Micafungin and caspofungin pharmacodynamics in patients with candidemia

Candidemia is an important cause of health care-associated bloodstream infections. Appropriate antifungal therapy is crucial to reduce morbidity and mortality. A previous randomized clinical trial compared the clinical effectiveness of caspofungin 50 mg, micafungin 100 mg (M100), and micafungin 150 mg (M150) in treating invasive candidiasis and candidemia. M100 treatment success was noninferior to caspofungin, while M150 yielded less favourable clinical and mycological responses. Pharmacodynamic (PD) variables were explored to assess the effect of caspofungin, M100 and M150 on the following outcomes: mycological eradication, clinical success, days to eradication, recurrence, and emergent infection for cases of candidemia only from the aforementioned clinical trial. Univariate and multivariate analyses were performed to evaluate the ratio of area under the concentration curve to minimum inhibitory concentration [AUC:MIC] for the treatment arms. AUC:MIC >1500 conferred better mycological eradication, but an inverse relationship between micafungin and mycological eradication based on an AUC:MIC ≥7500 was found. M150 required more days to eradication than M100 (p<0.03) and caspofungin (p<0.02). A correlation was also noted for days to eradication with several other risk factors, including race, Candida parapsilosis, and neutropenia.


INTRODUCTION
Candidemia is the most common invasive fungal infection encountered in the hospital setting (Pfaller and Diekema, 2007).Candidemia previously ranked fourth among causes of nosocomial bloodstream infections but is the third most common cause of such infections in the intensive care unit (Bow et al., 2010;Kuhn et al., 2002;Ostrosky-Zeichner et al., 2007;Wisplinghoff et al., 2006).However, a recent point prevalence study of nosocomial infections in 2015 in the United States underscored its importance, as it was the second most common cause of nosocomial bloodstream infection (Magill et al., 2018).Its incidence is rising and is associated with increased costs for patient management through increased length of stay, as well as with significant morbidity and mortality (Pfaller and Diekema, 2007;Andes et al., 2011;Fridkin, 2005;Nucci et al., 2010;Pappas et al., 2003;Wisplinghoff et al., 2004;Morgan et al., 2005;Zilberberg et al., 2008).Studies have shown that inadequate antifungal therapy is associated with higher mortality (Bow et al., 2010;Wisplinghoff et al., 2006;Ibrahim et al., 2000;Kumar et al., 2009;Kumar et al., 2006;Pappas et al., 2009).Efforts to optimize antifungal therapy for the treatment of candidemia are well founded.
Two echinocandins were assessed in the first head-tohead echinocandin randomized clinical trial for invasive candidiasis, conducted by Pappas et al. (2007).This double blind multicenter randomized clinical trial compared caspofungin with micafungin at two different dosages (100 and 150 mg) and demonstrated the noninferiority of micafungin to caspofungin at four time points: the end of intravenous (IV) therapy (the primary endpoint), the end of all antifungal therapy, two weeks after the end of therapy, and six weeks after the end of therapy.Of note, micafungin proved to be more efficacious than caspofungin against Candida parapsilosis candidemia (Pappas et al., 2007).
Although no significant difference in overall efficacy was observed between micafungin and caspofungin, surprisingly, the higher dose of micafungin (150 mg, M150) was somewhat less effective than the lower dose (100 mg, M100) at primary end point of the end of blinded intravenous therapy.M150 produced slightly lower response rates, slower time to eradication of Candida bloodstream infections (median time to eradication of three versus two days), and higher rates of persistently positive blood cultures (Pappas et al., (2007).One explanation for these observations of higher efficacy, faster eradication of candidemia and lower rate of persistence of candidemia may be the pharmacodynamic (PD) properties of M100 compared to those of M150.
The aim of this study was to explore PD modelling for candidemia of M100, compared to M150, and secondarily of caspofungin, in order to correlate the PD effect of these agents with clinical outcome and eradication of candidemia.It was hoped that this assessment may provide a rationale for the somewhat greater clinical efficacy of M100 compared to M150 in treating candidemia thus providing a clinical correlate to the invitro paradoxical effect noted with the echinocandins (Chamilos et al., 2007;Clemons et al., 2006;Paderu et al., 2007).It was also aimed to verify the comparable clinical response rates and eradication times of candidemia for micafungin and caspofungin.PD variables of area under the inhibitory curve (AUC) and minimum inhibitory concentrations (MIC) as well as the AUC:MIC ratio of the drugs were also examined to assess differences in success rates among the Candida species causing candidemia in the original trial.

PATIENTS AND METHODS
Patient data were extracted from the previously referenced clinical trial conducted by Pappas et al. (2007.As described in the original study, patients were ≥18 years of age and stratified by Acute Physiology and Chronic Health Evaluation (APACHE) II score and region.The population for the present study consisted only of patients with candidemia from the original study, defined as having at least one positive blood culture for Candida organisms.All bloodstream non-Candida yeast infections were excluded, as well as those patients with Candida isolated from a sterile body and invasive candidiasis only.Individuals with candidemia and concomitant invasive candidiasis were included, but only the candidemia infection was analyzed.
Baseline demographic data (age, sex, race, weight) and patient risk factors for candidemia were recorded from the original study to be included in this analysis.Baseline information was defined as the day of study enrolment.Five study outcomes were evaluated: mycological eradication, clinical success at end of therapy (EOT), recurrence, emergent fungal infection and days to eradication (DTE).Clinical response at the end of intravenous study drug therapy (EOT) as previously defined was used (Pappas et al., 2007).
Mycological response was categorized as Candida eradication or persistence.Eradication was defined as an absence of the infecting Candida spp.from the bloodstream as documented by two negative blood culture samples obtained at least 24 h apart.Persistence implied the continued presence of the baseline fungal pathogen in follow up blood cultures after the initiation of antifungal therapy or when cultures could not be obtained (e.g.death).Emergent fungal infection was defined as a proven invasive fungal infection diagnosed more than 72 h after the first dose of intravenous study drug caused by another species in a blood culture than the one or ones identified at baseline.
A fungal infection with the same species as the baseline infection, but at a different site, was counted as an emergent invasive infection only if it was diagnosed during blinded intravenous therapy.Recurrent fungal infection was defined as having positive blood cultures of the same species as the infection diagnosed at enrollment (baseline) that required additional systemic antifungal therapy.For the EOT response as defined earlier, patients were divided into clinical successes versus failures as per the original study.Clinical success, partial response, stabilization and failure were as previously published (Pappas et al., 2007).Adjudication of response for candidemia was independently performed by a Data Review Panel (DRP).As previously reported, patients were randomized in a 1:1:1 ratio to one of three daily intravenous treatment groups: M100, M150, or caspofungin (70 mg on day 1 and 50 mg daily thereafter) (Pappas et al., 2007).
In vitro susceptibility testing for all isolates was performed (National Committee for Clinical Laboratory Standards, 2002) and these data were recorded for the following Candida spp.categories: Candida albicans, Candida tropicalis, C. parapsilosis, Candida glabrata, Candida krusei, and "others".MICs were determined after 24 h of incubation.The 24-h MIC was utilized to calculate the AUC:MIC for each of the study drugs.The AUC:MIC was equal to the 24-h area under the curve (AUC) over the 24-h MIC for a given species.AUC:MICs were calculated based on estimated steady state 24-h AUC and MIC for the Candida isolate causing the bloodstream infection.The predicted steady state AUC values were estimated based on the patient creatinine clearance and body weight in a fashion similar to pharmacokinetic data previously published by Amsden et al. (1993).
Only patients with available susceptibility data performed for the baseline Candida pathogen and sufficient data to generate AUC:MIC were utilized in the present analysis.Patients with missing microbiological data were excluded.Two patients, one in the caspofungin arm and one in the M100 arm, with missing weights precluded AUC calculation and were excluded.AUC:MIC values were calculated employing the population pharmacokinetic methods utilized by Amsden et al. (1993).
Briefly, the AUC:MIC was equal to the 24 h estimated steady state area under the concentration curve over the 24 h MIC of the organism.The estimated AUC was based on the patient's calculated creatinine clearance and weight.For the purpose of AUC:MIC calculation, height for all patients was assumed to be 161 cm for women and 170 cm for men as individual patients from North America as well as AUC:MIC Europe and South America were enrolled in the original study.Patients with missing follow up blood cultures were adjudicated as failures.For the AUC:MIC calculation, creatinine clearance was estimated by means of the Cockcroft-Gault formula (Cockcroft and Gault, 1976) .Real body weights were employed rather than ideal body weight.
Statistical analyses were performed as follows: categorical variables expressed as percentages were compared using the chisquare and Fisher's exact tests, while normally distributed continuous variables (presented as means with standard deviations) were analyzed with a one way ANOVA test.Nonnormally distributed variables such as APACHE II scores, were assessed by the Kruskal Wallis test.
Univariate logistic regression models were created from which the odds ratios for eradication (ORE) for the following risk factors were estimated: study drug; organism (C.parapsilosis vs. other species); type of infection and catheter removal.Univariate logistic regression was also used to examine the effect of AUC:MIC of the study medications (M100 and M150) on mycological response for the DRP assessed outcome data.
Subsequently, multivariable logistic regression models were used to investigate the effect of treatment arm in conjunction with risk factors on clinical response, mycological response, recurrence and emergent fungal infection.A backward stepwise logistic regression procedure was employed to select significant risk factors from the following list: age, gender, race, weight, hematologic malignancy, solid tumor, chemotherapy within 30 days, radiation within 2 weeks, solid organ transplantation, hematopoietic stem cell transplant, graft versus host disease, hemodialysis, treatment arm, Apache II score, corticosteroid use within 2 weeks, neutropenia at baseline, diabetes, type of infection, catheter removal, organism type and average AUC:MIC.Thereafter, multivariate logistic regression was also undertaken to examine the effect of treatment on mycological and EOT responses with the various correlated risk factors.The analysis focused on an AUC:MIC threshold of 3000 (<3000 versus ≥3000).
Univariate and multivariable generalized linear Poisson regression was used to determine the effects of treatment on days to eradication (DTE).For all the models, estimates with a Wald test P value less than 0.05 was considered as statistically significant (used to determine the significance of an exploratory variable in a statistical model).Kaplan-Meier plots were employed to assess cumulative eradication probability versus DTE for various AUC:MIC cut points from 50 to 1000 for caspofungin and 50 to 17500 for M100 and M150.A Log-rank p value less than 0.05 was considered as statistically significant.Finally, the probability of having eradication in the groups for the issue of catheter removal versus no catheter removal was also compared using Kaplan-Meier plots.Stata

RESULTS
The disposition of patients is as shown in Figure 1.The intention to treat (ITT) population consisted of 593 patients randomized to one of the 3 arms of the study (Reboli et al., 2007).Thereafter, the modified intention to treat (MITT-those who received one dose of the test drug) population was further reduced by assessing only patients with candidemia and available AUC:MIC data.The final allocated subject number (436 patients) was as follows: M100 (n=149), M150 (n=152), and caspofungin (n=135).
The patient demographic characteristics and risk factors for candidemia in the treatment arms are shown in Table 1.There was a statistically significant weight difference in the patient groups, with the M100 group having the lowest mean weight (p=0.002).With regards to risk factors for candidemia, no statistically significant differences between groups were noted, except for: surgery within two weeks of enrolment (more frequent in the caspofungin arm, p=0.04); kidney disease (most common in the M100 arm, p=0.03); and bacteremia (most common in the M150 arm, p=0.02).Finally, there was a significant difference in the AUC:MIC distribution across the 3 arms as noted in Table 1.The caspofungin arm had the fewest number of patients with higher AUC:MICs, ≥1500 (1 patient; p<0.001).
A specific breakdown of Candida isolates producing candidemia in each arm and their susceptibilities are included in Table 2. C. parapsilosis was singled out due to its decreased susceptibility to echinocandins.When comparing C. parapsilosis vs. other Candida spp., the distribution differed significantly, with the caspofungin group possessing the highest frequency of C. parapsilosis (p=0.01).
Five study outcomes were evaluated: mycological eradication, clinical success at EOT, recurrence, emergent fungal infection and DTE.Assessments of the aforementioned outcomes were independently undertaken by the DRP.The outcomes and death rates across the three treatment arms are shown in Table 3.
To examine the effect of AUC:MIC on mycological eradication, AUC:MIC cut points ranging from 150 to 12500 were explored with univariate logistic regression (Table 4).Keeping in mind that caspofungin had only 1 patient with an AUC:MIC ≥1500 as noted in Table 1 (and in fact this one caspofungin patient had an AUC:MIC ≥3000), these data really reflect the AUC:MIC values for M100 and M150.An odds ratio of eradication (ORE) of 2.19 (95% confidence interval [CI] 1.06-4.55)for patients with an AUIC ≥1500 versus those with an AUIC <1500, indicating that patients with an AUIC ≥1500 had a twofold greater chance for successful Candida eradication (p=0.04).No significant dichotomy between higher AUC:MICs and poorer mycological eradication was observed for micafungin.However, as indicated in Table 4, decreased OREs were noted when a threshold of AUC:MIC ≥7500 for micafungin, thus signifying inferior  mycological eradication.Also, using a univariate logistic regression model, there was a statistically significant difference in the mycological eradication for all Candida spp.versus C. parapsilosis, in the DRP assessment (ORE 0.39, p=0.004).An ORE of < 1.00 indicated that the odds of pathogen eradication for patients with C. parapsilosis were less than the odds for patients with other Candida spp.Adding catheter removal into the model showed that catheter removal had no impact on mycological eradication.
Multivariate logistic regression analysis was used to model the effect of other risk factors on mycological eradication per treatment group with an AUC:MIC threshold of 3000.The antifungal treatment used proved to have a significant impact on Candida eradication, but the M150 group had a lower chance of eradication compared to the caspofungin group (p=0.02).Other risk factors having a significant impact on the ORE included: race (others vs. white) (ORE 0.30, CI 0.14-0.64,p=0.002); diabetes (ORE 2.73, CI 1.24-6.02,p=0.01); C. parapsilosis (ORE 0.33, CI 0.14-0.79,p=0.01); and liver disease (ORE 0.47, CI 0.25-0.88,p=0.02) with either of the assessments.
Although univariate logistic regression showed that catheter removal had no direct correlation with mycological eradication, we further explored the impact of catheter removal by incorporating it as a variable in multivariate models for mycological eradication.According to the DRP outcome adjudication, when catheter removal was included in the multivariate modelling, the treatment arm was again significant with M150 having a 68% lower odds of eradication compared to the caspofungin arm (p=0.04).
Univariate logistic regression was used to examine the effect of treatment on recurrence and emergent fungal infection.With multivariate model evaluations, the DRP data for recurrence employed an AUC:MIC threshold of 3000 and yielded a multivariate model result where weight (every 1 kg increase) was a significant risk factor producing an odds ratio of recurrence (ORR) of 1.03 (CI 1.02-1.06,p=0.02).For emergent infections, with the DRP outcomes, significant correlated risk factors were: race (black vs. white) (odds ratio of emergent infection [OREI 0.08, p=0.03); solid organ transplant (OREI 8.09,

DISCUSSION
This retrospective analysis of the previously published first head-to-head clinical trial comparison of echinocandins for invasive candidiasis (Pappas et al., 2007) was undertaken to evaluate PD parameters of caspofungin, M100, and M150 for the treatment of candidemia alone.
The original double blind trial used two independent evaluations from the Investigators and DRP.The DRP assessments were utilized only for the purpose of this study for three treatment arms with regards to five study outcomes: mycological eradication, clinical success at the end of the study drug IV therapy, DTE, recurrence, and emergent fungal infections.
With regards to mycological eradication, univariate logistic regression of the DRP outcomes showed that the AUC:MIC threshold of 1500 was a significant factor.Patients with AUC:MICs equal to or above this level exhibited a higher ORE of eradication as compared to patients with AUC:MICs below this threshold.This was only true for the two micafungin arms, as the caspofungin arm had only one patient who attained an AUIC ≥1500.Thus, higher doses of micafungin lead to better eradication for susceptible organisms as noted by the shorter DTEs.Although, one might then have anticipated that higher doses of the study drug would always correlate with improved Candida eradication, the exploratory analyses noted a reversal of the ORE at AUC:MICs >7500.This may indicate a possible paradoxical effect with micafungin.Similarly, an analysis of mycological eradication showed that starting from very high AUC:MIC thresholds of about 7500, M150 was inferior to M100.
The species also impacted mycological eradication.Univariate and multivariate models corroborated the finding that patients with C. parapsilosis infections had a decreased chance of eradication compared to patients with infections caused by other species.This was expected, as echinocandins have lower in vitro activity and clinical efficacy against C. parapsilosis (Pfaller and Diekema, 2007;Walker et al., 2010;Pfaller et al., 2008;Ostrosky-Zeichener et al., 2003;Chen et al., 2011;Barchiesi et al., 2006).Although race was consistently correlated with mycological eradication, the impact of race on the treatment of candidemia is unclear.
Additionally, catheter removal was a significant factor for time to eradication.This was not surprising, as early catheter removal is generally the standard practice for patients with candidemia (Mermel et al., 2009).It is noteworthy that catheter removal had no significant effect on the other outcomes.
The AUC:MIC data analysis support the clinical response results of the original study, whereby M100 was associated with greater success than the M150 arm.This finding is further corroborated by the DTE results in the analysis where the DRP evaluation found treatment arm to be a significant factor affecting DTE: the M150 arm required more DTE than the M100 arm.Finally, APACHE II score was a significant factor in treatment success.For each unit increase in the median score at baseline, there was 3 to 4% less chance of clinical success.
The sole risk factor that correlated with recurrence was weight.With each 1 kg increase in weight, patients had 3% higher odds of recurrent infection per the DRP assessment (ORR 1.03, p=0.02).This may be explained by the following.There is evidence that total body weight may affect volume of distribution (Hanley et al., 2010).Furthermore, there is a statistically significant relationship between weight and drug clearance.Andes et al. (2011) found that patients with weights over 100 kg had clearance values that were approximately 30% higher than the clearance values for the overall population, while patients with weights under 45 kg had clearance values approximately 30% lower than the overall population (Andes et al., 2011).With faster clearance, the drug is more rapidly removed, rendering it unavailable to produce its pharmacological effect.
Thus, it is plausible that an increase in weight may impact PK/PD factors such as volume of distribution and clearance such that serum levels are diminished, thus reducing the drug's effect.Perhaps larger doses of the drugs may be necessary in patients above 100 kg to prevent recurrence.However, it is unknown why the effect of weight was only observed with regards to recurrence.
The present study, however, possesses some limitations.The data evaluations were conducted in a retrospective fashion after the publication of the clinical trial.This may have curtailed some of the data collection.Some differences were also noted between the investigator assessment reported in the original study and the DRP assessment.The DRP evaluation was favoured as it was deemed that it would be more robust and less subject to bias.They were also hampered by the use of generalized population pharmacokinetic parameters rather than individualized parameters based on individual antifungal agent serum levels.
In addition, one may question our exclusion of noncandidemic invasive candidiasis.Such cases may have prolonged DTE, but were excluded because the DTE could not be determined with certainty as follow up cultures were not available.Also, the rationale for using a threshold of AUC:MIC 3000 may seem subject to bias.But, this threshold seemed appropriate as caspofungin group did not attain AUC:MIC thresholds above this level.There is however another caveat worth mentioning with regard to the caspofungin MICs that were obtained.It is known that caspofungin MICs to Candida spp.exhibit considerable variability (Espinel-Ingroff et al., 2013).
Thus, the present reported AUC:MIC results for caspofungin may not reflect the true state of affairs for this variable.Furthermore, any pharmacodynamic data on the other echinocandin and anidulafugin was not included.This echinocandin was not part of the data that was assessed nor was there access to data of anidulafungin's pharmacodynamic effects for eradicating candidemia.Precise data on the DTE are not available for anidulafungin (Reboli et al., 2007).These deficiencies notwithstanding were successful in exploring differences between M100 and M150.
Of import, patients treated with micafungin attaining AUC:MICs ≥3000 exhibited less time to eradication than those with AUC:MICs below this threshold.These findings parallel those of Andes et al. (2018).This trend, however, was not consistent at higher AUC:MIC cut points (Table 4).A paradoxical effect with M150 may have been present.Mycological eradication and DTE analyses both showed that M150 was frequently associated with a lower success rates than M100 in the treatment of candidemia.
The present study demonstrated the value of PD analyses in the evaluation of the treatment of candidemia.The clinical response findings were corroborated at the end of therapy as noted by Pappas et al. (2007).The results for pathogen eradication and DTE noted, that M100 was superior to M150.Our multivariate models featuring PD variables also helped to identify correlated risk factors that could serve as predictors of favourable treatment.Finally, AUC:MIC values of ≥3000 but <7000 were associated with optimal outcomes for both micafungin doses.Future PD studies are warranted to explore the optimal dosing for echinocandins with respect to achieving mycological eradication, DTE, and deterring recurrent and emergent infections.
This work was supported in part by grant funds provided by Astellas Pharma Canada Inc.
version 12 (StataCorp 2011, Release 12, College Station, Texas StataCorp LP) was used for all the analyses.

Table 1 .
Demographic characteristics of the treatment groups.

Table 2 .
In vitro susceptibilities of Candida isolates per treatment group.Minimum inhibitory concentration to inhibit 50% of isolates.+ Minimum inhibitory concentration to inhibit 90% of isolates.ǂ Min MIC = minimum MIC.**Max MIC = maximum MIC. *

Table 3 .
Mycological and clinical outcomes for the treatment groups.

Table 3 .
Contd.One patient had a missing value in mycological response (Investigator).+ Data review panel.ǂ One patient had missing values in clinical response (Investigator & DRP).**Min = minimum days.

Table 4 .
Univariate logistic regression for the effect of AUC:MIC on mycological eradication for micafungin.