Diagnostic accuracy of MMBV in predicting bacterial infection in febrile children: a systematic review and meta-analysis

Background Distinguishing viral from bacterial infections in febrile children remains challenging despite available biomarkers. MMBV is a promising host-protein biosignature that computationally integrates TNF-related apoptosis-induced ligand, interferon-γ-induced protein-10, and C-reactive protein. MMBV uses predefined score thresholds for viral/non-bacterial (0–35), equivocal (36–65) and bacterial (66–100) etiologies. We conducted a systematic review and meta-analysis to evaluate MMBV diagnostic accuracy in febrile children. Methods Embase, MEDLINE, the Cochrane Library, Scopus, and CINAHL were searched until July 1, 2024. Studies evaluating MMBV diagnostic accuracy in febrile children, comparing the results of the MMBV score with a predefined reference standard established by an expert panel, were included. Two investigators independently screened, extracted data and evaluated the quality of eligible/included studies using the QUADAS-2 tool. Aggregate estimates of sensitivities and specificities were calculated with 95% confidence intervals (CI), overall and by type of infections. Results Of 486 studies identified, 14 met inclusion criteria, and nine were included in the meta-analysis. The pooled sensitivity and specificity estimates were 0.87 (95%CI 0.79–0.92, I2 55.6%) and 0.93 (95%CI 0.92–0.94, I2 0.0%), respectively. Pooled sensitivity and specificity for studies on febrile children, regardless of source, were 0.86 (95%CI 0.82–0.89, I2 0.0%) and 0.93 (95%CI 0.92–0.94, I2 0.0%). For studies on respiratory infections and fever without source, they were 0.89 (95%CI 0.79–0.95, I2 75.4%) and 0.93 (95%CI 0.91–0.94, I2 15.0%) Conclusions MMBV showed high diagnostic accuracy, and results remained consistently elevated across sub-analyses. Additional studies are needed to evaluate its effectiveness in reducing unnecessary antibiotic prescriptions in febrile children.