We enrolled patients with purulent meningitis and hospitalized at our hospital between July 31, 2014 and June 30, 2016 after approval from institutional ethics committee. The inclusion criteria included: (i) The clinical characteristics including irritability, poor feeding, respiratory distress, marbling of skin and hyper- or hypotonia in neonates or very young babies [8]; fever, seizures, fontanell bulge, neck stiffness and vomiting in infants; and headache accompanied by fever in old children [8]; and (ii) mainly polymorphic leukocytes in CSF, elevated protein level, low glucose concentration, low CSF to blood glucose ratio [8]. Patients who had blood-tinged CSF that may affect the test results were excluded. From each patients, 4-5 mL CSF specimens were collected and divided in three portions: 1.0-1.5 mL for culture, 1-1.5 mL for cytology, and 1.5-2 mL for biochemistry, pneumococcal antigen detection and PCR. Microorganism identification and antimicrobial susceptibility test were performed by using the Vitek system (Mérieux, France). Pneumococcal antigen was detected by using the BinaxNOW Streptococcus pneumonia antigen detection kit (Alere, ME, USA).

Bacterial DNA was extracted from CSF, and bacterial 16S rDNA V3-V4 region was amplified by PCR using primer pairs: 341F: CCT AYG GGR BGC ASC AG and 806R: GGA CTA CNN GGG TAT CTA AT. The PCR products with sufficient quantity were collected and purified. An OTU clustering analysis was carried out after high-throughput sequencing. The results on culture and pneumococcal antigen detection in CSF in patients with or without previous antibiotics were compared.

Statistical analysis: The collected data were compared using the chi-square test. P<0.05 was considered to be of statistical significance. Diagnostic accuracy testing was described by calculating sensitivity and specificity.

In this study, CSF samples were collected from 184 patients (36.4% neonates), aged from 1 day to 13 years and 8 months (median age of 1.92 months). Only 46 (25%) had positive culture results; with isolated bacteria being Escherichia coli (15 isolates), pneumococci (10 isolates), Streptococcus agalactiae (7 isolates), by Staphylococcus aureus (3 isolates), Enterococcus faecium (3 isolates, Candida famata was isolated in one of them), Streptococcus mitis (2 isolates), Listeria monocytogenes (2 isolates), Streptococcus sanguis (1 isolate), Enterobacter cloacae (1 isolate), Haemophilus influenzae (1 isolate), Acinetobacter baumannii (1 isolate), and C. famata (1 isolate). Fewer positive culture results were found in patients who had received previous antibiotics when compared with those who had not (P<0.001); but this difference was not seen for pneumococcal culture (P =0.08).

Pneumococcal antigen was tested positive in 34 specimens (18.5%), which included these 10 positive pneumococci cultures. No difference in the positivity rate of antigen detection was found between those with history of previous antibiotics and those without previous antibiotics (P=0.09) (Table I). 46 (25%) had positive results for culture and 10 (5.4%) were pneumococci; 34 (18.5%) were pneumococcal antigen positive. The sensitivity and specificity of pneumococcal antigen detection were 100% (95% CI: 89.4%-100%) and 86.2% (95% CI: 96.4%-99.9%), respectively (Table II).

TABLE I	Pneumococcal Antigen Detection in Cerebrospinal Fluid of Patients With or Without Previous Antibiotics (N=184)

TABLE II	Pneumococcal Antigen Detection in Cerebrospinal Fluid in Patients With Suspected Pneumococcal Meningitis 

Twenty-one CSF specimens were selected for 16s rDNA-PCR product sequence analysis. Among these were 13 positive and 8 negative for pneumococcal antigen testing. The distribution of the bacteria at the species level based on OTU is shown in Fig. 1. In one case with positive pneumococcal antigen, it was also positive for E. faecium and C. famata in the CSF culture and the abundance of pneumococcal OTU was low. As the sample with many species of OTU may be contaminated, pneumococcal infection could not be confirmed in this case.

Fig. 1 The distribution of bacterial species in 21 CSF samples by 16S rDNA-PCR high-throughput sequencing and OTU clustering.

The pneumococcal antigen test is a rapid diagnosis method in the diagnosis of pneumococcal meningitis [7,10-11]; Its advantage is the simplicity, rapidity, and usefulness in cases that have already received prior antibiotics. The sensitivity and specificity in our study were high, which was in accordance with the results from previous studies that have evaluated the diagnostic accuracy by detecting pneumococcal antigen in CSF specimens [7,9,10]. In the present study, 72.7% of all (24/33) patients with pneumococcal meningitis were missed when the diagnosis was based on the CSF culture, which was mainly attributed to the fact that a majority of these patients had received antibiotics before the sampling. The introduction of the pneumococcal antigen test signi-ficantly improved the diagnosis of pneumococcal meningitis in our study. One advantage of pneumococcal antigen test is that the pneumococcal antigen might degrade slowly; it usually persists in vivo until 7 days (90%) to 4-6 weeks (40-48%) after recovery [11,12].

The pneumococcal antigen test was confirmed by PCR as a method with high accuracy. Bacterial DNAs are still detectable by PCR within several months after being killed by antibiotics; therefore, the diagnosis of pathogen based on the pneumococcal antigen and nucleic acid detection should be suggested in conjunction with clinical manifestations. However, both positive results of pneumococcal antigen detection and nucleic acid detection only provide the evidence of pneumococcal infection, rather than ongoing infection. There was one case positive for pneumococcal antigen testing but also positive for E. faecium and C. famata in the CSF culture. As contamination by E. faecium may lead to a false positive result of the pneumococcal antigen test [6] and the abundance of penumococcal DNA was not high, the exact pathogen in this case could not be determined and was not considered as PnM.

There are certain limitations in this study. The CSF specimen used for antigen detection and PCR were the same for biochemical tests and thus had a certain risk of contamination. The 16S rDNA sequencing could not be performed in all samples because of the inadequate CSF volume, which may cause biased results. Further studies are needed to confirm our conclusion with more patients.

Acknowledgment: Dr Ying-Jie Lu, Boston Children’s Hospital, for critical reading of the manuscript.

Contributors: YPX: collecting of clinical data; PCR of 16S rDNA; test the pneumococcal antigen; analysis and interpretation of data; drafting the article; CZH: design of the study; diagnosis of meningitis; analysis on 16S rDNA OTU; revising the article critically for important intellectual content; HJW: diagnosis of meningitis; collecting of cerebrospinal fluid (CSF) and clinical data; ANS: isolation and identification of the bacteria; drug-sensitivity test. JS: acquisition of consent and collecting specimen from the patients, collection and analysis of the clinical data, revising the article.

Funding: None; Competing interest: None stated.