Introduction: Critically ill cirrhosis patients exhibit heterogeneity, often leading to therapeutic inertia. We hypothesized that integrating baseline comorbidity endotypes with dynamic biomarker trajectories would identify clinical phenotypes with actionable “treatable traits,” providing a framework for timely intervention. Methods: We characterized under IRB exemption 1,053 cirrhosis patients stratified by severity (Mild, n=701; Moderate-Severe, n=352). Clustering of Elixhauser Comorbidity Index (v2025.1, n=39) domains defined comorbidity endotypes. Biomarker trajectory (eg, Systemic Immune-Inflammation Index SII, estimated Plasma Volume Status/ Albumin ratio) over first two ICU days (D) captured dynamic host response. These data streams were integrated using sequential machine learning approaches to define clinical phenotypes and treatable traits. Results: Similar intergroup demographics included age 60 years among females (36%), males (64%), Whites (86%), Blacks (5%) and other races (9%). Mild cohort expressed distinct “Neuro-Psychiatric,” “Hemorrhagic-Anemic,” and “Cardio-Metabolic” endotypes. Though Moderate-Severe cohort was dominated by “Systemic Decompensation 8.2 mg/dL and 1720) coupled with worsening core organ function (INR increased +0.56), indicative of “immuno-nutritional paralysis.” Mild group expressed a robust but resolving inflammatory response (higher D1 CRP and SII; 11.5 mg/dL and 2844) and improving organ function (INR decreased -0.24). Integrating comorbidities with dynamic biomarker data, three clinical phenotypes emerged linked to an actionable treatable trait: “Hyperinflammatory Responder” (resolving inflammation), “Immuno-Nutritional Collapse” (immune paralysis), and “Progressive Endotheliopathy” (capillary leak). Conclusions: Integrated analysis revealed distinct clinical phenotypes with actionable treatable traits. Translating these phenotypes into clinical practice offers a strategy to overcome therapeutic inertia by providing objective, real-time triggers for titrating care, initiating targeted therapies, or avoiding harmful interventions.
Rivero et al. (Sun,) studied this question.