- Email: [email protected]
Antibiotic therapy for Pseudomonas aeruginosa bacteremia: Outcome correlations in a prospective study of 200 patients
Antibiotic Therapy for Pseudomonas aerugjnosa Bacteremia: Outcome Correlations in a Prospective Study of 200 Patients MEGAN HILF, M.s., VICTOR L. Yu, M.D., JOANN SHARP, B.S., JEFFREYJ. ZURAVLEFF, M.D., JOYCE A. KORVICK, M.D., ROBERT R. MUDER, M.D. Pittsburgh, Pennsylvania
PURPOSEAND PATIENTSAND MJTl’HODs:We performed a prospective clinical study of 200 consecutive patients with Pseudomonasaeruginosa bacteremias to analyze iu V&O susceptibility and synergistic testing of antibiotics the patients received and clinical parameters to assesstheir relationship to SurviVaL RESULTS:No significant correlation between in vitro susceptibility testing (minimal inhibitory concentrations/minimal bactericidal concentrations) and outcome could be demonstrated. Similarly, improved outcome could not be demonstrated for patients receiving antibiotic combinations that were synergistic in vitro (either time-kill or checkerboard) versus those combinations that were not. There was also no correlation between results obtained by time-kill curve and checkerboard synergistic testing, i.e., combinations found to be synergistic by one method were not necessarily synergistic by the other method. Clinical parameters associated with improved survival were a urinary portal of entry and absence of neutropenia. Conversely, survival was significantly decreased when the portal was the respiratory tract. The mortality rate between patients receiving combination therapy (27% ) and monotherapy (47% ) was significant (p cO.02); this significant relationship held true for most subgroups including malignancy, nosocomial infection, and infection site. CONCLUSION: Increasing effort should be placed on ensuring timely administration of combination therapy to patients with I? aeruginosa bacteremia since the use of combination therapy was even more important in determining outcome than was underlying disease.
From the Infectious Disease and Special Pathogens Sections, Veterans Administration Medical Center and University of Pittsburgh, Pittsburgh, Pennsylvania. Requests for reprints should be addressed to Victor L. Yu. M.D., University of Pittsburgh. School of Medicine, Scaife 968, Pittsburgh, Pennsylvania 15261. Manuscript submitted August 15.1988. and accepted in revised form July 10. 1989. For a copy of the 22-page Appendix cited in this article, please enclose a stamped, self-addressed, large (at least 6 X IO%-inch) envelope.
540
November
1989
The American
Journal
of Medicine
Volume
seudomonas aeruginosa continues as a cause of challenging infections, most notably in the comP promised and hospitalized patient. Mortality due to P. aeruginosa bacteremia remains ominously high [l-9]; prognosis is consistently poor for patients with leukemia and lymphoma (mortality of 67%to 100%) [1,4,9]. As a result, in vitro tests have been touted as a guide for choosing optimal antibiotic therapy in patients infected by P. aeruginosa. However, reproducibility and interpretations remain controversial, correlation of the results of in vitro antibiotic susceptibility and synergistic testing and clinical outcome has been notably absent. Hence, it has become almost reflex for investigators to acknowledge the need for such correlation in any publication reporting in uitro results. Therefore, we initiated a prospective, multihospital study of P. aeruginosa bacteremias. We made a concerted effort to addressthose areasof weaknessesthat have prevented a definitive resolution to this ongoing controversy, including prospective rather than retrospective study, adequate sample size for statistical analysis (200 patients), definitive evidence of infection (bacteremia), selection of a single organism (P. aeruginosa) rather than “gram-negative” organisms, blinded performance of in vitro tests without knowledge of outcome, and use of objective endpoints (survival at fixed number of days). Clinical parameters, ‘together with in uitro susceptibility and synergistic test results for antibiotics the patient received, were assessedin their relationship to patient survival. PATIENTS AND METHODS Clinical Study Two hundred consecutive patients from whom P. aeruginosa was isolated in blood culture were followed prospectively from June 1982 to June 1986, at eight Pittsburgh area hospitals. Organisms were isolated and identified at the respective clinical microbiologic laboratories and transported to a central laboratory. Susceptibility testing was performed without knowledge of clinical outcome. Data collected prospectively included patient demographics, hospital course, and antibiotics administered. All treatment decisionswere made by the attending physicians without consultation or intervention by the investigators except to ensure that aminoglycoside serum concentrations were measured. In Vitro Antibiotic Susceptibility Testing The antipseudomonal agent(s) selected for in uitro testing was the antipseudomonal agent that the patient received for the bacteremia. If a number of antipseudomonal agents were used over the course of the bacteremia, the agent(s) selected for in uitro testing wasthe agent received for at least two days within the first three days of bacteremia. Susceptibility testing 87
ANTIBIOTICS
was performed by microdilution techniques as previously described [lo]. Dilutions of antibiotics were made in Mueller-Hinton broth supplemented with 60 mg/L of calcium and 20 mg/L of magnesium. The inoculum was 5 x lo5 to lo6 CFU (colony-forming units)/mL. Each set of minimum inhibitory concentration (MIC) and minimum bactericidal concentration (MBC) calculations were performed in duplicate on different days. In Vitro Synergisitic Testing Synergistic testing was performed on those isolates taken from patients who received combination therapy. A breakdown of antibiotic combinations prescribed is given in the Results section. CHECKERBOARDTECHNIQUE: Druginteractionswere determined in an 8 X 8 column pattern using 8 X 12 microtiter plates. The inoculum was 5 X lo5 to lo6 CFU/mL. The fractional inhibitory concentration index (FIC) and fractional bactericidal concentration index (FBC) were calculated as previously described
Pll.
KILL CURVES: Time-kill curves were performed in 25 X loo-mm tubes as previously described [12]. An inoculum size of 5 X lo5 to lo6 CFU/mL was used. The final concentration of the two agents used was one fourth of the previously determined MBC for those agents. Quantitation of growth was measured from samples taken at four and 24 hours and inoculated onto a 100 X 15-mm agar plate using a spiral plater (Spiral Systems, Inc., Bethesda, Maryland). The interaction was defined as synergistic if colony counts using the combination showed a greater than or equal to loo-fold decrease when compared with the single most active agent. Complete technical details of susceptibility testing, checkerboard methodology, time-kill curve method, and experiments in reproducibility are given in a detailed manual contained in an Appendix available upon request from the authors.
Statistical Methods Patient and laboratory data were entered into a computer databank (Prophet System, Division of Research Resources, National Institutes of Health, Bethesda, Maryland). Fisher’s exact test or the chisquare test (two-tailed) was used to assessstatistical significance for various outcome measures. The Mann-Whitney rank sum test was used to assessthe association of neutrophil count and outcome. A stepwise logistic regressionanalysis was used for multivariate analysis (BMDP, University of California). The dependent binary variable was mortality at 10 days, and the independent variables were the prospectively monitored clinical factors and the in vitro test results. A 30-day survival curve was calculated for each therapy using a Kaplan-Meier estimate [13]. The differencesbetween the curves were then analyzed using a Mantel-Cox log rank [14] and the Gehan-Breslow test
[151RESULTS Patient Demographics The study population was composedof 200 patients with a mean age of 59 (range: one week to 96 years). Twenty-three percent (46 of 200) and 77%(154 of 200) of patients had community-acquired and nosocomial
AND PSEUDOMONASBACTEREMIA
/ HILF ET Al.
bacteremia, respectively. Bacteremias were classified asnosocomial if they occurred after 72 hours of hospitalization; for these patients, the mean was 20 days of hospitalization before development of bacteremia. Thirty-four percent (67 of 200) of the patients were in intensive-care units when P. aeruginosa bacteremia developed. Fifty-eight percent (115 of 200) of the patients were deemed to be immunosuppressed at the time of bacteremia by the following criteria: cytotoxic drugs (69 patients), corticosteroid therapy (80 patients with 88% receiving the equivalent of 20 mg of prednisone or greater), radiation therapy (26 patients), hematologic malignancy (48 patients), and neutropenia (neutrophil count lessthan 3,OOO/mL)(53 patients); of the 53 neutropenic patients, 27 had neutrophil counts of less than lOO/mL. Eighty-four percent (168 of 200) of the patients had undergone invasive procedures within seven days before the onset of bacteremia. Eighty-six percent (144 of 168) had placement of intravascular catheters, 55% (110 of 168) had urinary tract catheters, 45% (76 of 168) required mechanical ventilatory assistance, and 34% (57 of 168) had undergone surgery within seven days prior to the positive blood culture result. Microbiology EXTRA-BLOOD SITES: P. aeruginosa was cultured concomitantly from urine (29%, 58 of 200 patients), respiratory tract (34%, 67 of 200 patients), wound (15%, 29 of 200 patients), and tips of indwelling vascular catheters (14%, 28 of 200 patients). P. aeruginosa was also isolated from other sites: abdominal drainage (six) and biliary drainage (eight), aswell asaspirates of kidney (one), liver (one), and bone marrow (two).
Outcome Seventy-six percent (151 of 200) of the patients were alive at five days and 59% (117 of 200) were alive 10 days after isolation of P. aeruginosa from blood culture. Survival, for the purposes of this study, was arbitrarily defined as surviving 10 days after the onset of bacteremia, whereas mortality was defined as death within 10 days after the onset of bacteremia. The correlation for other endpoints including outcome at five days, 14 days, or at time of death/discharge was esentially identical to the lo-day endpoint (data in Appendix available from authors upon request). Clinical Parameters and Outcome There was a significant association between increased mortality and presence of neutropenia (neutrophil count less than 3,OOO/mL) (p <0.05, Fisher’s exact test, data not shown) in the 143 patients receiving combination therapy. Site of infection was significantly associated with outcome. Patients whose portal of entry was the urinary tract had a significantly improved survival rate compared with those patients with non-urinary tract portals (p = 0.001, data not shown). Patients with pneumonia as the primary site had a significantly higher mortality rate than patients without pneumonia (p cO.05, Appendix). Since severity of illness at the time of bacteremia was expected to be a prognostic factor that could confound the effect of antipseudomonal antibiotic therapy, we also controlled for this factor. Patients were deemed “critically ill” if they were receiving mechani-
November
1999 The American
Journal
of Medlclne
Volume
87
541
ANTIBIOTICS
AND PSEUDOMONAS
BACTEREMIA
/ HILF ET AL
TABLE I Clinical Parameters and Outcome in 143 patients with Bacteremia Receiving Combination Therapy (Patients Receiving Monotherapy or No Antlpseudomonal Agent . Therapy Are Excluded from This Analysis) Numberof Patients
Clinical Parameter
Percent Mortality
‘26
Neutrophils3,OOO/mL
31 (a/26) 47 (8/17) 22 (22/100)
NS (0.08)*
21(13/63) 28(16/57) 39 (9/23)
NS’
47 (18/37) 19 (20/ 106)
0.001t
1:; 20 16 7 100
35 (7/20) 13 (2/16) 43 (3/7) 26 (26/100)
NS (0.09)’
1:;
McCabe criteria [57] Non-fatal Ultimately fatal Rapidly fatal
63 ::
Portal5 Pneumonia Urinary tract Intravascular Other
S = not significant. Chi-square test (two-tatled). ‘isher exact test (two-tailed). \s defined by mechanical respiratory assistance, or coma. s Portal defined by physician using clinical criterra simultaneously from site of infection.
TABLE II Correlation Outcome
p Value
acute
hypotensive
episode.
and Pseudomonas
isolated
of Results of in Vitro Synergy Testing with
trations according to the literature, and the MC. NO correlation could be demonstrated for MICs or for MBCs (Appendix). Only 7% (nine of 131) of organisms were resistant to the beta-lactam prescribed and 4% (seven of 158) were resistant to the aminoglycoside prescribed. Combination therapy was administered to 143 patients and synergy was evaluated by checkerboard assay and time-kill curve for 123 bacteremic organisms involved (18 organisms were not available for testing and two others were excluded due to the combination received). Results of the checkerboard assaywere analyzed with two break points: FIC lessthan or equal to 0.50 (17 organisms) (Table II) and FIC less than or equal to 1.0 (110 organisms). An FIC lessthan or equal to 0.5 denotes that the MICs of the two agents decreasefourfold or more when the agents are combined, an arbitrary but well-accepted criterion for synergy. This takes into account an inherent twofold dilution error accepted for the method. An FIC less than or equal to 1.0 denotes that the MICs of the two agents decrease twofold or more when the agents are combined, a strict mathematical interpretation of synergy [17]. Results from 71 time-kill curves fulfilled the criterion for synergy. No significant correlation with patient outcome was seen (p = 0.10) (Table II). There was also no correlation between results obtained by time-kill curve and checkerboard assay (using either FIC less than or equal to 0.50 or FIC less than or equal to 1.0 as the criterion for checkerboard synergy) (p >0.05, Fisher’s exact test). Concordance for synergy and nonsynergy by the two methods was only 49% (Appendix).
Microbiologic Parametersand Outcome In vitro susceptibility testing (MICs) could not be correlated with patient outcome. Since achievable serum concentrations of each antibiotic are also pertinent, we calculated the therapeutic index [16], defined asthe ratio of expected peak antibiotic serum concen-
Antibiotics and Patient Outcome Ninety-three percent (186 of 200) of the patients received antipseudomonal antibiotics. Ninety percent (179 of 200) received an aminoglycoside: gentamicin (50 patients), tobramycin (108), or amikacin (21). Seventy-five percent (149 of 200) received a beta-la&am agent (one patient received two beta-lactam agents): piperacillin (72), ticarcillin (49), mezlocillin (ll), cefoperazone (nine), carbenicillin (five), imipenem (one), cefotaxime (two), and ceftazidime (one). Seventy percent (143 of 200) received combination therapy; in 142 patients, therapy consisted of an antipseudomonal beta-la&am agent plus an aminoglycoside and, in one patient, therapy wasmezlocillin plus ceftazidime. The most common combinations given were piperacillin plus tobramycin (25%, 36 of 143) and ticarcillin plus tobramycin (24%, 35 of 143). Seven percent (14 of 200) received no antipseudomonal antibiotic therapy; all of these patients died within 10 days of the onset of bacteremia. The most striking factor associated with patient outcome was the receipt of combination therapy versus monotherapy (Table III). The 143 patients who received a combination of an aminoglycoside and an antipseudomonal beta-la&am agent had a 27% (38 of 143) mortality rate as compared with the patients receiving monotherapy, who had a mortality rate of 47% (20 of 43) (p = 0.023). The trend for improved survival in patients given combination therapy was also seen for every patient subgroup studied, although not all were statistically significant (Table III). Thirty-day survival curves were also significantly different for the two treatment groups (Appendix). The probability of survival during the 30-day period
542
87
Percent Dled Checkerboard Synergistict Nonsynergistic Kill curve Synergistic Nonsynergistic
Percent Survived
p Value*
3 (4/123) 24 (29/123)
11 (13/123) 63 (77/123)
NS
12(15/123) 15(18/123)
46 (56/123) 28 (34/123)
NS (0.10)
NS = not significant. * Fisher exact test (two-tailed). r Defined as FIC index 50.5. If synergy was defined value still would not have been significant.
as FIC Index
51.0.
the p
cal respiratory assistance, had experienced an acute hypotensive episodewithin 72 hours prior to the positive blood culture result, or were comatose.As expected, patients who were “critically ill” had a significantly lower survival rate than those who were not critically ill (p
November
1999
The American
Journal
of Medicine
Volume
ANTIBIOTICS
AND PSEUDOMONASBACTEREMIA
/ HILF ET AL
TABLE III Mortality Rates for Patient Subgroups Receiving Combination Combination Therapy Was Superior to Monotherapy)
Patient Subgroup All patients Nosocomial origin Pneumonia Critically ill* lmmunosuppressed Malignancy5 Non-critically ill Neutropenia (<3,0OO/mL)
Number of Patients
Antibiotic
Therapy and Monotherapy
Percent Mortality Combination
186t 143 :9” 107 85 137 49
27 (38/143) 32 (34/108) 35 (7/iO) ’ 47 (18/37) 33 (28/86) 32 (21/66) 19 (20/106) 37 (16/43)
(For All Subgroups,
Percent Mortality Single Agent
p Value*
47 (20/43) 5Uw;8/;5)
0.023 0.04 0.033 0.016 NS NS
92(11/U) 48 (10/21) 47 (9/19)
2
Ki = not significant. ’ Fisher exact test (two-talled). t Fourteen patients received no antipseudomonal antibiotics (since therapy was often empiric prior to knowledge t “Critically ill” was defined as mechanical ventilatory support. acute hyootensive episode, or coma. . 5 Includes hematologic malignancies and solid tumors.
of Pseudomonas
in the blood
culture);
all died.
more specific area of P. aeruginosa infections, controversy exists regarding the necessity for double combination antibiotic therapy versus monotherapy [1,20221.The techniques commonly used to measuresynergy are the checkerboard and time-kill curve methods, and each in oitro method has its supporters [23]. Advocates of the checkerboard method cite its likenessto single-agent susceptibility testing and consider it a more practical assay to perform and interpret [24]. Advocates of time-kill curve methodology point out that it represents a more dynamic picture of bacterialantimicrobial interaction [25]. Comparisons of the methodologies are few and contradictory [26,27]. In addition, it is agreed that performance of these tests requires considerable technical experience, and standardization of the methods has been insufficient [24,25,2%30]. Multivariate Analysis The issueof synergy has been further explored in in Data for all 200patients were analyzed by a stepwise logistic regression model (BMDP, University of Cali- uiuo animal studies but results have been conflicting. Combination antibiotic therapy hasimproved survival fornia). The dependent variable was outcome (survival or death at 10 days), and the independent variables in normal and neutropenic animals infected with P. were the prospectively monitored risk factors of noso- aeruginosa (31-381. If the combination were synergistic in vitro, survival was further enhanced. On the comial versus community-acquired infection, neutrophi1 count, urinary portal (yes or no), respiratory por- other hand, results from other animal studies have failed to show significant improvements in outcome tal (yes or no), and antibiotic therapy (antipseudomonal beta-lactam single agent or amino- for combination therapy versus monotherapy [39-42]. Human studies are also conflicting, with studies supglycoside single agent versus combination therapy). The probabilities to enter and remove variables were porting both the superiority of two antibiotics [7,43set at 0.10 and 0.15, respectively. Combination antibi49] and the efficacy of a single antibiotic [21,50,51]. otic therapy, urinary tract portal, and absenceof neu- Use of combinations that are synergistic in vitro has tropenia were found to be significantly associatedwith also been credited with improved survival in immunosurvival; the p values to remove were 0.004,0.052, and suppressed patients [44,52-541. 0.03, respectively. A subset of the study group who A review of the clinical studies suggests possible received combination antibiotic therapy (143 pa- explanations for the conflicting results. In some studtients) was also evaluated with the stepwise logistic ies, the definition of combination therapy was not regression analysis. The aforementioned clinical fac- clearly stated. Combinations considered non-synergistors of nosocomial versus community-acquired infec- tic included aminoglycoside and beta-la&am agents tion, neutrophil count, urinary portal, and respiratory without antipseudomonal activity (i.e., cephalothin). portal were entered plus the following measuresof in. It has been cogently argued that this regimen is not a uitro interaction of antibiotic combinations: FIC, true combination but equivalent to single-agent amiFBC, and kill-curve result (synergy or nonsynergy). noglycoside therapy [55]. Most studies were retrospecNo parameters achieved statistical significance in this tive [8,21], had a small sample size (fewer than 100 model. patients) [4,6-91, and lacked any in vitro correlation. Studies conducted at individual institutions utilized COMMENTS data gathered over many years [21], such that evoluSince the 19709,there has been an ongoing but unre- tion of therapy for underlying diseasesand the develsolved debate on the utility of in vitro synergy testing‘ opment of new antimicrobial agents during the exin selecting optimal antibiotic therapy [l&19]. In the tended study period became confounding factors. following a positive blood culture result was better for patients receiving combination therapy (p = 0.041, Mantel-Cox log rank test). This difference was more pronounced when analyzed by the Gehan-Breslow method (p = 0.005), which places greater emphasison early deaths. The patient population that received combination therapy was comparable to those receiving monotherapy (Appendix). There was no significant difference for the following clinical parameters: underlying disease,portal of entry, and degree of illness. There was, however, a higher frequency of neutropenic patients among those who received combination therapy. Since neutropenic patients had a significantly worse prognosis, this was a potential bias in favor of monotherapy.
November 1989 The American Journal of Medicine
Volume
87
543
ANTIBIOTICS
AND /‘SEUDDMDMlS
BACTEREMIA
/ HlLF ET AL
A major problem has been the imprecise criteria for defining the endpoint of therapy. ‘Cure” has been loosely defined as the eradication of all signs and symptoms of pseudomonal infection, and “SurvivaI” has been defined by “bacteriologic cure,” ‘tdefervescence,” and “elimination of infection” [7,21,47,49,56]. It should be noted that these definitions require some degree of clinical judgment. In a severely ill patient, death may be a result of a combination of factors, only one of which may be infection. Finally, the primary reason that these important issues remain unresolved is the difficulty in conducting a human trial that could specifically resolve these points-the limiting factor is the sheer number of patients required for statistical validation. For example, in a recent study of 550 febrile neutropenic hosts, only 17% had bacteremia (an infection that can be diagnosed by objective criteria) and only 2% were caused by P. aeruginosa [56]. This observational study attempted to address these issues. Our sample size of 200 patients is the largest and the most comprehensive prospective study of P. aeruginosa bacteremia ever performed. Given the multiplicity of factors that might play a role in outcome, we performed univariate analysis on all 200 patients, aswell as subsetsof patients, controlling for clinical factors, rank-sum analysis for continuous variables (neutrophil count), multivariate analysis for clinical and laboratory parameters, and survival curve analysis for patients receiving combination therapy versus monotherapy. For the purposes of our study, it was believed that an objective criterion was necessary to define the study endpoint of outcome. Thus, an arbitrary cutoff of 10 days post-discovery of bacteremia was selected. It wasbelieved that if death occurred within 10 days of P. aeruginosa bacteremia, it washighly likely that the infection would be a contributory cause. [Our study results were essentially unchanged if five days and 14 days were usedinstead of 10 days (Appendix).] In any study in which outcome is an endpoint, a notable deficiency has been failure to stratify patients by severity of illness at the time therapy is initiated. Obviously, a comatose, hypotensive patient requiring mechanical ventilation will have a poorer prognosis than will a patient whose condition is stable, regardless of the type of therapy administered. We controlled for severity of illness in two ways: the first was stratifying by the McCabe criteria [57] and the second was stratifying by a definition of being “critically ill” (hypotensive, requiring mechanical respiratory support, or comatose). We confirm that underlying diseaseand severity of illness at the onset of bacteremia are critical determinants of outcome. It is noteworthy that those patients previously defined as“critically ill” and more likely to die (Table I) still had an improved outcome if they received combination therapy versus monotherapy (Table III). The McCabe criteria [57] for underlying diseasemay require redefinition in modern times to maximize its predictive capabilities. In oitro susceptibility and synergistic studies were performed in a blinded fashion without knowledge of patient outcome using the organism isolated from the patient and the antibiotics actually administered to the patient. The in vitro methods were rigorously standardized and performed in duplicate on different
days. Despite extensive analysis, we failed to demonstrate any significant correlation between results of in uitro susceptibility and synergistic testing versus outcome. Antibiotic failure or successcould not be correlated with in uitro susceptibility test results (MIC, MBC) of the infecting organism to the antibiotic(s) the patient received (Appendix). Interestingly, Flick and Cluff [8], in a retrospective study of 108 casesof pseudomonalbacteremia, also found that in vitro SUSceptibility test results did not correlate with outcome. Results of synergistic testing by either checkerboard or kill curve did not correlate with outcome (Table II). No improved correlation was seen even if the analyses for synergy were restricted to other subgroups, including neutropenic patients, immunosuppressed patients, or patients who were not “critically ill” (Appendix). Furthermore, no correlation could be demonstrated for results derived from checkerboard methods versus time-kill curve methods (Appendix). That is, combinations found to be synergistic by one method were not necessarily synergistic by the other method. The large sample size also afforded us the opportunity to addressthe ongoing controversy of monotherapy versus combination antibacterial agent therapy for pseudomonal bacteremia. Combination therapy was defined as simultaneous use of an antipseudomonal beta-la&am agent plus an aminoglycoside. This study strongly supported the use of combination antibacterial agent therapy rather than monotherapy for bacteremia causedby P. aeruginosa. Improved survival was noted for all patient subgroups that received combination therapy (Table III); statistical significance (p <0.05) was achieved for the total patient study group, patients with pneumonia, patients with bacteremia of a nosocomial origin, and critically-ill patients. What are the weaknessesof this study? Prospective evaluation of serum bactericidal titers was not performed, in part becauseof inadequate standardization for timing of blood sampling when two agents of varying half-lives were to be given. Although aminoglycoside serum concentrations were monitored for most patients and dosagealtered accordingly, there was no set protocol becauseof a lack of consensuswith respect to timing of serum concentrations and duration of intravenous infusion [58]. Beta-la&am serum concentrations were not monitored. We choseobjective criteria for assessing success or failure of antibiotic therapy; these criteria, however, may not allow one to determine with certainty the precise contribution of pseudomonal bacteremia to mortality. Clinical criteria, including defervescence, that might have been revealing for individual patients were not used (in part because of a lack of standardization). On the other hand, our useof objective criteria allows comparison of results from this large-scale study to future studies to be conducted by other investigators. There have been many proposals for improvements in methodology or interpretive criteria for synergy [17,28,59], and it might be validly argued that a significant correlation between in vitro testing and clinical outcome would have been obtained in our study had these or other “improvements” been incorporated. We point out, however, that the methodology used in this study is the standard one recommended by authorities and used by most academic medical centers, clinical
544
87
November
1989
The American
Journal
of Medicine
Volume
ANTIBIOTICS
microbiology laboratories, and pharmaceutical testing firms throughout the world [26,60-621. Should other investigators wish to evaluate any alternative methodolog)i, the organisms in this study are available for further evaluation and blinded correlations between laboratory tests and clinical outcome could be reassessed. Although the study was prospective, allocation of patients for combination therapy versus monotherapy was not controlled. Nevertheless, the two patient groups were comparable (Appendix). The one potential bias was that neutropenic patients tended to receive combination therapy more frequently than monotherapy. However, since neutropenia was also an adverse risk factor for outcome, this should have created a bias for poorer outcome in the patients receiving combination therapy! Thus, our contention that combination therapy is superior to monotherapy is understated. What are the implications of this study? In the antibiotic-bacterial interaction in a bacteremic patient, there are undoubtedly a multiplicity of factors that affect the ultimate outcome, including the host’s immune system. Susceptibility and synergy (as measured in uitro) are only two of many factors that are operative. Nevertheless, we failed to demonstrate that these in vitro tests are useful in predicting outcome in pseudomonal bacteremia even when the test results were confined to specific high-risk patient populations (e.g., neutropenic patients, patients with non-urinary tract infections, or patients with bacteremia of nosocomial origin). Although it may be reasonable to perform these in vitro tests in selected patients (e.g., those showing a poor response to presumably appropriate therapy), the results of this study failed to justify the routine use of these in uitro tests in the management of an individual patient. The time-kill curve might be considered superior to the checkerboard technique (as it is currently performed and interpreted) in that a trend was seen for improved outcome in those patients infected by organisms designated as synergistic by time-kill curves (Table II). Although the association was not statistically significant (p = O.lO), a larger sample size of 250 to 400 patients might have been required to bring out a significant difference should one have existed (type II error). Finally, increasing effort should be focused on ensuring timely administration of combination therapy to patients with P. aeruginosa bacteremia, since the use of combination therapy was even more important in determining ultimate outcome than were underlying disease and degree of neutropenia.
ACKNOWLEDGMENT We thank Bruce Farber and Ying Yee for participating in our confirmatory in vitro test protocol; Floyd B. Taylor, Marilyn Wagener. and Katherine Godfrey for assistance with the statistical analysis; Vincent Andriole and Aldona Baltch for critical review; and Shirley Brinker for secretarial assistance.
REFERENCES 1. Baltch AL, Smith RP: Combinations of antibiotics against Pseudomonasaerugh osa Am J Med 1985: 79: 8-14. 2. Bennett JV: Hospital acquired infections and the altered hosts: nosocomial infections due to Pseudomonas. J Infect Dis 1974; 130: 54-57. 3. Cross A, Allen JR, Burke J. et al: Nosocomial infections due to Pseudomonas aeruginosa: review of recent trends. Rev Infect Dis 1983; 5: %37X%45. 4. Fishman LS, Armstrong D: Pseudomonas aeruginosa bacteremia in patients with neoplastic disease. Cancer 1972; 30: 764-773.
AND /‘SfUDOAfOMSBACTEREMlA
/ HILF ET AL .
5. Allen JR, Hightower AW. Martin SM. Dixon RE: Secular trends in nosocomial infections: 1970-1979. Am J Med 1981: 70: 389-391. 6. Baltch AL, Hammer M. Smith RP, Sutphen N: Pseudomonas aerugihosa bacteremia: susceptibility of 100 blood culture isolates to seven antimicrobial agents and its clinical significance. J Lab Clin Med 1979; 94: 201-214. 7. Tapper ML, Armstrong D: Bacteremia due to Pseudomonas aerugihosa complicating neoplastic disease: a progress report. J Infect Dis 1974; 130: S14-S23. 8. Flick MR. Cluff LE: Pseudomonas bacteremia: review of 108 cases. Am J Med 1976; 60: 501-508. 9. Whitecar JP. Luna M. Bodey Gl? Pseudomonas bacteremia in patients with malignant disease. Am J Med Sci 1970; 260: 216-223. 10. Fass RJ, Rotilie CA, Prior RB: Interaction of clindamycin and gentamicin in vitro. Antimicrob Agent Chemother 1974; 6: 582-587. 11. Elion GB, Singer S, Hitchings GH: Antagonist of nucleic acid derivatives. VIII. Synergism in combination of biochemically related metabolites. J Viol Chem 1954; 208: 477-488. 12. Bulger RJ. Nielson K: Effect of different media on in vitro studies of antibiotic combination. Appl Microbial 1968; 16: 890-895. 13. Kaplan EL, Meier P: Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958; 53: 457-481. 14. Mantel N: Evaluation of survival data and two new rank order statistics arising in its consideration. Cancer Chemother Rep 1966; 50: 163-170. 15. Breslow N: A generalized Kruskal-Wallis test for comparing K samples subject to unequal patterns of censorship. Biometrika 1970; 57: 579-594. 16. Ellner PD. Neu HC: The inhibitory quotient: a method for interpreting minimum inhibitory concentration data. JAMA 1981; 246: 1575-1578. 17. Berenbaum MC: Correlations between methods for measurement of synergy. J Infect Dis 1980; 142: 476-479. 18. Comber KR, Basker MJ. Osborne CD, Sutherland R: Synergy of carbenicillin and gentamicin in experimental infections with Pseudomonas. J Infect Dis 1971; 124: 546-555. 19. Giamarellou H: Aminogiycoside plus beta-lactam against gram-negative organisms-evaluation of in vitro synergy and chemical interactions. Am J Med 1986; 80: S12GS137. 20. Chandrasekar PH, Crane LR, Bailey EJ: Comparison of the activity of antibiotic combinations in vitro with clinical outcome and resistance emergence in serious infections by Pseudomonas aeruginosa in nonneutropenic patients. J Antimicrob Chemotner 1987; 19: 321-329. 21. Bodey GP, Jadeja L, Elting L: Pseudomonas bacteremia; retrospedie analysis of 410 episodes. Arch Intern Med 1985: 145: 1621-1629. 22. Young LS: Empirical antimicrobial therapy in the neutropenic host. N Engl J Med 1986; 313: 58C-581. 23. Moellering RC: Antimicrobial synergism-an elusive concept. J Infect Dis 1979; 140: 639-641. 24. Marymont JH. Marymont J: Laboratory evaluation of antibiotic combinations: a review of methods and problems. Lab Med 1981; 12: 47-55. 25. Krogstad DJ. Moellering RC: Antibiotics in laboratory medicine. Baltimore: Williams and Wilkins, 1986; 537-595. 26. Weinstein RJ. Young LS. Hewitt WL: Comparison of methods for assessing in vitroantibiotic synergism against Pseudomonasand Serralia. J Lab Clin Med 1975; 86: 853-862. 27. Norden CW. Wentzel H. Keleti E: Comparison of techniques for measurement of in vitro antibiotic synergism. J Infect Dis 1979; 140: S629-S633. 28. Woolfrey BF. Fox JMK. Lally RT. Quall CO: Broth microdilution testing of Pseudomonas aeruginosa and aminoglycosides: need for employing dilutions differing by small arithmetic increments. J Clin Microbial 1982; 16: 663667. 29. Hallander HO, Dornbusch K. Gezelius L. Jacobson K. Karlson I: Synergism between aminoglycosides and cephalosporins with antipseudomonal activity: interaction index and killing curve method. Antimicrob Agents Chemother 1982; 22: 743-752. 30. Young LS: Significance of in vitro tests. In: Combination antibiotic therapy in the compromised host. New York: Klastersky and Staquet. 1982; 1-15. 31. Andriole VT: Synergy of carbenicillin and gentamicin in experimental infection with Pseudomonas. J Infect dis 1971; 124: S46-S55. 32. Runsack MG. Orake TA, Hackbarth CJ, Sande MA: Single versus combination antibiotic therapy for pneumonia due to Pseudomonas aeruginosa in neutropenic guinea pigs. J Infect Ois 1984: 149: 98C-985. 33. Lumish RM. Norden CW: Therapy of neutropenic rats infected with Pseudomo nas aeruginosa. J Infect Ois 1976; 133: 538-547. 34. Gordin FM, Rusnak MG. Sande MA: Evaluation of combination chemotherapy in a lightly anesthetized animal model of Pseudomonas pneumonia. Antimicrob Agents Chemother 1987; 31: 398-403. 35. Shyu WC, Nightingale CH. Quintiliani R: Pseudomonas peritonitis in neutropenic rats treated with amikacin. ceftazadime and ticarcillin alone and in combination. J Antimicrob Chemother 1987; 19: 807-814. 36. Peterson LR, Gerding DN. Moody JA. et al: Comparison of azlocillin, ceftizoxime. cefotitan. and amikacin alone and in combination against Pseudomonas aeruginosa in a neutropenic-site rabbit model. Antimicrob Agents Chemother 1984: 25: 545-552. 37. Khakoo RA. Kluge RM: Effectiveness of gentamicin and carbenicillin in a rat model against infection with Pseudomonas aerughosa resistant to gentamicin or gentamicin and carbenicillin. J Antimicrob Chemother 1979; 5: 53-59. 38. SCOtt RE, Robson HG: Synergistic activity of carbenicillin and gentamicin in experimental Fseudomonas bacteremia in neutropenic rats. Antimicrob Agents
November
1989
The American
Journal
of Medicine
Volume
87
545
ANTIBIOTICS
AND PSEUDOMONASBACTEREMIA
I HlLF ET AL
Chemother 1976; 10: 646651. 39. Pennington JE. Stone RM: Comparison of antibiotic regimens for treatment of experfmental pneumonia due to Pseudomonas. J Infect Dis 1979; 140: 881-889. 40. Chusid MJ. Davis SD, Sarff LD: Experimental Pseudomonas aerugirtosa sepsis: absence of synergy between ticarcillin and tobramycin. J Lab Clin Med 1983: 4: 441-449. 41. Saslaw S. Carlisle HN. Moheimani M: Comparison of gentamicin. carbenicillin and gentamicin. and carbenicillin in Pseudomonas sepsis in monkeys. Antimicrob Agents Chemother 1973; 3: 274-278. 42. Norden C, Schaffer M: Activities of tobramycin and azlocillin alone and in combination against experimental osteomyelitis caused by Pseudomonasaerugin osa. Antimicrob Agents Chemother 1982; 21: 62-65. 43. Klatersky J: Combination of antibiotics for therapy of severe infections in cancer patients. Eur J Cancer 1979; 15: S3-S13. 44. Klatersky J, Meunier-Carpenter F, Prevost J: Significance of antimicrobial synergism for the outcome of gram-negative sepsis. Am J Med Sci 1977: 273: 157167. 45. KeatingMJ. Bodey GP. Valdivieso M. Rodriguez V: A randomized comparative trial of three aminoglycosides-comparison of continuous infusions of gentamicin, amikacin, and sisomicin combined with carbenicillin on the treatment of infections in neutropenic patients with malignancies. Medicine (Baltimore) 1979; 58: 159170. 46. Love U. Schimpff SC, Schiffer CA. Wrernik PH: Improved prognosis for granulocytopenic patients with gram-negative bacteremia. Am J Med 1980; 68: 643-648. 47. Schimpff S. Satterlee W. Young VM. eta/: Empiric therapy with carbenicillin and gentamicin in febrile cancer patients with cancer and granulocytopenia. N Engi J Med 1977; 284: 1061-1065. 48. Jackson GG. Riff U: Pseudomonas bacteremia: pharmacologic and other basesforfailureof treatment withgentamicin. J Infect Dis 1971; 124: S185-S191. 49. The EORTC International Antimicrobial Therapy Cooperative Group: Ceftazidime combined with a short or long course of amikacrn for empirical therapy of gram-negative bacteremia in cancer patients with granulocytopenia. N Engl J Med
546
November
1989
The American
Journal
of Medicine
Volume
1987: 317: 1692-1698. 90. Baltch AL, Griffin PE: Pseudomonas aeruginosa bacteremia: a clinical study of 75 patients. Am J Med Sci 1977; 274: 119-129. 51. Bodey CP: Carbenicillin therapy for Pseudomonasinfections. JAMA 1971: 218: 62-66. 52. Lau WK. Young LS. Black RE, et a/:Comparative efficacy and toxicity of amikacin/carbenicillin versus gentamicin/carbenicillin in leukopenic patients-a randomized prospective trial. Am J’ Med 1977; 62: 959-966. 53. Young LS: Review of the clinical significance of synergy in gram-negative infections at the University of California, Los Angeles, Hospital. Infection 1978; 6: 547552. 54. Anderson ET, Young LS, Hewitt W: Antimicrobial synergism in the therapy of gram-negative rod bacteremia. Chemotherapy 1978; 24: 45-54. 55. Bodey G. Bolivar R. Fainstein V. Jadeja L: Infections caused by Pseudomonas aeruginosa. Rev Infect Dis 1983; 5: 279-313. 56. Pizzo PA, Hathorn JW. Hiemenz J. efaltA randomized trial comparing ceftazidime alone with combination antibiotic therapy in cancer patients with fever and neutropenia. N Engl J Med 1986; 315: 552-558. 57. McCabe WR, Jackson CG: Gram-negative bacteremia. I. Etiology and ecology. Arch Intern Med 1962; 110: 847-855. 58. Barza ht. Lauermann M: Why monitor serum levels of gentamicin! Clin Pharmacokinet 1978; 3: 202-215. 59. King TC, Schlesfgner D. Krogstad D: The assessment of antimicrobial combinations. Rev Infect Dis 1981; 3: 627633. 60. Schoenknecht FD. Sabath LD, Thornsberry C: Susceptibility tests: special tests. In: Manual of clinical microbiology, 4th ed. chap 105. American Society of Microbiology, 1985; lWO-1008. 61. Methods for dilution. Antimicrobial susceptibility tests for bacteria that grow aerobically. Presented at the National Committee for Clinical Laboratory Standards vol 5. no. 22. 1985. 62. Barry A: The antimicrobic susceptibility test: In: Principles and practices, chap 6-7. Philadelphia: Lea and Febiger, 1976.
87
PURPOSEAND PATIENTSAND MJTl’HODs:We performed a prospective clinical study of 200 consecutive patients with Pseudomonasaeruginosa bacteremias to analyze iu V&O susceptibility and synergistic testing of antibiotics the patients received and clinical parameters to assesstheir relationship to SurviVaL RESULTS:No significant correlation between in vitro susceptibility testing (minimal inhibitory concentrations/minimal bactericidal concentrations) and outcome could be demonstrated. Similarly, improved outcome could not be demonstrated for patients receiving antibiotic combinations that were synergistic in vitro (either time-kill or checkerboard) versus those combinations that were not. There was also no correlation between results obtained by time-kill curve and checkerboard synergistic testing, i.e., combinations found to be synergistic by one method were not necessarily synergistic by the other method. Clinical parameters associated with improved survival were a urinary portal of entry and absence of neutropenia. Conversely, survival was significantly decreased when the portal was the respiratory tract. The mortality rate between patients receiving combination therapy (27% ) and monotherapy (47% ) was significant (p cO.02); this significant relationship held true for most subgroups including malignancy, nosocomial infection, and infection site. CONCLUSION: Increasing effort should be placed on ensuring timely administration of combination therapy to patients with I? aeruginosa bacteremia since the use of combination therapy was even more important in determining outcome than was underlying disease.
From the Infectious Disease and Special Pathogens Sections, Veterans Administration Medical Center and University of Pittsburgh, Pittsburgh, Pennsylvania. Requests for reprints should be addressed to Victor L. Yu. M.D., University of Pittsburgh. School of Medicine, Scaife 968, Pittsburgh, Pennsylvania 15261. Manuscript submitted August 15.1988. and accepted in revised form July 10. 1989. For a copy of the 22-page Appendix cited in this article, please enclose a stamped, self-addressed, large (at least 6 X IO%-inch) envelope.
540
November
1989
The American
Journal
of Medicine
Volume
seudomonas aeruginosa continues as a cause of challenging infections, most notably in the comP promised and hospitalized patient. Mortality due to P. aeruginosa bacteremia remains ominously high [l-9]; prognosis is consistently poor for patients with leukemia and lymphoma (mortality of 67%to 100%) [1,4,9]. As a result, in vitro tests have been touted as a guide for choosing optimal antibiotic therapy in patients infected by P. aeruginosa. However, reproducibility and interpretations remain controversial, correlation of the results of in vitro antibiotic susceptibility and synergistic testing and clinical outcome has been notably absent. Hence, it has become almost reflex for investigators to acknowledge the need for such correlation in any publication reporting in uitro results. Therefore, we initiated a prospective, multihospital study of P. aeruginosa bacteremias. We made a concerted effort to addressthose areasof weaknessesthat have prevented a definitive resolution to this ongoing controversy, including prospective rather than retrospective study, adequate sample size for statistical analysis (200 patients), definitive evidence of infection (bacteremia), selection of a single organism (P. aeruginosa) rather than “gram-negative” organisms, blinded performance of in vitro tests without knowledge of outcome, and use of objective endpoints (survival at fixed number of days). Clinical parameters, ‘together with in uitro susceptibility and synergistic test results for antibiotics the patient received, were assessedin their relationship to patient survival. PATIENTS AND METHODS Clinical Study Two hundred consecutive patients from whom P. aeruginosa was isolated in blood culture were followed prospectively from June 1982 to June 1986, at eight Pittsburgh area hospitals. Organisms were isolated and identified at the respective clinical microbiologic laboratories and transported to a central laboratory. Susceptibility testing was performed without knowledge of clinical outcome. Data collected prospectively included patient demographics, hospital course, and antibiotics administered. All treatment decisionswere made by the attending physicians without consultation or intervention by the investigators except to ensure that aminoglycoside serum concentrations were measured. In Vitro Antibiotic Susceptibility Testing The antipseudomonal agent(s) selected for in uitro testing was the antipseudomonal agent that the patient received for the bacteremia. If a number of antipseudomonal agents were used over the course of the bacteremia, the agent(s) selected for in uitro testing wasthe agent received for at least two days within the first three days of bacteremia. Susceptibility testing 87
ANTIBIOTICS
was performed by microdilution techniques as previously described [lo]. Dilutions of antibiotics were made in Mueller-Hinton broth supplemented with 60 mg/L of calcium and 20 mg/L of magnesium. The inoculum was 5 x lo5 to lo6 CFU (colony-forming units)/mL. Each set of minimum inhibitory concentration (MIC) and minimum bactericidal concentration (MBC) calculations were performed in duplicate on different days. In Vitro Synergisitic Testing Synergistic testing was performed on those isolates taken from patients who received combination therapy. A breakdown of antibiotic combinations prescribed is given in the Results section. CHECKERBOARDTECHNIQUE: Druginteractionswere determined in an 8 X 8 column pattern using 8 X 12 microtiter plates. The inoculum was 5 X lo5 to lo6 CFU/mL. The fractional inhibitory concentration index (FIC) and fractional bactericidal concentration index (FBC) were calculated as previously described
Pll.
KILL CURVES: Time-kill curves were performed in 25 X loo-mm tubes as previously described [12]. An inoculum size of 5 X lo5 to lo6 CFU/mL was used. The final concentration of the two agents used was one fourth of the previously determined MBC for those agents. Quantitation of growth was measured from samples taken at four and 24 hours and inoculated onto a 100 X 15-mm agar plate using a spiral plater (Spiral Systems, Inc., Bethesda, Maryland). The interaction was defined as synergistic if colony counts using the combination showed a greater than or equal to loo-fold decrease when compared with the single most active agent. Complete technical details of susceptibility testing, checkerboard methodology, time-kill curve method, and experiments in reproducibility are given in a detailed manual contained in an Appendix available upon request from the authors.
Statistical Methods Patient and laboratory data were entered into a computer databank (Prophet System, Division of Research Resources, National Institutes of Health, Bethesda, Maryland). Fisher’s exact test or the chisquare test (two-tailed) was used to assessstatistical significance for various outcome measures. The Mann-Whitney rank sum test was used to assessthe association of neutrophil count and outcome. A stepwise logistic regressionanalysis was used for multivariate analysis (BMDP, University of California). The dependent binary variable was mortality at 10 days, and the independent variables were the prospectively monitored clinical factors and the in vitro test results. A 30-day survival curve was calculated for each therapy using a Kaplan-Meier estimate [13]. The differencesbetween the curves were then analyzed using a Mantel-Cox log rank [14] and the Gehan-Breslow test
[151RESULTS Patient Demographics The study population was composedof 200 patients with a mean age of 59 (range: one week to 96 years). Twenty-three percent (46 of 200) and 77%(154 of 200) of patients had community-acquired and nosocomial
AND PSEUDOMONASBACTEREMIA
/ HILF ET Al.
bacteremia, respectively. Bacteremias were classified asnosocomial if they occurred after 72 hours of hospitalization; for these patients, the mean was 20 days of hospitalization before development of bacteremia. Thirty-four percent (67 of 200) of the patients were in intensive-care units when P. aeruginosa bacteremia developed. Fifty-eight percent (115 of 200) of the patients were deemed to be immunosuppressed at the time of bacteremia by the following criteria: cytotoxic drugs (69 patients), corticosteroid therapy (80 patients with 88% receiving the equivalent of 20 mg of prednisone or greater), radiation therapy (26 patients), hematologic malignancy (48 patients), and neutropenia (neutrophil count lessthan 3,OOO/mL)(53 patients); of the 53 neutropenic patients, 27 had neutrophil counts of less than lOO/mL. Eighty-four percent (168 of 200) of the patients had undergone invasive procedures within seven days before the onset of bacteremia. Eighty-six percent (144 of 168) had placement of intravascular catheters, 55% (110 of 168) had urinary tract catheters, 45% (76 of 168) required mechanical ventilatory assistance, and 34% (57 of 168) had undergone surgery within seven days prior to the positive blood culture result. Microbiology EXTRA-BLOOD SITES: P. aeruginosa was cultured concomitantly from urine (29%, 58 of 200 patients), respiratory tract (34%, 67 of 200 patients), wound (15%, 29 of 200 patients), and tips of indwelling vascular catheters (14%, 28 of 200 patients). P. aeruginosa was also isolated from other sites: abdominal drainage (six) and biliary drainage (eight), aswell asaspirates of kidney (one), liver (one), and bone marrow (two).
Outcome Seventy-six percent (151 of 200) of the patients were alive at five days and 59% (117 of 200) were alive 10 days after isolation of P. aeruginosa from blood culture. Survival, for the purposes of this study, was arbitrarily defined as surviving 10 days after the onset of bacteremia, whereas mortality was defined as death within 10 days after the onset of bacteremia. The correlation for other endpoints including outcome at five days, 14 days, or at time of death/discharge was esentially identical to the lo-day endpoint (data in Appendix available from authors upon request). Clinical Parameters and Outcome There was a significant association between increased mortality and presence of neutropenia (neutrophil count less than 3,OOO/mL) (p <0.05, Fisher’s exact test, data not shown) in the 143 patients receiving combination therapy. Site of infection was significantly associated with outcome. Patients whose portal of entry was the urinary tract had a significantly improved survival rate compared with those patients with non-urinary tract portals (p = 0.001, data not shown). Patients with pneumonia as the primary site had a significantly higher mortality rate than patients without pneumonia (p cO.05, Appendix). Since severity of illness at the time of bacteremia was expected to be a prognostic factor that could confound the effect of antipseudomonal antibiotic therapy, we also controlled for this factor. Patients were deemed “critically ill” if they were receiving mechani-
November
1999 The American
Journal
of Medlclne
Volume
87
541
ANTIBIOTICS
AND PSEUDOMONAS
BACTEREMIA
/ HILF ET AL
TABLE I Clinical Parameters and Outcome in 143 patients with Bacteremia Receiving Combination Therapy (Patients Receiving Monotherapy or No Antlpseudomonal Agent . Therapy Are Excluded from This Analysis) Numberof Patients
Clinical Parameter
Percent Mortality
‘26
Neutrophils
31 (a/26) 47 (8/17) 22 (22/100)
NS (0.08)*
21(13/63) 28(16/57) 39 (9/23)
NS’
47 (18/37) 19 (20/ 106)
0.001t
1:; 20 16 7 100
35 (7/20) 13 (2/16) 43 (3/7) 26 (26/100)
NS (0.09)’
1:;
McCabe criteria [57] Non-fatal Ultimately fatal Rapidly fatal
63 ::
Portal5 Pneumonia Urinary tract Intravascular Other
S = not significant. Chi-square test (two-tatled). ‘isher exact test (two-tailed). \s defined by mechanical respiratory assistance, or coma. s Portal defined by physician using clinical criterra simultaneously from site of infection.
TABLE II Correlation Outcome
p Value
acute
hypotensive
episode.
and Pseudomonas
isolated
of Results of in Vitro Synergy Testing with
trations according to the literature, and the MC. NO correlation could be demonstrated for MICs or for MBCs (Appendix). Only 7% (nine of 131) of organisms were resistant to the beta-lactam prescribed and 4% (seven of 158) were resistant to the aminoglycoside prescribed. Combination therapy was administered to 143 patients and synergy was evaluated by checkerboard assay and time-kill curve for 123 bacteremic organisms involved (18 organisms were not available for testing and two others were excluded due to the combination received). Results of the checkerboard assaywere analyzed with two break points: FIC lessthan or equal to 0.50 (17 organisms) (Table II) and FIC less than or equal to 1.0 (110 organisms). An FIC lessthan or equal to 0.5 denotes that the MICs of the two agents decreasefourfold or more when the agents are combined, an arbitrary but well-accepted criterion for synergy. This takes into account an inherent twofold dilution error accepted for the method. An FIC less than or equal to 1.0 denotes that the MICs of the two agents decrease twofold or more when the agents are combined, a strict mathematical interpretation of synergy [17]. Results from 71 time-kill curves fulfilled the criterion for synergy. No significant correlation with patient outcome was seen (p = 0.10) (Table II). There was also no correlation between results obtained by time-kill curve and checkerboard assay (using either FIC less than or equal to 0.50 or FIC less than or equal to 1.0 as the criterion for checkerboard synergy) (p >0.05, Fisher’s exact test). Concordance for synergy and nonsynergy by the two methods was only 49% (Appendix).
Microbiologic Parametersand Outcome In vitro susceptibility testing (MICs) could not be correlated with patient outcome. Since achievable serum concentrations of each antibiotic are also pertinent, we calculated the therapeutic index [16], defined asthe ratio of expected peak antibiotic serum concen-
Antibiotics and Patient Outcome Ninety-three percent (186 of 200) of the patients received antipseudomonal antibiotics. Ninety percent (179 of 200) received an aminoglycoside: gentamicin (50 patients), tobramycin (108), or amikacin (21). Seventy-five percent (149 of 200) received a beta-la&am agent (one patient received two beta-lactam agents): piperacillin (72), ticarcillin (49), mezlocillin (ll), cefoperazone (nine), carbenicillin (five), imipenem (one), cefotaxime (two), and ceftazidime (one). Seventy percent (143 of 200) received combination therapy; in 142 patients, therapy consisted of an antipseudomonal beta-la&am agent plus an aminoglycoside and, in one patient, therapy wasmezlocillin plus ceftazidime. The most common combinations given were piperacillin plus tobramycin (25%, 36 of 143) and ticarcillin plus tobramycin (24%, 35 of 143). Seven percent (14 of 200) received no antipseudomonal antibiotic therapy; all of these patients died within 10 days of the onset of bacteremia. The most striking factor associated with patient outcome was the receipt of combination therapy versus monotherapy (Table III). The 143 patients who received a combination of an aminoglycoside and an antipseudomonal beta-la&am agent had a 27% (38 of 143) mortality rate as compared with the patients receiving monotherapy, who had a mortality rate of 47% (20 of 43) (p = 0.023). The trend for improved survival in patients given combination therapy was also seen for every patient subgroup studied, although not all were statistically significant (Table III). Thirty-day survival curves were also significantly different for the two treatment groups (Appendix). The probability of survival during the 30-day period
542
87
Percent Dled Checkerboard Synergistict Nonsynergistic Kill curve Synergistic Nonsynergistic
Percent Survived
p Value*
3 (4/123) 24 (29/123)
11 (13/123) 63 (77/123)
NS
12(15/123) 15(18/123)
46 (56/123) 28 (34/123)
NS (0.10)
NS = not significant. * Fisher exact test (two-tailed). r Defined as FIC index 50.5. If synergy was defined value still would not have been significant.
as FIC Index
51.0.
the p
cal respiratory assistance, had experienced an acute hypotensive episodewithin 72 hours prior to the positive blood culture result, or were comatose.As expected, patients who were “critically ill” had a significantly lower survival rate than those who were not critically ill (p
November
1999
The American
Journal
of Medicine
Volume
ANTIBIOTICS
AND PSEUDOMONASBACTEREMIA
/ HILF ET AL
TABLE III Mortality Rates for Patient Subgroups Receiving Combination Combination Therapy Was Superior to Monotherapy)
Patient Subgroup All patients Nosocomial origin Pneumonia Critically ill* lmmunosuppressed Malignancy5 Non-critically ill Neutropenia (<3,0OO/mL)
Number of Patients
Antibiotic
Therapy and Monotherapy
Percent Mortality Combination
186t 143 :9” 107 85 137 49
27 (38/143) 32 (34/108) 35 (7/iO) ’ 47 (18/37) 33 (28/86) 32 (21/66) 19 (20/106) 37 (16/43)
(For All Subgroups,
Percent Mortality Single Agent
p Value*
47 (20/43) 5Uw;8/;5)
0.023 0.04 0.033 0.016 NS NS
92(11/U) 48 (10/21) 47 (9/19)
2
Ki = not significant. ’ Fisher exact test (two-talled). t Fourteen patients received no antipseudomonal antibiotics (since therapy was often empiric prior to knowledge t “Critically ill” was defined as mechanical ventilatory support. acute hyootensive episode, or coma. . 5 Includes hematologic malignancies and solid tumors.
of Pseudomonas
in the blood
culture);
all died.
more specific area of P. aeruginosa infections, controversy exists regarding the necessity for double combination antibiotic therapy versus monotherapy [1,20221.The techniques commonly used to measuresynergy are the checkerboard and time-kill curve methods, and each in oitro method has its supporters [23]. Advocates of the checkerboard method cite its likenessto single-agent susceptibility testing and consider it a more practical assay to perform and interpret [24]. Advocates of time-kill curve methodology point out that it represents a more dynamic picture of bacterialantimicrobial interaction [25]. Comparisons of the methodologies are few and contradictory [26,27]. In addition, it is agreed that performance of these tests requires considerable technical experience, and standardization of the methods has been insufficient [24,25,2%30]. Multivariate Analysis The issueof synergy has been further explored in in Data for all 200patients were analyzed by a stepwise logistic regression model (BMDP, University of Cali- uiuo animal studies but results have been conflicting. Combination antibiotic therapy hasimproved survival fornia). The dependent variable was outcome (survival or death at 10 days), and the independent variables in normal and neutropenic animals infected with P. were the prospectively monitored risk factors of noso- aeruginosa (31-381. If the combination were synergistic in vitro, survival was further enhanced. On the comial versus community-acquired infection, neutrophi1 count, urinary portal (yes or no), respiratory por- other hand, results from other animal studies have failed to show significant improvements in outcome tal (yes or no), and antibiotic therapy (antipseudomonal beta-lactam single agent or amino- for combination therapy versus monotherapy [39-42]. Human studies are also conflicting, with studies supglycoside single agent versus combination therapy). The probabilities to enter and remove variables were porting both the superiority of two antibiotics [7,43set at 0.10 and 0.15, respectively. Combination antibi49] and the efficacy of a single antibiotic [21,50,51]. otic therapy, urinary tract portal, and absenceof neu- Use of combinations that are synergistic in vitro has tropenia were found to be significantly associatedwith also been credited with improved survival in immunosurvival; the p values to remove were 0.004,0.052, and suppressed patients [44,52-541. 0.03, respectively. A subset of the study group who A review of the clinical studies suggests possible received combination antibiotic therapy (143 pa- explanations for the conflicting results. In some studtients) was also evaluated with the stepwise logistic ies, the definition of combination therapy was not regression analysis. The aforementioned clinical fac- clearly stated. Combinations considered non-synergistors of nosocomial versus community-acquired infec- tic included aminoglycoside and beta-la&am agents tion, neutrophil count, urinary portal, and respiratory without antipseudomonal activity (i.e., cephalothin). portal were entered plus the following measuresof in. It has been cogently argued that this regimen is not a uitro interaction of antibiotic combinations: FIC, true combination but equivalent to single-agent amiFBC, and kill-curve result (synergy or nonsynergy). noglycoside therapy [55]. Most studies were retrospecNo parameters achieved statistical significance in this tive [8,21], had a small sample size (fewer than 100 model. patients) [4,6-91, and lacked any in vitro correlation. Studies conducted at individual institutions utilized COMMENTS data gathered over many years [21], such that evoluSince the 19709,there has been an ongoing but unre- tion of therapy for underlying diseasesand the develsolved debate on the utility of in vitro synergy testing‘ opment of new antimicrobial agents during the exin selecting optimal antibiotic therapy [l&19]. In the tended study period became confounding factors. following a positive blood culture result was better for patients receiving combination therapy (p = 0.041, Mantel-Cox log rank test). This difference was more pronounced when analyzed by the Gehan-Breslow method (p = 0.005), which places greater emphasison early deaths. The patient population that received combination therapy was comparable to those receiving monotherapy (Appendix). There was no significant difference for the following clinical parameters: underlying disease,portal of entry, and degree of illness. There was, however, a higher frequency of neutropenic patients among those who received combination therapy. Since neutropenic patients had a significantly worse prognosis, this was a potential bias in favor of monotherapy.
November 1989 The American Journal of Medicine
Volume
87
543
ANTIBIOTICS
AND /‘SEUDDMDMlS
BACTEREMIA
/ HlLF ET AL
A major problem has been the imprecise criteria for defining the endpoint of therapy. ‘Cure” has been loosely defined as the eradication of all signs and symptoms of pseudomonal infection, and “SurvivaI” has been defined by “bacteriologic cure,” ‘tdefervescence,” and “elimination of infection” [7,21,47,49,56]. It should be noted that these definitions require some degree of clinical judgment. In a severely ill patient, death may be a result of a combination of factors, only one of which may be infection. Finally, the primary reason that these important issues remain unresolved is the difficulty in conducting a human trial that could specifically resolve these points-the limiting factor is the sheer number of patients required for statistical validation. For example, in a recent study of 550 febrile neutropenic hosts, only 17% had bacteremia (an infection that can be diagnosed by objective criteria) and only 2% were caused by P. aeruginosa [56]. This observational study attempted to address these issues. Our sample size of 200 patients is the largest and the most comprehensive prospective study of P. aeruginosa bacteremia ever performed. Given the multiplicity of factors that might play a role in outcome, we performed univariate analysis on all 200 patients, aswell as subsetsof patients, controlling for clinical factors, rank-sum analysis for continuous variables (neutrophil count), multivariate analysis for clinical and laboratory parameters, and survival curve analysis for patients receiving combination therapy versus monotherapy. For the purposes of our study, it was believed that an objective criterion was necessary to define the study endpoint of outcome. Thus, an arbitrary cutoff of 10 days post-discovery of bacteremia was selected. It wasbelieved that if death occurred within 10 days of P. aeruginosa bacteremia, it washighly likely that the infection would be a contributory cause. [Our study results were essentially unchanged if five days and 14 days were usedinstead of 10 days (Appendix).] In any study in which outcome is an endpoint, a notable deficiency has been failure to stratify patients by severity of illness at the time therapy is initiated. Obviously, a comatose, hypotensive patient requiring mechanical ventilation will have a poorer prognosis than will a patient whose condition is stable, regardless of the type of therapy administered. We controlled for severity of illness in two ways: the first was stratifying by the McCabe criteria [57] and the second was stratifying by a definition of being “critically ill” (hypotensive, requiring mechanical respiratory support, or comatose). We confirm that underlying diseaseand severity of illness at the onset of bacteremia are critical determinants of outcome. It is noteworthy that those patients previously defined as“critically ill” and more likely to die (Table I) still had an improved outcome if they received combination therapy versus monotherapy (Table III). The McCabe criteria [57] for underlying diseasemay require redefinition in modern times to maximize its predictive capabilities. In oitro susceptibility and synergistic studies were performed in a blinded fashion without knowledge of patient outcome using the organism isolated from the patient and the antibiotics actually administered to the patient. The in vitro methods were rigorously standardized and performed in duplicate on different
days. Despite extensive analysis, we failed to demonstrate any significant correlation between results of in uitro susceptibility and synergistic testing versus outcome. Antibiotic failure or successcould not be correlated with in uitro susceptibility test results (MIC, MBC) of the infecting organism to the antibiotic(s) the patient received (Appendix). Interestingly, Flick and Cluff [8], in a retrospective study of 108 casesof pseudomonalbacteremia, also found that in vitro SUSceptibility test results did not correlate with outcome. Results of synergistic testing by either checkerboard or kill curve did not correlate with outcome (Table II). No improved correlation was seen even if the analyses for synergy were restricted to other subgroups, including neutropenic patients, immunosuppressed patients, or patients who were not “critically ill” (Appendix). Furthermore, no correlation could be demonstrated for results derived from checkerboard methods versus time-kill curve methods (Appendix). That is, combinations found to be synergistic by one method were not necessarily synergistic by the other method. The large sample size also afforded us the opportunity to addressthe ongoing controversy of monotherapy versus combination antibacterial agent therapy for pseudomonal bacteremia. Combination therapy was defined as simultaneous use of an antipseudomonal beta-la&am agent plus an aminoglycoside. This study strongly supported the use of combination antibacterial agent therapy rather than monotherapy for bacteremia causedby P. aeruginosa. Improved survival was noted for all patient subgroups that received combination therapy (Table III); statistical significance (p <0.05) was achieved for the total patient study group, patients with pneumonia, patients with bacteremia of a nosocomial origin, and critically-ill patients. What are the weaknessesof this study? Prospective evaluation of serum bactericidal titers was not performed, in part becauseof inadequate standardization for timing of blood sampling when two agents of varying half-lives were to be given. Although aminoglycoside serum concentrations were monitored for most patients and dosagealtered accordingly, there was no set protocol becauseof a lack of consensuswith respect to timing of serum concentrations and duration of intravenous infusion [58]. Beta-la&am serum concentrations were not monitored. We choseobjective criteria for assessing success or failure of antibiotic therapy; these criteria, however, may not allow one to determine with certainty the precise contribution of pseudomonal bacteremia to mortality. Clinical criteria, including defervescence, that might have been revealing for individual patients were not used (in part because of a lack of standardization). On the other hand, our useof objective criteria allows comparison of results from this large-scale study to future studies to be conducted by other investigators. There have been many proposals for improvements in methodology or interpretive criteria for synergy [17,28,59], and it might be validly argued that a significant correlation between in vitro testing and clinical outcome would have been obtained in our study had these or other “improvements” been incorporated. We point out, however, that the methodology used in this study is the standard one recommended by authorities and used by most academic medical centers, clinical
544
87
November
1989
The American
Journal
of Medicine
Volume
ANTIBIOTICS
microbiology laboratories, and pharmaceutical testing firms throughout the world [26,60-621. Should other investigators wish to evaluate any alternative methodolog)i, the organisms in this study are available for further evaluation and blinded correlations between laboratory tests and clinical outcome could be reassessed. Although the study was prospective, allocation of patients for combination therapy versus monotherapy was not controlled. Nevertheless, the two patient groups were comparable (Appendix). The one potential bias was that neutropenic patients tended to receive combination therapy more frequently than monotherapy. However, since neutropenia was also an adverse risk factor for outcome, this should have created a bias for poorer outcome in the patients receiving combination therapy! Thus, our contention that combination therapy is superior to monotherapy is understated. What are the implications of this study? In the antibiotic-bacterial interaction in a bacteremic patient, there are undoubtedly a multiplicity of factors that affect the ultimate outcome, including the host’s immune system. Susceptibility and synergy (as measured in uitro) are only two of many factors that are operative. Nevertheless, we failed to demonstrate that these in vitro tests are useful in predicting outcome in pseudomonal bacteremia even when the test results were confined to specific high-risk patient populations (e.g., neutropenic patients, patients with non-urinary tract infections, or patients with bacteremia of nosocomial origin). Although it may be reasonable to perform these in vitro tests in selected patients (e.g., those showing a poor response to presumably appropriate therapy), the results of this study failed to justify the routine use of these in uitro tests in the management of an individual patient. The time-kill curve might be considered superior to the checkerboard technique (as it is currently performed and interpreted) in that a trend was seen for improved outcome in those patients infected by organisms designated as synergistic by time-kill curves (Table II). Although the association was not statistically significant (p = O.lO), a larger sample size of 250 to 400 patients might have been required to bring out a significant difference should one have existed (type II error). Finally, increasing effort should be focused on ensuring timely administration of combination therapy to patients with P. aeruginosa bacteremia, since the use of combination therapy was even more important in determining ultimate outcome than were underlying disease and degree of neutropenia.
ACKNOWLEDGMENT We thank Bruce Farber and Ying Yee for participating in our confirmatory in vitro test protocol; Floyd B. Taylor, Marilyn Wagener. and Katherine Godfrey for assistance with the statistical analysis; Vincent Andriole and Aldona Baltch for critical review; and Shirley Brinker for secretarial assistance.
REFERENCES 1. Baltch AL, Smith RP: Combinations of antibiotics against Pseudomonasaerugh osa Am J Med 1985: 79: 8-14. 2. Bennett JV: Hospital acquired infections and the altered hosts: nosocomial infections due to Pseudomonas. J Infect Dis 1974; 130: 54-57. 3. Cross A, Allen JR, Burke J. et al: Nosocomial infections due to Pseudomonas aeruginosa: review of recent trends. Rev Infect Dis 1983; 5: %37X%45. 4. Fishman LS, Armstrong D: Pseudomonas aeruginosa bacteremia in patients with neoplastic disease. Cancer 1972; 30: 764-773.
AND /‘SfUDOAfOMSBACTEREMlA
/ HILF ET AL .
5. Allen JR, Hightower AW. Martin SM. Dixon RE: Secular trends in nosocomial infections: 1970-1979. Am J Med 1981: 70: 389-391. 6. Baltch AL, Hammer M. Smith RP, Sutphen N: Pseudomonas aerugihosa bacteremia: susceptibility of 100 blood culture isolates to seven antimicrobial agents and its clinical significance. J Lab Clin Med 1979; 94: 201-214. 7. Tapper ML, Armstrong D: Bacteremia due to Pseudomonas aerugihosa complicating neoplastic disease: a progress report. J Infect Dis 1974; 130: S14-S23. 8. Flick MR. Cluff LE: Pseudomonas bacteremia: review of 108 cases. Am J Med 1976; 60: 501-508. 9. Whitecar JP. Luna M. Bodey Gl? Pseudomonas bacteremia in patients with malignant disease. Am J Med Sci 1970; 260: 216-223. 10. Fass RJ, Rotilie CA, Prior RB: Interaction of clindamycin and gentamicin in vitro. Antimicrob Agent Chemother 1974; 6: 582-587. 11. Elion GB, Singer S, Hitchings GH: Antagonist of nucleic acid derivatives. VIII. Synergism in combination of biochemically related metabolites. J Viol Chem 1954; 208: 477-488. 12. Bulger RJ. Nielson K: Effect of different media on in vitro studies of antibiotic combination. Appl Microbial 1968; 16: 890-895. 13. Kaplan EL, Meier P: Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958; 53: 457-481. 14. Mantel N: Evaluation of survival data and two new rank order statistics arising in its consideration. Cancer Chemother Rep 1966; 50: 163-170. 15. Breslow N: A generalized Kruskal-Wallis test for comparing K samples subject to unequal patterns of censorship. Biometrika 1970; 57: 579-594. 16. Ellner PD. Neu HC: The inhibitory quotient: a method for interpreting minimum inhibitory concentration data. JAMA 1981; 246: 1575-1578. 17. Berenbaum MC: Correlations between methods for measurement of synergy. J Infect Dis 1980; 142: 476-479. 18. Comber KR, Basker MJ. Osborne CD, Sutherland R: Synergy of carbenicillin and gentamicin in experimental infections with Pseudomonas. J Infect Dis 1971; 124: 546-555. 19. Giamarellou H: Aminogiycoside plus beta-lactam against gram-negative organisms-evaluation of in vitro synergy and chemical interactions. Am J Med 1986; 80: S12GS137. 20. Chandrasekar PH, Crane LR, Bailey EJ: Comparison of the activity of antibiotic combinations in vitro with clinical outcome and resistance emergence in serious infections by Pseudomonas aeruginosa in nonneutropenic patients. J Antimicrob Chemotner 1987; 19: 321-329. 21. Bodey GP, Jadeja L, Elting L: Pseudomonas bacteremia; retrospedie analysis of 410 episodes. Arch Intern Med 1985: 145: 1621-1629. 22. Young LS: Empirical antimicrobial therapy in the neutropenic host. N Engl J Med 1986; 313: 58C-581. 23. Moellering RC: Antimicrobial synergism-an elusive concept. J Infect Dis 1979; 140: 639-641. 24. Marymont JH. Marymont J: Laboratory evaluation of antibiotic combinations: a review of methods and problems. Lab Med 1981; 12: 47-55. 25. Krogstad DJ. Moellering RC: Antibiotics in laboratory medicine. Baltimore: Williams and Wilkins, 1986; 537-595. 26. Weinstein RJ. Young LS. Hewitt WL: Comparison of methods for assessing in vitroantibiotic synergism against Pseudomonasand Serralia. J Lab Clin Med 1975; 86: 853-862. 27. Norden CW. Wentzel H. Keleti E: Comparison of techniques for measurement of in vitro antibiotic synergism. J Infect Dis 1979; 140: S629-S633. 28. Woolfrey BF. Fox JMK. Lally RT. Quall CO: Broth microdilution testing of Pseudomonas aeruginosa and aminoglycosides: need for employing dilutions differing by small arithmetic increments. J Clin Microbial 1982; 16: 663667. 29. Hallander HO, Dornbusch K. Gezelius L. Jacobson K. Karlson I: Synergism between aminoglycosides and cephalosporins with antipseudomonal activity: interaction index and killing curve method. Antimicrob Agents Chemother 1982; 22: 743-752. 30. Young LS: Significance of in vitro tests. In: Combination antibiotic therapy in the compromised host. New York: Klastersky and Staquet. 1982; 1-15. 31. Andriole VT: Synergy of carbenicillin and gentamicin in experimental infection with Pseudomonas. J Infect dis 1971; 124: S46-S55. 32. Runsack MG. Orake TA, Hackbarth CJ, Sande MA: Single versus combination antibiotic therapy for pneumonia due to Pseudomonas aeruginosa in neutropenic guinea pigs. J Infect Ois 1984: 149: 98C-985. 33. Lumish RM. Norden CW: Therapy of neutropenic rats infected with Pseudomo nas aeruginosa. J Infect Ois 1976; 133: 538-547. 34. Gordin FM, Rusnak MG. Sande MA: Evaluation of combination chemotherapy in a lightly anesthetized animal model of Pseudomonas pneumonia. Antimicrob Agents Chemother 1987; 31: 398-403. 35. Shyu WC, Nightingale CH. Quintiliani R: Pseudomonas peritonitis in neutropenic rats treated with amikacin. ceftazadime and ticarcillin alone and in combination. J Antimicrob Chemother 1987; 19: 807-814. 36. Peterson LR, Gerding DN. Moody JA. et al: Comparison of azlocillin, ceftizoxime. cefotitan. and amikacin alone and in combination against Pseudomonas aeruginosa in a neutropenic-site rabbit model. Antimicrob Agents Chemother 1984: 25: 545-552. 37. Khakoo RA. Kluge RM: Effectiveness of gentamicin and carbenicillin in a rat model against infection with Pseudomonas aerughosa resistant to gentamicin or gentamicin and carbenicillin. J Antimicrob Chemother 1979; 5: 53-59. 38. SCOtt RE, Robson HG: Synergistic activity of carbenicillin and gentamicin in experimental Fseudomonas bacteremia in neutropenic rats. Antimicrob Agents
November
1989
The American
Journal
of Medicine
Volume
87
545
ANTIBIOTICS
AND PSEUDOMONASBACTEREMIA
I HlLF ET AL
Chemother 1976; 10: 646651. 39. Pennington JE. Stone RM: Comparison of antibiotic regimens for treatment of experfmental pneumonia due to Pseudomonas. J Infect Dis 1979; 140: 881-889. 40. Chusid MJ. Davis SD, Sarff LD: Experimental Pseudomonas aerugirtosa sepsis: absence of synergy between ticarcillin and tobramycin. J Lab Clin Med 1983: 4: 441-449. 41. Saslaw S. Carlisle HN. Moheimani M: Comparison of gentamicin. carbenicillin and gentamicin. and carbenicillin in Pseudomonas sepsis in monkeys. Antimicrob Agents Chemother 1973; 3: 274-278. 42. Norden C, Schaffer M: Activities of tobramycin and azlocillin alone and in combination against experimental osteomyelitis caused by Pseudomonasaerugin osa. Antimicrob Agents Chemother 1982; 21: 62-65. 43. Klatersky J: Combination of antibiotics for therapy of severe infections in cancer patients. Eur J Cancer 1979; 15: S3-S13. 44. Klatersky J, Meunier-Carpenter F, Prevost J: Significance of antimicrobial synergism for the outcome of gram-negative sepsis. Am J Med Sci 1977: 273: 157167. 45. KeatingMJ. Bodey GP. Valdivieso M. Rodriguez V: A randomized comparative trial of three aminoglycosides-comparison of continuous infusions of gentamicin, amikacin, and sisomicin combined with carbenicillin on the treatment of infections in neutropenic patients with malignancies. Medicine (Baltimore) 1979; 58: 159170. 46. Love U. Schimpff SC, Schiffer CA. Wrernik PH: Improved prognosis for granulocytopenic patients with gram-negative bacteremia. Am J Med 1980; 68: 643-648. 47. Schimpff S. Satterlee W. Young VM. eta/: Empiric therapy with carbenicillin and gentamicin in febrile cancer patients with cancer and granulocytopenia. N Engi J Med 1977; 284: 1061-1065. 48. Jackson GG. Riff U: Pseudomonas bacteremia: pharmacologic and other basesforfailureof treatment withgentamicin. J Infect Dis 1971; 124: S185-S191. 49. The EORTC International Antimicrobial Therapy Cooperative Group: Ceftazidime combined with a short or long course of amikacrn for empirical therapy of gram-negative bacteremia in cancer patients with granulocytopenia. N Engl J Med
546
November
1989
The American
Journal
of Medicine
Volume
1987: 317: 1692-1698. 90. Baltch AL, Griffin PE: Pseudomonas aeruginosa bacteremia: a clinical study of 75 patients. Am J Med Sci 1977; 274: 119-129. 51. Bodey CP: Carbenicillin therapy for Pseudomonasinfections. JAMA 1971: 218: 62-66. 52. Lau WK. Young LS. Black RE, et a/:Comparative efficacy and toxicity of amikacin/carbenicillin versus gentamicin/carbenicillin in leukopenic patients-a randomized prospective trial. Am J’ Med 1977; 62: 959-966. 53. Young LS: Review of the clinical significance of synergy in gram-negative infections at the University of California, Los Angeles, Hospital. Infection 1978; 6: 547552. 54. Anderson ET, Young LS, Hewitt W: Antimicrobial synergism in the therapy of gram-negative rod bacteremia. Chemotherapy 1978; 24: 45-54. 55. Bodey G. Bolivar R. Fainstein V. Jadeja L: Infections caused by Pseudomonas aeruginosa. Rev Infect Dis 1983; 5: 279-313. 56. Pizzo PA, Hathorn JW. Hiemenz J. efaltA randomized trial comparing ceftazidime alone with combination antibiotic therapy in cancer patients with fever and neutropenia. N Engl J Med 1986; 315: 552-558. 57. McCabe WR, Jackson CG: Gram-negative bacteremia. I. Etiology and ecology. Arch Intern Med 1962; 110: 847-855. 58. Barza ht. Lauermann M: Why monitor serum levels of gentamicin! Clin Pharmacokinet 1978; 3: 202-215. 59. King TC, Schlesfgner D. Krogstad D: The assessment of antimicrobial combinations. Rev Infect Dis 1981; 3: 627633. 60. Schoenknecht FD. Sabath LD, Thornsberry C: Susceptibility tests: special tests. In: Manual of clinical microbiology, 4th ed. chap 105. American Society of Microbiology, 1985; lWO-1008. 61. Methods for dilution. Antimicrobial susceptibility tests for bacteria that grow aerobically. Presented at the National Committee for Clinical Laboratory Standards vol 5. no. 22. 1985. 62. Barry A: The antimicrobic susceptibility test: In: Principles and practices, chap 6-7. Philadelphia: Lea and Febiger, 1976.
87