Day-to-day variation in physiological network connectivity predicts outcomes in decompensated cirrhosis

Abstract Cirrhosis is a complex multisystem disease in which physiological network disruption is associated with adverse outcomes. Network mapping of routine biomarkers predicts mortality independently of established prognostic scores, but previous studies have assessed networks only at a single time point. We aimed to determine whether day-to-day network stability and variation during hospitalisation are associated with outcomes in decompensated cirrhosis. In this sub-study of the multicentre ATTIRE trial, daily albumin, bilirubin, creatinine, CRP and WCC measurements from 777 patients were analysed using parenclitic network mapping, with standard-care survivors as the reference population. Mean parenclitic deviation and day-to-day variation in network deviations during the first week of hospitalisation were compared between 15-day survivors and non-survivors and tested for associations with mortality after adjustment for MELD and age. Three network connections were identified. Mean parenclitic deviation was greater in non-survivors along the CRP–WCC (P = 9×10⁻⁵) and WCC–bilirubin (P = 7×10⁻⁷) axes. Day-to-day variation was also greater in non-survivors (mean absolute successive difference: P = 3×10⁻⁶; P = 8×10⁻¹², respectively) and predicted survival independently of mean deviation, MELD and age. Temporal network dynamics provide prognostic information beyond static measurements, highlighting dysregulated immune–hepatic interactions and reduced physiological network stability as determinants of survival in decompensated cirrhosis.