A Scoring Model with Simple Clinical Parameters to Predict Successful Discontinuation of Continuous Renal Replacement Therapy

The furosemide stress test predicts successful discontinuation of continuous renal replacement therapy in critically ill patients with acute kidney injury

PURPOSE

There is still no good method for predicting renal recovery and successful discontinuation of continuous renal replacement therapy (CRRT). This study assessed the ability of the furosemide stress test (FST) to predict successful discontinuation of CRRT.

MATERIALS AND METHODS

This prospective single-center study included patients with acute kidney injury who underwent an initial attempt at discontinuation of CRRT. Successful discontinuation was defined as alive without renal replacement therapy for 7 days after discontinuation. Furosemide 1.0 mg/kg was administered intravenously within 2 h after discontinuation of CRRT. Urine output was recorded for the next 2 h. Receiver-operating characteristic curve and logistic regression analyses were performed to determine the best discriminative variable and to identify independent risk factors.

RESULTS

Discontinuation of CRRT was successful in 30 of 55 patients. The area under the curve for prediction of successful discontinuation was significantly greater for urine output in the 2 h following the FST (0.913) than for 24-h urine output on the previous day (0.739, P = 0.003) and urine neutrophil gelatinase-associated lipocalin (0.725, P = 0.020). A 2-h urine output of 188 mL had optimal sensitivity (0.800) and specificity (0.920). Multivariate analysis showed that 2-h urine output independently predicted successful discontinuation.

CONCLUSIONS

A urine output >188 mL in the first 2 h after FST predicted successful discontinuation of CRRT.

Validation of prediction model for successful discontinuation of continuous renal replacement therapy: a multicenter cohort study

BACKGROUND

Continuous renal replacement therapy (CRRT) has become the standard modality of renal replacement therapy (RRT) in critically ill patients. However, consensus is lacking regarding the criteria for discontinuing CRRT. Here we validated the usefulness of the prediction model for successful discontinuation of CRRT in a multicenter retrospective cohort.

METHODS

One temporal cohort and four external cohorts included 1,517 patients with acute kidney injury who underwent CRRT for >2 days from 2018 to 2020. The model was composed of four variables: urine output, blood urea nitrogen, serum potassium, and mean arterial pressure. Successful discontinuation of CRRT was defined as the absence of an RRT requirement for 7 days thereafter.

RESULTS

The area under the receiver operating characteristic curve (AUROC) was 0.74 (95% confidence interval, 0.71-0.76). The probabilities of successful discontinuation were approximately 17%, 35%, and 70% in the low-score, intermediate-score, and highscore groups, respectively. The model performance was good in four cohorts (AUROC, 0.73-0.75) but poor in one cohort (AUROC, 0.56). In one cohort with poor performance, attending physicians primarily controlled CRRT prescription and discontinuation, while in the other four cohorts, nephrologists determined all important steps in CRRT operation, including screening for CRRT discontinuation.

CONCLUSION

The overall performance of our prediction model using four simple variables for successful discontinuation of CRRT was good, except for one cohort where nephrologists did not actively engage in CRRT operation. These results suggest the need for active engagement of nephrologists and protocolized management for CRRT discontinuation.

Machine Learning-Aided Decision-Making Model for the Discontinuation of Continuous Renal Replacement Therapy

INTRODUCTION

Continuous renal replacement therapy (CRRT) is a primary form of renal support for patients with acute kidney injury in an intensive care unit. Making an accurate decision of discontinuation is crucial for the prognosis of patients. Previous research has mostly focused on the univariate and multivariate analysis of factors in CRRT, without the capacity to capture the complexity of the decision-making process. The present study thus developed a dynamic, interpretable decision model for CRRT discontinuation.

METHOD

The study adopted a cohort of 1,234 adult patients admitted to an intensive care unit in the MIMIC-IV database. We used the eXtreme Gradient Boosting (XGBoost) machine learning algorithm to construct dynamic discontinuation decision models across 4 time points. SHapley Additive exPlanation (SHAP) analysis was conducted to exhibit the contributions of individual features to the model output.

RESULT

Of the 1,234 included patients with CRRT, 596 (48.3%) successfully discontinued CRRT. The dynamic prediction by the XGBoost model produced an area under the curve of 0.848, with accuracy, sensitivity, and specificity of 0.782, 0.786, and 0.776, respectively. The performance of the XGBoost model was far superior to other test models. SHAP demonstrated that the features that contributed most to the model results were the Sequential Organ Failure Assessment score, serum lactate level, and 24-h urine output.

CONCLUSION

Dynamic decision models supported by machine learning are capable of dealing with complex factors in CRRT and effectively predicting the outcome of discontinuation.

When to discontinue renal replacement therapy. what do we know?

PURPOSE OF REVIEW

Acute kidney injury is common in intensive care patients. Supportive care involves the use of renal replacement therapies as organ support. Initiation of renal replacement therapy has been the subject of much interest over the last few years with several randomised controlled studies examining the optimal time to commence treatment. In contrast to this, little evidence has been generated regarding cessation of therapy. Given that this treatment is complex, not without risk and expensive it seems timely that efforts should be expended at examining this vexing issue.

RECENT FINDINGS

Although several studies have been reported examining the successful discontinuation of renal replacement therapies all studies reported to-date are observational in nature. Conventional biochemical criteria have been used as well as physiological parameters including urine output. More recently, more novel biomarkers of renal function have been studied. Although to-date no optimal variable nor threshold for discontinuation can be established.

SUMMARY

Several variables have been described which may have a role in determining which patients may be successfully weaned from renal replacement therapy. However, few have been exposed to vigorous examination and evidence is sparse in support of any potential approach although urine output currently is the most often described. More recently novel biomarkers have also been examined but again are limited by study design and heterogeneity. Further research is clearly needed focussing on proposed variables preferably in multivariate models to improve predictive ability and successful cessation of therapy.

Different Roles of Functional and Structural Renal Markers Measured at Discontinuation of Renal Replacement Therapy for Acute Kidney Injury

INTRODUCTION

Severe acute kidney injury (AKI) requiring renal replacement therapy (RRT) has been associated with an unacceptably high mortality of 50% or more. Successful discontinuation of RRT is thought to be linked to better outcomes. Although functional and structural renal markers have been evaluated in AKI, little is known about their roles in predicting outcomes at the time of RRT discontinuation.

METHODS

In this prospective single-center cohort study, we analyzed patients who received continuous RRT (CRRT) for AKI between August 2016 and March 2018 in the intensive care unit of the University of Tokyo Hospital (Tokyo, Japan). Clinical parameters and urine samples were obtained at CRRT discontinuation. Successful CRRT discontinuation was defined as neither resuming CRRT for 48 h nor receiving intermittent hemodialysis for 7 days from the CRRT termination. Major adverse kidney events (MAKEs) were defined as death, requirement for dialysis, or a decrease in the estimated glomerular filtration rate (eGFR) of more than 25% from the baseline at day 90.

RESULTS

Of 73 patients, who received CRRT for AKI, 59 successfully discontinued CRRT and 14 could not. Kinetic eGFR, urine volume, urinary neutrophil gelatinase-associated lipocalin (NGAL), and urinary L-type fatty acid binding protein were predictive for CRRT discontinuation. Of these factors, urine volume had the highest area under the curve (AUC) 0.91 with 95% confidence interval [0.80-0.96] for successful CRRT discontinuation. For predicting MAKEs at day 90, the urinary NGAL showed the highest AUC 0.76 [0.62-0.86], whereas kinetic eGFR and urine volume failed to show statistical significance (AUC 0.49 [0.35-0.63] and AUC 0.59 [0.44-0.73], respectively).

CONCLUSIONS

Our prospective study confirmed that urine volume, a functional renal marker, predicted successful discontinuation of RRT and that urinary NGAL, a structural renal marker, predicted long-term renal outcomes. These observations suggest that the functional and structural renal makers play different roles in predicting the outcomes of severe AKI requiring RRT.

Delivering optimal renal replacement therapy to critically ill patients with acute kidney injury

Critical illness is often complicated by acute kidney injury (AKI). In patients with severe AKI, renal replacement therapy (RRT) is deployed to address metabolic dysfunction and volume excess until kidney function recovers. This review is intended to provide a comprehensive update on key aspects of RRT prescription and delivery to critically ill patients. Recently completed trials have enhanced the evidence base regarding several RRT practices, most notably the timing of RRT initiation and anticoagulation for continuous therapies. Better evidence is still needed to clarify several aspects of care including optimal targets for ultrafiltration and effective strategies for RRT weaning and discontinuation.