Sepsis Phenotyping and Differential Response to Fluid Resuscitation

• In Sepsis
• Thu, 4 Sep 2025

Sepsis is a life-threatening, dysregulated immune response often leading to hypotension and multisystem organ dysfunction, with acute kidney injury (AKI) being the most common organ failure in critically ill patients. Traditional broad therapies for sepsis have largely failed due to the heterogeneity of the syndrome. Recent approaches focus on identifying biologically distinct host-response subphenotypes to enable targeted therapies. 

Using latent class analysis on ICU patients with sepsis-induced AKI, two subphenotypes (SP1 and SP2) were identified, differing in molecular profiles, outcomes, and response to early vasopressin therapy. A simplified three-biomarker model (Angiopoietin-1, Angiopoietin-2, sTNFR-1) reflects endothelial function and inflammation and can distinguish these subphenotypes.

Next steps included validating these subphenotypes in broader sepsis populations, identifying them early in emergency settings, and testing tailored interventions such as fluid management. The CLOVERS trial, which compared restrictive versus liberal fluid resuscitation in sepsis-induced hypotension, found no overall outcome differences, but it was hypothesised that SP2, characterised by high endothelial injury and inflammation, might respond worse to liberal fluids than SP1.

In a study, a secondary analysis of the CLOVERS trial, participants were randomised within 4 hours of meeting inclusion criteria to either a restrictive or liberal fluid resuscitation strategy for 24 hours. The restrictive strategy prioritised vasopressors, with rescue fluids allowed for severe hypovolaemia, while the liberal strategy began with a 2-L isotonic crystalloid infusion, followed by fluid boluses guided by clinical triggers, with rescue vasopressors permitted.

Plasma concentrations of Ang-1, Ang-2, and sTNFR-1 were used to classify patients into two subphenotypes (SP1 and SP2). A change in the sTNFR-1 assay was addressed through calibration, showing a high correlation with the previous assay. Patients with a probability ≥0.5 were classified as SP2. Additionally, a validated three-variable model incorporating IL-6, sTNFR-1, and serum bicarbonate was used to classify ARDS/sepsis subphenotypes.

AKI was defined using KDIGO criteria as an increase in serum creatinine >50% or >0.3 mg/dL above baseline at randomisation; urine output was not used. Baseline creatinine was taken from the lowest outpatient or inpatient value in the past year when available (n=909). If unavailable and no chronic kidney disease (CKD) history existed, baseline creatinine was imputed assuming an eGFR of 90 mL/min/1.73 m² (n=517). For patients with CKD but no prior creatinine, the lowest in-hospital value was used (n=29). Patients were staged per KDIGO guidelines, and 75 patients were on maintenance haemodialysis before enrolment.

The primary outcomes were 28-day and 90-day mortality. Secondary outcomes included ventilator-free days, renal replacement therapy–free days, and vasopressor-free days within 28 days.

Among 1,563 CLOVERS participants, 1,289 had available biospecimens and creatinine measurements. Using plasma Ang-1, Ang-2, and sTNFR-1, 1,016 (79%) were classified as SP1 and 273 (21%) as SP2. SP2 patients had higher rates of congestive heart failure, chronic kidney disease, and more severe AKI at enrolment. Treatment arm allocation was similar across subphenotypes.

After adjusting for demographics, comorbidities, AKI stage, SOFA scores, and baseline norepinephrine dose, SP2 had higher 28-day and 90-day mortality than SP1. SP2 patients also had fewer ventilator-free, renal replacement therapy–free, and vasopressor-free days at 28 days.

SP1 patients in the liberal fluid group received ~2 L more fluids than those in the restrictive group; SP2 received ~2 L more than restrictive, with greater vasopressor use in restrictive groups for both subphenotypes.

SP1 showed no mortality difference between liberal and restrictive fluids (28-day mortality 9% vs. 9%). SP2 had higher 28-day (41% vs. 27%) and 90-day mortality (48% vs. 36%) with liberal fluids compared to restrictive, with significant interaction. Greater probability of SP2 correlated with larger differences in 28-day mortality between fluid strategies. Secondary outcomes in SP2 improved with restrictive fluids: +3.2 vasopressor-free days, +2.7 renal replacement therapy–free days, +3.4 ventilator-free days. No differences were seen in SP1.

In 846 patients with AKI or ESRD, SP2 continued to show worse outcomes with liberal fluids, consistent with overall findings. Among patients without AKI/ESRD, SP2 prevalence was low (7%), and no treatment effect heterogeneity was observed.

Cross-tabulation of SP1/SP2 with hypoinflammatory/hyperinflammatory ARDS/sepsis subphenotypes showed moderate overlap. About 81% were concordant, 19% discordant. No differential response to fluid strategy was observed by ARDS/sepsis subphenotypes.

Overall, SP2, characterised by endothelial injury and inflammation, had worse outcomes, particularly with liberal fluid resuscitation, while SP1 outcomes were unaffected by fluid strategy.

The findings suggest that endothelial dysfunction–based phenotyping may identify patients most likely to benefit from targeted interventions, such as restrictive fluids and early vasopressors, whereas hypoinflammatory or hyperinflammatory classifications did not predict fluid responsiveness. Overall, these results support a precision medicine approach in sepsis-induced hypotension, moving beyond a one-size-fits-all strategy to optimise outcomes and resource use.

Source: AJRCCM
Image Credit: iStock