Article Text
PDF
Abstract
Objective To assess incidence and prognostic significance of bacterial infections (BIs) occurring in compensated viral cirrhosis.
Design This prospective study involved 35 French centres. Inclusion criteria were biopsy-proven HCV or HBV cirrhosis, Child–Pugh A and no previous hepatic complications. Cumulative incidence (CumI) of events was estimated in a competing risks framework.
Results 1672 patients were enrolled (HCV 1323, HBV 318, HCV-HBV 31). During a median follow-up of 43 months, 234 BIs occurred in 171 patients (5 year CumI: 12.9%), among whom 14.6% had septic shock. Main localisations included the urinary tract (27.4%), lung (25.2%) and peritoneum (10.7%) (other, 86 (36.7%)). Most BIs occurred as a first event prior to liver decompensation (n=140, 81.8%) and were community-acquired (CA, 84.2%). The risk of BI was higher in patients with HCV than in patients with HBV (5 year CumI: 15.2% vs 5.5%, p=0.0008). Digestive localisation, concomitant interferon-based treatment, isolation of resistant bacteria and non-CA BIs were associated with lowest probability of resolution. The occurrence of a first BI impaired survival in patients infected with HCV (5 year survival: 60.2% vs 90.4%, p<0.001) and patients infected with HBV (5 year survival: 69.2% vs 97.6%, p<0.001). BIs represented the third cause of death (14.1%) after liver failure and liver cancer. BI risk factors comprised older age, lower albumin, proton pump inhibitor intake and absence of virological eradication/control.
Conclusion BI mostly occurs as a first complication and represents a turning point in the course of compensated viral cirrhosis. Its occurrence impacts long-term prognosis and may define a subgroup of patients in whom adaptation of management is warranted.
- CIRRHOSIS
- HEPATITIS B
- HEPATITIS C
- BACTERIAL INFECTION
Statistics from Altmetric.com
Significance of this study
What is already known on this subject?
The occurrence of bacterial infection (BI) in the course of cirrhosis is one of the leading causes of death.
However, most available data on BI have thus far been restricted to patients with impaired liver function.
The incidence and prognostic significance of BI in compensated viral cirrhosis are unknown.
What are the new findings?
BI occurs as frequently as liver-related complications in patients with viral-compensated cirrhosis.
Patients experiencing a first BI have a higher probability of long-term hepatic decompensation and death.
BI in compensated viral cirrhosis is favoured by virological replication and proton pump inhibitors intake.
How might it impact on clinical practice in the foreseeable future?
BI is a turning point that may define a specific subgroup of compensated patients in whom adaptation of long-term therapeutic management (including transplantation) could be discussed.
Identification of novel biomarkers may help in identifying individuals at higher risk of BI, in whom personalised management constitutes an exciting challenge.
The occurrence of bacterial infection (BI) in the course of cirrhosis is common and is one of the leading causes of hospitalisation and death in these patients.1 BI often triggers hepatic complications and dramatically increases mortality in patients with decompensated cirrhosis.2 Not surprisingly, most of these BIs occur in patients with impaired hepatocellular function in whom repeated hospitalisation for liver-related complications further increases the risk of infections caused by multidrug-resistant organisms (MDROs).3 However, most available data on BI have thus far been restricted to patients admitted for an episode of sepsis or hepatic decompensation, in whom previous liver status is generally unknown.4 Furthermore, all studies conducted in this field incorporated very heterogeneous populations and assessed only short-term prognosis. In other terms, the characteristics of these infections have thus far been examined only through the prism of acute-on-chronic liver failure (ACLF) or end-stage liver disease, clinical settings that do not cover the entire spectrum of the course of patients with cirrhosis. The outcome in patients with cirrhosis has substantially improved over the past few decades due to widespread use of broad-spectrum antibiotics, and continuous development of invasive procedures such as endoscopic or interventional radiology treatments; such progress results in longer survival and also major changes in the epidemiology of BI in this population.4 The history of BI occurring in compensated patients with cirrhosis, its specific burden and its interaction with other critical events (liver-related or not), along with comorbidities impacting the long-term prognosis of these patients, remain to be described.
The ANRS CO12 CirVir cohort was specifically designed to prospectively assess the incidence of, burden and predictive factors for complications occurring in a large multicentre population of HCV-infected or HBV-infected compensated patients with cirrhosis with biopsy-proven cirrhosis. Baseline characteristics of the ANRS CO12 CirVir cohort and a brief description of the first events occurring during follow-up have been reported.5 In the present study, specific focus on BI occurring during a longer follow-up of this population was undertaken. The main objectives were to report the incidence and characteristics of BI, their predictive factors and their influence on long-term prognosis, as well as their burden on mortality, taking into account competing risks of death in this population. The impact of virological control or clearance at end point as modifiers of BI incidence was also examined.
Methods
Patient selection
The present work is an ancillary study derived from the CirVir cohort5 with specific goals and objectives redefined according to STROBE statement.6 Patients were recruited in 35 French clinical centres between 2006 and 2012. Selection criteria were: (1) age older than 18 years; (2) histologically proven cirrhosis; (3) chronic infection with HCV or HBV; (4) absence of previous complications of cirrhosis and (5) patients belonged to Child–Pugh class A.
Follow-up
Patients were seen every 6 months in the setting of hepatocellular carcinoma (HCC) screening. All events occurring during follow-up, liver-related or not, were recorded based on information obtained from medical files. Likely causes of death were established. Patients who underwent liver transplantation were censored at the date of transplantation for analysis. All treatments were recorded at inclusion. Antiviral therapy and response criteria were defined according to international recommendations.7–9 All medical diagnoses of events occurring during follow-up were confirmed by two senior hepatologists (authors VB and PN) and a bacteriologist (author ML) and include an extensive microbiological description.
Bacterial infection
All episodes of BI occurring during follow-up of patients were recorded, with criteria for diagnosis of infections as follows:4 urinary tract infection (UTI): more than 10 leucocytes per high-power field in urine and positive urine cultures or significant leucocyte count per field without positive cultures.10 Pneumonia was defined as the presence of radiologic evidence of consolidation, plus at least two of the following criteria: fever higher than 38°C or hypothermia less than 35°C, dyspnoea, cough and purulent sputum, pleuritic chest pain or signs of consolidation on physical examination.11 Digestive tract infections comprised spontaneous bacterial peritonitis (SBP, polymorphonuclear cell count in ascitic fluid >250/mm3).12 Other systemic infection encompassed infections with systemic damage such as catheter-related infection (positive blood and catheter cultures) or osteoarticular infections (positive blood and/or bone biopsy or joint aspiration). Diagnosis of other infections was made according to conventional criteria.13 When bacteraemia was detected in a patient with UTI, pneumonia, digestive infection, SBP or other BI, it was interpreted as secondary to these infections and defined by the primary infection. All episodes of BI were classified, according to clinical severity and biological results, as follows: (1) infections defined by the absence of sepsis (eg, cystitis or bronchic super-infection); (2) infections with no recorded fever or positive blood culture but possibly complicated by sepsis; (3) infections with isolated fever; (4) infections with positive blood culture and (5) septic shock, as defined by international criteria.14 Non-community-acquired (non-CA) BI encompassed hospital-acquired infections when the infection was declared >48 h after hospitalisation, and healthcare-associated infections if related to a medical procedure; in all other cases, BI was considered to be CA.15 When a bacterium was isolated, the monomicrobial or polymicrobial nature of the infection was indicated. Susceptibility to antibiotics was assessed by the disc diffusion method, as recommended (http://www.sfm.asso.fr/). We evaluated the number of MDRO in the following species: Enterobacteriaceae family members, Pseudomonas aeruginosa, Acinetobacter baumannii, Staphylococcus aureus, Enterococcus faecalis and Enterococcus faecium, for which the information was available as described elsewhere.16 We also evaluated the number of quinolone-resistant bacteria (QRB). Resolution of BI was defined as the absence of death directly attributable to the episode of BI.
Statistical analyses
Statistical analyses were performed using Stata (V.13.0, StataCorp, College Station, Texas, USA). All analyses were performed in the entire CirVir cohort and because of their specific characteristics in each subgroup of patients infected with HCV or patients infected with HBV. BI occurrence and overall death were the main outcomes. The cumulative incidence (CumI) of BI was estimated by the Kaplan–Meier method. This incidence was graphically compared with the CumI function (CIF) built into a competing risk framework, where deaths free of BI were considered as competing with the infectious event. Since differences in CumIs from the two methods quoted above are very low, predictive analysis of factors associated with risk of BI were tested using univariate and multivariate Cox models. Similar analyses were used for modelling risk of overall death. A p value ≤0.05 was considered statistically significant. Factorial analyses of correspondence (FAC) were computed in order to study: (1) relationships between clinical features (localisation, severity, concomitant interferon (IFN) regimen-treatment, CA or non-CA characteristics and resolution or not) and isolation of bacteria or not and (2) then restricting these explorations in BI with available microbiological data (monomicrobial or polymicrobial character, type of bacteria, QRB or not and MDRO or not), according to the method developed by Greenacre.17
Results
Inclusion period and data analysis
A total of 1822 patients were included. Among them, 150 were subsequently excluded from analysis after revision of individual data due to either non-compliance with inclusion criteria (n=141) or consent withdrawal (n=9). Final analyses were performed in 1672 patients. For analysis, the reference date was 16 January 2014. Median follow-up was 43 months (range: 16.4–63.2).
Patient baseline characteristics
Patients with HCV cirrhosis had a higher prevalence of comorbidities. Although most patients were undergoing antiviral therapy at inclusion or had previously undergone it, patients with HCV had higher rates of detectable viral load at the time of inclusion compared with patients with HBV (table 1 and see online supplementary table S1). Median HBV viral load in viraemic patients infected with HBV was 774 (68–34 125) IU/mL.
Characteristics of patients at inclusion according to cause of cirrhosis
Characteristics and incidence of BI
A total of 234 BIs occurred in171 patients (table 2). The 5 year CumI of a first episode of BI was 13.1% in the whole cohort. The risk of developing a first episode of BI was higher in patients with HCV than in patients with HBV (5 year CumI: 15.2% vs 5.5%, p=0.0008, figure 1A). The main locations of the 234 BIs were UTI and pulmonary infections, followed by digestive infections (characterised by a majority of SBP) and cutaneous infections. Other sites of infections were reported in 44 cases (18.8%). Among them, 23 cases (9.8%) were other systemic infections (five osteoarticular infections, five cases of endocarditis and five catheter-related infections) and 21 cases were other miscellaneous infections (including four dental abscesses and three bronchial super-infections). These distributions did not differ according to the cause of cirrhosis (table 2). The majority (84.2%) were CA infections. A total of 109 bacteria were isolated in 98 culture-positive infections (41.9%). Bacterial isolation was similar in CA and non-CA infections. The rate of positive culture was 54.7% in UTI, 30.5% in pulmonary infections and 52.3% in digestive infections. Among the 98 infections with positive cultures, 89 (90.8%) were monomicrobial and 9 (9.2%) were polymicrobial. Among the 89 bacteria isolated in monomicrobial infections, 50 (56.2%) were Gram-negative bacilli (GNB) and 30 (33.7%) were Gram-positive cocci (GPC). More specifically, among GNB infections, 41 (46.1%) belonged to the Enterobacteriaceae family with 35 being Escherichia coli; among GPC infections, 21 belonged to the Staphylococcaceae family, with 19 being S. aureus. The other bacteria were mainly of the Streptococcaceae or Enterococcaceae family (n=9). Among UTI, GNB (46.9%, mostly characterised by Enterobacteriaceae family) were predominant, whereas other systemic infections were dominated by GPC (68.4%). Absence of isolated bacteria was more frequent in pulmonary and cutaneous infections (69.5% and 87.0%, respectively). Digestive BIs were characterised by the absence of positive culture in 47.7% of cases. When a bacterium was retrieved in digestive BI, GNB and GPC were the most prevalent (20.5% and 13.6%, respectively). We then calculated the resistance rates among all bacteria retrieved in monomicrobial and polymicrobial infections for which in vitro antibiotic sensitivity was available. We observed seven GNB (17.5%) resistant to third generation cephalosporins and seven methicillin-resistant S. aureus (MRSA) (30.4%). Finally, among Enterobacteriaceae, P. aeruginosa, S. aureus, E. faecalis and E. faecium species for which the information was available (n=66), 21 (31.8%) were classified as MDRO (mainly extended spectrum β-lactamase producer Enterobacteriaceae members, MRSA or vancomycin-resistant Enterococcus). Within the same species for which the information was available (n=72), 20 (27.8%) were resistant to quinolones.
Characteristics of BI occurring during follow-up according to cause of cirrhosis
Incidence of bacterial infection (BI) during follow-up. (A) Patients with HCV-related cirrhosis had a higher incidence of BI than patients infected with HBV (global p value). (B) Cumulative incidence function (CIF) analyses did not display major competing risk effects on the incidence of a first BI in patients infected with HCV or patients infected with HBV with cirrhosis.
Relationship with episodes of liver decompensation and survival
During the same period, 170 patients presented at least one episode of liver decompensation, defined by the occurrence of either ascites, hepatic encephalopathy or gastrointestinal bleeding (related to portal hypertension in 29 out of 46) (table 2), with a corresponding 5 year CumI of 13.7%. Most of the 171 first BIs (n=140, 81.8%) occurred in compensated patients who had not previously experienced any episode of decompensation during follow-up; therefore, all analyses focused on subsequent episodes of decompensation or death were restricted to these patients. The probability of liver decompensation was higher in patients infected with HCV than in patients with HBV (5 year CumI: 16.2% vs4.8%, p<0.001, figure 2A) as well as the occurrence of a first BI before decompensation (see online supplementary figure S1A). CIF analyses did not reveal any major competing risk effect on the incidence of BI (figure 1B). Analyses focused on Model for End-stage Liver Disease (MELD) are provided in the online supplementary material section.
Impact of a first bacterial infection (BI) on subsequent hepatic decompensation and overall survival. (A) The incidence of hepatic decompensation was higher in patients infected with HCV than in patients infected with HBV with cirrhosis. (B) The occurrence of a first BI in compensated patients infected with HCV with cirrhosis was associated with higher risk of subsequent liver decompensation. (C) The occurrence of a first bacterial infection in compensated patients infected with HCV with cirrhosis impaired survival. (D) The occurrence of a first BI in compensated patients infected with HBV with cirrhosis impaired survival.
In patients infected with HCV, the occurrence of a first episode of BI increased the probability of hepatic decompensation (5 year CumI: 45.2% vs 14.7%, p<0.001, figure 2B). In the 17 patients infected with HBV who experienced a first episode of BI, an incidental peak of 5.9% for liver decompensation was observed after 2 months, before regaining the 5 year CumI of 5.1% of patients with HBV without BI (p=0.79, data not shown).
During follow-up, patients who experienced a first episode of BI did not have a higher risk of HCC (p=0.65), cardiovascular event (p=0.81) or extrahepatic cancer (p=0.062).
Overall, 136 patients died during follow-up (table 2), corresponding to 5 year survival of 89.6% for the entire cohort. Patients infected with HCV had lower probability of survival than patients infected with HBV (5 year CumI: 87.7% vs 97.4%, p<0.001). During follow-up, 30 patients were transplanted, 15 for terminal hepatic failure and 15 for HCC. BI was considered to be the direct cause of death in 20 patients (16.5%). When considering all infections except SBP, BI was the third cause of death in the cohort after terminal liver failure and primary liver cancer (PLC) progression.
The occurrence of a first BI impaired survival in patients infected with HCV (5 year probability of survival: 60.2% vs 90.4%, p<0.001, figure 2C) and patients infected with HBV (5 year probability of survival: 69.2% vs 97.6%, p<0.001, figure 2D). The negative impact of BI on long-term survival was still observed after exclusion of patients with septic shock (see online supplementary figure S1B).
Probability of BI resolution
We then assessed overall relationships between clinical and microbiological characteristics of these episodes and probability of resolution (defined by the absence of BI-related death), by carrying out a FAC in the 191 BIs with all available data (figure 3A). Projections of variables on the plane F1/F2, which accounted for 70.1% of the total variance, particularly distinguished UTI, non-systemic/miscellaneous BI, pulmonary and cutaneous BI, absence of isolated bacteria and success of specific BI therapy, which were projected on the positive values of factor F1. In contrast, digestive including SBP and other systemic BIs, highly severe BIs, concomitant IFN treatment, non-CA characteristics and BI-related death, were all projected on the negative values of F1. We secondarily performed a FAC on 40 monomicrobial BIs for which all microbiological data were available (figure 3B). Projections of the variables on the plane F1/F2, which accounted for 63.7% of the total variance, particularly distinguished as associated with a worse prognosis the following characteristics: pulmonary BI with microbiological data, other systemic data and SBP, Staphylococcaceae and Streptococcaceae species and other bacterium, multidrug resistance and quinolone resistance.
Factorial analyses of correspondence of bacterial infection (BI) variables. (A) Projections of variables of 191 BIs with all available data: (isolation of bacteria in a sample (blue), type of BI (red), severity of BI (green), success of therapy (orange), community-acquired or non-community-acquired (CA) (brown) and concomitant interferon (IFN) treatment (grey)) on F1/F2 plane computed from factorial analyses of correspondence. (B) Projections of variables of 40 BIs with isolated bacteria in monomicrobial infections with available microbiological data (main bacterial families (blue), resistance to quinolone (QR) (turquoise), multidrug-resistant (MDR) (purple) type of BI (red), severity of BI (green), success of therapy (orange), community-acquired or non-community-acquired (brown) and concomitant IFN treatment (grey)), on F1/F2 plane computed from factorial analyses of correspondence. SBP, spontaneous bacterial peritonitis.
We finally performed χ2 or Fisher's exact tests to identify the risk factors of BI severity. Risk factor for the severity of BI was concomitant IFN-based treatment. Risk factors for the absence of BI resolution were the presence of QRB (p=0.039), strong severity of infection and non-CA BI (all p<0.01).
Features associated with occurrence of BI
Sustained virological response (SVR) was achieved in 440 patients infected with HCV (33.3%) and was associated with decreased risk of BI occurrence (figure 4A, HR=0.25, p<0.001). All treatments consisted of an IFN-based regimen (with (n=92) or without a first-generation direct antiviral agents (DAAs)). Of note, 42 of the 208 BIs occurring in patients with HCV (20.2%) were registered during the course of an IFN-based regimen. Most patients with HBV were treated by oral drugs (entecavir or tenofovir), with only nine IFN-based regimen (2.9%) that did not impact the risk of BI (HR 95% CI 2.52 (0.33 to 19.36), p=0.37). At end point, 35 patients (12.6%) had a positive HBV viral load, a parameter that did not influence the risk of BI (HR 95% CI 0.69 (0.09 to 5.26), p=0.72).
Features associated with greater risk of bacterial infection. (A) Patients infected with HCV with compensated cirrhosis who did not achieve sustained virological response (SVR) during follow-up had a higher incidence of bacterial infection. (B) In the entire cohort, proton pump inhibitors (PPIs) were associated with higher risk of bacterial infection occurrence. (C) In the overall cohort, β-blocker intake was not associated with higher risk of bacterial infection.
We finally focused on drugs previously reported as being potentially associated with higher risk of BI in patients with cirrhosis. When considering the entire cohort, 262/1589 (16.5%) (MD: 83) patients received a proton pump inhibitor (PPI), either at inclusion or during follow-up, prior to the occurrence of a first BI. Among them, 46 (17.6%) developed a first BI versus 109 (8.2%) in patients without any PPI prescription, corresponding to a HR of 1.71 (p=0.003, figure 4B). The same observation was made when considering only patients infected with HCV (HR=1.59, p=0.014). In contrast, there was no difference in BI risk according to non-selective β-blocker intake (HR=1.02, p=0.92, figure 4C) in the entire cohort. Both drugs were more frequently prescribed in patients with oesophageal varices at inclusion (β-blockers: 84 (92.2%) vs 121 (34.2%) p<0.001 and PPI: 77 (20.9%) vs 149 (16.2%), p=0.04). Finally, antibiotics intake was reported in 40 patients (2.4%) of the CirVir cohort before and outside any episode of BI. These drugs were not associated with BI in the entire cohort or subgroups.
Table 4 summarises positive associations between all parameters under study and BI. The multivariate model selected lower albuminaemia and absence of SVR as independent predictors of BI in patients with HCV. The benefit of viral clearance was not impacted by the deleterious effect of IFN (see online supplementary material and supplementary figure S2). In patients infected with HBV, higher BMI was the main feature associated with greater probability of developing a first BI during follow-up (table 3). PPI persisted as an independent predictor of BI when taking into account the entire population. Low serum albumin levels and lack of SVR remained associated with BI after excluding SBP (see online supplementary table S2).
Outcome and other events occurring during follow-up according to cause of cirrhosis
Discussion
Because of the high morbidity and mortality of BI in patients with cirrhosis, better knowledge of their characteristics and burden in compensated cirrhosis is of paramount importance. Present data were obtained in a large multicentre cohort of compensated patients with biopsy-proven cirrhosis, in whom rigorous protocol-driven systematic data collection and analysis of predefined outcomes were undertaken in a competing risk framework.18 This approach, by covering the entire spectrum of BI occurring during the course of cirrhosis, was able to provide original and pivotal knowledge that might impact long-term management of patients with cirrhosis without overt liver failure in clinical practice.
The 5 year incidence of BI reached 13.1% in this population and was higher in patients infected with HCV, even after excluding infections occurring during IFN-based treatments. These incidences were similar to those of critical liver-related complications such as decompensation (13.7%) or PLC (13.6%), without evidence of competing effects (figure 1B). Most of these BI occurred as a first event prior to occurrence of a decompensation episode or PLC. These findings suggest that BI is a frequent complication in this population, independently of other well-known liver-related events, as observed in other chronic diseases such as diabetes, renal failure or malignancies.19–21 Unfortunately, large prospective cohorts of patients with compensated non alcoholic steato-hepatitis (NASH)- or alcohol-related liver disease are lacking; in this context the incidences and characteristics of BI observed in the CirVir cohort should not be extrapolated to populations with non-viral Child–Pugh A cirrhosis. Of note, although nearly one-third of the CirVir population had a past history of excessive alcohol consumption (table 1), this feature did not influence the risk of BI (data not shown).
Microbiological data are in accordance with those reported by Fernandez et al4 in a European monocentric cohort of mainly decompensated patients (table 2). Such replication underlines the quality of the CirVir cohort based on a prospective design ensuring an unbiased record of events during follow-up.5 FAC (figure 3) combining clinical and microbiological data enabled better classification of these BIs by identifying those most likely to be associated with high probability of resolution (mainly UTI, monomicrobial with enterobacteriaceae or Streptococcaceae, CA) compared with those associated with worse short-term prognosis (polymicrobial or monomicrobial with Staphylococcaceae, non-CA).
Although nearly 80% of cases of BI were diagnosed before any episode of hepatic decompensation, their occurrence dramatically impacted the long-term prognosis of this compensated population. Indeed, patients infected with HCV who experienced BI had a nearly fivefold-higher incidence of subsequent liver decompensation (figure 2B), leading to 5 year probability of survival as low as 60.2% (figure 2C). The impact on survival was even more striking in patients infected with HBV, in whom overall rates of liver-related complications were observed during follow-up, probably due to the fact that most of them were virologically controlled by oral agents (figure 2D). When considering causes of the 136 deaths registered thus far in the CirVir cohort, direct imputability of BI (excluding SBP) was attributed in 20 (16.5%) (table 2), thus ranking this event as the third cause of death in this compensated population, after terminal liver failure and PLC progression. The complex interplay leading to a vicious circle between BI and liver decompensation is well known. Hospitalised patients with cirrhosis have a high risk of developing infection, In turn, BI may be a triggering factor in occurrence of gastrointestinal bleeding, ascites, hepatic encephalopathy, kidney failure and development of ACLF.22 Previous studies focused on patients with a predictable poor outcome, namely those with impaired liver function, variceal bleeding, low ascitic fluid protein levels, prior SBP and repeated hospitalisation for hepatic complications. In this context, the impact of BI on mortality was evaluated only over the short term, and concluded that there was an increase of nearly fourfold for death probability in patients with decompensated cirrhosis.23 The present report extends this data to patients with compensated cirrhosis in whom the occurrence of a first BI, before clear-cut worsening of liver function, should be considered a turning point of their liver disease.
In light of these findings, decreasing the risk of BI in compensated cirrhosis would constitute a major step towards improvement of patient prognosis. The CirVir cohort included patients with HBV-related and HCV-related cirrhosis so as to clearly identify differences in outcome and specific risk factors for complications. Because of differences in baseline characteristics and rates of virological control/clearance (together with possible underpowered analyses performed in the small-sample size population of patients with HBV), rates and predictive factors of BI differed according to the cause of cirrhosis. In particular, the higher rates of BI observed in patients infected with HCV compared with patients infected with HBV suggest that better control of comorbidity and improvement in virological clearance would lead to a decrease in BI. Better control of comorbidity is obviously essential in management of patients with cirrhosis, in whom features of metabolic syndrome or addiction are known to strongly favour the occurrence of all liver-related complications and thus impact prognosis.24 However, the link between viral eradication and BI was not clear up until now. In the first description of the CirVir cohort,5 baseline viral load was associated with an increased incidence of all complications, including BI. The present report, with the advantage of a longer follow-up and by studying virological clearance at end point, now clearly shows that achieving SVR in patients infected with HCV leads to a decrease in BI occurrence (figure 4A), a feature that persisted in multivariate analysis (table 4). This improvement led to an annual incidence tending to reach that of patients infected with HBV in whom virological control at end point was obtained for most patients via oral antiviral agents (figure 1A). When considering the prognostic impact of occurrence of a first BI, it is tempting to speculate that the clinical benefit of viral eradication over the long term might be exerted through slower worsening of liver function25 and by disruption of a vicious circle triggered by end organ dysfunction-related BI, including for non-SBP infection. This constitutes another strong argument favouring access to second-generation anti-HCV DAAs in compensated patients with cirrhosis,26 as well as present data confirming higher risk of severe BI during the course of an IFN-based regimen.27 ,28 Finally, indications for long-term therapies commonly adopted in patients with cirrhosis such as β-blockers or PPI should be discussed (figure 4B, C). Several studies, usually retrospective, have reported a statistical association between these drugs and BI in patients with cirrhosis,29 ,30 hypothetically through modification of gut microbiota and intestinal barrier dysfunction favouring bacterial translocation.31 While the deleterious role of β-blockers remains subject to debate,32 converging evidence has established a possible epidemiological link between PPI intake and BI in decompensated patients with cirrhosis.30 However, probably because of their apparent safety, PPI have become one of the most frequently prescribed long-term drugs in a variety of non-approved indications, including cirrhosis.33 Analysis of the CirVir population is another strong argument for clarification of PPI indications in patients with cirrhosis, as these drugs remained an independent feature associated with BI regardless of potential signs of portal hypertension (table 3).
Features associated with occurrence of BI in patients with compensated HCV-related or HBV-related cirrhosis according to Cox proportional hazards model
In summary, BI occurs as frequently as liver-related complications in patients with viral compensated cirrhosis, and often as a first critical event having strong prognostic significance. Patients experiencing a first BI, even in the absence of prior hepatic decompensation, have a higher probability of long-term hepatic decompensation and death, placing this event as a turning point that may define a specific subgroup of compensated patients in whom adaptation of long-term therapeutic management (including transplantation) could be discussed. Control of comorbidities and achievement of virological control, as well as avoidance of potentially deleterious drugs such as PPI, may decrease their incidence and thus improve prognosis.
Acknowledgments
Scientific committee: J-C Trinchet and P Nahon (principal investigators), R Layese and F Roudot-Thoraval (data management), P Bedossa, M Bonjour, V Bourcier, I Durand-Zaleski, H Fontaine, A Laurent, P Marche, D Salmon, V Thibault, V Vilgrain, F Zoulim, J Zucman-Rossi, S Allam (ANRS), V Petrov-Sanchez (ANRS). Clinical Centres (ward/participating physicians): CHU Jean Verdier, Bondy (JC Trinchet, V Bourcier, P Nahon); CHU Cochin, Paris (S Pol, H Fontaine); CHU Pitié-Salpétrière, Paris (Y Benhamou); CHU Saint-Antoine, Paris (L Serfaty); CHU Avicenne, Bobigny (D Roulot); CHU Beaujon, Clichy (F Degos); CHU Henri Mondor (A Mallat); CHU Kremlin Bicêtre (C Buffet); CHU Tenon, Paris (JD Grangé); CHRU Hôpital Nord, Amiens (D Capron); CHU Angers (P Calès); Hôpital Saint-Joseph, Marseille (M Bourlière); CHU Brabois, Nancy (JP Bronowicki); Hôpital Archet, Nice (A Tran); Institut Mutualiste Montsouris, Paris (F Mal, C Christidis); CHU Poitiers (C Silvain); CHU Pontchaillou, Rennes (D Guyader); CH Pays d'Aix, Aix-en-Provence (C Wartelle); CHU Jean Minjoz, Besancon (V Di Martino); CHU Bordeaux—Hôpital Haut-Leveque, Pessac (V de Ledinghen); CHU Bordeaux—Hôpital Saint-André, Bordeaux (JF Blanc); CHU Hôtel Dieu, Lyon (C Trepo, F Zoulim); CHU Clermont-Ferrand (A Abergel); Hôpital Foch, Suresnes (S Hillaire); CHU Caen (T Dao); CHU Lille (P Mathurin); CH Le Mans (C Pilette); CHU Michallon, Grenoble (JP Zarski); CHU St Eloi, Montpellier (D Larrey); CHU Reims (B Bernard-Chabert); CHU Rouen (O Goria, G Riachi); Institut Arnaud Tzanck, St Laurent-du-Var (D Ouzan); CHU Purpan, Toulouse (JM Péron); service medicine interne CHU Purpan, Toulouse (L Alric) and CHU Tours (Y Bacq).
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
- Data supplement 1 - Online supplement
Footnotes
Collaborators Microcir group: Marie Kempf (CHU Angers, laboratoire de Bactériologie, Angers), Fréderic Bert (APHP Hôpital Beaujon, Service de Microbiologie, Clichy), Alexandra Doloy (APHP, Hôpital Cochin, Laboratoire de Bactériologie, Paris), Isabelle Poilane (AP-HP, Hôpital Jean Verdier, Service de Microbiologie, Bondy), Olivia Peuchant (CHU Haut Lévêque, Laboratoire de Bactériologie, Bordeaux), Etienne Carbonnelle (AP-HP Hôpital Avicenne, Laboratoire de Bactériologie-Virologie Hygiène, Bobigny), Bertrand Picard (AP-HP Hôpital Avicenne, Laboratoire de Bactériologie-Virologie Hygiène, Bobigny), Christophe Burucoa (CHU Poitiers, Laboratoire de Bactériologie, Poitiers), Vincent Cattoir (CHU Caen, Service de Microbiologie, Caen), Dominique Decré (AP-HP Hôpital Saint Antoine, Paris), Nicolas Degand (CHU Nice, Laboratoire de Bactériologie, Nice), Laurent Dortet (AP-HP Hôpital Bicêtre, Laboratoire de Bactériologie Le Kremlin Bicêtre), Samer Kayal (CHU Pontchaillou, Service de Bactériologie et Hygiène Hospitalière, Rennes), Véronique Vernet-Garnier (CHU Robert Debré, Laboratoire de Bactériologie, Reims), Alain Lozniewski (CHU Nancy, Laboratoire de Bactériologie, Nancy), Edouard Tuaillon (CHU-Hôpital Saint Eloi, Laboratoire de Bactériologie, Montpellier), Sophie Vimont (APHP, Hôpital Tenon, Laboratoire de Bactériologie, Paris), Emilie Bessède (CHU Pellegrin, Laboratoire de Bactériologie, Bordeaux), Isabelle Patry (CHU Jean Minjoz, Laboratoire de Bactériologie, Besançon), Nadine Lemaitre (CHRU Lille, Laboratoire de Bactériologie, Lille) and Christine Pachetii (CH Aix, Laboratoire de Diagnostic des Maladies Infectieuses, Aix en Provence).
Contributors PN, J-CT, ML, VB and FR-T had full access to all data in the study and take responsibility for the integrity of data and the accuracy of data analysis. Study concept and design: PN, J-CT, VB and FR-T. Acquisition of data: PN, VB, ML, RL, NT, PM, DG, SP, D, VDL, DO, FZ, DR, AT, J-PB, J-PZ, OG, PC, J-MP, LA, MB, PM, J-FB, AA, LS, AM, J-DG, Buffet, YB, CW, TD, YB, CP, CS, CC, DC, BB-C, SH, VDM, J-CT and FR-T. Analysis and interpretation of data: PN, VB, RL, ML, J-CT, RM and FR-T. Drafting of the manuscript: PN, VB, RL, ML and FR-T. Critical revision of the manuscript for important intellectual content: PN, VB, RL, ML, PM, DG, SP, DL, VDL, DO, FZ, DR, AT, J-PB, J-PZ, OG, PC, J-MP, LA, MB, PM, J-FB, AA, LS, AM, J-DG, Buffet, YB, CW, TD, YB, CP, CS, CC, DC, BB-C, SH, VDM, RM and FR-T. Statistical analysis: RL and FR-T. Funding obtained: J-CT. Administrative, technical and material support: PN, VB, RL, NT, J-CT and FR-T. Study supervision: PN, VB, RL, ML, J-CT and FR-T.
Funding The promoter of the study was the ANRS (France REcherche Nord and sud Sida-HIV Hépatites-FRENSH), with the cohort funded by the same institution.
Competing interests None declared.
Patient consent Obtained.
Provenance and peer review Not commissioned; externally peer reviewed.
Ethics approval The study protocol was in conformity with ethical guidelines of the 1975 Declaration of Helsinki and was approved by the Institutional review board (Comité de Protection des Personnes, Aulnay-sous-Bois, France).