Abstract

Background: Systemic inflammation is known as a major role in pathogenesis of acute-on-chronic liver failure (ACLF) from cross-sectional studies, but its detailed impacts on the dynamic trajectory of this disease from the absence of organ dysfunction (OD) to onset of OD, organ failure (OF) and eventual ACLF remains poorly understood due to challenges in collecting data and samples from individuals. Multinomial and ordinal regression analyses are useful tools to circumvent the limitation intrinsic to cross-sectional studies in cirrhosis. They identify significant predictors of different outcome categories and provide insights into the transition between stages towards ACLF.

Methods: We investigated inflammatory signals, including blood cell counts and 21 serum biomarkers in 238 inpts with cirrhosis assigned to one of the following four stages: no OD (n=110), OD (n=40), OF without ACLF (n=54), and ACLF (n=34). The association of signals with each stage was assessed with multivariable models using multi-category outcomes without (Softmax regression) and with order (adjacent-category logit model). Prognostic ability of the signals was assessed with competing risk analysis followed by the evaluation of integrated discrimination improvement (IDI) and net reclassification improvement (NRI) of these signals to the exiting prognostic models. Another 156 cirrhosis inpts were used for validation.

Results:

Descriptive analysis: The progression of cirrhosis towards ACLF is characterized by an increasingly maladapted immune response as indicated by: 1) blunted type 1 immunity (type 1 effector IFN-γ did not change across stages); 2) abortion of type 2 immunity (production blocked between first-line cytokine, IL-25 and IL-33 second-line cytokines IL-4, IL-13); 3) imbalance of type 3 immunity (IL-22 dominant phenotype) ; 4) enhanced immunosuppression (continuous increases of anti–inflammatory markers (IL-10, sCD163), decreases in chemokines [e.g., IL-8, IP-10], and a reduction in lymphocyte count)

Multivariable analyses: Higher levels of IL-6, IL-10, sCD163, MCP-1, IL-25, IL-33, IL-22, or higher white-cell or neutrophil counts were associated with increased risk of getting more severe from OD towards ACLF at any disease stage (Figure 1).

Prognostic analysis: The 28–day all–cause mortality can be independently predicted by the baseline levels of IL–6, sCD163, and MIP–3α, which together with blood immune cell, significantly enhanced the prognostic accuracy of CLIF-C OF (C-index [95%CI], 0.88 [0.82-0.94] vs. 0.84 [0.77-0.92], both P=<0.01 by IDI and NRI). The results were validated in an independent cohort with 156 patients.

Conclusion: Future therapies targeting the activation of type 1 immunity, restoration of type 2 immunity and balance of type 3 immunity would be important to prevent ACLF. Markers identified in this study are potential targets for developing new or improving prognostic scores currently available.