Serotype specific pneumococcal vaccine effectiveness in children with sickle cell disease: A two-decade analysis

Abstract

Objectives:

Sickle cell disease (SCD) is the most common genetic hematologic disease globally and children with SCD are at increased risk for pneumococcal disease.

Methods:

We utilized data from population-based enhanced surveillance for invasive pneumococcal disease (IPD) in children <18 years of age in Massachusetts from 2002 to 2020. We calculated incidence rates (IR) among children with SCD using bootstrapping resampling and incidence rate ratios (IRR) for pre- and post-PCV13 periods. Vaccine effectiveness (VE) was calculated as 100*(1-IRR), and PCV13 vaccine failure probability was predicted using a random forest model.

Results:

Children with SCD had higher IR during both pre− /post-PCV13 periods compared with otherwise healthy children 240.0/100,000 versus 4.6/100,000 in pre-PCV13 period (2002–2009); 172.7/100,000 versus 1.9/100,000 in post-PCV13 period (2011− 2020), respectively. After widespread use of PCV7 for a decade, a modest reduction of 28.1 % (95% CI 25.9–37.2%) in the incidence of overall IPD during the post-PCV13 period was observed in children with SCD, whereas a more substantial 59.5% (96% CI 57.8–61.4%) reduction was observed in otherwise healthy children. There was a 60.8% (95% CI 55.2%-NA) reduction in the incidence of VST13 IPD in children with SCD and an 83.0% (95% CI 80.67–85.63%) reduction in children without underlying health condition. Overall, 61.1% of the remaining IPD among children with SCD were due to non-PCV13 serotypes (8, 10A, 15A,15B, 22F, 23B), many of which are included in expanded valency vaccines.

Conclusion:

Children with SCD continue to have higher rates of IPD compared with otherwise healthy children despite vaccination. Majority of the remaining disease is due to serotypes not included in vaccine formulations that have been used for the last two decades. Our study highlights the potential value of expanded valency vaccines and importance of risk-based vaccination strategies tailored for this vulnerable population.

1. Introduction

Streptococcus pneumoniae is a major cause of mortality and morbidity worldwide causing approximately 1 million global deaths annually [1–6]. Although over 100 serotypes have been identified so far, the majority of invasive pneumococcal disease (IPD) is caused by a limited number of those serotypes [7,8]. Pneumococcal colonization in the nasopharynx (NP) is a prerequisite for every invasive disease episode and is reported in up to 20 to 40% of healthy children [5,9].

Pneumococcal conjugate vaccines (PCVs) have been widely effective in both vaccinated and unvaccinated populations, with substantial reduction in colonization, transmission, and invasive disease caused by the serotypes included in the vaccine [10]. Seven valent pneumococcal conjugate vaccine (PCV7) was approved in the United States in 2000 and targeted serotypes 4, 6B, 9V, 14, 18C, 19F, and 23F. Since then, expanded valency conjugated vaccines, including PCV13 (in 2010, PCV7 serotypes plus 1, 3, 5, 6 A, 7F, and 19A), PCV15 (in 2022, PCV13 serotypes plus 22F and 33F) [11] and PCV20 (in 2023, PCV15 serotypes plus 8, 10A, 11A, 12F, and 15B) have been approved [12]. New vaccine candidates in the pipeline, PCV21/PCV24, include additional serotypes 2, 9N, 15A, 15C, 16F, 17F, 20A, 20B, 23A, 23B, 24F, 31 and 35B.

Sickle cell disease (SCD) is the most common genetic hematologic disease globally [13]. In the United States, SCD disproportionately affects African Americans, occurring in 1 out of every 365 births [14]. Individuals with SCD are at increased risk for invasive disease with pneumococci due to multiple immunologic defects such as functional/anatomic asplenia, impaired complement activation, opsonization and phagocytosis, and skewed T-helper immune response [14–17]. S. pneumoniae was the leading infectious cause of mortality among SCD patients <20 years of age in the United States before implementation of penicillin prophylaxis and pneumococcal vaccination [18,19]. Even in high-income settings with high vaccine coverage, children with SCD remain at approximately 50 times higher risk of developing IPD [2,20]. In addition to the PCV vaccination schedule recommended by the Advisory Committee on Immunization Practices (ACIP) for otherwise healthy children, those with SCD are recommended to receive an additional dose of 23-valent pneumococcal polysaccharide vaccine (PPSV23) starting at two years of age and a second dose five years following the first, or one dose of PCV20 at least eight weeks after the last dose of PCV13 or PCV15 [21–23]. Even with vaccination, children with SCD remain at an increased risk of developing IPD and, in general, do not generate as strong a response to the PCV vaccines compared to healthy children [2,24]. Here we report the impact of existing and future candidate conjugated vaccines by assessing their effectiveness against IPD in children with and without SCD in a high-income setting with high vaccine coverage using a 3 + 1 schedule.

2. Methods

2.1. Data preparation

A population-based enhanced passive surveillance for IPD in children in the state of Massachusetts was initiated in October 2001 [25]. A case of IPD was defined as isolation of S. pneumoniae from a normally sterile site from a Massachusetts resident <18 years of age. All clinical microbiology laboratories in Massachusetts submit isolates of S. pneumoniae from blood, cerebrospinal fluid, or other normally sterile body fluids to Massachusetts Department of Public Health (MDPH). Demographic and clinical data, including immunization records, are confirmed through follow-up phone interviews with each case’s primary care provider or parents/guardians by epidemiologists. All isolates are then sent to the Maxwell Finland Laboratory for Infectious Diseases where identification is confirmed using standard microbiological methods according to guidelines of Clinical and Laboratory Standards Institute [26], and serotyping is performed using antisera from Statens Serum Institute (Copenhagen, Denmark) [27].

A total of 1306 cases were identified between January 01, 2002, and December 31, 2020. Out of the total, 4 cases were excluded for missing gender information. This study evaluated 1302 cases where race/ethnicity, age, gender, vaccine status, underlying health condition, year of diagnosis, and disease serotype were recorded for each case. Study period was divided into two: (i) pre-PCV13 period, 2002 to 2009, (ii) post-PCV13 period, 2011 to 2020. We excluded cases identified in 2010 from the analysis to mitigate potential biases arising from the transition period between PCV7 and PCV13.

2.2. Sickle cell disease data and estimating the number of SCD population at risk

We used SCD data obtained from the New England Newborn Screening Program and estimated mortality among those with SCD rates published in the literature [28–31]. To estimate the number of cases of SCD in children in Massachusetts during the study period, we employed a previously described methodology using ExpertFit [32]. We identified the binomial distribution as the best fit for the SCD data. The maximum likelihood estimates (MLE) for the parameters of the binomial distribution were determined to be N = 169, representing the number of trials, and p = 0.2071, indicating the probability of success in each trial. These estimates were derived based on the observed data and reflect the values that maximize the likelihood function for the binomial model. We used this distribution to simulate the number of newborns diagnosed with SCD each year from 1984 to 1999. Taking the sum of the simulated count of yearly newborns and the SCD data we received, we obtained the number of children with SCD in the state of Massachusetts under the age of 18 each year during 2002–2020 period.

2.3. Pneumococcal serotypes

We categorized serotypes into four distinct sets based on vaccine serotypes (VST): i) VST13 includes serotypes in PCV13 and 6C (1, 3, 4, 5, 6A, 6B, 6C, 7F, 9V, 14, 18C, 19A, 19F, and 23F), ii) VST7 includes serotypes in PCV7 (4, 6B, 9V, 14, 18C, 19F, and 23F), iii) VST13–7 includes the additional 6 serotypes in PCV13 but not those in PCV7 and 6C, and iv) VST20 includes serotypes in PCV20 (1, 3, 4, 5, 6A, 6B, 7F, 8, 9V, 10A, 11A, 12F, 14, 15B, 18C, 19A, 19F, 22F, 23F, and 33F). Note that 6C was included in the analysis of PCV13 due to the cross-protection of PCV13 on 6C [33].

2.4. Outcome metrics

The incidence rate (IR) was defined both annually and for the study periods (pre- and post-PCV13 periods). IR was calculated using bootstrapping resampling. Specifically, 10,000 random samples were drawn from the original data with replacement. For each sample, the IR was computed by dividing the number of IPD cases caused by a specific serotype or a group of serotypes by the total population at risk. The IR, along with its 95% confidence interval, was estimated based on the set of all sample IRs. See Supplemental Material Fig 1 and 2 for bootstrapping results of incidence rates. Averted (prevented) cases was estimated by multiplying the difference of pre-PCV13 IR and post-PCV13 IR by the total population at risk during the post-PCV13 period. The incidence rate ratio (IRR) was computed as the ratio of the pre-PCV13 IR and post-PCV13 IR. Vaccine effectiveness (VE) was calculated as 100* (1 minus IRR).

PCVx failure (x = 7, 13) was defined as the diagnosis of VSTx IPD in a child who completed the recommended PCVx vaccination regimen. Completeness of vaccination is defined according to the recommendation of ACIP [22]. A child is considered completely vaccinated if they have received the required PCVx doses at least 14 days before the IPD diagnosis. Given the exclusion of the transition year 2010 from the analysis, we assumed that children diagnosed with IPD beginning 2011 only benefited from the impact of PCV13. To evaluate the proportion of PCVx failure (x = 7, 13), two metrics were calculated using the bootstrap method (see Supplemental Material Fig. 3 for bootstrapping results):

$$VSTxIPDProportion=\frac{ThenumberofpatientswithVSTxIPD}{ThetotalnumberofIPDcasesduringPCVxperiod}$$ (1)

$$VSTxIPDFailureProportion=\frac{ThenumberofpatientswithVSTxIPD}{ThetotalnumberofIPDcaseswithcompletevaccinationduringPCVxperiod}$$ (2)

All outcome metrics were estimated using a bootstrapping approach, and the results were visualized using boxplots (refer to Supplemental Material). The boxplots exhibit symmetry around the median with minimal outliers, suggesting that the outcome metrics are stable and robust.

2.5. Prediction of probability of IPD caused by PCV13 serotypes

2.5.1. Data preprocessing

Of the 1302 cases initially included in the study, 270 lacked available isolates for serotyping and were therefore excluded from the predictive analysis. The final dataset used for model development thus comprised 1032 cases with complete serotype information. Underlying comorbidities (with, without, unknown), categorized according to the 2021 Report of the Committee on Infectious Diseases, were transformed into the numerical encoding of 0 – unknown (N = 94), 1 – with underlying condition (N = 220), and 2 – without underlying condition (N = 718) [23]. Similarly, vaccine status was also transformed into the numerical encoding of 0 – incomplete (N = 414), 1 – complete (N = 587), and 2 – unknown (N = 30). Serotypes were categorized as binary outcome variables based on their presence in VST13 IPD. This encoding is a standard preprocessing step required by many machine learning algorithms, which cannot directly process categorical text inputs, while ensuring that the categorical nature of these variables is preserved. To address the imbalanced distribution of the data, the Random Over Sampling (ROS) method was employed [34]. With ROS, cases were chosen randomly with replacements from the minority group and duplicated to augment the size of the minority group. After resampling, the size of the two groups, VST13 IPD and non-VST13 IPD, were evenly distributed, with 740 cases in each group.

2.5.2. Prediction model

Several machine learning models, including XGBoost, logistic regression, and tree-based algorithms, were developed and evaluated to predict PCV13 vaccine failure. Among these, Random Forest demonstrated the best predictive performance. Detailed performance metrics for each model are provided in the Supplemental Material. Random Forest (RF) was subsequently utilized to predict PCV13 vaccine failure, employing a replicative approach [35]. The number of replications was set to be 10. Within each replication, 90% of the data was randomly selected to train the model, while another 10% of the data was randomly selected to make predictions and test the performance of the model. Moreover, fine tuning method was applied to each replication to optimize essential parameters, details of which can be found in the Supplemental Material file for further reference [36].

2.5.3. Predictive performance

We trained our model using 90% of the dataset and assessed its performance using the remaining unseen/blinded data. Confidence intervals were applied using the 10 replications, and the probability of vaccine failure was reported with a 95% confidence level. We assessed the predictive efficacy of these models by using model accuracy as detailed in the Supplemental Materials. Alongside the model’s predictions for vaccine failure probabilities, we presented the observed prevalence of cases with vaccine failure.

3. Results

In total, there were 1306 IPD cases during the study period among children less than 18 years of age. To uphold the precision of our analyses and mitigate potential biases in our predictions, four cases were omitted from consideration due to missing or unknown gender information. The remaining 1302 cases comprise the study population for IR, IRR, and vaccine failure analysis. For the predictive model, another 270 cases were further excluded due to the lack of available isolates for serotyping.

3.1. Incidence rate (IR) and incidence rate ratio (IRR)

Incidence of IPD varied over the study period with highest IR recorded in 2009 at 8.7/100,000 (95% CI 7.2–10.1/100,000), and lowest IR observed in 2020 at 1.6/100,000 (95% CI 0.9–2.2/100,000) (Fig. 1). Substantial reductions in IPD rates were observed in post-PCV13 period with evident relative stability after the second year of implementation of the vaccine spanning 2013 to 2020, with the IPD rates fluctuating modestly within the range of 1.6/100,000 to 3.4/100,000. IPD rates in pre-PCV13 period and post-PCV13 period were 6.69/100,000 and 2.93/100,000, respectively reflecting a round 60% (IRR 44 %) additional reduction in post-PCV13 period after a decade of PCV7 use in Massachusetts (2000− 2010). The number of averted overall IPD cases was 519. There was an 80% reduction in rate of IPD caused by the serotypes included in PCV13 in post-PCV13 period (3.43/100,000 pre-PCV13 versus 0.68/100,000 post-PCV13; IRR 0.20) representing 379 averted VST13 IPD cases over a decade. Similar trends of VST13–7 IPD were observed with the IR of VST13 serotype IPD. The pre-PCV13 IR of VST13–7 IPD was 2.92, while the post-PCV13 IR of VST13–7 IPD was 0.56. The VST13–7 IRR was 0.19 and the number of averted VST13–7 IPD cases was 326.

Fig. 1.

Annual incidence rate for invasive pneumococcal disease of different serotype groups in children <18 years of age in pneumococcal conjugate vaccine period, Massachusetts, 2002 to 2020 (95% confidence intervals are depicted in error lines)*.

* PCV21–20 includes the 12 serotypes that are included in PCV21 but not PCV20.

Out of 1302 reported IPD cases, 299 (23%) cases were in children with at least one underlying condition and 24 cases were among children with SCD. Among children with SCD, higher incidence rates were observed during both pre- and post-PCV13 periods compared with otherwise healthy children 240.0/100,000 versus 4.6/100,000 in pre-PCV13 period; 172.7/100,000 versus 1.9/100,000 in post-PCV13 period, respectively (Table 1). To quantify the reduction in incidence after PCV13 implementation, percent reductions between pre- and post-PCV13 periods were estimated, with values in parentheses derived from the upper and lower bounds of the 95% confidence intervals for the incidence rates (Table 1). Despite high uptake of PCV7 and subsequently PCV13 in the pediatric, a modest reduction of 28.1% (25.9–37.2%) in the incidence of overall IPD during the post-PCV13 period was observed in children diagnosed with SCD, whereas a more substantial 59.5% (57.8–61.4%) reduction was observed in children without any underlying health conditions. Since SCD disproportionally impacts black individuals in the U.S. who historically had higher rates of IPD before PCVs and since all our SCD cases were black, we also estimated the reduction rates for otherwise healthy black and otherwise healthy non-black populations. The reductions in the incidence of IPD were similar in these two groups and higher than the SCD population (Table 1). The reduction in VST13 IPD observed for both children with SCD and children without any underlying condition was greater than the overall reduction in IPD. A 60.8% (55.2% - NA) reduction in the incidence of VST13 IPD in children with SCD and an 83.0% (80.7% - 85.6%) reduction in children without underlying health condition was observed. Of note otherwise healthy black children had the highest reduction in VST13 IPD in our study (93.5%, 90.2% - 103.3%) (Table 1). Overall, around 61 % of the remaining IPD among children with SCD were due to the non-PCV13 serotypes [8 (n = 1), 10A (n = 1), 15A (n = 1), 15B (n = 3), 22F (n = 4), 23B (n = 1), no isolates were available for serotyping in six cases] included in the expanded valency vaccines.

Table 1.

Pre-/Post-PCV13 incidence rates along with the 95 % confidence intervals and incidence rate ratios for children with sickle cell disease and children with no reported underlying condition by vaccine groups.

	Incidence rate		Incidence rate ratio Pre- vs Post-PCV13	% Reduction after PCV13 implementation*
	Pre-PCV13 period (2002–2009)	Post-PCV13 period (2011–2020)
Overall IPD
SCD**	240.00 (100.00, 360.00)	172.66 (62.78, 266.83)	0.72	28.01 % (25.88 %, 37.21 %)
No Underlying Condition	4.57 (4.28, 4.87)	1.85 (1.65, 2.06)	0.40	59.52 % (57.77 %, 61.41 %)
No Underlying Condition-Non-Black	4.41 (4.09, 4.73)	1.76 (1.54, 1.97)	0.40	60.04 % (58.35 %, 62.29 %)
No Underlying Condition-Black	6.51 (4.80, 8.11)	2.84 (1.81, 3.79)	0.44	56.40 % (53.32 %, 62.34 %)
VST13 IPD
SCD**	80.00 (0.00, 140.00)	31.39 (−15.70, 62.78)	0.39	60.76 % (55.15 %, NA)
No Underlying Condition	2.53 (2.27, 2.77)	0.43 (0.32, 0.54)	0.17	83.00 % (80.67 %, 85.63 %)
No Underlying Condition-Non-Black	2.41 (2.14, 2.67)	0.45 (0.33, 0.56)	0.19	81.26 % (79.00 %, 84.46 %)
No Underlying Condition-Black	4.00 (2.63, 5.26)	0.26 (−0.09, 0.52)	0.06	93.54 % (90.18 %, 103.27 %)

IPD Invasive pneumococcal disease; SCD: Sickle Cell Disease; VST13: Serotypes included in PCV13 plus 6A

^*The interval in parenthesis for % reduction represents the % reduction of the lower and upper bound of IRs.

^**All cases self identified race/ethnicity as black

3.2. PCV-13 vaccine failure overall

Both the VST7 IPD proportion and VST7 IPD failure proportion were significantly lower than the VST13 IPD proportion and VST13 IPD failure proportion, respectively (Table 2). During the post-PCV7 period (2002–2009), approximately two-thirds of the VST7 cases were vaccine failures. During the post-PCV13 period (2011–2020), almost all the VST13 IPD cases were vaccine failures.

Table 2.

Vaccine failure estimates and 95 % confidence intervals during PCV7 and PCV13 periods.

	PCV7 period (2002–2009)*	PCV13 period (2011–2020)**
All Children
Vaccine serotype disease proportion	0.0653 (0.0447, 0.0842)	0.190 (0.144, 0.233)
Vaccine serotype disease failure proportion	0.0462 (0.0245, 0.0652)	0.181 (0.124, 0.237)
Children with SCD
Vaccine serotype disease proportion	0.0909 (0.0447, 0.0842)	0.400 (0.000, 0.800)
Vaccine serotype disease failure proportion	0.000 (0.000, 0.000)	0.000 (0.000, 0.000)
Children without underlying condition
Vaccine serotype disease proportion	0.0673 (0.0424, 0.0898)	0.195 (0.137, 0.247)
Vaccine serotype disease failure proportion	0.0502 (0.0232, 0.0734)	0.195 (0.122, 0.260)

^*Vaccine serotype disease during PCV7 period include IPD caused by serotype 4, 6B, 9V, 14, 18C, 19F, and 23F;

^**Vaccine serotype disease during PCV13 period include IPD caused by serotype 1, 3, 4, 5, 6A, 6B, 6C, 7F, 9V, 14, 18C, 19A, 19F, and 23F

3.3. Predictive results: PCV-13 vaccine failure of subgroups

Table 3 presents the predicted probability of IPD from our Random Forest (RF) model. Our results were organized by subgrouping the population based on complete versus incomplete vaccination and considering for age, race, gender, and underlying conditions. Both the predicted and observed probabilities of PCV13 serotype disease were consistently higher among children with incomplete vaccination compared to those with complete vaccination across all subgroups. Among those with complete vaccination, children aged 5–18 year exhibited a higher predicted likelihood of vaccine failure, with a mean of 0.72 compared to 0.59 for children less than 2 years of age and 0.69 for children between 2 and 5 years of age. The highest predicted probabilities of vaccine failure (0.82 incomplete vaccination vs. 0.67 complete vaccination) were among Hispanic children. Due to the limited number of SCD cases in our dataset, the predicted probabilities for the SCD population lacked statistical significance and were therefore excluded from the table. The final prediction performance on the test dataset is summarized in Table 4.

Table 3.

Estimated probability of PCV13 serotype-specific IPD using random forest by vaccination status, underlying condition, and demographics. (Number of replications: 10).

Characteristics	Incomplete Vaccination			Complete Vaccination
	Predicted Probability of Vaccine Serotype Disease (95 % CI) ¥	Observed Probability of Vaccine Serotype Disease (95 % CI) ¥	Prediction Accuracy (%)	Predicted Probability of Vaccine Serotype Disease (95 % CI)	Observed Probability of Vaccine Serotype Disease (95 % CI)	Prediction Accuracy (%)
Age (years)
0- < 2 (N = 457)	0.59 (0.47, 0.71)	0.56 (0.45, 0.67)	68.62 %	0.59 (0.55, 0.63)	0.39 (0.33, 0.46)	66.04 %
2–5 (N = 289)	0.69 (0.60, 0.77)	0.59 (0.46, 0.72)	73.42 %	0.61 (0.49, 0.72)	0.54 (0.46, 0.61)	74.01 %
5–18 (N = 286)	0.72 (0.63, 0.81)	0.60 (0.55, 0.64)	72.67 %	0.45 (0.32, 0.58)	0.51 (0.44, 0.58)	54.03 %
Underlying Condition
Yes (N = 220)	0.71 (0.59, 0.83)	0.67 (0.58, 0.76)	70.21 %	0.66 (0.58, 0.74)	0.52 (0.46, 0.59)	62.82 %
No (N = 718)	0.68 (0.59, 0.77)	0.58 (0.51, 0.65)	74.91 %	0.58 (0.53, 0.63)	0.47 (0.45, 0.50)	66.74 %
Race/Ethnicity
White (N = 458)	0.67 (0.59, 0.76)	0.60 (0.53, 0.67)	74.09 %	0.61 (0.55, 0.66)	0.50 (0.44, 0.56)	67.36 %
Black (N = 124)	0.66 (0.46, 0.85)	0.69 (0.55, 0.82)	70.37 %	0.46 (0.29, 0.64)	0.47 (0.35, 0.60)	72.60 %
Hispanic (N = 115)	0.82 (0.60, 1.04)	0.59 (0.38, 0.80)	62.07 %	0.67 (0.55, 0.80)	0.36 (0.25, 0.47)	60.29 %
Asian (N = 52)	0.50 (0.11, 0.89)	0.65 (0.28, 1.01)	78.57 %	0.40 (0.21, 0.59)	0.44 (0.33, 0.55)	68.75 %
Sex
Female (N = 462)	0.72 (0.64, 0.81)	0.61 (0.52, 0.69)	76.24 %	0.52 (0.46, 0.58)	0.44 (0.39, 0.49)	68.27 %
Male (N = 668)	0.61 (0.50, 0.71)	0.57 (0.49, 0.65)	67.21 %	0.61 (0.57, 0.66)	0.48 (0.43, 0.52)	65.33 %

^¥The column signifies instances of PCV13 vaccine failure.

Table 4.

95 % confidence intervals (CI) of performance metrics on the test dataset.

	Mean	Lower 95 % CI	Upper 95 % CI
Accuracy	68.50 %	64.36 %	72.63 %
Sensitivity	77.39 %	72.43 %	82.35 %
Specificity	59.77 %	53.21 %	66.33 %
Precision	65.74 %	60.65 %	70.83 %
AUC	74.31 %	71.24 %	77.38 %

4. Discussion

In our study, overall IPD among individuals under 18 years of age exhibited a further decline of 56.2% during the post-PCV13 period (2011–2020) compared with the post-PCV7 period (2002–2009). The post-PCV13 incidence rates reduced to zero for serotypes 1, 4, 6B, 9V, 18C, and 23F, resulting in a 100% reduction rate. The incidence rate of VST13 IPD during the post-PCV13 period experienced a substantial 80.2% decrease from the post-PCV7 period, mirroring a similar trend in VST13–7 IPD, which decreased by approximately 81%. These findings align with studies conducted with similar cohorts in Canada [37], Sweden [38], Israel [39] and California [40] and underscore the considerable impact of the PCV13 vaccination on reducing IPD incidence, especially in serotype-specific cases.

However, the reduction in IPD incidence among children with SCD was less pronounced (~2-fold less) compared to otherwise healthy children. Adamkiewicz et al. [41] describes similar findings, identifying a significant reduction in IPD, meningitis, and death in pediatric populations aged 0–9 when comparing the post-PCV13 period to the post-PCV7 period. When comparing children with SCD to a race- and age-matched reference population during the same time periods, the increased likelihood of children with SCD developing IPD, contracting meningitis, and dying from IPD persisted [41].

Remarkably, in our study this reduction in otherwise healthy children surpassed by more than half the corresponding reduction observed in children with SCD. A similar trend was detected in VST13 IPD. The incidence rate of VST13 IPD in children with SCD decreased by 60.8% while the incidence rate of VST13 IPD in children without any underlying health condition decreased by 83.0% during the post-PCV13 period. Other studies have also identified that certain groups with underlying comorbid conditions such as SCD are at a greater risk of developing IPD and for vaccine failure [42–44]. These findings emphasize the nuanced impact of the PCV13 vaccination on different pediatric populations based on underlying health conditions, indicating the varying efficacy in specific subgroups. A systematic review focused on studies evaluating effectiveness of PCV-10 and PCV-13 also noted serotype 3 which is included in PCV13 was a specific cause for vaccine failure and breakthrough infections [45].

This study also assessed the vaccine effectiveness of both PCV7 and PCV13 by examining the proportion of serotype specific IPD cases and the associated vaccine failure rate. In the PCV7 period, 6.5% of patients were diagnosed with VST7 IPD, and among those who received complete PCV7 vaccination, 4.6% were diagnosed with VST7 IPD. Contrastingly, in the post-PCV13 period, 19.0% of all IPD patients were diagnosed with VST13 IPD, and among those with complete PCV13 vaccination, 18.1% were diagnosed with VST13 IPD. Notably, the vaccine failure rate of PCV13 was observed to be more than twice that of PCV7. These findings highlight a higher likelihood of vaccine failure associated with PCV13 in comparison to its predecessor, PCV7. As confirmed by other studies, this may be due to the fact that herd protection was more pronounced post-PCV7 [38] and that PCV13 serotypes 3 and 19A are increasing and causing most of the breakthrough IPD cases [46].

Additionally, this study includes a random forest model that was used to make prediction on the probability of VST13 serotype specific IPD in different subpopulation based on vaccination status, underlying condition, and demographics, including age, gender, and race/ethnicity. The prediction accuracy exceeded 60.0% for all subgroups, except for the subgroup of children aged 5 to 18 years. Across all subgroups, the average prediction accuracy was 68.5%. Our model predicted probabilities of VST13 IPD of different subgroups follow the trend of the observed probabilities but are slightly overestimated in most subgroups. In the existing body of literature, there is a scarcity of studies exploring model-based predictions related to vaccine failure. In alignment with our findings, particular research indicates a higher prevalence of PCV13 vaccine failure in older children as opposed to their younger counterparts [42].

One limitation of this study arises in the computation of the incidence rate for overall IPD and VST13 IPD in the pre-PCV13 and post-PCV13 periods for children without underlying health conditions. In this calculation, the entire population is utilized instead of specifically isolating the subpopulation of individuals without underlying health conditions. This approach may result in an underestimation of the true incidence rate. Another limitation arises with the group labeled PCV21–20 in Fig. 1 which encompasses the 12 additional serotypes present in PCV21 but absent in PCV. The results of this group do not directly imply the additional protection offered by PCV21 over PCV20, as PCV21 lacks certain serotypes that are included in PCV20. Other limitations include the rarity of cases of SCD and the fact that the data is only from one state. These limitations should be taken into consideration when interpreting the findings and extrapolating them to broader populations.

Given the ongoing development and implementation of higher-valency pneumococcal vaccines (PCV15, PCV20, PCV21, and PCV24), it is important to consider their potential impact on disease burden. Although we did not estimate the serotype coverage of PCV15 or PCV20 in this study, future analyses evaluating the potential impact of these higher-valency vaccines on IPD incidence—particularly among children with and without SCD—will be important to inform vaccine policy and risk-based prevention strategies.

5. Conclusions

In conclusion, although a reduction in incidence rates was detected after the implementation of PCV13 in 2010 in vaccine serotype IPD and overall IPD, we saw a more modest reduction in children with SCD. There are two components to the reduced effectiveness in children with SCD; one is a reduced effectiveness against vaccine serotypes and second an increased incidence of disease due to serotypes that are not included in the vaccines. Moreover, PCV13 had a higher rate of vaccine failure than PCV7, indicating lower effectiveness. This finding highlights the importance of maintaining immunogenicity and subsequent vaccine effectiveness while increasing valency in pneumococcal vaccines especially in vulnerable populations with high risk of disease burden.

Abbreviations:

IPD

invasive pneumococcal disease

SCD

sickle cell disease

MA

Massachusetts

MDPH

Massachusetts Department of Public Health

NP

nasopharynx

PCV

pneumococcal conjugate vaccine

PCV7

7-valent pneumococcal conjugate vaccine

PCV10

10-valent pneumococcal conjugate vaccine

PCV13

13-valent pneumococcal conjugate vaccine

PCV15

15-valent pneumococcal conjugate vaccine

PCV20

20-valent pneumococcal conjugate vaccine

PCV21

21-valent pneumococcal conjugate vaccine

PCV24

24-valent pneumococcal conjugate vaccine

PPSV23

23-valent pneumococcal polysaccharide vaccine

ROS

Random Over Sampling

RF

random forest

IR

incidence rate

IRR

incidence rate ratio

CI

confidence interval

VE

vaccine effectiveness

VT-IPD

vaccine-type invasive pneumococcal disease

VST7

serotypes included in PCV7

VST13

serotypes included in PCV13

VST13–7

serotypes included in PCV13 but not in PCV7

VST20

serotypes included in PCV20

VST21

serotypes included in PCV21

VST24

serotypes included in PCV24

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.vaccine.2025.127193.

Data availability

The authors do not have permission to share data.