AMB indicates conventional amphotericin B; AMBL, liposomal amphotericin B; KTCZ, ketoconazole; FLCZ, fluconazole; ITCZ, itraconazole; VOCZ, voriconazole; POCZ, posaconazole; CASP, caspofungin; and MCFG, micafungin.
The dotted line indicates null effect; diamonds, relative risk (RR); and black whiskers, 95% CI; red whiskers, 95% predicted interval (PrI); AMB, conventional amphotericin B; AMBL, liposomal amphotericin B; KTCZ, ketoconazole; FLCZ, fluconazole; ITCZ, itraconazole; VOCZ, voriconazole; POCZ, posaconazole; CASP, caspofungin; and MCFG, micafungin.
eAppendix. Search Strategy
eFigure 1. PRISMA Flowchart
eFigure 2. Quality Assessment
eFigure 3. Network Geometry of All Outcomes
eFigure 4. Forest Plots of All Outcomes
eFigure 5. League Tables of All Outcomes
eFigure 6. Cluster Rank Combined the SUCRA
eFigure 7. Inconsistency Plot of All Outcomes
eFigure 8. Funnel Plots of All Outcomes
Customize your JAMA Network experience by selecting one or more topics from the list below.
Wang J, Zhou M, Xu J, Zhou R, Chen B, Wan Y. Comparison of Antifungal Prophylaxis Drugs in Patients With Hematological Disease or Undergoing Hematopoietic Stem Cell Transplantation: A Systematic Review and Network Meta-analysis. JAMA Netw Open. 2020;3(10):e2017652. doi:10.1001/jamanetworkopen.2020.17652
What primary antifungal prophylaxis drugs for patients with hematological disease or undergoing hematopoietic stem cell transplantation perform best in randomized clinical trials?
In this systematic review and network meta-analysis of 69 randomized clinical trials that performed comparisons of individual antifungal agents in 14 789 patients, voriconazole was recommended for patients undergoing HSCT and posaconazole was recommended for patients with acute myeloid leukemia or myelodysplastic syndrome.
These findings may help clinicians to make antifungal prophylaxis treatment decisions.
Several antifungal drugs are available for antifungal prophylaxis in patients with hematological disease or who are undergoing hematopoietic stem cell transplantation (HSCT).
To summarize the evidence on the efficacy and adverse effects of antifungal agents using an integrated comparison.
Medline, EMBASE, and the Cochrane Central Register of Controlled Clinical Trials were searched to collect all relevant evidence published in randomized clinical trials that assessed antifungal prophylaxis in patients with hematological disease. Sources were search from inception up to October 2019.
Studies that compared any antifungal agent with a placebo, no antifungal agent, or another antifungal agent among patients with hematological disease or undergoing HSCT were included. Of 39 709 studies identified, 69 met the criteria for inclusion.
Data Extraction and Synthesis
The outcome from each study was estimated using the relative risk (RR) with 95% CIs. The Mantel–Haenszel random-effects model was used. The reliability and validity of the networks were estimated by addressing inconsistencies in the evidence from comparative studies of different treatments. Data were analyzed from December 2019 to February 2020. Reporting followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses for Network Meta-analysis (PRISMA-NMA) guideline.
Main Outcomes and Measures
The primary outcomes were invasive fungal infections (IFIs) and mortality. The secondary outcomes were fungal infections, proven IFIs, invasive candidiasis, invasive aspergillosis, fungi-related death, and withdrawal owing to adverse effects of the drug.
We identified 69 randomized clinical trials that reported comparisons of 12 treatments with at total of 14 789 patients. Posaconazole was the treatment associated with the best probability of success against IFIs (surface under the cumulative ranking curve, 86.7%; mean rank, 2.5). Posaconazole treatment was associated with a significant reduction in IFIs (RR, 0.57; 95% CI, 0.42-0.79) and invasive aspergillosis (RR, 0.36; 95% CI, 0.15-0.85) compared with placebo. Voriconazole was associated with a significant reduction in invasive candidiasis (RR, 0.15; 95% CI, 0.09-0.26) compared with placebo. However, posaconazole was associated with a higher incidence of withdrawal because of the adverse effects of the drug (surface under the cumulative ranking curve, 17.5%; mean rank, 9.2). In subgroup analyses considering efficacy and tolerance, voriconazole might be the best choice for patients undergoing HSCT, especially allogenic HSCT; however, posaconazole was ranked as the best choice for patients with acute myeloid leukemia or myelodysplastic syndrome.
Conclusions and Relevance
These findings suggest that voriconazole may be the best prophylaxis option for patients undergoing HSCT, and posaconazole may be the best prophylaxis option for patients with acute myeloid leukemia or myelodysplastic syndrome.
Invasive fungal infections (IFIs) have emerged as important causes of morbidity and mortality in patients receiving myelosuppressive chemotherapy, immunosuppressive therapy, or hematopoietic stem cell transplantation (HSCT). Because of the difficulty in obtaining a timely diagnosis as well as the high morbidity and mortality associated with IFIs, antifungal prophylaxis remains a high priority in these populations at high risk of IFIs.1 Over the past decade, clinical benefits from antifungal prophylaxis have been demonstrated.2-4 However, there is no clear consensus on antifungal prophylaxis treatment between different centers and groups, particularly in the choice of the antifungal prophylaxis agents. Conventional pairwise meta-analyses based on a direct comparison are relatively limited and difficult to use to investigate antifungal prophylaxis agents. We performed a systematic review and network meta-analysis5 to gain a better understanding of the outcomes associated with and tolerance to current antifungal agents.
This systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses for Network Meta-analysis (PRISMA-NMA) reporting guideline.6 This protocol has been registered at PROSPERO under registration number CRD42020161748.
Medline, EMBASE, and the Cochrane Central Register of Controlled Clinical Trials were searched to collect all published evidence from randomized clinical trials from inception to October 2019 that assessed primary antifungal prophylaxis in patients with hematological disease or undergoing HSCT. The search strategy is detailed in the eAppendix in the Supplement. The reference lists from all included studies and reviews were screened to identify potentially relevant evidence.
All available randomized clinical trials that aimed to compare any antifungal agent with a placebo, no antifungal agent, or another antifungal agent for prophylaxis in patients with hematological disease or undergoing HSCT were included. In this analysis, we assumed that there was no difference between placebo and no antifungal agent.
From each relevant study, the following data were extracted: authors’ names, year of publication, number of patients, age, use and dosage of drugs, categories of disease. Extracted outcomes included (1) incidence of fungal infections (superficial and IFI); (2) incidence of IFIs (possible, probable, and proven IFIs); (3) incidence of proven IFIs (positive histological results on biopsy from deep tissue); (4) incidence of invasive candidiasis; (5) incidence of invasive aspergillosis; (6) fungi-related death; and (7) withdrawal because of adverse effects of the drug.
Two of us (M.Z. and J.-Y.X.) independently participated in the quality assessment, and disagreements were resolved by a third reviewer (B.C.) until consensus was obtained. The quality of the evidence was assessed using the revised tool for risk of bias in randomized trials.7
We compared different agents through network meta-analyses performed under a frequentist framework using a random-effects model. The analysis was performed using the network and mvmeta packages in Stata statistical software version 14.0 (StateCorp).8,9 We estimated the outcome from each study using the relative risk (RR) with 95% CIs. A 95% CI of an RR not covering 1 indicated a statistically significant association. Forest plots and league tables were used to visually present the results of the network meta-analysis. For each outcome, the surface under the cumulative ranking curve (SUCRA) was used to separately rank each agent.10 The larger the SUCRA value, the better the rank. The reliability and validity of the networks were estimated by addressing the inconsistencies and heterogeneity in the evidence from comparative studies of different treatments.11 The overall and loop inconsistencies were evaluated.8,12 Heterogeneity was estimated by the restricted maximum likelihood method. A τ2 value less than 0.1 indicated a very low level of heterogeneity, and a τ2 value from 0.1 to 0.5 indicated a reasonable level; a τ2 value greater than 0.5 was considered to indicate high heterogeneity.13 Additional subgroup analyses were performed restricted to data from different patient populations. Small-study effects were described with a funnel plot. Each funnel plot was tested using the Begg test to assess the small-study effects. A 2-sided P < .05 was considered statistically significant. Data were analyzed from December 2019 to February 2020.
The flowchart of study selection for this network meta-analysis is shown in eFigure 1 in the Supplement. In total, 69 trials with 14 789 patients were included,14-82 including 12 groups: placebo, polyene, conventional amphotericin B, liposomal amphotericin B, miconazole, ketoconazole, fluconazole, itraconazole, voriconazole, posaconazole, caspofungin, and micafungin. The basic characteristics of the included studies are summarized in Table 1. The randomization process and selection of the reported results were not reported clearly in most trials (eFigure 2 in the Supplement).
The network geometry for each outcome is shown in eFigure 3 in the Supplement: fungal infections included 12 groups, 69 studies, and 14 789 patients; IFIs included 12 groups (Figure 1), 64 studies, and 12 943 patients; proven IFIs included 11 groups, 37 studies, and 7179 patients; invasive candidiasis included 12 groups, 45 studies, and 9838 patients; invasive aspergillosis included 12 groups, 40 studies, and 7958 patients; mortality included 12 groups, 69 studies, and 14 789 patients; fungi-related deaths included 12 groups, 45 studies, and 8636 patients; and withdrawal included 11 groups, 39 studies, and 9056 patients. Indirect and mixed-treatment comparisons are shown as forest plots (eFigure 4 in the Supplement).
The SUCRA value and rank of each agent for each outcome are shown in Table 2. Regarding IFIs, posaconazole was the approach with the highest ranking (SUCRA, 86.7%; mean rank, 2.5). The 2 approaches with the next-highest rankings were caspofungin (SUCRA, 84.2%) and micafungin (SUCRA, 76.4%). Posaconazole was associated with a significant reduction in IFIs (RR, 0.57; 95% CI, 0.42-0.79) and invasive aspergillosisus infections (RR, 0.36; 95% CI, 0.15-0.85) compared with placebo (Figure 2). Regarding mortality, the treatment ranked highest was micafungin (SUCRA, 90.0%; mean rank, 2.1). Voriconazole ranked second (SUCRA, 73.8%), and posaconazole ranked third (SUCRA, 68.5%).
Caspofungin (SUCRA, 84.9%) treatment ranked the highest for reducing fungal infections. Posaconazole ranked highest in preventing invasive aspergillosis (SUCRA, 87.8%). Caspofungin was ranked highest for preventing invasive candidiasis (SUCRA, 88.5%), and liposomal amphotericin B ranked the highest for reducing fungi-related deaths (SUCRA, 78.8%). Voriconazole was associated with a significant reduction in invasive candidiasis (RR, 0.15; 95% CI, 0.09-0.26) compared with placebo (eFigure 5 in the Supplement). Voriconazole was ranked highest for having the lowest incidence of withdrawal (SUCRA, 78.1%). Posaconazole was associated with a higher incidence of withdrawal because of the adverse effects of the drug (SUCRA, 17.5%; mean rank, 9.2) (eFigure 6 in the Supplement).
Because extensive categories of patients were included, we evaluated whether the prophylactic outcomes and tolerance of agents varied in different patient populations (patients with acute myeloid leukemia [AML] or myelodysplastic syndrome [MDS] or undergoing HSCT or allo-HSCT). Considering efficacy and tolerance, voriconazole was ranked as the best choice for patients undergoing HSCT; this result was also found in the allo-HSCT population. However, posaconazole was ranked as the best choice for patients with AML or MDS.
Heterogeneity and inconsistency are shown in Table 3. Heterogeneity was low for IFIs and mortality. In contrast, heterogeneity was reasonable for the other outcomes (τ2 values from 0.1 to 0.4). Loop inconsistency for placebo, amphotericin B, and fluconazole was found for invasive candidiasis (indirect effect estimate, 2.53; 95% CI, 1.07-3.98; P = .001) (eFigure 7 in the Supplement). We used a funnel plot to visually demonstrate small-study effects (eFigure 8 in the Supplement).
In this systematic review and network meta-analysis, we combined direct and indirect evidence to compare antifungal prophylaxis options for patients with hematological disease or undergoing HSCT. Our analysis may provide some important information for clinical decision-making for antifungal prophylaxis in these patients. We derived 2 principal findings from our analysis: voriconazole may be the best choice for patients undergoing HSCT, and posaconazole may be the best prophylactic option for patients with AML or MDS. Posaconazole is recommended for IFIs during remission induction chemotherapy for AML and MDS, according to the Guidelines from the Infectious Diseases Working Party of the German Society for Haematology and Medical Oncology.83 Overall, posaconazole and voriconazole are recommended as the most reasonable options for the prevention of IFIs. The difference between agents may be meaningful and is not available from single trials, to our knowledge. For instance, voriconazole has not been directly compared with other drugs except for placebo, fluconazole, and itraconazole; however, this network meta-analysis compared voriconazole, as well as posaconazole, with other drugs indirectly.
Posaconazole is recommended for the prevention of IFIs regardless of tolerance. The most commonly reported treatment-related adverse effects of oral posaconazole were digestive tract symptoms, including nausea, vomiting, and gastrointestinal upset.84 The most common cause for discontinuation was severe nausea or gastrointestinal upset.85 We noticed that the incidence of withdrawal was different in patients with AML or MDS and patients undergoing HSCT. We assumed that in patients undergoing HSCT, especially allo-HSCT, a high-dose pretreatment scheme, the use of cyclosporin and the incidence of gastrointestinal acute graft-vs-host disease (GVHD) would decrease the tolerance of posaconazole. The rate of adverse events that led to the discontinuation of posaconazole was 40% in the study by Chatter et al.22 In the posaconazole group, the rate of diarrhea was 67%, of nausea was 67%, and of vomiting was 29%. The rate of gastrointestinal adverse events was similar between the liposomal amphotericin B and posaconazole groups. A new route of administration through injection may be an option for patients who are unable to swallow or are intolerant of oral posaconazole.
Caspofungin is recommended to prevent invasive candidiasis, and the same results were confirmed in patients with AML or MDS. There were no relevant data from patients undergoing HSCT. An echinocandin drug is recommended as the initial therapy for candidemia, according to the clinical practice guidelines for the management of candidiasis from the Infectious Diseases Society of America.86 In our analysis, posaconazole was ranked the best choice for preventing invasive Aspergillus infections, followed by caspofungin. According to the guidelines for the diagnosis and management of aspergillosis from the Infectious Diseases Society of America, posaconazole, voriconazole, and micafungin are recommended for invasive aspergillosis prevention.87 However, in our analysis, it appeared that caspofungin treatment was associated with a better outcome than micafungin and voriconazole in preventing aspergillosis. Concerning the prevention of fungi-related death, treatment with liposomal amphotericin B might be associated with a better outcome, followed by posaconazole and voriconazole. There was no significant difference in fungi-related death with posaconazole.
Some previous studies have summarized the data from the literature on antifungal prophylaxis in hematological disease.88-90 These network meta-analyses did not take into account antifungal prophylaxis in other high-risk groups, such as patients with GVHD.91 With regard to GVHD, the risk of IFIs appears particularly prominent in patients with high-grade acute GVHD or steroid-dependent chronic GVHD.91 Our network meta-analysis has taken into account antifungal prophylaxis in patients with GVHD. A high number of patients with solid tumor without HSCT therapy were included in previous network meta-analyses. In the study by Ninane et al,92 solid tumors were present in more than 20% of patients. We did not consider this study appropriate for a network meta-analysis for antifungal prophylaxis, as routine antifungal prophylaxis is not recommended in patients with solid tumors.91 The evidence shown by Leonart et al88 focused on double-blind trials, and the study by Zhao et al89 focused on triazole agents. Therefore, the conclusions of the comparisons in those reports cannot be compared with the results in this meta-analysis.
Inconsistency refers to the differences between direct and various indirect effect estimates for the same comparison. Stata tests for inconsistency have 2 levels12: overall inconsistency, in which the level of inconsistency is computed according to the type of between-treatment comparison for all cases and a local approach, in which each treatment is individually examined. Local inconsistency because of loop inconsistency in placebo, amphotericin B, and fluconazole was found for invasive candidiasis in our analysis. There are 4 causes of inconsistency: chance, bias in head-to-head comparisons, bias in indirect comparisons, and genuine diversity.93 According to Higgins et al11 loop inconsistency refers to a difference between direct and indirect comparisons. The study by Behre et al16 may be the source of the loop inconsistency because aerosol amphotericin B inhalation treatment was used. When this study was excluded from the analysis, the difference between direct and indirect comparisons of the treatments for invasive candidiasis was not significant.
The interpretations of the results were based on SUCRA values and ranking in our study. Although rankings are appealing, they may be incorrectly emphasize particular treatments as being clinically useful. The uncertainty present in the rankings may be neglected in considering the best treatment, and the rankings may give a false sense that some interventions are superior to others. A PRISMA-NMA statement has suggested that more attention should be paid to the relative effect estimates, rather than the rankings, because a good rank does not necessarily translate to a clinically relevant effect.6 Although the usefulness of rankings is currently debated, rankings will probably continue to be reported. Reporting all probabilities for each intervention with each possible rank is one way to convey the uncertainty in the rank ordering.94 The treatment effects and rankings also depend on the number of treatments and trials in the network.95
Our study has some limitations. First, an improved understanding of antifungal pharmacology, pharmacokinetics, and pharmacodynamics has resulted in therapeutic drug monitoring becoming a valuable adjunct to the administration of some antifungal agents. We could not perform an analysis of therapeutic drug monitoring to evaluate the efficacy and adverse effects of antifungal prophylaxis or justify why therapeutic drug monitoring should not be performed with antifungal prophylaxis if it is strongly recommended with antifungal treatment because of the limited data. Second, the follow-up time of most studies was too short to determine the survival benefits from antifungal prophylaxis. Third, a limited number of head-to-head trials have investigated posaconazole and voriconazole. Fourth, the characteristics of patients and treatments were heterogeneous among the various randomized clinical trials. Although our subgroup analyses found different results among different patient populations, the data did not allow us to perform more detailed analyses, such as those for different ages and races/ethnicities. We also could not perform a more stratified analysis taking into consideration the dosage form and dose of agents. Therefore, the evidence derived from this meta-analysis should be used with caution for shared decision-making. However, our study provides important data from which future practice-changing prospective trials can be designed.
This network meta-analysis assessed the performance of various antifungal prophylaxis treatment in patients with hematological disease or undergoing HSCT. Our findings suggest that, in terms of the prevention of IFIs and tolerance, voriconazole may be the best prophylactic option for patients undergoing HSCT, and posaconazole may be the best prophylactic option for patients with AML or MDS.
Accepted for Publication: July 3, 2020.
Published: October 8, 2020. doi:10.1001/jamanetworkopen.2020.17652
Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2020 Wang J et al. JAMA Network Open.
Corresponding Authors: Bing Chen, MD, PhD, Department of Hematology, Affiliated Drum Tower Hospital of Nanjing University Medical School, 321 Zhongshan Rd, Nanjing, 210008, China (email@example.com); Yuan Wan, PhD, The Pq Laboratory of Micro/Nano BiomeDx, Department of Biomedical Engineering, Binghamton University – SUNY, Binghamton, Biotechnology Building, 65 Murray Hill Rd, Vestal, NY 13850 (firstname.lastname@example.org).
Author Contributions: Drs Wang and Wan had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Chen, Wan.
Acquisition, analysis, or interpretation of data: Wang, M. Zhou, Xu, R.-F. Zhou, Wan.
Drafting of the manuscript: M. Zhou, Wan.
Critical revision of the manuscript for important intellectual content: Wang, Xu, R.-F. Zhou, Chen, Wan.
Statistical analysis: Wang, M. Zhou, Wan.
Obtained funding: M. Zhou, Xu, Wan.
Administrative, technical, or material support: Wang, Xu, R.-F. Zhou, Wan.
Supervision: Chen, Wan.
Conflict of Interest Disclosures: None reported.
Funding/Support: This work was partially supported by Binghamton University Faculty Start-up Fund (grant No. 910252-35), Binghamton University S3IP Award (No. ADLG195), and Nanjing Medical Science and Technique Development Foundation (grant Nos. YKK15068 and YKK17074).
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.