Biomarkers for the Early Diagnosis of Sepsis in Burns: Systematic Review and Meta-analysis.

OBJECTIVE: The aim of this study was to evaluate the diagnostic performance of all biomarkers studied to date for the early diagnosis of sepsis in hospitalized patients with burns. BACKGROUND: Early clinical diagnosis of sepsis in burns patients is notoriously difficult due to the hypermetabolic nature of thermal injury. A considerable variety of biomarkers have been proposed as potentially useful adjuncts to assist with making a timely and accurate diagnosis. METHODS: We searched Medline, Embase, Cochrane CENTRAL, Biosis Previews, Web of Science, and Medline In-Process to February 2020. We included diagnostic studies involving burns patients that assessed biomarkers against a reference sepsis definition of positive blood cultures or a combination of microbiologically proven infection with systemic inflammation and/or organ dysfunction. Pooled measures of diagnostic accuracy were derived for each biomarker using bivariate random-effects meta-analysis. RESULTS: We included 28 studies evaluating 57 different biomarkers and incorporating 1517 participants. Procalcitonin was moderately sensitive (73%) and specific (75%) for sepsis in patients with burns. C-reactive protein was highly sensitive (86%) but poorly specific (54%). White blood cell count had poor sensitivity (47%) and moderate specificity (65%). All other biomarkers had insufficient studies to include in a meta-analysis, however brain natriuretic peptide, stroke volume index, tumor necrosis factor (TNF)-alpha, and cell-free DNA (on day 14 post-injury) showed the most promise in single studies. There was moderate to significant heterogeneity reflecting different study populations, sepsis definitions and test thresholds. CONCLUSIONS: The most widely studied biomarkers are poorly predictive for sepsis in burns patients. Brain natriuretic peptide, stroke volume index, TNF-alpha, and cell-free DNA showed promise in single studies and should be further evaluated. A standardized approach to the evaluation of diagnostic markers (including time of sampling, cut-offs, and outcomes) would be useful.