Community-acquired Invasive Bacterial Disease in Urban Gambia, 2005-2015: A Hospital-based Surveillance.

BACKGROUND
Invasive bacterial diseases cause significant disease and death in sub-Saharan Africa. Several are vaccine preventable, although the impact of new vaccines and vaccine policies on disease patterns in these communities is poorly understood owing to limited surveillance data.


METHODS
We conducted a hospital-based surveillance of invasive bacterial diseases in The Gambia where blood and cerebrospinal fluid (CSF) samples of hospitalized participants were processed. Three surveillance periods were defined in relation to the introduction of pneumococcal conjugate vaccines (PCVs), before (2005- 2009), during (2010-2011) and after (2012-2015) PCV introduction. We determined the prevalences of commonly isolated bacteria and compared them between the different surveillance periods.


RESULTS
A total of 14 715 blood and 1103 CSF samples were collected over 11 years; overall, 1045 clinically significant organisms were isolated from 957 patients (972 organisms [6.6%] from blood and 73 [6.6%] from CSF). The most common blood culture isolates were Streptococcus pneumoniae (24.9%), Staphylococcus aureus (22.0%), Escherichia coli (10.9%), and nontyphoidal Salmonella (10.0%). Between the pre-PCV and post-PCV eras, the prevalence of S. pneumoniae bacteremia dropped across all age groups (from 32.4% to 16.5%; odds ratio, 0.41; 95% confidence interval, .29-.58) while S. aureus increased in prevalence, becoming the most prevalent bacteria (from 16.9% to 27.2%; 1.75; 1.26-2.44). Overall, S. pneumoniae (53.4%), Neisseria meningitidis (13.7%), and Haemophilus influenzae (12.3%) were the predominant isolates from CSF. Antimicrobial resistance to common antibiotics was low.


CONCLUSIONS
Our findings demonstrate that surveillance data on the predominant pathogens associated with invasive disease is necessary to inform vaccine priorities and appropriate management of patients.

Invasive bacterial diseases (IBDs) are a leading cause of disease and death, especially among children <5 years of age [1,2]. Sub-Saharan Africa (SSA) carries a disproportionate burden of these diseases and associated deaths [3]. In addition, microbiological facilities and expertise are scarcely available in many SSA settings [4]. Limited etiological data, necessary to inform prevention strategies, show that several pathogens are associated with IBDs in SSA with diverse distribution profiles across different age groups [2,5]. The routine use of childhood vaccines, such as Haemophilus influenzae type b (Hib) conjugate vaccine and pneumococcal conjugate vaccines (PCVs) [6,7], has substantially modified the epidemiological profile of owing due to the steep reduction in the prevalence of these 2 bacteria in different age groups [7,8].
In The Gambia, with the introduction of routine infant immunization with the Hib vaccine in 1997 [9], invasive Hib disease was reduced to negligible levels [10], although a small resurgence was described more than a decade later [11][12][13]. Subsequently, Streptococcus pneumoniae, Staphylococcus aureus, E. coli, and nontyphoidal Salmonella (NTS) became the leading causes of bacteremia in the country [14]. In August 2009, a 7-valent PCV was introduced as part of the Expanded Programme on Immunisation (EPI); it was replaced by the 13-valent PCV (PCV13) in 2011. The introduction of PCV13 reduced the incidence of invasive pneumococcal disease by 55% among young children in The Gambia [15]. The impact of PCV introduction in The Gambia on bacterial diseases in older children and adults, owing to the vaccine's herd effect, has not yet been described.
The increasing threat from antimicrobial resistance remains a global challenge, resulting in longer durations of illness, mortality, and prophylaxis failure [16]. This is particularly so in SSA, where the burden is substantial, the choice of effective antimicrobials is limited, and surveillance data for invasive infections are lacking [17][18][19]. In addition to data on the prevalent bacterial pathogens associated with community and hospital-acquired invasive disease, hospital-based surveillance in resource limited settings should therefore include resistance patterns to commonly used antibiotics.
This study provides data on the main causes of postneonatal IBD in The Gambia and antibiotic susceptibility patterns over an 11-year period, between 2005 and 2015, during which PCV was introduced in the country. The data presented here are part of ongoing facility-based IBD surveillance.

Study Setting and Population
The Gambia is a subtropical country in West Africa with a single wet season from June to October. Malaria is endemic, and peak transmission occurs from July to November during the rains; however, because of scaling up of malaria control interventions, a substantial decline has been observed in recent years [20]. Malnutrition remains a problem, with the prevalence of underweight, stunting, and wasting among children <5 years old estimated at 16.4%, 25.0%, and 4.3%, respectively [21]. The prevalence of human immunodeficiency virus (HIV) among adults aged 15-49 years remains low and was estimated at 2.1% in 2015 [22].Vertical transmission is low among mothers with HIV receiving prophylaxis and treatment [23]. Infant EPI vaccine coverage is high, above 95% for the BCG vaccine and above 90% and 80%, respectively, for a single dose and 3 doses of the diphtheria-pertussis-tetanus vaccine in all regions [24]. The Hib vaccine is given at birth and at 2, 3, 4, and 16 months of age, and the PCV at 2, 3, and 4 months of age. The vaccine against Neisseria meningitidis group A (MenAfriVac) has been used in mass campaigns during outbreaks but has not been rolled out as part of the EPI schedule [10,25,26].
This hospital-based surveillance was conducted at the Clinical Services Department (CSD) of the Medical Research Council (MRC) Unit The Gambia (MRCG) at the London School of Hygiene and Tropical Medicine, situated 12 km from the capital, Banjul. The CSD has provided primary-and secondary-level care to sick individuals from the surrounding population, MRCG research study participants, ad a small number of patients referred from other clinics since the late 1950s. Approximately 50 000 patients of all ages are seen each year in the outpatient department, and 1400 are hospitalized in the 42-bed ward. It is the only health facility in The Gambia where microbiological cultures are routinely obtained in patients with suspected IBD. Blood and cerebrospinal fluid (CSF) samples are routinely collected for bacterial culture from patients with suspected sepsis and meningitis.
Patients with suspected sepsis are treated empirically with ampicillin and gentamicin, and those with suspected meningitis are treated empirically with ceftriaxone. Treatment is subsequently modified by clinical response and laboratory results. MRCG research study participants are recruited from other surrounding health facilities, with clinical samples sent to the MRCG clinical laboratories for processing using similar methods to those used for non-research study patients. Blood and CSF samples are collected only from referred patients who require admission on the MRC ward.

Microbiological Procedures
As part of this surveillance, bacterial isolates were obtained from blood using an automated blood-culture system (BACTEC 9050; Becton Dickinson), following the manufacturer's instructions for quality control and blood volume requirements. Commercially produced BD BACTEC PEDS/Plus/F culture bottles were used for specimens obtained from children (aged 1 month to 15 years) and BD BACTEC Plus Aerobic/F* and Plus Anaerobic/F* culture bottles for specimens from adults (aged >15 years), as described elsewhere [14,27]. CSF samples were processed according to World Health Organization protocol [28]. Standard microbiological procedures were performed, as described elsewhere for all pathogens [10]. In summary, pneumococcal isolates were identified using optochin disk susceptibility tests on blood agar in 5% carbon dioxide and bile solubility tests to confirm resistant isolates [27]. H. influenzae were serotyped by latex agglutination [10], and S. aureus were identified using coagulase and mannitol. For other isolates, further identification was done as appropriate for the pathogen. All normal skin flora isolates (coagulase-negative staphylococci, Bacillus species, Micrococcus spp., diphtheroids, Propionibacterium spp., and Bacillus spp. other than Bacillus anthracis) were regarded as clinically nonsignificant.
Antimicrobial sensitivity patterns were determined by means of Kirby-Bauer disk diffusion on Mueller-Hinton agar and interpreted according to the relevant Clinical Laboratory Standard Institute guidelines on antimicrobial agents [29]. Antibiotics tested as relevant to each pathogen included ampicillin, gentamicin, tetracycline, cotrimoxazole, chloramphenicol, ciprofloxacin, cefoxitin (to infer susceptibility to methicillin), and cefotaxime (BD Oxoid). Appropriate American Type Culture Collection controls were consistently used for the antibiotic susceptibility testing. Invasive bacteria isolates (blood and CSF) were stored at −70°C as part of routine microbiological surveillance. Samples were processed at the clinical microbiology laboratory, which is Good Clinical Laboratory Practice (2010) and ISO (International Organization for Standardization) 15189 (2015) accredited and submits to the external quality control assessment of the Kenya Accreditation Service in accordance with international quality systems for laboratories.

Statistical Analysis
Data were extracted from clinic and laboratory databases for research study participants and non-research study patients. All relevant invasive bacterial isolates within the study period were included in the analysis. We assumed each presentation of a patient as independent but considered patients with multiple positive cultures with the same pathogen, obtained within 4 weeks of each other, as the same episode, which was therefore only reported once. We also considered an observation to be repeated if the same patient was associated with different specimen types (blood and CSF sample).
Bacterial etiology patterns and trends were compared between surveillance periods defined by the introduction of PCVs, as follows: before PCV introduction (January 2005 to December 2009), PCV introduction and rollout (January 2010 to December 2011), and after PCV introduction (January 2012 to December 2015). Crude odds ratios for IBD, comparing post-PCV and pre-PCV periods, were obtained using logistic regression. To address confounding by age, we stratified the analyses by age. Missing values were excluded likewise, and because <1% of variables had missing values, imputation methods were not considered, and a complete case analysis was deemed adequate. The descriptive analysis for this study were carried out using Stata 13.1 software and the R Version 3.4.4 Plotly package for the graphs.

Ethical Review and Approval
Clinical samples were collected for standard clinical management, and the results were anonymized for analytical purpose. The surveillance received ethical approval from the joint MRC/ Gambia government ethics committee.  (Figure 2A). No distinct seasonal pattern was observed with other organisms.

Between
Overall and age-specific prevalences of all blood culture isolates by vaccine surveillance period re summarized in Table 1. In addition to the significantly decreased odds of S. pneumoniae bacteremia (odds ratio, 0.41; 95% confidence interval, .29-.58) and increased odds of S. aureus bacteremia (1.75; 1.26-2.44) among all age groups in the post-PCV surveillance period, we observed a 9-fold increase in the odds of bacteremia due to N. meningitidis. We observed no change in the E. coli but did note a decrease in NTS. When results were stratified by age, the most significant reduction in S. pneumoniae bacteremia was noted among children aged 2-23 months and adults ≥15 years. Likewise, the increases in S. aureus bacteremia was among children aged 2-23 months. No differences were observed between study periods for other common bacteria. The prevalence of the predominant pathogens by year is summarized in Figure 2A and shows a sustained increase on S. aureus bacteremia over the study periods along with a decline in S. pneumoniae.

Meningitis
Eighty-three (7.5%) of CSF samples cultured were positive, the majority (87.9% [73 of 83]) of which were considered clinically significant, representing 6.6% of the overall samples. Overall, the most common clinically significant isolates were S. pneumoniae  Table 2). We observed a considerably higher number of CSF samples processed in the pre-PCV period, during which there was a corresponding increase in the isolation of S. pneumoniae ( Figure 2B). S. pneumoniae was the predominant cause of meningitis in the pre-PCV period (67% [33 of 49] compared with 26% [5 of 19] after PCV introduction; P = .002). However, the prevalence of S. pneumoniae meningitis gradually declined from 2008, before the introduction of PCV. N. meningitidis increased in the post-PCV compared with the pre-PCV period (from 6.1% to 31.6%; odds ratio, 7.08; 95% confidence interval, 1.55-32.24), becoming the most prevalent bacteria in this latter period.

Antimicrobial Resistance
Antimicrobial resistance was low for the clinically relevant antimicrobial. as recommended by the World Health Organization [30]. S. pneumoniae was highly sensitive to penicillin, ampicillin, and gentamicin ( Figure 3A). Similarly, S. aureus was sensitive to cefoxitin, the surrogate for methicillin and gentamicin ( Figure 3B). Only 4 cases of invasive methicillinresistant S. aureus (1.9%) were found. In addition, resistance to gentamicin and cephalosporins was low for NTS ( Figure 3C) and E. coli ( Figure 3D). There was no increase in resistance over the study periods.

DISCUSSION
This study provides 11 years of data on the trends of the major clinically significant pathogens responsible for IBD among inpatients in urban Gambia. For the first time we have shown how these major bacteria have changed in relation to the introduction of PCV, with S. aureus and N. meningitidis replacing S. pneumoniae as the major causes of bacteremia and meningitis, respectively. The findings of the current study underscore the need for routine hospital microbiological surveillance.
The decline of S. pneumoniae as the main cause of bacteremia mirrors the corresponding rise in the predominance of S. aureus. Population-based surveillance in rural Gambia has shown a 55% decrease in the incidence of invasive pneumococcal disease among infants aged 2-23 months after the introduction of PCV [15]. Our hospital-based surveillance data support  the previous denominator-based findings and, in addition, show a decrease in the prevalence of invasive disease among nonvaccinated older children and adults. Interestingly, the decrease in S. pneumoniae prevalence, along with the increase in S. aureus, started before the introduction of PCVs. The reason for these changes remain unclear, given that no major public health interventions that could have resulted in a rapid epidemiological shift were introduced around the same time. Nevertheless, years after the introduction of the wider serotype-covering PCV13, S. pneumoniae remains an important cause of invasive disease across all age groups (second most cause of both bacteremia and meningitis). Recent data from The Gambia have shown persistence of nasopharyngeal carriage of pneumococci of vaccine serotypes along with an increased carriage of nonvaccine serotypes [31]. Because carriage is a prerequisite for invasive disease, those data, along with our results, suggest that further improvement to widen the coverage of serotypes and schedule of PCV could have an additional impact.
S. aureus was prevalent over the surveillance period but emerged as the primary cause of bacteremia in the post-PCV era being isolated in 30% of proven bacteremia cases among infants 2-23 months old. Although our study design precludes documenting an increase incidence of S. aureus bacteremia, there is a global increase in incidence with a changing epidemiology attributed to a range of factors such as vaccines, virulence, invasive procedures, antibiotic resistance, and immune suppression [32,33], underlying an urgent need for further studies and improved preventive strategies. In particular, the underlying source of S. aureus bacteremia cases in low-and middle-income countries warrants further investigation, because these may be distinct from the prevalent causes in highincome countries, such as intravenous drug use and nosocomial infections related to intravenous devices and catheters. Other factors, such as skin and soft-tissue infections, bone and joint infections, and endocarditis [34], are implicated in methicillinsusceptible S. aureus bacteremia.
E. coli and NTS were also important causes of bacteremia in our setting. Although the prevalence of E. coli has remained stable, that of NTS significantly declined after PCV introduction, with the overall number of cases relatively small (n = 97). This is consistent with previous reports from rural Gambia [35], in which it has been associated with the decline in malaria infection, because malaria transmission is relatively low in the study setting [20]. In addition, multiple host risk factors described for invasive NTS, such as HIV infection and poor sanitation, are low in prevalence, and with documented improvement in the latter [22,36]. Only 10 cases of meningitis due to N. meningitidis were observed over the entire surveillance period, and these coincided with an outbreak in the Eastern region. More than 2 decades after the Hib vaccine was introduced, with near-elimination of invasive disease in The Gambia and neighboring Senegal [10,37], sporadic cases still occur, reinforcing the need for continued surveillance.
Antimicrobial resistance was generally low, including very few methicillin-resistant S. aureus isolates compared with some SSA countries [17,38]. Antimicrobial resistance for NTS in our setting has been described elsewhere [39], and resistance to third-generation cephalosporins was low for E. coli. However, overall ciprofloxacin resistance for E. coli may warrant monitoring resistance quinolones for probable emergence of multidrug resistance [40]. Although resistance did not significantly increase over the study period, overprescription of antibiotics needs to be monitored, and stringent control measures should be in place to encourage the use of guidelines [41], because alternative antibiotics strategies are limited in resourcelimited countries such as The Gambia.
The current study had several limitations intrinsic to the surveillance design. Because our data are hospital based, the  observed changes in pathogen prevalence were dependent on sampling and may not reflect changes in disease incidence. It is a retrospective analysis, and any changes in case ascertainment, amount of sample volume collected, or health-seeking behavior in the population may have modified the overall yield of bacteria. Although the case ascertainment for patients seen at the MRCG CSD did not change over the surveillance period, there were several research studies during the period, each with different age inclusion criteria. Although this might not have changed the distribution of pathogens, it probably increased the number of isolates in some years compared with others. The linking of clinical data associated with microbiological findings is an another limitation. In addition, we could not determine what proportion of patients had received the PCV. Finally, we determined antimicrobial sensitivity using disk diffusion without final confirmation by Etest, which may have resulted  in an overestimation of the prevalence of antibiotic resistance levels. Still, results show a low prevalence of antibiotic resistance and no trends for an increase. Our long surveillance data have shown that S. aureus and N. meningitis have emerged as the leading causes of bacteremia and meningitis, respectively, in urban Gambia, while S. pneumoniae remains a leading cause of IBD, even after the introduction of PCV13. The changing epidemiology of IBD makes a compelling case for regular microbiological and antimicrobial surveillance data, which is lacking in SSA. Not only are such data necessary for healthcare workers to inform appropriate antibiotic prescribing practices, but they are also vital for prioritizing vaccine development, for emerging pathogens such as S. aureus and optimizing schedules of current vaccines. New vaccines targeting S. aureus should focus on young infants and older children, among whom the prevalence of IBD is highest.