Diagnostic accuracy of serum (1,3)-beta-d-glucan for neonatal invasive candidiasis: systematic review and meta-analysis

Abstract

Objectives

Neonatal invasive candidiasis (NIC) is a leading cause of infection-related morbidity and mortality in preterm neonates. Several studies have shown that (1,3)-Beta-d-glucan (BDG) was accurate in detecting invasive fungal infection in adults, but studies in neonates are scarce. The aim was to obtain summary estimates of the accuracy of BDG detection in serum for the diagnosis of NIC.

Methods

We searched Medline, Embase, Clinicaltrials.gov, and Google Scholar (inception to July 2019). We checked the reference lists of included studies, clinical guidelines, and review articles. We included studies that assessed the accuracy of BDG against a reference standard that defined groups of patients with ordinal levels of NIC probability (e.g. proven, probable, possible) and included fungal blood culture. Participants were neonates suspected of having NIC. The intervention was BDG measurement in serum (Fungitell® assay). We assessed risk of bias and applicability using QUADAS-2. We used bivariate meta-analysis to produce summary estimates of diagnostic accuracy at prespecified positivity thresholds of 80 and 120 pg/mL. This study was registered with PROSPERO (CRD42018089545).

Results

We included eight studies (465 participants). Of these, two were judged at low overall risk of bias. There was substantial variability across studies in the reference standards used. At a positivity threshold of 80 pg/mL, summary estimates of sensitivity and specificity of BDG were 89% (95% CI: 80–94%) and 60% (53–66%), respectively; summary sensitivity for detecting proven cases of NIC was 99% (93–100%). At a positivity threshold of 120 pg/mL, summary estimates of sensitivity and specificity were 81% (71–88%) and 80% (67–88%), respectively.

Conclusions

Because of high sensitivity, BDG seems promising to rule-out NIC. It might be too early to recommend its use because of the scarcity of reliable clinical data, heterogeneity in case definitions, and unstable accuracy estimates.