Comparison of albicans vs. non-albicans candidemia in French intensive care units

Introduction Candidemia raises numerous therapeutic issues for intensive care physicians. Epidemiological data that could guide the choice of initial therapy are still required. This analysis sought to compare the characteristics of intensive care unit (ICU) patients with candidemia due to non-albicans Candida species with those of ICU patients with candidemia due to Candida albicans. Methods A prospective, observational, multicenter, French study was conducted from October 2005 to May 2006. Patients exhibiting candidemia developed during ICU stay and exclusively due either to one or more non-albicans Candida species or to C. albicans were selected. The data collected included patient characteristics on ICU admission and at the onset of candidemia. Results Among the 136 patients analyzed, 78 (57.4%) had candidemia caused by C. albicans. These patients had earlier onset of infection (11.1 ± 14.2 days after ICU admission vs. 17.4 ± 17.7, p = 0.02), higher severity scores on ICU admission (SOFA: 10.4 ± 4.7 vs. 8.6 ± 4.6, p = 0.03; SAPS II: 57.4 ± 22.8 vs. 48.7 ± 15.5, P = 0.015), and were less often neutropenic (2.6% vs. 12%, p = 0.04) than patients with candidemia due to non-albicans Candida species. Conclusions Although patients infected with Candida albicans differed from patients infected with non-albicans Candida species for a few characteristics, no clinical factor appeared pertinent enough to guide the choice of empirical antifungal therapy in ICU.


Introduction
The importance of fungal infections in Intensive Care Units (ICUs) was recently underlined by the EPIC Study II, since fungal agents represented 19% of positive isolates [1]. Moreover, candidemia still raises numerous therapeutic issues to Intensive Care physicians. The relationship between prognosis and early initiation of the adequate antifungal therapy is well established [2][3][4]. Ideally, adequate therapy must be started much before candidemia is ascertained, therefore much before the causative Candida species is identified and its susceptibility to antifungals is known. Broad spectrum antifungals have enriched the therapeutic arsenal in the past few years. The above therapeutic constraints might tempt clinicians to use these agents widely. However, apart from financial aspects, an excessive usage could become dele-terious by resulting in the selection of strains with reduced susceptibility.
In a prospective multicenter observational study named AmarCand performed to assess the current epidemiology, management and prognosis of invasive Candida infections in French ICUs, we demonstrated that 95.6% of C. albicans strains were susceptible to fluconazole whereas only 68% of non-albicans Candida were susceptible [5]. Thus, among data that could guide the choice of initial therapy, the availability of elements that would allow a binary distinction, with sufficient liability, between albicans or non-albicans Candida could represent an interesting first step.
In the present paper, the characteristics of patients from the AmarCand study, on ICU admission and at the onset of candidemia, are described according to whether candidemia was due to an albicans or a non-albicans Candida strain.

The AmarCand ("Analyse du Management en Anesthésie et Réanimation des Candidoses invasives") study
This study has already been described in two publications from the Group [5,6]. Briefly, AmarCand was a prospective, multicentre, national and observational study. Adult ICU patients with invasive Candida infection requiring a systemic antifungal therapy were included. Criteria used for diagnosis were those proposed in 2002 by the members of the European Organization for Research and Treatment of Cancer/Invasive Fungal Infections Cooperative Group and the National Institute of Allergy and Infectious Diseases Mycoses Study Group [7]. In accordance with the French law, approval of an Ethics Committee was not required. However, all patients gave informed consent to participate. Approval of the "Commission Nationale de l'Informatique et des Libertés" was obtained, ensuring that patient data were kept confidential according to the French regulation.
For each episode of invasive Candida infection, demographic characteristics, underlying diseases, current hospitalization, severity of illness, and process of care were recorded by each investigator on a standardized report form. Identification of the Candida isolates was performed in mycology and microbiology laboratories using the routine methods of each hospital. Isolates were classified as susceptible (S), susceptible-dose dependent (S-DD), or resistant (R) to antifungals according to CLSI interpretive categories [8].

Comparison of candidemia due to Candida albicans vs. nonalbicans Candida species
For this comparison, we identified patients from the AmarCand study with a candidemia acquired in ICU and exclusively due either to C. albicans or to one or more non-albicans Candida species. These groups were compared for the patients' characteristics on ICU admission and at the onset of candidemia.

Statistics
Data were analyzed using SAS ® 8.2 (SAS Institute Inc., Cary, NC, USA). Variables were expressed as mean values ± standard deviation for numerical variables and as frequencies and percentages for categorical variables. Groups were compared using the Chi-square and Fisher's exact tests. Continuous variables were compared using the Student's t test. Statistical significance was accepted at the 5% level. For the purposes of the present paper, we excluded from the 271 evaluable patients: 87 patients with invasive candidiasis but no candidemia, 13 patients with mixed candidemia due to albicans and non-albicans Candida species, and 35 patients who acquired candidemia before admission in ICU. Therefore, 136 patients were included in the present analysis.
The time from ICU admission to onset of candidemia was 13.8 ± 16.1 days. It was significantly shorter in the case of candidemia due to C. albicans: 11.1 ± 14.2 days vs. 17.4 ± 17.7 days with non-albicans Candida species (P = 0.02). Candidemia developed within six days after admission in ICU for 59 patients. Such an early infection was significantly more frequent when candidemia was due to C. albicans than when it was due to non-albicans Candida species (40/78 vs. 19/58, P = 0.03). Neutropenia (absolute neutrophil count <500 cells/mm 3 ) was concomitant to candidemia in nine patients. It was significantly more frequent when candidemia was caused by non-albicans Candida species than when it was caused by C. albicans (7/58 vs. 2/78, P = 0.04).
The main features of the patients' care at the onset of candidemia are shown in Table 2. There were no significant differences between the two patient groups, notably for previous exposure to azole agents.

Discussion
In this study, the characteristics of patients with candidemia caused by non-albicans Candida species versus Candida albicans in ICU were compared. The main result is that only a few significant differences were observed: severity of the disease, the time to candidemia onset and the rate of underlying neutropenia. So, we did not identify a parameter pertinent enough to allow a binary distinction between albicans or non-albicans Candida and to guide the choice of empirical antifungal therapy.
Studies that compare the epidemiological characteristics of patients with candidemia caused by non-albicans Candida species versus Candida albicans are scarce and their results are disparate.
Three studies included patients who were not all adults and/or not all admitted in ICU. Cheng et al. retrospectively analyzed 130 cases of fatal candidemia due to either a C. albicans (n = 68) or a non-albicans C. species (n = 62) [9]. Multivariate analyses showed that factors independently associated with C. albicans infection were the age ≥65 years, hyperleukocytosis (>15,000 cells/mm 3 ), and immunosuppressant therapy. Shorr et al. retrospectively analyzed the files of 245 candidemic patients from two different hospitals [10]. C. albicans represented 52% of the causative species. None of the parameters describing severity of the disease and previous exposition to azole agents was significantly predictive of a candidemia due to a non-albicans Candida strain. The third report arises from a large registry of 2,019 patients included between July 2004 and March 2008 in 23 North American hospitals [11]. Underlying hematological malignancy and bone marrow grafting were less common in patients with a candidemia due to C. albicans. Prior antifungal therapy was reported in 43% of the 2,019 patients. It was significantly less frequent in patients with C. albicans infection (38.8% vs. 46.5%, P < 0.001).
Three other studies were performed exclusively in ICU. In Australia Playford et al. carried out a three-year prospective, national survey [12]. A total of 179 episodes of candidemia were studied, of which 62% were related to C. albicans. Factors associated independently with a candidemia not related to C. albicans were recent intraabdominal surgery and recent exposition to systemic antifungal therapy. Chow et al. compared 79 patients Results are expressed as mean ± SD values or numbers (%) of patients. HIV: human immunodeficiency virus; SAPS: simplified acute physiology score; SOFA: sepsis-related organ failure assessment *Classified according to the criteria proposed by McCabe and Jackson [17] with candidemia due to C. albicans and 67 patients with candidemia due to non-albicans Candida species [13]. Previous exposition to azole agents, duration of central venous catheter implantation and the number of antimicrobial agents per day were associated with non-albicans Candida infection in multivariate analyses. Conversely, the duration of parenteral nutrition was associated with a reduced risk of non-albicans Candida infection. Finally, 189 candidemic patients (C. albicans: 56%) were included in the international, multicenter, retrospective study of Holley et al. [14]. Factors associated independently with candidemia due to non-albicans Candida species were female gender and duration of central venous catheter implantation using multivariate analysis. Our results obtained in 136 episodes of candidemia contrast with those of the three above studies performed in ICU [12][13][14]. Indeed, neither intra-abdominal surgery, previous exposure to azole agents, duration of central venous catheter implantation, nor female gender were associated with non-albicans Candida infection. Our results are however in line with those of Shorr et al., who could not identify clearly any parameter associated with the Candida species causative of candidemia [10].
Several consensual recommendations for the management of invasive Candida infections have been published in the last few years. In the French recommendations, which go back to 2004, the algorithm takes into account previous exposition to azole agents [15]. Empirical treatment based on fluconazole is proposed for patients who had not been exposed previously to azole agents. The 2009 North American recommendations propose empirical broad-spectrum treatment with an echinocandin for all ICU patients irrespective of previous exposition to azole agents [16]. Both recommendations propose to continue therapy with de-escalation if the causative strain is susceptible to fluconazole. Our results showing a high incidence (38.3%) of fluconazole non-susceptible Candida, and no pertinent parameter able to predict the species of causative Candida suggest that, in France and probably the rest of Europe, the use of the North American recommendation may now be more adequate. The use of echinocandin as first-line treatment (before the identification of the causative Candida strain and determination of its susceptibility) for all ICU patients suffering from candidemia could decrease the incidence of inappropriate empirical antifungal therapy and thus improve outcome of patients with such invasive candidiasis [3]. A de-escalation could be proposed in patients who are clinically stable when minimum inhibitory concentration (MIC) to fluconazole is ≤8 mg/L [16].

Conclusions
Comparison of the characteristics of ICU patients with candidemia caused by non-albicans Candida species versus Candida albicans did not allow identifying any parameter pertinent enough to guide the choice of empirical antifungal therapy.