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

Acute kidney injury (AKI) is common among patients in the intensive care unit (ICU), leading to complications in 40–60% of all ICU patients. Patients with AKI who need continuous renal replacement therapy (CRRT) have a particularly poor prognosis [1‒4]. Although multiple randomized controlled studies have tried to reveal the question of when to start CRRT, only a small number of observational studies have been conducted so far to answer the question of when to stop CRRT [5‒8]. Further, the guidelines from the Kidney Disease: Improving Global Outcome (KDIGO) organization similarly do not clearly state an answer to the “when to stop CRRT” question. The concept called “CRRT trauma” has been claimed, meaning the adverse effects of CRRT itself, such as continuous anticoagulation, constant restraint, excessive deprivation of medications, and essential nutrients [9, 10]. Finding reliable predictors of successful CRRT discontinuation could be a solution. We had previously stated that the kinetic estimated glomerular filtration rate (kinetic eGFR) in addition to the volume of urine as a promising early indicator for successful CRRT discontinuation based on a retrospective observational study [11]. Another issue is that there is insufficient data to forecast the non-recovery of RRT requiring AKI. Although there are several AKI biomarkers to detect kidney damage leading to early diagnosis and risk stratification in AKI, they have not much been used for prognosis prediction. Based on the results obtained in our previous retrospective study, we therefore planned a prospective study to reveal the utility of existing clinical markers, including AKI biomarkers, in forecasting the prognosis of patients undergoing CRRT at the time of CRRT discontinuation.

We prospectively enrolled all patients diagnosed with AKI and treated with CRRT at the University of Tokyo Hospital from August 2016 to March 2018. Patients who were under the age of 18, on CRRT for less than 12 h, or voluntarily discontinued intensive treatment were excluded from the analysis. Written informed consent was obtained from participants (or – if it applies – their parent/legal guardian/next of kin) to participate in the study. This study protocol was reviewed and approved by the Ethics Committee of the University of Tokyo, approval number 2810-(13) and 11239-(2). On CRRT stop, blood tests were performed and within 24 h of CRRT stop, urinary neutrophil gelatinase-associated lipocalin (NGAL), L-type fatty acid binding protein (L-FABP), sediments, and chemistries were obtained. Urinary NGAL and L-FABP were measured at the central laboratory using ARCHITECT urine NGAL chemiluminescent immunoassay (Abbott Japan, Tokyo, Japan), and Lumipulse Presto L-FABP chemiluminescent enzyme immunoassay (Fuji Revio, Tokyo, Japan), respectively. Urinary sediments were scored as previously described [12].

We reviewed the patients’ clinical and laboratory variables on the days of CRRT initiation and stop as well as their outcome variables (duration of CRRT, mortality, and renal function [eGFR and RRT dependence]) at 90 days after CRRT initiation. We also reviewed daily weight, urine volume, and serum creatinine (sCr) to calculate the kinetic eGFR. Urine volume was recorded hourly, and daily laboratory tests were performed daily at 6:00 a.m. We defined the blood test at CRRT stop as day 0, and the results of the next day’s 6:00 a.m. tests as day 1. The interval between the two tests, in hours, was designated as T.

Successful CRRT discontinuation was defined as no treatment with CRRT within 48 h nor with intermittent hemodialysis within 7 days after the last CRRT discontinuation [11]. Major adverse kidney events (MAKEs) at 90 days were defined as eGFR decline of more than 25% compared with baseline, RRT dependence, or death within 90 days after CRRT initiation.

Descriptive statistics were presented separately for the patients in the CRRT discontinued and non-discontinued groups. Continuous variables were presented as medians with interquartile range and were compared using the Wilcoxon rank-sum test. Categorical variables were compared using Fischer’s exact test. The effectiveness of each variable for predicting CRRT discontinuation was ascertained using the receiver operating characteristic (ROC) curve from a univariate logistic regression analysis. Multiple-variable combinations were examined using multivariate logistic regression analyses. Optimal cut-off values were defined using the Youden index (sensitivity + specificity −1), which is a typical ROC curve summary measure indicating the maximum potential effectiveness of a marker [15]. The ROC curves were compared using a previously reported method [16]. All statistical analyses were conducted in JMP Pro version 13 (SAS Institute Inc., Cary, NC). A p value of <0.05 was deemed to indicate statistical significance.

From August 1, 2016, to March 31, 2018, 109 patients received CRRT for AKI in the ICU of the University of Tokyo Hospital. We excluded 26 patients who chose to withdraw from intensive care, 1 patient on CRRT for less than 12 h, 2 patients not consenting to participate, and 7 patients with insufficient data for the analysis. Of 73 patients included in the statistical analysis, 14 were designated as “CRRT non-discontinued” because they failed to meet the criteria for successful CRRT discontinuation and 26 were designated as experiencing of MAKEs (Fig. 1). We could not follow-up 12 patients who transferred to another hospital before 90 days. The RRT trajectory revealed that the RRT dependency rate at 90 days after CRRT start was 6% (6/97), and all were on intermittent RRT (Fig. 2; online suppl. fig. S1; for all online suppl. material, see https://doi.org/10.1159/000532034).

At CRRT initiation, sCr and blood urea nitrogen were lower in the discontinued group than those in the non-discontinued group (online suppl. Table S1). Online supplementary Table S1 also presents patient outcomes. In the discontinued group, the CRRT duration was significantly shorter, MAKEs were fewer, and RRT dependence at 90 days was less frequent.

Parameters that could predict CRRT discontinuation were evaluated by univariate analysis for the non-discontinued and the discontinued groups (Table 1; online suppl. S2). Previously reported predictive parameters, including urine volume and kinetic eGFR, were all significantly different between the groups. Urinary NGAL, L-FABP, protein, N-acetyl-β-d-glucosaminidase, and osmolality were also significantly different. The urinary sediments score did not show significant difference. Table 3 and online suppl. S3 represents the results of the ROC curve analysis, which confirmed that urine volume is most accurate predictor (area under the curve [AUC] 0.91 [0.80–0.96]) and could not be significantly improved by combination with any other parameter (Table 3; online suppl. Fig. S2). A previously proposed discontinued index was also evaluated (online suppl. Table S4) [11] and was not found to be of added utility (AUC 0.76 [0.62–0.86]) compared with the single parameter of urine volume (online suppl. Table S4). The use of diuretics was also confirmed to have no significant effect on the utility of the predictive parameters (online suppl. Table S5). In this prospective study, urine volume was confirmed to be the most accurate predictor of RRT discontinuation.

To identify parameters predictive for MAKEs within 90 days, the likely parameters were assessed by univariate analysis (Table 2; online suppl. S6). Most predictive parameters for CRRT discontinuation failed to show significance, but urinary NGAL and L-FABP were distinct for the non-MAKEs and the MAKEs groups. Table 3 presents the outcomes of the ROC curve analysis for MAKEs prediction. Compared to urine volume, urinary NGAL was a significant predictor with AUC 0.76 (0.62–0.86) (Table 3; online suppl. Table S7; online suppl. Fig. S3). Diuretic use had no significant effect on the AUC of urinary NGAL (online suppl. Table S8).

In the present prospective observational study, we examined markers at CRRT discontinuation that could predict successful CRRT discontinuation and MAKEs within 90 days. Urine volume is reported to predict successful CRRT discontinuation as summarized in online suppl. Table S9 includes four meta-analyses [5, 6, 8, 17] and multiple other studies [18‒35]. Even though we have confirmed that urine volume was the strongest predictive parameter for CRRT discontinuation, it did not demonstrate a significant ability to predict long-term outcomes, such as MAKEs within 90 days. In contrast, urinary NGAL had a significant predictive ability for MAKEs within 90 days. AKI injury markers such as urinary NGAL and L-FABP have different properties and utilities compared with those of functional markers such as urine volume and sCr.

In the present prospective study, we measured two AKI biomarkers, urinary NGAL and L-FABP, at CRRT discontinuation. Despite extensive research on these biomarkers as early diagnostic markers of AKI, the evidence for their utility as renal recovery markers is limited. Zeng et al. [36] measured L-FABP and NGAL multiple times in postsurgical patients with AKI and reported that an AUC of 0.70 was the maximum for a prediction of AKI non-recovery. During Acute Renal Failure Trial Network study in patients receiving CRRT for AKI, Srisawat et al. [37] demonstrated that urinary NGAL measured on days 1, 7, and 14 were useful in predicting a renal recovery in 60 days with an AUC of 0.70. Our findings are consistent with the previous reports and more focused on variables measured at CRRT discontinuation. Yang et al. [38] reported that serum osteopontin and interleukin 6 at CRRT discontinuation predict 60-day renal recovery and survival. Our findings indicate that urinary NGAL at CRRT discontinuation predicts MAKEs within 90 days, which will be clinically beneficial because urinary NGAL can be used as a tool for predicting renal recovery and categorizing patients with AKI who require dialysis.

We comprehensively evaluated the potential renal recovery predictive ability of markers such as sCr, eGFR, kinetic eGFR, creatinine clearance, other urinary chemistry measurements, and urinary sediments. Despite the clinical utility of urinary sediments in AKI diagnosis [12, 39], no predictive value of the urinary sediments score for either CRRT discontinuation or MAKEs within 90 days was revealed. We were also unable to validate the utility of kinetic eGFR as an additive to urine volume in either a multivariate model or a discontinuation index score as reported in our previous study [11]. One possible reason is that the kinetic eGFR was overestimated in patients receiving CRRT as a result of a decrease in creatinine production because of sepsis [40] or long-duration CRRT [41]. Diuretic use did not affect the main findings in the present study, but the furosemide stress test is another functional test that is a candidate predictor of renal recovery and therefore requires further investigation [42, 43].

This study has several limitations. First, the generalizability of the data is unknown because this prospective cohort study was conducted at a single health center. To demonstrate reproducibility, our findings must be replicated in another cohort. Second, the study did not define any specific criteria for CRRT discontinuation or resumption; we therefore cannot rule out the possibility that the predictive parameters influenced the decision to discontinue or resume CRRT and must acknowledge the possible presence of circular arguments. To exclude this bias, prospective multicenter studies with predefined CRRT discontinuation criteria should be conducted.

In conclusion, our study demonstrated that urine volume, a functional renal marker, predicts successful discontinuation of RRT. On the other hand, urinary NGAL, a structural renal marker, predicts long-term renal outcomes. These observations indicate that the renal structural and functional markers at CRRT discontinuation have different roles in predicting the outcomes of severe AKI requiring CRRT.

We are thankful to all physicians, nurses, and medical engineers in particular who have taken care of patients involved in this study in the University of Tokyo Hospital.

Written informed consent was obtained from participants (or – if it applies – their parent/legal guardian/next of kin) to participate in the study. This study protocol was reviewed and approved by the Ethics Committee of the University of Tokyo, approval number 2810-(13) and 11239-(2).

The authors have no conflicts of interest to declare.

The authors have no funding sources to declare.

T.Y. and K.D. conceived the study design. T.Y., R.M., Y.K., Y.M., and K.Y., conducted patient care and obtained consent. T.Y. and K.D. wrote the manuscript. Y.H., E.N., M.N., and K.D. reviewed the manuscript critically.

The data that support the findings of this study are not publicly available due to the privacy of research participants but are available from the authors.