Pharmacokinetics/Pharmacodynamics of Caspofungin in Plasma and Peritoneal Fluid of Liver Transplant Recipients

ABSTRACT

The weaker diffusion of echinocandins in the peritoneal fluid (PF) could promote Candida-resistant isolates. The aim of this study was to analyze the pharmacokinetics (PK)/pharmacodynamics (PD) of caspofungin in plasma and PF samples from liver transplant recipients. Liver transplant patients received caspofungin as postoperative prophylaxis. Caspofungin concentrations were quantified in plasma and PF samples on days 1, 3, and 8. Data were analyzed using nonlinear mixed-effect modeling and Monte Carlo simulations. Area under the curve (AUC) values for plasma and PF were simulated under three dosing regimens. Probabilities of target attainment (PTAs) were calculated using area under the unbound plasma concentration-time curve from 0 to 24 h at steady state (fAUC_0-24)/MIC ratios, with MICs ranging from 0.008 to 8 mg/L. All of the patients included were monitored weekly for Candida colonization and for Candida infections. Twenty patients were included. The median daily dose of caspofungin was 0.81 mg/kg. Plasma (n = 395) and PF (n = 50) concentrations at steady state were available. A two-compartment model with first-order absorption and elimination was described. Our two-compartment model with first-order absorption and elimination produced an effective PK/PD relationship in plasma, achieving a PTA of ≥90% with MICs ranging from 0.008 to 0.12 mg/L for Candida albicans and Candida glabrata. In PF, PTAs at D8 were optimal only for a MIC of 0.008 mg/L in patients weighing 60 kg under the three dosing regimens. Among the 16 patients colonized, all MIC values were below the maximal concentration (C_max) in plasma but not in PF. PF concentrations of caspofungin were low. Simulations showed that the PTAs for Candida spp. in PF were not optimal, which might suggest a potential risk of resistance.

TEXT

Invasive fungal infections (IFIs) are of increasing concern in patients undergoing liver transplantation (LT), with an incidence ranging from 7% to 42% (1). IFIs are associated with increased lengths of stay and death (2–4). Although some of the risk factors for IFIs are now well known (4), others have emerged in recent years (5, 6).

In this population, invasive candidiasis is the most common IFI (60 to 80%) (7), mainly candidemia or intraabdominal candidiasis (8). The most common strains are Candida albicans (>50% of cases) and Candida glabrata (9–11). Echinocandins are recommended as first-line curative therapy for intraabdominal candidiasis, due to their high in vitro activity against Candida spp., their excellent safety profile, and their favorable pharmacokinetics (PK) (1, 9–12). These drugs inhibit the synthesis of β-1,3-glucan in the fungal cell wall. Moreover, the European ESCMID guidelines recommend administering echinocandins as prophylactic antifungal therapy for LT recipients (13).

The PK of antimicrobial agents in these transplant recipients may vary considerably and differ from those in the general populations (14–17). The PK of caspofungin are influenced by hepatic impairment, body weight, and hypoalbuminemia (18–20). Moreover, the efficacy of echinocandins is concentration dependent and mainly related to the ratio between the area under the unbound plasma concentration-time curve from 0 to 24 h at steady state (fAUC_{0 −24}) and the MIC of the microorganism (fAUC_0–24/MIC ratio) (21, 22). Weaker diffusion of echinocandins in the abdomen may promote Candida-resistant isolates, mainly for haploid C. glabrata, around or in digestive sites (23). Thus, abdominal candidiasis can be a hidden reservoir for the emergence of echinocandin-resistant Candida strains. Shields et al. showed that FKS mutant Candida isolates were recovered from 24% of abdominal candidiasis patients (6/25 patients) exposed to echinocandins. Mutations were associated with prolonged echinocandin exposure (P = 0.01), breakthrough infections (P = 0.03), and therapeutic failures despite source control interventions (100%) (24).

Literature is scarce regarding on the diffusion of echinocandins into peritoneal fluid (PF). In critically ill patients, the penetration of micafungin into PF is low to moderate (25), and similar results have been obtained with all three echinocandins in patients with peritonitis (26). Steady-state caspofungin concentrations in serum and PF were evaluated in only eight recipients with intraabdominal infections (26). Furthermore, no correlations could be established between PK/pharmacodynamic (PD) target attainment and clinical responses. The aim of this study was to explore the PK/PD features of caspofungin in plasma and PF from LT recipients.

RESULTS

Patients.

During the study period, 48 patients underwent LT; 20 of them received antifungal prophylactic therapy and were included. Their demographic and clinical data are detailed in Table 1. The median Model for End-Stage Liver Disease (MELD) score was 29 (interquartile range [IQR], 19.5 to 36), and the median Sequential Organ Failure Assessment (SOFA) score was 13 (IQR, 10.5 to 17.25). IFI risk factors that justified prophylactic antifungal therapy were a MELD score of >30 (n = 9), renal failure requiring or not requiring replacement therapy (n = 8), early reintervention (n = 2), fulminant hepatic failure (n = 2), choledochojejunostomy (n = 2), and multifocal colonization by Candida spp. All recipients received a dose of 70 mg on day 1 (D1), after which the median daily dose of caspofungin was normalized to body weight at 0.81 mg/kg (IQR, 0.75 to 0.89 mg/kg) for a median period of 20 days (IQR, 13 to 25.5 days).

TABLE 1.

Demographic, clinical, and laboratory characteristics of LT recipients (n = 20)

Variablea	Finding
Age (median [IQR]) (yr)	45 [40.7–50]
Gender (no. male/female [% male])	9/11 [45]
Body weight (median [IQR]) (kg)	72 [62–81]
BMI (median [IQR]) (kg/m2)	25.6 [24–29.6]
Indication for LT (no. [%])
Acute liver failure	2 [10]
Hepatocellular carcinoma	3 [15]
End-stage liver disease	15 [75]
Etiology of cirrhosis (no. [%])
Alcohol	7 [35]
Viral	3 [15]
Autoimmune	3 [15]
Cholestatic liver disorder	2 [10]
Other	3 [15]
Prognostic scores
SOFA score (median [IQR])	13 [10.5–17.25]
MELD score (median [IQR])	29 [19.5–36]
MELD score of >30 (no. [%])	9 [45]
ACLF grade (no.)
Grade 0	5
Grade 1	4
Grade 2	6
Grade 3	5
Charlson score (median [IQR])	5 [4–5.25]
Comorbidities (no. [%])
Hepatorenal syndrome	9 [45]
Hepatic encephalopathy	11 [55]
Preoperative laboratory data
PT (median [IQR]) (%)	34 [30–50]
Total bilirubin level (median [IQR]) (μmol/L)	90 [29–150]
Urea level (median [IQR]) (mmol/L)	11.1 [8.8–17.4]
Serum creatinine level (median [IQR]) (μmol/L)	128 [87–227]
ALT level (median [IQR]) (IU/L)	43 [31–95]
AST level (median [IQR]) (IU/L)	93 [60–144]
Leukocyte count (median [IQR]) (109 cells/L)	8.1 [5.3–12.3]
C-reactive protein level (median [IQR]) (mg/L)	45 [23–71]
Albuminemia (median [IQR]) (g/L)	28 [24–33]
Perioperative resuscitation fluids
Fluid infused (median [IQR]) (mL)	3,750 [3,500–4,175]
Received transfusion (no. [%])	19 [95]
Red blood cells received (median [IQR]) (units)	6 [5.5–10]
Blood loss (median [IQR]) (mL)	1,600 [900–5,100]

^aContinuous variables are presented as medians with interquartile ranges [IQR], and categorical variables are expressed as n (%). Abbreviations: ACLF, acute-on-chronic liver failure; MELD, model for end-stage liver disease; SOFA, sequential organ failure assessment; ICU, intensive care unit; RTT, renal replacement therapy; PT, prothrombin time; ALT, alanine transaminase; AST, aspartate transaminase.

The median volumes of PF drained were 1,085 mL (IQR, 589 to 1,360 mL) on D1, 1,080 mL (IQR, 589 to 1,706 mL) on D3, and 849 mL (IQR, 425 to 1,246 mL) on D8 (see Table S1 in the supplemental material). Eight patients experienced renal failure during the postoperative period, 3 of whom required replacement therapy. Two patients had serum bilirubin levels of >50 μmol/L, but no patient had a prothrombin time (PT) of <50% on D5. No patients experienced graft failure or primary nonfunction.

Caspofungin PK.

At steady state, 395 plasma and 50 PF caspofungin level values were available. Figure 1 presents the individual concentration-time curves in plasma on D1, D3, and D8 and in PF on D8. A two-compartment model with first-order absorption and elimination and an effect compartment linked to the central compartment was able to satisfactorily describe plasma and PF concentrations (see Fig. S1). For PF, the best fits and Bayesian information criterion (BIC) were obtained by estimating different input and output rate constants for the effect compartment. Adding a transit compartment between the central and effect compartments did not improve the fit. The PK parameters of this model were clearance (CL), central volume of distribution (V₁), intercompartmental clearance (Q), peripheral volume of distribution (V₂), plasma-to-PF transfer rate constant (k_eff13), and PF-to-plasma transfer rate constant (k_eff31).

FIG 1.

Concentration-time curves for caspofungin in plasma (A) and PF (B).

Residual variability was described using a proportional error model for both caspofungin plasma concentrations and caspofungin PF concentrations. Interindividual variability was retained for CL, V₁, V₂, and k_eff13. For plasma concentration modeling, allometry improved the model; for PF concentration modeling, the effects of covariates on k_eff13 did not significantly decrease the BIC. Caspofungin PK parameters were not influenced by the covariates tested. Table 2 summarizes the final population PK estimates for the model, including the relative standard errors (RSEs). All parameters were well estimated, given the low RSE values. Goodness of fit plots are presented in Fig. S2 in the supplemental material. A predicted-corrected visual predictive check of the final model showed that the 5th, 50th, and 95th percentiles of observed data were clearly included within the 90% confidence interval (CIs) of the 5th, 50th, and 95th simulated percentiles for both plasma and PF caspofungin concentrations (see Fig. S3).

TABLE 2.

Caspofungin PK parameters

Parameter	Estimate	RSE (%)a
Structural model
CL (L/h)	0.38	8.16
V1 (L)	6.24	15.8
Q (L/h)	2.58	0.2
V2 (L)	6.44	28.5
keff 13 (h−1)	0.08	13.6
keff 31 (h−1)	0.26
Statistical modelb
ωCL	0.33	18.6
ωV1	0.59	21.6
ωV2	1.07	21.9
ωkeff13	0.54	18.8
σprop1	0.36	4.4
σprop2	0.60	10

^aRSE, standard error of the estimate divided by the estimate and multiplied by 100.

^bω, coefficient of variation for between-subject variability; σ, parameters of error model.

The median plasma maximal concentration (C_max) and minimal concentration (C_min) were 9.2 mg/L (IQR, 7.2 to 11.5 mg/L) and 2.2 mg/L (IQR, 1.4 to 2.6 mg/L), respectively, at D3 and 11.2 mg/L (IQR, 8.3 to 13.4 mg/L) and 2.7 mg/L (IQR, 1.8 to 3.2 mg/L), respectively, at D8. The median caspofungin plasma AUC was 131 mg · h/L (IQR, 108 to 189 mg · h/L) at D3 and 165 mg · h/L (IQR, 122 to 204 mg · h/L) at D8. The median accumulation ratio, AUC_D8/AUC_D3, was 1.26 for plasma, while the median caspofungin PF AUC was 27 mg · h/L (IQR, 11 to 54 mg · h/L) at D8. The median PF AUC/plasma AUC ratio on D8 was 0.18 (IQR, 0.08 to 0.23).

Monte Carlo simulations and PTAs.

Monte-Carlo simulations were performed to assess the daily exposure obtained with licensed and alternative dosing regimens. The caspofungin AUC in plasma on D8 and the simulated AUCs using regimens I, II, and III are presented in Fig. 2, and regimens II and III are compared in Table 3. The probability of target attainment (PTA) values simulated at D8 are described in Fig. 3. Using the recommended regimen, the PTAs in plasma were optimal at a MIC of 0.06 mg/L for C. albicans and C. glabrata. For these two species, the PTAs in plasma were optimal at a MIC of 0.12 mg/L for patients weighing 60 or 80 kg using regimen II and for patients weighing 100 kg using regimen III. The PTAs in PF were optimal on D8 only at a MIC of 0.008 mg/L for C. albicans and C. glabrata for patients weighing 60 kg, under the three regimens. In the other cases, the PTAs were below 90%, with MICs of less than 0.12 mg/L. In the case of Candida parapsilosis, plasma PTAs were optimal at a MIC of 0.06 mg/L for patients weighing 60, 80, or 100 kg using regimen II, while regimen III achieved optimal PTAs at a MIC of 0.12 mg/L for patients weighing 100 kg. In PF, the PTAs ranged from 60% to 80% for patients whatever their weight and the regimen employed.

FIG 2.

Plasma AUC_0–24 values on D8 for our patients and after Monte Carlo simulations for two regimens, i.e., 70/70 mg and 100/100 mg (loading dose/maintenance dose). The horizontal dotted line represents the mean AUC_0–24 described for ICU patients.

TABLE 3.

PK parameters in plasma and PF simulated under regimens II and III

Regimen	Median (IQR) for:
	Plasma			PF
	AUC0-24 (mg · h/L)	Cmax (mg/L)	Cmin (mg/L)	AUC0-24 (mg · h/L)
Regimen II (70 mg/70 mg)	165 (131–210)	11 (7–15)	1.5 (0.5–2.6)	23 (10–53)
Regimen III (100 mg/100 mg)	236 (187–298)	14 (9–18)	2.2 (0.8–3.4)	33 (14–75)

FIG 3.

PTAs depending on the dosage used. (A) PTAs for Candida albicans (AUC/MIC of >25.9) in plasma (left) and PF (right) under the two regimens with body weights of 60, 80, and 100 kg. (B) PTAs for Candida glabrata (AUC/MIC of >13.5) in plasma (left) and PF (right) under the two regimens with body weights of 60, 80, and 100 kg. (C) PTAs for Candida parapsilosis (AUC/MIC of >35.5) in plasma (left) and PF (right) under the two regimens with body weights of 60, 80, and 100 kg. 70/50, 70 mg loading dose/50 mg maintenance dose; 70/70, 70 mg loading dose/70 mg maintenance dose; 100/100, 100 mg loading dose/100 mg maintenance dose.

Mycological findings.

Among the 20 recipients included, 16 were colonized by Candida spp. during the treatment with caspofungin; the species identified were C. albicans (45%), C. glabrata (35%), Candida nivariensis (1%), Candida tropicalis (1%), and Candida dubliniensis (1%). Etest MIC values ranged from 0.032 to 0.19 mg/L for all species except for C. parapsilosis (Table 4). Only two patients developed an infection, i.e., candidemia (patient P1) and peritonitis (patient P9), both due to C. parapsilosis. For patient P1, the prophylactic treatment was given for 11 days, during which colonization with C. nivariensis was found (low MICs for echinocandins: caspofungin, 0.19 mg/L; micafungin, 0.016 mg/L; anidulafungin, 0.012 mg/L). A blood culture with C. parapsilosis (MICs: micafungin, 1 mg/L; anidulafungin, 0.75 mg/L) appeared at D11 of the prophylaxis. This blood culture was treated with liposomal amphotericin B for 14 days. Colonization screening was continued during curative treatment, with positive colonization results for C. nivariensis and C. krusei and a positive abdominal collection with C. nivariensis (MICs: micafungin, 0.016 mg/L; anidulafungin, 0.008 mg/L). No follow-up monitoring of the patient after completion of curative treatment was performed. For patient P9, prophylaxis was prescribed for 21 days, with weekly colonization monitoring results being negative. At D22, the PF sample was positive, as were the patient's colonization screening results for C. parapsilosis, with high MICs for echinocandins (micafungin, 1 mg/L; anidulafungin, 2 mg/L). The curative treatment instituted was fluconazole for 21 days at D23; at D24, a blood culture was positive for C. parapsilosis, with high MICs for echinocandins (micafungin, 1 mg/L; anidulafungin, 1.5 mg/L). At D45, the patient's colonization screening results were negative.

TABLE 4.

Etest MICs for caspofungin measured in patients

Patient	Day of administration	Candida species	MIC (mg/L)
P1	D8	C. nivariensis	0.19
	D10	C. parapsilosis (candidemia)	0.75
P2	D4	C. albicans	0.064
P3	D2	C. glabrata	0.19
	D9	C. glabrata	0.19
P5	D2	C. glabrata	0.25
	D9	C. glabrata	0.19
P6	D0	C. glabrata	0.125
	D6	C. glabrata	0.19
P7	D10	C. glabrata	0.125
P8	D6	C. glabrata	0.125
		C. albicans	0.047
P10	D7	C. albicans	0.125
P13	D1	C. albicans	0.094
		C. glabrata	0.125
P18	D4	C. albicans	0.032
		C. glabrata	0.094
P19	D4	C. albicans	0.032
		C. tropicalis	Not available

DISCUSSION

This study reports the development of a population PK model for caspofungin in plasma and PF from LT recipients. This model, which included two compartments with first-order absorption and elimination and an effect compartment linked to the central compartment, was successful in simulating different caspofungin dosing regimens. Thus, this model makes it possible to predict the probability of reaching the therapeutic objective.

We were able to report caspofungin PK parameters in plasma that were higher than those published for critically ill patients (14–17, 26). Our study also supported findings regarding lower PF concentrations of caspofungin. Moreover, the simulations showed that the PTAs for Candida spp. in PF were not optimal.

In plasma, the AUCs obtained with our model were higher than those described for intensive care unit (ICU) patients, i.e., 130.9 mg · h/L (IQR, 107.7 to 189.0 mg · h/L) versus 88.7 mg · h/L (IQR, 72 to 98 mg · h/L) and 78 mg · h/L (IQR, 61 to 129 mg · h/L) at D3 and 164.9 mg · h/L (IQR, 121.9 to 204.4 mg · h/L) versus 107.2 mg · h/L (IQR, 90 to 125 mg · h/L) at D8 (27, 28). It should be noted that only 6 patients received the high-dose (70/70 mg) regimen because of body weight of >80kg, and this did not explain the higher AUC found in our cohort. Although other authors (the CASPOLOAD study) proposed a 140-mg loading dose for 24 h in ICU patients in order to obtain an AUC of 80 mg · h/L (29), our data showed that this was not necessary for posttransplant patients. Consequently, the CL estimated in our model was lower than that calculated for ICU patients (21, 28, 30).

Several hypotheses can be advanced to explain this difference in CL. First, in ICU patients, caspofungin PK parameters do not remain stable over the first 3 days of treatment, because of the increase in CL and V between the first and third doses (31, 32). Indeed, PK parameters are likely to be influenced by sepsis or organ failure (33). The posttransplant period also accentuates the PK instability of caspofungin. Second, higher AUC and lower CL values might be linked to impaired liver function during the first days after transplantation. In our cohort, impaired liver function (PT, aspartate transaminase [AST] level, alanine transaminase [ALT] level, and total bilirubin level) increased PK parameter stability failure and metabolic failure at D8 posttransplant. Furthermore, in ICU patients, there were a bilirubin-driven negative correlation with the caspofungin rate constant (34). No patients developed primary nonfunction or hepatic graft dysfunction that could explain these results. Third, the concomitant administration of caspofungin and immunosuppressive drugs might explain an increase of the caspofungin AUC. Caspofungin is a poor substrate for cytochrome P450 (CYP) enzymes, and the coadministration with powerful CYP inducers or inhibitors could result in impaired clearance (35). In our cohort, only 6/20 recipients received tacrolimus on D1, and all patients received tacrolimus on D8. It is also of note that caspofungin PK values were not affected by coadministration with tacrolimus (36). Finally, caspofungin binds very strongly to albumin (22, 30). Some authors reported a threshold of 25 g/L, below which echinocandin CL increases (19, 33), while others reported that albumin concentrations did not significantly affect the caspofungin PK in ICU patients (31). In our cohort, only 3 patients presented with such hypoalbuminemia (<25 g/L) at D1 and just 1 patient at D3. The absence of severe hypoalbuminemia partly contributed to the reduction in CL.

In PF, the interindividual variability of caspofungin AUC values was very high in our cohort at D1, D3, and even D8. This was due to the variability of PK in plasma, due to fluid shifts resulting in concentration variability and probably the presence of a drain facilitating the direct elimination of PF. It was recently reported that, in surgical patients, PF caspofungin concentrations ranged from 0.2 to 0.46 mg/L, below the concentration for the selection of resistant mutants (26). Few studies concerning intraperitoneal diffusion are available in ICU patients regarding the use of micafungin and anidulafungin (23, 37, 38). At 100 mg/day micafungin, PK/PD targets were achieved on D1 in plasma and PF (approximately 30 to 40%) in ICU patients with severe peritonitis. At conventional doses, PF concentrations of anidulafungin (about 30% of the plasma concentrations) exceeded the MIC for usual Candida spp. (39). The results were similar in our study, in which caspofungin had low intraperitoneal diffusion. Since caspofungin is highly protein bound and its antifungal activity is concentration dependent, the fAUC/MIC ratio is a better predictor of PK/PD characteristics (21, 30). Our model shows a good PK/PD relationship in plasma, reaching a PTA of ≥90% and MIC values ranging from 0.008 to 0.06 mg/L for C. albicans, C. glabrata, and C. parapsilosis using conventional dosing regimens. Regimens II and III increase the PTA for MICs of 0.25 mg/L for C. albicans and C. glabrata and 0.15 mg/L for C. parapsilosis. Our results were very close to those obtained by Yang et al., who concluded that caspofungin is a good choice to achieve PK/PD targets in ICU patient plasma for C. albicans and C. glabrata (40). In PF, however, a PTA of ≥90% was observed only for a MIC of 0.008 mg/L, whatever the body weight and Candida spp. The targets were not achieved despite the low elimination half-life of caspofungin, compared to other echinocandins. This might promote the resistance of Candida spp. to caspofungin, as shown in the study by Shields et al., which demonstrated that abdominal candidiasis is a hidden reservoir of echinocandin resistance, and in the study by Prigent et al., in which resistance to echinocandins appeared in 8% of treated LT patients within 1 month (23, 24). In the present study, more than one-half of the patients were colonized by various yeasts during the course of prophylaxis (Table 4), essentially due to C. glabrata and C. albicans. Only the 2 patients colonized with C. parapsilosis developed an infection, including peritonitis for 1 patient and positive blood culture results for both patients. These infections were probably due to high echinocandin MICs for C. parapsilosis, which may require an alternative choice of treatment for C. parapsilosis infections. Increasing the dose did not restore the PTA in PF (37, 38, 41). No yeast resistance was detected after this prophylactic treatment with caspofungin, perhaps because the 15-day prophylaxis is not sufficient to induce resistance. Limitations of our study concern the small number of patients included in a single center. Furthermore, our study included a heterogeneous population with several covariates. PK data are not necessarily relevant to treatment of peritonitis, in which inflammation is present, or of intraabdominal abscesses, which are different clinicopathological sites. The PK of caspofungin in LT recipients show both intraindividual and interindividual variability. Nevertheless, the PK of this molecule seem to be higher than those observed in critically ill patients.

To our knowledge, this is the first study to have involved extensive PK scheduling on separate days in plasma and PF from LT recipients. The caspofungin PK parameters were not influenced to a relevant extent by covariates. We conclude that LT patients do not require higher doses, compared with other reference groups, regarding plasma PK parameters. We also strongly recommend further debate regarding the peritoneal diffusion of caspofungin and the risks of the development of secondary resistance to this antifungal drug.

MATERIALS AND METHODS

Study design and population.

This was an open-label, phase IV, monocentric, prospective study in LT recipients admitted to the liver ICU at Henri Mondor University Hospital between January 2017 and September 2017. Antifungal prophylactic therapy with caspofungin (MSD, France) was initiated for patients at increased risk of IFIs (13). These patients received caspofungin via intravenous infusion over 1 h, with a loading dose of 70 mg and then 50 mg per day (or 70 mg/day if the recipient weighed >80 kg).

The study protocol was approved by the local ethics committee (Comité de Protection des Personnes [Boulogne-Billancourt, France], approval number VS16-345); written information was given to each patient, and their consent was obtained. The database was officially registered with the French Data Protection Authority (Commission Nationale Informatique et Liberté, accession number 1699340).

The study inclusion criteria were age of >18 years, undergoing LT, and with increased risk of IFI. The exclusion criteria were previous antifungal therapy (including caspofungin), a diagnosis of IFI before transplantation, and development of primary nonfunction or hepatic graft dysfunction (42). The following demographic and clinical data were recorded for all included patients: age, gender, weight, body mass index (BMI), indication for LT, underlying comorbidities, risk factors for IFIs, immunosuppressive drugs and other treatments, severity scores (MELD score, SOFA score, and Acute-On-Chronic Liver Failure [ACLF] grade), biological parameters (PT, serum albumin level, urea level, serum creatinine level, total bilirubin level, ALT level, AST level, leukocyte count, and C-reactive protein level), data regarding resuscitation fluids (intraoperative fluids infused, transfusion, number of red blood cells, and blood loss), and data regarding organ support measures and outcomes.

Collection of plasma and PF samples.

Plasma and PF samples were collected simultaneously just before the start of caspofungin treatment and then 1, 2, 4, 8, 12, and 24 h thereafter, after the first dose (on D1) and on D3 and D8 of treatment. A drain with a closed suction device (in situ as part of standard clinical management) was used to collect PF samples, while blood samples were collected in heparinized tubes. All blood and PF samples were centrifuged at 3,000 × g for 15 min at 4°C and then stored at −80°C until analysis.

Sample analysis.

Caspofungin concentrations in plasma and PF samples were determined using a validated high-performance liquid chromatography method with fluorescence detection, with excitation and emission wavelengths of 220 and 304 nm, respectively (FLD-3100; Thermo Fisher Scientific). Analytes were separated on a Kinetex C₁₈ column (250 by 4.6 mm, 5 μm) at 40°C using an isocratic mobile phase (0.1% trifluoroacetic acid in water/acetonitrile, 55:45 [vol/vol] [pH 2.5]) at 1 mL/min. The internal standard was micafungin. Extraction consisted of protein precipitation using cold acetonitrile. The calibration curve was linear from 0.5 to 30 mg/L. The lower limit of quantification was 0.5 mg/L. Intraday precision ranged from 2.0% to 10.4% for plasma samples and from 6.4% to 9.5% for PF samples, while interday precision ranged from 1.2% to 7.4% for plasma samples and from 1.0% to 7.1% for PF samples. Accuracy ranged from 99.0% to 92.6% for plasma and PF samples, respectively.

Population PK model.

Caspofungin data were analyzed using nonlinear mixed-effect modeling software (MONOLIX version 2019R2), together with the stochastic approximation expectation-maximization (SAEM) algorithm. In a first step, caspofungin plasma concentrations were modeled and the PK parameters were estimated. These parameters were then fixed in order to investigate the caspofungin PF PK modeling. In a final step, all plasma and PF parameters were estimated. Various structural models were tested, including one- or two-compartment distribution with first-order absorption and elimination rate constants. The combination of both an additional compartment linked to the central compartment by a first-order process and an effect compartment was tested. Several models were explored, namely, (i) the effect compartment was linked to the central compartment by a first-order process with the same or different input and output constants, (ii) a transit compartment was inserted between the central and effect compartments, and (iii) the effect compartment was linked to the peripheral compartment. Categorical covariates were tested as follows:

$${\theta }_{\text{i}}={\theta }_{\mathit{pop}}\text{X}{\theta }^{\mathit{COV}}$$

where θ_i is the individual parameter for the ith patient, θ_pop is the typical value of the parameter, θ^COV is the covariate parameter, and COV is category 0 or 1. Continuous covariates were associated with PK parameters as follows:

$${\theta }_i={\theta }_{\mathit{pop}}\times {\left(\frac{\mathit{Co}{v}_i}{\left(\mathit{Median}\left(\text{C}\mathit{ov}\right)\right)}\right)}^{\mathit{PWR}}$$

where Cov_i is the covariate value for the ith patient and PWR is the exponent. An adult value of 70 kg was taken as the reference value for body weight, and the exponents (PWR) were 0.75 for CL and Q and 1 for V, according to the allometric rule. Other covariates were also tested, including demographic characteristics (age, gender, weight, and BMI), hepatic function findings (albumin level, PT, and total bilirubin level), renal function findings (urea and serum creatinine levels), inflammatory parameters (leukocyte count and C-reactive protein level), and severity scores (MELD score and SOFA score).

The effect of a covariate on a structural parameter was retained if it caused a decrease in the BIC and/or reduced the corresponding between-subject variability with a P value of <0.05 using a likelihood ratio test. Diagnostic graphs were used to evaluate the goodness of fit. Concentration profiles were simulated and compared with the observed data using a visual predictive check in order to validate the model.

Monte Carlo simulations of PK and target attainment.

The PK model was used to perform a Monte Carlo simulation of 10,000 individuals achieving steady-state AUC values. The predicted exposure to caspofungin was assessed in plasma and PF on D8. Simulated fAUC_0-24 values were obtained by means of covariate distributions similar to those used for our population and assuming 97.0% plasma protein binding. AUC values were simulated under three dosing regimens, as follows: regimen I, 70/50 mg/day (loading dose/maintenance dose); regimen II, 70/70 mg/day; regimen III, 100/100 mg/day. Estimated fAUC_0-24/MIC ratios were based on the MICs for Candida spp. ranging from 0.008 to 8 mg/L.

A Monte Carlo simulation was performed to calculate the PTA, defined as the percentage of subjects who achieved the requisite PD exposure. Our study used preclinical targets determined in neutropenic murine models; the fAUC/MIC ratios were 25.9, 13.5, and 35.5 for C. albicans, C. glabrata, and Candida parapsilosis, respectively (27). A PTA of ≥90% was considered to be optimal. The body weight covariate was evaluated for 60, 80, and 100 kg.

Mycological analysis.

All patients were monitored weekly for Candida colonization and screened, if necessary, for Candida infections in blood cultures and at other sites as a function of clinical signs. As part of this routine surveillance for Candida colonization, swab samples were taken systematically from five superficial sites (mouth, nose, axillary surface, inguinal fold, and anus) on the day of admission to undergo LT and then once a week thereafter until discharge from the ICU or death. Colonization was defined as the presence of Candida species isolates at least one of the monitored sites.

Swab samples, PF samples, and clinical samples collected in the event of suspected infections were cultured on CHROMagar plates (Becton Dickinson) and incubated for at least 48 h at 37°C. Candida isolates were identified by matrix-assisted laser desorption ionization–time of flight mass spectrometry (MALDI-TOF MS) (Microflex; Brucker) using the MALDI BioTyper database version 3.0. All isolates were initially stored at −20°C on cryobeads (bioMérieux).

Caspofungin Etest strips (bioMérieux) were used to screen for the susceptibility of Candida isolates at sites of infection and at the anus, the abdominal site considered closest to the peritoneal site (23); these isolates were tested according to the manufacturer’s instructions. MIC values were read after 48 h of incubation at 37°C. An 80% inhibitory endpoint was applied when determining these MIC values. Because there are currently no Etest-specific breakpoints, Etest MIC values were interpreted according to CLSI breakpoints (43).

Statistical analysis.

Continuous data are expressed as medians with IQRs, and categorical variables are expressed as numbers and percentages.