Association between proteinuria trajectories and outcomes in critically ill patients with sepsis or shock

BACKGROUND: Proteinuria results from kidney damage and can be a predictor of illness severity and mortality in the intensive care unit (ICU). However, the optimal timing of proteinuria measurements and the reference values remain undetermined. Our objective was to identify the patterns of proteinuria change associated with mortality in ICU patients with sepsis or shock. METHODS: This monocentric retrospective cohort study performed from April 2010 to April 2018 involved all ICU patients with sepsis or shock and at least two measurements of proteinuria from a 24h-urine collection during the first 10 days of ICU stay, the first of which was made within 48h after ICU admission. We identified proteinuria trajectories by a semi-parametric mixture model and analysed the association between the trajectories and the mortality at day 28 by Cox proportional-hazards model. RESULTS: A total of 3,344 measurements of proteinuria from 659 patients were analysed. Four proteinuria trajectories were identified. Trajectories 1, 2, 3 and 4 comprised 127, 421, 60 and 51 patients, and were characterized by a first proteinuria of 1.14 [0.66–1.55], 0.52 [0.26–0.91], 2.92 [2.38–3.84] and 2.58 [1.75–3.32] g/24h (p<0.001) and a mortality of 24.4%, 38%, 20% and 43% (p = 0.002), respectively. Trajectories 3 and 4 had a high first proteinuria (>2g/24h). Only, the proteinuria of trajectory 4 increased within 3 days following the first measurement and was associated with increased mortality at day 28 (hazard ratio: 2.36 95%CI [1.07–5.19], p = 0.03), regardless of acute renal failure. The factors associated with trajectory 4 were cancer (relative risk: 8.91 95%CI [2.09–38.02], p = 0.003) and use of inotropic drugs (relative risk: 0.17 95%CI [0.04–0.69], p = 0.01). CONCLUSION: This exploratory study of ICU patients with sepsis or shock identified four proteinuria trajectories with distinct patterns of proteinuria change over time and mortality rates. These results provide novel insights into renal pathophysiology and may be helpful to investigate subphenotypes of kidney injury among ICU patients in future studies.