We undertook a large prospective population-based study to evaluate the controversial effectiveness of the PPV23 against myocardial infarction and stroke. To our knowledge, this is the first prospective study using validated clinical data that provides population based assessment of pneumococcal vaccination effectiveness against cardiovascular events. Although we did not conduct a randomized controlled trial, the large size of our study population together with adjustment for important covariables in the multivariable analysis, provides an adequate basis for assessing this major public health issue in the general population over 60 years (main target population where PPV23 is recommended).
As main finding, our data shows a marginally significant reduction in the risk of ischaemic stroke, suggesting a possible protective role of vaccination against some acute thrombotic events. Considering the total study population, pneumococcal vaccination was associated with a 35% (95% CI: 1% to 58%) reduction in the adjusted risk of ischaemic stroke. This result seems to be robust considering that this significant protective effect remains when people with history of prior stroke are excluded from the analysis. The observed protective effect of vaccination against myocardial infarction (17%) was weaker and did not reach statistical significance (-22% to 44%).
Prior studies that have tried to examine this question have reported conflicting results [10–12, 18]. An earlier case-control study that included 335 case patients with myocardial infarction and 199 control subjects reported a non significant protective effect of PPV23 against myocardial infarction (odds ratio [OR]: 0.89 (95% CI: 0.60-1.33) . In a Canadian hospital-based case-control study, 999 case patients who had been admitted for treatment of myocardial infarction were compared with 3,996 control patients reporting that cases were less likely than controls to have received PPV23 (7.1% vs 11.6%; adjusted OR: 0.53; 95% CI: 0.40-0.70) . Importantly, the study population only included persons at risk for myocardial infarction (persons with hypertension, diabetes mellitus and/or dyslipidemia) and "healthy user" bias possibly occurred in selecting controls [19, 20].
In a recent retrospective large cohort study involving 84,170 men aged 45 to 69 years in California, Tseng et al.  examined the relationship between pneumococcal vaccination and risk of myocardial infarction and stroke, concluding that receipt of pneumococcal vaccine was not associated with subsequent reduced risk of both events. However, some major methodological distinctions may explain the different results observed in the present study and the Tseng's report. This study used validated clinical data to establish outcome events rather than billing data used in Tseng's report. In addition, this study includes all population over 60 years, whereas the study by Tseng et al. only included men 45-69 years-old without prior history of coronary artery disease or stroke. In addition, as it was noted,[21, 22] it is not clear whether the protective effect of pneumococcal vaccination against vascular events in individuals aged 65 years or older (main target group for vaccination) was properly assessed in Tseng's report given that 78% of sample size were men less than 65 years-old. In fact, the adjusted Hazard Ratios for acute myocardial infarction (HR: 0.89; 95% CI: 0.80-1.01; p = 0.10) and stroke (HR: 0.85; 95% CI: 0.70-1.03; p = 0.10) observed in Tseng's report, although not reaching conventional significance levels, suggested that there might be a protective effect of pneumococcal vaccination against vascular events in people over 65 years-old [11, 22].
In other large observational study using health databases in Hong Kong, Hung et al.  have recently reported significant lower risks of acute myocardial infarction (HR: 0.52; 95% CI: 0.38-0.71) and ischaemic stroke (HR: 0.67; 95% CI: 0.54-0.83) among elderly persons who received dual vaccination with PPV23 and influenza vaccine. Our estimates of protective effect of vaccination against ischaemic stroke is essentially similar to data observed in the Hung's report.
Multiple mechanisms could contribute to the potential cardiovascular protective effect of pneumococcal vaccination. Possible direct effects include a reduction risk for pathogenic cross reaction antibodies induced by persistent pneumococcal antigens. Animal experiments have shown that pneumococcal vaccination reduces the extent of atherosclerotic lesions, and it has been hypothesized that antibodies directed against pneumococus would also cross react with oxidized low-density lipoprotein and impede the formation of foam cells . The role of acute infection in triggering acute coronary or cerebrovascular events would be complex and multifactorial (e.g., increased inflammatory activity, prothrombotic conditions, and biomechanical stress on the arteries, disrupting and triggering thrombosis in a preexisting advanced artery lesion) .
With regard to community acquired pneumonia, controversy about PPV efficacy against pneumonia exists [4–6]. The last Cochrane Review Reported a pooled vaccine efficacy of 29% (95% CI: 3% to 48%) against all-cause pneumonia. Our results showing a non-statistically significant 15% reduction in the adjusted risk of pneumonia among vaccinated subjects fits with data reported in Cochrane review and other meta-analysis and they do not exclude the possibility of a little, but non insignificant effect against pneumonia.
A major strength of this study is that it was population-based and included all target people for pneumococcal vaccination in a well defined geographical area. Other important strengths were the prospective design, the validation of outcome events by checking clinical records that protects against biases related to recall and the use of survival analysis methods to estimate vaccine effectiveness adjusted by important covariables such as age, sex, influenza vaccine status, presence of main comorbidities and cardiovascular risk factors. Two possible confounding factors (dietary habits and physical activity) were not measured in the present study. However, given multicolinearity exists between these factors and other multiple factors measured in our study (e.g., body mass index, blood pressure and cholesterol level), which were considered in the multivariable analyses, is seems unlikely that these two factors would introduce a systematic bias on the vaccination effectiveness estimates.
Given the large size of our study population, many statistically significant differences appear when comparing vaccinated and unvaccinated groups, but most of them were not substantial. The authors performed multivariable adjustment for potential confounders to account for these differences in the statistical analysis. However, as with all observational studies, the possible influence of residual confounding or a healthy user effect on the estimates of vaccine effectiveness cannot be completely excluded considering that vaccination was non-randomized.
The relatively high proportion of unvaccinated subjects in the present cohort study (67%) is largely due to criteria used to define cohort members as adequately immunized against pneumococcus (at least one dose of PPV23 within 5 years prior study start). According this restrictive definition, many elderly individuals (without criteria for revaccination) were classified as unvaccinated because they had received PPV23 more than five years before study start and this fact contributed to the relatively high size of the unvaccinated group. If we consider all individuals who had received a dose of PPV23 at any time, the proportion of cohort members never vaccinated against pneumococcus would be only 44%, which fits with data reported in other studies assessing PPV effectiveness .
On other hand, it must be noted that criteria used to classify individuals as adequately immunized against pneumococcus (at least one dose of PPV23 in prior 60 months before study start) can impact the results evaluating vaccination effectiveness. Although it has been reported that antibody level declines 3-5 years after vaccination, it is possible that a protective effect of PPV23 (if exists) could remain over time.
Future analyses (involving a considerable higher number of person-years followed) should evaluate the effects of the vaccine according to time elapsed since vaccination. These analyses should investigate the possible vaccination effectiveness among those individuals who received the PPV23 in prior 6-10 years as compared with those who received the vaccine in prior 5 years or those who were never vaccinated.
Given the demonstrated efficacy of PPV23 against invasive pneumococcal disease, commencing new Randomized Controlled Trials in populations at risk would create ethical difficulties. Thus, prospective cohort studies using validated clinical data are an acceptable alternative to estimate vaccine effectiveness against different clinically relevant outcomes among different populations at risk.