Abstract
Background
This study aimed to develop and validate a novel nomogram that incorporates the neutrophil percentage-to-albumin ratio (NPAR) for predicting 30-day and 90-day unfavorable outcomes in acute-on-chronic liver failure (ACLF) patients.
Methods
We retrospectively enrolled 641 ACLF patients from the First Affiliated Hospital of Anhui Medical University between January 2017 and June 2024, randomly assigning them to training (n = 448) and validation (n = 193) cohorts. Univariate Cox regression, the Least Absolute Shrinkage and Selection Operator (LASSO) regression, and multivariate Cox analysis identified seven independent predictors: bacterial infection, hepatic encephalopathy, age, prothrombin time (PT), total bilirubin (TBIL), lymphocyte count, and NPAR.
Results
The nomogram demonstrated superior predictive performance in the training cohort, with area under the curve (AUC) values of 0.874 (95% CI 0.838–0.910) and 0.877 (95% CI 0.845–0.909) for 30-day and 90-day outcomes, respectively, and a concordance index (C-index) of 0.825 (95% CI 0.796–0.853), significantly outperforming the MELD, MELD-Na, CLIF-C ACLFs and COSSH–ACLF II scores. Calibration curves showed strong concordance between predicted and observed survival probabilities, and decision curve analysis confirmed broad clinical applicability. We have also validated the predictive value of the new model for ACLF in a validation cohort (n = 193). Compared with COSSH–ACLF II, the model exhibited significant improvements in net reclassification index (NRI) and integrated discrimination improvement (IDI) (P < 0.001). Risk stratification effectively categorized patients into low-, intermediate-, and high-risk groups, revealing varied survival rates. In addition, a user-friendly dynamic web-based calculator was developed.
Conclusions
This nomogram that integrates inflammatory and nutritional indicators provides a high-precision tool for short-term prognosis assessment in ACLF patients and is expected to guide clinical management.
Supplementary Information
The online version contains supplementary material available at 10.1186/s40001-026-03988-8.
Keywords: Acute‐on‐chronic liver failure, Neutrophil percentage-to-albumin ratio, Prognosis, Nomogram, Risk factor
Introduction
Acute-on-chronic liver failure (ACLF) is a distinct clinical syndrome arising from underlying chronic liver disease, often triggered by factors, such as infection. Its development is typically marked by intense systemic inflammation, multiorgan failure, and adverse short-term outcomes [1, 2]. Reported mortality data under protocol-driven monitoring reveal 28-day (38.9%) and 90-day (55.7%) fatality rates in ACLF, reflecting its dismal prognosis [3]. Although the artificial liver support system (ALSS) provides temporary hepatic support for some patients and can improve short-term survival rates by 10–30%, it serves as a bridge therapy and cannot fundamentally reverse liver failure. Consequently, approximately 25–40% of treated patients still succumb to the disease, underscoring its persistently grave prognosis [4, 5]. Timely recognition of patients at elevated risk enables clinicians to optimize therapeutic interventions without delay, including considering liver transplantation [6]. Researchers worldwide have developed multiple prognostic scoring models for ACLF based on distinct diagnostic criteria, including the CLIF-C ACLF score [7], COSSH–ACLF score [8], COSSH–ACLF II score [9], APASL–AARC score [10], and MELD score with its derivative systems [11]. Although these tools are highly effective and valuable for assessing disease severity and prognosis, the evaluation of novel biomarkers for predictive value remains indispensable.
The pathophysiological mechanisms underlying ACLF remain incompletely elucidated. However, the progression of ACLF is consistently characterized by systemic inflammatory responses, oxidative stress, and immune dysregulation. These pathophysiological alterations create a complex cascade that exacerbates hepatic injury and multi-organ dysfunction, ultimately influencing patient prognosis. In particular, the hyperactivation of innate immune cells triggers the excessive release of pro-inflammatory cytokines, which further intensify tissue damage and systemic inflammation [12]. This intensity may escalate in response to precipitating factors and subside with ACLF amelioration [13, 14]. A recent study has identified that elevated blood cytokine levels, in conjunction with neutrophilia and lymphocytopenia, are substantially linked to a higher incidence of ACLF [15]. Immunometabolic dysregulation of circulating immune cells, especially monocytes, has been shown in a different study using single-cell RNA sequencing to significantly influence prognostic trajectories in ACLF [16]. Furthermore, the role of gut bacterial translocation in ACLF is widely recognized; however, the underlying active regulatory mechanisms remain unclear [17]. Emerging evidence reveals that the host can actively regulate pathogenic bacterial colonization through specific signaling axes, such as the KRAS-mediated miRNA3655/SURF6/IRF7/IFNβ pathway in colorectal cancer [18]. This provides a novel paradigm for understanding how the host actively shapes its microecology in ACLF, thereby influencing prognosis. malnutrition frequently complicates end-stage liver disease, leading to a vicious cycle through bidirectional interactions with progressive hepatic functional deterioration [19, 20]. A recent randomized controlled trial (RCT) demonstrated that intensive ambulatory nutritional support substantially enhances survival outcomes in patients with alcohol-related ACLF [21]. The study’s findings suggest that tailored nutritional strategies should be integrated into standard ACLF management protocols. Therefore, monitoring inflammatory markers and assessing nutritional status are critical for evaluating the severity of hepatic injury and predicting therapeutic efficacy. Certain inflammatory and nutritional markers, including C-reactive protein-to-albumin ratio (CAR) and neutrophil percentage-to-albumin ratio (NPAR), have been linked to unfavorable outcomes in both ACLF and colorectal cancer liver metastases [22–24]. However, the complex interplay between inflammatory scores, nutritional status, and ACLF progression requires further exploration and validation.
Compared with conventional predictive models, nomograms demonstrate superior advantages, including enhanced visual interpretability, user-friendly applicability, and intuitive representation of multifactorial influences on outcomes, thereby facilitating timely and accurate clinical decision-making by clinicians. Furthermore, real-world studies provide crucial evidence for analyzing the therapeutic value and risks within complex clinical scenarios [25]. Herein, we developed a novel nomogram integrating inflammatory biomarkers to predict adverse outcomes in ACLF patients.
Materials and methods
Patients
854 patients with ACLF were the subject of a retrospective cohort analysis at the First Affiliated Hospital of Anhui Medical University’s Department of Infectious Diseases between January 2017 and June 2024. Inclusion criteria were: (1) ACLF diagnosis per the Asia–Pacific Association for the Study of the Liver (APASL) guidelines [26], specifically defined as acute hepatic injury in patients with underlying chronic non-cirrhotic liver disease or compensated cirrhosis, characterized by elevated total bilirubin (TBil ≥ 5 mg/dL; 1 mg/dL = 17.1 umol/L) and coagulopathy [international normalized ratio (INR) ≥ 1.5 or prothrombin activity (PTA) ≤ 40%], accompanied by ascites and/or hepatic encephalopathy (HE) within 4 weeks, and is associated with high 28-day mortality and (2) age ≥ 18 years. Exclusion criteria comprised: (1) age < 18 years; (2) history of liver transplantation or undergoing liver transplantation during follow-up; (3) hepatocellular carcinoma or other malignant tumors; (4) pregnancy; (5) incomplete clinical documentation; and (6) presence of other severe comorbidities that may interfere with prognosis prediction, such as: myocardial infarction, stroke, or chronic end-stage renal disease (requiring long-term dialysis). Ultimately, a total of 641 patients were enrolled. The data set was separated at random into training and validation cohorts at a ratio of 7:3 using R functions, ensuring balanced event distribution between groups (Fig. 1). All patients received standardized treatment per current clinical guidelines after admission, including antiviral/antibiotic therapy, coagulation management, and nutritional support. Eligibility for ALSS therapy was assessed by evaluating disease severity and obtaining patient consent.
Fig. 1.
Flow diagram of study profile
The study followed the principles of the Declaration of Helsinki and received approval from the Hospital Ethics Review Committee (Approval No. PJ2022-07–18). This study rigorously adhered to the TRIPOD statement, employing a multivariable predictive model to assess individualized prognosis and diagnostic outcomes.
Data collection
The hospital’s electronic medical record system served as the source of the data for this investigation. We extracted patients demographic details, the underlying etiology of chronic liver disease (viral, alcoholic, cholestatic, and other), complications (ascites, gastrointestinal bleeding, bacterial infection, hepatic encephalopathy, and hepatorenal syndrome), and collected their first post-admission laboratory findings. The laboratory indicators included prothrombin time (PT), INR, fibrinogen (FIB), D-dimer, total bilirubin (TBIL), total bile acid (TBA), albumin (ALB), alanine aminotransferase (ALT), aspartate aminotransferase (AST), alkaline phosphatase (ALP), creatinine (Cr), serum urea, blood sodium (Na), C-reactive protein (CRP), blood routine, alpha-fetoprotein (AFP), and other test results. The diagnostic criteria for bacterial infection have been summarized in the supplementary materials. In addition, we calculated several inflammatory ratio indicators using the corresponding formulas. The details are as follows:
NLR = Neutrophil-to-lymphocyte ratio;
MLR = monocyte-to-lymphocyte ratio;
PLR = Platelet-to-lymphocyte ratio;
CAR = C-reactive protein-to-albumin ratio;
NPAR = neutrophil percentage × 100/albumin (g/dL);
SII = (platelets × neutrophils)/lymphocytes;
SIRI = (monocytes × neutrophils)/lymphocytes.
Based on electronic medical records and 90-day follow-up, we determined patients’ survival status and developed a clinical prognostic model to evaluate 30-day and 90-day outcomes.
Statistical analysis
Statistical analyses were performed as follows: normally distributed continuous variables underwent unpaired t tests, non-normal data were assessed with Mann–Whitney U tests, and categorical variable comparisons were examined using chi-square tests. For the selection of model variables, we initially conducted univariate Cox regression analysis to screen factors linked to adverse outcomes in ACLF patients. Significant variables (P < 0.05) then underwent least absolute shrinkage and selection operator (LASSO) regression, with 11 predictors identified through tenfold cross-validation using the lambda.1se threshold. Finally, a multivariate Cox stepwise backward regression based on the Akaike Information Criterion (AIC) was performed to select the final variables, which were then incorporated into the nomogram [27].
In addition, we assessed variable collinearity via the Variance Inflation Factor (VIF) and evaluated model discrimination using the area under the receiver operating characteristic curve (AUC–ROC) and the Concordance Index (C-index). Calibration (500 bootstraps) curves examined prediction accuracy. Decision Curve Analysis (DCA) measured clinical benefit. Predictive improvements over COSSH–ACLF II model were measured using the Net Reclassification Index (NRI) and the Integrated Discrimination Improvement (IDI). X-tile software identified optimal nomogram cutoffs for 30- and 90-day ACLF prognosis [28]. Patients were stratified into low, intermediate, and high risk. Kaplan–Meier and log-rank tests assessed survival differences. P < 0.05 was considered statistically significant.
Results
Patients and clinical characteristics
Using a 7:3 ratio, 641 eligible patients were randomly assigned to a training cohort (n = 448) and a validation cohort (n = 193). The demographic and clinical features of ACLF patients are summarized in Table 1. In the entire cohort, the median age was 50.00 years [interquartile range (IQR): 42.00–59.00], with 78% of the patients being male and 81.75% having viral hepatitis as the etiology. The mortality rates without liver transplantation at 30 days and 90 days were 22% and 35.57%, respectively, highlighting the substantial short-term risk. No significant differences were observed when comparing the training and validation cohorts in their baseline variables encompassing age, sex, etiology, comorbidities, underlying chronic diseases, laboratory indicators, severity scores, inflammatory markers, and mortality rates without liver transplantation. This indicates that the training and validation cohorts were essentially homogeneous and comparable, thereby supporting the robustness of subsequent predictive modeling.
Table 1.
Baseline characteristics of the patients
| Variables | Total (n = 641) | Training cohort (n = 448) |
Validation cohort (n = 193) |
P value |
|---|---|---|---|---|
| Age (years) | 50.00 (42.00, 59.00) | 51.00 (43.00, 60.00) | 49.00 (38.00, 59.00) | 0.243 |
| Male, n (%) | 500 (78.00) | 344 (76.79) | 156 (80.83) | 0.257 |
| Cause of chronic liver disease, n (%) | ||||
| Viral | 524 (81.75) | 371 (82.81) | 153 (79.27) | 0.442 |
| Alcoholic | 73 (11.39) | 48 (10.71) | 25 (12.95) | |
| Cholestatic | 34 (5.30) | 24 (5.36) | 10 (5.18) | |
| Other | 10 (1.56) | 5 (1.12) | 5 (2.59) | |
| Complications, n (%) | ||||
| Ascites | 371 (57.88) | 264 (58.93) | 107 (55.44) | 0.412 |
| Gastrointestinal bleeding | 24 (3.74) | 17 (3.79) | 7 (3.63) | 0.918 |
| Bacterial infection | 205 (31.98) | 145 (32.37) | 60 (31.09) | 0.750 |
| Hepatic encephalopathy | 96 (14.98) | 63 (14.06) | 33 (17.10) | 0.323 |
| Hepatorenal syndrome | 55 (8.58) | 37 (8.26) | 18 (9.33) | 0.658 |
| Common chronic diseases, n (%) | ||||
| Hypertension | 61 (9.52) | 38 (8.48) | 23 (11.92) | 0.174 |
| Diabetes | 68 (10.61) | 50 (11.16) | 18 (9.33) | 0.489 |
| Laboratory parameters | ||||
| PT (s) | 22.50 (19.60, 27.50) | 22.55 (19.60, 27.13) | 22.50 (19.40, 27.90) | 0.698 |
| INR | 1.99 (1.65, 2.50) | 1.99 (1.66, 2.50) | 1.99 (1.64, 2.51) | 0.888 |
| FIB (g/L) | 1.65 (1.38, 1.88) | 1.65 (1.42, 1.89) | 1.62 (1.37, 1.87) | 0.520 |
| D dimer (ug/L) | 2.10 (1.12, 2.99) | 2.13 (1.13, 3.16) | 2.07 (1.11, 2.79) | 0.284 |
| TBIL (umol/L) | 285.80 (210.71, 376.10) | 283.40 (211.30, 379.92) | 289.60 (210.50, 363.13) | 0.932 |
| TBA (umol/L) | 182.90 (121.20, 254.40) | 177.80 (119.27, 253.63) | 189.20 (124.00, 257.50) | 0.329 |
| ALB (g/L) | 31.10 (28.00, 34.50) | 31.00 (27.98, 34.40) | 31.50 (28.10, 34.60) | 0.719 |
| ALT (U/L) | 225.00 (71.00, 693.00) | 217.50 (71.00, 686.75) | 232.00 (71.90, 733.70) | 0.737 |
| AST (U/L) | 189.00 (106.00, 498.00) | 191.50 (104.00, 517.25) | 184.00 (107.00, 414.00) | 0.622 |
| ALP (U/L) | 145.00 (116.00, 178.00) | 145.00 (114.00, 177.00) | 148.00 (119.40, 184.00) | 0.223 |
| Cr (umol/L) | 62.00 (51.00, 76.20) | 62.00 (51.00, 76.45) | 62.60 (51.00, 75.50) | 0.989 |
| Serum urea (mmol/L) | 4.32 (3.18, 6.56) | 4.33 (3.12, 6.75) | 4.29 (3.28, 6.28) | 0.755 |
| Na (mmol/L) | 135.90 (132.90, 138.60) | 135.95 (132.67, 138.60) | 135.50 (133.20, 138.40) | 0.732 |
| Ca (mmol/L) | 2.12 (2.01, 2.22) | 2.12 (2.01, 2.23) | 2.10 (1.98, 2.20) | 0.138 |
| K (mmol/L) | 3.85 (3.50, 4.28) | 3.85 (3.51, 4.24) | 3.85 (3.43, 4.34) | 0.74 |
| CRP (mg/L) | 8.90 (4.67, 15.20) | 8.98 (4.62, 15.50) | 8.70 (5.11, 13.75) | 0.828 |
| WBC (109/L) | 6.74 (5.00, 9.08) | 6.74 (5.12, 9.02) | 6.84 (4.65, 9.61) | 0.872 |
| Neutrophil (109/L) | 4.79 (3.26, 6.74) | 4.76 (3.38, 6.69) | 4.88 (3.02, 7.09) | 0.713 |
| Lymphocyte (109/L) | 1.09 (0.77, 1.53) | 1.11 (0.78, 1.52) | 1.04 (0.75, 1.58) | 0.732 |
| Monocyte (109/L) | 0.58 (0.39, 0.81) | 0.58 (0.39, 0.80) | 0.58 (0.37, 0.85) | 0.848 |
| Haemoglobin (g/L) | 122.00 (107.00, 136.00) | 120.00 (105.00, 135.00) | 124.00 (109.00, 139.00) | 0.182 |
| Platelet (109/L) | 103.00 (69.00, 139.00) | 104.00 (69.00, 139.00) | 99.00 (67.00, 140.00) | 0.560 |
| Blood ammonia (umol/L) | 55.80 (39.00, 72.00) | 54.00 (38.50, 69.00) | 58.00 (42.00, 76.00) | 0.144 |
| AFP (ng/mL) | 23.00 (9.50, 52.60) | 24.00 (9.88, 54.00) | 21.00 (9.00, 49.00) | 0.452 |
| Severity scores | ||||
| MELD score | 22.30 (19.66, 26.24) | 22.42 (19.75, 26.37) | 22.25 (19.32, 25.79) | 0.475 |
| MELD-Na score | 23.67 (20.39, 30.05) | 23.91 (20.42, 30.00) |
23.28 (20.31, 31.27) |
0.584 |
| CLIF-C ACLFs | 29.79 (24.46, 36.39) | 29.64 (24.45, 36.59) | 30.03 (24.38, 36.00) | 0.985 |
| COSSH‐ACLF II score | 7.25 (6.55, 7.99) | 7.27 (6.55, 8.01) | 7.20 (6.54, 7.97) | 0.613 |
| Inflammatory indicators | ||||
| NLR | 4.19 (2.90, 6.81) | 4.11 (2.95, 6.63) | 4.36 (2.85, 6.96) | 0.578 |
| MLR | 0.54 (0.35, 0.80) | 0.54 (0.35, 0.79) | 0.54 (0.35, 0.81) | 0.893 |
| PLR | 97.98 (64.68, 144.00) | 97.17 (66.20, 146.35) | 100.14 (61.74, 139.89) | 0.908 |
| CAR | 0.28 (0.15, 0.48) | 0.29 (0.14, 0.50) | 0.27 (0.17, 0.46) | 0.799 |
| NPAR | 22.85 (19.87, 26.05) | 22.86 (19.68, 26.08) | 22.83 (20.06, 25.99) | 0.992 |
| SII | 438.36 (240.65, 759.17) | 428.30 (242.32, 759.63) | 452.40 (233.71, 754.36) | 0.969 |
| SIRI | 2.51 (1.30, 4.70) | 2.51 (1.30, 4.63) | 2.51 (1.31, 5.21) | 0.864 |
| LT-free mortality, n (%) | ||||
| 30 days | 141 (22.00) | 101 (22.54) | 40 (20.73) | 0.610 |
| 90 days | 228 (35.57) | 159 (35.49) | 69 (35.75) | 0.950 |
PT prothrombin time, INR international normalized ratio, FIB fibrinogen, TBIL total bilirubin, TBA total bile acids, ALB albumin, ALT alanine transaminase, AST aspartate aminotransferase, ALP alkaline phosphatase, Cr creatinine, Na sodium, Ca calcium, K potassium, CRP C-reactive protein, AFP alpha‐fetoprotein, MELD model for end‐stage liver disease, CLIF-C ACLFs CLIF–Consortium Acute-on-Chronic Liver Failure score, COSSH Chinese Group on the Study of Severe Hepatitis, NLR neutrophil-to-lymphocyte ratio, MLR monocyte-to-lymphocyte ratio, PLR platelet-to-lymphocyte ratio, CAR C-reactive protein-to-albumin ratio, NPAR neutrophil percentage × 100/albumin, SII systemic inflammation index, SIRI systemic inflammatory response index I
Variable selection for the nomogram
Table S1 shows a comprehensive comparison of demographic characteristics, clinical features, and other outcomes among ACLF patients with different prognoses in the training cohort. Compared with survivors, non-survivors were older and had higher rates of comorbidities (bacterial infections, hepatic encephalopathy, and hepatorenal syndrome). Baseline assessments revealed more severe liver function impairment, characterized by elevated TBIL and TBA, reduced ALB, and worse coagulation profiles, including prolonged PT and increased INR in non-survivors. In addition, several comprehensive inflammatory markers, including NLR, MLR, NPAR, and SII, were found to be elevated. First, we conducted univariate Cox regression on 44 variables and identified significant associations for age, bacterial infections, hepatic encephalopathy, hepatorenal syndrome, PT, INR, TBIL, TBA, ALB, CR, NA, CRP, neutrophil, lymphocyte, hemoglobin, blood ammonia, AFP, NLR, MLR, CAR, NPAR, SII, and SIRI (P < 0.05). Details are shown in Table S2. To further refine variable selection and address multicollinearity, we performed LASSO regression on the identified variables, selecting 11 predictive factors (Fig. S1). Finally, multivariable Cox backward stepwise regression revealed that bacterial infections (hazard ratio [HR]: 1.480, 95% confidence interval [CI] 1.060–2.065), hepatic encephalopathy (HR: 1.972, 95% CI 1.349–2.882), AGE (HR: 1.027, 95% CI 1.013–1.040), PT (HR: 1.077, 95% CI 1.061–1.093), TBIL (HR: 1.002, 95% CI 1.001–1.003), lymphocyte (HR: 0.601, 95% CI 0.428–0.844), and NAPR (HR: 1.102, 95% CI 1.066–1.138) had significant impacts on ACLF prognosis (Fig. 2). Moreover, all VIF values were < 4, further confirming that multicollinearity was within an acceptable range (Table S3).
Fig. 2.
Results of the multivariate Cox regression analysis
Development and validation of the nomogram
Utilizing multivariate COX regression results, a nomogram depicted in Fig. 3 was developed to predict ACLF patients’ 30- and 90-day survival probabilities. This nomogram achieves prediction by assigning weights to each variable. In particular, each variable’s value correlates to a certain score on the scale. For example, the nomogram score for a patient with the following characteristics was calculated as follows: bacterial infection (yes) = 10 points; hepatic encephalopathy (no) = 0 points; age (60 years) = 30 points; PT (25 s) = 17 points; TBIL (600 µmol/L) = 27 points; lymphocyte count (1.5 × 10⁹/L) = 44 points; NPAR (25) = 41 points. The sum of these points yielded a total score of 169, which corresponds to predicted 30-day and 90-day survival probabilities of approximately 62% and 37%, respectively. In addition, we have created a web-based calculator (https://death-predict.shinyapps.io/DynNomapp/) that allows clinicians to automatically and instantly calculate the total score and display the predictive survival probabilities by simply inputting the corresponding patient parameters. A screenshot of the interface is provided in Supplementary Fig. 2. In the training cohort, the model demonstrated AUC values of 0.874 (95% CI 0.838–0.910) for 30-day and 0.877 (95% CI 0.845–0.909) for 90-day prognosis. The validation cohort’s corresponding AUC values were 0.870 (95% CI 0.811–0.929) and 0.857 (95% CI 0.803–0.911), which were all noticeably superior to MELD, MELD-Na, CLIF-C ACLFs, and COSSH–ACLF II scores (Fig. 4). Furthermore, the training cohort’s C-index was 0.825 (95% CI 0.796–0.853), while the validation cohort’s was 0.810 (95% CI 0.762–0.857) (Table 2), further confirming the model’s robust discriminative ability. The calibration curves illustrated that the 30-day and 90-day survival rates observed and those predicted by the model agreed quite well (Fig. 5). DCA indicated favorable practical application over an extensive spectrum of threshold probabilities (Fig. 6). The nomogram demonstrated marked improvement compared to the COSSH–ACLF II scoring system, as demonstrated by robust NRI and IDI metrics. In the training cohort, NRI reached 0.554 (95% CI 0.122–0.554, P < 0.001) for 30-day and 0.547 (95% CI 0.391–0.627, P < 0.001) for 90-day mortality prediction. Corresponding IDIs were 0.143 (95% CI 0.017–0.205, P < 0.001) and 0.179 (95% CI 0.113–0.240, P < 0.001), respectively. This predictive superiority persisted in the validation cohort (Table 2), confirming the model’s enhanced discriminative capacity for short-term clinical outcomes in ACLF populations.
Fig. 3.
Nomogram for 90-day survival prediction of ACLF
Fig. 4.
ROC curves comparing nomogram, MELD, MELD-NA, CLIF-C ACLFs, and COSSH–ACLF II scores for 30- and 90-day overall survival in ACLF patients: training cohort (A, B) and validation cohort (C, D)
Table 2.
C-index, NRI, and IDI of the nomogram and COSSH–ACLF II in survival prediction for ACLF patients
| Index | Training cohort | Validation cohort | ||||
|---|---|---|---|---|---|---|
| Estimate | 95%CI | P value | Estimate | 95%CI | P value | |
| NRI (vs. COSSH‐ACLF II score) | ||||||
| For 30-day OS | 0.554 | 0.122–0.554 | < 0.001 | 0.462 | 0.087–0.545 | < 0.001 |
| For 90-day OS | 0.547 | 0.391–0.627 | < 0.001 | 0.465 | 0.246–0.580 | < 0.001 |
| IDI (vs. COSSH‐ACLF II score) | ||||||
| For 30-day OS | 0.143 | 0.017–0.205 | < 0.001 | 0.149 | 0.027–0.207 | < 0.001 |
| For 90-day OS | 0.179 | 0.113–0.240 | < 0.001 | 0.165 | 0.086–0.246 | < 0.001 |
| C-index | ||||||
| COSSH‐ACLF II score | 0.747 | 0.711–0.783 | 0.754 | 0.698–0.810 | ||
| Nomogram | 0.825 | 0.796–0.853 | 0.81 | 0.762–0.857 | ||
| Change | 0.078 | 0.050–0.106 | < 0.001 | 0.056 | 0.015–0.096 | 0.007 |
OS overall survival
Fig. 5.
Calibration curves of the nomogram for 30-day and 90-day overall survival in ACLF: training cohort (A, B) and validation cohort (C, D). The ideal reference line is shown by the grey dashed line, which is when the observed survival rate and the anticipated probability coincide. The greater the proximity of the solid lines to the grey dashed line, the higher the accuracy of the model in predicting overall survival
Fig. 6.
Decision curve analysis of the nomogram for 30-day and 90-day overall survival in ACLF: training cohort (A, B), validation cohort (C, D)
Risk stratification based on the nomogram
We used the total score from the nomogram to determine two optimal cutoff values (148 and 177) with the xtile software, stratifying ACLF patients into three tiers: low risk (< 148), medium risk (148–177), and high risk (≥ 177). In the training cohort, the 30-day survival probabilities for the low-, medium-, and high-risk groups were 96.50%, 59.56%, and 32.73%, respectively, while the 90-day survival probabilities were 88.72%, 39.71%, and 10.91%, respectively. This descending trend was confirmed in the validation cohort, where the corresponding 30-day survival probabilities for the three risk groups were 93.75%, 70.69%, and 30.43%, and the 90-day survival probabilities were 83.93%, 44.83%, and 17.39%, respectively. Cox proportional hazards regression analysis showed significantly elevated mortality risks for the intermediate-risk (HR = 7.79, 95% CI 5.09–11.93, P < 0.001) and high-risk groups (HR = 19.63, 95% CI 12.29–31.34, P < 0.001) compared to the low-risk group (Fig. 7A). Kaplan–Meier curves further demonstrated that the three groups’ 90-day survival rates differed significantly (log-rank test P < 0.001). In addition, the risk stratification method exhibited excellent consistency in the validation cohort (Fig. 7B).
Fig. 7.
Risk stratification of the nomogram. A Kaplan–Meier curves of 90-day survival in the training cohort stratified by the nomogram classification rule (low/medium/high risk: nomogram < 148/148–177/≥ 177). B Kaplan–Meier curves of 90-day survival in the validation cohort stratified by the nomogram classification rule (low/medium/high risk: nomogram < 148/148–177/≥ 177)
Discussion
ACLF signifies a pivotal clinical deterioration in chronic liver disease, triggered by acute insults and frequently progressing to multi-organ failure within a narrow timeframe [29]. Given the high mortality and limited curative interventions associated with ACLF, early recognition of critically ill individuals and timely prioritization for transplantation are essential to optimize survival opportunities.
This study developed a novel and concise nomogram based on NPAR, integrating inflammatory and nutritional indicators, to predict 30- and 90-day prognoses in ACLF patients. Internal validation demonstrated that the nomogram achieved significantly higher ROC–AUC values compared to conventional models, including MELD, MELD-Na, CLIF-C ACLFs, and COSSH–ACLF II. Furthermore, well-fitted calibration curves and substantial net clinical benefit confirmed its clinical utility. When compared to COSSH–ACLF II, the nomogram showed superior performance with higher NRI, IDI, and C-index values, indicating strong predictive accuracy. Kaplan–Meier curves effectively stratified patients into low-, intermediate-, and high-risk subgroups, providing robust support for clinical decision-making.
In this research, our nomogram incorporated multiple factors that significantly impact the outcomes in ACLF patients, involving advanced age, bacterial infection, hepatic encephalopathy, TBIL, PT, lymphocyte count, and NPAR. Notably, advanced age, as a strong predictor of adverse prognosis, aligns with the risk stratification system established in the COSSH–ACLF II study [9]. This finding supports the pathogenic pathways of immunosenescence and reduced organ reserve that underlie the development of ACLF in older patients. In addition, this study confirmed that bacterial infection is a standalone predictor for higher death rates among individuals with ACLF (HR = 1.480, 95% CI 1.060–2.065), consistent with findings from the European CLIF Consortium [30] and a global multicenter cohort study [14]. Bacterial infections may lead to systemic inflammatory response and multiorgan dysfunction in ACLF patients, thereby significantly increasing their mortality risk. Studies indicate that bacterial infections may activate the TLR4/MyD88 pathway within Kupffer cells by interacting with pathogen-associated molecular patterns (PAMPs), thereby inducing the release of pro-inflammatory cytokines, such as IL-6 and TNF-α, with serum levels increasing by 4–6 times. These factors directly impair hepatocyte mitochondrial function and inhibit regeneration [31]. Systemic inflammation may set off a vicious cycle of immune dysregulation, which in turn increases the risk of secondary infection and serves as a separate predictor of death [32]. This study indicates that lymphopenia (HR: 0.601, 95% CI 0.428–0.844) is closely linked to adverse outcomes in ACLF patients, a finding that is consistent with a recent clinical study [15]. Moreover, a basic research study reported that B- and T-lymphocyte attenuator (BTLA) may induce CD4⁺T cell exhaustion, thereby increasing the risk of infection and mortality in patients with ACLF [33]. These findings collectively underscore the pivotal role of immune dysfunction in the pathogenesis of ACLF.
Recent studies have demonstrated that cytokine and inflammatory ratio indices are associated with high mortality risk in ACLF patients [34, 35]. These indicators, including MLR, provide a quantitative measure of systemic inflammatory burden driven by diverse immune-activating factors, such as viral hepatitis or chronic alcohol abuse. They offer an effective means of predicting ACLF, a syndrome characterized by acute hepatic decompensation and associated multiorgan dysfunction. A study of 347 HBV–ACLF patients showed that MLR and NLR are significantly related to 90-day mortality (OR 6.738, 95% CI 3.188–14.240, P < 0.001). The AUC for predicting outcomes using combined MLR and NLR analysis was 0.694. While these indicators are easy to obtain, focusing solely on inflammation based on routine blood test parameters has limitations [34]. In particular, such measures may not fully account for the complex interplay between liver function, immune regulatory pathways, and ongoing tissue injury. Moreover, single-dimensional inflammatory biomarkers can be confounded by acute bacterial infections or transient fluctuations in circulating leukocyte populations. However, this study innovatively introduces the neutrophil percentage-to-albumin ratio (NPAR), a comprehensive marker of inflammation and nutrition. It reflects neutrophil-driven inflammatory intensity and hepatic synthetic reserve (indicated by albumin synthesis). Nutritional status, a critical factor affecting patient’s quality of life and survival, significantly increases mortality risk by weakening intestinal barrier function, exacerbating endotoxemia, and suppressing tissue repair. However, it is often overlooked in traditional ACLF severity scoring systems. This study innovatively establishes the first multidimensional prognostic assessment system incorporating nutritional metabolic parameters, addressing this deficiency. Through this integrated model, clinicians can better identify high-risk patients who may benefit from aggressive nutritional interventions, ultimately aiming to reduce complications and improve outcomes in ACLF management.
This study confirms that hepatic encephalopathy (HE) is a separate risk factor for ACLF patients prognosis (HR: 1.972, 95% CI 1.349–2.882). Its complex pathogenesis may involve neuroinflammatory cascades driven by inflammatory mediators (e.g., TNF-α and IL-6), ammonia metabolic disorders, and oxidative stress-induced blood–brain barrier damage [36]. This finding aligns with the conclusions of the COSSH–ACLF II scoring system [9], indicating that HE not only reflects cerebral decompensation but also serves as an early warning sign of multi-organ failure. Notably, HE patients often exhibit mitochondrial dysfunction in astrocytes, further exacerbating metabolic crises and neuronal apoptosis [37]. Clinically, these changes correlate with elevated intracranial pressure, impaired neurotransmitter balance, and increased susceptibility to secondary infections. Therefore, early identification and targeted management of HE—such as ammonia-lowering therapies, anti-inflammatory interventions, and intracranial pressure monitoring—are crucial to improving outcomes in ACLF.
Meanwhile, the nomogram model in this study integrates widely recognized ACLF prognostic factors, such as TBIL and PT [38, 39]. In summary, these included indicators are readily available in clinical practice, with relatively low testing costs, facilitating their application across various healthcare settings. By comprehensively incorporating these risk factors that are significantly associated with adverse outcomes in ACLF, the model not only enhances prognostic accuracy but also balances practical feasibility and cost-effectiveness, thereby potentially offering more practical and effective decision support for the clinical management of ACLF patients.
However, this study has several limitations. First, as a single-center retrospective study, the results may be influenced by the specific patient population and local treatment practices. The lack of an independent external validation cohort also limits the model’s generalizability and external validity. Second, this study only validated the model’s predictive value for short-term outcomes at 30 and 90 days, without assessing its performance for longer term endpoints (such as 180-day survival rate) or incorporating patient-reported outcomes like quality of life. Furthermore, the assessment of nutritional status relied solely on serum albumin and did not include more comprehensive nutritional parameters, such as prealbumin or body composition, which may not fully capture the integrated impact of nutritional status on prognosis. Finally, this study did not systematically analyze the acute insults of ACLF, and future prospective studies are needed to further clarify their relationship with prognosis.
Conclusion
Based on the following identified risk factors: bacterial infection, hepatic encephalopathy, age, PT, TBIL, lymphocyte count, and NPAR, we developed and validated a novel nomogram. It demonstrated favorable predictive performance for short-term outcomes in ACLF patients during internal validation, suggesting its potential for clinical application. However, further validation in larger scale, multi-center studies is warranted given the single-center retrospective nature of this study.
Supplementary Information
Acknowledgements
Not applicable.
Author contributions
Z.L. conceived the study, gathered and analyzed the data, created tables and figures, and drafted the initial manuscript. Z.W. and J.Y. collected the data. S.D. reviewed the statistical methods. C.Z. and J.L. supervised the study. Y.G. conceived the study and reviewed the final version. The article’s final version was approved by all authors who contributed to it.
Funding
This work was supported by Natural Science Foundation of China (No. 82370608), Anhui Provincial Natural Science Foundation (No. 2208085MH204), Anhui Education Department Natural Science Foundation (No. 2022AH040160), and Anhui Provincial Special Fund for Clinical Medical Research Transformation (No. 202304295107020040).
Data availability
The original contributions presented in the study are included in the article/supplementary material, the raw data have been uploaded to Related files. Further inquiries can be directed to the corresponding author.
Declarations
Ethics approval and consent to participate
The studies involving humans were approved by the Ethics Committee of the First Affiliated Hospital of Anhui Medical University (Approval No: PJ2022-07-18). The studies were conducted in accordance with the local legislation and institutional requirements. The ethics committee/institutional review board waived the requirement of written informed consent for participation from the participants or the participants’legal guardians/next of kin, given that it is a retrospective study with anonymized data analysis.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Arroyo V, Moreau R, Jalan R. Acute-on-chronic liver failure. N Engl J Med. 2020;382(22):2137–45. [DOI] [PubMed] [Google Scholar]
- 2.Kulkarni AV, Sarin SK. Acute-on-chronic liver failure – steps towards harmonization of the definition! J Hepatol. 2024;81(2):360–6. [DOI] [PubMed] [Google Scholar]
- 3.Aggarwal A, Biswas S, Arora U, Vaishnav M, Shenoy A, Swaroop S, et al. Definitions, etiologies, and outcomes of acute on chronic liver failure: a systematic review and meta-analysis. Clin Gastroenterol Hepatol. 2024;22(11):2199-2210.e25. [DOI] [PubMed] [Google Scholar]
- 4.Yao J, Li S, Zhou L, Luo L, Yuan L, Duan Z, et al. Therapeutic effect of double plasma molecular adsorption system and sequential half‐dose plasma exchange in patients with HBV‐related acute‐on‐chronic liver failure. J Clin Apher. 2019;34(4):392–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Maiwall R, Bajpai M, Singh A, Agarwal T, Kumar G, Bharadwaj A, et al. Standard-volume plasma exchange improves outcomes in patients with acute liver failure: a randomized controlled trial. Clin Gastroenterol Hepatol. 2022;20(4):e831–54. [DOI] [PubMed] [Google Scholar]
- 6.Xu W, Zhu S, Yang L, Li Z, Wu L, Zhang Y, et al. Safety and efficacy of double plasma molecular adsorption system with sequential low‐volume plasma exchange in intermediate‐stage hepatitis B virus‐related acute‐on‐chronic liver failure. J Med Virol. 2023;95(3):e28650. [DOI] [PubMed] [Google Scholar]
- 7.Jalan R, Saliba F, Pavesi M, Amoros A, Moreau R, Ginès P, et al. Development and validation of a prognostic score to predict mortality in patients with acute-on-chronic liver failure. J Hepatol. 2014;61(5):1038–47. [DOI] [PubMed] [Google Scholar]
- 8.Wu T, Li J, Shao L, Xin J, Jiang L, Zhou Q, et al. Development of diagnostic criteria and a prognostic score for hepatitis B virus-related acute-on-chronic liver failure. Gut. 2018;67(12):2181. [DOI] [PubMed] [Google Scholar]
- 9.Li J, Liang X, You S, Feng T, Zhou X, Zhu B, et al. Development and validation of a new prognostic score for hepatitis B virus-related acute-on-chronic liver failure. J Hepatol. 2021;75(5):1104–15. [DOI] [PubMed] [Google Scholar]
- 10.Choudhury A, Jindal A, Maiwall R, Sharma MK, Sharma BC, Pamecha V, et al. Liver failure determines the outcome in patients of acute-on-chronic liver failure (ACLF): comparison of APASL ACLF research consortium (AARC) and CLIF-SOFA models. Hepatol Int. 2017;11(5):461–71. [DOI] [PubMed] [Google Scholar]
- 11.Kim WR, Mannalithara A, Heimbach JK, Kamath PS, Asrani SK, Biggins SW, et al. MELD 3.0: the model for end-stage liver disease updated for the modern era. Gastroenterology. 2021;161(6):1887-1895.e4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Trebicka J, Fernandez J, Papp M, Caraceni P, Laleman W, Gambino C, et al. The PREDICT study uncovers three clinical courses of acutely decompensated cirrhosis that have distinct pathophysiology. J Hepatol. 2020;73(4):842–54. [DOI] [PubMed] [Google Scholar]
- 13.Arroyo V, Angeli P, Moreau R, Jalan R, Clària J, Trebicka J, et al. The systemic inflammation hypothesis: towards a new paradigm of acute decompensation and multiorgan failure in cirrhosis. J Hepatol. 2021;74(3):670–85. [DOI] [PubMed] [Google Scholar]
- 14.Wong F, Piano S, Singh V, Bartoletti M, Maiwall R, Alessandria C, et al. Clinical features and evolution of bacterial infection-related acute-on-chronic liver failure. J Hepatol. 2021;74(2):330–9. [DOI] [PubMed] [Google Scholar]
- 15.Cao Z, Yao Y, Cai M, Zhang C, Liu Y, Xin H, et al. Blood markers for type-1, -2, and -3 inflammation are associated with severity of acutely decompensated cirrhosis. J Hepatol. 2025;82(5):836–50. [DOI] [PubMed] [Google Scholar]
- 16.Feio-Azevedo R, Boesch M, Radenkovic S, Van Melkebeke L, Smets L, Wallays M, et al. Distinct immunometabolic signatures in circulating immune cells define disease outcome in acute-on-chronic liver failure. Hepatology. 2025;81(2):509–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Trebicka J, Hernaez R, Shawcross DL, Gerbes AL. Recent advances in the prevention and treatment of decompensated cirrhosis and acute-on-chronic liver failure (ACLF) and the role of biomarkers. Gut. 2024;73(6):1015–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Chen Y, Liu S, Tan S, Zheng Y, Chen Y, Yang C, et al. KRAS mutations promote the intratumoral colonization of enterotoxigenic bacteroides fragilis in colorectal cancer through the regulation of the miRNA3655/SURF6/IRF7/IFNβ axis. Gut Microbes. 2024;16(1):2423043. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Song R, Wang X, Li Z, Wu H, Tan J, Tan J, et al. ALTA: a simple nutritional prognostic score for patients with hepatitis B virus-related acute-on-chronic liver failure. Front Nutr. 2024;9(11):1370025. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Chen Y, Jia H, Ye R, Zhou Z, Chen W, Zheng M, et al. Low visceral adipose tissue predicts the outcome of neoadjuvant chemotherapy for colorectal liver metastases: a multicentre real‐world study. J Cachexia Sarcopenia Muscle. 2025;16(2):e13785. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Giri P, Taneja S, Sahni N, Bhujade H, Padhi BK, Karki T, et al. Outpatient Intensive Nutrition Therapy Improves Survival and Frailty in Males With Alcohol-related ACLF - Randomized Controlled Trial. Clin Gastroenterol Hepatol. 2024. S1542356524009595. [DOI] [PubMed]
- 22.Zhu ZY, Yan LJ. Prognostic value of dynamic changes in C-reactive protein to albumin ratio in patients with acute-on-chronic liver failure. World J Hepatol. 2025;17(9):110652. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Zheng T, Zheng Y, Zhou Z, Ye R, Jia H, Chen W, et al. C-reactive protein-to-albumin ratio predicts outcome of neoadjuvant chemotherapy for colorectal liver metastases: a multicenter prediction model. Surgery. 2025;182:109334. [DOI] [PubMed] [Google Scholar]
- 24.Ma Y, Du L, Bai L, Tang H. Association between neutrophil percentage to albumin ratio and short term prognosis of acute on chronic liver failure treated with artificial liver support system. Sci Rep. 2025;15(1):5042. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Chen Y, Zheng Y, Wu J, Ye R, Jia H, Zhou Z, et al. Neoadjuvant chemotherapy combined with bevacizumab for resectable colorectal liver metastasis with risk factors for recurrence: a multicenter real-world study. Ther Adv Med Oncol. 2025;17:17588359251328457. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Sarin SK, Choudhury A, Sharma MK, Maiwall R, Al Mahtab M, Rahman S, et al. Acute-on-chronic liver failure: consensus recommendations of the Asian Pacific association for the study of the liver (APASL): an update. Hepatol Int. 2019;13(4):353–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Balachandran VP, Gonen M, Smith JJ, DeMatteo RP. Nomograms in oncology: more than meets the eye. Lancet Oncol. 2015;16(4):e173–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Camp RL, Dolled-Filhart M, Rimm DL. X-Tile: a new bio-informatics tool for biomarker assessment and outcome-based cut-point optimization. Cancer Res. 2004;10(21):7252–9. [DOI] [PubMed] [Google Scholar]
- 29.Chen T, Chen G, Wang G, Treeprasertsuk S, Lesmana CRA, Lin HC, et al. Expert consensus on the diagnosis and treatment of end-stage liver disease complicated by infections. Hepatol Int. 2024;18(3):817–32. [DOI] [PubMed] [Google Scholar]
- 30.Fernández J, Acevedo J, Wiest R, Gustot T, Amoros A, Deulofeu C, et al. Bacterial and fungal infections in acute-on-chronic liver failure: prevalence, characteristics and impact on prognosis. Gut. 2018;67(10):1870. [DOI] [PubMed] [Google Scholar]
- 31.Xiang X, Feng D, Hwang S, Ren T, Wang X, Trojnar E, et al. Interleukin-22 ameliorates acute-on-chronic liver failure by reprogramming impaired regeneration pathways in mice. J Hepatol. 2020;72(4):736–45. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Khamri W, Abeles RD, Hou TZ, Anderson AE, El-Masry A, Triantafyllou E, et al. Increased expression of cytotoxic T-lymphocyte−associated protein 4 by T cells, induced by B7 in sera, reduces adaptive immunity in patients with acute liver failure. Gastroenterology. 2017;153(1):263-276.e8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Yu X, Yang F, Shen Z, Zhang Y, Sun J, Qiu C, et al. BTLA contributes to acute-on-chronic liver failure infection and mortality through CD4+ T-cell exhaustion. Nat Commun. 2024;15(1):1835. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Wang N, He S, Zheng Y, Wang L. The value of NLR versus MLR in the short-term prognostic assessment of HBV-related acute-on-chronic liver failure. Int Immunopharmacol. 2023;1(121):110489. [DOI] [PubMed] [Google Scholar]
- 35.Zhu B, Gao F, Li Y, Shi K, Hou Y, Chen J, et al. Serum cytokine and chemokine profiles and disease prognosis in hepatitis B virus-related acute-on-chronic liver failure. Front Immunol. 2023;27(14):1133656. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Gallego-Durán R, Hadjihambi A, Ampuero J, Rose CF, Jalan R, Romero-Gómez M. Ammonia-induced stress response in liver disease progression and hepatic encephalopathy. Nat Rev Gastroenterol Hepatol. 2024;21(11):774–91. [DOI] [PubMed] [Google Scholar]
- 37.Cheon SY, Kim MY, Kim J, Kim EJ, Kam EH, Cho I, et al. Hyperammonemia induces microglial NLRP3 inflammasome activation via mitochondrial oxidative stress in hepatic encephalopathy. Biomed J. 2023;46(5):100593. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Xue R, Yang J, Wu J, Wang Z, Meng Q. Novel prognostic models for predicting the 180-day outcome for patients with hepatitis-B virus-related acute-on-chronic liver failure. J Clin Transl Hepatol. 2021;9:514. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Xiao L, Chen J, Zhao S, Zhoudi W, He K, Qian X, et al. The 90‐day survival threshold: a pivotal determinant of long‐term prognosis in HBV‐ACLF patients – insights from a prospective longitudinal cohort study. Adv Sci. 2024;11(16):2304381. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The original contributions presented in the study are included in the article/supplementary material, the raw data have been uploaded to Related files. Further inquiries can be directed to the corresponding author.