Initiation of Inappropriate Antimicrobial Therapy Results in a Fivefold Reduction of Survival in Human Septic Shock FREE TO VIEW

With a mortality risk of 40 to 70%, septic shock is the most common cause of death in the modern ICU.^1,2 Research efforts in this area over the last 2 decades have focused primarily on the study of resuscitative modalities and experimental anti-inflammatory agents. However, several studies^3,4 have indicated that the effective use of antimicrobial agents is central to the optimization of outcome in life-threatening infections in critically ill patients. We have demonstrated in a large retrospective cohort study³ that delay in the initiation of effective antimicrobial therapy following the onset of hypotension is the critical determinant of outcome in septic shock. The initiation of treatment with inappropriate antimicrobial agents (in relation to the subsequently demonstrated sensitivity of the pathogen) as the initial empiric therapy may be the single most common cause of prolonged delays in the introduction of effective therapy.

A large body of literature^4–27 has addressed the issue of inappropriate antimicrobial therapy in patients with bacteremia and other serious infections. Equivalent data on septic shock are very limited.^18,26 Accordingly, a retrospective multicenter study was undertaken to examine the relationship between the appropriateness of initial empiric antimicrobial therapy and survival in patients with septic shock and major clinical subgroups of patients with septic shock.

A retrospective review of adult patients (≥ 18 years of age) who had received a diagnosis of septic shock was performed. A waived consent protocol was approved by the health ethics boards of the University of Manitoba and at each individual participating center. Consecutive adult patients with septic shock from 22 medical institutions from Canada (17 sites), the United States (4 sites), and Saudi Arabia (1 site) were collected from discrete periods between 1996 and 2005. Each institution contributed a minimum of 100 cases. Potential cases were initially identified by using a combination of internal ICU registries/databases and/or International Classification of Diseases, Ninth Revision, Clinical Modification, coding strategies. Each potential case was screened to determine whether the case met the specific criteria for septic shock, as described by the 1991 Society of Critical Care Medicine/American College of Chest Physicians Consensus statement on sepsis definitions.²⁸ As per that definition, case patients were required to have documented or suspected infection, persistent hypotension requiring therapy with pressors, and two of the following four elements: (1) a heart rate of > 90 beats/min; (2) a respiratory rate of > 20 breaths/min or Pco₂ of < 32 mm Hg; (3) a core temperature of < 36°C or > 38°C; and (4) a WBC count of < 4,000/μL, >12,000/μL, or > 10% immature (bands) forms.

Clinical infection definitions were adapted from previous recommendations or studies.^3,29,30 In order to qualify as potential pathogens causing shock, isolates from both the local site and/or blood cultures were required to have been obtained within 48 h of the onset of shock. The use of appropriate antimicrobial therapy (ie, with in vitro activity appropriate to isolated pathogenic organisms or, if a pathogenic organism was not isolated, appropriate for the underlying clinical syndrome) was determined for all cases. Questionable cases or data elements were reviewed by the principal investigator for adjudication.

Predetermined rules were created to assess the appropriateness of antimicrobial therapy (see Appendices 1 and 2). For culture-negative septic shock, appropriate therapy was deemed to be initiated when antimicrobial agents consistent with broadly accepted norms for empiric management of the typical pathogens for the clinical syndrome (in the context of host immune/health status, environmental factors, and local flora) were administered. At each participating institution, infectious diseases physicians/microbiologists were consulted to establish the requirement for appropriate empiric coverage in view of local community and nosocomial flora during the time period covered by data collection. Otherwise, appropriate empiric therapy of culture-negative infections leading to septic shock was defined by the recommendations enumerated in Table 1 of the “Clinical Approach to Initial Choice of Antimicrobial Therapy” in the Sanford Guide to Antimicrobial Therapy 2004 (34th ed),³¹ which encompassed the official guidelines that were available for that edition. For unanticipated scenarios not covered by the predetermined rules, data were reviewed independently by two infectious disease/critical care medicine physicians who were blinded to outcome. Agreement allowed data entry. Discordant assessments were reviewed by a third similarly trained physician whose decision was determinative. A similar adjudication approach was used for other issues where clinical judgment was required.

Key definitions and data collection methods were enumerated in a previous publication.³ Appropriate antimicrobial therapy was considered to have been initiated if an antimicrobial with in vitro activity appropriate for the isolated pathogen or pathogens (or in the case of culture-negative septic shock, an antimicrobial or antimicrobial agent concordant with accepted international norms for empiric therapy and modified to local flora) was either the first new antimicrobial agent with which therapy was started after the onset of recurrent or persistent hypotension or was initiated within 6 h of the administration of the first new antimicrobial agent. Otherwise, inappropriate therapy was considered to have been initiated.

The primary outcome variable was survival to hospital discharge, including discharge to a chronic health-care facility. The actual (culture positive) or probable (culture negative) appropriateness of the initial antimicrobial regimen was the primary independent variable.

A χ² analysis was used to compare survival in subgroups of patients receiving appropriate vs inappropriate initial empiric therapy. A multivariable logistic regression model was used to examine the independent impact of a variety of clinical and therapeutic variables (including the appropriateness or inappropriateness of initial antimicrobial therapy) on survival to hospital discharge in all patients, in culture-positive patients and in blood culture-positive patients. These statistical analyses were performed using a statistical software package (Statistical Analysis Software, version 9.0; SAS Institute, Inc; Cary, NC). The data are expressed as the mean ± SD for normally distributed variables and median (interquartile range) for others.

The χ² tests were used to examine the distribution of comorbidities and clinical infection sites among those patients receiving appropriate or inappropriate initial therapy. Similarly, the Fisher exact test was used to examine distributions of organisms in the same groups.

In total, 5,715 cases were found to fit the diagnostic criteria for septic shock. The mean age of case patients with septic shock was 62.6 ± 16.3 years, with 56.3% male patients and 43.7% female patients. The mean acute physiology and chronic health evaluation (APACHE) II score³² determined from the most abnormal results within 24 h of shock onset was 25.2 ± 9.7. Treatment with drotrecogin-alfa (activated)³³ was used in 124 cases (outside of randomized trials), while low-dose steroid therapy was utilized in 1,548 cases. An indication for source control (either open surgical or percutaneous) existed in 2,480 cases.

Community-acquired infections accounted for 55.0% of cases, while 45.0% of cases were deemed to be of nosocomial origin. Table 1 describes the frequency of chronic comorbidities among patients with septic shock. Table 2 lists the frequency of clinically defined infection sites.

Documented infections were present in 4,698 of the total of 5,715 cases (82.2%). The remaining 1,017 cases (17.8%) were suspected infections without either a plausible bacterial pathogen isolated, or definitive radiologic, surgical, autopsy, or biopsy evidence of infection. The survival rate in these groups was similar at 43.3% and 45.9%, respectively. A plausible primary microbial pathogen was identified in 4,056 cases (71.0%) and was isolated from the blood in 2,012 cases (35.2%). The breakdown of the specific organisms isolated is shown in Table 3.

Overall, the rate of survival to hospital discharge was 43.7%. Survival was similar whether the infection was documented or suspected, whether a plausible pathogen was identified, and whether bacteremia was present or absent.

Of the entire septic shock study population, 4.8% of patients did not receive therapy that was either proven (defined pathogen) or adjudicated (undefined pathogen) to be effective for the infection thought to underlie septic shock prior to death. The fraction of the total number of patients in whom septic shock developed after they had received apparently effective antimicrobial therapy was 21.8%. Among the remaining 73.4% of patients, the median time to the administration of effective antimicrobial therapy was 5.82 h (interquartile range, 2.2 to 14.3).

Appropriate empiric antimicrobial therapy was initiated in 80.1% of all cases of septic shock. There was substantial variation between subgroups in the fraction of patients in whom appropriate empiric therapy had been initiated (Table 4, Fig 1). There were also significant variations in the appropriateness of initial antimicrobial therapy between groups of clinical infections and organisms (Fig 1). Urinary tract infection (84.8%) and skin and soft-tissue infection (86.9%) appeared to have significantly better initial antimicrobial coverage than pneumonia (75.5%) or intraabdominal infection (75.8%; p < 0.0001). Catheter-associated infection (69.8%) and primary bloodstream infection (68.6%) appeared to have initial coverage that was inferior to that for the other major clinical syndromes (p < 0.0001). Similarly, major differences in the appropriateness of initial antimicrobial coverage existed depending on whether the isolated organism was a Gram-positive organism (77.8% initially appropriate antimicrobial therapy), Gram-negative organism (83.7%), anaerobic organism (84.6%), or fungal organism (43.6%; p < 0.0001 vs other groups) [Fig 1]. Variations in the appropriateness of initial antimicrobial coverage with different microbial species also occurred (p < 0.0001) [Fig 1]. Initial antimicrobial coverage was highest with Streptococcus species, followed by Gram-negative organisms, Staphylococcus aureus, enterococci, and fungi (p < 0.0001).

Overall, there were significant variations in the distribution of patients who had received appropriate vs inappropriate initial empiric antimicrobial therapy when stratified by comorbidities (Table 1), clinical infection syndromes (Table 2), and pathogens (Table 3) [each p < 0.0001]. In some cases, even when such variations did not appear to exist in larger groups (eg, Gram-positive infection), they were nonetheless present in the distribution of individual organisms within that group (Table 3).

Survival to hospital discharge with appropriate and inappropriate initial therapy was 52.0% and 10.3%, respectively (odds ratio [OR], 9.45; 95% CI, 7.74 to 11.54; p < 0.0001), an approximately fivefold decrease. Similar differences in survival were seen in all major epidemiologic, clinical, and organism subgroups (Figs 2–5). The decrease in survival with inappropriate initial therapy ranged from 2.3-fold for pneumococcal infection (Fig 5) to 17.6-fold with primary bloodstream infection-associated septic shock (Fig 3).

A multivariable logistic regression analysis was performed with other factors that might be potentially associated with outcome. Variables encompassed in the model included age, sex, hospital admission APACHE II score, the presence of major comorbid illnesses (AIDS; lymphoma; leukemia; metastatic malignancy; neutropenia; immunosuppression including chemotherapy, major organ transplant, clinically significant liver cirrhosis or failure; heart failure; severe chronic obstructive lung disease; chronic renal failure; end-stage renal disease with long-term dialysis; medication-dependent diabetes mellitus; recent history of elective or emergency surgery or major trauma; substance abuse; systemic autoimmune disease; organic brain syndrome; and major neuromuscular disorder), infection acquisition site (community or nosocomial), hospital, clinical infection site, the presence of a surgical/source control-requiring infection, magnitude of early fluid resuscitation, and choice and rapidity of initiation of initial vasopressor/inotropic support. After adjustment for all variables, the appropriateness of the initial antimicrobial therapy remained most strongly associated with outcome (OR, 8.99; 95% CI, 6.60 to 12.23; p < 0.0001) among all the risk factors assessed (Table 5). Similar results were obtained when the analysis was restricted to culture-positive cases (OR, 6.95; 95% CI, 3.11 to 15.52; p < 0.0001) and blood culture-positive cases (OR, 7.19; 95% CI, 4.90 to 10.55; p < 0.0001) [Table 5].

These data suggest that the provision of inappropriate initial empiric antimicrobial therapy has a very substantial adverse effect on survival in patients with septic shock and that this effect is present across all major clinical subgroups. Further, the study confirms that the occurrence of the initiation of organism-inappropriate empiric therapy is surprisingly common.

The degree to which the administration of inappropriate initial empiric therapy adversely affected survival in the study group was startling. Overall, survival dropped fivefold from 52.0% to 10.3% (OR, 9.45; 95% CI, 7.74 to 11.54; p < 0.0001) with inappropriate initial therapy (Fig 2). Even after adjustment for multiple other putative risk factors, the association of poor outcome with the initiation of inappropriate therapy remained singularly powerful for the overall data set (OR, 8.99; 95% CI, 6.60 to 12.23; p < 0.0001) but also for the culture-positive subgroups (OR, 6.95; 95% CI, 3.11 to 15.52; p < 0.0001) and blood culture-positive subgroups (OR, 7.19; 95% CI, 4.89 to 10.55; p < 0.0001) [Table 5].

Similar associations between mortality and inappropriate initial antimicrobial therapy were shown for every major clinical subgroup. Patients with and without documented bacteremia had comparable reductions in survival depending on whether they received appropriate or inappropriate initial antimicrobial therapy (4.8-fold and 4.7-fold decrease in survival percentage, respectively) [Fig 2]. The relative reduction in survival percentage for appropriate vs inappropriate initial empiric therapy was significantly greater in culture-negative cases than in culture-positive cases (9.2-fold vs 4.7-fold difference in survival, respectively) [Fig 2]. This may be a consequence of the fact that appropriateness must be adjudicated without knowledge of antimicrobial isolate sensitivity for culture-negative cases (ie, we are assessing concordance with international recommendations or guidelines for antimicrobial therapy of these conditions). This may lead to a bias, overestimating the fraction of patients receiving appropriate therapy and underestimating those receiving initial inappropriate therapy.

As noted, nosocomial infection–associated septic shock had a significantly higher overall mortality risk than community-acquired infection-associated septic shock (Fig 1). However, the relative decrease in the fraction of survivors in those patients receiving appropriate vs inappropriate initial therapy in the two groups was similar (5.1-fold vs 4.1-fold, respectively). All major clinical infections (pneumonia, intraabdominal infections, urinary tract infections, skin and soft-tissue infections, catheter-associated infections, and primary bloodstream infections without an overt clinical focus) demonstrated a similar benefit of appropriate therapy over inappropriate initial empiric therapy. The relative decrease in the fraction of survivors in these groups with inappropriate initial empiric therapy ranged from 3.4 to 4.5. The only exception was primary bloodstream infections without an obvious clinical source. In that subgroup (mostly neutropenic, Candida, and S aureus septic shock), appropriate therapy was associated with 17.6-fold better survival than inappropriate initial therapy (47.5% vs 2.7%, respectively).

Patients with septic shock associated with Gram-positive cocci and Gram-negative bacilli infection had relatively similar survival and survival fraction changes with inappropriate initial empiric therapy (52.8% and 15.3%, respectively; Gram-positive cocci, 57.9% and 14.7%, respectively; Gram-negative bacilli infection, 3.5-fold and 3.9-fold decrease, respectively). Patients with both anaerobic and yeast infections had substantially greater differences in outcome depending on whether the initial therapy had been appropriate (decrease in the fraction of survivors of 10.6-fold and 7.8-fold, respectively). With regard to specific organisms, Streptococcus pneumoniae demonstrated the smallest fractional change in survival when inappropriate initial empiric therapy was given (55.1% appropriate vs 23.5% inappropriate, 2.3-fold change). At the other end, septic shock caused by Candida albicans demonstrated a 24.6% survival rate with initial appropriate therapy, but only a 4.6% survival rate without it (ninefold decrease).

Though it might be assumed that the initiation of appropriate antimicrobial therapy should have a beneficial effect on outcome, there are a surprising number of studies^{24,27,34–43} that have failed to demonstrate such an effect. Apart from insufficient sample size, there are plausible scientific rationales to support the possibility that initial empiric antimicrobial therapy might not significantly impact survival in patients with sepsis and other severe infections. For example, in vitro sensitivity testing of pathogenic microbial isolates may not reflect clinical efficacy, as has been suggested in some studies of pneumococcal infection.³⁷ Sepsis may represent a manifestation of an inflammatory/coagulation cascade that advances independent of the initial infectious trigger.^44,45 If so, then the syndrome may be expected to progress independently of the control of the underlying infection through the initiation of appropriate antimicrobial therapy. This may suggest that the most critically ill patients with shock could fail to demonstrate sensitivity to the beneficial effects of appropriate antimicrobial therapy, as has been noted by Rello et al⁴⁶ in their study of nosocomial bacteremia.

The primary weakness of the study is that it was observational in nature. All such studies are susceptible to several forms of bias, most notably selection bias. The large study population obviates this problem to some extent in that the key observation is replicated consistently across all subgroups. As in any observational study, outcomes may also be influenced by unmeasured quality elements. For example, institutions with a high degree of antimicrobial appropriateness may also have other quality-of-care advantages, resulting in a fallacious link between a high degree of initial antimicrobial appropriateness and survival. Against this possibility is that the survival advantage held even when the patient's institution was added to the multivariate analysis. The inclusion of patients with culture-negative septic shock can be criticized. Their inclusion leads to a requirement for a relative judgment as to the appropriateness of the initial empiric therapy (ie, concordance with international recommendations for empiric therapy) in these patients. However, the use of a predefined algorithm for the assessment of appropriateness should minimize the problem of subjective judgment. The fact that the association of survival with appropriate therapy for the subgroups of culture-positive and blood culture-positive cases substantially parallels the response for the entire group (inclusive of the culture-negative cases) suggests a relatively consistent assessment of the appropriateness therapy between the groups. Notably, exclusion of the subgroup could itself be criticized as not being representative of clinical reality.

We have demonstrated that the time to the initiation of effective antimicrobial therapy is the critical determinant of outcome in patients with septic shock.³ The choice of an appropriate empiric antimicrobial agent is one of the key elements in effecting the rapid initiation of effective antimicrobial therapy. This study supports other investigations emphasizing that the selection of an initial antimicrobial regimen with coverage against the inciting isolated or presumed pathogens is required for the best possible survival from septic shock.²⁶ The empiric selection of antimicrobial agents must be a central element in any efforts to address the suboptimal management of this condition.

Author contributions: Study design: Dr. Anand Kumar. Data collection: members of the Cooperative Antimicrobial Therapy of Septic Shock Database Research Group. Data analysis and interpretation: Drs. Anand Kumar and Chateau. Manuscript development: Drs. Anand Kumar, Ellis, Arabi, Roberts, Light, Parrillo, Dodek, Wood, Aseem Kumar, Ahsan, and Simon, and Ms. Peters. Dr. Anand Kumar had full access to all of the data in the study, and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Financial/nonfinancial disclosures: Dr. Anand Kumar has received honoraria for lectures from Eli-Lilly and Co, Merck, Wyeth, Pfizer, and Astra-Zeneca; he has also received grant support for this project as noted. Dr. Light is a consultant for Eli-Lilly and Co. Dr. Parrillo has received research grants from GlaxoSmithKline for porcine/human research on sepsis; holds grants from Arginox, DeepBreeze, and Minimitter; and is also on advisory boards for Edwards, GlaxoSmithKline, and OrthoBioTech (Johnson & Johnson). Drs. Ellis, Arabi, Roberts, Dodek, Wood, Aseem Kumar, Simon, Ahsan, and Chateau, and Ms. Peters have reported to the ACCP that no significant conflicts of interest exist with any companies/organizations whose products or services may be discussed in this article.

Other contributions: Special thanks to Christine Mendez, Sheena Ablang, Debbie Friesen, and Lisa Halstead for data entry.

Role of sponsors: Companies that provided partial grant support for this project had no role in study design; collection, analysis, or interpretation of the data; writing of the report; or the decision to submit the work for publication.

Kenneth E. Wood, MD, University of Wisconsin Hospital and Clinics, Madison, WI; Kevin Laupland, MD, Foothills Hospital, Calgary, AB, Canada; Andreas Kramer, MD, Brandon General Hospital, Brandon, MB, Canada; Bruce Light, MD, Winnipeg Regional Health Authority, Winnipeg, MB, Canada; Satendra Sharma, MD, Winnipeg Regional Health Authority, Winnipeg, MB, Canada; Steve Lapinsky, MD, Mount Sinai Hospital, Toronto, ON, Canada; John Marshall, MD, St. Michael's Hospital, Toronto, ON, Canada; Sandra Dial, MD, Jewish General Hospital, Montreal, QC, Canada; Ionna Skrobik, MD, Hôpital Maisonneuve Rosemont, Montreal, QC, Canada; Gourang Patel, PharmD, and Dave Gurka, MD, Rush-Presbyterian-St. Luke's Medical Center, Chicago, IL; Sergio Zanotti, MD, and R. Phillip Dellinger, MD, Cooper Hospital/University Medical Center, Camden, NJ; Dan Feinstein, MD, St. Agnes Hospital, Baltimore, MD; Jorge Guzman, MD, Harper Hospital, Detroit, MI; and Nehad Al Shirawi, MD, and Ziad Al Memish, MD, King Saud Bin Abdulaziz University for Health Sciences, Saudi Arabia.

John Ronald, MD, Nanaimo Regional Hospital, Nanaimo, BC, Canada; Mustafa Suleman, MD, Concordia Hospital, Winnipeg, MB, Canada; Harleena Gulati, MD, Erica Halmarson, MD, Robert Suppes, MD, Cheryl Peters, RN, Katherine Sullivan, Rob Bohmeier, Sheri Muggaberg, and Laura Kravetsky, University of Manitoba, Winnipeg, MB, Canada; Amrinder Singh, MD, Winnipeg, MB, Canada; Lindsey Carter, BA, Winnipeg, MB, Canada; Kym Wiebe, RN, and Laura Kolesar, RN, St. Boniface Hospital, Winnipeg, MB, Canada; Jody Richards, Camosun College, Victoria, BC, Canada; Danny Jaswal, MD, Harris Chou, BSc, Tom Kosick, MD, Winnie Fu, Charlena Chan, and Jia Jia Ren, University of British Columbia, Vancouver, BC, Canada; Mozdeh Bahrainian, MD, Madison, WI; Ziaul Haque, MD, Montreal, QC, Canada; Omid Ahmadi Torshizi, MD, Montreal, QC, Canada; Heidi Paulin, University of Toronto, Toronto, ON, Canada; Farah Khan, MD, Toronto, ON, Canada; Runjun Kumar, University of Toronto, Toronto, ON, Canada; Johanne Harvey, RN, Hôpital Maisonneuve Rosemont, Montreal, QC, Canada; Christina Kim, Jennifer Li, and Latoya Campbell, McGill University, Montreal, QC, Canada; Leo Taiberg, MD, Rush Medical College, Chicago, IL; Christa Schorr, RN, Cooper Hospital/University Medical Center, Camden, NJ; Ronny Tchokonte, MD, Wayne State University Medical School, Detroit, MI; and Catherine Gonzales, RN, Norrie Serrano, RN, and Sofia Delgra, RN, King Saud Bin Abdulaziz University for Health Sciences, Saudi Arabia.