Abstract: FR-PO100
Urinary Proteomics Profiles Classify AKI Subphenotypes
Session Information
- AKI: Diagnosis and Outcomes
October 25, 2024 | Location: Exhibit Hall, Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Acute Kidney Injury
- 102 AKI: Clinical, Outcomes, and Trials
Authors
- Stanaway, Ian Byrell, University of Washington School of Medicine, Seattle, Washington, United States
- Zelnick, Leila R., University of Washington School of Medicine, Seattle, Washington, United States
- Mabrey, Frances Linzee, University of Washington School of Medicine, Seattle, Washington, United States
- Sathe, Neha A., University of Washington School of Medicine, Seattle, Washington, United States
- Bailey, Zoie A., University of Washington School of Medicine, Seattle, Washington, United States
- Lo, Jordan J., University of Washington School of Medicine, Seattle, Washington, United States
- Himmelfarb, Jonathan, University of Washington School of Medicine, Seattle, Washington, United States
- Mikacenic, Carmen, Benaroya Research Institute at Virginia Mason, Seattle, Washington, United States
- Evans, Laura, University of Washington School of Medicine, Seattle, Washington, United States
- Wurfel, Mark M., University of Washington School of Medicine, Seattle, Washington, United States
- Morrell, Eric D., University of Washington School of Medicine, Seattle, Washington, United States
- Bhatraju, Pavan K., University of Washington School of Medicine, Seattle, Washington, United States
Background
Acute kidney injury (AKI) is a common form of organ failure in the intensive care unit. AKI has two distinct sub-phenotypes (SP1 and SP2) that differ in clinical outcomes and response to treatment, with SP2 having worse outcomes. Identification of AKI sub-phenotypes using urine biomarkers could overcome invasive blood sampling and highlight kidney specific pathology.
Methods
We prospectively enrolled 173 ICU patients admitted with a suspected respiratory infection. Of these, 86 patients had AKI and 66 were classified as SP1 and 20 as SP2 using plasma markers of endothelial dysfunction (ANGPT1, ANGPT2) and inflammation (TNFR1). We modeled risk of renal replacement therapy (RRT) adjusting for sex, age, BMI, diabetes, CKD, and COVID-19. Somalogic aptamers assessed 5,212 urine protein abundances collected on admission. We developed urinary proteomic models to predict AKI sub-phenotypes by randomly splitting the data into training (75%) and test sets (25%) and sampling 1000 bootstrap data splits. LASSO 10-fold cross validation was applied to top proteins associated with AKI sub-phenotype (FDR<0.2) in training sets with sex and age as covariates. Iterations over 1000 random data splits obtained a mean area under the curve (mAUC) in test sets. We predicted patients with SP2 among all patients, and predicted SP2 versus SP1 among patients with AKI. We combined training and test sets for final models of selected proteins. For comparison, we also tested a clinical model, which included age, sex, and plasma creatinine.
Results
SP2 patients were more likely to develop incident RRT (OR=12.5 (95% CI 1.9-82.4)). Among all patients, a 32 urinary protein model predicted SP2: mAUC=0.86 (95% CI 0.69-0.99). Among the subset of patients with AKI, a 26 urinary protein model predicted SP2 from SP1: mAUC=0.79 (95% CI 0.58-0.98). The clinical model predicted SP2 from all patients: mAUC=0.77 (95% CI 0.53-0.94) and SP2 from SP1: mAUC=0.68 (95% CI 0.46-0.89).
Conclusion
These results suggest that urinary proteomics can detect AKI SP2 with clinically useful accuracy. Urine tests for AKI sub-phenotypes may be developed for use in hospitals for bedside diagnostics and in low resource settings such as low-middle income countries, military field hospitals, rural and remote regions.
Funding
- NIDDK Support