Pneumococcal Carriage in Sub-Saharan Africa—A Systematic Review

Background Pneumococcal epidemiology varies geographically and few data are available from the African continent. We assess pneumococcal carriage from studies conducted in sub-Saharan Africa (sSA) before and after the pneumococcal conjugate vaccine (PCV) era. Methods A search for pneumococcal carriage studies published before 2012 was conducted to describe carriage in sSA. The review also describes pneumococcal serotypes and assesses the impact of vaccination on carriage in this region. Results Fifty-seven studies were included in this review with the majority (40.3%) from South Africa. There was considerable variability in the prevalence of carriage between studies (I-squared statistic = 99%). Carriage was higher in children and decreased with increasing age, 63.2% (95% CI: 55.6–70.8) in children less than 5 years, 42.6% (95% CI: 29.9–55.4) in children 5–15 years and 28.0% (95% CI: 19.0–37.0) in adults older than 15 years. There was no difference in the prevalence of carriage between males and females in 9/11 studies. Serotypes 19F, 6B, 6A, 14 and 23F were the five most common isolates. A meta-analysis of four randomized trials of PCV vaccination in children aged 9–24 months showed that carriage of vaccine type (VT) serotypes decreased with PCV vaccination; however, overall carriage remained the same because of a concomitant increase in non-vaccine type (NVT) serotypes. Conclusion Pneumococcal carriage is generally high in the African continent, particularly in young children. The five most common serotypes in sSA are among the top seven serotypes that cause invasive pneumococcal disease in children globally. These serotypes are covered by the two PCVs recommended for routine childhood immunization by the WHO. The distribution of serotypes found in the nasopharynx is altered by PCV vaccination.


Introduction
The human nasopharynx is the main reservoir for pneumococci. The bacteria which adhere to pharyngeal epithelial cells through epithelial receptor molecules may be acquired very early in life [1,2], and in most children the pneumococcus is present in the nasopharynx at some point in the first few years of life [3]. Carriage is generally higher in developing countries and among economically deprived populations [4,5]. The prevalence of carriage might also vary between developing countries. In one study, Abdullahi et al suggested that colonisation prevalence in East and Southern Africa is substantially lower than in the Gambia [6]. High prevalence have however been reported in Ethiopia and Mozambique.
Carriage is a prerequisite for disease [3,7] and because it is much more common than a disease outcome, it may be a valuable measure of the efficacy of new pneumococcal vaccines [8]. The relation between carriage and disease was first demonstrated in a cohort of infants [9]. Subsequent studies showed that carriage is a risk factor for acute and recurrent otitis media in children [10,11]. Other studies have shown that bacterial carriage densities may be related to the risk of disease in adults and children [12,13], and O'Brien et al have suggested that PCV may reduce carriage density in children [14].
Since the introduction of PCV, several studies have reported a reduction in invasive pneumococcal disease (IPD). However, this is frequently accompanied by a change in the distribution of circulating serotypes. A decrease in vaccine type (VT) IPD and an increase in non-vaccine type (NVT) IPD have been reported in America [15], Spain [16], Canada [17] and Australia [18]. In particular, serotype 19A has been isolated more frequently after the introduction of PCV 7 [19][20][21][22].
This review of pneumococcal carriage in sSA aims to: 1) describe the variability in carriage prevalence across countries in sSA; 2) describe the distribution of serotypes, and 3) assess the impact of pneumococcal vaccination on carriage of VT and NVT serotypes.

Methods
A comprehensive literature search strategy was developed to identify published articles describing pneumococcal carriage in sSA (Appendix S1). The search was conducted in December 2011 using the electronic databases MEDLINE (from 1950), EMBASE (from 1947) and African Index Medicus (AIM). To ensure the retrieval of relevant articles, the search was performed by exploring and combining medical subject headings (MeSH) and free search terms relating to carriage, nasopharyngeal, oropha-  ryngeal, Streptococcus pneumoniae, serotypes, pneumococcal vaccine and specific names of the African countries. Titles and abstracts were reviewed and duplicates, non-relevant studies, and those involving streptococcal infections other than S. pneumoniae were excluded ( Figure 1). The full texts of potential papers were then screened for eligibility.

Inclusion and exclusion criteria
The review was limited to studies from countries within the sSA region that reported the prevalence of carriage with or without serotyping of the pneumococcal isolates. We used data from both hospital and community based studies that collected swabs from either the nasopharynx or oropharynx. The search was limited to human subjects but there was no restriction on the age of participants, study design or language of publication. Bibliographies of relevant papers and review papers were searched to identify articles that may have been missed in the electronic search.

Data analysis
Data were obtained for the following variables: prevalence of S. pneumoniae, country, first author, year the study was conducted (or year of publication if the study year was not reported), age of participants, number of swabs collected per individual, health of the population swabbed, rural or urban setting, and season or months of the year when the study was conducted. The data were entered in an Excel spread sheet and Stata version 12 was used for all analyses. For studies with multiple swabs per individual, only results from the first swab were included in the analysis and for those with interventions, either PCV or other interventions such as antibiotics, only the control arm was included in the analysis. To assess the impact of PCV we used data from randomised trials where PCV was the intervention. The extracted data were reviewed independently by a second reviewer who checked the data to ensure completeness using the template prepared for data extraction.
To describe the prevalence of carriage by age, the studies were grouped as: ,5 years (children), 5-15 years (children), and .15 years (adults). Studies that recruited children in both age groups were assigned to the age group 5-15 years. Studies where participants were recruited across child and adult age groups, and where suitable stratified results were unavailable were excluded from the analysis of carriage by age. A random-effects model was used to summarise carriage by age group across the different studies. Studies where the standard error of the prevalence could not be computed were excluded from this analysis. The effects of region, season and urban/rural location on carriage were examined by comparing between studies using random effects model (meta-regression). For each study that reported carriage by gender, the absolute difference in prevalence between males and females (risk difference) was calculated and statistical significance was determined using Fisher's exact test.  Data from four of the studies were pooled to assess the impact of PCV on overall carriage and the carriage of VT and NVT serotypes among children 9 to 24 months. In all four studies children who received no PCV were compared with children who received at least three doses of PCV. Random effects models were used to estimate the average effect of PCV (DerSimonian-Laird estimate) across studies and to assess the degree of heterogeneity between studies.
Serotypes isolated in each study were ranked in order of prevalence and the five most prevalent serotypes in each study were identified. For each serotype, we determined the proportion of studies in which it was among five most prevalent serotypes.

Quality of the studies
The studies were reviewed for quality using the WHO guidelines for conducting nasopharyngeal studies. The guidelines are for the material used for sample collection, the technique of sample collection, and the transport media [23]. For each study, we identified potential sources of bias in the method of selection of study subjects.

Characteristics of the studies
A total of 57 studies were included in this review (Table 1). Southern Africa contributed the most studies, 23(40.3%). Twenty studies (35.1%) were from West Africa with more than half of these from The Gambia. There were 12(21.1%) and 2(3.5%) studies from East and Central Africa respectively.
The majority of the studies (87.7%) collected nasopharyngeal swabs, only 4(7.0%) collected oropharyngeal swabs, and 2(3.5%) studies collected both. In one study, the anatomical site of sampling was not reported. Calcium alginate was the most  24.5% were conducted in outpatients, 6.9% in HIV positive populations and in 1.7% the population was not stated. In 15(26.3%) studies, participants were swabbed more than once.

Pneumococcal carriage by age and geographic region
Carriage was highest for children less than 5 years and decreased with age (Table 2 and Figures S1, S2 & S3). High prevalence (.85%) in children were recorded in Ethiopia, Mozambique (only one study) and The Gambia. The Gambia also had the highest prevalence in adults ( Table 2). The prevalence of carriage varied considerably between studies. The I 2 index, which assesses heterogeneity between studies, was greater than 99% in all the age categories. In children less than 5 years the prevalence was higher in studies conducted in a rural, rather than urban setting. Carriage was not associated with season, population health, swab type or year (Table 3).

Pneumococcal carriage and gender
Eleven studies reported the prevalence of carriage by gender. Three of these studies reported no association, one study reported a higher prevalence in males compared to females (p = 0.05), and one study reported a higher prevalence in females (OR = 0.61; 95% CI: 0.39-0.95; p = 0.02). From our analysis, there was no significant difference in the risk of carriage between males and females in any of the remaining six studies (Table 4).

PCV and pneumococcal carriage
Seven studies from three countries (The Gambia, Kenya and South Africa) assessed the association between PCV and carriage. One study from the Gambia was a village cluster randomised trial with adults and older children in 10 villages receiving one dose of PCV 7, and adults and older children in 11 control villages receiving meningococcal serogroup C vaccine. In both arms of the trial, infants aged between 2 and 11 months received three doses of the vaccine given at monthly intervals, and children aged between 12 and 30 months received two doses at one month interval between doses. Infants born during the study received three doses of the vaccine given monthly at the ages 2, 3, and 4 months [24]. The other six studies compared carriage in vaccinated and unvaccinated children using data from individually randomised control trials (RCTs). Four of these studies collected carriage data on all children that participated in the trial [25][26][27][28] while two studies used data from a subsample of children enrolled in the original trial [29,30] (Table 5).
The participants in all the PCV studies were children from the general population presenting at infant welfare clinics for immunisation, except for the village cluster randomized trial which recruited children and adults in the community. Carriage of vaccine type (VT) serotypes was reduced by vaccination, while carriage of non-vaccine type serotypes was greater among vaccinated children (Figure 2 & Table 5). The prevalence of overall carriage was not affected by vaccination.
In The Gambia, the carriage study during the PCV 9 trial found that serotype 19F (11.5%) was again the most common VT serotype isolated and serogroup 15 was the most common NVT, and more prevalent in the PCV 9 vaccinated children (11.8% versus 8.5%, P,0.05). When the study children were swabbed a second time about 10 months later, NVT serotypes 10, 21 and 35B were isolated more frequently from PCV 9 vaccinated children than controls. There was no longer any difference between the groups for carriage of serotype 19F [29]. In the cluster randomised trial conducted in The Gambia, serotypes 23F, 6A, 6B, 3, 11 and 7C were the most common serotypes before vaccination and serotypes 3, 11, 19F and 6A were the most common serotypes after vaccination [24].

Discussion
This systematic review of pneumococcal carriage in sSA summarises the prevalence of carriage, distribution of serotypes and the effect of PCV on carriage. The majority of the studies were from Southern and West Africa, particularly South Africa and the Gambia. There were only two studies from Central Africa.
We found that the prevalence of pneumococcal carriage in sSA is generally high but there is much variation between countries, particularly among older age groups. Carriage was higher in children than adults as reported outside sSA [31,32].
A small number of studies conducted outside sSA have reported a higher prevalence of carriage in males compared to females [33][34][35]. However, in this review gender was not associated with  carriage in 9/11 studies, and in two studies carriage was more common among females [36,37]. One of these studies was conducted in South Africa among children 1-59 months with severe pneumonia, 67.3% of whom were HIV positive and another was in Nigeria in children 1-4 years who presented for regular check-up or immunisation, some of whom had a cough and a cold. The five serotypes that were most common in this review are among the seven that cause most global IPD in children; PCV 10 and PCV 13 will cover at least 70% of the cases of IPD caused by these serotypes [38,39]. The other two serotypes, serotypes 1 and 5, are rarely isolated from carriage studies, although they are often associated with pneumococcal disease epidemics [40][41][42]. Serotypes also differ in their ability to cause invasive pneumococcal disease [43].
It has been suggested that the impact of PCV on disease can be determined by pneumococcal carriage studies because it is newly acquired serotypes that lead to disease [9]. In this review, studies that assessed the impact of PCV on carriage generally showed a decrease in carriage of VT and an increase in NVT serotypes, with no change in the overall prevalence of carriage. One study in this review, and one in native Indians have shown a gradual decrease in overall carriage following vaccination [24,35]. Continuous surveillance of circulating serotypes will be important as countries introduce PCV.
Nasopharyngeal swabs are more sensitive for S.pneumoniae than the oral swabs [44,45]. The prevalence of carriage is therefore likely to be underestimated in the four studies that used oral swabs. The different lab methods used might also have been responsible for some of the variability in the prevalence of carriage reported in this review. WHO recommends calcium alginate or Dacron polyester swabs since cotton swabs suppress the pneumococcus [23]. However, only half of the studies (50.9%) followed the WHO guidelines, and in 13 (22.4%) the type of swab was not stated.
Another source of variation between studies is the prevalence of antibiotic use, since antibiotics might reduce carriage [46]. Some studies excluded those individuals who had taken antibiotics from their analysis. However, even when these individuals were excluded, often different periods were used to define prior use.
We have combined results from all available published studies irrespective of the study population (unpublished studies were not included in this review). Study participants were recruited from the community, day care centres, schools and outpatient clinics and hospital wards. Hospital patients may have higher carriage than the rest of the population, particularly if they were admitted for pneumonia. Generally, we expect selection bias to be less in studies conducted in the community compared with studies that use outpatient clinics. In this review, 57.9% of the studies were conducted in hospital/clinic settings and 35.1% of the studies in the community.
We have summarised available data on pneumococcal carriage in sub Saharan Africa. There remain unexplained differences in carriage within the region, and multi centre studies may provide reasons for some of the differences seen. Pneumococcal carriage studies can show indirect effects of PCV by showing changes in unvaccinated age groups and can supplement disease surveillance studies as PCV is introduced in the region. w-weeks, m-months, y-years, RD-Risk difference-Risk in the PCV vaccinated group minus the risk in the control group calculated in Stata, PPV-Polyvalent polysaccharide vaccine, ns-not stated, a children who received PCV 5 in an RCT and controls matched with age and place of residence who did not receive PCV, PCV b+c received b+c doses of PCV doses with b for the primary series and c for booster dose, d includes vaccine associated serotypes. e received PCV7 at 6, 10 and 14 weeks, f received PCV7 at 6, 10 and 14 weeks or at 0, 10 and 14 weeks, g Overall carriage for both groups, h 4-6 months after vaccination, i 22 months after vaccination; *these three studies were not included in the meta-analysis (in these 3 studies, both groups received PCV).P-value obtained using Fisher's exact test. doi:10.1371/journal.pone.0085001.t005