Secondary peritonitis is a severe compartmentalized infectious insult characterized by a rapid response of innate immunity leading to a major inflammatory process. The initial response is usually followed by a post-injury depression of innate immunity in various types of sepsis 1234. However, there is a paucity of data regarding systemic and local innate immune responses during peritonitis in humans and on their relation to prognosis 5678. The outcome of secondary peritonitis has been shown to be influenced by several clinical and bacteriological features of the disease. We recently reported that two or more micro-organisms in peritoneal fluid culture, intra-peritoneal anaerobes, yeasts, or Enterococcus species were associated with worse prognosis. Other groups have suggested that postoperative peritonitis was also associated with worse outcome 91011, a finding that our own observations did not corroborate 12. Whether or not these features are associated with differential immune responses is unknown. Therefore, the aims of this study were: 1) to characterize the inflammatory mediator response (pro-inflammatory: IL-1, TNFα, IL-6, IFNγ ; and anti-inflammatory: IL-10) in the peritoneal and blood compartments of patients with secondary peritonitis; 2) to evaluate the potential differences in mediator profiles between shocked and non-shocked patients, survivors and non survivors and according to the type of peritonitis (community-acquired or post-operative); and 3) to look at the potential impact of peritoneal microbiological findings on mediator release.

Over a period of two years, all patients with secondary peritonitis admitted in our unit (surgical intensive care unit, Lariboisière University Hospital) were prospectively screened for peritoneal and plasma cytokine measurements. The study was approved by our Institutional Review Board (N° IRB00006477, Comité d'Ethique de la Recherche Biomédicale du GHU Nord).

Plasma cytokine measurements were performed on leftovers of blood from routine daily samples and patients were informed during follow-up, but the need for written informed consent was waived. This cohort was a subgroup of a larger cohort of secondary peritonitis, which was previously reported 12. Patients were included if they were older than 18 years and if the diagnosis of secondary generalized peritonitis (community-acquired and postoperative peritoneal infection) was confirmed surgically. They were not included if they had secondary peritonitis as a result of penetrating trauma, tertiary peritonitis defined as recurrent postoperative peritonitis, primary peritonitis (medical cause of intra-abdominal infection that did not require surgery) or if they received steroids as part of their treatment.

Sixty-six patients with secondary peritonitis were studied prospectively. Patient characteristics are presented in Table 1. Twenty three patients (34%) had septic shock and eight of them died while in ICU (ICU mortality: 34%). Laparotomy was always performed within 24 hrs of septic shock onset. The incidence of shock did not reach a difference between community-acquired and post-operative peritonitis (24% versus 48%, respectively, P = 0.08). The origin of peritonitis was colon (n = 20), gastro-duodenal (n = 13), post-duodenal small bowel (n = 18), biliary tract (n = 3), appendix (n = 10) and others (n = 2). The proportion of patients with polymicrobial cultures (≥2 germs) was 51% (n = 34 patients): anaerobes, 18% (n = 12), Enterococcus species, 20% (n = 13), and yeasts, 14% (n = 9).

Plasma and peritoneal fluid IL-1, TNFα, IL-6, IFNγ and IL-10 at day one (day of surgical procedure) and the ratio of peritoneal to plasma concentrations are presented in Table 2. Concentrations of cytokines were always higher in peritoneum than in plasma, with great variability of the peritoneal fluid/plasma ratio among cytokines (from 5 for IFNγ to more than 1,300 for IL-1). There was no correlation between plasma and peritoneal fluid concentration of any cytokine.

In the plasma compartment, the IL-6 profile over the first three days differed between shock and no-shock patients (P < 0.03), whereas profiles of TNFα, IFNγ and IL-10 did not differ significantly (that is, no significant interaction) but inter-group and overtime differences were noted between the two groups (Figure 1). Comparing survivors versus non-survivors, TNFα plasma profiles had a significant interaction (P < 0.008) while IL-6, IFNγ and IL-10 exhibited only intergroup and overtime differences (Figure 2). Regardless of the clinical situation, plasma IL-1 did not differ among groups: shock versus no shock (5 (5 to 8) versus 5 (5 to 8) pg/ml, P = 0.54) and survival versus non survival (5 (5 to 8) versus 5 (5 to 6) pg/ml, P = 0.73). There was no difference in plasma cytokine concentrations between community-acquired and post-operative peritonitis. Plasma cytokine levels according to bacteriological findings (mono- versus poly-microbial cultures of peritoneal fluid, presence or not of anaerobes, Enterococcus, or yeast) are presented in Table 3. Plasma TNFα was significantly higher in the case of polymicrobial culture or when anaerobes or Enterococcus were present in the peritoneal fluid. Plasma IL-10 was significantly higher in the case of polymicrobial culture or when anaerobes were present in the peritoneal fluid.

We did not find any difference in peritoneal cytokines with regard to the occurrence of shock. Only peritoneal levels of IL-10 were higher in survivors versus non-survivors (1,505 pg/ml (450 to 3,130) versus 102 pg/ml (9 to 710), respectively, P = 0.04). Table 4 shows peritoneal cytokine levels in relation to bacteriological findings. Anaerobes and Enterococcus species were associated with higher IFNγ in peritoneal fluid. The presence of yeast did not result in higher plasma or peritoneal cytokine levels, but only a few patients from our cohort (n = 9) had yeast in their peritoneal fluid.

In this cohort of 66 patients with secondary generalized peritonitis, we observed a large gradient between peritoneal fluid and plasma concentrations of cytokines, with no correlation between peritoneal and plasma levels. Plasma TNFα, IL-6, IFNγ and IL-10 were higher in patients with shock and in non survivors and decreased over the first three postoperative days in a parallel manner. Plasma cytokines did not differ between community-acquired and postoperative peritonitis. The presence of peritoneal anaerobes or Enterococcus species was associated with a higher plasma level of TNFα. IL-10 plasma levels were also higher when anaerobes or polymicrobial positive culture were found. In the peritoneal compartment, there was no difference in cytokine concentration between shock and no shock patients, but there was a greater IL-10 concentration in survivors. The presence of anaerobes and Enterococcus species was associated with higher peritoneal IFNγ.

There are very limited data on innate immune response assessed by plasma and peritoneal inflammatory mediators in patients with secondary peritonitis. To our knowledge, only one study addressed this issue and reported a large concentration gradient for TNFα and IL-6 between peritoneal fluid and plasma in a small group (n = 17) of patients with peritonitis 6. Our results confirm that IL-1, TNFα, IL-6, IL-10 and IFNγ are present at high concentrations in the peritoneal fluid of patients with peritonitis. This observation fits well with the concept of the 'tip of the iceberg' which suggests that plasma levels increase only after saturation of tissues within the abdominal compartment 1516. However, a direct intravascular stimulation of cytokine release by monocytes cannot be ruled out in some patients, especially when bacteremia occurred, since this condition is associated with higher plasma cytokines 17. In this cohort, the proportion of positive blood cultures within the first 48 hours of peritonitis was 11/66 (16%).

We also observed that plasma levels of pro- and anti-inflammatory mediators were higher in patients with shock and in non-survivors, a finding that has been reported by several authors 16781819. The fact that pro- and anti-inflammatory mediators are released simultaneously during the early phase of peritoneal sepsis confirms what was recently noted in animal models of sepsis 20 and septic patients 42122. The only data available regarding IFNγ and outcome during abdominal infection come from experimental studies and the results are conflicting. According to some authors, low plasma IFNγ is associated with increased mortality 2324 while others reported that prophylactic inhibition of IFNγ improves survival 2526. Our values for plasma and peritoneal IFNγ measurements represent the first set of data in humans with GP. The plasma concentration of IFNγ was very low compared to other cytokines, in accordance with what has been observed in experimental models of peritonitis 4.

Similarly, no human data were available on IL-10 during peritonitis. Experimental studies suggest that intra-peritoneal or subcutaneous IL-10 injections reduce mortality 27, which fits well with our observation that the intra-peritoneal IL-10 concentration was greater in survivors. Since IL-10 is considered an anti-inflammatory mediator, our findings corroborate the fact that both pro- and anti-inflammatory mediators are produced simultaneously in the peritoneum of patients with peritonitis 28.

In a previous study, we found that polymicrobial cultures or anaerobes in the peritoneal fluid are associated with more frequent septic shock 12. The role of the infecting pathogen on the innate immune response of patients with peritonitis has been poorly investigated. Controversy exists regarding a specific role of some pathogens on the pattern of the immune response. Some authors support the concept of a 'generic septic response' in which an identical immune response is triggered by any type of bacteria 2930. This theory has been contradicted by others who suggested that different types of germs may elicit various immune responses, despite a common pathway of activation. For example, distinct host responses to Streptococcus pneumoniae or Pseudomonas aeruginosa have been reported in experimental models of pneumonia 31. In humans with sepsis of various origins, Gogos et al. recently observed that HLA-DR expression and the apoptosis rate of CD-4 and CD-8 within 24 hours of sepsis onset varied according to the underlying pathogen 32. In rats with peritonitis, Montravers et al. showed that adjunction of Enterococcus faecalis is associated with increased mortality as well as higher levels of TNFα and IL-6 in peritoneal fluid 3334. In the present study, we observed that patients in whom anaerobes or Enterococcus species were isolated from peritoneal fluid cultures released more TNFα in their plasma than those who were infected with other strains. One of the limitations of our study is that the number of patients with specific characteristics or bacteriological findings in each subgroup is rather small, preventing any definitive conclusion to be drawn from these results. In addition, small differences must be considered cautiously as multiple comparisons were performed to compare various subgroups. However, some of the observations we made are consistent with previous experimental findings and, therefore, provide the first human data to support laboratory hypotheses.

Secondary peritonitis is responsible for an acute local and systemic innate immune response involving a simultaneous release of both pro- and anti-inflammatory mediators. The most severely affected patients had higher plasma levels of all cytokines but this difference was not observed with peritoneal fluid cytokines. A large gradient in cytokine concentration was noted between the peritoneal cavity and plasma, highlighting the importance of compartmentalization of innate immune response during peritonitis.

• Compartmentalization of innate immune response during peritonitis is reflected by a large gradient in cytokine concentration between peritoneum and plasma.

• Higher plasma levels of pro and anti-inflammatory cytokines are associated with shock and mortality.

• Innate immune response was not different between patients with post-operative and community-acquired peritonitis.

• The infecting pathogen might influence host immune response.

CD: cluster of differentiation; ELISA: enzyme-linked immunosorbent assay; HLA-DR: human leucocyte antigen-DR; IL: interleukin; INFγ: interferon gamma; SAPS: Simplified Acute Physiologic Score; TNFα: tumor necrosis factor alpha.

The authors declare that they have no competing interests, except DP who acknowledges receiving lecturing fees from Eli-Lilly, Torey Medical, Vitech Italy and Meditor in the past five years.

FR conceived the study and its design, and drafted the manuscript. BC analyzed data and drafted the manuscript. EG performed the statistical analyses. CC and JML carried out the cytokine dosage. PV participated in study design and coordination. MJL participated in enrollment and correct routing of blood samples. JM and DP helped to interpret data and participated in manuscript drafting. All authors read and approved the final manuscript.