Impact of 10-Valent Pneumococcal Conjugate Vaccine Introduction on Pneumococcal Carriage and Antibiotic Susceptibility Patterns Among Children Aged <5 Years and Adults With Human Immunodeficiency Virus Infection: Kenya, 2009–2013

Background. Kenya introduced 10-valent pneumococcal conjugate vaccine (PCV10) among children <1 year in 2011 with catch-up vaccination among children 1–4 years in some areas. We assessed changes in pneumococcal carriage and antibiotic susceptibility patterns in children <5 years and adults. Methods. During 2009–2013, we performed annual cross-sectional pneumococcal carriage surveys in 2 sites: Kibera (children <5 years) and Lwak (children <5 years, adults). Only Lwak had catch-up vaccination. Nasopharyngeal and oropharyngeal (adults only) swabs underwent culture for pneumococci; isolates were serotyped. Antibiotic susceptibility testing was performed on isolates from 2009 and 2013; penicillin nonsusceptible pneumococci (PNSP) was defined as penicillin-intermediate or -resistant. Changes in pneumococcal carriage by age (<1 year, 1–4 years, adults), site, and human immunodeficiency virus (HIV) status (adults only) were calculated using modified Poisson regression, with 2009–2010 as baseline. Results. We enrolled 2962 children (2073 in Kibera, 889 in Lwak) and 2590 adults (2028 HIV+, 562 HIV−). In 2013, PCV10-type carriage was 10.3% (Lwak) to 14.6% (Kibera) in children <1 year and 13.8% (Lwak) to 18.7% (Kibera) in children 1–4 years. This represents reductions of 60% and 63% among children <1 year and 52% and 60% among children 1–4 years in Kibera and Lwak, respectively. In adults, PCV10-type carriage decreased from 12.9% to 2.8% (HIV+) and from 11.8% to 0.7% (HIV−). Approximately 80% of isolates were PNSP, both in 2009 and 2013. Conclusions. PCV10-type carriage declined in children <5 years and adults post–PCV10 introduction. However, PCV10-type and PNSP carriage persisted in children regardless of catch-up vaccination.

compare the impact of different PCV10 introduction strategies on carriage, 2) to evaluate differences in pneumococcal carriage prevalence among adults living with and without human immunodeficiency virus (HIV) infection, and 3) to evaluate changes in the proportion of antibiotic-nonsusceptible pneumococci among carried strains.

Study Setting
We utilized 2 ongoing surveillance platforms to conduct annual pneumococcal carriage surveys: the Population-based Infectious Disease Surveillance (PBIDS) and the Western Kenya Health and Demographic Surveillance System (HDSS). PBIDS is conducted in 2 geographically distinct regions, Kibera and Lwak [8,9]. Catch-up vaccination was implemented in Lwak but not in Kibera. HDSS has been in place in western Kenya; and since 2005, residents of 33 of the HDSS villages have also been enrolled in Lwak PBIDS [10] (see Supplementary Methods).

Cross-sectional Survey in Children
Annual cross-sectional pneumococcal carriage surveys were performed in Kibera and Lwak among children aged less than 5 years during 2009-2013, October-December. Children aged less than 5 years were sampled differently in the 2 sites: in Lwak (with catch-up vaccination targeting children aged 1-4 years), children aged less than 5 years were randomly selected; in Kibera (no catch-up vaccination), age-stratified (<1 year and 1-4 years) random sampling was performed (see Supplementary Figure). Survey methods are described in detail elsewhere [11]. Upon enrollment, trained surveyors collected information on household characteristics, respiratory illness, and antibiotic use using standardized forms. PCV10 vaccination history was obtained for children enrolled during 2011-2013. Surveyors were instructed to verify each reported vaccine dose with the child's vaccination card. Vaccination history was validated through PBIDS and HDSS records whenever possible.

Cross-sectional Survey in Adults
In Lwak, pneumococcal carriage surveys among adults living with children aged less than 5 years were conducted in 2009 and annually from 2011 to 2013 during October-December. Details of the survey methods were previously described [10]. Briefly, HDSS records were used to identify compounds where at least 1 HIV-positive parent of a child aged less than 5 years resided. HDSS records are linked to home-based counseling and testing for HIV that occurred widely in the area during 2008-2009 [12]. To maintain confidentiality of HIV testing status, all adults with children aged less than 5 years living in the same compound were invited to participate, regardless of HIV status. For each adult enrolled, trained survey staff used a standardized questionnaire to collect information [10]. participants were serotyped by Quellung reaction. Isolates recovered from participants enrolled during 2010-2013 were serotyped by multiplex polymerase chain reaction-based testing. Antibiotic susceptibility testing was performed on pneumococcal isolates from 2009 and 2013 by broth microdilution (Trek Diagnostics). Antibiotic susceptibility was determined using the 2012 Clinical and Laboratory Standards Institute criteria for minimum inhibitory concentrations [11,14,15]. Intermediate and resistant isolates were considered nonsusceptible to the antibiotic tested (see Supplementary Methods).

Data Management and Analysis
We performed descriptive analyses of participants by site, year, and age group (<1 year, 1-4 years, and adults). We calculated unadjusted prevalence ratios using classic methods for estimation of risk ratios with 2009-2010 (2009 only in adults) as the reference period. Potential confounders associated with pneumococcal carriage identified through previous studies were explored, and all models were adjusted for respiratory illness within the past 30 days, antibiotic use within 7 days, and area used for cooking [10,[16][17][18]. Adjusted prevalence ratios (aPRs) were calculated using Poisson regression with robust error variance [19]. Changes in carriage prevalence by serotype were compared between 2009 and 2013 for children aged less than 5 years and for adults. Changes in the proportion (among pneumococcal isolates) and carriage prevalence (among survey participants) of antibioticnonsusceptible pneumococci between 2009 and 2013 were tested. Chi-square test or Fisher's exact test was used to compare categorical variables and Wilcoxon rank-sum test was used for continuous variables. Analyses were performed using SAS software (version 9.4; SAS Institute) (see Supplementary Methods).

Ethical Considerations
The study was approved by the ethics committees at the Kenya Medical Research Institute and the Centers for Disease Control and Prevention. Written informed consent was obtained from all adult participants and the parent or guardian of all participating children.

Overall Characteristics in Children
During 2009-2013, 2073 children in Kibera and 889 children in Lwak were enrolled (Table  1). Kibera children were more likely to sleep in a crowded room (median number of people sleeping in the same room, 5 vs 3; P < .001) and report recent (≤30 days of the survey) respiratory illness than Lwak children (67.3% vs 56.7%, P < .001). Kibera children were less likely than Lwak children to be exposed to tobacco smoke at home (10.2% vs 15.8%, P < .001) and have recent (≤7 days before the survey) antibiotic use (18.3% vs 22.4%, P = .01). In both sites, about one-third of participants reported taking antibiotics within 30 days of the survey; cotrimoxazole and penicillin/amoxicillin were the most frequently reported antibiotics. Differences in types of fuel used and area used for cooking were also noted between the sites. Characteristics of children by survey year are summarized in Supplementary Tables 1 and 2.

Overall Characteristics in Adults
In 2009 and during 2011-2013, a total of 3547 adults were recruited in Lwak. Of those, 2590 (73.0%) had known HIV status (2028 [78.3%] HIV positive and 562 [21.7%] HIV negative) and were further analyzed (Table 2). Overall, median age was 33 years (interquartile range, 28-38 years) and 67.0% were female. HIV-positive adults were older than HIV-negative adults (median age, 34 vs 30 years; P < .001) and more likely to report antibiotic use within 30 days of the survey, particularly cotrimoxazole (68.7% vs 15.5%, P < .001). Characteristics of HIV-positive adults by survey year are summarized in Supplementary Table 3.

Vaccine Coverage in Children
Vaccination history was available for 96.0% (1734 of 1806) of children enrolled during 2011-2013 (Supplementary Tables 1 and 2), of which 60.8% (1055 of 1734) was validated. Thus, we considered any vaccine dose given to children prior to the survey as valid regardless of when it was administered. Among children aged less than 1 year, coverage for 2 or more PCV10 doses increased from 87% (2011) to 95% (2013) in Kibera and from 77% (2011) to 100% (2013) in Lwak (Table 3). Among children aged 1-4 years, only 10% in Kibera had received 2 or more PCV10 doses in the 2011 survey compared with 59% in Lwak (Table 3). By 2013, the proportion who received 2 or more PCV10 doses increased to 61% and 82% in Kibera and Lwak, respectively.

Changes in PCV13-Unique Type and NVT Pneumococcal Carriage Prevalence in Children and Adults
Compared with baseline, PCV13-unique type (serotypes contained in PCV13 but not in PCV10-3, 6A, 19A) carriage prevalence in 2013 increased in children in Kibera and HIVpositive adults, but the change was significant only in Kibera children aged 1-4 years (aPR, 1.70; 95% CI, 1.21-2.41) ( Table 3). NVT carriage prevalence increased in all groups of children, although was nonsignificant in Lwak children aged less than 1 year. In contrast, NVT carriage prevalence decreased in HIV-positive adults (aPR, 0.76; 95% CI, 0.60-0.96) ( Table 3).
In adults, the predominant serotypes in 2009 included PCV10 serotypes 19F and 23F along with PCV13-unique types 3 and 6A, similar to Lwak children ( Figure 2C, D). In 2013, some of the PCV10 types were not detected in HIV-positive or HIV-negative adults. In HIVpositive adults, serotype 3 (PCV13-unique type) carriage remained stable and was the most commonly identified serotype in both years (4.9% vs 4.5%, P = .76) ( Figure 2C); the increase in serotype 3 was not significant among HIV-negative adults (2.0% vs 3.7%, P = .37).

Changes in the Proportion of Antibiotic Nonsusceptible Pneumococcal Isolates
Antimicrobial susceptibility testing results were available from a total of 1637 out of 1714 (95.5%) pneumococcal isolates collected in 2009 and 2013 (944 from Kibera, 344 from Lwak children, and 349 from HIV-positive adults) ( Table 4). In both sites, age groups, and years, more than 95% of pneumococcal isolates were nonsusceptible to cotrimoxazole, and approximately 80% were penicillin nonsusceptible; the majority of penicillin-nonsusceptible isolates had minimum inhibitory concentrations in the intermediate range.  Table 5). The reduction was offset by a significant increase in PISP among NVT in children but not in HIV-positive adults.

DISCUSSION
In children, we observed a 52% (Kibera, aged 1-4 years) to 60% (Kibera, aged <1 year; Lwak, aged 1-4 years) reduction in PCV10-type pneumococcal carriage prevalence approximately 2 years after PCV10 introduction, similar to the report from Kilifi, Kenya (site with catch-up vaccination targeting children aged 1-4 years) [4]. Reductions were also observed among adults, including those who were HIV positive. Despite reductions, PCV10type carriage remained common in children in 2013 (15-19% in Kibera, 10-14% in Lwak). These figures are consistent with the results from Kilifi, Kenya [4], but higher than from countries that used a PCV10 schedule with 3 primary doses plus a booster (3 + 1) [20][21][22]. No significant reduction in penicillin nonsusceptible pneumococci (PNSP) carriage was observed except in HIV-positive adults. WHO currently recommends both the 3 + 0 and 2 primary doses and a booster (2 + 1) vaccination schedule for PCV administration for infants [23]. Although the 3 + 0 schedule is widely used in sub-Saharan Africa, data from carriage [24] and IPD [25] studies suggest that the booster dose in the 2 + 1 schedule might be required to achieve a sustained reduction in vaccine-type colonization. While this might explain the residual PCV10-type carriage that we observed in children, the long-term significance of this finding is unknown. In Kilifi, PCV10-type carriage in children aged less than 5 years remained at 6% 6 years after PCV10 introduction [26], yet a significant reduction in PCV10-type IPD in all age groups including unvaccinated adults was observed (92% reduction in children aged <5 years, 74% in children aged 5-14 years, and 81% in those aged ≥15 years) [27]. Whether this impact was blunted due to persistent circulation of PCV10-type pneumococci or waning mucosal immunity due to lack of a booster PCV dose is unknown. Of note, our results showed that PCV10-type carriage in 2013 was close to elimination in HIV-negative adults.
Of the PCV13-unique types, we observed a significant increase in serotype 19A carriage prevalence post-PCV10 introduction in Kibera. Although the serotype 19F antigen contained in both the 7-valent PCV (PCV7) and PCV10 was thought to provide some crossprotection against serotype 19A [28], cross-protection has not been proven and increases in serotype 19A IPD incidence post-PCV7 introduction were reported in multiple countries [29][30][31]. In our study, the increase in carriage of 19A was nonsignificant among children in Lwak and was not observed in adults. Carriage studies conducted in Kilifi [4] and Brazil [20] within 3 years of PCV10 introduction (both with catch-up vaccination targeting the studied age groups) also reported nonsignificant increases in serotype 19A carriage post-PCV10 introduction. Although Brazil reported an increase in IPD due to non-PCV10 serotypes (3, 6C, and 19A) 5 years after PCV10 introduction, results from Kilifi showed nonsignificant results [27]; and so far, longer-term impact of PCV10 against serotype 19A has been undetermined [32, 33]. Since our study was not powered to assess changes in individual serotypes, follow-up and correlation with IPD trends are needed.
As previously described [11], a high proportion of pneumococcal carriage isolates were nonsusceptible to penicillin and cotrimoxazole, and the proportion was essentially unchanged after PCV10 introduction. PISP carriage prevalence among PCV10 types declined significantly, but this decline was balanced by a significant increase in PISP NVT carriage in children. Similar findings were reported both in US children [7,34] and adults [35] post-PCV7 introduction. However, isolates with intermediate susceptibility to ceftriaxone were detected for the first time in the post-PCV10 period, most of which were serotype 3. In the United States, IPD caused by NVT antibiotic-nonsusceptible S. pneumoniae remained below the pre-PCV period [36,37]. In Kenya, however, the higher proportion of overall pneumococcal carriage and antibiotic use in the community might result in more sustained transmission of antibiotic-nonsusceptible S. pneumoniae. Although antibiotic susceptibility results were not available, data from Kilifi showed a nonsignificant increase in NVT-IPD in all age groups aged 2 or more months, despite significant reductions in overall and PCV10-type IPD incidence [27].
We observed a reduction in overall pneumococcal carriage prevalence and PNSP carriage prevalence among HIV-positive adults post-PCV10 introduction. This is likely due to the significant reduction in NVT carriage prevalence observed in 2013, but not in 2011 or 2012, compared with baseline. Similar findings have been reported [38] and attributed to different dynamics in pneumococcal carriage (HIV-positive mothers are more likely to carry vaccinetype pneumococci) [39], changes in HIV management over time, and widespread use of antibiotic prophylaxis [40,41]. Despite the observed reductions, the high proportion of serotype 3 carriage, a serotype known to be associated with severe outcomes [42], and the higher carriage prevalence compared with HIV-negative adults indicate that HIV-positive adults will continue to be at a higher risk of IPD than HIV-negative adults, as seen in other countries [40,43].
Our study is subject to several limitations. First, we were not able to assess differences in indirect effects between Lwak and Kibera due to the small number of unvaccinated Lwak children aged 1-4 years. We used children who received 1 or less PCV10 dose as a proxy for unvaccinated children, but the number of children in this group was also small. Since adults were only recruited in Lwak, we could not compare indirect effects among adults between the 2 sites. Second, although we were not able to confirm PCV10 vaccination dates for all children, we considered any vaccine dose reported prior to the survey regardless of the timing; therefore, some reported doses might not have been valid. Third, misclassification of HIV-negative adults might have occurred since some adults had a test date more than 1 year before the survey. Last, we did not have data on CD4 counts or HIV treatment history when assessing the impact of PCV10 among HIV-positive adults.
Despite these limitations, our study addresses key questions related to PCV10 introduction in sub-Saharan Africa. We observed significant declines in PCV10-type carriage in children and adults approximately 2 years after PCV10 introduction in Kenya. Although not statistically significant, there was a greater reduction in PCV10-type carriage in the site with catch-up vaccination targeting older children compared with the site without catch-up vaccination. However, as additional birth cohorts receive vaccination and as vaccinated children age, the differences between the 2 sites might be less noticeable. We did not observe changes in the proportion of isolates that were PNSP in children post-PCV10; this is likely due to rapid replacement of PCV10-type isolates by NVT PNSP. Very few studies have comprehensively addressed the impact of PCV10 introduction in sub-Saharan Africa, where 3 + 0 is widely used. The most recent WHO position statement on PCV states the potential benefits of the 2 + 1 schedule over the 3 + 0 schedule in providing longer protection, although supporting data are currently limited [23]. Continued monitoring of PCV10-type pneumococcal carriage in children and antibiotic-nonsusceptible pneumococci and their association with disease rates could help inform pneumococcal vaccination policies.

Supplementary Material
Refer to Web version on PubMed Central for supplementary material.