Burden of Pneumococcal Disease in Northern Togo before the Introduction of Pneumococcal Conjugate Vaccine

Background S. pneumoniae is a leading cause of meningitis morbidity and mortality in the African meningitis belt, but little is known of its contribution to the burden of pneumonia in the region. We aimed to estimate the incidence of pneumococcal disease in children and adults in northern Togo, before the introduction of pneumococcal conjugate vaccine (PCV). Methods and findings From May 1st 2010 to April 30th 2013, we systematically enrolled all hospitalized patients meeting a case definition of suspected meningitis or clinical pneumonia, residing in Tone or Cinkasse districts, northern Togo and providing informed consent. We collected clinical data and tested biological specimens according to standardized procedures, including bacteriology and PCR testing of cerebro-spinal fluid for meningitis patients and blood cultures and whole blood lytA PCR for pneumonia patients. Chest X-rays (CXR) were interpreted using the WHO methodology. We included 404 patients with meningitis (104 <5 years of age) and 1550 with pneumonia (251 <5 years) over the study period. Of these, 78 (19%) had pneumococcal meningitis (13 <5 years), 574 (37%) had radiologically-confirmed pneumonia (83 <5 years) and 73 (5%) had culture-confirmed pneumococcal pneumonia (2 <5 years). PCV13 serotypes caused 79% (54/68) of laboratory-confirmed pneumococcal meningitis and 83% (29/35) of culture-confirmed pneumococcal pneumonia. Serotype 1 predominated in meningitis (n = 33) but not in pneumonia patients (n = 1). The incidence of pneumococcal disease was 7.5 per 100,000 among children <5 years of age and 14.8 in persons 5 years of age and above in the study area. When considering CXR-confirmed and blood PCR-positive pneumonia cases as likely pneumococcal, incidence estimates increased to 43.7 and 66.0 per 100,000 in each of these age groups, respectively. Incidence was at least 3-fold higher when we restricted the analysis to the urban area immediately around the study hospitals. Conclusions Our findings highlight the important role of S. pneumoniae as a meningitis and pneumonia-causing pathogen in the African meningitis belt. Pneumococcal disease incidence in our population was substantially lower than expected from global models; we hypothesize that poor access to hospital care led us to substantially underestimate the burden of disease. Surveillance is ongoing and will enable an evaluation of PCV impact, providing novel, high quality data from the region.

Introduction Streptococcus pneumoniae (Sp or pneumococcus) has been the leading cause of pediatric bacterial meningitis and severe pneumonia since the introduction of Haemophilus influenzae type b into routine immunization programs [1]. The burden of disease is greatest in low-income countries with high infant and child mortality. In randomized controlled trials, pneumococcal conjugate vaccines (PCVs) were found to be safe and efficacious against invasive pneumococcal disease (IPD), pneumonia with WHO-defined alveolar consolidation or pleural effusion (CXR-confirmed pneumonia), and death in young children [2][3][4][5]. Further, post-introduction studies have shown that these vaccines prevent IPD and pneumonia hospitalizations in both immunized and unimmunized age groups [6][7][8]. The latter are protected through reductions in nasopharyngeal carriage among vaccinated children, who represent the age group most likely to transmit disease [9,10].
The African meningitis belt, first described by Lapeyssonie in the 1960s [11], is characterized by hyperendemic bacterial meningitis during the dry season (approximately December through May), sporadic localized epidemics and large-scale epidemics every 4-10 years [12]. Pneumococci are responsible for the majority of bacterial meningitis cases outside epidemics and contribute to the pronounced meningitis seasonality [13,14]. Pneumococcal meningitis affects all age groups, with incidence rates constant across the age range; and is highly lethal, with case fatality ratios of nearly 50% in patients treated at health facilities [15][16][17]. A variety of pneumococcal serotypes are seen in children less than 5 years of age, while serotype 1 dominates in older children and adults [13]. In contrast to high income countries, the prevalence of pneumococcal carriage has been found to be high up to adult age [13]. While pneumococcal meningitis epidemiology has been well described in this region, little is known of the burden and etiology of other invasive pneumococcal disease syndromes, particularly pneumonia. Pneumonia appears to be a climate-dependant disease [18][19][20][21] and Sp is the most common cause of pneumonia hospitalizations and deaths in the elderly in developed countries [22]. It therefore may also play an important role in both young and older adults in the meningitis belt.
In this study, we aimed to describe the epidemiology of pneumococcal pneumonia and meningitis among children and adults in northern Togo, in the African meningitis belt, before the introduction of pneumococcal conjugate vaccine. We implemented systematic hospitalbased surveillance and characterized the incidence, severity and etiology of disease based on clinical, radiological, and laboratory confirmation of cases.

Study area and calendar period
This study was conducted in the Tone and Cinkasse districts of northern Togo, a 900 km 2 area bordering Burkina Faso to the north and Ghana to the west. The study area had an estimated population of 358,568 (18% under 5 years of age, 4% under 1 year of age) (INSEED Togo) in 2010, increasing at a 3% annual rate. We implemented surveillance at all six health facilities that provide inpatient care for the two districts (Centre Hospitalier Regional in Dapaong (CHR-D), Yendube pediatric mission hospital, Centre Medical des Armees, Polyclinique-Dapaong and Win'Pang private clinic, all located in Dapaong, and the Centre Hospitalier Prefectoral in Cinkasse). Primary health care centers routinely refer severely ill patients to one of these facilities for hospitalization. We enrolled patients of all ages presenting with clinically suspected meningitis [23] or pneumonia (tachypnea for age or difficulty breathing, see case definitions below), who were admitted to hospital, resided in one of the districts and consented to participate in the study. We did not apply any exclusion criteria. Patients with repeat admissions could be enrolled in the study as many times as they were eligible and were assigned a new participant ID at each inclusion. Patients with both pneumonia and meningitis were enrolled in either the meningitis or the pneumonia group depending on their primary cause of hospitalization. The study covered the costs of hospitalization and treatment for all participants. Here, we report on data for the first three years of surveillance, from May 1 st 2010 through April 30 th 2013. Hib conjugate vaccine was introduced in Togo in 2008, while PCV and meningococcal conjugate vaccine were not in use during the study period.
Clinical data, chest X-rays and biological specimens Standardized case-report forms were used to collect demographic and clinical data from all patients, including medical history, signs and symptoms on admission, treatment, and disease outcome. During routine hospital care, pneumonia and meningitis patients are presumptively treated with ceftriaxone, unless culture identifies a drug-resistant pathogen, and oxygen is available to ventilate hypoxic patients. For our study, meningitis patients had cerebro-spinal fluid collected in two tubes, one for routine bacteriological testing and one for PCR. Pneumonia patients had a chest radiograph taken upon admission. X-rays were digitized and read by a pediatrician and a radiologist trained to apply the WHO guidelines for pediatric chest X-ray interpretation [24]. A second pediatrician arbitrated in case of discordance between readers. Pneumonia patients also had two blood draws, based on guidelines of the French Society for Microbiology. The first blood draw was done immediately at the time of enrollment, the second one 30-60 minutes later, following initiation of antibiotic treatment. Blood was inoculated into aerobic and anaerobic hemoculture bottles (for each of the two blood draws: 10 mL in FN and SA bottles for adults, 6 mL in PF bottle for children) for use in the BacT/ALERT semiautomated system (BioMerieux, 69280 Marcy l'Etoile, France), EDTA tubes for whole blood PCR (both blood draws) and serum separator tubes for C-reactive protein evaluation (first blood draw only). CSF and blood were transported from the clinical site to the laboratory by the project courier within 2 hours of collection, with the exception of CSF specimens from the CHP-Cinkasse or from overnight admissions to Yendube hospital, which were inoculated into Trans-Isolate medium and transported to CHR-D as soon as possible.

Laboratory testing
All CSF specimens and BacT/ALERT-positive blood cultures were sub-cultured on sheep blood agar and chocolate agar plates and processed according to standard bacteriology techniques. CSF also underwent cytology, Gram stain and latex agglutination testing (Pastorex, Biorad, France) at CHR-D and conventional multiplex PCR for detection of S. pneumoniae (lytA gene), H. influenzae (bexA gene) and N. meningitidis (crgA gene) at the Centre Muraz laboratory in Bobo-Dioulasso, Burkina Faso. PCR-positive CSF specimens were serotyped (40 Sp serotypes; Hi type b vs. Hi non-b) or genogrouped (serogroups A, B, C, X, Y, W) according to algorithms published elsewhere [25,26]. Pneumococcal isolates from CSF and blood had antimicrobial susceptibility testing by E-test at CHR-D, were serotyped by PCR at Centre Muraz and subsequently sent to the German National Reference Laboratory for Streptococci in Aachen, Germany for confirmatory testing and Quellung. In addition, a random sample of CSF specimens was sent to the WHO Collaborating Center for bacterial meningitis at Institut Pasteur, Paris, France for quality control testing by PCR. Meningitis cases with discrepant culture, latex and/or PCR results had their final etiology determined by the study team after considering quality control and reference laboratory findings: precedence was given to reference laboratory results; in case of discordance between reference laboratories, we examined all results together to determine the most likely etiology. Similarly, Quellung held precedent over PCR-based serotyping for pneumococcal isolates with discrepant findings.
Whole blood was tested by real-time multiplex PCR at the Fondation Merieux Emerging Pathogens Laboratory in Lyon, France. In brief, nucleic acids were extracted from 200 ul of whole blood with the QIAamp DNA blood minikit (Qiagen, Netherlands) and eluted in 100 ul of elution buffer. Five microliters of extracted DNA was then tested with a triplex rtPCR for pneumococcus (lytA), Haemophilus influenzae type b, and Staphylococcus aureus using the iQ Multiplex Powermix (Bio-Rad, USA). LytA-positive whole blood specimens were further serotyped by multiplex PCR [27]. Finally, serum C reactive-protein (CRP) was measured locally by spectrophotometer.

Case definitions
We applied the following case definitions in the analysis.
Meningitis. Suspected meningitis: clinical diagnosis of meningitis according to the admitting clinician.
Confirmed pneumococcal meningitis: suspected meningitis with S pneumoniae identified from CSF by culture or latex agglutination or PCR.
Chest X-ray (CXR) confirmed pneumonia: clinical pneumonia with lobar consolidation or pleural effusion according to standardized radiographic interpretation criteria reported above.
Pneumococcal pneumonia: clinical pneumonia with blood real-time PCR or blood culture positive for pneumococcus.

Data management and analysis
Data were entered into an EpiInfo 3.5 database and cleaned on a monthly basis. Data analysis was conducted quarterly using Stata 12.0 (StataCorp, College Station, Texas). We described clinical features of study patients including disease severity, treatment and outcome, calculating mean and standard deviation or median and inter-quartile range for continuous variables and proportions for categorical variables. We used t-tests for comparison of means and chi-square tests for proportions, with p = 0.05 as the cut-off for statistical significance. We calculated incidence and case-fatality ratios for pneumonia and meningitis endpoints and investigated the serotype distribution and antimicrobial resistance patterns of pneumococcal cases, by age group, time period (month or year) and geographic location of residence (urban vs. rural).

Ethics statement
This study was approved by the Togolese national Ethical Review Committee (Comité de Bioéthique pour la Recherche en Santé) and the French data privacy regulator (Commission National Informatique et Libertés). It was also granted a positive opinion by a French Institutional Review Board (Comité de Protection des Personnes de Saint Germain en Laye). All participants provided written informed consent before enrollment into the study.

Patient enrollment and clinical characteristics
From May 1 st 2010 to April 30 th 2013, we enrolled 404 patients with meningitis (104 age <5 years, 25.7%) and 1550 patients with pneumonia (251 age <5 years, 16.2%). An additional 6 meningitis patients and 57 pneumonia patients were eligible for enrollment in the study but did not provide informed consent to participate. We have no further data on these patients. Tables 1 and 2 show demographic and clinical characteristics of study participants.
The annual incidences of suspected, probable and confirmed Sp meningitis were 36.1, 18.8 and 7.0 per 100,000 over the study period, respectively, with monthly incidences ranging from 3.2 to 77.1, 3.2-45.0 and 0-28.9 per 100,000 and peaking during the dry season each year (Figs  1 and 2). Sp meningitis incidence varied from year to year (2.9-11.3 per 100,000) and was higher in infants (17.9/100,000) than in persons !1 year of age (6.5/100,000; p = 0.02). We identified a serotype for 68 pneumococcal meningitis cases; an additional 6 were non-typeable by PCR, 2 were negative by culture and did not have sufficient CSF remaining for serotyping  by PCR and 2 were positive by latex agglutination only (Table 3). In children <5 years of age, 12 of 12 serotyped cases were PCV13 vaccine-type (100%), 1 of which was due to serotype 1 (8.3%) whereas in persons !5 years of age, 42 of 56 cases were vaccine-type (75.0%) and 32 were serotype 1 (57.1%). The number of serotype 1 cases dropped from 23 in year 1 to 9 in year 2 and 1 in year 3 of the project, whereas the number of non-serotype 1 cases was more stable over time with 18, 13 and 10 cases, respectively. All CSF pneumococcal isolates were sensitive to ceftriaxone and penicillin; 37 (69.8%) were resistant to cotrimoxazole.  The kappa statistic for agreement between blood culture and real-time PCR for pneumococcal detection was only 0.32. Thus, pneumococcal yield from culture and PCR combined was 2 times higher overall and 8 times higher in children <5 than from culture alone. The most common serotypes among 35 Sp isolates that could be recovered for testing by PCR and Quellung were serotypes 4 (n = 8), 3 (n = 5) and 19A (n = 5); 29 (83%) were vaccinetype. Comparatively, serotyping of 124 whole blood specimens by PCR identified serotypes 3 (n = 16), 5 (n = 12) and 9V (n = 9) most frequently, with occasional mixed serotypes and 50 cases of indeterminate serotype. Serotypes identified from isolates and whole blood were concordant for all but one patient (Table 4). Pneumococci isolated from blood had very low resistance to ceftriaxone (n = 1, 1.4%) and penicillin (n = 2, 2.7%) but high resistance to cotrimoxazole (n = 53, 72.6%).

Pneumonia incidence
The incidences of clinical, CXR-confirmed and culture-confirmed Sp pneumonia were 138.3, 51.2 and 6.5 per 100,000 over the project period, respectively (Figs 3 and 4). Pneumonia peaked during the dry season in persons !5 years of age but had no clear seasonal pattern in children <5. Clinical and CXR-confirmed pneumonia incidences were highest in adults !30 years of age followed by infants <1 year of age and lowest in children 5-14 years old. In contrast, culture-confirmed Sp pneumonia incidences increased steadily with age ( Fig 5).
The observed annual incidence of clinical pneumonia was significantly higher in the town of Dapaong than in the rest of the study area, with estimates of 546.9 vs. 63.0 per 100,000 among children age <5 years and 778 vs. 46.3 per 100,000 among persons age 5+ years (p<0.01 for both comparisons). Among children age <5 years, clinical pneumonia incidence remained stable over time in Dapaong but decreased 8-fold over the study period outside Dapaong; in older children and adults, a sharp decline was observed in both the urban and rural areas, with 3.3-fold and 2-fold reductions, respectively. The ratio of urban to rural incidence was also high for CXR-confirmed pneumonia and culture-confirmed pneumococcal pneumonia (Table 2). For each of these endpoints, disease incidence decreased over time in both age groups in the urban and rural areas.

Incidence of pneumococcal disease
The incidence of laboratory-confirmed pneumococcal disease was 7.5 per 100,000 among children <5 years of age and 14.8 in persons 5 years of age and above in the study area. When considering CXR-confirmed and blood PCR-positive pneumonia cases as likely pneumococcal, incidence estimates increased to 43.7 and 66.0 per 100,000 in each age group, respectively; these numbers were 154.5 and 265.3 per 100,000 after restricting to the analysis to the town of Dapaong.

Discussion
This study describes the epidemiology of meningitis and pneumonia in all age groups in the African meningitis belt, using state-of-the-art diagnostic methods and a population-based approach. We calculated the incidence of pneumococcal disease by age group, over time and Serotypes are ranked in order of their frequency among blood culture isolates. Mixed serotypes appear twice in the table so the total number of results in the table is greater than the total number of patients with a serotype identified. In sum, results from whole blood yielded 60 VT, 2 mixed VT/NVT, 12 NVT and 50 non-typeable by PCR. for both urban and rural settings. For children under 5, our estimates of incidence ranged from 7.5 to 154.5 per 100,000, depending on the sensitivity of the case definition chosen and the geographic area considered; for persons 5 years of age and above, the range was 14.8 to  ; this is more than 20-fold higher than our maximum estimate. The low rates of pneumococcal disease we observed reflect very low rates of hospitalized pneumonia i.e. 124 per 100,000 as compared to the meta-estimate of 2,200 per 100,000 derived from 14 hospitalized pneumonia surveillance studies in Africa (incidence range: 800-10,000 per 100,000) [28]. Reported under-5 mortality in the Savanes region where our study was conducted was 107 per 1,000 live births over the study period [29], so the expected number of under-5 deaths for our study area was approximately 1,500 per year, of which an estimated 300 would be attributable to pneumonia [30]. Only 3 in-hospital pneumonia deaths were observed each year in our study, so if WHO estimates are accurate, this suggests that 99% of fatal pneumonia cases occurred out of hospital. In contrast, UNICEF estimated that in 2013-14, approximately 40% of rural and 40% of lowest wealth quintile children with pneumonia were taken to a health care provider [31]. Several explanations exist for the discrepancy including inaccurate estimates of population size, under-5 mortality, access to care for pneumonia, or the fraction of deaths attributable to pneumonia. Investigating reasons for the enormous gap between expected and observed hospital pneumonia deaths should be a public health priority for Togo, and meningitis belt countries in general, where few data such as ours exist. As with children, the incidence of adult hospitalized pneumonia in our population also appeared low compared to rural African settings, though to our knowledge evidence from the region is limited to Kenya [32,33] and the low HIV prevalence in Togo could explain its lower disease burden [29]. Because the project covered the cost of hospitalization and treatment for all enrolled patients, we consider that direct medical costs were unlikely to limit access to care in our study. However, disease rates were 8-fold higher in children <5 and 16-fold higher in persons !5 in the town of Dapaong than outside Dapaong, suggesting that distance to hospital was a major impediment to accessing care, as seen elsewhere [34]. The decline in incidence observed over the project period is another partial explanation for our low overall rates. This decline may be due to a community health worker program implemented in the region in 2012, which included regular home visits for all children <5 and amoxicillin treatment for those with respiratory infections [35]. Alternately, though we used a sensitive case definition for identification of pneumonia cases and trained surveillance staff to apply this definition systematically, it is possible that clinicians were more likely to categorize patients as malaria or malnutrition than pneumonia, leading to decreased enrollment into the study [36]. Finally, the drop in serotype 1 incidence observed among laboratory-confirmed cases over time may have contributed to the broad declines in all-cause pneumonia.
Our data on meningitis confirm earlier findings from the region, with a stable incidence of disease beyond infancy, annual peaks during the dry season, and a predominance of S. pneumoniae in the absence of a large meningococcal outbreak [15][16][17]. Pneumococcus was the leading confirmed meningitis etiology over the three-year period and occurred at similar rates as in earlier studies. Over three-quarters of pneumococcal meningitis cases were preventable by PCV13. Serotype 1 was uncommon in young children and represented >50% of cases in older children and adults. The incidence of serotype 1 Sp fluctuated over time, with a decline in Sp1 coinciding with an increase in other pathogens. Similar waves of serotype 1 meningitis have been described in Ghana [14,37]. As observed previously, over one-third of Sp meningitis cases died in hospital despite receiving appropriate treatment with ceftriaxone; this highlights the importance of disease prevention to reduce morbidity and mortality.
The prevalence of CXR-confirmed pneumonia and of pneumococcal bacteremia suggests that the pneumococcus plays an important role in pneumonia etiology across the age range. The prevalence of pneumococcal bacteremia among pneumonia patients increased with age, from <1% in children under 5 to 8% in adults !50 years of age. Despite the implementation of strict procedures for blood culture collection, transport and testing, the low blood volumes obtained from sick children and frequency of prior antimicrobial use likely affected yields [38]. The increase in Sp yield after age 15 years may be due to an increasing specificity of the clinical diagnosis of pneumonia, which is less likely to be confused with other illnesses in adults than in children. Real-time PCR on whole blood allowed identification of additional possible Sp cases: in young children, culture and PCR together detected 8 times more cases than culture alone. Recent data from the PERCH study suggest that lytA PCR may have limited specificity (personal communication, MD Knoll): in several PERCH sites, DNA-emia was seen in healthy controls, which could be explained by the transfer of genetic material from the nasopharynx to the blood in colonized children; however, these results run counter to earlier analyses showing high specificity for blood PCR [39,40]. Elsewhere, we reported findings from latent class analysis models combining laboratory and radiological data from our patients (Blake et al., accepted for publication): we showed that standardized CXR interpretation was the best approach to diagnosing Sp pneumonia and estimated that one third of pneumonia cases were attributable to pneumococcus, a proportion that increased with age from 20 to 50% and was consistent with clinical trials and observational data [5]. These results justify our use of CXR-confirmed pneumonia as a proxy for pneumococcal pneumonia in the estimation of disease incidence.
Our study provides novel information on the incidence and characteristics of pneumococcal meningitis and pneumonia among children and adults in the African meningitis belt, before the introduction of pneumococcal conjugate vaccine. Ongoing surveillance will enable a rigorous evaluation of PCV impact, including direct and indirect effects.
Hospitalier Regional-Dapaong, Hopital Pediatrique Yendube, Clinique Win'Pang, Centre Medical des Armees, Polyclinique-Dapaong and Centre Hospitalier Prefectoral-Cinkasse; Pr Ndakena, Pr Atakouma and Pr Prince-David for expert input on all scientific aspects of the study; B Chandler for chest X-ray interpretation; JN Telles, G Paranhos-Baccala, A Rajoharison, M Van der Linden, M Kheir-Taha and M Slack for molecular testing of blood and nasopharyngeal specimens and characterization of bacterial strains. This study was funded by an Investigator-Initiated Research grant from Pfizer.