Systematic review and network meta-analysis of clinical outcomes associated with isavuconazole versus relevant comparators for patients with invasive aspergillosis

Abstract

Objectives: Voriconazole, amphotericin B (AmB) formulations, and isavuconazole are all included in guideline recommendations for treatment of patients with invasive aspergillosis (IA) but the relative efficacy of isavuconazole versus AmB formulations has not been directly compared. We aimed to estimate the relative efficacy of isavuconazole compared with AmB deoxycholate (AmB-D), liposomal AmB (L-AmB), and voriconazole for the treatment of patients with proven/probable IA.

Methods: Nine literature databases were screened for randomized controlled trials comparing treatments with any of voriconazole, AmB-D, L-AmB and isavuconazole for treatment of proven/probable IA. Articles meeting the criteria were included in a meta-analysis to determine the efficacy of AmB-D, L-AmB and voriconazole relative to isavuconazole based on all-cause mortality (ACM) and overall response using a fixed-effects model.

Results: Four articles were identified that compared L-AmB with AmB-D (Study 1), standard-dose L-AmB (3–5 mg/kg/day) with high-dose L-AmB (10 mg/kg/day; Study 2), voriconazole with AmB-D (Study 3), and isavuconazole with voriconazole (Study 4). In the network meta-analysis, isavuconazole was statistically superior to AmB-D on both ACM (odds ratio [95% credible intervals] shown as natural log, 1.00 [0.26, 1.74]) and overall response (−1.39 [−2.21, −0.63]). Differences between isavuconazole, and standard-dose L-AmB, high-dose L-AmB and voriconazole were not statistically significant for either ACM (0.18 [−1.17, 1.53], 0.50 [−1.11, 2.13] and 0.32 [−0.19, 0.84], respectively) or overall response (−0.99 [−2.21, 0.29], −0.89 [−2.41, 0.65] and 0.06 [−0.43, 0.57], respectively).

Conclusions: This data suggests that the efficacy of isavuconazole for treatment of IA is superior to AmB-D and comparable with both L-AmB and voriconazole.

Keywords:

• Amphotericin B
• invasive aspergillosis
• isavuconazole
• network meta-analysis
• voriconazole

Introduction

Invasive aspergillosis (IA) is a major cause of morbidity and mortality in immunocompromised patients¹. Although mortality rates in patients with IA have declined within the last two decades with the substitution of amphotericin B deoxycholate (AmB-D) with voriconazole as the standard of care¹, treatment remains sub-optimal for patients due to adverse events and drug–drug interactions with immunosuppressive drugs^2,3. More recently, isavuconazole was shown to have non-inferior efficacy compared with voriconazole as well as significantly fewer study drug-related adverse events and significantly fewer disorders of the hepatobiliary system, eye, and skin and subcutaneous tissues⁴. Guidelines for treatment of IA from the Infectious Diseases Society of America (IDSA) recommend voriconazole as primary therapy (strong recommendation; high-quality evidence), and as alternatives for primary therapy, isavuconazole or liposomal AmB (L-AmB) (3–5 mg/kg) (strong recommendation; moderate-quality evidence for both medications)⁵. The guidelines indicate that AmB-D should only be used in resource-limited settings in which no alternative agents are available, and that triazoles are generally the preferred agents for treatment and prevention of IA in most patients. Newly issued IA treatment recommendations from the 6th European Conference on Infections in Leukemia (ECIL-6) indicate that the strongest evidence supports the use of voriconazole or isavuconazole for first-line treatment of IA in leukemia and hematopoietic stem cell transplant patients and there is weaker evidence to support the use of lipid-based formulations of AmB⁶. Recent joint guidelines from the European Society for Clinical Microbiology and Infectious Diseases, European Confederation of Medical Mycology and the European Respiratory Society strongly support the use of voriconazole or isavuconazole as first-line treatment of pulmonary IA, whereas voriconazole is the only pharmacologic agent with a strong recommendation for extrapulmonary IA; lipid-based formulations of AmB are only moderately supported⁷.

Although comparable efficacy has been demonstrated for voriconazole and isavuconazole in the treatment of IA in a randomized controlled trial (RCT), there are no studies that directly compare the efficacy of isavuconazole with that of AmB formulations in IA. Nevertheless, information regarding the relative efficacy of different therapies can sometimes be obtained indirectly using network meta-analysis, also called mixed treatment comparison meta-analysis^8,9. Network meta-analyses have become an established method to compare outcomes for interventions that have not been tested in head-to-head studies¹⁰. In this approach, the efficacy of an intervention relative to others against which it has not been directly compared is inferred by analyses of RCTs that have comparators in common and that are identified via a systematic literature review. The current analysis used this approach to identify RCTs that assessed clinical outcomes in patients with proven and probable IA treated with isavuconazole, L-AmB, AmB-D or voriconazole, and use data from those studies to assess the relative efficacy of isavuconazole (effects on all-cause mortality [ACM] and overall response) compared with those agents.

Methods

The systematic literature review and network meta-analysis was performed in accordance with standards set out by the National Institute for Health and Care Excellence, the Centre for Reviews and Dissemination (CRD), and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines^11–14.

Systematic review: database searches and study selection

Search strategies were designed and performed for each database screened (Supplementary Material), which included Embase, MEDLINE, CDSR, DARE, NHS EED, HTA, CENTRAL, ICTRP and NGC, and searches were supplemented by hand searching reference lists. The initial search was conducted on 6 January 2015; criteria are summarized in Table 1. Strategies were designed to identify RCTs that had used isavuconazole or any of the comparators as monotherapy for the treatment of probable or proven IA. Inclusion criteria were RCTs with human subjects published in English since 1995, treatment of patients with invasive fungal disease (IFD) caused by Aspergillus spp., and measures of clinical effectiveness of isavuconazole or any of the comparators (mortality, overall response, rate of hospitalization, discontinuation [due to adverse events or any reason], adverse events or serious adverse events). Studies excluded were those not available in English, published before 1 January 1995, not conducted as RCTs, or not including comparators or interventions of interest. Eligible studies were assessed for bias as per guidance provided by the Centre for Reviews and Dissemination¹².

Table 1. Parameters and criteria for the systematic review.

Parameter	Criteria
Population	Adult patients with proven, probable or possible invasive fungal disease caused by Aspergillus species
Intervention	Isavuconazole
Comparators	Voriconazole, liposomal amphotericin B, amphotericin B deoxycholate
Outcomes	Mortality (n, %) Overall response (success; n, %) Rate of hospitalization (including intensive care unit) and length of stay Discontinuation due to adverse event (treatment emergent and drug related) Discontinuation due to any reason Adverse events (treatment emergent and drug related) Serious adverse events (treatment emergent and drug related)
Study types	Randomized controlled trials
Language	English language only
Years	Last 20 years
Exclusions	Combination treatment studies Reviews, discussion papers, non-research letters and editorials Qualitative studies Case studies, case series Animal studies Conference abstracts

Network meta-analysis

Data extracted from RCTs were subjected to network analysis within a Bayesian framework using Markov-chain Monte Carlo methods as per Lu and Ades¹⁵ and Dias et al. (2011¹⁶ and 2013¹⁷). The models were run with OpenBUGS version 3.2.3 (http://www.openbugs.net/w/FrontPage). Where possible, outcome data was obtained from patients with proven or probable IFDs as defined using the 2008 criteria of the European Organisation for Research and Treatment of Cancer (EORTC) and the Mycoses Study Group (MSG)¹⁸. For studies performed and published prior to the publication of the 2008 EORTC/MSG guidelines, IFDs were categorized using the 2002 EORTC/MSG guidelines¹⁹. For studies that had been reanalyzed and republished after the publication of the 2008 guidelines, data was extracted from the parental study when not available from the re-analysis. Because only one study was found for each comparison of treatments, only a fixed-effects model was used. The outcomes assessed and compared were ACM and overall response (complete and partial responses, as defined within each study). The rates of adverse events or discontinuations (overall or due to adverse events) were not assessed because there was a lack of comparable information in the identified trials with which to perform the analysis. To address the potential role of the certainty of diagnosis, sensitivity analyses were performed that included patients previously excluded due to lack of proven or probable IFD. Statistical significance was assessed using 95% posterior credible intervals (CrIs; a Bayesian statistical term analogous to confidence intervals in frequentist statistics). Odds ratios were calculated for the main outcomes using a logistic regression model.

Results

Systematic review

The initial search identified 455 potentially eligible articles; two additional articles published after the initial search was performed were also included. After screening of titles, abstracts, and keywords to exclude duplicates and articles that did not meet inclusion criteria (Table 1), 42 articles were included for full-text screening (Figure 1; Supplementary Material). Of these, eight articles covering four clinical trials were identified as meeting the eligibility criteria (Table 2). For trials with multiple published articles in which different diagnostic criteria had been applied, articles selected for inclusion in the meta-analysis were those that analyzed outcomes in patients assessed as having probable or proven IFD or IA using 2008 criteria set out by the EORTC/MSG¹⁸. Thus, the final data set included one article each from four studies (Figure 2). Study 1 was a clinical trial that compared AmB-D (1 mg/kg) with L-AmB (5 mg/kg) in patients with neutropenia-associated IFD, including most with documented or highly suspected IA (AmB-D, 76.5%; L-AmB, 78.1%)²⁰. Study 2 was a clinical trial that compared standard and high doses of L-AmB (3 mg/kg vs. 10 mg/kg) for treatment of patients with IFD and for which two articles were found^21,22. In the parent study, 97% of patients met the protocol-specific criteria for probable or proven IA. Outcomes for patients with proven and probable IFD using 2008 EORTC/MSG criteria were published in a re-analysis of the original study data²². Study 3 was a clinical trial that compared voriconazole (4 mg/kg BID maintenance dose) with AmB-D (1.0–1.5 mg/kg) in patients with IA and for which three articles were found^23–25. Outcomes for patients with proven and probable IA using 2008 EORTC/MSG criteria were published in a re-analysis of the original study data²⁵. Study 4 was a clinical trial of isavuconazole (200 mg QD maintenance dose) versus voriconazole (4 mg/kg BID maintenance dose) in patients with IFD and for which two articles were found^4,26. In the primary publication, most patients with proven or probable IFD (determined by an independent data review committee) had documented IA or positive serum galactomannan test results (isavuconazole group, 86.7%; voriconazole group, 83.7%)⁴. For the purposes of this analysis, all patients from Study 1 (AmB-D vs. L-AmB) were included, whereas for Studies 2–4, only patients with probable or proven IFD or IA who received at least one dose of study drug (modified intent-to-treat [mITT] populations) were included.

Figure 1. Flow diagram of studies included and excluded at each step. ^aTwo articles published after the performance of the systematic search were also included: Herbrecht et al. 2015²⁵ and Maertens et al. 2016⁴.

Figure 2. Comparisons in studies identified in the systematic review included in the network analysis. Abbreviations. AmB-D, amphotericin B deoxycholate; ISAV, isavuconazole; L-AmB, liposomal amphotericin B; VRC, voriconazole. ^aMaertens et al. 2016⁴. ^bHerbrecht et al. 2015²⁵. ^cLeenders et al. 1998²⁰. ^dCornely et al. 2011²².

Table 2. Studies identified for the analysis.

Study	Study drugs (patient numbers)	Primary end point (populations, time points)	Selected secondary end points	Time points
Study 1: Leenders et al. 1998^20	AmB-D 1 mg/kg (n = 34) L-AmB 5 mg/kg (n = 32)	Therapeutic response (CE, 2 weeks)	Clinical success and improvement Mortality rate	Completion of therapy During therapy and 4 weeks after completion of therapy
Study 2: Cornely et al. 2011 (AmBiLoad)^22	L-AmB 3 mg/kg (n = 45) L-AmB 10 mg/kg (n = 38)	Overall response (ITT, mITT; 12 weeks)	Survival	12 weeks
Study 3: Herbrecht et al. 2015^25	Voriconazole 4 mg/kg IV or 200 mg oral BID^a (n = 124) AmB-D 1.0–1.5 mg/kg^a (n = 113)	Successful outcome (ITT, mITT; 12 weeks)	Successful response (favorable response) Survival	End of initial therapy End of therapy 12 weeks
Study 4: Maertens et al. 2016 (SECURE)^4	Isavuconazole 200 mg QD^a (n = 143) Voriconazole 4 mg/kg or 200 mg oral BID^a (n = 129)	All-cause mortality (ITT, mITT; 42 days)	All-cause mortality Efficacy	84 days, EOT 42 days, 84 days, EOT

Abbreviations. AmB-D, amphotericin B deoxycholate; BID, twice daily; CE, clinical evaluable population; EOT, end of treatment; ITT, intent-to-treat population; IV, intravenous; L-AmB, liposomal amphotericin B; mITT, modified intent-to-treat population; QD, once daily.

^a Maintenance dose.

Baseline and treatment details

The average ages of treatment groups (means or medians) were all within a range of 42–52.5 years (Supplementary Table S1). Study 1 included predominantly females, whereas more than half of patients in Studies 2–4 were males. Data on body weight were only available for Studies 2 and 3, and were similar (Supplementary Table S1).

Available data regarding certainty of diagnosis was variable due to the evolution of diagnostic criteria over time and approaches to analysis of data across Studies 1–4 (Table 3). Data regarding infection sites and underlying diseases was largely similar. In more than half of patients in each study, the most common infection site was the lungs and the most common underlying disease was hematologic malignancy. In Study 1, neutropenia within 14 days of enrollment was a study inclusion criterion, and so almost all patients were neutropenic at baseline. In Studies 2–4, between approximately half and two-thirds of patients in each study group were neutropenic (Table 3).

Table 3. Infection sites and underlying diseases in patients with proven or probable invasive fungal diseases^a.

Study	Treatment arm	Site of infection	Primary underlying disease	Baseline neutropenia
Study 1: Leenders et al. 1998^20	L-AmB	Pulmonary, 56% Not reported, 44%	Acute non-lymphocytic leukemia/myelodysplastic syndromes, 56%; ALL, 19%; chronic leukemia, 6%; other, 19%	94%^b
	AmB-D	Pulmonary, 65% Not reported, 35%	Acute non-lymphocytic leukemia/myelodysplastic syndromes, 59%; ALL, 6%; chronic leukemia, 9%; other, 26%	88%^b
Study 2: Cornely et al. 2011^22	L-AmB 3 mg/kg	Pulmonary only, 80%	Hematologic malignancies, 89%; HSCT, 20%	62%^b
	L-AmB 10 mg/kg	Pulmonary only, 79%	Hematologic malignancies, 92%; HSCT, 16%	63%^b
Study 3: Herbrecht et al. 2015^25	VRC	Pulmonary only, 85% Sinus, 6% Cerebral, 3% Disseminated, 3% Other, 3%^c	AML, 36%; HSCT, 29%; ALL, 8%; other hematologic malignancies, 12%; solid organ cancer, 1%; SOT, 6%; other non-malignant disease, 8%^d	50%^b,^d
	AmB-D	Pulmonary only, 88% Sinus, 5% Cerebral, 4% Disseminated, 1% Other, 2%^c	AML, 38%; HSCT, 26%; ALL, 7%; other hematologic malignancies, 15%; SOT, 4%; other non-malignant disease, 10%^d	49.4%^b,^d
Study 4: Maertens et al. 2016^4	ISAV	Pulmonary only, 81% Pulmonary and other organ, 8% Non-pulmonary only, 10%	AML, 38%; ALL, 12%; lymphoma, 13%; myelodysplastic syndrome, 9%; CLL, 4%; aplastic anemia, 3%; CML, 2%; multiple myeloma, 2%; COPD, 2%; Hodgkin’s disease, 1%; diabetes mellitus, 2%	63.2%^b
	VRC	Pulmonary only, 83% Pulmonary and other organ, 12% Non-pulmonary only, 5%	AML, 49%; ALL, 9%; lymphoma, 9%; myelodysplastic syndrome, 5%; CLL, 5%; aplastic anemia, 3%; CML, 3%; multiple myeloma, 3%; COPD, 1%; Hodgkin’s disease, 1%	67.8%^b

Abbreviations. ALL, acute lymphoblastic leukemia; AmB-D, amphotericin B deoxycholate; AML, acute myeloid leukemia; CLL, chronic lymphocytic leukemia; CML, chronic myeloid leukemia; COPD, chronic obstructive pulmonary disease; HSCT, hematopoietic stem cell transplantation (includes both allogeneic and autologous); IA, invasive aspergillosis; ISAV, isavuconazole; L-AmB, liposomal amphotericin B; SOT, solid organ transplantation; VRC, voriconazole.

^a For Study 1, certainty of diagnosis was categorized as “documented” and “highly suspected”²⁰; Study 3 was an analysis of patients with proven or probable IA.

^b Defined as absolute neutrophil count < 0.5 × 10⁹/L.

^c Percentage of patients with definite or probable IA in the parent study²³.

^d Percentage of patients with proven, probable or possible IA in the re-analysis.

Data regarding the average dose of study drug and concomitant medications were not consistently reported (Supplementary Table S2). For AmB-D and L-AmB treatment arms in Studies 1–3, the median treatment duration ranged from 10–15 days, whereas the treatment duration with triazole antifungal drugs in Studies 3 and 4 (voriconazole and isavuconazole) ranged from 58–87 days.

Risk of bias assessment within individual studies

In the risk of bias assessments for each study, randomization was appropriate for all studies and therefore posed a low risk (Supplementary Table S3). Treatment groups in each study were reported as having similar baseline characteristics, suggesting a low risk of bias due to imbalances. Risks of bias were identified for Study 1 because of the open-label trial design (no blinding or concealment) and primary analysis of the per-protocol population instead of the ITT population. For Study 3, although assessors of outcomes were blinded to treatment allocation, the study was otherwise unblinded. In addition, baseline characteristics were reported for the mITT population rather than the ITT population, so it could not be established unambiguously that there was not an imbalance in study dropouts. There was no evidence that any of the studies had a risk of bias attributable to unreported outcomes, and no other risks of bias were detected.

Efficacy outcomes of individual studies and network meta-analysis

Studies 2–4 included data on ACM at 12 weeks (84 days)^4,22,25, whereas Study 1 included ACM data at the end of study (treatment period plus 4 week follow-up period) and used logistic regression to factor in the influence of malignancy status²⁰. A significant difference between treatment arms was reported in Study 1 (favoring L-AmB over AmB-D) and in Study 3 (favoring voriconazole over AmB-D), whereas the treatment difference was not significant in Study 4 (isavuconazole vs. voriconazole) and significance was not reported in Study 2 (L-AmB 3 mg/kg vs 10 mg/kg) (Figure 3A). In the network meta-analysis, isavuconazole demonstrated a statistically significant advantage over AmB-D, whereas differences with L-AmB (3 mg/kg and 10 mg/kg) and voriconazole did not reach significance (Figure 3B). To determine whether the certainty of diagnosis affected comparisons, sensitivity analyses were performed that also included patients with possible IFD (Studies 2 and 3) or the ITT population (Study 4; all randomized patients who received at least one dose of study medication, irrespective of diagnosis) (Supplementary Table S4). Overall treatment comparisons were similar to those for patients with proven or probable IFD/IA (Supplementary Figure S1).

Figure 3. All-cause mortality in individual studies (A) and odds ratios^a for comparisons with isavuconazole in the network meta-analysis (B). Odds ratio shown as natural log. ^aCalculated using exact logistic regression factoring in malignancy status. Abbreviations. AmB-D, amphotericin B deoxycholate; CrI, credible interval; ISAV, isavuconazole; L-AmB, liposomal amphotericin B; NR, not reported; NS, not significant; VRC, voriconazole.

Table 4. Clinical/pharmacologic aspects of isavuconazole, voriconazole, amphotericin B deoxycholate and liposomal amphotericin B in the treatment of invasive mold disease^a.

Characteristic	Isavuconazole	Voriconazole	Amphotericin B deoxycholate	Liposomal amphotericin B
Indication for IMD	IA, mucormycosis^b	IA, serious IMD due to Scedosporium apiospermum, Fusarium spp.	IA, mucormycosis	IA, mucormycosis
Guideline recommendations for treatment of IA
ECIL-6^6	Grade A I^c: as effective as voriconazole and better tolerated	Grade A I^c	Grade A I^c against use as first-line treatment: less effective and more toxic	Grade B I^d
IDSA^5	Alternative to voriconazole	Primary recommendation	Should be reserved for use in resource-limited settings in which no alternative agents are available	Alternative to voriconazole; use should be considered in settings in which azoles are contraindicated or not tolerated
ESCMID/ECMM/ERS^7	Grade A I^c (pulmonary IA)	Grade A I^c (pulmonary IA); Grade A II^e (extrapulmonary IA)	Grade D I^f	Grade B II^g (pulmonary IA); Grade A II^e (extrapulmonary IA)
Available formulations (guideline recommendations for dosing in IA)	IV or oral (200 mg daily)^h	IV (4 mg/kg twice daily) or oral (200 mg twice daily)^i	IV (1–1.5 mg/kg)^j	IV (3–5 mg/kg)^k
Clinical guideline recommendations for TDM
ECIL-6	No recommendation	Indicated	Not applicable	Not applicable
IDSA	Further studies needed to know whether it is helpful or necessary	Moderate amount of evidence to support value for enhancing efficacy	Not applicable	Not applicable
ESCMID/ECMM/ERS	Potentially useful	Indicated	Not applicable	Not applicable
Drug–drug interactions	•Moderate inhibitor of CYP3A4•Mild inhibitor of P-gp, OCT2•Substrate for CYP3A4/5•Co-administration of strong inhibitors or inducers of CYP3A4 contraindicated	•Strong inhibitor of CYP3A4•Substrate for CYP2C19, CYP2C9, CYP3A4•Co-administration of strong inhibitors or inducers of CYP3A4 contraindicated•Monitoring and potential dose adjustments required with inhibitors of CYP2C19, CYP2C9	•Potential for interactions with antineoplastic agents, corticosteroids and corticotropin, digitalis glycosides, flucytosine, imidazoles, other nephrotoxic medications, skeletal muscle relaxants, leukocyte transfusions	•Potential for interactions with antineoplastic agents, corticosteroids and corticotropin, digitalis glycosides, flucytosine, azoles, other nephrotoxic medications, skeletal muscle relaxants, leukocyte transfusions

Abbreviations. CYP, cytochrome P450; ECIL-6, 6th European Conference on Infections in Leukemia; ECMM, European Confederation of Medical Mycology; ERS, European Respiratory Society; ESCMID, European Society for Clinical Microbiology and Infectious Diseases; IA, invasive aspergillosis; IDSA, Infectious Diseases Society of America; IMD, invasive mold disease; IV, intravenous; OCT2, organic cation transporter 2; P-gp, P-glycoprotein; TDM, therapeutic drug monitoring.

^a From ECIL-6, ESCMID/ECMM/ERS, and IDSA clinical guidelines and prescribing information.

^b Approval of isavuconazole for mucormycosis by the European Medicines Agency stipulates its use in patients for whom amphotericin B is not appropriate.

^c Good evidence to support a recommendation for use; evidence from ≥1 properly randomized, controlled trial.

^d Moderate evidence to support a recommendation for use; evidence from ≥1 properly randomized, controlled trial.

^e Good evidence to support a recommendation for use; evidence from ≥1 well designed non-randomized trial, case–control study, multiple time series or dramatic results of uncontrolled experiments.

^f Good evidence to support a recommendation against use; evidence from ≥1 properly randomized, controlled trial.

^g Moderate evidence to support a recommendation for use; evidence from ≥1 well designed non-randomized trial, case–control study, multiple time series or dramatic results of uncontrolled experiments.

^hMaintenance dose, following loading dose of 200 mg every 8 h for 48 h.

ⁱ Maintenance dose, following loading dose of 6 mg/kg IV every 12 h for 24 h.

^j Dose recommendation not provided in ECIL-6 guidelines.

^k IDSA recommendation, 3–5 mg/kg; ECIL-6, ESCMID/ECMM/ERS recommendation, 3 mg/kg.

For assessments of overall response, Studies 1, 2 and 4 included assessments at end of treatment, whereas detailed data on complete and partial responses for Study 3 was available only at 12 weeks. A significant difference favoring voriconazole over AmB-D was reported in Study 3, whereas the treatment difference between isavuconazole and voriconazole was not significant in Study 4, and significance of differences for combined complete and partial responses was not reported for Studies 1 (L-AmB vs. AmB-D) and 2 (L-AmB 3 mg/kg vs. 10 mg/kg) (Figure 4A). In the network meta-analysis, isavuconazole again demonstrated a statistically significant advantage over AmB-D but not L-AmB (3 mg/kg and 10 mg/kg) or voriconazole (Figure 4B). In sensitivity analyses to assess the potential role of the certainty of diagnosis, patients with possible IFD/IA from Studies 2 and 3 were also included (Supplementary Table S5). Again, isavuconazole was found to have a significant advantage over AmB-D but not L-AmB or voriconazole (Supplementary Figure S2).

Figure 4. Overall response in individual studies (A) and odds ratios for comparisons with isavuconazole in the network meta-analysis (B). Odds ratio shown as natural log. Abbreviations. AmB-D, amphotericin B deoxycholate; CrI, credible interval; ISAV, isavuconazole; L-AmB, liposomal amphotericin B; NR, not reported; NS, not significant; VRC, voriconazole.

An analysis of the risk of bias across studies could not be performed because of the limited number of studies available for comparison.

Discussion

In this study, network meta-analysis of RCTs allowed the clinical outcomes of isavuconazole relative to AmB-D, L-AmB and voriconazole to be assessed. Outcomes of mortality and overall response were statistically significantly better for isavuconazole than AmB-D, whereas differences between isavuconazole and other comparators were not statistically significant. Thus, these results support the comparable efficacy of voriconazole, isavuconazole and L-AmB as first-line treatment for patients with IA.

Given the similar efficacy of these agents for treatment of IA demonstrated in this network meta-analysis, selection of an appropriate therapy may be best guided by consideration of other factors associated with each, including flexibility in the route of administration (intravenous and oral), safety/tolerability and potential for drug–drug interactions (Table 4). The availability of oral formulations for voriconazole and isavuconazole may be an advantage over L-AmB, particularly for patients requiring prolonged treatment courses or outpatient treatment. The median duration of therapy with AmB formulations in Studies 1, 2 and 3 (range, 10.0–16.5 days)^20,22,23 was considerably shorter than those for voriconazole or isavuconazole in Studies 3 and 4 (range, 45–77 days)^4,23, but approximately 80% of patients who received AmB-D in Study 3 switched to other licensed antifungal agents due to intolerance or insufficient response²⁴. Although L-AmB is less nephrotoxic than AmB-D and has fewer infusion-related reactions, its use is associated with potential hepatotoxicity²⁷. Triazoles are recognized as preferred treatment options in both IDSA and ECIL-6 guidelines based in large part on better safety/tolerability^5,6, but differences exist in the safety/tolerability of the different triazoles, including for potential hepatotoxicity. Although clinical experience with isavuconazole is more limited than with voriconazole, the IDSA guidelines recognize the lower rates of photosensitivity, skin disorders, and hepatobiliary and visual disorders with isavuconazole, and the ECIL-6 guidelines cite its better safety profile compared with voriconazole⁶. Isavuconazole also has a lesser propensity than voriconazole for drug–drug interactions mediated by cytochrome P450 3A4²⁸. Voriconazole also displays highly variable, non-linear pharmacokinetics and variability in plasma concentrations associated with polymorphisms in the cytochrome P450 2C19 enzyme, which necessitates frequent therapeutic drug monitoring to maintain plasma concentrations in a safe and efficacious range²⁹.

This analysis has some important limitations. The limited number of eligible studies did not permit an assessment of the risk of bias across studies. The time span over which these studies were performed was likely to have introduced some differences in trial characteristics. For example, the 1998 study comparing AmB-D and L-AmB did not use EORTC/MSG criteria for diagnosis of IFD, and included 10 patients with infections due to Candida spp. or other yeast²⁰, which were excluded from our analysis. Some differences between studies were evident with respect to sites of infection. For example, Studies 2–4 included a greater percentage of patients with pulmonary infections (approximately 80–90%) compared with Study 1 (approximately 55–65%), and that may have affected comparisons. Standards of clinical trial reporting have evolved over the last 25 years and so the baseline data available from past studies does not always permit comparison. In addition, changes in healthcare related to underlying diseases probably introduced differences in concomitant medications and may have had different effects on outcomes of ACM between studies. Nevertheless, comparisons of risk ratios in the current analysis are unlikely to have been unduly influenced by absolute differences in ACM. Finally, recommendations regarding time points for efficacy assessments have also been transformed over the years, and were not identical in all of these studies. Nevertheless, all studies in the analysis were RCTs with sufficient numbers of enrolled patients to permit an estimation of relative efficacy. Additionally, the similarities observed in this analysis for the overall response (a measurement subject to changes over time) and ACM (an objective, unchanging measurement) support the validity of the results.

Conclusion

Notwithstanding any limitations, the results of this analysis support the notion that the comparable efficacy of isavuconazole and voriconazole observed in the SECURE trial might be extended to imply superiority of isavuconazole to AmB-D. The results also suggest that isavuconazole is at least as efficacious as L-AmB, although any comparisons with AmB formulations would need to be confirmed in an RCT. Taken together, these findings tend to support the use of isavuconazole as primary therapy for IA and might help to inform clinicians’ decisions when considering therapeutic options.

Transparency

Declaration of funding

This study was funded by Basilea Pharmaceutica International Ltd. The funder was involved in the study design, data collection and data analysis.

Author contributions: N.P. performed data analysis; all authors were involved in the design of the study, interpretation of the data, drafting the paper and revising it critically for intellectual content. All authors agree to be accountable for all aspects of the work. All authors had full access to all study data and had final responsibility for the decision to submit for publication.

Declaration of financial/other relationships

R.H. has disclosed that he has received honoraria from Astellas, Basilea, Gilead, MSD and Pfizer, and a research grant from Pfizer. D.K. and J.P. have disclosed that they are full-time employees of Basilea Pharmaceutica International Ltd, Basel, Switzerland and both hold stock options with Basilea. C.E. has disclosed that he was a full-time employee of Basilea Pharmaceutica International Ltd at the time the analysis was performed. N.P. has disclosed that he was a full-time employee of PHMR, London, UK, at the time the analysis was performed (PHMR was commissioned by Basilea Pharmaceutica International Ltd to carry out statistical analyses).

CMRO peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Acknowledgements

Medical writing support was provided by Ed Parr PhD CMPP of Envision Scientific Solutions and funded by Basilea Pharmaceutica International Ltd.