Impact of empirical treatment in extended-spectrum beta-lactamase-producing Escherichia coli and Klebsiella spp. bacteremia. A multicentric cohort study

We studied cases of ESBL-E bacteremia diagnosed from January 2004 through December 2008 in 19 Spanish Hospitals, included in the Infectious Diseases Research Group of the Spanish Society of Internal Medicine. Cases were identified using the microbiology laboratory databases and reviewed by internal medicine/infectious diseases physicians.

Patients were identified retrospectively and reviewed using a standardized data collection sheet. Only the first episode of mono-microbial bacteremia of each patient was selected for the study. All charts were reviewed. A multi-institutional database was developed including the following variables: age, sex, origin of bacteremia, dates of hospital admission and discharge, comorbidities, antimicrobial therapy, surgical procedures, hospitalization in an intensive care unit, and in-hospital mortality.

Bacteremia was considered to have been nosocomially acquired if appearing 48 hours after admission and no evidence of infection was present on admission. Episodes were considered health care–associated according to the criteria of Friedman et al [7]. The source of bacteremia was determined by clinical assessment and appropriate cultures (urine, sputum…) when available. Renal failure was defined by a creatinine value >2.0 mg/dL. Neutropenia was defined as an absolute neutrophil count of <500 cells/mm3 at the onset of the bacteremia. Immunosuppression was defined as the presence of neutropenia or HIV infection (with CD4 count <350 cells/mm3), or immunosuppressive therapy. Co-morbidities were assessed by using the Charlson co-morbidity score [8]. Sepsis, severe sepsis and septic shock were defined according to the American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference [9]. Empirical antimicrobial therapy was defined as adequate in terms of in vitro susceptibility of an organism isolated, and if antibiotic treatment was started within 24 hours after drawing blood cultures. The correction for definitive therapy was defined as that implemented after blood cultures results was obtained in patients with inadequate empirical therapy, and also was considered adequate or inadequate in terms of in vitro susceptibility of an organism isolated. Oxyimino-beta-lactams (cefuroxime, cefotaxime, ceftriaxone, ceftazidime and aztreonam) were considered to be inappropriate, regardless of the MIC, for the treatment of infections caused by ESBL-producing Enterobacteriaceae. Therapy with urinary antiseptics such as norfloxacin, fosfomycin, pipemidic acid or nalidixic acid was considered inadequate [10]. Organisms reported as intermediate in susceptibility to a particular antibiotic were classified as non-susceptible for this report. ESBL-E showing resistance to at least 1 representative of 2 other families of antibiotics (fluoroquinolones, beta-lactam/beta-lactam combinations, or aminoglycosides) were considered multidrug resistant (MDR). The dosage of the antibiotic was not taken into account when assessing adequacy. All study definitions were established before data analysis. The study was approved by the Regional Institutional Review Board of Cantabria.

ESBL-E isolation and identification were performed following standard methods at each centre. ESBL production was screened at the same time by the double disc synergy method as a complementary test for automated systems, at each centre. Minimal inhibitory concentrations were interpreted according to Clinical and Laboratory Standards Institute guidelines [11].

Based on the results of the literature a sample size calculation was performed with the openepi program (http://​www.​openepi.​com). A mortality of 20% was considered in the adequate empirical treatment group and of 35% in the inadequate treatment group. For a power of 80% and a level of significance of 0.05, an overall sample of 270 patients was required (50% with adequate empirical treatment).

Categorical data were compared by the x2 or Fisher’s exact tests. Quantitative data were compared by Student’s t-test or the Mann–Whitney U-test, as appropriate. All reported p values are two-sided and have not been adjusted for multiple testing. Variables with a p crude value <0.1 in the univariate analysis were candidates for multivariate analysis and possible interactions were evaluated. The analysis was performed with the stepwise logistic–regression model of the SPSS software package. All subgroup analysis was a priori defined.

The study included 399 patients detected in the participating centres. Excluded from the study were 12 (3%) patients due to insufficient data for the analysis. Among the remaining 387 patients, in hospital mortality was 20.9%. E. coli was isolated from blood cultures in 343 (88.6%) cases and K. pneumoniae in 44 cases (11.4%). The mean ± standard deviation patient age was 70.3 ± 16 years. Three patients were neonates (one with bacteremia caused by E. coli and two by K. pneumoniae). Seventy-two patients (18.6%) had severe sepsis and fifty-three (13.7%) had septic shock as a manifestation of the ESBL-E bacteremia. Median Charlson index was 3 (range 0-12). Table 1 shows the characteristics of the patients. Percentage of non-susceptibility of ESBL-E to antimicrobial was: quinolones: 82.9%, amoxicillin-clavulanate: 40.3%, gentamicin: 27.1%, piperacillin-tazobactam: 16.8% and carbapenem: 0%. In 45.7% of the cases the ESBL-E was multidrug resistant. In 189 (48.8%) episodes of bacteremia, the empirical antibiotic treatment was inadequate according to the microbiological results. In 44 (11.4%) episodes the definitive antibiotic treatment was inadequate. Of these patients with definitive inadequate therapy 24 (54.5%) patients died in the first three days after blood cultures extraction.

Of the 387 episodes of bacteremia 84 (21.7%) were detected in the period 2004-2005, and 303 (78.3%) in the period 2006-2008. When we compared the characteristics of the patients between these two periods we detected significant differences proportion in the frequency of empirical antibiotic treatment with piperacillin-tazobactam (7.3% vs. 14.9%, p = 0.03). Respect to E-ESBL resistance pattern, a trend to decrease in susceptibility from the first to the second period to amoxicillin-clavulanate (66.7% vs. 57.8%, p = 0.08) and decrease to ciprofloxacin (25% vs. 14.9%, p = 0.02) was detected and an increase in the proportion of multiresistant infections (35.7% vs. 48.5%, p = 0.02). A trend to decrease in the proportion of patients with community acquired infections was also detected (35.7% vs 26.4%, p = 0.06).

The results of the univariate analysis of the risk factors associated with in hospital mortality are reflected in Table 2. Variables associated with mortality in the multivariate analysis were: hospital acquisition [adjusted OR (95% CI): 3.41 (1.89-6.18); P < 0.001], presence of metastatic disease [adjusted OR (95% CI): 4.2 (1.78-9.95); P = 0.001], adequacy of empirical antibiotic therapy [adjusted OR (95% CI): 0.39 (0.31-0.97); P = 0.04], presence of severe sepsis or shock [adjusted OR (95% CI): 11.66 (6.28-21.65) P < 0.001].

A multivariate analysis selecting patients without severe sepsis or shock showed that adequate empirical therapy was not associated with mortality in this group, but it was associated with mortality in patients with severe sepsis or shock (adjusted OR, 0.42; 95% CI, 0.19–0.92; P = 0.03).

The most frequent empirical antibiotic treatments in the entire cohort were carbapenem (18.1%), amoxicillin/clavulanate (17.6%), ciprofloxacin or levofloxacin (16%), ceftriaxone or cefotaxime (12.9%) and piperacillin/tazobactam (12.9%).

A stratified analysis of mortality considering adequacy with different empirical antibiotic treatments is reflected in Table 3. For the group of patients who received adequate empirical treatment with beta-lactam/beta-lactam inhibitor combinations or with carbapenem mortality was 13.1% (11 of 84) and 24.6% (17 of 69) respectively (p = 0.05).