Inadequate Empiric Antibiotic Therapy among Canadian. Hospitalized Solid-Organ Transplant Patients: Incidence and Impact on Hospital Mortality

Transcription

1 Inadequate Empiric Antibiotic Therapy among Canadian Hospitalized Solid-Organ Transplant Patients: Incidence and Impact on Hospital Mortality by Bassem Hamandi A thesis submitted in conformity with the requirements for the degree of Master of Science Graduate Department of Pharmaceutical Sciences University of Toronto Copyright by Bassem Hamandi (2008)

2 Inadequate Empiric Antibiotic Therapy among Canadian Hospitalized Solid-Organ Transplant Patients: Incidence and Impact on Hospital Mortality Master of Science (2008) Bassem Hamandi Graduate Department of Pharmaceutical Sciences, University of Toronto ABSTRACT Background: The incidence of inadequate empiric antibiotic therapy (IET) and its clinical importance as a risk factor for hospital mortality in Canadian solid-organ transplant patients remains unknown. Methods: This retrospective cohort study evaluated all patients admitted to a transplant unit from May/2002-April/2004. Therapy was considered adequate when the organism cultured was found to be susceptible to an antibiotic administered within 24 hours of the index sample collection time. Univariate and multivariate regression analyses were conducted to determine associations between potential determinants, IET, and mortality. Results: IET was administered in 169/312 (54%) transplant patients. Regression analysis demonstrated that an increasing duration of IET (adjusted OR at 24h, 1.33; p < 0.001), ICUassociated infections (adjusted OR, 6.27; p < 0.001), prior antibiotic use (adjusted OR, 3.56; p = 0.004), and increasing APACHE-II scores (adjusted OR, 1.26; p < 0.001), were independent determinants of hospital mortality. Conclusions: IET is common and appears to be an important determinant of hospital mortality in the Canadian transplant population. ii

3 ACKNOWLEDGEMENTS I would like to thank my supervisor, Dr. Anne Holbrook for her support and guidance during my graduate studies. Her constructive criticism and comments from the initial conception to the end of this work were highly appreciated. Thank you to my committee members, Dr. James Brunton and Dr. Manny Papadimitropoulos for their expert opinions, advice, and invaluable feedback at our meetings. Dr. Michael Gardam s expertise and feedback were much appreciated. I would also like to thank Dr. Lehana Thabane for his statistical insight and advice. A special thanks to Dr. Atul Humar for his unique expertise in transplant infectious diseases. I am indebted to Mr. Gary Wong for his support, suggestions, and feedback during the early planning and throughout the implementation of the study. I would also like to thank my colleagues in the Pharmacy Department at The University Health Network for their help and support. Finally, I would like to thank my family and friends for their encouragement and support throughout my studies. iii

4 TABLE OF CONTENTS ABSTRACT...II ACKNOWLEDGEMENTS...III LIST OF ABBREVIATIONS...VI LIST OF TABLES...VII LIST OF FIGURES...VIII LIST OF APPENDICES...IX 1.0 INTRODUCTION STATEMENT OF THE PROBLEM Infection after Solid-Organ Transplantation Antibiotic Resistance Inadequate Empiric Antibiotic Therapy PURPOSE Objective Objective STATEMENT OF RESEARCH HYPOTHESIS RATIONALE FOR HYPOTHESIS REVIEW OF THE LITERATURE METHODS STUDY LOCATION AND POPULATION Inclusion Criteria Exclusion Criteria STUDY DESIGN MICROBIOLOGY DEFINITIONS Infection vs. Contamination vs. Colonization Infectious Episodes Healthcare vs. ICU vs. Community Associated Infections Primary Site of Infection Empiric Antibiotic Therapy Adequate vs. Inadequate Empiric Antibiotic Therapy Previous Antibiotic Therapy Previous Graft Rejection and Immunosuppressant Use Multi-Drug Resistance...31 iv

5 2.5 OUTCOMES STATISTICAL ANALYSIS RESULTS PATIENTS INADEQUATE EMPIRIC ANTIBIOTIC THERAPY CHARACTERISTICS RELATED TO HOSPITAL MORTALITY LOGISTIC REGRESSION ANALYSIS SECONDARY OUTCOMES DISCUSSION STUDY LIMITATIONS IMPLICATIONS OF INADEQUATE EMPIRIC THERAPY CONCLUSIONS REFERENCES PUBLICATIONS AND ABSTRACTS TO DATE APPENDICES...59 APPENDIX I LITERATURE REVIEW SEARCH STRATEGY...59 APPENDIX II RAW DATA...60 APPENDIX III MULTIVARIATE LOGISTIC REGRESSION MODELLING...66 v

6 LIST OF ABBREVIATIONS AET APACHE ATG BAL CDC CLSI CMV CNS CVC HAI ICU IET IL-2 MDR MIC MOT MRSA NNIS OR RR SOT UTI VAP VRE Adequate empiric antibiotic therapy Acute Physiology and Chronic Health Evaluation Anti-thymocyte globulin Bronchoalveolar lavage Centers for Disease Control Clinical and Laboratory Standards Institute Cytomegalovirus Coagulase-negative staphylococci Central venous catheter Healthcare-associated infections Intensive care unit Inadequate empiric antibiotic therapy Interleukin-2 Multi-drug resistant Minimum inhibitory concentration Multi-organ transplant Methicillin-resistant Staphylococcus aureus National Nosocomial Infections Surveillance Odds ratio Relative risk Solid-organ transplant Urinary tract infection Ventilator-associated pneumonia Vancomycin-resistant enterococci vi

7 LIST OF TABLES TABLE 1. A SUMMARY OF 22 NON-RANDOMIZED COMPARATIVE COHORT STUDIES ASSESSING THE ASSOCIATION BETWEEN IET AND MORTALITY...11 TABLE 2. CHARACTERISTICS OF PATIENTS RECEIVING ADEQUATE OR INADEQUATE EMPIRIC ANTIBIOTIC THERAPY...34 TABLE 3. CHARACTERISTICS OF TRUE INFECTIOUS EPISODES TREATED WITH ADEQUATE OR INADEQUATE EMPIRIC ANTIBIOTIC THERAPY...36 TABLE 4. ORGANISMS CULTURED FROM PATIENTS WITH INADEQUATELY TREATED INFECTIOUS EPISODES...38 TABLE 5. MULTI-DRUG RESISTANT ORGANISMS CULTURED AMONG PATIENTS RECEIVING ADEQUATE OR INADEQUATE EMPIRIC ANTIBIOTIC THERAPY...39 TABLE 6. INTRA-ABDOMINAL ORGANISMS CULTURED AMONG 17 PATIENTS RECEIVING ADEQUATE AND 30 PATIENTS RECEIVING INADEQUATE EMPIRIC ANTIBIOTIC THERAPY...39 TABLE 7. REASONS FOR ADMINISTRATION OF INADEQUATE EMPIRIC THERAPY...40 TABLE 8. PATIENT CHARACTERISTICS AMONG HOSPITAL SURVIVORS AND NONSURVIVORS...41 TABLE 9. CULTURE CHARACTERISTICS AMONG HOSPITAL SURVIVORS AND NONSURVIVORS...42 TABLE 10. MOST COMMONLY CULTURED ORGANISMS AND THEIR ASSOCIATED MORTALITY AMONG PATIENTS RECEIVING ADEQUATE OR INADEQUATE EMPIRIC ANTIBIOTIC THERAPY...42 TABLE 11. LOGISTIC REGRESSION ANALYSIS PREDICTING HOSPITAL MORTALITY...44 TABLE 12. OUTCOMES OF PATIENTS RECEIVING ADEQUATE VS. INADEQUATE EMPIRIC ANTIBIOTIC THERAPY...44 vii

8 LIST OF FIGURES FIGURE 1. INDIVIDUAL MULTIVARIABLE REGRESSION ANALYSIS OR (95% CI) OF 20 NON-RANDOMIZED COMPARATIVE STUDIES ILLUSTRATING THE ASSOCIATION BETWEEN INADEQUATE ANTIBIOTIC THERAPY AND MORTALITY AT END OF FOLLOW-UP. STUDIES CONDUCTED IN CRITICALLY ILL UNITS ARE SHOWN SEPARATELY NEAR THE BOTTOM...13 FIGURE 2. AN ILLUSTRATION DEPICTING THE TIMELINE FOR DETERMINING INADEQUATE EMPIRIC ANTIBIOTIC THERAPY. THERAPY WAS CONSIDERED ADEQUATE WHEN, FOR A GIVEN INFECTIOUS EPISODE, THE ORGANISM CULTURED WAS SUBSEQUENTLY FOUND TO BE SUSCEPTIBLE TO AN ANTIBIOTIC THAT WAS ADMINISTERED WITHIN 24 HOURS OF THE SAMPLE COLLECTION TIME...30 FIGURE 3. INCIDENCE AND RELATIVE RISK OF MORTALITY FOR INADEQUATE VERSUS ADEQUATE EMPIRIC ANTIBIOTIC THERAPY AMONG HOSPITALIZED SOLID-ORGAN TRANSPLANT RECIPIENTS...34 FIGURE 4. DELAY IN ADMINISTRATION OF ADEQUATE EMPIRIC ANTIBIOTIC THERAPY AND ASSOCIATED MORTALITY RATES...43 viii

9 LIST OF APPENDICES APPENDIX I LITERATURE REVIEW SEARCH STRATEGY...59 APPENDIX II RAW DATA...60 APPENDIX III MULTIVARIATE LOGISTIC REGRESSION MODELLING...66 ix

10 1 1.0 INTRODUCTION Healthcare associated infections (HAIs) are infections that patients acquire in a healthcare setting, during the course of receiving treatment for another condition (1). HAIs have imposed significant burdens on our healthcare system, leading to increased morbidity, mortality, and healthcare costs (2-4). As a result, the optimal management of HAIs has become an important healthcare concern. HAIs typically affect patients who are immunocompromised, either because of their age, underlying disease, or as a result of medical or surgical treatments (2). Along with an aging population, the growing use of medical and surgical interventions, including invasive devices and organ transplantation, have resulted in patients who are quite susceptible. The pathogens involved and the body sites of infection are often related to the treatments and devices used in intensive care units (ICUs). As a result, the highest infection rates are found among ICU patients, who experience approximately three times higher rates than patients found elsewhere in the hospital (2). Upon clinical suspicion of infection, antibiotic therapy is often started early and empirically, before pathogen identification, and before antibiotic susceptibilities are known. Deciding on the use of an antibiotic requires a balance between the benefits of more potent broad-spectrum antibiotics against their costs, including the potential for increased antibiotic resistance rates caused by their overuse. Once the decision to initiate antibiotic therapy has been made, it should ideally be directed at the most likely causative pathogens, taking into account local antibiotic susceptibility patterns. As antibiotic resistance rates continue to increase, it appears that the likelihood of administrating inadequate empiric antibiotic therapy (IET) also increases (5;6). Most clinicians consider therapy to be inadequate when the antibiotic agent initiated

11 2 demonstrates poor or no in vitro activity against the causative pathogen at the tissue site of infection (7). Several studies concentrating on the consequences of IET have been conducted in the ICU setting (5;6;8-11), however little information exists concerning outcomes of inadequate empiric therapy among hospitalized solid-organ transplant (SOT) patients. 1.1 Statement of the Problem Infection after Solid-Organ Transplantation Infection in SOT patients is an important determinant of clinical outcomes (12), consequently, the treatment of acquired bacterial infections with antibiotic therapy is recognized as being an essential component in improving outcomes (13). Kidney, liver, heart, lung, pancreas, and small bowel transplantation has become a therapeutic option for many end-stage organ diseases. Advances in surgical techniques, medical management, and immunosuppressants have enhanced both graft and patient survival rates and quality of life, however, infection continues to be a major cause of morbidity and mortality among SOT recipients (14-19). The use of newer and more potent immunosuppressants, particularly induction therapy with agents such as antithymocyte globulin (ATG) and interleukin-2 receptor antagonists, increases the level of immunosuppression and leads to increased susceptibility to infection in the early post-transplant period (20). The incidence of bacterial infections in SOT recipients ranges from 21 to 68% depending on the organ(s) transplanted, although the severity of the infection can vary among the different SOT groups (20). In liver transplant patients, bacterial infections of the liver, peritoneal cavity, biliary tree, bloodstream, and surgical wound are common (18). Lung and heart transplant recipients are susceptible to pulmonary infections and bacteremias, of which 50% are of pulmonary origin during the first post-transplant year (21). Infections among kidney transplant patients include wound, bloodstream, and more commonly, urinary tract infections (UTIs)

12 3 (15;17;19;22;23). Severe infections, such as bacteremia, continue to pose an increased risk of death in transplant patients, with 14-day mortality rates ranging from 11% in kidney, 24% in liver, 33% in heart recipients (14), and a 28-day mortality rate of 25% in lung recipients (21) Antibiotic Resistance The spread and rapid increase in antibiotic resistance rates has become a serious worldwide healthcare concern (13). In 1946, Sir Alexander Fleming suggested that, It seems likely that in the next few years a combination of antibiotics with different antibacterial spectra will furnish a cribrum therapeuticum from which fewer and fewer infecting bacteria will escape. Despite the advent of these potent antibiotics, the emergence and spread of antibiotic-resistant bacteria has become a tremendous burden on our healthcare system. In 1970, the Centers for Disease Control (CDC) established the National Nosocomial Infections Surveillance (NNIS) system, which receives monthly reports of nosocomial infections from a non-random sample of hospitals in the United States (24). With nearly 300 institutions currently reporting, data from the NNIS system shows that the nosocomial infection rate remains relatively unchanged. However, the gradual decline in the duration of inpatient stays has increased the rate of nosocomial infections per 1,000 patient days by 36%, from 7.2 in 1975 to 9.8 in 1995 (2). In 2002, nosocomial infections accounted for nearly 1.7 million infections, resulting in excess of 98,000 deaths in the United States (2;25). More than 70 percent of the bacteria that cause these infections are resistant to at least one antibiotic that is commonly used to treat them (3). Drug-resistant infections can be significantly more expensive to treat than non-resistant infections because they tend to result in a longer duration of hospitalization, increased rates of readmission, higher drug costs, more posthospital care, lost work days, and increased mortality (3). Hospital-treated infections are

13 4 estimated to cost $ million each year in Canada, however, resistant infections may add 2.8 times more than what a drug-susceptible infection adds to the direct cost of care (26). The excessive and inappropriate use of antibiotic agents continues to be one of the most important factors affecting antibiotic resistance patterns (13;27). In more and more cases, bacteria are becoming resistant to multiple drugs, leaving clinicians with few effective therapies, if any. Bacteria demonstrating multiple drug resistance were found to be responsible for 48% of bloodstream infections in a cohort of lung transplant recipients (21). Antibiotic resistance in hospitals may be increasing as a result of several factors, including the proliferation and prolongation of broad-spectrum antibiotic use, grouping of patients with higher disease acuity in segregated wards or units, and decreased staffing leading to increased person-to-person transmission (28). Several studies have shown an association between previous antibiotic use and the development of resistance in both gram-negative and gram-positive bacteria, especially in specialized settings such as ICUs (29-31;31-33). Conversely, colonization and infection with antibiotic resistant bacteria, increases the likelihood of administering IET (5), leading one to believe that a circular and confounding relationship may exist between antibiotic use, resistance and IET. Moreover, for some patients who receive IET, altering their antibiotic therapy later in the course of infection, based on subsequent culture susceptibility results, may yield little benefit with respect to in-hospital mortality, suggesting that adequate early treatment is vital (8). Overall, it seems that infections caused by antibiotic resistant bacteria are more difficult to treat and are associated with higher mortality rates and hospital costs (13).

14 Inadequate Empiric Antibiotic Therapy Pharmacological treatment with antibiotics demonstrating bacteriostatic or bactericidal activity against the causative pathogen remains the cornerstone of managing infectious diseases. The aim of in vitro antimicrobial susceptibility testing is to predict the in vivo success or failure of a panel of antibiotics at standard concentrations. The results of antimicrobial susceptibility testing combined with clinical information and experience, allows clinicians to select the most appropriate antibiotic. The safety and efficacy of antimicrobial agents in treating infections caused by specific pathogens must be established in well-controlled clinical trials or studies. The degree to which in vitro susceptibility results may or may not correlate with the in vivo efficacy of antimicrobials has been previously studied (34). Apart from the minimum inhibitory concentration (MIC) for a particular isolate, clinicians must consider patient, antimicrobial and pathogen-specific factors in determining how to best treat an infection. Achieving levels at or above the MIC by itself does not provide any information on persistent effects of antibacterial agents, such as the post-antibiotic effect. For moderate or severe infections, clinicians commonly initiate antibiotic therapy early and empirically, before the results of cultures and their respective antibiotic susceptibilities are known. Empiric therapy for patients with suspected or confirmed infections should be prescribed after considering patient symptoms, laboratory findings and the patient s past medical history, in the context of appropriate local and wider antibiotic resistance trends. As a consequence, there is a possibility then that situations may arise where the chosen antimicrobial agent demonstrates poor or no in vitro activity against the identified causative pathogen. This condition can be considered to be one of inadequate empiric antibiotic therapy. Thus, prescribing empiric therapy demands a balance between the benefits of using agents that have a broader spectrum of in vitro susceptibilities that may correspond to the isolated

15 6 pathogen s profile, against the current financial costs, potential side-effects, and future costs of developing resistance. Several factors have been shown to be problematic in the selection of adequate antimicrobial therapy. First, the complexity of the drug selection process may seem confusing, presenting a difficult challenge to many clinicians. Selecting adequate therapy involves early recognition of infection in a patient that may present with several confounding signs and symptoms, identification of the causative pathogen, and prescribing of an antimicrobial regimen that is efficacious, cost-effective and poses minimal toxicity. In addition, the types of pathogens, along with antibiotic resistance patterns, have been shown to vary among different hospitals and even within in-hospital units, suggesting the need to develop unit-specific reporting systems (35;36). But until the susceptibility profile of the pathogen is known, antibiotic selection occurs through an empiric process, based on local sensitivity patterns and the patient s clinical presentation. Infections caused by antibiotic resistant bacteria can lead to the problem of IET (37;38). To further this dilemma, some organisms have become resistant to a point where few or no treatment alternatives exist (39). The escalating concern with regard to antimicrobial resistance in the hospital setting has led several investigators to examine how this and other factors have influenced the prescribing of inadequate treatment. However, even after controlling for other contributing risk factors, it may still be difficult to determine whether delayed or inadequate therapy or antibiotic resistance has led to poor outcomes. To complicate issues further, humans who are able to produce an innate immune response, may rid themselves of the infection and thus seem to respond to inadequate therapy or even no treatment at all. Unfortunately, few studies to date have addressed the issue of IET use in SOT recipients and its relationship to clinical outcomes.

16 7 1.2 Purpose To date, little work has been done with respect to the incidence and clinical importance of IET as a risk factor for hospital mortality in SOT recipients. The purpose of this study was two-fold. Firstly, to determine the scale of the problem of IET among a cohort of Canadian hospitalized SOT recipients. Secondly, this study aimed to examine the extent to which IET contributes to inhospital mortality among SOT patients. Determining the incidence and impact of IET in this population may help in the decision-making process of prescribing empiric antibiotic therapy, and perhaps justify the use of broad-spectrum empiric antibiotics in this population Objective 1 Our first objective was to determine the incidence of inadequate empiric antibiotic therapy among Canadian hospitalized SOT patients Objective 2 Our second objective was to determine whether IET and the duration of IET are clinically important risk factors for in-hospital mortality in SOT patients. 1.3 Statement of Research Hypothesis Null Hypothesis: There is no statistically significant difference (p < 0.05) in hospital mortality between SOT recipients receiving adequate versus inadequate empiric antibiotic.

17 8 1.4 Rationale for Hypothesis Although the view that early adequate therapy should improve survival seems plausible, few studies exist to support this assumption outside the ICU setting. Many of these studies specifically evaluated bacteremic ICU patients who had been prescribed IET, a possible confounder given that bloodstream infections have been found to be an independent predictor for mortality (5). The ICU setting includes a variety of pressures that are not as prevalent in other settings, influencing the emergence and spread of antibiotic resistance. One of these pressures includes patients with prolonged hospitalization who may harbour these organisms for the duration of their stay. The presence of invasive devices, such as urinary catheters and endotracheal tubes, along with prolonged mechanical ventilation may also promote infections with resistant bacteria (24;40). Severity of illness may also be an important confounder among ICU patients, and as such, use of broad-spectrum antibiotics early in the course of infection may have a greater impact in this population. While SOT recipients are immunocompromised and may share some of the qualities of ICU patients, in general, they are not as acutely ill, suggesting that IET may not contribute to an excess risk for in-hospital mortality among this cohort. 1.5 Review of the Literature We performed a literature search of the National Library of Medicine using the OVID MEDLINE database to find original English-language articles published from 1950 to April The search strategy is outlined in Appendix I. The aim was to find publications that included IET as a primary independent variable of interest, and mortality as a dependent variable. We used terms related to antibiotic use or infection in addition to transplantation, critical illness, and hospitalization to define our population of interest. We limited the search

18 9 with keywords for adequate or inadequate therapy and outcomes related to morbidity or mortality. Additional articles referenced in publications found in the MEDLINE search were also included. IET was considered to be a primary exposure of interest if it was explicitly stated as such in the study objectives and it was forced into a multivariate statistical analysis. We only included publications that accounted for the administration of empiric therapy, as defined by the receipt of the final antibiotic sensitivity profile of the organism isolated from the index culture. No studies that met the search criteria with SOT recipients as their primary population of interest were found. We did find two studies that described the administration of discordant initial and inactive antibiotic therapy among SOT patients (21;41). In a prospective cohort of 56 bacteremic lung transplant recipients, discordant therapy (defined as therapy that was inactive in vitro for the first two days following the index blood culture) occurred in 12/56 (21%) of the patients (21). Six of the 12 patients receiving discordant died within 28-days, compared to a mortality rate of 8/44 (18%) among patients receiving concordant therapy (21). Another prospective examination of 66 SOT recipients who developed septic shock, revealed that empiric therapy was inactive in vitro in 14/66 (21%) of the cases, with a mortality rate of 64% versus 52% for those receiving active empiric therapy (41). Neither study included inadequate empiric therapy as a primary exposure of interest nor did they include the term in the final multivariable analysis. Table 1 summarizes 22 non-randomized comparative cohort studies that did assess the association between IET and mortality (5;6;10;11;38;42-58). Two of these studies did not report the 95% OR obtained from the multivariate analysis (53;58). The reported incidence of IET among these studies ranges from 10-80%. Many studies assessing the impact of inadequate

19 10 empiric antibiotic usage have focused on the ICU population and have yielded more or less similar conclusions about the importance of adequate therapy in bloodstream infections, including sepsis and septic shock, and ventilator-associated pneumonia (VAP). The majority of these studies have demonstrated that hospital mortality for critically ill patients receiving inadequate antibiotic treatment is significantly greater than those receiving adequate therapy (5;6;10;11;42;56). However, one study did not find adequate antibiotic treatment to be associated with a significant mortality benefit in critically ill patients (58). We extracted the reported multivariable regression analysis odds ratio along with the associated 95% confidence interval from each individual study. Figure 1 depicts a list of the individual studies multivariable regression analysis OR (95% CI), illustrating the association between inadequate antibiotic therapy and mortality at end of follow-up. Compared to non-critically ill settings, studies conducted in critically ill patients had a stronger trend towards favouring the use of adequate therapy.

20 Table 1. A summary of 22 non-randomized comparative cohort studies assessing the association between IET and mortality. 11 Study Design 1 Patient population Byl et al. (38), 1999 Clec h et al. (42), 2004 Fraser et al. (43), 2006 Harbarth et al. (44), 2003 P P P P 417 episodes 361 patients Tertiary care hospital 196 episodes 142 patients Six ICUs 895 patients 3 tertiary care hospitals 904 patients 108 hospitals Infection/ pathogens Inadequate therapy definition Inadequate therapy incidence Bacteremia 24 h 159/428 (37%) VAP 24 h 109/196 (56%) All bacterial infections Severe sepsis or septic shock 24 h 319/895 (36%) 24 h 211/904 (23%) Mortality definition Attributable in-hospital mortality In-hospital mortality 30-day mortality 28-day mortality Mortality (adequate vs. inadequate therapy) 33/258 (13%) vs. 24/159 (15%) 30/63 (48%) vs. 41/79 (52%) 68/576 (12%) vs. 64/319 (20%) 168/693 (24%) vs. 82/211 (39%) Multivariable regression analysis OR (95% CI) 2.22 ( ) 7.24 ( ) 1.58 ( ) 1.8 ( ) Hyle et al. (45), 2005 Ibrahim et al. (6), 2000 Iregui et al. (11), 2002 Kang et al. (46), 2005 Kim et al. (47), 2006 Kollef et al. (5), 1999 Leibovici et al. (10), 1998 Lodise et al. (48), 2003 R P P R R P P R 187 patients 2 Tertiary care hospitals 492 patients ICU 107 patients ICU 286 patients Tertiary care hospital 238 patients Tertiary care hospital 655 patients ICU 3413 patients Tertiary care hospital 167 patients Level 1 trauma centre ESBL E. coli & Klebsiella species 48 h 112/187 (60%) Bacteremia >72 h 147/492 (30%) In-hospital mortality In-hospital mortality VAP 24 h 33/107 (31%) Attributable in-hospital mortality Gram-negative bacteremia S. aureus bacteremia All bacterial infections 24 h 151/286 (53%) 48 h 117/238 (49%) >72 h 169/655 (26%) Bacteremia 48 h 1255/3440 (36%) S. aureus bacteremia 30-day mortality 12-week attributable mortality Attributable in-hospital mortality In-hospital mortality 48 h 48/167 (29%) Attributable mortality 8/75 (11%) vs. 24/112 (21%) 98/345 (28%) vs. 91/147 (62%) 8/74 (11%) vs. 13/33 (39%) 37/135 (27%) vs. 58/151 (38%) 34/121 (28%) vs. 45/117 (39%) 59/486 (12%) vs. 88/169 (52%) 436/2158 (20%) vs. 432/1255 (34%) 23/119 (19%) vs. 16/48 (33%) 0.69 ( ) 6.86 ( ) 7.68 ( ) 3.64 ( ) 1.39 ( ) 4.26 ( ) 1.6 ( ) 3.8 ( )

21 12 Study Design 1 Patient population Lujan et al. (49), 2004 Micek et al. (50), 2005 Osih et al. (51), 2007 Paterson et al. (52), 2003 Roghmann (53), 2000 Scarsi et al. (54), 2006 Schramm et al. (55), 2006 Valles et al. (56), 2003 Vidal et al. (57), 1996 Zaragoza et al. (58), 2003 P R R P R R R P P P 100 patients Tertiary care hospital 305 patients Tertiary care hospital 167 episodes 159 patients Tertiary care hospital 85 patients 12 hospitals 132 episodes 125 patients Tertiary care hospital 884 patients Tertiary care hospital 549 patients Tertiary care hospital 339 patients 30 ICUs 189 episodes 182 patients Tertiary care hospital 166 patients ICU Infection/ pathogens S. pneumoniae bacteremia P. aeruginosa bacteremia P. aeruginosa bacteremia ESBL K. pneumoniae S. aureus bacteremia Gram-negative bacteremia MRSA sterilesite infections Communityacquired bacteremia P. aeruginosa bacteremia Inadequate therapy definition Inadequate therapy incidence Mortality definition 24 h 10/100 (10%) 28-day mortality >72 h 75/305 (25%) In-hospital mortality 24 h 68/167 (41%) In-hospital mortality >72 h 11/82 (13%) 14-day mortality 48 h 105/132 (80%) 24 h 125/884 (14%) 24 h 380/549 (69%) 30-day mortality In-hospital mortality In-hospital mortality 24 h 49/339 (14%) In-hospital mortality >72 h 19/189 (10%) All-cause mortality Bacteremia >72 h 39/166 (23%) Attributable mortality Mortality (adequate vs. inadequate therapy) 13/90 (14%) vs. 5/10 (50%) 41/230 (18%) vs. 23/75 (31%) 35/99 (35%) vs. 26/68 (38%) 10/71 (14%) vs. 7/11 (64%) 28/102 (27%) vs. 5/23 (22%) 122/759 (16%) vs. 17/125 (14%) 28/169 (17%) vs. 99/380 (26%) 107/290 (37%) vs. 34/49 (69%) 24/170 (14%) vs. 10/19 (53%) 29/127 (23%) vs. 12/39 (31%) Multivariable regression analysis OR (95% CI) 5.72 ( ) 2.04 ( ) 0.93 ( ) ( ) Not reported 0.61 ( ) 1.92 ( ) 4.11 ( ) 6.53 ( ) Not reported 1 P = Prospective; R = Retrospective

22 Figure 1. Individual multivariable regression analysis OR (95% CI) of 20 non-randomized comparative studies illustrating the association between inadequate antibiotic therapy and mortality at end of follow-up. Studies conducted in critically ill units are shown separately near the bottom. 13

23 14 One of the earliest studies assessing the relationship between inadequate antibiotic treatment and hospital mortality, evaluated a prospective cohort of 2,000 patients admitted over an 8-month period to the medical or surgical ICU of a large urban teaching hospital in St. Louis, Missouri. Inadequate antimicrobial treatment was defined as the microbiological documentation of a pathogen causing infection, which was not effectively treated at the time of its identification. This included both the absence of antimicrobial agents and the administration of an agent to which the pathogen was resistant. Comparisons were made between patients receiving inadequate and adequate therapy and hospital survivors to non-survivors. Multiple logistic regression analysis was used to evaluate the relationship between the dependent variable of hospital mortality and the independent variable of inadequate treatment and to identify independent risk factors for the administration of inadequate treatment. Of the 655 patients with a nosocomial or community-acquired infection, 169 (25.8%) were found to have received inadequate antibiotic treatment. The infection-related mortality rate for infected patients receiving inadequate therapy (42.0%) was significantly greater than patients receiving adequate antibiotic treatment (17.7%) (RR, 2.37; 95% CI, 1.83 to 3.08; p < 0.001). In addition, a logistic regression model demonstrated that inadequate antibiotic treatment was the most important independent determinant of hospital mortality (adjusted OR, 4.27; 95% CI, 3.35 to 5.44; p < 0.001). The incidence of inadequate antimicrobial treatment was most common among patients with nosocomial infections, which developed after treatment of a community-acquired infection (45.2%), followed by patients with nosocomial infections alone (34.3%) and patients with community-acquired infections alone (17.1%) (p < 0.001). Among patients with nosocomial infections, inadequate therapy occurred most commonly as a result of Gram-negative bacteria that were resistant to third-generation cephalosporins. Inadequate treatment for methicillinresistant Staphylococcus aureus (MRSA), Candida species and vancomycin-resistant enterococci

24 15 (VRE) were also commonly found among nosocomial infections. Multiple logistic regression analysis revealed that the prior administration of antibiotics (adjusted OR, 3.39; 95% CI, 2.88 to 4.23; p < 0.001), presence of a bloodstream infection (adjusted OR, 1.88; 95% CI, 1.52 to 2.32; p = 0.003), increasing Acute Physiology and Chronic Health Evaluation (APACHE) II scores in 1- point increments (adjusted OR, 1.04; 95% CI, 1.03 to 1.05; p = 0.002), and decreasing patient age (adjusted OR, 1.01; 95% CI, 1.01 to 1.02; p = 0.012) were independently associated with the administration of inadequate antibiotic treatment. This study established an association between the prescribing of inadequate therapy and hospital mortality, and demonstrated that previous antibiotic use may be an important risk factor for inadequate therapy among ICU patients. Continuing the work of Kollef et al., Ibrahim et al. prospectively evaluated the relationship between the adequacy of antimicrobial treatment for bloodstream infections and the primary outcome of hospital mortality among a cohort of patients at the same university-affiliated urban teaching hospital. All patients admitted to the medical or surgical ICU were eligible for enrolment. Inadequate antimicrobial treatment was defined as the microbiological documentation of a pathogen causing infection, both bacterial and fungal, which was not effectively treated at the time the pathogen and its susceptibility profile were known. The primary analysis compared hospital survivors to non-survivors. Multiple logistic regression analysis was used to evaluate the relationship between the dependent variable of hospital mortality and the independent variable of inadequate antimicrobial treatment and to identify independent risk factors for the administration of inadequate treatment. Over a two-year period, 4913 critically ill patients were admitted, of whom 492 (10.0%) were found to have a bloodstream infection. Of the 492 patients, 147 (29.9%) received inadequate treatment. Furthermore, the hospital mortality for these patients was significantly greater than those receiving adequate therapy (61.9% vs. 28.4%; RR, 2.18; 95% CI,

25 to 2.69; p < 0.001). Multiple logistic regression analysis identified the administration of inadequate antibiotic treatment as an independent risk factor hospital mortality (adjusted OR, 6.86; 95% CI, 5.09 to 9.24; p < 0.001). The most commonly identified bloodstream pathogens along with their rates of inadequate antimicrobial treatment included: VRE (n = 17; 100%), Candida species (n=41; 95.1%), MRSA (n = 46; 32.6%), coagulase-negative staphylococci (CNS) (n = 96; 21.9%), and Pseudomonas aeruginosa (n = 22; 10.0%). A statistically significant correlation was found between the rates of inadequate antimicrobial treatment for individual micro-organisms and their associated rates of hospital mortality (Spearman s correlation coefficient ; p=0.006). However, some organisms, such as Escherichia coli and Klebsiella species, were found to be associated with relatively low rates of inadequate antimicrobial therapy, though their associated hospital mortality rates were greater than 30%. Multiple logistic regression analysis also demonstrated that the following criteria were independently associated with the administration of inadequate antimicrobial treatment: Bloodstream infection attributable to Candida species (adjusted OR, 51.86; 95% CI, to ; p < 0.001); Prior administration of antibiotics during the current hospital stay (adjusted OR, 2.08; 95% CI, 1.58 to 2.74; p = 0.008); Decreasing serum albumin concentrations (adjusted OR, 1.37; 95% CI, 1.21 to 1.56; p = 0.014); Increasing central catheter duration (adjusted OR, 1.03; 95% CI, 1.02 to 1.04; p = 0.008). The study demonstrated that ICU patients with bloodstream infections receiving inadequate antimicrobial treatment were at an increased risk of death compared to patients receiving adequate treatment. The authors recommended initial empiric therapy with vancomycin for MRSA and CNS, along with combination therapy for the treatment of P. aeruginosa.

26 17 Most studies with bacteremia seem to support the importance of adequate empiric therapy, however, there are a few studies that do not come to this conclusion (51;54;58). These studies may include certain groups of organisms that may be less virulent and thus it may be more problematic in determining their role in morbidity or mortality. One such cohort study in Spain was conducted among 166 prospectively followed patients with bacteremia, 39 (23.5%) of which received inadequate antibiotic treatment, while 127 (76.5%) received adequate treatment (58). Bacteremia was determined to be nosocomial in nature in 92.3% of the inadequately treated cohort, and 79.5% of adequately treated group. The occurrence of coagulase-negative staphylococci isolates (OR, 2.62; 95% CI, 1.10 to 6.21; p = 0.015), and absence of a respiratory or abdominal source of infection (OR, 0.35; 95% CI, 0.12 to 0.97; p = 0.04) was greater in the cohort with inadequate treatment than in the group with adequate treatment. Neither crude mortality rates (56.4% vs. 50.3%; p = 0.512) nor bacteremia-related mortality rates (30.8% vs. 22.8%; p = 0.315) were significantly different among the inadequately and adequately treated groups, respectively. Multivariate analysis did not reveal inadequate treatment to be an independent predictor of mortality. The authors suggested that this was likely a result of microbiological factors and clinical features, such as the types of micro-organisms isolated and the sources of the bacteremia. Cultures from patients in the inadequately treated arm were more likely to have grown less virulent isolates and to have been sampled from sources of infections with typically better prognoses. As a result, this tended to dilute the effects on mortality, and produce a statistically non-significant difference. In an application to non-critically ill patients, Leibovici et al. set out to determine whether empiric antibiotic treatment matching the in vitro susceptibility of the pathogen, which they termed as being appropriate treatment, improved survival in hospitalized patients with

27 18 bloodstream infections. This prospective and observational cohort study examined patients with bloodstream infections identified between 1988 and 1994 in an urban hospital in Israel. Empirical antibiotic treatment was defined as appropriate if it was started within two days of the first positive blood culture, and the infecting micro-organism was subsequently found to be susceptible to an intravenously administered drug. However, the authors decided that treatment of a pseudomonal infection with just an aminoglycoside would be considered inappropriate. This study analyzed the benefit presented by appropriate empiric treatment in stratified subgroups of patients defined by a set of other mortality risk factors. Logistic regression analysis was used to measure the independent contribution of inappropriate treatment to hospital mortality. Of the 3415 patients identified with bloodstream infections, 2158 (63.2%) were given appropriate empiric antibiotic treatment, of which 436 (20.2%) died, compared with the death of 432 (34.4%) of 1255 patients who were given inappropriate treatment (OR, 2.1; 95% CI, 1.8 to 2.4; p = ). The median duration of hospital stay for survivors was 9 days when given appropriate treatment and 11 days when given inappropriate treatment (p = ). The greatest relative reduction in the mortality rate associated with appropriate treatment versus inappropriate treatment in patients was seen most commonly in: Pediatric patients (4% vs. 17%; OR, 5.1; 95% CI, 2.4 to 10.7); Intra-abdominal infections (12% vs. 34%; OR, 3.8; 95% CI, 2.0 to 7.1); Skin and soft tissue infections (23% vs. 49%; OR, 3.1; 95% CI, 1.8 to 5.6); Infections caused by Klebsiella pneumoniae (17% vs. 39%; OR, 3.0; 95% CI, 1.7 to 5.1), and Streptococcus pneumoniae (22% vs. 42%; OR, 2.6; 95% CI, 1.1 to 5.9). Multivariable logistic regression analysis revealed that inappropriate empiric treatment was associated with a significant risk for in-hospital mortality (adjusted OR, 1.6; 95% CI, 1.3 to 1.9) independent of other risk factors. The investigators concluded that in this relatively large cohort of hospitalized patients with

28 19 bloodstream infections, inappropriate empiric treatment was associated with an increased risk of death, regardless of concomitant risk factors for mortality. Infections resulting from certain groups of organisms, such as antibiotic-resistant gram-negative bacilli, have become more of a concern in recent years, as patients infected by these relatively virulent isolates may be at a higher risk of receiving inadequate therapy (45;46;52;55). To evaluate the effect of inappropriate initial antimicrobial therapy on mortality, Kang et al. retrospectively reviewed 286 hospitalized patients in Seoul, South Korea. They identified and included patients with nosocomial antibiotic-resistant gram-negative bacteremia: 61 patients with E. coli, 65 with K. pneumoniae, 74 with P. aeruginosa, and 86 with Enterobacter species. Initial antibiotic therapy was considered to have been appropriate if a patient received at least one agent within 24 hours of blood culture collection to which the causative pathogens were susceptible. Only the first bacteremic episode per patient was included. Antibiotic resistance was defined as in vitro resistance to either cefotaxime or ceftazidime, except for P. aeruginosa, which was required to be resistant to either piperacillin, ciprofloxacin, ceftazidime, or imipenem. High-risk sources of bacteremia were defined as the lung, peritoneum, or an unknown source. The main outcome measure was 30-day mortality. Of the 286 patients, 135 (47.2%) received appropriate initial empirical antimicrobial therapy, with the remaining 151 (52.8%) patients receiving inappropriate therapy. The inadequately treated group had a significantly greater mortality rate compared to the adequately treated cohort (38.4% vs. 27.4%; p=0.049). Multivariate analysis demonstrated that septic shock, a high-risk source of bacteremia, P. aeruginosa infection, and an increasing APACHE II score, were independent risk factors for mortality. In a subgroup analysis of patients with a high-risk source of bacteremia (n = 132), inappropriate initial antimicrobial therapy was independently associated with decreased survival (adjusted OR, 3.64; 95% CI, 1.13

29 20 to 11.72; p = 0.030). The results of this study suggest that inappropriate initial antimicrobial therapy is associated with adverse outcomes in patients diagnosed with antibiotic-resistant gramnegative bacteremia, specifically in those that are severely ill, have a high-risk source of bacteremia, or have isolates that are especially virulent. Some evidence suggests that when adequate empiric antibiotic therapy (AET) is initiated early in the course of the infection and prior to the availability of susceptibility results, the associated mortality rates are significantly lower (8;11;48;55;59). Iregui et al. fixed their efforts on the clinical importance of delayed initial appropriate antibiotic treatment in critically ill patients with clinically diagnosed VAP. Their goals were to identify the occurrence of initially delayed appropriate antibiotic treatment for VAP, and to determine its effects on patient outcomes. They prospectively observed a cohort of patients requiring mechanical ventilation and admitted to the medical ICU of a large urban teaching hospital in St. Louis, Missouri. The primary outcome compared hospital mortality among those receiving initially delayed appropriate antibiotic treatment to all other patients in the cohort. Adequate initial treatment was defined as an antibiotic with in vitro susceptibility to the pathogen isolated in respiratory samples. Delayed therapy was defined as a time period of >24 hours between the time of VAP diagnosis, until the time that appropriate antibiotic therapy was administered. Thirty-three of 107 (30.8%) patients received appropriate antibiotic therapy that was delayed >24 hours after the clinical diagnosis of VAP. The most common cause of inadequate initial treatment was a delay in writing the medical orders (n = 25; 75.8%). The presence of a resistant micro-organism (n = 6) accounted for 18.2% of cases. Among these patients, the mean time delay between VAP diagnosis and the administration of an appropriate antibiotic was 28.6 ± 5.8 hours, compared to 12.5 ± 4.2 hours for all other patients (p < 0.001). Patients with initially delayed treatment had a significantly

30 21 greater hospital mortality compared to the other patients in the cohort (69.7% vs. 28.4%; p < 0.01). It is important to note that diagnostic delays were not counted, and that the authors utilized a clinical VAP diagnosis as opposed to using bronchoscopically obtained cultures. It was suggested that clinicians avoid delaying the administration of appropriate antibiotic therapy to patients with VAP to minimize their mortality risk. The extent to which the timely use of adequate therapy impacts clinical outcomes in hospitalized patients may depend on the causative pathogens and populations studied, but may also depend on the actual time delay itself. In a study of episodes of nosocomial S. aureus bacteremia, Lodise et al. attempted to determine the effect of delayed therapy on morbidity and mortality in a retrospective cohort of 167 hospitalized patients at a trauma centre in Detroit, Michigan. If a patient had more than one episode of S. aureus bacteremia during a hospitalization, only the first episode was included. Classification and regression tree analysis was utilized to select the time interval (from the time the culture result was obtained until administration of adequate therapy) that classified patients as having either a low-risk or high-risk of infection-related mortality. The time breakpoint between delayed and early treatment was determined to be hours. Accordingly, 48 (28.7%) patients did not receive appropriate treatment prior to the breakpoint time and were deemed to have received delayed treatment. The remaining 119 (71.3%) patients did receive appropriate treatment within hours, and were classified into the early treatment group. A comparison of the infection-related mortality rates between the two groups revealed a 1.7-fold increase among patients receiving delayed therapy versus early treatment (33.3% vs. 19.3%; p = 0.05). A multivariate analysis revealed that delayed treatment was an independent predictor of infection-related mortality (adjusted OR, 3.8; 95% CI, 1.3 to 11.0; p = 0.01) and was associated with a longer hospital stay than early treatment (20.2 vs days; p = 0.05). For

31 22 patients with an APACHE II score >15.5 and a high-risk source of infection (non-iv catheter related), mortality was 86.7% in the delayed treatment group compared with 44.7% in the early treatment group (p = 0.006). However, among patients with an APACHE II score <15.5, the mortality rate was not significantly different, indicating that delayed treatment had more of an adverse effect on those that were more severely ill. The results of this study point towards a delay in adequate therapy in the realm of hours as being an important factor in clinical outcomes of patients hospitalized with S. aureus nosocomial bacteremias, particularly those who are severely ill or have a high-risk source of infection. To date, studies determining the impact of initially delayed adequate therapy on the clinical outcomes of infections, have utilized observational cohorts in their design. Randomized controlled trials would provide a methodological advantage in reducing bias and the effect of confounders, however, this design is neither practical nor ethical. The propensity score is an analysis that utilizes the probability of exposure to a specific treatment conditional on observed variables and is increasingly being used in observational studies. This analysis attempts to compensate for selection bias by creating strata in which subjects are matched on the propensity score, balancing the covariables between patients receiving adequate or inadequate therapy. Kim et al. studied 238 hospitalized patients with S. aureus bacteremia who received either inappropriate or appropriate empirical therapy, and compared them by using two risk stratification models. The first model used a cohort study with a propensity score to adjust for confounding by treatment allocation, and the second, used a propensity-matched case-control study. Inappropriate therapy was modeled on the basis of patient characteristics, and included in the multivariate model to adjust for confounding. For the case-matching analysis, patients with inadequate empiric treatment (cases) were matched to those with adequate empiric treatment

32 23 (controls) on the basis of the propensity score. The cohort study revealed that the bacteremiarelated mortality rate was 38.4% (45/117) among those inappropriately treated versus 28.1% (34/121) for those appropriately treated (adjusted OR 1.60; 95% CI, ; p = 0.09). Conducting a multivariate analysis to adjust for independent predictors for mortality and the propensity score, demonstrated that inappropriate empiric therapy was not associated with mortality (adjusted OR, 1.39; 95% CI, ). The matched case-control study analyzed 50 pairs, with mortality rates of 32% (16/50) for the case group and 28% (14/50) for the control group (OR, 1.15; 95% CI, ; p = 0.85). Once more, these authors indicated that this may have been a result of microbiological factors, specifically, the fact that most gram-positive pathogens tend not to be as virulent as gram-negative ones. In addition, they point to the methodology of the study and the propensity score, including the possibility of it being underpowered and not being able to control for detection bias. In summary, these results suggest an association between the time to administration of IET and mortality, promoting the principle of utilizing broad-spectrum antibiotics early in the course of infection. However, the patients were not homogeneous, with considerable variation with respect to their sites of infection, organism virulence, and severity of illness. With a few exceptions, the evidence favours the early use of adequate empiric therapy in relatively ill patients who are at risk of an infection with a virulent organism. The Infectious Diseases Society of America and the Society for Healthcare Epidemiology of America has developed guidelines and a set of recommendations for enhancing antimicrobial stewardship. With respect to combination therapy, they state that there are insufficient data to recommend the routine use of combination therapy to prevent the emergence of resistance (Grade C-II) (60). However, combination therapy does have a role in certain clinical contexts, including increasing the breadth of empiric coverage and the