Background. Despite the widespread acceptance of the 23-valent pneumococcal capsular polysaccharide vaccine (PPV), its protective effect continues to be disputed. We describe a novel approach to examine the protective effect of this vaccine.
Methods. We recorded the vaccination status of every patient for whom a culture yielded Streptococcus pneumoniae during a 4.5-year period, comparing rates of prior PPV administration in patients with (1) bacteremic pneumococcal pneumonia, (2) all-invasive pneumococcal disease, (3) nonbacteremic pneumococcal pneumonia, (4) acute exacerbation of chronic bronchitis (AECB) due to S. pneumoniae, and (5) pneumococcal colonization. The principal comparisons were with patients who had bacteremic pneumonia or any invasive pneumococcal disease and those with nonbacteremic pneumococcal pneumonia. We also compared vaccination rates in patients who had nonbacteremic pneumonia with vaccination rates in patients with AECB or pneumococcal colonization.
Results. The rate of prior PPV vaccination was lower among patients with bacteremic pneumococcal pneumonia (39.7%) or any invasive pneumococcal disease (38.0%) than among patients with nonbacteremic pneumonia (57.6%), AECB (60.0%), or pneumococcal colonization (57.8%). PPV conferred a 54% protection rate against bacteremic versus nonbacteremic pneumococcal pneumonia. There was no apparent protection against nonbacteremic pneumonia compared, for example, with colonized persons or with those who had AECB.
Conclusions. PPV provides moderate protection against invasive pneumococcal disease but does not protect against nonbacteremic pneumococcal pneumonia. These findings suggest the importance of a continued search for a better pneumococcal vaccine.
A variety of methods have been used to estimate the protective effect of 23-valent pneumococcal polysaccharide vaccine (PPV), including prospective clinical trials , case control studies , cohort studies [3, 4], and comparison of infecting serotypes in vaccinated and nonvaccinated subjects [5–8]. Although most authorities agree that PPV is protective at some level [7, 8], conflicting results have been obtained and controversy continues to surround the issue .
At our large Veterans Affairs Medical Center (Houston, Texas), where we prospectively study every patient from whom Streptococcus pneumoniae is isolated, a mandate to increase the use of PPV in 1996 coincided with the implementation of fully computerized patient records that document all vaccinations. We hypothesized that we could examine the protective effect of PPV by applying a case-control analysis  to compare rates of vaccine administration in patients with bacteremic pneumococcal pneumonia, nonbacteremic pneumococcal pneumonia, acute exacerbation of chronic bronchitis (AECB) due to S. pneumoniae, or pneumococcal colonization. We herewith present the results of this novel method of analysis.
Site of study. The Michael E. DeBakey Veterans Affairs Medical Center (Houston, TX) and its satellite clinics provide care for veterans of the US Armed Forces. In the past decade, the patient population has shifted to include substantial numbers of persons from the middle and upper middle classes who have entered this system because they are entitled to benefits and either have lost, or can no longer afford, private medical insurance. Many of the veteran patients rely entirely on this center for their medical care. Because the southwestern United States is an area to which people tend to retire—rather than one from which they retire—the patient population is growing, but it tends to be otherwise quite stable. Since 1996, the medical records of our patients have been computerized, with complete electronic documentation of every encounter between a patient and the medical system; all information on vaccinations in the medical record is contained in a specially designated field. At least once each year, nurses take a patient's vaccination history and record the results in that field. They then offer appropriate vaccine(s) and document whether the patient had received or declined them.
Inclusion of patients. During the period 1 January 2001 through 31 July 2005, we monitored all isolates of S. pneumoniae obtained in our microbiology laboratory, reviewed Gram stain results, and actually saw and evaluated many of the patients with pneumonia. For the purposes of this article, complete electronic records were reviewed before assigning a final clinical diagnosis; only then were vaccination fields probed to determine vaccination status.
During the time of the study, S. pneumoniae was isolated by culture from 390 patients. We included in our study patients who met the following definitions. Bacteremic pneumococcal pneumonia was diagnosed in patients who had a clinical syndrome consistent with pneumonia, radiologic documentation of an infiltrate, and isolation of S. pneumoniae from blood specimens . Nonbacteremic pneumococcal pneumonia was diagnosed in patients who had a typical clinical syndrome of pneumonia, radiologic evidence of a new pulmonary infiltrate, sputum Gram stain showing large numbers of inflammatory cells and gram-positive diplococci , sputum culture that yielded pneumococci with no other likely bacterial pathogens, and ⩾1 negative blood culture result(s) obtained before antibiotics were administered . AECB due to pneumococcus was defined by a constellation of symptoms and signs  that included increased cough and sputum production, fever and/or leukocytosis, absence of a pulmonary infiltrate on a chest radiograph, a characteristic Gram-stained sputum finding, and a culture yielding pneumococci without other likely bacterial pathogens. Colonized patients were those in whom S. pneumoniae was isolated from sputum or bronchial lavage fluid during evaluation of some other pulmonary condition, such as pulmonary fibrosis, lung cancer, tuberculosis, or heart failure, and in whom no acute infection was attributable to pneumococcus. Patients with invasive pneumococcal infections were those from whom pneumococcus was isolated from a normally sterile site ; this group included patients who had bacteremic pneumococcal pneumonia.
Exclusion of patients. We excluded 90 patients, including 29 who had pneumococcal pneumonia but for whom blood cultures were not performed, 39 with pneumonia whose sputum contained other likely bacterial pathogens (such as Haemophilus influenzae or Staphylococcus aureus), 8 in whom the clinical significance of the pneumococcal sputum isolate could not be determined with confidence, 4 whose blood cultures were positive for another likely pathogen, and 10 from whom pneumococci were isolated from miscellaneous nonsterile sites, such as skin, soft tissue, conjunctiva, or urine.
Comorbid conditions. Electronic medical records were searched for documentation of 7 conditions that have consistently been shown to be associated with pneumococcal disease [14, 15] and are indications for administering PPV [7, 8].
Statistical analysis. The rate of vaccination in patients with pneumococcal pneumonia was compared to the rate of vaccination in other groups by OR using χ2 analysis (table 1). Further comparison of the likelihood that a group of patients had received pneumococcal vaccine with the same likelihood of vaccination for all other groups was made by logistic regression (table 2). The presence of preexisting comorbid conditions in each group (table 3) was studied first by univariate analysis; for results in which P ⩽.20, the multivariate model was applied. Data are presented using 95% CIs. The aggregate protective effect of PPV was calculated as a percentage, as previously described . In all analyses, statistical significance was evaluated based on the standard P ⩽.05 using SAS, version 9.1 (SAS Institute), and Excel 2003 (Microsoft).
Authorization of study. This study was approved by the Infection Control and Research and Development Committees of the Michael E. DeBakey Veterans Affairs Medical Center and the Institutional Review Board of the Baylor College of Medicine.
Of the 300 patients, 58 (19.3%) had bacteremic pneumococcal pneumonia (table 1). An additional 13 patients (4.3%) had invasive pneumococcal disease without recognized pneumonia, including 5 with bacteremia and no apparent focus, 3 with spontaneous bacterial peritonitis, 2 each with septic arthritis and meningitis, and 1 with endocarditis; thus, 71 patients (23.7%) had invasive pneumococcal disease. Nonbacteremic pneumococcal pneumonia was diagnosed in 85 patients (28.3%) and AECB in 35 (11.7%). One hundred nine patients (36.3%) were colonized. The mean age of all patients was 63.9 years, and 293 (97.7%) were male; there were no significant differences in age or sex among the groups. In accordance with earlier reports , African American persons appeared to be somewhat more susceptible than white persons to invasive pneumococcal disease, but differences in our study did not reach statistical significance (P =.15). In contrast, the likelihood of having received PPV among African American persons in our patient group was higher than among white persons or Hispanic persons (64 [71.9%] of 89 vs. 119 [59.8%] of 199 and 7 [58.3%] of 12, respectively; difference significant by χ2 analysis, P <.01).
As shown in table 1, only 23 (39.7%) of 58 of those with bacteremic pneumococcal pneumonia and 27 (38.0%) of 71 of those who had any invasive pneumococcal disease received PPV within the previous 5 years, compared with 49 (57.6%) of 85 of those who had nonbacteremic pneumococcal pneumonia, 21 of 35 (60.0%) of those who had AECB, and 63 of 109 (57.8%) of those who were colonized. When patients who had bacteremic pneumococcal pneumonia were compared with those who had nonbacteremic pneumococcal pneumonia, the OR for having received PPV was 0.44 (95% CI, 0.17–1.12; P =.035). Similarly, the OR for prior receipt of PPV, comparing any invasive pneumococcal disease with nonbacteremic pneumonia, was 0.45 (95% CI, 0.22–0.90; P =.015). ORs for the comparison of bacteremic pneumonia with AECB and colonization were 0.44 (95% CI, 0.17–1.12; P =.057) and 0.48 (95% CI, 0.24–0.96; P =.026), respectively.
Compared with patients who had nonbacteremic pneumococcal pneumonia, the likelihood of having been vaccinated in the previous 5 years, as determined by logistic regression analysis, was lower among patients with bacteremic pneumococcal pneumonia (0.50; 95% CI, 0.24–1.04) and significantly lower among all patients with invasive pneumococcal disease (0.46; 95% CI, 0.23–0.90) (table 2). The likelihood of vaccination among patients who were colonized and those with nonbacteremic pneumococcal pneumonia was nearly identical (0.99; 95% CI, 0.55–1.83). Using colonized patients as the referent group, the ORs for having received PPV were 0.51 (95% CI, 0.25–1.02) for patients with bacteremic pneumonia and 0.46 (95% CI, 0.24–0.91) for those with any invasive pneumococcal disease. These data showed that PPV conferred a 54% protection rate against invasive pneumococcal disease compared with protection in nonbacteremic pneumonia or colonized persons. They also revealed the lack of apparent protection against nonbacteremic pneumonia when patients with that condition were compared with those who were colonized. Patients with AECB were significantly more likely to have received PPV than those with bacteremic pneumonia (2.68; 95% CI, 1.00–7.18) or all those with invasive disease (3.01; 95% CI, 1.17–7.76). When data were analyzed using any history of prior vaccination (rather than confining the definition of prior vaccination to the previous 5 years), results were similar but generally did not reach statistical significance (data not shown).
Comparability among groups of subjects was examined using logistic regression, by documenting the presence of 7 diseases that have been implicated in susceptibility to pneumococcal infection [14, 15] (table 3). By univariate analysis, HIV infection (the only comorbid condition with a specific defect in antibody response to capsular polysaccharides ) and liver disease were more common in patients with bacteremic pneumonia than in those with nonbacteremic pneumonia (OR, 2.95 [95% CI, 1.06–8.22] and 2.33 [95% CI, 1.07–5.08], respectively); chronic obstructive pulmonary disease was less common in bacteremic than in nonbacteremic pneumonia (OR, 0.46; 95% CI, 0.21–0.84). In the multivariate model, only the association of HIV infection with all invasive pneumococcal disease remained significant (OR, 7.17; 95% CI, 1.47–35.07). When we reanalyzed the data after removing results for HIV-infected subjects, the effectiveness of PPV remained significant against bacteremic pneumonia (55%; P =.05) and against all invasive disease (58%; P =.025), compared with nonbacteremic pneumococcal pneumonia.
In this study, we compared the rates of pneumococcal vaccination among patients with pneumococcal infections (bacteremic pneumonia, all invasive disease, nonbacteremic pneumonia, and AECB) with the rates of vaccination among those who were colonized by S. pneumoniae without apparent infection. Applying the method of case-control analysis  to this data set, we observed that PPV provides a 54% protection rate against invasive pneumococcal disease, compared with protection against nonbacteremic pneumonia or in colonized persons. Vaccination rates were similar in patients with nonbacteremic pneumonia or AECB and in those who were colonized, indicating that PPV offers no greater protection against nonbacteremic pneumococcal pneumonia than it does against acute bronchitis or colonization. Except for the higher prevalence of HIV disease among patients with invasive pneumococcal disease, the groups were generally similar; when HIV-infected patients were excluded from analysis, the protective effect of vaccine against all invasive disease was essentially the same (58%).
Our original hypothesis was that PPV should protect against nonbacteremic as well as bacteremic pneumonia. Accumulation of plasma constituents in the alveoli is the histological hallmark of pneumonia, and it seemed reasonable to believe that antibodies in the early inflammatory exudate would opsonize bacteria for phagocytosis, thereby offering protection against pneumonia, just as they would opsonize invasive bacteria, thereby clearing them from the bloodstream. Previous studies, as summarized in a Cochrane review , have suggested, however, that protection is limited to invasive disease. The results of the present study, which used an entirely different approach to analysis, supported those earlier findings. Our estimated 54% protective effect of PPV against invasive disease is remarkably similar to the 53% reached in the Cochrane review .
The principal strengths of this study include the accuracy of the clinical diagnoses—especially of nonbacteremic pneumococcal pneumonia—that were made on the basis of published criteria that were uniformly and prospectively applied. Fully computerized medical records assure that all data were available for collection. The availability of data is particularly important in documenting vaccination histories; by policy of the Department of Veterans Affairs, data on vaccinations are recorded in a single designated field in the medical record. Each fall, the vaccination records are updated so that any new patient who enters the system, as well as all old patients, are reinterviewed to determine whether they have received or declined PPV.
Limitations of this study include the fact that, despite the similarities among the groups, this was still a post hoc analysis. The higher proportion of HIV-infected persons (a group known to respond poorly to PPV [17, 19]) among patients with bacteremic disease might have affected the interpretation of the results; when these subjects were removed from the data set, however, the protective effect of the vaccine remained unchanged. In the present study, we focused on patients with bacteremic pneumonia versus those with nonbacteremic pneumonia because we hypothesized that they would be the most similar in terms of predisposing factors and eligibility for vaccine. Patients in whom pneumococci were isolated during workup of other pulmonary conditions and patients with AECB were more likely to have chronic obstructive pulmonary disease, an obvious indication for pneumococcal vaccination, and PPV administration was more common in this group. The generalizability of our results is limited by the disproportionate number of male patients in the veteran population. Finally, we did not include an analysis of influenza vaccine patterns; however, because patients who receive pneumococcal vaccine nearly always receive influenza vaccine, and because those who decline pneumococcal vaccine decline influenza vaccine, these factors would bias the data to favor protection by PPV.
Shapiro et al.  found that immunocompromised status, age, and time since vaccination influenced determination of vaccine effectiveness. In our study, age was nearly identical to their findings, and rates of immunocompromising diseases were similar, except as noted above. Our study supports others [3, 18] by showing that PPV protects recipients against invasive pneumococcal disease. The putative mechanism of protection is opsonic removal of organisms that migrate from the alveoli into the bloodstream. Because mortality from bacteremic pneumococcal disease exceeds that from nonbacteremic pneumococcal disease, this finding indicates benefit from receiving PPV. There is, however, substantial morbidity and mortality from nonbacteremic pneumococcal pneumonia , and the absence of apparent protection against this condition provides further impetus to the ongoing search for a better vaccination strategy.
Financial support. The Department of Veterans Affairs, through the Merit Review Program.
Potential conflicts of interest. D.M.M.: from 1997 through 2001, D.M.M.'s laboratory received funding from Merck for participating in a multicenter study of pneumococcal polysaccharide vaccine. All other authors: no conflicts.