Multiplex proteomics as risk predictor of infection in patients treated with hemodialysis—A prospective multicenter study
Hemodialysis International, 2021
Glerup R., Svensson M., Jakobsen L., Fellstrøm B., Jensen J., Christensen J.
| Disease area | Application area | Sample type | Products |
|---|---|---|---|
Infectious Diseases Hematology | Patient Stratification | Plasma |  Olink Target 96 |
Abstract
Introduction
Severe infection is a major problem in hemodialysis patients. Multiplex proteomics might reveal novel insights into disease mechanisms increasing the risk of infection and might also be used as a risk prediction tool. The aims of this study were (1) to evaluate associations between 92 proteins assessed by a proximity extension assay and the development of severe infection in patients on hemodialysis and (2) to develop a risk prediction model for severe infection using prespecified clinical variables and proteomics.
Methods
Prospective, observational multicenter cohort study with 5‐year follow‐up. Patients receiving in‐center hemodialysis in five facilities in Denmark were included. The primary composite endpoint was death caused by infection, bacteremia, and infections requiring hospitalization of at least 2 days or prolonging a hospital stay.
Findings
Of 331 patients included 210 patients reached the primary endpoint during follow‐up. In adjusted Cox regression analyses, 14 plasma proteins were associated with severe infection. Correcting for multiple testing revealed only cathepsin‐L1 and interleukin‐6 significantly associated with the primary outcome. Cathepsin‐L1—hazard ratio: 1.64 (95% confidence interval [CI] 1.24–2.17) and interleukin‐6—hazard ratio: 1.16 (95% CI 1.05–1.29). Apparent C‐statistics of the risk prediction model using clinical variables was 0.605, addition of cathepsin‐L1 and interleukin‐6 to the model improved discrimination slightly: C = 0.625.
Discussion
Proteomic profiling identified cathepsin‐L1 and interleukin‐6 as markers for infectious risk in hemodialysis patients. Further studies are needed to replicate the results and to examine possible causality. The developed risk prediction models need considerable improvement before implementation in clinical practice is meaningful.