Heterogeneity of Treatment Effect by Molecular Endotype in Acute Respiratory Distress Syndrome: A Systematic Review

Despite numerous randomized controlled trials (RCTs) in acute respiratory distress syndrome (ARDS), few have identified effective therapies. One reason may be heterogeneity of treatment effect (HTE), in which distinct subgroups respond differently to the same intervention, potentially reflective of underlying biological endotypes.

We conducted a systematic review to evaluate effect modification by molecular endotype across randomized ARDS trials and assess credibility of reported subgroup effects. In doing so, we aim to expose current knowledge gaps in how subgroups are defined and inform future clinical practice guideline recommendations.

Following PRISMA guidelines, we searched OVID Medline, Embase, and the Cochrane Central Register of Controlled Trials for secondary and post-hoc analyses of RCTs in adult critically-ill patients with ARDS that evaluated treatment effects stratified by molecular endotype. We evaluated subgroup credibility using the Instrument to assess the Credibility of Effect Modification Analyses (ICEMAN).

Eleven secondary analyses from ten RCTs (5,514 participants) were included. Most studies identified two reproducible endotypes, hyperinflammatory and hypoinflammatory, using small panels of circulating protein biomarkers. Of 11 analyses, four demonstrated statistically significant effect modification (higher PEEP, liberal fluid strategy, vv-ECCO₂R, simvastatin). The hyperinflammatory endotype drove every observed interaction, though the direction varied by intervention. The most credible subgroup effect was observed with liberal fluid strategy in hyperinflammatory ARDS (high credibility), followed by simvastatin and vv-ECCO₂R (moderate credibility).

Endotyping in ARDS reliably identifies biologically distinct patient groups, but evidence that endotypes consistently modify treatment response is sparse and of generally low credibility. Progress toward endotype-guided therapy will require multi-omic molecular characterization, prospective trials with pre-specified interaction hypotheses, and real-time endotype assignment.