Intervention Review

You have free access to this content

Beta-lactam versus beta-lactam-aminoglycoside combination therapy in cancer patients with neutropenia

  1. Mical Paul1,*,
  2. Yaakov Dickstein2,
  3. Agata Schlesinger3,
  4. Simona Grozinsky-Glasberg4,
  5. Karla Soares-Weiser5,
  6. Leonard Leibovici3

Editorial Group: Cochrane Gynaecological, Neuro-oncology and Orphan Cancer Group

Published Online: 29 JUN 2013

Assessed as up-to-date: 7 JUN 2013

DOI: 10.1002/14651858.CD003038.pub2


How to Cite

Paul M, Dickstein Y, Schlesinger A, Grozinsky-Glasberg S, Soares-Weiser K, Leibovici L. Beta-lactam versus beta-lactam-aminoglycoside combination therapy in cancer patients with neutropenia. Cochrane Database of Systematic Reviews 2013, Issue 6. Art. No.: CD003038. DOI: 10.1002/14651858.CD003038.pub2.

Author Information

  1. 1

    Rambam Health Care Campus, Division of Infectious Diseases, Haifa, Israel

  2. 2

    Rambam Health Care Center, Medicine A and Unit of Infectious Diseases, Haifa, Israel

  3. 3

    Beilinson Hospital, Rabin Medical Center, Department of Medicine E, Petah Tikva, Israel

  4. 4

    Dept of Medicine, Hadassah-Hebrew University Medical Center, Neuroendocrine Tumors Unit, Endocrinology & Metabolism Service, Jerusalem, Israel

  5. 5

    Cochrane, Cochrane Editorial Unit, London, UK

*Mical Paul, Division of Infectious Diseases, Rambam Health Care Campus, Ha-aliya 8 St, Haifa, 33705, Israel. paulm@post.tau.ac.il. m_paul@rambam.health.gov.il.

Publication History

  1. Publication Status: Edited (no change to conclusions)
  2. Published Online: 29 JUN 2013

SEARCH

 
Characteristics of included studies [ordered by study ID]
Agaoglu 2001

MethodsRandomisation: "systematic sampling".
Blinding: single.
Intention-to-treat: yes.
Interim analysis: none.
Exclusions from analysis: 0/82.


ParticipantsTurkey: 82 children.
<18 years with 87 episodes randomly assigned to 3 arms. Underlying haematological cancer in 74/87 episodes. Neutropenia < 1000/mm³ and fever > 38.5°.


InterventionsMeropenem 20 mg/kg × 3 versus Cefepime 33 mg/kg × 3 + netilmicin 2.5 mg/kg × 2 versus Ceftazidime 33 mg/kg × 3 + amikacin 7.5 mg/kg × 2.


OutcomesAll cause mortality; infection related mortality.
Treatment failure.
Adverse events.


NotesJournal publication.
No outcomes in subgroups.
Additional outcome in study: cost.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
Low riskAll episodes included in analysis.

Ahmed 2007

MethodsRandomisation: sealed opaque envelopes drawn at a central phone.
Location.
Blinding: no information.
Intention to treat: possible.
Interim analysis: none.
Exclusions from analysis: 10/129 episodes.
Follow-up period: no information.


ParticipantsEgygt: 129 episodes among children < 18 years with haematological cancer mainly (80%), neutropenia < 500/mm³ expected to last > 6 days and fever.


InterventionsImipenem 20-25 mg/kg × 4 versus Ceftriaxone 100 mg/kg × 1 + amikacin 15 mg/kg × 1.


OutcomesAll cause mortality.
Infection related mortality.
Treatment failure.
Adverse events.


NotesJournal publication and correspondence with author.
Outcome in subgroups: bacteraemia, Gram-negative bacteraemia,Pseudomonas aeruginosa bacteraemia, haematological cancer.
Additional outcome in study: cost.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskCentral and sealed opaque envelopes.

Allocation concealment (selection bias)Low riskA-Adequate.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk10 patients excluded from analysis.

Akova 1999

MethodsRandomisation: sealed envelopes (opaque not mentioned), containing numbers from a computer-generated list in balanced blocks.
Blinding: none.
Intention-to-treat: possible.
Interim analysis: none.
Exclusions from analysis: 12/83 (in study).
Follow-up period: mean 14 ± 9 days.


ParticipantsTurkey: Multicentre.
83 adults > 18 years with cancer, neutropenia < 500/mm³ and fever. Patients with life expectancy < 24 hours were excluded.


InterventionsMeropenem 1 gr × 3 versus Ceftazidime 2 gr × 3 + amikacin 1 gr × 1.


OutcomesAll cause mortality; infection related mortality.
Treatment failure.
Adverse events.
Dropouts after randomisation.


NotesJournal publication.
No outcomes in subgroups.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Adequate.

Allocation concealment (selection bias)Low riskA-Adequate.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk6 patients excluded from analysis.

Alanis 1983

MethodsRandomisation: random selection of sealed envelopes (opaque not mentioned).
Blinding: none.

Intention-to-treat: no.
Interim analysis: none.
Exclusions from analysis: 14/108.
Follow-up period: no information.


ParticipantsUSA: 108 febrile episodes in 86 cancer patients (9-74 years) with neutropenia < 1000/mm³ . Included were 3 participants with neutropenia unrelated to malignancy.


InterventionsMoxalactam 50- to 70 mg/kg × 2-3 (max 14 gr/d) versus Nafcillin 30 mg/kg × 6 (max 12 gr/d) + tobramycin 1.7 mg/kg × 3.


OutcomesTreatment failure.
Bacterial and fungal super-infections.
Colonisation.
Adverse events.
Dropouts after randomisation.


NotesJournal publication.
Outcome in subgroups.
Bacteraemia,
Severe neutropenia.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Adequate.

Allocation concealment (selection bias)Low riskA-Adequate.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk14 episodes excluded from analysis.

Antmen 2001

MethodsRandomisation: no information.
Blinding: no information.
Intention-to-treat: yes.
Interim analysis: none.
Exclusions from analysis: 0/67 episodes.
Follow-up period: no information.


ParticipantsTurkey: 67 febrile episodes in 52 children (11 months-15 years) with haematological cancer mainly and neutropenia < 1000/mm³ .


InterventionsMeropenem versus ceftazidime + amikacin (no data on doses).


OutcomesTreatment failure.


NotesConference proceeding: no outcomes in subgroups.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskB-Unclear.

Blinding of outcome assessment (detection bias)
All outcomes
Unclear riskB-Unclear.

Incomplete outcome data (attrition bias)
All outcomes
Low riskAll episodes included in analysis.

Au 1994

MethodsRandomisation: no information.
Blinding: no information.
Intention-to-treat: no.
Interim analysis: none.
Exclusions from analysis: 5/56.
Follow-up period: no information.


ParticipantsSingapore: 50 cancer patients > 14 years with neutropenia < 1000/mm³ and fever > 38°. Patients with life expectancy < 24 hours were excluded.


InterventionsImipenem/cilastatin 500 mg × 4 versus Ceftriaxone 2 gr × 1 + gentamicin 1 mg/kg × 3.


OutcomesTreatment failure.
Adverse events.


NotesJournal publication.
Outcome in subgroups.

Documented infections: bacteraemia.
Documented Gram-negative and Pseudomonas infections.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk6 participants excluded from analysis.

Behre 1998

MethodsRandomisation: pre-formed randomisation lists, provided by study centre.
Blinding: none.
Intention-to-treat: yes.
Interim analysis: none.
Exclusions from analysis: 7/78 (for efficacy analysis).
Follow-up period: 14 days following end of study medication.


ParticipantsGermany, multicentre: 78 episodes in 71 adults > 18 years with cancer (excluding allogeneic BMT), neutropenia < 500/mm³ and fever.


InterventionsMeropenem 1 gr × 3 versus Ceftazidime 2 gr × 3 + amikacin 5-7.5 mg/kg × 2-3.


OutcomesAll cause mortality; infection related mortality.
Treatment failure.
Adverse events.
Dropouts after randomisation.


NotesJournal publication and author correspondence.
Outcomes in subgroups.
Documented infections: bacteraemia.
Documented Gram-negative and Pseudomonas infections.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Adequate.

Allocation concealment (selection bias)Low riskA-Adequate.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk7 episodes excluded from analysis.

Bezwoda 1985

MethodsRandomisation: no information.
Blinding: no information.
Intention-to-treat: no.
Interim analysis: none.
Exclusions from analysis: 3/63.
Follow-up period: no information.


ParticipantsSouth Africa: 60 adult cancer patients with neutropenia < 1000/mm³ and fever > 39°.


InterventionsMoxalactam 2 gr × 3 versus cephradine 2 gr × 3 + tobramycin 1.5 mg/kg × 3.


OutcomesAll cause mortality.
Treatment failure.
Adverse events.


NotesJournal publication.
Trial terminated because of increasing resistance to cephradine.
Outcomes in subgroups.

Documented infections: documented Gram-negative and Pseudomonas infections.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk3 participants excluded from analysis.

Borbolla 2001

MethodsRandomisation: no information.
Blinding: no information.
Intention-to-treat: no information.

Interim analysis: none.
Exclusions from analysis: no information (the study does not refer to excluded patients, 40 participants included).
Follow-up period: no information.


ParticipantsMexico: 40 acute leukaemia patients included. Neutropenia < 500/mm³ and fever > 38° or focal infection.


InterventionsCefepime 2 gr × 3 versus ceftriaxone 17 mg/kg × 3 + amikacin 15 mg/kg × 1.


OutcomesTreatment failure.
Adverse events.


NotesJournal publication.
Outcomes in subgroups.
Haematological cancer patients.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
Low riskAll participants included in analysis.

Cometta 1996

MethodsRandomisation: central computer-generated randomisation using minimisation technique for 2 stratifications: leukaemia or BMT and centre.
Blinding: single (assessing committee).
Intention-to-treat: no (performed according to study on eligible evaluable patients).
Interim analysis: 2.
Exclusions from analysis: 76/1034.
Follow-up period: 30 days.


ParticipantsEurope, multicentre: 1034 cancer or BMT patients aged > 3 months with neutropenia < 1000/mm³, fever and a presumed infection.


InterventionsMeropenem 1 gr × 3 or 20 mg/kg × 3 versus ceftazidime 2 gr × 3 or 35 mg/kg × 3 + amikacin 20 mg/kg × 1.


OutcomesAll cause mortality.

Infection related mortality.
Treatment failure.
Adverse events.
Super-infections.
Dropouts after randomisation.


NotesJournal publication.
Envelopes used for randomisation in case of computer/connection failure.
Outcomes in subgroups.

Documented infections: documented Gram-negative, resistant Gram-negative and Pseudomonas infections.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Adequate.

Allocation concealment (selection bias)Low riskA-Adequate.

Blinding of participants and personnel (performance bias)
All outcomes
High riskNo blinding.

Blinding of outcome assessment (detection bias)
All outcomes
Low riskOutcome assessor blinded.

Incomplete outcome data (attrition bias)
All outcomes
High risk76 participants excluded from analyses.

Conte 1996

MethodsRandomisation: no information.
Blinding: none.
Intention-to-treat: yes.
Exclusions from analysis: 0/40 episodes.
Follow-up period: no information.


ParticipantsChile: 40 episodes in 25 participants with haematological cancer and high-risk febrile neutropenia (median count < 100/mm³).


InterventionsCefoperazone-sulbactam 3 gr × 2 versus ceftazidime 1 gr × 3 + amikacin 7.5 mg/kg × 2.


OutcomesAll cause mortality.
Treatment failure.
Adverse events.


NotesConference proceedings.

Outcomes in subgroups: haematological malignancies.
Vitamin K added to cefoperazone-sulbactam group.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
Low riskAll participants included in analysis.

Corapcioglu 2005

MethodsRandomisation: consecutive with no further details.
Blinding: none.
Intention-to-treat: no

Interim analysis: 0.
Exclusions from analysis: 10/60 episodes; follow-up period: not specified.


ParticipantsTurkey: 60 episodes among 29 children 11 months-17 years, mainly with haematological cancer (74%), with neutropenia < 500/mm³ or < 1000/mm³ and expected to decline to < 500/mm³ within 24-48 hours and fever.


InterventionsCefepime 50 mg/kg × 3 versus Ceftazidime 50 mg/kg × 3 + amikacin 15 mg/kg × 1.


OutcomesInfection related mortality.
Treatment failure.
Adverse events.
Dropouts after randomisation.


NotesJournal publication.
No outcomes in subgroups.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)High riskC-Inadequate.

Allocation concealment (selection bias)High riskC-Inadequate.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk10 episodes excluded from analysis.

Cornelissen 1992

MethodsRandomisation: no information.
Blinding: none.
Intention-to-treat: no.
Exclusions from analysis: 6/100.
Follow-up period: no information.


ParticipantsNetherlands: 100 episodes in 93 adult cancer patients with neutropenia < 500/mm³ and fever. Allogeneic BMT patients excluded.


InterventionsImipenem 500 mg × 4 versus Cefuroxime 1.5 gr × 3 or cephalotin 1 gr × 6 + gentamicin 80 mg × 3.


OutcomesTreatment failure.
Colonisation.
Adverse events.
Dropouts after randomisation.


NotesJournal publication.
Treatment modification suggested by protocol differs between the two treatment groups.
Outcomes in subgroups.

Documented infections: Gram-negative, resistant Gram-negative and Pseudomonas infections; bacteraemia.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk6 participants excluded from analysis.

Cornely 2001

MethodsRandomisation: block randomisation with a block length of six (3 arms for first randomisation × 2 arms for second randomisation). Consecutively numbered sealed opaque envelopes, non-used envelopes to be returned (envelopes glued to the binding of the CRF).
Blinding: none.
Intention-to-treat: yes for mortality.
Exclusions from analysis: 73/602 participants for failure patients.
Follow-up period: up to 42 days (mortality data given up to 30 days).


ParticipantsMC, Germany: 602 adult participants with high-risk haematological malignancies, neutropenia < 500 mm³ and fever > 38.5° once or > 38 twice within 24 hours. Randomly assigned to 3 arms.


InterventionsMeropenem 1 gr × 3 versus Cefepime 2 gr × 3 versus Piperacillin-tazobactam 4.5 gr × 3 + aminoglycoside once daily. For the purposes of the meta-analysis, the two monotherapy arms (cefepime and meropenem) were joined.


OutcomesAll cause mortality.

Infection related mortality.
Treatment failure.
Adverse events.


NotesConference proceedings: Full methods and results supplied by the author.

Outcomes in subgroups: haematological malignancies.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Computer-generated block randomisation.

Allocation concealment (selection bias)Low riskA-Consecutively numbered sealed opaque envelopes.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk73 participants excluded from analysis.

De la Camara 1997

MethodsRandomisation: consecutive computer-generated, concealed by sealed, opaque envelopes, stratified by centre.
Blinding: none.

Intention-to-treat: possible for mortality.
Interim analysis: none.
Exclusions from analysis: 29/122 episodes (for failure).
Follow-up period: 1 month following end of treatment.


ParticipantsSpain, multicentre: 122 episodes in 103 participants > 16 years with neutropenia < 500/mm³ and fever. All participants with underlying haematological malignancy, of which 49% had BMT and an additional 29% acute leukaemia.


InterventionsMeropenem 1 gr × 3 versus Ceftazidime 2 gr × 3 + Amikacin 5-7.5 mg/kg × 2-3.


OutcomesAll cause mortality.

Infection related mortality.
Treatment failure.
Adverse events.
Bacterial and fungal super infections.
Dropouts after randomisation.


NotesJournal publication and author correspondence.
Outcomes in subgroups.
Documented infections: bacteraemia; haematological cancer patients.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Adequate.

Allocation concealment (selection bias)Low riskA-Adequate.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk10 participants excluded from analysis.

De Pauw 1983

MethodsRandomisation: no information.
Blinding: none.
Intention-to-treat: no.
Interim analysis: no.
Exclusions from analysis: 4/78.
Follow-up period: no information.


ParticipantsNetherlands: 78 haematological or BMT cancer patients > 15 years with neutropenia < 1000/mm³ and fever.


InterventionsCeftazidime 2 gr × 3 versus Cefotaxime 2 gr × 4 + gentamicin 80 mg × 3.


OutcomesTreatment failure.
Fungal super infections.
Adverse events.


NotesJournal publication.
Surveillance cultures performed.
Outcomes in subgroups: Gram-negative and Pseudomonas infections; haematological cancer patients.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk4 participants excluded from analysis.

De Pauw 1994

MethodsRandomisation: central in blocks of eight, concealed by sealed envelopes (opaque not mentioned).
Blinding: none.
Intention-to-treat: no.
Interim analysis: no.
Exclusions from analysis: 74/1086.
Follow-up period: end of treatment.


ParticipantsAustralia, Canada, Europe, multicentre: 968 episodes in 872 participants > 14 years evaluated, with neutropenia < 500/mm³ and fever. Underlying haematological cancer in 83% of participants.


InterventionsCeftazidime 2 gr × 3 versus Piperacillin 3-4 gr × 4-6 + Tobramycin 1.7-2 mg/kg × 3.
Supplemented as indicated by Vancomycin 1 gr × 2 or Metronidazole 500 × 3-4.


OutcomesAll cause mortality.

Infection related mortality.
Treatment failure.
Super-infections.
Adverse events.
Dropouts after randomisation.


NotesJournal publication, conference proceedings and author correspondence.
Participants with suspected Gram-positive (n = 113) or anaerobic (n = 71) infections were given vancomycin or metronidazole, respectively, in addition to the randomly allocated antibiotic/s.
Outcomes in subgroups.
Documented infections: haematological cancer patients.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Adequate.

Allocation concealment (selection bias)Low riskA-Adequate.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk118 episodes excluded from analysis.

Del Favero 2001

MethodsRandomisation: computer-generated, central, stratified by center and underlying malignancy.
Blinding: double, placebo-controlled.
Intention-to-treat: possible for failure.
Interim analysis: no.
Exclusions from analysis: 27/760 (for efficacy analysis and mortality).
Follow-up period: 30 days.


ParticipantsItaly, multicentre: 760 participants > 13 years with neutropenia < 500/mm³ and fever > 38.5°. Underlying haematological cancer in 81%. of participants, and BMT in 52%.


InterventionsPiperacillin-tazobactam 4.5 gr × 3 versus piperacillin-tazobactam 4.5 gr × 3 + amikacin 7.5 mg/kg × 2.


OutcomesAll cause mortality.

Infection related mortality.
Treatment failure.
Dropouts after randomisation.
Adverse events.


NotesJournal publication.
Outcomes in subgroups.
Documented infections: bacteraemia.
Documented Gram-negative and Pseudomonas infections.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Adequate.

Allocation concealment (selection bias)Low riskA-Adequate.

Blinding of participants and personnel (performance bias)
All outcomes
Low riskDouble-blinded.

Blinding of outcome assessment (detection bias)
All outcomes
Low riskDouble-blinded.

Incomplete outcome data (attrition bias)
All outcomes
High risk27 participants excluded from analysis.

Dincol 1998

MethodsRandomisation: no information.
Blinding: none.
Intention-to-treat: yes.
Interim analysis: no.
Exclusions from analysis: none.
Follow-up period: end of treatment.


ParticipantsTurkey: 150 episodes in 97 cancer patients > 14 years, with neutropenia < 500/mm³ and fever > 38.5°. Underlying haematological cancer in 43%.


InterventionsImipenem 500 mg × 4 versus cefoperazone-sulbactam 2 gr × 2 + amikacin 15 mg/kg × 1.


OutcomesAll cause mortality.

Infection related mortality.
Treatment failure.
Adverse events.


NotesJournal publication and author correspondence.
Outcomes in subgroups.

Documented infections: bacteraemia.
Documented Gram-negative and Pseudomonas infections.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Adequate.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
Low riskAll episodes included in analysis.

Doyen 1983

MethodsRandomisation: no information.
Blinding: none.
Intention-to-treat: possible for mortality.
Interim analysis: 1.
Exclusions from analysis: 7/104 (for failure).
Follow-up period: no information.


ParticipantsBelgium: 104 episodes in 83 adult haematological cancer with neutropenia < 500/mm³ and fever > 38°.


InterventionsCeftazidime 30 mg/kg × 3 versus ceftazidime 30 mg/kg × 3 + amikacin 5 mg/kg × 3.


OutcomesAll cause mortality.

Infection related mortality.
Treatment failure.
Bacterial and fungal super infections.


NotesConference proceeding and author correspondence: Study not completed, all randomly assigned participants included in the review.

Outcomes in subgroups.

Documented infections: bacteraemia; haematological cancer patients.
Documented Gram-negative infections.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk7 episodes excluded from analysis.

Duzova 2001

MethodsRandomisation: no information.
Blinding: single-blinded.
Intention-to-treat: unknown.
Interim analysis: no.
Exclusions from analysis: no information, 90 episodes included.
Follow-up period: no information.


ParticipantsTurkey: 90 episodes in children < 16 years with lymphomas or solid tumours (leukaemia excluded), with neutropenia < 500/mm³ and fever > 38.3°.


InterventionsMeropenem 50 mg/kg × 3 versus Piperacillin 200 mg/kg × 4 + Amikaciin 15 mg/kg × 1.


OutcomesAll cause mortality.

Infection related mortality.
Treatment failure.


NotesJournal publication and author correspondence.

Outcomes in subgroups: haematological cancer patients.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskNo blinding.

Blinding of outcome assessment (detection bias)
All outcomes
Low riskOutcome assessor blinded.

Incomplete outcome data (attrition bias)
All outcomes
Low riskAll episodes included in analysis.

El Haddad 1995

MethodsRandomisation: no information.
Blinding: none.
Intention-to-treat: no.
Interim analysis: no.
Exclusions from analysis: unknown, only number of evaluated participants reported.
Follow-up period: unknown.


ParticipantsEgypt: 46 episodes in children < 15 years with leukaemia or lymphoma and neutropenia < 500/mm³ with fever > 38.5° once or > 38° thrice during 24 hours.


InterventionsRandomization 2:1 to cefoperazone-sulbactam 67 mg/kg × 3 versus Piperacillin 100 mg/kg × 4 + Amikacin 5 mg/kg × 3.


OutcomesInfection related mortality.

Treatment failure.
Adverse events.


NotesJournal publication.

Outcomes in subgroups: haematological cancer.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
Unclear riskB-Unclear.

Erjavec 1994

MethodsRandomisation: no information.
Blinding: none.
Intention-to-treat: yes.
Interim analysis: no.
Exclusions from efficacy and subgroup analysis: 36/179.
Follow-up period: resolution of neutropenia.


ParticipantsNetherlands: 179 episodes in 127 haematological cancer patients > 16 years, with neutropenia < 500/mm³ and fever.


InterventionsImipenem 12.5 mg/kg -1gr × 4 versus Cefuroxime 15 mg/kg × 3 + Tobramycin 2 mg/kg × 2 following a loading dose of 2.5 mg/kg.


OutcomesTreatment failure.
Bacterial and fungal super-infections.


NotesJournal publication.
Outcomes in subgroups.
Documented infections: documented Gram-negative, Pseudomonas and resistant Gram-negative infections; haematological cancer patients.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk36 episodes excluded from analysis.

Esteve 1997

MethodsRandomisation: no information.
Blinding: none.
Intention-to-treat: unknown.
Interim analysis: unknown.
Exclusions analysis: only the number of treated participants reported.
Follow-up period: unknown.


ParticipantsSpain: 85 episodes in 75 haematological cancer patients (excluding bone marrow transplantation) with neutropenia < 1000/mm³ and fever.


InterventionsPiperacillin-tazobactam 4 gr × 4 versus Piperacillin-tazobactam 4 gr × 4 + Amikacin 15 mg/kg × 1.


OutcomesAll cause mortality.

Treatment failure.
Adverse events.


NotesConference proceeding.

Outcomes in subgroups: haematological cancer patients.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
Unclear riskB-Unclear.

Gaytan-Martinez 2002

MethodsRandomisation: no information.
Blinding: none.
Intention-to-treat: no.
Interim analysis: unknown.
Exclusions from analysis: unknown (only number of evaluated episodes reported).
Follow-up period: unknown.


ParticipantsMexico: 117 evaluated episodes in participants with non-Hodgkin lymphoma or acute leukaemia with neutropenia and fever > 38.3°.


InterventionsCefepime 2 gr × 2 versus Ceftazidime 2 gr × 3 + Amikacin 1 gr × 1.


OutcomesTreatment failure.


NotesConference proceeding.

No outcomes in subgroups.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
Unclear riskB-Unclear.

Gibson 1989

MethodsRandomisation: computer-generated random numbers, concealed by sealed envelopes, which were taken in consecutive order (opaque not mentioned).
Blinding: none.
Intention-to-treat: yes.
Interim analysis: none.
Exclusions from analysis: 0.
Follow-up period: no information.


ParticipantsAustralia: 102 adults > 14 years. All with underlying haematological malignancy. Neutropenia < 1000/mm³ and fever or clinically localised site of infection.


InterventionsCeftazidime 2 gr × 3 versus Azlocillin 4 gr × 4 + Amikacin 5 mg/kg × 3.


OutcomesAll cause mortality.

Infection related mortality.
Treatment failure.


NotesJournal publication.
Additional empirical treatment with flucloxacillin allowed.
Outcomes in subgroups: haematological cancer patients.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Adequate.

Allocation concealment (selection bias)Low riskA-Adequate.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
Low riskAll participants included in analysis.

Gorschluter 2003

MethodsRandomisation: computer-generated, concealed by sealed, opaque envelopes.
Blinding: none.
Intention-to-treat: modified (performed on all eligible participants).
Interim analysis: 2.
Exclusions from analysis: 29/212 episodes.
Follow-up period: 21 days after treatment initiation.


ParticipantsGermany: 212 episodes in 130 adults > 18 years. All with underlying haematological malignancy, 90% acute leukaemia. Leukopaenia < 1000/mm³ or neutropenia < 500/mm³ and fever > 38.5° (rectal) or > 38° (axillary).


InterventionsPiperacillin-tazobactam 4.5 gr × 3 versus Ceftriaxone 2 gr × 1 + gentamycin 5 mg/kg × 1.


OutcomesAll cause mortality.
Infection related mortality.
Treatment failure.


NotesJournal article.
Study discontinued at second interim analysis by protocol because of a significant advantage for monotherapy.
Outcomes in subgroups.
Documented infections: bacteraemia.
Documented: Gram-negative, Pseudomonas and resistant Gram-negative infections; haematological cancer patients.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Computer generated lists.

Allocation concealment (selection bias)Low riskA-Sealed opaque envelopes.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk29 episodes excluded from analysis.

Gribble 1983

MethodsRandomisation: computer-generated, concealment not specified.
Blinding: none.
Intention-to-treat: no.
Interim analysis: none.
Exclusions from analysis: 4/54.

Follow-up period: 14 days after treatment cessation.


ParticipantsCanada: 50 episodes in 38 adults > 16 years evaluated. Of these, 30 episodes were in neutropenic participants with neutropenia < 1000/mm³ and fever > 38.3°.


InterventionsPiperacillin 75 mg/kg × 4 versus Carbenicillin 125 mg/kg × 4 (could be replaced by ticarcillin) + Gentamicin 1.5 mg/kg × 3 (could be replaced by tobramycin).


OutcomesTreatment failure: super-infections.
Adverse events.


NotesJournal publication.
Additional empirical treatment with cloxacillin allowed.
Study includes both neutropenic and non-neutropenic participants, and only outcomes that can be separated were extracted.
No outcomes in subgroups (for neutropenic participants only).


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Adequate.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk4 participants excluded from analysis.

Hansen 1986

MethodsRandomisation: no information, stratification according to cytotoxic therapy.
Blinding: none.
Intention-to-treat: no.
Interim analysis: none.
Exclusions from analysis: 7/40.
Follow-up period: no information.


ParticipantsDenmark: 80 episodes in 70 solid tumour cancer patients with neutropenia < 1500/mm³ and fever, randomised to 4 arms, of which 2 arms and 40 episodes are included in the review.


InterventionsLatamoxef 2 gr × 3 versus Carbenicillin 10 gr × 3 +gentamicin 80 mg × 3.


OutcomesTreatment failure: dropouts after randomisation.


NotesJournal publication.
Outcomes in subgroups: severe neutropenia, bacteraemia, documented Gram-negative infections.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk7 episodes excluded from analysis.

Hense 2000

MethodsRandomisation: no information.
Blinding: none.
Intention-to-treat: yes for mortality, possible for failure.
Interim analysis: none.
Exclusions from analysis: in study 3/87.
Follow-up period: no information.


ParticipantsGermany: 88 adult patients with haematological malignancy, with neutropenia < 500/mm³ and fever.


InterventionsMeropenem 1 gr × 3 (given either as bolus or infusion-2 arms merged for this review) versus Ceftazidime 2 gr × 3 + Amikacin 5 mg/kg × 3.


OutcomesAll cause mortality.
Infection related mortality.
Treatment failure: adverse events; dropouts after randomisation.


NotesConference proceeding and results from author.
Outcomes in subgroups: haematological malignancy.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk3 participants excluded from analysis.

Hess 1998

MethodsRandomisation: consecutively numbered sealed envelopes (opaque not mentioned) in randomly permuted blocks, by 24 hours' service.
Blinding: none.
Intention-to-treat: no.
Interim analysis: none.
Exclusions from analysis: 11/107.
Follow-up period: 30 days after inclusion.


ParticipantsSwitzerland: 107 episodes in 83 cancer patients > 13 years with neutropenia < 500/mm³ and fever or documented infection without fever.


InterventionsPiperacillin-tazobactam 4.5 gr × 3 versus Ceftazidime 2 gr × 3 + Amikacin 15 mg/kg × 1.


OutcomesAll cause mortality.

Infection related mortality.
Treatment failure: adverse events; dropouts after randomisation.


NotesJournal publication.
Additional empirical treatment with vancomycin allowed by protocol for non-responders after 48 hours.
Discrepancy between tables and text concerning dropouts, disabling analysis by intention-to-treat.
Outcomes in subgroups: bacteraemia.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Adequate.

Allocation concealment (selection bias)Low riskA-Adequate.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk11 episodes excluded from analysis.

Hung 2003

MethodsRandomisation: no information, stratified by haematological malignancy.
Blinding: none.
Intention-to-treat: no.
Interim analysis: none.
Exclusions from analysis: 24/100 episodes.
Follow-up period: no information.


ParticipantsChina: 76 episodes in 51 children < 14 years with neutropenia < 500/mm³, fever and suspected infection.


InterventionsMeropenem 40 mg/kg × 3 versus Ceftazidime 50 mg/kg × 3 + Amikacin 5 mg/kg × 3,


OutcomesAll cause mortality.

Infection related mortality.
Treatment failure: adverse events; dropouts after randomisation.


NotesJournal publication.
Outcomes in subgroups: severe neutropenia.

Documented infections.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk24 episodes excluded from analysis.

Jacobs 1993

MethodsRandomisation: no information, stratified by centre.
Blinding: none.
Intention-to-treat: no.
Interim analysis: none.
Exclusions from analysis: 16/107.
Follow-up period: no information.


ParticipantsUSA: multicentre, 107 episodes in 92 children < 18 years treated for cancer with fever > 38° and neutropenia > 500/mm³.


InterventionsCeftazidime 50 mg/kg × 3 versus Ceftazidime 50 mg/kg × 3 + Tobramycin 2.5 mg/kg × 3.


OutcomesTreatment failure: bacterial super-infections; adverse events; dropouts after randomisation.


NotesJournal publication.
Outcomes in subgroups.

Documented infections: bacteraemia.
Documented Gram-negative and Pseudomonas infections.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk16 episodes excluded from analysis.

Jimeno 2006

MethodsRandomisation: computer-generated, concealed by sealed opaque envelopes.

Blinding: none.
Intention-to-treat: yes.
Interim analysis: study stopped early because cancer treatment protocol changed.
Exclusions from analysis: 0/51 episodes.
Follow-up period: no information.


ParticipantsSpain: 51 episodes in 49 adults with solid malignancies treated with high-dose chemotherapy and peripheral blood stem cell support, with fever > 38.3° (or > 38° lasting > 1 hour) and neutropenia < 500/mm³ .


InterventionsCefepime 2 gr × 3 versus Ceftazidime 2 gr × 3 + Amikacin 500 mg × 2.


OutcomesAll cause mortality.
Infection related mortality.
Treatment failure: adverse events; hospitalisation duration.


NotesJournal publication and correspondence with author.
Outcomes in subgroups.

Documented infections.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Computer generated.

Allocation concealment (selection bias)Low riskA-Sealed opaque envelopes.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
Low riskAll episodes included in analysis.

Kiehl 2001

MethodsRandomisation: no information.
Blinding: none.
Intention-to-treat: yes.
Interim analysis: unknown.
Exclusions from analysis: 0/71 participants.
Follow-up period: no information.


ParticipantsGermany: 71 patients following autologous stem cell transplantation with febrile neutropenia.


InterventionsPiperacillin-tazobactam 4.5 gr × 3 versus Piperacillin-tazobactam 4.5 gr × 3 + Netilmicin 5 mg/kg × 1.


OutcomesTreatment failure.


NotesConference proceeding.

Outcomes in subgroups: haematological cancer patients.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
Low riskAll participants included in analysis.

Kinsey 1990

MethodsRandomisation: no information.
Blinding: none.
Intention-to-treat: no
Interim analysis: none.
Exclusions from analysis: 45/205.
Follow-up period: no information.


ParticipantsUK: 205 episodes in 139 haematological cancer patients, aged 9-74 years with neutropenia < 500/mm³ and fever.


InterventionsCeftazidime versus ceftazidime + gentamicin (no dosing information).


OutcomesAll cause mortality.

Infection related mortality.
Treatment failure.


NotesJournal publication.
No outcomes in subgroups.
11 deaths in 45 excluded participants.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk45 episodes excluded from analysis.

Kliasova 2001

MethodsRandomisation: no information.
Blinding: none.
Intention-to-treat: unknown.
Interim analysis: unknown.
Exclusions from analysis: only number evaluated provided.
Follow-up period: 60 days after treatment.


ParticipantsRussia: 43 episodes in 42 haematological cancer patients > 15 years after bone marrow transplantation, with fever and neutropenia.


InterventionsMeropenem 1 gr × 3 versus Ceftazidime 2 gr × 3 + Amikacin 500 mg × 3.


OutcomesAll cause mortality.
Infection related mortality.
Treatment failure: adverse events.


NotesConference proceeding.

Outcomes in subgroups: haematological cancer patients.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
Unclear riskB-Unclear.

Koehler 1990

MethodsRandomisation: no information.
Blinding: none.
Intention-to-treat: yes.
Interim analysis: none.
Exclusions from analysis: 0/120 episodes.
Follow-up period: no information.


ParticipantsPoland: 76 children with haematological cancer with 120 episodes of neutropenia <= 1000/mm³ and fever >= 38° for > 3 hours.


InterventionsCeftazidime 50 mg/kg × 3 versus Ampicilin or Amoxycillin 100 mg/kg × 3 + Tobramycin 4 mg/kg × 3.


OutcomesTreatment failure: bacterial and fungal super-infections; adverse events.


NotesJournal publication.
Outcomes in subgroups: haematological cancer.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
Low riskAll episodes included in analysis.

Kojima 1994

MethodsRandomisation: stratified by use of G-CSF, no further information.
Blinding: none.
Intention-to-treat: no.
Interim analysis: none.
Exclusions from analysis: 2/70.
Follow-up period: 72 hours after completion of treatment.


ParticipantsJapan: 70 episodes in 60 solid tumour cancer adult patients with neutropenia < 1000/mm³ and fever. No BMT patients.


InterventionsImipenem 500 mg × 4 versus Imipenem 500 mg × 4 + Amikacin 200 mg/m2 × 2.


OutcomesTreatment failure: adverse events; fungal colonisation; dropouts after randomisation.


NotesJournal publication and author correspondence. Study terminated prematurely because of excess failures in monotherapy.
Outcomes in subgroups:
Documented infections: documented Gram-negative and Pseudomonas infections.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Low riskA-Adequate.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk3 participants excluded from analysis.

Leyland 1992

MethodsRandomisation: generation not specified, concealment by sealed envelopes (opaque not mentioned).
Blinding: single (outcome assessor).
Intention-to-treat: no.
Interim analysis: none.
Exclusions from analysis: 96/312.
Follow-up period: 14 days.


ParticipantsUK , multicentre: 312 episodes in 234 adults > 18 years with haematological cancer, neutropenia < 1000/mm³ and fever.


InterventionsImipenem 0.5-1 gr × 4 versus Piperacillin 4 gr × 4 + Gentamycin 80 mg/kg × 3.


OutcomesOverall mortality.
Treatment failure: bacterial super-infections; adverse events; dropouts after randomisation.


NotesJournal publication.
Exclusion rate 30.8%, with 3 patients not accounted for.
Outcomes in subgroups.

Documented infections: bacteraemia.
Documented Gram-negative infections: haematological cancer patients.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Low riskA-Adequate.

Blinding of participants and personnel (performance bias)
All outcomes
High riskNo blinding.

Blinding of outcome assessment (detection bias)
All outcomes
Low riskOutcome assessor blinded.

Incomplete outcome data (attrition bias)
All outcomes
High risk96 episodes excluded from analysis.

Lieschke 1990

MethodsRandomisation: generated through centre at drug company, concealed by envelopes (sealed or opaque not mentioned).
Blinding: none.
Intention-to-treat: yes.
Interim analysis: 1.
Exclusions from analysis: none.
Follow-up period: end of fever, infection or neutropenia.


ParticipantsAustralia: 182 episodes in 150 adult febrile neutropenic cancer patients. Neutropenia < 1000/mm³ and fever > 38°.


InterventionsImipenem 500 mg × 4 versus Piperacillin 4 gr × 4 + Tobramycin 1 mg/kg × 3.


OutcomesAll cause mortality.

Infection related mortality.
Treatment failure: bacterial super-infections; adverse events.


NotesData from manuscript of unpublished trial supplied by author. Published as an abstract at an interim analysis.
No outcomes in subgroups.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Low riskA-Adequate.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
Low riskAll episodes included in analysis.

Liu 1989

MethodsRandomisation: no information.
Blinding: none.

Intention-to-treat: no.
Interim analysis: none.
Exclusions from analysis: 1/28.
Follow-up period: no information.


ParticipantsChina: 28 adults > 18 years with cancer, neutropenia < 500/mm³ and fever. Randomised to 3 treatment arms.


InterventionsImipenem 500 mg × 4 versus Ceftriaxone 2 gr × 1 + Amikacin 7.5 mg/kg × 2 versus Ceftazidime 2 gr × 3 + Amikacin 7.5 mg/kg × 2.


OutcomesTreatment failure: bacterial and fungal super-infections; adverse events; dropouts after randomisation.


NotesJournal publication.
Outcomes in subgroups: documented infections.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk1 patient excluded from analysis.

Marie 1991

MethodsRandomisation: "triage a sort" concealed by sealed envelopes (opaque not mentioned).

Blinding: none.

Intention-to-treat: unknown.
Interim analysis: yes.
Exclusions from analysis: no information (the study does not refer to excluded participants).
Follow-up period: no information.


ParticipantsFrance, bi-centre: 146 episodes in adult cancer patients with cancer, neutropenia < 500/mm³ and fever. Randomised to 3 treatment arms, of which 2 are relevant for the comparison in the review.


InterventionsCeftazidime 1 gr × 3 versus Ceftazidime 1 gr × 3 + Amikacin 7.5 mg/kg × 2 versus Ceftazidime 1 gr × 3 + Vancomycin 500 mg × 3 (third treatment arm excluded).


OutcomesTreatment failure: bacterial and fungal super-infections; adverse events.


NotesJournal publication.
No outcomes in subgroups.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Adequate.

Allocation concealment (selection bias)Low riskA-Adequate.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
Unclear riskB-Unclear.

Matsui 1991

MethodsRandomisation: computer generated, concealed with opaque envelopes.
Blinding: single.

Intention-to-treat: possible.
Interim analysis: none.
Exclusions from analysis: 3/101 (in study).
Follow-up period: end of treatment.


ParticipantsJapan: 101 episodes in 98 adults with chemotherapy treated lung cancer, leukopenia < 3000/mm³ and fever (80% with neutropenia < 1000/mm³ ). All participants with clinically or microbiologically documented infection.


InterventionsImipenem 1 gr × 2 versus Moxalactam 2 gr × 2 + Tobramycin 90 mg × 2.


OutcomesAll cause mortality.

Infection related mortality.
Treatment failure.


NotesJournal publication and author correspondence.
Outcomes in subgroups.
Documented infections: documented Gram-negative andPseudomonas infections; severe neutropenia.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Adequate.

Allocation concealment (selection bias)Low riskA-Adequate.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOnly patient blinded.

Blinding of outcome assessment (detection bias)
All outcomes
High riskNo blinding.

Incomplete outcome data (attrition bias)
All outcomes
High risk3 episodes excluded from analysis.

Miller 1993

MethodsRandomisation: no information.
Blinding: none.
Intention-to-treat: no, modified ITT possible for failure.
Interim analysis: none.
Exclusions from analysis: 45/131 episodes.
Follow-up period: end of treatment.


ParticipantsUSA: 131 episodes in 106 adult patients with haematological or solid cancer from three hospitals, with neutropenia < 500/mm³ and fever > 38° or a clinically or microbiologically documented source of infection.


InterventionsImipenem 500 mg × 4 versus Ceftazidime 2 gr × 3 + Tobramycin 1-1.5 mg/kg × 3-4.


OutcomesAll cause mortality.

Infection related mortality.
Treatment failure: bacterial super-infections; adverse events.


NotesJournal publication.
Outcomes in subgroups.

Documented infections: bacteraemia.
Documented Gram-negative infections; severe neutropenia.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk45 episodes excluded from analysis.

Morgan 1983

MethodsRandomisation: no information.

Blinding: none.
Intention-to-treat: no.
Interim analysis: none.
Exclusions from analysis: 3/50.
Follow-up period: no information.


ParticipantsEngland: 50 episodes in 34 children < 15 years with malignancy. Neutropenia < 1000/mm³ and fever.


InterventionsCeftazidime 30 mg/kg × 3 versus Azlocillin 50 mg/kg × 3 + Tobramycin 2 mg/kg × 3.


OutcomesAll cause mortality.

Infection related mortality.
Treatment failure: dropouts after randomisation.


NotesJournal publication.
Outcomes in subgroups:Pseudomonas infections.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk3 episodes excluded from analysis.

Norrby 1987

MethodsRandomisation: concealed by sealed envelopes (opaque not mentioned), stratified by centre. No information on allocation generation.
Blinding: none.
Intention-to-treat: yes.
Interim analysis: none.
Exclusions from analysis: none.
Follow-up period: 7 days after treatment.


ParticipantsEurope+Canada, multicentre: 210 participants > 16 years with neutropenia < 1000/mm³ and fever. Participants with high probability of death within 48 hours excluded.


InterventionsImipenem 1 gr (or 125 mg/kg) × 4 versus Piperacillin 4 gr × 4 (or 75 mg/kg × 4-6) + Amikacin 5-7.5 mg/kg × 2-3.


OutcomesAll cause mortality.

Infection related mortality.
Treatment failure: bacterial and fungal super-infections; colonisation and resistant colonisation; adverse events.


NotesJournal publication and author correspondence.

Outcomes in subgroups.
Documented infections: bacteraemia
Documented Gram-negative and Pseudomonas infections; severe neutropenia.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Low riskA-Adequate.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
Low risk78 participants excluded from efficacy analysis in study, but ITT analysis also given.

Novakova 1990

MethodsRandomisation: computer-generated allocation with sealed envelope concealment (opaque not mentioned).
Blinding: none.
Intention-to-treat: possible.
Interim analysis: none.
Exclusions from analysis: 21/90 (in study).
Follow-up period: no information.


ParticipantsNetherlands: 90 episodes in 83 adult patients > 15 years with neutropenia < 500/mm³ and fever. All with underlying haematological malignancy.


InterventionsCeftazidime 2 gr × 3 versus Piperacillin 4 gr × 4 + Amikacin 500 mg × 3.


OutcomesAll cause mortality.

Infectionvrelated mortality.
Treatment failure: bacterial and fungal super-infections; dropouts after randomisation.


NotesJournal publication.
Outcomes in subgroups: bacteraemia; haematological cancer patients.
Participants nursed in reverse isolation.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Adequate.

Allocation concealment (selection bias)Low riskA-Adequate.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk21 episodes excluded from analysis.

Novakova 1991

MethodsRandomisation: computer-generated, concealed by sealed envelopes (opaque not mentioned).
Blinding: none.
Intention-to-treat: possible.
Interim analysis: none.
Exclusions from analysis: 21/90 (in study).
Follow-up period: no information.


ParticipantsNetherlands: 90 episodes in 82 adults > 15 years with neutropenia < 1000/mm³, fever and signs of a local infection. All with underlying haematological malignancy or BMT.


InterventionsCeftazidime 2 gr × 3 versus Ceftazidime 2 gr × 3 + Amikacin 500 mg × 3.


OutcomesAll cause mortality.

Infection×related mortality.
Treatment failure: bacterial and fungal super-infections; dropouts after randomisation.


NotesJournal publication.
Outcomes in subgroups.

Documented infections: bacteraemia; haematological cancer patients.
Participants nursed in reverse isolation.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Adequate.

Allocation concealment (selection bias)Low riskA-Adequate.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk21 episodes excluded from analysis.

Ozyilkan 1999

MethodsRandomisation: central randomisation with random file number.
Blinding: double.
Intention-to-treat: yes.
Interim analysis: none.
Exclusions from analysis: none.
Follow-up period: 12 months.


ParticipantsTurkey: 30 adult cancer patients with neutropenia < 1000/mm³ and fever. 93% with underlying haematological malignancy.


InterventionsImipenem 500 mg × 4 versus Cefoperazone-sulbactam 2 gr × 2 + Amikacin 7.5 mg/kg × 2.


OutcomesAll cause mortality.

Infection related mortality.
Treatment failure: bacterial and fungal super-infections; dropouts after randomisation.


NotesJournal publication and author correspondence.
Outcomes in subgroups.

Documented infections.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Adequate.

Allocation concealment (selection bias)Low riskA-Adequate.

Blinding of participants and personnel (performance bias)
All outcomes
Low riskDouble-blinded.

Blinding of outcome assessment (detection bias)
All outcomes
Low riskDouble-blinded.

Incomplete outcome data (attrition bias)
All outcomes
Low riskAll participants included in analysis.

Papachristodoulou 96

MethodsRandomisation: no information.
Blinding: none.
Intention-to-treat: yes.
Interim analysis: unknown (results obtain from conference proceeding).
Exclusions from analysis: 0/85 episodes.
Follow-up period: no information.


ParticipantsGreece: 85 episodes in 77 cancer patients with neutropenia <= 1000/mm³ and fever >= 38°.


InterventionsCeftazidime 6 gr/day versus Ceftazidime 6 gr/day + Amikacin 1 gr/day.


OutcomesAll cause mortality.
Treatment failure.


NotesConference proceeding.
No outcomes in subgroups.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
Low riskAll episodes included in analysis.

Pegram 1984

MethodsRandomisation: no information.
Blinding: none.
Intention-to-treat: unknown.
Interim analysis: yes.
Exclusions from analysis: no information.
Follow-up period: no information.


ParticipantsUSA: 140 episodes in cancer patients. with neutropenia < 1000/mm³ and fever.


InterventionsMoxalactam 4 gr × 3 versus Ticarcillin 50 mg/kg × 6 + Tobramycin 1.5 mg/kg × 3.


OutcomesTreatment failure.


NotesConference proceeding.
Outcomes in subgroups.
DocumentedPseudomonas infections.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
Unclear riskB-Unclear.

Pellegrin 1988

MethodsRandomisation: allocation by table of 20 numbers; no reference to concealment.
Blinding: none.
Intention-to-treat: no information.
Interim analysis: none.
Exclusions from analysis: no information (157 participants evaluated, the study does not refer to the number of randomly assigned participants).
Follow-up period: no information.


ParticipantsFrance: 157 patients with acute leukaemia newly diagnosed or in first remission, with neutropenia < 500/mm³ for 21 or more days, and fever. All participants > 16 years. BMT patients excluded.


InterventionsCeftazidime 2 gr × 2 versus Cefotaxime 2 gr × 2 + Tobramycin 1 mg/kg × 2.


OutcomesTreatment failure: bacterial and fungal super-infections.


NotesJournal publication: French language.
Outcomes in subgroups

Documented infections: bacteraemia; haematological cancer patients.

Documented Gram-negative infections.
Participants treated in a protected environment.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Adequate.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
Unclear riskB-Unclear.

Pereira 2009

MethodsRandomisation: allocation by number list, no reference to concealment.

Blinding: none.

Intention-to-treat: no.

Interim analysis: no.

Exclusion from analysis: 5/130 episodes.

Follow-up period: not described.


ParticipantsBrazil: 130 episodes in 57 patients with haematological malignancy all ≤ 18 years.

Neutropenia < 500 cells/mm3 or between 500 and 1000 cells/mm3 before the nadir of chemotherapy and fever with axillary temperature > 38.0° C or 3 measurements between 37.5° C and 38.0° C.


InterventionsCefepime 50 mg/kg × 3 versus Ceftriaxone 50 mg/kg × 2 + Amikacin 15 mg/kg × 1.


OutcomesTreatment failure: infection related mortality; bacterial super-infection.


NotesJournal publication.

Outcome in subgroup: first episode of neutropenic fever.

Added AMP-B after 5 days with continued neutropenic fever, vanco for CR-BSI, skin and pulmonary infections and hypotension.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskAllocation by number list.

Allocation concealment (selection bias)Unclear riskNot described.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk5 episodes excluded from analysis.

Perez 1995

MethodsRandomisation: allocation by balanced table, no reference to concealment.
Blinding: none.
Intention-to-treat: yes.
Interim analysis: none.
Exclusions from analysis: 0/60.
Follow-up period: no information.


ParticipantsChile: 60 episodes in 52 cancer patients > 16 years, of whom 88% had underlying haematological malignancy. Neutropenia < 500/mm³ and fever.


InterventionsImipenem 500 mg × 4 versus Ceftazidime 1-1.5 gr × 4 + Amikacin 7.5 mg/kg × 2.


OutcomesAll cause mortality.

Infection related mortality.
Treatment failure: bacterial super-infections; dropouts after randomisation.


NotesJournal publication,
Spanish language.
Outcomes in subgroups.

Documented infections: bacteraemia.

Documented Gram-negative infections.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Adequate.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
Low riskAll episodes included in analysis.

Petrilli 2003

MethodsRandomisation: allocation by table of random numbers concealed by sealed opaque envelopes.
Blinding: none.
Intention-to-treat: no.
Interim analysis: none.
Exclusions from analysis: 2/138 episodes.
Follow-up period: no information.


ParticipantsBrazil: 138 episodes in 70 children or adolescents with leukaemia or grade III-IV lymphoma, neutropenia < 500/mm³ (or < 1000/mm³ expected to decline) and fever.


InterventionsTicarcillin-clavulanic acid 62.5 mg/kg × 4 versus Ceftriaxone 100 mg/kg × 1 + Amikacin 7.5 mg/kg × 2,


OutcomesAll cause mortality.

Infection related mortality;
Treatment failure: bacterial super-infections; adverse events; dropouts after randomisation.


NotesJournal publication.
Outcomes in subgroups.

Documented Gram-negative infections: haematological cancer patients; bacteraemia.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskA-Random number table.

Allocation concealment (selection bias)Low riskA-Sealed opaque envelopes.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk2 episodes excluded from analysis.

Piccart 1984

MethodsRandomisation: no information.
Blinding: none.
Intention-to-treat: no.
Interim analysis: none.
Exclusions from analysis: 5/49.
Follow-up period: no information.


ParticipantsBelgium: 154 cancer patients > 17 years randomly assigned, of which 49 patients with neutropenia < 1000/mm³, and with fever > 38.5° are included in the review.


InterventionsCefoperazone 6 gr × 2 versus Cefoperazone 2 gr × 2 + Amikacin 500 mg × 2.


OutcomesAll cause mortality (in bacteraemia only).

Infection related mortality.
Treatment failure: super-infections; dropouts after randomisation.


NotesJournal publication.
Outcomes in subgroups.

Documented infections: Gram-negative and resistant Gram-negative infections; haematological cancer patients; severe neutropenia.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk5 participants excluded from analysis.

Pickard 1983

MethodsRandomisation: random number table supplied by sponsor, concealed by sealed envelopes (opaque not mentioned).
Blinding: none.
Intention-to-treat: possible for mortality.
Interim analysis: 2.
Exclusions from analysis: 5/80.
Follow-up period: no information.


ParticipantsUSA: 80 episodes in cancer patients > 18 years with neutropenia < 1000/mm³ and fever.


InterventionsMoxalactam 2-4 gr × 3 versus Ticarcillin 3 gr × 4 + Tobramycin 1.66 mg/kg × 3.


OutcomesAll cause mortality.
Treatment failure: bacterial super-infections; adverse events; dropouts after randomisation.


NotesAuthor correspondence and conference proceedings.
Outcomes in subgroups.

Documented infections: documented Gram-negative and Pseudomonas infections.
Study not by Intention-to-treat but permits re-analysis by Intention-to-treat.
Participants nursed in reverse isolation.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Adequate.

Allocation concealment (selection bias)Low riskA-Adequate.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk5 episodes excluded from analysis.

Piguet 1988

MethodsRandomisation: envelope selection (sealed or opaque not mentioned).
Blinding: none.
Intention-to-treat: no.
Interim analysis: none.
Exclusions from analysis: 5/174.
Follow-up period: until resolution of neutropenia.


ParticipantsFrance, multicentre: 169 episodes evaluated in participants > 16 years with underlying haematological malignancy, neutropenia < 1000/mm³ and fever. BMT patients excluded.


InterventionsCeftazidime 2 gr × 3 versus Cefotaxime 2 gr × 3 + Amikacin 5 mg/kg × 3.


OutcomesAll cause mortality.

Infection related mortality.
Treatment failure.


NotesJournal publication: French language.
Outcomes in subgroups.

Documented infections: haematological cancer patients.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk5 episodes excluded from analysis.

Rodjer 1987

MethodsRandomisation: by envelope (sealed or opaque not mentioned).
Blinding: none.
Intention-to-treat: no.
Interim analysis: none.
Exclusions from analysis: 16/61 for failure; 3/61 for death.
Follow-up period: no information.


ParticipantsSweden: 61 febrile episodes in 52 immunocompromised cancer patients > 16 years. Neutropenia not part of inclusion criteria, but 70% of included patients were neutropenic < 1000/mm³.


InterventionsCeftazidime 1-2 gr × 2-3 versus Cefuroxime 1.5 gr × 2-3 + Tobramycin 1.5 mg/kg × 2-3.


OutcomesAll cause mortality.
Treatment failure: bacterial and fungal super-infections.


NotesJournal publication.
Outcomes in subgroups.

Documented infections: bacteraemia.

Documented Gram-negative and Pseudomonas infections.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk3 episodes excluded from analysis.

Rodriguez 1995

MethodsRandomisation: table of random numbers, concealment not specified.
Blinding: none.
Intention-to-treat: no.
Interim analysis: none.
Exclusions from analysis: 14/150 participants.
Follow-up period: no information.


ParticipantsPeru: 136 participants > 15 years evaluated, with solid cancer or non-Hodgkin lymphoma, neutropenia < 1000/mm³ expected to last less than 10 days and fever or suspected infection using specific criteria.


InterventionsCefotaxime 1 gr × 3 versus Cephalotin 1gr × 4 + Gentamicin 4 mg/kg × 1.


OutcomesInfection related mortality.
Treatment failure.


NotesJournal publication.
Outcomes in subgroups.

Documented infections: bacteraemia; Pseudomonas infections; severe neutropenia.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Table of random numbers.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk14 participants excluded from analysis.

Rolston 1992

MethodsRandomisation: computer-generated sequence of numbers.
Blinding: single (outcome assessor).
Intention-to-treat: no.
Interim analysis: none.
Exclusions from analysis: 158/908.
Follow-up period: median 8-9 days.


ParticipantsUSA: 750 episodes in 567 participants > 16 years in 4 arms. 67% underlying haematological malignancy. Neutropenia < 1000/mm³ and fever or documented infection.


InterventionsCeftazidime 1 gr × 6 versus Ceftazidime 1 gr × 6 + Amikacin and Imipenem 12.5 mg/kg × 4 versus Imipenem 12.5 mg/kg × 4 + Amikacin.
Amikacin given continuously 800 mg/m2 per day after 200 mg/m2 loading dose.


OutcomesTreatment failure: bacterial and fungal super-infections; adverse events.


NotesJournal publication.
Outcomes in subgroups.

Documented infections.

Documented Gram-negative, resistant Gram-negative and Pseudomonas infections; severe neutropenia.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Adequate.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskNo blinding.

Blinding of outcome assessment (detection bias)
All outcomes
Low riskOutcome assessor blinded.

Incomplete outcome data (attrition bias)
All outcomes
High risk158 episodes excluded from analysis.

Schuchter 1988

MethodsRandomisation: no information.
Blinding: double-blind.
Intention-to-treat: unknown (assumed yes).
Interim analysis: unknown.
Exclusions from analysis: no information.
Follow-up period: no information.


ParticipantsUSA: 133 patients following bone marrow transplantation between 2 and 57 years (median, 27 years) with neutropenia < 500/mm³ and fever > 38.5° once or > 38° more than once during 24 hours..


InterventionsCeftazidime 50 mg/kg or 2 gr × 3 versus Ticarcillin 45 mg/kg × 6 + Gentamicin 2 mg/kg × 4.


OutcomesTreatment failure: bacterial and fungal super-infections.


NotesConference proceeding.
Outcomes in subgroups.
Documented infections: haematological cancer patients.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
Low riskDouble-blinded.

Blinding of outcome assessment (detection bias)
All outcomes
Low riskDouble-blinded.

Incomplete outcome data (attrition bias)
All outcomes
Unclear riskB-Unclear.

Smith 1990

MethodsRandomisation: no information.
Blinding: none.
Intention-to-treat: unknown.
Interim analysis: none.
Exclusions from analysis: no information.
Follow-up period: 7 days.


ParticipantsUK: 100 episodes in 63 children < 16 years old.
Neutropenia < 500/mm³ and fever > 39°.


InterventionsCeftriaxone 50 mg/kg × 1 versus Azlocillin 75 mg/kg × 3 + Netilmicin 2.5 mg/kg × 3.


OutcomesAll cause mortality.

Infection related mortality.
Treatment failure.


NotesJournal publication.
No outcomes in subgroups.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
Low riskAll episodes included in analysis.

Tamura 2002

MethodsRandomisation: table of random numbers concealed with sealed opaque envelopes.

Blinding: none.
Intention-to-treat: yes for mortality and possible for failure
Interim analysis: yes, number not specified, trial stopped when the interim analysis demonstrated that the number of participants was sufficient for analysis.

Exclusions from analysis: 12/206 participants for failure.
Follow-up period: 30 days.


ParticipantsJapan, multicentre: 206 adult cancer patients with neutropenia < 1000/mm³ and fever >= 37.5°. Nearly all patients with haematological cancer.


InterventionsCefepime 1-2 gr × 2 versus Cefepime 1-2 gr × 2 + Amikacin (28 participants) or Isepamicin (36 participants) or Tobramycin or Netilmicin (12 participants).


OutcomesAll cause mortality.
Treatment failure: adverse events; dropouts after randomisation.


NotesJournal publication and author correspondence.
Outcomes in subgroups.

Documented infections.
An additional arm of carbapenem monotherapy is not included in this review.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Adequate.

Allocation concealment (selection bias)Low riskA-Adequate.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk12 participants excluded from analysis.

Tamura 2004

MethodsRandomisation: computer-generated, central.

Blinding: none.
Intention-to-treat: no.
Interim analysis: no information.
Exclusions from analysis: 12/201 participants.
Follow-up period: 30 days.


ParticipantsJapan, multicentre: 201 haematological cancer patients with neutropenia < 1000/mm³ and fever >= 37.5°.


Interventionscefepime 1-2 gr × 2 versus Cefepime 1-2 gr × 2 + Amikacin 100-200 mg × 2.


OutcomesAll cause mortality.

Infection related mortality.
Treatment failure: adverse events; dropouts after randomisation.


NotesJournal publication.
Outcomes in subgroups: haematological malignancy; bacteraemia.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Computer generated.

Allocation concealment (selection bias)Low riskA-Central.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk12 participants excluded from analysis.

Wade 1987

MethodsRandomisation: no information.
Blinding: double.
Intention-to-treat: unknown.
Interim analysis: none.
Exclusions from analysis: no information.
Follow-up period: no information.


ParticipantsUSA: 460 evaluable episodes in cancer patients with neutropenia < 500/mm³ and fever > 38°.


InterventionsImipenem 4 gr/qd versus Piperacillin 300 mg/kg/qd + Amikacin 24 mg/kg/qd.


OutcomesTreatment failure.


NotesConference proceeding + review.
Outcomes in subgroups.

Documented infections.

Documented Gram-negative and Pseudomonas infections.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskB-Unclear.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
Low riskDouble-blinded.

Blinding of outcome assessment (detection bias)
All outcomes
Low riskDouble-blinded.

Incomplete outcome data (attrition bias)
All outcomes
Unclear riskB-Unclear.

Wrzesien-Kus 2001

MethodsRandomisation: coin toss performed after participants' recruitment into the trial.
Blinding: none.
Intention-to-treat: yes.
Interim analysis: none.
Exclusions from analysis: 0/40 participants.
Follow-up period: no information.


ParticipantsPoland: 40 adults with cancer (haematological in 95%), neutropenia <= 1000/mm³ and fever >= 38°.


InterventionsCefepime 2 gr × 3 versus Cefepime 2 gr × 3 + Amikacin 500 mg × 2.


OutcomesAll cause mortality.
Infection related mortality.

Treatment failure.


NotesJournal publication in Polish.
No outcomes in subgroups.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Coin toss.

Allocation concealment (selection bias)Low riskA-Performed after patient recruitment.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
Low riskAll participants included in analysis.

Yamamura 1997

MethodsRandomisation: computer-generated code, concealed by sealed, opaque envelopes.

Blinding: none.
Intention-to-treat: possible.
Interim analysis: none.
Exclusions from analysis: 11/111 (in study).
Follow-up period: no information.


ParticipantsUSA, multicentre: 111 cancer patients > 18 years.
Neutropenia < 1000/mm³ and fever > 38.5°.


InterventionsCefepime 2 gr × 3 versus Piperacillin 3 gr × 6 + Gentamicin 1.5 mg/kg × 3.


OutcomesTreatment failure: bacterial and fungal super-infections; adverse events; dropouts after randomisation.


NotesJournal publication and author correspondence.
Vancomycin addition after 72 hours permitted by protocol, not counted as failure (27 participants).
Outcomes in subgroups.
Documented infections: bacteraemia.

Documented Gram-negative and Pseudomonas infections; haematological cancer patients.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Adequate.

Allocation concealment (selection bias)Low riskA-Adequate.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk12 participants excluded from analysis.

Yildirim 2008

MethodsRandomisation: computer-generated random number list.

Blinding: open.
Intention-to-treat: no.
Interim analysis: no.
Exclusions from analysis: 12/99 participants.
Follow-up period: no information.


ParticipantsTurkey: 99 episodes in 99 participants with haematological malignancy all ≤ 16.

Neutropenia < 500 or < 1000 and expected to decline and fever > 38.5° or 2 measurements > 38°.


InterventionsImipenem/meropenem 20 mg/kg × 3; Piperacillin-tazobactam 80 mg/kg × 4 + Amikacin 7.5 mg/kg × 2.


OutcomesTreatment failure: duration of fever; neutropenia; hospitalisation; mortality; need for additional antibiotics or antifungal drugs.


NotesJournal publication.

Added glycopeptide after 72 hours with persistent fever, added AMP-B after 5 days with continued fever and neutropenia. In participants with monotherapy added aminoglycoside after 72 hours with persistent fever.

No outcomes in subgroups.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskA-Computer-generated random number list.

Allocation concealment (selection bias)Unclear riskB-Unclear.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk12 participants excluded from analysis.

Zengin 2011

MethodsRandomisation: no mention.

Blinding: open,
Intention-to-treat: no,
Interim analysis: no.
Exclusions from analysis: 7/79 episodes.
Follow-up period: no information.


ParticipantsTurkey: 79 episodes in 43 participants with haematological malignancy all ≤ 19 years.

Neutropenia ≤ 500 or ≤ 1000 with decrease to ≤ 500 within 48 hours and fever ≥ 38.5° once or ≥ 8° for longer than 1 hour.


InterventionsPiperacillin-tazobactam 90 mg/kg × 4; Piperacillin-tazobactam 90 mg/kg × 4 + Amikacin 15 mg/kg × 1.


OutcomesTreatment failure.


NotesJournal publication.

Added teicoplanin after 96 hours with persistent fever, added AMP-B or LipAMP-B or fluconazole after 120 hours with persistent neutropenia and fever.

Subgroup analyses: episodes with and without catheter; high-dose cytosine arabinocide in the previous chemotherapy.


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)High riskNo random sequence-participants randomly assigned by presentation.

Allocation concealment (selection bias)High riskNo concealment.

Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label.

Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label.

Incomplete outcome data (attrition bias)
All outcomes
High risk7 episodes excluded from analysis.

 
Characteristics of excluded studies [ordered by study ID]

StudyReason for exclusion

Berezin 2003Retrospective comparative study comparing cefepime monotherapy versus ceftriaxone combined with gentamicin or amikacin.

Bodey 1976The randomisation to beta-lactam versus beta-lactam + aminoglycoside was semi-empirical, after pathogen identification. Empirically the participants all received carbenicillin and cephalotin. Following pathogen isolation, participants with Pseudomonas sp. or Proteus mirabilis infection (analysed together) were randomly assigned to carbenicillin monotherapy versus carbenicillin-gentamicin combination therapy, and those with other Gram-negative infections were randomly assigned to cephalotin versus cephalotin with gentamicin.

Bru 1986Study randomly assigned participants to ticarcillin-clavulanate versus ticarcillin-clavulanate + amikacin; however, several problems preclude its inclusion: Participants following bone marrow transplantation were allocated to combination therapy only, over-riding the random allocation; only the number of evaluated episodes is reported, numbers of randomly assigned participants/episodes unknown; number of episodes described in results is larger than the number of evaluable episodes; most results are reported as percentages and the denominator is unknown.

Cetto 1983Participants with haematological malignancies were randomly assigned to receive cefuroxime or tobramycin plus ampicillin. However, all neutropenic participants also received carbenicillin with both regimens.

D'Antonio 1992Study included non-neutropenic, haematological cancer patients with altered immune defences.

Drusano 1985Study includes monotherapy and combination treatment groups compatible with the protocol, but randomisation was not performed between these groups. Non-neutropenic cancer patients received empirically beta-lactam monotherapy, while neutropenic cancer patients were randomly assigned to double beta-lactam combination therapy versus beta-lactam-aminoglycoside combination therapy.

EORTC 1987Randomisation to monotherapy versus beta-lactam-aminoglycoside combination therapy semi-empirical. Empirically all participants received beta-lactam-aminoglycoside combination therapy. After 3 days, participants were randomly assigned to continue the combination, or to discontinue the aminoglycoside (beta-lactam monotherapy).

Fainstein 1983Randomized study comparing ceftazidime versus ceftazidime + tobramycin. The study randomly assigned 321 episodes in 253 cancer patients with or without neutropenia. A subgroup of participants with neutropenia and documented infection were analysed separately. The number of neutropenic participants per group is not known, only the denominator for the subgroup of participants with neutropenia and documented infections is given. The outcome assessed in the subgroup is failure but does not include the non–infection related deaths. Author contacted without response.

Hauer 1990Non-randomised controlled clinical trial.

Hazel 1998Randomised trial presented as a conference proceeding comparing piperacillin-tazobactam + tobramycin versus imipenem + tobramycin for participants with febrile neutropenia, colonised with ESBL+ Enterobactericeae.

Hoepelman 1988Study includes data on neutropenic and non-neutropenic participants combined. Data on neutropenic participants are not separated.

Karthaus 1998Study not randomised: prospective observational design.

Moreno-Sanchez 1992Randomized trial comparing imipenem versus ceftazidime + amikacin presented in conference. The abstract states that the study is in progress, but no further publications were identified. Results in abstract are given for 31 participants, but the number of dropouts is unknown; only evaluable participants are discussed. Author contacted without response.

Moroni 1987Incompatible comparator regimens: ceftazidime + amikacin versus ceftazidime + vancomycin.

Pegram 1989Randomisation to monotherapy versus beta-lactam-aminoglycoside combination therapy semi-empirical. Empirically all participants were treated with combination therapy. At 4 days, participants were randomly assigned to continue the combination, or to discontinue the aminoglycoside (beta-lactam monotherapy).

Petrilli 1991Non-randomised study describing treatment with imipenem and ceftriaxone monotherapy for high-risk and low-risk febrile neutropenic children, respectively.

Pizzo 1986Randomised trial comparing ceftazidime monotherapy versus double beta-lactam-aminoglycoside combination therapy.

Reilly 1983Study not randomised: patient groups were studied sequentially.

Sampi 1987Study compares two combination regimens: cefmenoxime + amikacin versus piperacillin + amikacin.

Sanz 2005Study not randomised: prospective observational matched cohort study comparing imipenem monotherapy versus piperacillin-tazobactam + amikacin for febrile neutropenia.

Sawae 1996Study randomly assigned participants to imipenem monotherapy or combination therapy. The combination arm included several different combinations (beta-lactam-aminoglycoside combinations, beta-lactam-beta-lactam combinations and other combinations), but the decision as to which combination therapy the patient received was left to the physician's decision. (Personal communication with author.)

Wrzesien-Kus 2000Comparison between cefepime and ceftazidime, both combined with amikacin.

 
Characteristics of studies awaiting assessment [ordered by study ID]
Li 1998

MethodsNo information

ParticipantsFebrile neutropenia

InterventionsCeftazidime and netilmicin

OutcomesNo information

Notes

 
Characteristics of ongoing studies [ordered by study ID]
Bilgir 2005

Trial name or titleThe comparison of imipenem with piperacillin/tazobactam and amikacin combination in participants with haematological malignancies in the treatment of febrile neutropenia.

Methods

ParticipantsTurkey: 40 participants with haematological malignancies.

InterventionsImipenem versus piperacillin-tazobactam + amikacin.

OutcomesTreatment failure and adverse events reported only as percentages, without a denominator per group.

Starting dateUnknown. Results presented in EHA 2005.

Contact informationDr. O. Bilgir, Okmeydani Hastanesi, Izmir, Turkey.

NotesAuthor's address unknown.

 
Comparison 1. Overall effectiveness

Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size

 1 All cause mortality447186Risk Ratio (M-H, Fixed, 95% CI)0.87 [0.75, 1.02]

    1.1 same beta-lactam
111718Risk Ratio (M-H, Fixed, 95% CI)0.74 [0.53, 1.06]

    1.2 different beta-lactam
335468Risk Ratio (M-H, Fixed, 95% CI)0.91 [0.77, 1.09]

 2 Infection-related mortality416872Risk Ratio (M-H, Fixed, 95% CI)0.80 [0.64, 0.99]

    2.1 same beta-lactam
81403Risk Ratio (M-H, Fixed, 95% CI)0.68 [0.43, 1.10]

    2.2 different beta-lactam
335469Risk Ratio (M-H, Fixed, 95% CI)0.83 [0.65, 1.06]

 3 Treatment failure71Risk Ratio (M-H, Fixed, 95% CI)Subtotals only

    3.1 same beta-lactam
162833Risk Ratio (M-H, Fixed, 95% CI)1.11 [1.02, 1.20]

    3.2 different beta-lactam
557736Risk Ratio (M-H, Fixed, 95% CI)0.92 [0.88, 0.97]

 
Comparison 2. Superinfections

Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size

 1 Bacterial superinfections294961Risk Ratio (M-H, Fixed, 95% CI)1.02 [0.87, 1.19]

 2 Fungal superinfections203437Risk Ratio (M-H, Fixed, 95% CI)0.70 [0.49, 1.00]

 
Comparison 3. Adverse events

Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size

 1 Any adverse event (monotherapy)497412Risk Ratio (M-H, Fixed, 95% CI)0.87 [0.81, 0.94]

    1.1 imipenem monotherapy
121429Risk Ratio (M-H, Fixed, 95% CI)0.99 [0.86, 1.13]

    1.2 meropenem monotherapy
92003Risk Ratio (M-H, Fixed, 95% CI)0.95 [0.84, 1.06]

    1.3 ceftazidime monotherapy
91941Risk Ratio (M-H, Fixed, 95% CI)0.64 [0.53, 0.76]

    1.4 moxalactam monotherapy
5421Risk Ratio (M-H, Fixed, 95% CI)0.71 [0.51, 0.97]

    1.5 cefepime monotherapy
81079Risk Ratio (M-H, Fixed, 95% CI)0.94 [0.75, 1.17]

    1.6 other monotherapy
7539Risk Ratio (M-H, Fixed, 95% CI)0.93 [0.61, 1.44]

 2 Discontinuation due to adverse event164051Risk Ratio (M-H, Fixed, 95% CI)0.61 [0.40, 0.93]

 3 Any nephrotoxicity - Ag dosing regimen396608Risk Ratio (M-H, Fixed, 95% CI)0.45 [0.35, 0.57]

    3.1 Once daily
81707Risk Ratio (M-H, Fixed, 95% CI)0.31 [0.15, 0.63]

    3.2 Multiple daily
314901Risk Ratio (M-H, Fixed, 95% CI)0.47 [0.36, 0.61]

 4 Severe nephrotoxicity - Ag dosing regimen204199Risk Ratio (M-H, Fixed, 95% CI)0.16 [0.05, 0.49]

    4.1 Once daily
61526Risk Ratio (M-H, Fixed, 95% CI)0.20 [0.03, 1.14]

    4.2 Multiple daily
142673Risk Ratio (M-H, Fixed, 95% CI)0.14 [0.03, 0.60]

 
Comparison 4. Documented infections (subgroup analysis)

Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size

 1 All cause mortality131188Risk Ratio (M-H, Fixed, 95% CI)0.81 [0.56, 1.17]

 2 Treatment failure35Risk Ratio (M-H, Fixed, 95% CI)Subtotals only

    2.1 same beta-lactam
81043Risk Ratio (M-H, Fixed, 95% CI)1.05 [0.93, 1.19]

    2.2 different beta-lactam
272740Risk Ratio (M-H, Fixed, 95% CI)0.89 [0.82, 0.95]

 
Comparison 5. Bacteraemia (subgroup analysis)

Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size

 1 All cause mortality14676Risk Ratio (M-H, Fixed, 95% CI)0.74 [0.46, 1.18]

 2 Treatment failure26Risk Ratio (M-H, Fixed, 95% CI)Subtotals only

    2.1 same beta-lactam
6395Risk Ratio (M-H, Fixed, 95% CI)1.05 [0.90, 1.23]

    2.2 different beta-lactam
201149Risk Ratio (M-H, Fixed, 95% CI)0.86 [0.78, 0.95]

 
Comparison 6. Gram-negative infections (subgroup analysis)

Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size

 1 All cause mortality16376Risk Ratio (M-H, Fixed, 95% CI)0.64 [0.37, 1.11]

 2 Treatment failure29Risk Ratio (M-H, Fixed, 95% CI)Subtotals only

    2.1 same beta-lactam
7261Risk Ratio (M-H, Fixed, 95% CI)1.34 [1.03, 1.74]

    2.2 different beta-lactam
22603Risk Ratio (M-H, Fixed, 95% CI)0.74 [0.60, 0.90]

 
Comparison 7. Pseudomonas infections (subgroup analysis)

Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size

 1 All cause mortality971Risk Ratio (M-H, Fixed, 95% CI)0.87 [0.34, 2.24]

 2 Treatment failure16Risk Ratio (M-H, Fixed, 95% CI)Subtotals only

    2.1 same beta-lactam
349Risk Ratio (M-H, Fixed, 95% CI)1.41 [0.90, 2.22]

    2.2 different beta-lactam
1399Risk Ratio (M-H, Fixed, 95% CI)0.89 [0.60, 1.31]

 
Comparison 8. Haematological cancer patients (subgroup analysis)

Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size

 1 All cause mortality223463Risk Ratio (M-H, Fixed, 95% CI)0.88 [0.68, 1.13]

 2 Treatment failure32Risk Ratio (M-H, Fixed, 95% CI)Subtotals only

    2.1 same beta-lactam
8778Risk Ratio (M-H, Fixed, 95% CI)1.04 [0.91, 1.20]

    2.2 different beta-lactam
243671Risk Ratio (M-H, Fixed, 95% CI)0.94 [0.88, 1.01]

 
Comparison 9. Severe neutropenia (subgroup analysis)

Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size

 1 All cause mortality6737Risk Ratio (M-H, Fixed, 95% CI)0.68 [0.37, 1.24]

 2 Treatment failure11Risk Ratio (M-H, Fixed, 95% CI)Subtotals only

    2.1 same beta-lactam
2237Risk Ratio (M-H, Fixed, 95% CI)1.48 [1.12, 1.96]

    2.2 different beta-lactam
9871Risk Ratio (M-H, Fixed, 95% CI)0.96 [0.84, 1.10]

 
Comparison 10. Monotherapy

Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size

 1 All cause mortality43Risk Ratio (M-H, Fixed, 95% CI)Subtotals only

    1.1 ceftazidime monotherapy
101868Risk Ratio (M-H, Fixed, 95% CI)0.89 [0.70, 1.14]

    1.2 imipenem monotherapy
91164Risk Ratio (M-H, Fixed, 95% CI)1.01 [0.68, 1.50]

    1.3 meropenem monotherapy
91921Risk Ratio (M-H, Fixed, 95% CI)1.14 [0.77, 1.69]

    1.4 moxalactam monotherapy
2140Risk Ratio (M-H, Fixed, 95% CI)0.52 [0.26, 1.06]

    1.5 piperacillin-tazobactam monotherapy
51093Risk Ratio (M-H, Fixed, 95% CI)0.62 [0.40, 0.96]

    1.6 cefepime monotherapy
5802Risk Ratio (M-H, Fixed, 95% CI)1.08 [0.61, 1.93]

    1.7 other monotherapy
4320Risk Ratio (M-H, Fixed, 95% CI)0.65 [0.19, 2.25]

 2 Treatment failure65Risk Ratio (M-H, Fixed, 95% CI)Subtotals only

    2.1 same BL - ceftazidime
6647Risk Ratio (M-H, Fixed, 95% CI)1.07 [0.94, 1.23]

    2.2 same BL - imipenem
167Risk Ratio (M-H, Fixed, 95% CI)3.05 [0.92, 10.10]

    2.3 same BL - piperacillin-tazobactam
3911Risk Ratio (M-H, Fixed, 95% CI)1.04 [0.92, 1.18]

    2.4 same BL - cefepime
3343Risk Ratio (M-H, Fixed, 95% CI)1.08 [0.84, 1.39]

    2.5 same BL - other monotherapy
144Risk Ratio (M-H, Fixed, 95% CI)0.83 [0.30, 2.33]

    2.6 different BL - ceftazidime
101917Risk Ratio (M-H, Fixed, 95% CI)0.96 [0.89, 1.05]

    2.7 different BL - imipenem
141964Risk Ratio (M-H, Fixed, 95% CI)0.88 [0.78, 1.01]

    2.8 different BL - meropenem
81542Risk Ratio (M-H, Fixed, 95% CI)0.88 [0.79, 0.98]

    2.9 different BL - moxalactam
5402Risk Ratio (M-H, Fixed, 95% CI)0.77 [0.60, 0.99]

    2.10 different BL - piperacillin-tazobactam
2203Risk Ratio (M-H, Fixed, 95% CI)0.74 [0.53, 1.02]

    2.11 different BL - cefepime
5377Risk Ratio (M-H, Fixed, 95% CI)0.97 [0.77, 1.22]

    2.12 different BL - other
7575Risk Ratio (M-H, Fixed, 95% CI)1.03 [0.83, 1.28]

 
Comparison 11. Adults vs. children

Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size

 1 All cause mortality447186Risk Ratio (M-H, Fixed, 95% CI)0.87 [0.75, 1.02]

    1.1 children
9789Risk Ratio (M-H, Fixed, 95% CI)0.80 [0.39, 1.64]

    1.2 mixed/ undefined
62089Risk Ratio (M-H, Fixed, 95% CI)0.74 [0.52, 1.04]

    1.3 adults
294308Risk Ratio (M-H, Fixed, 95% CI)0.93 [0.77, 1.12]

 2 Treatment failure68Risk Ratio (M-H, Fixed, 95% CI)Subtotals only

    2.1 same BL - children
2163Risk Ratio (M-H, Fixed, 95% CI)1.34 [0.95, 1.90]

    2.2 same BL - mixed
3985Risk Ratio (M-H, Fixed, 95% CI)1.01 [0.90, 1.14]

    2.3 same BL - adults
111685Risk Ratio (M-H, Fixed, 95% CI)1.17 [1.04, 1.32]

    2.4 different BL - children
121086Risk Ratio (M-H, Fixed, 95% CI)1.02 [0.87, 1.18]

    2.5 different BL - mixed/ undefined
112263Risk Ratio (M-H, Fixed, 95% CI)0.93 [0.83, 1.04]

    2.6 different BL - adults
294160Risk Ratio (M-H, Fixed, 95% CI)0.90 [0.85, 0.96]

 
Comparison 12. Sensitivity analysis (outcome in parenthesis)

Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size

 1 Allocation concealment (mortality)447186Risk Ratio (M-H, Fixed, 95% CI)0.87 [0.75, 1.02]

    1.1 A
245489Risk Ratio (M-H, Fixed, 95% CI)0.88 [0.73, 1.05]

    1.2 B
191625Risk Ratio (M-H, Fixed, 95% CI)0.87 [0.61, 1.24]

    1.3 C
172Risk Ratio (M-H, Fixed, 95% CI)0.0 [0.0, 0.0]

 2 Intention-to-treat vs. efficacy analysis (mortality)447186Risk Ratio (M-H, Fixed, 95% CI)0.87 [0.75, 1.02]

    2.1 efficacy analysis
204432Risk Ratio (M-H, Fixed, 95% CI)0.88 [0.73, 1.06]

    2.2 intention-to-treat analysis
242754Risk Ratio (M-H, Fixed, 95% CI)0.87 [0.66, 1.15]

 3 Unit of randomisation (mortality)447186Risk Ratio (M-H, Fixed, 95% CI)0.87 [0.75, 1.02]

    3.1 patient analysis
193711Risk Ratio (M-H, Fixed, 95% CI)0.84 [0.66, 1.08]

    3.2 episode analysis
253475Risk Ratio (M-H, Fixed, 95% CI)0.90 [0.73, 1.11]

 4 Publication status (mortality)437110Risk Ratio (M-H, Fixed, 95% CI)0.88 [0.75, 1.03]

    4.1 jounal publication
345811Risk Ratio (M-H, Fixed, 95% CI)0.84 [0.71, 1.00]

    4.2 other publication or un-published
91299Risk Ratio (M-H, Fixed, 95% CI)1.07 [0.72, 1.59]

 5 Trial size (mortality)447186Risk Ratio (M-H, Fixed, 95% CI)0.87 [0.75, 1.02]

    5.1 number randomised>median 94p
195438Risk Ratio (M-H, Fixed, 95% CI)0.90 [0.75, 1.07]

    5.2 number randomised<median 94p
251748Risk Ratio (M-H, Fixed, 95% CI)0.78 [0.55, 1.11]

 6 Allocation concealment (failure)6910357Risk Ratio (M-H, Fixed, 95% CI)0.97 [0.93, 1.01]

    6.1 same beta-lactam - A
61310Risk Ratio (M-H, Fixed, 95% CI)1.10 [0.99, 1.22]

    6.2 same beta-lactam - B
91451Risk Ratio (M-H, Fixed, 95% CI)1.13 [0.99, 1.30]

    6.3 same beta-lactam - C
172Risk Ratio (M-H, Fixed, 95% CI)1.03 [0.73, 1.46]

    6.4 different beta-lactam - A
214422Risk Ratio (M-H, Fixed, 95% CI)0.94 [0.88, 1.00]

    6.5 different beta-lactam - B
313052Risk Ratio (M-H, Fixed, 95% CI)0.87 [0.79, 0.96]

    6.6 different beta-lactam - C
150Risk Ratio (M-H, Fixed, 95% CI)0.8 [0.48, 1.34]

 7 Intention to treat vs. efficacy analysis (failure)70Risk Ratio (M-H, Random, 95% CI)Subtotals only

    7.1 same BL - efficacy analysis
121884Risk Ratio (M-H, Random, 95% CI)1.11 [0.98, 1.26]

    7.2 same BL - ITT analysis
4949Risk Ratio (M-H, Random, 95% CI)1.04 [0.91, 1.19]

    7.3 different BL - efficacy analysis
386010Risk Ratio (M-H, Random, 95% CI)0.95 [0.88, 1.01]

    7.4 different BL - ITT analysis
161659Risk Ratio (M-H, Random, 95% CI)0.80 [0.71, 0.91]

 8 Intention to treat vs. efficacy analysis, assuming dropouts=failures (failure)68Risk Ratio (M-H, Fixed, 95% CI)Subtotals only

    8.1 same BL - efficacy analysis
51238Risk Ratio (M-H, Fixed, 95% CI)1.15 [1.02, 1.29]

    8.2 same BL - ITT analysis
101590Risk Ratio (M-H, Fixed, 95% CI)1.07 [0.96, 1.19]

    8.3 different BL - efficacy analysis
203037Risk Ratio (M-H, Fixed, 95% CI)0.96 [0.89, 1.04]

    8.4 different BL - ITT analysis
334922Risk Ratio (M-H, Fixed, 95% CI)0.92 [0.86, 0.97]

 9 Trial size (failure)70Risk Ratio (M-H, Fixed, 95% CI)Subtotals only

    9.1 same BL no. randomised>median
72210Risk Ratio (M-H, Fixed, 95% CI)1.10 [1.01, 1.21]

    9.2 same BL no. randomised<median
9623Risk Ratio (M-H, Fixed, 95% CI)1.14 [0.94, 1.39]

    9.3 different BL no. randomised>median
286032Risk Ratio (M-H, Fixed, 95% CI)0.98 [0.92, 1.03]

    9.4 different BL no. randomised<median
261637Risk Ratio (M-H, Fixed, 95% CI)0.75 [0.67, 0.84]

 10 Unit of randomisation (failure)71Risk Ratio (M-H, Fixed, 95% CI)Subtotals only

    10.1 same beta-lactam - patient
61212Risk Ratio (M-H, Fixed, 95% CI)1.05 [0.93, 1.19]

    10.2 same beta-lactam - episode
101621Risk Ratio (M-H, Fixed, 95% CI)1.16 [1.04, 1.30]

    10.3 different beta-lactam - patient
203137Risk Ratio (M-H, Fixed, 95% CI)0.89 [0.82, 0.96]

    10.4 different beta-lactam - episode
364656Risk Ratio (M-H, Fixed, 95% CI)0.95 [0.89, 1.01]

 11 Blinding (failure)71Risk Ratio (M-H, Fixed, 95% CI)Subtotals only

    11.1 same beta-lactam - double blind
1754Risk Ratio (M-H, Fixed, 95% CI)1.04 [0.90, 1.20]

    11.2 same beta-lactam - other
152079Risk Ratio (M-H, Fixed, 95% CI)1.14 [1.04, 1.26]

    11.3 different beta-lactam - double blind
3623Risk Ratio (M-H, Fixed, 95% CI)1.14 [0.83, 1.55]

    11.4 different beta-lactam - other
527113Risk Ratio (M-H, Fixed, 95% CI)0.92 [0.87, 0.96]

 12 Publication status (failure)71Risk Ratio (M-H, Fixed, 95% CI)Subtotals only

    12.1 same beta-lactam - journal publication
122496Risk Ratio (M-H, Fixed, 95% CI)1.12 [1.02, 1.21]

    12.2 same beta-lactam - other
4337Risk Ratio (M-H, Fixed, 95% CI)1.05 [0.79, 1.41]

    12.3 different beta-lactam - journal publication
445866Risk Ratio (M-H, Fixed, 95% CI)0.90 [0.86, 0.96]

    12.4 different beta-lactam - other
111870Risk Ratio (M-H, Fixed, 95% CI)1.00 [0.89, 1.12]

 
Summary of findings for the main comparison. beta-lactam monotherapy compared to beta-lactam-aminoglycoside combination therapy for Febrile neutropenic cancer patients

Beta-lactam monotherapy compared with beta-lactam-aminoglycoside combination therapy for febrile neutropenic cancer patients

Patient or population: febrile neutropenic cancer patients.
Settings:
Intervention: beta-lactam monotherapy.
Comparison: beta-lactam-aminoglycoside combination therapy.

OutcomesIllustrative comparative risks* (95% CI)Relative effect
(95% CI)
No of Participants
(studies)
Quality of the evidence
(GRADE)
Comments

Assumed riskCorresponding risk

Beta-lactam-aminoglycoside combination therapyBeta-lactam monotherapy

All cause mortalityStudy populationRR 0.87
(0.75 to 1.02)
7186
(44 studies)
⊕⊕⊕⊕
high

83 per 100072 per 1000
(62 to 85)

Moderate

68 per 100059 per 1000
(51 to 69)

Any nephrotoxicity - Ag dosing regimen (Copy)Study populationRR 0.45
(0.35 to 0.57)
6608
(39 studies)
⊕⊕⊕⊕
high

57 per 100026 per 1000
(20 to 33)

Moderate

29 per 100013 per 1000
(10 to 17)

Treatment failure - same beta-lactamStudy populationRR 1.11
(1.02 to 1.2)
2833
(16 studies)
⊕⊕⊕⊝
moderate1

405 per 1000449 per 1000
(413 to 485)

Moderate

398 per 1000442 per 1000
(406 to 478)

Treatment failure - different beta-lactamStudy populationRR 0.92
(0.88 to 0.97)
7736
(55 studies)
⊕⊕⊝⊝
low1,2,3,4

426 per 1000392 per 1000
(375 to 413)

Moderate

432 per 1000397 per 1000
(380 to 419)

*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
CI: Confidence interval; RR: Risk ratio.

GRADE Working Group grades of evidence:
High quality: Further research is very unlikely to change our confidence in the estimate of effect.
Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.
Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.
Very low quality: We are very uncertain about the estimate.

 1 Outcome determined mainly by treatment modifications. Poor correlation with all cause mortality, the ultimate target of treating cancer patients.
2 Differences decreased with low risk of bias regarding allocation concealment.
3 Differences in effects between published and unpublished trials.
4 No explanation was provided.&&