Streptococcus pneumoniae (pneumococcus) is a major cause of morbidity and mortality among children globally [1]. Prior to the introduction of the seven-valent pneumococcal conjugate vaccine (PCV7) in 2000 in the United States, the incidence of invasive pneumococcal disease (IPD) among U.S. children aged < 5 years was 98.7 per 100,000 cases [2]. Implementation of PCV7 was associated with reductions in the burden of vaccine-type IPD (VT-IPD) followed by increases in disease caused by serotype 19A and other non-PCV7 serotypes [3,4]. In 2010, the 13-valent pneumococcal conjugate vaccine (PCV13) replaced PCV7 in the routine childhood immunization schedule, and a supplemental dose of PCV13 was recommended for children aged 14–59 months who had received four doses of PCV7 [5]. In 2018–2019, the incidence of IPD among children aged < 5 years in the United States was 7.2 cases per 100,000 [6].

PCV13 was licensed using safety and immunogenicity endpoints only; thus, post-licensure vaccine effectiveness (VE) studies are critical to generate estimates of protection conferred by PCV13 against IPD and monitor vaccine protection under real-world conditions of use [7]. A post-licensure matched case-control study conducted in the United States that enrolled IPD cases and population-based healthy controls, estimated that the VE of ≥ 1 doses of PCV13 against IPD caused by PCV13-type serotypes was 85 % [8]. However, this study was conducted shortly after PCV13 introduction when many children only received one catch-up PCV13 dose as opposed to the recommended four-dose series of PCV13 (3 + 1): three primary doses before 12 months of age followed by a booster dose at 12–15 months [9]. Additionally, protection conferred from PCV13 differs by serotype, and may be lower for serotypes 3 and 19A, although the reported estimates have varied [8,10–12]. Serotype 3 has a high invasive capacity and propensity to cause severe infections; therefore, real-world serotype 3 vaccine effectiveness data remains important to monitor for establishing PCV13 impact on invasive disease [13].

Using additional years of data available post-PCV13 introduction in U.S. children, we sought to estimate effectiveness of the protection conferred by ≥ 3 doses of PCV13 against IPD caused by VT serotypes among children aged < 5 years in the United States. We additionally assessed VE by PCV13 dosing schedule and individual serotype among the most common circulating VT serotypes.

The study included 1,161 children aged 2–59 months with IPD, including 524 children enrolled from a previous post-licensure case-control study of PCV13 VE conducted during the early PCV13 implementation period and 637 children identified through routine surveillance during the late PCV13 implementation period; 61 were excluded because vaccination history could not be verified (Fig. 1; Figure S1). For the full study period, there were 223 cases of IPD caused by PCV13 VT serotypes (VT-cases), and 938 cases of IPD caused by NVT serotypes (NVT-controls) (Table 1; Figure S2).

During the full study period, three VT serotypes (19A, 3, and 19F) were responsible for 90.1 % (201/223) of VT-IPD; serotype 19A was the most common VT-serotype causing IPD, responsible for 43.0 % (96/223) of VT-IPD throughout the study period, followed by serotype 3 and serotype 19F (Figure S2, Table S2). Whereas serotype 19F was responsible for 3.8 % during the early implementation period, 19F was responsible for 34.4 % of cases of VT-IPD during the late PCV13 implementation period (Fig. 2). The majority of NVT disease was caused by serotypes 22F and 33F (Figure S2; Table S3).

Among the 1,161 participants included, 176 VT-cases (78.9 %; 176/223) and 885 NVT-controls (94.3 %; 885/938) had received ≥ 1 dose of PCV13 regardless of prior PCV7 receipt; 47 VT-cases (21.1 %; 47/223) and 53 NVT-controls (5.7 %; 53/938) had received no PCV doses (Table 2; Table S4). For the primary analysis assessing VE of ≥ 3 PCV13 doses, we excluded 353 participants (68 VT cases and 285 NVT-controls) who received 1 or 2 doses of PCV13. Therefore, among the 808 participants included in the primary analysis (297 during the early PCV13 implementation period and 511 during the late PCV13 implementation period), 108 VT-cases and 600 NVT-controls had received ≥ 3 doses of PCV13 (Table 2).

For the full study period, effectiveness against VT-IPD afforded by ≥ 3 doses of PCV13 versus no doses of any pneumococcal vaccine was estimated to be 90.2 % (95 % confidence interval: 75.4–96.1 %) (Table 2). Point estimates of PCV13 VE for ≥ 3 doses were higher in the late PCV13 implementation period (90.5 % [68.1–97.2 %]) as compared to the early PCV13 implementation period (86.5 % [28.0–97.3 %]); however, the confidence intervals overlapped. For the early and late PCV13 implementation period, upon exclusion of VR serotypes from the analysis, VE of ≥ 3 doses of PCV13 was estimated to be 90.6 % (32.5–98.6 %) and 88.0 (57.6–96.5 %), respectively. VE of ≥ 3 doses for the full study period excluding children with at least one immunocompromising condition was 88.0 % (74.9–94.3 %). VE for ≥ 1 dose for the full study period was 81.7 % (69.1–89.1 %). In a sensitivity analysis including participants who received 1–2 doses in a separate exposure category, VE estimates were similar (90.2 % [75.4–96.1 %]).

VE for a single primary dose of PCV13 (1 + 0), two primary doses of PCV13 (2 + 0), and three primary does of PCV13 (3 + 0) were 58.3 % (9.7–80.3 %), 58.2 % (−35.1–86.8 %), and 87.6 % (67.9–95.2 %), respectively (Table 2). VE among participants who received three primary doses and a booster dose (3 + 1 schedule) or two primary doses and a booster (2 + 1) was 92.4 % (78.2–97.2 %) and 99.9 % (78.2–100.0 %), respectively.

For the full study period, ≥3 doses of PCV13 were highly effective for serotype 19A (86.8 % [73.7–93.3 %]), serotype 7F (94.5 % [13.5–99.6 %]), and serotype 19F (93.8 % [69.8–98.8 %]) (Table 3). However, VE for ≥ 3 doses of PCV13 against serotype 3 had a lower point estimate and was not statistically significant: 50.2 % (− 28.4–80.5 %). Owing to limited number of cases during the early PCV13 implementation period VE of ≥ 1 PCV13 dose against serotype 19F could not be estimated; however, VE of ≥ 1 PCV13 dose against 19F during the late PCV13 implementation period was 94.1 % (80.1–98.2 %) (Table 4).

Receipt of ≥ 3 doses of PCV13 conferred strong protection against VT-IPD among U.S. children aged 2–59 months. This aligns with recent estimates of PCV13 effectiveness from a multi-center indirect cohort study conducted in Europe during a similar time period, 2012–2018 [11] and the previously published case-control study in the United States [8]. VE increased with increasing number of primary doses among schedules without a booster although the confidence intervals overlapped, and the effectiveness of schedules with a booster dose (2 + 1 and 3 + 1) were both > 90 %. During the full study period, three VT serotypes (19A, 3, and 19F) were responsible for the majority of VT-IPD; we estimated significant VE for serotypes 19A, 19F, and 7F, but VE was not significant for serotype 3. Non-PCV13 types 22F and 33F were the most common NVTs contributing to IPD, both of which are included in higher valency PCVs [6,19].

While this study was not adequately powered to address serotype-specific VE for all VT serotypes, we did not find evidence of significant PCV13 protection against serotype 3 disease, adding to the real-world evidence suggesting PCV13 may not provide adequate protection against disease caused by this serotype [20]. In clinical trials, the serotype 3 immune response was lower than the other additional five PCV13 serotypes (i.e., serotypes included in PCV13 but not PCV7) [21]. Serotype 3 has unique capsular properties compared to other pneumococcal serotypes including a thick capsule, high invasive capacity, and resistance to opsonic killing, all of which may contribute to lower VE [22–25]. As higher valency and PCVs using new or different technology are implemented in the United States, it will be important to continue to monitor and evaluate VE against serotype 3 [19].

Cases of serotype 19F, a serotype contained in both PCV7 and PCV13, were observed at a higher frequency during the later study period as compared to the early study period among both recipients and nonrecipients of PCV13. Comparative estimates of 19F VE by study period are not available owing to the small number of 19F cases during the early PCV13 implementation period; however, VE of ≥ 3 doses against 19F estimated from the current study for the full study period (93 %) was similar to estimates from a U.S.-based case-control study conducted during the PCV7 era (87 %), and a recent multi-center indirect cohort study in Europe (84 %) [3,11]. However, evidence of persistent 19F carriage among children throughout the PCV13 era has been well documented across various settings despite early reports from a trial indicating PCV13 reduced 19F carriage [26–28]. Additionally, some immunogenicity studies have demonstrated lower 19F immunogenicity following PCV13 than PCV7 after the primary infant doses [29,30]. These data indicate that PCV13 provides direct protection against serotype 19F disease. However, protection against carriage is less complete. Thus, indirect effects against serotype 19F IPD could have decreased post-PCV13 introduction compared with post-PCV7 introduction. If so, this could explain the increase in 19F disease during the late PCV13 implementation period, particularly among the unvaccinated.

The majority of children enrolled in this study received four doses of PCV13 in a 3 + 1 schedule, in accordance with U.S. recommendations [31]. As such, we were not able to evaluate VE for reduced dosing schedules (e.g., 1 + 1), which have garnered global interest particularly in countries with mature PCV programs [32]. VE increased in schedules without booster doses with increasing number of primary doses, although the confidence intervals overlapped likely due to the small number of children who received PCV13 based on 1 + 0 or 2 + 0 schedules. Larger studies are needed to evaluate the impact of VE by number of primary doses. However, growing evidence regarding the two World Health Organization-recommended PCV schedules (2 + 1 or 3 + 0) suggests that administration of the booster dose is what provides durable protection leading to stronger protection among the unvaccinated population [33]. Point estimates from our schedule-specific analysis suggest VE was > 90 % for both schedules with a booster dose (2 + 1, 3 + 1). Given that the United States is one of the few countries in the world that continues to recommend a 3 + 1 PCV schedule, additional evidence that directly compares VE across vaccine schedules would help inform the discussion on a possible transition to a reduced-dose PCV schedule in United States in the future.

VE estimates of ≥ 1 PCV13 dose produced for the early PCV13 implementation period from our study (69.7 % [33.0–86.2 %]) were not statistically different from estimates produced in a previous case-control study (86 % [75.5–92.3 %]) that enrolled matched, population-based controls [8]. Point estimates of VE from the indirect cohort study were lower than the case control study; this is consistent with the expectation that case-control studies may bias VE estimates due to differential healthcare seeking behavior between cases and population-based controls [34]. In other words, vaccination coverage among controls may not represent the source population if enrolled controls are more likely to seek vaccination than the source population, resulting in biased overestimates of VE. The indirect cohort method mitigates potential selection biases that may arise due to differences in the characteristics between cases and controls as both are identified using the same case definition [15,35].

This study has limitations. First, while we attempted to adjust for confounders that may bias the causal pathway between pneumococcal vaccination status and IPD using a multivariate model, the possibility of unmeasured confounding persists. Second, inclusion of cases with VR serotypes as NVT, for which PCVs may confer some protection, may result in underestimates of VE [11,36]. Indeed, sensitivity analyses which excluded VR serotypes resulted in higher point estimates of VE, though differences were not meaningfully different from models that included the VR serotypes. Additionally, the indirect cohort method relies on the assumption that implementation of vaccination does not influence the risk of NVT-IPD. However, given that the US has not observed significant serotype replacement following PCV13 introduction, this assumption is likely not violated [37]. Third, exposure misclassification is possible despite robust efforts by surveillance sites to obtain detailed vaccination history. Fourth, while this is one of the largest published studies of pediatric PCV13 vaccine effectiveness in the United States as of writing to our knowledge, owing to the success of PCV implementation, IPD case numbers were too small to detect significant effects for some stratified analyses. Lastly, we were unable to estimate effectiveness of alternative dosing schedules such as one primary dose with a booster dose (1 + 1) as the majority of vaccinated children had received a four-dose schedule, in accordance with U.S. recommendations [9].

In summary, ≥3 doses of PCV13 continues to be effective in preventing VT-IPD among children aged < 5 years in the United States, although protection against serotype 3 IPD was non-significant. Indirect cohort studies are an efficient platform in a setting with robust IPD surveillance to both estimate VE and mitigate potential bias from healthcare seeking behaviors influencing estimates from traditional case-control studies. As new PCVs become available for use in children, additional data are needed to monitor effectiveness including serotype-specific protection and differences by dosing schedule.