P3506A novel urinary biomarker predicts 1 year mortality after discharge from Intensive Care

Abstract

Abstract Rationale Tools reflecting molecular processes predicting death after discharge from intensive care unit (ICU) are currently unavailable. Objectives To construct a urinary proteomic biomarker predicting 1-year post-ICU mortality. Methods In 1081 patients, enrolled in the French and European Outcome Registry in Intensive Care Unit (NCT01367093), the urinary proteome was measured at ICU discharge using capillary electrophoresis couple with mass spectrometry along with clinical variables; circulating biomarkers (NT-proBNP, hsTnT, biologically active adrenomedullin, soluble ST2, and NGAL) and urinary albumin were available in 886 patients. Measurements and main results In the discovery sample (60/256 deaths/survivors), support vector machine modelling identified ACM150, mainly consisting of collagen fragments, yielding an AUC of 0.676 (95% CI, 0.615–0.737). In the replication sample (151/608 deaths/survivors), AUC was 0.704 (0.659–0.750). While accounting for center and clinical risk factors, the hazard ratios in all available patients were 1.27 (1.18–1.37) for ACM150 (+1 SD), 1.20 (1.16–1.33) for the Charlson score (+1 point), and ≥1.30 (P≤0.0071) for the other biomarkers (+ 1 SD). Model performance assessed by adding ACM150 to a basic model including the aforementioned covariables, the Charlson score or any other biomarker confirmed the prognostic accuracy of ACM15 with significant increases (P≤0.0038) in integrated discrimination (≥ +0.50) and net reclassification improvement (≥ +53.7) and AUC (≥ +0.037). Interactome mapping (STRING) based on 72 sequenced peptides and 25 parental proteins gravitated around collagen nodes. Conclusions ACM150 is a urinary proteomic classifier predicting 1-year post-ICU mortality over and beyond other biomarkers and reflects dysregulation of collagen turnover as underlying pathophysiological process.