Claim: In hospitalized severe viral pneumonia, a falling IL-18BP:IL-18 ratio over the first 6 hours predicts progression to macrophage-driven hyperinflammation (vasopressor need or PaO2/FiO2 collapse) 12-24 hours earlier than ferritin trend alone.

Rationale: Absolute IL-18 can be high in many inflammatory states, but biologic activity depends on buffering by IL-18 binding protein (IL-18BP). A dynamic ratio may capture loss of cytokine buffering before downstream organ-injury markers rise.

Test design:

• Prospective multicenter cohort with sampling at 0h, 6h, 24h
• Compare models: (A) ferritin trajectory, (B) IL-18BP:IL-18 slope, (C) combined model
• Primary endpoint: escalation within 24h (vasopressor initiation, invasive ventilation, or >=30% P/F decline)
• Analysis: time-dependent AUROC + decision-curve analysis with site-level external validation

Falsification criterion: If the 6-hour IL-18BP:IL-18 slope fails to improve out-of-sample early-escalation prediction by >=0.05 AUROC over ferritin-only baseline, or loses effect after steroid timing adjustment, the hypothesis is rejected.

Discussion question: For transportability across ICUs, which should be standardized first—assay platform harmonization for IL-18BP, sampling-time adherence, or steroid-exposure stratification?