Scolaris Content Display Scolaris Content Display

Administration systématique versus sélective d'antifongiques dans la prévention d'infections fongiques chez les patients cancéreux

Collapse all Expand all

Résumé scientifique

Contexte

L'infection fongique systémique est considérée comme une cause importante de morbidité et de mortalité chez les patients cancéreux, et particulièrement chez ceux présentant une neutropénie. Les médicaments antifongiques sont souvent donnés à titre prophylactique, ou de manière empirique aux patients atteints de fièvre persistante.

Objectifs

Évaluer si les médicaments antifongiques utilisés couramment réduisent la mortalité chez les patients cancéreux présentant une neutropénie.

Stratégie de recherche documentaire

Nous avons effectué des recherches dans PubMed, de 1966 au 7 juillet 2014, et dans les références bibliographiques des articles identifiés.

Critères de sélection

Des essais cliniques randomisés comparant l'amphotéricine B, le fluconazole, le kétoconazole, le miconazole, l'itraconazole ou le voriconazole, au placebo ou à l'absence de traitement chez les patients cancéreux présentant une neutropénie.

Recueil et analyse des données

Les deux auteurs de la revue ont indépendamment évalué l'éligibilité des essais ainsi que les risques de biais et en ont extrait les données.

Résultats principaux

Trente‐deux essais impliquant 4 287 patients ont été inclus. Le traitement prophylactique ou empirique à l'amphotéricine B a réduit de façon significative la mortalité totale (risque relatif (RR) 0,69 ; intervalle de confiance (IC) à 95 % de 0,50 à 0,96), alors que les estimations de risque relatif pour le fluconazole, le kétoconazole, le miconazole et l'itraconazole étaient proches de 1,00. Aucun essai éligible avec le voriconazole n'a été trouvé. L'amphotéricine B et le fluconazole ont réduit la mortalité imputable aux infections fongiques, RR 0,45 (IC à 95 % de 0,26 à 0,76) et RR 0,42 (IC à 95 % de 0,24 à 0,73) respectivement. L'incidence des infections fongiques invasives a diminué significativement avec l'administration d'amphotéricine B (RR 0,41 ; IC à 95 % de 0,24 à 0,73), de fluconazole (RR 0,39 ; IC à 95 % de 0,27 à 0,57) et d'itraconazole (RR 0,53 ; IC à 95 % de 0,29 à 0,97), mais pas avec le kétoconazole ou le miconazole. L'estimation des effets était similaire pour les 13 essais effectués en aveugle avec une assignation secrète adéquate. Le compte rendu des préjudices causés variait beaucoup trop d'un essai à l'autre pour permettre une évaluation significative. Lors des mises à jour de 2011 et 2014, aucun essai supplémentaire n’a été identifié pour inclusion.

Conclusions des auteurs

L'amphotéricine B par voie intraveineuse a été le seul agent antifongique ayant réduit la mortalité totale. Elle devrait donc être préférée lors de la mise en place d'un traitement antifongique prophylactique ou empirique chez les patients cancéreux présentant une neutropénie.

PICOs

Population
Intervention
Comparison
Outcome

The PICO model is widely used and taught in evidence-based health care as a strategy for formulating questions and search strategies and for characterizing clinical studies or meta-analyses. PICO stands for four different potential components of a clinical question: Patient, Population or Problem; Intervention; Comparison; Outcome.

See more on using PICO in the Cochrane Handbook.

Prévention des infections fongiques au moyen de médicaments antifongiques chez les patients atteints de cancer

Les patients cancéreux recevant une chimiothérapie ou une greffe de moelle osseuse sont exposés aux infections fongiques. Celles‐ci peuvent être mortelles, surtout quand elles se propagent dans tout le corps. Les patients ayant une faible numération de globules blancs (neutropénie) sont particulièrement à risque. Les médicaments antifongiques sont souvent administrés de manière systématique à titre préventif, ou lorsque des personnes à risque ont de la fièvre. La revue a constaté que l'amphotéricine B par voie intraveineuse pourrait réduire la mortalité. Trois des médicaments (l'amphotéricine B, le fluconazole et l'itraconazole) ont permis de réduire les infections fongiques.

Authors' conclusions

Implications for practice

Intravenous amphotericin B was the only antifungal agent that reduced total mortality. It should therefore be preferred when prophylactic or empirical antifungal therapy is instituted in cancer patients with neutropenia.

Implications for research

It may be difficult to justify further placebo‐controlled trials whereas there is a need for unbiased trials comparing intravenous amphotericin B with other antifungal drugs. Such trials should comprise at least 1000 patients and they should include data on length of hospital stay and similar measures, allowing cost‐benefit analyses to be performed. Data on the incidence of bacterial infections should be collected since azole compounds might increase the risk of such infections.

Background

Bacterial infections are an important cause of death in cancer patients (Inagaki 1974), and patients with low white blood cell counts are particularly at risk (Estey 1982). Systemic fungal infection, with wide dissemination of infection throughout the body, is also considered to be an important cause of morbidity and mortality (Meyers 1990; Verfaillie 1991). This type of infection is mainly caused by Candida or Aspergillus species (Walsh 1991). The mortality rate in patients with Candida sepsis or deep tissue involvement is around 75% (Meyers 1990; Verfaillie 1991), and a positive blood culture or histologic signs of invasion were found before death in 37% of the patients in one series of patients (Estey 1982).

Antifungal agents are often given prophylactically in conjunction with chemotherapy or bone marrow transplantation, or empirically to patients without documented fungal infection but with persisting fever despite antibiotic treatment. The rationale for these approaches is to start therapy before it is too late, since it is difficult to diagnose an invasive fungal infection with certainty (Verfaillie 1991; Walsh 1990).

Studies with historical controls have shown a positive effect of antifungal agents on mortality (Stein 1986; Walsh 1991) but such non‐randomised comparisons have been shown to overestimate the effect of cancer treatments considerably (Berlin 1989). We performed a meta‐analysis of trials in which commonly used antifungal agents were compared with an untreated control group.

Objectives

The primary aim was to assess whether commonly used antifungal agents decrease mortality in cancer patients with neutropenia.

Methods

Criteria for considering studies for this review

Types of studies

Only randomised trials were included. We accepted reports in any language. Studies concerned with the treatment or prevention of oral candidiasis were excluded.

Types of participants

Cancer patients with neutropenia caused by chemotherapy or bone marrow transplantation, as defined by the researchers within each study.

Types of interventions

Experimental: amphotericin B (including lipid soluble formulations), fluconazole, ketoconazole, miconazole, itraconazole or voriconazole given orally or intravenously. Control: placebo or no treatment.

Types of outcome measures

  • Mortality

  • Mortality ascribed to fungal infection

  • Invasive fungal infection (defined as positive blood culture, oesophageal candidiasis, lung infection or microscopically confirmed deep tissue involvement)

  • Colonisation

  • Use of additional (escape) antifungal therapy

  • Harms

Search methods for identification of studies

Electronic searches

We searched PubMed from 1966 to 7 July 2014 and the reference lists of identified articles.

The search strategy used is in Appendix 1.

The search strategies have been developed and executed by the authors.

Searching other resources

This has not been carried out since 2007 as we have not found it worthwhile.

Data collection and analysis

Data extraction and management

Decisions on which trials to include, and also which variables to use when a number of options were available for the same outcome, were based on the methods sections of the trials. Details on diagnosis, drug, dose, rules for use of additional (rescue) antifungal therapy, average length of treatment with placebo, length of follow‐up, randomisation (Schulz 1995) and blinding methods, number of randomised patients, number of patients excluded from analysis, deaths, invasive fungal infections, colonisation and use of rescue drug were independently extracted by both review authors; differences in the data extracted were resolved by consensus.

We defined invasive fungal infection as a positive blood culture, oesophageal candidiasis, lung infection or microscopically confirmed deep tissue infection. We excluded cases of oropharyngeal and vulvovaginal candidiasis, skin infections, Candida in the urine and vaguely described infections.

We asked the trial authors to confirm the extracted information and to answer additional questions; numbers in the tables may therefore be different from those given in the published articles. To increase the response rate, the authors' most recent addresses were located in PubMed. In an attempt to increase the power of the meta‐analyses and avoid reporting bias, the trial authors were specifically asked for the three months mortality data for all randomised patients, also secondarily excluded patients. We asked the trial authors for details on the randomisation process, especially whether it was concealed and irreversible so that an allocation could not be known beforehand or changed later.

Data synthesis

The outcomes were weighted by the inverse variance. Since heterogeneity of the studies was expected because of various designs; diagnoses; drugs; doses and routes of administration; and criteria for fungal invasion, colonisation and use of rescue drug a random‐effects model was used. A fixed‐effect analysis was preferred, however, if the P value was greater than 0.10 for the test of heterogeneity. Ninety‐five per cent confidence intervals (CI) were presented.

Results

Description of studies

We identified 44 potentially eligible trials of which 12 were excluded: two were not randomised (Hiddemann 1991; Schaison 1990); one concerned oropharyngeal candidiasis only (Samonis 1990); one had a control group receiving active drugs (Wang 2003); in one only 14 of 146 patients had neutropenia and only data on oropharyngeal candidiasis were provided (Bodey 1990); one only reported a subgroup of 72 of 298 randomised patients who underwent autopsy (Ezdinli 1979); one used a surrogate outcome (resolution of fever) and only one patient developed candidaemia (Schiel 2006); one was published as an abstract without useful data and where no fungal diseases were found (Reed 1993); and four were excluded since they were unpublished and we could not obtain any data from them (Benhamou 1991b; Brincker 1990; Prentice 1989; Siegel 1982b).

Two of the included 32 trials were published only as abstracts (Acuna 1981; Siegel 1982a); one was an interim analysis (Goldstone 1994); and two were published in Japanese (Fukuda 1994; Suda 1980).

Finally, we found a conference abstract describing a small placebo‐controlled trial of voriconazole (25 patients) (Cornely 2006). This trial awaits assessment and may not be eligible as the primary outcome was lung infiltrates, which have other causes than fungal infection.

For the 2011 and 2014 updates no additional trials were identified for inclusion.

Risk of bias in included studies

We adopted broad quality assessment criteria and considered the risk of bias as low if the randomisation method was concealed, if central randomisation, use of sealed envelopes, a code provided by a pharmacy or a company was described; generation of the allocation sequence was adequate (for example random numbers); and the trial was placebo controlled and blinded. On one occasion what seemed to be a sound randomisation, provided by a pharmacy, on further questioning proved to result in medicine packages labelled A or B (Vreugdenhil 1993), which one would not have expected for a trial published in 1993. Such a procedure is risky as code breaking for just one patient would make it possible to predict all future allocations. Thirteen trials had adequate allocation concealment and were blinded (Brincker 1978; Brincker 1983; Goodman 1992; Kelsey 1999; Nucci 2000; Perfect 1992; Riley 1994; Rotstein 1999; Schaffner 1995; Tollemar 1993; Vreugdenhil 1993; Wingard 1987; Winston 1993). One trial maintained the blinding during data analysis (Schaffner 1995).

The antifungal agent was given prophylactically in 29 trials and empirically in three. Acute leukaemia was the most common indication in 21 trials, bone marrow transplantation in 11. The length of the follow‐up period was often not reported; it probably varied for different patients even within the same study since many trial authors stated that the trial drugs had been given till the neutropenia had resolved.

Effects of interventions

Thirty‐two trials involving 4287 patients were included. Prophylactic or empirical treatment with amphotericin B significantly decreased total mortality (relative risk (RR) 0.69, 95% confidence interval (CI) 0.50 to 0.96). The point estimate was the same for those four trials that had adequate allocation concealment and were blinded (RR 0.69, 95% CI 0.41 to 1.19). Intravenous administration of amphotericin B was used in the trials apart from one in which the drug was given orally (Suda 1980). RRs for the other drugs were close to 1.00: fluconazole, RR of 1.04 (95% CI 0.84 to 1.30); ketoconazole, RR of 0.97 (95% CI 0.63 to 1.49); miconazole, RR of 1.16 (95% CI 0.71 to 1.87); and itraconazole, RR of 0.94 (95% CI 0.63 to 1.40).

Amphotericin B and fluconazole decreased mortality ascribed to fungal infection (RR 0.45, 95% CI 0.26 to 0.76 and RR 0.42, 95% CI 0.24 to 0.73, respectively).

The incidence of invasive fungal infection decreased significantly with administration of amphotericin B (RR 0.41, 95% CI 0.24 to 0.73), fluconazole (RR 0.39, 95% CI 0.27 to 0.57) and itraconazole (RR 0.53, 95% CI 0.29 to 0.97); but not with ketoconazole (RR 1.32, 95% CI 0.68 to 2.54) or miconazole (RR 0.52, 95% CI 0.20 to 1.31). Effect estimates were similar for those 13 trials that had adequate allocation concealment and were blinded. We did not find any eligible trials with voriconazole.

The trial authors' definitions of fungal colonisation and their methods varied widely and there was considerable heterogeneity between the trials for the effects of the drugs. The overall reduction in fungal colonisation was statistically significant for prophylactic amphotericin B, fluconazole and ketoconazole.

Use of rescue antifungal agents tended to be more common in the untreated groups. There was considerable heterogeneity for those drugs that were used in most of the studies.

The reporting of harms was far too variable from trial to trial to allow a meaningful overview. Usually no treatment discontinuations because of harms were reported in the fluconazole trials, whereas 16% discontinued in one trial with itraconazole (Menichetti 1999) and only 3% in another trial with this drug (Nucci 2000). We have listed the most important harms that were reported in Table 1.

Open in table viewer
Table 1. Harms

Trial

Trial drug

Number of patients

Harms

EORTC 1989

amphotericin B

80 versus 77

Treatment discontinuations: 6 pts on trial drug (infusion‐related toxicity, allergic reactions); Nephrotoxicity: 8 versus 3, none required dialysis and renal function restored later; Hypokalaemia: 33 versus 16

Goldstone 1994

amphotericin B

64 versus 69

Treatment discontinuations: 4 on trial drug (chills, rigour, hypotension, rash and bronchospasm); Elevated liver function tests: 26 versus 32; Nephrotoxicity: 1 on trial drug that did not require withdrawal of therapy

Kelsey 1999

amphotericin B

74 versus 87

Treatment discontinuations: 5 on trial drug, 1 on placebo because of immediate reactions; Nephrotoxicity: 9 versus 6; Hypokalaemia: 1 versus 0; Clinical adverse events were very similar in the two groups

Penack 2006.

amphotericin B

75 versus 57

Treatment discontinuations: 2 (1 skin rash, 1 chills) versus 0; Other harms: "no differences in ... liver function tests, renal function parameters and hypokalaemia"

Perfect 1992

amphotericin B

91 versus 91

More infusion‐related harms on active drug, but no data provided; No significant differences in renal function, hepatic enzymes or electrolytes (no data provided)

Pizzo 1982

amphotericin B

18 versus 16

Treatment discontinuations: none; Rash: 2 versus 1; Azotaemia: 1 versus 0; liver enzyme elevations: 2 versus 3; Electrolyte abnormalities: 18 versus 16

Riley 1994

amphotericin B

17 versus 18

Treatment discontinuations: none; Renal function: no difference (P value 0.82 for blood urea, P value 0.63 for creatinine); Potassium supplements: no difference; Infusion reactions: none

Suda 1980

amphotericin B

39 versus 31

Article is in Japanese

Tollemar 1993

amphotericin B

42 versus 42

Treatment discontinuations: 4 versus 0 for infusion reactions; Potassium supplementation: no difference; Renal function: no useful data, but only small changes reported, compared to normal ranges

Fukuda 1994

fluconazole

37 versus 26

Article is in Japanese

Goodman 1992

fluconazole

179 versus 177

Treatment discontinuations: 1 on trial drug, 2 on placebo for clinical side effects, and 17 versus 11 for elevated liver function tests; in 7 versus 3, hepatic dysfunction contributed to death; Graft‐versus‐host disease or organ failure: 44 versus 24 deaths; Nausea: 13 versus 9; Skin rash: 9 versus 9; Eosinophilia in 6 versus 0

Kern 1998

fluconazole

36 versus 32

Bacteriaemia: 15 versus 7; Other harms similar: 5 versus 6 elevations in transaminases, 19 versus 17 nausea or vomiting, 3 versus 0 allergy

Rotstein 1999

fluconazole

141 versus 133

Treatment discontinuations: none reported; Elevated liver enzymes: 17 versus 19; Rash: 51 versus 59; Nausea: 106 versus 95; Vomiting: 68 versus 85

Schaffner 1995

fluconazole

76 versus 76 episodes

Treatment discontinuations: none reported; No data on harms ("no significant differences were found")

Slavin 1995

fluconazole

152 versus 148

Treatment discontinuations: 32 versus 31 for abnormal liver function; Graft‐versus‐host disease: 102 versus 85; Nausea: 33 versus 22; Seizures: 8 versus 9; Liver enzymes: no differences

Winston 1993

fluconazole

124 versus 133

Treatment discontinuations: none reported; Elevated liver enzymes: 25 versus 14; Nausea and vomiting: 9 versus 5; Rash: 13 versus 7; Other harms were similarly distributed

Yamac 1995

fluconazole

41 versus 29

Treatment discontinuations: none reported; Other harms: no data

Acuna 1981

ketoconazole

28 versus 24

No data

Benhamou 1991

ketoconazole

63 versus 62

Treatment discontinuations: 38 versus 14 (among them 2 patients with veno‐occlusive disease, 1 hepatitis, 3 skin rash in active group; none in placebo group); Severe gastrointestinal intolerance: 4 versus 7; Three‐fold greater values than normal for transaminases: 13 versus 8

Brincker 1983

ketoconazole

19 versus 19

One patient on trial drug stopped treatment because of universal exanthema

Estey 1984

ketoconazole

77 versus 73

Bacterial infections: 33 versus 24

Hansen 1987

ketoconazole

27 versus 29

Treatment discontinuations: 2 on trial drug (1 skin rash and 1 elevated liver function tests); Bacterial infections: 20% versus 15% of neutropenic episodes

Hughes 1983

ketoconazole

42 versus 22

Treatment discontinuations:1 on trial drug (nausea and anorexia); Other harms: 2 nausea and anorexia, 1 abdominal pain, 1 transient rash, all on trial drug

Palmblad 1992

ketoconazole

55 versus 61

Treatment discontinuations: 2 (elevated transaminases) versus 6 (1 elevated transaminases, 5 exanthema); Bacteriaemias: 37 versus 21

Siegel 1982a

ketoconazole

12 versus 13

No data

Brincker 1978

miconazole

15 versus 15

None ascribable to trial drug

Wingard 1987

miconazole

97 versus 111

Treatment discontinuations: 1 (pruritis and flushing) versus 2 (1 rash, 1 nausea); Severe hypotension: 2 on trial drug

Caselli 1990

itraconazole, amphotericin B, ketoconazole

30 versus 10

No data

Kaptan 2003

itraconazole

31 versus 24

Treatment discontinuations: 2 on trial drug (cardiac arrhytmia and gastric irritation); "there was a clinical impression that hypokalaemia occurred at a greater rate in patients with itraconazole"

Menichetti 1999

itraconazole

201 versus 204

Treatment discontinuations: 37 versus 27; Bacteriaemia: 47 versus 31; Elevated transaminases: 5 versus 3

Nucci 2000

itraconazole

104 versus 106

Treatment discontinuations: 3 versus 4; Skin rash: 3 versus 1; Elevated liver enzymes: 3 versus 4; Nausea: 2 on placebo

Vreugdenhil 1993

itraconazole

49 versus 49

Treatment discontinuations: 1 (nausea) versus 1 (liver function deterioration); Liver function deterioration: 28 versus 22 episodes; Renal function deterioration: 4 versus 2 episodes

Discussion

Amphotericin B was the only antifungal agent among those we studied that significantly decreased total mortality. The trials were relatively small but this finding is supported by another review in which we reported on three trials that compared intravenous lipid soluble amphotericin B (AmBisome) with smaller doses of standard intravenous amphotericin B (Johansen 2001). The relative risk (RR) for total mortality in these trials was 0.74 (95% CI 0.52 to 1.07).

The advantage of using total mortality as an outcome measure is not only that it is unbiased but it may also be the most relevant outcome since the drugs could have important harms leading to drug‐related mortality. Ketoconazole, for example, is immunosuppressive and in all three trials in which bacterial infections were reported they were more common with ketoconazole than with placebo; 37 versus 21 patients (Palmblad 1992), 33 versus 24 patients (Estey 1984), and 20% of neutropenic courses versus 15% (Hansen 1987). Interestingly this adverse effect could be a class effect related to azoles as an increased incidence of bacteraemias has also been reported with fluconazole, 27 versus 16 patients (Schaffner 1995) and 15 versus 7 patients (Kern 1998); and with itraconazole, 47 versus 31 patients (Menichetti 1999). Fluconazole has been associated with an excess of graft‐versus‐host disease or organ failure in the two large studies of bone marrow transplant recipients; 29 versus 16 deaths (Goodman 1992), or 44 versus 24 according to later correspondence (Goodman 1992), and 102 versus 85 cases of graft‐versus‐host disease (Slavin 1995). It is noteworthy that the largest trial of fluconazole found no difference in total mortality, 55 versus 46 deaths (RR 1.18, 95% CI 0.85 to 1.65), whereas fewer deaths were ascribed to acute systemic fungal infections in the group receiving fluconazole compared to the group receiving placebo (1 of 179 versus 10 of 177 patients, P value 0.01) (Goodman 1992). This indicates that fluconazole increases mortality from other causes (54 versus 36 deaths, P value 0.04). Itraconazole has been associated with congestive heart failure, which does not seem to be a class effect (Ahmad 2001).

In an eight year follow‐up of the only study that has reported a significant but possibly flawed effect of fluconazole on total mortality (Slavin 1995) the trial authors ascribe this finding to the fact that most of their patients (88%) had received allogeneic grafts. They write that the other large trial (Goodman 1992) primarily included autologous graft recipients, but in fact 48% of the patients in that trial had also received allogeneic grafts and the trial had three times as many deaths as the trial authors' own trial. Commentators (Slavin 1995) have suggested that the trial authors' positive result on mortality might represent a statistical aberration. We agree that this seems more likely than the tentative explanation based on type of graft offered by the trial authors. Furthermore, we suspect that a biased decision on length of follow‐up may have been made in this trial. A conference abstract notes a follow‐up period of 75 days (maximum length of treatment) plus an additional two weeks, that is 89 days (Slavin 1995). The final article gives no explanation why the follow‐up period had been extended to 110 days. We used the data after 89 days (21 versus 28 deaths), which came closest to the three months follow‐up we aimed for in the meta‐analysis and which we assume was also the time frame stipulated in the trial protocol for the study.

Another bias that involves reporting at favourable time points relates to informal interim analyses. Concern about bias in cancer trials that is caused by lax stopping rules has previously been raised (Pocock 1978) and a survey showed that the majority of cancer cooperative groups perform annual interim analyses of their trials without formal stopping rules (Buyse 1993). A trial author informed us that one of his studies was stopped prematurely after 30 patients when an interim analysis showed a significant effect, but the trial report did not describe that this had occurred or that the study was planned to include more patients (Brincker 1978). A third study report did not mention any interim analysis, although an earlier conference abstract reported an interim analysis (Riley 1994); a fourth study described interim analyses but gave no rules (Estey 1984); and a fifth study has only been published as an interim analysis (Goldstone 1994).

Publication bias (Stern 1997), which is a well documented phenomenon in cancer trials (Berlin 1989; Simes 1986), is also of concern. A study on fluconazole was stopped by the company in 1990 when 32 patients had been entered; the investigator never learned about the results (Brincker 1990, personal communication). Our efforts to make contact with the medical companies were disappointing; Pfizer and Janssen‐Cilag did not wish to share their unpublished reports with us, or even to give us a list of the trials so that we could approach the investigators. This secrecy, which we have previously reported for Pfizer's trials that compared fluconazole with amphotericin B (Johansen 1999), and which other meta‐analysts have reported when they approached Janssen (Huston 1996), is unethical and must be changed (Chalmers 1990). Clinical trial data can only be assembled through the patients' willingness to contribute to science for the benefit of future patients; the data should therefore be regarded as public property to be used for the common good. It has been amply documented that the withholding of data that are not favourable for the drug in question may be harmful to patients.

It should be noted that a possible effect on total mortality may be overlooked if rescue antifungal therapy is instituted too quickly in the control group. The positive effect of amphotericin B was seen despite the fact that rescue antifungal therapy was introduced rather quickly in the control group, for example after an average of 10, 10 and 15 days in three of the most positive trials (Perfect 1992; Pizzo 1982; Riley 1994); there were no data on this in the trial by Penack 2006. This problem was hardly the reason for the lack of effect of fluconazole as the three largest studies with fluconazole 400 mg daily reported the use of rescue antifungal therapy after an average of 14 days (Winston 1993), 20 days (Goodman 1992) and 55 days (Slavin 1995).

In previous versions of this review we did not include death attributed to fungal infection as an outcome measure. It is difficult to determine the cause of death in these severely ill patients and we therefore suspected that disease‐specific mortality would be unreliable. As we could not confirm this suspicion in a study of disease‐specific mortality (Due 2006), we have now included this outcome.

The harms we have listed in Table 1 should be interpreted cautiously. Biased reporting of harms is very common (Chan 2004; Gøtzsche 1989), and we suspected that biased reporting had occurred in several of the trials we reviewed. For example, arbitrary cut‐offs delineating what constitutes a clinically relevant increase in creatinine and liver enzymes can have a major impact on what is found. We have described this type of biased reporting in a pivotal trial that compared voriconazole with liposomal amphotericin B (Jørgensen 2006). The investigators found that 29 versus 32 patients had a two‐fold increase in S‐creatinine. They also found that 43 versus 80 patients experienced a 1.5‐fold increase (P value < 0.001), which is the result they reported in the abstract. This case was unusual as the bias was so obvious. We have never seen a 1.5‐fold increase in S‐creatinine reported in other trials, and the lack of clinical relevance of the trivial difference in S‐creatinine was underlined by the fact that two patients in the voriconazole group and one in the amphotericin B group died from renal failure.

Amphotericin B, fluconazole and itraconazole all had an effect on invasive fungal infection. The data on mortality that we have presented in this review suggest that amphotericin B is a better drug than fluconazole. The comparisons are indirect, however. Unfortunately it is not possible to make a direct judgement on whether fluconazole is of similar effectiveness as amphotericin B on mortality or invasive infection despite the fact that more than a dozen comparative trials have been published (Johansen 1999; Johansen 2002a). First, most of the trials have used oral amphotericin B, which is not absorbed from the gastrointestinal tract and which is poorly documented. Second, the results for amphotericin B were combined with those for nystatin in large three‐armed trials despite the fact that nystatin is no better than placebo when given to patients with severe immunodeficiency, such as those with cancer complicated by neutropenia (Johansen 1999; Johansen 2002b). It is also of concern that widespread use of fluconazole could lead to the development of resistance or to infection with inherently resistant species of fungi, for example Candida krusei and Candida (Torulopsis) glabrata (Working Party 1995). For itraconazole, little data is available at present to judge its efficacy compared with amphotericin B.

Comparison 1 Antifungals versus placebo or no treatment, Outcome 1 Death.
Figures and Tables -
Analysis 1.1

Comparison 1 Antifungals versus placebo or no treatment, Outcome 1 Death.

Comparison 1 Antifungals versus placebo or no treatment, Outcome 2 Death related to fungal infection.
Figures and Tables -
Analysis 1.2

Comparison 1 Antifungals versus placebo or no treatment, Outcome 2 Death related to fungal infection.

Comparison 1 Antifungals versus placebo or no treatment, Outcome 3 Invasive infections.
Figures and Tables -
Analysis 1.3

Comparison 1 Antifungals versus placebo or no treatment, Outcome 3 Invasive infections.

Comparison 1 Antifungals versus placebo or no treatment, Outcome 4 Colonisation.
Figures and Tables -
Analysis 1.4

Comparison 1 Antifungals versus placebo or no treatment, Outcome 4 Colonisation.

Comparison 1 Antifungals versus placebo or no treatment, Outcome 5 Use of escape drug.
Figures and Tables -
Analysis 1.5

Comparison 1 Antifungals versus placebo or no treatment, Outcome 5 Use of escape drug.

Table 1. Harms

Trial

Trial drug

Number of patients

Harms

EORTC 1989

amphotericin B

80 versus 77

Treatment discontinuations: 6 pts on trial drug (infusion‐related toxicity, allergic reactions); Nephrotoxicity: 8 versus 3, none required dialysis and renal function restored later; Hypokalaemia: 33 versus 16

Goldstone 1994

amphotericin B

64 versus 69

Treatment discontinuations: 4 on trial drug (chills, rigour, hypotension, rash and bronchospasm); Elevated liver function tests: 26 versus 32; Nephrotoxicity: 1 on trial drug that did not require withdrawal of therapy

Kelsey 1999

amphotericin B

74 versus 87

Treatment discontinuations: 5 on trial drug, 1 on placebo because of immediate reactions; Nephrotoxicity: 9 versus 6; Hypokalaemia: 1 versus 0; Clinical adverse events were very similar in the two groups

Penack 2006.

amphotericin B

75 versus 57

Treatment discontinuations: 2 (1 skin rash, 1 chills) versus 0; Other harms: "no differences in ... liver function tests, renal function parameters and hypokalaemia"

Perfect 1992

amphotericin B

91 versus 91

More infusion‐related harms on active drug, but no data provided; No significant differences in renal function, hepatic enzymes or electrolytes (no data provided)

Pizzo 1982

amphotericin B

18 versus 16

Treatment discontinuations: none; Rash: 2 versus 1; Azotaemia: 1 versus 0; liver enzyme elevations: 2 versus 3; Electrolyte abnormalities: 18 versus 16

Riley 1994

amphotericin B

17 versus 18

Treatment discontinuations: none; Renal function: no difference (P value 0.82 for blood urea, P value 0.63 for creatinine); Potassium supplements: no difference; Infusion reactions: none

Suda 1980

amphotericin B

39 versus 31

Article is in Japanese

Tollemar 1993

amphotericin B

42 versus 42

Treatment discontinuations: 4 versus 0 for infusion reactions; Potassium supplementation: no difference; Renal function: no useful data, but only small changes reported, compared to normal ranges

Fukuda 1994

fluconazole

37 versus 26

Article is in Japanese

Goodman 1992

fluconazole

179 versus 177

Treatment discontinuations: 1 on trial drug, 2 on placebo for clinical side effects, and 17 versus 11 for elevated liver function tests; in 7 versus 3, hepatic dysfunction contributed to death; Graft‐versus‐host disease or organ failure: 44 versus 24 deaths; Nausea: 13 versus 9; Skin rash: 9 versus 9; Eosinophilia in 6 versus 0

Kern 1998

fluconazole

36 versus 32

Bacteriaemia: 15 versus 7; Other harms similar: 5 versus 6 elevations in transaminases, 19 versus 17 nausea or vomiting, 3 versus 0 allergy

Rotstein 1999

fluconazole

141 versus 133

Treatment discontinuations: none reported; Elevated liver enzymes: 17 versus 19; Rash: 51 versus 59; Nausea: 106 versus 95; Vomiting: 68 versus 85

Schaffner 1995

fluconazole

76 versus 76 episodes

Treatment discontinuations: none reported; No data on harms ("no significant differences were found")

Slavin 1995

fluconazole

152 versus 148

Treatment discontinuations: 32 versus 31 for abnormal liver function; Graft‐versus‐host disease: 102 versus 85; Nausea: 33 versus 22; Seizures: 8 versus 9; Liver enzymes: no differences

Winston 1993

fluconazole

124 versus 133

Treatment discontinuations: none reported; Elevated liver enzymes: 25 versus 14; Nausea and vomiting: 9 versus 5; Rash: 13 versus 7; Other harms were similarly distributed

Yamac 1995

fluconazole

41 versus 29

Treatment discontinuations: none reported; Other harms: no data

Acuna 1981

ketoconazole

28 versus 24

No data

Benhamou 1991

ketoconazole

63 versus 62

Treatment discontinuations: 38 versus 14 (among them 2 patients with veno‐occlusive disease, 1 hepatitis, 3 skin rash in active group; none in placebo group); Severe gastrointestinal intolerance: 4 versus 7; Three‐fold greater values than normal for transaminases: 13 versus 8

Brincker 1983

ketoconazole

19 versus 19

One patient on trial drug stopped treatment because of universal exanthema

Estey 1984

ketoconazole

77 versus 73

Bacterial infections: 33 versus 24

Hansen 1987

ketoconazole

27 versus 29

Treatment discontinuations: 2 on trial drug (1 skin rash and 1 elevated liver function tests); Bacterial infections: 20% versus 15% of neutropenic episodes

Hughes 1983

ketoconazole

42 versus 22

Treatment discontinuations:1 on trial drug (nausea and anorexia); Other harms: 2 nausea and anorexia, 1 abdominal pain, 1 transient rash, all on trial drug

Palmblad 1992

ketoconazole

55 versus 61

Treatment discontinuations: 2 (elevated transaminases) versus 6 (1 elevated transaminases, 5 exanthema); Bacteriaemias: 37 versus 21

Siegel 1982a

ketoconazole

12 versus 13

No data

Brincker 1978

miconazole

15 versus 15

None ascribable to trial drug

Wingard 1987

miconazole

97 versus 111

Treatment discontinuations: 1 (pruritis and flushing) versus 2 (1 rash, 1 nausea); Severe hypotension: 2 on trial drug

Caselli 1990

itraconazole, amphotericin B, ketoconazole

30 versus 10

No data

Kaptan 2003

itraconazole

31 versus 24

Treatment discontinuations: 2 on trial drug (cardiac arrhytmia and gastric irritation); "there was a clinical impression that hypokalaemia occurred at a greater rate in patients with itraconazole"

Menichetti 1999

itraconazole

201 versus 204

Treatment discontinuations: 37 versus 27; Bacteriaemia: 47 versus 31; Elevated transaminases: 5 versus 3

Nucci 2000

itraconazole

104 versus 106

Treatment discontinuations: 3 versus 4; Skin rash: 3 versus 1; Elevated liver enzymes: 3 versus 4; Nausea: 2 on placebo

Vreugdenhil 1993

itraconazole

49 versus 49

Treatment discontinuations: 1 (nausea) versus 1 (liver function deterioration); Liver function deterioration: 28 versus 22 episodes; Renal function deterioration: 4 versus 2 episodes

Figures and Tables -
Table 1. Harms
Comparison 1. Antifungals versus placebo or no treatment

Outcome or subgroup title

No. of studies

No. of participants

Statistical method

Effect size

1 Death Show forest plot

26

3902

Risk Ratio (M‐H, Fixed, 95% CI)

0.94 [0.81, 1.09]

1.1 Amphotericin

9

988

Risk Ratio (M‐H, Fixed, 95% CI)

0.69 [0.50, 0.96]

1.2 Fluconazole

7

1470

Risk Ratio (M‐H, Fixed, 95% CI)

1.04 [0.84, 1.30]

1.3 Ketoconazole

4

429

Risk Ratio (M‐H, Fixed, 95% CI)

0.97 [0.63, 1.49]

1.4 Miconazole

2

238

Risk Ratio (M‐H, Fixed, 95% CI)

1.16 [0.71, 1.87]

1.5 Itraconazole

4

777

Risk Ratio (M‐H, Fixed, 95% CI)

0.94 [0.63, 1.40]

2 Death related to fungal infection Show forest plot

23

3490

Risk Ratio (M‐H, Fixed, 95% CI)

0.52 [0.38, 0.71]

2.1 Amphotericin

9

988

Risk Ratio (M‐H, Fixed, 95% CI)

0.45 [0.26, 0.76]

2.2 Fluconazole

6

1213

Risk Ratio (M‐H, Fixed, 95% CI)

0.42 [0.24, 0.73]

2.3 Ketoconazole

3

304

Risk Ratio (M‐H, Fixed, 95% CI)

1.49 [0.55, 4.04]

2.4 Miconazole

1

208

Risk Ratio (M‐H, Fixed, 95% CI)

0.13 [0.01, 2.33]

2.5 Itraconazole

4

777

Risk Ratio (M‐H, Fixed, 95% CI)

0.70 [0.31, 1.56]

3 Invasive infections Show forest plot

30

4044

Risk Ratio (M‐H, Fixed, 95% CI)

0.50 [0.39, 0.64]

3.1 Amphotericin

8

855

Risk Ratio (M‐H, Fixed, 95% CI)

0.41 [0.24, 0.73]

3.2 Fluconazole

8

1539

Risk Ratio (M‐H, Fixed, 95% CI)

0.39 [0.27, 0.57]

3.3 Ketoconazole

7

562

Risk Ratio (M‐H, Fixed, 95% CI)

1.32 [0.68, 2.54]

3.4 Miconazole

2

238

Risk Ratio (M‐H, Fixed, 95% CI)

0.52 [0.20, 1.31]

3.5 Itraconazole

4

810

Risk Ratio (M‐H, Fixed, 95% CI)

0.53 [0.29, 0.97]

3.6 Itraconazole/ketoconazole/amphotericin

1

40

Risk Ratio (M‐H, Fixed, 95% CI)

0.0 [0.0, 0.0]

4 Colonisation Show forest plot

22

Risk Ratio (M‐H, Random, 95% CI)

Subtotals only

4.1 Amphotericin

3

378

Risk Ratio (M‐H, Random, 95% CI)

0.51 [0.33, 0.77]

4.2 Fluconazole

6

1393

Risk Ratio (M‐H, Random, 95% CI)

0.55 [0.33, 0.90]

4.3 Ketoconazole

8

626

Risk Ratio (M‐H, Random, 95% CI)

0.67 [0.51, 0.87]

4.4 Miconazole

2

238

Risk Ratio (M‐H, Random, 95% CI)

0.92 [0.37, 2.24]

4.5 Itraconazole

2

503

Risk Ratio (M‐H, Random, 95% CI)

0.91 [0.57, 1.45]

4.6 Itraconazole/ketoconazole/amphotericin

1

40

Risk Ratio (M‐H, Random, 95% CI)

0.57 [0.31, 1.04]

5 Use of escape drug Show forest plot

24

Risk Ratio (M‐H, Random, 95% CI)

Subtotals only

5.1 Amphotericin

6

636

Risk Ratio (M‐H, Random, 95% CI)

0.60 [0.35, 1.03]

5.2 Fluconazole

7

1469

Risk Ratio (M‐H, Random, 95% CI)

0.88 [0.76, 1.02]

5.3 Ketoconazole

6

412

Risk Ratio (M‐H, Random, 95% CI)

0.92 [0.58, 1.44]

5.4 Miconazole

2

238

Risk Ratio (M‐H, Random, 95% CI)

1.17 [0.94, 1.46]

5.5 Itraconazole

3

712

Risk Ratio (M‐H, Random, 95% CI)

0.74 [0.57, 0.95]

Figures and Tables -
Comparison 1. Antifungals versus placebo or no treatment