early combination antibiotic therapy yields improved ......early combination antibiotic therapy...

Early combination antibiotic therapy yields improved survival comparedto monotherapy in septic shock: A propensity-matched analysis

Anand Kumar, MD; Ryan Zarychanski, MD; Bruce Light, MD; Joseph E. Parrillo, MD; Dennis Maki, MD;Dave Simon, MD; Denny Laporta, MD; Steve Lapinsky, MD; Paul Ellis, MD; Yazdan Mirzanejad, MD;Greg Martinka, MD; Sean Keenan, MD; Gordon Wood, MD; Yaseen Arabi, MD; Daniel Feinstein, MD;Aseem Kumar, PhD; Peter Dodek, MD; Laura Kravetsky, BSc; Steve Doucette, MSc; the Cooperative AntimicrobialTherapy of Septic Shock (CATSS) Database Research Group

From Section of Critical Care Medicine (AK, LK),Health Sciences Centre/St. Boniface Hospital, Univer-sity of Manitoba, Winnipeg, Manitoba, Canada; CancerCare Manitoba (RZ), University of Manitoba, Winnipeg,Manitoba, Canada; Cooper Hospital/University MedicalCenter (BL), Robert Wood Johnson Medical School,UMDNJ, Camden, New Jersey; Section of InfectiousDiseases (DM), University of Wisconsin, Madison, Mad-ison, WI; Section of Infectious Diseases (DS), RushUniversity, Chicago, Illinois; Section of Critical CareMedicine (DL), Jewish General Hospital, McGill Univer-sity, Montreal, Quebec City, Canada; Section of CriticalCare Medicine (SL), Mount Sinai Hospital, University ofToronto, Toronto, Ontario Canada; Department ofEmergency Medicine (PE), University Health Network,University of Toronto, Toronto, Ontario, Canada; SurreyMemorial Hospital (YM), Surrey, British Columbia, Can-ada; Richmond General Hospital (GM), Vancouver, Brit-ish Columbia, Canada; Royal Columbian Hospital (SK),Vancouver, British Columbia, Canada; Royal JubileeHospital/Victoria General Hospital (GW), University ofBritish Columbia, Victoria, British Columbia, Canada;Intensive Care Department (YA), King Saud Bin Abdu-laziz University for Health Sciences, Riyadh, Saudi

Arabia; Moses H. Cone Memorial Hospital (DF),Greensboro, North Carolina; Laurentian University (AK),Biomolecular Sciences Program and Department ofChemistry and Biochemistry, Sudbury, Ontario, Can-ada; Section of Critical Care Medicine (PD), St. Paul’sHospital, University of British Columbia, Vancouver,British Columbia, Canada; Ottawa Health ResearchCentre (SD), Ottawa, Ontario, Canada.

Dr. Kumar has received grants from Wyeth, Astra-Zeneca, Pfizer, and Rocheu. Dr. Parrillo consulted withSangart, Artisan, Philips, and Immunetrics. He receiveda grant from the Robert Wood Johnson Foundation. Dr.Mirjanezad consulted for the advisory boards of Scher-ing-Plough Corporation and Pfizer. He received hono-raria/speaking fees from Merck, Schering-Plough Cor-poration, Bayer, and Wyeth. He also received grantsfrom Par101, C.diff, and INC Research. All other au-thors have no potential conflicts of interest to disclose.

None of the authors have financial or personalrelationships or affiliations that could influence (orbias) the decision regarding the analysis or manuscriptin any regard.

Funding for this work was provided by the Mani-toba Health Research Council, Health Sciences Centre

Foundation, and the Alfred Deacon Foundation. Addi-tional support was provided by through unrestrictedgrants from Eli-Lilly, Pfizer, Astellas Pharma, Merck,Wyeth, Bayer, Bristol-Myers-Squibb, and Astra-Zeneca. Funding sources had no role in the design andconduct of the study; collection, management, analy-sis, and interpretation of the data; and preparation,review, or approval of the manuscript.

Dr. Kumar had full access to all the data in thestudy is responsible for the integrity of the databaseand the accuracy of the data analysis.

This specific research concept, the septic shockdatabase, and manuscript were developed by Dr. Ku-mar. Dr. Kumar and Mr. Doucette were responsible forthe methodological design issues and data analysis. Allauthors assisted with data interpretation and manu-script revisions.

For information regarding this article, E-mail:[email protected]

Copyright © 2010 by the Society of Critical CareMedicine and Lippincott Williams & Wilkins

DOI: 10.1097/CCM.0b013e3181eb3ccd

Background: Septic shock represents the major cause of infec-tion-associated mortality in the intensive care unit. The possibilitythat combination antibiotic therapy of bacterial septic shock im-proves outcome is controversial. Current guidelines do not recom-mend combination therapy except for the express purpose of broadeningcoverage when resistant pathogens are a concern.

Objective: To evaluate the therapeutic benefit of early combi-nation therapy comprising at least two antibiotics of differentmechanisms with in vitro activity for the pathogen in patients withbacterial septic shock.

Design: Retrospective, propensity matched, multicenter, co-hort study.

Setting: Intensive care units of 28 academic and communityhospitals in three countries between 1996 and 2006.

Subjects: A total of 4662 eligible cases of culture-positive,bacterial septic shock treated with combination or monotherapyfrom which 1223 propensity-matched pairs were generated.

Measurements and Main Results: The primary outcome ofstudy was 28-day mortality. Using a Cox proportional hazardsmodel, combination therapy was associated with decreased 28-

day mortality (444/1223 [36.3%] vs. 355/1223 [29.0%]; hazardratio, 0.77; 95% confidence interval, 0.67–0.88; p � .0002). Thebeneficial impact of combination therapy applied to both Gram-positive and Gram-negative infections but was restricted to pa-tients treated with �-lactams in combination with aminoglyco-sides, fluoroquinolones, or macrolides/clindamycin. Combinationtherapy was also associated with significant reductions in inten-sive care unit (437/1223 [35.2%] vs. 352/1223 [28.8%]; odds ratio,0.75; 95% confidence interval, 0.63–0.92; p � .0006) and hospitalmortality (537/1223 [53.6%]vs. 424/1223 [46.4%]; odds ratio,0.69; 95% confidence interval, 0.59–0.81; p < .0001). The use ofcombination therapy was associated with increased ventilator(median and [interquartile range], 10 [0–25] vs. 17 [0–26]; p �.008) and pressor/inotrope-free days (median and [interquartilerange], 23 [0–28] vs. 25 [0–28]; p � .007) up to 30 days.

Conclusion: Early combination antibiotic therapy is associatedwith decreased mortality in septic shock. Prospective randomizedtrials are needed. (Crit Care Med 2010; 38:000–000)

KEY WORDS: antibiotic; combination; monotherapy; mortality;sepsis; septic shock

1Crit Care Med 2010 Vol. 38, No. 9

Several studies have shown thatappropriate antimicrobial ther-apy, defined as the use of atleast one agent with in vitro

activity for the isolated pathogen, leads tolower mortality rates in life-threateninginfections associated with sepsis (1–5).The incremental benefit of combinationas opposed to single-agent antimicrobialtherapy in such situations is controver-sial (6–10).

Some clinical studies of bacterial in-fection, including endocarditis, Gram-negative bacteremia, and neutropenicsepsis (11–13), and animal models of se-vere infection (14–16) have supportedthe possibility of clinically relevant anti-microbial synergism with appropriatecombinations of antibiotics. However,two separate meta-analyses have failed todemonstrate any consistent benefit withcombination therapy in immunocompe-tent patients with sepsis and/or Gram-negative bacteremia (17, 18).

The beneficial effect of early appropri-ate antimicrobial therapy appears to bemost important in critically ill patients,particularly those with septic shock (5).Given these data and in view of a recentmeta-regression study that suggests thatthe beneficial effect of combination ther-apy may be restricted to patients withlife-threatening infection at highest riskfor death (19), we performed a propensi-ty-matched study examining the impactof combination antibiotic therapy inadult patients with bacterial septic shockusing a multinational septic shock data-base. The hypothesis of this study wasthat early combination antimicrobialtherapy of septic shock as defined by theuse of at least two antibiotics (of differentmechanistic classes) with in vitro activityagainst the isolated pathogen is associ-ated with reduced mortality.

SUBJECTS AND METHODS

A retrospective review of adult (age 18 yrsor older) patients with septic shock was per-formed. A waived consent protocol was ap-proved by the Health Ethics Board of the Uni-versity of Manitoba and at each individualparticipating center. Consecutive adult septicshock patients from 28 medical institutions inCanada, United States, and Saudi Arabia forperiods between 1996 and 2007 were retro-spectively identified using internal intensivecare unit registries/databases and/or Interna-tional Classification of Diseases (9 or 10) cod-ing strategies. Each institution contributed aminimum of 50 cases. Each potential case wasscreened to determine whether the case met

specific criteria for septic shock as describedby the 1991 Society of Critical Care Medicine/American College of Chest Physicians Consen-sus Statement on Sepsis Definitions (20). Theprocess used to identify the final study popu-lation is outlined in Figure 1.

Data Elements and Definitions

Clinical infection definitions were adaptedfrom previous recommendations or studies(21–23). To qualify as potential pathogenscausing shock, isolates from anatomical sitesand/or blood cultures were required to havebeen obtained within 48 hrs of onset of shock.

A priori criteria were developed to uni-formly determine the primary pathogen/pathogens and to assess the appropriateness ofantimicrobial therapy across participating in-stitutions (Appendices 1 and 2). The first useof any appropriate antimicrobial therapy (i.e.,with in vitro activity for the primary isolatedpathogen or pathogens) was determined for allcases. For the purposes of this study, antibi-otic monotherapy was defined as the adminis-tration of any single, appropriate, intravenous,preferably bactericidal (see Appendix 2 for ex-ceptions) antibiotic at any point after the on-set of recurrent or persistent hypotension.Combination therapy was defined as the con-comitant use of two or more such antibioticsof different mechanistic classes for at least 24hrs after the onset of hypotension or until

death if the patient survived �24 hrs afterhypotension documentation. The secondagent had to be started within 24 hrs of thefirst antibiotic or within 24 hrs of the onset ofhypotension (if the first agent was initiatedbefore hypotension was documented). Thecombination of two �-lactams or a �-lactamand a glycopepide was not considered to rep-resent antibiotic combination therapy (be-cause all are cell-wall–active agents with sim-ilar mechanisms of action) for purposes of thisanalysis. Typical permutations of antibioticsmeeting criteria for combination therapy in-clude any two of a cell-wall–active agent (i.e.,a �-lactam or glycopeptide), aminoglycoside,fluoroquinolone, or macrolide/clindamycin.

Questionable cases or data elements werereviewed and adjudicated by the principal in-vestigator. Cases of septic shock associatedwith negative or absent cultures and thosecaused by yeast/fungi, anaerobes, and atypicalpathogens, such as Mycobacterium tuberculo-sis and Legionella species, were excluded (Fig.1). Patients who did not receive any appropri-ate antimicrobial therapy before death werealso excluded.

Subsets of subject data examined in thisstudy have been utilized for several earlierpublications (5, 21, 24, 25). Data collectionmethods have been described in those previ-ous studies (5, 21). Data were collected bytrained research nurses/medical students us-

Figure 1. Subject selection flow diagram. SCCM, Society of Critical Care Medicine; ACCP, AmericanCollege of Chest Physicians; ICU, intensive care unit. *Clostridia species, Bacteroides species, pep-tostreptococci, Clostridium difficile-associated septic shock, miscellaneous anaerobes. **Mycobacte-rium tuberculosis, Legionella species, Listeria, Bacillus species, Corynebacterium jeikeium.

2 Crit Care Med 2010 Vol. 38, No. 9

ing a standardized and piloted data extractiontemplate. Variables collected included patientdemographics, baseline comorbidities, AcutePhysiology and Chronic Health Evaluation IIscore (26), physiologic/laboratory parameters,and the need for hemodynamic or ventilatorysupport. The site of infection, microbiologicalculture results, and the time to appropriateantimicrobial therapy from the onset of hypo-tension were also recorded.

Outcome Measures

The primary outcome variable was mortal-ity over 28 days. Mortality stratified by severityof illness (Acute Physiology and ChronicHealth Evaluation II score), time from initialdocumentation of hypotension to first appro-priate antimicrobial, and time between initia-tion of the first and second appropriate anti-biotic were identified a priori as secondaryoutcome measures. Other predetermined sec-ondary end points included hospital and in-tensive care unit mortality and vasopressor/inotropic support-free days in the first 30 daysafter shock. Exploratory analyses were per-formed to examine the association of any po-tential benefit of combination therapy accord-ing to antibiotics utilized, clinical syndrome,and primary pathogen.

Statistical Analysis

Baseline characteristics between patientsreceiving monotherapy and combination anti-biotic therapy were compared using Student’st test or Wilcoxon’s rank sum test for contin-uous variables, or the chi-square test for cat-egorical variables. All reported p values weretwo-tailed. Because combination therapy wasnot randomly assigned, a propensity analysiswas undertaken to account for potential con-founding factors and selection biases. The pro-pensity matching and analytic methods usedin this study incorporated aspects from severalreference sources (24, 27, 28). A propensityscore for combination antibiotic therapy usewas developed using multivariable logistic re-gression. This score represents the probabilitythat a patient would receive combination ther-apy based on variables that were known orsuspected to influence group assignment(monotherapy or combination therapy) or toaffect mortality risk. Variables included in thederivation of the propensity score are shownin Table 1. Among these were age, gender,Acute Physiology and Chronic Health Evalua-tion II score, the number of day 1 organ fail-ures, occurrence of shock before initiation orduring therapy with an appropriate antibiotic,time to the initial dose of the first appropriateantibiotic (after documentation of sepsis-associated hypotension), site of infection ac-quisition (community, nosocomial), preexist-ing medical conditions, infecting organism,

presence of bacteremia, primary antibiotictherapy, anatomical site of infection, volumeof fluid resuscitation in the first hour of hy-potension, use of therapies including sourcecontrol, activated protein C, corticosteroids,mechanical ventilation, and a variety of labo-ratory data including white blood cell count,platelet count, the international normalizedratio, serum creatinine, and serum bicarbon-ate. To account for temporal and geographicpractice variability, the date of intensive careunit admission and hospital sites (region/academic vs. nonacademic) were also incorpo-rated as matching variables.

One of the two antibiotics in each combi-nation therapy regimen had to be designatedas the primary agent in the combination formatching purposes. The required mono-therapy match was prioritized as follows:�-lactam/vancomycin, fluoroquinolones, mac-rolides/clindamycin, and other. The priority-matched drug was considered the primary an-tibiotic and the additional drug of thecombination regimen was considered the sec-ondary/supplemental antibiotic. Aminoglyco-sides were always considered secondary/supplemental.

Propensity scores were used to match pa-tients who received combination therapy to acontrol patient receiving monotherapy using aSAS macro (SAS, Cary, NC). A greedy match-ing procedure selected match pairs initiallyidentical to five decimal places of probability(29). If no match existed at five decimal places,then matching would occur at four decimalplaces, and so on. If no match existed at onedecimal place, then that patient receivingcombination therapy was excluded from thestudy. To compensate for immortal time bias(30), matching was restricted so that the min-imum duration of survival (from hypotension)of the patient with the monotherapy case wasconsistently longer (range, 1 min to 48 hrs asa consequence of this restriction) than theduration of time between hypotension onsetand administration of the second appropriateantibiotic for the patient with the combinationtherapy case (i.e., the matching patient withthe monotherapy case always lived longenough to have had the same or greater op-portunity to have received a second appropri-ate antibiotic). This was accomplished by en-suring the patient with the monotherapy casesutilized for matching always lived at least 1day beyond the point that the patient with thecombination therapy case received the firstdose of the second appropriate antibiotic. Us-ing this strategy, 1223 of 1714 (71.4%) pa-tients who received combination therapy wereable to be suitably matched using propensityscores.

Mortality over 28 days was assessed using aCox proportional hazards model. Hazard mod-els incorporated survival data over the com-

plete duration of the study period (28 days) oruntil the time of censoring (i.e., death). Mor-tality estimates stratified by the delay fromhypotension to the first antibiotic and the firstto second antibiotic (in combination regi-mens) were assessed by the addition of aninteraction term to the hazard model (31). Ahazard or odds ratio �1 signifies decreasedmortality in the combination therapy groupcompared to the monotherapy group. Statisti-cal analyses were conducted using SAS version9.1 (SAS Institute, Cary, NC). The confidencelimits and p values reported reflect � level of0.05.

RESULTS

Baseline Characteristics BeforePropensity Matching

Antibiotic combination therapy wasadministered to 1714 of 4662 (36.8%)patients with eligible bacterial septicshock (Fig. 1). The remaining 63.2% re-ceived monotherapy. Baseline demo-graphics, preexisting medical conditions,and relevant clinical, physiologic, andlaboratory parameters in the unmatchedstudy population are summarized in Ta-ble 1. Males comprised 56.8% and 58.7%of the combination and monotherapygroups. The age and admission AcutePhysiology and Chronic Health Evalua-tion II score in the unmatched study pop-ulation was 61.9 (�16.2) and 25.1 (�8.0),respectively. Age was significantlyyounger in the combination therapygroup compared with the monotherapygroup; Acute Physiology and ChronicHealth Evaluation II scores also trendedlower. The median time to appropriateantibiotic therapy was significantlyshorter in the combination therapygroup, although there were no differ-ences in fluid resuscitation volumes be-tween groups in the first hour of hypo-tension.

Several clinical differences betweenthe groups existed in the unmatched co-hort. The baseline prevalence of liver fail-ure/cirrhosis, diabetes, and chronic renalfailure was higher in the monotherapygroup, whereas the prevalence of inva-sive/metastatic malignancy, neutropenia,and other immunosuppression washigher in the combination therapy group(Table 1). Recent elective or emergencysurgery/trauma was associated with anincreased probability of receiving mono-therapy. The documented presence ofbacteremia, in contrast, was associatedwith combination therapy. The platelet


Table 1. Unmatched and propensity score-matcheda baseline characteristics

Unmatched Cohort Propensity Matched Cohort

Monotherapy,n � 2948

Combined Therapy,n � 1714 p



Male, n 1730 (58.7%) 974 (56.8%) .22 697 (57.0%) 686 (56.1%) .65Age, yr, mean � SD 62.8 � 16 60.3 � 16.4 �.0001 61.6 � 16.3 61.8 � 16.1 .75Shock date, median May 21, 2001 September 5, 2000 �.0001 October 31, 2001 November 13, 2001 .65Time to first antibiotic, hrs, median (IQR) 3.75 (0.4–10.3) 1.53 (0–5.1) �.0001 2.15 (0.02–6) 2.0 (0.08–6.2) .94Acute Physiology and Chronic Health Evaluation

II Score, mean � SD

23.9 � 10 23.3 � 9.2 .07 23.4 � 9.5 23.7 � 9.6 .46

Total day 1 organ failures, median (IQR) 3 (2–5) 3 (2–4) .03 3 (2–5) 3 (2–4) .97Duration of hospitalization before shock (IQR) 1 (0–9) 1 (0–3) �.0001 1 (0–5) 1 (0–5) .85Infection acquisition site

Community 1659 (56.3%) 1247 (72.8%) �.0001 841 (68.8%) 822 (67.2%) .41Hospital (nosocomial) 1289 (43.7%) 467 (27.2%) 382 (31.2%) 401 (32.8%)

Hospital/regional distributionCentral Canada (academic) 924 (31.3%) 568 (33.1%) �.0001 407 (33.3%) 407 (33.3%) .9Central Canada (community) 323 (11%) 221 (12.9%) 163 (13.3%) 160 (13.1%)Eastern Canada (mixed) 447 (15.2%) 192 (11.2%) 146 (11.9%) 154 (12.6%)West Coast Canada (academic) 375 (12.7%) 171 (10%) 107 (8.7%) 121 (9.9%)West Coast Canada (community) 386 (13.1%) 181 (10.6%) 145 (11.9%) 131 (10.7%)East Coast United States (mixed) 172 (5.8%) 131 (7.6%) 89 (7.3%) 86 (7%)Central United States (academic) 144 (4.9%) 164 (9.6%) 85 (7%) 93 (7.6%)Outside North America (academic) 177 (6%) 86 (5%) 81 (6.6%) 71 (5.8%)

Preexisting medical conditions, nLiver failure/cirrhosisb 264 (9%) 87 (5.1%) �.0001 72 (5.9%) 76 (6.2%) .73Chronic obstructive pulmonary diseasec 417 (14.1%) 219 (12.8%) .19 165 (13.5%) 169 (13.8%) .81Diabetes mellitusc 850 (28.8%) 444 (25.9%) .03 341 (27.9%) 332 (27.1%) .68Chronic renal insufficiencyd 462 (15.7%) 227 (13.2%) .02 178 (14.6%) 184 (15%) .73Dialysis dependence 228 (7.7%) 128 (7.5%) .74 101 (8.3%) 104 (8.5%) .83Malignancye 424 (14.4%) 300 (17.5%) .005 209 (17.1%) 205 (16.8%) .83Immunosuppressionf 374 (12.7%) 271 (15.8%) .003 185 (15.1%) 181 (14.8%) .82Neutropenia (�1000 neutrophils/�L) 95 (3.2%) 96 (5.6%) �.0001 61 (5%) 55 (4.5%) .57

Recent surgical history, nElective surgery 491 (16.7%) 216 (12.6%) .0002 172 (14.1%) 182 (14.9%) .57Emergency surgery 258 (8.8%) 88 (5.1%) �.0001 82 (6.7%) 76 (6.2%) .62No elective or emergency surgery 2231 (75.7%) 1424 (83.1%) �.0001 981 (80.2%) 977 (79.9%) .84

Physiologic and laboratory parameters onadmission, median (IQR)a

White blood cells, �108 cells/L 13.9 (5.9–21.3) 13.5 (4.6–21.3) .23 14 (4.8–21.9) 13.7 (5.0–21.0) .30Platelet count, �1011 cells/L 162 (88–257) 151 (80–237) .001 159 (87–251) 155 (84–244) .39Serum creatinine 141.4 (80–247) 141.4 (88–239) .24 141 (83–245) 139 (85–240) .61Serum bicarbonate 16.2 (0–22) 16.7 (0–21.4) .58 16.0 (0–21.2) 16.4 (0–21.6) .87Serum bilirubin 14 (6–30.8) 14 (6.8–27) .96 14 (6–29) 14 (6.8–28) .77International normalized ratio 1.4 (1.1–1.8) 1.3 (1.1–1.7) .06 1.3 (1.1–1.7) 1.3 (1.1–1.7) .57

Bacteremia, n 1305 (44.3%) 951 (55.5%) �.0001 622 (50.9%) 627 (51.3%) .84Cointerventions, n

Activated protein C 132 (4.5%) 100 (5.8%) .04 50 (4.1%) 60 (4.9%) .21Steroids 937 (31.8%) 518 (30.2%) .27 355 (29%) 361 (29.5%) .79Source control 1141 (38.7%) 662 (38.6%) .96 478 (39.1%) 475 (38.8%) .90Ventilator support (admission day) 2183 (74.1%) 1160 (67.7%) �.0001 853 (69.7%) 852 (69.7%) .96First hr fluid resuscitation, L, mean � SD 0.67 � 0.96 0.68 � 0.94 .81 0.69 � 0.87 0.67 � 0.93 .71

Site of Infection, nPrimary bloodstream infection 152 (5.2%) 113 (6.6%) �.0001 68 (5.6%) 70 (5.7%) 1.00Catheter-related infection 124 (4.2%) 72 (4.2%) 50 (4.1%) 52 (4.3%)Respiratory infection 1214 (41.2%) 639 (37.3%) 449 (36.7%) 439 (35.9%)Urinary tract infection 379 (12.9%) 304 (17.7%) 224 (18.3%) 229 (18.7%)Intra-abdominal infection 681 (23.1%) 332 (19.4%) 272 (22.2%) 275 (22.5%)Central nervous system infection 41 (1.4%) 13 (0.8%) 13 (1.1%) 12 (1%)Soft tissue infection 281 (9.5%) 201 (11.7%) 119 (9.7%) 116 (9.5%)Surgical site infection 48 (1.6%) 21 (1.2%) 16 (1.3%) 16 (1.3%)Nonrespiratory intrathoracic infection 17 (0.6%) 8 (0.5%) 7 (0.6%) 8 (0.7%)Other infection 11 (0.4%) 11 (0.6%) 5 (0.4%) 6 (0.5%)


count was significantly lower and the in-ternational normalized ratio trendedlower in the combination therapy group.All patients required vasoactive medica-tions because of hypotension. The needfor mechanical ventilation on admissionwas higher in the monotherapy group.Nosocomial and healthcare-associated in-fections were more likely than communi-ty-acquired infections to be in themonotherapy groups. There were nostatistically significant differences inthe use of stress dose steroids and theprovision of source control between

groups, but the use of activated proteinC was significantly higher with combi-nation therapy. There were significantdifferences between groups in distribu-tion of the anatomical site of infection,primary pathogenic organism, primaryantibiotic used, and hospital sites (Ta-ble 1).

Baseline Characteristics AfterPropensity Matching

Suitable propensity matches werefound for 1223 (71.4%) of 1714 patients

receiving combination therapy. The cstatistic for the propensity derivationmodel was 0.785. The range of propen-sity scores was similar in both the com-bination and the monotherapy groups(each, 0.01– 0.96). The matching pro-cess eliminated all significant differ-ences that existed between the combi-nation and monotherapy groupsregarding patient demographics, epide-miologic factors, preexisting medicalconditions, or relevant clinical, physio-logic and laboratory parameters (Table1). Penicillin or carbapenem therapy of

Table 1.—Continued

Unmatched Cohort Propensity Matched Cohort





Primary pathogen, nStaphylococcus pyogenes (group A streptococci) 47 (1.6%) 141 (8.2%) �.0001 45 (3.7%) 49 (4%) 1.00Non-group A b-hemolytic streptococci 50 (1.7%) 57 (3.3%) 33 (2.7%) 30 (2.5%)Viridans streptococci 56 (1.9%) 39 (2.3%) 30 (2.5%) 30 (2.5%)Staphylococcus pneumoniae 170 (5.8%) 281 (16.4%) 141 (11.5%) 141 (11.5%)Staphylococcus aureus 803 (27.2%) 129 (7.5%) 139 (11.4%) 128 (10.5%)Enterococcus species 179 (6.1%) 33 (1.9%) 29 (2.4%) 30 (2.5%)Other Gram-positivesg 4 (0.1%) 3 (0.2%) 2 (0.2%) 2 (0.2%)Escherichia coli 681 (23.1%) 490 (28.6%) 373 (30.5%) 386 (31.6%)Klebsiella species 269 (9.1%) 179 (10.4%) 144 (11.8%) 139 (11.4%)Enterobacter species 118 (4%) 68 (4%) 55 (4.5%) 51 (4.2%)Other Enterobacteriaceaeh 150 (5.1%) 97 (5.7%) 67 (5.5%) 71 (5.8%)Pseudomonas aeruginosa 226 (7.7%) 125 (7.3%) 102 (8.3%) 98 (8%)Haemophilus species 76 (2.6%) 32 (1.9%) 32 (2.6%) 31 (2.5%)Other non-Enterobacteriaceaei 81 (2.7%) 28 (1.6%) 22 (1.8%) 25 (2%)Neisseria meningitidis 27 (0.9%) 6 (0.4%) 4 (0.3%) 6 (0.5%)Moraxella catarrhalis 11 (0.4%) 6 (0.4%) 5 (0.4%) 6 (0.5%)

Primary antibiotic, nPenicillinsj 86 (2.9%) 188 (11%) �.0001 62 (5.1%) 71 (5.8%) 1.00Anti-staphylococcal penicillinsb,0 117 (4%) 40 (2.3%) 40 (3.3%) 36 (2.9%)�-lactam/�-lactamase inhibitorsb,b 603 (20.5%) 357 (20.8%) 289 (23.6%) 293 (24%)Cephalosporins, 1st generationb,c 51 (1.7%) 28 (1.6%) 21 (1.7%) 18 (1.5%)Cephalosporins, 2nd generationb,d 146 (5%) 77 (4.5%) 60 (4.9%) 62 (5.1%)Cephalosporins, 3rd generationb,e 513 (17.4%) 520 (30.3%) 339 (27.7%) 332 (27.1%)Cephalosporins, anti-pseudomonal/monobactamb,f 172 (5.8%) 173 (10.1%) 119 (9.7%) 116 (9.5%)Carbapenemsb,g 437 (14.8%) 172 (10%) 154 (12.6%) 152 (12.4%)Vancomycin 483 (16.4%) 85 (5%) 82 (6.7%) 76 (6.2%)Fluorquinolonesb,h 266 (9%) 67 (3.9%) 50 (4.1%) 60 (4.9%)Macrolidesb,i/clindamycin 55 (1.9%) 6 (0.4%) 6 (0.5%) 6 (0.5%)

Otherb,j 19 (0.6%) 1 (0.1%) 1 (0.1%) 1 (0.1%)Secondary antibiotic, n

Aminoglycosides 683 (39.9%) 526 (43.0%)Fluoroquinolones 651 (38.0%) 498 (40.7%)Macrolidesb,i/clindamycin 350 (20.4%) 174 (14.2%)Otherb,j 30 (1.8%) 25 (2.0%)

IQR, interquartile range.avariables included in propensity derivation but not shown include the occurrence of shock while using appropriate antimicrobial therapy, appropri-

ateness of initial antibiotic therapy, the presence of congestive heart failure, and admission serum lactate and albumin concentrations; bdefined byappropriate history in context of clinical symptoms/signs of liver dysfunction per attending physicians; cmedication-dependent; dcreatinine 1.5� normalvalue; einvasive or metastatic; facquired immune deficiency syndrome, malignancy or autoimmune-related chemotherapy, �20 mg/day chronic prednisoneequivalent or major organ transplant; gstaphylococcus lugdenesis, Leuconostoc, and Micrococcus species; hserratia, Proteus, Citrobacter, Morganella,Salmonella, Providencia, and Hafnia species; iacinetobacter, Stenotrophomonas, Aeromonas, Burkholderia, and Acaligenes; jpenicillin, ampicillin,piperacillin, ticarcillin, and mezlocillin; kcloxacilin, nafcillin, and oxacillin; lpiperacilln/tazobactam, ticarcilin/clavulanate, and ampicillin/sulbactam;mcefazolin; ncefuroxime, cefoxitin, and cefotetan; ocefotaxime, and ceftriaxone; pceftazidime, cefepime, and aztreonam; qmeropenem, imipenem/cilastatin,and ertapenem; rlevofloxacin, ciprofloxacin, gatifloxacin, trovofloxacin, and ofloxacin; sazithromycin, erythromycin, and clarithromycin; tcolistin, tri-methoprim/sulfamethoxazole, linezolid, daptomycin, quinipristin/dalfopristin, and rifampin. Acute Physiology and Chronic Health Evaluation. Patientswere assessed on the day of onset of shock. The range of scores for this test is 0 to 71.


enterococci accounted for 29 of the 33cases in which a static regimen wasincluded in the matched cohorts.

Combination Antibiotic Therapyand Mortality

In the propensity-adjusted Cox model,mortality over 28 days was significantlyreduced with combination therapy (444/1223 [36.3%] vs. 355/1223 [29.0%]; haz-ard ratio, 0.77; 95% confidence interval,0.67–0.88; p � .0002) (Fig. 2). Stratifica-tion by Acute Physiology and ChronicHealth Evaluation II score also revealedconsistently reduced 28-day mortality ineach assessed tertile, with significant dif-ferences in the middle and highest risktertile groups (Table 2).

Stratification by time from hypoten-sion to first appropriate antimicrobialdemonstrated evidence of an increasedeffect with greater delays (test of interac-tion p � .39). Stratification by the delaybetween the first and second antimicro-bial in combination therapy revealed re-duced efficacy of combination therapy,with increasing delays out to 24 hrs (testfor interaction p � .03) (Table 2). Theabsolute reduction in mortality in pa-tients with the most rapid administrationof the second drug (delay of 0.0–1.0 hr)was 11.9%, with a corresponding hazardratio of 0.68 (95% confidence interval,0.53–0.89; p � .004), whereas the abso-lute reduction in mortality for the mostdelayed second drug (maximum delay, 24hrs) was 2.5%, with a hazard ratio of 0.89(95% confidence interval, 0.89–1.15; p �.37) (Table 2). Combination therapy re-duced mortality associated with death at-

tributable to refractory shock, sepsis-related organ failure, and apparentnonsepsis-related causes (Table 3).

In the propensity matched cohort, theproportion of ventilated patients success-fully liberated from mechanical ventilationwas higher in the combination therapygroup compared with the monotherapygroup (67.8% vs. 61.8%; odds ratio, 1.34;95% confidence interval, 1.12–1.61; p �.001). Successful discontinuation of vaso-pressor/inotrope support was also higher inthe combination therapy group (80.1% vs.75.3%; odds ratio, 1.32; 95% confidenceinterval, 1.09- 1.61; p � .005). Similarly thenumber of ventilator and pressor/inotrope-free days (in the first 30 days) was signifi-cantly greater in patients receiving combi-nation therapy (Table 4). The medianhospital length of stay in survivors (totaln � 1647) was significantly shorter in thecombination therapy group (22 days; inter-quartile range [IQR], 13–44; vs. 26 days;IQR, 14–49; p � .01), but the median in-tensive care unit length of stay was not(monotherapy: 7 days; IQR, 4–15; vs. com-bination therapy: 7 days; IQR, 4–13; p �.29). Among survivors, no significant differ-ences between monotherapy and combina-tion therapy in duration of pressor infusion(3 days; IQR, 2–4; vs. 3 days; IQR 1–4; p �.67) and duration of mechanical ventilation(6 days; IQR, 3–12.5; vs. 6 days; IQR,3–11.5; p � .88) were observed.

Of the eight hospital/regional group-ings shown in Table 1, six demonstratedevidence of combination therapy benefitin the unadjusted analysis, with five re-taining significance (p � .05) after pro-pensity matching. The remaining threedid not demonstrate a significant advan-

tage to either combination therapy ormonotherapy.

The beneficial effect of combinationtherapy was predominantly seen with�-lactams as primary therapy includingboth penicillins and cephalosporins (Fig.3). When �-lactams were included as theprimary agent, combination therapy withaminoglycosides, fluoroquinolones, andmacrolides/clindamycin as secondary/supplemental agents was associated witha superior outcome compared to mono-therapy with the �-lactam (Fig. 4).

The benefit of combination therapywas significant for bacteremic and non-bacteremic infections (Fig. 5). Similarly,both pneumonia and nonpneumonia in-fections demonstrated superiority ofcombination therapy (Fig. 5). A signifi-cant benefit was also seen in both Gram-positive and Gram-negative organismsand in Staphylococcus pneumoniae andStaphylococcus enterobacteriaceae in-fections in particular (Fig. 6). For bothclinical infections and specific organisms,most groups trended in favor of combi-nation therapy.

DISCUSSION

In this retrospective, propensity-matched cohort study of septic shock, theuse of combination antibiotic therapy wasassociated with reduced 28-day mortality.Intensive care unit and hospital mortalitywere similarly reduced with combinationtherapy. The benefit of combination ther-apy was greatest with shorter periods be-tween administration of the two drugs.Mortality differences narrowed with in-creasing delay from the first to the sec-ond antibiotic (Table 2). A shorter periodbetween documentation of hypotensionand administration of the first drug wasassociated with improved survival but arelative reduction in the benefit of com-bination therapy (Table 2). Combinationtherapy was associated with more pres-sor-free and ventilator-free days, an in-crease in successful liberation of ventila-tory support, and discontinuation ofvasoactive medications. The benefit ofcombination therapy appeared to be as-sociated primarily with combinations of�-lactams with aminoglycosides, fluoro-quinolones, and macrolides/clindamycin(Fig. 4). However, the effect was consis-tent across most clinical infections andpathogens.

Although the use of combinations ofantibiotics is relatively common in septicshock, the rationale is typically to ensure

Figure 2. Adjusted Cox proportional hazards of mortality associated with combination antibiotictherapy of septic shock.


that the pathogenic organism is coveredby at least one active drug in the initialregimen (as per current guidelines andrecommendations) (32, 33). An effort toensure that the probable pathogen orpathogens are covered by more than oneactive agent (effective combination ther-apy) is not advised based on several meta-analyses that have failed to show a benefit

of combination therapy in severe infec-tions, including sepsis and Gram-nega-tive bacteremia (17, 18).

This study and a recently completedmeta-regression/meta-analysis (19) suggestthat combination antibiotic therapy may beadvantageous in septic shock. The basis onwhich combination therapy provides a sur-vival benefit at higher levels of mortality

risk can potentially be related to severalmechanisms. These include an increasedlikelihood that the infecting pathogen willbe susceptible to at least one of the com-ponents of the regimen; prevention ofemergence of resistant superinfection (34–36); potential beneficial immunomodula-tory effect of the secondary agent (37, 38);and generation of an additive or even syn-

Table 2. Mortality outcomes

SampleSize, n

Mortality Raten of Deaths/Total n of Patients

Hazard Ratio (95%Confidence Interval) pMonotherapy Combination Rx

28-Day mortalityUnadjusted 4662 1277/2948 (43.3%) 461/1714 (26.9%) 0.57 (0.51–0.63) �.0001Propensity score adjusted 2446 444/1223 (36.3%) 355/1223 (29.0%) 0.77 (0.67–0.88) .0002Matched cohort 28-day mortality analysis

stratified by Acute Physiology andChronic Health Evaluation II

5–20 755 66/366 (18.0%) 52/389 (13.4%) 0.72 (0.50–1.04) .0821–27 747 114/393 (39.0%) 78/354 (22.0%) 0.73 (0.55–0.98) .0328 818 242/398 (60.8%) 213/420 (50.7%) 0.78 (0.64–0.93) .007

Matched cohort 28-day mortalitystratified by delay betweendocumented hypotension and 1st

appropriate antibiotica

0.000–1.99 600 57/297 (19.2%) 57/303 (18.8%) 0.97 (0.67–1.40) .872–5.99 610 98/312 (31.4%) 69/298 (23.2%) 0.71 (0.52–0.96) .036 642 177/316 (56.0%) 150/326 (46.0%) 0.78 (0.62–0.96) .02

Matched cohort 28-day mortalitystratified by time from 1st to 2nd

appropriate antibiotic incombination therapya

0.000–1 1186 332/925 (35.8%) 67/261 (25.7%) 0.68 (0.53–0.89) .0041.001–4 1147 332/925 (35.8%) 68/222 (30.6%) 0.84 (0.65–1.09) .194.001–10 1147 332/925 (35.8%) 67/222 (30.2%) 0.82 (0.63–1.07) .14�10 1147 332/925 (35.8%) 74/222 (33.3%) 0.89 (0.69–1.15) .37

Matched cohort 28-day mortalitystratified by community/nosocomialinfection acquisition

Community 1663 278/841 (33.1%) 221/822 (26.9%) 0.79 (0.66–0.94) .008Nosomial 783 166/382 (43.5%) 134/401 (33.4%) 0.73 (0.58–0.91) .006

Matched cohort 28-day mortalitystratified by whether the initialappropriate antibiotic wasadministered before or after firstdocumented hypotension

Before 594 112/298 (37.6%) 79/296 (26.7%) 0.67 (0.50–0.89) .007After 1852 332/925 (35.9%) 276/927 (29.8%) 0.80 (0.69–0.94) .007

aExcluding those where first appropriate antibiotic was administered before hypotension.

Table 3. Mortality outcomes

SampleSize, n

Mortality Rate by Therapyn of Deaths/Total n of Patients

Odds Ratio (95%Confidence Interval) pMonotherapy Combination Rx

Intensive care unit mortality, n (%) 2446 437/1223 (35.2%) 352/1223 (28.8%) 0.75 (0.63–0.88) 0.0006Hospital mortality, n 2446 537/1223 (53.6%) 424/1223 (46.4%) 0.69 (0.59–0.81) �0.0001Death from:

Refractory shock 2446 311/1223 (25.4%) 258/1223 (21.1%) 0.78 (0.65–0.95) 0.01Sepsis-related organ failure 2446 184/1223 (15.0%) 137/1223 (11.2%) 0.71 (0.56–0.90) 0.005Nonsepsis-related organ failure 2446 89/1223 (7.3%) 62/1223 (5.1%) 0.68 (0.49–0.95) 0.02


ergistic antimicrobial effect of the combi-nation (i.e., more rapid killing of the patho-gen) (11, 15, 39–41).

With regard to this analysis, the pos-sibility of increased breadth of coverage isimmaterial. The study was restricted toculture-positive infections so that all pa-tients were known to have been treatedwith at least one antibiotic appropriate totheir primary pathogen. Although it isdifficult to rule out the possibility thatincreased late development of infectionswith resistant organisms could contrib-ute to increased mortality in the mono-therapy group, the magnitude of the ef-fect required to produce such a largechange in mortality appears improbable.

Figure 3. Primary antibiotic and outcome. The use of �-lactams as part of combination therapy wasassociated with reduced hazard ratio of death. A significant association with survival was seen with useof most penicillins and cephalosporins in such therapy. The benefit did not extend to antipseudomonal�-lactamase inhibitors, cephalosporins, and carbapenems. Combinations in which an antibiotic otherthan a �-lactam was the primary agent also did not show evidence of benefit. Monotherapy representsthe reference group. MT, monotherapy; CT, combination therapy; staph, Staphylococcus; gen, gen-eration; ceph, cephalosporin; Ps, pseudomonal; HR, unadjusted hazard ratio; CI, confidence interval.For antibiotics belonging to each group, refer to Table 1.

Figure 4. Supplemental antibiotic and outcome. The use of aminoglycoside, fluoroquinolone, or amacrolide/clindamycin in addition to a �-lactam was associated with a reduced hazard ratio for deathcompared to �-lactam alone. No other drug combinations demonstrated evidence of significantbenefit. Primary antibiotic group in italics. Supplemental agents listed with “” in normal text.Monotherapy with primary antibiotic represents reference. The numbers of deaths/total number ofcases in the primary monotherapy group and in each supplemental combination therapy group aredenoted. AG, aminoglycoside; FQ, fluoroquinolone; ML/CL, macrolide/clindamycin; HR, unadjustedhazard ratio; CI, confidence interval. For antibiotics belonging to each group, refer to Table 1.

Figure 5. Association of combination therapywith outcome stratified by clinical syndrome andcharacteristics. Among clinical syndromes, a sig-nificantly reduced hazard ratio for death was seenamong respiratory tract infections treated withcombination therapy; similar, but nonsignificant,trends were seen with septic shock attributable toother infections. In aggregate, all nonrespiratoryinfections also demonstrated a reduced hazardratio with combination therapy. Similar resultswere found when the analysis was restricted tobacteremic, nonbacteremic, septic shock patientsrequiring source control, and septic shock pa-tients not requiring source control. MT, mono-therapy; CT, combination therapy; PBSI, primarybloodstream infection (no apparent anatomicalsource); CRI, intravascular catheter-associatedinfection; RTI, respiratory tract infection; UTI,urinary tract infection; IAI, intra-abdominal in-fection; CNSI, central nervous system infection;SSTI, skin and soft tissue infection; SSI, surgicalsite infection; ITI, intrathoracic infection (exclu-sive of respiratory tract); SC inf, source controlrequiring infection; SC inf, nonsource controlrequiring infection; bact, infection associatedwith documented bacteremia; bact, infectionnot associated with documented bacteremia; HR,unadjusted hazard ratio; CI, confidence interval.

Table 4. Secondary outcomes

SampleSize Monotherapy

CombinedTherapy p

Within 30 daysNonmechanical intubation days,

median (interquartile range)2446 10 (0–25) 17 (0–26) .008

Nonvasoactive medication days,median (interquartile range)

2446 23 (0–28) 25 (0–28) .007

Days alive out of hospital 2446 0 (0–9) 0 (0–13.6) �.0001Days alive out of intensive care unit 2446 14 (0–24) 19 (0–25) .0003


Finally, immunomodulatory activity hasbeen ascribed to macrolides (38, 42–45)and, to a lesser extent, fluoroquinolones(37, 38, 46). Data supporting direct im-munomodulatory effects of aminoglyco-sides are very limited (47, 48). Even so,clinically relevant immunomodulatoryeffects of the secondary agent would seemto be improbable as a cause of the bene-ficial effect because such effects wouldseem unlikely to exist across all groups ofsecondary agents (aminoglycosides, fluoro-quinolones, and macrolides/clindamycin).In addition, an immunomodulatory basis ofbenefit of combination therapy might alsobe expected to extend across all �-lactamsrather than only the least powerful agents(in terms of standard pharmacokinetic in-dices, i.e., time above MIC).

Given the known benefit of synergistictherapy in certain infections, includingbacterial endocarditis, it would seemmore likely that the beneficial effect isrelated to faster clearance of organisms inmany of the combination therapy pa-tients. Antimicrobial synergy appears to

be best-established for �-lactam/amino-glycoside combinations (49 –52). How-ever, similar data on synergistic activityhave emerged for combinations of a�-lactam and fluoroquinolone (53–59).There are also data suggesting additiveeffects (60) or even potential synergism(61–64) for �-lactam/macrolides combi-nations in certain circumstances. Al-though speculative, synergistic antimi-crobial combinations may eradicate theunderlying bacterial pathogens morequickly than a single drug, resulting inmore rapid resolution of infection-associated physiologic instability, leadingto improved clinical cure and survival.

Interestingly, carbapenems (almostentirely imipenem/cilastatin and mero-penem), extended-spectrum �-lactam/�-lactamase inhibitor combinations, andanti-pseudomonal cephalosporins, whichtend to demonstrate optimal pharmaco-kinetic indices (with presumably maxi-mal kill rates) for most septic shockpathogens, yielded the weakest evidenceof benefit of combination therapy. Thiseffect cannot be explained on the basis ofimmune modulation but may be explain-able on the basis of a second agent’s in-ability to substantially further increasecidality when the most potent �-lactamsare part of a combination regimen.

As with any retrospective analysis, thisstudy has limitations and weaknesses thatmerit attention. There are at least twopotential time-dependent biases associ-ated with this study. Immortal time biasis, in essence, a mathematical problem inwhich survival duration may be linked toan inappropriate reference point yieldingan inaccurate time-dependent survivalprobability (30). A retrospective study ofcritically ill patients with a very highearly mortality risk may also demonstratea survival duration-related selection bias.In this case, the group of patients who areknown to have lived long enough to re-ceive a second antibiotic during septicshock may be selected to more likely liveto a given temporal end point than thegroup who cannot be known to have livedlong enough to receive the second drug.Several statistical approaches can be usedto control for immortal time bias. Oneconservative approach to control for bothtypes of potential biases is to ensure thatsubjects in both groups live long enoughto potentially have received the secondantibiotic. For that reason, each patientwith a combination therapy case wasmatched to a patient with monotherapycase who lived, at a minimum, to the

equivalent time point (after initial docu-mentation of hypotension) at which thepatient with the combination therapycase received the initial dose of the sec-ond appropriate antibiotic, i.e., thematched patient always lived longenough to have received the same or bet-ter opportunity to have been adminis-tered a second drug. In fact, because thematching of each monotherapy case pa-tient was limited to subjects who lived atleast 1 day past the point at which thecombination therapy case patients re-ceived the second drug (to be conserva-tive), a consistent survival bias in favor ofmonotherapy is created.

Despite sophisticated methods to ac-count for individual patient differences,retrospective studies also may be con-founded by indication. Some comorbidconditions may indicate or contraindicatethe addition of a second antibiotic andalso could be associated with clinical out-comes. For example, the use of aminogly-cosides and, less so, fluoroquinolonesmay be problematic in patients withacute renal failure who would also beexpected to have an increased mortalityrisk. Increased severity of illness mightotherwise increase the probability of ad-ministration of a second agent. Other bi-ases may be less obvious. We attemptedto control for potential confoundersthrough propensity matching of a varietyof variables (epidemiologic factors in-cluding hospital site, laboratory mea-sures, severity of illness, specific patho-gen, clinical infection syndrome, andantibiotic used). Nonetheless, it is possi-ble that residual confounders not re-corded in the dataset could exist. Propen-sity methods are unable to account forthese unknown factors. An obvious weak-ness of any retrospective design is thatthe allocation of patients and the use ofcombination therapy cannot be random-ized, nor can the antibiotic regimens bestandardized. Further, in this study, indi-cations for the use of combination ther-apy could not be clearly defined.

This study also has importantstrengths. A comprehensive clinical data-base allowed for the identification of alarge number of patients eligible for anal-yses in this study. Baseline differencesbetween the combination and mono-therapy groups existed, which could havethe potential to bias mortality estimates;however, the large sample size allowedfor a rigorously conducted propensitymatched analyses whereby patients weresuccessfully matched for �30 clinically

Figure 6. Association of combination therapywith outcome stratified by bacterial pathogen. Inaggregate, both Gram-positive and Gram-negative pathogens yielded evidence of reducedhazard ratio for death with combination therapy.Similarly, combination therapy of enterobacteri-aceae group pathogens was associated with im-proved survival, whereas nonenterobacteriaceaegroup pathogens trended in a similar direction.The only individual pathogen that yielded signif-icant evidence of benefit of combination therapywas Staphylococcus pneumoniae. GAS, group A�-hemolytic streptococci; nonGAS strep, �-he-molytic streptococci other than group A; sp, spe-cies; Ps, Pseudomonas; EB, Enterobacteriaceae;HR, unadjusted hazard ratio; CI, confidence in-terval. For organisms in each group, refer toTable 1.


relevant parameters. Although no retro-spective method can replace the advan-tage of prospective randomization, pro-pensity analyses have been demonstratedto be an effective means of reducing biasin baseline characteristics when assessingtreatment effects (28, 65). In this analy-sis, all significant baseline differences be-tween study groups were adequately rec-onciled using this method. The inclusionof patients from multiple hospitals with avariety of clinical infections and patho-gens adds further to the applicability ofthe findings as they relate to septic shock.

Our data strongly suggest that earlyempirical combination antibiotic therapywith two antibiotics of different mecha-nisms of action is associated with supe-rior outcomes than monotherapy in thetreatment of bacterial septic shock. Theseobservational data point to the need for alarge, randomized, controlled trial to ex-amine this issue in patients with septicshock.

ACKNOWLEDGMENTS

We thank Christine Mendez, SheenaAblang, Debbie Friesen, and Lisa Halsteadfor data entry.

REFERENCES

1. McCabe WR, Jackson GG: Gram-negativebacteremia II. Clinical, laboratory and ther-apeutic observations. Arch Intern Med 1962;110:856–864

2. Kreger BE, Craven DE, McCabe WR: Gram-negative bacteremia IV. Re-evaluation ofclinical features and treatment in 612 pa-tients. Am J Med 1980; 68:344–355

3. Kollef MH, Sherman G, Ward S, et al: Inad-equate antimicrobial treatment of infections:A risk factor for hospital mortality amongcritically ill patients. Chest 1999; 115:462–474

4. Parkins MD, Sabuda DM, Elsayed S, et al:Adequacy of empirical antifungal therapy andeffect on outcome among patients with inva-sive Candida species infections. J AntimicrobChemother 2007; 60:613–618

5. Kumar A, Ellis P, Arabi Y, et al: Initiation ofinappropriate antimicrobial therapy resultsin a five-fold reduction of survival in humanseptic shock. Chest 2009; 136:1237–1248

6. Chow JW, Yu VL: Combination antibiotictherapy versus monotherapy for gram-negative bacteraemia: A commentary. Int JAntimicrob Agents 1999; 11:7–12

7. Mufson MA, Stanek RJ, Mufson MA, et al:Revisiting combination antibiotic therapy forcommunity-acquired invasive Streptococcuspneumoniae pneumonia. Clin Infect Dis2006; 42:304–306

8. Patel SM, Saravolatz LD: Monotherapy ver-

sus combination therapy. Med Clin NorthAm 2006; 90:1183–1195

9. Paul M, Leibovici L: Combination antibiotictherapy for Pseudomonas aeruginosa bacter-aemia. Lancet Infect Dis 2005; 5:192–193

10. Weiss K, Tillotson GS: The controversy ofcombination vs monotherapy in the treat-ment of hospitalized community-acquiredpneumonia. Chest 2005; 128:940–946

11. Anderson ET, Young LS, Hewitt WL: Antimi-crobial synergism in the therapy of Gram-negative rod bacteremia. Chemotherapy1978; 24:45–54

12. De Jongh CA, Joshi JH, Newman KA, et al:Antibiotic synergism and response in gram-negative bacteremia in granulocytopeniccancer patients. Am J Med 1986; 80:96–100

13. Bouza E, Munoz P: Monotherapy versuscombination therapy for bacterial infections.Med Clin North Am 2000; 84:1357–1389

14. Darras-Joly C, Bedos JP, Sauve C, et al: Syn-ergy between amoxicillin and gentamicin incombination against a highly penicillin-resistant and -tolerant strain of Streptococ-cus pneumoniae in a mouse pneumoniamodel. Antimicrob Agents Chemother 1996;40:2147–2151

15. Calandra T, Glauser MP: Immunocompro-mised animal models for the study of antibi-otic combinations. Am J Med 1986; 80:45–52

16. Kumar A, Mensing J, Zelenitsky S, et al:Effect of antibiotic sequence on blood bacte-rial counts in a rat model of E. coli peritoni-tis/septic shock. ICAAC Proceedings, 2004,pp 26, A-1296

17. Safdar N, Handelsman J, Maki DG, et al: Doescombination antimicrobial therapy reducemortality in Gram-negative bacteraemia? Ameta-analysis. Lancet Infect Dis 2004;4:519–527

18. Paul M, Benuri-Silbiger I, Soares-Weiser K,et al: �lactam monotherapy versus �lactam-aminoglycoside combination therapy for sep-sis in immunocompetent patients: System-atic review and meta-analysis of randomisedtrials. BMJ 2004; 328:668–679

19. Kumar A, Safdar N, Reddy S, et al: The sur-vival benefit of combination antibiotic ther-apy for serious infections associated withsepsis and septic shock is contingent on therisk of death: A meta-analytic/meta-regres-sion study. 2010; (under review)

20. Bone RC, Balk R, Cerra FB, et al: ACCP/SCCM Consensus Conference: Definitions forsepsis and organ failure and guidelines foruse of innovative therapies in sepsis. Chest1992; 101:1644–1655

21. Kumar A, Roberts D, Wood KE, et al: Dura-tion of hypotension before initiation of effec-tive antimicrobial therapy is the critical de-terminant of survival in human septic shock.Crit Care Med 2006; 34:1589–1596

22. McGeer A, Campbell B, Emori TG, et al:Definitions of infection for surveillance inlong-term care facilities. Am J Infect Control1991; 19:1–7

23. Garner JS, Jarvis WR, Emori TG, et al: CDC

definitions for nosocomial infections, 1988.Am J Infect Control 1988; 16:128–140

24. Zarychanski R, Doucette S, Fergusson D, etal: Early intravenous unfractionated heparinand mortality in septic shock. Crit Care Med2008; 36:2973–2979

25. Bagshaw SM, Lapinsky S, Dial S, et al: Acutekidney injury in septic shock: Clinical out-comes and impact of duration of hypotensionprior to initiation of antimicrobial therapy.Intensive Care Med 2009; 35:871–881

26. Knaus WA, Draper EA: APACHE II: A severityof disease classification system. Crit CareMed 1985; 13:818–829

27. Glynn RJ, Schneeweiss S, Sturmer T: Indica-tions for propensity scores and review oftheir use in pharmacoepidemiology. BasicClin Pharmacol Toxicol 2006; 98:253–259

28. Luellen JK, Shadish WR, Clark MH: Propen-sity scores: An introduction and experimen-tal test. Eval Rev. 2005; 29:530–558

29. Parsons LS: SUGI 26: Reducing bias in apropensity score matched-pair sample usinggreedy matching techniques. SAS Institute,Cary, NC, 2001

30. Suissa S: Immortal time bias in observa-tional studies of drug effects Pharmacoepi-demiol Drug Saf 2007; 16:241–249

31. Breslow NE, Day NE: Statistical methods incancer research. Analysis of case-controlstudies. Volume 1. Lyon, France, Interna-tional Agency for Research on Cancer, IARCScientific Publications No. 32, 1980, pp138–146

32. Bochud PY, Glauser MP, Calandra T, et al:Antibiotics in sepsis, Intensive Care Med2001; 27(Suppl 1):S33–S48

33. Feldman C, Smith C, Levy H, et al: Klebsiellapneumoniae bacteraemia at an urban generalhospital. J Infect 1990; 20:21–31

34. Gribble MJ, Chow AW, Naiman SC, et al:Prospective randomized trial of piperacillinmonotherapy versus carboxypenicillin-aminoglycoside combination regimens in theempirical treatment of serious bacterial in-fections. Antimicrob Agents Chemother1983; 24:388–393

35. EORTC International Anti-microbial TherapyCooperative Group: Ceftazidime combinedwith a short or long course of amikacin forempirical therapy of gram-negative bactere-mia in cancer patients with granulocytope-nia. The EORTC International AntimicrobialTherapy Cooperative Group. N Engl J Med1987; 317:1692–1698

36. Milatovic D, Braveny I: Development of re-sistance during antibiotic therapy. Eur J ClinMicrobiol 1987; 6:234–244

37. Pasquale TR, Tan JS: Nonantimicrobial ef-fects of antibacterial agents. Clin Infect Dis2005; 40:127–135

38. Labro MT: Interference of antibacterialagents with phagocyte functions: immuno-modulation or “immuno-fairy tales”? ClinMicro Rev 2000; 13:615–650

39. De Jongh CA, Joshi JH, Thompson BW, et al:A double �lactam combination versus anaminoglycoside-containing regimen as em-


piric antibiotic therapy for febrile granulocy-topenic cancer patients. Am J Med 1986; 80:101–111

40. Klastersky J, Zinner SH: Synergistic combi-nations of antibiotics in gram-negative bacil-lary infections. Rev Infect Dis 1982;4:294–301

41. Giamarellou H: Aminoglycosides plus beta-lactams against gram-negative organisms.Evaluation of in vitro synergy and chemicalinteractions. Am J Med 1986; 80:126–137

42. Guchelaar HJ, Schultz MJ, van D, et al: Phar-macokinetic-pharmacodynamic modeling ofthe inhibitory effect of erythromycin on tu-mour necrosis factor-alpha and interleukin-6production. Fund Clin Pharmacol 2001; 15:419–424

43. Labro MT: Cellular and molecular effects ofmacrolides on leukocyte function. CurrPharm Des 2004; 10:3067–3080

44. Khan AA, Slifer TR, Araujo FG, et al: Effect ofclarithromycin and azithromycin on produc-tion of cytokines by human monocytes. Int JAntimicrob Agents 1999; 11:121–132

45. Oda H, Kadota J, Kohno S, et al: Erythromy-cin inhibits neutrophil chemotaxis in bron-choalveoli of diffuse panbronchiolitis. Chest1994; 106:1116–1123

46. Khan AA, Slifer TR, Remington JS: Effect oftrovafloxacin on production of cytokines byhuman monocytes. Antimicrob Agents Che-mother 1998; 42:1713–1717

47. Goscinski G, Lundholm M, Odenholt I, et al:Variation in the propensity to release endo-toxin after cefuroxime exposure in differentgram-negative bacteria: Uniform and dose-dependent reduction by the addition of to-bramycin. Scand J Infect Dis 2003; 35:40–46

48. Goscinski G, Lipcsey M, Eriksson M, et al:Endotoxin neutralization and anti-inflamma-tory effects of tobramycin and ceftazidime inporcine endotoxin shock. Crit Care 2004;8:R35–R41

49. Comber KR, Basker MJ, Osborne CD, et al:Synergy between ticarcillin and tobramycinagainst Pseudomonas aeruginosa and Enter-obacteriaceae in vitro and in vivo. Antimi-crob Agents Chemother 1977; 11:956–964

50. Archer G, Fekety FR Jr: Experimental endo-carditis due to Pseudomonas aeruginosa II.Therapy with carbenicillin and gentamicin.J Infect Dis 1977; 136:327–335

51. Yoshikawa TT, Shibata SA: In vitro antibac-terial activity of amikacin and ticarcillin,alone and in combination, against Pseudo-monas aeruginosa. Antimicrob Agents Che-mother 1978; 13:997–999

52. Kluge RM, Standiford HC, Tatem B, et al:Comparative activity of tobramycin, amika-cin, and gentamicin alone and with carbeni-cillin against Pseudomonas aeruginosa. An-timicrob Agents Chemother 1974; 6:442–446

53. Gimeno C, Borja J, Navarro D, et al: In vitrointeraction between ofloxacin and cefotaximeagainst gram-positive and gram-negativebacteria involved in serious infections. Che-motherapy 1998; 44:94–98

54. Gradelski E, Kolek B, Bonner DP, Valera L, et

al: Activity of gatifloxacin and ciprofloxacinin combination with other antimicrobialagents. Int J Antimicrob Agents 2001; 17:103–107

55. Zembower TR, Noskin GA, Postelnick MJ, etal: The utility of aminoglycosides in an era ofemerging drug resistance. Int J AntimicrobAgents 1998; 10:95–105

56. Neu HC: Synergy and antagonism of fluoro-quinolones with other classes of antimicro-bial agents. Drugs 1993; 45(Suppl 3):54–58

57. Pohlman JK, Knapp CC, Ludwig MD, et al:Timed killing kinetic studies of the interac-tion between ciprofloxacin and beta-lactamsagainst gram-negative bacilli. Diagnostic Mi-crobiol Infect Dis 1996; 26:29–33

58. Milatovic D, Wallrauch C: In vitro activity oftrovafloxacin in combination with ceftazi-dime, meropenem, and amikacin. Eur J ClinMicrobiol Infect Dis 1996; 15:688–693

59. Pankuch GA, Lin G, Seifert H, et al: Activityof meropenem with and without ciprofloxa-cin and colistin against Pseudomonas aerugi-nosa and Acinetobacter baumannii. Antimi-crob Agents Chemother 2008; 52:333–336

60. Lin E, Stanek RJ, Mufson MA: Lack of syn-ergy of erythromycin combined with penicil-lin or cefotaxime against Streptococcuspneumoniae in vitro. Antimicrob AgentsChemother 2003; 47:1151–1153

61. MacGowan AR, Bowker K, Bedford KA, et al:Synergy testing of macrolide combinationsusing the chequerboard technique. J Antimi-crob Chemother 1993; 32:913–915

62. Furuya R, Nakayama H, Kanayama A, et al: Invitro synergistic effects of double combinationsof beta-lactams and azithromycin against clin-ical isolates of Neisseria gonorrhoeae. J InfectChemother 2006; 12:172–176

63. Maioli E, Debbia EA, Gualco L, et al: In vitrointeraction between mecillinam and pipera-cillin-tazobactam in the presence of azithro-mycin against members of the Enterobacte-riaceae family and Pseudomonas aeruginosa.New Microbiol 2008; 31:37–46

64. Allen NE, Epp JK: Mechanism of penicillin-erythromycin synergy on antibiotic-resistantStaphylococcus aureus. Antimicrob AgentsChemother 1978; 13:849–853

65. Connors AF Jr, Speroff T, Dawson NV, et al:The effectiveness of right heart catheteriza-tion in the initial care of critically ill pa-tients. JAMA 1996; 276:889–897

APPENDIX 1

Rules to assign clinical significance tomicrobial isolates:

1) Clinically significant isolates from ei-ther local site and/or blood cultureswere required to have been obtainedwithin 48 hrs of onset of shock.

2) The following were considered to rep-resent clinically significant isolates:a) Positive blood culture for any patho-

gen other than coagulase-negative

staphylococci or other skin contam-inants.

b) Any growth from a normally ster-ile site (e.g., gall bladder, bron-chial lavage, peritoneal, pleuralfluid, operative tissue specimen)apart from coagulase-negativestaphylococci other skin contami-nants.

c) Growth of a pathogen in sputumsample of a patient with respira-tory signs and symptoms or a newinfiltrate on chest radiography,with no other likely source of in-fection.

d) Growth of a pathogen in a urinesample (�108 organisms/L) witheither local clinical symptoms orin the absence of a more plausibleclinical infection site.

e) Growth from a deep biopsy or adeep aspirate of a finding in softtissue or skin.

f) Concurrent, congruent, positive,semiquantitative catheter coloni-zation (�15 colonies) with bloodculture or clinical evidence of siteinfection.

g) A positive direct measurement ofStreptococcus pneumoniae, Neis-seria meningitidis, or Haemophi-lus influenzae antigen in the spu-tum or cerebrospinal fluid.

3) Enterococci were considered to be clin-ically significant only in the absence ofother more plausible pathogens.

4) Staphylococcus epidermidis was uni-formly considered to be incapable ofcausing septic shock. Other coagu-lase-negative staphylococci similarlywere considered to be unlikely tocause septic shock unless present as asole isolate in multiple blood culturesin the absence of evidence of endovas-cular infection.

APPENDIX 2

Rules regarding antimicrobial therapy

1) The following organisms were consid-ered sensitive to listed antibiotics (i.e.,appropriate therapy) even in the ab-sence of specific sensitivity testing:group A, B, and G streptococcus to all�-lactams; all Gram-positive bacteriaexcept enterococci to vancomycin;and organisms to �-lactamase inhibi-tor combinations if sensitive to the�-lactam alone.

2) The following organisms were consid-ered resistant to listed antibiotics (i.e.,


inappropriate therapy): Enterococci toall cephalosporins and trimethoprim/sulfamethoxazole and Enterococcusfaecalis to quinupristin-dalfopristin.

3) Clindamycin, macrolides, and third-generation cephalosporins were notconsidered appropriate for Staphylo-coccus aureus septic shock irrespec-tive of listed sensitivity.

4) Cefotaxime and ceftriaxone were notconsidered appropriate therapy ofPseudomonas aeruginosa irrespectiveof listed sensitivity.

5) Treatment with bacteriostatic antibi-otics was considered appropriate ther-apy only in the case of sensitive en-terococci treated with a penicillin orcarbapenem and for any pathogen forwhich no commonly available bacteri-cidal primary or supplemental agentsexisted.

6) Use of aminoglycosides as the supple-mental/secondary antibiotic for treat-ment of group A streptococci (Staph-ylococcus pyogenes) or viridansstreptococci infections with an appro-priate �-lactam or vancomycin wasconsidered to be appropriate combina-tion therapy even in the absence ofspecific sensitivity data.

7) In case of multiple isolates at a localsite, appropriate therapy was consid-ered to have been delivered if thedensest pathogen was covered. If mul-tiple pathogens were isolated at simi-lar density, then all pathogens wererequired to have been covered.

8) For multiple simultaneous blood iso-lates, appropriate therapy had to coverall pathogens.

APPENDIX 3

Additional Members of the Coopera-tive Antimicrobial Therapy of SepticShock (CATSS) Database Research Groupinclude: Kenneth E. Wood, MD, Univer-sity of Wisconsin Hospital and Clinics,

Madison, WI, USA; Kevin Laupland, MD,Foothills Hospital, Calgary AB, Canada;Andreas Kramer, MD, Brandon GeneralHospital, Brandon MB, Canada; CharlesPenner, MD, Brandon General Hospital,Brandon MD, Canada; Bruce Light, MD,Winnipeg Regional Health Authority,Winnipeg MB, Canada; Satendra Sharma,MD, Winnipeg Regional Health Author-ity, Winnipeg MB, Canada; Steve Lapin-sky, MD, Mount Sinai Hospital, TorontoON, Canada; John Marshall, MD, St. Mi-chael’s Hospital, Toronto ON, Canada;Sandra Dial, MD, Jewish General Hospi-tal, Montreal QC, Canada; Sean Bagshaw,MD, University of Alberta Hospital, Edm-onton AB, Canada; Ionna Skrobik, MD,Hopital Maisonneuve Rosemont, Mon-treal QC, Canada; Gourang Patel,PharmD, Rush-Presbyterian-St. Luke’sMedical Center, Chicago IL, USA; DaveGurka, MD, Rush-Presbyterian-St. Luke’sMedical Center, Chicago, IL, USA; SergioZanotti, MD, Cooper Hospital/UniversityMedical Center, Camden, NJ, USA; PhillipDellinger, MD, Cooper Hospital/Univer-sity Medical Center, Camden, NJ, USA;Dan Feinstein, MD, St. Agnes Hospital,Baltimore, MD, USA; Jorge Guzman, MD,Harper Hospital, Detroit, MI, USA; NehadAl Shirawi, MD, King Abdulaziz MedicalCity, Riyadh, Saudi Arabia; Ziad Al Mem-ish, MD, King Abdulaziz Medical City,Riyadh, Saudi Arabia; John Ronald, MD,Nanaimo Regional Hospital, NanaimoBC, Canada.

Associate Members of the CATSS Data-base Research Group include: MustafaSuleman, MD, Concordia Hospital, Win-nipeg, MB Canada; Harleena Gulati, MD,University of Manitoba, Winnipeg, MB Can-ada; Erica Halmarson, MD, University ofManitoba, Winnipeg MB, Canada; RobertSuppes, MD, University of Manitoba, Win-nipeg MB, Canada; Cheryl Peters, Univer-sity of Manitoba, Winnipeg MB, Canada;Katherine Sullivan, University of Manitoba,Winnipeg MB, Canada; Rob Bohmeier, Uni-

versity of Manitoba, Winnipeg MB, Canada;Sheri Muggaberg, University of Manitoba,Winnipeg MB, Canada; Laura Kravetsky,University of Manitoba, Winnipeg MB, Can-ada; Muhammed Wali Ahsan, MD, Win-nipeg MB; Canada, Amrinder Singh, MD,Winnipeg, MB Canada; Lindsey Carter, BA,Winnipeg MB, Canada; Kym Wiebe, RN, St.Boniface Hospital, Winnipeg MB, Canada;Laura Kolesar, RN, St. Boniface Hospital,Winnipeg MB, Canada; Jody Richards,Camosun College, Victoria BC, Canada;Danny Jaswal, MD, University of British Co-lumbia, Vancouver BC, Canada; HarrisChou, BSc, of British Columbia, VancouverBC, Canada; Tom Kosick, MD, University ofBritish Columbia, Vancouver BC, Canada;Winnie Fu, University of British Columbia,Vancouver BC, Canada; Charlena Chan,University of British Columbia, VancouverBC, Canada; Jia Jia Ren, University of Brit-ish Columbia, Vancouver BC, Canada; Moz-deh Bahrainian, MD, Madison WI, ZiaulHaque, MD, Montreal QC, Canada; OmidAhmadi Torshizi, MD, Montreal QC, Can-ada; Heidi Paulin, University of Toronto,Toronto ON, Canada; Farah Khan, MD, To-ronto ON, Canada; Runjun Kumar, Univer-sity of Toronto, Toronto ON, Canada; Jo-hanne Harvey, RN, Hopital MaisonneuveRosemont, Montreal QC, Canada; ChristinaKim, McGill University, Montreal QC, Can-ada; Jennifer Li, McGill University, Mon-treal QC, Canada; Latoya Campbell, McGillUniversity, Montreal QC, Canada; LeoTaiberg, MD, Rush Medical College, Chi-cago, IL, USA; Christa Schorr, RN, CooperHospital/University Medical Center, Cam-den, NJ, USA; Ronny Tchokonte, MD,Wayne State University Medical School, De-troit, MI, USA; Catherine Gonzales, RN,King Abdulaziz Medical City, Riyadh, SaudiArabia, Norrie Serrano, RN, King AbdulazizMedical City, Riyadh, Saudi Arabia, SofiaDelgra, RN, King Abdulaziz Medical City,Riyadh, Saudi Arabia.


early combination antibiotic therapy yields improved ......early combination antibiotic therapy...

Documents