A dynamic prediction model for prognosis of acute-on-chronic liver failure based on the trend of clinical indicators

Abstract

Acute-on-chronic liver failure (ACLF) is a dynamic syndrome, and sequential assessments can reflect its prognosis more accurately. Our aim was to build and validate a new scoring system to predict short-term prognosis using baseline and dynamic data in ACLF. We conducted a retrospective cohort analysis of patients with ACLF from three different hospitals in China. To construct the model, we analyzed a training set of 541 patients from two hospitals. The model's performance was evaluated in a validation set of 130 patients from another center. In the training set, multivariate Cox regression analysis revealed that age, WGO type, basic etiology, total bilirubin, creatinine, prothrombin activity, and hepatic encephalopathy stage were all independent prognostic factors in ACLF. We designed a dynamic trend score table based on the changing trends of these indicators. Furthermore, a logistic prediction model (DP-ACLF) was constructed by combining the sum of dynamic trend scores and baseline prognostic parameters. All prognostic scores were calculated based on the clinical data of patients at the third day, first week, and second week after admission, respectively, and were correlated with the 90-day prognosis by ROC analysis. Comparative analysis showed that the AUC value for DP-ACLF was higher than for other prognostic scores, including Child-Turcotte-Pugh, MELD, MELD-Na, CLIF-SOFA, CLIF-C ACLF, and COSSH-ACLF. The new scoring model, which combined baseline characteristics and dynamic changes in clinical indicators to predict the course of ACLF, showed a better prognostic ability than current scoring systems. Prospective studies are needed to validate these results.

Figure 1

Flowchart of ACLF training and validation group cases screening.

Figure 2

(a) Univariate Cox regression analysis showing that the factors related to ACLF prognosis were age, WGO type, basic etiology, bacterial infection, ascites, gastrointestinal bleeding, hepatic encephalopathy stage, glutamate transpeptidase, albumin, total bilirubin, cholinesterase, prothrombin activity, blood urea nitrogen, creatinine, serum sodium, neutrophil ratio, and platelets. (b), Multivariate Cox regression analysis confirmed that age, WGO type, alcoholic etiology, total bilirubin, creatinine, prothrombin activity, and hepatic encephalopathy stage remained associated with prognosis of ACLF. (Assignment of indexes in COX regression analysis was as follows: WGO type A, B, and C of ACLF were assigned 1, 2, and 3 respectively; basic etiology was assigned 1 for the alcoholic liver disease alone, and 0 for others; positive complications including bacterial infection, gastrointestinal bleeding, ascites were all assigned 1, no complications 0; hepatic encephalopathy was assigned 0, 1, 2, 3, and 4 according to the West Haven stages 0, I, II, III, and IV, respectively; artificial liver system support (ALSS) was assigned 1, no ALSS 0).

Figure 3

(a) Elevated levels of total bilirubin at the third day and first, second, third, and fourth weeks after admission were risk factors for death. The Harrell’s C index of model fitting degree showed an increasing trend with the extension of time. (b) Elevated levels of creatinine at the third day and first, second, third, and fourth weeks after admission were risk factors for death. The Harrell’s C index of model fitting degree increased to the highest value at the second week, and decreased slightly at other time points. (c) Declined levels of prothrombin activity at the third day and first, second, third, and fourth weeks after admission were all risk factors for death. The Harrell’s C index of model fitting degree showed an increasing trend with the extension of time.

Figure 4

(a–c) DP-ACLF scores based on the patients of the training set at the third day, first week, and second week after admission, respectively, and comparison of ROC curves with other prognostic scores at the same time point. (d) ROC curve comparisons of DP-ACLF scores at different time points.

Figure 5

The DP-ACLF score was compared with the baseline data of other prognostic scoring systems, and ROC curves were performed to determine prognostic accuracy. (a) Training group. (b) Validation group.

Figure 6

According to the DP-ACLF scores, C-support vector classifier function was used to predict the decision function value of each case in the validation group. Python software with matplotlib and sklearn packages was used to show that 90 cases of survivors (blue) (survival time ≥ 90 days) and 21 cases of non-survivors (red) (survival time < 90 days) were distributed in the green and yellow areas, respectively. The darker the color of the region in which the dots were located, the more accurate the prediction was.

Figure 7

Kaplan–Meier survival analysis showing a significant difference in 28-day (χ² = 238.411, p < 0.001) and 90-day cumulative survival rates (χ² = 270.194, p < 0.001) in patients with ACLF who were graded according to the DP-ACLF score.