Global Case Fatality of Bacterial Meningitis During an 80-Year Period: A Systematic Review and Meta-Analysis

Abstract

Importance: The impact of vaccination, antibiotics, and anti-inflammatory treatment on pathogen distribution and outcome of bacterial meningitis over the past century is uncertain.

Objective: To describe worldwide pathogen distribution and case fatality ratios of community-acquired bacterial meningitis.

Data sources: Google Scholar and MEDLINE were searched in January 2022 using the search terms bacterial meningitis and mortality.

Study selection: Included studies reported at least 10 patients with bacterial meningitis and survival status. Studies that selected participants by a specific risk factor, had a mean observation period before 1940, or had more than 10% of patients with health care-associated meningitis, tuberculous meningitis, or missing outcome were excluded.

Data extraction and synthesis: Data were extracted by 1 author and verified by a second author. The study followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses. Random-effects models stratified by age (ie, neonates, children, adults), Human Development Index (ie, low-income or high-income countries), and decade and meta-regression using the study period's year as an estimator variable were used.

Main outcome and measure: Case fatality ratios of bacterial meningitis.

Results: This review included 371 studies performed in 108 countries from January 1, 1935, to December 31, 2019, describing 157 656 episodes. Of the 33 295 episodes for which the patients' sex was reported, 13 452 (40%) occurred in females. Causative pathogens were reported in 104 598 episodes with Neisseria meningitidis in 26 344 (25%) episodes, Streptococcus pneumoniae in 26 035 (25%) episodes, Haemophilus influenzae in 22 722 (22%), other bacteria in 19 161 (18%) episodes, and unidentified pathogen in 10 336 (10%) episodes. The overall case fatality ratio was 18% (95% CI, 16%-19%), decreasing from 32% (95% CI, 24%-40%) before 1961 to 15% (95% CI, 12%-19%) after 2010. It was highest in meningitis caused by Listeria monocytogenes at 27% (95% CI, 24%-31%) and pneumococci at 24% (95% CI, 22%-26%), compared with meningitis caused by meningococci at 9% (95% CI, 8%-10%) or H influenzae at 11% (95% CI, 10%-13%). Meta-regression showed decreasing case fatality ratios overall and stratified by S pneumoniae, Escherichia coli, or Streptococcus agalactiae (P < .001).

Conclusions and relevance: In this meta-analysis with meta-regression, declining case fatality ratios of community-acquired bacterial meningitis throughout the last century were observed, but a high burden of disease remained.

Figure 1.. Causative Pathogens

Proportions of identified causative pathogens per decade (interval indicates the study periods' mean observation year; before 1961: 3684 identified episodes; 1961-1970: 4082 identified episodes; 1971-1980: 22 533 identified episodes; 1981-1990: 15 739 identified episodes; 1991-2000: 17 567 identified episodes; 2001-2010: 24 305 identified episodes; after 2010: 6352 identified episodes.).

Figure 2.. Meta-Analysis Results

Case fatality ratios (CFR) in bacterial meningitis (Forest plot with individual studies suppressed) indicating the overall pooled estimate and the estimates of the time intervals as subgroups (before 1961: 34 study periods; 1961-1970, 29 study periods; 1971-1980: 47 study periods; 1981-1990: 92 study periods; 1991-2000: 114 study periods; 2001-2010: 73 study periods; after 2010: 38 study periods) with prediction intervals.

Figure 3.. Meta-Regression Results

Case fatality ratios of patients with bacterial meningitis using a meta-regression model with the studies’ mean observation year as estimator variable, stratified according to age groups in high-income countries (A, neonates, aged younger than 2 months; B, children, aged 2 months to 16 years; C, adults, aged older than 16 years). Solid lines are regression lines, dashed lines indicate 95% CI, dotted lines indicate 95% prediction interval. Circles indicate individual study periods with the circles' sizes corresponding to the model weights.