease of 46% overall (Supplementary Table S5).

A previous publication [11] qualified the methods applied here by demonstrating that it is possible to reliably predict the VE of novel vaccines for serotypes shared with PCV13. A related manuscript [13] then used the methods to predict the VE of the first new pediatric vaccine to market, V114. The method has been applied here to predict the VEs of PCV20 and V114 for a 2 + 1 recommended dosing schedule in infants; while a previous manuscript [25] applied the method to these same vaccines but in populations with a 3 + 1 recommended dosing schedule.

The present work has provided substantial insights into the potential strengths and challenges of predicting VE for pneumococcal vaccination in infants. First, it is critical to identify appropriate inputs for this type of modeling. Models and predictions are limited by the available data, require that assays be appropriately normalized, and require that populations are generally similar in terms of age, geography, etc. For example, changing geography and the time period can, for IPD, change background (pre-vaccination and placebo) immunity for each serotype. Changes in background immunity may result in partial or full protection of placebo subjects, with additional potential cross-protection from circulating vaccine or non-vaccine serotypes. While changes in the placebo population do not impact correlates of risk, they impact both apparent efficacy and effectiveness. Specifically, Andrews et al. [15], used a similar method to ours to predict C_ps and VEs for PCV13. During the assessment of potential input data sources for the analyses presented here, the input data sources used in Andrews et al. were considered but were not selected based on the considerations outlined above. Further, comparisons of this analysis to Andrews et al. must be made cautiously given, for example, the differences in geography and demographics which may have resulted in differences in exposure (and concomitant differences in titers and protection) of unvaccinated infants.

Second, the fact that antibody concentration is a CoP (accepted even by regulatory agencies) [14] means that antibody concentration can be used for making predictions (for the observation period of interest here) even though the input data come from different sources, differing in particular with respect to number of doses. For example, the titer data used in building the model for this analysis is from infants who received zero (placebo) or two doses of PCV, whereas the available VE data include infants with zero, one, two, or three doses of PCV, reflecting the overall population of children under five years of age. However, CoPs are defined as a measurement (here antibody concentration) that predicts protection (reduction of an individual’s risk) regardless of how many doses of vaccine have been received (i.e. even apply to placebo or those partially vaccinated, and typically also even for different vaccines). Thus, the connection between titer and VE that the model quantifies does not depend on the number of doses received.

Third, the immunobridging studies (comparing to PCV13) for V114 and for PCV20 measured IgG GMCs at two timepoints- one-month post-primary series (i.e. infants at five months of age), and one-month post-toddler dose (i.e. toddlers aged anywhere from 13–16 months). The post-primary series time point was selected as titer input for several reasons, primarily because this time point is most sensitive for detecting differences between vaccines and has been shown to predict future protection [26,27]. This sensitivity is because infants <1 year old have immature immune systems as compared to toddlers and older children, and thus are at higher risk for IPD and its complications. As a result, the bar is highest for this age group in terms of achieving vaccine-induced antibodies sufficient to prevent disease.

An additional factor that led to the selection of post-primary series titers is the availability of suitable placebo titer data. This analysis uses placebo titer data from 2-month-old infants. The method is impacted by titers from placebo data, i.e. those from unvaccinated children. Due to both exposure and immune system maturity, it is not tenable to assume that the available placebo IgG GMCs from 2-month-old infants are similar to unvaccinated toddlers. Given the pervasiveness of infant PCV vaccination programs throughout the world, there are not enough post-toddler series IgG GMC placebo data available to enable an analysis using post-toddler series IgG GMCs. Thus, the post-primary series titers are used.

Fourth, a potentially unfamiliar phenomenon is that the VE depends both on the GMC and the variability in titer. A vaccine inducing a very narrowly distributed titer with GMC just above the C_p can be 100% effective. If that GMC is the same, but the variance is larger, many subjects will have titers below C_p and the effectiveness can drop to nearly 50% (with no change in GMC). This phenomenon can also produce a lower VE even when GMC and variability are both higher, as is seen comparing GMCs and VEs (and the respective RCDCs and C_ps) for serotypes 18C, 1, and 19A. Furthermore, it is possible for variability in source data (due to, for example, sample size limitations) to result in challenges in interpreting VE predictions. This circumstance occurred for serotype 6B. Here, as mentioned earlier in section 2.2, the PCV13 titers for 6B are low, and the observed VE for 6B comes from data combined with those for 6A. The resulting titer distribution-VE combination makes it impossible to reliably estimate the C_p (and, hence, the VE) for 6B. This is a unique circumstance for 6B and highlights the reliance of the method on the precision of the input data.

Fifth, in the modeling of VE for new PCVs, it must be recognized that serotype 3 is unique in terms of its substantial geographic variability in epidemiology and VE. The relatively high effectiveness of PCV13 against serotype 3 observed in the US (79.5%) [28] and EU (71.8%) [2] may not be reflective of the epidemiology in some regions of the world. The lower VE for serotype 3 estimated by Andrews et al. [15] may be closer to the true VE of PCV13 against serotype 3 given the persistence of serotype 3 IPD in current epidemiology.

Sixth, the C_p predicted with this methodology are for IPD. It is understood that the C_ps for mucosal disease and carriage are higher [29,30]. Although predictions here are of VE for IPD, these data may shine light on how well novel vaccines may protect against mucosal diseases and carriage when comparing the IgG GMC concentrations elicited by the vaccines to the estimated C_ps. A vaccine that elicits a response 2- or 3-fold higher than the predicted C_p for many serotypes may protect significantly better against mucosal disease than a vaccine that elicits a lesser response.

Finally, VEs of novel PCVs need to be confirmed with real-world data. In the interim, despite its limitations, this method offers a perspective on how novel vaccines may perform in the broader population. This perspective is needed in an environment that has mature infant PCV immunization programs, where establishing vaccine efficacy in clinical trials is not possible and real-world effectiveness data will take years to generate due to the overall low incidence of IPD. New pneumococcal vaccines are tested in clinical trials that rely on immunobridging through multiple ‘bridges’ to the original PCV7 efficacy trials, and it may be increasingly challenging to solely rely on bridged data for regulatory and National Immunization Technical Advisory Group decisions. The VE prediction method used here provides an additional tool for the decision makers’ toolbox when evaluating new pneumococcal vaccines.

5. Limitations

Our modeling method is subject to several limitations. In addition to the methodological and data limitations described in previous publications [11,13,25] and discussed above, the VE predictions using this method are specific to the population and dosing schedule that were used for the VE and GMC inputs. As such, results presented here should not be extrapolated to, for example, other disease outcomes or populations with substantially different force of infection and/or and vaccination coverage rate. Another limitation is that the modeling method cannot make predictions for novel serotypes for which effectiveness data are not yet available, i.e. serotypes 22F and 33F in V114 and serotypes 8, 10A, 11A, 12F, and 15B in PCV20.

6. Conclusions

In conclusion, modeling can predict, for a 2 + 1 regimen, serotype-specific VE values for the serotypes V114 and PCV20 share with PCV13. V114 is predicted to have substantially higher protection against pneumococcal serotype 3, a key contributor to the burden of disease and health economics of IPD. It is also predicted to have similar protection to PCV13 for 11 of the other serotypes in common. For PCV20, VE is predicted to decrease across the serotypes shared with PCV13.

Declaration of interest

All authors were employees of Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, NJ, U.S.A. at the time of the study. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or material discussed in the manuscript apart from those disclosed.

Reviewer disclosures

Peer reviewers on this manuscript have received an honorarium for their review work. A reviewer on this manuscript has disclosed that they are a member of the WHO SAGE Working Group on Pneumococcal Vaccines, the JCVI Pneumococcal Vaccines Working Group and the Standing Committee on Vaccination.

Author contributions

Conception and design: JR, JRS, NB, TW, JW. Analysis and interpretation of the data: JR, JRS, NB, TW, MA, KLY, JW. Drafting the manuscript: JR, NB, TW, KLY, JW. Revising the manuscript for intellectual content: JR, JRS, NB, TW, MA, KLY, JW. All authors approve the final version of the manuscript and agree to be accountable for all aspects of the work.

Acknowledgments

The authors thank Melissa Stauffer, PhD, in collaboration with ScribCo, for medical writing assistance.

Data availability statement

The authors confirm that the data supporting the findings of this study are available within the article or its supplementary materials.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/14760584.2024.2335323

Additional information

Funding

This study was funded by Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, NJ, USA.