When evaluating empiric therapy in this setting, the criteria for assessment have a major effect on the outcome [20]. Regardless of the method of assessment, the principal goal of empiric antibacterial therapy is to ensure that the patient survives both the infection and the neutropenic episode. Consequently, we used the criteria proposed by Pizzo and colleagues [17], in which the outcome of treatment is only scored as a failure if the patient dies of infection. Importantly, this method of assessment takes account of the initial infective event as well as any event that might subsequently develop during the same neutropenic episode. We also wished to reflect the common clinical practice of adjusting therapy whenever it is proved or perceived to be inadequate. Therefore, the protocol allowed modification of the empiric regimen by addition or substitution, but the reasons for altering therapy had to be reported in the case report form. By so doing, we were then able to subdivide the category of success with modification as defined by Pizzo and colleagues [17] and discriminate between true failures, such as persistence of the initial pathogen, and changes in therapy made simply because the clinician lacked confidence in the regimen without any objective sign of clinical deterioration caused by a bacterial infection.

Several alternatives were available for the comparative arm, but the combination of piperacillin and tobramycin was established as the standard empiric therapy for the febrile, neutropenic adult patient in most centers. We therefore selected this combination to compare its efficacy and toxicity with that of ceftazidime alone as empiric therapy for fever in the neutropenic patient, particularly focusing on infections caused by gram-negative bacteria.

Methods

Top Methods Results Discussion Author & Article Info References

The study was designed as a multicenter, nonblinded, randomized, parallel-group comparison of ceftazidime and a combination of piperacillin and tobramycin for the empiric management of fever defined as a single axillary temperature of 38.5 °C or at least two readings of more than 38 °C taken 2 hours apart (Figure 1). Patients with cancer who were older than 12 years were included provided that intensive chemotherapy induced neutropenia with granulocyte counts of less than 500 cells/mm³ within 3 days of starting antibiotic treatment. Those who received only palliative therapy for their malignant disease were not eligible. Patients were excluded if they were allergic to any of the trial antibiotics, if their serum creatinine exceeded 265 µmol/L (3 mg/dL), or if they had received previous parenteral antibiotic therapy during the same neutropenic episode. Any prophylactic antibacterial agents were to be withdrawn at the onset of fever with the exception of tuberculostatics and sulfamethoxazole-trimethoprim when administered twice weekly for the prevention of Pneumocystis carinii infection [21]. As is common practice in such trials [22], patients could re-enter the protocol for multiple, separate episodes of neutropenia provided that each episode was followed by an interval of at least one week with granulocyte counts in excess of 1000/mm³ and the patient also satisfied all inclusion criteria. The protocol was approved by the ethical committees of the participating institutions.

View larger version (30K): [in this window] [in a new window]   Figure 1. Flow diagram of the study.

Randomization Procedure

After giving informed consent, patients were assigned to one of six strata. They were first stratified according to underlying disease, namely leukemia and bone marrow transplants or solid tumor and lymphoma, following which, they were stratified according to the infection at presentation, skin or soft-tissue infection likely to be caused by gram-positive bacteria (for example exit-site infections related to venous access devices), anaerobic infection associated with the alimentary tract, or other infectious complications. The last group comprised the principal group of patients who were to be treated with either ceftazidime alone or a combination of piperacillin and tobramycin, whereas patients with skin and soft-tissue infections and those with anaerobic infections were to receive vancomycin and metronidazole, respectively, in addition to these core regimens. We do not report results obtained for episodes treated with these additional antibiotics because the data were not germane to the primary objective of the study and only a limited number of patients were involved.

Treatment Allocation

After having been placed in one of the three strata, patients were sequentially and randomly assigned to receive one of the regimens by means of opening sealed envelopes that contained the treatment allocation that had been generated centrally for each participating institution in blocks of eight to ensure balance.

All drugs were given intravenously: ceftazidime as a 20- to 30-minute infusion of 2 g every 8 h; piperacillin by slow injection of 16 to 18 g/d in 4 to 6 divided doses, except in Finland and Sweden, where the regulatory authorities recommended not exceeding 12 g daily. Tobramycin was given at an initial dose of 1.7 to 2 mg per kilogram body weight every 8 h. Serum concentrations of tobramycin were to be determined at least once during the first 3 days of therapy to certify that adequate trough and peak levels were achieved. Dosages were reduced accordingly for patients with impaired renal function. Treatment was discontinued once the granulocyte count had risen to greater than 500/mm³ and all signs and symptoms of infection had resolved for at least 2 days or after at least 5 consecutive days if neutropenia persisted provided that the patient had been free of all signs or symptoms of infection during this time. All patients who were still neutropenic when antibiotic treatment was stopped remained under observation for the occurrence of relapses or new fevers until neutropenia had resolved.

Therapy was started immediately after a comprehensive clinical examination and collection of specimens for microbiological investigation [12]. Ten millilitres of blood was obtained for culture from two separate veins; another 10 mL was taken via the central venous catheter when present. Appropriate cultures were also obtained from other local sites of infection when present. Isolates were identified using standard techniques.

Patients were examined daily and biochemical and hematologic profiles were done once and three times weekly, respectively. Blood was cultured after 48 to 72 hours of treatment, before any change in antibiotic therapy, and only when indicated thereafter. Chest radiographs were done weekly in persistently febrile patients, and invasive diagnostic procedures, such as bronchoalveolar lavage, were done on a case-by-case basis according to local practice. Appropriate investigations were repeated whenever changes in therapy were deemed necessary.

Modification of Initial Therapy

Changes in therapy were recommended when a primary pathogen appeared to be resistant or if a presumably causative pathogen persisted after at least 48 hours of therapy. In addition, continued deterioration or progression of the presenting infection, evidence of a nonbacterial infection, or a serious adverse event were considered reasons to modify therapy. The choice of the agent for modification was to be directed by the in vitro susceptibility when available, although vancomycin was recommended for infections involving gram-positive cocci and amphotericin B was the drug of choice in patients with either evidence of fungal infection or unexplained fever persisting for 6 days during neutropenia. Acyclovir was to be given to treat infections caused by herpes simplex virus and varicella zoster. Finally, erythromycin and sulfamethoxazole-trimethoprim were the treatment of choice for patients with pulmonary infiltrates suggestive of infection with Legionella species and P. carinii, respectively.

Definitions of Infections and Infectious Death

Primary infection was classified as microbiologically proven when cultures yielded organisms from a site of infection or bacteremia was detected; as clinically documented when a putative site of infection such as pulmonary infiltrate or cellulitis was identified without confirmation by culture; and as unexplained fever when no site was identified and microbiological evaluation failed to yield any potential pathogen within the first 72 hours after inclusion in the study. Bacteremia was defined as the occurrence of at least one positive blood culture of any organism with the exception of coagulase-negative staphylococci and skin coryneforms, for which two identical positive cultures were necessary. Defervescence was defined as a return of the temperature to below 37.5 °C for at least 2 days. A recurrent fever was one that developed during treatment, at least 2 days after the original episode had subsided. The development of a focus, a new episode of bacteremia, or fever was distinguished from the primary infection and was only considered a superinfection with or without positive cultures when it occurred during treatment. Any infection that occurred after stopping treatment while the patient remained neutropenic was considered to be a subsequent infection and was regarded as a relapse when the original pathogen was again isolated; otherwise it was considered to represent a new infection. Death was attributed to infection when it occurred as a direct consequence of the presenting infection, new infection, or superinfection.

Evaluation of Outcome

Clinical Evaluation

Patients were evaluated after 72 hours of therapy for survival. At the end of neutropenia, the clinical outcome was evaluated according to the criteria of Pizzo and colleagues [17] by the local investigator. All cases with documented infections as well as all deaths were reviewed by one of the members of the Data Review Committee, who was blinded to the treatment regimen. If the committee member disagreed with the assessment made by the local investigator, a second opinion was obtained from one of the other committee members. A case was considered to be "not assessable" when treatment had to be discontinued because of immediate hypersensitivity, noninfectious death, or other reasons unrelated to infection. Patients who died of infection were regarded as "failures". "Success (without modification)" was applied to any neutropenic episode in which the patient survived and became free of all signs and symptoms of infection without adjustment of the initial empiric scheme. "Success with modification" referred to any patient who survived the neutropenic episode but for whom empiric therapy had been changed in any way other than a dose adjustment. This category included patients who required additional antibacterial agents to survive the neutropenic episode, those who were changed to prophylaxis after resolution of the presenting infection, and those who received additional antifungal or antiviral therapy only. For the purposes of analysis, all cases that had been classified as "success with modification" were reviewed retrospectively by members of the Data Review Committee to permit discrimination between changes to therapy made because of an inadequate response of a primary bacterial infection and those made for other reasons.

A response was considered satisfactory if episodes were treated successfully with the study regimen alone, if therapy was stopped and changed to prophylaxis, or if the primary infection was later shown to be nonbacterial in nature. Likewise, antibacterial agents added within 3 days for fever without signs of clinical deterioration that was later shown to be of nonbacterial origin and therapy given to manage a new infective event were not considered to represent a failure of the study regimen; these were also classified as a satisfactory response. In contrast, the response was deemed to have been unsatisfactory when the patient died of infection or relapses occurred, when antibiotics were added or changed because the pathogen was either presumed or proven to be resistant, or if the response was considered inadequate because of clinical deterioration or continuing fever beyond 3 days, and when a recurrent fever or superinfection occurred.

Microbiological Evaluation

"Eradication" was defined by the occurrence of negative blood cultures taken either before changing therapy or stopping it altogether. Clearance was attributed to the study treatment whenever a modification was made with any agent lacking activity against the initial pathogen or pathogens, for example, the addition of amphotericin B in cases of bacteremia. Cases in which follow-up cultures had not been taken were deemed "not assessable." Antimicrobial susceptibility tests were done by agar disc diffusion, and any isolate that developed zones of inhibition of less than 18 mm in diameter for ceftazidime and piperacillin (equivalent to minimum inhibitory concentrations of 16 and 128 mg/L) or less than 15 mm in diameter for tobramycin (equivalent to a minimum inhibitory concentration of 8 mg/L) was considered resistant to the antibiotic. A strain was only considered resistant to the combination if it showed reduced susceptibility to both antibiotics.

Evaluation of Toxicity

All patients were assessed for toxicity, and all adverse events were assessed for their relation to the study treatment by the investigator. A skin rash was considered evidence of an allergic reaction when it was temporally related to the administration of antimicrobial agents. Clinically apparent loss of hearing, renal toxicity (increase of at least 50% of the baseline serum creatinine level), or substantial elevations of liver enzyme levels (twice the baseline values for bilirubin, transaminases, or alkaline phosphatase) were assumed to be related to the study drugs even though other antimicrobial agents may have been added.

Sample Size Calculation and Statistical Analysis

Our primary hypothesis was whether ceftazidime was equally effective as the combination of piperacillin and tobramycin for empiric treatment of febrile, neutropenic patients. A satisfactory response rate of 60% was expected requiring a minimum sample size of 700 evaluable episodes to detect a 10% difference in the regimens studied with a power of 80% at the 5% level of significance. We further assumed that approximately 10% of episodes would prove not assessable and therefore a total of 770 episodes were required.

Log-linear analysis (GLIM version 3.77; University of London, London, England) was used to compare treatment groups with respect to both efficacy and toxicity involving categorical variables (for example, demographic characteristics), whereas quantitative variables (for instance, age) were subjected to a one-way analysis of variance. To test for any interaction of the study centers on outcome at the end of treatment, centers were grouped as follows in order to achieve large enough groups: Canada and Australia; the United States; the British Isles; the Netherlands; Finland and Sweden; France and Spain; and Germany, Israel, and Korea.

All treatment episodes with at least one dose of study regimens were included in the assessment of toxicity, which took account of all adverse events likely to be related to the study treatment.

The Fisher exact test was used to analyze data sets where any of the values was small.

Rates of satisfactory response for all treatment episodes and the corresponding microbiological outcome were expressed as the odds ratios of ceftazidime to the combination when a value of more than 1 indicated superiority of ceftazidime, whereas a value below unity indicated the converse. Lack of statistical significance was inferred when the 95% CIs encompassed unity, although formal analysis was also undertaken by calculating chi-square without a correction factor.

The time to defervescence was analyzed using the Wilcoxon test and the time to first modification was presented as a Kaplan-Meier curve and was analyzed using the log-rank test. All analyses were done using a two-sided significance level of 5%.

Results

Top Methods Results Discussion Author & Article Info References

Demographic Characteristics

From 1988 to 1989, 35 institutions enrolled 696 neutropenic patients for 876 episodes of fever. Although 266 more episodes were entered into the parallel strata of the protocol in which either vancomycin or metronidazole was given concomitantly at the start of the study treatment, they are not included in this report. Ninety-two episodes were excluded from analysis, comprising 49 episodes allocated to ceftazidime and 43 allocated to combination therapy with piperacillin and tobramycin (not meeting the entry criteria, 27 compared with 26; given systemically active antiviral or antifungal agents on the day of entry, 21 compared with 14; and major protocol violations, 1 compared with 3). Treatment groups were well balanced with respect to age, sex, underlying disease, degree and duration of granulocytopenia, and the use of oral prophylaxis and central venous lines (Table 1). Three hundred eighty-nine episodes treated with ceftazidime alone and 395 episodes treated with piperacillin and tobramycin were evaluable. The distribution of infections, including bacteremias, was similar for both groups, as was the mean temperature at entry: 39.5 °C (CI, 39 to 40 °C) for ceftazidime and 39.2 °C (CI, 38.8 to 39.6 °C) for the other treatment group. Ninety-eight percent of the gram-negative isolates were susceptible to ceftazidime and 75% were susceptible to one or both of the antibiotics in the combination. Of the gram-positive isolates, 73% were susceptible to ceftazidime and 78% were susceptible to one or both of the antibiotics in the combination. Patients were treated with ceftazidime for a mean of 17.0 days (CI, 15.9 to 18.1 days) and with the combination for 16.6 days (CI, 15.5 to 17.7 days).

Clinical Results

Within the first 72 hours of treatment, 3 (0.8%) patients with episodes treated with ceftazidime died compared with 6 (1.7%) treated with the combination therapy. At this stage, therapy had already been modified in 32% of 386 episodes treated with ceftazidime and in 33% of 389 episodes treated with the combination. Defervescence had occurred in only 115 of the 244 episodes (47%) that ultimately responded without any modification to the study regimen.

At the end of treatment, 22 episodes treated with ceftazidime and 40 given the combination were considered not assessable according to the classification of Pizzo and colleagues [17]. Of 367 assessable episodes of treatment with ceftazidime, 35% were registered as successes compared with 33% of 355 episodes treated with piperacillin and tobramycin Table 2. Twenty-one episodes (6%) were fatal during treatment with ceftazidime compared with 29 (8%) during treatment with the combination therapy. Seven of the deaths occurring in the ceftazidime group were caused by superinfection compared with 14 of those in the combination therapy group. Both regimens produced better results in patients who had solid tumors or lymphoma compared with patients with acute leukemia or bone marrow transplants (Table 2).

Changes in therapy were made at the same rate in both groups (Figure 2). Reasons for modification (n = 219 in the ceftazidime group, n = 209 in the combination therapy group) were as follows: continuing or recurrent fever, 94 (43%) compared with 91 (44%); suspected nonbacterial infection, 36 (16%) compared with 27 (13%); presumed or proven resistance of the primary pathogen, 23 (11%) compared with 39 (18%); new fever, clinical deterioration, clinical infection, or pathogen, 52 (24%) compared with 42 (20%); treatment discontinued and prophylaxis started, 11 (5%) compared with 10 (5%); and adverse events, 3 (1%) compared with none.

View larger version (9K): [in this window] [in a new window]   Figure 2. Kaplan-Meier plot of the time to first modification. The plots are effectively superimposed (P > 0.2).

The most frequent antimicrobial modification in both groups involved vancomycin, constituting 96 of the 219 (44%) modifications in the ceftazidime group and 94 of the 209 (45%) in the piperacillin-tobramycin group. Systemically active antifungal compounds comprised 55 (25%) of the modifications to the single agent and 48 (23%) of the combination therapy, whereas the figures for acyclovir were 24 (11%) and 17 (8%), respectively. The addition of an aminoglycoside accounted for 31 (14%) of the modifications in the ceftazidime group, whereas in the combination group, tobramycin was replaced by another aminoglycoside in 4 (2%) cases. Other ß-lactam antibiotics were used to modify 15 cases (7%) in the ceftazidime group and to modify 13 cases (6%) in the combination group.

No differences were found between the two treatment groups in any subgroup of infection. Thirty-three of 118 (28%) clinically assessable episodes of bacteremia were classified as successes without modification to ceftazidime therapy, whereas 25 of 132 (19%) were classified as successes without modification to piperacillin and tobramycin; 76 (64%) and 89 (67%), respectively, were classified as successes with modification, and 9 (8%) and 18 (14%), respectively, were considered failures. Among the episodes of clinically documented infection or microbiologically proven infection without bacteremia, the results were more favorable: Twenty-four of 63 (38%) episodes treated with ceftazidime were scored as successes compared with 26 of 65 (40%) treated with piperacillin and tobramycin. Because of a high number of modifications for persistent fever only, similar response rates were found for episodes of unexplained fever: Seventy of 186 (38%) in the ceftazidime group were registered as successes without modification compared with 66 of 158 (42%) in the piperacillin-tobramycin group.

Outcome in Terms of Satisfactory Response

A satisfactory response was obtained in 230 of 367 episodes (62.7%) treated with ceftazidime compared with 217 of 355 (61.1%) treated with the combination therapy (Table 3). The odds ratio of ceftazidime over the combination was 1.07 (CI, 0.79 to 1.44; P > 0.2), showing the equivalence of the regimens. The rate of defervescence was also similar for both treatment groups (Figure 3). Likewise, no differences between the treatment groups were observed with respect to the underlying diseases and infection categories; the corresponding odds ratios were as follows: documented infections, 1.01 (CI, 0.68 to 1.52; P > 0.2); unexplained fever, 1.03 (CI, 0.65 to 1.65; P > 0.2); acute leukemia and bone marrow transplants, 1.19 (CI, 0.86 to 1.66; P > 0.2); solid tumor and lymphoma, 0.74 (CI, 0.33 to 1.66; P > 0.2). No difference in outcome was seen between the comparative regimens among the different regions, but the proportion of satisfactory responses varied widely (Figure 4). The percentage of satisfactory responses for episodes treated with ceftazidime was 50% in France and Spain compared with 52% for those treated with piperacillin plus amikacin, whereas the corresponding figures for the other regions were as follows: United States, 60% compared with 65%; Australia and Canada, 60% compared with 65%; the British Isles, 64% compared with 63%; Germany, Israel, and Korea, 62% compared with 58%; The Netherlands, 74% compared with 58%; and Finland and Sweden, 66% compared with 63%.

View larger version (10K): [in this window] [in a new window]   Figure 3. Cumulative rate of defervescence. Only around one third of patients had become afebrile by day 3 of treatment, whereas rates of defervescence were virtually identical (P > 0.2).

View larger version (17K): [in this window] [in a new window]   Figure 4. Comparison of satisfactory responses among the different regions. Odds ratios are shown together with their 95% CIs, indicating that neither regimen was favored.

In episodes characterized by profound granulocytopenia (<100 cells per mm³ for at least 2 weeks during treatment), the response was considered satisfactory in 52 of 103 (50%) episodes treated with ceftazidime compared with 55 of 96 (57%) treated with the combination therapy (odds ratio, 0.76; CI, 0.43 to 1.33; P > 0.2); one patient in the ceftazidime group and two in the combination therapy group died of infection during this period of profound granulocytopenia.

Microbiological Outcome for Bacteremia

Overall, eradication of 72 of 91 (79%) microorganisms was achieved with ceftazidime and 73 of 107 (68%) with piperacillin and tobramycin (odds ratio 1.76; CI, 0.92 to 3.38; P = 0.08) (Table 4). The corresponding odds ratios for the eradication of gram-negative bacteria and gram-positive bacteria were 5.25 (CI, 1.0 to 27.5; P = 0.03) and 1.37 (CI, 0.6 to 3.13; P > 0.2). When all signs and symptoms related to gram-negative infection resolved, neutropenia with granulocyte counts of 500/mm³ or less persisted in 31 episodes treated with ceftazidime (82%) and in 25 of those treated with the combination (80%). Resistance to the study regimen emerged once in each treatment group with organisms that initially had been responsible. The patient given ceftazidime relapsed with bacteremia due to Enterobacter cloacae whereas both Escherichia coli and Klebsiella pneumoniae bacteremia recurred in the patient treated with the combination. No relapses with susceptible organisms were observed.

Tobramycin Levels

At least one tobramycin assay was done during the first 3 days of treatment in 354 patients (90%). The mean 1-hour peak and corresponding trough levels were 5.3 ±0.26 mg/L and 1.2 ±0.1 mg/L, respectively. The dose of tobramycin was adjusted in 173 cases (44%) during the course of treatment to bring levels within the therapeutic range.

Adverse Events

These problems were observed 2.9 times more frequently (CI, 1.9 to 4.6) in the piperacillin-tobramycin group compared with ceftazidime (P < 0.001), with allergic reactions, mostly skin rash, occurring most often (Table 5). The incidence of renal toxicity was also significantly higher in the combination therapy group (P < 0.001); five patients required hemodialysis and one patient died with renal insufficiency. Clinically apparent ototoxicity was limited to the group who received the combination therapy, with five patients reporting hearing loss or tinnitus or both.

Subsequent Infections

Seven episodes in the monotherapy group and 14 in the combination group were fatal because of superinfection. In addition, a microbiologically proven superinfection in the blood or lung occurred in 31 episodes in the ceftazidime group and in 24 episodes in the combination therapy group. The organisms involved are shown in Table 6; during some episodes more than one microorganism was isolated.

Enterocolitis associated with Clostridium difficile occurred in seven episodes treated with ceftazidime compared with four treated with piperacillin and tobramycin.

One relapse occurred in the ceftazidime group with bacteremia caused by E. cloacae and one in the combination group with bacteremia caused by E. coli and K. pneumoniae. In addition, a new bacteremia caused by Enterococcus faecalis occurred once in the ceftazidime group after treatment had been discontinued and twice in the combination therapy group involving coagulase-negative staphylococci. All patients with relapses and new bacteremias survived after receiving appropriate therapy.

Discussion

Top Methods Results Discussion Author & Article Info References

We designed our study to formally address the issue of monotherapy, given the objections to previously published investigations such as inadequate numbers of patients and the inclusion of those at relatively low risk. The response rates obtained matched those that had been anticipated; therefore, we can conclude with sufficient power that the two regimens are equally effective. The results are also applicable to high-risk patients because most patients had been treated for acute leukemia that had resulted in prolonged, profound neutropenia. Moreover, of 322 microbiologically documented infections, 262 (81%) were bacteremias. Seventy-seven episodes were caused by single gram-negative bacilli, which substantially exceeds those reported previously for other trials of empiric therapy in febrile neutropenic patients with cancer. Both empiric regimens were shown to be at least equivalent in terms of rates of satisfactory response or success without modification and in ensuring survival for patients with these microbiologically documented infections as well as those in whom the cause of the initial infection was not established. The rates of defervescence underscored the similarity in efficacy as did the bacteriologic eradication rates.

Multiple entry into the study was permitted for separate neutropenic episodes and involved 23% of patients. Although this may deviate somewhat from other therapeutic trials, it is common practice in those involving neutropenic patients [22]. It has been argued that only a single exposure to a regimen is statistically valid. However, the drugs we investigated were chosen precisely because they were the most widely used, and the first entry into the study was not necessarily the patient's first exposure to these antibiotics nor was it their first infective episode. Each episode in this study also corresponded with a distinct period of neutropenia induced at a different stage of the underlying disease. It therefore seems reasonable to assume that such episodes are biologically independent because each cycle of chemotherapy induces different degrees of mucosal damage and myelotoxicity which, in turn, produce different infective risks. There also seems only a remote possibility of a "carry-over" effect because the minimum interval between treatment episodes was 1 week and, given the short-half lives of the drugs, any residual tissue concentrations would be neglible. Finally, each patient had to satisfy the criteria for eligibility on each occasion that excluded anyone who had developed hypersensitivity or chronic organ dysfunction after exposure to one or another of the study drugs during the preceding entry into the study.

Previously, single agents have only been acknowledged to constitute adequate empiric therapy for unexplained fevers and not for documented infections or episodes with prolonged neutropenia for which a combination was considered necessary [1, 18, 22, 23]. However, not only did microbiologically documented infections respond equally well to both regimens, but ceftazidime actually produced more favorable results in bacteremias caused by gram-negative bacteria, accomplishing an eradication rate of 95% compared with 77% for the combination therapy. It could be argued that a better test of the feasibility of monotherapy might have been provided if ceftazidime, rather than piperacillin, had been combined with an aminoglycoside. However, both the microbiological outcome and satisfactory response rates for ceftazidime alone corresponded with those reported for the long course of amikacin plus ceftazidime in the study performed by the European Organization for Research and Treatment of Cancer (EORTC) International Antimicrobial Therapy Cooperative Group [1], and the survival rates we found were also similar to those reported by the National Cancer Institute [17] and most other trials [24-29]. When ceftazidime is used alone at a dose of 2 g every 8 h for the empiric treatment of febrile episodes in prolonged neutropenia, the development of resistance does not appear to be a major concern. During our study, this occurred only in one of 38 (3%) cases with gram-negative bacteremia, involving E. cloacae. Therefore, ceftazidime alone appears to be as feasible an option as a standard combination particularly as, at the onset of fever, it is virtually impossible to distinguish between microbiologically documented and other categories of infection because the results of cultures are not available until 2 or 3 days later.

In contrast to most trials, the protocol for the present study allowed the local investigator considerable discretion in modifying initial therapy to conform as much as possible to normal clinical practice [30]. The liberal guidelines for modifying the study regimens undoubtedly were responsible for the higher-than-usual number of unnecessary modifications in our trial, but no differences were found between both treatment groups because therapy was modified at essentially the same rate for similar reasons, using an almost identical choice of antimicrobial agents. Of 428 episodes in which therapy was adjusted, 41% of modifications were made because of continuing fever without any evidence of clinical deterioration. More than half of the patients who were ultimately scored as successes were still febrile after 3 days of therapy, and treatment was modified as frequently for episodes of documented infection as for unexplained fever. These data suggest that continuing fever alone provides an unsatisfactory basis for altering antibacterial therapy and seems more a reflection of the lack of confidence on the part of the investigator rather than any deficiency of the regimens. On the other hand, the development of subsequent infective events, such as new infection, emphasizes that a genuine need for modification does, in fact, frequently exist independent of the initial regimen used. The diversity in satisfactory response rates among the various regions is hard to explain but may be related to local differences in the approach to managing the persistently febrile neutropenic patient.

The combination of piperacillin and tobramycin was associated with significant toxicity compared with ceftazidime, particularly with respect to renal function. Aminoglycoside agents were added in only 14% of cases in the ceftazidime group, of only two patients had persistent gram-negative infection. Therefore, as many as 86% of febrile, neutropenic patients do not appear to require aminoglycosides, which are known to be toxic to the ear [31] and kidney [32, 33]. The use of a ß-lactam-aminoglycoside combination would only seem justifiable if there is a high chance of ß-lactam resistance such as occurs with E. cloacae [34]. Concerns about toxicity have become important with the increasing tendency to co-administer drugs like cyclosporin A, cis-platinum, and amphotericin B in this group of patients. Any reduction in toxicity would help manage costs. Another advantage of monotherapy is that it is also less labor intensive to administer [25]. Once-daily administration of a combination of antibiotics is another option to cut costs, but one should be careful in using a regimen where Pseudomonas aeruginosa is not appropriately covered by a ß-lactam [35]. Indeed, the results obtained with the combination of ceftriaxone and amikacin for this pathogen, whose incidence probably temporarily and inexplicably has decreased, were not encouraging because only one of five episodes showed a response [22].

The results of our study extend the findings of a meta-analysis of previous trials with ceftazidime [36] in showing that monotherapy with this compound is as effective as a standard combination with ß-lactam agents. Because ceftazidime is safer, it is therefore at least as appropriate a choice as the combination of piperacillin and tobramycin for the empiric treatment of the febrile patient without evidence of staphylococcal or anaerobic infection, even in those with profound and prolonged granulocytopenia.

Appendix 1. Participating Centers and Committees

The Intercontinental Antimicrobial Study Group comprised 35 participating centers, mainly in Europe and North America, and the following committees: Protocol Writing Committee: B.E. De Pauw, J.P. Donnelly, and E.F. Lane-Allman; Study Coordinators and Data Review Committee: B.E. De Pauw, Chair, S.C. Deresinski, J.P. Donnelly, and R. Feld; Project Coordinator for Glaxo Group Research: E. F. Lane-Allman; Data Assembly and Validation: Data were computerized by Glaxo Research, Ltd., and validated externally by an independent company (Vital Information Ltd., London, United Kingdom). In addition, key items in the files of 5% of episodes, selected at random, were assessed for accuracy by J.P. Donnelly; Statistical Analysis: Mr. N. Elahi, Glaxo Group Research, Greenford, United Kingdom, and J.P. Donnelly (except for the Kaplan-Meier analysis, which was done by Professor A.L. Verbeek).

Appendix 2. Principal Investigators and Participating Centers

Europe: B.E. De Pauw, University Hospital St. Radboud, Nijmegen, the Netherlands; B. Oppenheim, Christie Hospital, Manchester, United Kingdom; J. Harrouseau, Hospital St. Jacques, Nantes, France; M. Valtonen, University Central Hospital, Helsinki, Finland; J. Westin, Sahlgrenska Hospital, Gothenburg, Sweden; R. Horn, Royal Victoria Hospital, Montreal, Canada; S. Rodjer, Ostra Hospital, Gothenburg, Sweden; B. Boughton, Queen Elizabeth Hospital, Birmingham, U.K.; D. Maraninchi, Institut J Poali-1 Calamettes, Marseille, France; S. McCann, St. James Hospital, Dublin, Ireland; J. Briere, Hospital Morvan, Brest, France; G. Ehninger, Medizinische Universitatsklinik, Tubingen, Germany; M. Westerhausen, St. Johannis Hospital, Duisburg, Germany; N. Russell, City Hospital, Nottingham, U.K.; E. Barrio, General De Galicia Hospital, Santiago de Compostela, Spain. Canada: R. Feld, Princess Margaret Hospital, Toronto; G.E. Garber, Ottawa General Hospital, Ottawa; L. Mandell, McMasters Medical Center, Toronto; R. Saginur, Ottawa Civic Hospital, Ottawa; M. Laverdiere, Hospital Maisonneuve-Rosemont, Montreal; P. Phaneuf, Hotel-Dieu de Montreal, Montreal; E. Bow, St. Boniface General Hospital, Winnipeg; S.D. Shafran, University Hospital, Saskatoon; A.R. Rachlis, Sunnybrook Medical Center, Toronto; W. Lofters, Kingston Regional Cancer Center, Ontario; D. Haase, Victoria General Hospital, Halifax. United States: C.A. Linker, University of California, San Francisco; S. Deresinski, L. Levitt, Stanford University School of Medicine, Palo Alto; P. Geiseler, City of Hope National Medical Center, Duarte; R. Chinn, Sharp Hospital, San Diego; W. Lau, St. Francis Hospital, Honolulu. Other countries: K.W. Choe, National University Hospital, Seoul, South Korea; T. Sacks, Hadassah Medical Center, Jerusalem, Israel; R. Zaizov, Beilinson Medical Center, Petal Tikva, Israel; J. Biggs, St. Vincents Hospital, Sydney, Australia.

Presented as abstract number 116 at the Sixth International Symposium of Infections in the Immunocompromised Host, 3-6 June 1990, Peebles, Scotland, The United Kingdom.