Intended for healthcare professionals


Safety and toxicity of amphotericin B in glucose 5% or intralipid 20% in neutropenic patients with pneumonia or fever of unknown origin: randomised study

BMJ 1998; 317 doi: (Published 08 August 1998) Cite this as: BMJ 1998;317:379
  1. Patrick Schöffski, physician (Patrick_Schoffski{at},
  2. Mathias Freund, professor of medicined,
  3. R Wunder, physiciana,
  4. D Petersen, physicianb,
  5. C H Köhne, physiciand,
  6. H Hecker, professor of biometricsc,
  7. U Schubert, documentation assistanta,
  8. A Ganser, professor of medicinea
  1. aDepartment of Haematology/Oncology, Hanover University Medical School, D-30625 Hanover, Germany
  2. bDepartment of Clinical Chemistry
  3. cBiometrical Department
  4. dDepartment of Haematology/Oncology, Rostock University, D-18055 Rostock, Germany
  1. Correspondence to: Dr Schöffski
  • Accepted 5 May 1998


Objective : To compare the feasibility of treatment, safety, and toxicity of intravenous amphotericin B deoxycholate prepared in either glucose or intralipid for empirical antimycotic treatment of neutropenic cancer patients.

Design : Single centre stratified, randomised non-blinded phase II study.

Setting : University hospital providing tertiary clinical care.

Subjects : 51 neutropenic patients (leukaemia (35), lymphoma (11), solid tumours (5)) with refractory fever of unknown origin (24) or pneumonia (27).

Interventions : Amphotericin B 0.75 mg/kg/day in 250 ml glucose 5% solution or mixed with 250 ml intralipid 20%, given on eight consecutive days then alternate days, as a 1-4 hour infusion.

Main outcome measures : Feasibility of treatment, subjective tolerance (questionnaire), and objective toxicity (common toxicity criteria of the National Cancer Institute).

Results : Study arms were balanced for age, sex, underlying malignancy, renal and liver function, and pre- and concomitant treatment with antibiotics and nephrotoxic agents. No statistically significant or clinically relevant differences were found between the treatment groups for: daily or cumulative dose and duration of treatment with amphotericin B; incidence and time of dose modifications or infusion duration changes related to toxicity; dose or duration of symptomatic support with opiates, antipyretics, or antihistamines; renal function; subjective tolerance; most common toxicity scores; course of infection; and incidence of treatment failures. Patients treated with amphotericin B in intralipid were given fewer diuretics (P<0.05) and therefore had more peripheral oedema (P<0.01) and needed less potassium supplementation (P<0.05) than patients given amphotericin in glucose. Acute respiratory events were more common in the intralipid arm (P<0.05).

Conclusions : Amphotericin B 0.75 mg/kg/day in intralipid given on eight consecutive days then alternate days provides no benefit and is associated with potential pulmonary side effects possibly because of fat overload or an incompatibility of the two drugs.

Key messages

  • Intralipid does not decrease the renal toxicity and subjective side effects associated with intravenous amphotericin B at a dose of 0.75 mg/kg/day in neutropenic cancer patients

  • Use of amphotericin B in intralipid 20% can be associated with transient pulmonary toxicity, possibly related to fat overload

  • 250 ml of intralipid 20% should not be given as a short term infusion or mixed with agents other than those recommended by the manufacturer

  • On the basis of published data, amphotericin B and intralipid 20% should be regarded as chemically incompatible

  • The combination of amphotericin B and intralipid should not be given to patients


Amphotericin B is regarded as the agent of choice for treatment of life threatening mycoses in neutropenic patients because of its broad antimycotic activity.1 It is conventionally given intravenously in glucose 5%, as a colloidal suspension with the detergent sodium deoxycholate. Amphotericin B is associated with a high incidence of renal toxicity, potassium loss, fever, and chills. Attempts have been made to overcome its dose limiting renal toxicity.2 Well tolerated, highly expensive liposomal formulations are commercially available, and can be used in patients developing renal toxicity after exposure to amphotericin B deoxycholate.

Non-liposomal lipid emulsions are also known to reduce toxic effects of amphotericin in vitro and in vivo and have been given to patients.311 Intralipid, a solution of soya bean oil, phosphatidylcholine, glycerol, and water, has shown promise as a carrier for amphotericin, and in clinical trials has been observed to reduce some side effects associated with amphotericin. 8 1214 We carried out a prospective randomised phase II trial to determine the safety and toxicity of amphotericin in glucose 5% or intralipid 20% in neutropenic cancer patients with either fever of unknown origin or pneumonia.

Subjects and methods

Patient selection criteria — Neutropenic cancer patients with pulmonary infiltrates or patients with refractory fever of unknown origin during treatment with broad spectrum antibiotics qualified for the trial. Empirical amphotericin was given to high risk patients not being treated with antifungal agents within seven days before entry to the study.

Randomisation and stratification — Patients were stratified according to presence of refractory fever of unknown origin or pneumonia, and underlying malignancy, then block randomised for treatment with amphotericin in either glucose or intralipid.

Baseline and follow up examinations — All patients underwent a standardised clinical evaluation including radiography, ultrasonography, blood counts, serum analyses, microbiology, and serology. Bronchoscopy was done whenever possible.

Treatment — Amphotericin in glucose was prepared as recommended by the manufacturer (Bristol-Myers Squibb, Munich, Germany), and amphotericin in intralipid as described.13 Amphotericin was dissolved in glucose 5% (10 mg amphotericin/ml) then mixed with 250 ml intralipid 20% (Kabi-Pharmacia, Erlangen, Germany) and infused without delay. The non-blinded infusions were prepared and given in a uniform manner. Both arms had the same amphotericin dose and schedule (0.75 mg/kg/day for eight days then alternate days). Dose escalation was prohibited, but the infusion duration could be prolonged from 1-4 hours.

Concurrent treatment — Piperacillin or third generation cephalosporins plus aminoglycosides were given as first line treatment and changed to an imipenem and glycopeptide combination in patients with refractory infections after three days.15 Amphotericin was part of the first line treatment in patients with pulmonary infiltrates. Patients with refractory fever of unknown origin received amphotericin from the fourth day. The concomitant use of 5-flucytosine was allowed. Opioids, antihistamines, or antiemetics were not given before the first dose, growth factors were not routinely given, and granulocyte transfusions were not given.


Baseline patient characteristics at study entry (data are numbers of patients unless stated otherwise)

View this table:

Assessment of feasibility and toxicity — The feasibility of treatment was evaluated with the protocol compliance, incidence of infusion duration changes or delays related to toxicity, and use of supportive drugs. A standardised questionnaire was completed by the patients on days 0, 1, 4, and 8, assessing 16 subjective side effects. Objective toxicities were evaluated according to modified criteria of the National Cancer Institute.

Assessment of efficacy — The patients were treated empirically without proof of systemic mycosis, and received a variety of concomitant antimicrobial agents. The study was thus not designed to determine antifungal efficacy, but to re-evaluate the feasibility of treatment with amphotericin and intralipid and its safety and toxicity.

Study ethics — The protocol was approved by our ethics committee and performed according to the Declaration of Helsinki. All patients gave their informed consent.

Statistics — The trial's design was based on a prior power calculation for the endpoints of serum creatinine concentration and creatinine clearance. The possible difference between the study arms was estimated from the literature.12 We planned to randomise 50 patients for each arm, to achieve a power of 90% for the detection of an effect size of δ=d/s=0.667 when applying a two sided t test at the 5% level. For creatinine clearance, this corresponds to a difference of d=20 ml/min between the treatments assuming a SD of s=30 ml/min within each group. After observation of severe side effects not previously reported, we performed a non-scheduled interim analysis of toxicity data resulting in premature termination of the trial for patient safety. With the sample size (n=51) the minimal effect size under the same conditions is δ=d/s=0.93. Discrete variables were compared by χ2 analysis, Mantel-Haenszel, or Fisher's exact test depending on the expected frequencies, and continuous variables were compared with a two sample t test or Mann-Whitney U test depending on the skewness. A P value of <0.05 was considered significant (table 1).


Twenty four patients were randomised to receive amphotericin in glucose and 27 to receive amphotericin in intralipid. The study arms were balanced for age, sex, malignancy, and all relevant laboratory values at baseline (table 1). The concurrent use of nephrotoxic and antimicrobial agents including 5-flucytosine was identical in both arms; opiates, antipyretics, and antihistamines were required in comparable dosages irrespective of treatment assignment (data not shown). Prednisolone was used significantly more in the glucose arm as part of the anticancer treatment (mean cumulative dose 285 (SD 479.0) versus 95 (155.8) mg; P=0.007, t test).


Administration of amphotericin B

View this table:

The daily and cumulative dose of amphotericin, overall duration of treatment, incidence of infusion duration changes, or dose reductions due to toxicity were not significantly different (table 2). The study arms were balanced for sodium chloride supplementation known to be capable of preventing renal toxicity; the sodium content of all drugs was taken into account (data not shown).

Renal tubular damage was obvious in both groups. Mean serum potassium concentration fell uniformly during the course of amphotericin treatment irrespective of treatment assignment (data not shown). The patients were efficiently supplemented. The cumulative potassium requirement was significantly lower in the intralipid arm (1750 (1480) versus 1194 (972) mmol; P=0.037). This observation was related to unexplained differences in the use of frusemide (furosemide) (59.4 (18.0) versus 19.0 (21.9) mg/day; P=0.028). Peripheral oedema was therefore noted more often in the intralipid group (P=0.009, Mantel-Haenszel χ2 test). Complications possibly related to electrolyte imbalance, such as arrhythmia, were rare and occurred in three controls and six patients treated with the lipid emulsion.

Effects of amphotericin on renal function were evident (fig). Compared with baseline, serum creatinine and blood urea nitrogen concentrations increased during the course of treatment. Significant differences between the study arms, however, were not found for any of the variables of renal function (table 3; data for blood urea nitrogen concentration not shown). Four patients—one in the control group and three in the experimental arm — had severe renal dysfunction. One patient in the conventional group and two patients in the intralipid group underwent haemodialysis. Clogging of the dialysis membranes, possibly related to hyperlipidaemia, occurred repeatedly in the two patients in the intralipid group.


Mean (SE) values for creatinine clearance (normal range 80-170 ml/min) (two upper curves) and serum creatinine concentration (50-80 μmol/l) (two lower curves) in patients treated with amphotericin B in glucose 5% (n=24) or intralipid 20% (n=27) (differences between two randomisation arms not significant)


Renal function

View this table:

Aspartate transaminase concentration was mildly elevated and significantly higher in the glucose group throughout treatment (mean 17.9 (16.0) versus 10.0 (5.1) U/l; P=0.028, t test). No differences were noted for alanine transaminase or bilirubin concentrations (data not shown). Liver dysfunction was not clinically evident.

Other non-haematological toxicities occurred frequently in both groups — most were related to the use of amphotericin, and others to the underlying malignancy and its treatment (table 4). Common grade 1-2 toxicities were chills, fever, and sweating. These occurred in up to 69% of patients and showed little difference between the study arms. Severe grade 3-4 toxicity was not uncommon, and discomfort and dyspnoea were the most frequent observations. A variety of renal, gastrointestinal, cardiovascular, and respiratory toxicities were documented. These were evenly distributed in both study arms, the exception being acute pulmonary events.

Eleven patients in the glucose group (four with refractory fever of unknown origin and seven with pneumonia) and 17 patients in the intralipid group (five with refractory fever of unknown origin and 12 with pneumonia) had pulmonary symptoms. Three patients in the glucose group and two in the experimental group had grade 1-2 dyspnoea. Grade 3-4 acute dyspnoea occurred in four control patients and in 11 patients in the intralipid group (P=0.083, Mantel-Haenszel test). Mild coughing was present in nine and 13 patients respectively, and grade 3-4 coughing was observed in one patient in each group. Other pulmonary events (eg, respiratory pain, pleuritis, fibrosis, adult respiratory distress syndrome, respirator treatment) occurred in one patient in the conventional arm and in eight patients in the intralipid group (graded 3-4 in one glucose patient and five intralipid patients; P=0.029, Mantel-Haenszel test).

We repeatedly observed acute respiratory distress after initiation of the lipid infusion, sometimes associated with coughing, tachypnoea, agitation, cyanosis, and deterioration of oxygen saturation. Pulmonary events occurred when amphotericin in intralipid was repeatedly given to patients with either pneumonia or fever of unknown origin, with either one or four hour infusions. This toxicity seemed unrelated to pre-existing pulmonary conditions, state of infection, or extent of infiltrates, and was not attributable to the use of specific agents or possible predisposing factors such as fluid overload. The events were evaluated with bedside tests and radiography, and treated with discontinuation of the infusion, oxygen supplementation, and monitoring of vital signs. The events were of sudden onset and reversible in minutes or hours. We were not able to perform specific pulmonary function tests, bronchoscopy, or biopsy to elucidate this toxicity because of the poor condition of the neutropenic and thrombocytopenic patients and the acute, self limiting nature of the events. The patients' case records and radiological findings were reviewed in detail without showing further information. We stopped amphotericin and intralipid treatment early in two patients because of pulmonary events.


Evaluation of maximum non-haematological toxicity per patient based on questionnaire data and physician toxicity ratings according to modified common toxicity criteria of National Cancer Institute (grade 1-2=mild to moderate, grade 3-4=severe or life threatening events)

View this table:

Pulmonary biopsy showed two cases of invasive candidiasis in each study arm, which were diagnosed during treatment. Six deaths related to cancer or chemotherapy occurred in the glucose group, none attributable to proved mycotic infection. Three of the four patients who died in the intralipid group had respiratory failure—one due to pulmonary candidiasis and one due to multiorgan aspergillosis—otherwise, the course of infection and incidence of treatment failures were identical in both arms.


Systemic mycotic infections are a major threat to immunocompromised individuals because of their increasing incidence and high mortality.16 The current strategy for prevention and treatment of these infections in neutropenic patients is based on the early empirical use of toxic antifungal agents such as amphotericin.

Our study was initiated in 1993 when randomised trials suggested that amphotericin and intralipid combined was less toxic and better tolerated than conventional amphotericin (table 5). We tried to confirm these observations in the context of a larger randomised trial, since the amphotericin and intralipid combination was not licensed by health authorities and its feasibility and safety inadequately evaluated.17


Prospective trials comparing toxicity of amphotericin B in glucose or intralipid

View this table:

Renal toxicity, potassium loss, fever, and chills were evident in both of our study arms. A reduced cumulative need for potassium supplementation was confounded by differences in individual diuretic treatment and did not translate into a decreased incidence of cardiac arrhythmia. The prospective analysis of various measures of feasibility and toxicity did not show a clinically relevant or statistically significant advantage of the lipid emulsion, and questionnaire data did not indicate a subjective benefit for our patients. Though the study had to be stopped prematurely for patient safety, the statistical power of our data was good enough to provide strong evidence that, in contrast with other observations, intralipid did not reduce the toxicity of amphotericin.

Pulmonary toxicity from conventional amphotericin is rare and usually related to bronchospasms or acute febrile reactions during the early phase of treatment. Respiratory distress with sudden dyspnoea, hypoxaemia, haemoptysis, and interstitial infiltrates has been observed in patients treated simultaneously with granulocyte transfusions.18 Severe dyspnoea, agitation, and chest tightness did occur in patients treated with liposomal amphotericin.19 Pulmonary events have never been reported in patients receiving amphotericin and intralipid. They were, however, frequent in our trial and unrelated to granulocyte transfusions, underlying medical conditions, or known risk factors. A significant weakness of this study is that pulmonary toxicity was not evaluated in a highly standardised manner. Blood-gas analysis or measurement of peripheral oxygen saturation was not routinely done, as there was no reported evidence of potential respiratory events. Bedside tests were performed only in patients with pulmonary distress. Because of the sudden, transient nature of the symptoms, however, a routine evaluation might not have been helpful. The interpretation of our pulmonary toxicity data is biased by the fact that pneumonia by chance was more common in the intralipid arm, although this difference was not statistically significant.

The temporal relation between the infusion of amphotericin in intralipid and the clinical manifestation of pulmonary symptoms suggests a causal association. The respiratory events were most likely related to the unconventional use of intralipid, or its incompatibility with amphotericin.

Several studies have reported varying intensity of pulmonary dysfunction when fat emulsions were given to seriously ill patients.20 In septicaemic patients rapid infusions of intralipid induce temporary, dose dependent changes in diffusion capacity and arterial oxygenation and an increase in pulmonary artery pressure. Short term infusions of intralipid 10-20% at a rate of 1-4 ml/min do result in a transient decrease in pulmonary diffusion capacity, even in healthy volunteers. 21 22 For this reason, although not exclusively, nutrition based on soya bean oil is given cautiously as a 12-24 hour infusion. A more specific complication is the fat overload syndrome, a rare condition related to repeated infusions of concentrated soya bean oil preparations, and associated with variable end organ dysfunction. This syndrome can present clinically with cough, dyspnoea, tachypnoea, or cyanosis, but has never been reported with antimycotic treatment. 23 24

The pulmonary toxicity could have also been related to an interaction between intralipid and amphotericin. The product information of intralipid highlights the increased risk of incompatibility with other drugs. None of the manufacturers of soya bean oil solutions, including the manufacturer of intralipid, recommends the use of lipid amphotericin. On the basis of theoretical considerations, clinical observations, and several recent pharmacological studies, amphotericin and intralipid should be regarded chemically incompatible. Several different concentrations of amphotericin in either intralipid 10 or 20% or other soya bean oil solutions, and both the Caillot and Moreau type of emulsion, have been studied under various experimental conditions.25 Amphotericin lipid mixtures are unstable, show an increase in particle size in emulsion over a short period, and do precipitate. Patients are therefore at risk of discontinuous administration of amphotericin, and possibly embolism. Self made lipid emulsions of amphotericin B should be regarded as unsafe until more pharmacological data are available.

The antimycotic efficacy of amphotericin B in intralipid is another controversial issue. Some evidence shows activity of this combination against superficial infections like HIV associated oropharyngeal candidiasis, or candidaemia in neutropenic cancer patients. The antimycotic activity of the lipid emulsion in systemic organ mycosis caused by candida or aspergillus, however, is still unproved.


Amphotericin and intralipid should be regarded as a highly experimental drug combination with an unclear pharmacological profile, unproved antimycotic efficacy, and potential risks. On the basis of the findings of our prospective randomised phase II study and a review of the current literature we conclude that the use of amphotericin in intralipid is not an acceptable way to give the antifungal agent to neutropenic cancer patients.



We thank Antje Luttmann and Radoslaw Kowalski for their help with the data acquisition and analysis, and Dr Ulrike Rapp-Bernhard, Department of Diagnostic Radiology, Hannover University Medical School, for independently reviewing all xray films and computed tomograms. Parts of this work were presented orally at the 36th interscience conference on antimicrobial agents and chemotherapy, 1996, New Orleans, Louisiana (organised by the American Society of Microbiology). Other parts were presented orally and as a poster at the 8th European congress of clinical microbiology and infectious diseases, 1997, Lausanne, Switzerland (organised by the European Society of Clinical Microbiology and Infectious Diseases).

Contributors: PS had the original idea for this study, prepared, coordinated, performed, and analysed the clinical trial, and wrote the paper; he will act as guarantor for the paper. MF coinitiated the research, discussed the study hypothesis, and participated in the protocol design. RW was responsible for data acquisition, documentation, and analysis, and contributed to the paper. DP helped with the interpretation of clinical chemistry data. CHK was involved in the treatment of study patients and discussed the main study results. HH was responsible for the statistical design and the power calculation. US did the statistical analysis. AG discussed core ideas, gave an interpretation of the findings, and contributed to the paper.


  • Funding None.

  • Conflict of interest None.


View Abstract