Streptococcus pneumoniae (pneumococcus) is a leading cause of bacterial pneumonia, sepsis, and meningitis, and an important cause of morbidity and mortality globally. Use of pneumococcal conjugate vaccines in the United States childhood immunization program since the year 2000 led to dramatic declines in invasive pneumococcal disease (IPD) in both children (direct effect and indirect effect from herd immunity) and adults (indirect effect) [1,2]. Despite these declines, burden of pneumococcal disease in adults ≥65 years of age continued to be high [3]. In 2010, the Advisory Committee on Immunization Practices (ACIP) recommended that 7-valent pneumococcal conjugate vaccine (PCV7) be replaced by 13-valent pneumococcal conjugate vaccine (PCV13) for children in the United States [4]. In 2013, three years after PCV13 replaced PCV7, incidence of IPD was almost three times higher in individuals ≥65 years of age (24.8 cases per 100,000) compared to children [5]. Similar to IPD, the incidence of pneumococcal community acquired pneumonia (CAP) was five times higher among adults ≥65 years of age than in younger adults in 2012 [3]. Approximately a quarter of IPD cases in persons ≥65 years of age in 2016 and 6% of CAP cases in adults ≥18 years in 2012 are caused by the serotypes included in PCV13 and are potentially vaccine preventable [6,7].

In August 2014, ACIP recommended routine use of PCV13 in adults ≥65 years of age [8]. This decision was based in part on the results of a large randomized placebo-controlled trial (CAPiTA trial) in the Netherlands that showed a PCV13 efficacy of 45% (95% CI: 14.2–65.3%) against vaccine-type (VT) non-bacteremic pneumococcal pneumonia, and an efficacy of 75% (95% CI: 41.4–90.8%) against VT IPD [9].

Pneumococcus is often carried in the upper respiratory tract. Although pneumococcal colonization is generally asymptomatic, carriage is considered to be a precursor of IPD [10]. Host factors, such as immune response, age, and underlying medical conditions, play a role in whether an individual develops disease following colonization [11]. Use of PCV13 in adults ≥65 years of age has the potential to impact pneumococcal colonization rates in this age group and therefore disease. However, colonization decreases with age [12,13] and the prevalence of pneumococcal colonization in older Americans after introduction of PCV13 in children in 2010 has not been well evaluated. In the United Kingdom, after PCV7 and prior to PCV13 introduction into the national infant immunization program, a pneumococcal carriage survey conducted in adults ≥ 65 years of age found a 2.2% pneumococcal carriage rate [14]. A study done in adults with HIV infection (19–66 years of age) in the United States from 2005 to 2007 (post PCV7 introduction) found 3.4% pneumococcal carriage [15,16]. In Alaska, immunization of children with PCV13 led to substantial declines in PCV13-type pneumococcal colonization among adults ≥45 years of age through indirect effects [16].

Given the high IPD and CAP incidence in the U.S. elderly population, we sought to define pneumococcal carriage prevalence and serotype distribution and to describe risk factors for pneumococcal colonization among adults ≥65 years of age in the United States before widespread use of PCV13 in this age group. These results can help to understand the potential impact of the ACIP recommendation for routine PCV13 use in U.S. adults ≥65 years of age on pneumococcal colonization rates.

We enrolled 3011 adults; 2989 (99.3%) had good specimen quality and were included in the analysis (Fig. 1). Median age was 74 years (range 65–102 years), 1927 (64.5%) were female, and 2179 (72.9%) were white (Table 1). Most participants reported living alone or with one other person. Few participants (<1%) reported any children under five living in the same household. Approximately 7% of participants reported smoking, and 48% reported having at least one chronic medical condition. Less than 1% of enrolled participants reported having pneumonia or influenza in the month prior to enrollment, and 6% of participants reported taking any antibiotic in the prior 2 weeks. Most participants reported having both Medicare and private supplemental health insurance.

Following PCV13 introduction in adults ≥65 years of age, carriage of pneumococcus was rare among those who were not nursing homes residents or did not have severe immunosuppressing conditions. Only presence of respiratory symptoms was associated with carriage, which is consistent with other studies [12,22,23]. Our study enrolled a large sample size across four U.S. states, which allowed for a greater opportunity to detect VT serotypes. PCV13 serotypes accounted for 14% of those with pneumococcus detected and an overall 0.3% carriage rate.

Indirect effects from years of pediatric vaccination likely contributed to the very low carriage rate. Consecutive carriage surveys among children in Georgia pre- and post- PCV13 introduction into the national childhood immunization schedule showed a dramatic decline in VT pneumococcal colonization from 29% in 2010 to 3% in 2013, following PCV13 introduction [24]. Introduction of pneumococcal conjugate vaccines into the routine infant immunization schedule has been shown to decrease VT carriage in adults in the United States and in Kenya [16,25]. It is also possible that direct effects of PCV13 vaccination contributed to the observed low VT carriage rates, given the PCV13 coverage of 45.3% among study participants. Therefore, it is not possible to determine the individual contributions of direct effect versus indirect effects of PCV13 vaccination in this study.

The prevalence of overall and VT pneumococcal carriage in our study was similar to what the United Kingdom reported in the same age group (2.2% and 0.5%, respectively) after PCV7 introduction and before PCV13 introduction into their national infant immunization program [14]. However, more recent studies show higher carriage prevalence. A study among older adults with influenza-like illness found carriage rates as high as 36% [26]. A recent longitudinal study among U.S. adults ≥65 years of age found a monthly pneumococcal carriage prevalence ranging from 0 to 17% [27]. Different laboratory methods have been utilized for detection of pneumococci in adults making comparison between studies difficult. Most studies showing high colonization rates in adults include real-time PCR targeting lytA on either OP swabs or saliva, increasing detection rate by at least five fold. It has been demonstrated that non-pneumococcal streptococci (e.g., S. mitis, S. oralis, S. infantis) carry homolog genes of pneumococci, which could lead to false-positive detection and PCR serotyping results [20,28]. This is more problematic with OP or saliva specimens because of the diversity of streptococci that can be found in the oropharynx [29]. Due to this diversity, a lytA detection from an OP or saliva specimen may not be from the same organisms DNA as the DNA from which the serotype was detected, and most published studies do not report Ct values to allow for comparison. In general, a reported Ct value from the serotyping should not be lower than that of the lytA due to the fact that the methodology for serotyping uses multiplex assays which require more DNA copies in order to detect while the lytaA assay is a singleplex requiring less copies for detection. Therefore, it is possible that pneumococcal carriage surveys relying on PCR of oral specimens may have overestimated pneumococcal carriage rates [14], even though molecular methods may be better suited to detecting lower density colonization which is commonly found in the adult population when compared to culture. Also, less than 1% of our adult participants had children living in the same household, and the median number of household members was one. It is well known that children are the main reservoir of S. pneumoniae, and that residing in a crowded household is a risk factor for pneumococcal colonization [30]; none of those were common among our study population.

Among the eight participants with VT pneumococcal carriage, the serotypes isolated (19F, 19A and 3) are also the most common vaccine serotypes causing IPD among adults ≥65 years of age in the United States [31]. Antimicrobial susceptibility profiles of the 45 pneumococcal isolates were similar to what is reported for IPD cases in the United States, with high prevalence of erythromycin and tetracycline resistance. We observed a high prevalence of penicillin non-susceptible isolates (28.9%) using the non-parenteral CLSI breakpoint for resistant isolates. However, if we had used the non-meningitis parenteral breakpoint, the prevalence of non-susceptible isolates would decrease to 2.2%, which is similar to that of isolates from IPD cases in the United States (4.3% of 2597 invasive pneumococcal isolates) [32] and from healthy children 6–59 months of age in Georgia (3.0% of 99 NP carriage isolates) [24].

Our study provides insights on factors associated with pneumococcal carriage and with PCV13 receipt among study participants. For pneumococcal carriage, we found an association of respiratory symptoms in the past 2 weeks, which is not surprising given that the risk of pneumococcal acquisition increases following acute respiratory illness with influenza or parainfluenzae [33]. PCV13 receipt was associated with supplementary health insurance, health care exposures since the ACIP PCV13 recommendation in August 2014, and chronic medical conditions. Participants with chronic medical conditions and supplemental insurance likely had higher numbers of healthcare encounters, providing more opportunities to receive PCV13. When compared to vaccination coverage estimates using Centers for Medicare & Medicaid Services data, our study showed a higher rate of PCV13 vaccination among older Americans (45.3% vs 31.5%) [34]. However, we also found lower rates of vaccine coverage among non-white individuals when compared to white individuals, which is consistent with other reports [35].

There are some limitations to our study. First, vaccination with PCV13 was quite high and colonization was very low at the time of enrollment, precluding us from obtaining a true baseline carriage estimate prior to PCV13 introduction in adults. High vaccine coverage could have contributed to the low observed colonization rates; however, when we stratified overall and VT pneumococcal carriage by PCV13 vaccination status, the observed rates were similar. Second, we did not enroll participants with severe immunosuppressing conditions or residents of nursing homes, which does not allow us to make inferences about pneumococcal colonization rates in these populations. Based on a number of factors, it is possible that pneumococcal carriage rates are higher in nursing home residents than what we observed in adults in the community [36,37]. Third, even though we included both NP and OP swabs to measure the pneumococcal colonization rate in adults, we relied solely on a culture-based method for OP specimens, given that current molecular methods for pneumococcal detection may lead to false-positive results [20,28]. Therefore, it is possible that we underestimated pneumococcal colonization. Finally, our findings might not be generalizable to the entire U.S. population ≥65 years of age. Compared to the 2015 American Community Survey from the U.S. Census Bureau [38], the median age of our study participants and the percentage with health insurance were similar to those of the U.S. population. However, we enrolled a higher percentage of African Americans and females and a far smaller percentage of Hispanics compared to the U.S. population.

This is the largest pneumococcal carriage survey done among adults ≥65 years of age in the United States. Pneumococcal colonization of the upper respiratory tract among older Americans was rare after over a decade of PCV use in children and after PCV13 introduction in older adults. Both vaccinated and unvaccinated participants in our study had a carriage rate of less than 0.5% for VT pneumococcal serotypes. Many years of PCV vaccination in pediatric populations likely led to significant indirect effects in adults. However, recent PCV13 recommendation for all U.S. adults ≥65 years of age may have also contributed to the low observed rates.