Introduction
Liver cirrhosis represents the end stage of chronic liver disease, regardless of its etiology. Globally, complications of cirrhosis account for more than one million deaths annually.1,2 Patients with cirrhosis, particularly those with decompensated disease, are highly susceptible to bacterial infections, with an incidence ranging from 25% to 46% among hospitalized patients.3–7 These infections often lead to severe clinical consequences, including a fourfold increase in mortality among patients with decompensated cirrhosis,8 as well as a detrimental impact on long-term outcomes even in those with compensated disease.9 Moreover, bacterial infection is the most common trigger of acute-on-chronic liver failure (ACLF), resulting in a more severe clinical course and poorer outcomes compared to ACLF precipitated by other causes.4
Accurate epidemiological data on bacterial infections in patients with cirrhosis are critical for optimizing prevention, clinical management, and public health strategies, particularly in the context of increasing multidrug-resistant (MDR) bacteria.3,5,10–13 Furthermore, substantial geographic variability in infection profiles and treatment practices underscores the need for region-specific data, as emphasized by two recent global studies.3,14 However, such data remain scarce in China.
To address this gap, we conducted a multicenter cohort study to investigate the demographic, clinical, microbiological, and antibiotic treatment characteristics of patients with cirrhosis and bacterial infections in China. We also compared these findings with international data to identify potential region-specific differences.
Methods
Patient
This multicenter, retrospective study consecutively enrolled 1,438 hospitalized adult patients with cirrhosis and bacterial or fungal infections across 24 hospitals in China between January 2018 and September 2024. A detailed flowchart of patient selection is presented in Supplementary Figure 1. Patients were excluded if they met any of the following criteria: (1) hepatocellular carcinoma; (2) extrahepatic malignancy; (3) severe extrahepatic comorbidities, including congestive heart failure (New York Heart Association class ≥ III), chronic obstructive pulmonary disease (Global Initiative for Chronic Obstructive Lung Disease stage ≥ III), or chronic kidney disease requiring renal replacement therapy; (4) history of solid organ transplantation; (5) human immunodeficiency virus infection; or (6) use of immunosuppressive agents (excluding corticosteroids for liver-related indications) within one month prior to admission.
The study adhered to the Declaration of Helsinki and was approved by the Ethics Committee of the First Affiliated Hospital, School of Medicine, Zhejiang University (IIT20230123B-R1). Written informed consent was waived by the Ethics Committee.
Data collection and follow-up
Demographic, clinical, laboratory, microbiological, and treatment-related data at the time of infection diagnosis were retrieved from the electronic medical record systems of each participating hospital using a standardized, pre-specified data collection form. Additional in-hospital data included the occurrence of new bacterial or fungal infections, development of septic shock, ACLF, transfer to the intensive care unit (ICU), use of vasopressors, mechanical ventilation, and renal replacement therapy. For patients who developed a second infection during hospitalization, repeat microbiological cultures and antibiotic susceptibility testing were performed using previously described approaches.15,16
Pre-admission data were also collected, including: (1) antibiotic use within the preceding three months; (2) recent use of medications such as rifaximin, β-blockers, proton pump inhibitors (PPIs), or quinolone prophylaxis within the past month; (3) history of invasive procedures (e.g., surgery, central venous catheterization, indwelling urinary catheter, or paracentesis) in the prior three months; (4) ICU admission within the past week; and (5) any infection occurring within three months prior to index hospitalization. Patients were followed until death, liver transplantation, or their last available visit within 90 days of admission, whichever occurred first.
Definitions
Cirrhosis was diagnosed based on radiological evidence of a nodular liver contour, endoscopic signs of portal hypertension, or clinical evidence of hepatic decompensation.17 ACLF was defined according to the European Association for the Study of the Liver (EASL)–Chronic Liver Failure Consortium criteria.18
Diagnostic criteria for each type of infection are provided in Supplementary Table 1. A second infection was defined as a new nosocomial infection occurring after the initial infection during the same hospitalization.14,19 The diagnostic criteria were applied consistently across infection episodes.
A positive quick Sequential Organ Failure Assessment score was defined by the presence of at least two of the following: (1) altered mental status; (2) respiratory rate ≥ 22 breaths/m; and (3) systolic blood pressure ≤ 100 mmHg.20 Systemic inflammatory response syndrome was defined by the presence of at least two of the following: (1) body temperature < 36°C or > 38°C; (2) heart rate > 90 beats/m; (3) respiratory rate > 20 breaths/m; (4) white blood cell count < 4,000/mm3 or > 12,000/mm3; or (5) immature neutrophil count > 10%.21 Septic shock was diagnosed as sepsis with hypotension requiring vasopressors.22
MDR bacteria were defined as isolates resistant to at least one agent in three or more antimicrobial classes.10 Extensively drug-resistant (XDR) bacteria were defined as isolates susceptible to only one or two antimicrobial classes, while pan-drug-resistant bacteria were resistant to all currently available antibiotics.10
Empirical antibiotic regimens were categorized into two strategies: (1) classical strategies, including first- to third-generation cephalosporins, amoxicillin-clavulanic acid, cloxacillin, or quinolones; and (2) MDR-covered strategies, including piperacillin-tazobactam, carbapenems, or ceftazidime/cefepime with or without glycopeptides (or linezolid/daptomycin).23 Clinical response to empirical therapy was assessed by the treating physician based on symptom resolution, laboratory improvement, and microbiological results. Antibiotic regimens were considered “adherent” to EASL guidelines24 if at least one recommended antibiotic was used (Supplementary Table 2). Non-adherent regimens were further classified as “weaker” (narrower spectrum than recommended) or “broader” (wider spectrum than recommended). Antibiotic escalation was defined as the addition of at least one new agent or a switch to a broader-spectrum agent within five days. De-escalation was defined as a reduction in the number of antibiotics or a switch to a narrower-spectrum regimen within the same timeframe.
Statistical analysis
Continuous variables were expressed as mean ± standard deviation or median (IQR), and compared using Student’s t test, one-way ANOVA, or the Kruskal–Wallis test, as appropriate. Categorical variables were expressed as counts (percentages) and compared using Pearson’s chi-square test or Fisher’s exact test.
Multivariate Cox proportional hazards models were used to identify independent factors associated with in-hospital, 28-day, and 90-day mortality.25 Candidate variables included: age, sex (female as reference), diabetes, hypertension, cirrhosis etiology (hepatitis B virus as reference), recent infection, prior transjugular intrahepatic portosystemic shunt (TIPS), recent medication use, prior antibiotic exposure, recent invasive procedures, ICU admission, presence of ascites, hepatic encephalopathy (HE), model for end-stage liver disease (MELD) score, culture positivity (vs. negative), MDR isolates (vs. non-MDR), and empirical antibiotic strategy.
All statistical tests were two-tailed, and P-values < 0.05 were considered significant. Statistical analyses were performed using SPSS 26.0 (Chicago, IL) and R 4.3.1 (Vienna, Austria).
Results
Characteristics of the patients
A total of 1,438 patients with cirrhosis and bacterial or fungal infections were enrolled across 24 centers (Supplementary Fig. 2). As shown in Table 1, the median age was 61.0 years (IQR, 52.0–72.0), and 63.6% were male. A total of 362 (25.2%) and 377 (26.2%) patients had diabetes mellitus and hypertension, respectively. Hepatitis B virus was the predominant etiology of cirrhosis (44.6%), followed by alcohol-related liver disease (17.9%). Overall, 529 patients (36.8%) had recent infections, and 3.8% had a history of TIPS. The median hospital stay was 11.0 (IQR, 8.0–17.0) days. ACLF was diagnosed in 288 patients (20.0%). The median MELD and MELD-Na scores were 14.0 (IQR, 9.0–19.0) and 16.0 (IQR, 10.0–23.0), respectively. A total of 22 (1.5%), 142 (9.9%), and 230 (16.0%) patients died during hospitalization, at 28-day, and 90-day follow-up, respectively.
Table 1Baseline characteristics of the overall cohort of the first infection
| Characteristics | n = 1,438 |
|---|
| Age (y) | 61.0 (52.0–72.0) |
| Sex, n (%) |
| Male | 914 (63.6) |
| Female | 524 (36.4) |
| Diabetes mellitus, n (%) | 362 (25.2) |
| Hypertension, n (%) | 377 (26.2) |
| Etiology, n (%) |
| HBV | 641 (44.6) |
| Alcohol | 257 (17.9) |
| MASLD | 80 (5.6) |
| Cryptogenic | 131 (9.1) |
| Others | 329 (22.9) |
| Recent* infection, n (%) | 529 (36.8) |
| History of TIPS, n (%) | 54 (3.8) |
| Recent* medications, n (%) |
| Rifaximin | 24 (1.7) |
| β-blockers | 65 (4.5) |
| PPIs | 273 (19.0) |
| Quinolone prophylaxis | 77 (5.4) |
| Recent* antibiotic use, n (%) | 332 (23.1) |
| Invasive procedures, n (%) | 226 (15.7) |
| Recent* ICU admission, n (%) | 24 (1.7) |
| Length of hospital stay (day) | 11.0 (8.0–17.0) |
| MAP (mmHg) | 81.3 (72.7–92.0) |
| HR (bpm) | 92.0 (83.0–105.0) |
| Body temperature (°C) | 37.8 (37.1–38.6) |
| RR (breath/min) | 20.0 (19.0–21.0) |
| SpO2 | 97.0 (95.0–98.0) |
| Leukocytes (109/L) | 6.2 (3.6–10.5) |
| NLR | 5.0 (2.0–11.1) |
| C-reactive protein (mg/L) | 29.1 (10.5–68.0) |
| Albumin (g/L) | 28.5 (24.9–32.6) |
| TB (µmol/L) | 40.7 (19.7–101.2) |
| Cr (µmol/L) | 77.0 (60.0–110.4) |
| Serum sodium (mmol/L) | 136.8 (133.0–139.8) |
| INR | 1.4 (1.2–1.7) |
| Ascites, n (%) | 658 (45.8) |
| Hepatic encephalopathy, n (%) |
| Grades 1/2 | 147 (10.2) |
| Grades 3/4 | 133 (9.2) |
| Use of vasopressors, n (%) | 162 (11.3) |
| Transfer to ICU, n (%) | 140 (9.7) |
| Mechanical ventilation, n (%) | 86 (6.0) |
| Renal replacement therapy, n (%) | 40 (2.8) |
| ACLF, n (%) | 288 (20.0) |
| MELD score | 14.0 (9.0–19.0) |
| MELD-Na score | 16.0 (10.0–23.0) |
| Second infection, n (%) | 99 (6.9) |
| In-hospital mortality, n (%) | 22 (1.5) |
| 28-day mortality, n (%) | 142 (9.9) |
| 90-day mortality, n (%) | 230 (16.0) |
Characteristics of the first infection
Among the first documented infections (Table 2), 86.6% were community-acquired and 13.4% were nosocomial. Pneumonia was the most prevalent infection (26.7%), followed by spontaneous bacterial peritonitis (SBP, 19.5%) and spontaneous bacteremia (14.1%). At diagnosis, 54.2% of patients met the criteria for systemic inflammatory response syndrome, 7.2% had a quick Sequential Organ Failure Assessment score ≥ 2, and 8.1% presented with septic shock.
Table 2Clinical and microbiological characteristics of the first infection
| Characteristics | n = 1,438 |
|---|
| Type of infection, n (%) |
| Community acquired | 1,245 (86.6) |
| Nosocomial | 193 (13.4) |
| Site of infection per infection, n (number of infections/%) |
| SBP | 326 (19.5) |
| Pneumonia | 445 (26.7) |
| UTI | 134 (8.0) |
| Spontaneous bacteremia | 236 (14.1) |
| Skin and soft tissue | 96 (5.8) |
| Bacterial entero-colitis | 44 (2.6) |
| Cholangitis | 118 (7.1) |
| Others | 154 (9.3) |
| Unproven bacterial infection | 115 (6.9) |
| Severity of infection, n (%) |
| SIRS | 780 (54.2) |
| qSOFA | 104 (7.2) |
| Septic shock | 117 (8.1) |
| Patients with positive cultures, n (%) | 620 (43.1) |
| Isolates per patient, n (%) |
| 1 | 506 (35.2) |
| >1 | 114 (7.9) |
| Type of strains isolated, n (%)† |
| Gram-negative | 441 (58.5) |
| Gram-positive | 212 (28.1) |
| Fungi | 101 (13.4) |
| Most frequently isolated bacteria, n (%)‡ |
| Escherichia coli | 142 (21.7) |
| Klebsiella pneumoniae | 131 (20.1) |
| Staphylococcus aureus | 58 (8.9) |
| Acinetobacter baumannii | 31 (4.7) |
| Enterococcus faecium | 28 (4.3) |
| Pseudomonas aeruginosa | 22 (3.4) |
| Enterococcus faecalis | 14 (2.1) |
| MDR, n (%)‡ | 268 (41.0) |
| XDR, n (%)‡ | 16 (2.5) |
| Type of empirical antibiotic strategies, n (%)§ |
| Classical | 484 (34.1) |
| MDR coverage | 936 (65.9) |
| Change of antibiotic treatment, n (%) |
| Escalation | 426 (30.0) |
| De-escalation | 119 (8.4) |
| No change | 877 (61.7) |
| Clinical resolution, n (%) | 913 (63.5) |
Culture tests identified 754 microorganisms from 620 patients, with more than one species isolated in 7.9% of cases. Detailed isolate data by center are provided in Supplementary Table 3. Overall, gram-negative bacteria were the most common (58.5%), followed by gram-positive bacteria (28.1%), and fungi accounted for approximately 13%. The most frequent isolates were Escherichia coli (21.7%), Klebsiella pneumoniae (20.1%), and Staphylococcus aureus (8.9%). The prevalence of MDR and XDR isolates was 41.0% and 2.5%, respectively.
Initial empirical antibiotic regimens included MDR coverage in 65.9% of cases, while 34.1% received classical regimens. During treatment, 61.7% of patients remained on the initial regimen, 30.0% required escalation, and 8.4% underwent de-escalation. The clinical resolution rate was 63.5%.
The 24 participating centers were categorized into three tiers: central, regional, and county-level hospitals (Supplementary Table 4). Among them, central hospitals had the highest rate of positive cultures (47.7%), followed by regional (40.0%) and county-level hospitals (37.4%). Despite differences in culture positivity, the prevalence of MDR and XDR organisms did not significantly differ across hospital tiers.
Characteristics of the second infection
A total of 115 second infection episodes occurred in 99 patients (6.9%) (Supplementary Table 5). The most common types were pneumonia (21.7%), urinary tract infection (UTI, 19.1%), and spontaneous bacteremia (19.1%). The culture positivity rate was 69.7%, with polymicrobial infections identified in 17.2% of cases. The most frequently isolated pathogens included Enterococcus faecium (20.1%), Klebsiella pneumoniae (19.0%), and Escherichia coli (11.4%). MDR bacteria were detected in 51.9% of cases, and XDR bacteria in 5.1%. Notably, pan-drug-resistant bacteria were identified in 97 cases of second infections.
Prevalence and types of MDR/XDR bacteria
A total of 284 MDR and XDR bacterial isolates were identified from 259 patients across 284 infection episodes (Table 3). Among gram-negative organisms, extended-spectrum β-lactamase (ESBL)-producing Escherichia coli was the most frequently detected MDR pathogen, accounting for 7.7% of all isolates, 8.9% of patients, and 8.5% of infection episodes. This was followed by ESBL-producing Klebsiella pneumoniae, which represented 2.3% of isolates, 2.7% of patients, and 2.6% of episodes. Among gram-positive organisms, vancomycin-susceptible enterococci and methicillin-resistant Staphylococcus aureus (MRSA) were the most common. Vancomycin-susceptible enterococci was identified in 26 isolates (4.0%), affecting 26 patients (4.6%) and corresponding to 26 infection episodes (4.2%). MRSA was found in 20 isolates (3.1%), involving 20 patients (3.6%) and 22 infection episodes (3.5%).
Table 3Prevalence of MDR/XDR bacteria and specific types
| Strains | Number of isolates | Number of patients | Number of episodes of infections | Incidence (%)
|
|---|
| Per isolate (n = 653) | Per patient (n = 560) | Per episode (n = 627) |
|---|
| Total | 284 | 259 | 284 | 43.5 | 46.3 | 45.3 |
| MDR | 268 | 243 | 265 | 41.0 | 43.4 | 42.3 |
| ESBL-E | 50 | 50 | 53 | 7.7 | 8.9 | 8.5 |
| ESBL-KP | 15 | 15 | 16 | 2.3 | 2.7 | 2.6 |
| CRKP | 5 | 5 | 5 | 0.8 | 0.9 | 0.8 |
| CRPA | 4 | 4 | 4 | 0.6 | 0.7 | 0.6 |
| CRE | 6 | 6 | 6 | 0.9 | 1.1 | 1.0 |
| CRAB | 4 | 4 | 4 | 0.6 | 0.7 | 0.6 |
| MRSA | 20 | 20 | 22 | 3.1 | 3.6 | 3.5 |
| VSE | 26 | 26 | 26 | 4.0 | 4.6 | 4.2 |
| XDR | 16 | 16 | 19 | 2.5 | 2.9 | 3.0 |
Factors associated with mortality
A total of 22, 142, and 230 deaths occurred during hospitalization and at 28-day and 90-day follow-up, respectively. In the multivariable Cox regression analysis (Table 4), the following were independently associated with in-hospital mortality: hypertension (adjusted hazard ratio [aHR], 3.888; 95% CI, 1.403–10.770; P = 0.009), recent infection (aHR, 5.567; 95% CI, 1.689–18.346; P = 0.005), recent quinolone prophylaxis (aHR, 5.489; 95% CI, 1.045–28.837; P = 0.044), higher MELD score (aHR, 1.086; 95% CI, 1.048–1.126; P < 0.001), and HE grade 3/4 (aHR, 5.425; 95% CI, 1.890–15.565; P = 0.002).
Table 4Factors associated with in-hospital/28-day/90-day death in patients with cirrhosis and infection
| Variables | Univariate analysis
| Multivariate analysis
|
|---|
| HR | 95% CI | P-value | HR | 95% CI | P-value |
|---|
| In-hospital |
| Hypertension | 2.373 | 1.024–5.497 | 0.044 | 3.888 | 1.403–10.770 | 0.009 |
| Recent infection | 3.523 | 1.436–8.644 | 0.006 | 5.567 | 1.689–18.346 | 0.005 |
| Recent quinolone prophylaxis | 2.355 | 0.692–8.018 | 0.170 | 5.489 | 1.045–28.837 | 0.044 |
| Recent antibiotic use | 1.533 | 0.639–3.679 | 0.338 | 0.229 | 0.059–0.885 | 0.033 |
| HE grade 3/4 | 5.863 | 2.344–14.665 | <0.001 | 5.425 | 1.890–15.565 | 0.002 |
| MELDs | 1.081 | 1.048–1.116 | <0.001 | 1.086 | 1.048–1.126 | <0.001 |
| 28-day |
| Cryptogenic etiology | 1.480 | 0.887–2.469 | 0.134 | 2.252 | 1.258–4.030 | 0.006 |
| History of TIPS | 1.627 | 0.797–3.321 | 0.181 | 2.193 | 1.043–4.612 | 0.038 |
| Recent use of PPIs | 0.808 | 0.517–1.263 | 0.349 | 0.586 | 0.348–0.985 | 0.044 |
| HE | 1.914 | 1.144–3.202 | 0.013 | 1.874 | 1.102–3.186 | 0.020 |
| MELDs | 1.082 | 1.071–1.093 | <0.001 | 1.092 | 1.046–1.139 | <0.001 |
| 90-day |
| Other etiology | 1.078 | 0.777–1.497 | 0.652 | 1.483 | 1.009–2.179 | 0.045 |
| History of TIPS | 1.681 | 0.961–2.943 | 0.069 | 1.984 | 1.104–3.563 | 0.022 |
| Little or no ascites | 0.471 | 0.280–0.793 | 0.005 | 0.535 | 0.313–0.913 | 0.022 |
| HE | 2.395 | 1.661–3.455 | <0.001 | 2.138 | 1.464–3.122 | <0.001 |
| MELDs | 1.070 | 1.061–1.080 | <0.001 | 1.076 | 1.064–1.088 | <0.001 |
| Positive cultures | 1.720 | 1.167–2.534 | 0.006 | 1.772 | 1.002–3.134 | 0.049 |
Independent risk factors for 28-day mortality included: cryptogenic cirrhosis (aHR, 2.252; 95% CI, 1.258–4.030; P = 0.006), a history of TIPS (aHR, 2.193; 95% CI, 1.043–4.612; P = 0.038), recent use of PPIs (aHR, 0.586; 95% CI, 0.348–0.985; P = 0.044), HE grade 3/4 (aHR, 1.874; 95% CI, 1.102–3.186; P = 0.020), and MELD score (aHR, 1.092; 95% CI, 1.046–1.139; P < 0.001). At 90-day follow-up, independent predictors of mortality included: other cirrhosis etiologies (aHR, 1.483; 95% CI, 1.009–2.179; P = 0.045), a history of TIPS (aHR, 1.948; 95% CI, 1.104–3.563; P = 0.022), presence of minimal or no ascites (aHR, 0.535; 95% CI, 0.313–0.913; P = 0.022), HE (aHR, 2.138; 95% CI, 1.464–3.122; P < 0.001), MELD score (aHR, 1.076; 95% CI, 1.064–1.088; P < 0.001), and a positive microbiological culture result (aHR, 1.772; 95% CI, 1.002–3.134; P = 0.049).
Comparison of clinical and microbiological characteristics between culture-positive and culture-negative infections
Among the 1,188 patients who underwent culture testing, 620 (52.2%) had positive results. Supplementary Table 6 summarizes the clinical characteristics of patients with culture-positive versus culture-negative infections at the time of their first documented infection. Compared to those with culture-negative results, culture-positive patients had a higher prevalence of diabetes mellitus (29.2% vs. 22.0%, P = 0.005), more frequent recent infections (41.8% vs. 34.3%, P = 0.008), and a greater history of prior antibiotic exposure (26.3% vs. 20.6%, P = 0.021). In terms of infection severity, culture-positive patients were more likely to present with ACLF (25.7% vs. 15.8%, P < 0.001) and had significantly higher MELD and MELD-Na scores.
As shown in Supplementary Table 7, community-acquired infections were more common among culture-negative patients (88.2% vs. 83.4%, P = 0.018), who also had a significantly lower incidence of septic shock compared to those with positive cultures (5.1% vs. 16.6%, P < 0.001). Although empirical antibiotic strategies at baseline did not differ significantly between groups, patients with culture-positive infections were more likely to undergo antibiotic escalation (38.8% vs. 25.8%, P < 0.05), and their clinical resolution rates were notably lower (53.5% vs. 70.1%, P < 0.001).
Patients with positive cultures had significantly higher in-hospital mortality (2.6% vs. 0.7%, P = 0.012). Similarly, culture-negative patients demonstrated improved survival at both 28-day and 90-day follow-up (Supplementary Fig. 3B and C).
Comparison of clinical and microbiological characteristics of bacterial infections: Chinese versus global cohorts
As shown in Table 5, the proportion of community-acquired infections in the Chinese cohort was significantly higher than that reported in the global study (86.6% vs. 74.0%, P < 0.001). Compared with patients in the global cohort, Chinese patients were more likely to present with pneumonia (28.7% vs. 18.9%, P < 0.05) and spontaneous bacteremia (15.2% vs. 7.8%, P < 0.05), but were less likely to develop SBP (21.0% vs. 27.6%, P < 0.05) or urinary tract infections (8.6% vs. 22.5%, P < 0.05). Culture-positive infections were significantly less frequent in the Chinese cohort (43.1% vs. 56.8%, P < 0.001). Klebsiella pneumoniae was more commonly isolated in China (20.1% vs. 15.5%, P < 0.05), while Escherichia coli (21.7% vs. 28.9%, P < 0.05) and Enterococcus faecalis (2.1% vs. 5.6%, P < 0.05) were less frequently detected.
Table 5Comparison of the characteristics of infections between patients in the study and those in the global study
| Characteristics | China (n = 1,438) | Global (n = 1,302)† | P-value |
|---|
| Type of infection, n (%) | <0.001 |
| Community acquired | 1,245 (86.6) | 964 (74.0) | |
| Nosocomial | 193 (13.4) | 338 (26.0) | |
| Site of infection, n (%) | <0.001 |
| Unproven bacterial infection | 115 (6.9) | 20 (1.5) | |
| Proven site | 1,553 (93.1) | 1,282 (98.5) | |
| SBP | 326 (21.0) | 354 (27.6) | <0.05 |
| Pneumonia | 445 (28.7) | 242 (18.9) | <0.05 |
| UTI | 134 (8.6) | 289 (22.5) | <0.05 |
| Spontaneous bacteremia | 236 (15.2) | 100 (7.8) | <0.05 |
| Skin and soft tissue | 96 (6.2) | 101 (7.9) | >0.05 |
| Bacterial entero-colitis | 44 (2.8) | 31 (2.4) | >0.05 |
| Cholangitis | 118 (7.6) | 37 (2.9) | <0.05 |
| Others | 154 (9.9) | 128 (10.0) | >0.05 |
| Severity of infection, n (%) |
| SIRS | 780 (54.2) | 405 (36.2) | <0.001 |
| qSOFA | 104 (7.2) | 255 (22.8) | <0.001 |
| Septic shock | 117 (8.1) | 174 (13.4) | <0.001 |
| Patients with positive cultures, n (%) | 620 (43.1) | 740 (56.8) | <0.001 |
| Isolates per patient, n (%) | <0.001 |
| 1 | 506 (35.2) | 592 (45.5) | |
| >1 | 114 (7.9) | 148 (11.4) | |
| Type of strains isolated, n (%) |
| Gram-negative | 441 (58.5) | 561 (58.5) | >0.05 |
| Gram-positive | 212 (28.1) | 360 (37.5) | <0.05 |
| Fungi | 101 (13.4) | 38 (4.0) | <0.05 |
| Most frequently isolated bacteria, n (%) |
| Escherichia coli | 142 (21.7) | 266 (28.9) | <0.05 |
| Klebsiella pneumoniae | 131 (20.1) | 143 (15.5) | <0.05 |
| Staphylococcus aureus | 58 (8.9) | 78 (8.5) | >0.05 |
| Enterococcus faecalis | 14 (2.1) | 52 (5.6) | <0.05 |
| Enterococcus faecium | 28 (4.3) | 53 (5.8) | >0.05 |
| MDR bacteria, n (%) | 268 (41.0) | 322 (35.0) | 0.014 |
| ESBL- Enterobacteriaceae | 50 (18.7) | 89 (27.6) | <0.05 |
| CRE | 6 (2.2) | 35 (10.9) | <0.05 |
| Acinetobacter baumannii | 7 (2.6) | 19 (5.9) | >0.05 |
| MRSA | 20 (7.5) | 14 (4.3) | >0.05 |
| VRE | 0 (0.0) | 16 (5.0) | <0.05 |
| XDR bacteria, n (%) | 16 (2.5) | 73 (7.9) | <0.05 |
Although the overall prevalence of MDR bacteria was higher in China (41.0% vs. 35.0%, P = 0.014), the incidence of XDR organisms was significantly lower (2.5% vs. 7.9%, P < 0.05). Regarding specific resistance patterns, carbapenem-resistant Enterobacteriaceae and vancomycin-resistant enterococci were less common in China, whereas MRSA was more frequently isolated.
A detailed comparison across geographic regions—including China, the United States, Asia (excluding China), and Europe—is presented in Supplementary Table 8 and yielded consistent findings regarding the epidemiological characteristics of Chinese patients.
Comparison of antibiotic treatment practices: Chinese versus global cohorts
As shown in Table 6, adherence to the EASL guidelines for empirical antibiotic therapy was significantly lower in the Chinese cohort compared with the global cohort (21.5% vs. 61.2%, P < 0.001). Adherence rates were highest in the United States (65.0%) and Europe (64.0%) (Supplementary Table 9). Chinese patients were more likely to receive broader-spectrum empirical antibiotics beyond guideline recommendations (75.1% vs. 35.5%, P < 0.001). Specifically, β-lactamase inhibitors such as piperacillin–tazobactam and cefoperazone–sulbactam (39.6%) and carbapenems (17.6%) were more frequently used in China. In contrast, classical β-lactamases/β-lactamase inhibitor combinations (e.g., amoxicillin–clavulanic acid or ampicillin–sulbactam, 5.5%) and third-generation cephalosporins (19.7%) were prescribed less commonly (Supplementary Table 10). Notably, the clinical resolution rate was significantly lower in the Chinese cohort compared to the global population (63.5% vs. 79.8%, P < 0.001).
Table 6Comparison of antibiotic treatment practices in Chinese patients with cirrhosis versus those in the global study
| Characteristics | China (n = 1,438) | Global (n = 1,302)† | P-value |
|---|
| In vitro susceptibility to empirical antibiotic treatment, n (%) | 0.008 |
| Susceptible | 400 (76.8) | 500 (69.9) | |
| Non-susceptible | 121 (23.2) | 215 (30.1) | |
| Adherence to the EASL empirical antibiotic treatment recommendations, n (%) |
| Adherence | 267 (21.5) | 796 (61.2) | <0.001 |
| Non-adherence | 976 (78.5) | 504 (38.8) | <0.001 |
| Weaker | 243 (24.9) | 325 (64.5) | <0.001 |
| Broader | 733 (75.1) | 179 (35.5) | <0.001 |
| Change of antibiotic treatment, n (%) |
| Escalation | 426 (30.0) | 477 (36.6) | <0.05 |
| De-escalation | 119 (8.4) | 102 (7.8) | >0.05 |
| No change | 877 (61.7) | (55.5) | <0.05 |
| Clinical resolution, n (%) | 913 (63.5) | 1,038 (79.8) | <0.001 |
Discussion
In this large, multicenter cohort study, we characterized the demographic, clinical, and microbiological profiles of patients with cirrhosis and bacterial infections in China and compared these findings with global data to highlight regional disparities. Patients in the Chinese cohort exhibited a high prevalence of community-acquired infections, with non-SBP infections being the predominant type. Notably, the emergence of Klebsiella pneumoniae as a leading pathogen and the high rate of MDR isolates (41.0%) were concerning trends. Even more alarming was the low adherence to clinical practice guidelines for empirical antibiotic therapy, which was associated with a suboptimal clinical resolution rate. These findings underscore the urgent need for region-specific strategies for infection prevention and treatment optimization.
Compared to the global cohort, Chinese patients were more likely to develop pneumonia, a trend also observed in other Asian populations when data were stratified by region.26 This higher incidence aligns with previous studies3,14,26,27 and may be attributed to factors such as high population density, environmental or climatic conditions, and lifestyle behaviors like smoking.14,26 Additionally, Chinese patients had a significantly higher proportion of spontaneous bacteremia compared to other regions, a concerning finding given the strong association between bloodstream infections and poor outcomes in patients with cirrhosis.28 In contrast, UTIs were markedly less frequent in this cohort, a trend consistent with prior Chinese studies.27 This may be partly explained by fewer ICU admissions and, consequently, less frequent use of indwelling urinary catheters, a known risk factor for UTIs, especially in Western countries.6,27
Several microbiological trends warrant attention. Gram-negative bacteria were the most common pathogens, with Enterobacteriaceae being the predominant isolate, consistent with global data. Notably, Klebsiella pneumoniae emerged as the second most frequent isolate, nearly as common as Escherichia coli. Infections caused by Klebsiella pneumoniae are associated with increased mortality, prolonged hospital stays, and higher healthcare costs due to its virulence, resistance, and transmissibility.29–31 Of particular concern is the emergence of carbapenem-resistant hypervirulent Klebsiella pneumoniae, which is both MDR and highly transmissible.32–36 This poses a significant public health threat, especially for cirrhotic patients who are inherently more susceptible to severe infections.
Timely surveillance and targeted intervention strategies are urgently needed to curb its spread.
Another major concern was the high prevalence of MDR bacteria, which reached 41.0%, exceeding global estimates, despite the predominance of community-acquired infections in our cohort. Among second infections, exclusively nosocomial, MDR isolates accounted for 51.9% of cases. This likely contributes to the substantial negative impact of secondary infections on survival among cirrhotic patients.19 Regarding resistance patterns, ESBL-producing Enterobacteriaceae remained the most common, consistent with global data. However, we also observed a significantly higher proportion of MRSA isolates in our study population. The increasing prevalence of Staphylococcus aureus has been linked to the widespread use of invasive medical procedures and exogenous sources of infection.37 This is particularly concerning given MRSA’s broad array of virulence factors, its ability to acquire resistance, and its potential to generate novel clones.38 A prior study has shown that MRSA infections are associated with significantly higher mortality in patients with cirrhosis compared to infections caused by other bacterial species.39
Our findings also highlighted important gaps in clinical practice. Adherence to EASL guidelines for empirical antibiotic therapy was substantially lower in the Chinese cohort, with a strong preference for broader-spectrum agents. Paradoxically, this more aggressive antibiotic approach did not translate into improved clinical outcomes, as evidenced by a significantly lower clinical resolution rate compared to the global cohort. While this discrepancy may reflect differences in patient characteristics or definitions of clinical resolution, it nonetheless underscores the urgent need to optimize empirical antibiotic strategies. For example, the antibiotic piperacillin–tazobactam, which was frequently used in our cohort, is known to be suboptimal against ESBL-producing Enterobacteriaceae,40 the most common MDR bacteria identified. Second, the culture-positive rate in our cohort was 43.1%, which was lower than that reported in the global study. This discrepancy may be attributed to differences in infection types. For instance, compared with the global cohort, our study included a lower proportion of UTIs and a higher proportion of pneumonia and unproven infections—conditions that are often diagnosed clinically without a confirmed microbiological culture. Third, the widespread use of PPIs, which has been associated with an increased risk of bacterial infections,41,42 along with the underutilization of rifaximin, β-blockers, and quinolone prophylaxis, which are known to reduce microbial translocation and the risk of bacterial infections,43–46 may have contributed to the high infection rate observed in this predominantly community-acquired setting.
This study has several limitations. First, its retrospective design may introduce selection bias. Additionally, heterogeneity in diagnostic and therapeutic practices across participating centers, along with the absence of centralized microbiological testing, could have led to misclassification of infection sources and inconsistencies in sample quality and clinical management.47 Second, the lack of long-term follow-up data limits our ability to evaluate extended outcomes. This is particularly relevant given prior evidence indicating that up to 63% of cirrhotic patients with infections may die within one year.8 Third, we did not adjust for potential confounders such as age, sex, and liver disease etiology when comparing the epidemiological patterns in our cohort to those in the global study. Lastly, variations in local healthcare infrastructure, resource availability, and prescribing practices across different countries and hospitals may influence adherence to international guidelines. Furthermore, the local epidemiology of infections may influence empirical antibiotic choices, which could confound the interpretation of guideline adherence.