Serum Metabolite Profiles Predict Outcomes in Critically Ill Patients Receiving Renal Replacement Therapy

Abstract Acute kidney injury (AKI) requiring renal replacement therapy (RRT) increases the incidence of dialysis dependence and portends high mortality in critically ill patients. At the early stage of RRT, serum metabolic biomarkers might identify patients with a high risk of mortality or permanent kidney injury from those who can recover. Serum samples from participants enrolled in the Veteran’s Affairs / National Institutes of Health Acute Renal Failure Trial Network study were collected on day 1 (n = 97) and day 8 (n = 105) of randomized RRT. The samples were further evaluated using LC/MS-based metabolic profiling. The model predicting mortality by day 8 from analyzing samples collected on day 1, was based on four metabolites with an area under the curve (AUC) of 0.641. The model most predictive of mortality by day 28 was built from the levels of 11 serum metabolites on day 8 with an AUC of 0.789. Both day 1 and day 8 biomarkers showed lower serum levels of 1-arachidonoyl-lysoPC and 1-eicosatetraenoyl-lysoPC (involved in anti-inflammatory processes) in the critically ill patients who died by day 8 or day 28. Higher levels of amino acids and amino acid metabolites in the day 8 model predicting