Systemic inflammation is increasingly recognised as a driver of acute deterioration in cirrhosis, including acute-on-chronic liver failure (ACLF). Interleukin-6 (IL-6) is a proximal inflammatory cytokine implicated in the acute-phase response and has been associated with mortality in prior cirrhosis cohorts. Critical gaps addressed by this study were whether admission IL-6 adds independent prognostic information beyond established severity scores in a real-world hospitalized population, whether its prognostic role varies by etiology (notably ALD versus MASLD), and whether infection or other pathways mediate any observed associations. The investigators predefined aims to characterise IL-6 distribution, evaluate independent prognostic value for multiple mortality endpoints, quantify incremental discrimination over existing models, identify subgroups of added value, explore mediators, and determine candidate cutoffs for clinical use.
This was a prospective, observational cohort conducted at a tertiary hepatology centre. Consecutive adult patients with cirrhosis admitted between April 2022 and October 2023 were enrolled if serum IL-6 was measured within 24 hours of admission as part of routine admission labs. The final sample comprised 489 patients. The protocol adhered to STROBE guidance and received Institutional Ethics Committee approval with a waiver of written informed consent because measurements and care were per routine practice and data were anonymised prior to analysis.
Cirrhosis diagnosis used clinical, biochemical, imaging and histology data when available. Decompensation and ACLF were defined using contemporary consensus criteria (Baveno VII for decompensation; EASL‑CLIF criteria for ACLF and CLIF-C Organ Failure scoring). Disease severity was quantified with Child–Turcotte–Pugh, MELD-3.0, and CLIF‑C AD scores. Alcohol-associated liver disease (ALD) and metabolic dysfunction-associated steatotic liver disease (MASLD) were defined using established alcohol-consumption thresholds and the 2023 Delphi consensus nomenclature respectively. Infection at admission was rigorously defined by site-specific criteria and confirmed by microbiological, radiological or clinical indices within 48 hours.
The primary outcome was in-hospital all-cause mortality. Secondary outcomes included mortality at 6 months, 12 months, and overall mortality during the observation period. Time-to-event analyses were performed from admission to death or last contact; in-hospital mortality was analysed as a binary endpoint.
Continuous data were summarised as medians with interquartile ranges and categorical data as counts and percentages. Log-transformation of IL-6 was used for parametric models and IL-6 was also analysed by tertiles and quartiles. Associations with mortality were estimated using univariable and multivariable logistic regression (adjusting primarily for age, sex and disease severity), Kaplan–Meier and Cox proportional hazards methods for time-to-event outcomes, and landmark analyses at 30 and 90 days. Incremental predictive value of IL-6 over established scores was quantified via likelihood-ratio tests, ΔAUC, and integrated discrimination improvement with internal validation by bootstrap resampling. Subgroup discrimination and interaction testing were performed across predefined strata. Mediation analyses used a Baron and Kenny framework. A CART model provided exploratory, non-parametric assessment of variable importance. Bayesian logistic regression with weakly informative priors and Hamiltonian Monte Carlo sampling complemented frequentist inference, yielding posterior odds ratios, 95% credible intervals, posterior probabilities, and model comparison metrics.
Median admission IL-6 was 23.8 pg/mL. In-hospital mortality in the cohort was 9.0%; overall mortality during follow-up was 70.6%. In multivariable logistic regression adjusting for age, sex and disease severity, IL-6 independently predicted in-hospital mortality (adjusted odds ratio 1.34, 95% CI 1.08–1.67). Bayesian analysis supported this association with a posterior probability of 99.6%. Combining IL-6 with conventional scores produced modest but statistically significant improvements in discrimination for in-hospital death; the CLIF-C Organ Failure score plus IL-6 model achieved the highest reported AUC of 0.802.
For longer-term endpoints (6-month, 12-month and overall mortality), IL-6’s independent prognostic signal attenuated, indicating a predominantly time-limited association centred on the in-hospital window.
The prognostic impact of IL-6 was substantially stronger in alcohol-associated liver disease compared with MASLD. Sub-cohort analyses showed that the etiology‑specific effect remained after adjustment for presence of infection and after excluding patients with severe alcohol-associated hepatitis. The incremental prognostic value of IL-6 over MELD-3.0 was confined to ALD patients and to those with infection.
Mediation analysis estimated that infection mediated 17.3% of the overall IL-6–mortality association. Stratified mediation showed infection accounted for approximately 23% of the association in ALD but only 2% in MASLD. These findings suggest that infection contributes to—but does not fully explain—the association between elevated IL-6 and in-hospital death, particularly in alcohol-related disease.
Although IL-6 added statistically significant discrimination for in-hospital mortality when combined with established severity measures, it did not outperform those scores alone. MELD-3.0 remained the dominant prognostic determinant across analyses. The added value of IL-6 was modest and concentrated in the short-term, etiology-specific and infection‑positive subgroups.
The report notes that data are patient-level clinical records not publicly shareable due to institutional and national data-governance rules; a de-identified dataset may be available to qualified researchers on request under a data‑use agreement and Ethics Committee approval. No formal a priori sample-size calculation was performed; the cohort represents consecutive eligible admissions over the study period. The observational design precludes causal inference, and the findings require prospective validation and economic appraisal before recommending routine IL-6 measurement.
Admission IL-6 in this large prospective cohort provided an independent, time-limited prognostic signal for in-hospital mortality that was concentrated in alcohol-associated liver disease and amplified when infection was present. IL-6 produced modest incremental discrimination over existing severity scores but did not supplant them, and its predictive value weakened for longer-term outcomes. The authors conclude that routine IL-6 testing in all hospitalized cirrhosis admissions is not supported by the current data; instead, targeted prospective validation and cost-effectiveness evaluation are warranted for ALD patients—especially those with infection or intermediate MELD-3.0 scores—where IL-6-informed risk stratification may be most valuable.