Pharmacokinetic/Pharmacodynamic Target Attainment of Different Antifungal Agents in De-escalation Treatment in Critically Ill Patients: a Step toward Dose Optimization Using Monte Carlo Simulation

ABSTRACT

Differences in pharmacokinetics/pharmacodynamics (PK/PD) target attainment are rarely considered when antifungals are switched in critically ill patients. This study intends to explore whether the antifungal de-escalation treatment strategy and the new intermittent dosing strategy of echinocandins in critically ill patients are able to achieve the corresponding PK/PD targets. The published population PK models of antifungals in critically ill patients and a public data set from the MIMIC-III database (n = 662) were employed to evaluate PK/PD target attainment of different dosing regimens of antifungals. Cumulative fraction of response (CFR) was calculated for each dosing regimen. Most guideline-recommended dosing regimens of fluconazole and voriconazole could achieve target exposure as de-escalation treatment in critically ill patients. For initial echinocandin treatment, achievement of the target exposure decreased as body weight increased, and the intermittent dosing strategy had a slightly higher CFR value in most simulations compared to conventional dosing strategy. For Candida albicans and Candida glabrata infection, caspofungin at the lowest dose achieved a CFR of >90%, while micafungin or anidulafungin required almost the highest doses simulated in this study to achieve the same effect. None of the echinocandins other than 150 mg every 24 h (q24h) or 200 mg q48h of caspofungin achieved the target CFR for Candida parapsilosis infection. These findings support the guideline-recommended dose of triazoles for antifungal de-escalation treatment and confirm the insufficient dosage of echinocandins in critically ill patients, indicating that a dosing regimen based on body weight or intermittent dosing of echinocandins may be required.

INTRODUCTION

Invasive candidiasis (IC) is the most common fungal disease among critically ill patients (1). Despite the increasing opportunities for antifungal treatment, IC-related prognosis is still poor, with a crude mortality rate approaching 40% in these patients (2, 3).

Several guidelines recommended echinocandins as the first-line treatment option for IC in nonneutropenic critically ill adults based on their superb efficacy, favorable safety profile, broader spectrum compared to fluconazole, and limited drug interactions (4). Meanwhile, it has become common practice for clinicians treating patients with IC to initiate an echinocandin, then switch to a triazole (i.e., fluconazole or voriconazole) once the patient has turned to clinically stable (5).

Pharmacokinetics/pharmacodynamic (PK/PD) target attainment is closely associated with the efficacy of antifungals. The probability of target attainment (PTA) is a commonly used indicator to assess overdose or underdose for different antimicrobial regimens. There is increasing evidence indicating that antifungal treatment is frequently underdosed in the treatment of IC, especially in critically ill patients (6). Although several observational studies have demonstrated that an antifungal de-escalation treatment strategy could reduce treatment costs without affecting outcomes among critically ill patients (7, 8), the possible change in PTA of antifungal de-escalation treatment strategy has not yet been investigated. A previous study had reported that the PTA for narrow-spectrum β-lactam antibiotics applied as de-escalation treatment were not optimal for a selection of microorganisms (9), and whether this will occur for antifungal de-escalation treatment is still unknown.

Meanwhile, echinocandin therapy with conventional dosing fails in approximately 40% of cases (10). The property of echinocandins (i.e., concentration-dependent killing and a prolonged postantifungal effect) suggests that a new dosing strategy with the administration of higher doses with extended dosing intervals (i.e., intermittent dosing) might be sufficient to eradicate Candida spp. The effect of intermittent dosing of the echinocandins was tested in experimental animal models (11) and clinical studies (12). Nevertheless, the comparison of PTA between the conventional and the new intermittent dosing strategy has not been investigated.

Hence, this study would like to focus on antifungal de-escalation treatment strategy, comparing the PTA with conventional dosing of echinocandins and triazoles for different Candida spp. Furthermore, the second purpose of this study was to determine whether an intermittent dosing strategy of echinocandins could improve target attainment.

RESULTS

De-escalation treatment.

As shown in Table 1, 6 mg/kg of body weight or 400 mg of fluconazole, regardless of the bodyweight of patients or the route of administration, achieved target attainment at 100% for Candida albicans, Candida parapsilosis, and Candida tropicalis. As for Candida glabrata, only when fluconazole was administered at high doses (i.e., 12 mg/kg or 800 mg) as an injection form could PTA achieve a value of >95% for all simulations in different weights (except 800 mg given for patients weighing 100 kg), which was in line with the guideline recommendations (5). However, high-dose fluconazole administered orally failed to achieve optimal exposure in patients of any weight for IC caused by C. glabrata.

TABLE 1.

Probability of target attainment of de-escalation antifungal treatment against Candida spp. in critically ill patients with different weights using epidemiologic cutoff values as pharmacodynamics index

Antifungal	fAUC0–24/MIC	Dosing regimena	Probability of target attainment (%) for Candida species according to wt (kg)b
			C. albicans			C. glabrata			C. parapsilosis			C. tropicalis			C. krusei
			50	75	100	50	75	100	50	75	100	50	75	100	50	75	100
Fluconazole	25	6 mg/kg QD i.v.	100	100	100	0.2	0	0	100	100	100	100	100	100
		400 mg QD i.v.		100	100		0	0		100	100		100	100
		12 mg/kg QD i.v.				100	99.9	99.8
		800 mg QD i.v.					98.3	6.6
		6 mg/kg QD p.o.	100	100	100	0	0	0	100	100	100	100	100	100
		400 mg QD p.o.		100	100		0	0		100	100		100	100
		12 mg/kg QD p.o.				0	0	0.5
		800 mg QD p.o.				38.6
Voriconazolec	25	150 mg q12h i.v.		100			100			100			100			86.9
		150 mg q12h p.o.		100			100			100			100			60.3
		200 mg q12h i.v.		100			100			100			100			99.8
		200 mg q12h p.o.		100			100			100			100			97.8

^aQD, once a day; i.v., intravenously; p.o., orally.

^bBoldface indicates PTA of ≤90%. Body weights are given as 50, 75, and 100 kg.

^cPTA values for voriconazole are calculated regardless of weight.

When voriconazole was selected as a de-escalation treatment for IC caused by Candida krusei, patients weighing 50 kg administered a dose of 3 mg/kg every 12 h (q12h) (150 mg) might not achieve a therapeutic exposure. For any other Candida species, 150 mg q12h of voriconazole could achieve 100% PTA regardless of route of administration (Table 1).

Additionally, changing the PD index to MIC breakpoints or setting the fluconazole target value to 100 gave results similar to the original scenario in the sensitivity analyses, except that the efficacy of high-dose fluconazole (i.e., 12 mg/kg or 800 mg) against C. glabrata and oral fluconazole treatment against C. parapsilosis and C. tropicalis (see Tables S1 and S2 in the supplemental material).

Initial echinocandin treatment.

The results of the simulations for three echinocandins with several dosing regimens (conventional and intermittent dosing) in different weights are displayed in Table 2. Achievement of the cumulative fraction of response (CFR) generally decreased with an increase in body weight. The CFR for echinocandins with intermittent dosing strategy was equal to or slightly higher than that of the conventional dosing at an equivalent dose.

TABLE 2.

Cumulative fraction of response (CFR) of echinocandins against Candida spp. in critically ill patients with different weights using net fungal stasis target

Antifungal	Dosing regimen (mg/day)	Cumulative fraction of response (%) against Candida spp. according to wt (kg)a
		C. albicans			C. glabrata			C. parapsilosis			C. tropicalis			C. krusei			C. auris
		50	75	100	50	75	100	50	75	100	50	75	100	50	75	100	50	75	100
Caspofungin	70/50b	90.4	88.5	88.5	99.0	99.0	99.0	64.6	61.8	19.3
	100 q48h	88.7	88.5	88.4	99.0	99.0	99.0	63.2	60.8	14.5
	100/70	98.5	98.2	93.2	99.7	99.7	99.2	64.7	64.6	64.2
	150 q48h	98.5	98.4	95.3	99.7	99.7	99.3	65.1	64.6	64.3
	100	98.5	98.5	98.4	99.7	99.7	99.7	89.2	86.5	64.7
	200 q48h	98.5	98.5	98.5	99.7	99.7	99.7	90.1	88.7	65.9
	150	99.4	99.4	98.8	99.7	99.7	99.7	90.8	90.8	90.7
Micafungin	100b	62.3	18.3	10.2	71.7	20.5	11.2	11.0	7.9	3.6	41.1	10.5	4.9	2.9	0.5	0.1	94.3	94.3	94.2
	200 q48h	63.0	25.6	10.4	72.5	29.0	11.3	11.0	9.0	3.9	41.6	15.6	5.0	2.9	0.9	0.1	94.3	94.3	94.3
	300 q72h	63.0	28.3	10.5	72.5	32.2	11.4	11.0	9.5	4.5	41.7	17.5	5.1	2.9	1.0	0.1	94.3	94.3	94.3
	150	65.6	62.9	57.6	74.5	72.4	66.2	22.2	11.0	10.9	45.3	41.6	37.9	3.7	2.9	2.6	94.3	94.3	94.3
	300 q48h	67.3	63.1	59.8	75.8	72.6	68.7	25.1	11.2	10.9	47.7	41.7	39.4	4.3	2.9	2.7	94.3	94.3	94.3
	200	88.3	71.1	63.6	91.7	78.6	72.9	34.3	25.6	13.1	78.0	53.2	42.4	11.2	5.5	3.1	94.3	94.3	94.3
	250	88.5	87.7	75.7	91.8	91.2	82.1	34.5	34.2	28.6	78.3	77.1	59.8	11.3	11.0	7.1	94.3	94.3	94.3
	300	88.9	88.4	87.1	92.0	91.8	90.7	55.9	34.4	33.9	79.3	78.2	76.2	14.2	11.3	10.8	94.3	94.3	94.3
Anidulafunginc	200/100b		68.2			69.0			4.0			94.8
	250 q48h		72.5			75.2			4.1			95.1
	350 q72h		71.4			73.7			4.1			95.1
	200/150		75.6			79.6			4.9			95.6
	300 q48h		75.5			79.4			4.8			95.6
	200		81.8			86.3			6.1			96.8
	250		89.7			93.9			6.7			97.4
	300		92.0			95.7			7.6			97.7

^aBoldface indicates CFR of ≤90%. Body weights are given as 50, 75, and 100 kg.

^bConventional dose for echinocandins.

^cCFR values for anidulafungin are calculated regardless of weight.

For C. albicans, only caspofungin at any simulated dose (except for 100 mg q48h for patients in any weight or 70/50 mg for patients weighing >75 kg) or 300 mg q24h of anidulafungin achieved a CFR of >90%. Similarly, caspofungin at the lowest dose (70/50 mg or 100 mg q48h) or micafungin or anidulafungin administered at a high dose (300 mg or 250 mg q24h of micafungin or anidulafungin, respectively) could achieve a CFR of >90% for patients in any weight regarding C. glabrata infection. None of the echinocandins other than 150 mg q24h or 200 mg q48h of caspofungin for patients in any weight or patients weighing 50 kg, respectively, achieved the target CFR for C. parapsilosis infection. The simulated lowest dose of micafungin and anidulafungin was sufficient to guarantee a CFR of >90% for IC caused by Candida auris and C. tropicalis, respectively. Meanwhile, the simulated maximum dose of micafungin still failed to achieve the optimal exposure in the treatment of C. tropicalis (CFR < 80%) and C. krusei (CFR < 15%) infection.

Almost the same findings as the base were revealed when 1-log fungal kill was used as the PD targets in the sensitivity analysis (see Table S3 in supplemental material).

DISCUSSION

Antifungal de-escalation therapy to fluconazole or voriconazole is strongly recommended in several guidelines (13) to reduce echinocandin consumption, preventing FKS mutation-related Candida nonsusceptibility and saving costs (14). De-escalation therapy generally seems to be safe and possible and has been associated with lower crude mortality rates in an observational study (7). This study, as far as we know, is the first of its kind to quantify the range of potential exposures correlated with different weight-based doses of antifungal treatment (initial echinocandin treatment and de-escalation triazole treatment) based on published population PK models from critically ill patients. All of the population PK models of antifungals that we employed are deliberately selected. The variability of PK parameters was adequately described in these models and allowed other researchers to simulate the corresponding PK profiles of critically ill patients.

We have found that most guideline-recommended dosing regimens of fluconazole and voriconazole could achieve the target exposure of de-escalation treatment, which conflicts with other reports that current fluconazole dosing regimens do not achieve adequate target attainment in critically ill adults (15). There are two possible reasons for this disparate finding. First, the susceptibility of the microorganism might play an important role. We used epidemiologic cutoff values (ECVs) of fluconazole and voriconazole for Candida spp., which are much lower than clinical breakpoints, to calculate the corresponding PTA of de-escalation treatment. The ECV is set at the upper limit of the wild-type distribution, and it encompasses 97.5% of the modeled wild-type population. Such an organism with a MIC higher than the ECV is presumed to express resistance mutations that may affect treatment response (16). However, sensitivity analyses using the MIC breakpoints found no significant change from baseline results. Second, de-escalation treatment is generally performed for critically ill patients who are clinically stable and have a lower fungal burden, and therefore, we set the PD target of fluconazole as 25 rather than 100 (15, 17). Of note, high-dose fluconazole (i.e., 12 mg/kg or 800 mg) against C. glabrata and oral fluconazole treatment against C. parapsilosis and C. tropicalis were no longer effective when the PD target was set at 100.

PK data on echinocandins in critically ill patients are very limited. All of the population PK models of echinocandins that we applied have identified weight-related covariates to explain PK variation (weight for caspofungin and micafungin; BMI for anidulafungin), and we found that achievement of the target exposure decreased as body weight increased. However, echinocandins are generally administered in a fixed-dosing manner in clinical practice. On the basis of safety and efficacy considerations, the weight-based dosing method may be more appropriate for achieving adequate exposure in critically ill patients where weight could explain PK variability as a significant covariate, which is consistent with the previous report on caspofungin (18). Besides weight-related covariates, sequential organ failure assessment (SOFA) score and albumin level have also been found to affect the variation of PK parameters of micafungin (SOFA and albumin level on clearance [CL] and albumin level on volume of distribution [V]) (19) and anidulafungin (SOFA on CL) (20). SOFA is a hybrid score, consequently reflecting the general status of the organs and predicting outcomes in critically ill patients. A higher SOFA score is associated with lower CL, which may indicate that there is no need to increase the dose of micafungin or anidulafungin in critically ill patients with a higher SOFA score. In the present study, the simulations were undertaken using the same population from the MIMIC-III database, which makes the comparison of PK profiles between different antifungals more homogeneous.

In this simulation study, caspofungin showed the best PK/PD target attainment among three different echinocandins, which was based upon its highest CFR value regardless of the species of Candida (even if the lowest dose of caspofungin is compared with the highest dose of micafungin or anidulafungin). Meanwhile, except for micafungin against C. auris and anidulafungin against C. tropicalis, the CFR values of both micafungin and anidulafungin were not able to achieve the target at conventional doses or relatively low doses. The low susceptibility of C. parapsilosis to echinocandins dramatically reduced the corresponding CFR value of echinocandins. Only high-dose caspofungin (150 mg q24h for patients in any weight or 200 mg q48h for patients weighing 50 kg) seemed to be effective against C. parapsilosis. The results of the present study were partly consistent with our previous simulation study of echinocandins, which applied the noncompartmental PK data that could only be used to simulate the actual dose (21).

Due to pathophysiological changes in critically ill patients, such as an augmented renal clearance or increased V, conventional dosing might not constantly result in optimal target attainment. Our study has observed the comparable 6-day dose-normalized exposure over time for higher doses with extended dosing intervals (i.e., intermittent dosing) and conventional dosing of echinocandins (with the intermittent dosing strategy having a slightly higher CFR value in most simulations), which was in line with other real-world data (22, 23). Notably, proportionally higher peak concentrations for the intermittent dosing were also demonstrated in our study. Andes et al. found the intermittent dosing of micafungin improved cure rates to close to 90% (87% on every-other-day dosing versus 79% with daily doses [P = 0.056]) and halved relapse rates (6% versus 12%, respectively [P = 0.051]) in patients with esophageal candidiasis (12). The areas under the concentration-time curve (AUCs) were identical between the dosing regimens; however, the peak concentrations of the 300-mg every-other-day dosing is 2 times higher than 150 mg q24h, but the trough concentration is lower, indicating that the larger peak concentration and earlier adequate exposure might have driven efficacy (12). Administering intermittent dosing of anidulafungin to achieve better target attainment was also reported in patients with acute leukemia (24). Nevertheless, the advantages of this dosing method might not be fully shown in critically ill patients, considering that these patients themselves are in critical condition and have to receive daily fluid therapy. Possible increased rates of hepatic and cardiovascular toxicity are the major concern in patients treated with echinocandins at high doses. Nonetheless, the safety concerns are very limited, as high dosages of echinocandins have been investigated to be well-tolerated in multiple clinical trials (23, 25, 26).

Some limitations of the present study should be discussed. First, this simulation study could not obtain the measured concentrations of real patients. However, all of the population PK models of antifungals that we deliberately selected could adequately describe the variability of PK parameters in critically ill patients; therefore, the accuracy of simulated concentrations and relevant results could be assumed to be acceptable. Of note, due to the lack of existing evidence, we could only apply the population PK model of oral fluconazole suspension and intravenous voriconazole to simulate the PK profiles of oral fluconazole and voriconazole, respectively. Second, the PK/PD target is not well established in clinical trials, especially in critically ill patients, and the currently used targets are based on murine models only on day 1 or 2 of antifungal treatment except micafungin (27). These targets should be examined with clear outcome measures in prospective patient cohorts and for whether or not the target changes over time also need to be investigated. Third, although this study focused on antifungal de-escalation treatment, it is hard to investigate the subsequent impact of the switch of treatment regimen on individuals; hence, we only initially explored the targets attainment of different antifungal agents in critically ill patients when they were used independently. Fourth, hypoproteinemia is very common in critically ill patients, but our study did not consider its effect on drug concentrations. In fact, in patients with hypoalbuminemia, the unbound proportion of highly protein-bound drugs (e.g., echinocandins) increases due to the decrease in available binding sites. As a result, there will be increased elimination of the drugs due to an increase in CL of the unbound proportion, which will further exacerbate the underexposure of echinocandins (28). Finally, for the intermittent dosing strategy of echinocandins, we examined the 6-day total exposure instead of 24-h exposure. It is still unknown which timing of the target attainment is more important for antifungal treatment.

In conclusion, the results of this simulation study demonstrated that most guideline-recommended dosing regimens of fluconazole and voriconazole could achieve the target exposure as de-escalation treatment in critically ill patients. For initial echinocandin treatment, achievement of the target exposure decreased as body weight increased, and the intermittent dosing strategy has a slightly higher CFR value in most simulations compared to that of the conventional dosing strategy. Moreover, caspofungin appeared to be the most effective of the three echinocandins. Future work should be focused on the clinical outcomes of the intermittent dosing strategy of echinocandins in real-world critically ill patients.

MATERIALS AND METHODS

PK data.

We searched the PubMed database in June 2021 to identify studies providing echinocandins and triazoles PK data using the search terms “caspofungin OR micafungin OR anidulafungin OR fluconazole OR voriconazole,” “population pharmacokinetics OR population PK OR population model,” and “ICU OR intensive OR critical OR critically.” Published population PK models with the largest sample sizes or numbers of patients for echinocandins (caspofungin [29], micafungin [19], and anidulafungin [20]) and triazoles (fluconazole [30, 31] and voriconazole [32]) in critically ill patients were collected to obtain the parametric equations regarding clearance (CL) and volume of distribution (V) (Table 3). Protein binding values of 97%, 99.75%, 99%, 12%, and 58% for caspofungin, micafungin, anidulafungin, fluconazole, and voriconazole were used in all simulations to calculate free drug concentrations (28).

TABLE 3.

Population pharmacokinetic models for different antifungal agents^a

Antifungal	Yr	Population	Dialysis or RRT or ECMO	No. patients (no. samples)	Compartments (no.)	Equation	Reference
Caspofungin	2016	ICU patients	Not mentioned	21 (419)	2	CL = 0.55 · (BW/70)0.75; V = 8.98 · (BW/70)	29
Micafungin	2017	ICU patients with sepsis and mechanical ventilation	19 patients were on intermittent hemodialysis and 11 patients were on ECMO	99 (436)	2	CL = 1.34 · (BW/84)0.59 · 1.14 (if albumin ≤ 25 g/L) · 0.75 (if SOFA ≥10); Vc = 11.8 · (BW/84)0.61 · 1.14 (if albumin ≤25 g/L); Vp = 7.68 · (BW/84)0.67 · 1.14 (if albumin ≤25 g/L)	19
Anidulafungin	2020	Critically ill patients	Not mentioned	13 (205)	2	CL = 0.778 · (SOFA/12)-0.924; V =10.2 · (BMI/25)2.74	20
Fluconazole (i.v.)	2012	Critically ill patients	Not mentioned	57 (295)	1	CL = 0.799 · (CLCR/92.7)0.685; V = 48.1 · (BW/65) 1.40	30
Fluconazole (p.o.)	2003	Patients in the surgical ICU	Not mentioned	110 (409)	1	Cl = 1.19 · CLCR/(45.4 + CLCR); V = 109 · (BW/80)·{1 + [−0.0141 · (age − 60)]}	31
Voriconazole	2015	Critically ill patients with pulmonary disease	Not mentioned	62 (240)	1	CL = 4.28·(DBIL/2.6) − 0.40; V = 93.4	32

^aBMI, body mass index; BW, body weight; CL_CR, creatinine clearance; DBIL, direct bilirubin; ECMO, extracorporeal membrane oxygenation; ICU, intensive care unit; SOFA, sequential organ failure assessment; RRT, renal replacement therapy; Vc, volumes of central distribution; Vp, volumes of peripheral distribution.

MIC distribution for Candida spp.

The MIC distribution data of echinocandins for Candida spp. were obtained from the literature (33, 34). The microorganisms used in this simulation study were C. albicans, C. glabrata, C. parapsilosis, C. tropicalis, C. krusei, and C. auris (Table 4). The cumulative fraction of response (CFR) value would be calculated using the susceptibility rates of Candida spp. as a measure of the MIC distribution.

TABLE 4.

Pharmacodynamics targets and MIC distributions of echinocandins for Candida spp.

Candida species	Antifungal agent	PK/PD targets (fAUC0–24/MIC)	PK/PD targets in sensitivity analysis	Percentage of isolates at an MIC (μg/mL) of:											ECV	MIC breakpoint
				0.008	0.016	0.031	0.063	0.125	0.25	0.5	1	2	4	8
C. albicans	Caspofungin	22.4	25.9	0.7	16.2	45.7	25.9	10.0	0.9	0.4	0.1	0.1
	Micafungin	12.5	25.2	9.5	53.6	25.4	5.9	2.3	1.9	1.0	0.3	0.1
	Anidulafungin	27.8	91	9.2	31.4	34.4	17.2	5.0	1.6	0.9	0.3	0.1
	Fluconazole	25	100												0.5	2
	Voriconazole	25													0.03	0.064
C. glabrata	Caspofungin	2.9	13.5		1.2	58.5	23.3	13.0	2.5	0.5	0.7		0.2	0.2
	Micafungin	12.5	9.45	10.3	62.3	19.2	4.0	1.3	0.8	0.6	0.5	0.6	0.4
	Anidulafungin	13.7	32	0.1	2.5	27.2	49.1	17.0	1.5	0.4	0.9	0.8	0.4
	Fluconazole	25	100												8
	Voriconazole	25													0.25
C. parapsilosis	Caspofungin	16.8	35.5		0.4	1.4	10.4	52.4	26.2	8.8	0.4
	Micafungin	0.7125	7.73		0.7	0.2	0.4	1.4	8.3	23.3	45.4	19.3	0.9
	Anidulafungin	11.5	22		0.4	1.7	1.9	2.7	4.9	13.6	30.2	38.3	6.1	0.1
	Fluconazole	25	100												1	2
	Voriconazole	25													0.03	0.125
C. tropicalis	Anidulafungin	6.1		7.4	22.0	38.4	20.5	6.9	2.3	1.6	0.5	0.4
	Micafungin	12.5		4.4	37.3	36.6	16.8	3.4	0.9	0.1	0.4		0.1	0.1
	Fluconazole	25	100												1	2
	Voriconazole	25													0.12	0.125
C. krusei	Micafungin	12.5			2.9	8.4	48.1	32.1	7.6	0.5	0.3
	Voriconazole	25													0.5	0.5
C. auris	Micafungin	0.13	0.33		3.3	3.3	38.2	39.8	7.3	1.6	0.8			5.7

For antifungal de-escalation treatment, we used the epidemiologic cutoff values (ECVs) of fluconazole and voriconazole for Candida spp. to calculate the corresponding PTA (Table 4), considering that using ECVs as the PD target for dosing purposes was proposed in an important recommendation when the measured MIC for the bacterial strain is within the wild-type distribution (35). The MIC breakpoints of fluconazole and voriconazole were also employed in sensitivity analyses.

PK/PD targets for Candida spp.

In this study, the PK/PD indices evaluated were the free-drug 24-h area under the plasma concentration-time curve/MIC (fAUC_0–24/MIC) for Candida spp. The PK/PD targets of echinocandins were derived from previous studies (27, 36–38) (Table 4). We used net fungal stasis for Candida spp. as the PK/PD targets since that previous analysis had determined the consistency between mycological efficacy and this preclinical endpoint in patients with candidemia and IC who were treated with echinocandin (27). The targets of fAUC_0–24/MIC of fluconazole and voriconazole were both set as 25 (39, 40). In addition, a 1-log fungal kill 24-h fAUC_0–24/MIC for echinocandins and an fAUC_0–24/MIC of 100 for fluconazole were set as PK/PD target values in the sensitivity analysis.

Given that the guidelines recommend that the initial echinocandin therapy be for 5 to 7 days and then transition to triazoles, we examine the cumulative AUC exposure over 6 days of echinocandins.

Monte Carlo simulation.

The simulations were performed using Berkeley Madonna v.9.0.127 (Macey R & Oster G, Berkeley, CA). For each antifungal agent to Candida spp., the fAUC_0–24/MIC at the relevant MIC was calculated. Then, 1,000 Monte Carlo simulations were undertaken using the parameters from the published covariate model, and a patient data set (n = 662) was employed to calculate PTA and further corresponding CFR based on the MIC strata as shown below.

$$\text{CFR}={\displaystyle \sum }_{i=1}^n\text{PTA}\left({\text{MIC}}_i\right)\cdot \text{p}({\text{MIC}}_i)$$

A CFR value of ≥90% was regarded as the minimum for achieving optimal empirical therapy.

The patient data set was obtained from the Medical Information Mart for Intensive Care III (MIMIC-III; v1.4) database. The MIMIC-III is maintained by the Laboratory for Computational Physiology at MIT. It contains information for more than 46,000 patients who stayed in critical care units of the Beth Israel Deaconess Medical Center between 2001 and 2012. The database is accessible to researchers who have completed “protecting human subjects” training. X. Han obtained access to this database (certification number 35842213) and was responsible for data extraction. Data extraction was performed using PostgreSQL tools v.1.12.3.

Patients were eligible for enrollment if they were more than 16 years old and were infected by a Candida species isolate detected by positive specimen culture taken from a sterile site (tissue or blood) (Fig. 1). The patient characteristics are summarized in Table 5.

FIG 1.

Flow diagram of the reviewed randomized controlled trials.

TABLE 5.

Patient characteristics

Characteristic	Valuea
No. of males/no. of females	383/279
Age (yr)	62.5 (16–89)
Wt (kg)	80 (32–187)
Ht (cm)	168 (140–188)
BMI (kg/m2)	28.3 (16.5–53.0)
SOFA score	6 (0–21)
Creatinine clearance (mL/min)	67.6 (9.3–292.0)
Serum creatinine concn (mg/dL)	1.2 (0.2–10.8)
Serum albumin concn (g/L)	25 (11–52)
Direct bilirubin (mg/dL)	3.9 (0.2–10)

^aData are reported as the median (range).