Background: Differentiating bacterial from nonbacterial community-acquired pneumonia in children is difficult. Although several studies have evaluated serum concentrations of C-reactive protein (CRP) as a predictor of bacterial pneumonia in this patient population, the utility of this test remains unclear.
Objective: The purpose of this meta-analysis was to quantitatively define the utility of serum CRP as a predictor of bacterial pneumonia in acutely ill children.
Methods: Multiple databases were searched, bibliographies reviewed, and 2 authorities in the field were queried. Studies were included if: (1) the patient population was between 1 month and 18 years of age; (2) CRP was quantified in all subjects as part of the initial evaluation of a suspected, infectious, pulmonary process; (3) a cutoff serum CRP concentration between 30 and 60 mg/dL was used to distinguish nonbacterial from bacterial pneumonia; (4) some criteria were applied to differentiate bacterial from nonbacterial or viral pneumonia; (5) all patients were acutely ill; and (6) a chest radiograph was obtained as part of the initial evaluation. The quality of each included study was determined across 4 metrics: diagnostic criteria; study design; exclusion of chronically ill or human immunodeficiency virus infected subjects; and exclusion of patients who recently received antibiotics. Data was extracted from each article; the primary outcome measure was the odds ratio of patients with bacterial or mixed etiology pneumonia and serum CRP concentrations exceeding 30-60 mg/L. Heterogeneity among the studies was determined by Cochran's Q statistic; the methods of both Mantel and Haenszel, and DerSimonian and Laird were used to combine the study results.
Results: Eight studies fulfilled inclusion criteria. Combining all of the studies demonstrated a pooled study population of 1230 patients with the incidence of bacterial infection of 41%. Children with bacterial pneumonia were significantly more likely to have serum CRP concentrations exceeding 35-60 mg/L than children with nonbacterial infections (odds ratio = 2.58, 95% confidence interval = 1.20-5.55). Sensitivity analysis demonstrated that this difference was robust. There was significant heterogeneity among the 8 studies (Q = 37.7, P < 0.001, I2 = 81.4) that remained throughout the sensitivity analysis.
Conclusions: In children with pneumonia, serum CRP concentrations exceeding 40-60 mg/L weakly predict a bacterial etiology.