How nasopharyngeal pneumococcal carriage evolved during and after a PCV13-to-PCV10 vaccination programme switch in Belgium, 2016 to 2018

Background The current carriage study was set up to reinforce surveillance during/after the PCV13-to-PCVC10 switch in Belgium. Aim This observational study monitored carriage of Streptococcus pneumoniae (Sp) serotypes, particularly those no longer covered (3, 6A, 19A), as well as Haemophilus influenzae (Hi), because PCV10 contains the non-typeable Hi protein D. Methods A total of 2,615 nasopharyngeal swabs from children (6–30 months old) attending day care were collected in three periods over 2016–2018. Children’s demographic and clinical characteristics and vaccination status were obtained through a questionnaire. Sp and Hi were identified by culture and PCR. Pneumococcal strains were tested for antimicrobial (non-)susceptibility by disc diffusion and serotyped by Quellung-reaction (Quellung-reaction and PCR for serotypes 3, 6A, 19A). Results The carriage prevalence of Sp (> 75%) remained stable over the successive periods but that of Hi increased (87.4%, 664 Hi-carriers/760 in 2016 vs 93.9%, 895/953 in 2017–2018). The proportion of non-PCV13 vaccine serotypes decreased (94.6%, 438 isolates/463 in 2016 vs 89.7%, 599/668 in 2017–2018) while that of PCV13-non-PCV10 vaccine serotypes (3 + 6A + 19A) increased (0.9%, 4 isolates/463 in 2016 vs 7.8%, 52/668 in 2017–2018), with serotype 19A most frequently identified (87.9%, 58/66 isolates). Non-susceptibility of pneumococci against any of the tested antibiotics was stable over the study period (> 44%). Conclusions During and after the PCV13-to-PCV10 vaccine switch, the proportion of non-PCV13 serotypes decreased, mainly due to a serotype 19A carriage prevalence increase. These results complement invasive pneumococcal disease surveillance data, providing further basis for pneumococcal vaccination programme policy making.

The primary virulence factor of Sp is its polysaccharide capsule, which also determines the serotype. More than 95 serotypes exist and they vary in their capacity to activate the host immune system and to invade [12][13][14][15]. PCVs provide direct protection to the vaccinated individuals against a number of clinically relevant serotypes [12]. In addition, the wider population experiences indirect protection against pneumococcal disease through reduced nasopharyngeal carriage of pneumococcal vaccine serotypes (VTs). However, the observed magnitude of this indirect effect varies in different contexts, and it is eroded by the rising incidence of non-VT-(NVT-)related diseases [16]. Several studies on carriage or IPD in the pre-and post-PCV era reported on serotype replacement, i.e. VTs being largely replaced by NVTs [17,18]. Furthermore, co-colonisation with other pathogens such as Haemophilus influenzae (Hi), Moraxella catarrhalis (Mc), Staphylococcus aureus (Sa), and Streptococcus pyogenes (GAS) may be changed after PCV-introduction because of mutual interactions [19][20][21].
The current carriage study was set up to reinforce surveillance after the PCV13-to-PCV10 vaccination programme switch, in order to monitor the three pneumococcal serotypes that were no longer covered (3, 6A, 19A), as well as Hi, because PCV10 contains the non-typeable Hi (NTHi) protein D. To this end, we studied nasopharyngeal carriage of Sp and Hi in children between 6 and 30 months of age attending day care centres (DCCs) during three consecutive periods between 2016 and 2018. High pneumococcal carriage rates (range: 21-89%) have been reported in young children attending day care [17,[27][28][29]. As such, the   impact of the PCV-programme change was monitored in a random sample of this target population, to complement sentinel laboratory-reported IPD-surveillance.
In this paper, we focus on pneumococcal serotype distribution and antimicrobial (non-)susceptibility during and after the PCV13-to-PCV10 vaccination programme switch.

Ethical statement
The

Study design
The design of this observational study was previously described in detail and is summarised here for the complete study period (from Period 1 in 2016 up to Period 3 in 2017-2018) [30,31].
Nasopharyngeal sampling was performed between March and July in Period 1 (2016) and between November and March in the consecutive periods (Period 2: 2016-2017 and Period 3: 2017-2018). Healthy children were recruited in DCCs randomly selected over the three Belgian regions (Wallonia, Flanders, Brussels), according to a population-proportionate distribution at regional level based on Belgian Federal Government Statistics for the 0-4 year population. In the consecutive periods, 85, 112, and 102 DCCs participated in the study, of which 66 DCCs participated in all three periods, 44 in two periods, and 24 in one period. A population-proportionate sample at the regional level was achieved from 2016 to 2017 onwards, after deliberate over-recruitment in Wallonia in 2016, in order to include a maximum of children who received PCV13 for both primary vaccine doses and their booster, since at that time, PCV10 was not yet introduced in Wallonia. The inclusion criteria were: no treatment with oral antibiotics (ABs) in the 7 days before sampling and age between 6 and 30 months included. An additional age criterion (date of birth before 1 January 2015) was applied in Flanders and Brussels for 2016, in order to exclusively include children who received PCV13 for their primary vaccine doses [30]. In this way, children recruited in the first period were on average older compared with children recruited in the subsequent periods.
Trained nurses and physicians conducted a questionnaire collecting demographic and clinical characteristics of the study participants. The vaccination status of the participating child was based on vaccination documentation or parental reporting. A single nasopharyngeal swab was taken with a flocked nylon fibre swab, transported in 1 mL skim milk-tryptone-glucoseglycerol (STGG) and cultured or stored at − 80 °C within 24 hours.

Culture analyses
At the National Reference Centre (NRC) for pneumococci, nasopharyngeal samples were plated on blood agar plates for identification of Sp, Mc, Sa, GAS and a selective plate for identification of Hi, following overnight enrichment in brain-heart infusion (BHI) broth (entire study period for blood agar plates, 2016 only for selective plate) and directly (from 2016-2017 onwards for blood agar plates and selective plate). Sp-strains were serotyped via Quellung-reaction. Antimicrobial (non-)susceptibility of the Sp-strains for erythromycin, levofloxacin, penicillin, tetracycline and trimethoprim/ sulfamethoxazole was determined by disc diffusion according to the guidelines of Clinical and Laboratory Standards Institute (CLSI; 2016 and 2016-2017) [32] and European Committee on Antimicrobial Susceptibility Testing (EUCAST; 2017-2018) [33]. If non-susceptibility for penicillin or levofloxacin was identified by disc diffusion, the minimum inhibitory concentration (MIC) was determined by Etest (Biomérieux, Craponne, France). A MIC of > 0.06 mg/L for penicillin or > 2 mg/L for levofloxacin was interpreted as non-susceptible. Samples were considered positive for any of the respective pathogens if either culture or PCR was positive.

Molecular analyses
The presented carriage prevalences of Sp and Hi were based on culture and PCR-results, whereas overall serotype distribution and antimicrobial (non-)susceptibility were based on Quellung-reaction and culture results respectively. The reported carriage prevalence of serotypes 3, 6A, and 19A was based on Quellungreaction and PCR-results.

Statistics
Sample size and power were calculated using the R-package 'power' [30]. A sample size of 700 children in 2016 and 900 children from the second period onwards allows the detection of 4% changes in carriage prevalence of Sp-serotypes 19A or 6A over the observation period of the study period with 80% power and assuming a starting carriage prevalence below 2%.
In IBM SPSS Statistics 25, the chi-squared (Chi 2 ) or Fisher's Exact Test (FET) and the Mann-Whitney U Test (MWU) were used to test for significance at a level of 5%. To identify predictors of carrying Sp, PCV13-non-PCV10-VTs, and Hi in DCC-children (three periods pooled), univariate and multiple binary logistic regressions were performed and adjusted using generalised estimating equations (GEEs) with an exchangeable correlation structure since 148 children (303 isolates) contributed more than one sample over the 3-year study period. The GEE model analyses were performed using the statistical software R (version 3.6.1) with the geepack package (version 1.2-1). Variables with a p value < 0.1 in the univariate analysis were included in the multiple regression analysis. Since no children were sampled in the youngest age category in Flanders and in Brussels in 2016, no adjustments can be made for the different sampling probabilities in the different study years. A continuity correction was applied for 95% confidence intervals (95% CI) on proportions.
Missing values were not replaced.

Study population
Over the three successive periods, nasopharyngeal samples from 2,883 children attending DCCs were collected. In total, 2,615 samples (760 in  ) originated from children who contributed more than one sample, but never in the same period. The univariate and multiple binary logistic regression models were adjusted for this through GEEs with an exchangeable correlation structure.
The main demographic and clinical characteristics of the child population over the study period are shown in Table 1, with the majority of these characteristics remaining similar over the study period. Nevertheless, a decreasing trend (p < 0.001) was observed for history of acute otitis media (AOM-history) and AB-use in the 3 months before sampling. The proportions of children being breastfed for more than 6 months (p = 0.026) and with symptoms (runny nose and/or cough) of common cold (p < 0.001) increased over the study period.     In a multiple regression analysis using pooled data of the three periods, positive predictors for Sp-carriage among DCC-children (   Figure 1).

Streptococcus pneumoniae and its antimicrobial non-susceptibility
The proportion of Sp-strains that were non-susceptible against any of the five tested antibiotics remained stable over the study period ( were non-susceptible over the entire study period. The serotypes (based on Quellung-reaction) that were most often found to be non-susceptible against at least one of the tested antibiotics among Sp-carriers are shown in Figure 4.
Over the 3-year study period, 23.2% (13/56 19A-strains) of the 19A-strains were non-susceptible to more than one of the tested antibiotics, whereas two 19A-strains (3.6%; 2/56 19A-strains) were non-susceptible to only one of the tested antibiotics; erythromycin and trimethoprim/sulfamethoxazole for the respective strains.

Discussion
In this pneumococcal carriage study, we evaluated during 3 years any changes in the nasopharyngeal carriage prevalence, serotype distribution and antimicrobial (non-)susceptibility of Sp in healthy children (aged 6-30 months) attending DCCs in Belgium, from 2016 onwards, i.e. during and immediately after a PCV13-to-PCV10 vaccination programme switch. Common co-colonising bacteria were followed as well, with a special focus on Hi.
Demographics and clinical characteristics of the study population showed higher percentages of common cold symptoms in 2016-2017 and 2017-2018 compared with 2016, which might be due to differences in sampling period. In 2016, nasopharyngeal samples were taken during spring (March-July), whereas the subsequent periods encompassed autumn and winter (November-March), during which common cold frequently occurs. The higher percentages of AOM-history in 2016 compared with the other periods might be due to recruitment of (on average) older children in this year compared with the subsequent periods and in 2016 more AB-treatments in the 3 months before sampling were observed. With regard to breastfeeding, it is unclear why it is more frequent in 2017-2018 compared with the previous periods. It is likely that several factors contributed to fluctuations in breastfeeding practices, one of which may be the coinciding extensive campaigns (personal communication: Marc Hainaut, 24 Oct 2018) on the importance of breastfeeding in Brussels' maternity hospitals (in 2017-2018, > 54% of the included Brussels children were breastfed for more than 6 months).
During the study period, the vaccination status of the child populations gradually changed; from mainly PCV13-vaccinated children in 2016 to mainly PCV10vaccinated children in 2017-2018. Sp-carriage prevalence was consistently high (> 75%) over the study period. Real-life carriage in DCC might be slightly lower since children who were treated with oral ABs in the 7 days before sampling could not take part in the study. In contrast, the carriage prevalence of Hi increased significantly over the study period (especially in the two oldest age categories), despite the increasing number of children vaccinated with PCV10, containing the NTHi Error bars indicate 95% confidence intervals, numbers added to the respective line charts indicate the total number of isolates for the respective serotype during the indicated study period.
The predictors of Sp-carriage identified in our study include study period, sex, siblings, common cold symptoms, use of antibiotics, Hi-carriage, Mc-carriage, Sa-carriage and confirm the findings of other reports (besides study period and sex) [5,19,20,42,43].
Based on Quellung-reaction, the proportion of non-PCV13-VTs decreased significantly over the study period, associated with an increase in the proportion of the three PCV13-non-PCV10-VTs 3, 6A, 19A (mainly 19A). We verified for confounders (preterm delivery, previous hospitalisation, age-appropriate vaccination, GAS-carriage, parental smoking, breastfeeding, and the variables shown in Table 2) that could have caused increased PCV13-non-PCV10-VT-carriage, but could not identify any besides study period, which strengthens the hypothesis that the increase was caused by the vaccine switch.
Serotype 19A became the most frequent vaccine serotype in 2017-2018 and its PCR-based prevalence rose from 0.4% in 2016 to 6.4% in 2017-2018. According to surveillance data on IPD from the NRC in 2017, serotype 19A was the second most frequent serotype (after serotype 12F) among IPD-isolates of children younger than 2 years of age. While no increase in serotype 19A frequency had been noted since the introduction of PCV13 in 2011, an increase was observed for the first time in 2016, when the frequency changed from 2.1% in 2016 to 14.2% in 2017 [44]. These findings were confirmed in 2018, 2 to 3 years after the switch from PCV13 to PCV10 [45]. In addition to being reported as an invasive serotype, 19A has also been reported as a serotype that is frequently non-susceptible to antimicrobials [46]. Nevertheless, in our study other serotypes dominated  a Non susceptibility to antibiotics was assessed using cultures and the overall proportions of non-susceptible isolates (against ≥ 1 and > 1 antibiotic) are represented by lines.
b Dominating serotypes among strains of Streptococcus pneumoniae non-susceptible to ≥ 1 antibiotic as well as all the other remaining serotypes are shown within the chart bars. For each particular serotype presented, the serotype number, followed by the proportion and number of isolates for this serotype is indicated. Serotypes were determined by Quellung-reaction.
among the non-susceptible strains (23B, 11A, 15B) as they were more prevalent than 19A. Since we excluded children who were treated with oral ABs in the 7 days before sampling, we possibly missed some non-susceptible strains (including 19A). Besides for serotype 19A, a consistently increasing proportion was also observed for the carriage of serotype 6C, which increased from 0.9% in 2016 to 5.8% in 2017-2018. Sweden, where PCV13 is used in some regions, while PCV10 is used in others, also reported an increase in serotype 6C in PCV10-regions [47]. This could have implications for the non-vaccinated elderly, as reported in Finland, where serotypes 19A and 6C are frequently isolated from IPD in adults older than 65 years [48].
Our results should be interpreted in the context of several limitations. First, a decreasing age trend was introduced by recruiting older children in 2016 in order to include a maximum of children vaccinated with PCV13 and for the same reason, Wallonia was over-recruited in 2016. Since this is intrinsic to our study design, we cannot adjust for this: the sampling probabilities did not allow re-weighting for any of the analyses because no children were sampled in the youngest age category in Flanders and in Brussels in 2016.
Furthermore, the Sp-carriage prevalence was stable within the different age categories (6-12, 13-24, 25-30 months) over the study period, but the Hi-carriage prevalence in the two oldest age categories increased over the study period. Second, we over-recruited in Wallonia in 2016 to enlarge the PCV13-vaccinated population, but in the first season no regional differences in overall Hi-carriage or pneumococcal carriage, and vaccine type carriage were found. Third, a comparative analysis based on the children's vaccination schedule was not performed due to the small size of these subpopulations in either 2016 (few PCV10-vaccinated children) or 2017-2018 (few PCV13-vaccinated children). Finally, a 3-year study period is short to completely exclude natural fluctuations in the carriage prevalence of specific serotypes; a follow-up study should confirm whether or not a new equilibrium is reached.
Despite these limitations, our study allowed to monitor the impact of the PCV13-to-PCV10 vaccine switch on nasopharyngeal carriage, serotype distribution, and antimicrobial (non-)susceptibility of Sp. This is a complementary activity to the analysis of IPD-surveillance data, thus providing further basis for policy making on pneumococcal vaccination programme options.

Conclusions
As the proportion of children vaccinated exclusively with PCV10 increased, the proportion of the serotypes not included in PCV13 decreased over the 3-year study period, mainly due to an increase in the carriage prevalence of serotype 19A.