Therapeutic Plasma Exchange in COVID-19-Associated Sepsis: IL-6 Dynamics, Inflammatory Phenotypes, and Short-Term Organ-Failure Trajectories in a Real-World Cohort

<b>Background and Objectives:</b> In severe COVID-19-associated sepsis, therapeutic plasma exchange (TPE) is used as a rescue strategy to modulate cytokine and coagulation derangements, but its biomarker and organ-failure effects remain incompletely characterised. We evaluated peri-procedural changes in interleukin-6 (IL-6), other inflammatory markers, and Sequential Organ Failure Assessment (SOFA) scores according to TPE intensity, timing, and inflammatory phenotypes. <b>Methods:</b> We conducted a single-centre retrospective cohort study including 102 mechanically ventilated adults with COVID-19-associated sepsis who received ≥1 TPE session. Patients were grouped by number of sessions (1, 2, ≥3), timing (≤14 vs. >14 days from symptom onset), IL-6 responder status (≥50% reduction), and two unsupervised inflammatory-thrombotic clusters. Peri-procedural changes (Δ) in biomarkers and SOFA were compared using non-parametric tests, with multivariable logistic and linear regression exploring predictors of IL-6 response and ΔSOFA. <b>Results:</b> Baseline severity was similar across TPE-intensity groups, with median APACHE II scores of 11-12 and SOFA scores around 7 in all strata. Median IL-6 concentrations declined after TPE in each group (e.g., Δ -59.4 pg/mL after 1 session and Δ -65.1 pg/mL after ≥3 sessions), but between-group differences in ΔIL-6 were not statistically significant (<i>p</i> = 0.276). By contrast, D-dimer exhibited a marked decline only in the 1-session group (median Δ -1.7 mg/L vs. ~0.0 mg/L in the 2- and ≥3-session groups; <i>p</i> < 0.001). Timing (early vs. late TPE) did not materially affect ΔIL-6, ΔCRP, ΔSOFA (median 0.0 in both), or ΔD-dimer. Overall, 50% of patients were IL-6 responders; baseline IL-6 was the only independent predictor (adjusted OR 1.9 per doubling, 95% CI 1.3-2.8). A hyperinflammatory-thrombotic cluster (<i>n</i> = 44) exhibited higher baseline IL-6 (612.3 vs. 92.4 pg/mL), more ≥3-session TPE (65.9% vs. 29.3%), and higher IL-6 responder rates (75.0% vs. 31.0%), but similar 28-day mortality (40.9% vs. 29.3%). <b>Conclusions:</b> In this real-world TPE programme, biochemical improvements-particularly IL-6 and D-dimer reductions in hyperinflammatory-thrombotic patients-were not consistently accompanied by short-term SOFA or survival benefits, underscoring the need for phenotype-guided and trial-based use.