Severe sepsis and septic shock

Feature Articles Antibiotic strategies in severe nosocomial sepsis: Why do we not de-escalate more often?* Sarah Heenen, MD; Frédérique Jacobs, MD; Jean-Louis Vincent, MD, PhD, FCCM Objectives: To assess the use of antibiotic de-escalation in patients with hospital-acquired severe sepsis in an academic setting. Design: We reviewed all episodes of severe sepsis treated over a 1-yr period in the department of intensive care. Antimicrobial therapy was considered as appropriate when the antimicrobial had in vitro activity against the causative microorganisms. According to the therapeutic strategy in the 5 days after the start of antimicrobial therapy, we classified patients into four groups: de-escalation (interruption of an antimicrobial agent or change of antibiotic to one with a narrower spectrum); no change in antibiotherapy; escalation (addition of a new antimicrobial agent or change in antibiotic to one with a broader spectrum); and mixed changes. Setting: A 35-bed medico-surgical intensive care department in which antibiotic strategies are reviewed by infectious disease specialists three times per week. Patients: One hundred sixty-nine patients with 216 episodes of severe sepsis attributable to a hospital-acquired infection who required broad-spectrum -lactam antibiotics alone or in association with other anti-infectious agents. Measurements and Main Results: The major sources of infection were the lungs (44%) and abdomen (38%). Microbiological data were available in 167 of the 216 episodes (77%). Initial antimicrobial therapy was inappropriate in 27 episodes (16% of culture-positive episodes). De-escalation was applied in 93 episodes (43%), escalation was applied in 22 episodes (10%), mixed changes were applied in 24 (11%) episodes, and there was no change in empirical antibiotic therapy in 77 (36%) episodes. In these 77 episodes, the reasons given for maintaining the initial antimicrobial therapy included the sensitivity pattern of the causative organisms and previous antibiotic therapy. The number of episodes when the chance to de-escalate may have been missed was small (4 episodes [5%]). Conclusion: Even in a highly focused environment with close collaboration among intensivists and infectious disease specialists, de-escalation may actually be possible in <50% of cases. (Crit Care Med 2012; 40: 1404 1409) Key Words: antimicrobials; empiric antibiotics; infection; microbiology Severe sepsis and septic shock are common causes of morbidity and mortality, especially in the intensive care unit (ICU) (1). Infection control with effective antibiotic therapy and source elimination whenever indicated is a cornerstone in the management of sepsis. Early effective antibiotic therapy has been shown to decrease mortality rates (2 5) so that rapid institution of broad-spectrum antibiotic therapy is recommended if bacteriologic information is not available. However, broad-spectrum antibiotic *See also p. 1645. From the Departments of Intensive Care (SH, JLV) and Infectious Disease (FJ), Erasme Hospital, Université Libre de Bruxelles, Brussels, Belgium. Institutional funds only were received. The authors have not disclosed any potential conflicts of interest. For information regarding this article, E-mail: jlvincen@ulb.ac.be Copyright 2012 by the Society of Critical Care Medicine and Lippincott Williams & Wilkins DOI: 10.1097/CCM.0b013e3182416ecf therapy can favor the emergence of resistant organisms (6, 7). Despite no strong evidence (8), antibiotic de-escalation, in which wide-spectrum empirical antibiotic therapy is started initially but the spectrum is narrowed as soon as microbiological results are available, therefore has been recommended to reduce selection pressure and possibly to reduce toxicity and limit costs (9 11). Concerns about the risk of recurrent infection in de-escalation therapy are probably unjustified (12, 13). De-escalating strategies are recommended particularly in hospital-acquired infections, in which the number of potential types of causative microorganism is relatively large and the risk of resistance is particularly high. Antibiotic de-escalating strategies have been studied primarily in hospital-acquired pneumonia (9, 12, 14 18). In these studies, the rate of de-escalation varied considerably, from 6% to 74%. The aim of our study was to assess how de-escalation is applied in patients with hospital-acquired severe sepsis in our intensive care department in which close collaboration with infectious disease specialists is associated with frequent reassessment of antibiotic regimens. We also assessed the impact of de-escalation on outcome and determined which factors were associated with de-escalation in our cohort. PATIENTS AND METHODS Patients The study was conducted in the 35-bed medico-surgical intensive care department of the academic hospital of the University of Brussels. We reviewed the files of all adult (older than 18 yrs) patients over a 1-yr period (2007) who were treated for severe sepsis (with or without septic shock) because of a hospital-acquired infection and required broad-spectrum b-lactam antibiotics, including piperacillin-tazobactam, ceftazidime, cefepime, and meropenem alone or in association with other anti-infectious agents. As a retrospective study, ethics committee approval was requested but patient consent was waived. Severe sepsis and septic shock were defined by standard criteria (19). 1404 Crit Care Med 2012 Vol. 40, No. 5

Patients who died within the first 3 days of therapy, before de-escalation could be instituted, were excluded. The choice of empirical treatment was made according to local guidelines and results of any previous microbiological data. Piperacillin-tazobactam was preferred as the first-line therapy in proven or suspected intra-abdominal infections. Ceftazidime and cefepime were used as first-line therapy in other cases, including ventilatoracquired and hospital-acquired pneumonia. Meropenem was used as second-line therapy (i.e., failure of piperacillin-tazobactam or cephalosporins) or in suspected or previous colonization by extended-spectrum b-lactamase (ESBL)-producing Gram-negative bacteria. Amikacin was usually added for 1 3 days in cases of severe sepsis (with high suspicion of Gram-negative bacteria involvement). Vancomycin was added when infection with methicillin-resistant Staphylococcus aureus or methicillin-resistant Staphylococcus epidermidis was suspected. An antifungal drug, generally fluconazole, was added in patients highly colonized by fungi, mostly in cases of sepsis resistant to broad-spectrum antibiotics. Demographic data, Sequential Organ Failure Assessment score (20) recorded on day 1 of the septic episode, source of infection, bacteriologic data, and antibiotic therapy for the 5 first days of treatment were noted for each episode. The site of infection, documented or suspected, was assessed according to standard criteria (21). Urine, bronchoalveolar lavage fluid, and catheters required quantitative cultures for confirmation of site: 100,000 colonies/ml for urine (semiquantitative), 10,000 colony-forming units/ml for bronchoalveolar lavage, and 15 colonies/ml for catheter cultures (22, 23). Cultures of other fluids (pleural, peritoneal, abscesses, tracheal aspirates, and blood) were nonquantitative. We considered isolated organisms to be colonizing if there was no clinical evidence of infection or no leukocytes in the samples (except in samples from normally sterile sites or from leukopenic patients). Microbiology results were considered as inconclusive when there was no growth of organism. For our study, multiresistant organisms were ESBL-producing Gram-negative rods, methicillin-resistant Staphylococcus aureus, and methicillin-resistant Staphylococcus epidermidis. An episode of severe sepsis was considered to start on the day of sepsis diagnosis and to end with either resolution of sepsis (significant decrease in white blood cell count or C-reactive protein concentration, resolution of cardiovascular, and other organ dysfunctions), the patient s death, or a new episode of sepsis. Worsening of a patient s condition (hemodynamic, respiratory, neurologic, renal, or hepatic, or a new increase in blood lactate concentration or a coagulopathy) in association with signs of sepsis was considered as a new episode of sepsis only if the patient had transiently improved before the degradation and the degradation occurred 5 days after the previous episode. In our unit, antimicrobial therapy is prescribed by the ICU doctor after discussion with infectious disease consultants (a separate specialty from the microbiologists at our hospital) who are available 24 hrs per day. Rounds with infectious disease specialists are held at least three times per week to discuss all relevant aspects of management. The ICU team also includes a clinical pharmacist who reviews all drug prescriptions. Antimicrobial therapy was considered appropriate when the antimicrobial agent had in vitro activity against the incriminated organisms. De-escalation was defined either as the discontinuation of an antimicrobial agent (antibiotic or anti-fungal) or as a change from one antibiotic to another, i.e., from meropenem to any other b-lactam, from ceftazidime, cefepime, or piperacillintazobactam to amoxicillin-clavulanic acid or any other type of penicillin, or from vancomycin to any type of penicillin. Escalation was defined as the addition of a new antiinfectious agent or a change in antibiotic therapy in the reverse direction to that described. Changes among cefepime, ceftazidime, and piperacillin-tazobactam were not considered a significant alteration in the spectrum of cover. For the purposes of this study, we grouped episodes according to the antimicrobial strategy in the 5 days after diagnosis: de-escalation (group I); no change in the empirical treatment (group II); escalation (group III); and mixed strategy, in which changes in therapy were made in both directions (group IV). In each case, we identified the reasons that led to the specific strategy. Statistical Analyses Statistical analyses were performed using IBM SPSS Statistics 19 for windows (SPSS Inc, Chicago, IL). The Kolmogorov-Smirnov test was used, and histograms and normal-quantile plots were examined to verify if there were significant deviations from the normality assumption of continuous variables. Difference testing among groups was performed using analysis of variance, Student t test, chi-square test, or Fisher exact test, as appropriate. The Bonferroni correction was made for multiple comparisons. To examine differences in outcomes among groups, we studied only the first episodes of sepsis. Logistic regression analysis was used to identify the factors associated with deescalation. Variables associated with de-escalation (p.2) on a univariate basis were introduced into the multivariable analysis. Colinearity between variables was excluded before modeling. A Hosmer-Lemeshow goodness-of-fit test was performed, and classification tables and odds ratios with 95% confidence intervals were computed. Data are presented as mean ± sd or number (%) as appropriate. All tests were twotailed, and p<.05 was considered statistically significant. RESULTS We included a total of 241 episodes of severe sepsis, of which 25 were excluded because the patients died within the first 3 days of treatment. Demographic and infection-related data concerning the 216 evaluable episodes in 169 patients are presented in Table 1; 136 patients had 1 episode, 24 had 2 episodes, and nine had 2 episodes of severe sepsis or septic shock. Microbiological cultures provided a conclusive bacteriologic diagnosis in 163 episodes (75%). Table 2 presents the organisms found in these episodes and in four additional episodes in which only surveillance swabs (throat/nose and rectum swabs) were positive; multiple organisms were found in 86 of these episodes (51%). At least one Gramnegative bacillus was involved in 121 of the culture-positive episodes (72%); 19 of the isolated Gram-negative bacillus (11%) had ESBL activity. Enterobacter spp. and ESBL-producing organisms were more common in episodes in which escalation was used than in other groups. At least one Gram-positive coccus was isolated in 87 episodes, of which 33 strains were methicillin-resistant staphylococci (38%). Fungi were present in 29 of the culture-positive episodes (17%). Initial antimicrobial therapy was appropriate in 140 of the 167 culturepositive episodes (84%). It was more frequently appropriate in patients who underwent de-escalation than in patients in whom treatment was escalated or who had mixed strategies (Table 2). The empirical antibiotics are shown in Table 3 according to therapeutic strategy. Glycopeptides were more commonly used in episodes for which de-escalation was performed (group I) than in the episodes in which escalation was used (group III). Use of amikacin was more common in group I than in the other groups. The use of monotherapy was also significantly more common in the escalation and unchanged treatment groups (groups II and III) than in group I. Figure 1 summarizes the therapeutic strategy for all episodes according to the Crit Care Med 2012 Vol. 40, No. 5 1405

Table 1. Demographic and infection-related data Total Episodes microbiological data. In the microbiologically documented infections (167 episodes), initial treatment was appropriate in 140 episodes (84%). De-escalation was possible in 89 episodes and was performed in 72 of these 89 episodes (81%). In the 49 episodes in which microbiological documentation was inconclusive, Group I (De-Escalation) Group II (No Change) Group III (Escalation) Group IV (Mixed) N 216 93 77 22 24 Male, n (%) 120 (56%) 51 (55%) 44 (57%) 14 (64%) 11 (46%) Age, yr (mean ± sd) 61 ± 14 61 ± 13 61 ± 15 62 ± 12 59 ± 12 Surgical admission, n (%) 106 (49%) 48 (52%) 35 (45%) 11 (50%) 12 (50%) Sequential Organ Function 9 ± 3 9 ± 4 9 ± 4 10 ± 4 9 ± 3 Assessment score (mean ± sd) Septic shock, n (%) 128 (59%) 52 (56%) 47 (61%) 15 (68%) 14 (58%) Septic source, n (%) Chest 94 (44%) 35 (38%) 32 (42%) 11 (50%) 16 (67%) Abdomen 83 (38%) 34 (37%) 31 (40%) 10 (45%) 8 (33%) Urinary tract infection 24 (11%) 11 (12%) 6 (8%) 3 (14%) 4 (17%) Soft tissue 35 (16%) 15 (16%) 13 (17%) 5 (23%) 2 (8%) Catheter 10 (5%) 4 (4%) 3 (4%) 2 (9%) 1 (4%) Primary bacteremia 11 (5%) 6 (6%) 3 (4%) 0 (0%) 2 (8%) Secondary bacteremia 46 (21%) 24 (26%) 6 (8%) a 5 (23%) 11 (46%) b Multiple 40 (19%) 11 (12%) 17 (22%) 5 (23%) 7 (29%) Unidentified 18 (8%) 8 (9%) 8 (10%) 1 (5%) 1 (4%) Positive microbiological documentation n (%) 167 (77%) 73 (78%) 53 (69%) 18 (81%) 23 (96%) b a Statistically different at 5% level vs. group I; b statistically different at 5% level vs. group II. Table 2. Detailed microbiological findings for the most frequently isolated microorganisms Total Episodes (Culture-Positive) Group I Group II Group III (De-Escalation) (No Change) (Escalation) Group IV (Mixed) N 167 73 53 18 23 Gram-negative bacilli, n (%) 121 (72%) 58 (79%) 36 (68%) 12 (62%) 15 (65%) Escherichia coli 43 (26%) 24 (33%) 12 (23%) 4 (22%) 3 (13%) Pseudomonas aeruginosa 30 (18%) 11 (15%) 10 (19%) 6 (33%) 3 (13%) Enterobacter sp. 20 (12%) 5 (7%) 7 (13%) 6 (33%) 2 (9%) c Klebsiella sp. 17 (10%) 6 (8%) 4 (8%) 2 (11%) 5 (22%) Nonfermenting Gram-negative 6 (4%) 3 (4%) 0 (0%) 0 (0%) 3 (13%) a bacteria Acinetobacter sp. 4 (2%) 3 (4%) 1 (2%) 0 (0%) 0 (0%) Extended spectrum 19 (11%) 6 (8%) 4 (8%) 7 (39%) a,b 2 (9%) beta-lactamase Gram-positive cocci 87 (52%) 33 (45%) 32 (60%) 7 (39%) 15 (65%) Methicillin sensitive 23 (14%) 12 (16%) 6 (11%) 1 (6%) 4 (17%) Staphylococcus aureus Methicillin resistant 11 (7%) 6 (8%) 3 (6%) 1 (6%) 1 (4%) Staphylococcus aureus Methicillin resistant 22 (13%) 6 (8%) 10 (19%) 1 (6%) 5 (22%) Staphylococcus epidermidis Enterococci sp. 31 (19%) 10 (14%) 14 (26%) 4 (22%) 3 (13%) Fungi 29 (17%) 7 (10%) 9 (17%) 5 (28%) 8 (35%) a Clostridium difficile 8 (5%) 2 (3%) 2 (4%) 1 (6%) 3 (13%) Polymicrobial infection 86 (51%) 36 (49%) 27 (51%) 10 (56%) 13 (57%) Appropriate initial anti-microbial therapy 140 (84%) 72 (97%) 53 (100%) 8 (44%) a 7 (30%) a a Statistically different at 5% level vs. group I; b statistically different at 5% level vs. group II; c statistically different at 5% level vs. group III. therapy was de-escalated in 20 (41%), mostly because no resistant bacteria were found in any samples. Of the 93 total episodes of de-escalation, 61 involved stopping an anti-infectious agent, 20 involved a change in antibiotic, and 12 involved both an interruption and a change. Taking a closer look at the 77 group II episodes to determine why there was no change in antimicrobial strategy, we found that in 35 of the episodes (46%), de-escalation was not possible because of the sensitivity of the microorganisms; the spectrum was already as narrow as it could be. In 25 episodes (32%), there was no change in therapy because of lack of microbiological data (a late culture returned positive in six episodes); in most of these cases, patients had already been receiving antibiotics when sepsis developed and had only improved by increasing the spectrum of cover, so physicians were then reluctant to reduce it without microbiological data. In ten episodes (13%), the microbiological data were considered inconclusive by the physician and the patient was not considered a candidate for de-escalation; in these episodes, patients had either been receiving antibiotic therapy when the sepsis occurred and microbiological documentation showed microorganisms sensitive to the previous treatment (six episodes), or there were multiples sites of infection with negative cultures for one or more of them (four episodes). In three episodes (4%), the patient was colonized with multiresistant organisms and the physician was not comfortable to deescalate and to potentially lose coverage of the multiresistant organisms. In four episodes (5%), there was no good reason why the antibiotherapy was not changed and the opportunity to de-escalate was simply missed. In 46 episodes, therapy was escalated or mixed strategies were used (groups III and IV), including five episodes in which the microbiological data were negative. Some escalations were performed even when the causative organisms were covered by the empirical treatment (15 episodes [33%]). The major reasons for escalation were: in 28 episodes (61%), microbiological findings revealed a microorganism that was potentially not covered by the empirical therapy (pending antibiogram and late cultures included); in 11 (24%) episodes, there was an inadequate response to the initial empirical treatment (degradation within the 5 first days of treatment); and in seven episodes (15%), there was a re-evaluation within 12 24 hrs. The organisms that were not covered were Candida albicans and nonalbicans (eight episodes), ESBL-producing bacteria (six episodes), methicillinresistant Staphylococcus epidermidis (three episodes), Clostridium difficile 1406 Crit Care Med 2012 Vol. 40, No. 5

Table 3. Initial empirical antibiotic therapy according to therapeutic strategy Total Episodes Group I (De-Escalation) Group II (No Change) Group III (Escalation) Group IV (Mixed) N 216 93 77 22 24 Used agent, n (%) Piperacillin-tazobactam 81 (38%) 32 (34%) 33 (43%) 11 (50%) 5 (21%) Ceftazidime 5 (2%) 1 (1%) 2 (3%) 1 (5%) 1 (4%) Cefepime 55 (25%) 28 (30%) 12 (16%) 6 (27%) 9 (38%) Meropenem 75 (35%) 32 (34%) 30 (39%) 4 (18%) 9 (38%) Amikacin 33 (15%) 29 (31%) 0 (0%) a 0 (0%) a 4 (17%) b Vancomycin/linezolid 119 (55%) 62 (67%) 38 (49%) 5 (23%) a 14 (58%) Fluconazole 34 (16%) 19 (20%) 11 (14%) 2 (9%) 2 (8%) Cotrimoxazole 10 (5%) 6 (6%) 3 (4%) 1 (5%) 0 (0%) Other 63 (29%) 28 (30%) 21 (27%) 7 (32%) 7 (29%) Regimen, n (%) 1 agent 55 (25%) 11 (12%) 29 (38%) a 10 (45%) a 5 (17%) 2 agents 90 (40%) 40 (43%) 28 (36%) 10 (45%) 12 (54%) 3 agents 51 (24%) 27 (29%) 16 (21%) 1 (5%) 7 (29%) 3 agents 20 (9%) 15 (16%) 4 (5%) 1 (5%) 0 (0%) a Statistically different at 5% level vs. group I; b statistically different at 5% level vs. group II. aureus (three), methicillin-resistant Staphylococcus epidermidis (ten), and Enterococcus with either resistance or intermediate sensitivity to penicillin (seven). The remaining 18 episodes in which vancomycin was not de-escalated with no obvious microbiological support to keep it were mainly soft tissue infections for which vancomycin was kept for its good soft tissue penetration. There were also three episodes of previous methicillin-resistant Staphylococcus aureus/methicillin-resistant Staphylococcus epidermidis colonization. In cases of negative culture (49 episodes), de-escalation was performed in 20 episodes, escalation was performed in 5 episodes, and no change was performed in 24 episodes. Among these 24 episodes, the reason for no de-escalation was: monotherapy (12); soft tissue infections (four); septic episodes (three) while the patient was using antibiotics; two cases of positive Gram-negative stain tests with no growth; and three cases in which the patients were immunosuppressed and using large-spectrum anti-microbial therapy. Considering only the 169 first episodes of sepsis, de-escalation was performed in 79, no change was performed in 58, escalation was performed in 15, and mixed changes were performed in 17. Forty-four of these 169 patients died, 13 (16%) in group I, 15 (26%) in group II, 9 (60%) in group III, and 7 (41%) in group IV (p =.002). In the multivariable analysis, appropriate initial antimicrobial therapy (odds ratio 2.7; 95% confidence interval 1.27 5.73; p =.01) was the only factor associated with increased odds of de-escalation; monotherapy compared to multitherapy (odds ratio 0.17; 95% confidence interval 0.06 0.5, p =.001) were associated with decreased odds of de-escalation. DISCUSSION Figure 1. Overview of the therapeutic strategy in the 216 episodes classified according to positive microbiological results, effectiveness of initial antibiotherapy, and possibility of de-escalation according to sensitivity of isolated organisms. (two episodes), and other organisms (nine episodes). Because the rate of ESBL is relatively low in our hospital, meropenem is reserved as a second-line treatment for nosocomial infection or for patients known to carry an ESBL-producing Gram-negative rods. In these conditions, microbiological results are often inconclusive and meropenem was rarely de-escalated in the absence of documentation. In contrast, vancomycin is frequently prescribed and de-escalated after 2 or 3 days in our hospital. In the present study, it was prescribed in 119 episodes and maintained in just 38. Taking a closer look at these 38 episodes, vancomycin was maintained in 20 episodes for methicillin-resistant Staphylococcus Current guidelines on the management of severe sepsis recommend early broad-spectrum antibiotic therapy with de-escalation as soon as possible (11). Our ICU team benefits from a close collaboration with infectious disease specialists in addition to microbiologists; nevertheless, 16% (27/167 episodes) of patients had organisms that were not covered by the empirical antibiotic treatment and our de-escalation rate was only 43%. In previous studies of ventilator-associated pneumonia (VAP) or hospitalacquired pneumonia, de-escalation rates Crit Care Med 2012 Vol. 40, No. 5 1407

have varied broadly, from 6% to 74% (9, 12, 14 18, 24). Comparisons among studies are made difficult by the lack of precise definitions of de-escalation, differences in empirical regimens, and the differences in patient populations and local microbioloigical epidemiology. One of the highest de-escalation rates was reported by Eachampati et al (12) in 135 cases of VAP. These authors used a wider empirical antibiotic spectrum than in our present study, with a 23% rate of Gram-negative bacillus-targeted monotherapy and an overall rate of bitherapy of 77% (including association of vancomycin with piperacillin-tazobactam, quinolones, carbapenem, or cefepime). Their rate of appropriate antibiotic therapy was 93% with these regimens. Only ten patients required antibiotic escalation (7%). In 2006, the Canadian Critical Care Trials groups (18) performed a large randomized study of techniques to diagnose VAP. They included patients from 28 ICUs in Canada and had a large-spectrum initial antibiotherapy with meropenem alone and meropenem in association with ciprofloxacin. Their de-escalation rate was high (74%), largely because 21% of the enrolled patients were not infected and treatment was therefore stopped, and because the wide spectrum of the empirical treatment easily allowed streamlining. Despite their initial broad antibiotherapy, their adequacy rate was only 89%. The lowest de-escalation rate of 6% was reported by Rello et al in 113 patients, with changed therapy (not always de-escalation) reported in 43 patients (38%) (14). These authors had a rate of inadequate initial antibiotic therapy of 25%. The low de-escalation rate can be explained, at least in part, by the use of monotherapy in 47% of cases. In 2004, the same group evaluated the practice of de-escalation in patients with VAP (15). They analyzed 121 episodes of VAP and found a total rate of de-escalation of 31%, increasing to 38% when isolates were sensitive. They used monotherapy in 41% of the cases. However, their rate of de-escalation cannot be compared to ours because 12% of their patients were treated with amoxicillin/clavulanate, so that they could not de-escalate in these patients. Morel et al (13) recently assessed antibiotic strategies in 133 episodes of suspected infection requiring empirical antibiotic treatment. These authors reported a de-escalation rate of 45% in their overall patient population, but this included patients in whom antibiotics were stopped because they were later found to be not infected. They assessed the risks of re-escalation and of re-infection in their patients and reported a significant reduction in recurrent infection in patients in whom de-escalation was used. In another recent study, Shime et al (25) analyzed the rate of de-escalation in immunocompetent patients with bacteremia attributable to antibiotic-sensitive pathogens. This group studied a narrow population because they only included patients with positive blood cultures, a single causative microorganism covered by the empirical antibiotic therapy, and patients in whom it was possible to deescalate the empirical treatment. Despite this selective group, they recorded a rate of de-escalation of just 39%, whereas in our study if we limit our population to the same group of patients, then we achieve a de-escalation rate of 81%. The authors do not provide any reason for not performing de-escalation but, interestingly, show a trend toward reduced mortality and treatment failures when comparing a de-escalation group with a group without de-escalation, although the group without de-escalation included patients with unchanged and escalated treatments. In our outcomes analysis, there was higher mortality in groups III and IV (escalation and mixed strategies) compared to groups I and II (de-escalation and no change in treatment). It is not possible to say whether this simply reflects the severity of patient condition or if it is a direct consequence of the antibiotic strategy, but, importantly, it demonstrates that deescalation did not increase mortality in this cohort. Even with an initial broad-spectrum antibiotic therapy that included meropenem, piperacillin-tazobactam, or cephalosporins, all organisms in our study were not covered. Hence, antibiotic escalation was needed in 22 episodes (10%), and a combination of de-escalation and escalation was used in 24 episodes (11%). In the 77 episodes in which therapy was not changed, a reason was identified in most cases and the number of cases in which the chance to de-escalate may have been missed was actually small (4 cases [5%]). Although the rates of escalation and not covered organisms in our study may seem high, the population we studied (nosocomial severe sepsis and septic shock) is a more complex subgroup of patients than patients with VAP or community-acquired infections with more resistant organisms and potentially less microbiological information. Despite this difficulty, our results are within the middle of the range compared to previous studies. There was no difference in the severity of infection between the episodes in which de-escalation was performed and other episodes. The choice of empirical antibiotic therapy is influenced by local epidemiology with different degrees of bacterial resistance, justifying different initial strategies and different possibilities for de-escalation among units; therefore, rates of de-escalation can vary considerably. Belgium is a country with mixed resistant patterns, with more resistance than in Northern European countries but less resistance than in some southern countries (26). Rates of appropriate empirical antimicrobial treatment may be improved by using a wider-spectrum microbiological policy at the beginning of the septic episode and de-escalation could be improved by daily review, with the help of microbiology or infectious disease consultants, of the used regimen in the light of the microbiological results. Use of more invasive diagnostic strategies also has been associated with an increase in de-escalation rates (17). Our study may have the disadvantage of having been conducted in a single center, but interregional variation in microbial patterns is so large that interpretation of multiregional or international data may be difficult. Nevertheless, it is important to emphasize that our results may not apply everywhere. Although the retrospective design of our study made it more difficult to retrieve the reasons leading to each strategy, a prospective study design may have influenced therapeutic changes. CONCLUSION We demonstrated that in our academic environment with close collaboration with infectious disease specialists, de-escalation was possible in 43% of episodes of severe nosocomial sepsis. Antibiotic de-escalation is a strategy that is promoted for its potential advantages for the patient and for the hospital community by ensuring adequate coverage of causal infective agents but limiting selection pressure for multiresistant bacteria. Further study is needed to better-define 1408 Crit Care Med 2012 Vol. 40, No. 5

appropriate empirical antimicrobial therapies and the use of de-escalation strategies in various groups of ICU patients. REFERENCES 1. Osmon S, Warren D, Seiler SM, et al: The influence of infection on hospital mortality for patients requiring 48 h of intensive care. Chest 2003; 124:1021 1029 2. Kollef MH, Sherman G, Ward S, et al: Inadequate antimicrobial treatment of infections: A risk factor for hospital mortality among critically ill patients. Chest 1999; 115:462 474 3. Iregui M, Ward S, Sherman G, et al: Clinical importance of delays in the initiation of appropriate antibiotic treatment for ventilator-associated pneumonia. Chest 2002; 122:262 268 4. Garnacho-Montero J, Ortiz-Leyba C, Herrera- Melero I, et al: Mortality and morbidity attributable to inadequate empirical antimicrobial therapy in patients admitted to the ICU with sepsis: A matched cohort study. J Antimicrob Chemother 2008; 61:436 441 5. Gaieski DF, Mikkelsen ME, Band RA, et al: Impact of time to antibiotics on survival in patients with severe sepsis or septic shock in whom early goal-directed therapy was initiated in the emergency department. Crit Care Med 2010; 38:1045 1053 6. Lautenbach E, Patel JB, Bilker WB, et al: Extended-spectrum beta-lactamaseproducing Escherichia coli and Klebsiella pneumoniae: risk factors for infection and impact of resistance on outcomes. Clin Infect Dis 2001; 32:1162 1171 7. Yates RR: New intervention strategies for reducing antibiotic resistance. Chest 1999; 115:24S 27S 8. Gomes Silva BN, Andriolo RB, Atallah AN, et al: De-escalation of antimicrobial treatment for adults with sepsis, severe sepsis or septic shock. Cochrane Database Syst Rev 2010; CD007934 9. Alvarez-Lerma F, Alvarez B, Luque P, et al: Empiric broad-spectrum antibiotic therapy of nosocomial pneumonia in the intensive care unit: A prospective observational study. Crit Care 2006; 10:R78 10. Niederman MS: Impact of antibiotic resistance on clinical outcomes and the cost of care. Crit Care Med 2001; 29:N114 N120 11. Dellinger RP, Levy MM, Carlet JM, et al: Surviving Sepsis Campaign: International guidelines for management of severe sepsis and septic shock: 2008. Crit Care Med 2008; 36:296 327 12. Eachempati SR, Hydo LJ, Shou J, et al: Does de-escalation of antibiotic therapy for ventilator-associated pneumonia affect the likelihood of recurrent pneumonia or mortality in critically ill surgical patients? J Trauma 2009; 66:1343 1348 13. Morel J, Casoetto J, Jospe R, et al: Deescalation as part of a global strategy of empiric antibiotherapy management. A retrospective study in a medico-surgical intensive care unit. Crit Care 2010; 14:R225 14. Rello J, Gallego M, Mariscal D, et al: The value of routine microbial investigation in ventilator-associated pneumonia. Am J Respir Crit Care Med 1997; 156:196 200 15. Rello J, Vidaur L, Sandiumenge A, et al: De-escalation therapy in ventilatorassociated pneumonia. Crit Care Med 2004; 32:2183 2190 16. Leone M, Garcin F, Bouvenot J, et al: Ventilator-associated pneumonia: Breaking the vicious circle of antibiotic overuse. Crit Care Med 2007; 35:379 385 17. Giantsou E, Liratzopoulos N, Efraimidou E, et al: De-escalation therapy rates are significantly higher by bronchoalveolar lavage than by tracheal aspirate. Intensive Care Med 2007; 33:1533 1540 18. Canadian Critical Care Trials Group: A randomized trial of diagnostic techniques for ventilator-associated pneumonia. N Engl J Med 2006; 355:2619 2630 19. Levy MM, Fink MP, Marshall JC, et al: 2001 SCCM/ESICM/ACCP/ATS/SIS International Sepsis Definitions Conference. Crit Care Med 2003; 31:1250 1256 20. Vincent JL, Moreno R, Takala J, et al: The SOFA (Sepsis-related Organ Failure Assessment) score to describe organ dysfunction/failure. On behalf of the Working Group on Sepsis- Related Problems of the European Society of Intensive Care Medicine. Intensive Care Med 1996; 22:707 710 21. Calandra T, Cohen J, International Sepsis Forum Definition of Infection in the ICU Consensus Conference: The international sepsis forum consensus conference on definitions of infection in the intensive care unit. Crit Care Med 2005; 33:1538 1548 22. Garner JS, Jarvis WR, Emori TG, et al: CDC definitions for nosocomial infections, 1988. Am J Infect Control 1988; 16:128 140 23. Cooper GL, Hopkins CC: Rapid diagnosis of intravascular catheter-associated infection by direct Gram staining of catheter segments. N Engl J Med 1985; 312:1142 1147 24. Hoffken G, Niederman MS: Nosocomial pneumonia: the importance of a de-escalating strategy for antibiotic treatment of pneumonia in the ICU: Chest 2002; 122:2183 2196 25. Shime N, Satake S, Fujita N: De-escalation of antimicrobials in the treatment of bacteraemia due to antibiotic-sensitive pathogens in immunocompetent patients. Infection 2011; 39:319 325 26. European Center for Disease Prevention and Control: Antimicrobial resistance surveillance in Europe 2009. Annual Report of the European Antimicrobial Resistance Surveillance Network (EARS-Net). Available at: http://www.ecdc.europa.eu/en/publications/ Publications/1011_SUR_annual_EARS_ Net_2009.pdf. Accessed December 2, 2011 Crit Care Med 2012 Vol. 40, No. 5 1409