Skip to main content

Foundational Trials

ImmunoSep

Image: Shutterstock / ChaNaWiT

Publication

  • Title: Precision Immunotherapy to Improve Sepsis Outcomes: The ImmunoSep Randomized Clinical Trial
  • Acronym: ImmunoSep
  • Year: 2025
  • Journal published in: JAMA
  • Citation: Giamarellos-Bourboulis EJ, Kotsaki A, Kotsamidi I, et al; ImmunoSep Study Group. Precision immunotherapy to improve sepsis outcomes: the ImmunoSep randomized clinical trial. JAMA. Published online December 8, 2025.

Context & Rationale

  • Background
    • Sepsis is biologically heterogeneous, with host responses spanning fulminant hyperinflammation and profound immune suppression; this heterogeneity is a plausible contributor to repeated neutral immunomodulation trials in unselected populations.
    • Two “actionable” immune states have been prioritised in translational work: macrophage activation-like syndrome (MALS), commonly operationalised by extreme hyperferritinaemia; and sepsis-induced immunoparalysis, often operationalised by low monocyte HLA-DR expression and secondary infection susceptibility.
    • IL-1 pathway blockade in broad sepsis cohorts was historically neutral, yet post hoc analyses suggested benefit signals in patients with macrophage activation syndrome-like biology.45
    • Immunostimulatory strategies (including IFN-γ) have restored aspects of innate immune function in small mechanistic and early clinical studies of sepsis-associated immune suppression, motivating phenotype-targeted testing.67
  • Research Question/Hypothesis
    • Whether a biomarker-defined “precision immunotherapy” strategy (anakinra for MALS; recombinant human interferon-γ for sepsis-induced immunoparalysis), added to standard care, improves early organ dysfunction trajectory compared with standard care plus placebo.
    • Whether this strategy improves key downstream outcomes (mortality, immune dysfunction reversal, and infection status).
  • Why This Matters
    • Tests (in a blinded, multinational RCT) a central precision-medicine thesis in sepsis: “treat the right biology in the right patient at the right time”, rather than applying a single immunotherapy across a heterogeneous syndrome.
    • If reproducible, a feasible “immune triage + targeted drug” workflow could reframe sepsis immunotherapy from repeated neutral mega-trials towards biomarker-stratified programmes of care.

Design & Methods

  • Research Question: In adults with sepsis and either MALS or sepsis-induced immunoparalysis, does a phenotype-guided precision immunotherapy strategy (anakinra or recombinant human interferon-γ) improve early organ dysfunction (SOFA trajectory) versus placebo, when added to standard care?
  • Study Type: Randomised (1:1), multicentre (33 sites), multinational (6 countries), double-blind, double-dummy, placebo-controlled, investigator-initiated strategy trial conducted in emergency department and ICU/hospital settings; protocol published separately.2
  • Population:
    • Setting: Hospitalised adults with sepsis due to community-acquired pneumonia, hospital-acquired pneumonia, ventilator-associated pneumonia, or primary bacteraemia; enrolment and treatment initiation within 72 hours of Sepsis-3 sepsis classification.
    • Key inclusion: Age ≥18 years; written informed consent; Sepsis-3 sepsis (SOFA ≥2 if presenting with infection, or SOFA increase ≥2 if already hospitalised); immune state consistent with MALS or sepsis-induced immunoparalysis per prespecified biomarker thresholds.
    • Key exclusions: Acute pyelonephritis, intra-abdominal infection, meningitis, or skin infection; stage IV malignancy; neutropenia (absolute neutrophil count <1500/mm3); “do not resuscitate” decision; active tuberculosis; HIV infection or primary immunodeficiency; systemic corticosteroids ≥0.4 mg/kg prednisone-equivalent for >15 days; anti-cytokine biologics within 1 month; systemic lupus erythematosus; multiple sclerosis/demyelinating disease; pregnancy or lactation; bacteraemia exclusions for likely contaminants/commensals or catheter-related infection.
    • Screening yield: 672 assessed; 281 randomised; 355 screened patients excluded because immune state was “unclassified” (ferritin ≤4420 ng/mL and monocyte HLA-DR ≥5000 receptors on CD45/CD14 monocytes).
  • Intervention:
    • Strategy: Standard care plus “precision immunotherapy” selected by immune phenotype.
    • MALS phenotype: Anakinra 200 mg IV (in 20 mL 0.9% saline) every 8 hours, plus subcutaneous dummy injections every 48 hours, for up to 15 days.
    • Immunoparalysis phenotype: Recombinant human interferon-γ 100 μg subcutaneously (final volume 0.5 mL) every 48 hours, plus IV dummy injections every 8 hours, for up to 15 days.
  • Comparison:
    • Strategy: Standard care plus placebo immunotherapy (double-dummy).
    • Dosing schedule: IV placebo (20 mL 0.9% saline) every 8 hours and subcutaneous placebo every 48 hours, for up to 15 days.
  • Blinding: Double-blind with double-dummy matching (participants, clinicians, and investigators blinded to assignment and phenotype-targeted active vs dummy route); primary outcome based on SOFA mitigates subjectivity-related detection bias.
  • Statistics: Powered for the primary endpoint assuming attainment in 40% (precision immunotherapy) vs 20% (placebo); α=0.05; power 90%; required 117 patients per group; inflated to 280 total for ~15% dropout; primary analysis included all randomised patients who did not withdraw consent and request removal of all data (modified intention-to-treat); deaths before day 9 imputed with SOFA=24 from death to day 15; discharges/losses before day 9 used last-observation-carried-forward SOFA.
  • Follow-Up Period: Primary endpoint through day 9; key secondary endpoints through day 15 and day 28 (immune restoration); mortality through day 90; safety captured as treatment-emergent adverse events.

Key Results

This trial was not stopped early. Planned enrolment was achieved (281 randomised; 276 included in the primary analysis after post-randomisation consent withdrawal with data removal requests).

Outcome Precision immunotherapy Placebo Effect p value / 95% CI Notes
Primary endpoint: ≥1.4-point decrease of mean SOFA score (days 2–9) from baseline 46/131 (35.1%) 26/145 (17.9%) OR 2.48 95% CI 1.42 to 4.32; P=.002 Absolute difference 17.2% (95% CI 6.8 to 27.2).
28-day mortality 57/131 (43.5%) 72/145 (49.7%) OR 0.78 95% CI 0.49 to 1.26; P=.34 Absolute difference 6.1% (95% CI −5.6 to 17.6).
90-day mortality 90/131 (68.7%) 98/145 (67.6%) OR 1.05 95% CI 0.63 to 1.75; P=.90 Absolute difference 1.1% (95% CI −9.8 to 11.9).
≥1.4-point decrease of mean SOFA score (days 2–15) from baseline 52/131 (39.7%) 34/145 (23.4%) OR 2.15 95% CI 1.28 to 3.61; P=.004 Absolute difference 16.3% (95% CI 5.3 to 26.8).
Reversal of sepsis-induced immune dysfunction 46/59 (78.0%) 32/66 (48.5%) OR 3.76 95% CI 1.72 to 8.22; P=.001 Subset with serial blood draws; MALS reversal defined by ≥15% ferritin decrease; immunoparalysis reversal defined by >8000 increase in absolute HLA-DR receptors per cell.
Infection status by day 15 (ordinal model) — — OR 0.59 95% CI 0.38 to 0.91; P=.02 Distribution: resolution 58/131 (44.3%) vs 46/145 (31.7%); intermediate 11/131 (8.4%) vs 9/145 (6.2%); failure 32/131 (24.4%) vs 44/145 (30.3%); superinfection 30/131 (22.9%) vs 46/145 (31.7%). OR reflects odds of worse outcome in precision immunotherapy vs placebo.
Primary endpoint in MALS stratum (SOFA decrease days 2–9) 12/25 (48.0%) 4/23 (17.4%) Not reported P=.04; 95% CI Not reported Phenotype-targeted active drug: anakinra vs placebo within the MALS subgroup.
Primary endpoint in immunoparalysis stratum (SOFA decrease days 2–9) 34/106 (32.1%) 22/122 (18.0%) Not reported P=.02; 95% CI Not reported Phenotype-targeted active drug: recombinant human interferon-γ vs placebo within the immunoparalysis subgroup.
Serious adverse event (probably or possibly related to study drug) 8/131 (6.1%) 3/145 (2.1%) Not reported Not reported Serious adverse reactions were uncommon; suspected unexpected serious adverse reactions occurred in 5 patients (3 placebo; 2 interferon-γ).
Anaemia (serious TEAEs >2% incidence; safety set) 12/25 (48.0%) 4/23 (17.4%) Not reported Not reported MALS stratum: anakinra vs placebo subgroup comparison; overall safety signal highlighted by investigators.
Haemorrhage (serious TEAEs >2% incidence; safety set) 5/106 (4.7%) 0/122 (0.0%) Not reported Not reported Immunoparalysis stratum: recombinant human interferon-γ vs placebo subgroup comparison; most haemorrhagic events occurred with concomitant thrombocytopenia.
  • Precision immunotherapy increased the proportion of patients achieving the prespecified SOFA improvement endpoint by day 9 (35.1% vs 17.9%; OR 2.48; 95% CI 1.42 to 4.32; P=.002), with consistent signal through day 15 (39.7% vs 23.4%; OR 2.15; 95% CI 1.28 to 3.61; P=.004).
  • Despite organ dysfunction and infection-status improvements, 28-day mortality was not significantly different (43.5% vs 49.7%; OR 0.78; 95% CI 0.49 to 1.26; P=.34), and 90-day mortality was similar (68.7% vs 67.6%; OR 1.05; 95% CI 0.63 to 1.75; P=.90).
  • Biological separation was supported by higher immune dysfunction reversal rates in the precision immunotherapy group (78.0% vs 48.5%; OR 3.76; 95% CI 1.72 to 8.22; P=.001) within the subset with serial immune monitoring.

Internal Validity

  • Randomisation and allocation: 1:1 randomisation using separate, site-specific computer-generated sequences; allocation concealment mechanism not reported in detail, but double-dummy drug preparation and blinding reduce plausibility of foreknowledge-driven selection.
  • Dropout or exclusions:
    • 281 randomised (135 precision immunotherapy; 146 placebo).
    • Post-randomisation consent withdrawal with request to remove all data: 4/135 in precision immunotherapy; 1/146 in placebo.
    • Primary analysis population: 131 (precision immunotherapy) vs 145 (placebo).
  • Performance/detection bias: Double-blind, double-dummy design; primary endpoint based on SOFA score (comparatively objective); infection status by day 15 incorporated investigator assessment using prespecified criteria (greater subjectivity than SOFA, but analysed ordinally).
  • Protocol adherence:
    • Included as treated-as-randomised: 131 vs 145.
    • Intervention discontinuation: 62/131 (47.3%) vs 72/145 (49.7%).
    • Reasons for discontinuation: death 35 vs 51; discharge before completion 18 vs 19; therapy-related serious adverse events 8 vs 2; therapy-related non-serious adverse event 0 vs 1.
  • Baseline characteristics:
    • Age: 69 (SD 13) vs 70 (SD 14) years.
    • Female sex: 42/131 (32.1%) vs 51/145 (35.2%).
    • Baseline SOFA: median 10 (IQR 7–12) vs 9 (IQR 6–11).
    • APACHE II: median 20 (IQR 15–26) vs 19 (IQR 14–25).
    • Charlson Comorbidity Index: median 4 (IQR 2–6) vs 5 (IQR 3–6).
    • Immune strata: MALS 25/131 (19.1%) vs 23/145 (15.9%); immunoparalysis 106/131 (80.9%) vs 122/145 (84.1%).
    • Baseline organ support: invasive mechanical ventilation 116/131 (88.5%) vs 129/145 (89.0%); continuous venovenous haemofiltration 23/131 (17.6%) vs 24/145 (16.6%).
  • Heterogeneity: The intervention strategy contained two distinct biologics (anakinra vs interferon-γ) assigned by immune phenotype; internal consistency was supported by prespecified phenotype definitions and prespecified within-phenotype analyses, but the “strategy” design can complicate attribution to a single mechanism or agent.
  • Timing: Treatment initiation within 72 hours of Sepsis-3 classification; biomarker classification required ferritin and quantitative monocyte HLA-DR measurements using standardised reagents and protocols across sites.
  • Dose: Anakinra 200 mg IV every 8 hours; interferon-γ 100 μg subcutaneously every 48 hours; maximum treatment duration 15 days; optimality of dosing for sepsis biology remains uncertain, but separation in immune restoration supports biological activity in at least a subset.
  • Separation of the variable of interest:
    • Primary endpoint separation: 35.1% vs 17.9% achieved ≥1.4-point mean SOFA improvement days 2–9.
    • Immune restoration (subset): 78.0% vs 48.5% met prespecified immune dysfunction reversal criteria.
    • Infection outcome distribution favoured precision immunotherapy (resolution 44.3% vs 31.7%; superinfection 22.9% vs 31.7%).
  • Key delivery aspects: High screen-to-randomise attrition driven by “unclassified” immune state; feasibility depends on rapid and standardised immunomonitoring and prompt delivery of phenotype-matched therapy.
  • Crossover: Not reported.
  • Adjunctive therapy use: Not reported in detail in the main paper; baseline organ support use was similar between groups.
  • Outcome assessment: Primary endpoint relied on SOFA trajectory; prespecified imputation rules applied (death assigned SOFA=24; early discharge/loss carried forward last SOFA).
  • Statistical rigor: Power target met for the primary endpoint; primary analysis was modified intention-to-treat (excluding patients with consent withdrawal and data removal requests); prespecified sensitivity analysis excluding early discharge before day 9 performed (results not detailed in the abstract).

Conclusion on Internal Validity: Overall, internal validity appears strong given double-blinding with double-dummy matching, prespecified phenotype definitions and endpoints, objective primary outcome metrics, and minimal post-randomisation exclusion (5/281), although reliance on imputation rules and a strategy design (two agents) complicate fine-grained causal attribution.

External Validity

  • Population representativeness:
    • Enrolled patients were older (mean ~70 years) with substantial baseline organ dysfunction (median SOFA 9–10) and high mortality, typical of many ICU sepsis cohorts, but not representative of the full sepsis spectrum.
    • Eligibility restricted to pneumonia (CAP/HAP/VAP) and primary bacteraemia, excluding common sources such as intra-abdominal infection and urinary tract infection/pyelonephritis.
    • Only patients with biomarker-defined MALS or immunoparalysis were eligible; a large proportion of screened patients had an “unclassified” immune state and were excluded (355/672), limiting applicability to broader sepsis populations.
  • Applicability:
    • Implementation requires rapid access to ferritin and quantitative monocyte HLA-DR flow cytometry with standardisation across laboratories, which may be difficult outside high-resource centres.
    • The strategy may translate best to systems with established immunomonitoring infrastructure and capacity for early phenotype-guided biologic administration.

Conclusion on External Validity: Generalisability is moderate to limited: findings are most applicable to high-risk, hospitalised/ICU adults with pneumonia or primary bacteraemia who meet stringent biomarker criteria for MALS or immunoparalysis and can access timely immunophenotyping.

Strengths & Limitations

  • Strengths:
    • Multinational, multicentre, double-blind, double-dummy randomised design, reducing performance and detection bias.
    • Explicit biomarker-defined enrichment strategy operationalising immune phenotypes (MALS vs immunoparalysis) into a bedside treatment algorithm.
    • Objective, prespecified primary endpoint with prespecified missing-data rules and achieved sample size target.
    • Evidence of biological and clinical separation (immune dysfunction reversal and infection-status improvement) alongside the organ dysfunction signal.
  • Limitations:
    • Primary endpoint is a surrogate (SOFA trajectory) rather than a patient-centred endpoint; mortality was not the powered endpoint and did not differ at 28 days.
    • Strategy design includes two biologics, complicating attribution and raising the possibility that effects differ substantially by phenotype subgroup.
    • High screening attrition driven by “unclassified” immune state; feasibility and generalisability depend on availability and speed of immunomonitoring.
    • Restricted infection sources and extensive exclusions constrain applicability to broader sepsis populations.

Interpretation & Why It Matters

  • Clinical meaning
    • In a biomarker-selected sepsis population, phenotype-guided immunotherapy improved early organ dysfunction and infection-status outcomes but did not demonstrate a mortality benefit.
    • The absolute SOFA-endpoint difference (17.2% by day 9) suggests a clinically relevant shift in early trajectory for some patients, while the absence of mortality separation highlights the need for adequately powered, patient-centred phase 3 evaluation.
  • How this changes assumptions
    • Supports the plausibility that “precision immunotherapy” can produce measurable organ dysfunction and immune restoration signals in sepsis when patients are selected by immunophenotype rather than syndrome labels alone.
    • Reinforces the concept that failure of prior immunotherapy trials may reflect patient selection more than therapeutic class futility, aligning with contemporary immunobiology frameworks.3

Controversies & Subsequent Evidence

  • Endpoint selection and interpretability:
    • Primary outcome was a prespecified SOFA trajectory threshold rather than mortality, enabling a feasible sample size but raising debate about how reliably organ dysfunction surrogates predict patient-centred benefit in sepsis trials.
    • Mortality was not significantly different at day 28, and late mortality was similar, emphasising that the trial should be read as a precision-medicine proof-of-concept rather than definitive practice-changing mortality evidence.
  • Phenotype definitions and feasibility:
    • The strategy depends on operational biomarker thresholds (ferritin for MALS; quantitative monocyte HLA-DR for immunoparalysis) and timely testing logistics; external replication will hinge on reproducible assays and rapid turnaround.
    • Most screened patients were excluded because their immune state was “unclassified”, implying that the tested strategy addresses a high-risk subset rather than “most sepsis”.
  • Drug-class antecedents and biological plausibility:
    • Neutral IL-1 receptor antagonist trials in unselected sepsis populations provide important context for the biomarker-guided anakinra signal in MALS-like biology.4
    • A mechanistically aligned reanalysis associated IL-1 blockade with lower mortality in sepsis patients with macrophage activation syndrome features, supporting the enrichment rationale.5
    • Foundational mechanistic work demonstrated restoration of monocyte function with IFN-γ in sepsis, and a randomised pilot study showed in vivo reversal of immunoparalysis, providing biological support for the immunostimulatory component of the ImmunoSep strategy.67
  • Positioning within the precision-immunotherapy landscape:
    • Earlier sepsis precision-immunotherapy efforts (e.g., PROVIDE) support feasibility of phenotype-guided approaches but also illustrate the challenges of operationalising immune biology into scalable trial designs.8
    • Other biomarker-enriched immunomodulation programmes in septic shock (e.g., nangibotide in ASTONISH) demonstrate the field’s shift toward enrichment strategies, though clinical outcome signals remain inconsistent across agents and endpoints.10
    • A recent phase 2 trial of IFN-γ-1b to prevent hospital-acquired pneumonia in the critically ill contextualises safety/biological plausibility for IFN-γ strategies, though it addresses a different clinical question than ImmunoSep’s organ dysfunction endpoint.9
  • Editorial perspective:
    • The accompanying editorial frames ImmunoSep as an important “end of the beginning” moment for sepsis precision therapeutics, while emphasising that the path to practice change requires replication, scalability of immunophenotyping, and demonstration of patient-centred benefit.1

Summary

  • ImmunoSep tested a phenotype-guided precision immunotherapy strategy in adult sepsis (MALS → anakinra; immunoparalysis → interferon-γ) using a double-blind, double-dummy, placebo-controlled design.
  • The primary organ dysfunction endpoint was met: 35.1% vs 17.9% achieved a ≥1.4-point mean SOFA improvement by day 9 (OR 2.48; 95% CI 1.42 to 4.32; P=.002).
  • Mortality did not differ significantly at day 28 (43.5% vs 49.7%) or day 90 (68.7% vs 67.6%).
  • Biological separation was supported by higher immune dysfunction reversal (78.0% vs 48.5%) and better infection-status distribution by day 15 (ordinal OR 0.59; 95% CI 0.38 to 0.91; P=.02).
  • Safety signals were phenotype-linked: anaemia was more frequent with anakinra in the MALS stratum (48.0% vs 17.4%), and haemorrhage occurred with interferon-γ in the immunoparalysis stratum (4.7% vs 0.0%).

Overall Takeaway

ImmunoSep is a landmark proof-of-concept that phenotype-guided immunotherapy can shift early organ dysfunction trajectory and biological endpoints in selected sepsis patients under rigorous blinding. Its neutral mortality findings, high screening attrition, and operational complexity mean it should be viewed as a platform for phase 3, patient-centred precision trials rather than immediate broad adoption.

Overall Summary

  • Biomarker-guided precision immunotherapy improved early SOFA-based organ dysfunction outcomes.
  • No statistically significant mortality benefit at 28 or 90 days in this phase 2b strategy trial.
  • Feasibility and impact depend on scalable, rapid immunophenotyping and confirmatory trials with patient-centred endpoints.

Bibliography