EFFICACY AND SAFETY OF TRIAZOLES VERSUS ECHINOCANDINS IN THE TREATMENT OF INVASIVE ASPERGILLOSIS: META-ANALYSIS

Backgroun/Aim. Although majority of guidelines recommend triazoles (voriconazole, posaconazole, itraconazole and isavuconazole) as first-line therapeutic option for treatment of invasive aspergillosis, echinocandins (caspofungin, micafungin and anidulafungin) are also used for this purpose. However, head-to-head comparison of triazoles and echinocandins for invasive aspergillosis was rarely target of clinical trials. The aim of this meta-analysis was to compare efficacy and safety of triazoles and echinocandins when used for treatment of patients with invasive aspergillosis. Methods. This meta-analysis was based on systematic search of literature and selection of high-quality evidence according to pre-set inclusion and exclusion criteria. The literature search was made for comparison of treatment with any of triazoles (isavuconazole, itraconazole, posaconazole or voriconazole) versus any of echinocandins (caspofungin, anidulafungin or micafungin). The effects of triazoles (itraconazole, posaconazole or voriconazole) and echinocandins (caspofungin, anidulafungin or micafungin) were summarized using RevMan 5.3.5 software, and heterogeneity assessed by the Cochrane Q test and I? values. Several types of bias were assessed, and publication bias was shown by the funnel plot and Egger?s regression. Results. Two clinical trials and three cohort studies were included in this meta-analysis. Mortality in patients with invasive aspergillosis who were treated with triazoles was significantly lower than in patients treated with echinocandins [odds ratio 0.29 (0.13, 0.67)], and rate of favorable response (overall treatment success) 12 weeks after the therapy onset was higher in patients treated with triazoles [3.05 (1.52, 6.13)]. On the other hand, incidence of adverse events was higher with triazoles than with echinocandins in patients treated for invasive aspergillosis [3.75 (0.89, 15.76)], although this difference was not statistically significant. Conclusion.Triazoles (voriconazole in the first place) could be considered as more effective and somewhat less safe therapeutic option than echinocandins for invasive aspergillosis: However, due to poor quality of studies included in this meta-analysis, definite conclusion should await results of additional, well designed clinical trials.


Introduction
Invasive aspergillosis (IA) is the most frequent invasive mold infection caused by fungi belonging to the genus Aspergillus.It is a potentially life-threatening infection (usually taking place in respiratory tract) with high mortality rate (80-90%) in high risk patients such as patients with hematological malignancies and patient undergoing hematopoietic stem cell transplant (HSCT) [1] .Without adequate therapy, invasive pulmonary aspergillosis is further complicated as a result of hematogenous dissemination or direct extension leading to infection of other tissues as well, CNS or cardiovascular system [2] .Invasive aspergillosis is the most common type of infection among stem cell transplant recipients, and the second most common type of fungal infection in organ transplant recipients.One-year survival in patients with invasive aspergillosis was 59% in organ transplant recipients [3] and 25% among recipients of stem cells [4] .A major barrier to successful treatment of invasive aspergillosis is delayed diagnosis.Due to the lack of reliable and feasible diagnostic techniques, over one third of Aspergillus infections still remain undiagnosed [5] .Members of the European Organization for Research in Treatment of Cancer/Invasive Fungal Infection Cooperative Group and National Institute of Allergy and Infectious Diseases Mycoses Study Group (EORTC/MSG) formed a Consensus Committee in order to develop standard definitions for invasive fungal infections for clinical research [6]   .According to them, three levels of certainty of invasive aspergillosis were defined: proven, probable, and possible.From the year 2008 the same study group recommended detection of serum biomarker galactomannan as one of the criteria of probable invasive aspergillosis, which is very much helpful when diagnosing this infection in neutropenic patients before characteristic chest x-ray signs of aspergillosis become visible.
There are only four major classes of antifungal agents (polyenes, flucytosine, azoles and echinocandins) which could be used for systemic treatment of invasive mycoses.Primarily amphotericin B and flucitosine were exploited, but due to their high toxicity, triazoles as efficient and safer drugs were later on usually recommended as the first-line choice in medical literature; however this recommendation was not based on comparative studies between triazoles and echinocandins [7,8] .Triazoles are isomeric chemical compounds containing a five-membered ring with two carbon atoms and three nitrogen atoms.These drugs (posaconazole, isavuconazole, itraconazole and voriconazole) primarily inhibit synthesis of ergosterol by inhibition of lanosterol 14α-demethylase enzymes in the fungal membrane, but not in host cells.Currently, they are successfully used in clinical management of invasive mycoses, including prophylaxis, pre-emptive, empiric and targeted therapy.On the other hand, echinocandins, which were developedin the early 2000s, are also frequently used in the treatment of invasive mycoses (including aspergillosis)due to their low host toxicity and good efficacy, especially as salvage therapy for invasive aspergillosis [9] .The three echinocandins, caspofungin, micafungin and anidulafungin, were the first antifungals that were created to selectively target the fungal cell wall.Echinocandinscause disruption in the b-(1,3)-D-glucan synthesis andincrease permeability of cell wall that leads to adisbalance of the intracellular osmotic pressure of the fungal cell and the fungal cell lysis [10] .
Although clinical trials comparing triazoles and echinocandins for curing invasive apergillosis were published, neither meta-analysis nor systematic review were performed on this topic up to date.Summarizing the available evidence about efficacy and safety of triazoles vs. echinocandins in this indication will be helpful for planning future clinical trials or observational studies with this drugs for invasive aspergillosis.The aim of this meta-analysis was to compare efficacy and safety of triazoles and echinocandins when used for treatment of patients with invasive aspergillosis.

Methods
Our study was registered at PROSPERO register of systematic reviews and metaanalyses under the number CRD42017081282 prior to commencement of the research.
The following criteria to include studies for this review were used: (1) types of studies -both randomized, controlled clinical trials and observational studies which compare any of triazoles with any of echinocandins in patients with IA; (2) types of participants -patients of both sex and any age with proven or probable invasive Aspergillosis (proven IA is characterised by documented histopathological and microbiological evidence of Aspergillus spp.infection, either at autopsy or in biopsied tissue or culture samples from a normally sterile site; probable IA is characterised by the presence of radiological [nodules, cavities, halos or air crescent signs on chestradiography or computed tomography (CT)] and microbiological (direct microscopy, culture) features in an immune-suppressed patient [absolute neutrophil count (ANC) < 500 cells/mm3, prolonged steroid therapy, use of a T-cell suppressor or allogeneic stem transplantation]; (3) types of interventions -intravenous treatment with any of triazoles (isavuconazole,itraconazole, posaconazole or voriconazole) versus any of echinocandins (caspofungin, anidulafungin or micafungin) for at least 7 days.
Search methods for identification of studies primarily included electronic databases, and collection of journal articles and books of University Library, University of Kragujevac, Kragujevac, Serbia.).There were no restrictions on publication date, format or language in the search strategy.The references of the retrieved articles were searched for further similar studies ("snowball search").The collection of journal articles and books of University Library, University of Kragujevac washand searched for relevant studies by one author (RZZ).

Data collection and analysis
The data collection sheet was created and the articles included in the review were assessed for: (1)   22) partial response at end of treatment for each treatment group; (23) favorable response (overall treatment success) at 12 weeks after the start of treatment; (24) failure to respond at end of treatment; (25) failure at end of treatment; and (26) stable disease at end of treatment.Values provided as percentages were converted into actual patient numbers for analysis, as well as standard errors into standard deviations using number of patients, when reported as such.

Selection of studies
Based on the searching strategy, all titles and abstracts retrieved were independently scanned by four authors (SU, RZZ, MR and SJ).Eligibility of the retrieved articles was assessed at first from the title and the abstract, and if it was not possible, the full text of the articles was retrieved and searched.An article was included for review if all authors (SU, RZZ, MR and SJ) agreed that eligibility criteria had been met.In case that the reviewers had different opinions about eligibility of a study for inclusion, the matter was resolved by the corresponding author (SJ).

Data extraction and management
The data were extracted from eligible studies using the data collection sheet described previously (under the "data collection and analysis" subheading).The data collection sheet was made in electronic form, using an Excel 2007 worksheet.The data were extracted by threeinvestigators independently (SU, RZZ and MR) and then collating of the four tables was done by another investigator (SJ), who produced the final extraction table.Meta-analysis was made for the following head-to-head comparisons found in the literature: itraconazole, posaconazole or voriconazole versus caspofungin, anidulafungin or micafungin.

Assessment of risk of bias in included studies
Risk of bias was assessed by two investigators independently (RZZ and MR), and collating the assessments was done by the corresponding investigator (SJ).The following sources of bias were assessed: (1) randomization if any; (2) sequence generation; (3) allocation sequence concealment; (4) blinding; (5) performance bias; (6) detection bias; (7)  attrition bias; and (8) reporting bias.Although some of the studies had high risk of bias, none was excluded from further analysis due to small number of eligible studies (only five).

Measures of treatment effect
All of the outcomes used in the studies were dichotomous: mortality for each treatment group; complete response at end of treatment for each treatment group; partial response at end of treatment for each treatment group; favorable response (overall treatment success) at 12 weeks after the start of treatment; failure to respond at end of treatment; failure at end of treatment; stable disease and adverse events frequency.For these outcomes the treatment effect was measured by risk ratio.

Unit of analysis issues
Unit of analysis in the clinical trials or cohort studies that were included in this meta-analysis were individual patients.Individual participants wereeither randomized or simply allocated to one of two parallel intervention groups, and a single measurement for each outcome from each participant was collected and analyzed.

Dealing with missing data
Missing data were requested directly from the original investigators, however they did not respond to our requests except with courtesy.The missing data were then searched foramong the results presented on ClinicalTrials.gov,when available.Finally, the potential impact of missing data on the findings of the meta-analysis will be addressed in the Discussion section.

Assessment of heterogeneity
Between-study heterogeneity was assessed with the Cochrane Q test using a chisquared function (p values < 0.10 were considered significant).I² values were calculated to quantify inconsistency across studies.I² values of 30% or less may represent low heterogeneity, values from 30 to 50% may represent moderate heterogeneity, values from 50 to 90% substantial heterogeneity and values of 90% or more may represent considerably heterogeneity.An I² value > 30% was considered significant in this meta-analysis.

Assessment of reporting biases
The possibility of within-study selective outcome reporting was examined for each study included in this meta-analysis.First, by constructing matrix of the outcomes for all studies, we identified studies and specific outcomes that were not reported.Then we searched for published protocols of such studies at ClinicalTrials.gov and other forms of publications of the same studies, in order to find the missing outcomes.Finally, the authors were contacted with a request to provide the missing data, but they did not send us the data.The possibility of between-study publication bias was examined by construction of funnel plotsfor continuous outcomes and by Egger's regression for discrete outcomes [11] .Klein's number was also calculated for all outcomes [12] .

Data synthesis
The random effects model (which includes both within-study and between-study variations in calculation of the weighted average) was used to combine the results from the studies.The Mantel-Haenszel method (fixed effect model) was also used to estimate how our conclusions could be influenced by assumptions about the model and by the study heterogeneity.Since significant heterogeneity of the studies was not found, subgroup analysis was not performed.All calculations were done by Review Manager (RevMan) software version 5.3.5 [13].

Sensitivity analysis
Sensitivity analysis was performed by excluding individual trials one at a time and recalculating the pooled odds ratio and mean difference estimates for the remaining studies.In this way we got insight how each of the included studies influenced our conclusions.

Results
Results of the literature search are shown in the Figure 1.Only fivestudies fulfilled all inclusion and missed all exclusion criteria which were set prior the study commencement (two of the trials were published in the same publication, Raad et al [14] , and one trial was published in two publications [15,16] ).Characteristics of the included studies with risk of bias are shown in detail in the Table 1 [14,15,16,17,18] .Summaries of differences in effects of triazoles vs. echinocandins for the main outcomes (using random effects model) were as following: triazoles were associated with lower mortality (odds ratio 0.29), higher complete and partial response rate at end of treatment(odds ratios 2.38 and 2.83, respectively), more favorable response (overall treatment success) at 12 weeks after the start of treatment (odds ratio 3.05), less failure to respond at end of treatment (odds ratio 0.38) and more stable disease at end of treatment (odds ratio 1.16), but treatment discontinuation due to side effects and incidence of adverse events were higher with triazoles than with echinocandins (odds ratios 3.89 and 3.75, respectively).Details of the summaries of differences in effects are shown in the Table 2, expressed as relative risk.Sensitivity analysis did not show significant changes with exclusion of single trials.Summaries of differences in effects of triazoles andechinocandins for the most important outcomes (mortality, complete response at end of treatment and incidence of adverse effects) with heterogeneity estimates are shown by the Forest plots (Figures 2, 3 and 4).
The reporting bias was assessed by Klein's number, Egger's regression and Funnel Plot, using "trim and fill" method for mortality as outcome.The central symmetry axis of Funnel Plotfor mortality rate did not change place significantly after "trim and fill" exercise.In the Figure 5 Funnel Plots are shown before and after "trim and fill" exercise for mortality outcome.Klein's number for mortality rate was 9.63, however the Egger's regression showed significant correction of the summary effect estimate: from OR = 0.29 to OR = 0.001 (Figure 6).High: Observational design.

Risk of allocation concealment bias
Low: Randomized study High: Observational design.

Risk of blinding of patients and personnel bias
High: There was no blinding

High: There was no blinding
High: There was no blinding

High: There was no blinding
High: There was no blinding

Risk of blinding of outcome assessment bias
High

Risk of other bias
High: Efficacy outcomes were not reported for entire intention-totreat population.
High: Efficacy outcomes were not reported for entire intention-totreat population.
Low: Efficacy outcomes were reported for entire intention-totreat population.
Low: Efficacy outcomes were reported for entire intention-totreat population.
Low: Efficacy outcomes were reported for entire intention-totreat population.
* external control group -since a control treatment could not have been compared with posaconazole in the same study, the patients from participating study sites who were treated by the control drugs, but not enrolled in the study, were used as controls if fulfilling pre-specified criteria.The control patients were matched with patients receiving posaconazole for important prognostic factors to allow for fair comparison between the treatments.Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.

Very low quality:
We are very uncertain about the estimate.

Discussion
Our study showed that mortality in patients with invasive aspergillosis who were treated with triazoles was significantly lower than in patients treated with echinocandins.However, among the other efficacy outcomes, only rate of favorable response (overall treatment success) 12 weeks after the therapy onset was significantly different between the patients treated with triazoles and echinocandins, triazoles being favored.Other efficacy outcomes invariably were more beneficial in triazoles group, but significance could not be reached because not all included studies recorded every outcome, and certain of them (e.g.failure at end of treatment or stable disease at end of treatment) were mentioned in only one or two studies.On the other hand, incidence of adverse events was higher in the groups of patients receiving triazoles.Systematic reviews and meta-analyses of clinical studies including patients with invasive aspergillosis are rare, and mostly focused on comparison of combination therapy (triazoles or amphotericin B plus an echinocandin) with non-echinocandin-based monotherapy (i.e.triazoles) after first-line antifungals were ineffective (salvage therapy) [19,20] .Although the authors of these studies at first concluded that combination therapy had increased efficacy, later on they questioned their own conclusion and limited it to situations where antifungal drug resistance is suspected or adequate blood levels could not be achieved [19] .Good efficacy of triazoles (mostly voriconazole) was observed in these studies, as well as relatively high rate of their adverse reactions, but triazoles and echinocandins were not compared head-to-head as monotherapy.Our meta-analysis confirmed good efficacy of triazoles against invasive aspergillosis and relatively high adverse events rate in both first-line and salvage settings, when used as monotherapy and compared with echinocandins.Voriconazole and posaconazole penetrate to tissues to high extent (especially to lungs, voriconazole6.26 µg/g and posaconazole87.7 µg/mL), while among the echinocandins only anidulafungin has comparable penetration (17.9 µg/g of lung tissue); however, in the studies included in our meta-analysis only caspofungin and micafungin were used, which could additionally explain superior efficacy of triazoles [21] .Resistance of aspergillus is less frequent to triazoles (from no resistance of isolated Aspergillus spp. to posaconazole and voriconazole to 17% resistance of isolated Aspergillus fumigatus to voriconazole) [22,23] than to echinocandins (22% resistance of Aspergillus fumigates to caspofungin) [23] , making the first more reliable therapeutic option especially for second-line treatment of invasive aspergillosis.
Increased incidence of adverse events in patients with invasive aspergillosis treated by triazoles in comparison to those treated by echnocandins that was found in our study related mostly to increased incidence of hepatic adverse effects [24] .Although both triazoles and echinocandins may cause either hepatocellular orcholestatic liver injury, frequency is higher with voriconazole,itraconazole, posaconazole or isavuconazole than with caspofungin, anidulafungin or micafungin (up to 24% vs. up to 9%, respectively).However, majority of patients experience only laboratory abnormalities, i.e. elevation of serum levels of aspartate aminotransferase, alanine aminotransferase and bilirubin, and serious liver injuries are rare with both drug groups [24] .Our study confirmed these findings, as none of the patients exposed to either triazoles or echinocandins in included studies had fulminant hepatitis or acute liver failure, yet adverse events rate was significantly higher in the groups exposed to triazoles.Additionally, all triazoles interact with cytochrome P450, especially with CYP3A4 and CYP3A5, while voriconazole interacts also with CYP2C19 [25] , and their potential to inhibit elimination of other drugs metabolized through the same enzymes is much higher than that of echinocandins [26] .Echinocandins are not metabolized through cytochromes (except micafungin in minor extent) and therefore do not influence elimination of other drugs that are oxidized at these enzymes in liver [21] .
Our results should be taken conditionally, since some of the important clinical outcomes were reported in only one of the included studies (e.g.failure at end of treatment or stable disease at end of treatment), and overall number of the included studies was low, even after widening of inclusion criteria to encompass cohort studies, which are less reliable than clinical trials,due to inherent limitations of observational design.Since clinical trials with triazoles in patients with invasive aspergillosis are likely to be initiated in close future, new meta-analysis should be made to challenge our results.

Conclusion
In conclusion, on the basis of published clinical trials and cohort studiestriazoles (voriconazole in the first place) could be considered as more effective and somewhat less safe therapeutic option than echinocandins for invasive aspergillosis for the time being.Future studies which would include new clinical trials are necessary to confirm this conclusion.Accepted on October 11 Online First October, 2018.

Figure 1 .
Figure 1.Selection of studies included in meta-analysis.

Figure 2 .
Figure 2. Summary of differences in mortality rate of patients with invasive aspergillosis treated by triazoles or echinocandins.

Figure 3 .
Figure 3. Summary of differences in complete response rate at end of treatment of patients with invasive aspergillosis treated by triazoles or echinocandins.

Figure 4 .
Figure 4. Summary of differences in adverse effects rate observed in patients with invasive aspergillosis treated by triazoles or echinocandins.

Figure 5 .
Figure 5. Funnel Plots before and after "trim and fill" exercise for mortality rate.

Figure 6 .
Figure 6.Egger's regression for mortality in the included studies.

Figure 1 .
Figure 1.Selection of studies included in meta-analysis.

Figure 2 .
Figure 2. Summary of differences in mortality rate of patients with invasive aspergillosis treated by triazoles or echinocandins.

Figure 3 .
Figure 3. Summary of differences in complete response rate at end of treatment of patients with invasive aspergillosis treated by triazoles or echinocandins.

Figure 4 .
Figure 4. Summary of differences in adverse effects rate observed in patients with invasive aspergillosis treated by triazoles or echinocandins.

Figure 5 .
Figure 5. Funnel Plots before and after "trim and fill" exercise for mortality rate.

Figure 6 .
Figure 6.Egger's regression for mortality in the included studies.

Table 1 .
Characteristics of the studies included in meta-analysis.

Table 2 .
Summary of findings of studies included in the meta-analysis/

Triazoles (itraconazole, posaconazole and voriconazole) compared with echnocandins (caspofungin, anidulafungin or micafungin) for treatment of invasive aspergillosis Patient or population: both sex and any age withproven or probable invasive Aspergillosis Settings: hospitalized patients. Intervention: Triazoles (itraconazole, posaconazole and voriconazole) Comparison: Echnocandins (caspofungin, anidulafungin or micafungin) Outcomes Illustrative comparative risks* (95% CI)
The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes.The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).Risk Ratio; [other abbreviations, eg.OR, etc] GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect.