Inflammatory mediators are predictors of clinical outcomes in cirrhotic patients with non-acute decompensation

Abstract

Background

Non-acute decompensation (NAD) has been proposed as a distinct pathway of decompensation. This study evaluated the prognostic value of inflammatory mediators for subsequent clinical outcomes in patients with NAD.

Methods

Outpatients with NAD were included in this ambispective cohort study (Cohort 1 retrospectively; Cohort 2: prospectively). Patients were followed for 24 months. The primary endpoints were the occurrence of acute decompensation (AD), acute-on-chronic liver failure (ACLF), and liver-transplantation-free mortality.

Results

Cohort 1 ultimately comprised 373 patients. Cohort 2 included 192 patients. In multivariable analysis of cohort 1, both IL-10 and IL-6 levels independently predicted AD (log-transformed-IL-10, HR: 1.26; 95% confidence interval, CI: 1.16–1.36; P < 0.001; log-transformed-IL-6, HR: 1.13; 95% CI: 1.01–1.27; P = 0.03) and ACLF (log-transformed-IL-10, HR: 2.03; 95% CI: 1.57–2.64; P < 0.001; log-transformed-IL-6, HR: 2.21; 95% CI: 1.57–3.11; P < 0.001). Patients were stratified into three groups: no systemic inflammation (SI), SI without immunodeficiency (ID), and SI with ID (elevated IL-6 + IL-10). While SI alone increased AD risk, the coexistence of ID markedly amplified the risk of severe outcomes. Compared to SI alone, patients with SI + ID had a substantially higher risk of ACLF (Cohort 1 (training) HR: 28.02; 95% CI: 12.54–62.62; Cohort 2 (validating) HR: 39.78; 95% CI: 10.97–144.30). Nearly all deaths occurred in the SI + ID subgroup [Cohort 1: 10/10; Cohort 2: 4/5].

Conclusions

A combined biomarker profile of SI (IL-6) and ID (IL-10) identifies a distinct high-risk subgroup among NAD outpatients, characterized by a markedly elevated risk of progression to AD, ACLF and liver-transplantation-free mortality.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12876-026-05105-9.

Background

Cirrhosis represents the end stage of progressive liver fibrosis and is a major contributor to global morbidity and mortality [1]. The clinical course of cirrhosis is characterized by a dynamic transition from a stable, compensated state to acute decompensation (AD), which develops rapidly, is frequently triggered by a precipitating event, and is marked by complications such as ascites, hepatic encephalopathy (HE), or variceal bleeding (VB) that usually require hospitalization [2–4]. In its most severe form, AD may evolve into acute-on-chronic liver failure (ACLF), a syndrome of acute hepatic decompensation superimposed on chronic liver disease that is accompanied by one or more extrahepatic organ failures and carries a high short-term mortality [5–8]. More recently, non-acute decompensation (NAD) has been proposed as a distinct pathway of decompensation, complementing the current spectrum, and is defined by mild ascites, mild grade 1 or 2 HE, or jaundice that does not require hospitalization [9]. In contrast to the rapid, hospitalization-requiring course of AD, NAD develops more slowly and insidiously, often appearing in patients experiencing their first decompensation event and/or in the early stages of chronic liver disease [10]. Although successful treatment may lead to clinical stabilization or even a return to a compensated state, NAD may also progress to AD and, ultimately, to ACLF [9–11].

Patients with NAD therefore face a significant risk of clinical worsening—including recurrent AD, ACLF, and death—yet identifying which patients will follow this adverse trajectory remains a major clinical challenge, particularly for vulnerable outpatients in a non-acutely decompensated state, including those with compensated cirrhosis or those who have recovered from a prior decompensation event [11–12]. At present, no validated model predicts the clinical course of decompensated cirrhosis. Established prognostic tools such as the Child-Pugh score and the MELD (model for end-stage liver disease) score quantify the severity of liver disease, primarily reflecting hepatic synthetic function and portal hypertension [13–14]; however, they were not designed to forecast the trajectory of decompensation and may not capture the underlying pathogenic drivers of disease progression—most notably systemic inflammation and immune dysfunction, which are increasingly recognized as pivotal in the pathogenesis of AD and ACLF [15–16].

According to the systemic inflammation hypothesis, the progression of hepatic decompensation is driven by a state of chronic, maladaptive systemic inflammation that mechanistically links the host immune response to clinical deterioration [16–18]. Within this framework, cirrhosis-associated immune dysfunction encompasses both excessive systemic inflammation and a coexisting immunodeficiency that predisposes patients to infection and organ failure [16–18]. Cytokines, as key mediators of inflammation and immune regulation, have therefore emerged as promising biomarkers of this process [16, 19]. Interleukin-6 (IL-6), a pro-inflammatory cytokine and major driver of the acute-phase response, has been strongly associated with disease severity, systemic inflammation, and poor prognosis in cirrhosis [19–22], whereas interleukin-10 (IL-10), a potent anti-inflammatory cytokine, acts to dampen immune responses and maintain homeostasis [23]. The balance between pro-inflammatory (e.g., IL-6) and anti-inflammatory (e.g., IL-10) signals is thought to be crucial in determining the clinical trajectory [17, 22]. Indeed, patients with end-stage liver disease may exhibit concurrent systemic inflammation (SI) and immunodeficiency (ID), signaling a risk of impending deterioration even in the absence of acute symptoms [24–26]. Taken together, these observations lead to the hypothesis that the pro- and anti-inflammatory cytokine balance could serve to predict the course of decompensated cirrhosis.

Therefore, we conducted an ambispective cohort study involving outpatients with NAD to evaluate whether inflammatory mediators measured at enrollment predicted the subsequent occurrence of AD, ACLF, and liver-transplantation-free mortality during a 24-month follow-up period. Our findings aim to contribute to the development of more precise, biologically-informed risk assessment tools for the management of cirrhotic outpatients with NAD.

Methods

Study design

We conducted an ambispective cohort study comprising two cohorts (Fig. 1). Cohort 1 (retrospective) included all patients with NAD who attended the Outpatient Center of our hospital and underwent cytokine testing as part of routine clinical care between January 1, 2020 and March 31, 2023. Cohort 2 (prospective) consecutively enrolled 217 patients with NAD attending the outpatient clinic of the Department of Hepatology from September 17, 2023 to December 22, 2023, and 25 patients were lost to follow-up.

Fig. 1.

Flow chart of patient screening, enrollment, and inclusion in Cohorts 1 and 2. HCC, hepatocellular carcinoma

Demographic information, clinical data, laboratory results, imaging findings, and cytokine assay results were collected for all patients.

In both cohorts, the blood sample for cytokine measurement was obtained at the same outpatient visit at which NAD was diagnosed and the patient was enrolled, before the initiation of any treatment for the current presentation; this timepoint was designated as Month 0. The sampling timing and pre-treatment condition were therefore identical across the two cohorts.

For Cohort 1, follow-up was conducted by telephone and required patient recall. For Cohort 2, routine follow-up was scheduled at 3, 6, 12, and 24 months and was conducted primarily by telephone; additional follow-up by telephone and WeChat was performed whenever events such as AD, ACLF, or re-hospitalization occurred. During follow-up, any condition requiring active treatment was managed accordingly.

Cytokine measurement

A 12-plex cytokine detection kit (based on magnetic-particle chemiluminescence immunoassay) was used for the quantitative in vitro determination of serum concentrations of interleukin (IL)-1β, IL-2, IL-4, IL-5, IL-6, IL-8, IL-10, IL-12p70, IL-17a, tumor necrosis factor-α (TNF-α), interferon-γ (IFN-γ), and interferon-α (IFN-α). All assays were performed in strict accordance with the manufacturer’s instructions.

For all marker-based stratifications, the reference upper limit of normal (ULN) provided by the assay manufacturer / established by the clinical laboratory was used as the cutoff (hsCRP 2.95 mg/L; IL-6 10.56 pg/mL; IL-10 10.89 pg/mL). hsCRP was further categorized according to the ULN.

Patient selection

Inclusion criteria are as follows: (1) Age 18 years and above, 80 years and below; (2) Diagnosed as patients with liver cirrhosis through histological, biochemical, ultrasound or radiological examinations; (3) With ascites or mild grade 1 or 2 HE or jaundice, not requiring hospitalization; (4) Cohort 1 obtained ethical committee approval for exempted informed consent, and all included patients in Cohort 2 obtained informed consent.

Exclusion criteria are as follows: (1) Patients requiring hospitalization, within 2 weeks of receiving inpatient treatment; (2) Acute infection within the past week or using prophylactic antibiotics within one week; (3) Human immunodeficiency virus infection; (4) Using corticosteroids, immunosuppressants or cytotoxic drugs before cytokine testing; (5) Hepatocellular carcinoma (HCC) or non-liver-related malignant tumors; (6) Having undergone liver lobectomy, LT surgery or portosystemic shunt surgery; (7) Final diagnosis considering special infections; (8) Final diagnosis considering blood system diseases; (9) Having severe heart failure (New York Heart Association cardiac function classification of grade 4), chronic obstructive pulmonary disease (Global Chronic Obstructive Pulmonary Disease Initiative classification of grades 3–4), or renal dysfunction (Chronic Kidney Disease stage 5) and these conditions were diagnosed before the onset of liver cirrhosis; (10) Pregnant or lactating; and (11) Lack of follow-up data.

Definitions of outcomes

NAD was defined as the slow formation of ascites (< 3 cm on ultrasound), grade 1–2 HE, or progressive jaundice (> 2 × ULN) in non-cholestatic cirrhosis [9]. AD was defined as the first occurrence or recurrence of any of the following within a 2-week period: grade 2 or 3 ascites, acute HE (first episode or recurrence) in a patient with previously normal consciousness, acute VB, or any acute bacterial infection (BI) [9]. ACLF was defined according to the established EASL criteria, based on failure of the liver, kidneys, brain, coagulation, circulation, and lungs; the number of organ failures was used to stratify ACLF into grades 1, 2, and 3, corresponding to 1, 2, or ≥ 3 organ failures, respectively [6].

Endpoints

The objective of this study was to evaluate the prognostic value of inflammatory mediators for subsequent clinical outcomes in patients with NAD. The primary endpoints were the occurrence of AD, ACLF, and liver-transplantation-free mortality during the 24-month follow-up, with each endpoint analyzed separately. Liver-transplantation-free mortality was defined as all-cause death before liver transplantation; patients who underwent liver transplantation were censored on the date of transplantation. Rehospitalization was a secondary endpoint.

Ethics

The study was performed in accordance with the ethical principles of the Declaration of Helsinki (as revised in 2013) and the Declaration of Istanbul, and was approved by the Ethics Committee of the Second Hospital of Nanjing (Approval No. 2023-LY-ky-056). For the retrospective cohort (Cohort 1), the requirement for informed consent was waived by the Ethics Committee due to the retrospective design and the use of anonymized data. Written informed consent was obtained from all participants in the prospective cohort (Cohort 2).

Sample size calculation

The sample size was estimated using the Wald (normal-approximation) precision method for a single proportion, as implemented in the samplingbook package. With a two-sided significance level of 0.05, an expected sensitivity of 0.83, and a permissible error (half-width of the 95% confidence interval, CI) of 0.05, a minimum of 217 patients was required. Because this approach is based on the precision of a single proportion, an assumed specificity and an expected event rate were not required as input parameters. Loss to follow-up was not explicitly incorporated into the calculation, which we acknowledge as a limitation.

Quality control and data management

Three physicians periodically reviewed the data to identify errors and assess accuracy. Laboratory personnel were blinded to the patients’ clinical status. Data collection and patient follow-up were performed by separate research personnel. The datasets were subsequently merged by two researchers, and statistical analysis was initiated only after verification of all collected clinical variables.

Statistical analysis

Data are presented as frequencies and percentages, means and standard deviations, or medians and interquartile ranges, as appropriate. The normality of continuous variables was assessed using the Shapiro–Wilk test, supplemented by visual inspection of Q–Q plots. Normally distributed continuous variables are presented as mean ± standard deviation and were compared between groups using the independent-samples Student’s t-test, whereas non-normally distributed continuous variables are presented as median (interquartile range) and were compared using the Wilcoxon rank-sum (Mann–Whitney U) test. Categorical variables are presented as frequencies (percentages) and were compared using the chi-square test or Fisher’s exact test, as appropriate. Kaplan–Meier curves are presented by group. All statistical tests were two-sided, and a P value < 0.05 was considered statistically significant. For the time-to-event analyses, univariable Cox proportional hazards regression was first performed for each candidate variable; only variables with a univariable P value < 0.05 were entered into the multivariable Cox model, which was then constructed using forward stepwise selection based on the Wald statistic (entry and removal probabilities of 0.05 and 0.10, respectively). Hazard ratios are reported with 95% CIs. The same model-building strategy was applied to the analyses of AD and ACLF. For the liver-transplantation-free mortality analysis, only 10 deaths occurred in Cohort 1. A conventional multivariable Cox model would therefore be at substantial risk of overfitting and unstable estimation. We consequently fitted a penalized Firth Cox model restricted to the two prespecified biomarkers, and these estimates were regarded as exploratory. Statistical analyses were performed using GraphPad (version 9.0), R (version 4.3) using the survival and coxphf packages and SPSS (version 22.0).

Results

Baseline characteristics of the cohorts

Based on the inclusion and exclusion criteria, Cohort 1 ultimately comprised 373 eligible NAD patients (mean age: 57.73 ± 11.45 years; 150 males); Cohort 2 included 192 eligible NAD patients (mean age: 58.41 ± 11.75 years; 74 males) (Fig. 1).

The baseline characteristics of the patients are presented in Table 1. Notable differences in baseline data were observed between Cohort 1 and Cohort 2, including the distribution of NAD types, etiologies, and cytokine levels.

Table 1.

Baseline characteristics of the cohorts

	Cohort 1 (N = 373)	Cohort 2 (N = 192)	P-value
Demographic variables
Age (years)	57.73 ± 11.45	58.41 ± 11.75	0.51
Sex (males/females)	150/223	74/118	0.70
Etiology of cirrhosis, n (%)			0.001
AIH or OS	23 (6.2)	10 (5.2)
Alcoholic	30 (8.0)	29 (15.1)
HBV or HCV	190 (50.9)	102 (53.1)
MASH	2 (0.5)	0 (0.0)
PBC	28 (7.5)	10 (5.2)
combined	37 (9.9)	14 (7.3)
others or unspecific	63 (16.9)	27 (14.1)
Clinical presentation of NAD
Jaundice (yes/no)	71/302	37/155	0.94
Grade 1 ascites (yes/no)	320/53	158/34	0.28
Grade 1–2 HE (yes/no)	37/336	38/154	0.001
Biochemical data
hsCRP (mg/L)	2.60 (1.40, 5.90)	2.80 (1.00, 10.50)	0.29
White blood cell count (*10^9/L)	4.25 ± 2.48	4.39 ± 2.68	0.54
Neutrophils percentage (%)	57.66 ± 13.48	55.88 ± 13.20	0.13
Haemoglobin (g/L)	109.72 ± 23.79	109.24 ± 25.26	0.82
Platelet count (*10^9/L)	92.21 ± 97.04	82.83 ± 58.49	0.22
International normalized ratio	1.29 ± 0.25	1.29 ± 0.25	0.99
Total bilirubin (µmol/L)	23.90 (15.10, 43.20)	27.40 (16.35, 42.25)	0.18
Albumin (g/L)	34.19 ± 5.28	33.30 ± 5.82	0.07
Alanine aminotransferase (U/L)	22.00 (16.40, 36.10)	23.20 (16.05, 34.30)	0.77
Aspartate aminotransferase (U/L)	31.60 (23.50, 44.90)	35.35 (25.50, 49.10)	0.08
Gamma-glutamyl transpeptidase (U/L)	44.50 (26.00, 98.00)	38.55 (24.35, 88.05)	0.19
Alkaline phosphatase (U/L)	103.00 (75.00, 142.00)	92.50 (69.50, 123.50)	0.01
Creatinine (µmol/L)	73.73 ± 32.29	72.08 ± 29.91	0.56
Blood urea nitrogen (mmol/L)	5.88 ± 4.29	6.72 ± 5.72	0.05
Serum sodium (mmol/L)	139.10 ± 3.59	138.72 ± 3.48	0.23
Inflammatory mediators
IFN-α (pg/mL)	0.01 (0.01, 0.47)	0.01 (0.01, 0.01)	< 0.001
IL-1β (pg/mL)	1.97 (0.01, 17.70)	0.01 (0.01, 0.01)	< 0.001
IL-5 (pg/mL)	0.08 (0.01, 0.18)	0.01 (0.01, 0.01)	< 0.001
IL-12p70 (pg/mL)	0.01 (0.01, 0.56)	0.01 (0.01, 0.01)	< 0.001
IL-8 (pg/mL)	46.00 (16.40, 58.16)	15.38 (0.01, 38.56)	< 0.001
IL-2 (pg/mL)	0.01 (0.01, 0.30)	0.01 (0.01, 0.01)	< 0.001
IL-4 (pg/mL)	0.01 (0.01, 0.33)	0.01 (0.01, 0.01)	< 0.001
IL-6 (pg/mL)	136.06 (9.09, 262.64)	6.52 (0.01, 33.73)	< 0.001
IL-10 (pg/mL)	1.94 (0.01, 22.73)	0.01 (0.01, 1.77)	< 0.001
TNF-α (pg/mL)	0.01 (0.01, 3.28)	0.01 (0.01, 0.01)	< 0.001
IFN-γ (pg/mL)	0.01 (0.01, 1.27)	0.01 (0.01, 0.01)	< 0.001
IL-17a (pg/mL)	0.01 (0.01, 0.28)	0.01 (0.01, 0.01)	< 0.001

The P value for etiology represents the overall comparison of the distribution of etiological categories between Cohort 1 and Cohort 2

AIH autoimmune hepatitis, OS overlapping syndrome, HBV hepatitis B virus, HCV hepatitis C virus, MASH metabolic dysfunction-associated steatohepatitis, PBC primary biliary cholangitis, NAD non-acute decompensation, HE hepatic encephalopathy, hsCRP high sensitivity C-reactive protein, ULN upper limit of normal, IFN interferon, IL interleukin, TNF tumor necrosis factor

Bold P values indicate statistical significance at P < 0.05

Clinical outcomes

During the 2-year follow-up period, the following outcomes were observed: In Cohort 1, 84 patients experienced AD, presenting with ascites (n = 45), acute VB (n = 16), HE grade 3 or 4 (n = 13), or BI (n = 34). 24 patients developed ACLF. 10 patients died. ACLF was the leading cause of death (7/10, 70.0%), while the cause remained undetermined for the remaining 3 patients (although they occurred after ACLF, the direct cause of death was not clearly recorded in the medical records). 103 patients required hospitalization within 24 months after the initial visit. The primary reason was AD (77/103, 74.8%). Other admission causes included treatment for suspected hepatorenal syndrome (HRS) (n = 7), hepatic venous pressure gradient (HVPG) measurement with primary prophylaxis for varices (n = 2), diagnosis of portal hypertensive gastropathy following persistent melena (n = 1), hepatic angiography for characterization of liver space-occupying lesions (n = 5), neurological consultation for persistent minimal HE (n = 2), and progressive jaundice (n = 9). In Cohort 2, 40 patients experienced AD, comprising ascites (n = 24), acute VB (n = 4), HE grade 3 or 4 (n = 9), or BI (n = 25). 13 patients developed ACLF. 5 deaths occurred, all attributable to ACLF (5/5, 100%). One patient underwent liver transplantation. 47 patients were hospitalized during follow-up. Reasons included AD (n = 37, 78.7%), suspected HRS (n = 2), suspected HCC (n = 1), HVPG measurement (n = 3), and jaundice (n = 4).

The proportions of patients with (1) hospitalization due to AD (Cohort 1 20.6% vs. Cohort 2 19.3%, P = 0.74), (2) ACLF (Cohort 1 6.4% vs. Cohort 2 6.8%, P = 0.86), and (3) death (Cohort 1 2.7% vs. Cohort 2 2.6%, P = 1.00) were similar (Fig. S1).

Primary endpoints

The 373 NAD patients in Cohort 1 were categorized into three groups based on their 24-month outcomes: Group 1 (remained clinically stable), Group 2 (progressed to AD but did not develop ACLF or die), and Group 3 (developed ACLF or died).

Levels of high-sensitivity C-reactive protein (hsCRP), IL-1β, IL-8, IL-6, IL-10, and TNF-α were compared among these groups. Significant differences across the three groups were observed for hsCRP, IL-6, and IL-10 levels (P < 0.05) (Fig. S2 A/B/D). Elevated levels of hsCRP (> 1 × ULN) and IL-6 (> 1 × ULN) were associated with an increased probability of developing AD within 24 months (Fig. 2A/B).

Fig. 2.

Acute decompensation probability by high-sensitivity C-reactive protein, interleukin-6 and interleukin-10 in Cohort 1. A. Comparison of the probability of acute decompensation among patients with different levels of high-sensitivity C-reactive protein; B. Comparison of the probability of acute decompensation between patients with different levels of interleukin-6; C. Comparison of the probability of acute decompensation between patients with different levels of interleukin-10. AD: acute decompensation; NAD: non-acute decompensation; hsCRP: high sensitivity C-reactive protein; ULN: upper limit of normal; IL: interleukin

Importantly, after adjusting for covariates using a multivariable Cox proportional hazards regression model, several factors were identified as independent predictors of AD (Table 2): international normalized ratio (INR) (Hazard Ratio, HR: 4.02; 95% CI: 1.90–8.54; P < 0.001), albumin (ALB) (HR: 0.92; 95% CI: 0.88–0.97; P = 0.003), serum sodium (Na) (HR: 0.94; 95% CI: 0.89–0.99; P = 0.03), hsCRP (> 1 × ULN: HR: 2.01; 95% CI: 1.23–3.29; P = 0.005), IL-10 (log-transformed, HR: 1.26; 95% CI: 1.16–1.36; P < 0.001), and IL-6 level (log-transformed, HR: 1.13; 95% CI: 1.01–1.27; P = 0.03).

Table 2.

Cox regression evaluation of factors associated with acute decompensation for non-acute decompensation patients

Variable	Univariable analysis		Multivariable analysis
	Hazard ratio	P-value	Hazard ratio	P-value
INR	9.86 (4.87–19.96)	< 0.001	4.02 (1.90-8.54)	< 0.001
AST	1.01 (1.00-1.01)	< 0.001	…	…
TB	1.01 (1.00-1.01)	< 0.001	…	…
ALB	0.87 (0.83–0.92)	< 0.001	0.92 (0.88-0.97)	0.003
Cr	1.01 (1.01–1.01)	< 0.001	…	…
BUN	1.06 (1.03–1.08)	< 0.001	…	…
Na	0.88 (0.84–0.93)	< 0.001	0.94 (0.89-0.99)	0.03
log_IL-1β	0.86 (0.80–0.93)	< 0.001	…	…
log_IL-12p70	0.70 (0.61–0.82)	< 0.001	…	…
log_TNF-ɑ	0.73 (0.65–0.82)	< 0.001	…	…
log_IFN-γ	0.64 (0.55–0.75)	< 0.001	…	…
log_IL-17a	0.53 (0.41–0.68)	< 0.001	…	…
hsCRP_group ˃1 × ULN	2.96 (1.92–4.57)	< 0.001	2.01 (1.23-3.29)	0.005
log_IFN-ɑ	0.74 (0.64–0.86)	< 0.001	…	…
log_IL-10	1.12 (1.05–1.20)	0.001	1.26 (1.16-1.36)	< 0.001
log_IL-5	0.75 (0.63–0.90)	0.002	…	…
log_IL-2	0.78 (0.67–0.91)	0.002	…	…
WBC	1.11 (1.04–1.19)	0.003	…	…
ALT	1.01 (1.00-1.01)	0.01	…	…
Hb	0.99 (0.98-1.00)	0.01	…	…
log_IL-6	1.12 (1.03–1.23)	0.01	1.13 (1.01-1.27)	0.03
NEUT-pct	1.02 (1.00-1.04)	0.03	…	…
log_IL-4	0.86 (0.75–0.99)	0.03	…	…

AD acute decompensation, NAD non-acute decompensation, INR international normalized ratio, AST aspartate aminotransferase, TB total bilirubin, ALB albumin, Cr creatinine, BUN blood urea nitrogen, Na serum sodium, IL interleukin, TNF tumor necrosis factor, IFN interferon, hsCRP high sensitivity C-reactive protein, ULN upper limit of normal, WBC white blood cells, ALT alanine transaminase, Hb haemoglobin, NEUT-pct neutrophils percentage

Furthermore, patients were stratified based on the occurrence of specific AD events within 24 months, creating subgroups according to the presence or absence of ascites (45 vs. 328), VB (16 vs. 357), HE (13 vs. 360), and BI (34 vs. 339). Using hsCRP and IL-6 as markers of SI and IL-10 as an indicator of ID, differences between these subgroups were assessed. Patients who experienced VB, HE, or BI within 24 months typically exhibited a concurrent state of SI and ID (Fig. S3). In contrast, for patients who developed rapid-onset ascites, levels of hsCRP and IL-6 showed strong, albeit statistically non-significant, differences, while IL-10 levels remained comparable between groups (Fig. S3).

Elevated levels of hsCRP (> 1 × ULN) and IL-10 (> 1 × ULN) were associated with an increased probability of developing BI within 24 months (Fig. S4A/C). While IL-6 levels differed between the groups, its HR for predicting the occurrence of VB, HE or ascites did not reach statistical significance (Fig. S4B). This suggests that re-evaluating and selecting more appropriate cutoff values for IL-6 specific to each endpoint may be necessary.

During the follow-up period, bacterial infection was the most common precipitating factor for ACLF (16/24, 66.7%), followed by VB (5/24, 20.8%). One case of ACLF was triggered by an HBV flare, another was attributed to herbal medicine induction, and the precipitating factor remained undetermined in one patient.

After adjusting for covariates using a multivariable Cox proportional hazards regression model, the following factors were identified as independent predictors of ACLF (Table 3): ALB (HR: 0.83; 95% CI: 0.73–0.94; P = 0.004), TNF-α (log-transformed, HR: 0.72; 95% CI: 0.58–0.90; P = 0.003), IL-10 (log-transformed, HR: 2.03; 95% CI: 1.57–2.64; P < 0.001), and IL-6 (log-transformed, HR: 2.21; 95% CI: 1.57–3.11; P < 0.001). hsCRP was not significantly associated with ACLF in the univariable Cox analysis (> 1 × ULN vs. ≤1 × ULN: HR, 1.19; 95% CI, 0.90–1.46; P = 0.123) and was therefore not entered into the multivariable model (Table 3).

Table 3.

Cox regression evaluation of factors associated with acute-on-chronic liver failure for non-acute decompensation patients

Variable	Univariable analysis		Multivariable analysis
	Hazard ratio	P-value	Hazard ratio	P-value
INR	18.37(5.46–61.77)	< 0.001	…	…
AST	1.01 (1.01–1.01)	< 0.001	…	…
TB	1.01 (1.01–1.01)	< 0.001	…	…
log_IL-6	2.42 (1.90–3.08)	< 0.001	2.21 (1.57–3.11)	< 0.001
log_IL-10	1.94 (1.64–2.31)	< 0.001	2.03 (1.57–2.64)	< 0.001
Na	0.85 (0.79–0.92)	< 0.001	…	…
NAD_Jaundice	7.90 (3.45–18.05)	< 0.001	…	…
NAD_Ascites	0.26 (0.11–0.58)	0.001	…	…
ALT	1.01 (1.00-1.01)	0.005	…	…
log_TNF-ɑ	0.74 (0.59–0.93)	0.009	0.72 (0.58–0.90)	0.003
ALB	0.90 (0.82–0.98)	0.01	0.83 (0.73–0.94)	0.004
log_IFN-γ	0.74 (0.57–0.94)	0.02	…	…
hsCRP_group ˃1 × ULN	1.19 (0.90–1.46)	0.12	…	…
log_IL-17ɑ	0.68 (0.48–0.97)	0.03	…	…
WBC	1.14 (1.01–1.28)	0.03	…	…
lactate	1.80 (1.03–3.12)	0.04	…	…

ACLF acute-on-chronic liver failure, NAD non-acute decompensation, INR international normalized ratio, AST aspartate aminotransferase, TB total bilirubin, IL interleukin, Na serum sodium, ALT alanine transaminase, TNF tumor necrosis factor, ALB albumin, IFN interferon, hsCRP high sensitivity C-reactive protein, ULN upper limit of normal, WBC white blood cells

The probability of developing ACLF differed significantly across the three hsCRP strata (< 1 × ULN, 1–3 × ULN, and > 3 × ULN), with the highest probability observed in the > 3 × ULN group. Elevated IL-6 (> 1 × ULN) and IL-10 (> 1 × ULN) were also associated with a higher probability of developing ACLF within 24 months (Fig. 3).

Fig. 3.

Acute-on-chronic liver failure probability by high-sensitivity C-reactive protein, interleukin-6 and interleukin-10 in Cohort 1. A. Comparison of the probability of acute-on-chronic liver failure among patients with different levels of high-sensitivity C-reactive protein; B. Comparison of the probability of acute-on-chronic liver failure between patients with different levels of interleukin-6; C. Comparison of the probability of acute-on-chronic liver failure between patients with different levels of interleukin-10. ACLF: acute-on-chronic liver failure; NAD: non-acute decompensation; hsCRP: high sensitivity C-reactive protein; ULN: upper limit of normal; IL: interleukin

During the 24-month follow-up, 10 deaths occurred in Cohort 1. Figure 4 shows that elevated IL-10 was associated with a higher probability of liver-transplantation-free mortality. Univariable Cox analyses of liver-transplantation-free mortality for the full set of candidate variables are presented in Table S1. In the penalized Firth Cox model restricted to the two prespecified biomarkers, log IL-6 (HR: 1.15, 95% penalized-likelihood-CI: 0.80–1.59; P = 0.053) was not statistically significant, whereas log IL-10 (HR: 2.15, 95% penalized-likelihood-CI: 1.61–2.91; P < 0.001) was associated with liver-transplantation-free mortality.

Fig. 4.

Liver-Transplantation-free mortality by high-sensitivity C-reactive protein, interleukin-6 and interleukin-10 in Cohort 1. A. Comparison of the probability of liver-transplantation-free mortality among patients with different levels of high-sensitivity C-reactive protein; B. Comparison of the probability of liver-transplantation-free mortality between patients with different levels of interleukin-6; C. Comparison of the probability of liver-transplantation-free mortality between patients with different levels of interleukin-10. LT: liver transplantation; NAD: non-acute decompensation; hsCRP: high sensitivity C-reactive protein; ULN: upper limit of normal; IL: interleukin

Using IL-6 as a marker for SI and IL-10 as a marker for ID, patients were stratified into three groups: no SI, SI without ID, and SI with ID, to assess their prognoses. The analysis revealed that the presence of ID (indicated by a significant increase in IL-10) stratified the prognostic outcomes among patients with SI.

In Cohort 1, compared to the no SI group, both the SI without ID group (HR: 1.75; 95% CI: 1.00-3.05; P = 0.06) and the SI with ID group (HR: 1.86; 95% CI: 1.05–3.28; P = 0.03) had a higher probability of developing AD. However, the presence of ID did not further increase the risk of AD compared to SI alone (i.e., the risk increase was driven by SI, not augmented by ID) (Fig. S5A). Regarding ACLF, compared to the no SI group, the SI without ID group showed a comparable risk. In stark contrast, the SI with ID group exhibited a markedly higher risk of ACLF compared to the SI without ID group (HR: 28.02; 95% CI: 12.54–62.62; P < 0.001) (Fig. S5B). Notably, all 10 patients who died belonged to the SI with ID subgroup.

Validation Cohort 2

In Cohort 2, compared to the no SI group, both the SI without ID group (HR: 10.36; 95% CI: 4.17–25.71; P < 0.001) and the SI with ID group (HR: 36.72; 95% CI: 9.62–140.20; P < 0.001) had a significantly higher probability of developing AD. Notably, the coexistence of ID further increased the risk of AD compared to SI alone (HR: 4.63; 95% CI: 1.97–10.89; P < 0.001) (Fig. 5A).

Fig. 5.

Acute decompensation and acute-on-chronic liver failure, liver-transplantation-free mortality and transitions by different phenotypes in Cohort 2. A. Comparison of the probability of acute decompensation among patients with no SI, SI without ID, SI with ID; B. Comparison of the probability of acute-on-chronic liver failure among patients with no SI, SI without ID, SI with ID; C. Comparison of the probability of liver-transplantation-free mortality among patients with no SI, SI without ID, SI with ID; D. Sankey diagram illustrating the transitions between different disease states across the study timeline. AD: acute decompensation; ACLF: acute-on-chronic liver failure; NAD: non-acute decompensation; LT: liver transplantation; SI: systemic inflammation; ID: immunodeficiency; HE: hepatic encephalopathy

Regarding ACLF, the risk in the SI without ID group was comparable to that in the no SI group. However, the presence of ID dramatically increased the risk of ACLF compared to the SI without ID group (HR: 39.78; 95% CI: 10.97–144.30; P < 0.001) (Fig. 5B).

Similarly, mortality in the SI without ID group was comparable to that in the no SI group. In contrast, the coexistence of ID significantly elevated the mortality risk compared to the SI without ID group (HR: 13.82; 95% CI: 2.26–84.57; P = 0.002) (Fig. 5C).

Fig. 5D and Fig. S6 present a Sankey diagram illustrating the transitions between different disease states across the study timeline.

Discussion

In this ambispective cohort study of cirrhotic outpatients with NAD, inflammatory mediators measured in a clinically stable, non-hospitalized state predicted disease progression over the subsequent two years. Markers of systemic inflammation, including hsCRP and IL-6, together with IL-10, were associated with subsequent clinical deterioration. IL-6 and IL-10 independently predicted AD and ACLF, whereas IL-10 was associated with liver-transplantation-free mortality in exploratory analyses. Most importantly, combining a pro-inflammatory signal (IL-6) with an anti-inflammatory signal (IL-10) defined a discrete SI + ID phenotype that carried a disproportionately high risk of ACLF and liver-transplantation-free mortality, and this relationship was reproduced in the prospective validation cohort.

These observations are biologically coherent within the framework of cirrhosis-associated immune dysfunction, in which a state of persistent, maladaptive systemic inflammation coexists with a progressive immunodeficiency that compromises effective host defence and predisposes to infection and organ failure [18]. Within this paradigm, IL-6 reports the pro-inflammatory arm and IL-10 the counter-regulatory, immunodeficiency arm; the clinical trajectory is determined less by either signal alone than by the balance between them [18, 27]. Our data are consistent with this concept: an isolated rise in IL-6 (SI alone) was sufficient to raise the risk of AD, but it was the concurrent elevation of IL-10 (SI + ID) that distinguished the patients destined for ACLF and liver-transplantation-free mortality.

The prognostic implication of stratifying patients by inflammatory phenotype was therefore substantial. In both cohorts, the SI-without-ID group behaved similarly to patients without SI with respect to ACLF, whereas the SI + ID group showed a many-fold increase in ACLF risk, and nearly all deaths occurred within this subgroup. This pattern indicates that a high anti-inflammatory tone arising on a background of ongoing inflammation is not a benign counter-regulatory epiphenomenon but a marker of failing immune homeostasis—the immunodeficiency component of cirrhosis-associated immune dysfunction that leaves patients vulnerable to infection-driven organ failure [18].

The information carried by these mediators also varied by the type of decompensating event. Bacterial infection, the dominant precipitant of ACLF in our cohort, was best captured by hsCRP and by IL-10, whereas IL-6 did not reach significance for isolated ascites, variceal bleeding, or HE. This is consistent with the view that infection and infection-related organ failure are the clinical events most tightly coupled to the inflammation–immunodeficiency axis [18, 27–28], and it suggests that endpoint-specific cutoffs—rather than a single threshold—may be required if IL-6 is to be used to anticipate individual complications.

Several baseline characteristics differed significantly between the retrospective Cohort 1 and the prospective Cohort 2, including the distribution of cirrhosis etiologies, the frequency of grade 1–2 HE, hsCRP strata, and the absolute concentrations of most cytokines. These differences are largely explained by how Cohort 1 was assembled: the 12-cytokine panel was ordered selectively on clinical grounds rather than systematically in all outpatients, which enriched the retrospective cohort for patients with higher measured cytokine levels and accounts for the markedly higher baseline medians of IL-6, IL-10, and IL-8. Importantly, these baseline differences do not appear to have materially affected the central findings. The clinically relevant event rates were comparable across cohorts and the prognostic relationship of interest was not merely preserved but reproduced with an even larger effect size in the prospective cohort. Because the SI/ID classification is anchored to the upper limit of normal rather than to absolute concentrations, the dichotomized phenotype remained robust despite the inter-cohort differences in raw cytokine values. The consistency of the SI + ID signal across two cohorts that differ in etiology, encephalopathy frequency, and absolute cytokine levels argues that it reflects a generalizable biological state rather than a feature of any single cohort’s composition.

Taken together, these data point to a practical, mechanism-based approach to risk assessment in NAD. Measuring two pro-inflammatory mediators (IL-6 and hsCRP), which reflect systemic inflammation, together with an anti-inflammatory mediator (IL-10), which reflects immunodeficiency, may be a key strategy for detecting impending progression—specifically hospitalization for AD, ACLF, and liver-transplantation-free mortality—in cirrhotic patients following NAD. The associations first identified in the retrospective Cohort 1 were validated in the prospective Cohort 2, increasing confidence that they reflect a true biological relationship rather than chance or design-specific bias. We hope these findings will be confirmed by other centers, so that this combination of inflammatory mediators can be incorporated into routine prognostication and used to anticipate the clinical course after NAD in cirrhosis.

Our study has several strengths, including its ambispective design with an independent prospective validation cohort, a well-characterized NAD population, and a comprehensive cytokine panel. Limitations should also be acknowledged. As a retrospectively assembled cohort, Cohort 1 is subject to selection bias, because the cytokine panel was ordered on the basis of specific clinical considerations rather than in an unselected manner; this selective testing most likely underlies the baseline differences from the prospectively enrolled Cohort 2. The number of deaths was small, limiting multivariable modelling of liver-transplantation-free mortality. Cytokines were assessed at a single time point, and serial measurements might add dynamic prognostic information. We used the established upper limit of normal as the analytic cutoff, but the optimal thresholds for predicting different endpoints (AD, ACLF, liver-transplantation-free mortality) may differ and warrant dedicated calibration. Finally, the study was conducted at a single center, and external validation in cohorts of differing ethnicity and etiology is needed.

Conclusions

In conclusion, inflammatory mediators measured in cirrhotic patients with NAD were associated with subsequent clinical deterioration. IL-6 and IL-10 independently predicted the development of AD and ACLF, whereas IL-10 was associated with liver-transplantation-free mortality in exploratory analyses. The coexistence of elevated IL-6 and IL-10 identified a distinct SI-with-ID phenotype characterized by a particularly high risk of ACLF and mortality. These findings support further multicenter validation of inflammation-based risk stratification in patients with NAD.

Abbreviations

ACLF

Acute-on-chronic liver failure

AD

Acute decompensation

AIH

Autoimmune hepatitis

AKP

Alkaline phosphatase

ALB

Albumin

ALT

Alanine aminotransferase

AST

Aspartate aminotransferase

BI

Bacterial infection

BUN

Blood urea nitrogen

CI

Confidence interval

Cr

Creatinine

EASL

European Association for the Study of the Liver

Hb

Haemoglobin

HBV

Hepatitis B virus

HCC

Hepatocellular carcinoma

HCV

Hepatitis C virus

HE

Hepatic encephalopathy

HR

Hazard ratio

HRS

Hepatorenal syndrome

hsCRP

High-sensitivity C-reactive protein

HVPG

Hepatic venous pressure gradient

ID

Immunodeficiency

IFN

Interferon

IL

Interleukin

INR

International normalized ratio

LT

Liver transplantation

MASH

Metabolic dysfunction-associated steatohepatitis

MELD

Model for end-stage liver disease

Na

Serum sodium

NAD

Non-acute decompensation

NEUT-pct

Neutrophil percentage

OS

Overlapping syndrome

PBC

Primary biliary cholangitis

PT

Prothrombin time

SI

Systemic inflammation

TB

Total bilirubin

TNF-α

Tumor necrosis factor-α

ULN

Upper limit of normal

VB

Variceal bleeding

WBC

White blood cell

Authors’ contributions

Conceptualization (XW, JS, RL, YY), data curation, biological sample collection (JS, XW, JG, FY, SC, HR, HZ, XF, ZL), formal analysis (XW, JS, JG, RL), funding acquisition (XW, YY), investigation (XW, JS, JG, XD, FY, SC, HR, HZ, XF, ZL, DY), project administration (CZ, QX, RL, YY), resources (CZ, QX, YY), validation (XW, JS, JG, RL), visualization (XW), writing—original draft (XW, JS, JG), and writing—review & editing (RL, YY). All authors have approved the final version and publication of the manuscript.

Funding

This work was supported by the Innovation Center for Infectious Disease of Jiangsu Province (No. CXZX202232) and the SEU Innovation Capability Enhancement Plan for Doctoral Students (CXJH_SEU 24230).

Data availability

Data in this article can be requested by contacting Yongfeng Yang (E-mail: yyf1997@163.com) and Ruiqi Li (E-mail: 627364595@qq.com). Data anonymized to protect patient characteristics will be provided for studies whose aims and objectives align with the study protocols. Only proposals where data will be used for statistical and scientific studies will be considered. Data will be shared through Excel electronic forms after the signing of a data access agreement.

Declarations

Ethics approval and consent to participate

The study was performed in accordance with the ethical principles of the Declaration of Helsinki (as revised in 2013) and the Declaration of Istanbul, and was approved by the Ethics Committee of the Second Hospital of Nanjing (Approval No. 2023-LY-ky-056). For the retrospective cohort (Cohort 1), the requirement for informed consent was waived by the Ethics Committee due to the retrospective design and the use of anonymized data. Written informed consent was obtained from all participants in the prospective cohort (Cohort 2).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.