Dynamic changes of microbial flora and therapeutic consequences in persistent peritonitis

To the best of our knowledge, this study is the first to focus on susceptibility issues of the microorganisms isolated in patients with persistent peritonitis. Inadequate source control was identified as a leading cause of persistent peritonitis. Initial severity, the presence of Candida species and inadequate source control were the only risk factors for reoperation. A progressive shift in the cultured flora was observed with the number of reoperations, marked by extinction of susceptible streptococci and anaerobic strains, and the emergence of both Gram-negative and Gram-positive MDR strains and fungi. The interval between initial surgery and reoperation, and the presence of Candida appeared to play an important role as risk factors for emergence of microorganisms.

Our highly selected population and the retrospective study design could be considered to be a limitation, but a prospective multicentre study on this topic would appear to be unfeasible. Our findings on the patterns of empirical and definitive antibiotic usage, the characteristics of local microbial flora, or the hospital influence on the culture results have to be considered cautiously and cannot be directly generalised. The long study duration may have led to changes in antibiotic susceptibility patterns over time. However, such changes were not observed and environmental characteristics appeared to play only a minor role in our population. Our data did not distinguish between pathogenic strains and colonisation, which is particularly important for coagulase-negative staphylococci and yeasts. In clinical practice, prescribers usually target most, if not all, isolates. Another major limitation in the interpretation of our results is the lack of any pharmacokinetic monitoring in peritoneal fluid. We routinely monitored the plasma concentrations of antibiotics, but satisfactory plasma concentrations do not guarantee adequate local concentrations [23,24]. However, in the absence of data in the literature, we consider that our results provide an interesting overview of persistent abdominal sepsis in ICU patients, which could be helpful for physicians in the field.

Our failure to identify clear risk factors for persistent peritonitis is not surprising, as this study was not designed for this purpose. Several studies over the last decade have also failed to identify clear risk factors or surrogates for persistent peritonitis [4,25,26]. The time to reoperation varies considerably according to the literature, from 2 to 7 days [4,16,27], but some papers have reported delayed reoperations after up to 20 days [16]. In our cohort, times to reoperation were situated in the median range throughout the periods analysed. This variability suggests that time to reoperation is at least partially a physician-dependent variable, rather than a variable that reflects the course of the disease.

To our knowledge, the only study assessing the dynamics of peritoneal flora during persistent/recurrent abdominal sepsis used a different methodology [11]. Although our results are in agreement with those reported by these authors on the concept of changing peritoneal flora, we report different results in terms of Gram-negative strains and anaerobes. In addition, we provide a global analysis because of the high frequency of generalised peritonitis in most of our cases.

The present study presents the highest rate of MDR ever reported in the course of intra-abdominal infections [5,8,14,15,28-30]. The progressive extinction of highly susceptible strains such as streptococci and anaerobes [5,14,15,28,30] and the emergence of MDR strains [14,15,29] have been previously reported, but not to the extent observed here. Several risk factors promoting the emergence of MDR strains have been described, such as antibiotic therapy during the three months preceding surgery, prolonged hospital stay, previous and/or prolonged antibiotic therapy [12-15]. However, the timing of emergence of resistance strains has rarely been described. Seguin et al. suggested that a 5-day cut off in length of hospital stay or between the first operation and repeat laparotomy was the best predictive factor for the presence of MDR bacteria [13,14]. Our findings confirm this cut off and highlight the sustained and progressive increase in the proportions of MDR with the number of reoperations. This progression rate can be estimated to be about 15% between each reoperation, as illustrated by our sequential observations.

Among Gram-negative bacteria, MDR Enterobacteriaceae, mainly represented by Enterobacter spp. and E coli, were the predominant strains at the first reoperation. The importance of MDR strains of E. coli has been previously reported, but to a lesser extent, while the capacity of Enterobacter spp. to develop resistance is commonly observed [12,13,15,31]. The development of NFGNB, mainly P. aeruginosa, has been largely reported in the literature [12,13,15,31], although these organisms are usually observed in limited proportions [32]. Although their ability to develop resistance is well known, their progressive emergence from the first to the third reoperation to the detriment of Enterobacteriaceae is a noteworthy finding. These observations, well known in ventilator-associated pneumonia, are poorly documented in intra-abdominal infections.

The absence of MRSA among MDR Gram-positive cocci is another important finding compared to previous reports [15,31,33]. The decreased incidence of MRSA observed progressively over the last two decades in healthcare-associated infections in French institutions could explain this trend [34]. Coagulase-negative staphylococci, regularly mentioned in reports of tertiary peritonitis [2,5,7], were almost constantly observed in polymicrobial infections of the present study, suggesting that they only play a colonising role. In the absence of data concerning their pathogenicity, these strains were frequently taken into account in the choice of anti-infective therapy. The presence of E. faecium was largely expected, as these organisms are commonly reported in postoperative peritonitis, accounting for 30 to 50% of all Gram-positive cocci in previous reports [29,35], but the proportions resistant to amoxicillin and aminoglycosides were impressive.

Previous studies on tertiary peritonitis have reported the importance of yeasts [2,5,11]. The proportions of yeasts and the predominance of C. albicans remained stable from index surgery to the third reoperation. The great majority of patients treated by antifungal agents received fluconazole according to the recommendations available at the time. The use of fluconazole rather than a fungicidal agent might explain this high proportion of Candida isolates. We also observed a frequent association between Candida and staphylococci and NFGNB which may not be simply a coincidence, as mechanisms of synergy have been described experimentally between these organisms, [36,37]. The pathophysiology of persistent peritonitis remains poorly understood and some similarities with biofilms could be hypothesized, related to the presence of fibrin. The emerging concept of fungal prophylaxis in patients with recurrent abdominal perforation could be relevant in this setting [38,39].

Emerging organisms, especially MDR strains, have been commonly reported in the course of postoperative peritonitis [12-15,29,31] leading to the proposal of specific strategies targeting the largest number of organisms right from the empirical phase of treatment [12]. Interestingly, Gram-negative organisms were the predominant emerging strains. A progressive tapering in the emergence of Candida, and to a certain extent Gram-positive strains, was observed with no clear explanation. The potential protective role of Candida as a risk factor for emergence of microorganisms needs to be confirmed, as Candida peritonitis is also a factor for poor prognosis [28].

Finally, we did not identify any specific risk factors for emergence of MDR microorganisms compared to those previously reported [12-14]. Despite a cautious analysis of the relationships between antibiotic agents and emerging strains, we did not evidence a role of any specific class of antimicrobial agent. The only obvious finding was a decreased incidence of Gram-positive strains in the presence of vancomycin therapy. No recommendations can be proposed concerning the optimal antibiotic regimen on the basis of these findings. However, our observations led to a change in our approach to the use of anti-infective agents. Despite the absence of proof in the field of intra-abdominal infections, we now try to optimize the administration of antibiotic agents with increased dosing and prolonged or continuous infusion of beta-lactams [40]. In addition, we try to rapidly de-escalate our empirical therapy in order to preserve our antibiotic resources. Finally, we now use echinocandins as first-line therapy when fungal infection is suspected.