Association Between Daily Fluid Balance Trajectories and Dynamic Organ Failure Resolution (SOFA-2) in Sepsis: A Latent Class Growth Analysis of 16 262 Patients
Journal of Intensive Care Medicine · 21.11.2026
DOI: 10.1177/08850666261509926 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Intensive Care Medicine, online erschienen am 21.11.2026. Verfasst von Zhenzhou Zhai und 2 weiteren Autor:innen. Die Arbeit nennt 26 Literaturangaben.
Themen: Sepsis
Abstract (Originalsprache)
Background
Fluid management in sepsis has transitioned toward restrictive de-resuscitation during the intensive care unit (ICU) ebb phase (Days 3-7). The clinical heterogeneity of longitudinal fluid trajectories, their interaction with active therapeutic de-resuscitation, and their prognostic association with dynamic organ recovery evaluated under the 2025 SOFA-2 criteria require clarification in large multicenter databases.
Objective
To identify latent phenotypes of daily net fluid balance during the de-resuscitation phase and examine their independent associations with dynamic organ failure resolution and 28-day mortality after accounting for active de-resuscitation interventions.
Methods
We performed a retrospective observational cohort study of 16 262 adult septic patients from the MIMIC-IV v3.1 database who survived at least 72 h in the ICU. Daily net fluid balances (ml/kg/day) from Day 3 to Day 7 were modeled using group-based trajectory modeling without prior outcome stratification. Active de-resuscitation interventions, including cumulative intravenous loop diuretic dosing and renal replacement therapy (RRT), were extracted. Multivariable logistic regression and sensitivity analyses evaluated independent associations with 28-day mortality and longitudinal SOFA-2 trajectories.
Results
Three distinct trajectory phenotypes were identified: Class 1 (Rapid Recovery, 85.4%, n = 13 892), Class 2 (Late Response, 8.3%, n = 1354), and Class 3 (Persistent Failure, 6.3%, n = 1016). Class 1 maintained a steady, low net balance and experienced the lowest 28-day mortality (20.8%). Class 2 exhibited sustained high fluid retention until Day 5 before a delayed downward inflection, whereas Class 3 showed relentless fluid accumulation. Daily SOFA-2 scores resolved rapidly in Class 1 (decreasing from 4.24 on Day 3 to 1.18 by Day 7), whereas recovery was stalled in Class 2 (1.71 by Day 7) and persistently blunted in Class 3 (2.17 by Day 7). In multivariable logistic regression fully adjusted for baseline illness severity, loop diuretic exposure, and RRT, both Class 2 (adjusted odds ratio [aOR] 1.30, 95% CI 1.14-1.49; P < .001) and Class 3 (aOR 1.27, 95% CI 1.09-1.48; P = .002) were independently associated with increased 28-day mortality compared with Class 1. These associations remained consistent across sensitivity analyses restricted to non-RRT patients and loop diuretic users.
Conclusion
Discrete longitudinal fluid balance phenotypes identify clinically distinct sub-populations during the de-resuscitation window of sepsis. Rather than proving direct causality, these trajectories serve as clinical markers of underlying endothelial integrity, intrinsic excretory capacity, and physiological responsiveness to de-resuscitation.
Zitieren
APA
Zhai, Z., Jiang, X., & Gu, W. (2026). Association Between Daily Fluid Balance Trajectories and Dynamic Organ Failure Resolution (SOFA-2) in Sepsis: A Latent Class Growth Analysis of 16 262 Patients. Journal of Intensive Care Medicine. https://doi.org/10.1177/08850666261509926
Vancouver
Zhai Z, Jiang X, Gu W. Association Between Daily Fluid Balance Trajectories and Dynamic Organ Failure Resolution (SOFA-2) in Sepsis: A Latent Class Growth Analysis of 16 262 Patients. Journal of Intensive Care Medicine. 2026. doi:10.1177/08850666261509926
BibTeX
@article{Zhai2026,
author = {Zhai, Zhenzhou and Jiang, Xiaowen and Gu, Weifeng},
title = {Association Between Daily Fluid Balance Trajectories and Dynamic Organ Failure Resolution (SOFA-2) in Sepsis: A Latent Class Growth Analysis of 16 262 Patients},
journal = {Journal of Intensive Care Medicine},
year = {2026},
doi = {10.1177/08850666261509926}
}
EUVIMED