A new mathematical analysis of randomized trials applied to critical care syndromes: a review.

Guideline reversals and reversals of randomized controlled trial (RCT) treatment effect estimates have been repeatedly documented in critical care syndrome research over several decades. Critical care syndrome RCTs differ structurally from conventional disease-based RCTs. Rather than enrolling patients on the basis of a specific disease or causal mechanism, they typically employ consensus-derived eligibility criteria composed of disease-and cause-agnostic prognostic physiological thresholds, laboratory abnormalities, and/or nonspecific clinical features. These enrollment criteria select a heterogeneous mixture of distinct diseases and causal systems which are mixed into a single trial. Consequently, the resulting treatment-effect estimate is not the average treatment effect of a single disease or causal mechanism. Instead, it represents a second-order average: an overall average generated by averaging disease or causal system-specific average treatment effects across multiple distinct causal systems. The resulting estimate is therefore best understood as an artefactual "mean of causally unrelated means": a mixture-weighted average of different treatment effects arising from biologically distinct disease processes. The proposed framework implies that the prevalent structure may produce unstable effect estimates and a "cause-mixture paradox," in which observed treatment effects reverse to the null or opposite sign across trials as the disease mixture of each trial shifts. This framework comprises a new tool for the mathematical assessment of evidentiary strength of critical care trials under guideline consideration.