A Hierarchical End Point Analysis of the Worldwide Exploration of Renal Replacement Outcomes Collaborative in Kidney Disease (WE ROCK) Cohort

Visual Abstract

Abstract

Key Points

• Early trial of liberation from continuous renal replacement therapy was associated with improved patient-important clinical outcomes in children with preserved urine output.

• Hierarchical composite end point analysis demonstrated greater intensive care unit and hospital-free days among patients who underwent a liberation trial.

• These findings supported early assessment of readiness for liberation using urine output to guide decision making for children receiving continuous renal replacement therapy.

Background

Data to guide assessment of readiness for liberation from continuous renal replacement therapy (CRRT) in children are limited, and hierarchical composite end points have not been applied in this setting. We evaluated whether performing a trial of CRRT liberation among patients with preserved urine output is associated with improved outcomes using a hierarchical end point approach.

Methods

We conducted a retrospective cohort study of patients aged≤25 years from the multicenter Worldwide Exploration of Renal Replacement Outcomes Collaborative in Kidney Disease registry. Patients with urine output >0.5 ml/kg per hour for≥24 hours within the first 5 days after CRRT initiation were included. The primary exposure was a trial of CRRT liberation by day 5. The primary outcome was a hierarchical composite of: (1) 90-day mortality, (2) 90-day KRT dependence, (3) intensive care unit-free days, and (4) hospital-free days. Inverse probability weighting was used to balance baseline characteristics, and outcomes were compared using a win ratio analysis.

Results

Among 366 eligible patients, 192 (53%) underwent a liberation trial by day 5. These patients were older, had higher reference serum creatinine levels, and exhibited lower illness severity when compared with those who did not undergo a trial of liberation. In adjusted analyses, a liberation trial was associated with an improved hierarchical outcome (win ratio, 2.14 [95% confidence interval, 1.56 to 2.94], P < 0.001), driven primarily by greater intensive care unit- and hospital-free days.

Conclusion

Among CRRT recipients with preserved urine output, a trial of liberation by day 5 was associated with improved outcomes, supporting the early assessment of liberation readiness in patients with a urine output >0.5 ml/kg per hour.

Introduction

Continuous renal replacement therapy (CRRT) remains a cornerstone of supportive care for critically ill patients with severe AKI. Although CRRT can be lifesaving, it is an invasive and resource-intensive procedure. As such, timely and successful CRRT liberation represents an important clinical milestone with potential implications for AKI resolution and kidney recovery, intensive care unit (ICU) duration and associated costs, and long-term outcomes.^1–3 Previous findings from the Worldwide Exploration of Renal Replacement Outcomes Collaborative in Kidney Disease (WE-ROCK) cohort emphasized that successful CRRT liberation is associated with fewer major adverse kidney events at 90 days and improved ICU outcomes.^1,2 Observational studies across both pediatric and adult populations have identified a urine output >0.5 ml/kg per hour for 6–12 hours as a pragmatic indicator of readiness for CRRT liberation.^2–6 However, practice variation across pediatric centers remains substantial, reflecting uncertainty about the optimal timing and criteria for a trial of liberation.^7,8 Traditional randomized trials on pediatric CRRT liberation face challenges, highlighting the need for alternative approaches to rigorously assess outcomes.

The win ratio (WR) analytic approach leverages a hierarchical composite end point that prioritizes outcomes based on clinical and patient-important relevance. This framework may provide a more nuanced evaluation of CRRT liberation than traditional regression or time-to-first event models, which may obscure the relationships among interdependent outcomes. Outcomes such as mortality, KRT dependence, and ICU or hospital duration are inherently conditional. For example, KRT dependence cannot be assessed in patients who die before the end of the observation period. By accounting for this hierarchy, the WR enables comparisons that respect both clinical- and patient-level priorities. The WR method has been shown to enhance statistical power, clarify drivers of benefit, and better reflect real-world decision making.^9,10 It has been demonstrated to be an informative and feasible analytic framework for assessing outcomes in adult CRRT patients¹¹ and clinical trials of kidney disease progression.¹²

Recently, a multidisciplinary group of international experts in critical care nephrology proposed a modified framework for major adverse kidney events, which uses a hierarchical composite end point approach using the WR to prioritize clinically significant and patient-relevant outcomes.¹³ The WR approach, with its capacity to prioritize outcomes in a structured hierarchy, offers a compelling method for assessing the success of liberation in pediatric CRRT studies. The objective of this study was to apply a hierarchical composite end point analysis to data from the WE-ROCK cohort. Our primary aim was to determine differences in hierarchically ordered patient-important clinical outcomes between patients with preserved urine output who did and did not receive a trial of CRRT liberation by day 5. We hypothesized that those who received a trial of liberation by 5 days would have improved clinical outcomes.

Methods

Study Design and Setting

This was a retrospective cohort study using data from the WE-ROCK cohort, a multicenter international registry of patients aged 0–25 years receiving CRRT.¹⁴ Although the parent registry included older adolescents and young adults, the analytic cohort in this study was predominantly pediatric. The study protocol was reviewed and approved by the Institutional Review Boards at each participating institution and has therefore been performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments. A waiver of consent was obtained as appropriate at each site. The analysis adhered to the Strengthening the Reporting of Observational Studies in Epidemiology guidelines.

Participants

Patients included in the WE-ROCK cohort (2015–2021) were aged 0–25 years and initiated CRRT in the ICU for fluid overload and/or AKI. Patients with a history of maintenance dialysis, receiving extracorporeal membrane oxygenation, or initiated on CRRT for alternative indications were excluded from the original WE-ROCK cohort. A comprehensive methodology has been outlined previously.¹⁴ For this study, we included all patients who achieved an average urine output >0.5 ml/kg per hour for at least 24 hours within 5 days of CRRT initiation. The total volume of urine output was quantified every 24 hours for all patients, beginning at CRRT initiation. Daily urine output values were recalculated for each consecutive 24-hour period. Patients were excluded if they had missing data or died before reaching day 5 of CRRT.

Data Collection and Variable Definitions

The primary exposure was receipt of a trial of CRRT liberation by day 5, defined as any intentional trial of CRRT (successful or not). The 5-day cutoff was selected a priori based on previous analyses of the WE-ROCK cohort, which demonstrated that the median time to achieve sufficient kidney function to discontinue CRRT among those who do so is 5 days.^1,2 The primary hierarchical composite outcome consisted of four components, ordered by clinical and patient importance: (1) 90-day mortality, (2) KRT dependence at 90 days, (3) ICU-free days within the first 28 days after CRRT initiation, and (4) hospital-free days within the first 60 days after CRRT initiation. Patients who did not undergo a liberation trial by day 5 were not excluded from later liberation attempts and remained eligible to accrue ICU-free days within the 28-day follow-up period. Importantly, the inclusion of a patient representative (D. Duquette) as a co-author was done to help ensure that outcome selection was informed by patient-important priorities, emphasizing measures that align with how families and survivors might define meaningful recovery.

Baseline characteristics and clinical variables included age, admission weight, reference serum creatinine, presence of sepsis at ICU admission, admission diagnostic category, percentage fluid accumulation from ICU admission to CRRT initiation,¹⁵ and illness severity as assessed by the Pediatric Logistic Organ Dysfunction-2 (PELOD-2) score at CRRT initiation.¹⁶ The reference serum creatinine was defined as the lowest serum creatinine level within the 3 months preceding hospital admission. This included outpatient laboratory values as well as values obtained during prior inpatient encounters within the 3-month window. If no serum creatinine was available before hospital admission, we imputed an eGFR of 100 ml/min per 1.73 m² and back-calculated a reference serum creatinine level based on the bedside Schwartz equation.^17,18 These variables were selected as potential confounders because of their associations with both the decision to attempt a trial of CRRT liberation and the likelihood of recovery.^1,2

Statistical Analysis

Descriptive statistics were used to compare baseline characteristics between patients who did and did not receive a liberation trial within 5 days of CRRT initiation. Continuous variables were reported as medians with interquartile ranges (IQRs) and compared using the Wilcoxon rank-sum test. Categorical variables were summarized as frequencies and percentages and compared using chi-square or Fisher exact tests, as appropriate. To account for potential confounding, stabilized inverse probability of treatment weights (IPTWs) were generated using logistic regression to estimate propensity scores for receiving a trial of CRRT liberation by day 5, implemented with the Weightlt package in R. Covariates included in the propensity model were selected a priori based on their presumed associations with both the exposure (liberation attempt) and the outcomes of interest. After weighting, covariate balance between groups was assessed using standardized mean differences (SMDs), with SMDs <0.2 considered indicative of acceptable balance (visualized using the cobalt package). The weighted cohort was then analyzed using a hierarchical WR approach, comparing outcomes between patients who received a liberation trial by day 5 and those who did not. The WR prioritized outcomes in the following order: 90-day mortality, KRT dependence at 90 days, ICU-free days, and hospital-free days. The prespecified difference in ICU-free days and hospital-free days was set at 3 days to be considered a win. Using the win package in R, we calculated the number of pairwise comparisons, the distribution of wins, losses, and ties and estimated the WR, win odds, and net benefit with confidence intervals (CIs) and P values adjusted for the weighting.

In secondary analyses, each outcome was examined independently using IPTW-adjusted logistic regression. Binary outcomes (alive at 90 days; not requiring KRT at 90 days) were analyzed using logistic regression to estimate odds ratios (ORs) with 95% CIs. Count-based outcomes (ICU-free days and hospital-free days) were analyzed using ordinal regression, with results expressed as ORs and 95% CIs. All models were adjusted for age, reference serum creatinine, admission weight, sepsis at ICU admission, admission diagnosis, PELOD-2 score at CRRT initiation, and percentage fluid accumulation from ICU admission to CRRT initiation. All analyses were conducted using R (version 4.3.2), and statistical significance was defined as a two-sided P value < 0.05.

Results

A total of 1043 patients were enrolled in the WE-ROCK registry, of whom 407 had an average urine output >0.5 ml/kg per hour on at least 1 day within the first 5 days of CRRT initiation (Figure 1). The cohort was predominantly pediatric, with a median age of 8.18 years (IQR, 1.80–14.65). After excluding patients who died within 5 days of starting CRRT (n=33) and those with missing data (n=8), a total of 366 patients met study inclusion criteria. Of these 366 patients, 192 (52.5%) received a liberation trial, and 174 (47.5%) did not receive a liberation trial within 5 days of CRRT initiation. Among the 192 patients who received a liberation trial, 141 (73.4%) had unsuccessful liberation. Following the liberation attempt in these 141 patients, 73 (51.8%) transitioned to intermittent hemodialysis, 5 (3.5%) transitioned to peritoneal dialysis, and 63 (44.7%) resumed CRRT.

Figure 1.

Study cohort derivation for the analysis. Flow diagram outlining the derivation of the analytic cohort from the WE-ROCK study population. Of 1043 participants who received CRRT, 407 patients had a urine output >0.5 ml/kg per hour on at least 1 day by CRRT day 5. After excluding patients who died before day 5, had missing data regarding fluid accumulation, or 90-day RRT status, 366 patients comprised the final analytic cohort. Among these, 192 received a trial of CRRT liberation by day 5, and 174 did not. CRRT, continuous renal replacement therapy; n, number of patients; WE-ROCK, Worldwide Exploration of Renal Replacement Outcomes Collaborative in Kidney Disease.

As shown in Table 1, patients who received a liberation trial were older and heavier. The median reference serum creatinine for those receiving a liberation trial was higher (0.54 [0.34–0.69] mg/dl versus 0.39 [0.23–0.60] mg/dl, P < 0.001). Of the 366 patients in the analytic cohort, 161 patients (44%) had no serum creatinine value available within the 3-month preadmission window and therefore had a reference creatinine back-calculated from an assumed eGFR of 100 ml/min per 1.73 m² using the bedside Schwartz equation. Sepsis at ICU admission was less common in those who received a trial (33% versus 52%, P < 0.001), and diagnostic categories differed between groups, with a greater proportion of patients with congenital heart disease, heart failure, and/or cardiomyopathy undergoing a liberation trial. Patients who underwent a liberation trial had lower median PELOD-2 scores at CRRT initiation (5 [IQR, 3–7] versus seven,^5–9 P < 0.001) and a lower median percentage fluid accumulation at CRRT initiation (6% [0.83–11.6] versus 9% [3.1–24.3], P < 0.001).

Table 1.

Comparison of baseline variables by the receipt of a liberation trial within 5 days of continuous renal replacement therapy initiation

Variable	Overall N=366	Liberation Trial within 5 d of CRRT Initiation
		Yes n=192	No n=174
Age (yr)a	8.66 (6.86)	9.47 (6.55)	7.78 (7.10)
Sex b
Female	156 (43%)	82 (43%)	74 (43%)
Reference serum creatininec	0.48 (0.28–0.65)	0.54 (0.34–0.69)	0.39 (0.23–0.60)
Admission weightc	27.60 (12.00–52.90)	33.75 (14.75–58.80)	20.15 (10.14–45.80)
Sepsis at ICU admissionb	154 (42%)	64 (33%)	90 (52%)
Admission diagnosis b
CNS dysfunction	9 (3%)	4 (2%)	5 (3%)
Congenital heart disease	17 (5%)	10 (5%)	7 (4%)
Other	104 (28%)	67 (35%)	37 (21%)
Pain or sedation management	2 (0.5%)	1 (0.5%)	1 (0.6%)
Postsurgical or minor trauma	12 (3%)	6 (3%)	6 (3%)
Primary cardiac heart failure and/or cardiomyopathy	12 (3%)	8 (4%)	4 (2%)
Primary cardiac postsurgical	17 (5%)	10 (5%)	Seven (4%)
Respiratory failure	65 (18%)	23 (12%)	42 (24%)
Shock, infection, or major trauma	128 (35%)	63 (33%)	65 (37%)
PELOD-2 score at CRRT initiationc	6 (4,8)	5 (3–7)	Seven (5–9)
% Fluid accumulation (ICU admission to CRRT initiation)b	6.75 (1.77–17.72)	5.45 (0.83–11.63)	8.75 (3.06–24.31)
Primary comorbidities: noneb	76 (21%)	43 (22%)	33 (19%)
Primary comorbidities: cardiacb	71 (19%)	37 (19%)	34 (20%)
Primary comorbidities: endocrinologicb	21 (6%)	11 (6%)	10 (6%)
Primary comorbidities: gastrointestinalb	64 (17%)	32 (17%)	32 (18%)
Primary comorbidities: nephrologicb	35 (10%)	23 (12%)	12 (7%)
Primary comorbidities: neurologicb	47 (13%)	24 (13%)	23 (13%)
Primary comorbidities: respiratoryb	55 (15%)	27 (14%)	28 (16%)

CNS, central nervous system; CRRT, continuous renal replacement therapy; PELOD-2, pediatric logistic organ dysfunction.

^aMean (SD).

^bNo. (%).

^cMedian (Q1–Q3).

Based on these baseline differences, a WR analysis was conducted using IPTW to account for potential confounders, including age, weight, reference serum creatinine, sepsis at ICU admission, diagnostic admission category, PELOD-2 score, and percent fluid accumulation at CRRT initiation. Postweighting, balance was achieved across covariates, with all SMDs below 0.2 (Figure 2). Distributions of wins (favorable for the outcome), losses (unfavorable for the outcome), and ties are shown in Figure 3. The IPTW-weighted WR for the group that received a liberation trial within 5 days of CRRT initiation was 2.14 (1.56–2.94, P < 0.001), indicating that the group that received a liberation trial had twice the likelihood of a favorable outcome when assessed using the hierarchy of events: mortality (first), KRT dependence at 90 days (second), ICU-free days (third), and hospital-free days (fourth). Similarly, the IPTW-weighted win odds was 1.91 (95% CI, 1.46 to 2.51, P < 0.001). The net benefit, representing the absolute difference in the proportion of wins, was 0.31 (95% CI, 0.18 to 0.45, P < 0.001).

Figure 2.

Covariate balance before and after weighting for liberation trial by day 5 of CRRT. SMDs comparing baseline characteristics between patients who did and did not receive a liberation trial by day 5 of CRRT, before (red) and after (blue) weighting. The vertical dashed line at SMD=0.2 denotes the conventional threshold for acceptable balance. Weighting substantially improved covariate balance across a range of clinical and demographic factors. CNS, central nervous system; ICU, intensive care unit; IPTW, inverse probability of treatment weight; PELOD-2, pediatric logistic organ dysfunction; SMD, standardized mean difference.

Figure 3.

Hierarchical WR analysis comparing liberation trial by day 5 of CRRT versus no trial. This figure illustrates the hierarchical structure used to compare outcomes between patients who received a CRRT liberation attempt by day 5 (intervention n=192) versus those who did not (control, n=174). Each of the 192 intervention patients was compared with each of the 174 control patients, yielding 33,408 unique pairwise comparisons. For every pair, outcomes were evaluated sequentially in order of clinical priority: (1) survival at 90 days, (2) no requirement for renal replacement therapy at 90 days, (3) ICU-free days (≥3-day prespecified difference), and (4) hospital-free days (≥3-day prespecified difference). At each tier the pair was classified as a win (intervention) if the intervention patient had the more favorable outcome (e.g., at the mortality tier, the intervention patient was alive at 90 days while the control patient had died), a win (control) if the control patient had the more favorable outcome, or a tie if both patients had the same outcome at that level, in which case the comparison advanced to the next tier. Pairs that remained tied at all four tiers were classified as overall ties. Numbers within each node represent the count of pairwise comparisons resolved at that step. Of the 33,408 total comparisons, 25.6% were resolved at the mortality tier, 12.9% at renal replacement therapy dependence at 90 days, 45.0% at ICU-free days, and 2.6% at hospital-free days; 13.9% remained tied. Net win differences (intervention wins minus control wins) at each tier were +2,316, –346, +8,407, and –79, respectively. The IPTW-weighted WR combining all tiers was 2.14 (95% CI, 1.56 to 2.94, P < 0.001). CI, confidence interval; WR, win ratio.

Of the 20,546 pairwise comparisons that remained tied after the mortality and 90-day KRT-dependence tiers, 15,023 (73%) reached the prespecified ≥3-day difference at the ICU-free-days tier; the patient who received an early liberation trial had ≥3 more ICU-free days in 11,715 of these pairs (57%) and ≥3 fewer in 3308 (16%), with the remaining 5523 pairs (27%) cascading as ties to the hospital-free-days tier. At the hospital-free-days tier, only 871 of these 5523 pairs (16%) reached the ≥3-day difference (396 favoring intervention, 475 favoring control), and the remaining 4652 pairs (84%) were classified as overall ties in the WR. Together, the 3-day threshold for ICU- or hospital-free days resolved 15,894 of 33,408 pairwise comparisons (48%), making it the largest contributor to the hierarchical decision.

Table 2 shows the four outcomes of interest by the receipt of a liberation trial by 5 days individually, as estimated using logistic and ordinal regression. There was no difference between groups for being alive at 90 days or alive and not requiring KRT 90 days after CRRT initiation. Patients who received a liberation trial within 5 days of CRRT initiation had a greater number of ICU-free (median 12 [IQR, 0–19] versus zero [0–8]; OR, 0.29; 95% CI, 0.19 to 0.44, P < 0.0001) and hospital-free days (median zero [IQR, 0–7] versus zero [0–0]; OR, 0.43; 95% CI, 0.25 to 0.74, P = 0.002) when compared with those who did not receive a trial.

Table 2.

Association between continuous renal replacement therapy liberation trial by day 5 and clinical outcomes

Variable	Overalla N=366	Yesa n=192	Noa n=174	OR (95% CI, P Value)
Alive at 90 d	312 (85)	170 (89)	142 (82)	0.82 (0.41 to 1.61, 0.55)
Not requiring KRT at 90 d after CRRT initiation	281 (77)	152 (79)	129 (74)	0.82 (0.34 to 1.96, 0.65)
ICU-free d	4 (0–15)	12 (0–19)	0 (0–8)	0.29 (0.19 to 0.44, <0.001)
Hospital-free d	0 (0–1)	0 (0–7)	0 (0–0)	0.43 (0.25 to 0.74, 0.002)

CI, confidence interval; CRRT, continuous renal replacement therapy; ICU, intensive care unit; OR, odds ratio.

^aNo. (%); Median (Q1–Q3).

Discussion

In this international multicenter cohort study of critically ill pediatric patients receiving CRRT, we provide an assessment of outcomes of CRRT liberation using a novel hierarchical composite end point that prioritizes clinically significant and patient-relevant outcomes. In this study, we found that patients with preserved urine output who received a trial of liberation within 5 days of CRRT initiation were more likely to experience favorable outcomes, as assessed using a hierarchical composite end point. In adjusted analyses, early liberation attempts were associated with more than double the likelihood of improved outcomes.

These findings align with emerging adult evidence that prolonged exposure to KRT may itself delay recovery. In the recently published LIBERATE-D randomized clinical trial, adults with dialysis-requiring AKI managed with a conservative, indication-driven dialysis strategy experienced higher rates of kidney function recovery compared with those receiving conventional thrice-weekly dialysis.¹⁹ However, our study was observational, and therefore cannot determine whether earlier liberation attempts directly improve outcomes. Accordingly, these results should be viewed as hypothesis-generating and as supportive of the need for prospective studies and clinical trials to evaluate structured pediatric CRRT liberation strategies before such approaches are widely implemented. Over the past decade, numerous trials have evaluated the timing and impact of CRRT initiation, but recent efforts have shifted toward a better understanding of the timing and impact of CRRT liberation. This study shows that a trial of CRRT liberation within 5 days of initiation is associated with improved outcomes. In patients demonstrating adequate early urine output, an early trial of liberation may mitigate daily trauma, promoting kidney recovery. Although urine output is a readily available marker of readiness for CRRT liberation in both pediatric and adult populations, the decision to attempt CRRT liberation is a complex interplay of multiple variables.^2,4–6 These multiple interrelated variables, including illness severity, fluid status, hemodynamics, and clinician judgment, may introduce confounding in observational analyses. Traditional analyses using binary or time-to-event outcomes may fail to capture the interdependence and clinical prioritization of outcomes. The WR addresses this by preserving the hierarchy of outcomes and providing a more intuitive, patient-important interpretation of benefit. Its application here demonstrates the feasibility and interpretability of this approach in critically ill children.

ICU-free days was included as a patient-important measure of recovery after critical illness. Because the exposure of interest was receipt of a liberation trial by day 5, patients in the comparator group remained eligible for later CRRT discontinuation and could still accrue ICU-free days during follow-up. Nonetheless, ICU-free days is closely related to CRRT discontinuation and may partly represent a downstream consequence of liberation decisions. Importantly, our outcome framework did not treat an unsuccessful liberation attempt as a separate negative end point. In this analysis, failed liberation would only be captured indirectly through downstream measures such as ongoing KRT exposure and fewer ICU- or hospital-free days. However, from the perspective of patients and families, an unsuccessful trial may represent more than delayed recovery alone; it may also be experienced as a major setback associated with renewed uncertainty, additional procedures, and disruption in the recovery trajectory. As such, unsuccessful liberation may carry a distinct patient- and family-centered burden that is not fully reflected by the outcomes used here. Future prospective studies should therefore distinguish failed liberation from deferred liberation and evaluate whether unsuccessful attempts should be incorporated as a separate outcome, ideally alongside patient- and family-reported measures of recovery and experience.

Although the WR analysis showed that patients receiving an early liberation trial were more likely to experience an overall favorable outcome, only ICU- and hospital-free days maintained the association in the individual analyses. Reducing ICU duration is a valuable outcome in itself, associated with a decreased risk of nosocomial complications, improved neurodevelopmental trajectories in children, and lower health care costs.^20–22 Beyond these direct benefits, shorter ICU stays may also lessen the psychological burden on families, decrease exposure to invasive devices, and facilitate earlier initiation of rehabilitation and recovery services. These findings underscore the importance of incorporating patient-important measures such as ICU- and hospital-free days into future CRRT studies. Notably, the outcomes used in this study do not fully encompass the broader patient- and family-centered domains identified in prior pediatric critical care work, including physical, emotional, and cognitive functioning, health-related quality of life, family functioning, communication, and experience of care. Previous qualitative and consensus work in pediatric critical care has shown that although diagnosis, survival, and prognosis are key priorities during acute illness, functioning and quality of life become especially important during recovery.^23,24

There remains a critical need to identify objective and reproducible predictors, such as urine output, to guide CRRT liberation decisions. Although our findings suggest that early urine output may help identify patients in whom liberation readiness should be assessed, they should not be interpreted as establishing a clinically actionable liberation strategy. Rather, these data may help inform the design of future prospective studies aimed at testing standardized, evidence-based readiness assessments and liberation protocols. Currently, liberation practices vary widely across centers,^7,8 and no validated tools exist to guide decision making. These results suggest that it may be feasible to use objective clinical markers, such as time-averaged urine output, in conjunction with structured readiness assessments, to inform liberation attempts. This could mirror successful strategies employed in mechanical ventilation and sedation protocols,^25,26 thereby reducing unnecessary delays in a trial of liberation. Future work may also integrate potential novel biomarkers of AKI resolution, kidney recovery, and persistent injury, such as urinary neutrophil gelatinase-associated lipocalin and C-C motif chemokine ligand 14, which may enhance risk stratification.^27–30 Coupling biomarker-based insights with physiologic indicators like urine output could yield more precise and individualized liberation strategies.

This study has several limitations. As a retrospective analysis of observational data, causality cannot be inferred. Although IPTW was used to reduce confounding, unmeasured factors, such as trends in urine output, hemodynamic parameters, or provider decision-making heuristics, may still have biased the results. Patients who did not undergo a liberation trial by day 5 were generally younger and more severely ill. Although these factors were incorporated into adjustment, unmeasured differences in clinical trajectory, provider decision making, and center practice may still have influenced both the likelihood of liberation trial and subsequent outcomes. The exposure was defined using 24-hour average urine output; however, it is possible that shorter periods of urine output >0.5 ml/kg per hour could also be predictive of successful CRRT liberation, which was not evaluated in this analysis. In addition, we applied a single urine output across a heterogeneous pediatric age range, although emerging neonatal and younger-child data suggest that the optimal discriminatory threshold may vary by age, particularly in neonates.^31,32 Center-specific practice variation is another important limitation. Liberation strategies were not protocolized across participating centers, and differences in CRRT initiation thresholds, ultrafiltration practices, anticoagulation, and local readiness criteria may have influenced both exposure and outcomes. As such, unmeasured center-level confounding may remain despite adjustment for measured patient-level characteristics. Outcomes were assessed through a maximum of 90 days, and longer-term follow-up was not available. Finally, although the WR enables clinically meaningful comparisons, it has been recognized to have methodologic limitations. Specifically, the WR ignores ties, which may lead to overestimating treatment effects.³³ It assigns equal weight to all hierarchical components, although their relative importance may not align with the values and preferences of all patients and families.

In conclusion, among children receiving CRRT who demonstrated a preserved urine output during the first 5 days of CRRT, those who received a trial of liberation by day 5 had a significantly improved composite outcome, and in particular, greater ICU- and hospital-free days. These findings support the potential for structured readiness criteria and earlier liberation attempts to improve recovery trajectories. In this study, we present an analysis using a patient-informed composite outcome that may serve as a model for future research and clinical trials in critically ill children undergoing CRRT. Future studies should aim to validate these results and methodology prospectively, incorporate longer-term follow-up, and explore objective, generalizable tools to support individualized decisions regarding CRRT liberation in children.

Disclosures

Disclosure forms, as provided by each author, are available with the online version of the article at http://links.lww.com/CJN/C806.

Author Contributions

Conceptualization: Sean M. Bagshaw, D'Arcy Duquette, Dana Y. Fuhrman, Katja M. Gist, Stuart L. Goldstein, Matthieu Legrand, Andrew D. Shaw.

Data curation: Rashid Alobaidi, Ayse Akcan Arikan, Dana Y. Fuhrman, Katja M. Gist, Aadil Kakajiwala, Shina Menon, Nicholas J. Ollberding, Zaccaria Ricci, Indrani Sarkar, Emily See, David T. Selewski, Danielle E. Soranno.

Formal analysis: Dana Y. Fuhrman, Nicholas J. Ollberding, Indrani Sarkar.

Funding acquisition: Stuart L. Goldstein.

Investigation: Dana Y. Fuhrman, Katja M. Gist, Nicholas J. Ollberding.

Methodology: Sean M. Bagshaw, D'Arcy Duquette, Dana Y. Fuhrman, Katja M. Gist, Stuart L. Goldstein, Matthieu Legrand, Nicholas J. Ollberding, Indrani Sarkar, Andrew D. Shaw.

Project administration: Katja M. Gist.

Resources: Katja M. Gist.

Supervision: Katja M. Gist.

Validation: Katja M. Gist.

Visualization: Dana Y. Fuhrman, Katja M. Gist.

Writing – original draft: Dana Y. Fuhrman.

Writing – review & editing: Rashid Alobaidi, Ayse Akcan Arikan, Sean M. Bagshaw, D'Arcy Duquette, Dana Y. Fuhrman, Katja M. Gist, Stuart L. Goldstein, Aadil Kakajiwala, Matthieu Legrand, Shina Menon, Nicholas J. Ollberding, Zaccaria Ricci, Indrani Sarkar, Emily See, David T. Selewski, Andrew D. Shaw, Danielle E. Soranno.

Funding

D.Y. Fuhrman: Cincinnati Children's Hospital Medical Center.

Declarative Statements

This study includes clinical experimentation and received Institutional Review Board or Ethics Committee approval. The need to obtain informed patient consent was waived. This study includes clinical experimentation and complies with the Declaration of Helsinki.

Data Availability Statements

Original data generated for the study will be made available upon reasonable request to the corresponding author. Data Type: Aggregated Data. Reason for Restricted Access: The underlying dataset was not placed in a public repository because it includes human participant clinical data from a specialized cohort, for which de-identification may not fully eliminate the risk of participant re-identification. Data sharing is also subject to institutional privacy, IRB, and data-governance requirements. Therefore, data will be made available by the corresponding author upon reasonable request and, when required, institutional approval and a data use agreement. The datasets generated and/or analyzed for this study are available through the WE-ROCK registry group (werockstudy.org). Requests for data can be submitted via ancillary proposals via the aforementioned website.