The global burden of community-acquired pneumonia (CAP), in general, and infections with Streptococcus pneumoniae (the pneumococcus) in particular, have remained considerable [1, 2], albeit of unequal geographical distribution [3, 4]. Epidemiologic studies have emphasized that the impact of pneumococcal infections is substantial in older adults, with and without comorbid disease and in younger adults with comorbid conditions, and indicated that as the world’s population continues to age, these infections are likely to become even more problematic. Even with appropriate treatment, pneumococcal pneumonia has both acute [5] and long-term [6, 7] morbid effects. Thus, prevention should remain the backbone of our efforts. The aim of this manuscript is to review the history of pneumococcal vaccination and the impact of different pneumococcal vaccines and vaccination strategies that have been utilized in adults over the past 111 years.

The early years of pneumococcal vaccine development

More than one hundred years ago, the South African gold mining industry asked British physician, Almroth Wright, who had previously worked on a vaccine against typhoid fever, to study vaccination against pneumococcal pneumonia in men working in gold mines located on the Witwatersrand [8]. Whether practical or altruistic, the motivation of the mine owners was simple. Potential miners seeking employment came from all around Africa and were housed in barracks. During the first four months of arrival, the monthly attack rate of pneumonia – nearly all of which was proven to be due to pneumococcus – was between 1 and 2%, and in men who were proven to be infected, the death rate ranged between 25 and 56%. Between 5 and 10% of men who began working in the mines died each year [8]. The observations of Wright and his colleagues [8] showed a substantial reduction in cases of pneumonia and in deaths following inoculation of varying doses of killed pneumococci [9].

Initial experiments with pneumococcus as a vaccine

What was the basis for recommending inoculation with killed pneumococci? Remarkably, two decades earlier, and in a single year (1891), three different groups of investigators reported observations on induction of immunity by repeated inoculation of experimental animals with pneumococci [10]. Most influential was the work of Georg and Felix Klemperer [11], an uncle and nephew who showed that repeated inoculation of rabbits with killed pneumococci rendered them immune to subsequent challenge by that organism. Almost as if added as an afterthought, their article states that when they transferred serum, the humoral substance, from immunized to unimmunized rabbits, these animals were also protected against pneumococcal challenge. These seminal observations documented: (i) the induction of immunity by repeated inoculation with killed organisms; and (ii) the transferability of immunity with serum—the basic principle of humoral immunity.

The possible existence within organisms identified as pneumococci, of antigenically distinct groups, was suggested by Besancon and Griffon [12], and in 1910, Neufeld and Handel (cited in White [13], p.361) reported that type 1 and type 2 pneumococci could be distinguished by serologic reactions. In a meticulous study of all patients with lobar pneumonia treated at the Rockefeller Institute during three years, the great majority of whom were infected with pneumococcus, Dochez and Gillespie, in 1913 [14], used these serologic techniques to stratify the pneumococci they isolated into three groups (later called serotypes or types), numbered 1, 2 and 3 (78% of patients); they classified the remaining 22% of pneumococci as group (type) 4. Based on protection against pneumococcal challenge in mice, they found that types 1, 2 and 3 were serologically distinct; protection against pneumococcal challenge in mice was specific to each type and did not extend to challenge with other types.

Initial vaccine studies in humans

Unfortunately, the absolute importance of type specificity was apparently not considered by the first investigators who tried to apply immunization in human subjects. The massive work of Wright et al., included a total of at least 60,802 miners. In most experiments, subjects were randomized 1:1 either to receive or not to receive vaccine; in some, the randomization was 2:1 [9, 15]. Overall, these investigators correctly interpreted their data as showing a nearly 50% reduction both in cases and in deaths due to pneumonia, results that we find impressive. Criticisms about methods of randomization are trivial when compared to the essential flaw in this enormous body of work, namely, that the investigators failed to consider which type(s) of pneumococci were used in the vaccine and which were prevalent as infecting organisms at the time of the vaccination program. Wright et al. directed most of their attention to different doses of killed organisms and tried to explain inconsistent results in some experiments on dosage, but there is simply no mention of pneumococcal types, which is understandable considering that Neufeld and Handler had only reported their results in 1910 [9, 15]. Attempts at vaccination by Maynard [16], also in South African miners, very likely yielded inconsistent results because of serotype mismatch between inoculated and infecting types. In 1913, Dochez and Gillespie [14] commented on these studies, stating that “heterogeneous collections of organisms are used, and difference in origin is the only standard of differentiation of the various strains employed”.

Type-specific protection by pneumococcal vaccine

Working only a few years later, Lister, a protégé of Wright, supported the criticism by Dochez and Gillespie when he showed, in a study of 10,866 newly arrived miners, that type-specific vaccine was 100% effective in preventing infection due to pneumococcus types 1 and 2 [17]. When the camps were closed and then reopened, a new major outbreak was abruptly terminated with the use of a trivalent whole-pneumococcal vaccine.

Working in the United States, in 1917, Cecil and Austin [18] identified cases of pneumonia at military Camp Upton due to types 1, 2 or 3 pneumococcus. They inoculated varying doses of killed pneumococci from these types into a small number of subjects and studied agglutination of pneumococci and protection against mouse challenge by serum of vaccinees, as well as adverse effects of the vaccine. Based on these results, they chose to administer four weekly injections of 1–3 × 109 killed pneumococci to 12,519 subjects. An additional 19,481 subjects, who were unvaccinated were also studied. About 14,610 of the controls matched the vaccinated subjects in having been present at the camp (although the durations of their stay were unstated). The remaining control subjects were newly arrived recruits. The authors did not state how they selected men who were to be vaccinated, but it seems fair to assume that no particular selection bias might have affected the outcome. Strikingly, no cases of type 1, 2 or 3 pneumococcal infections occurred amongst the 12,519 vaccinated subjects (ignoring a single case in whom infection developed the day after the first vaccination) compared to 18 of 14,610 of the ‘matched’ unvaccinated group, again showing the efficacy of type-specific pneumococcal vaccine. It is important to note that some of the efficacy shown in studies using whole, killed pneumococci may have been due to cell constituents other than capsular, although the early failure to demonstrate a beneficial effect argues against that possibility.

Vaccination with purified capsular polysaccharides

In a remarkably logical series of experiments over the next two decades, pneumococcal capsule was shown to be the immunizing substance. In 1917, Dochez and Avery reported a “soluble specific substance” in urine and serum of infected patients [19] and in broth cultures of pneumococcus [20]; this substance precipitated when exposed to serotype-specific antiserum. Heidelberger and Avery tentatively suggested that this substance was the capsular polysaccharide (CPS) [21] and appeared to confirm that suggestion by showing that different types of pneumococcus had capsules with chemically different polysaccharides [22]. In an elegant series of experiments, Felton [23] showed that: (i) CPS was responsible for serotypes; (ii) antibody to this polysaccharide was responsible for the agglutination reaction of pneumococci; and (iii) immunization with CPS induced immunity to pneumococcal infection.

Large clinical trials of pneumococcal vaccine followed. In the winter of 1933–4, Ekwurzel, Felton et al. [24], vaccinated 3,126 members of the Civilian Conservation Corps with a preparation of equal parts of CPS from type 1 and type 2 pneumococcus. No pneumonia occurred, compared to eight cases among approximately 9,000 unvaccinated individuals (Table 1). Over the next three winters, 39,621 additional Corps members were vaccinated and 44,494 were not. Each year, the investigators continued to carefully observe their subjects, but recruitment and follow-up times were not uniform, and all of these studies were carried out over relatively short time periods. In the winter of 1937–8, Smillie and colleagues [25] used Felton’s recommended dose of type 1 pneumococcal polysaccharide to abort an outbreak of type 1 pneumococcal pneumonia at a state hospital in Worcester. Shortly thereafter, they repeated this maneuver at an annex to the main hospital. They properly stated that, “We cannot affirm, of course, that the antigen stopped the outbreak… [it] might have stopped simultaneously”.

Table 1 Vaccination of members of the civilian conservation corps, 1934–1937

During the Second World War, outbreaks of pneumococcal pneumonia in US army camps created an opportunity to do well-controlled studies of pneumococcal vaccine. In 1944, MacLeod et al. [26], used a vaccine that contained polysaccharides from pneumococcus types 1, 2, 5 and 7 in a clinical trial involving 17,035 soldiers at a US Army Technical School. With self-randomization, 8,586 men received vaccine, and 8,449 served as controls. They were followed for an average of about 6 months, and the occurrence of all cases of pneumococcal pneumonia was documented (Table 2). The incidence of pneumonia due to types 2 and 7 was significantly reduced (statistical analysis is added; this was not reported in the original paper); pneumonia caused by nonvaccine types was unchanged.

Table 2 Vaccination of soldiers at a US army technical school, 1942–4Pneumococcal vaccination in the antibiotic era

With the discovery of penicillin, interest in vaccines to prevent pneumonia waned; a systematic review of the etiology of pneumonia disclosed only four publications on the subject between 1946 and 1970 [27]. The assumption was that the problem would largely be eliminated by use of this antibiotic. Austrian and Gold [5], however, showed that, despite treatment with penicillin, deaths from pneumococcal pneumonia were unchanged in the first 96 h of therapy. In other words, penicillin reduced, but, by no means eliminated deaths due to pneumococcal pneumonia, and Austrian made it his life’s work to reactivate interest in pneumococcal vaccination [28]. He and his colleagues developed two pneumococcal polysaccharide capsular vaccines (PPSV) preparations, one containing six and one containing 12 capsular types [29], which Smit et al. [30] studied in two separate randomized controlled studies in South African miners. The vaccine efficacy for types contained in the vaccine averaged 84%. In response to this work, in 1977, Merck licensed a vaccine containing capsular polysaccharides from 14 pneumococcal types and followed this in 1983 with one containing 23 capsular polysaccharides (PPSV23 [Pneumovax®]), which is still marketed today.

In the two ensuing decades, many investigations of the efficacy of pneumococcal vaccine were carried out – including randomized control trials (RCTs), observational, case control, and indirect cohort studies. These were summarized in two widely cited meta-analyses published in 2008 by Moberley et al. (a Cochrane Review [31]), and in 2009 by Huss et al. [32]. Moberley and her colleagues included in their final analysis 22 RCTs involving 48,566 subjects and 7 non-RCTs involving 62,294 subjects. These authors concluded that PPSV reduced the risk for bacteremic pneumococcal pneumonia, non-bacteremic pneumococcal pneumonia, and all invasive pneumococcal disease (IPD) by 87%, 73% and 82%, respectively. Huss [32], a statistician, excluded all but five studies of PPSV for statistical reasons, and concluded that this vaccine produced no significant benefit; this analysis gave heavy weight to two studies that met all statistical criteria for design, but used non-validated methods for diagnosing pneumococcal infection. A second Cochrane Review by Moberley et al., in 2013 [33], reached conclusions similar to those in their report of 2008. It is worthwhile however, to give examples of studies that showed no benefit from PPSV23. In a carefully designed and executed RCT, Simberkoff et al. [34] found that, in an older male population with many comorbid conditions, there was no apparent protection from PPSV23, but this study was underpowered.

Using a case–control design, Shapiro et al. [35], showed that protection of young adults by PPSV23 at 3 years exceeded 90% and remained at > 85% for 5 years of observation. However, with each decade of increase in age of the subjects, protection declined, such that in the oldest group, no protection was observed five years post-vaccination.

Recent studies of PPSV23 have often shown much less protection than did earlier ones (Table 3) [36,37,38,39,40,41]. Possible reasons for reduced vaccine efficacy of PPSV23 in recent years include: (i) persistence of types for which vaccine (PPSV or protein-conjugate vaccine [see below]) does not appear to stimulate good protection, e.g., serotypes 3, 19A; (ii) decreasing prominence of PCV13 strains from the population with opening of an ecologic niche for other pneumococcal types; and (iii) the consequent emergence of strains not covered by PPSV23, e.g., types 15A, 23B, 35 and 38 [34].

Table 3 Recent studies of vaccine efficacy of PPSV23Protein conjugate vaccines

The principal problem with PPSV, as had been the case with the polyribosyl ribitol phosphate (PRP) capsular vaccine for Haemophilus influenzae type b, was that it failed to immunize infants and toddlers, the subjects who were also at highest risk for pneumococcal disease. To overcome this problem, Schneerson et al. [42], in 1980, showed that chemical linking or conjugation of PRP to a protein (they used CRM197, a non-toxic recombinant variant of diphtheria toxin) rendered it immunogenic. Within a few years, widespread vaccination of infants with CRM197-conjugated PRP almost eliminated H. influenzae type b meningitis from the population. Inspired by this success investigators at Wyeth, conjugated CPS of the seven most common types of Streptococcus pneumoniae to CRM197. In a seminal study [43], 37,868 infants in the Kaiser Permanente health care system were randomized to receive the 7-valent conjugate pneumococcal vaccine (Prevnar7®) or meningococcal polysaccharide (as a placebo). The incidence of pneumococcal meningitis was reduced by > 95% in infants who were fully vaccinated, and the type-specific vaccine efficacy against all pneumococcal invasive disease exceeded 97%. The lesser effect on all-cause pneumonia or otitis media reflects the multiplicity of etiologic organisms in these infections.

As had been suggested by some observers [44, 45] the establishment of mucosal immunity by PCV greatly reduced pneumococcal carriage and, therefore, pneumococcal transmission in vaccine recipients. This led to a striking secondary effect, namely, reducing or nearly eliminating infection due to vaccine strains in the population at large (Fig. 1) [46]. A similar result would later follow widespread use of a 13-valent conjugate vaccine [47].

Fig. 1

Effect of PCV7 on IPD in children < 5 years (direct effect; upper panel) and adults > 65 years (herd effect; lower panel). Note increase in non-PCV7 types, especially 19A. * From reference [46], by permission of Oxford University Press

Between 2008 and 2010, in a large and well-designed clinical trial, Bonten et al. [48], randomized 84,496 Dutch adults ≥ 65 years of age to receive a conjugate vaccine that contained CPS from 13 pneumococcal serotypes conjugated to CRM197 (PCV13) or placebo. This study excluded adults with immunocompromising conditions and subjects who were placed on an immunosuppressive drug from the final analysis. PCV13 reduced IPD and non-bacteremic pneumococcal pneumonia due to vaccine-type pneumococci (diagnosed using a serotype-specific urine antigen detection test) by 75% and 45%, respectively [48]. At the time of this trial, < 2% of Dutch adults had received a pneumococcal vaccine, childhood vaccination with PCV7 was in its early stages, and PCV13 was not yet approved for use in children, rendering comparison with the situation in the US problematic. The trial did not compare PCV13 with PPSV23, so any conclusion about the relative benefits remains inferential. While some patients in this study did develop immunosuppressing conditions, the study was not designed to show efficacy in such patients, and the number of these patients was too low to draw any meaningful conclusions regarding PCV13 efficacy in this population.

An additional population-based study of effectiveness of PCV13 vaccination in preventing hospitalization for vaccine-type CAP in adults > 65 years was undertaken in Louisville, Kentucky during 2015–2016 [49], at which time a substantial proportion of infants and children had received PCV13. Cases were those hospitalized with CAP due to PCV13 serotypes (based on culture or the serotype-specific urinary antigen detection test), whereas controls were those hospitalized with CAP, without infections caused by any of the PCV13 serotypes. This study included adults with or without compromised immunity and cases we would previously have labeled as healthcare-associated pneumonia. Vaccine effectiveness was 73% for all PCV13-type CAP and 68% for non-bacteremic PCV13-type CAP. To our knowledge, no clinical trial has directly compared vaccine efficacy of PCV13 to that of PPSV23.

Immunogenicity and safety comparisons between these two vaccines, including at least one systematic review and meta-analysis, have indicated that PCV13 elicits a better immune response among adults compared with PPSV23, while having a similar safety profile [50]; however, these studies all presented data at one month after vaccination. Supplementary data in Jackson et al. [51], however, showed that at 12-months post vaccination, there was no discernible difference in opsonophagocytic effect of serum from recipients of PCV13 or PPSV23.

There have also been cost-effectiveness studies of PCV13 versus PPSV23. Earlier studies, before the full impact of PCV13 on non-bacteremic pneumococcal pneumonia, or the magnitude of PCV13-associated indirect herd effects were known, indicated, respectively, that adding PCV13 to adult pneumococcal vaccination was favored compared with PPSV23 [52], and that cost-effectiveness of PCV13 in adults aged 50 years and older was comparable to other vaccine interventions [53]. Other studies, including one in a population in which a significant number of adults had pre-existing HIV infection, showed PCV13 to be cost-effective compared with PPSV23, both in the population as a whole and in particular in the HIV-infected population [54]. It seems highly unlikely that PCV13 followed by PPSV23 will be cost effective in countries where PCV13 has been used widely in infants and young children, since these strains have largely disappeared from the adult population.

The advisory committee for immunization practices (ACIP) recommendations for pneumococcal vaccinations in adults

The milestones in the ACIP recommendations for pneumococcal vaccination in adults are shown in Table 4, starting with updated recommendations for use of PPSV23 in adults in 2010, followed thereafter by recommendations initially for use of PCV13 in adults aged 50 years and older (this recommendation has now been archived) and then later recommendations for use of PCV13 followed by PPSV23, based on risk factors and then on age [55,56,57,58,59,60].

Table 4 Milestones in the ACIP recommendations for use of pneumococcal vaccines in adults

In 2015, the ACIP recommended PCV13 followed a year later by PPSV23 for all adults ≥ 65 years of age. The recommendation was based on several assumptions that, as Musher [61] pointed out elsewhere, may not have been well-based in fact; (i) PCV13 stimulates higher levels of anti-CPS antibody than PPSV23; (ii) this antibody persists for longer intervals; (iii) PCV13 primes the immune system for a booster response by PPSV23; and (iv) PCV13 more effectively protects immunocompromised and elderly adults than does PPSV23. Perhaps the most important reason not to follow such a recommendation in the US was the consideration that once a conjugate polysaccharide vaccine was routinely used in children, those strains would disappear from the population; but this has not been consistent in many countries outside the US, as discussed further below.

In 2019, the ACIP down-graded the recommendation, indicating that in the elderly without CSF leak, cochlear implant or immunocompromising conditions, only PPSV23 should be given routinely, leaving the matter of whether PCV13 should be given in addition, to shared decision-making by patients and doctors. However, in the case of the elderly with CSF leak, cochlear implant or immunocompromising conditions, the recommendation for the use of PCV13 followed by PPSV23 persists [60]. The decision in 2019 was based on a review of the circumstances that existed in the US regarding residual PCV13-type pneumococcal pneumonia in adults aged ≥ 65 years of age. The use of PCV13 in children had resulted in sharp declines in, and historically low levels of PCV13-type pneumonia in the elderly, with little added benefit from additional direct vaccination. However, it was recognized by the ACIP authors that in certain circumstances the elderly may be at potentially greater risk of exposure to PCV13 serotypes. These would include those living in nursing homes or long-term care facilities, or in settings with low PCV13 uptake, or those intending to travel to an area with no PCV13 immunization program, all of whom may benefit additionally from direct vaccination [60].

Some additional factors need to be considered regarding these recommendations (Table 5) [62,63,64,65,66,67,68,69,70,71,72,73,