Advertisement

Evaluation of first-line therapies for the treatment of candidemia in ICU patients: A propensity score analysis

Open AccessPublished:January 23, 2020DOI:https://doi.org/10.1016/j.ijid.2020.01.037

      Highlights

      • Echinocandins recommended for candidemia and fluconazole, as an alternative.
      • Emergence of Candida resistance to echinocandins.
      • Local guidelines for candidemia according to clinical practice and local epidemiology.
      • Fluconazole suggested to be a reasonable alternative to echinocandins in ICU.
      • To be incorporated in local guidelines through antifungal stewardship activities.

      Abstract

      Background

      Candidemia is a major cause of mortality in the intensive care unit (ICU). According to the Infectious Diseases Society of America (IDSA), an echinocandin is recommended as initial therapy and fluconazole as an alternative. In a context of echinocandin resistance development, the question arising is whether azoles are a suitable alternative to echinocandins for the treatment of candidemia in critically ill patients.

      Methods

      A 3-year (2015–2017) retrospective multicentric cohort study was conducted. Adult patients with a diagnosis of candidemia during the ICU stay and treated with echinocandins or azoles were included. Demographic, clinical data, mycological data, and antifungal treatments were collected. Kaplan–Meier survival analysis, univariate analysis, and a multivariate logistic regression analysis using a propensity score with the inverse probability of treatment weighting method were performed.

      Findings

      Seventy-nine patients (n = 79) were analyzed. Treatment success, as well as survival on day 90 (Kaplan–Meier survival analysis, log rank test, p = 0.542), were comparable between patients who received echinocandins (caspofungin (n = 47)) or azoles (fluconazole (n = 29) or voriconazole (n = 3)). A multivariable analysis demonstrated that higher SOFA score on the day of candidemia diagnosis and absence of adequate Candida source control were independently associated with a greater risk of 90-day mortality, whereas azoles treatment was not associated with an excess 90-day mortality.

      Interpretation

      This study confirms that the use of azoles recommended for candidemia, mostly fluconazole, as a first-line therapy is a reasonable alternative to caspofungin for ICU patients in our institution. This needs to be included in local guidelines through antifungal stewardship programs.

      Keywords

      Introduction

      Invasive fungal infections are responsible for over 1.6 million deaths each year (
      • Almeida F.
      • Rodrigues M.L.
      • Coelho C.
      The still underestimated problem of fungal diseases worldwide.
      ). In intensive care units (ICU), invasive candidiasis predominates with an estimated incidence of 2.1 to 6.7 per 1000 admissions (
      • Colombo A.L.
      • de Almeida Júnior J.N.
      • Slavin M.A.
      • Chen S.C.A.
      • Sorrell T.C.
      Candida and invasive mould diseases in non-neutropenic critically ill patients and patients with haematological cancer.
      ). Candidemia, known as the most common manifestations of invasive candidiasis, has an attributable mortality up to 40% and is recognized as a major cause of mortality in the ICU (
      • Kullberg B.J.
      Arendrup MC: invasive candidiasis.
      ,
      • Baldesi O.
      • Bailly S.
      • Ruckly S.
      • Lepape A.
      • L’Heriteau F.
      • Aupee M.
      • et al.
      ICU-acquired candidaemia in France: epidemiology and temporal trends, 2004-2013 – a study from the REA-RAISIN network.
      ).
      According to the Clinical Practice Guideline for the Management of Candidiasis by the Infectious Diseases Society of America (IDSA) (
      • Pappas P.G.
      • Kauffman C.A.
      • Andes D.R.
      • Clancy C.J.
      • Marr K.A.
      • Ostrosky-Zeichner L.
      • et al.
      Clinical practice guideline for the management of candidiasis: 2016 Update by the Infectious Diseases Society of America.
      ), an echinocandin (caspofungin, micafungin, or anidulafungin) is recommended as initial therapy. Fluconazole is an acceptable alternative to an echinocandin as initial therapy in selected patients, including those who are not critically ill and who are considered unlikely to have a fluconazole-resistant Candida species (
      • Pappas P.G.
      • Kauffman C.A.
      • Andes D.R.
      • Clancy C.J.
      • Marr K.A.
      • Ostrosky-Zeichner L.
      • et al.
      Clinical practice guideline for the management of candidiasis: 2016 Update by the Infectious Diseases Society of America.
      ). Voriconazole is effective for candidemia and offers little advantage over fluconazole as initial therapy for patients who have fluconazole-resistant isolates of C. krusei, C. guillermondii or C. glabrata that are voriconazole-susceptible (
      • Pappas P.G.
      • Kauffman C.A.
      • Andes D.R.
      • Clancy C.J.
      • Marr K.A.
      • Ostrosky-Zeichner L.
      • et al.
      Clinical practice guideline for the management of candidiasis: 2016 Update by the Infectious Diseases Society of America.
      ). Interestingly, despite favorable characteristics of echinocandins compared to azoles including fungicidal activity, broad spectrum, safety profile, and few drug interactions (
      • Bassetti M.
      • Poulakou G.
      • Kollef M.H.
      The most recent concepts for the management of bacterial and fungal infections in ICU.
      ), azoles are still widely used in ICU as demonstrated by a prospective observational cohort study that showed echinocandin accounting for 63.4% of the prescriptions and azoles for 34.5% in ICU (
      • AmarCAND2 study group
      • Leroy O.
      • Bailly S.
      • Gangneux J.P.
      • Mira J.P.
      • Devos P.
      • et al.
      Systemic antifungal therapy for proven or suspected invasive candidiasis: the AmarCAND 2 study [Internet].
      ). Then, the use of echinocandins versus azoles in ICU patients is still a matter of debate. This is particularly true in a context of emergence of C. auris and Candida resistance to echinocandins (
      • Arendrup M.C.
      • Perlin D.S.
      Echinocandin resistance: an emerging clinical problem?.
      ). Importantly, echinocandins resistance in C. glabrata can be associated with cross-resistance to azoles thus leading to multi-drug resistant strains (
      • Arendrup M.C.
      • Perlin D.S.
      Echinocandin resistance: an emerging clinical problem?.
      ). This raises the question of the large use of echinocandins to avoid the development of resistant strains. Moreover, the demonstration that most of Candida isolates recovered in the ICU were susceptible to fluconazole gives the opportunity to use azoles for candidemia treatment of patients admitted in the ICU (
      • Baldesi O.
      • Bailly S.
      • Ruckly S.
      • Lepape A.
      • L’Heriteau F.
      • Aupee M.
      • et al.
      ICU-acquired candidaemia in France: epidemiology and temporal trends, 2004-2013 – a study from the REA-RAISIN network.
      ,
      • Bassetti M.
      • Poulakou G.
      • Kollef M.H.
      The most recent concepts for the management of bacterial and fungal infections in ICU.
      ).
      In this context, the question arising is whether azoles, including fluconazole and voriconazole, are a suitable alternative to echinocandins for the treatment of candidemia in critically ill patients with the aim of implementing local guidelines as recently suggested (
      • Wang Y.
      • McGuire T.M.
      • Hollingworth S.A.
      • Dong Y.
      • Van Driel M.L.
      Antifungal agents for invasive candidiasis in non-neutropenic critically ill adults: what do the guidelines recommend?.
      ). We thus conducted a retrospective multicenter cohort study in Lyon University Hospital to investigate the local epidemiology of Candida and antifungal treatments used in patients suffering from candidemia in ICU. This study was part of an antifungal stewardship program implemented in our institution whose major aim is to contain resistance through the optimal use of antifungals on an individual patient basis (
      • Pappas P.G.
      • Lionakis M.S.
      • Arendrup M.C.
      • Ostrosky-Zeichner L.
      • Kullberg B.J.
      Invasive candidiasis.
      ).

      Materials and methods

      Study design

      This study was a 3-year (2015–2017) retrospective multicenter cohort study using Hospital databases. Six ICUs including medical (two units), surgical (one unit), and medical surgical (three units) ICUs were involved and located in four medical centers. Adult patients >19 years-old with a diagnosis of candidemia during the ICU stay and treated with echinocandins or azoles were included in the study. Candidemia was defined as at least one blood culture positive for Candida. The date of candidemia was the day of the first blood culture positive for Candida. There was no age limit for inclusion. Neutropenic patients, patients who did not receive antifungals, patients with a fungal co-infection, patients who received antifungal therapy within 7 days before candidemia diagnosis, and patients who received liposomal amphotericin b or multiple antifungal agents as first-line therapy were not included. Patients were categorized into two groups: patients receiving first-line therapy with echinocandins, i.e., caspofungin, and patients receiving azoles, i.e. fluconazole or voriconazole. First-line therapy was defined as the first antifungal agent administered at the onset of candidemia regardless of treatment duration. During the study period, major diagnostic, therapeutic, and infection control standards remained unchanged. The study was registered in Clinicaltrials.gov (NCT03799172).

      Ethical approval

      All procedures performed in studies involving human participants were in accordance with the ethical standards of our institutional research committee and with The Code of Ethics of the World Medical Association (Declaration of Helsinki). The study was approved by the institutional Ethics Committee (N°19–37). Due to the retrospective nature of the study, formal consent was not required. Electronic records were under the auspice of the French National Committee for Data Protection and Freedom of Information.

      Data collection

      The following data were collected: age, gender, main underlying disease, previous abdominal surgery within 30 days before candidemia, dates of ICU admission and discharge, dates of hospital admission and discharge, SAPSII score on ICU admission, SOFA score on the day of candidemia, septic shock (clinically identified by a vasopressor requirement) on the day of candidemia, prescription of an antifungal within 30 days before candidemia diagnosis, antifungal regimen (molecule, dose, duration), occurrence of complications (endocarditis, thrombophlebitis, endophthalmitis), adequate source control of Candida including removal of central venous catheter or surgical procedures to drain abscesses, mycological results (date of positive blood culture, Candida species, susceptibility profile, and if applicable, date of negative blood culture), and all cause hospital mortality at the end of antifungal treatment, on day 30, and on day 90. To ensure reproducibility and completeness of data extraction, an Excel spread sheet (Microsoft Corp., Redmond, WA, USA) compiling all variables to be extracted was used. Clinicians were in charge of clinical data collection and pharmacists, of biological data, as well as collection of antifungal regimen. Data extraction was double-checked by ALB. Disagreements over data extraction were resolved by discussion. Provided data were centrally checked by an independent operator for completeness, plausibility, and integrity before synthesis.

      Evaluation of outcome

      The primary outcome variable was all cause hospital mortality on day 90. The secondary outcome was treatment success on day 30. Treatment success was assessed according to Mycoses Study Group and European Organization for Research and Treatment of Cancer Consensus Criteria (
      • Segal B.H.
      • Herbrecht R.
      • Stevens D.A.
      • Ostrosky-Zeichner L.
      • Sobel J.
      • Viscoli C.
      • et al.
      Defining responses to therapy and study outcomes in clinical trials of invasive fungal diseases: mycoses Study Group and European Organization for Research and Treatment of Cancer consensus criteria.
      ). Treatment success was defined as a complete response if the following two criteria were fulfilled: survival and resolution of all attributable symptoms and signs of disease, and mycological success (documented clearance of pathogen from the blood). Treatment success was defined as a partial response if the following two criteria were fulfilled: survival and improvement of attributable symptoms and signs of disease, and documented clearance of blood. In case of persistent or recurrent fever despite clearance of blood, this situation should equate with a complete response (
      • Segal B.H.
      • Herbrecht R.
      • Stevens D.A.
      • Ostrosky-Zeichner L.
      • Sobel J.
      • Viscoli C.
      • et al.
      Defining responses to therapy and study outcomes in clinical trials of invasive fungal diseases: mycoses Study Group and European Organization for Research and Treatment of Cancer consensus criteria.
      ,
      • Reboli A.C.
      • Rotstein C.
      • Pappas P.G.
      • Chapman S.W.
      • Kett D.H.
      • Kumar D.
      • et al.
      Anidulafungin versus fluconazole for invasive candidiasis.
      ). Treatment failure included stable response, progression of disease, and death. Stable response was defined as survival and minor or no improvement in attributable symptoms and signs of disease, and persistent isolation of Candida species from blood. Progression of disease was defined as persistent isolation of Candida species from blood associated with worsening of clinical symptoms or signs of disease (e.g., septic shock, progression of hematogenous cutaneous candidiasis).

      Statistical analysis

      Categorical variables were analyzed with Chi-square test or with Fisher’s exact test, as appropriate. Continuous variables were analyzed by Mann–Whitney U test. A univariable analysis of groups, followed by a univariable analysis for 90-day mortality including potentially relevant variables for our purpose, was performed. To face the lack of randomisation, a propensity score was calculated for each patient including the most significant (p < 0.2) and/or relevant variables between echinocandin and azoles groups (i.e. age, diabetes mellitus, cirrhosis, solid organ transplantation (SOT), solid cancer, hematological cancer, chronic heart failure, abdominal surgery within 30 days of candidemia, LOS in ICU before candidemia, SAPSII, SOFA, septic shock and mechanical ventilation). Model discrimination was analyzed by evaluating the area under the Receiver Operating Characteristic (ROC) curve: the logistic regression model used for the propensity score calculation gave an area under the ROC curve of 0.848 (95% CI 0.766–0.931).
      A multivariable logistic regression analysis of 90-day mortality was performed to adjust for differences in the most relevant and significant variables (p < 0.05) between alive and deceased patients at day 90 (i.e., SOT, SOFA on the day of candidemia, time elapsed to treatment initiation, adequate Candida source control, first-line antifungal). Factors contributing to multicollinearity (i.e. SAPSII, septic shock) were excluded from the multivariable analysis. The multivariable model was weighted using the Inverse Probability of Treatment Weighing (IPTW) method based on the propensity score. Adjusted odds ratios were calculated with their respective 95% CIs. Kaplan–Meier survival analyses were performed to show the relationship between therapeutic strategies and 90-day mortality; the differences in survival were assessed using the log-rank test (Mantel–Cox). R 3.5.2 software (R Foundation for Statistical Computing, Vienna, Austria) was used for propensity score and multivariable analyses, and SPSS V21.0 software (IBM SPSS, Chicago, USA), for Kaplan–Meier survival analyses. P-value < 0.05 was considered statistically significant.

      Results

      Baseline characteristics

      A total of 104 patients had a candidemia during their ICU stay from 2015 to 2017 (Figure 1) corresponding to a candidemia incidence of 3.5/1000 and 4.6/1000 admissions, in medical and surgical ICUs respectively. Of these, 79 patients were included in the study. Forty-seven (n = 47) patients received an echinocandin (caspofungin) and 32 received azoles (fluconazole n = 29 and voriconazole n = 3) as first-line therapies. Twenty-five patients were not included: 13 patients did not receive any antifungal treatment, six patients were treated with antifungals before candidemia, five patients received an antifungal other than azoles or echinocandin, and one patient had a co-infection with Aspergillus. Baseline characteristics were comparable for both echinocandin and azoles-treated patients, except for previous abdominal surgery, which was more prevalent in the echinocandin group, and length of stay (LOS) in ICU before candidemia, which was longer in echinocandin group compared to azoles group (Table 1).
      Table 1Baseline characteristics and outcomes according to antifungal treatment.
      CharacteristicsEchinocandins

      (n = 47)
      Azoles

      (n = 32)
      p
      Clinical features
       Age (years)62.3 ± 10.963.0 ± 15.30.473
       Male sex, n (%)

       Medical ICU
      36 (76.6)

      15 (31.9)
      24 (75.0)

      14 (43.8)
      0.870

      0.284
       Diabetes mellitus, n (%)18 (38.3)6 (18.8)0.064
       Cirhhosis, n (%)7 (14.9)3 (9.4)0.731
       Necrotizing pancreatitis, n (%)3 (6.4)1 (3.1)0.643
       Chronic heart failure, n (%)6 (12.8)9 (28.1)0.087
       Chronic kidney disease, n (%)10 (21.3)6 (18.8)0.783
       Solid cancer, n (%)8 (17.0)10 (31.3)0.139
       Solid organ transplantation, n (%)5 (10.6)3 (9.4)1
       Hematological cancer, n (%)8 (17.0)2 (6.3)0.188
       Glucocorticoids >1 mg/kg/j, n (%)9 (19.1)7 (21.9)0.767
       SAPSII on ICU admission58.0 ± 18.055.0 ± 20.90.379
       SOFA score on the day of candidemia8.1 ± 4.57.9 ± 5.30.830
       Septic shock on the day of candidemia, n (%)42 (89.4)28 (87.5)0.798
      Clinical course
       Previous abdominal surgery, n (%)18 (38.3)4 (12.5)0.012
       Previous antifungal, n (%)10 (21.3)2 (6.3)0.067
       LOS in ICU before candidemia (days)8 [0–102]1.5 [0–66]0.030
       LOS in ICU (days)
      All patients25.0 [1–152]11.0 [2–92]0.012
      Alive38.0 [4–152]14.5 [5–92]0.239
      Dead24.0 [1–142]11.0 [2–45]0.045
       LOS in Hospital (days)
      All patients35.0 [1–224]26.0 [2–144]0.192
      Alive61.0 [13–224]81.0 [15–144]0.575
      Dead25.0 [1–142]12.5 [2–45]0.067
      Mycological results
      C. albicans, n (%)17 (36.2)15 (46.9)0.447
      C. glabrata, n (%)13 (27.6)8 (25.0)0.793
      C. tropicalis, n (%)7 (14.9)5 (15.7)1
      C. parapsilosis, n (%)2 (4.3)2 (6.2)1
      C. krusei, n (%)3 (6.4)1 (3.1)0.643
      C. kefyr, n (%)1 (2.1)1 (3.1)1
       Multiple Candida4 (8.5)0 (0.0)0.142
      Candida with decreased susceptibility, n (%)16 (34.0)9 (28.1)0.892
       Time to negative blood culture (days)4.0 [1–32]3.0 [1–25]0.878
      Treatment
       Time elapsed to treatment initiation (days)2 [0–10]2 [0–6]0.499
       Duration of first-line antifungal (days)7.0 [1–190]5.5 [2–37]0.441
       Total duration of antifungals (days)14.5 [2–541]11.0 [2–47]0.336
       Patients who switched to oral, n (%)3 (6.4)3 (9.4)0.622
       Adequate Candida source control, n (%)39 (83)21 (65.6)0.076
      Outcomes
       30-day mycological success, n (%)43 (91.5)26 (81.3)0.179
       30-day complete response, n (%)25 (53.2)13 (40.6)0.272
       Mortality at the end of antifungal treatment, n (%)17 (36.2)17 (53.1)0.135
       30-day mortality, n (%)22 (46.8)17 (53.1)0.581
       90-day mortality, n (%)24 (51.1)18 (56.3)0.650
       Time to death (days)12.0 [1–148]15.0 [3–201]0.698
       Complications, n (%)4 (8.5)6 (18.8)0.307
      Data are number (%), mean ± standard deviation, or median [range], as appropriate. ICU: intensive care unit; LOS: length of stay; SAPSII: simplified acute physiologic score; SOFA: sepsis-related organ failure assessment.

      Mycological findings

      The distribution of Candida species and the proportion of Candida with decreased susceptibility were similar in the two groups (Table 1). The most common species causing candidemia were C. albicans and C. glabrata accounting respectively for 40.5% (32/79) and 26.6% (21/79) (Table 1). Multiple Candida isolates were recovered for 8.5% (4/47) patients in echinocandin group, including a co-infection of C. albicans/C. kefyr, C. albicans/C. glabrata, C. krusei/C. kefyr, and C. dubliniensis/C. glabrata (Table 1). The median times to negative blood culture for overall Candida, as well as C. albicans, were similar for echinocandin and azoles groups (p = 0.878 and p = 0.546, respectively for overall Candida and C. albicans) (Table 1).

      Antifungal treatment and length of stay

      All first-line therapies were considered adequate according to Candida antifungal susceptibility profile. Durations of first-line and overall antifungal treatment were comparable between echinocandin and azoles groups, as well as time elapsed to treatment initiation and adequate control of Candida source including removal of central venous catheter and surgical procedures to drain abscesses (Table 1). As expected, duration of antifungal treatment was longer in alive patients compared to deceased patients (Table 2). LOS in ICU before the onset of candidemia was longer for patients treated with echinocandin compared to azoles, as demonstrated by medians of 8 and 1.5 days respectively (p = 0.030)) (Table 1). LOS in ICU after the onset of candidemia was prolonged for patients who were treated with echinocandin, but were deceased on day 90, compared to azoles (25.0 ± 26.5 days and 13.3 ± 12.3 days (p = 0.045), respectively for caspofungin and azoles groups) (Table 1).
      Table 2Univariable analysis of factors associated with 90-day mortality.
      CharacteristicsDead

      (n = 42)
      Alive

      (n = 37)
      p
      Clinical features
       Age (years)62.8 ± 14.759.6 ± 13.20.067
       Male sex, n (%)

       Medical ICU
      33 (78.6)

      17 (40.5)
      27 (73.0)

      12 (32.4)
      0.561

      0.744
       Diabetes mellitus, n (%)13 (31.0)11 (29.7)0.906
       Cirhhosis, n (%)8 (19.0)2 (5.4)0.094
       Necrotizing pancreatitis, n (%)1 (2.4)3 (8.1)0.336
       Chronic heart failure, n (%)9 (21.4)6 (16.2)0.556
       Chronic kidney disease, n (%)7 (16.7)9 (24.3)0.398
       Solid cancer, n (%)11 (26.2)7 (18.9)0.442
       Solid organ transplantation, n (%)1 (12.4)7 (18.9)0.023
       Hematological cancer, n (%)6 (14.3)4 (10.8)0.643
       Glucocorticoids >1 mg/kg/j, n (%)8 (19.0)8 (21.6)0.776
       SAPSII on ICU admission58.4 ± 21.851.0 ± 17.90.010
       SOFA score on the day of candidemia9.9 ± 5.56.2 ± 4.9<0.001
       Septic shock on the day of candidemia, n (%)42 (100.0)28 (75.7)<0.001
      Clinical course
       Previous abdominal surgery, n (%)10 (23.8)12 (32.4)0.394
       Previous antifungal, n (%)7 (16.7)5 (13.5)0.697
       LOS in ICU before candidemia (days)7 [1–142]5 [4–152]0.044
      Mycological results
      C. albicans, n (%)18 (42.9)14 (37.8)0.650
      C. glabrata, n (%)11 (26.2)10 (27.0)0.933
      C. tropicalis, n (%)7 (16.7)5 (13.5)0.697
      C. parapsilosis, n (%)0 (0.0)4 (10.8)0.044
      C. krusei, n (%)2 (4.8)2 (5.4)0.877
      C. kefyr, n (%)1 (2.4)1 (2.7)1
       Multiple Candida3 (7.1)1 (2.7)0.618
      Candida with decreased susceptibility, n (%)13 (31.0)12 (32.4)0.887
       Time to negative blood culture (days)3 [1–18]4 [1–32]0.271
      Treatment
       Time elapsed to treatment initiation (days)2 [0–6]2 [0–10]0.016
       Antifungal0.650
      Caspofungin24 (57.1)23 (62.2)
      Azoles18 (42.9)14 (37.8)
       Duration of first-line antifungal (days)4 [1–190]13 [3–119]<0.001
       Total duration of antifungals (days)10 [2–190]16 [3–541]0.003
       Patients who switched to oral, n (%)1 (2.4)5 (13.5)0.093
       Adequate Candida source control, n (%)27 (64.3)33 (89.2)0.010
      Data are number (%), mean ± standard deviation, or median [range], as appropriate. ICU: intensive care unit; LOS: length of stay; RRT: renal replacement therapy; SAPSII: simplified acute physiologic score; SOFA: sepsis-related organ failure assessment

      Treatment success and mortality

      Complete response on day 30, as well as mycological success on day 30, were comparable between echinocandin and azoles groups (Table 1). In the overall population, the mortality rates at the end of antifungal treatment, on day 30, and on day 90, were respectively 43.0% (34/79), 49.4% (39/79), and 53.3% (42/79). The all cause hospital mortality rates were similar in echinocandin and azoles-treated patients at the end of antifungal treatment (p = 0.135), on day 30 (p = 0.581), and on day 90 (p = 0.650) (Table 1). These results were confirmed by Kaplan–Meier survival analysis that demonstrated that survival on day 90 was comparable between patients who received echinocandin (caspofungin) and those who received azoles (fluconazole or voriconazole) (log rank test, p = 0.542, Figure 2). Candidemia with a median time to death of 13 days is rapidly fatal despite antifungal treatment initiation in a median time of 2 days in the overall population. It has to be noted that mean time elapsed to antifungal initiation in deceased patients was shorter compared to alive patients, respectively 1.7 ± 1.5 and 2.7 ± 2.1 days (p = 0.016) (Table 2). Moreover, higher SAPSII or higher SOFA score was demonstrated in deceased compared to alive patients (p = 0.010 and p < 0.001, respectively) and all patients who were deceased on day 90 presented with a septic shock on the day of candidemia, compared to 75.7% for alive patients (p < 0.001) (Table 2).
      Figure 2
      Figure 2Kaplan–Meier analysis of survival to 90 days among patients with candidemia (log-rank test, p = 0.542).
      A multivariable logistic regression analysis based on propensity score with IPTW revealed that azoles treatment was not associated with a greater risk of all cause hospital 90-day mortality (p = 0.196) (Table 3). Multivariable analysis also demonstrated that higher SOFA score on the day of candidemia, a shorter time to antifungal initiation, and absence of adequate Candida source control were independently associated with a greater risk of 90-day mortality in the overall population (p < 0.001) (Table 3). Whereas SOT was more prevalent in alive compared to deceased patients on day 90 (p = 0.023) (Table 2), multivariable logistic regression failed to demonstrate that SOT was associated with a reduction in 90-day mortality (p = 0.147, Table 3).
      Table 3Multivariable analysis of risk factors for 90-day mortality.
      Risk factorAdjusted odds ratio95%CIP
      Solid organ transplantation0.2510.037–1.6240.147
      SOFA on the day of candidemia1.3631.214–1.530<0.001
      Time elapsed to treatment initiation0.5640.406–0.783<0.001
      Adequate Candida source control0.0480.011–0.211<0.001
      Azoles first-line therapy1.8980.719–5.0060.196
      CI: confidence interval; ICU: intensive care unit; SOFA: sepsis-related organ failure assessment.

      Discussion

      In a context of echinocandins resistance development, we studied the local epidemiology of Candida and addressed the issue of azoles use (mostly fluconazole) in ICU patients suffering candidemia through a retrospective multicenter study using a propensity score analysis. Fluconazole is indeed an acceptable alternative to an echinocandin as initial therapy in selected patients, including those who are not critically ill and who are considered unlikely to have a fluconazole-resistant Candida species according to IDSA guidelines (
      • Pappas P.G.
      • Kauffman C.A.
      • Andes D.R.
      • Clancy C.J.
      • Marr K.A.
      • Ostrosky-Zeichner L.
      • et al.
      Clinical practice guideline for the management of candidiasis: 2016 Update by the Infectious Diseases Society of America.
      ). This recommendation is based on a study that showed anidulafungin to be non-inferior to fluconazole in the treatment of invasive candidiasis: anidulafungin was successful in 75.6% of patients, as compared with 60.2% for fluconazole (
      • Reboli A.C.
      • Rotstein C.
      • Pappas P.G.
      • Chapman S.W.
      • Kett D.H.
      • Kumar D.
      • et al.
      Anidulafungin versus fluconazole for invasive candidiasis.
      ). Candidemia is a common disease in ICU patients as demonstrated by the estimated incidence of 3.5/1000 and 4.6/1000 admissions, respectively in medical and surgical ICUs. The main underlying diseases in this cohort of ICU patients suffering candidemia included diabetes, solid cancer, chronic kidney disease, and chronic heart disease. Baseline characteristics of patients treated with echinocandin and azoles were comparable, except for LOS in ICU before candidemia and previous abdominal surgery. The median LOS of 8 days in ICU before candidemia onset was indeed longer in the echinocandin group compared to a median of 1.5 days in the azoles group; this may account for the longer ICU LOS of patients treated with echinocandin compared to azoles. The longer LOS of patients in the echinocandin group may be explained by a higher proportion of patients who underwent abdominal surgery compared to the azoles group. Twenty-two (n = 22) patients had previous abdominal surgery with most of them (n = 18) receiving caspofungin; the more frequent use of echinocandin for patients with previous abdominal surgery was in agreement with IDSA guidelines that recommend an echinocandin as initial therapy for patients who had recent abdominal surgery (
      • Pappas P.G.
      • Kauffman C.A.
      • Andes D.R.
      • Clancy C.J.
      • Marr K.A.
      • Ostrosky-Zeichner L.
      • et al.
      Clinical practice guideline for the management of candidiasis: 2016 Update by the Infectious Diseases Society of America.
      ). Of note, previous abdominal surgery was not associated with a poorer outcome on day 90.
      Despite an antifungal treatment initiated with a median time of 2 days, candidemia was still a highly fatal disease in the ICU associated with a mortality rate of 43.0% on day 30 and 53.3% on day 90, which is consistent with a recent European study on invasive candidiasis in the ICU that reported a 30-day mortality of 42% (
      • Bassetti M.
      • Giacobbe D.R.
      • Vena A.
      • Trucchi C.
      • Ansaldi F.
      • Antonelli M.
      • et al.
      Incidence and outcome of invasive candidiasis in intensive care units (ICUs) in Europe: results of the EUCANDICU project.
      ). Short first-line antifungal treatments were not an exclusion criteria as this may exclude the most critically ill patients more prone to early death and consequently, to receive short treatments. The shorter mean time elapsed to antifungal initiation shown in deceased compared to alive patients may be in relation with the critical illness of deceased patients who had higher SOFA and SAPSII scores compared to alive patients. In this cohort of patients suffering candidemia during their ICU stay, Kaplan–Meier survival analysis and a multivariable logistic regression model based on propensity score with IPTW revealed that azoles treatment, mostly fluconazole, was not associated with a greater risk of 90-day mortality in ICU patients compared to caspofungin. Of note, factors that may impact 90-day mortality such as SAPSII, SOFA score, septic shock, durations of antifungal treatments, time elapsed to treatment initiation, and adequate control of Candida source, were comparable between echinocandin and azoles groups. This result is in line with an observational study that demonstrated no significant association between initial antifungal treatment with echinocandins and 28-day mortality (
      • Bailly S.
      • Leroy O.
      • Azoulay E.
      • Montravers P.
      • Constantin J.M.
      • Dupont H.
      • et al.
      Impact of echinocandin on prognosis of proven invasive candidiasis in ICU: a post-hoc causal inference model using the AmarCAND2 study.
      ). But, this is not the case in another study that showed the use of echinocandins for candidemia to be a protective factor (
      • Garnacho-Montero J.
      • Díaz-Martín A.
      • Cantón-Bulnes L.
      • Ramirez P.
      • Sierra R.
      • Arias-Verdu D.
      • et al.
      Initial antifungal strategy reduces mortality in critically Ill patients with Candidemia: a propensity score-adjusted analysis of a multicenter study.
      ): 58 patients deceased on day 90 in the fluconazole group compared to 51 in the echinocandin group. The protective advantage of echinocandins may be explained by the high number of patients (n = 10) in the fluconazole group that received inappropriate antifungal therapy, whereas all patients received appropriate therapy in the echinocandin group. In our series, all azoles treatments were considered appropriate.
      With respect to previous studies (
      • AmarCAND2 study group
      • Leroy O.
      • Bailly S.
      • Gangneux J.P.
      • Mira J.P.
      • Devos P.
      • et al.
      Systemic antifungal therapy for proven or suspected invasive candidiasis: the AmarCAND 2 study [Internet].
      ,
      • Reboli A.C.
      • Shorr A.F.
      • Rotstein C.
      • Pappas P.G.
      • Kett D.H.
      • Schlamm H.T.
      • et al.
      Anidulafungin compared with fluconazole for treatment of candidemia and other forms of invasive candidiasis caused by Candida albicans: a multivariate analysis of factors associated with improved outcome.
      ,
      • Andes D.R.
      • Safdar N.
      • Baddley J.W.
      • Playford G.
      • Reboli A.C.
      • Rex J.H.
      • et al.
      Impact of treatment strategy on outcomes in patients with candidemia and other forms of invasive candidiasis: a patient-level quantitative review of randomized trials.
      ,
      • Puig-Asensio M.
      • Pemán J.
      • Zaragoza R.
      • Garnacho-Montero J.
      • Martin-Mazuelos E.
      • Cuenca-Estrella M.
      • et al.
      Impact of therapeutic strategies on the prognosis of candidemia in the ICU.
      ), multivariable analysis identified higher SOFA score and absence of adequate Candida source control to be independent predictors of mortality. The need for an early adequate source control in critically ill patients with invasive candidiasis was recently recalled by the ESICM/ESCMID task force (
      • Martin-Loeches I.
      • Antonelli M.
      • Cuenca-Estrella M.
      • Dimopoulos G.
      • Einav S.
      • De Waele J.J.
      • et al.
      ESICM/ESCMID task force on practical management of invasive candidiasis in critically ill patients [Internet].
      ). Whereas SOT was more prevalent in alive compared to deceased patients, multivariable analysis failed to demonstrate a decrease in 90-day mortality for SOT patients; this may be in relation with the low number of patients who received SOT (n = 8). Anidulafungin was previously demonstrated to be associated with significantly faster clearance of blood cultures compared to fluconazole for patients with C. albicans infections (
      • Reboli A.C.
      • Shorr A.F.
      • Rotstein C.
      • Pappas P.G.
      • Kett D.H.
      • Schlamm H.T.
      • et al.
      Anidulafungin compared with fluconazole for treatment of candidemia and other forms of invasive candidiasis caused by Candida albicans: a multivariate analysis of factors associated with improved outcome.
      ); this was not the case in our series, in which no difference was observed in median time to negative blood culture between echinocandin and azoles groups for overall Candida, as well as for C. albicans.
      This study has some limitations. This is a retrospective study that did not allow a randomization of patients, but this was overcome by the use of a propensity score with the aim to reach the conditions of a randomized controlled trial. The area under the ROC curve was 0.848, indicating good ability of the score to accurately predict group assignment. Monitoring of caspofungin and azoles levels was not available for all patients, thus rendering drug exposure of patients non-evaluable. Caspofungin was the only echinocandin used in this multicenter study, but most experts agree that these agents are sufficiently similar to be considered interchangeable (
      • Kohno S.
      • Izumikawa K.
      • Yoshida M.
      • Takesue Y.
      • Oka S.
      • Kamei K.
      • et al.
      A double-blind comparative study of the safety and efficacy of caspofungin versus micafungin in the treatment of candidiasis and aspergillosis.
      ,
      • Pappas P.G.
      • Rotstein C.M.F.
      • Betts R.F.
      • Nucci M.
      • Talwar D.
      • De Waele J.J.
      • et al.
      Micafungin versus caspofungin for treatment of candidemia and other forms of invasive candidiasis.
      ).
      This study suggests that the use of recommended azoles for candidemia, mostly fluconazole, is a reasonable alternative to caspofungin for ICU patients suffering candidemia in regard to the local epidemiology of Candida. In a context of increasing Candida resistance, this study highlights the need to evaluate the clinical practice; it supports indeed the implementation of local guidelines through antifungal stewardship programs that promote the optimal use of antifungals.

      Authors’ contributions

      A.L.B., C.G., F.A., J.L.F., A.F., J.M., C.M., T.R., V.P., C.C., L.A., and G.L., designed the study. A.L.B., A.F., C.G., R.H., C.M., S.P., and G.L., contributed to data extraction. A.L.B., P.P., C.G., L.A., and G.L., analysed the data. All authors contributed to the interpretation of results and revised the different versions of the manuscript.

      Funding

      None.

      Conflict of interest

      None.

      Ethical approval

      Ethical approval was obtained from the Ethics Committee (N°19-37) of Hospices Civils de Lyon, France.

      Acknowledgments

      We thank the Antifungal group, the Anti-Infective Committee, and the Committee for Medicinal Products and Medical Devices of Lyon University Hospital.

      References

        • Almeida F.
        • Rodrigues M.L.
        • Coelho C.
        The still underestimated problem of fungal diseases worldwide.
        Front Microbiol. 2019; 10
        • AmarCAND2 study group
        • Leroy O.
        • Bailly S.
        • Gangneux J.P.
        • Mira J.P.
        • Devos P.
        • et al.
        Systemic antifungal therapy for proven or suspected invasive candidiasis: the AmarCAND 2 study [Internet].
        Ann Intensive Care. 2016; 6 ([cited 2019 Feb 7] Available from:)
        • Andes D.R.
        • Safdar N.
        • Baddley J.W.
        • Playford G.
        • Reboli A.C.
        • Rex J.H.
        • et al.
        Impact of treatment strategy on outcomes in patients with candidemia and other forms of invasive candidiasis: a patient-level quantitative review of randomized trials.
        Clin Infect Dis. 2012; 54: 1110-1122
        • Arendrup M.C.
        • Perlin D.S.
        Echinocandin resistance: an emerging clinical problem?.
        Cur Opin Infect Dis. 2014; 27: 484-492
        • Bailly S.
        • Leroy O.
        • Azoulay E.
        • Montravers P.
        • Constantin J.M.
        • Dupont H.
        • et al.
        Impact of echinocandin on prognosis of proven invasive candidiasis in ICU: a post-hoc causal inference model using the AmarCAND2 study.
        J Infect. 2017; 74: 408-417
        • Baldesi O.
        • Bailly S.
        • Ruckly S.
        • Lepape A.
        • L’Heriteau F.
        • Aupee M.
        • et al.
        ICU-acquired candidaemia in France: epidemiology and temporal trends, 2004-2013 – a study from the REA-RAISIN network.
        J Infect. 2017; 75: 59-67
        • Bassetti M.
        • Poulakou G.
        • Kollef M.H.
        The most recent concepts for the management of bacterial and fungal infections in ICU.
        Intensive Care Med. 2018; 44: 2000-2003
        • Bassetti M.
        • Giacobbe D.R.
        • Vena A.
        • Trucchi C.
        • Ansaldi F.
        • Antonelli M.
        • et al.
        Incidence and outcome of invasive candidiasis in intensive care units (ICUs) in Europe: results of the EUCANDICU project.
        Crit Care. 2019; 23
        • Colombo A.L.
        • de Almeida Júnior J.N.
        • Slavin M.A.
        • Chen S.C.A.
        • Sorrell T.C.
        Candida and invasive mould diseases in non-neutropenic critically ill patients and patients with haematological cancer.
        Lancet Infect Dis. 2017; 17: e344-e356
        • Garnacho-Montero J.
        • Díaz-Martín A.
        • Cantón-Bulnes L.
        • Ramirez P.
        • Sierra R.
        • Arias-Verdu D.
        • et al.
        Initial antifungal strategy reduces mortality in critically Ill patients with Candidemia: a propensity score-adjusted analysis of a multicenter study.
        Crit Care Med. 2018; 46: 384-393
        • Kohno S.
        • Izumikawa K.
        • Yoshida M.
        • Takesue Y.
        • Oka S.
        • Kamei K.
        • et al.
        A double-blind comparative study of the safety and efficacy of caspofungin versus micafungin in the treatment of candidiasis and aspergillosis.
        Eur J Clin Microbiol Infect Dis. 2013; 32: 387-397
        • Kullberg B.J.
        Arendrup MC: invasive candidiasis.
        N Engl J Med. 2015; 373: 1445-1456
        • Martin-Loeches I.
        • Antonelli M.
        • Cuenca-Estrella M.
        • Dimopoulos G.
        • Einav S.
        • De Waele J.J.
        • et al.
        ESICM/ESCMID task force on practical management of invasive candidiasis in critically ill patients [Internet].
        Intensive Care Med. 2019; ([cited 2019 Apr 3] Available from:)
        • Pappas P.G.
        • Rotstein C.M.F.
        • Betts R.F.
        • Nucci M.
        • Talwar D.
        • De Waele J.J.
        • et al.
        Micafungin versus caspofungin for treatment of candidemia and other forms of invasive candidiasis.
        Clin Infect Dis. 2007; 45: 883-893
        • Pappas P.G.
        • Kauffman C.A.
        • Andes D.R.
        • Clancy C.J.
        • Marr K.A.
        • Ostrosky-Zeichner L.
        • et al.
        Clinical practice guideline for the management of candidiasis: 2016 Update by the Infectious Diseases Society of America.
        Clin Infect Dis. 2015; : civ933
        • Pappas P.G.
        • Lionakis M.S.
        • Arendrup M.C.
        • Ostrosky-Zeichner L.
        • Kullberg B.J.
        Invasive candidiasis.
        Nat Rev Dis Primers. 2018; 4
        • Puig-Asensio M.
        • Pemán J.
        • Zaragoza R.
        • Garnacho-Montero J.
        • Martin-Mazuelos E.
        • Cuenca-Estrella M.
        • et al.
        Impact of therapeutic strategies on the prognosis of candidemia in the ICU.
        Crit Care Med. 2014; 42: 1423-1432
        • Reboli A.C.
        • Rotstein C.
        • Pappas P.G.
        • Chapman S.W.
        • Kett D.H.
        • Kumar D.
        • et al.
        Anidulafungin versus fluconazole for invasive candidiasis.
        N Engl J Med. 2007; 356: 2472-2482
        • Reboli A.C.
        • Shorr A.F.
        • Rotstein C.
        • Pappas P.G.
        • Kett D.H.
        • Schlamm H.T.
        • et al.
        Anidulafungin compared with fluconazole for treatment of candidemia and other forms of invasive candidiasis caused by Candida albicans: a multivariate analysis of factors associated with improved outcome.
        BMC Infect Dis. 2011; 11
        • Segal B.H.
        • Herbrecht R.
        • Stevens D.A.
        • Ostrosky-Zeichner L.
        • Sobel J.
        • Viscoli C.
        • et al.
        Defining responses to therapy and study outcomes in clinical trials of invasive fungal diseases: mycoses Study Group and European Organization for Research and Treatment of Cancer consensus criteria.
        Clin Infect Dis. 2008; 47: 674-683
        • Wang Y.
        • McGuire T.M.
        • Hollingworth S.A.
        • Dong Y.
        • Van Driel M.L.
        Antifungal agents for invasive candidiasis in non-neutropenic critically ill adults: what do the guidelines recommend?.
        Int J Infect Dis. 2019; 89: 137-145