Disease Burden of Meningitis Caused by Streptococcus pneumoniae Among Under-Fives in China: A Systematic Review and Meta-analysis

This systematic review was conducted in accordance with the Cochrane Collaboration and Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (Table S1) [11]. The literature search was restricted to articles published between January 1980 and August 2022. Studies on bacterial meningitis were identified using standard search algorithms in both English (PubMed, Ovid-EMBASE, Biosis, Web of Science and Cochrane) and Chinese (CNKI, Wanfang and ViP) databases, which were constructed based on the MeSH term Meningitis and the keywords China, child, mortality, death, incidence, prevalence, morbidity, and distribution. Detailed search algorithms for each database are listed in Appendix S1 in the Supplementary Material. The reference lists of the retrieved articles were reviewed to identify possibly relevant studies.

This article is based on previously conducted studies and does not contain any new studies with human participants or animals performed by any of the authors.

Based on the recommended case definition from the World Health Organization (WHO), patients were labeled as suspected bacterial meningitis (SBM) cases if they presented a sudden onset of fever and meningeal signs including neck stiffness and altered consciousness [12]. Patients with cerebrospinal fluid (CSF) abnormalities meeting at least one of the following standards were classified as probable bacterial meningitis (PBM) cases: CSF with a turbid appearance; leukocytosis (> 100 cells/mm³); leukocytosis (10–100 cells/mm³ with an increased protein concentration of higher than 100 mg/dL or a decreased glucose concentration of lower than 40 mg/dL) [12]. Confirmed PM cases were those with CSF specimens positive for S. pneumoniae. Data from each article were extracted using a structured data collection form that included authors, publication year, province, study site, study design, etc. Adjusted incidence or mortality rates were recorded rather than raw rates if provided. Studies reporting the pathogen distribution were all based on CSF or blood specimens. For PBM cases, if either the blood culture or the CSF culture tested positive, the case was considered to be positive for the specific pathogen. It is important to note that the diagnosis of confirmed bacterial meningitis (CBM) relied not only on the results of pathogen detection but also considered the clinical symptoms manifested by the patient. Reference management and data extraction were conducted by EndNote X9.1 (Tomson, Inc., Philadelphia, USA).

The validity of the studies reporting on epidemiologic and etiologic characteristics of bacterial meningitis cases was independently assessed by two reviewers in terms of study design, length of surveillance period, reliability of diagnosis methods, and the possibility of leaving out potential cases. Based on these criteria, included studies were labeled as one of the three categories representing article quality: “A” papers, in which both reviewers judged that both criteria were met; “B” papers, in which only one reviewer judged that each criterion was met; “C” papers, in which both reviewers judged that either criterion was not met or that insufficient data were available to make a judgment. Prior to inclusion in the final dataset, a third quality assessment was performed on studies classified as “C” and all included studies. Studies remaining “C” in quality after the third assessment were discarded.

Pooled incidence rates along with 95% confidence intervals (CIs) were calculated for different types of meningitis. The test for heterogeneity was performed by calculating I², which showed the proportion of variation across studies. An I² of less than 50% implied that a fixed-effects model could be used; otherwise, a random effects model was employed. When estimating pooled CFRs, the double-arcsine method with a correction factor of 0.5 was used to handle zero events. Subgroup analyses were conducted to identify the source of heterogeneity and summarize point estimates. Poisson regression was performed to assess between-group differences in the distributions of outcomes. A P value of < 0.05 was considered significant. All analyses were conducted using R (v4.2.1, R Foundation for Statistical Computing, Vienna, Austria).

PM cases and deaths of under-fives in China were estimated by a multiplication model based on pooled results of meta-analysis and the 2020 China census data (Figure S1). Assuming that most but not all of the meningitis cases were hospitalized, the proportion of patients with acute meningitis or encephalitis (AME) who received in-hospital treatment was applied as the ratio of admitted cases of bacterial meningitis [13]. Along with the adjusted hospitalization rate assuming no vaccine use, we calculated the pooled incidence rate of PM using random-effects models. Combined with the 2020 China census data on 77,883,888 under-fives, we derived the total number of PM cases prior to the vaccines [14]. Because reported CFR values only reflect mortality in children who sought care at a health facility, we adjusted them to account for the higher CFR assumed for those not accessing care. The adjusted CFR of pathogen-specific meningitis was multiplied by the vaccine-adjusted PM cases to calculate the number of PM deaths. There is no measure of health-care seeking for children with meningitis in standardized surveys from China. Therefore, we used the proportion of children seeking care for minor illnesses (such as stomachache or diarrhea) from the China Health and Nutrition Survey (CHNS) as a proxy [15]. We applied different indicators based on the CHNS database as proxies for access to care and conducted sensitivity analysis to compare the estimated disease burdens. In order to adjust for the effect of vaccine use in China, the effective PCV7 coverage obtained by Lai et al. was applied to calculate PM incidences prior to vaccines [16]. It is worth noting that the hospitalization rate in Hong Kong reported by Pak et al. was ruled out because Hong Kong introduced PCV7 into the childhood immunization program in September 2009 [17, 18].

Six studies regarding the meningitis incidence rate were included, along with four studies on mortality rate (Tables S3 and S4). Pooled estimates of different meningitis types were calculated. The pooled annual incidence rates of PBM, CBM, and PM in China were 12.25 (95% CI: 5.48–19.02), 11.17 (95% CI: 1.34–20.99), and 2.10 (95% CI: 0.59–7.46) cases per 100,000 children, respectively (Table 1 and Fig. S6). It was estimated that 20.69 deaths per 100,000 children each year (95% CI: 12.10–29.29/100,000/year) were associated with all-cause meningitis (Fig. S6D).

The majority of the records on CFR from bacterial meningitis were published after the year 2008, when PCV7 and PPSV23 were introduced to China (Fig. S3). The pooled CFR from bacterial meningitis across studies was 3.85% (95%CI: 0.98–8.12%) in children aged between 1 month and 5 years in China (Fig. S7A). Similar age distribution patterns were observed in bacterial and PM, with the highest CFR occurring in children aged between 3 and 5 years, followed by those aged between 1 and 12 months and between 1 and 3 years (Fig. 2). Subgroup analysis by study year showed a significant decrease in estimated CFR from 1980–2005 (13.47%, 95% CI: 7.11–21.36%) to 2006–2015 (0.79%, 95% CI: 0.00–2.98%) (P < 0.001). Disparity was observed across geographic regions, and bacterial meningitis in northeast China yielded the highest CFR (5.65%, 95% CI: 0.00–22.89%), while the lowest estimate (0.23%, 95% CI: 0.00–9.09%) was from two studies conducted in southwest China. Unlike bacterial meningitis, studies on PM were all performed during the period 2006–2015 and the overall CFR was 24.59% (95% CI: 19.25–30.28%). The highest estimated CFR from PM was found in northern China (33.95%, 95% CI: 24.19–44.40%), and the lowest was found in eastern China (0.00%, 95% CI: 0.00–60.24%).

Culture or plus PCR was employed for the detection of pathogens in the included studies. Based on studies reporting the etiologic distribution of SBM, the weighted mean positive rate across pathogens was 0.17% (95% CI: 0.11–0.28%). Positive detection of S. pneumoniae was observed in all included studies on SBM, and 32 out of 8114 biological specimens were identified as S. pneumoniae positive (0.63%, 95% CI: 0.11–3.68%), ranking second in positive rate (Fig. 3 and Fig. S8A). In addition, S. pneumoniae accounted for 15.91% (95% CI: 7.15–31.74%) of all positive cases (Fig. S8B).

Thirteen studies on PBM cases yielded a pooled positive rate of 2.41% (95% CI: 1.95–2.98%). With a positive rate of 5.01% (95% CI: 3.09–8.04%), S. pneumoniae ranked second among the 31 identified pathogens (Fig. S9A). This dominant distribution pattern remained after stratifying the available records into four age groups (< 3 months, < 1 year, < 3 years, and < 5 years), except for the  < 3 months group, as none of the studies of this group reported S. pneumoniae detection (Fig. 4). The estimation of pathogen-specific proportions in CBM cases, including laboratory-positive cases of PBM, was conducted. Similarly, subgroup analysis of the proportions of CBM cases in which pathogens were positively detected showed that S. pneumoniae ranked the first in the < 3 years group and second in the < 1 year and < 5 years groups (29.86%, 95% CI: 23.87%–36.64%; 21.68%, 95% CI: 16.14%–28.48%; 21.07%, 95% CI: 12.77%–32.74%) (Fig. S4).

The pooled proportion of PM among confirmed bacterial meningitis cases was 22.05% (95% CI: 17.83–26.27%) (Fig. S9B). Subgroup analysis by age revealed variation in the proportion of PM among CBM cases and in the positive rate of S. pneumoniae among PBM cases. The proportion of PM was the highest in the < 3 years group (P < 0.001), while the positive rate was the highest in the < 1 year group (Fig. 5). Most of the studies were conducted between 2008 and 2017, namely the period between the introduction of PCV7 and PCV13 (Fig. S5). The positive rate and the proportion of S. pneumoniae displayed different distribution patterns across different periods, and the proportion reached its highest pooled estimates in the period 2006–2015 (P < 0.05) (Fig. 5). The estimated proportion of PM in CBM cases was highest in eastern China, while the positive rate was highest in multi-center sites.

There were seven studies containing data on the antimicrobial resistance of S. pneumoniae (Table S5). A total of 104 S. pneumoniae isolates were tested and 27 antibiotics were involved. Among them, S. pneumoniae displayed more than 90% resistance towards ampicillin, clindamycin, and tetracycline separately (Fig. 6). None of the S. pneumoniae isolates were resistant to vancomycin. Also, S. pneumoniae was susceptible to nine antibiotics, including moxifloxacin and linezolid.

Only two studies provided information about the serotype distribution of S. pneumoniae. Sixty-four isolates of S. pneumoniae were successfully serotyped, and serotypes 6B, 14, 19F, 19A, 23F, and 15B/C were found in both studies. The pooled serotype coverages of pneumococcal polysaccharide vaccines were 78.95% (95% CI: 69.03–88.87%) for PCV7, 91.10% (95% CI: 80.05–100.00%) for PCV13, and 100.00% (95% CI: 97.12–100.00%) for PPSV23 (Table S5).

Based on the 2020 China census data and the pooled PM incidence rate, we calculated that 1635.56 (95% CI: 459.51–5810.14) S. pneumoniae-associated meningitis cases occurred in under-fives without considering the impact of PCV in China in 2020. After adjusting the PCV impact for the PCV coverage reported by Lai et al. and a vaccine effectiveness of 73.5%, we estimated that there were 1617.16 (95% CI: 454.35–5744.78) PM cases [16, 19]. Multiplied by the adjusted CFR, we estimated that there were 548.86 (95% CI: 474.80–627.62) deaths related to PM among children younger than 5 years in China in 2020.