Secondary Logo

Journal Logo

Infectious Diseases

Competition Between Streptococcus Pneumoniae Strains: Implications for Vaccine-Induced Replacement in Colonization and Disease

Mehtälä, Juha; Antonio, Martin; Kaltoft, Margit S.; O’Brien, Katherine L.; Auranen, Kari

Author Information
doi: 10.1097/EDE.0b013e318294be89

Streptococcus pneumoniae (pneumococcus) is a bacterium with more than 90 known serotypes.1 Pneumococcal colonization of the nasopharynx, which only occasionally progresses to disease, is especially common among young children,2 and during the first year of life the prevalence of colonization can reach as high as 90%.3 Pneumococcal conjugate vaccines are currently directed against up to 13 serotypes and are highly efficacious in protecting against the most serious forms of pneumococcal disease caused by serotypes included in the vaccine as well as reducing acquisition of vaccine-type colonization.4–7 The reduced vaccine-type colonization may open a niche in the nasopharynx allowing for increases in the acquisition and prevalence of pneumococcal nonvaccine serotype colonization and subsequent disease.8 Such replacement has been reported in many vaccinated populations9–13 implying that pneumococcal serotypes do not colonize human hosts independently but competitively interact with each other.

Further evidence for between-strain competition has been provided by experimental inoculation studies in mice14 and in vitro studies,15 which have also indicated possible differences in the competitive strength across pneumococcal serotypes. The biological basis for intraspecies competition in pneumococci could be competition for contact sites or resources. Another explanation is direct killing mediated by bacteriocins (pneumocins) targeting other members of the species and enhancing lateral gene transfer.16,17 However, many questions about pneumococcal strain competition and replacement remain unanswered. In particular, it is not clear to what extent the magnitude of replacement could be predicted by the specific mechanism of competitive interactions between pneumococcal strains (serotypes).

The mechanism of competition may work through reduced acquisition of new serotypes due to a person’s current colonization by other serotypes. In addition, clearance of colonization may be enhanced or its density decreased by concurrent colonization by other serotypes. Another type of interaction may be mediated by immunity acquired at colonization and protecting against subsequent colonization by the specific serotype18 or pneumococci in general.19 The mechanism of competition may influence the dynamics of colonization and, subsequently, the impact of vaccination on disease. Most importantly, as new acquisitions pose the highest risk for subsequent development of disease,20,21 competition in acquisition is the mechanism that leads most directly to replacement in disease by the nonvaccine serotypes.

Longitudinal data among humans are essential to learn about competition between pneumococcal strains in their natural habitat. Analyses of repeated colonization measurements have provided clear evidence for between-serotype competition and for differences across serotypes in the strength of competition.22,23 A recent analysis based on longitudinal data with sensitive measurement of multiple colonization indicated that the most likely mechanism of competition would be reduced acquisition due to current colonization, rather than enhanced clearance of simultaneously colonizing strains.24

Empirical data of pneumococcal colonization may suffer from limitations that hinder the analysis of between-strain competition. Colonization data usually consist of observations of a single serotype per subject at a time.22,23 The analysis of competition is then limited to “knock-out” events in which one serotype instantaneously replaces another as the only colonizing strain. Models addressing the possibility of multiple colonization are more realistic but require empirical observations. In studies of multiple colonization, approximately 10% of colonized persons have harbored more than one strain simultaneously.3,24,25 However, the sensitivity to detect multiple colonization may not be perfect. Recently, it has been shown that conventional methods in particular underestimate the prevalence of multiple colonization.26 The analysis of competition is further complicated by the discrete nature of observations. In particular, the longer the time interval between consecutive observations the more difficult it becomes to describe the underlying dynamics due to missing episodes of colonization in the observed data.

We use three datasets, each with repeated measurements of pneumococcal colonization among infants or young children, and with special efforts to detect simultaneously colonizing serotypes. The data from Danish day-care attendees have been analyzed earlier to investigate between-strain competition.24 Here we reanalyze these data by taking into account the discrete nature of observations with a refined method, and allowing the possibility of imperfect sensitivity to detect multiple colonization. The two other datasets from studies among American Indian and Gambian infants are evaluated to assess similarities or differences in competition across different epidemiological settings. We also address the question of possible differences in competition between individual serotypes. In general, we characterize potential biases in estimation of competition due to insensitive detection of multiple colonization and discrete observations. Finally, we discuss the importance of competition and its mechanism on serotype replacement in pneumococcal colonization and disease.


We analyze three datasets of longitudinal measurements of pneumococcal colonization in infants and young children. Written informed consent was obtained from the parents or guardians of each child. The studies are summarized in Table 1, described briefly below and in more detail elsewhere.3,24,25

Summary of Three Datasets of Longitudinal Colonization in Children

The Danish study involved nasopharyngeal samples collected with a monthly interval from children who attended day care, their caretakers, and relatives of the children. Here we use data from the day-care children with at least two successive samples. In the statistical analysis, we modeled nine strains: the eight most common serotypes (accounting for 79% of positive samples) and the remaining serotypes combined into a single category.

The American Indian study involved scheduled samples of pneumococcal conjugate–vaccinated and control-vaccinated infants at 7, 12, and 18 months of age. In addition, two samples were taken 1 and 3 months after the scheduled samples from infants who were themselves colonized, or for whom any other child in their household was found to be colonized. Here we use data from the infants in the control arm of the study. As above, we modeled the eight most common serotypes, accounting for 56% of positive samples, and the rest combined into a single category.

In the Gambian study, participants consisted of infants who were observed with a 2-week sampling interval until 6 months of age and with a bimonthly interval from 6 months to 1 year of age. Here we modeled 14 strains: the 13 most common serotypes, accounting for 75% of positive samples, and the rest combined into a single category.

Different approaches were used in the three studies with regard to transport medium and identification of multiple serotypes. Otherwise, standardized methodologies were used for identification and growth of pneumococci. The standard method for detecting colonization (including skim milk–tryptone-glucose-glycerol transport medium) was used both in the American Indian and Gambian studies.27 In the American Indian study, detection of multiple colonization was done by an immunoblot method and confirmed by Quellung.28 In the Gambian study, the sweep serotyping method was applied for detection of multiple serotypes. In the Danish study, direct serotyping of an enrichment broth was applied,29 and if more than one serotype was detected, several colonies from a secondary plate were serotyped in order to find and isolate the pneumococcal strains.

The Model of Between-Strain Competition

We assume a person can be colonized with up to two strains simultaneously. In a model for n strains, the possible states for any person thus are: noncolonized, colonized with one of the n strains, and colonized with any two strains simultaneously. The total number of states is n(n + 1)/2 + 1. Figure 1 presents the model for 3 strains for conciseness; in the analysis there are 9 or 14 strains depending on the dataset (see above).

The epidemiologic states of colonization and transition rates between the states in a model for 3Streptococcus pneumoniae strains (x, y, and s). State (0) denotes noncolonization and state (x, y) simultaneous colonization with strains x and y. In the analysis, strain s is the target strain with 4 strain-specific competition parameters (θ(x),(x,s) = θ(y),(y,s), θ(s),(x,s)= θ(s),(y,s), φ(x,s),(x) = φ(y,s),(y), φ(x,s),(s) = φ(y,s),(s)). For the nontarget strains (x, y), the competition parameters are assumed to be equal (θ(x),(x,y) = θ(y),(x,y) = θ and φ(x,y),(x) = φ(x,y),(y) = φ). Generally, the number of states in the model for n strains is (n [n + 1]/2) + 1.

Let the acquisition and clearance rates of the n strains be λ(0),(k) and λ(k),(0), k = 1, …, n. If the host is already colonized with strain x, the acquisition rate of strain y is denoted by λ(x),(x,y). Similarly, if the host is simultaneously colonized with another strain x, the clearance rate of strain y is λ(x,y),(x). We quantify competition in acquisition and clearance in term of relative rates as follows:

The parameter θ(x),(x,y) is the relative rate of acquisition of strain y in a person currently colonized with another strain(x), compared with being noncolonized. Values of θ(x),(x,y) below 1 indicate competition in acquisition. Likewise, φ(x,y),(x) is the relative rate of clearance of strain y in someone simultaneously colonized with another strain (x), compared with singly colonized with strain y. Values of φ(x,y),(x) above 1 indicate competition in clearance.

For strain y, we define its strength of competition against strain x as 1 − θ(x),(x,y)(x,y),(x). It is the relative reduction in the expected time spent doubly colonized per time unit spent singly colonized with strain y, compared with no competition (eAppendix A, The stronger competition is, the closer the strength of competition is to 1. Furthermore, we define the mode of competition for y and x as whether competition appears in acquisition (θ(x),(x,y) <1) or in clearance (φ(x,y),(x) > 1). In presence of competition, the mode describes the mechanism of reducing double colonization as either reduced acquisition or reduced duration of double colonization.

Model Parameterization and Statistical Analysis

We allow each of the n strains to have its own acquisition and clearance rates. To avoid too many parameters in the model, however, the competition parameters are assumed to be shared, that is, θ(x),(x,y) = θ(y),(x,y) = θ and φ(x,y),(x) = φ(x,y),(y) = φ, for all strains x and y except for one target strain s∈{1,…, n}. The two parameters θ and φ represent the average competition among all strains excluding s.

For the target strain s, competition is described in terms of four parameters:

where x is any strain in the model, distinct of s. These strain-specific parameters represent competition for the target strain s (strain s against the nontarget strains overall, and the nontarget strains against s). Possible differences in competition between strains are addressed by comparing the strain-specific competition parameters to the overall competition parameters.

Using the above model parameterization, each of the three datasets is analyzed separately. In the analyses there are n strains (Denmark and American Indian: n = 9; The Gambia: n= 14) of which n−1 are nontarget strains. The target strains are Denmark, 23F; American Indian, 6A; and The Gambia, 19F. In the Gambian dataset, an additional analysis is performed with serotype 6B as the target strain. Altogether 6 competition parameters and 2n rate parameters (the strain-specific acquisition and clearance rates) are estimated.

We employ a hidden Markov model that allows imperfect observation of doubly colonized states. Specifically, the model allows any doubly colonized state to be detected as either the true state or as a singly colonized state with one of the two strains involved. The sensitivity to detect both strains is denoted by ν∈[0,1]. In case a doubly colonized sample is observed as singly colonized (with probability 1-ν), we assume that either of the two strains is detected with the same probability 0.5. This is a neutral assumption in the absence of information of how sensitivity depends on density of colonization. The detection sensitivity of double colonization (ν) is not subject to estimation. In addition to value ν = 1.0 (perfect sensitivity), ν = 0.5 is used as a scenario to investigate how estimates of competition change if the sensitivity is not nearly perfect. In a complementary study, ν = 0.15 is employed in a model with a reduced number of parameters (eAppendix B, With simulation experiments, multiple combinations of the true detection sensitivity and the possibly misspecified sensitivity used in the actual analysis are studied to characterize the direction of bias due to model misspecification.

Statistical inference is based on the Bayesian posterior distribution of the model parameters. The posterior mean and 90% credible interval (CI) are reported. To approximate the posterior distribution, a Markov chain Monte Carlo algorithm is employed. Detailed methods are presented in eAppendix A (


Figure 2 shows the observed serotype distributions in the three datasets. In each of them, the same pediatric serotypes (23F, 19F, 6A, 6B, 14) are among the most prevalent ones. For most serotypes, the distributions among all isolates were similar to those among isolates in doubly colonized samples. Clear exceptions were serotypes 6B and 3 in the Gambian dataset, with 6B being less prevalent and 3 being more prevalent, among the doubly colonized samples. Table 2 presents the estimated competition parameters for the three datasets and Table 3 the results for the additional analysis for serotype 6B as the target strain in the Gambian dataset. eAppendix B and eAppendix C ( present the complete results and the model assessment, respectively.

Estimates of Competition Parameters
Estimates of Competition Parameters for Serotype 6B in the Gambian Data
Serotype distribution among all isolates (black) and among isolates detected in doubly colonized samples (grey). The total numbers of isolates and the numbers of isolates in doubly colonized samples are (515, 100; Denmark), (1,191, 144; American Indian), and (2,752, 452; The Gambia).

Competition Assuming Perfect Detection Sensitivity

Assuming perfect sensitivity (ν = 100%), the posterior mean of the overall competition strength (1 − θ/φ) was estimated at 0.89 (90% CI = 0.84–0.93) in the Danish data, 0.92 (90% CI = 0.88–0.95) in the American Indian, and 0.96 (90% CI = 0.95–0.97) in the Gambian data.

In the Danish data, competition was identified completely in acquisition, that is, θ was approximately 0.1 whereasφ was close to 1. In the American Indian data, competition was identified in acquisition (θ ≈ 0.28) as well as in clearance (φ ≈ 2.6), and in the Gambian data, competition was identified in acquisition (0.5-fold) but it was considerably stronger (14-fold) in clearance.

The estimates of the target-specific competition parameters for serotype 23F in the Danish dataset as well as for serotype 19F in the Gambian dataset were similar to the overall competition. For serotype 6B in the Gambian dataset, competition in clearance was stronger than the overall competition. The American Indian data were too sparse to identify serotype-specific competition parameters.

Competition Assuming Imperfect Detection Sensitivity

Assuming imperfect sensitivity (ν = 50%), the overall competition strength was estimated weaker than with perfect sensitivity in all three datasets: Denmark 0.65 (90% CI = 0.46–0.81), American Indian 0.73 (0.63–0.81), and The Gambia 0.88 (0.86–0.90).

In each dataset, competition in acquisition (θ) was approximately at the same level as when assuming perfect detection sensitivity. In contrast, the parameters for competition in clearance (φ) were estimated to be smaller: 0.3 in the Danish data, 1.0 in the American Indian data, and 3.2 in the Gambian data.

For the target serotypes in the Danish dataset (23F) and the Gambian dataset (19F), the serotype-specific estimates of competition parameters were again similar to the overall competition. For serotype 6B in the Gambian dataset, competition in clearance was estimated stronger than the overall competition. Assuming very poor sensitivity (ν = 15%) in a model with a reduced number of parameters, competition in acquisition was still estimated below 1 (eAppendix B,

Simulation Experiment

To investigate how estimation of competition parameters depends on the sensitivity to detect double colonization and the sampling interval, we performed a simulation study. Specifically, we investigated how robust inferences about between-strain competition are if the detection sensitivity in the analysis is misspecified. Figure 3 shows the biases in the posterior means of the two competition parameters (θ and φ), if the detection sensitivity used in estimation (νest) differs from its true value (νtrue). The biases are shown for two different sampling intervals: short enough so that repeated observations of any double colonization episode are likely, and long so that repeated observations are not likely.

Estimates of how competition parameters (θest for acquisition and φest for clearance) depend on the relationship between the true detection sensitivity (νtrue) and the detection sensitivity used in estimation (νest). Ratios comparing the posterior means of the competition parameters to their true values (θest/θ, φest/φ) are shown for data simulated under three parameters sets and four values of νtrue/νest (see the eAppendix A, The vertical and horizontal lines indicate where the posterior mean is equal to its true value (θest = θ, φest=φ). The data on the left were collected using a short sampling interval and likely include repeated observations of double colonization episodes. The data on the right were collected with a long sampling interval and are thus less likely to include repeat observations.

With the short sampling interval, the estimation of competition in acquisition is robust to discrepancies between νtrue and νest. In contrast, competition in clearance appears artificially strong if νtrue is smaller than νest, that is, if the data are recorded with poorer sensitivity than accounted for in the analysis.

With the long sampling interval, either or both of the competition parameters could be estimated incorrectly. If νtrue is smaller than νest there is a tendency for both competition in acquisition and in clearance to appear artificially strong. This indicates the general finding that it is more difficult to identify the mode of competition with long sampling intervals.


Insensitive detection of multiple simultaneously colonizing strains (serotypes) poses a serious challenge for the statistical analysis of between-strain competition in pneumococci. The problem is exacerbated by the fact that the actual times of acquiring and clearing strains cannot be observed, so that the analysis needs to rely on repeated measurements of the current status of colonization. We employed a statistical model that addresses the possibility of imperfect detection of double colonization as well as the discrete nature of observations. In each of the three longitudinal datasets (Danish, American Indian, and Gambian), between-strain competition was found to be strong when assuming perfect detection sensitivity and remained considerable even when assuming clearly imperfect sensitivity. Regardless of the assumed level of sensitivity, the mode of competition was always identified in acquisition. Inferences about competition in clearance were more susceptible to assumptions about the detection sensitivity.

In our analysis, the information for inferring the strength of between-strain competition derives from the proportion of double colonization among positive samples. Assuming that the serotypes compete only weakly or not at all, this proportion should be above 40% in each of the three datasets (eAppendix C,—clearly in excess to what was actually observed. Poor sensitivity to detect double colonization would artificially lower the observed proportions of double colonization among positive samples and could thus explain what appears as competition. Nevertheless, as the finding of strong competition remained even when allowing only 50% detection sensitivity, it should be clearly lower than this to fully account for the low observed level of multiple colonization.

Identification of the actual mechanism of competition is sensitive to the ability to capture the duration of double colonization with sufficient accuracy. The extent to which this is a problem depends greatly on the sampling interval with which the data are collected. With too long a sampling interval (so that repeated observations of the same underlying episode of double colonization are not likely) poor sensitivity causes some episodes to remain completely unobserved. Competition in both acquisition and clearance may then appear artificially strong (Figure 3; right panel). In contrast, if the sampling interval is sufficiently short to allow repeated observations from the same episode of double colonization, the most likely effect of poor sensitivity is to shorten the duration of these episodes. Because the frequency of acquisition events could still be observed only competition in clearance would appear artificially strong (Figure 3; left panel).

The simulation results (Figure 3) may well correspond to the differences in our empirical estimates of competition between scenarios with 100% and 50% sensitivities. In particular, assuming perfect sensitivity, enhanced clearance of simultaneously colonizing serotypes was not found in the Danish data, was present in the American Indian data, and was strong in the Gambian data. Assuming imperfect sensitivity (50%), such competition in clearance was estimated weaker in each dataset. Based on the Danish data with assumed 50% sensitivity, simultaneous colonization of two serotypes appeared to decelerate the clearance of both. This seems biologically implausible, suggesting that at least in this setting the sensitivity could not have been very low.

Although one should be careful to draw firm conclusions regarding differences in frequency of double colonization based on the methods used in the three studies, it can be expected that sweep serotyping (The Gambia) and the direct serotyping of an enrichment broth culture (Denmark) had a higher sensitivity than the method applied in the American Indian study. Thus, unaccounted poor sensitivity does not seem to explain all differences in the estimated competition in clearance between the datasets. An alternative explanation could be that the strength of competition in clearance depends on age, as the study subjects were oldest in the Danish data and youngest in the Gambian data.

Clearance rates have previously been shown to be lower for younger age groups.30,31 In agreement with this, the clearance rates were estimated to be highest in the Danish data and lowest in the Gambian data. The fact that the prevalence of colonization in the Gambian study quickly increased from very low levels up to 90% could bias the estimation of clearance rates because only few events of clearance could be identified in the observed data. Although the clearance rates estimated from the Gambian data agreed with previous studies, we reanalyzed these data fixing the clearance rates at 0.75 per month, that is, clearly higher than what was initially estimated. In this analysis, competition was identified mostly in acquisition (data not shown).

The estimation of competition in acquisition could be confounded if some unadjusted background variables were associated with both the colonization status (colonized/noncolonized) and exposure to acquisition. For instance, persons could belong to subpopulations dominated by different serotypes (microepidemics).32 An analysis not adjusted for exposure could then give too strong estimates for competition in acquisition. However, for example, for the relevant subpopulations in the Danish data (day-care centers), the presence of any particular serotype was not associated with other types’ absence at the same time (eAppendix C, Another possible confounder is age. For example, newborns are less likely colonized than older children. If older children were more exposed, an analysis not adjusted for age could lead to too weak estimates of competition in acquisition. This could be the case with the analysis of the Gambian dataset, in which the subjects were noncolonized at birth and after the first acquisition virtually always were found to be carriers of some serotype.

We found no differences in competition between individual serotypes and the average behavior of the nontarget types, expect for 6B in the Gambia. In general, the serotype-specific data were scarce, and the analysis in the Danish and the American Indian datasets could be made only for the most prevalent types. Type 6B, which indicated strong competition in clearance in the Gambian dataset, has shown differences in competition also in earlier studies.22 In all three datasets, type 6B had a lower proportion of doubly colonized isolates than its proportion of all isolates, in contrast to most other types. An opposite relation was observed for serotype 3.These observed disparities could be related to the capsule size of the serotypes, which has been related to the ability to compete.15 Alternatively, the sensitivity to detect serotypes from doubly colonized samples could be different due to the capsule size. For example, in the case of serotype 3, the Quellung test often reveals a large capsular reaction and therefore may be easier to observe.

Most previous epidemiological studies of competition between pneumococcal strains have allowed only singly colonized states (eg, references 22 and 23). In such models, competition cannot be described in terms of the (relative) frequency of double colonization but rather in terms of the (relative) frequency of “displacement events” in which serotypes “knock out” each other as the one type carried. Moreover, previous studies vary in how they account for the possibility of unobserved episodes of colonization from discrete data. Straightforward comparisons of inferences about competition are therefore problematic. If competition in acquisition was very strong, that is, if most transitions from one serotype to another involved a noncolonized state in between, as implied by our analysis, even models not allowing doubly colonized states should be able to identify competition in acquisition. However, if the model treats the “displacement events” as such and not as transitions through the noncolonized state, competition in acquisition is likely to be underestimated.

The question of the mechanism of competition is of great significance because the dynamics of colonization are different depending on whether competition occurs in acquisition or clearance. If competition occurs predominantly in acquisition, the underlying rate of acquisition events maintaining a certain prevalence of double colonization is smaller than if competition occurs in clearance. In case some strains are eliminated (eg, by vaccination) under such strong competition as found in our analyses, the noneliminated strains replace colonization regardless of the mode of competition, but the number of acquisition events by the noneliminated strains increases more if the mode is in acquisition. Because each acquisition poses a specific risk event for subsequent development of disease,20,21 replacement in different disease manifestations could be milder if competition appeared mainly in clearance.

In summary, our analysis supports rather strong between-strain competition in pneumococci even when assuming 50% sensitivity in detection of double colonization. In addition, the strongest evidence here indicates that once a strain has established its status as a colonizing strain, it prevents other strains from further colonizing the same host. It remains unclear how strong competition in clearance is in different settings and whether it weakens with age.


1. Park IH, Pritchard DG, Cartee R, Brandao A, Brandileone MC, Nahm MH. Discovery of a new capsular serotype (6C) within serogroup 6 of Streptococcus pneumoniae. J Clin Microbiol. 2007;45:1225–1233
2. Principi N, Marchisio P, Schito GC, Mannelli S. Risk factors for carriage of respiratory pathogens in the nasopharynx of healthy children. Ascanius Project Collaborative Group. Pediatr Infect Dis J. 1999;18:517–523
3. Hill PC, Cheung YB, Akisanya A, et al. Nasopharyngeal carriage of Streptococcus pneumoniae in Gambian infants: a longitudinal study. Clin Infect Dis. 2008;46:807–814
4. Black S, Shinefield H, Fireman B, et al. Efficacy, safety and immunogenicity of heptavalent pneumococcal conjugate vaccine in children. Northern California Kaiser Permanente Vaccine Study Center Group. Pediatr Infect Dis J. 2000;19:187–195
5. Klugman KP, Madhi SA, Huebner RE, Kohberger R, Mbelle N, Pierce NVaccine Trialists Group. . A trial of a 9-valent pneumococcal conjugate vaccine in children with and those without HIV infection. N Engl J Med. 2003;349:1341–1348
6. O’Brien KL, Moulton LH, Reid R, et al. Efficacy and safety of seven-valent conjugate pneumococcal vaccine in American Indian children: group randomised trial. Lancet. 2003;362:355–361
7. Cutts FT, Zaman SM, Enwere G, et al.Gambian Pneumococcal Vaccine Trial Group. Efficacy of nine-valent pneumococcal conjugate vaccine against pneumonia and invasive pneumococcal disease in The Gambia: randomised, double-blind, placebo-controlled trial. Lancet. 2005;365:1139–1146
8. Lipsitch M. Vaccination against colonizing bacteria with multiple serotypes. Proc Natl Acad Sci U S A. 1997;94:6571–6576
9. Eskola J, Kilpi T, Palmu A, et al.Finnish Otitis Media Study Group. Efficacy of a pneumococcal conjugate vaccine against acute otitis media. N Engl J Med. 2001;344:403–409
10. Huang SS, Platt R, Rifas-Shiman SL, Pelton SI, Goldmann D, Finkelstein JA. Post-PCV7 changes in colonizing pneumococcal serotypes in 16 Massachusetts communities, 2001 and 2004. Pediatrics. 2005;116:e408–e413
11. Lexau CA, Lynfeld R, Danila R, et al. Changing epidemiology of invasive pneumococcal disease among older adults in the era of pediatric pneumococcal conjugate vaccine. JAMA. 2005;294:2043–2051
12. Hicks LA, Harrison LH, Flannery B, et al. Incidence of pneumococcal disease due to non-pneumococcal conjugate vaccine (PCV7) serotypes in the United States during the era of widespread PCV7 vaccination, 1998-2004. J Infect Dis. 2007;196:1346–1354
13. Weinberger DM, Malley R, Lipsitch M. Serotype replacement in disease after pneumococcal vaccination. Lancet. 2011;6736:62225–62228
14. Lipsitch M, Dykes JK, Johnson SE, et al. Competition among Streptococcus pneumoniae for intranasal colonization in a mouse model. Vaccine. 2000;18:2895–2901
15. Weinberger DM, Trzciński K, Lu YJ, et al. Pneumococcal capsular polysaccharide structure predicts serotype prevalence. PLoS Pathog. 2009;5:e1000476
16. Dawid S, Roche AM, Weiser JN. The blp bacteriocins of Streptococcus pneumoniae mediate intraspecies competition both in vitro and in vivo. Infect Immun. 2007;75:443–451
17. Johnsborg O, Eldholm V, Bjørnstad ML, Håvarstein LS. A predatory mechanism dramatically increases the efficiency of lateral gene transfer in Streptococcus pneumoniae and related commensal species. Mol Microbiol. 2008;69:245–253
18. Weinberger DM, Dagan R, Givon-Lavi N, Regev-Yochay G, Malley R, Lipsitch M. Epidemiologic evidence for serotype-specific acquired immunity to pneumococcal carriage. J Infect Dis. 2008;197:1511–1518
19. Granat SM, Ollgren J, Herva E, Mia Z, Auranen K, Mäkelä PH. Epidemiological evidence for serotype-independent acquired immunity to pneumococcal carriage. J Infect Dis. 2009;200:99–106
20. Gray BM, Turner ME, Dillon HC Jr. Epidemiologic studies of Streptococcus pneumoniae in infants. The effects of season and age on pneumococcal acquisition and carriage in the first 24 months of life. Am J Epidemiol. 1982;116:692–703
21. Syrjänen RK, Auranen KJ, Leino TM, Kilpi TM, Mäkelä PH. Pneumococcal acute otitis media in relation to pneumococcal nasopharyngeal carriage. Pediatr Infect Dis J. 2005;24:801–806
22. Melegaro A, Choi Y, Pebody R, Gay N. Pneumococcal carriage in United Kingdom families: estimating serotype-specific transmission parameters from longitudinal data. Am J Epidemiol. 2007;166:228–235
23. Lipsitch M, Abdullahi O, D’Amour A, et al. Estimating rates of carriage acquisition and clearance and competitive ability for pneumococcal serotypes in Kenya with a Markov transition model. Epidemiology. 2012;23:510–519
24. Auranen K, Mehtälä J, Tanskanen A, et al. Between-strain competition in acquisition and clearance of pneumococcal carriage - epidemiologic evidence from a longitudinal study of day-care children. Am J Epidemiology. 2010;171:169–176
25. O’Brien KL, Millar EV, Zell ER, et al. Effect of pneumococcal conjugate vaccine on nasopharyngeal colonization among immunized and unimmunized children in a community-randomized trial. J Infect Dis. 2007;196:1211–1220
26. Turner P, Hinds J, Turner C, et al. Improved detection of nasopharyngeal cocolonization by multiple pneumococcal serotypes by use of latex agglutination or molecular serotyping by microarray. J Clin Microbiol. 2011;49:1784–1789
27. O’Brien KL, Nohynek HWorld Health Organization Pneumococcal Vaccine Trials Carriage Working Group. . Report from a WHO Working Group: standard method for detecting upper respiratory carriage of Streptococcus pneumoniae. Pediatr Infect Dis J. 2003;22:e1–11
28. Bronsdon M, O’Brien KL, Whitney CG, et al. Detection of multiple serotypes of Streptococcus pneumoniae in nasopharyngeal secretions using an immunoblot method. J Clin Microbiol. 2004;42:1596–1600
29. Kaltoft MS, Skov Sørensen UB, Slotved HC, Konradsen HB. An easy method for detection of nasopharyngeal carriage of multiple Streptococcus pneumoniae serotypes. J Microbiol Methods. 2008;75:540–544
30. Högberg L, Geli P, Ringberg H, et al. Age- and serogroup-related differences in observed durations of nasopharyngeal carriage of penicillinresistant pneumococci. J Clin Microbiol. 2007;45:948–952
31. Sleeman KL, Griffiths D, Shackley F, et al. Capsular serotype-specific attack rates and duration of carriage of Streptococcus pneumoniae in a population of children. J Infect Dis. 2006;194:682–688
32. Fraser C, Hanage WP, Spratt BG. Neutral microepidemic evolution of bacterial pathogens. Proc Natl Acad Sci U S A. 2005;102:1968–1973

Supplemental Digital Content

Copyright © 2013 Wolters Kluwer Health, Inc. All rights reserved.