Biomarkers in lower respiratory tract samples in the diagnosis of ventilator-associated pneumonia: a systematic review.

Abstract

Background: Ventilator-associated pneumonia (VAP) is the most common intensive care unit-acquired infection, yet its diagnosis is complicated by the lack of reliable diagnostic criteria and validated biomarkers. Due to the compartmentalisation of the immune response, host proteins in respiratory tract samples are more likely than serum proteins to accurately identify VAP. However, a reliable biomarker is still missing and it is generally agreed that >90% sensitivity and specificity are required for the introduction of a VAP biomarker into clinical routine.

Methods: A structured database search was performed to identify publications aimed at deriving or verifying human respiratory tract VAP biomarkers. The results were screened by two independent reviewers and summarised using statistical and narrative synthesis.

Results: 40 articles were identified, focusing on 29 unique biomarkers with clinical and microbiological diagnoses of VAP as the reference standard. The most frequently studied biomarker was soluble triggering receptor expressed on myeloid cell 1 (sTREM-1) (n=16), followed by various interleukins (n=7), neutrophil-related proteins (n=8) and amylase as a surrogate for microaspiration (n=4). The target accuracy of >90% specificity and sensitivity for VAP was reported in four publications on sTREM-1, one on pentraxin-3 (PTX3) and one on heparin-binding protein (HBP). Meta-analysis of sTREM-1 resulted in a sensitivity of 78% (95% CI 61-89%) and specificity of 76% (95% CI 49-91%).

Discussion: This systematic review found that no biomarker can currently be recommended for clinical use due to performance below 90% specificity or sensitivity, or insufficient data (PTX3 and HBP). Accurate clinical phenotyping into VAP subcategories may enable the discovery of VAP biomarkers with higher accuracy.