Abstract
INTRODUCTION: Renal replacement therapy (RRT) is commonly used in critically ill patients with acute kidney injury (AKI). However, optimal timing of RRT liberation remains controversy. This meta-analysis evaluates novel biomarkers to predict successful RRT liberation in critically ill AKI patients. METHODS: The systematic review reported following PRISMA guidelines, PubMed, Embase, and Scopus were searched up to May 2, 2025, and were screened using predefined criteria. Methodological quality was assessed using the Newcastle-Ottawa scale. Pooled ROC-AUCs with 95% CIs were calculated; heterogeneity was evaluated using I(2) statistics. RESULTS: Sixteen studies (3020 patients) involving 23 biomarkers were included. Urinary neutrophil gelatinase-associated lipocalin (uNGAL) demonstrated fair predictive performance with 4 studies (AUC 0.766, I(2) = 39.8%). When excluding a study focused on long-term outcomes, the result showed a better predictive ability with low heterogeneity (AUC 0.801, I(2) = 0%). Plasma proenkephalin A (PENK) and serum NGAL also showed potential, but quantitative synthesis was limited by study number and heterogeneity. The cut-off value also varied widely, complicating clinical translation. In addition, multivariable models combining novel biomarkers with clinical indicators have also demonstrated promising predictive potential. However, due to the limited number of studies and inconsistent conclusions, further exploration is needed. CONCLUSION: uNGAL moderately predicts short-term RRT liberation, while other biomarkers (e.g., PENK) require further validation. Standardizing definitions of successful liberation and integrating dynamic biomarker changed with clinical indicators (e.g., urine output) may enhance predictive accuracy. Further large-scale, prospective, and multicenter studies are needed to validate these findings.