A clinical model to predict successful renal replacement therapy (RRT) discontinuation in patients with Acute Kidney Injury (AKI)

Abstract

Introduction: Ideal timing of Renal Replacement Therapy (RRT) discontinuation in Acute Kidney Injury (AKI) is still unknown. We aimed to study the role of creatinine-related variables in predicting RRT successful discontinuation and to propose a clinical predictive score.

Methods: In this single-centre retrospective study, we evaluated all AKI patients in whom RRT was interrupted for at least 48 hours. Patients who were still RRT-independent 7 days after initial RRT cessation were included in the "Success" group and opposed to the "Failure" group. We evaluated baseline characteristics and variables collected at the time of RRT interruption, as well as the Kinetic estimated Glomerular Filtration Rate (KeGFR), the simple variation in serum Creatinine (ΔsCr), and the incremental creatinine ratio on the first three days after RRT interruption. Multivariable analysis was performed to evaluate prediction of success. Internal validation using a simple binomial generalized regression model with Lasso estimation and 5-fold cross validation method was performed.

Results: We included 124 patients, 49 in the "Failure" group and 75 in the "Success" group. All creatinine-related variables predicted success in simple and multiple logistic regression models. The best model generated a clinical score based on the odds ratio obtained for each variable and included urine output, non-renal SOFA score, fluid balance, serum urea, serum potassium, blood pH, and the variation in sCr values after RRT discontinuation. The score presented an area under the ROC of 0.86 (95% CI 0.76‒1.00).

Conclusion: Creatinine variation between the first 2 consecutive days after RRT discontinuation might predict success in RRT discontinuation. The developed clinical score based on these variables might be a useful clinical decision tool to guide hemodialysis catheter safe removal.

Keywords: Acute kidney injury; Catheter removal; Discontinuation; Hemodialysis; Score.

Fig. 1

Patient recruitment's flow diagram.

Fig. 2

Trajectories of serum creatinine-related variables over the days following RRT discontinuation, by success. Trajectories of serum creatinine-related variables over the days following RRT discontinuation, by success. (A) sCr (mg/dL), (B) ΔsCr (mg/dL), (C) KeGFR (mL/min), (D) sCr ratio (over the first sCr after RRT discontinuation). Values presented as Median [IQR]. *p < 0.05 vs variable on first day, in the same group; # p < 0.05 vs variable on second day in the same group; & p < 0.05 between groups.

Fig. 3

Multivariable model comparison to predict success in RRT discontinuation.

Fig. 4

Training and internal validation of the proposed clinical score to predict success in RRT discontinuation. (A) Training curve ‒ AUC 0.87 (95% CI 0.74‒0.92); (B) Validation curve ‒ AUC 0.86 (95%CI 0.76‒1.00).