Plasma proteomics defines two reproducible subphenotypes of sepsis-associated acute kidney injury with distinct outcomes.

Sepsis is the leading cause of acute kidney injury in critically ill patients, and this complication carries a high risk of death. The biology underlying it varies between patients, which may explain why treatments have not succeeded. Here we show, using an ensemble method that groups patients by patterns across hundreds of blood proteins in three independent groups of patients, that sepsis-associated acute kidney injury comprises two reproducible subphenotypes. One subphenotype shows widespread activation of inflammation, metabolism, and oxidative stress; the other shows a quieter profile. Patients in the inflammatory subtype have higher mortality, more complications, and fewer days alive and out of the intensive care unit. A simple seven-protein test reproduces these groupings and identifies them accurately in separate cohorts. In a randomized trial, the drug ilofotase alfa appears to benefit the inflammatory subtype but not the other, suggesting that this classification could guide future treatment.