Population effect of 10-valent pneumococcal conjugate vaccine on nasopharyngeal carriage of Streptococcus pneumoniae and non-typeable Haemophilus influenzae in Kilifi, Kenya: findings from cross-sectional carriage studies

Summary Background The effect of 7-valent pneumococcal conjugate vaccine (PCV) in developed countries was enhanced by indirect protection of unvaccinated individuals, mediated by reduced nasopharyngeal carriage of vaccine-serotype pneumococci. The potential indirect protection of 10-valent PCV (PCV10) in a developing country setting is unknown. We sought to estimate the effectiveness of introduction of PCV10 in Kenya against carriage of vaccine serotypes and its effect on other bacteria. Methods PCV10 was introduced into the infant vaccination programme in Kenya in January, 2011, accompanied by a catch-up campaign in Kilifi County for children aged younger than 5 years. We did annual cross-sectional carriage studies among an age-stratified, random population sample in the 2 years before and 2 years after PCV10 introduction. A nasopharyngeal rayon swab specimen was collected from each participant and was processed in accordance with WHO recommendations. Prevalence ratios of carriage before and after introduction of PCV10 were calculated by log-binomial regression. Findings About 500 individuals were enrolled each year (total n=2031). Among children younger than 5 years, the baseline (2009–10) carriage prevalence was 34% for vaccine-serotype Streptococcus pneumoniae, 41% for non-vaccine-serotype Streptococcus pneumoniae, and 54% for non-typeable Haemophilus influenzae. After PCV10 introduction (2011–12), these percentages were 13%, 57%, and 40%, respectively. Adjusted prevalence ratios were 0·36 (95% CI 0·26–0·51), 1·37 (1·13–1·65), and 0·62 (0·52–0·75), respectively. Among individuals aged 5 years or older, the adjusted prevalence ratios for vaccine-serotype and non-vaccine-serotype S pneumoniae carriage were 0·34 (95% CI 0·18–0·62) and 1·13 (0·92–1·38), respectively. There was no change in prevalence ratio for Staphylococcus aureus (adjusted prevalence ratio for those <5 years old 1·02, 95% CI 0·52–1·99, and for those ≥5 years old 0·90, 0·60–1·35). Interpretation After programmatic use of PCV10 in Kilifi, carriage of vaccine serotypes was reduced by two-thirds both in children younger than 5 years and in older individuals. These findings suggest that PCV10 introduction in Africa will have substantial indirect effects on invasive pneumococcal disease. Funding GAVI Alliance and Wellcome Trust.


Introduction
Introduction of pneumococcal conjugate vaccine (PCV) into the routine immunisation schedule of developed countries over the past 13 years has resulted in a dramatic reduction in the incidence of invasive pneumococcal disease caused by vaccine serotypes. [1][2][3][4][5] Additionally, vaccinated individuals are less likely to be carriers of vaccine-serotype pneumococci, and are therefore less likely to transmit the infection, than non-vaccinated individuals. At a population level, vaccination leads to a reduction in the carriage prevalence of vaccine-serotype pneumococci in both vaccinated and unvaccinated individuals and a reduction in the incidence of invasive pneumococcal disease caused by vaccine-serotype pneumococci in the whole population (ie, herd protection). Within 4 years after the introduction of PCV into the childhood immunisation schedule in the USA, the incidence of vaccine-serotype invasive pneumococcal disease in people aged 5 years or older had fallen by 62%. 6 The indirect protection provided by PCV was greater than its direct protection, and this factor had a profound eff ect on estimates of the cost-eff ectiveness of the vaccine. 7 The nasopharyngeal niche vacated by vaccine-serotype pneumococci is rapidly occupied by non-vaccine-type pneumococci, which has led to serotype replacement disease of varying magni tude in diff erent populations. 5, [8][9][10] Many low-income countries will introduce PCV in the coming decade. 11 The paucity of robust longitudinal surveillance systems for invasive pneumococcal disease in developing countries poses a challenge in identifying the programmatic eff ectiveness of PCV in these settings. However, studies of nasopharyngeal carriage, in which the anatomical locus of indirect vaccine eff ects are investigated, are logistically feasible in developing countries and can help monitor the population eff ect of PCV. [12][13][14] In 2011, Kenya became one of the fi rst countries in Africa to introduce PCV and the fi rst country to use the 10-valent PCV conjugated to non-typeable Haemophilus infl uenzae protein-D (PCV10). 15 Because PCV10 uses protein-D from non-typeable H infl uenzae as its carrier protein, it might induce protection against infections caused by non-typeable H infl uenzae, an important cause of otitis media and respiratory tract infection. 16 The eff ect of vaccination on pneumococcal carriage prevalence, and possibly on nontypeable H infl uenzae carriage prevalence, might aff ect other bacteria in the nasopharynx. An inverse relation between carriage of pneumococcus and Staphylococcus aureus has been described, leading to speculation that PCV use might result in an increase in S aureus carriage in children, possibly leading to staphylococcal disease. [17][18][19][20] With support from the GAVI Alliance, PCV10 was introduced into the routine infant vaccination programme in Kenya in January, 2011, with a catch-up campaign for infants. In Kilifi County, a catch-up campaign was also done for all children aged younger than 5 years, which accelerated the population eff ect of cohort introduction by several years. We aimed to assess the programmatic eff ects of PCV10 introduction on nasopharyngeal carriage of Streptococcus pneumoniae, non-typeable H infl uenzae, and S aureus at an early stage.

Study design and participants
The study took place in the Kilifi Health and Demographic Surveillance System (KHDSS), a 891-km² area within Kilifi County, a poor rural district on the Indian Ocean coast of Kenya. The KHDSS has a population of about 260 000 people who have been under surveillance for vital events and migration through 4-monthly household visits since 2000. 21 H infl uenzae type b conjugate vaccine was introduced into this area in 2001; coverage for three doses of H infl uenzae type b vaccine was 95% at 12 months of age among residents of the KHDSS in 2007. 22 In January, 2011, the government of Kenya introduced PCV10 into the national immunisation schedule, administered simultaneously with pentavalent vaccine (diphtheria, whole cell pertussis, tetanus, hepatitis B, and H infl uenzae type b combined vaccine) at age 6, 10, and 14 weeks. In 2011, all infants were encouraged to present for a three-dose catch-up schedule 4 weeks apart. In Kilifi County, an additional catch-up campaign was undertaken to provide up to two doses of PCV10 to children aged 12-59 months in two outreach campaigns, beginning on Jan 31, 2011, and March 21, 2011, and lasting 1-2 weeks. These campaigns were managed by the Ministry of Public Health and Sanitation at 45 community health facilities. We did annual cross-sectional studies of nasopharyngeal carriage in the KHDSS in the 2 years before and 2 years after introduction of PCV10. For each year of the study, we used a Stata program to randomly select 50 residents in each of ten age strata (0, 1-2, 3-4, 5-9, 10-14, 15-19, 20-39, 40-59, 50-59, and ≥60 years) from the KHDSS population register. Using the same method, we randomly selected 30 additional residents in each age strata to serve as a back-up list to cater for people who were lost to follow-up or declined to participate. Participants included in the fi rst year were not excluded from future selection. During June-October, fi eldworkers visited the home of each potential participant, explained the study, and obtained written informed consent from each adult participant or from the parent or guardian of each participant aged younger than 18 years. The protocol was approved by the Oxford Tropical Ethical Review Committee (number 30-10) and the Kenya National Ethical Review Committee (SSC1433).

Procedures
Fieldworkers administered a short questionnaire eliciting risk factors for carriage, documented vaccination history from the immunisation cards of children, and then collected a nasopharyngeal swab specimen. Residents who had moved out of the KHDSS, could not be located, or declined to participate were replaced by choosing the fi rst remaining name from a back-up random selection of residents in each age stratum. A nasopharyngeal rayon swab (Medical Wire, Corsham, UK) specimen was collected from each participant. Specimens were collected by passing the swab through the nostril, along the fl oor of the nasal cavity until it touched the posterior nasopharyngeal wall, where it was left for 2-3 s, rotated, and removed. Swabs were placed in skim-milk tryptone glucose glycerol media and processed at the Kenya Medical Research Institute-Wellcome Trust Research Programme Laboratory (Kilifi , Kenya), in accordance with WHO recommendations. 23 Isolates of S pneumoniae were identifi ed from gentamicin-blood agar by optochin susceptibility testing; serotyping was done by latex agglutination and the Quellung reaction (including separate antisera for serotypes 6A and 6C). If pneumococcal colonies of varying appearance occurred, only those of the dominant colony morphology were serotyped. All serogroup 6 isolates were retested by PCR for confi rmation of serotype. Isolates of H infl uenzae were identifi ed from bacitracin-chocolate agar by gram stain and X and V factor dependence. Typing of H infl uenzae isolates was done by multiplex PCR using an IgA target that discriminates between H infl uenzae and Haemophilus haemolyticus, a bexA target, to identify encapsulation and specifi c targets for each capsular type. 24,25 Isolates of suspected S aureus identifi ed from mannitol salt agar were subcultured and identifi ed by gram stain and biochemical testing.

Statistical analysis
Vaccine serotypes were defi ned as those contained in PCV10 (1, 4, 5, 6B, 7F, 9V, 14, 18C, 19F, and 23F). Nasopharyngeal carriage prevalence was estimated in four broad age strata for each bacterial group (vaccine-serotype and non-vaccine-serotype pneumococci, non-typeable H infl uenzae, and S aureus). Unadjusted prevalence ratios were calculated for the vaccine period (2011-12) compared with the baseline period (2009-10) using classic methods for estimation of risk ratios. To identify potential confounders, we tested the association between all questionnaire variables and the vaccine period. For consistency across models of diff erent age groups and bacterial groups, we adjusted all models for each of the confounding variables that was signifi cant at a p value of less than 0·1. To obtain unconfounded estimates of prevalence ratios in the vaccine period compared with the baseline period, we explored both secular changes in carriage prevalence and secular changes in potential confounders. Prevalence ratios were modelled using logbinomial regression; if the models failed to converge, we used Poisson regression with robust 95% CIs. 26 Variation in the eff ect of PCV10 on carriage prevalence with age was tested as an interaction term. Changes in prevalence over time, adjusted for vaccine period, were tested per year of study. Adjusted prevalence ratios were age standardised in ten strata, to represent the stratifi ed sampling scheme, by the inverse of the sampling ratio as population weights; the reference was the KHDSS population register at the midpoint of the study (Jan 1, 2011).
The signifi cance of vaccine eff ect on carriage of 25 individual serotypes was tested using a Bonferroni correction (ie, 0·05/25). Vaccine eff ectiveness against carriage (VE carr ) was calculated as 1 minus the agestandardised, adjusted prevalence ratio. Estimates of vaccine eff ectiveness against acquisition were calculated as 1 minus the age-standardised, adjusted odds ratio. 27 All statistical analyses were done using Stata version 11.2.

Role of the funding source
This work was done under a collaborative arrangement with the PenumoADIP at Johns Hopkins Bloomberg School of Public Health and funded by the GAVI Alliance. This study was done at a research unit funded by the Wellcome Trust of Great Britain. The funders of the study had no role in study design, data collection, data analysis, writing of the report, or the decision to submit manuscript for publication. LLH had full access to all the data in the study, takes responsibility for the integrity of the data and the accuracy of the data analysis, and had fi nal responsibility for the decision to submit for publication.

Results
Overall   isolates per year were identifi ed as possible pneumococci on the basis of optochin testing but were non-typeable and were excluded from this analysis. Two serotype 6C isolates were detected: one in 2009 that was detected by PCR alone and one in 2012 that was detected by both the Quellung reaction and PCR. Among isolates of H infl uenzae, 588 (94%) were non-typeable H infl uenzae by PCR. Figure 1 shows the pneumococcal carriage prevalence for each of the 4 years of the study among participants in the four age groups. Figure 2 shows results for nontypeable H infl uenzae and S aureus. After adjusting for vaccine period and month of sampling within a given year, there were no signifi cant secular changes in carriage prevalence for any of the fi ve bacterial groups tested either in participants younger than 5 years or aged 5 years or older. Among ten epidemiological variables tested for association with vaccine period (table 1), three were signifi cant (p<0·1): month of sampling within a given year (p<0·0001), number of people with whom the participant shared a bed (p=0·0004), and self-reported use of antibiotics within the 14 days before swab collection (p=0·08). All subsequent prevalence ratios were adjusted for these three factors. Table 2 details the carriage prevalence in the baseline and vaccine periods, crude prevalence ratios, and agestandardised, adjusted prevalence ratios for each of the fi ve bacterial classifi cations. The adjusted prevalence ratios did not vary signifi cantly by age specifi ed in four strata (<5, 5-17, 18-49, ≥50 years). Consequently, the data are presented for simplicity in two age strata: those targeted for vaccination (age <5 years) and those who were not targeted for vaccination (age ≥5 years). Even in these two strata, the adjusted prevalence ratios did not diff er signifi cantly with age for any of the bacterial groups examined.
We also examined the eff ect of vaccination at an individual level in an exploratory post-hoc anaylsis. During the vaccine period, among children 1-4 years old, the adjusted prevalence ratio for those who received at least  two doses of PCV10 compared with those who received zero or one doses was 0·47 (95% CI 0·21-1·03) for vaccine-serotype pneumococci and 1·22 (0·87-1·70) for non-typeable H infl uenzae. Because of high vaccine uptake, we were not able to do this analysis among infants, or to compare children who received no doses with those who received at least one dose. Figure 3 shows the serotype-specifi c carriage prevalence among children younger than 5 years for the baseline and vaccine periods. The diff erences in carriage prevalence were signifi cant only for serotypes 6B (10% vs 3%; p=0·0003) and 19F (12% vs 5%; p=0·002). No eff ects on carriage of the vaccine-related serotypes 6A or 19A were noted.

Discussion
We report rapid, signifi cant reductions in vaccine serotype nasopharyngeal carriage at a population level in Kilifi , Kenya, after introduction of PCV10 into the routine infant vaccination schedule accompanied by a catch-up campaign for children younger than 5 years. To our knowledge, this is the fi rst study to report the eff ects on carriage of a national PCV vaccination programme in a GAVI Allianceeligible developing country (panel). Although pneumococcal carriage is often asymptomatic and benign, it is a necessary precursor in the development of invasive disease. Because of this causal link, vaccine eff ect on carriage is an important marker of vaccine-induced protection against disease in children and adults. 13 About 18 months after the introduction of PCV10, we noted a 64% reduction in vaccine-serotype S pneumoniae carriage among children younger than 5 years, 79% of whom had received at least one dose of PCV10. In comparison, in Alaska, USA, 3 years after introduction of 7-valent PCV (PCV7), a 91% reduction in vaccine serotype carriage was noted among Alaska Native children aged 5 years or younger, more than 99% of whom had received at least one dose of PCV7. 33 In both settings, PCV was introduced with a catch-up campaign. In a large clusterrandomised study in The Gambia, 39 in which widespread PCV7 vaccination was undertaken, there was a 56% reduction in vaccine serotype carriage in children aged 2-5 years living in villages where children younger than 30 months were vaccinated and a 74% reduction in villages where all residents receive at least one dose of PCV7. Within 2 years after PCV7 was introduced into the public immunisation programme for infants without a catch-up campaign in a South African community with high HIV prevalence, vaccine serotype carriage was reduced by 50% among children younger than 2 years, 51% of whom had received three doses of PCV7. 40 In the Netherlands, 3 years after introduction of PCV7, without a catch-up campaign, vaccine serotype carriage was reduced by 80-90% among vaccinated children aged between 11 months and 24 months. 34 In Portugal, 4 years after PCV7 became available, in children aged 4 months to 6 years, 57% of whom had received PCV7, vaccine    serotype carriage was reduced by 78%. 35 The abovementioned studies show a substantial eff ect on vaccine serotype carriage 1·5-4 years after introduction of PCV with and without a catch-up campaign, with vaccine coverage in the target age group ranging between 50% and 100%. Reductions in carriage have been matched by reductions in invasive pneumococcal disease in settings where such data are available, including the KHDSS, where 1 year after PCV10 introduction, the vaccine eff ectiveness was estimated to be 72% (95% CI 34-88) against vaccine-serotype invasive pneumococcal disease in children aged younger than 5 years. [36][37][38]41 In addition to the eff ect in the vaccine target age group, we noted a 66% reduction in vaccine serotype carriage adjusted prevalence in individuals aged 5 years or older. This fi nding is consistent with fi ndings among Alaska Native people, in whom a 68% reduction in vaccine serotype carriage was reported among people aged at least 18 years about 3 years after introduction of PCV7. 33 The reductions in vaccine serotype carriage in the nontarget age group in Kilifi were apparent in the fi rst post-PCV10 survey (2011), when coverage with at least one dose of PCV10 among children younger than 5 years was 63% in the KHDSS and 69% among study participants. Reasons for the diff erence in the vaccine coverage estimates for study participants compared with all KHDSS children are as follows: (1) the time lag between the random selection of potential participants and their enrolment in the study caused our sample of the very youngest age group (0-12 months) to be skewed towards the upper end of this bracket when children were more likely to have been vaccinated; (2) our study captured vaccinations given outside the area covered by the vaccine registry; and (3) some of the migrant population-who generally have lower levels of vaccine coverage-would have been lost after random selection in our study. Nonetheless, our fi ndings suggest that substantial indirect eff ects occur when two-thirds of children younger than 5 years are vaccinated and imply that the indirect protection against invasive pneumococcal disease noted in the USA and UK can probably be replicated in developing countries.
In other settings, the reduction in vaccine serotype carriage prevalence after programmatic introduction of PCVs has been matched by a reciprocal increase in carriage prevalence of non-vaccine serotypes (ie, serotype replacement carriage) such that the overall pneumococcal carriage prevalence, typically, is unchanged from baseline. However, serotype replacement in the nasopharynx has had a variable eff ect on invasive pneumococcal disease. 5,10 In most settings, serotype replacement invasive pneumococcal disease has been minimal, whereas in some settings it has almost negated the benefi cial eff ect of PCVs in some subgroups of the population. We noted a signifi cant increase in carriage of non-vaccine-serotype pneumococci among children younger than 5 years; however, because the magnitude of the decline in vaccine serotype carriage was greater, there was a slight decline in overall S pneumoniae carriage prevalence in the PCV10 period. The reduction in overall pneumococcal carriage in children is likely to be attributable to the vaccine itself, rather than to underlying variations in carriage, because analyses of the change in carriage prevalence for all pneumococci over the 4 years did not identify a signifi cant decline in prevalence after adjusting for vaccine eff ect. Although non-vaccine serotype carriage increased signifi cantly in children younger than 5 years, the increase was not statistically signifi cant in people aged 5 years or older. Children are likely to experience more rapid, direct clearance of vaccine serotype carriage and subsequent replacement carriage, whereas adults experience delayed, indirect clearance. Thus, replacement carriage in adults is probably delayed. Two other studies in Africa-a cluster-randomised trial of PCV in The Gambia 39 and an observational ecological study after  programmatic introduction of PCV7 in South Africa 40 -found that non-vaccine serotype carriage declined in adolescents and adults after PCV use in children. However, these fi ndings are subject to several limitations including the short period of follow-up and changes in HIV treatment regimens in South Africa, and an intercurrent community-wide azithromycin campaign in The Gambia. Long-term surveillance is essential to understand PCV-induced changes in nonvaccine serotype carriage and disease.
In serotype-specifi c analyses, we noted no eff ect on carriage of the serotypes 6A or 19A in the target age group. This fi nding is consistent with data from clinical trials that show that PCV10 does not induce a robust antibody response against these strains. 42 Only the predominant colony appearance of pneumococcus was serotyped from each nasopharyngeal swab, so serotypespecifi c variations do not account for changes that might have occurred among the non-dominant strains carried by an individual. However, assuming that the probability of sampling a strain is proportional to the frequency of that serotype in the nasopharynx then the present study is of a random sample of strains in a random sample of individuals and this limitation should not aff ect our conclusions about vaccine eff ectiveness. Carriage is expected to be in fl ux in the fi rst few years of vaccine use and carriage prevalence will probably continue to change before reaching equilibrium. 43,44 We noted a signifi cant reduction in nasopharyngeal carriage of non-typeable H infl uenzae among participants younger 5 years and at least 5 years old in the vaccine period compared with baseline. However, the role of PCV10 as the causative agent of this change is questionable since non-typeable H infl uenzae carriage prevalence seemed to rebound in year 2 of the vaccine period and we did not fi nd an association between an individual's vaccination status and carriage of non-typeable H infl uenzae (by comparing individuals with at least two doses to those with zero or one dose). Findings from early clinical trials suggested that use of an 11-valent protein-D conjugate vaccine reduced the carriage prevalence of vaccine-serotype and non-typeable H infl uenzae, although the decline in non-typeable H infl uenzae carriage was not signifi cant when molecular methods were used to diff erentiate non-typeable H infl uenzae from the closelyrelated H haemolyticus. 16,45 In long-term follow-up, lower non-typeable H infl uenzae carriage prevalence in vaccine recipients compared with controls was documented at about 2 years of age but at no other timepoint. 46 Other clinical trials of PCV10 have not documented a signifi cant, consistent eff ect of vaccination on carriage of non-typeable H infl uenzae. [47][48][49] Although the prevalence of non-typeable H infl uenzae might have been higher had we collected an oropharyngeal swab in addition to a nasopharyngeal swab, our methods were similar across years of the study, thus allowing comparison between periods before and after vaccination. 50 We reported no change in the carriage prevalence of S aureus after introduction of PCV10. By contrast, fi ndings from several studies have suggested an inverse relation between carriage of S pneumoniae and S aureus, [17][18][19][20] and one population-level assessment in the Netherlands reported an increase in S aureus carriage after introduction of PCV7. 51 Potential explanations for this diff erence include variations in the nasopharyngeal microbiome across populations and the competition dynamics that ensue after reductions in pneumococcal carriage in a rural developing country setting. S aureus was cultured in our study, and in most of the comparator studies cited earlier, from the posterior nasopharynx. Cultures of the anterior nares might be more appropriate to fully characterise the eff ect of the vaccine on S aureus. The period after vaccine surveillance in the present study is brief and the sustainability of eff ects (or absence of eff ects) on carriage of various diff erent bacteria can only be identifi ed after a longer period of surveillance. We intend to extend surveillance for at least 3 more years, but have reported early results because the catch-up campaign provided additional maturity to the programme and the vaccine eff ects are large.
In presenting measures of VE carr , we used the term eff ectiveness to describe the magnitude of the eff ect of the vaccine in the total population (that we sampled randomly) under the short-term conditions of rapid introduction with high coverage. The eff ect of the programme will evolve over time and will be determined

Systematic review
After trials of pneumococcal conjugate vaccine (PCV) in the USA, The Gambia, and South Africa showed an excellent vaccine effi cacy against vaccine-serotype invasive pneumococcal disease, [29][30][31][32] WHO recommended that PCV should be included in the routine immunisation schedules of developing countries and several funding agencies pledged support for this introduction. Kenya was chosen as one of the fi rst countries in Africa to receive support for vaccine introduction from the GAVI Alliance. The Kenya PCV Impact Study 28 was designed to assess the eff ectiveness and cost-eff ectiveness of PCV in a setting where it was possible to investigate the eff ect of PCV against a background of longitudinal surveillance. Several studies in developed countries have established a strong association between vaccine eff ect on carriage and vaccine eff ect on invasive pneumococcal disease. [33][34][35][36][37][38] The nasopharyngeal carriage study of the Kenya PCV Impact Study was designed to provide an early assessment of vaccine eff ect in a developing country setting. A formal systematic review was not done as part of the study.

Interpretation
In this study, introduction of PCV10 in a developing country setting, with a catch-up campaign, led to a two-thirds reduction in carriage prevalence of vaccine-serotype pneumococci both in children targeted for vaccination and in older people who were not vaccinated. The eff ect reported in children provides convincing functional evidence that the vaccine is inducing immunological protection at a level suffi cient to prevent invasive disease. The eff ect in older children and adults suggest that the childhood PCV10 programme is reducing transmission of vaccine-serotype pneumococci within the population and this is likely to lead to a reduction in vaccine-serotype invasive pneumococcal disease across all age groups (ie, herd protection).
by both the coverage in infants and the age structure of coverage among the total carrier population. In presenting the prevalence ratio, we have assumed that the key risk for disease is total carriage prevalence. If the key risk is acquisition of carriage, then the odds ratio might provide a more accurate estimate. However, the two methods (prevalence ratio and odds ratio) yielded similar results in this analysis.
PCVs are being introduced rapidly across developing countries, although there is, as yet, limited evidence of their operational eff ect. PCV10, in particular, has not been studied in any national vaccination programme. This study has shown that introduction of PCV10 in a developing country setting, with a catch-up campaign, has led to a two-thirds reduction in carriage prevalence of vaccine-serotype pneumococci both in children targeted for vaccination and in older people who were not vaccinated. The eff ect reported in children provides convincing functional evidence that the vaccine is inducing immunological protection at a level suffi cient to prevent invasive disease. The eff ect in older children and adults suggests that the childhood PCV10 programme is reducing transmission of vaccine-serotype pneumococci within the population and this is likely to lead to a reduction in vaccine-serotype invasive pneumococcal disease across all age groups (ie, herd protection).

Contributors
LLH and JAGS were involved in the design and conduct of the study, data analysis, data interpretation, and writing of the manuscript. DOA was involved in data collection, data analysis, and laboratory analysis. SCM was involved in laboratory analysis, data interpretation, and writing of the manuscript. AK and SN were involved in laboratory analysis and data analysis. NK was involved in data collection and data analysis. TB, EM, TK, and SKS were involved in design and conduct of the study.

Declaration of interests
LLH has received research funding from GlaxoSmithKline Biologicals and Pfi zer. JAGS has received research funding form GlaxoSmithKline Biologicals and support for travel or accommodation at a scientifi c meeting sponsored by Merck. All other authors declare no competing interests.