Factors associated with pneumococcal nasopharyngeal carriage: A systematic review

Pneumococcal disease is a major contributor to global childhood morbidity and mortality and is more common in low- and middle-income countries (LMICs) than in high-income countries. Pneumococcal carriage is a prerequisite for pneumococcal disease. Pneumococcal conjugate vaccine reduces vaccine-type carriage and disease. However, pneumococcal carriage and disease persist, and it is important to identify other potentially modifiable factors associated with pneumococcal carriage and determine if risk factors differ between low, middle, and high-income countries. This information may help inform pneumococcal disease prevention programs. This systematic literature review describes factors associated with pneumococcal carriage stratified by country income status and summarises pneumococcal carriage rates for included studies. We undertook a systematic search of English-language pneumococcal nasopharyngeal carriage studies up to 30th June 2021. Peer-reviewed studies reporting factors associated with overall pneumococcal nasopharyngeal carriage in healthy, community-based study populations were eligible for inclusion. Two researchers independently reviewed studies to determine eligibility. Results are presented as narrative summaries. This review is registered with PROSPERO, CRD42020186914. Eighty-two studies were included, and 46 (56%) were conducted in LMICs. There was heterogeneity in the factors assessed in each study. Factors positively associated with pneumococcal carriage in all income classification were young age, ethnicity, symptoms of respiratory tract infection, childcare attendance, living with young children, poverty, exposure to smoke, season, and co-colonisation with other pathogens. Breastfeeding and antibiotic use were protective against carriage in all income classifications. Median (interquartile range) pneumococcal carriage rates differed by income classification, ranging from 51% (19.3–70.2%), 38.5% (19.3–51.6%), 31.5% (19.0–51.0%), 28.5% (16.8–35.4%), (P = 0.005) in low-, lower-middle, upper-middle, and high-income classifications, respectively. Our findings suggest that where measured, factors associated with pneumococcal nasopharyngeal carriage are similar across income classifications, despite the highest pneumococcal carriage rates being in low-income classifications. Reducing viral transmission through vaccination and public health interventions to address social determinants of health would play an important role.

Thank you for inviting me to review this paper. A well written and comprehensive review. However, I suggest minor revision for this review. Find below my comments: 1. Line 48: Write out the full meaning of HIV

Authors' response
We thank Reviewer 1 for their time and consideration of our manuscript. We have amended the manuscript accordingly.
2. Line 62: Any reason why facility-based studies were not included?

Authors' response
To minimise the non-comparability of included studies, we restricted studies to those of communitybased populations. Inclusion of hospital or facility-based studies may introduce bias. Firstly, pneumococcal carriage rates among community-based populations and those attending health care facilities (particularly to treat respiratory infections) have previously been found to differ (Coles CL et al. Trop Med Int Health 2009;Kotler L et al. Vaccine. 2021). Secondly, although there are approximately 100 distinct pneumococcal serotypes, a relatively small proportion of the serotypes is responsible for most invasive pneumococcal disease (Hausdorf WP et al. Clin Infect Dis. 2000). As such, there may be a bias towards more invasive vaccine-serotypes among hospitalised populations than community-based populations (Hausdorf WP et al. Clin Infect Dis. 2000). Compared with community-based populations, hospital/facility-based populations have higher rates of exposure to factors that are relevant to facility-based populations, but less so to community-based populations, such as length of stay and receipt of antibiotics (Fortinksky RH et al. J Appl Gerontol.2014;Grant CC et al. J Paediatr Child Health. 2012). Additionally, hospitalised populations have higher rates of exposure to risk factors found to be associated with pneumococcal carriage risk factors, including crowding, exposure to cigarette smoke, comorbidities, receipt of antibiotics, and co-colonisation with other pathogens (e.g., Haemophilus influenzae and influenza type A and B) (Fortinksky RH et al. J 4. Line 88: Full meaning of WHO.

Authors' response
We have amended our manuscript accordingly. 5. Line 97: Give the basis for the use Kruskal-Wallis method for this analysis.

Authors' response
We have amended the Methods section to clarify that "We used a Kruskal-Wallis test to compare overall pneumococcal nasopharyngeal carriage rates by income classification (low, lower-middle, upper-middle, and high), as this is a rank-based, nonparametric method for testing differences between two or more categorical, independent groups on a continuous (or ordinal) outcome (where study-level carriage rates were treated as continuous)." 6. Line 120: In the discussion section can you give likely reasons for having just only one RCT in this study.

Authors' response
RCTs are considered the gold standard for investigating causal relationships between exposures/interventions and outcomes. However, this review focuses on factors associated with pneumococcal carriage rather than interventions to reduce pneumococcal carriage. As such, RCTs are less relevant to our research question. 7. Line 147: Any statement on the poor-quality studies? Were they included in the final analysis? Reasons for inclusion or non-inclusion.

Authors' response
In our original submission, supplementary file S Table 4 (Quality assessment of included studies) included a statement on poor-quality studies. In our revised manuscript, we have moved the statement on poor-quality studies to the Results section under the heading Quality Assessment. It reads, "Underlying reasons for poor quality related to a lack of sufficient detail in study methods. A cross-sectional study conducted in Hong Kong was considered poor quality as the methods were unclear, including a lack of clarity around the inclusion criteria and insufficient descriptions of pneumococcal carriage detection and statistical methods (Sung RY et al. Acta Paediatr. 1995). In this study, it was unclear how multivariable logistic regression models were built, there was no discussion regarding variable selection, and the only indication that a multivariable model had been used was in the abstract (Sung RY et al. Acta Paediatr. 1995). A cross-sectional study from Bolivia was considered poor quality, as it lacked a clear research question, the inclusion and exclusion criteria were unclear, no sample size calculation was included, and the description of statistical methods was insufficient (Inverarity D et al. Trans R Soc Trop Med Hyg. 2011). Studies were not excluded based on poor quality to ensure transparency and completeness of reporting from all studies identified as relevant to the review (Shea BJ et al. BMJ. 2017)." As a narrative summary has been presented rather than a meta-analysis, and the limitations have been outlined, no sensitivity analyses have been undertaken.
8. Discussion -Need to make specific public health recommendations based on the findings of this review.

Authors' response
In our revised Discussion, we suggest, we suggest "Identifying factors associated with pneumococcal carriage in certain settings may help inform other public health interventions that may be needed. Some risk factors are not modifiable, such as age, living with young children (however this is most likely due to increased viral transmission in this age group), and ethnicity. However, the risk of pneumococcal carriage, transmission, and disease may be reduced by public health programs and policies that target particular age groups (Berical AC et al Ann Am Thorac Soc.2016;Weinberger DM et al. Am J Epidemiol. 2018), to reduce transmission, such as , such as increased access to improved sanitation and hygiene (Mattos KJ et al. Env Eng Sci. 2021;Gudnason T et al. Scand J Infect Dis. 2014), or that are tailored to address socio-economic differences and social determinants of health which promote tranmission (Dunne EM et al. Pneumonia. 2016). Reducing environmental risk factors for pneumococcal carriage and viral transmission includes improving breastfeeding, reducing malnutrition, preventing overcrowding, enhancing respiratory etiquette, and reducing smoke exposure. Public health programs that promote birth spacing (which may reduce the number of young siblings living in the same household), breastfeeding, interventions to reduce poverty, and which ensure high coverage of infant vaccination, may reduce the risk of pneumococcal disease (Danino D et al. Clin Infect Dis. 2021;Rybak A et al. Pathogens. 2021;Kim DH et al. Int J Environ Res Pub Health. 2021). Many of these modifiable factors are included in the WHO integrated Global Action Plan for the Prevention and Control of Pneumonia and Diarrhoea (World Health Organisation. GAPPD. 2013). Having programs to address these factors would also help prevent other infectious diseases that are a common cause of child morbidity and mortality in LMICs. (Gastaldi A et al. Children. 2021;Di Mattia G et al. Pediatr Pulmonol. 2021). In Israel, pneumonia admissions declined despite pneumococcal carriage and density remaining unchanged. However, the circulation of other co-colonizing viruses which are known to increase the virulence of pneumococci declined substantially during the lockdown periods (Danino D et al. Clin Infect Dis. 2021;Smith CM et al. Am J Respir Crit Care Med. 2014). Although not assessed in primary studies in this review, pneumococcal carriage is more frequent in young children during infection with RSV than with other viruses (Brealey JC et al. Respirology. 2018;Sender V et al. Front Cell Infect Microbiol. 2021). Further, RSV stimulates substantial growth of pneumococci, and co-colonization with RSV is associated with increased pneumococcal density and severity of acute respiratory tract infections (Brealey JC et al. Respirology. 2018;Weinberger DM et al. PLoS Med. 2015;Morpeth SC et al. Sci Rep. 2018). Additionally, co-colonization with influenza and parainfluenza have been found to increase the probability of pneumococcal acquisition (Grijalva CG et al. Clin Infect Dis. 2014). This suggests that public health interventions that modify the transmission of viral respiratory pathogens are also very important in preventing pneumococcal disease, including vaccines against RSV and influenza, and other interventions that reduce viral pathogen transmission. Co-colonization with H. influenzae and M. catarrhalis were also found to be risk factors for pneumococcal carriage in this review." 9. Line 384: There is need to include the "References" heading.

Authors' response
We have amended our manuscript accordingly.
10. Figure 1 -This is blur. Include a clearer picture.

Authors' response We have uploaded our figure files to the Preflight Analysis and Conversion Engine to ensure figures meet PLOS requirements.
Reviewer 2: This systematic review describes some factors associated with pneumococcal carriage in children by classification of the settings on the income of country. It also describes carriage rates in children across those country level income categories. The manuscript is well structured and clearly written. The methods are acceptable and clearly described in adequate detail. I have the following comments to make.
A systematic review to answer this question need to rely on what RFs are assessed in the primary studies and that will not obviously be similar across studies. This needs to be properly highlighted and needs to be addressed in the methods with mention of any attempt to select homogenous studies in this regard somehow (noting studies in LMIC are few).

Authors' response
We thank Reviewer 2 for their time and review of our manuscript. An important point has been raised regarding the comparability of risk factors across studies. Due to the observational nature of all but one of the included studies, variables collected for assessment with pneumococcal carriage were dictated by primary study focus, design, and context. Comparability of risk factors across studies was low, such that meta-analysis was not appropriate. In our revised manuscript, we have amended the Eligibility criteria of the Methods section to include "As risk factors assessed in primary studies were unlikely to be similar across all pneumococcal carriage studies, we limited our review to studies of healthy, community-based populations in an attempt to select studies with as similar as possible exposures. For transparency and completeness, we also present the results for all factors assessed for association with pneumococcal carriage by each included primary study." How representative are these studies for each income category setting in particular in the low income settings. Even within these countries there could be variation between regions and population groups.

Authors' response
This review included primary studies which were most frequently based in high-income countries. Low and lower-middle-income countries were less represented. We agree with Reviewer 2 that there is likely variation between regions and within population groups. It is also very likely that there is underrepresentation from low-and middle-income countries, limiting the representativeness of these studies for these income settings. In addition, most studies used convenience sampling, and therefore their study populations may not be representative of their general populations. In our revised manuscript, we have amended the Limitations sections to include: "Although articles from low-and lower-middle-income countries were included in this review, most primary studies were conducted in high-income countries. Low-and lower-middle-income countries were proportionally underrepresented, limiting the potential representativeness of studies for these income settings. Further, most studies used convenience sampling. For these reasons, the studies for which pneumococcal nasopharyngeal carriage rates were available may not be representative of regional, country, or within-country populations. Therefore, we caution against using the reported rates by income classification as population or sub-population rates." I get the sense that authors somewhat downplay the impact of PCV use in changing the pneumococcal carriage levels. It is true that in some settings NVTs have replaced VTs in carriage with not a substantial decline overall. However, there is overwhelming evidence that in all settings there is large reductions in disease particularly that of severe end of the spectrum (IPD). The classification of these study settings needs to be considered in terms of PCV use, schedule and duration of program.