Revisiting post-ICU admission fluid balance across pediatric sepsis mortality risk strata: A secondary analyses from a prospective observational cohort study

Introduction: Post-ICU admission cumulative positive fluid balance (PFB) is associated with increased mortality among critically ill patients. We sought to test whether this risk varied across biomarker-based risk strata upon adjusting for illness severity, presence of severe acute kidney injury (AKI), and use of renal replacement therapy (CRRT) in pediatric septic shock. Design: Ongoing multi-center prospective observational cohort. Setting: Thirteen pediatric ICUs in the United States (2003–2023). Patients: Six hundred and eighty-one children with septic shock. Interventions: None. Measurements and Main Results: Cumulative percent positive fluid balance between day 1–7 (Day 1–7%PFB) was determined. Primary outcome of interest was complicated course defined as death or persistence of ≥ 2 organ dysfunctions by day 7. PERSEVERE-II biomarkers were used to assign mortality probability and categorize patients into high (n = 91), intermediate (n = 134), and low (n = 456) mortality risk strata. Cox proportional hazard regression models with adjustment for PERSEVERE-II mortality probability, presence of sepsis associated acute kidney injury (SA-AKI) on Day 3, and any use of CRRT, demonstrated that time-dependent variable Day 1–7%PFB was independently associated with increased hazard of complicated course in the cohort. Risk stratified analyses revealed that each 10% increase in Day 1–7%PFB was independently associated with increased hazard of complicated course among patients with high mortality risk strata (adj HR of 1.24 (95%CI: 1.08–1.42), p = 0.002), but not among those categorized as intermediate- or low- mortality risk. Conclusions: Our data demonstrate the independent influence of cumulative %PFB on the risk of complicated course. Contrary to our previous report, this risk was largely driven by patients categorized as having a high-mortality risk based on PERSEVERE-II biomarkers. Further research is necessary to determine whether this subset of patients may benefit from targeted deployment of restrictive fluid management or early initiation of de-escalation therapies upon resolution of shock.


Introduction
Sepsis is associated with high morbidity and mortality among children and adults.(1) Driven by a dysregulated host immune response and consequent endothelial activation, sepsis is characterized by microvascular capillary leak. (2) A large increase in systemic vascular capacitance necessitates uid resuscitation to ensure adequate preload and systemic oxygen delivery in the initial phase of illness. However, uid shifts over the course of disease result in extravascular uid accumulation, consequent hypoxemia, and organ dysfunction(s). (3) Persistence of a positive uid balance (PFB) has been independently associated with poor clinical outcomes among patients with sepsis (4)(5)(6)(7)(8)(9). Complicating matters further, patients may develop sepsis associated acute kidney injury (SA-AKI) and oliguria. A subset of patients may require continuous renal replacement therapy (CRRT). Both presence of SA-AKI and CRRT use may serve to confound the association between cumulative PFB and clinical outcomes. Recent research has emphasized conservative uid management, de-escalation, and deresuscitation where appropriate to negate the detrimental effects of positive uid balance among critically ill patients. (10) It is suggested that such approaches may improve liberation of patients from intensive care units (ICU). A key limitation is that few studies have explicitly considered biological heterogeneity among critically ill patients. It is likely that subsets of patients may bene t from tailored uid management approaches. (11)(12)(13) Our group has previously utilized Pediatric Sepsis Biomarker Risk Model (PERSEVERE) -a prospectively validated prognostic enrichment tool (14) re ective of the host in ammatory response-to the test the association between post-ICU admission cumulative PFB across mortality risk strata in pediatric septic shock patients. (15) We reported that only in patients with low-mortality risk was cumulative PFB associated with an increased odds of complicated course-a composite of death and multiple organ dysfunctions. Limitations of our prior study included: 1) reporting of univariate analyses that did not account for potential confounders, 2) the lack of consideration of the competing in uence of early deaths on ICU length of stay, and consequently on the cumulative PFB among the most critically ill patients, and 3) the relatively small number of patients belonged to high-and intermediate-PERSEVERE mortality risk strata.
In this study, we sought to test the independent in uence of post-ICU admission PFB on the risk of complicated course in a large cohort of children with septic shock. We further conducted risk-strati ed analyses to test differences in this association across PERSEVERE-II mortality risk strata.

Methods
Study design and patient selection: The study protocol was approved by Institutional Review Boards (IRBs) of participating institutions (Cincinnati Children's Hospital IR, Genomic Analysis of Pediatric Systemic In ammatory Syndrome, IRB ID: 2008 − 0558, Continuous Review) as well as all participating institutions. Informed consent was obtained from parent or guardian of patients. All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional review boards of participating institutions and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.
Our ongoing multi-institutional prospective observational cohort study of pediatric septic shock has been extensively detailed previously. (14,16,17) Inclusion criteria for this study was meeting pediatric-speci c consensus criteria for septic shock.(18) Those with (1) missing admission height to estimate baseline serum creatinine or weight to estimate percent PFB per kilogram body weight, (2) those with missing uid balance details on day 1, and those without (3) PERSEVERE-II biomarker data were excluded. Baseline illness severity was based on PRISM-III score. (19) The percent positive uid balance was measured by dividing the net uid balance (intake minus output) measured in liters on any given day by the admission weight in kilograms and multiplying it by a factor of 100. The primary exposure variable of interest was the cumulative percent positive uid balance between days 1 and 7 (Day 1-7%PFB). The primary outcome of interest was complicated course -de ned as death within 7 days or the presence of 2 or more organ dysfunctions on day 7 of septic shock.(17) Severe SA-AKI was de ned as per Kidney Disease Improving Global Outcomes (KDIGO) stage 2 AKI or higher, which corresponds to a two-fold or greater increase in serum creatinine (SCr) relative to baseline. (20) Baseline SCr values were unknown for all patients in the cohort, and thus were imputed using their calculated body surface area (m2) and an eGFR of 120 ml/min per 1.73 m2, as validated in the literature. 19,20 Presence of persistent SA-AKI on day 3 (D3 SA-AKI) and use of any CRRT between days 1-7 were considered as co-variates.

Statistical analyses
Statistically analyses were conducted using R software. Demographic and clinical data were summarized with numbers with percentages or median with interquartile range. Differences between groups were determined by χ 2 squared test for categorical variables and by Kruskal Wallis test for continuous variables with post-hoc Dunn's test for multiple comparison testing between risk strata. Univariate and multi-variable Cox proportional regression were used to estimate hazard of complicated course in the cohort and across PERSEVERE-II risk strata. Models were adjusted for D3 SA-AKI, use of CRRT, with Day 1-7% PFB being considered a time-dependent variable. Day of study enrollment was considered as the starting time (Time 0). Survivors who were transferred out of the ICU before day 7 and those who remained in the ICU on day 7 were right censored. A p-value of 0.05 was used to test statistical signi cance throughout the study.

Results
A total of 681 children met inclusion criteria of this study. Figure 1 shows ow diagram detailing patients who were included in the study; 13 children with missing admission height or weight, 29 children with missing uid balance on day 1, and 678 children without PERSEVERE-II biomarker data were excluded from analyses. Demographic and clinical variables based on presence of complicated course are presented in Table 1. Patients with complicated course had higher 28-day mortality, hospital length of stay, presence of D3 SA-AKI, and use of any CRRT. Percent PFB was higher between days 1 and 4 among those with complicated course. Net daily even or negative uid balance was achieved later among patients with complicated course than to those without. Finally, among patients who received CRRT, duration of CRRT use was longer among those with complicated course than those without.

Discussion
We provide evidence that cumulative positive uid balance is independently associated with increased hazard of death and multiple organ dysfunctions among pediatric septic shock patients. Further, riskstrati ed analyses based on PERSEVERE-II biomarkers demonstrated that this increased risk is attributable primarily to patients categorized as having a high-risk.
Numerous studies among adults have sought to address the link between positive uid balance and clinical outcomes in the previous decade. Data from a recent meta-analyses of 31 observational studies and 3 randomized trials showed an independent association between positive cumulative uid balance and mortality.(22) However, there was substantial variation in adjustment for confounding factors with most accounting for illness severity and some adjusting for use and timing of CRRT. Few studies have adjusted for the concomitant presence of AKI. (23) Recent results from the assessment of worldwide acute kidney injury renal angina and epidemiology (AWARE) study from over 5,000 critically ill children, demonstrate that mild-to-moderate uid overload as early at the end of the ICU day 1 was associated with adverse outcomes including higher mortality and fewer ICU and ventilator free days.(9) Our current data from a prospective observational study from multiple-pediatric centers are corroborative.
Few studies have considered heterogeneous responses across critical illness subclasses or risk-strata with regard to cumulative uid balance and mortality outcomes. Through latent pro le analyses of electronic health record data, Zhang and colleagues showed that sepsis "pro le 3", characterized by shock and multiple organ dysfunctions, demonstrated a survival bene t with a higher cumulative uid balance in the rst 48 hours. In contrast, "pro le 4" demonstrated an increased odds of death with a more positive uid balance.(11) Wang et al. studied uid balance trajectories among critically ill patients a signi cant proportion of whom had sepsis. They demonstrated that relative to patients with a low uid balance, those with a high uid balance demonstrated an increased odds of mortality. In contrast, patients with a "decreasing" uid balance trajectory demonstrated a survival bene t. (24) Our data contradict those previously published by our group, where we reported worse outcomes with high positive uid balance among those categorized as having a low biomarker-based mortality risk strata alone. (15) There are several important contributors to the differences between direction of association including 1) larger sample size to allow for detection of statistically signi cant differences among patients with highmortality risk, 2) use of multivariable models that account for D3 SA-AKI and CRRT, and 3) use of Cox proportional model to account for the competing in uence of death on ICU survival and therefore on cumulative uid balance. An advantage of our data is that biomarkers were collected on day 1 of illness with subsequent assessment of cumulative uid balance. Our data show that patients with the highest burden of systemic in ammation as indicated by PERSEVERE-II biomarkers on day 1 of illness, were more likely to have had a positive uid balance in the post-resuscitative and stabilization phases of septic shock and the worst outcomes. Given the observational nature of our study, we cannot make any inferences on whether PFB causally mediates the in uences between biomarker-based risk strata assignment and worse clinical outcomes.
A recent randomized trial of restrictive vs. standard uid management in patients with septic shock demonstrated no mortality bene t. (25) It is conceivable, however, that given the heterogeneous nature of sepsis, a subset of patients with a higher degree of systemic in ammation and endothelial activation may in fact demonstrate a differential response to restrictive uid management. Accordingly, our data among patients with high-mortality risk based on PERSEVERE-II biomarkers warrant further investigation to determine whether these patients may selectively bene t from restrictive uid management with early initiation of vasoactive support or alternatively, such patients may bene t from early de-escalation or deresuscitation upon resolution of shock.
Our study has several limitations: 1) data on resuscitative uids received outside of the pediatric ICU were not available and thus not included, 2) uid choice (crystalloid vs colloid, isotonic vs non-isotonic) were not documented and could not be controlled for, 3) CRRT prescription data were not available, 4) given the relatively limited number of patients on CRRT, we did not adjust for duration and timing of CRRT in multivariate regression models. Each of these potential confounding variables have been shown in other studies to potentially in uence the association between post-ICU admission positive uid balance and clinical outcomes.

Conclusions
Cumulative positive uid balance is independently associated with worse clinical outcomes among children with septic shock after adjusting for illness severity, presence of severe AKI, and use of CRRT.
Risk-strati ed analyses demonstrated that patients with a high mortality risk, based on PERSEVERE-II biomarkers, primarily contributed to this association. Our data warrant further observation to determine whether this subset of patients may bene t from targeted deployment of restricted uid management coupled with early initiation of vasoactive support or early de-escalation approaches upon clinical resolution of shock.   Day 1-7 cumulative percent uid overload by presence of complicated course across PERSEVERE-II risk strata. Patients with complicated course had higher Day 1-7 %PFB than those without only among patients categorized as high PERSEVERE-II mortality risk.