Chondroitin Sulfate Predicts 90-day Mortality and Response to Fluid Resuscitation Strategy in CLOVERS

RATIONALE The CLOVERS trial demonstrated that randomization of patients with sepsis-associated hypotension to an intravenous fluid-liberal vs. fluid-restrictive resuscitation strategy had no impact on 90-day survival. It is unclear if this negative result reflects heterogeneity of treatment effect across distinct patient endotypes in CLOVERS. We sought to determine if circulating glycosaminoglycans such as chondroitin sulfate (CS, a glycosaminoglycan that circulates as part of the inter-alpha-inhibitor antiprotease during homeostasis) and hyaluronic acid (HA, a glycosaminoglycan that is shed into the plasma during endothelial glycocalyx degradation) could identify patients differentially responsive to resuscitation strategy. METHODS We used an enriched sampling strategy to define a cohort of 574 CLOVERS patients (with plasma samples collected at 0, 24, and 72h after enrollment) and conducted analyses with inverse probability of selection weight adjustments. We used liquid chromatography-mass spectrometry to quantify plasma CS and HA. We used IL-6, measured at 0 and 72h, as a non-glycosaminoglycan comparator biomarker. The primary outcome was all-cause 90-day mortality. RESULTS Baseline plasma CS (a circulating glycosaminoglycan from multiple sources, including but not limited to endothelial glycocalyx degradation) was both associated with 90-day survival and strongly predicted the subsequent response to randomized resuscitation strategy. On multivariate analysis (controlling for age, severity of illness, comorbidities, and race/ethnicity), patients with baseline CS < 8,103 ng/ml (52% of patients) demonstrated improved survival when randomized to liberal fluid resuscitation, while patients with baseline CS > 8,103 ng/ml (∼48% of patients) had improved survival when randomized to restrictive resuscitation (Figure a,b). Baseline plasma HA (a biomarker for endothelial glycocalyx degradation) was strongly and independently associated with 90-day survival but did not identify a differential response to fluid resuscitation strategy. IL-6 (a marker of inflammation previously studied as a prognostic biomarker in sepsis) demonstrated a weaker association with 90-day survival and did not identify a differential response to resuscitation strategy (Figure c,d). Randomization to a liberal or restrictive fluid resuscitation strategy at day 0 had no effect on later (24h, 72h) plasma CS or HA nor any effect on 72h IL-6. CONCLUSIONS Circulating CS at CLOVERS enrollment was both a prognostic (independently predicting 90-day mortality) and predictive (identifying differential response to a liberal vs. restrictive intravenous fluid resuscitation strategy) biomarker in patients with sepsis-associated hypotension. HA (an index of endothelial glycocalyx degradation) and IL-6 (an index of systemic inflammation) were prognostic biomarkers but did not predict an individual patient's responsiveness to resuscitation.

Duke Scholars

Author Christopher Lindsell Biostatistics & Bioinformatics, Division of Biostatistics