Probiotics and carriage of Streptococcus pneumoniae serotypes in Danish children, a double-blind randomized controlled trial

This study examined the carriage of Streptococcus pneumoniae in healthy Danish children aged 8–19 months and assessed the effect of the probiotics Lactobacillus rhamnosus GG and Bifidobacterium animalis subsp lactis on the pneumococcal carriage during daycare enrolment. Potential risk factors of pneumococcal carriage were analysed and the carriage study was compared with registered invasive pneumococcal disease (IPD) data. This study is a part of the ProbiComp study, which was a double-blind, randomized controlled trial, including 290 children allocated to probiotics or placebo for 6 months and recruited during two autumn seasons (2014/2015). Pneumococci were identified by optochin sensitivity, bile solubility, α-hemolysis and/or capsular reaction. Serotyping was performed by latex agglutination kit and Quellung reaction. The carriage rate of S. pneumoniae was 26.0% at baseline and 67.4% at the end of intervention. No significant difference was observed between the placebo group and the probiotics group (p = 0.508). Children aged 8–19 months were carriers of non-pneumococcal vaccine serotypes causing IPD in children aged 0–4 years. However, serotypes causing most IPD cases in Danish elderly were either not found or found with low prevalence suggesting that children are not the main reservoir of those serotypes and other age groups need to be considered as carriers.

in the incidences of IPD caused by non-vaccine serotypes has been observed 1,6,7 . Thus, it is important to continue surveillance of serotype distribution.
Only a few pneumococcal carriage studies in children have been conducted in Denmark, the most recent was prior to the introduction of PCV7 9 .
The purpose of this study was to assess the effect of probiotics on S. pneumoniae carriage in healthy Danish children aged 8-19 months, determine serotype distribution and analyse risk factors for pneumococcal carriage. Furthermore, to assess whether the carried serotypes are also isolated in IPD cases of all age groups as described by Slotved et al. (2016) 14,15 and Harboe et al. 16 .

Methods
Study population. This study is part of the ProbiComp study, which was a randomised, double-blind, placebo-controlled parallel study investigating the effect of probiotics on infections in young children starting daycare described in details in Laursen et al. 17 .
The ProbiComp study included 290 healthy children aged 8-13 months starting daycare within 12 weeks after start of intervention. They were randomly assigned to a combination of the probiotics Lactobacillus rhamnosus GG (LGG) and Bifidobacterium animalis subsp lactis (BB-12), administered orally, in a dose of 10 9 colony-forming units/day (CFU/day) of each or placebo (maltodextrin) for 6 months. Children were examined at baseline before start of intervention and 6 months later at the end of intervention. The children were recruited during two autumn seasons from mid-August to mid-December in 2014 and 2015.
LGG and BB-12 are registered trademarks of Chr. Hansen A/S. For exclusion criteria see Laursen et al. 17 . Study design, compliance, randomisation, data collection, outcome measurements and sample size calculation are described in Laursen et al. 17 . Briefly, both parents and study personnel were blinded to group allocation, and the placebo powder and the probiotics powder did not differ in smell, taste or colour. Parents registered daily whether the child had ingested the product and by the end of intervention the parents returned the registration sheets and unused sachets with powder to evaluate compliance. PCV13 is administered in a 2 + 1 dose schedule at the age of 3, 5 and 12 months 1 . At baseline 98.2% of the children examined were covered by at least one dose of PCV13, while 92.6% were covered by two doses. At the end of intervention 88.0% of the children were covered by all three doses of PCV13. Vaccination data were obtained at the Danish Vaccination Register (DDV) (record number 2015-57-0102). Information regarding the children's background and health were obtained by interviewing parents and during the intervention period occurrence of children's symptoms of illness, absence from daycare, doctor's visits, and doctor-diagnosed illnesses were registered by the parents in weekly and daily web-based questionnaires 17 .
Nasal swab sampling. Two autumn seasons were included in the study. At baseline, samples were analysed from 141 children from the first season (2014) and 144 children from the second season (2015). At the end of intervention, samples from 124 children from the first season and 134 children from the second season were analysed.
Nasal swab samples were collected by a modified version of Satzke et al. 18 . Minitip flocked swabs (FLOQSwabs ™ , Copan, Italy) were used to take the samples. The swabs were inserted as far as possible into the nasal cavity and rotated, although within the limitations of the children's comfort. They were then placed in 1 mL Luria-Bertani (LB) broth with 10% glycerol in cryotubes, and stored at −80 °C until analysis.
Identification of pneumococcal serotypes. Identification of pneumococcal serotypes was performed as previously described 4 . Briefly, 10 µL of each sample was added to 3 mL serum-ox broth and incubated overnight at 37 °C in 5% CO 2 , before plating. The following day 1 µL of each serum-ox broth was cultured on 10% horse blood agar plates, which were incubated overnight at 37 °C, 5% CO 2 . S. pneumoniae were identified based on optochin sensitivity, bile solubility, α-hemolysis and/or capsular reaction (Quellung reaction). Pneumotest latex agglutination kit (SSI Diagnostica, Hillerød, Denmark) was performed on the serum-broth to determine pneumococcal group. Serotypes were identified by the Quellung reaction (Neufeld test) using serotype specific antisera (SSI Diagnostica, Hillerød, Denmark). The specimens were screened by pneumotest latex agglutination kit for multiple serotypes. If multiple serotypes were found, they were isolated and serotyped.
Invasive pneumococcal disease data. Data  We examined several risk factors for pneumococcal carriage including sex, having siblings <5 years, breastfeeding at baseline examination, living with a dog or cat, passive smoking exposure in the household, having respiratory infections such as bronchitis, pneumonia or otitis media during the intervening period and the effect of receiving antibiotics during the intervening period of 6 months. We did not have access to dates on when antibiotics were received. Univariate logistic regression was not undertaken as confounders were evenly distributed during the randomization process, therefore multivariate logistic regression was not required. Crude odds ratio of pneumococcal carriage was estimated in each characteristic (sex, having siblings <5, breastfeeding, respiratory infections including bronchitis, pneumonia and otitis media, antibiotic use, passive smoking exposure and exposure to cats or dogs) using two-tailed Fisher's exact test.

Ethical considerations. The ProbiComp study protocol was approved by the Committees on Biomedical
Research Ethics for the Capital Region of Denmark (H-4-2014-032), and we hereby confirm that all methods were performed according to the guidelines and regulations approved by the Committees on Biomedical Research Ethics. The study was registered at clinicaltrials.gov (identifier NCT02180581 (Supplementary file), posted 02/07/2014) 17 . Informed consent from parents and legal guardians of the children was required. Participation in the study was voluntary and parents could withdraw their consent at any time 17 .
Regarding IPD data, no ethical approval or informed consent was required since data were collected routinely for national surveillance purposes. Using the data is approved by the Danish Data Protection Agency (record number 2007-41-0229).

Results
Carriage study. The carriage rate of S. pneumoniae at baseline of the two seasons were comparable with no statistical significant difference between the two seasons as seen in Table 1. Hence, the two seasons were combined as one baseline group for further analysis. The mean carriage rate at baseline therefore was 26.0% (CI 20.8-31.1%) ( Table 1).
At the end of intervention half of the children had received the probiotics LGG and BB-12 for 6 months, hence each season included both a placebo group and a probiotics given group ( Table 1). The two seasons were alike with no statistical significant difference between the two placebo groups and the two probiotics groups, hence data were combined into one placebo group and one probiotics group ( Table 1).
The carriage rate of S. pneumoniae in children receiving the probiotics LGG and BB-12 was 69.5% (CI 61.6-77.5%) while the carriage rate in the placebo group was 65.4% (CI 57.2-73.6%). Thus, probiotics do not have a significant effect on the pneumococcal carriage rate (p = 0.508) ( Table 1). Furthermore, the serotype diversity/ distribution in the probiotics group and placebo group were not significantly different (data not shown). The placebo group and the probiotics group were therefore combined into one group representing the end of intervention for further analysis. The carriage rate at the end of intervention was 67.4% (CI 61.7-73.2%), which is significantly higher than the carriage rate at baseline (p < 0.001) ( Table 1).
All the cases of multiple serotype carriage involved two serotypes except one child at the end of intervention, who carried three serotypes ( Table 2).
The 81 isolates found at baseline belonged to 21 different serotypes and 1 serogroup (where serotyping was not possible) (Fig. 1). The most common serotype isolated at baseline was serotype 21 constituting 13.6% of all isolates. Serotype 23B, 35F and 11A were also commonly isolated, constituting 12.3%, 11.1% and 11.1% of all isolates respectively (Fig. 1). The only PCV13-serotypes isolated at baseline were 19A and 19F, which both constituted 2.5% of all isolates.
Serotype 8 was only found in two children at baseline (Fig. 1). Risk factors. Sex did not influence the carriage rate of S. pneumoniae significantly (p = 0.280 at baseline and p = 0.187 at end of intervention) (Tables 3 and 4).
Having siblings <5 years significantly increased the carriage rate of S. pneumoniae in children at baseline (p < 0.001), but did not influence the carriage rate of children at the end of intervention (p = 0.883) (Tables 3 and  4).
Breastfeeding at the time of baseline examination, receiving systemic antibiotics during the intervention period, exposure to passive smoking or living with a dog or cat did not influence the carriage rate of S. pneumoniae (Tables 3 and 4).
We examined if children with respiratory infections during the intervening period had an increased risk of pneumococcal carriage, but no significant difference in carriage rate was observed between children having had respiratory infections or children not affected by respiratory infections (p = 0.425) ( Table 4). However, children who had experienced bronchitis during intervention seemed to have a lower carriage rate compared to children who had not experienced bronchitis (p = 0.031), but this result may not be reliable due to the low number of children in the cohort (Table 4). Serotype distribution in children who have experienced bronchitis, pneumonia or otitis media is shown in Fig. 2.

Registered invasive pneumococcal diseases in Denmark. Data regarding IPD-causing serotypes in
Denmark were obtained from the Danish laboratory surveillance system as mentioned in the methods section and Fig. 3 shows the data. Serotype

Discussion
To our knowledge this is the first carriage study of S. pneumoniae in children after the introduction of PCV in Denmark in 2007. Oral administration of probiotics to children aged 8-13 months for 6 months had no effect on the carriage of pneumococcal serotypes as tested in two different years (2014 and 2015). Furthermore, we found an overall carriage rate in children attending daycare (end of intervention) of 67.4% as compared to a carriage rate of 69% before the introduction of PCV7 into the Danish Childhood Immunization Program 9,19 . Thus, PCV-use seems not to have influenced the overall carriage rate in children.
Living with siblings <5 years of age significantly increased the risk of pneumococcal carriage before starting in daycare (baseline) (p < 0.001) ( Table 3). This has also been observed by others [10][11][12] . We found no significant difference in carriage rate of S. pneumoniae among children receiving probiotics and children receiving placebo (Table 1). Nasopharyngeal carriage of S. pneumoniae is considered a prerequisite for developing pneumococcal associated diseases such as otitis media, pneumonia, bronchitis and other respiratory infections 5,23,24 . Wong et al. reported that the probiotic LGG inhibits the adherence of S. pneumoniae to human epithelial cells in vitro 25 . Clinical studies investigating the effect of administrating probiotics on the incidence of respiratory tract infections have shown inconclusive results. Laursen et al. 17 did not find LGG and BB-12 to reduce the number of days absent from daycare because of respiratory or gastrointestinal infections in the same children as included in our study, and the probiotics did not affect the endogenous gut microbiota composition in those children 26 . A Swedish study involving 248 healthy children aged 1-5 years and attending daycare showed a significantly lower incidence of otitis media in children who ingested milk with 10 7 CFU/mL Lactobacillus rhamnosus LB21 compared to placebo 27 . However, a study including 309 otitis media-prone children aged 10 months to 6 years showed no significant difference in the incidence of otitis media or nasopharyngeal carriage of S. pneumoniae between the group receiving capsules containing probiotics (8-9 × 10 9 CFU/capsule) and the placebo group 28 . Others found that LGG and Bifidobacterium lactis ingested orally reduced the risk of otitis media and recurrent respiratory infections in children <2 years 29 , as did milk containing LGG in healthy children aged 1-6 years 30 , while Skovbjerg et al. 31 report that nasal delivery of the probiotic Lactobacillus rhamnosus did not affect nasopharyngeal pneumococcal carriage in children aged 1-8 years with secretory otitis media 31 .
The carriage rate at the end of intervention was significantly higher than at baseline (p < 0.001) ( Table 1). The age difference between these two groups was small (8-13 months at baseline and 14-19 months at the end of intervention), and the use of probiotics during intervention did not affect the carriage rate (  Table 3. Potential risk factors for pneumococcal carriage in the baseline group and at the end of study. The table shows number of carriers, total number of children, carriage rates, p-values, odds ratios (OR) and confidence intervals (CI). * Odds ratios, confidence intervals and p-values were calculated by two-tailed Fisher's Exact Test.

**
No siblings at all *** including bronchitis, pneumonia, sore throat, sinusitis, croup and otitis media N/A = not available. CI: confidence interval, OR: odds ratio. factors could not be the reason for this difference. Baseline samples were collected during autumn season, while samples at the end of intervention were collected 6 months later (spring season), however the effect of seasonality on the carriage of S. pneumoniae is not clear 4 . The children were not attending daycare at baseline, while all the children attended daycare at the end of intervention. Attending daycare is considered an important risk factor in relation to acquisition of pneumococcal carriage 9,11,12 , hence the higher carriage rate at the end of intervention might be explained by daycare attendance.
Since we consider our data non-parametric and the general choice of statistical test within this field is two-tailed Fishers exact test we have used the univariate analysis by two-tailed Fisher's Exact Test to calculate the risk factors. This is not an age-matched group however the range of the age group is very small and we have not found studies showing such small age range to chance the carriage rate but external factors do chance the carriage rate such as daycare attendance 9,11,12 . Most studies include children aged 0-24 months as one group 3,4,7,9,20,29 and since the children included in our study are aged 8-19 months we believe the age range is not significant. The main differences between the baseline group (aged 8-13 months) and end of intervention group (aged 14-19 months) is daycare attendance and whether the children have received probiotics or not.
It is believed that children with high colonization rates of S. pneumoniae are the key transmitters of non-vaccine serotypes causing IPD to other age groups 6,8,22,32 . IPD caused by vaccine-serotypes has declined significantly after the implementation of PCV in children, while IPD caused by non-vaccine serotypes has increased 8,14,33 . This is consistent with the decline in carriage of vaccine-serotypes and an increase in carriage of non-vaccine serotypes in children which is supported here (Fig. 1) and by others 7,20,22,34 . Despite the use of effective vaccines, serotypes 3, 7F and 19A, which are included in PCV13, are still causing IPD among elderly in Denmark 13,14 (Fig. 3). The low carriage rate of serotype 8 among 8-19 months old Danish children in our study contrasts the high prevalence of serotype 8 seen among elderly patients with IPD in Denmark. The other 5 most common serotypes (3, 22F, 9N, 12F and 7F) isolated from elderly with IPD in 2014-2016 were either not found or found with low prevalence (Figs 1 and 3). It has been assumed that immunization of infants alone would result in herd immunity and hence protect adults and elderly against disease caused by vaccine-serotypes 21,35,36 . Our findings suggest that children aged 8-19 months are not the main reservoir of serotypes causing IPD in elderly, but suggest that other age groups need to be considered as main carriers of some pneumococcal serotypes.
The limitation of this study is that it included a narrow age group of children (8-19 months) and that the study only included children living in the capital region of Denmark. The strength of the study is, that we have  examined very young children who usually are difficult to collect swab samples from and were able to obtain detailed background information regarding the participants.
In conclusion, this study revealed that PCV-serotypes have disappeared in children and been replaced by non-PCV serotypes after the introduction of PCVs in Denmark. The probiotics LGG and BB-12 did not prevent carriage of S. pneumoniae in children attending daycare. Having a sibling <5 years before daycare increased the risk of pneumococcal carriage significantly. Children aged 8-19 months are carriers of pneumococcal serotypes causing IPD in children aged 0-4 years. However, the children seem not to be carriers of serotypes causing the majority of IPD cases in elderly in Denmark.  IPD incidences per 100.000. Serotypes are mentioned as PCV7-serotypes first, then PCV13-serotypes followed by non-PCV13-serotypes. Data were obtained from the Danish laboratory surveillance system at the National Neisseria and Streptococcus Reference Laboratory (NSR), Statens Serum Institut (SSI) as described by   14,15 .