ACLF can be triggered by ongoing HBV replication in patients with or without liver cirrhosis and can lead to an overwhelming immunological response against HBV infection [5]. HBV-ACLF is mainly characterized by jaundice and coagulopathy, which are the initial clinical manifestations of acute liver damage, and subsequently progresses to ascites, upper gastrointestinal hemorrhage, hepatorenal syndrome, spontaneous peritonitis and even hepatic encephalopathy [5]. As an effective tool to determine the survival rate of patients with HBV-ACLF, artificial liver support systems are widely used to treat liver failure in HBV-ACLF patients until their livers regenerate [27]. It has been reported that plasma exchange combined with DPMAS therapy is an effective treatment for improving liver function, coagulation function, and routine blood parameters and increasing the survival rate of ACLF patients [7, 8]. However, there are no specific effective methods, but primarily supportive treatment has been used. Since patients with ACLF have high short-term mortality [5], predicting and determining prognosis for prompt and early recognition is critical to guide clinical decision-making in patients with HBV-ACLF.
Nomograms, as a reliable prognostic tool for predicting waiting list mortality in candidates, are widely used to assess the clinical prognosis of various diseases [11]. Previously, multiple predictive models for the prognosis of patients with HBV-ACLF were established for various clinical conditions, such as bacterial infection or glucocorticoid medication [23, 24, 28]. According to predictive models for prognosis, systemic reviews of clinical data revealed that six independent risk factors (TB, INR, age, neutrophil count, HE level and urea level) appeared to be promising candidates for predicting poor prognosis, and multiple simplified prognostic scores were further developed for patients with HBV-ACLF [22–24]. In addition, upper gastrointestinal hemorrhage was also identified as an independent prognostic predictor for 6-week mortality in patients with liver cirrhosis [29]. However, there are no effective models for predicting disease prognosis in HBV-ACLF patients receiving artificial liver therapy.
In our study, we identified the clinical characteristics of patients with HBV-ACLF and developed a new noninvasive and simplified score that can accurately predict patient outcomes. Using univariate analysis and pertinent clinical data, we identified five independent high-risk predictors (age ≥ 40 years, middle stage of liver failure, end stage of liver failure, hepatic encephalopathy, upper gastrointestinal hemorrhage and artificial liver therapy with DPMAS + PE) that can influence individual outcomes. Although artificial liver support is now widely used for liver failure patients, no relevant study has compared the prognostic effects of DPMAS + PE versus PE + DPMAS. In our study, we found that PE + DPMAS was more effective than DPMAS + PE in improving the outcome of patients with HBV-ACLF. Previous studies have highlighted the importance of TB, hepatorenal syndrome and the INR in the prognosis of patients with HBV-ACLF [23, 25], which may not be directly reflected in our study. The discrepancy between these studies may be attributable to therapeutic and supportive approaches and even individual variations in patients in clinical circumstances.
It has been demonstrated that the MELD score is a strong predictor of outcomes in patients with liver failure [26]. Despite all of its advantages, 15–20% of cases cannot be accurately predicted by the MELD score [26, 30]. In our study, we established a prognostic nomogram that showed significantly better predictive value than the MELD and Child‒Pugh scores. The accuracy and sensitivity of any new prognostic score are essential for determining treatment strategies and predicting outcomes in patients with ACLF, which can be assessed using the area under the receiver operating characteristic curve (AUC). Furthermore, the calibration curve and DCA, which are tools for evaluating the nomogram, showed high consistency with the predicted results and had greater clinical practicability. As a result, our established nomogram indicated that this model also possessed outstanding discriminative ability. Therefore, this model might be promising for determining the outcome of patients with HBV-ACLF receiving artificial liver treatment.
There are several advantages in our study. First, artificial liver support PE + DPMAS was considered a potential prognostic factor in our study and can serve as a guide for the selection of artificial liver. Second, the prognosis may be determined in advance based on the basic characteristics at admission and to guide clinical decision-making. However, several limitations should be declared in this study. We should increase the number of cohorts for validating the predicted nomogram. In addition, the single-center and retrospective design of this study may have some inherent biases. Finally, detailed data on demographic characteristics (including region, nation and race) should be acquired and statistically analyzed.
Overall, we established and evaluated a nomogram that estimated the prognosis of hospitalized patients with HBV-ACLF. Our nomogram is extremely useful for clinical practice and will be helpful for guiding patient management and clinical use of artificial liver therapy in patients with HBV-ACLF.