Background: To determine the optimal time for discontinuing continuous renal replacement therapy (CRRT) by evaluating serum neutrophil gelatinase-associated lipocalin (NGAL) in critically ill patients with acute kidney injury (AKI). Methods: A prospective observational study was conducted from September 2015 to March 2018. AKI patients treated with CRRT for at least 24 h were divided into “success” and “failure” groups according to their RRT requirement within 7 days after the initial discontinuation of CRRT. The prefilter and effluent NGAL concentrations were measured to calculate the sieving coefficient (SC) of NGAL in all included subjects from 0 to 72 h. Results: In total, 110 patients were divided into success (n = 78) and failure groups (n = 32). The mean SC of NGAL during CRRT was less than 0.05. The patients in the failure group were associated with higher mortality compared with patients in the success group (37.5 vs. 12.8%, respectively, p = 0.013). There were significant differences in serum NGAL, creatinine, and urine output at discontinuation. In patients without sepsis (n = 70), serum NGAL and urine output were significant predictors of successful cessation. The area under the receiver operating characteristic to predict the successful discontinuation of CRRT was 0.88 for NGAL and 0.86 for urine output. An NGAL level of 403 ng/mL had the highest sensitivity (81%) and specificity (89%) and a urine output of 695 mL/day had the highest sensitivity (83%) and specificity (88%). However, in septic patients (n = 40), urine output but not serum NGAL (OR 0.999, p = 0.69) was a significant variable (OR 1.002, p = 0.005), with a cutoff of 796 mL/day (sensitivity 83%, specificity 88%). Conclusions: Serum NGAL was a significant factor for predicting successful CRRT discontinuation in nonseptic AKI patients. However, urine output, rather than serum NGAL, was a significant predictor in septic AKI patients.