Anti‐pseudomonal beta‐lactams for the initial, empirical, treatment of febrile neutropenia: comparison of beta‐lactams



Several beta‐lactams are recommended as single agents for the treatment of febrile neutropenia.


To compare the effectiveness of different anti‐pseudomonal beta‐lactams as single agents in the treatment of febrile neutropenia. To compare the development of bacterial resistance, bacterial and fungal superinfections during or following treatment with the different beta‐lactams.

Search methods

We searched the Cochane Register of Controlled Trials (CENTRAL), Issue 3, 2010. MEDLINE, EMBASE, LILACS, FDA drug applications, conference proceedings and ongoing clinical trial databases up to August 2010. References of included studies were scanned.

Selection criteria

Randomised controlled trials (RCTs) comparing an antipseudomonal beta‐lactam to another antipseudomonal beta‐lactam antibiotic, both given alone or with the addition of the same glycopeptide to both study arms, for the initial treatment of fever and neutropenia among cancer patients.

Data collection and analysis

Two review authors applied inclusion criteria and extracted the data independently. Missing data were sought. Risk ratios (RR) were calculated with 95% confidence intervals (CI), and pooled using the fixed effect model. The primary outcome was all‐cause mortality. Risk of bias was assessed using a domain‐based evaluation and its effect of results was assessed through sensitivity analyses.

Main results

Forty‐four trials were included. The antibiotics assessed were cefepime, ceftazidime, piperacillin‐tazobactam, imipenem and meropenem. Adequate allocation concealment and generation were reported in about half of the trials and only two trials were double‐blinded. The risk for all‐cause mortality was significantly higher with cefepime compared to other beta‐lactams (RR 1.39, 95% CI 1.04 to 1.86, 21 trials, 3471 participants), without heterogeneity and with higher RRs in trials at low risk for bias. There were no differences in secondary outcomes but for a non‐significantly higher rate of bacterial superinfections with cefepime. Mortality was significantly lower with piperacillin‐tazobactam compared to other antibiotics (RR 0.56, 95% CI 0.34 to 0.92, 8 trials, 1314 participants), without heterogeneity. Carbapenems resulted in similar all‐cause mortality and a lower rate of clinical failure and antibiotic modifications as compared to other antibiotics, but a higher rate of diarrhea caused by Clostridium difficile.

Authors' conclusions

Current evidence supports the use of piperacillin‐tazobactam in locations where antibiotic resistance profiles do not mandate empirical use of carbapenems. Carbapenems result in a higher rate of antibiotic‐associated and Clostridium difficile‐associated diarrhea. There is a high level of evidence that all‐cause mortality is higher with cefepime compared to other beta‐lactams and it should not be used as monotherapy for patients with febrile neutropenia.


Mical Paul, Dafna Yahav, Assaf Bivas, Abigail Fraser, Leonard Leibovici


Plain language summary

Single‐agent antibiotic treatment for cancer patients with fever and low white blood cell counts

Cancer patients develop neutropenia, a decrease in the neutrophil subset of the white blood cells, as a result of chemotherapy. Neutropenia exposes patients to infections, mainly bacterial. Without antibiotic treatment these infections may be fatal, therefore antibiotic treatment is administered when a patient with neutropenia develops fever. The objective of this review was to compare antibiotic treatments currently recommended in consensus guidelines for the initial treatment of cancer patients with fever and neutropenia.

We identified 44 studies comparing different antibiotics. Cefepime resulted in significantly higher mortality compared to all other antibiotics combined, at the end of patients' hospital stay or 30 days after entry into the study. The risk was 39% higher with cefepime, ranging from 4 to 86% increased risk. We did not find an explanation for this when looking into other outcomes reported in the primary studies. Piperacillin‐tazobactam resulted in lower mortality than other antibiotics. The other antibiotics (ceftazidime, imipenem and meropenem) showed comparable efficacy, with a lower rate of antibiotic changes for imipenem or meropenem and a higher rate of severe diarrhea with these two antibiotics.

We conclude that piperacillin‐tazobactam might be the preferred antibiotic for the treatment of cancer patients with fever and neutropenia and that cefepime should not be used. Antibiotic selection (other than cefepime) depends on the individual patient and the type of bacteria prevalent in the specific hospital.


Mical Paul, Dafna Yahav, Assaf Bivas, Abigail Fraser, Leonard Leibovici

Reviewer's Conclusions

Authors' conclusions 

Implications for practice 

The decision of the beta‐lactam used for the treatment of febrile neutropenia is dependent on local epidemiology and susceptibility patterns in the ward or hospital (Rolston 2006). Our review provides evidence on the comparative efficacy and toxicity of the available beta‐lactams when administered in a given epidemiological setting. Based on this review, cefepime should not be used for patients with febrile neutropenia. Piperacillin‐tazobactam should be the preferred beta‐lactam for use as single agent, with or without a glycopeptide, in settings where resistance of Gram‐negative bacteria to piperacillin‐tazobactam is not prevalent (lower than 25%). Ceftazidime, imipenem and meropenem can be used, considering that the carbapenems are associated with a higher rate of antibiotic‐associated and Clostridium difficile‐associated diarrhea. The decision whether to used a lower‐spectrum beta‐lactam (with or without an antipseudomonal spectrum of coverage) combined with an aminoglycoside to increase the spectrum of coverage or to select a broader‐spectrum beta‐lactam alone is difficult. Evidence shows that mortality is lower with the broader‐spectrum beta‐lactam and nephrotoxicity is significantly higher with combination therapy (Paul 2003; Paul 2004).

Implications for research 

Future trials assessing antibiotic treatment for febrile neutropenia should report outcomes that are important to the individual patient and for decision making. All cause mortality should be reported by intention to treat in all trials and also as‐treated in non‐inferiority trials (Piaggio 2006). Clinical failure, as currently defined, is not an appropriate surrogate outcome since it does not portend different in all‐cause mortality. Patients should be included only once in the trial or results reported for patients' first randomization, since analyses relying on episodes are incorrect. The outcomes of bacterial and fungal superinfections should be defined using established and replicable definitions. While definitions for fungal infections have been specifically devised (De Pauw 2008), definitions for other infections, including catheter‐related infections, have not been defined specifically for neutropenic patients. We suggest that the CDC/NHSN definitions for healthcare associated infections be used (CDC/NHSN 2008). Length of hospital stay is a highly relevant outcome to the individual patient and should be reported.

All randomized trials conducted should be registered in accessible trial registries and their results should be made publicly available if not published, to avoid situations such the one described herein for cefepime. As for cefepime, full data on trial methods, trial locations, inclusion criteria, patient characteristics and outcomes for the unpublished trials included in the FDA's analysis should be reported. Given the alternative antibiotics currently available, we see no further need or justification for trials assessing its efficacy/ safety among neutropenic cancer patient.

Get full text at The Cochrane Library