PT - JOURNAL ARTICLE AU - Maria E. Koster-Brouwer AU - Diana M. Verboom AU - Brendon P. Scicluna AU - Kirsten van de Groep AU - Jos F. Frencken AU - Davy Janssen AU - Rob Schuurman AU - Marcus J. Schultz AU - Tom van der Poll AU - Marc J.M. Bonten AU - Olaf L. Cremer AU - on behalf of the MARS consortium TI - Validation of a novel molecular host response assay to diagnose infection in hospitalized patients admitted to the ICU with acute respiratory failure AID - 10.1101/117853 DP - 2017 Jan 01 TA - bioRxiv PG - 117853 4099 - http://biorxiv.org/content/early/2017/03/17/117853.short 4100 - http://biorxiv.org/content/early/2017/03/17/117853.full AB - Purpose The discrimination between infectious and non-infectious causes of acute respiratory failure (ARF) in hospitalized patients admitted to the intensive care unit (ICU) is difficult. Using a novel diagnostic test measuring the expression of four RNA biomarkers in blood (SeptiCyte LAB) we aimed to distinguish between infection and inflammation in this setting.Methods We enrolled hospitalized patients with ARF requiring prompt intubation in the ICU from 2011 to 2013. We excluded patients having an established infection diagnosis or an evidently non-infectious reason for intubation. Blood samples were collected upon ICU admission. Test results were categorized into four probability bands (with higher bands indicating a higher probability of infection) and compared with the plausibility of infection as rated by post-hoc assessment using predefined definitions.Results Of 467 included patients, 373 (80%) were treated for a suspected infection at admission. Plausibility of infection was classified as ruled-out, undetermined, or confirmed in 41 (11%), 135 (36%), and 197 (53%) of these, respectively. Overall, the pre-test probability of infection was 42%. Test results correlated with the plausibility of infection (Spearman’s rho 0.332; p<0.001). After exclusion of undetermined cases, positive predictive values were 29%, 54%, and 76% for probability bands 2, 3, and 4, respectively, whereas the negative predictive value for band 1 was 76%. However, SeptiCyte LAB did not outperform CRP when comparing diagnostic discrimination (AUC 0.731; 95%CI 0.677-0.786 vs. 0.727; 95%CI 0.666-0.788).Conclusion In a setting of hospitalized patients admitted to the ICU with ARF, the diagnostic value of SeptiCyte LAB seems limited.