Distinct phenotyping of critical patients with demand-capacity imbalance initiating acute renal replacement therapy by consensus clustering

Abstract

Background

Certain patient subpopulations requiring dialysis initiation show varied survival rates and chances of ending renal replacement therapy (RRT). Consensus clustering can help identify these subgroups and their dialysis outcomes.

Methods

The study included patients who were over 18 years old with urine output above 400 ml per day and an estimated glomerular filtration rate over 15 ml/min/1.73 m². They underwent acute RRT because of systemic demand-capacity imbalance. Using consensus clustering with 33 clinical variables and urea:creatinine ratio (UCR) to the variables to investigate the catabolic demand. Endpoints were all-cause mortality and being dialysis-free at 180-day follow-up after RRT initiation.

Results

Of 946 patients (mean 63 ± 17 years and 649 men, 68.6 %) three distinct phenotypes were identified. 509 (53.8%) patients died and 364 (38.5%) patients were weaned off dialysis. Cluster 2 showed better survival (60.23% vs. 53.18% [cluster 1] vs. 45.85% [cluster 3], P < 0.01) and higher possibility to be weaned off RRT (45.24% vs. 38.44% [cluster 1] vs. 31.62% [cluster 3], P < 0.01). High UCR had increased mortality (59.16% vs. 47.75%, P < 0.01) and a lower weaning rates (33.27%; 45.72%, P < .01). UCR with the clustering phenotype improved risk stratification.

Conclusion

Among critical patients undergoing RRT due to systemic demand-capacity imbalance, more than half of the patients died. We identified distinct phenotypes in demand-capacity imbalance in a heterogeneous cohort of patients initializing RRT. Additionally, we found that pre-dialysis UCR as a novel predictor for mortality and the likelihood of being dialysis-free.