Patients diagnosed as cirrhosis with acute decompensation (AD) and hospitalized in the Department of Infectious Diseases at the First Affiliated Hospital, School of Medicine, Zhejiang University, Zhejiang, China, between February 2014 and March 2015 were eligible for inclusion in this study. Patients were excluded if they: (1) were <18 years old; (2) were pregnant; (3) had evidence of infection with human immunodeficiency virus, Epstein-Barr virus, or human cytomegalovirus; (4) had aplastic anemia, myelodysplastic syndrome, thrombocytopenia, hemophilia, or disseminated intravascular coagulation; (5) were currently using immunosuppressant medications or adrenocortical hormones; (6) had hepatocellular carcinoma or other malignant tumors; (7) had chronic renal disease or other serious comorbidity; or (8) had history of liver transplantation (19, 20). The diagnosis of liver cirrhosis was based on clinical evidence, endoscopic or radiologic examination, and liver biopsy (6, 21). Cirrhosis patients were diagnosed with AD if they had more than one of the following major complications: ascites, encephalopathy, gastrointestinal hemorrhage, severe liver injury, and infection (20). Severe liver injury was diagnosed if alanine aminotransferase (ALT) was ≥ five times the normal upper limit or more than twice the baseline value along with elevation of serum bilirubin to ≥ 85 μmol/l and international normalized ratio (INR) to ≥ 1.5 at any time during the preceding 1 month (21). The chronic liver failure-sequential organ failure assessment (CLIF-SOFA) score was used to diagnose organ failure (6). The diagnosis of ACLF was based on the criteria proposed by the Chinese Group on the Study of Severe Hepatitis B (COSSH) (22). The diagnosis of gastrointestinal bleeding and bacterial infections was made as previously described (1, 23, 24). PPI treatment was generally used in patients with gastrointestinal bleeding and gastric ulcer, but was extended in patients receiving endoscopic variceal ligation and those manifesting gastrointestinal disturbances such as hiccups, epigastric pain, nausea, or vomiting. For this study, “PPI use” was defined as intravenous administration of any PPI for at least 6 days with daily dose higher than that recommended by the WHO (13). Patients receiving lower doses were considered to be nonusers. PPI doses are classified according to the cumulative defined daily dose (cDDD) (25).

Informed consent for participation in this study was obtained from each participant or his or her legal surrogate. This study was approved by the Ethics Committee of the First Affiliated Hospital, School of Medicine, Zhejiang University, Zhejiang, China and all the procedures were in accordance with the latest version of the Declaration of Helsinki.

Demographic data, clinical history, reason for admission, physical examination findings, laboratory results, cirrhosis complications, events of organ failure, and prognosis were recorded. Potential precipitating factors for SBP and HE were noted. Survival at 28 days and 90 days following enrollment was recorded. Survival information was obtained by telephonic interview.

The Statistical Package for the Social Sciences (SPSS) version 26.0 (IBM Corporation, Armonk, New York, USA) was used for the data analysis. Categorical variables were summarized as numbers and percentages and compared by using the chi-squared test. Continuous variables were summarized as means ± SDs or medians (and interquartile ranges) and compared by using the Student's t-test or the Mann–Whitney U test. Variables found to be associated with SBP and HE on the univariate analysis (p < 0.10) were entered into the multivariate logistic regression analysis to identify the independent risk factors for SBP or HE. Baseline characteristics were compared between patients receiving and not receiving PPI therapy before and after propensity score matching (PSM). We performed PSM to adjust differences in baseline characteristics of decompensated cirrhosis and ACLF including gender, age, leukocyte count, platelet count, hemoglobin, serum bilirubin, albumin, aspartate aminotransferase (AST), ALT, INR, creatinine, serum sodium, the CLIF-SOFA score, and the model for end-stage liver disease (MELD) score by matching non-PPI users with comparable patients with PPI users in a ratio of 1:1 with a caliper of 0.1 of the SD. A standardized difference < 0.1 indicated a good balance between PPI and non-PPI groups. The multivariate Cox proportional hazards regression model was used to analyze the risk factors for 28-day and 90-day mortality. For the multivariate analysis, the probability for stepwise entry was at p = 0.05 and for removal at p = 0.10; only variables with p < 0.05 were retained in the final model. Statistical significance was at p ≤ 0.05.

A total of 489 acute decompensated patients with liver cirrhosis were selected in this study (Figure 1). Among these 489 patients (409 patients without ACLF and 80 patients with ACLF), PPI therapy was used in 299 (61.1%) patients: 256/409 (62.6%) patients without ACLF and 43/80 (53.8%) patients with ACLF (Figure 1; Table 1). The PPIs included omeprazole (n = 98; daily dose 40–80 mg), pantoprazole (n = 132; daily dose 40–80 mg), and lansoprazole (n = 69; daily dose 30–60 mg).

Proton pump inhibitor non-users (n = 190) and users (n = 299) did not differ significantly with respect to the baseline features (hepatitis B, 56.8 vs. 62.9%); (alcohol abuse, 20.0 vs. 16.7%); age (55.47 ± 13.21 vs. 53.60 ± 11.52 years); gender composition (male, 74.2 vs. 79.9%); prevalence of hypertension (20.0 vs. 15.7%); and history of HE (6.3 vs. 3.3%), bacterial infections (9.5 vs. 6.4%), or severe liver injury (12.1 vs. 8.0%) in the preceding 3 months. Gastrointestinal bleeding in the 3 months prior to admission was significantly more common in PPI users than in nonusers (24.4 vs. 15.3%, p = 0.015), while ascites in the 3 months prior to admission was significantly less common in PPI users (36.1 vs. 46.8%, p = 0.019). The incidence of HE (8.7 vs. 12.6%) and SBP (4.0 vs. 4.7%) during hospitalization was not differ significantly between PPI users and nonusers (Table 1). PPI users had significantly lower hemoglobin, serum bilirubin, ALT, INR, and the MELD score (p = 0, p = 0.041, p = 0.026, p = 0.007, and p = 0.017, respectively) (Table 1).

Then, we matched 489 patients with acute decompensated cirrhosis with 1:1. After PSM, there are 182 patients in PPI user group and non-user group, respectively (Table 1). Compared with the decompensated cirrhosis patients before PSM, some baseline characteristics have changed. PPI users were younger than nonusers (52.26 ± 11.17 vs. 55.48 ± 13.21 years, p = 0.020) included more male individuals (86.3 vs. 74.2%, p = 0.004) and fewer hypertension patients (11.5 vs. 20.3%, p = 0.022). Among PPI users, decompensated cirrhosis was more commonly caused by hepatitis B virus alone compared to PPI non-users (69.2 vs. 57.1%, p = 0.005) other than by other causes (14.3 vs. 22.5%, p = 0.043). In addition, PPI users had remarkably higher serum sodium level (p = 0.000), lower AST level, and the CLIF-SOFA score (p = 0.002 and p = 0.017). Other baseline characteristics did not change significantly.

We also analyzed 80 patients with ACLF in detail and matched them by the same method as above (Table 2). Among patients with ACLF, there were no significant differences between PPI users and non-users in mean age, gender composition, prevalence of hypertension, etiology of cirrhosis, incidence of previous decompensation, complications of cirrhosis during hospital stay, laboratory data, and scores of the MELD and the CLIF-SOFA. Similarly, PPI users and non-users had comparable 28-day mortality (27.8 vs. 45.9%, 0.482) and 90-day mortality rates (47.2 vs. 56.8%, 0.887). After PSM, other features did not change significantly, except for the lower prevalence of hypertension in PPI users (25.0 vs. 5.6%, p = 0.023).

The proportion of patients developing HE during hospitalization was comparable between PPI users and nonusers before and after PSM (8.7 vs. 12.6%, p = 0.162; 8.8 vs. 11.0%, p = 0.483) (Table 1). This was true even when patients with ACLF (14.0 vs. 16.2%, p = 0.779; 16.7 vs. 13.9%, p = 0.745) (Table 2). On the multivariate analysis, age [odds ratio (OR) = 1.07, 95% CI: 1.03–1.12, p = 0.001], albumin (OR = 1.13, 95% CI: 1.04–1.24, p = 0.006), history of previous HE (OR = 242.52, 95% CI: 40.17–1464.11, p < 0.001) and the CLIF-SOFA score (OR =2.89, 95% CI: 2.11–3.96, p < 0.001) were significantly associated with the risk of developing HE in decompensated cirrhosis patients during hospitalization (Table 3).

The proportion of patients developing SBP during hospitalization was comparable between PPI users and nonusers (4.0 vs. 4.7%, p = 0.701; 4.9 vs. 4.9%, p = 1) (Table 1). This was true even when patients with ACLF were considered separately (4.0 vs. 2.7%, p = 0.649; 5.6 vs. 2.8%, p = 0.558) (Table 2). On the multivariate analysis, history of previous severe liver injury (OR = 3.43, 95% CI: 1.16–10.17, p = 0.026) and previous bacterial infections (OR = 6.47, 95% CI: 2.29–18.29, p < 0.001) were independent risk factors for SBP in patients with decompensated cirrhosis (Table 3).

The cDDD of PPI in patients with decompensated cirrhosis and ACLF is summarized in Figure 2. The cDDDs of PPI in patients with decompensated cirrhosis and ACLF were divided into four groups: 1–7, 8–14, 15–30, and >30 patients. Among all the patients with decompensated cirrhosis, 25.15% were not treated with PPI. Especially, among all the patients with ACLF, 26.25% were not treated with PPI. Before and after PSM, the group of cDDD 15–30 patients accounted for the highest proportion of PPI users in both the decompensated cirrhosis group and ACLF group. But, the proportion of patients with cDDD > 30 patients was the smallest.

Then, we further discussed the relationship between the patients with HE and SBP during hospitalization and the cDDD of PPI (Figure 3). Before PSM, cDDD of 32% patients with HE and cDDD of 29% patients with SBP in decompensated cirrhosis were 15–30 patients, which had the highest proportion. The PPI non-user and cDDD 15–30 patients complicated with HE accounted for a high proportion of 33% in ACLF. After PSM of patients with decompensated cirrhosis, the PPI non-users and cDDD 15–30 patients complicated with HE accounted for a high proportion of 31%, while cDDD 1–7 patients and cDDD 15–30 patients with SBP accounted for a high proportion of 28%. ACLF patients with SBP had the same proportion in the three cDDD groups (1–7, 8–14, and 15–30 patients). Meanwhile, the type and dose of the PPI used were not related to the risk of developing HE or SBP (Table 3).

After median follow-up of 23 months (IQR 6–26 months), the 28-day and 90-day mortality rates were 11.7% (57/489) and 19.4% (95/489), respectively (Table 1; Figure 1).

The multivariate Cox proportional hazards regression model did not show significant association between the type and dose of the PPI and 28-day and 90-day mortality (Table 4; Figures 4, 5). The p-value of the cumulative incidence of mortality death between patients with decompensated liver cirrhosis and the dose of the PPI was 0.018 before PSM, but after PSM, there was no statistical significance. However, age, leukocyte count, and the high MELD score were independent predictors of 28-day and 90-day mortality in cirrhosis patients with AD (Table 4). In addition, previous severe liver injury was an independent predictor of 28-day mortality (Table 4).