Surveillance of bacteremia and meningitis etiology in children <5 years based at Dhaka Shishu Hospital (DSH) has been ongoing since January 2001. DSH is situated in Dhaka City (population 12 million) and is the only tertiary-care pediatric hospital in Bangladesh (population 150 million, the eighth most populous country in the world). Based on physician's judgment, blood culture was performed on suspected pneumonia and sepsis cases and lumbar puncture was undertaken on all cases of suspected meningitis. Through March 2004, data were recorded only for either blood or CSF cultures growing pneumococci.

Surveillance was extended to partial national pediatric surveillance of IPD on April 2004 by incorporating seven hospitals (DSH plus six new sites, one rural and five urban) in three districts [4], [5]. From 1 January 2009 onwards the surveillance was limited to the four high performing hospitals reporting the majority of cases. Pneumococcal isolates and any surplus CSF were sent to the microbiology laboratory at DSH. The age, gender, date of illness and residential addresses were collected on all cases. Locations of the case were collected by hand-held Magellan 350 Geographical Positioning System (GPS).

Pneumococcal meningitis was defined as isolation of S. pneumoniae from CSF, or from blood when the concurrent CSF was culture negative but contained ≥10 white cells ×10⁶/L. Additional cases of pneumococcal meningitis were ascertained by detection in CSF (with ≥10 white cells ×10⁶/L) of pneumococcal capsular antigens or pneumococcal-specific DNA sequence by PCR, see below. Cases of non-meningitis IPD consisted of those patients growing pneumococci from blood culture, who did not undergo lumbar puncture or on CSF examination had <10 white cells ×10⁶/L.

Blood and CSF were cultured as described previously [4]. Pneumococcal isolates were identified using standard methods [4], and preserved in media containing 2% skimmed milk, 3% tryptone, 10% glycerol and 0.5% glucose (STGG) at −70°C. Samples with surplus CSF were archived at −70°C. Pneumococcal isolates were serotyped by the conventional capsular swelling method as previously described [4]. Presence of antimicrobials in patients with ≥10 white cells ×10⁶/L was assayed on CSF using a disc-based agar plate bio-assay as previously described [4].

Pneumococcal antigen testing of surplus CSF was performed on all culture-negative CSF samples with ≥10 white cells ×10⁶/L, when blood cultures were also negative. The pneumococcal latex agglutination test (LAT) (Wellcogen Bacterial Antigen Kit; Remel Europe Ltd, Kent, UK) was initially used, and from 2004 onwards, if negative by LAT, CSF was next tested by immunochromatographic test (ICT) (Binax NOW Streptococcus pneumoniae test, Inverness Medical Professional Diagnostics, Princeton, NJ, USA), both according to the manufacturer's instructions.

PCR detection of pneumococcal sequences targeting the pneumolysin gene (ply-PCR) as previously described [6] was also performed on surplus CSF only for the specimens collected at DSH. PCR-based serotyping for the 38 most prevalent serotypes, using 38 pairs of primers as previously described [7], was performed on all non-culture pneumococcal positive specimens containing sufficient residual CSF.

Isolates were analyzed using pulsed-field gel electrophoresis (PFGE) according to standard protocols [8], [9], [10]. Multi-locus sequence type (MLST) was performed and analyzed using eBURST as previously described [11], [12], [13].

This study on invasive pneumococcal disease surveillance was approved by the ethics review committee of the Bangladesh Institute of Child Health, Dhaka Shishu Hospital, and the International Centre for Diarrhoeal Disease Research, Bangladesh (ICDDR,B). Written informed consents were taken from parents or legal guardian of all participants for this surveillance.

Data were entered into Epi-data and analyzed using STATA 9 (StataCorp, US). An empirical odds ratio (OR) for meningitis-causing potential was calculated for each serotype with ≥10 IPD cases relative to all the serotypes with <10 IPD cases, referred to here as ‘other serotypes’ (as the reference group), following the approach for calculating serotype-specific invasive ratios [14]. Serotype-specific ORs were calculated for each serotype causing culture-positive meningitis and total IPD using the following equation: OR = (ad)/(bc), where a is the number of meningitis-causing isolates for a specific serotype, b is the number of total IPD isolates for the same specific serotype, c is the number of meningitis-causing ‘other serotype’ isolates, and d is the number of total IPD ‘other serotype’ isolates. We interpret the odds ratio as a measure of serotype-specific propensity to cause meningitis.

The sponsors of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

S. pneumoniae was isolated from 342 children <5 years old; 230 were cases of meningitis and 112 were cases of non-meningitis IPD. Of these isolates, 333 (221 meningitis and 112 non-meningitis) were serotyped; 6 were lost and 3 were non-typable and were excluded (Table 1 and Table S1, Table S2). Of the four leading serotypes, serotype-2 was the most common serotype causing meningitis during the observation period (n = 45; 20.4%; 95% CI 15%–26%) followed by serotype 12A (n = 17; 7.7%, CI 4%–11%), serotype 1 (n = 16; 7.2%, CI 4%–11%) and serotype 5 (n = 14; 6.3%, CI 3%–10%). For 6 of the 9 years of surveillance, 2003–2008, serotype 2 was the leading serotype causing meningitis, although only one case was isolated in 2001 and in 2009 no culture positive cases were identified (Figure 1).

Of the 342 pneumococcal cases, 280 were recorded as part of the partial national pediatric surveillance commencing in 2004 for which denominator data was available. 45,437 patients were blood cultured and 10,618 suspected meningitis cases had lumbar puncture performed. Only 2,490(23%) of these yielded a cell count ≥10 white cells ×10⁶/L. Of these abnormal CSF specimens, 378(15%) grew a pathogen. Of these, 174 were S. pneumoniae, which was the leading species-specific cause of childhood meningitis, accounting for 46% (174/378; 95% CI 42%–50%) of cases that yielded a pathogen. Of the 45,473 potentially septic patients, 42,947 were considered non-meningitis cases, either on clinical grounds or because the CSF yielded <10 white cells ×10⁶/L. A pathogen was grown from 3,300 (7.7%) and of these, 106 (3.2%) were S. pneumoniae.

The median age for serotype-2 cases was 3 months compared to 7 months for other serotypes (P<0.00001; Wilcoxon rank test,). The age group difference between serotype-2 (median age 4 months) and other serotypes (median age 5 months) was similar for culture negative cases of pneumococcal meningitis; (P<0.007; Wilcoxon rank test) (Figure 2). Compared to other serotypes, the OR for serotype-2 causing meningitis (n = 45 isolates) versus total IPD (meningitis and bacteremia, n = 46 isolates) was 29.6 (95% CI: 3.4–256.3), p<0.001, the highest among all serotypes analyzed (Table 2).

Of the 2,112 culture-negative CSF samples from the partial national pediatric surveillance, 1,181 contained surplus CSF for further testing. These were processed by the LAT and ICT tests, of which 386 were positive for pneumococci, 84 were positive for Haemophilus influenzae type b and 22 for Neisseria meningitidis. Of the 687 samples negative by LAT and ICT tests, 17 were found to be positive by ply-PCR. Of the 403 LAT, ICT or ply-PCR positive CSF samples, 217 had sufficient amount for PCR serotyping; 124 were successfully typed, while 93 could not be typed.

Serotype-2, similar to culture positive cases, comprised the largest sub-set of the culture-negative cases, accounting for 33(27%; 95% CI 19%–34%) cases; 22 (18%, CI 11%–24%) were serotype-14, 21 were serotype-1 (17%, CI 10%–24%) and 48 consisted of 13 other serogroups/serotypes (Table S3). Fewer serotype-2 cases were detected in 2009, mirroring the decline observed in 2009 among culture positive cases (Figure 1).The large number of pneumococcal meningitis cases detected only by non-culture methods suggests that many had prior exposure to antibiotics. Sufficient CSF was available for antimicrobial substance testing from 639 CSF samples collected as part of the partial national pediatric hospital surveillance. Antimicrobial activity was detected in 348 (54%; CI 51%–58%) of these samples; 79%in the culture-negative group (n = 160) and 25% of the culture-positive group (n = 87) (Table S4).

The geographical distribution of the cases was mapped to place of residence using GPS technology for cases identified during 2001 through 2009. Of the 78 serotype-2 meningitis cases (45 culture-positive and 33 detected by non-culture methods), 69 were traced. These were scattered widely throughout the surveillance area, with the highest concentration in Dhaka city, the main referral base for DSH and possessing the highest population density (Figure S1). Community health workers' visits to households and interviews with families of cases did not reveal any clustering within Dhaka city.

Of the 46 serotype-2 isolates cultured, 41 were available for genotyping (Figure 3). PFGE identified three types (A, B & C) differing by only 2 or 3 bands and, therefore, would be regarded as closely related [9]. By MLST, these 41 isolates belonged to a single clonal complex, CC74, consisting of 3 sequence types (ST): ST 74 (n = 23) the ancestral type, a single locus variant ST 5083 (n = 3) and a double locus variant ST 5199 (n = 15). These latter two STs are unique to this study. This serotype-2 clonal complex is genetically unrelated to other rare serotype-2 sequence types on the http://www.mlst.net database; including the archetypal whole genome sequenced historic strain D39 designated ST 595(Figure 4 and Table S5).