Fax: (713) 792-8233
Treatment of febrile neutropenic patients with cancer who require hospitalization
A prospective randomized study comparing imipenem and cefepime
Version of Record online: 17 JUL 2003
Copyright © 2003 American Cancer Society
Volume 98, Issue 5, pages 1039–1047, 1 September 2003
How to Cite
Raad, I. I., Escalante, C., Hachem, R. Y., Hanna, H. A., Husni, R., Afif, C., Boktour, M. R., Whimbey, E. E., Kontoyiannis, D., Jacobson, K., Kantarjian, H., Levett, L. M. and Rolston, K. V. I. (2003), Treatment of febrile neutropenic patients with cancer who require hospitalization. Cancer, 98: 1039–1047. doi: 10.1002/cncr.11613
- Issue online: 20 AUG 2003
- Version of Record online: 17 JUL 2003
- Manuscript Accepted: 4 JUN 2003
- Manuscript Revised: 2 JUN 2003
- Manuscript Received: 17 MAR 2003
- Bristol-Myers Squibb (New York, NY)
- febrile neutropenia;
- cancer patients
The objective of the current study was to compare the efficacy and safety of imipenem and cefepime in the treatment of adult patients with cancer who had fever and neutropenia requiring hospitalization according to Infectious Disease Society of America criteria.
In the current prospective randomized clinical trial at a university-affiliated tertiary cancer center, adult patients with cancer who had fever (≥ 38.3 °C or ≥ 38.0 °C for > 2 hours) and neutropenia (≤ 500/mm3 or < 1000/mm3 but declining) requiring hospitalization were randomized to receive either cefepime or imipenem. Vancomycin or amikacin was added on suspicion of gram-positive or gram-negative bacterial infection, respectively.
Patients who received an imipenem regimen or a cefepime regimen were comparable in terms of age, gender, underlying malignancy, prior transplantation, degree and trend of neutropenia, and presence of central venous catheters (P ≥ 0.3). An intent-to-treat analysis showed a 68% response rate to the imipenem regimen, compared with a 75% response rate to the cefepime regimen (P = 0.2). The rates of antibiotic-related adverse events and superinfections also were comparable (P = 0.6). There was no difference in response among patients who received imipenem or cefepime alone compared with patients who also received vancomycin or amikacin (P = 1.0). Leukemia was the only independent risk factor associated with a poor outcome (odds ratio, 4.6; 95% confidence interval, 1.9–10.7; P < 0.0001).
Imipenem and cefepime had similar efficacy and safety profiles in the treatment of adult cancer patients with fever and neutropenia who required hospitalization. The addition of either vancomycin or amikacin may not be necessary. Cancer 2003;98:1039–47. © 2003 American Cancer Society.
When treated inappropriately, infections in high-risk patients with cancer who have neutropenia can follow a rapidly progressive, fulminant course that often is associated with a high mortality rate.1 Studies in the 1960s and 1970s showed that despite in vitro susceptibility, neither aminoglycosides nor first-generation and second-generation β-lactams could be used alone in neutropenic patients for successful treatment of gram-negative infections. This finding led to the development of combination therapies that usually involved an aminoglycoside and an antipseudomonal β-lactam to optimize empiric antiinfective therapy.2, 3 Over the past 2 decades, there has been a shift in the microbiology of infections in neutropenic febrile patients from a predominance of gram-negative organisms to gram-positive organisms.4–7 However, the emergence of vancomycin-resistant gram-positive organisms, particularly in patients with cancer, led the Infectious Disease Society of America (IDSA) to publish guidelines for the use of monotherapy or combination therapy in high-risk, neutropenic, febrile patients.8, 9 In hospitalized patients with fever and neutropenia, it has been shown that imipenem is an efficacious single agent.10, 11 Cefepime, as a fourth-generation cephalosporin, has broad-spectrum in vitro activity against some gram-positive organisms and most gram-negative organisms. It also has demonstrated effectiveness as a monotherapeutic agent in the treatment of patients with fever and neutropenia.12 In the 2002 IDSA guidelines, each of the two agents (imipenem or cefepime) was suggested as a frontline antibiotic (alone or in combination) in the treatment of high-risk patients with febrile neutropenia.9 The prospective randomized study described in the current report compares the efficacy and safety of these two agents with the goal of evaluating the validity of the IDSA guidelines regarding when to use monotherapy versus combination therapy to treat patients with cancer who have moderate-to-high-risk febrile neutropenia requiring hospitalization.8, 9
MATERIALS AND METHODS
Between May 1996 and April 2001, adult patients with cancer who became febrile during an episode of neutropenia and required hospitalization were evaluated for placement on the current study. The criteria for hospitalization were based on classifying these patients as at moderate risk or high risk for febrile neutropenia according to guidelines published by Rolston.13 Patients were considered to have significant fever if their oral temperature was ≥ 38.3 °C or if they had a temperature of ≥ 38.0 °C for > 2 hours during a 24-hour period that was not related to the administration of blood products, underlying tumor, or other pyrogenic substances. Neutropenia was defined as a neutrophil count < 500/mm3 or < 1000/mm3 if the patient's count was expected to fall below 500/mm3 neutrophils due to prior chemotherapy.5, 6 A medical history and a physical examination were obtained at study entry. The protocol for this study was approved by The University of Texas M. D. Anderson Cancer Center Institutional Review Board. Exclusion criteria included age younger than 18 years, low risk for neutropenia and fever as defined by Rolston,13 previous anaphylactic reaction to β-lactam antibiotics, prior antibiotic therapy for this febrile episode, severe renal insufficiency (creatinine clearance < 15 mL/minute), severe hepatic impairment (aspartate aminotransferase/alanine aminotransferase levels > 6-fold normal), infections caused by organisms known to be resistant to the study drugs, pregnancy (or being in the process of nursing), state of shock (systolic blood pressure < 80 mmHg) unresponsive to fluid replacement, previous treatment on the current study for the same episode, history of seizures, known brain metastasis or primary brain tumor, and current use of anticonvulsant therapy.
Before the initiation of antibiotics, culture specimens were obtained from the throat, urine, blood, sputum (if available), and other appropriate sites. A chest roentgenogram and a urinalysis were obtained within the first 24 hours after treatment initiation. Blood culture specimens were obtained daily as long as the patient remained febrile during the course of therapy. In vitro susceptibility of the infecting organisms to the study drugs was tested using a previously described microtiter broth dilution method.11 Fever of unknown origin was diagnosed when clinical, microbiologic, or radiographic evaluation failed to attribute the patient's fever to any infected site or microbial organism. A documented infection was diagnosed if the fever could be attributed to a clinical site of infection, such as a pulmonary infiltrate, or if the infecting organism(s) could be isolated.
The treatment plan was designed in accordance with the 1997 IDSA guidelines.8 Patients were assigned randomly to one of the following antibiotic regimens: 1) imipenem/cilastatin 500 mg administered intravenously (IV) over 1 hour every 6 hours or 2) cefepime 2.0 gm administered IV over 0.50 hours every 8 hours. In addition, vancomycin 1 gm administered IV every 12 hours was given to patients who presented with painful mucositis, clinically suspected catheter infection (catheter-site inflammation), prior colonization with penicillin-resistant pneumococci or methicillin-resistant Staphylococcus aureus, or known positive blood culture for gram-positive bacteremia. Amikacin at a dose of 15 mg/kg per day also was given in addition to either imipenem or cefepime to patients with pneumonia, documented Pseudomonas aeruginosa infection, and sepsis syndrome.
Appropriate dosage modifications were made on the basis of renal clearance (and weight for patients in the imipenem/cilastatin regimen) according to guidelines written in the package inserts for the respective antibiotics. Vancomycin was discontinued if cultures failed to show any vancomycin-susceptible organism at or before Day 4 of treatment. Randomization was done in an evaluator-blind fashion. The treatment assignments remained unknown to the investigators in charge of determining eligibility, establishing infectious diagnoses, and evaluating outcome and adverse events throughout the study. Responding patients received 7 days of antibiotics or 4 days of antibiotics after becoming afebrile, provided all signs and symptoms of infections had resolved. Patients who became afebrile after at least 48 hours on study regimen or who completed at least 4 days of study regimen with clinical improvement were considered eligible for discharge with an oral antibiotic regimen to complete 7 days of total antibiotic therapy. The oral antibiotic regimen consisted of ciprofloxacin 500 mg every 8 hours with amoxicillin or clavulanic acid 500 mg every 8 hours or, if the patient was allergic to penicillin, clindamycin 600 mg every 8 hours.
The published guidelines for the evaluation of new antiinfective drugs in the treatment of febrile episodes in patients with neutropenia were used as the basis for defining the endpoint assessment of therapeutic outcome.14 Response was defined as the eradication of all symptoms, signs, and microbiologic evidence of infection that could be attributed to the study regimen. Treatment failure was defined as no improvement in the infection or worsening of the infection while the patient was receiving the initial study regimen and necessitation of the addition of any antibacterial drug. Patients with neutrophil counts that failed to drop below 500 mm3, patients with documented fungal or viral infections, and patients who received concurrent antibiotics other than the study drugs were considered inevaluable. Superinfection was defined as an infection that occurred during treatment or 2 weeks thereafter and was caused by a different organism. Adverse events were considered drug-related if they had a temporal relation to the study drug with no apparent alternative etiology and if the effect improved after the drug was discontinued.
On the basis of results from a previous study conducted at our center, it was estimated that the worst response rate with either of these regimens would be 60%. A 25% improvement in the response rate would be considered significant. Detection of a 25% improvement in the response rate with a type I error (an α error) of 0.05 and with a type II error (or β error) of 0.02 would require approximately 108 evaluable patients in each arm. Based on our experience from similar studies conducted at our center, we estimated an evaluability rate of 86%. Hence, it was postulated that 251 patients needed to be entered into the current study. All statistical tests were performed as two-tailed tests. Differences between frequencies of categoric variables were determined using the chi-square test or the Fisher exact test, as appropriate. Continuous variables with a normal distribution were compared using the Student t test. Continuous variables that were not distributed normally were compared using the Mann–Whitney test. Factors significant at a P value > 0.25 in a univariate analysis of outcome related to specific characteristics of evaluable episodes were entered into a multivariate logistic regression model. Computations of 95% confidence intervals were performed for differences in outcomes between the two treatment regimens. A P value < 0.05 was considered significant.
Table 1 shows that there were no significant differences between the two groups of evaluable patients in terms of age, gender, underlying disease, bone marrow transplantation in the preceding year, prophylactic use of antibiotics within 1 week of study entry, neutropenia at onset of study, persistent neutropenia, and presence of a central venous catheter. Most patients (n = 169 [67%]) had an underlying hematologic malignancy (leukemia, lymphoma, or myeloma), and 137 patients (55%) were profoundly neutropenic at study entry (neutrophils < 100/mm3).
|Characteristic||No. of patients (%)||P value|
|Imipenem (n = 124)||Cefepime (n = 127)|
|Men||66 (53)||74 (58)||—|
|Women||58 (47)||53 (42)||0.4|
|Leukemia||55 (44)||62 (49)||—|
|Lymphoma/myeloma||27 (22)||25 (20)||—|
|Solid tumor||42 (34)||40 (32)||0.8|
|Bone marrow transplantation (in the preceding year)||7 (6)||4 (3)||0.3|
|Prophylaxis antibiotic (within 1 week of study)||12 (10)||11 (9)||0.8|
|Neutropenia at onset|
|< 100/mm3||69 (56)||68 (54)||—|
|100–499/mm3||41 (33)||42 (33)||—|
|500–1000/mm3||8 (7)||13 (10)||—|
|> 1000/mm3||6 (5)||4 (3)||0.7|
|Persistent neutropenia < 500/mm3||47 (38)||46 (36)||0.8|
|Central venous catheter||115 (93)||119 (94)||0.8|
Response to Therapy
During the 38-month study period, 251 patients were randomized to the treatment protocol, with 127 patients in the cefepime arm and 124 patients in the imipenem arm (Fig. 1). Most patients (n = 176 [70%]) received monotherapy with either cefepime (n = 90 [36%]) or imipenem (n = 86 [34%]). Combination therapy was administered as follows: 31 patients (12%) received cefepime plus vancomycin, 6 patients (2%) received cefepime plus amikacin, 30 patients (12%) received imipenem plus vancomycin, and 8 patients (3%) received imipenem plus amikacin. Two hundred fourteen episodes were considered evaluable.
The intent-to-treat analysis revealed a 68% response rate to the imipenem-containing regimen compared with a 75% response rate to the cefepime-containing regimen (P = 0.2). Response to therapy (Table 2), in all evaluable patients as well as broken down by underlying infection type, was comparable between the two groups. There was no difference in the general response to therapy between patients who had documented gram-positive infections and those who had gram-negative infections. The distribution of specific organisms also was comparable between the two groups, with the majority of gram-positive infections involving the coagulase-negative staphylococci S. aureus and S. viridans. The most common gram-negative organism was the Pseudomonas species.
|Variable||Imipenem (n = 102)||Cefepime (n = 112)||P value|
|No. of episodes||No. of responders (%)||No. of episodes||No. of responders (%)|
|Intent-to-treat analysis||124||84 (68)||127||95 (75)||0.2|
|Total evaluable patients||102||72 (71)||112||84 (75)||0.5|
|Fever of unknown origin||33||25 (76)||43||36 (84)||0.4|
|Documented infections||69||47 (68)||69||48 (70)||0.9|
|Organisms identified||46||34 (74)||54||42 (78)||0.7|
|No organisms identified||23||13 (57)||15||6 (40)||0.3|
|Monomicrobial gram-positive||25||17 (68)||37||29 (78)||0.4|
|Monomicrobial gram-negative||13||12 (92)||12||8 (67)||0.2|
|Polymicrobial||8||5 (63)||5||5 (100)||0.2|
|Documented infections (with or without organism identified)|
|Bacteremia||31||22 (71)||37||30 (81)||0.3|
|Pneumonia||16||7 (44)||18||8 (44)||1.0|
|Catheter-related||15||12 (80)||9||6 (67)||0.6|
|Other||7||6 (86)||5||4 (80)||1.0|
The response to antibiotic therapy was comparable irrespective of underlying disease and neutropenia status (Table 3). There was a trend toward a better response to cefepime for patients with persistent or declining neutropenia (P = 0.1). Patients with underlying leukemia had a lower response rate to antimicrobial therapy. Factors that may have influenced outcome, such as age, underlying disease, persistent or declining neutropenia, antimicrobial prophylaxis within 1 week of study entry, bone marrow transplantation within the year preceding study entry, presence of documented infection, site of infection, and type of treatment (cefepime or imipenem) all were evaluated by univariate analysis, which showed that underlying leukemia and persistent or declining neutropenia were associated significantly with poor outcome. All of the factors listed above were entered into a multivariate logistic regression model, and leukemia was the only risk factor found to be associated with poor outcome (odds ratio, 4.6; 95% confidence interval, 1.9–10.7; P < 0.0001).
|Variable||Imipenem (n = 102)||Cefepime (n = 112)||P value|
|No. of episodes||No. of responders (%)||No. of episodes||No. of responders (%)|
|Leukemia||44||24 (55)||58||39 (67)||0.2|
|Lymphoma/myeloma||24||18 (75)||21||15 (71)||0.8|
|Solid tumor||34||30 (88)||33||30 (91)||1.0|
|Persistent or declining neutropenia||42||24 (57)||44||32 (73)||0.1|
|Recovering neutropenia||60||48 (80)||68||52 (77)||0.6|
Monotherapy versus Combination Therapy
Table 4 shows that there was no difference in outcome between patients who received monotherapy with either cefepime or imipenem alone compared with patients who received a combination of cefepime or imipenem with either vancomycin or amikacin. Most of the documented gram-positive and gram-negative infections were not predicted on the basis of the IDSA criteria. Hence, most of the patients with documented gram-positive (71%) or gram-negative (96%) infections received monotherapy rather than combination therapy (Table 3). Patients with gram-positive infections who received monotherapy had responses comparable to the responses of patients with the same types of infections who received combination therapy containing vancomycin (P = 1.00). Similarly, patients with gram-negative infections who received monotherapy with either cefepime or imipenem had a high response rate (79%).
|Variable||No. of patients||No. of responders (%)||P value|
|Combination therapy||63||44 (70)||0.5|
|Vancomycin-containing regimen||18||13 (72)||1.00|
|Amikacin-containing regimen||1||1 (100)||1.00|
Toxicity and Superinfections
The antibiotic-related toxicities were comparable between the two groups (Table 5). Patients who received the imipenem-containing regimen had 17 episodes (14%) of antibiotic-related adverse events, compared with 13 episodes (10%) among patients who received the cefepime-containing regimen (P = 0.4). The most common adverse events were gastrointestinal symptoms (e.g., diarrhea, nausea, and emesis), which were distributed equally between the two groups. The number of rashes and seizures were higher in patients who received the imipenem-containing regimen; however, the number of rashes and seizures was not significantly different from what was observed in the cefepime group. It should be noted that there were two imipenem-related seizure events, whereas there were no central nervous system–related adverse events associated with cefepime.
|Event||No. of patients||P value|
|Antibiotic-related adverse events|
|Total (%)||17 (14)||13 (10)||NS|
|Total superinfections (%)a||14 (12)||12 (10)||NS|
The frequency of superinfections was comparable in the two study groups. There was a slightly higher rate of Clostridium difficile infection in patients who received imipenem compared with patients who received cefepime. In addition, all three of the Stenotrophomonous maltophilia superinfections were associated with cefepime treatment but not with imipenem; however, these differences were not statistically significant (Table 5).
Susceptibility of Isolated Organisms
The 73 gram-positive organisms and the 26 gram-negative organisms that were isolated from sterile body fluids of the neutropenic febrile patients in the current study had comparable susceptibility patterns to either cefepime or imipenem. Most of the observed gram-positive organisms were susceptible to cefepime (61 of 73 patients [84%]) and imipenem (60 of 73 patients [82%]; P = 0.8). Similarly, most gram-negative organisms in the current study also were susceptible to cefepime (61 of 73 patients [84%]) and imipenem (60 of 73 patients [82%]).
The current study demonstrated that imipenem and cefepime, when used in accordance with the IDSA guidelines for the treatment of moderate-to-high-risk, febrile, neutropenic patients who required hospitalization, were associated with high and comparable efficacy. In addition, these two antibiotic regimens were well tolerated and were associated with equivalent adverse events and frequency of superinfections.
Imipenem, as a monotherapeutic agent, has a long record of efficacy when used to treat febrile, neutropenic patients requiring hospitalization.10, 11, 15–28 At The University of Texas M. D. Anderson Cancer Center, a large prospective randomized trial showed that imipenem was more effective than ceftazidime alone and was as effective as imipenem combined with amikacin in the treatment of febrile, neutropenic patients requiring hospitalization.11 Other investigators have confirmed our results: Liang et al.20 demonstrated the superiority of imipenem over ceftazidime, and Freifeld et al.27 showed that when imipenem was compared with ceftazidime without modification, imipenem was associated with a significantly higher frequency of defervescence at 72 hours.
Like imipenem, cefepime has been recommended by the IDSA as a monotherapeutic alternative to ceftazidime in the treatment of high-risk, febrile, neutropenic patients.9 Cefepime, as a fourth-generation cephalosporin, has broad-spectrum activity against gram-positive organisms, such as methicillin-sensitive S. aureus and penicillin-resistant S. viridans,29, 30 and against gram-negative bacilli, such as Enterobacter and Citrobacter, against which third-generation cephalosporins have limited activity.31 Compared with ceftazidime as a monotherapeutic agent, cefepime was associated with a lower frequency of glycopeptide additions in the treatment of microbiologically documented infection32, 33; however, this difference was not significant. Given the rising frequency of ceftazdime-resistant S. viridans, S. aureus, and Enterobacter organisms, cefepime is considered to be a potentially improved monotherapeutic alternative.9, 34
Biron et al.35 compared cefepime with imipenem for use as monotherapeutic agents in the treatment of febrile patients with a short duration of neutropenia (median duration, 4 days). The schedule used for cefepime was 2 gm IV every 12 hours, whereas the imipenem schedule was 1 gm IV every 8 hours. The two regimens were equivalent in efficacy in a low-risk population that excluded patients with leukemia; however, imipenem was associated with a greater frequency of adverse events compared with cefepime (19% vs. 9%, respectively; P = 0.003), particularly nausea and emesis. The current study differed from the study by Biron et al. in several aspects. First, the current patient population consisted of moderate-to-high-risk, febrile, neutropenic patients who required hospitalization and were associated with a prolonged duration of neutropenia. The low-risk patient population described by Biron et al. could have been treated successfully on an outpatient basis with oral antibiotics. Second, our dosing schedules of imipenem and cefepime were different from those used by Biron et al. Third, Biron et al. used cefepime and imipenem as monotherapeutic agents and did not allow for the initial addition of vancomycin or amikacin, whereas the current study occasionally required the addition of vancomycin or amikacin in accordance with the IDSA guidelines.8, 9 Despite these differences, the current study confirmed the hypothesis that cefepime is as efficacious as imipenem in the treatment of moderate-to-high-risk, neutropenic, febrile patients, including patients with documented gram-positive or gram-negative infections and patients with underlying leukemia associated with prolonged neutropenia (Tables 2, 3).
The current study provides evidence-based support for the use of monotherapy in treating moderate-to-high-risk, febrile, neutropenic patients with cancer. Table 4 shows that monotherapy with either cefepime or imipenem alone produced a response similar to the response observed when a treatment regimen that combined vancomycin or amikacin with either of the two β-lactam agents was employed. Because patients were not randomized to receive monotherapy or combination therapy, caution should be exercised in making conclusions about these results, because the two patient populations receiving the therapies may have been different. The 2002 IDSA guidelines recommend cefepime or imipenem as monotherapeutic agents in the treatment of high-risk, neutropenic, febrile patients with the addition of vancomycin or amikacin in patients at high risk for gram-positive or gram-negative infections, respectively. In the current study, in which the IDSA guidelines were followed very carefully, 56 of 80 patients (70%) with documented gram-positive infections received either cefepime or imipenem alone, respectively (Table 4). Hence, the IDSA criteria used to predict gram-positive infections and, thus, to add vancomycin failed to do so in 70% of patients who had documented infections.
The data from the current study are consistent with data from a previous review of 858 patients from 4 prospective randomized studies conducted at our institution over a 10-year period. Those studies showed that high-risk, febrile, neutropenic patients who received imipenem therapy had a similar response rate to that of patients who received combination therapy with imipenem plus amikacin or imipenem plus vancomycin.10 The same trend in response seems to apply to cefepime. The IDSA guidelines promote the judicious initial use of vancomycin on the basis of the patient's risk of acquiring a gram-positive infection. However, data from the current study support the conclusion made in a recent review that vancomycin is not required to be included in the initial regimen for treating febrile neutropenia in patients with cancer, particularly if agents that may have better activity against gram-positive organisms (e.g., cefepime and imipenem) are used instead.36 Vancomycin may be added after 48 hours or 72 hours if the patient is not responding to monotherapy with either cefepime or imipenem or if the isolated organism is resistant to the monotherapeutic agent used.
In the current study, there was no significant difference in drug-related toxicities and the frequency of superinfections between the cefepime regimen and the imipenem regimen. Seizures remain a serious and unique adverse event associated with the use of imipenem. The frequency of seizures in the study using low-dose imipenem (500 mg administered IV every 6 hours) was 1.6%, similar to the frequency of seizures in patients receiving high-dose imipenem, which ranged from 0.3% to 1.5%.10, 11 Although imipenem was associated previously with S. maltophilia superinfections,37 no such infections were associated with the use of imipenem in the current study. Paradoxically, all three of the superinfections with S. maltophilia were associated with the use of cefepime. This observation confirms our previous observation that if imipenem is used as part of an initial regimen to treat patients who do not receive an antecedent course of broad-spectrum antibiotics, then the risk of superinfection with S. maltophilia is low.10
Although the two regimens were equivalent in terms of efficacy and safety, cefepime had a slight cost advantage compared with imipenem. The average wholesale price for a single dose of imipenem is $33.10, whereas the cost for cefepime is $32.75. Given the fact that imipenem is given 4 times per day and cefepime is given 3 times per day, the average wholesale price per day for imipenem treatment is $132.40, compared with $98.25 for cefepime.
In summary, imipenem used alone as a monotherapeutic agent or in combination with vancomycin or amikacin was similar in terms of efficacy and safety to cefepime in treating adult patients with cancer who had a moderate or high risk of developing fever and neutropenia. In addition, cefepime provided a cost advantage relative to imipenem. Our data suggest that initial monotherapy would be as effective as initial combination therapy that involves the addition of either vancomycin or amikacin with either of these β-lactams in this patient population. Additions and modifications may be made subsequently rather than initially. According to our multivariate analysis, the only risk factor for poor outcome in the current patient population was underlying leukemia.
- 3Fever and neutropenia: an historical perspective. In: RolstonKV, RubensteinEB, editors. Textbook of febrile neutropenia. London: Martin Dunitz Ltd., 2001: 1–26..
- 16Imipenem-cilastatin versus amikacin plus piperacillin in the treatment of infections in neutropenic patients: a prospective, randomized multi-clinic study. Scand J Infect Dis. 1987; 52: 65–78., , , et al.
- 33Cefepime versus ceftazidime as empiric therapy of febrile neutropenic patients. In: KlasterskyJ, editor. Febrile neutropenia. Berlin: Springer-Verlag, 1997: 63–74., , , et al.