Risk factors for multidrug resistant bacteria and optimization of empirical antibiotic therapy in postoperative peritonitis

From January 2001 to December 2004, all consecutive adult patients with a diagnosis of PP requiring admission to a surgical intensive care unit (ICU) were prospectively included in a database, and their medical charts were retrospectively reviewed. PP was defined as a peritoneal infection occurring after an initial abdominal surgery (S0), and confirmed by macroscopic findings and positive bacterial fluid culture yielding at least one microorganism (bacteria or yeast) at reoperation. In patients who required multiple reoperations, only the first one was considered. All types of abdominal surgery were included except cases of complicated acute pancreatitis. Patients with PP with pure fungal infection were not analysed. According to French law, because this observational study did not modify the physicians' laboratory or clinical practices, no informed consent was required. The Institutional Review Board of Paris North Hospitals, Paris 7 University, AP-HP, reviewed and approved the study.

Peritoneal fluid samples were systematically collected during reoperation and immediately sent to the bacteriology laboratory. Gram staining for direct examination and cultures were performed with identification and susceptibility testing for Gram-positive and Gram-negative bacteria. Antibiotic susceptibility was determined by the disk-diffusion method, according to the criteria of the Antibiogram Committee of the French Society for Microbiology [10]. In vitro susceptibility of nine antibiotics (amoxicillin/clavulanic acid (amox/clav); piperacillin/tazobactam (pip/taz); ceftazidime; imipenem/cilastatin; ciprofloxacin; gentamicin; amikacin and specifically metronidazole and vancomycin (for anaerobes and Gram-positive cocci)) was recorded for all bacteria. Results were expressed as proportions of susceptible bacteria for each antibiotic. Parenteral EA was systematically started at the time of reoperation according to the recommendations of our institutional protocol for PP. This protocol is based on treatment with a broad-spectrum beta-lactamin pip/taz or imipenem. Imipenem is selected for patients with severe peritonitis and/or previous antimicrobial therapy. The use of amikacin for spectrum broadening and synergistic combination is optional. The adjunction of vancomycin is considered in cases of prolonged hospital stay or methicillin-resistant staphylococcus or amoxicillin-resistant enterococcus carriage. Adequacy of EA was assessed according to the regimen used and the number of antibiotics in the case of combination therapy. Empirical antimicrobial therapy was considered adequate if, according to the susceptibility testing, all bacteria isolated were susceptible to at least one of the drugs administered. The antibiotic selection was considered to be adequate or inadequate strictly on the basis of the culture results obtained and did not reflect the authors' subjective assessment of appropriateness of care.

Analysis of antibiotic regimens classified as monotherapy or combination therapy (two-, three- and four-drug regimens) allowed the assessment of 17 potential regimens in order to determine suitable treatments providing adequate EA in the largest number of cases. This analysis was performed according to the presence or absence of MDR bacteria, and then according to the presence or absence of a risk factor for MDR strains found in our analysis. As the purpose of this study was to focus on antimicrobial therapy, fungi were not included in the definition of adequacy.

MDR bacteria were defined as: methicillin-resistant Staphylococcus aureus and coagulase-negative staphylococci (CNS); Enterobacteriaceae producing an extended-spectrum beta-lactamase or producing a cephalosporinase: and non-fermenting Gram-negative aerobes resistant to pip/taz, ceftazidime, or imipenem/cilastatin, or producing an extended-spectrum beta-lactamase (Pseudomonas aeruginosa and Acinetobacter baumanii). In line with the IDSA and SIS guidelines considering broad-spectrum agents active against P. aeruginosa, and methicillin-susceptible and amoxicillin-susceptible Enterococcus, we arbitrarily defined pip/taz, imipenem/cilastatin, and fluoroquinolones as broad-spectrum antibiotics. Interval antibiotics (IA) were defined as antimicrobial agents administered between S0 and reoperation, at during at least 24 hours and started at least 24 hours before reoperation. The use of all-types of IA and broad-spectrum IA during this period was recorded in every case and constituted new variables for the analysis. The reason for their prescription was recorded.

The patients' medical charts were reviewed and the following information was collected: age; gender; severity of the underlying medical condition [11]; presence of chronic diseases (such as malignancy; diabetes mellitus; steroid or immunosuppressive therapy for inflammatory bowel disease); previous hospitalization or antibiotic therapy within three months before S0; characteristics of S0, if performed in another institution, its type, route and wound class [12]; and use of IA. Parameters collected within the first 48 hours after ICU admission were: temperature; acute physiology and chronic health evaluation (APACHE) II score [13]; Sequential Organ Failure Assessment (SOFA) score [14]; organ failures assessed following Knauss definitions [15]; etiology and primary site (above or below transverse mesocolon) of the infection responsible for PP and time to reoperation; identification of pathogens in peritoneal fluid; and results of antimicrobial susceptibility tests.

Patient outcome was recorded as the number of reoperations, duration of mechanical ventilation, ICU length of stay, and ICU mortality. The prognosis was assessed by taking into account the presence of MDR organisms and the adequacy of EA.

Results are expressed as mean ± standard deviation, and as percentages for categorical variables. All analyses were performed using the Statview software package (version 5.0; SAS institute Inc, Cary, NC, USA). As the primary objective of the study was to determine risk factors and outcome of PP patients with MDR bacteria, the group of patients with MDR bacteria (called MDR group) was compared with the group of patients with 'other' bacteria (called other group). Secondly, the impact of broad-spectrum IA on susceptibility of microorganisms collected from peritoneal samples was analyzed. Univariate analysis was performed using Student's t-test or Wilcoxon's rank sum test, as appropriate for continuous variables, and the Chi squared or Fisher's exact test, as appropriate, for categorical variables. All variables with a P value less than 0.10 in the univariate analysis were entered into a multivariate logistic regression analysis. Odds ratio (OR) and 95% confidence intervals (CI) were calculated. Statistical significance was defined as P < 0.05.

During the study period, 107 patients with PP were admitted to our ICU. Seven patients were excluded because only fungi were found on culture. Epidemiologic characteristics, clinical status of the 100 patients on admission and clinical findings at the time of reoperation are shown in Tables 1 and 2. Initial surgery was digestive in 80 cases, hepatobiliary in 5 cases, urologic in 7, mixed urologic/digestive in 3 cases, and gynaecologic in 8 cases. In this study population, the presence of MDR bacteria was reported in 41 PP patients and 59 PP patients were free of MDR strains. According to univariate analysis, factors associated with the presence of MDR bacteria in peritoneal samples at the time of PP were emergent initial surgery, contaminated or infected initial surgery, prior antibiotic therapy before S0, IA and broad-spectrum IA. When these variables were entered into a logistic regression model, the use of broad-spectrum IA was the only significant risk factor for emergence of MDR bacteria (OR = 5.1; 95% CI = 1.7 to 15; P = 0.0031).

A total of 269 bacteria were cultured from peritoneal fluid (Table 3). Twenty five yeasts were isolated including Candida albicans (n = 12), Candida glabrata (n = 7) and Candida tropicalis (n = 4). Most patients (n = 68) received all-types of IA, and 35 of them received broad-spectrum IA. The main reasons for IA were contaminated or septic initial surgery, suspicion or occurrence of PP (n = 26), and new focus of infection (n = 21) including 12 cases of pneumonia. The distribution of bacteria according to the use of broad-spectrum IA therapy is presented in Table 4. The number of bacteria cultured from peritoneal fluid, was not different when broad-spectrum IA therapy had been administered (2.5 ± 1.7 vs 2.8 ± 2.1, P = 0.22). In these patients, we observed that cultures of peritoneal fluid samples exhibited a trend toward increased proportions of monomicrobial samples (20% vs 8% in patients without broad spectrum IA therapy, P = 0.18), with a higher number of MDR microorganisms, mainly due to resistant Enterobacteriaceae and methicillin-resistant CNS (P < 0.05 for both cases). All-types of IA were associated with a decreased number of bacteria (2.4 ± 1.5 vs 3.4 ± 2.4, P = 0.001) and PP was more often monomicrobial PP (28% vs 3%, P = 0.001).

Proportions of susceptible Gram-negative and Gram-positive strains have been evaluated. Among the various antibiotics tested, imipenem/cilastatin and amikacin were the most consistently active against aerobic Gram-negative bacteria in all patients, whereas the efficacy of pip/taz (87% vs 40%, P < 0.0001) and ceftazidime (87% vs 60%, P = 0.009) was markedly reduced in patients with broad-spectrum IA therapy. Vancomycin was the agent most frequently active against Gram-positive bacteria in all patients, except in one case of a naturally resistant Enterococcus casseliflavus strain. Following broad-spectrum IA therapy, staphylococci were resistant to beta-lactams and ciprofloxacin. The 36 cultured anaerobes had susceptibility rates of 87%, 93%, 93% and 100% toward amox/clav, pip/taz, metronidazole, and imipenem/cilastatin, respectively. Among the 20 Bacteroides strains, four were resistant to amox/clav, two to pip/taz and one to metronidazole.

We analysed EA prescribed at the time of reoperation in the 100 PP patients: monotherapy in 53 cases (45 pip/taz; 5 imipenem), double-drug combinations in 32 cases (13 based on pip/taz; 10 based on imipenem), and triple-drug combinations in 13 cases (4 based on pip/taz; 4 based on imipenem). Adequacy rates were 64%, 66%, and 62%, for monotherapies, double-drug combinations, and triple-drug combinations, respectively.

Pip/taz (n = 66) and imipenem/cilastatin (n = 23) were the main agents prescribed. Imipenem/cilastatin was more frequently administered than pip/taz in seriously ill patients (SOFA score 6 ± 4 vs 9 ± 3, P = 0.005), and in the case of prior broad-spectrum IA therapy between S0 and reoperation (87% for imipenem vs 65% for pip/taz; P = 0.04). A higher SOFA score was also associated with prescriptions of combinations rather than monotherapy (6 ± 4 for monotherapy vs 8 ± 4 for combination; p = 0.03). Three allergic patients received triple-drug combinations without beta-lactams. One patient with previous colonization by a multiresistant strain of P. aeruginosa received a four-drug combination (imipenem/cilastatin + vancomycin + aminoglycosides + colistin). One patient received antifungal therapy only because of previous fungal colonization and negative direct examination of peritoneal fluid.

Adequate EA was achieved in 64% of cases. Adequacy of EA decreased significantly in patients with MDR bacteria, as compared with patients with 'other bacteria' (39% vs 81%, P < 0.0001).

Evaluation of the adequacy rates of 17 theoretical regimens in the 100 episodes of PP according to the presence or absence of MDR bacteria, and according to the prescription of a broad-spectrum IA are shown in Figures 1 and 2, respectively. Only combination regimens including vancomycin achieved empirical therapy adequacy rates higher than 80%. Regimens based on imipenem/cilastatin obtained the highest adequacy rate. In patients with broad-spectrum IA, monotherapy with imipenem/cilastatin provided only poor adequacy rates, but was suitable for patients without broad-spectrum IA. Monotherapy with pip/taz gave poor results even in patients without broad-spectrum IA.

Forty-four patients had a reoperation after R1 (first repoperation at ICU admission) because of persistent peritonitis. ICU mortality rate was 31%. Mortality did not differ between patients with adequate EA and others (30% vs 31%, P = 0.9), and between patients with PP caused by MDR bacteria and other bacteria (29% for MDR group vs 35% for others, P = 0.69). The mean duration of antibiotic therapy (10 ± 4 days vs 12 ± 6 days, P = 0.07), mechanical ventilation (10 ± 9 days vs 11 ± 16 days, P = 0.6), length of ICU stay (16 ± 11 days vs 20 ± 19 days, P = 0.2), as well as the number of reoperations (0.8 ± 1.4 vs 0.8 ± 1, P = 0.9) were similar in patients with adequate EA and other patients, respectively. No outcome difference was observed between patients with MDR bacteria and patients with other microorganisms.