The overall incidence of severe sepsis and septic shock is steadily increasing due to the aging of the population and to the growing prevalence of underlying co-morbidities including chronic organ dysfunctions and immunosuppression 12. Several studies have highlighted the major influence of cirrhosis on the susceptibility to severe bacterial infections 34. Indeed, the overall mortality rate of septic shock remains particularly high in cirrhotic patients, ranging from 60% to 100% 567, raising the question of indications of aggressive and extensive organ failure support in such patients.

The overall outcome of septic shock has clearly improved over the recent years, related to improved supportive care and rapid and protocolized treatment interventions supported by international guidelines 8. Of note, the most significant improvements in survival from septic shock have been achieved in vulnerable subgroups including elderly patients, those with malignancies 910 or neutropenia 11. Cirrhotic patients are usually excluded from interventional trials in sepsis, but it is likely that such a high-mortality subgroup would particularly benefit from therapeutic advances in septic shock. In addition, great strides have also been recently achieved in the management of specific complications of cirrhosis. Whether the survival gain achieved by therapeutic advances in septic shock also extends to cirrhotic patients has not been assessed. In order to determine the trend in mortality of cirrhotic patients with septic shock and the impact of related therapeutic interventions, we performed a retrospective single-center study over a 14-year time period.

Septic shock represents a severe complication of cirrhosis with very low survival rates that question the relevance of life-sustaining therapies in this subgroup of patients. We report here that the current survival rate remains low but substantially improved over time, suggesting that advances in care of septic shock extended to this high-risk subgroup of patients. Most importantly, the short-term improvement in survival was sustained for at least 6 months, suggesting that ICU admission and extensive life support is justified in some patients. In addition to organ failures, the stage of liver disease as assessed by the Child-Pugh score appears to be an independent prognostic factor that should be taken into account in the decision-making process.

Cirrhosis is clearly associated with an increased predisposition to sepsis 4 and has been identified as an independent poor prognostic factor in patients with severe sepsis 24. In addition, chronic alcohol abuse by itself may contribute to worsening organ failures and has been shown to be an independent risk factor of septic shock 25. Multiple mechanisms concur to confer an increased susceptibility to bacterial infections and subsequently to multiple organ failure in cirrhotic patients. Bacterial translocation increases with the severity of liver disease and represents the main mechanism of spontaneous bacterial peritonitis. Furthermore, it may sustain the septic process in every type of infection as suggested by the increased levels of endotoxin observed in cirrhotic patients 26. In addition, innate immune cells display functional abnormalities such as increased production of pro-inflammatory cytokines in response to LPS 27 or alcohol-induced defective phagocytosis 28. As a consequence, plasma TNF-α and IL-6 levels are higher in cirrhotic patients with bacterial infection than in non-cirrhotic patients 29.

The overall outcome of severe sepsis and septic shock has been improved over the last decade concomitant to the emergence of adjuvant therapeutic interventions such as early-goal directed therapy 30, protective mechanical ventilation with low tidal volumes 13, low-dose corticosteroids 15, intensive insulin therapy 14, or activated protein C 16 supported by positive randomized controlled trials. Although the true benefit of some of these interventions have been addressed by additional studies, they have been included in the Surviving Sepsis Campaign guidelines 8. This international guideline-based performance improvement program promoted early recognition and management of sepsis and showed that increased compliance with the guidelines was associated with an improved survival rate 31. In the same way, a continuous decrease in the mortality of severe sepsis and septic shock patients has been reported in large unselected cohorts that mostly comprised non-cirrhotic patients 13233. Of note, some significant progress in septic shock has also been achieved in highly vulnerable subgroups of patients such as immunocompromised patients with malignancies 9.

Differences in survival between the two periods occurred early in the course of the disorder and were thereafter maintained. The rate of inadequate antibiotic treatment was low and similar between both periods. Most patients received an initial combination antimicrobial treatment of β-lactam associated with either aminoglycosides or fluoroquinolones. An initial single dose of aminoglycosides was commonly subsequently replaced by fluoroquinolones to limit harmful side-effects. Indeed, some retrospective studies suggest a benefit of combination antibiotherapy, most especially of betalactams and aminoglycosides, for septic shock in general cohorts as well as in cirrhotic patients 734. Survival improvement was more likely related to changes in the early management of shock and organ failures. With respect to fluid resuscitation strategies, the majority of patients from the recent period received albumin and consequently received less crystalloids. Albumin resuscitation was associated with higher Child-Pugh and MELD scores (P = 0.04 and P = 0.03 respectively, data not shown), suggesting that albumin was preferentially indicated to patients with advanced stages of cirrhosis. The role of albumin resuscitation in sepsis remains challenging. A meta-analysis suggested that fluid resuscitation with albumin compared to crystalloids was associated with improved survival in patients with severe sepsis 35. Specifically, albumin resuscitation has been shown to reduce mortality and renal impairment in cirrhotic patients treated for spontaneous bacterial peritonitis 36. Accordingly, frequent albumin resuscitation in the recent period of our study was also associated with less frequent recourse to renal replacement therapy. Altogether, these results and ours suggest a possible benefit of albumin resuscitation in cirrhotic patients with severe sepsis, that remains to be prospectively investigated. The frequent use of low-dose corticosteroids was also a hallmark of sepsis management during the second period. Indeed, adrenal dysfunction frequently occurs in patients with septic shock and is associated with hemodynamic instability, renal dysfunction, and increased mortality 37. However, the use of low-dose corticosteroids in septic shock remains controversial because of discrepant efficacy data and a possible higher risk of nosocomial infections 15. Of note, this treatment has been specifically addressed in cirrhotic patients with septic shock. In a case-control study, resolution of shock and survival were higher in hydrocortisone-treated cirrhotic patients 38. However a randomized controlled trial in septic shock cirrhotic patients was stopped for futility at interim analysis because hydrocortisone did not reduce mortality and was associated with an increase in adverse effects such as shock relapse and gastrointestinal bleeding 39. In addition to sepsis-specific therapies, general measures for ICU patients such as ventilation with low tidal volumes and glucose control with intravenous insulin therapy were also routinely implemented during the second period.

Cirrhotic patients are commonly perceived as poor candidates for ICU admission because of the very high mortality rates associated with organ failures 4041. A general improvement in outcome for cirrhotic patients in the ICU has been recently reported regardless of the type of acute complication 4243. Several factors may explain this progress, including a better selection of patients on the basis of previous functional status and advances in the management of acute complications such as variceal bleeding 44, hepatorenal syndrome 45, or septic shock as highlighted in the present study. The extent of organ failures clearly represents a major determinant of outcome, and the performance of SOFA score that nearly reached significance in our multivariate model is in accordance with previous studies 646. Most importantly, we also identified the stage of the underlying liver disease assessed by the Child-Pugh score as an independent prognostic factor of septic shock. Until now, the discrimination of the Child-Pugh score calculated at the time of ICU admission had remained inferior to organ failure scores 64647. As a matter of fact, we computed Child-Pugh score prior to the acute complication in order to reliably assess the stage of liver disease without any interference from new-onset organ failures. This finding carries major practical implications for the decision-making process. Indeed, this could allow a better selection of patients likely to benefit from intensive care, on the basis of underlying disease's status and realistic therapeutic options including liver transplantation. Nevertheless, an accurate individual prognosis prediction of cirrhotic patients is often difficult at the time of ICU admission, but can be markedly refined after a few days 646. In the light of improved outcomes and limited performance of initial prognostic prediction, critically-ill cirrhotic patients with a reasonable long-term prognosis should be offered a broad intensive care access policy with subsequent reappraisal based on the nature of the acute complication and the evolution of organ failures.

Our study has several limitations that we acknowledge. First, the design was retrospective despite most data were prospectively collected through computerized patient data management system. Therefore, we can only report an association between changes in care and patient outcome over the study period. Second, it was carried out in a single center, with a hepatology unit that is closely involved in the decision-making process and in the management of critically ill cirrhotic patients while in the ICU and after discharge. Third, the limited number of patients may limit the external validity of our findings. However, our survival rates in the latter period are similar to those reported by Arabi et al. (hospital survival 24%) 7 and Levesque et al. (ICU survival 36%) 6. Fourth, indications for ICU admission or end-of-life decisions might have evolved over the study period, and we cannot exclude that patients from the recent period were more carefully selected or referred earlier to the ICU. Nevertheless, the functional status, the stage of the underlying liver disease and the severity scores were similar between both periods. In addition, arterial blood lactate level as an indicator of prolonged systemic hypoperfusion also suggested similar duration and severity of shock before ICU admission. Altogether, these results suggest that improvements in survival were more likely related to changes in care. Fourth, we failed to link the recent improvement in survival with a single therapeutic change, suggesting that it is more likely related to a combination of interventions. Alternatively, some unrecognized or non-collected data might also influence the outcome. For instance, the functional status assessed by the performance status prior to the acute complication is a major prognostic factor in critically-ill cancer patients 48, and might be more accurate than the Knaus scale in this setting. In the same way, the nutritional status might also be of importance in these patients 49.

This study reports an encouraging improvement in survival in cirrhotic patients with septic shock that needs to be confirmed in a larger multicenter cohort. Implementation of therapeutic advances in sepsis probably accounted for this result. In addition, the stage of the underlying liver disease appears as an important prognostic factor. Delineation of the long-term prognosis of cirrhosis and the related therapeutic options thus appears essential in order to determine the indications for life-sustaining therapies.

• The current survival rate of septic shock in cirrhotic patients remains low but has improved over the recent years.

• Cirrhotic patients could have benefited from recent advances in the management of septic shock.

• The stage of liver disease prior to the acute complication as assessed by the Child-Pugh score appears to be an independent prognostic factor of hospital mortality.

ARDS: Acute Respiratory Distress Syndrome; CI: Confidence Interval; COPD: Chronic Obstructive Pulmonary Disease; ICU: Intensive Care Unit; IL: Interleukin; INR: International Normalized Ratio; MELD: Model for End-stage Liver Disease; OR: Odds Ratio; RRT: Renal Replacement Therapy; SAPS: Simplified Acute Physiology Score; SOFA: Sequential Organ Failure Assessment; TNF: Tumor Necrosis Factor.

JC is consultant for LFB and received lecture fees. JPM is consultant and member of the scientific board of LFB, and received lecture fees from LFB, Fresenius and Baxter. JC and JPM were the main investigators of an interventional trial on albumin resuscitation in septic shock. FP received lecture fees from LFB. The authors declare that they have no other competing interests relevant to the field of the manuscript.

BS, BC, and FP designed the study. BS, BC, AS, and NM extracted the data. BS and FP performed the statistical analysis. BS, BC, AS, NM, JC, AC, JDC, VM, JPM, and FP contributed to data analysis. BS and FP drafted the manuscript. All authors read and approved the final version of the manuscript.