Effect of Pneumococcal Vaccine on Mortality and Cardiovascular Outcomes: A Systematic Review and Meta-Analysis

Various studies have suggested the possible cardiovascular (CV) protective effects of the pneumococcal vaccine (PV). Therefore, we conducted a meta-analysis to assess the association between recipients of PV with mortality and CV outcomes among patients with and without established cardiovascular disease. We performed a systematic literature search in PubMed, Embase, and Scopus for studies evaluating the effect of PV on mortality and CV outcomes. A total of 15 studies with 347,444 patients were included in the meta-analysis: 111,784 patients received PV (32%) and 235,660 patients were in the unvaccinated group (68%). Recipients of PV were associated with decreased all-cause mortality (HR, 0.76 (95% CI: 0.66 to 0.87), p < 0.001). PV was associated with a decrease in the incidence of myocardial infarction (MI) (HR, 0.73 (95% CI: 0.56–0.96), p = 0.02), without significant reduction in CV mortality (HR, 0.87 (95% CI: 0.72–1.07), p = 0.18) and stroke (HR, 1.01 (95% CI: 0.93–1.10), p = 0.82). Our study found PV was associated with decreased risk of all-cause mortality and MI. Future RCTs will be necessary to confirm benefits associated with receipt of PV.


Introduction
Cardiovascular diseases (CVD) are the leading cause of death globally, accounting for an estimated 17.9 million deaths in 2019 [1]. Previous data suggested that preexisting chronic cardiac conditions may predispose to community acquired pneumonia (CAP). The odds of CAP and invasive pneumococcal disease (IPD) in patients with CVD are significantly increased [2,3]. Furthermore, patients with CAP have an increased risk of cardiovascular complications, such as heart failure, acute coronary syndrome, cardiac arrhythmias and strokes through multiple mechanisms [4,5]. Streptococcus pneumoniae is known to produce 100 types of biochemically and serologically distinct capsule types or serotypes. Antibodies against one serotype do not offer protection against another. Major adverse cardiovascular events, or MACE, are an essential factor for mortality in patients with CAP by streptococcus pneumoniae [6]. It is estimated that there is an increase of more than 14% of new-onset heart failure and more than 7% of myocardial infarctions 2 of 14 following CAP [7]. Furthermore, a previous meta-analysis reported an 18% occurrence of cardiovascular complications following CAP [8].
Various pathogen-driven mechanisms of cardiac damage as a result of S. pneumoniae pneumonia have been described. These include invasion of myocardium with the formation of microlesions [9,10], electrophysiological abnormalities [9,11,12], cardiomyocytes necrosis in a pneumolysin-dependent manner [8,10], enlargement and instability of atherosclerotic plaques [13], necroptosis in cardiomyocytes and macrophages infiltrating the heart [10,12,14], platelet activation and antibiotic-induced heart scarring [9,15]. Though most of the complications arising from these mechanisms manifest acutely, an increased cardiovascular risk may persist for as long as 10 years following infection [16].
The pneumococcal vaccine (PV) has been shown to reduce both the severity of CAP and reduce the subsequent risk of mortality associated with CAP [17]. The U.S. Centers for Disease Prevention and Control (CDC) recommends vaccination against pneumococcal disease for all adults with cardiovascular disease [18]. Similar recommendations in favor of pneumococcal vaccination were also encouraged for adult patients with heart failure in the 2021 European Society of Cardiology (ESC) guidelines and American Heart Association (AHA) 2013 guidelines for heart failure diagnosis and treatment [19,20].
Since PV has been shown to be efficacious, their protective effects amongst the elderly and in patients with CVD have been studied [21,22]. Several non-randomized studies (prospective and retrospective cohort studies) have been conducted in recent years and have shown conflicting results in terms of their effects on CV outcomes and mortality [21][22][23]. In addition, patients with prior CV diseases are at a higher risk of future CV events following pneumococcal infections. Therefore, we aim to perform a systematic review and meta-analysis to evaluate whether receipt of the pneumococcal vaccine among adults with and without history of cardiovascular disease is associated with a decreased risk of mortality and cardiovascular events.

Materials and Methods
This meta-analysis was conducted and reported following the Cochrane and PRISMA (Preferred reporting items for systematic review and Meta-analysis) 2020 guidelines and performed according to established methods, as described previously [24,25]. In addition, the pre-specified study protocol has been registered in PROSPERO: CRD42021277943.

Search Strategy
We conducted a systematic literature search in PubMed, Embase, Cochrane Library, and Scopus using predefined MESH terms using "AND" and "OR". The following search terms were used: "pneumococcal Vaccine" OR "pneumonia" AND "mortality" OR "cardiovascular disease" OR "myocardial infarction" OR "cerebrovascular". The search was performed from inception until February 2022 without any restrictions on the language of the studies. All the studies were carefully screened and exported to the Endnote 2020 library [26]. A manual check was carried out to crosscheck for any remaining duplicates. Two reviewers (SPA and AJ) reviewed the papers based on title and abstract. A third author (VJ) arbitrated discrepancies regarding the inclusion of studies.

Outcomes
The primary outcome for this meta-analysis was all-cause mortality between vaccinated and placebo/unvaccinated groups. The secondary outcomes were occurrence of myocardial infarction, stroke, and cardiovascular mortality. The definitions of all outcomes are provided in Table S1. 1.
Animal studies, abstracts, editorials, commentaries, systematic reviews, single patient case studies, letters, and studies with insufficient data were excluded.

2.
Studies where the pneumococcal and influenza vaccines were compared instead of pneumococcal alone with placebo or unvaccinated groups were also excluded.

Data Extraction and Quality Assessment
All the included articles were extracted for the following data: study type, author, country, number of patients in both groups, age, sex, history of influenza vaccination, follow-up period, comorbidities, risk factors, primary and secondary outcomes. Three investigators (SPA, JA, and AI) independently appraised the potential risk of bias for observational studies using the Newcastle-Ottawa scale for cohort studies (Table S2). We then classified studies into low, moderate, or high quality based on the scores after evaluation.

Statistical Analysis
The vaccinated group was defined as patients who have received any form of PV. In contrast, the comparison group was defined as patients who were either not vaccinated or had received a placebo during the study period. We pooled the effect sizes comparing these two groups based on our predefined primary and secondary outcomes and analyzed each outcome individually using the random-effects model and DerSimonian-Laird method to account for between-study variations [27].
The estimated effect size was reported as a point estimate and 95% confidence interval (CI). A two-tailed p-value of less than 0.05 is statistically significant. Considering potential confounding factors, maximally adjusted estimates, where available, were used in the analysis. Between-study heterogeneity was calculated using the I 2 test, with I 2 > 50% and >75% considered to be moderately and highly heterogeneous, respectively [28]. In addition, we performed subgroup analyses on all outcomes based on CV risk and follow-up period (≥3 years vs. <3 years).
The high CV risk group includes studies where at least 50% of patients have established atherosclerotic CV diseases (coronary artery diseases, cerebrovascular diseases, or peripheral vascular diseases), heart failure, end-stage chronic kidney disease with estimated GFR < 15 mL/min/1.73 m 2 or requiring dialysis, and diabetes mellitus with target end-organ damage. The remaining population fell under low or moderate CV risk. We performed sensitivity analyses by using the leave-out-one method, where studies were removed one at a time, and compared each overall effect size obtained with initial results [29]. Furthermore, meta-regression was performed to look for sources of heterogeneity and assess potential study-level covariates on the outcomes. The variables were as follows: age, gender, and comorbidities. We conducted all statistical analyses using STATA (StataCorp, Lakeway, TX, USA, version 17.0) [30].

Literature Search
A total of 683 articles were identified on the initial database screening by using the MESH terms. Forty-five duplicate articles were excluded manually based on author and titles of the study. A total of 605 articles were excluded based on outcomes, single-arm studies, mixed intervention groups, wrong drugs, and study designs that were not part of the inclusion criteria. Finally, 15 articles were included in the meta-analysis based on the eligibility criteria ( Figure 1).

Literature Search
A total of 683 articles were identified on the initial database screening by usin MESH terms. Forty-five duplicate articles were excluded manually based on autho titles of the study. A total of 605 articles were excluded based on outcomes, single studies, mixed intervention groups, wrong drugs, and study designs that were not p the inclusion criteria. Finally, 15 articles were included in the meta-analysis based o eligibility criteria ( Figure 1).

Studies Characteristics
A total of 347,444 patients were included in our analysis. The studies were u taken in seven different countries with most being conducted in North America (sev

Studies Characteristics
A total of 347,444 patients were included in our analysis. The studies were undertaken in seven different countries with most being conducted in North America (seven in the USA and two in Canada), three in Europe (two in Spain and one in the United Kingdom) and three in Asia (China, Japan and Taiwan). The follow-up period varied among the included studies with a range from 3 months to 13 years. Table 1 summarizes the characteristics of the included studies. The studies were performed in different patient population groups: five studies included patients that were at risk of developing or had a history of a MI or cerebrovascular accident, two studies included patients with or without a history of myocardial infarction (MI) or coronary artery disease (CAD), four studies included patients receiving hemodialysis, two studies included with chronic illness (including ischemic heart disease), and one study included all adults hospitalized for pneumonia while another included a cohort of men aged 45-69. In two studies, the patient population was divided into patients with high risk factors for CVD and those with low risk factors, while another included patients over >80 as one group and those 60-79 as another. Two studies used the same patient population but reported different outcomes, so they were included in the analysis.

Baseline Patient Demographics
A total of 111,784 patients received PV (32%) and 235,660 patients received a placebo or were unvaccinated (68%); 33% of all patients were male (25% in the PV while 36% in the unvaccinated group). The three most prevalent comorbidities were hypertension (29.5%; 29% in both groups separately), diabetes mellitus (29%; 23.9% in the PV group and 29.3% in the unvaccinated group) and hyperlipidemia (13.6%; 18% in the PV group and 10% in the unvaccinated group). The prevalence of smoking (either current or past) was also high in both groups: 24.1% in the PV group and 15.5% in the unvaccinated group. Table 2 describes the baseline characteristics of all included patients in the control and intervention group.
We performed a subgroup analysis based on CV risk stratification. It was found that the benefit of reduction in all-cause mortality was only in the high cardiovascular group (HR, 0.66 (95% CI: 0.56 to 0.79)) compared with low-to-moderate cardiovascular risk (HR, 0.96 (95% CI: 0.86 to 1.08)) and the subgroup difference was statistically significant (p < 0.01). The significant association seen in the total cohort seems to be due to weightage of the high CV risk cohort. Based on the subgroup analysis, only vaccinating high CV risk cohorts seem to have mortality benefit ( Figure S1). There was no subgroup difference (p = 0.15) in terms of follow-up duration using 3 years as a cut-off period ( Figure S2). Sensitivity analysis by the leave-one-out method showed that the odds of all-cause mortality remained significantly reduced in the vaccinated group compared to the control group, confirming the robustness of the results ( Figure S3).  [31,32,34,35,37,38,[41][42][43].
We performed a subgroup analysis based on CV risk stratification. It was found that the benefit of reduction in all-cause mortality was only in the high cardiovascular group (HR, 0.66 (95% CI: 0.56 to 0.79)) compared with low-to-moderate cardiovascular risk (HR, 0.96 (95% CI: 0.86 to 1.08)) and the subgroup difference was statistically significant (p < 0.01). The significant association seen in the total cohort seems to be due to weightage of the high CV risk cohort. Based on the subgroup analysis, only vaccinating high CV risk cohorts seem to have mortality benefit ( Figure S1). There was no subgroup difference (p = 0.15) in terms of follow-up duration using 3 years as a cut-off period ( Figure S2). Sensitivity analysis by the leave-one-out method showed that the odds of all-cause mortality remained significantly reduced in the vaccinated group compared to the control group, confirming the robustness of the results ( Figure S3).

Secondary Outcomes-Cardiovascular Mortality, Acute MI, and Stroke
Eight studies reported on acute MI with a total of 193,940 patients [32][33][34][35][36]40,43 PV was associated with a significant decreased incidence of MI when compared t vaccination with a hazard ratio of (HR, 0.73 (95% CI: 0.56-0.96), p = 0.02, I 2 = 84%) (Fi 3A). We performed a subgroup analysis based on CV risk stratification. Although was no statistically significant subgroup difference between high cardiovascular We performed a subgroup analysis based on CV risk stratification. Although there was no statistically significant subgroup difference between high cardiovascular risk against patients with low to moderate cardiovascular risk, the risk reduction effect of vaccination was lost when stratified. In the forest plot, it is evident that the association is statistically insignificant for the high CV risk cohort and low to moderate risk cohort; however, it is significant in the total cohort, suggesting the studies in each subgroup may be inadequate and thus lack statistical power to detect a significant beneficial effect of vaccination ( Figure S4). We found a significant subgroup difference based on follow-up duration (p = 0.04), with the reduction in acute MI being more profound in studies with a follow-up duration of less than 3 years (HR 0.60, 95% CI 0.41 to 0.89) compared to that of 3 years or more (HR 0.96, 95% CI 0.78 to 1.18) ( Figure S5). Sensitivity analysis via successive exclusion of each individual study showed that the results, in general, remained significant, except upon removal of studies by Zahid et al., Lamontagne et al. or Eurich et al., suggesting that results might be influenced by these studies ( Figure S6).

Meta-Regression
Meta-regression was performed for the outcomes of all-cause mortality and acute MI to explore for underlying effect modifiers. The following variables were considered: age, sex (male), specific comorbidities (HTN, DM, hyperlipidemia, prior MI or CVA, chronic kidney disease), smoking, and prior influenza vaccine. The results from the meta-regression are summarized in Table 3. Table 3. Meta-regression of potential effect modifiers for all study outcomes.

Meta-Regression Variables
All Meta-regression revealed male sex (coefficient −0.011, p = 0.027) ( Figure S8) as the sole significant negative predictor of the observed effect size on all-cause mortality.

Publication Bias
Publication bias was assessed using a funnel plot and Egger's regression test. Possible publication bias was detected through funnel plot asymmetry and Egger's test (p < 0.05) for all-cause mortality ( Figure S9). Overall estimates remain similar, with the vaccinated group showing a greater reduction in all-cause mortality after trim-and-fill analysis with the addition of three studies (HR, 0.85 (95% CI 0.74 to 0.99), p = 0.04). Otherwise, publication bias was not detected for the other outcomes including cardiovascular mortality (p = 0.76), MI (p = 0.09), and stroke (p = 0.45) (Figure S10A-C).

Discussion
To the best of our knowledge, this is the most comprehensive and the largest metaanalysis, consisting of 15 studies with 347,444 participants. The participants were followed for a mean period of 2.89 ± 3.25 years. Pooled results demonstrated that receipt of PV was associated with a 24% decrease in all-cause mortality and a 27% decrease in the risk of MI when compared to the unvaccinated group, with no significant interaction by baseline cardiovascular risk. Our findings support the recommendations of the Centers for Disease Prevention and Control (CDC) for vaccination against pneumococcal disease for all adults with cardiovascular disease and represent the best available current evidence.
The all-cause mortality benefit from PV in our study is in line with the findings of previous analyses. A previous meta-analysis revealed that there was a significant reduction in all-cause mortality in patients receiving PV; however, their analysis only considered patients with established CV disease (such as heart failure) or those with a very high CV risk [46]. On the other hand, our study demonstrated benefits in adult patients, regardless of the baseline cardiovascular risk. In contrast to previous meta-analyses, our study failed to show a protection effect on CV mortality after PV [47,48]. While most studies to date, including ours, showed no benefits in terms of stroke, a meta-analysis conducted by Vlachopoulos et al. revealed a decreased risk of stroke only in elderly populations [46][47][48].
With the currently available evidence, the cardioprotection offered by the PV could be associated with inhibition of pneumococcal infections and decreased inflammation or due to immunomodulatory mechanisms [49]. S. pneumoniae has been demonstrated to activate platelets, invade the myocardium, and cause cardiomyocyte necrosis, resulting in instability of the atherosclerotic plaques [23,[49][50][51]. Additionally, systemic inflammation during pneumococcal infections promotes a thrombogenic state and increases the metabolic demand of oxygen [52]. With the mismatch in supply and demand of oxygen, and the propensity towards rupture of extremely thin-capped fibroatheroma, the risk of both Type 1 and Type 2 MI is increased [50,53,54].
The immunomodulatory mechanism of PV may also be associated with suppression of atherosclerosis development/progression. It is based on the "antigen mimicry" between oxidized LDL (oxLDL), a component of atherosclerotic plaques, and phosphorylcholine lipid antigens expressed on the cell wall polysaccharide of Streptococcus pneumoniae [55,56]. Previous studies have reported a 40% reduction in atherosclerotic plaque in an animal model after pneumococcal vaccination [51]. Antibodies produced against S. pneumoniae after PV also bind to oxidized phospholipid epitopes on oxLDL and prevent uptake of oxLDL by macrophages, thus causing decreased progression of atherogenesis in pneumococcalimmunized mice [55,56]. These findings suggest that PV exerts beneficial effects by directly preventing CV events via the immunomodulatory mechanism. Therefore, it may be argued that current AHA and ESC recommendations for PV could be expanded for all patients with CVD risk factors and not just patients with heart failure [19,20], although this would require further prospective studies.
The protective effect of PV appears to be associated with follow-up duration. In a meta-analysis of cohort studies, the protective role of PV was found to be diminished over time, where PV offered higher protective value during shorter follow-up periods when compared to longer follow-up [48]. Similarly, our study showed that a shorter follow-up duration is associated with a larger magnitude of reduction in the incidence of MI in the vaccinated group, suggesting that the cardioprotective effect may wane over time.

Strengths and Limitations
We have used an all-inclusive search strategy to explore all the relevant studies comparing pneumococcal vaccine with placebo/control to date. Our analysis is the most up-to-date and included the largest sample size to check the real-world data and analyzed PV's effect on all adult populations. A major limitation is that the included studies were observational in nature, and thus, may be affected by other confounding factors which were not possible to account for. While our results had high heterogeneity, results after sensitivity analysis by leave-one-out analysis and exclusion of studies through guidance of Galbraith plot, in general, remained significant and in line with the main analysis. Whilst there are no RCTs, a multi-center phase 3 RCT with a sample of 4750 participants aged 55-60 years having two or more CVD risk factors but no known CVD events is ongoing (ACTRN12615000536561) [57].
In a study by Africano et al., MACE were independently associated with disease caused by serotypes 3 and 9n. The serotype 3 capsule is included in the current 13-valent conjugate vaccine (PCV13) formulation and pneumococcal polysaccharide vaccine (PPSV23). Still, it results in severe disease and causes MACE. Another serotype, 9N, is associated with invasive pneumonia and severe disease. The authors also found that patients infected with serotype 9n were more susceptible to MACE during invasive pneumonia. Serotype 9n is not covered by the currently available vaccine and was associated with mortality along with serotype 3 [6]. However, we did not have enough data to compare different types of vaccines against different serotypes. Information regarding different pneumococcal vaccines was limited; therefore, we were not able to analyze the outcomes by vaccine type. Lastly, limited or lack of data availability on certain endpoints such as MACE, stroke, cardiovascular mortality may have affected our assessment on these outcomes.

Conclusions
Pneumococcal infections in patients with cardiovascular disease are associated with an increased morbidity and mortality. The findings of this meta-analysis suggest that PV could prove to be cardioprotective as evidenced by the reduction in risk of MI and all-cause mortality. However, future randomized controlled trials are required to demonstrate the evidence suggested by our meta-analysis to inform future guidelines, and vaccination programs can be implemented in those with the greatest benefit.
Supplementary Materials: The following supporting information can be downloaded at: https://www. mdpi.com/article/10.3390/jcm11133799/s1, Table S1: Definitions of outcomes; Table S2: Newcastle-Ottawa scale for quality assessment and bias assessment of observational studies; Figure S1: Subgroup analyses based on cardiovascular risk for primary outcome: all-cause mortality; Figure S2: Subgroup analyses based on follow-up duration for primary outcome: all-cause mortality; Figure S3: Leaveone-out analysis for primary outcome: all-cause mortality; Figure S4: Subgroup analyses based on cardiovascular risk for secondary outcome: MI; Figure S5: Subgroup analyses based on follow-up duration for secondary outcome: MI; Figure S6: Leave-one-out analysis for secondary outcome: MI; Figure S7: Leave-one-out analysis for secondary outcome: (A) CV mortality (B) Stroke; Figure S8: Meta-regression of male as an effect modifier for all-cause mortality; Figure S9: Funnel plots for publication bias for primary outcome: all-cause mortality; Figure S10: Funnel plots after trim-and-fill for primary outcome: all-cause mortality.