ABSTRACT
Background and Aim
Recent studies show that patients with hepatitis B virus (HBV)‐related decompensated cirrhosis who achieve recompensation can still experience further decompensation, suggesting that recompensation status can change over time. This study aimed to classify patterns of recompensation and charaterize the clinical differences among these subgroups.
Methods
Eligible patients with HBV‐related decompensated cirrhosis were enrolled from two cohorts. Clinical characteristics and complications were assessed every 6 months for up to 5 years following their first episode of decompensation. Recompensation was defined according to the Baveno VII criteria and further categorised as stable (no subsequent decompensation) or unstable (recurrent decompensation or recompensation following multiple decompensation episodes).
Results
A total of 378 patients were included; 294 (77.8%) achieved recompensation, while 84 (22.2%) did not. After a median follow‐up of 5.3 years (IQR 4.4–5.8), recompensated patients were classified into stable recompensation (202/378, 53.4%) and unstable recompensation (92/378, 24.3%). The 5 year rate of hepatocellular carcinoma (HCC) or all‐cause mortality was higher in the unstable group than the stable group (14.7% vs. 10.1%, p = 0.038), yet remained lower than in patients with ongoing decompensation. Liver function improvement was intermediate in the unstable group compared with the stable recompensation and ongoing decompensation. Logistic regression yielded the highest accuracy for predicting recompensation (AUROC = 0.884), while support vector machine algorithms best predicted stable recompensation (AUROC = 0.911).
Conclusion
Recompensation is not a uniform condition and should be further subclassified. Unstable recompensation is a distinct state with poorer survival than stable recompensation, yet better outcomes than ongoing decompensation.
Keywords: cirrhosis, decompensation, hepatitis B virus, recompensation
Recompensation is a heterogeneous and dynamic state in HBV‐related decompensated cirrhosis; The prognosis for patients with unstable recompensation was found to be worse compared to those with stable recompensation, but still better than patients without recompensation; Subclassification of recompensation has significant implications for prognosis and tailored management.
1. Introduction
Chronic hepatitis B virus (HBV) infection is one of the most common causes of liver cirrhosis, which can be classified into two distinct clinical stages of compensated and decompensated cirrhosis according to the presence of clinical complications [1, 2, 3]. Decompensated cirrhosis significantly worsens patient prognosis, leading to a marked increase in mortality risk [4]. Antiviral therapy has been shown to reduce the risk of disease progression, ultimately reducing the occurrence of death and hepatocellular carcinoma (HCC) [5, 6].
Recently, the Baveno VII consensus referred to the reversal of decompensation as recompensation [2]. Three essential criteria are required to define recompensation cirrhosis: effective control of aetiology; resolution of decompensated events for at least 1 year; and stable improvement of liver function. Subsequently, increasing evidence has demonstrated that 53.0%–68.4% of patients with HBV‐related decompensated cirrhosis can achieve recompensation, and these patients showed significantly improved long‐term prognosis [7, 8, 9]. However, despite achieving recompensation, some patients still progressed and developed new decompensating events, illustrating recompensation may not be a completely stable stage [8]. Transient recompensation highlights the limitations of the current definition. It raises the question about its impact on patient outcomes. It also suggests that recompensation may exist in different forms, distinct from fully stable recompensation.
In this study, we aim to explore a new subtype of recompensation‐unstable recompensation and investigate how it differs from classical recompensation. Unstable recompensation was defined as patients either achieving recompensation after multiple episodes of decompensation or initially achieving recompensation but subsequently relapsing into decompensation. By identifying these new patterns of recompensation, we hope to refine patient management strategies for those with decompensated cirrhosis, enabling more accurate prognostic assessments and targeted interventions.
2. Methods
2.1. Study Population and Study Design
Patients with HBV‐related decompensated cirrhosis were enrolled from two cohorts. Cohort 1 (treatment‐naïve) was from a retrospective, observational, multi‐center study initiated in August 2021, which included eligible patients experiencing ascites and/or variceal bleeding as their first episode of decompensating complications and initiating anti‐HBV therapy within 3 months. Cohort 2 (on‐treatment) was from a prospective, observational, multi‐center study initiated in June 2012, which included patients with HBV‐related compensated cirrhosis [8, 10].
Inclusion criteria were (a) patients initiated and maintained continuous first‐line nucleos(t)ide analogues (NAs)‐based antiviral therapy and (b) patients with HBV‐related decompensated cirrhosis who presented with ascites and/or variceal bleeding as their first episode of decompensating, either at the initiation of or during antiviral therapy. Cirrhosis was diagnosed based on radiological features and/or clinical laboratory parameters. HBV was identified as the aetiology of cirrhosis based on the history of chronic HBV infection according to clinical laboratory parameters (i.e., positive hepatitis B surface antigen).
Exclusion criteria were (a) progression to HCC or death after the first decompensated events without recompensation; (b) no follow‐up for at least 1 year after decompensation; (c) absence of three consecutive follow‐up data; (d) loss to follow‐up or follow‐up duration less than 5 years; (e) underwent TIPS during the follow‐up period.
The studies were conducted following the ethical principles of the Declaration of Helsinki and its subsequent amendments, approved by the local ethics committee of the Beijing Friendship Hospital, Capital Medical University (2021‐P2‐224‐01, 2016‐P2‐022‐01), and each participating center. The studies were registered at clinicaltrials.gov (NCT05086536, NCT010720238). For retrospectively enrolled patients, written informed consent was waived, while all prospectively recruited patients provided written informed consent.
2.2. Follow‐Up and Clinical Evaluation
In the prospective cohort, patients were evaluated at 6‐month intervals during routine visits to the hepatology outpatient clinic. Clinical data during episodes of decompensation were also thoroughly documented.
In the retrospective cohort, baseline and follow‐up clinical data were collected every 6 months. All relevant data were retrieved from the electronic medical record system and recorded using the standard case report by study investigators in each participant center.
Clinical data included demographics, laboratory parameters, upper endoscopy, and imaging assessment. All enrolled patients were treated with NAs. Laboratory parameters were collected for each patient, including complete blood count, liver function tests (alanine aminotransferase [ALT], aspartate aminotransferase [AST], total bilirubin [TBIL], albumin [ALB]), alpha‐fetoprotein (AFP), coagulation profiles, and virologic tests. All patients were followed up for at least 5 years or until death/liver transplant or loss of follow‐up after the first episode of decompensated events.
2.3. Clinical Outcome
The primary outcome of this study was composite liver‐related endpoints, including further decompensation, HCC, or liver‐related death. Hepatic decompensation was defined as ascites > 1 cm on imaging examination, overt hepatic encephalopathy (West Haven grade ≥ II), and variceal bleeding [2]. Further decompensation was defined according to Baveno VII criteria without considering jaundice [2]. The diagnosis of HCC was based on histological and/or radiological findings. Death‐related information was acquired from the patient's electronic health records.
The secondary outcome was achieving recompensation. Recompensation was defined by (a) the removal of the aetiology (i.e., persistent suppression of HBV viral replication); (b) resolution of ascites and HE, defined by persistent discontinuation of diuretics and HE treatment, and no variceal bleeding event within the last 12 months; (c) stable improvement in liver function (Child‐Pugh grade A and/or Model for End‐Stage Liver Disease (MELD) scores <10; or TBIL <34 and ALB >35 in the absence of Child‐Pugh and MELD data).
2.4. Dataset Preparation and Model Training
The demographic and clinical features at baseline and 0.5 years were used to identify potential predictive factors. Missing data were employed by multiple imputations. Subsequently, 70% of the population was randomly allocated to a training set for model development to predict recompensation and stable recompensation, while the remaining 30% was used for testing and evaluation. Following rigorous performance evaluation on the testing set, the optimal predictive model was selected based on its predictive performance. Then, SHapley Additive exPlanations (SHAP) was applied to assess the importance of each clinical feature and interpret feature contributions.
2.5. Statistical Analysis
Continuous variables were reported as the median and interquartile range (IQR) and compared using Student's t‐test or Mann–Whitney U test. Categorical variables were reported as the number and corresponding percentage and compared via chi‐squared tests. Kaplan–Meier analysis was used to assess long‐term prognosis between different patterns of patients. Repeated measures variance analysis was used to assess dynamic change of clinical parameters during 1 year after first decompensation.
Univariate logistic regression (LR) evaluated parameters associated with recompensation. The missing data were imputed using the K‐Nearest Neighbours (KNN) algorithm. Predictive models were developed based on LR and machine learning algorithms encompassing Support Vector Machine (SVM). The area under the receiver operating characteristic curve (AUROC) was used to assess the discriminative ability of the models.
All analyses were conducted using SPSS 17.0 (SPSS Inc., Chicago, IL) and GraphPad Prism 8.0 (GraphPad Software Inc., La Jolla, CA, USA). Machine learning algorithms were performed by R programming software 4.2.2 (R‐project, New Zealand, AU). In this study, p‐values < 0.05 were considered significant. Sankey plot summarised the clinical course of patients since the first decompensation events using Origin 2021 (OriginLab, Northampton, MA).
3. Results
3.1. Patient Characteristics of the Study Cohort
A total of 550 patients with HBV‐related decompensated cirrhosis were screened for this study. Of them, 24 patients were excluded for progressing to HCC or death immediately after first decompensation without experiencing further decompensation or achieving recompensation. Additionally, 17 patients were excluded due to lack of follow‐up for at least 1 year after decompensation, 11 patients were excluded for missing three consecutive follow‐up data, and 16 patients were excluded for performing TIPS during the follow‐up period. Additionally, 104 patients were excluded due to loss of follow‐up or duration of follow‐up < 5 years. Finally, 378 were included in the study cohort (Figure 1).
FIGURE 1.
Flow chart of patient enrollment.
Demographic and baseline clinical characteristics of all enrolled patients are shown in Table 1. Patients were predominantly male (69.0%), with a median age of 58.1 years. Among 378 patients, the median follow‐up period from the first decompensation was 5.3 years (4.4, 5.8), and the most common cause of initial decompensated events was ascites in 276 patients (73.0%), followed by ≥ 2 concomitant events (15.3%) and VB (11.6%). The median platelet (PLT) count was 71.0 × 109/L and the patients with thrombocytopenia (PLT < 150 × 109/L) account for the majority (91.0%). Moreover, the median Child‐Pugh score was 8.0 points, and the MELD score was 13.0 points (Table 1).
TABLE 1.
The baseline clinical characteristics as well as the comparison of these characteristics among different subgroups.
| Characteristic | Overall | With recompensation (n = 294) | p a | Without recompensation (n = 84) | p b | ||
|---|---|---|---|---|---|---|---|
| Total | Stable recompensation (n = 202) | Unstable recompensation (n = 92) | Ongoing decompensation (n = 84) | ||||
| Age (years) | 58.1 (51.3, 65.9) | 57.7 (51.2, 65.4) | 56.9 (51.2, 65.8) | 60.1 (52.6, 64.8) | 0.579 | 59.9 (51.3, 69.4) | 0.226 |
| Male (n, %) | 261 (69.0%) | 204 (69.4%) | 136 (67.3%) | 68 (73.9%) | 0.256 | 57 (67.9%) | 0.789 |
| Initial decompensation (n, %) | |||||||
| Ascites | 276 (73.0%) | 219 (74.5%) | 160 (79.2%) | 59 (64.1%) | 0.013 | 57 (67.9%) | 0.319 |
| Bleeding | 44 (11.6%) | 34 (11.6%) | 20 (9.9%) | 14 (15.2%) | 10 (11.9%) | ||
| ≥ 2 concomitant events | 58 (15.3%) | 41 (13.9%) | 22 (10.9%) | 19 (20.7%) | 17 (20.2%) | ||
| WBC (109/L) | 3.9 (2.7, 5.2) | 4.1 (2.9, 5.3) | 4.3 (3.0, 5.5) | 3.7 (2.7, 4.9) | 0.030 | 3.1 (2.4, 4.6) | 0.008 |
| PLT (109/L) | 71.0 (50.0, 100.0) | 73.0 (53.0, 102.5) | 78.5 (57.0, 108.0) | 64.0 (43.5, 85.5) | < 0.001 | 59.0 (40.0, 88.5) | 0.003 |
| PLT < 150 (109/L) | 334 (91.0%) | 271 (92.2%) | 184 (91.1%) | 87 (94.6%) | 0.303 | 73 (86.9%) | 0.136 |
| ALT (IU/L) | 46.0 (28.0, 119.6) | 54.0 (29.8, 144.9) | 64.0 (30.0, 201.0) | 43.0 (29.0, 92.4) | 0.009 | 35.0 (26.0, 52.0) | < 0.001 |
| AST (IU/L) | 57.0 (35.0, 126.0) | 66.0 (36.6, 136.3) | 72.0 (38.7, 168.0) | 52.0 (32.4, 107.5) | 0.010 | 46.0 (32.0, 61.0) | < 0.001 |
| ALB (g/L) | 31.5 (26.9, 35.2) | 31.3 (26.6, 35.3) | 31.3 (26.9, 35.9) | 31.0 (25.8, 34.4) | 0.288 | 32.0 (28.2, 35.0) | 0.227 |
| TBlL (μmol/L) | 32.8 (20.1, 62.5) | 35.0 (20.6, 73.7) | 38.3 (21.5, 86.4) | 30.1 (18.8, 47.9) | 0.007 | 26.5 (17.3, 40.3) | 0.003 |
| INR | 1.4 (1.2, 1.6) | 1.4 (1.2, 1.6) | 1.4 (1.2, 1.6) | 1.4 (1.3, 1.7) | 0.852 | 1.4 (1.2, 1.5) | 0118 |
| AFP (ng/mL) | 14.3 (3.3, 97.6) | 24.4 (4.1, 115.1) | 34.3 (4.7, 149.3) | 12.1 (2.9, 61.4) | 0.028 | 5.3 (2.0, 14.5) | < 0.001 |
| Child‐Pugh score | 8.0 (7.0, 10.0) | 9.0 (7.0, 10.0) | 9.0 (7.0, 10.0) | 8.0 (7.0, 10.0) | 0.268 | 8.0 (7.0, 9.0) | 0.017 |
| MELD score | 13.0 (10.0, 17.0) | 13.0 (10.0, 17.0) | 14.0 (10.0, 18.0) | 13.0 (10.0, 16.0) | 0.045 | 12.0 (10.0, 14.0) | 0.032 |
| Spleen thickness (cm) | 1.2 (1.1, 1.4) | 4.8 (4.1, 5.5) | 4.6 (4.0, 5.3) | 5.3 (4.3, 5.9) | < 0.001 | 5.5 (4.9, 6.3) | < 0.001 |
| Portal width (cm) | 5.0 (4.3, 5.7) | 1.2 (1.1, 1.3) | 1.2 (1.1, 1.3) | 1.3 (1.2, 1.4) | < 0.001 | 1.3 (1.1, 1.4) | 0.007 |
Note: Data presented as median (IQR) or number (%); Quantitative differences were analysed by the Student's t‐test or Mann–Whitney U test for continuous variables, and by chi‐squared test or the Fisher exact test for categorical variables, as appropriate. The bold font represent statistically significant differences (P < 0.05).
Abbreviations: AFP, alpha‐fetoprotein; ALB, albumin; ALT, alanine transaminase; AST, aspartate aminotransferase; Child‐Pugh, Child‐Turcotte‐Pugh; INR, international normalised ratio; MELD, Model for End‐Stage Liver Disease; PLT, platelet; TBIL, total bilirubin; WBC, white blood cell count.
Stable recompensation vs. unstable recompensation.
With recompensation vs. without recompensation.
3.2. Achieving Recompensation Can Significantly Improve Long‐Term Prognosis
During the 5‐year follow‐up, 294 patients (77.8%) achieved recompensation as defined by the Baveno VII criteria, while 84 patients (22.2%) experienced repeated decompensation without recompensation. Approximately 74.8% (220/294) of patients achieved recompensation within the first 2 years following their initial decompensation, while 25.2% (74/294) of patients achieved recompensation after 2 years (Figure 2A).
FIGURE 2.
Comparison of recompensation and clinical endpoints in all patients. (A) Annual incidence of recompensation in all patients; (B) The proportion of HCC/All‐cause death, HCC, and Liver‐related death in patients with or without recompensation, respectively; (C) The cumulative incidence of clinical endpoints including HCC/All‐cause death, HCC and Liver‐related death.
Among these patients, 52 cases of HCC/all‐cause death were recorded during the 5‐year follow‐up periods, including 42 cases of HCC and 12 liver‐related deaths. The proportion of HCC/all‐cause death was significantly lower in patients with recompensation compared to those without recompensation (10.9% vs. 23.8%, p = 0.003), with similar results observed for HCC (10.2% vs. 14.3%, p = 0.064) and liver‐related death (1.0% vs. 10.7%, p = 0.023) (Figure 2B). The 5‐year cumulative incidence of HCC/all‐cause death was 14.9%, with 12.0% in HCC and 3.7% in liver‐related death (Figure 2C).
At baseline, patients with recompensation demonstrated higher ALT, AST, TBIL, Child‐Pugh, and MELD scores, as well as a lower degree of portal hypertension, indicated by PLT and spleen thickness, compared to those without recompensation (Table 1). Furthermore, given that the median time to recompensation was 73.0 weeks (54.0–104.0), the clinical characteristics at the time of recompensation (in patients with recompensation) compared to those at 78 weeks after initial decompensation (in patients without recompensation) were analysed. The results showed that patients with recompensation had higher WBC, PLT, ALB, and lower INR, Child‐Pugh score, MELD, portal vein width, and spleen thickness than those without recompensation (Table 2).
TABLE 2.
Recompensation characteristics of patients with stable or unstable recompensation (n = 294).
| Characteristics | With recompensation (n = 294) | p a | Without recompensation (n = 84) | p b | ||
|---|---|---|---|---|---|---|
| Total | Stable recompensation (n = 202) | Unstable recompensation (n = 92) | Ongoing decompensation (n = 84) | |||
| WBC (109/L) | 4.2 (3.3, 5.6) | 4.6 (3.6, 5.8) | 3.5 (2.7, 4.6) | < 0.001 | 2.5 (2.0, 3.3) | < 0.001 |
| PLT (109/L) | 87.0 (60.3, 124.5) | 91.0 (65.0, 130.0) | 74.0 (45.0, 107.0) | 0.004 | 51.0 (36.0, 70.0) | < 0.001 |
| ALT (IU/L) | 24.0 (18.0, 33.0) | 24.0 (18.0, 33.1) | 22.9 (14.5, 31.8) | 0.257 | 27.0 (19.5, 38.5) | 0.153 |
| AST (IU/L) | 28.6 (23.8, 36.0) | 28.0 (23.9, 36.0) | 30.0 (23.7, 37.8) | 0.472 | 30.0 (21.0, 45.9) | 0.383 |
| ALB (g/L) | 43.0 (39.2, 45.8) | 43.9 (40.3, 46.1) | 40.8 (36.8, 44.1) | < 0.001 | 37.6 (34.1, 40.8) | < 0.001 |
| TBIL (μmol/L) | 19.7 (14.7, 27.4) | 19.6 (14.8, 27.2) | 20.3 (13.9, 28.9) | 0.668 | 24.1 (16.9, 37.1) | 0.074 |
| INR | 1.1 (1.1, 1.3) | 1.1 (1.0, 1.2) | 1.2 (1.1, 1.3) | 0.083 | 1.3 (1.2, 1.4) | < 0.001 |
| AFP (ng/mL) | 2.9 (1.7, 4.3) | 2.8 (1.7, 4.3) | 3.1 (1.7, 5.3) | 0.723 | 3.1 (2.5, 3.9) | 0.561 |
| Child‐Pugh scores | 5.0 (5.0, 5.0) | 5.0 (5.0, 5.0) | 5.0 (5.0, 6.0) | 0.334 | 6.0 (5.0, 7.0) | < 0.001 |
| MELD scores | 9.0 (7.0, 10.0) | 9.0 (7.0, 10.0) | 9.0 (7.0, 11.0) | 0.681 | 11.0 (10.0, 13.0) | < 0.001 |
| Portal width (cm) | 1.2 (1.1, 1.3) | 1.2 (1.1, 1.2) | 1.2 (1.1, 1.4) | 0.064 | 1.4 (1.2, 1.6) | < 0.001 |
| Spleen thickness (cm) | 4.5 (4.0, 5.2) | 4.4 (3.8, 4.9) | 4.6 (4.4, 4.8) | 0.091 | 6.6 (5.9, 7.2) | < 0.001 |
Note: Data presented as median (IQR) or number (%); Quantitative differences were analysed by the Student's t‐test or Mann–Whitney U test for continuous variables, and by chi‐squared test or the Fisher exact test for categorical variables, as appropriate. Bold values indicate statistically significant differences (P < 0.05).
Abbreviations: AFP, alpha‐fetoprotein; ALB, albumin; ALT, alanine transaminase; AST, aspartate aminotransferase; Child‐Pugh, Child‐Turcotte‐Pugh; INR, international normalised ratio; MELD, Model for End‐Stage Liver Disease; PLT, platelet; TBIL, total bilirubin; WBC, white blood cell count.
Stable recompensation vs. unstable recompensation.
With recompensation vs. without recompensation.
3.3. Detailed Clinical Descriptions of All Patients Revealed Two Patterns of Recompensation
The clinical course of these patients experiencing recompensation and decompensation after first decompensated events was collected and illustrated using a Sankey diagram (Figure 3). Among these patients, 84 patients remain in a decompensated state throughout follow‐up, without ever achieving recompensation, classified as having ‘ongoing decompensation’. Among patients who fulfilled the criteria of recompensation, 202 (53.4%) patients meet recompensation criteria and do not experience any further episodes of decompensation after initial recompensation, defined as ‘stable recompensation’. The remaining 92 patients (24.3%) either achieve recompensation after multiple episodes of decompensation or initially achieve recompensation but subsequently relapse into decompensation, defined as ‘unstable recompensation’.
FIGURE 3.
Sankey plot presenting 5‐year clinical courses after the first decompensation in all patients. Patients with stable recompensation showing never experienced further decompensated events during follow‐up. Patients with unstable recompensation showing recompensation interweaved with decompensation after first events. Patients without recompensation showing experienced repeated decompensation during the clinical course and never achieved recompensation.
Patients with stable recompensation were more likely to have ascites as the initial decompensation event, whereas bleeding and combined events were significantly less common compared to those with unstable recompensation. However, those with unstable recompensation exhibited significantly more severe portal hypertension at baseline, as indicated by lower PLT, larger spleen thickness, and wider portal width. Moreover, patients with unstable recompensation exhibited lower baseline levels of ALT, AST, TBIL, and MELD scores (Table 1). At the time of recompensation, patients with unstable recompensation showed worse clinical profiles, with lower WBC, PLT, and serum ALB (Table 2).
For the long‐term prognosis, the cumulative incidence of HCC/all‐cause death in patients with stable recompensation (3‐, 5‐year incidence: 3.2%, 10.1%) was lower than that of patients with unstable recompensation (3‐, 5‐year incidence: 6.6%, 14.7%) and ongoing decompensation (3‐, 5‐year incidence: 21.1%, 27.3%) at all time points during the study period (Figure 4A). Notably, there was no significant difference in the 5‐year incidence of HCC between the three patterns of patients (10.9% vs. 12.5% vs. 17.9%, p = 0.063) (Figure 4B). The incidence of liver‐related death gradually increased in the three patterns of patients at all time points during the study period, respectively (5‐year incidence: 0% vs. 3.5% vs. 12.2%, p < 0.001) (Figure 4C).
FIGURE 4.
Cumulative incidence of HCC/All‐cause death, HCC, and Liver‐related death in different patterns of patients. Cumulative incidence of HCC/All‐cause death (A), HCC (B) and Liver‐related death (C) between patients with stable recompensation and unstable recompensation or ongoing decompensation.
We also compared the distribution of recompensation and long‐term prognosis based on the type of first episode events (Figure S1). Patients with single ascites as the first episode event had a higher proportion of stable recompensation than patients with ≥ 2 concomitant events (56.7% vs. 37.3%, p = 0.018), while no significant differences were observed between the bleeding and ≥ 2 concomitant events groups (Figure S1A). The 5‐year incidence of HCC/all‐cause death was highest in patients who did not achieve recompensation, regardless of the type of initial decompensated event. For patients with ascites (p = 0.009), stable recompensation was associated with the lowest event rates compared with unstable recompensation and ongoing decompensation. While in the bleeding group (p = 0.032) and ≥ 2 concomitant events (p = 0.102), the event rates were similar between unstable and stable recompensation, which might be due to the limited samples (Figure S1B–D).
3.4. Improvement Rates of Clinical Features Across Patients With Different Patterns of Clinical Course
The dynamic changes in clinical parameters were assessed to determine the improvement of clinical features during the 5‐year follow‐up. Patients with stable recompensation exhibited the most pronounced and sustained improvements compared to those with unstable recompensation and ongoing decompensation. Significant differences were observed in PLT (p < 0.001), ALT (p < 0.001), AST (p < 0.001), TBIL (p < 0.001), ALB (p < 0.001), INR (p = 0.042), Child‐Pugh scores (p < 0.001) and MELD scores (p = 0.023). In terms of the degree of improvement, patients with stable recompensation improved significantly more than those with unstable recompensation, who in turn showed better outcomes than patients with ongoing decompensation (Figure S2).
Since that recompensation requires at least 12 months of stability, the improvement rates of clinical parameters were assessed within 1 year after initial decompensation. Features were identified using univariate analysis showing significant differences among the three patient groups. The improvement within 1 year in liver function and portal hypertension parameters was similar with dynamic changes during 5 years. Patients with stable recompensation exhibited significantly the most pronounced improvement compared to other patients: ALT (p = 0.002), AST (p = 0.002), TBIL (p = 0.010), ALB (p < 0.001), INR (p = 0.018), Child‐Pugh scores (p < 0.001) and MELD scores (p = 0.044) (Figure 5B–H). However, the improvement of PLT showed no significant difference among the groups (Figure 5A, p = 0.479).
FIGURE 5.
The improvement rates of clinical parameters after initial decompensated events. The improvement rates of PLT (A), ALT (B), AST (C), TBIL (D), ALB (E), INR (F), Child‐Pugh scores (G) and MELD scores (H) during 1 year after initial decompensation.
Liver function improvement followed a similar tendency across all groups: rapid improvement during the first 0.5 year, followed by turning gently from 0.5 year to 1 year. The stable recompensation group exhibited the highest improvement rates in ALT, TB, ALB, Child‐Pugh score and MELD score, followed by the unstable recompensation group, with the ongoing decompensation group showing the least improvement.
3.5. Predictive Factors Influencing the Transition to Recompensation and Unstable Recompensation
Univariate analysis identified several baseline and follow‐up parameters associated with recompensation and unstable recompensation. For recompensation, predictors included baseline PLT, ALT, AST, AFP, Child‐Pugh and MELD at baseline; 0.5‐year values of WBC, ALB, INR, AFP and Child‐Pugh were identified as significant predictors (Table 3). For unstable recompensation, significant predictors were baseline PLT, TBIL and MELD; and WBC, PLT, ALB, INR, and MELD at 0.5‐year among patients who had achieved recompensation (Table 3).
TABLE 3.
Univariate analysis of associated factors for recompensation and stable recompensation.
| Variables | Univariate analysis | |||
|---|---|---|---|---|
| With recompensation | Unstable recompensation | |||
| OR (95% CI) | p | OR (95% CI) | p | |
| Baseline | ||||
| Sex (male) | 0.931 (0.553, 1.568) | 0.789 | 0.727 (0.419, 1.261) | 0.257 |
| Age (year) | 0.991 (0.968, 1.013) | 0.411 | 1.005 (0.982, 1.029) | 0.653 |
| WBC (109/L) | 1.059 (0.949, 1.182) | 0.305 | 0.885 (0.784, 1.000) | 0.051 |
| PLT (109/L) | 1.009 (1.001, 1.016) | 0.018 | 0.985 (0.977, 0.993) | < 0.001 |
| ALT (IU/L) | 1.006 (1.002, 1.010) | 0.003 | 0.999 (0.998, 1.000) | 0.052 |
| AST (IU/L) | 1.005 (1.001, 1.009) | 0.007 | 0.998 (0.997, 1.000) | 0.060 |
| ALB (g/L) | 0.983 (0.946, 1.021) | 0.371 | 0.980 (0.943, 1.018) | 0.304 |
| TBIL (μmol/L) | 1.003 (1.000, 1.007) | 0.074 | 0.996 (0.993, 0.999) | 0.035 |
| INR | 1.728 (0.774, 3.857) | 0.182 | 0.951 (0.466, 1.943) | 0.891 |
| AFP (ng/mL) | 1.010 (1.004, 1.016) | < 0.001 | 0.998 (0.997, 1.000) | 0.052 |
| Child‐Pugh score | 1.183 (1.031, 1.357) | 0.016 | 0.950 (0.837, 1.078) | 0.425 |
| MELD score | 1.062 (1.004, 1.124) | 0.037 | 0.946 (0.899, 0.996) | 0.034 |
| 0.5 year | ||||
| WBC (109/L) | 1.909 (1.284, 2.838) | 0.001 | 0.656 (0.504, 0.854) | 0.002 |
| PLT (109/L) | 1.012 (0.999, 1.026) | 0.070 | 0.982 (0.972, 0.993) | 0.001 |
| ALT (IU/L) | 1.006 (0.981, 1.032) | 0.630 | 1.003 (0.995, 1.011) | 0.461 |
| AST (IU/L) | 1.003 (0.984, 1.022) | 0.751 | 1.001 (0.994, 1.009) | 0.726 |
| ALB (g/L) | 1.147 (1.062, 11.239) | < 0.001 | 0.939 (0.886, 0.996) | 0.035 |
| TBIL (μmol/L) | 0.987 (0.962, 1.012) | 0.308 | 1.016 (0.996, 1.037) | 0.116 |
| INR | 0.024 (0.001, 0.397) | 0.009 | 3.667 (2.538, 9.942) | 0.011 |
| AFP (ng/mL) | 1.203 (1.015, 1.426) | 0.033 | 1.006 (0.972, 1.042) | 0.727 |
| Child‐Pugh score | 0.448 (0.295, 0.679) | < 0.001 | 1.250 (0.826, 1.893) | 0.291 |
| MELD score | 0.978 (0.832, 1.150) | 0.792 | 0.780 (0.630, 0.965) | 0.022 |
Note: Bold values indicate statistically significant differences (P < 0.05).
Abbreviations: AFP, alpha‐fetoprotein; ALB, albumin; ALT, alanine transaminase; AST, aspartate aminotransferase; Child‐Pugh, Child‐Turcotte‐Pugh; CI, confidence interval; decomp, decompensation; MELD, Model for End‐Stage Liver Disease; OR, odds ratio; PLT, platelet; TBIL, total bilirubin.
For predicting recompensation, 264 (70%) patients were allocated to the training set to develop models, while the remaining 114 (30%) patients constituted the testing set for evaluating model stability and reliability in the entire cohort. In the training and testing sets had similar distributions of patients with recompensation and without recompensation (recompensation, 78.4% vs. 76.3%; without recompensation, 21.6% vs. 23.7%) (Figure 3A). Table S1 shows comparative analyses of baseline and 0.5‐year clinical characteristics between patients with or without recompensation, stratified across both the training and testing sets.
For predicting recompensation, the AUROC was 0.896 and 0.936 in the training set based on LR and SVM models (Figure S4A,B). Overall, the LR and SVM models consistently showed excellent predictive performance in both sets for recompensation. At the LR model, the corresponding sensitivity (SE) in the training set was 0.828 with a specificity (SP) of 0.867, a positive prediction value (PPV) of 0.778 and a negative prediction value (NPV) of 0.860 (Table 4).
TABLE 4.
Predictive performance of the LR model and SVM model for recompensation and stable recompensation.
| AUROC | Sensitivity (SE) | Specificity (SP) | Positive predictive value (PPV) | Negative predictive value (NPV) | |
|---|---|---|---|---|---|
| Recompensation in the training set (n = 264) | |||||
| LR | 0.896 | 0.828 | 0.867 | 0.778 | 0.860 |
| SVM | 0.936 | 0.814 | 0.967 | 0.988 | 0.604 |
| Recompensation in the testing set (n = 114) | |||||
| LR | 0.775 | 0.911 | 0.583 | 0.672 | 0.721 |
| SVM | 0.778 | 0.578 | 0.875 | 0.945 | 0.356 |
| Unstable recompensation in the training set (n = 205) | |||||
| LR | 0.801 | 0.780 | 0.705 | 0.745 | 0.797 |
| SVM | 0.922 | 0.932 | 0.821 | 0.679 | 0.968 |
| Unstable recompensation in the testing set (n = 89) | |||||
| LR | 0.693 | 0.593 | 0.742 | 0.527 | 0.706 |
| SVM | 0.812 | 0.852 | 0.645 | 0.511 | 0.909 |
Abbreviations: LR, logistic regression; SVM, support vector machine.
Through SHAP analysis, several features were identified for predicting recompensation. Specifically, baseline MELD, Child‐Pugh, AFP, INR, AST, first decompensation types, and 0.5‐year PLT, ALB, WBC, AST were used to develop the LR model for predicting recompensation (Figure 6A).
FIGURE 6.
Shapley Additive Explanations plot: The impact of clinical features for predicting recompensation in the logistic regression and predicting unstable recompensation in the support vector machine algorithm. (A) Recompensation; (B) Unstable recompensation.
For predicting unstable recompensation, 207 (70%) patients were allocated to the training set and 87 (30%) patients constituted the testing set in these 294 patients with recompensation. Within the training cohort, 33.8% of patients achieved unstable recompensation while 66.2% manifested stable recompensation (Figure S3B). Table S2 shows comparative analyses of baseline and 0.5‐year clinical characteristics between patients with unstable and stable recompensation. The AUROC was 0.801 and 0.922 in the training set (Figure S4C,D). Overall, the SVM model showed the best predictive performance in both sets for unstable recompensation. The predictive performance for unstable recompensation was shown at Table 4.
Several features were identified for predicting unstable recompensation by using SHAP analysis. Specifically, 0.5‐year AFP, PLT, TBIL, ALT, WBC, ALB, AST, and baseline MELD, PLT, Child‐Pugh, AST, AFP, INR, ALT, ALB, WBC, TBIL, sex, age, and first decompensation types were used to develop the final SVM model for predicting unstable recompensation (Figure 6B).
4. Discussion
Recompensation following an effective cure or control of the aetiology has been confirmed in various different etiologies [8, 11, 12]. However, patients who achieve recompensation remain at risk for progression to decompensation, suggesting the possibility of different types or patterns of decompensation. Our findings reveal that patients with different recompensation patterns exhibit distinct clinical characteristics and outcomes after their first decompensation. Moreover, the baseline degree of portal hypertension and recovery of liver function are closely associated with different recompensation patterns and may serve as key indicators for predicting long‐term outcomes.
To our knowledge, the present study is the first to focus on the classification of recompensation subtypes according to clinical course. We innovatively defined a new recompensation subtype, ‘unstable recompensation’. Much like the PREDICT study found that patients with unstable decompensation have a distinctly worse prognosis [13]. Approximately one quarter of decompensated patients achieved unstable recompensation after their first decompensation, but their 5‐year prognosis, especially in terms of mortality, differed significantly from stable recompensation. Therefore, these patients who are unable to maintain a stable state were defined as unstable recompensation, which may be an underlying reason why the recompensation rate varies among different studies, ranging from 12.3% to 56.2% [7, 8, 11, 12, 14, 15].
So far, in all the recompensation‐related studies, recompensation was associated with a lower risk of mortality [16]. In patients with HBV‐related compensated cirrhosis, mortality rates following recompensation have been reported to range from 1% to 44% [8, 9, 14]. Our present study showed the patients achieving recompensation exhibited mortality of 1.3%, significantly lower than the 9.7% observed in patients without recompensation. However, the mortality rate still requires further investigation among patients achieving recompensation, as it varies widely across different studies. Classifying recompensation into these two subtypes will provide crucial prognostic information.
In addition, the clinical course of patients with unstable recompensation was complex. Some patients achieved recompensation initially before further decompensation, while others underwent multiple episodes of decompensation before recompensation. Interestingly, similar recompensation‐like trajectories have also been described in PBC patients responding to UDCA therapy, suggesting that the process of recompensation may represent a shared pathophysiological response across different etiologies of cirrhosis once the underlying cause is effectively treated [12]. However, patients experiencing unstable recompensation received comparatively less rigorous follow‐up than those with stable recompensation. This discrepancy arises due to decompensation occurring more than once and alternating with recompensation, disrupting continuous monitoring and management. The previous study has indicated that cirrhosis can reverse to virtually typical liver structure once the underlying cause is eliminated and adequate time is allowed for recovery [17]. The complexity of unstable recompensation suggests that clinical subtypes of recompensation should be differentiated in studies of recompensation mechanisms [16].
For patients maintained on etiological therapy, current studies fail to specify the timing for initiating etiological control when defining effective etiological therapy. This ambiguity undermines the reproducibility of recompensation criteria. To date, only one study demonstrates that patients who initiated antiviral therapy at least 3 months before the first decompensation can achieve recompensation [1]. Therefore, exploring the underlying mechanisms developing decompensation followed by recompensation during on‐treatment is crucial. We inferred that there was acute deterioration of liver function triggered by precipitating factors that may lead to decompensation events. Since the underlying etiological treatment remains ongoing, supportive care can restore hepatic recompensation.
The significantly lower PLT observed in these patients with unstable recompensation likely reflects more severe portal hypertension, as thrombocytopenia is a well‐established consequence of portal hypertension in advanced liver disease [2, 18, 19]. The persistence of severe portal hypertension even after achieving recompensation suggests a higher vulnerability to further decompensation. Similarly, ALB is exclusively produced by the liver and serves as a direct marker of hepatic synthetic function [20]. The combination of persistent portal hypertension and impaired synthetic function indicates a more advanced and irreversible disease stage in patients, rendering these patients more susceptible to re‐decompensation. Importantly, both PLT and ALB are easily measured in routine clinical practice and may serve as valuable prognostic biomarkers to stratify the risk of decompensation after recompensation.
As previously mentioned, patients experiencing different decompensated events have significantly varied recompensation rates [8, 9, 21]. Patients with ascites achieved a notably higher recompensation rate compared to those with bleeding events [8]. Although there was no statistical difference between ascites and bleeding groups, we observed a higher proportion of patients in the ascites group who achieved stable recompensation compared to other groups, illustrating that patients with ascites may be more likely to achieve stable recompensation. Moreover, there were different further decompensation rates between patients with ascites or bleeding, where patients with bleeding experienced a higher further decompensation rate [8, 22, 23].
Early recognition and prediction can facilitate timely intervention, potentially guiding patients towards a more favourable clinical course and prognosis. Our study found that clinical parameters associated with recompensation and its subtypes mainly fall into two categories: those related to the degree of portal hypertension and those related to liver function and liver regeneration. Similarly to previous studies, the baseline PLT level, which indicated the degree of portal hypertension, can predict whether recompensation can be achieved and whether it is stable [24, 25]. While baseline elevated AFP levels, which were associated with the ability of hepatocyte regeneration [26, 27, 28], indicate a higher likelihood of achieving recompensation. The increased ALB level across all subtypes of recompensation indicates robust restoration of hepatocellular function and lobular structure [9, 29, 30, 31].
Moreover, clinical parameters at 0.5 year provide superior predictive value for the transition to stable recompensation, which indicates the magnitude of early improvement serves as a critical prognostic marker for complete hepatic recovery. Furthermore, the liver function improvement within 0.5 year was more remarkable than that in the subsequent 0.5 year. This highlights the critical importance of early clinical management of decompensated events and the rapid repair of liver function. Early initiation of etiologic intervention to promote liver function recovery is essential to enhance the stability of recompensation and improve long‐term outcomes.
These findings underscore that recompensation represents a continuum of clinical stability. Distinguishing between stable and unstable recompensation is crucial for individualising management strategies. Patients identified as having unstable recompensation may benefit from closer monitoring, earlier intervention for complications, and expedited evaluation for liver transplantation.
The strength of our study includes a relatively large sample size of patients who experienced decompensation more than once, ensuring the robustness of the results. Additionally, sufficient follow‐up duration and multiple follow‐up data points allowed for a comprehensive exploration of the clinical course after recompensation by analysing the long‐term prognosis and dynamic change of liver function.
However, the study had several limitations. Firstly, most patients were derived from a retrospective cohort, with data collection relying on the retrospective method, which may have selection bias and outcome reporting bias. Secondly, although the median follow‐up period was 5 years, the occurrence of clinical outcomes such as HCC or death was relatively low, with only a small proportion of patients progressing to these outcomes during the study period. Further extended follow‐up is necessary to assess the durability of recompensation. Additionally, we enrolled only patients with HBV‐related cirrhosis; the findings may not be directly generalizable to patients with other etiologies of cirrhosis.
In summary, our study provides a comprehensive description of different patterns of recompensation, particularly highlighting the definition of unstable recompensation. The 5‐year prognosis for patients with unstable recompensation was found to be worse compared to those with stable recompensation, but still better than patients without recompensation. Subclassification of recompensation has significant implications for prognosis and tailored management.
Author Contributions
Shuai Xia: writing – original draft, data curation, formal analysis, investigation, software, visualization. Zhiying He: data curation, validation. Xiaoning Wu: supervision, writing – review and editing. Zhongjie Hu: supervision. Chunqing Zhang: data curation. Yanqin Hao: data curation. Yongfeng Yang: data curation. Yan Huang: data curation. Wei Rao: data curation. Xiaoqian Xu: methodology. Xinyu Zhao: methodology. Jialing Zhou: data curation. Yameng Sun: supervision. Shuyan Chen: supervision. Luqi Tang: data curation. Xiaojuan Ou: conceptualization. Xinyan Zhao: conceptualization. Jidong Jia: conceptualization, project administration, resources, funding acquisition. Bingqiong Wang: conceptualization, writing – review and editing, project administration, methodology, supervision. Hong You: conceptualization, funding acquisition, investigation, visualization, writing – review and editing, project administration, resources, supervision.
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting information
Figure S1: Distribution of recompensation status and long‐term outcome among patients with different type of inital decompensated events. (A) The proportion of subgroups stratified by their initial decompensated events; The cumulative incidence of HCC/All‐cause death among different subgroups in patients with ascites (B) bleeding (C) and ≥ 2 concomitant events (D) as initial events.
Figure S2: The change in key clinical parameters within 5 years after initial decompensated events. The changes of PLT (A), ALT (B), AST (C), TBIL (D), ALB (E), INR (F), Child‐Pugh scores (G) and MELD scores (H) during 5‐year after initial decompensation. Abbreviations: ALT, alanine aminotransferase; ALB, albumin; AST, aspartate aminotransferase; INR, international normalized ratio; MELD, Model for End‐Stage Liver Disease; PLT, platelet count; TBIL, total bilirubin.
Figure S3: The division of training set and testing set. (A) All patients; (B) Patients with recompensation.
Figure S4 The receiver operating characteristic curves for predicting recompensation and unstable recompensation in the logistic regression and support vector machine algorithm. (A) Recompensation in training set; (B) Recompensation in testing set; (C) Unstable recompensation in training set; (D) Unstable recompensation in testing set.
Table S1: The characteristics of patients with recompensation and without recompensation in training set and testing set.
Table S2: The characteristics of patients with stable recompensation and unstable recompensation in training set and testing set.
Handling Editor: Grace L.‐H. Wong
Funding: This work was supported by grants from the National Key Research and Development Program of China (2023YFC2306900).
Contributor Information
Bingqiong Wang, Email: 13031136358@163.com.
Hong You, Email: youhongliver@ccmu.edu.cn.
Data Availability Statement
The datasets generated for this study are available from the corresponding author Hong You.
References
- 1. He Z., Zhou J., Tian Y., et al., “Two‐Year Free of Complications During Antiviral Therapy Predicts Stable Re‐Compensation in Immediate‐Treatment HBV‐Related Decompensated Cirrhosis,” Scandinavian Journal of Gastroenterology 58 (2023): 403–411. [DOI] [PubMed] [Google Scholar]
- 2. de Franchis R., Bosch J., Garcia‐Tsao G., Reiberger T., Ripoll C., and Baveno VII Faculty , “Baveno VII ‐ Renewing Consensus in Portal Hypertension,” Journal of Hepatology 76 (2022): 959–974. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. D'Amico G., Bernardi M., and Angeli P., “Towards as New Definition of Decompensated Cirrhosis,” Journal of Hepatology 76 (2021): 202–207. [DOI] [PubMed] [Google Scholar]
- 4. Mandorfer M. and Simbrunner B., “Prevention of First Decompensation in Advanced Chronic Liver Disease,” Clinics in Liver Disease 25 (2021): 291–310. [DOI] [PubMed] [Google Scholar]
- 5. Yip T. C., Wong G. L., Chan H. L., et al., “HBsAg Seroclearance Further Reduces Hepatocellular Carcinoma Risk After Complete Viral Suppression With Nucleos(t)ide Analogues,” Journal of Hepatology 70 (2019): 361–370. [DOI] [PubMed] [Google Scholar]
- 6. Su T. H., Hu T. H., Chen C. Y., et al., “Four‐Year Entecavir Therapy Reduces Hepatocellular Carcinoma, Cirrhotic Events and Mortality in Chronic Hepatitis B Patients,” Liver International 36 (2016): 1755–1764. [DOI] [PubMed] [Google Scholar]
- 7. Wang Q., Zhao H., Deng Y., et al., “Validation of Baveno VII Criteria for Recompensation in Entecavir‐Treated Individuals With Hepatitis B‐Related Decompensated Cirrhosis,” Journal of Hepatology 77 (2022): 1564–1572. [DOI] [PubMed] [Google Scholar]
- 8. He Z., Wang B., Wu X., et al., “Recompensation in Treatment‐naïve HBV‐Related Decompensated Cirrhosis: A 5‐Year Multi‐Center Observational Study Comparing Patients With Ascites and Bleeding,” Hepatology International 58 (2023): 403–411. [DOI] [PubMed] [Google Scholar]
- 9. Kim T. H., Um S. H., Lee Y. S., et al., “Determinants of Re‐Compensation in Patients With Hepatitis B Virus‐Related Decompensated Cirrhosis Starting Antiviral Therapy,” Alimentary Pharmacology & Therapeutics 55 (2022): 83–96. [DOI] [PubMed] [Google Scholar]
- 10. Wu X., Zhou J., Xie W., et al., “Entecavir Monotherapy Versus de Novo Combination of Lamivudine and Adefovir for Compensated Hepatitis B Virus‐Related Cirrhosis: A Real‐World Prospective Multicenter Cohort Study,” Infection and Drug Resistance 12 (2019): 745–757. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Hofer B. S., Simbrunner B., Hartl L., et al., “Hepatic Recompensation According to Baveno VII Criteria Is Linked to a Significant Survival Benefit in Decompensated Alcohol‐ Related Cirrhosis,” Liver International 43 (2023): 2220–2231. [DOI] [PubMed] [Google Scholar]
- 12. Hofer B. S., Burghart L., Halilbasic E., et al., “Evaluation of Potential Hepatic Recompensation Criteria in Patients With PBC and Decompensated Cirrhosis,” Alimentary Pharmacology & Therapeutics 59 (2024): 962–972. [DOI] [PubMed] [Google Scholar]
- 13. Trebicka J., Fernandez J., Papp M., et al., “The PREDICT Study Uncovers Three Clinical Courses of Acutely Decompensated Cirrhosis That Have Distinct Pathophysiology,” Journal of Hepatology 73 (2020): 842–854. [DOI] [PubMed] [Google Scholar]
- 14. Hui V. W., Wong G. L., Wong V. W., et al., “Baveno VII Criteria for Recompensation Predict Transplant‐Free Survival in Patients With Hepatitis B‐Related Decompensated Cirrhosis,” JHEP Reports 5 (2023): 100814. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Tonon M., Gagliardi R., Zeni N., and Piano S., “Recompensation of Cirrhosis in Candidates of Transplant: Tips and Tricks for Delisting,” Liver Transplantation 1 (2024): 1181–1187. [DOI] [PubMed] [Google Scholar]
- 16. Piano S., Reiberger T., and Bosch J., “Mechanisms and Implications of Recompensation in Cirrhosis,” JHEP Reports 10 (2024): 101233. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Chen S., Zhou J., Wu X., et al., “Liver Fibrosis Showed a Two‐Phase Regression Rate During Long‐Term Anti‐HBV Therapy by Three‐Time Biopsies Assessments,” Hepatology International 18 (2024): 904–916. [DOI] [PubMed] [Google Scholar]
- 18. Wang B., Zhou J., Wu X., et al., “Screening Varices in Patients With HBV‐Related Cirrhosis on Antiviral Therapy: Platelet Alone or Together With LSM,” Liver International 41 (2021): 369–377. [DOI] [PubMed] [Google Scholar]
- 19. Zanetto A., Campello E., Bulato C., et al., “Increased Platelet Aggregation in Patients With Decompensated Cirrhosis Indicates Higher Risk of Further Decompensation and Death,” Journal of Hepatology 77 (2022): 660–669. [DOI] [PubMed] [Google Scholar]
- 20. Bernardi M., Angeli P., Claria J., et al., “Albumin in Decompensated Cirrhosis: New Concepts and Perspectives,” Gut 69 (2020): 1127–1138. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Li M., Zong Z., Xiong X., et al., “Ascites Re‐Compensation in HBV‐Related First Decompensated Cirrhosis After Anti‐Viral Therapy,” Frontiers in Cellular and Infection Microbiology 12 (2023): 1053608. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Balcar L., Tonon M., Semmler G., et al., “Risk of Further Decompensation/Mortality in Patients With Cirrhosis and Ascites as the First Single Decompensation Event,” JHEP Report 4 (2022): 100513. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Gupta A., Rana R., Agarwal S., et al., “Assessing the Risk of Further Decompensation and Survival in Patients With Cirrhosis With Variceal Bleeding as Their First Decompensation Event,” American Journal of Gastroenterology 118 (2022): 833–839. [DOI] [PubMed] [Google Scholar]
- 24. La Mura V., Garcia‐Guix M., Berzigotti A., et al., “A Prognostic Strategy Based on Stage of Cirrhosis and HVPG to Improve Risk Stratification After Variceal Bleeding,” Hepatology 72 (2020): 1353–1365. [DOI] [PubMed] [Google Scholar]
- 25. Jindal A., Bhardwaj A., Kumar G., and Sarin S. K., “Clinical Decompensation and Outcomes in Patients With Compensated Cirrhosis and a Hepatic Venous Pressure Gradient ≥ 20 Mm hg,” American Journal of Gastroenterology 115 (2020): 1624–1633. [DOI] [PubMed] [Google Scholar]
- 26. Di Bisceglie A. M., Sterling R. K., Chung R. T., et al., “Serum Alpha‐Fetoprotein Levels in Patients With Advanced Hepatitis C: Results From the HALT‐C Trial,” Journal of Hepatology 43 (2005): 434–441. [DOI] [PubMed] [Google Scholar]
- 27. Gopal P., Yopp A. C., Waljee A. K., et al., “Factors That Affect Accuracy of α‐Fetoprotein Test in Detection of Hepatocellular Carcinoma in Patients With Cirrhosis,” Clinical Gastroenterology and Hepatology 12 (2014): 870–877. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Kuhlmann W. D. and Peschke P., “Hepatic Progenitor Cells, Stem Cells, and AFP Expression in Models of Liver Injury,” International Journal of Experimental Pathology 87 (2006): 343–359. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Deng Y., Kang H., Xiang H., et al., “Durability and On‐Treatment Predictors of Recompensation in Entecavir‐Treated Patients With Hepatitis B and Decompensated Cirrhosis,” JHEP Report 6 (2024): 101091. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Wen S., Ruan J., Shen J., et al., “Development and Validation of a Nomogram to Predict Recompensation in HBV‐Related Cirrhosis With Ascites as the Single First Decompensating Event,” Scandinavian Journal of Gastroenterology 58 (2023): 915–922. [DOI] [PubMed] [Google Scholar]
- 31. Xu X., Wang H., Zhao W., Wang Y., Wang J., and Qin B., “Recompensation Factors for Patients With Decompensated Cirrhosis: A Multicentre Retrospective Case‐Control Study,” BMJ Open 23 (2021): e043083. [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
Figure S1: Distribution of recompensation status and long‐term outcome among patients with different type of inital decompensated events. (A) The proportion of subgroups stratified by their initial decompensated events; The cumulative incidence of HCC/All‐cause death among different subgroups in patients with ascites (B) bleeding (C) and ≥ 2 concomitant events (D) as initial events.
Figure S2: The change in key clinical parameters within 5 years after initial decompensated events. The changes of PLT (A), ALT (B), AST (C), TBIL (D), ALB (E), INR (F), Child‐Pugh scores (G) and MELD scores (H) during 5‐year after initial decompensation. Abbreviations: ALT, alanine aminotransferase; ALB, albumin; AST, aspartate aminotransferase; INR, international normalized ratio; MELD, Model for End‐Stage Liver Disease; PLT, platelet count; TBIL, total bilirubin.
Figure S3: The division of training set and testing set. (A) All patients; (B) Patients with recompensation.
Figure S4 The receiver operating characteristic curves for predicting recompensation and unstable recompensation in the logistic regression and support vector machine algorithm. (A) Recompensation in training set; (B) Recompensation in testing set; (C) Unstable recompensation in training set; (D) Unstable recompensation in testing set.
Table S1: The characteristics of patients with recompensation and without recompensation in training set and testing set.
Table S2: The characteristics of patients with stable recompensation and unstable recompensation in training set and testing set.
Data Availability Statement
The datasets generated for this study are available from the corresponding author Hong You.