Molecular Phenotyping of Patients with Sepsis and Kidney Injury and Differential Response to Fluid Resuscitation

Abstract

Purpose: Previous work has identified two AKI sub-phenotypes (SP1 and SP2) characterized by differences in inflammation and endothelial dysfunction. Here we identify these sub-phenotypes using biospecimens collected in the emergency department and test for differential response to restrictive versus liberal fluid strategy in sepsis-induced hypotension in the CLOVERS trial.

Methods: We applied a previously validated 3-biomarker model using plasma angiopietin-1 and 2, and soluble tumor necrosis factor receptor-1 to classify sub-phenotypes in patients with kidney dysfunction (AKI or end-stage kidney disease [ESKD]). We also compared a de novo latent class analysis (LCA) to the 3-biomarker based sub-phenotypes. Kaplan-Meier estimates were used to test for differences in outcomes and sub-phenotype by treatment interaction.

Results: Among 1289 patients, 846 had kidney dysfunction on enrollment and the 3-variable prediction model identified 605 as SP1 and 241 as SP2. The optimal LCA model identified two sub-phenotypes with high correlation with the 3-biomarker model (Cohen's Kappa 0.8). The risk of 28 and 90-day mortality was greater in SP2 relative to SP1 independent of AKI stage and SOFA scores. Patients with SP2, characterized by more severe endothelial injury and inflammation, had a reduction in 28-day mortality with a restrictive fluid strategy versus a liberal fluid strategy (26% vs 41%), while patients with SP1 had no difference in 28-day mortality (10% vs 11%) (p-value-for-interaction = 0.03).

Conclusion: Sub-phenotypes can be identified in the emergency department that respond differently to fluid strategy in sepsis. Identification of these sub-phenotypes could inform a precision-guided therapeutic approach for patients with sepsis-induced hypotension and kidney injury.

Keywords: acute kidney injury; sepsis; vasopressors; volume.

Figure 1

Makeup of sub-phenotypes by proportion of KDIGO Stage of AKI and ESKD Provided is the proportion of participants with Kidney Disease Improving Global Outcomes (KDIGO) Stage 1, 2 or 3 AKI and end-stage kidney disease (ESKD) within each sub-phenotype. Participants with SP2 had a higher proportion of participants with more severe AKI compared to participants with SP1.

Figure 2

Kaplan-Meier survival curve in sepsis-associated AKI sub-phenotypes stratified by resuscitation treatment group Curves for each sub-phenotype are stratified by liberal or restrictive resuscitation strategy.

Figure 3

Probability belonging to SP-2 and differential treatment response to a restrictive and liberal fluid resuscitation strategy Panel A is the probability of belonging to SP2 (x-axis) and the probability of 28-day mortality (y-axis). The lines plot the estimated mortality in either a liberal (red) or restrictive (blue) resuscitation strategy with 95% CIs over a range of probabilities of assignment to SP2. The p-value tests the null hypothesis of no interaction between treatment group and the continuous probability of SP2 membership, and excludes participants who were censored prior to day 28. Plots B, C and D below demonstrate the concentrations of sTNFR-1, Ang-1 and Ang-2 concentrations at different probabilities of belonging to SP2.